Philosophy & Psychology Writings of Sarah-Jane Leslie

Journal of Memory and Language 65 (2011) 1531
Contents lists available at ScienceDirect
Journal of Memory and Language

journal homepage: www.elsevier.com/locate/jml
Do all ducks lay eggs? The generic overgeneralization effect

Sarah-Jane Leslie a,, Sangeet Khemlani b, Sam Glucksberg b
a b
Department of Philosophy, Princeton University, Princeton, NJ 08544, USA Department of Psychology, Princeton University, Princeton, NJ 08544, USA
a r t i c l e
i n f o
a b s t r a c t
Generics are statements such as tigers are striped and ducks lay eggs. They express general, though not universal or exceptionless, claims about kinds (Carlson & Pelletier, 1995). For example, the generic ducks lay eggs seems true even though many ducks (e.g. the males) do not lay eggs. The universally quantied version of the statement should be rejected, however: it is incorrect to say all ducks lay eggs, since many ducks do not lay eggs. We found that adults nonetheless often judged such universal statements true, despite knowing that only one gender had the relevant property (Experiment 1). The effect was not due to participants interpreting the universals as quantifying over subkinds, or as applying to only a subset of the kind (e.g. only the females) (Experiment 2), and it persisted even when people judged that male ducks did not lay eggs only moments before (Experiment 3). It also persisted when people were presented with correct alternatives such as some ducks do not lay eggs (Experiment 4). Our ndings reveal a robust generic overgeneralization effect, predicted by the hypothesis that generics express primitive, default generalizations. 2010 Elsevier Inc. All rights reserved.
Article history: Received 19 November 2009 revision received 20 December 2010 Available online 24 February 2011 Keywords: Generics Semantics Linguistics Concepts Philosophy Quantiers
Introduction Statements such as tigers are striped, ravens are black, ducks lay eggs and ticks carry Lyme disease are known as generics. Insofar as they are not used to convey information about a particular individual but rather information about a kind, these statements express generalizations (Carlson, 1977; Carlson & Pelletier, 1995). Generics have been widely studied by linguists and philosophers, and have recently attracted the attention of psychologists (Chambers, Graham, & Turner, 2008; Cimpian, Brandone, & Gelman, 2010; Cimpian, Gelman, & Brandone, 2010; Gelman, 2003; Gelman & Bloom, 2007; Gelman, Goetz, Sarnecka, & Flukes, 2008; Gelman & Raman, 2003; Gelman, Star, & Flukes, 2002; Gelman & Tardif, 1998; Goldin-Meadow, Gelman, & Mylander, 2005; Hollander, Gelman, & Star, 2002; Khemlani, Leslie, & Glucksberg,
Corresponding author. Address: Department of Philosophy, 1879 Hall, Princeton University, Princeton, NJ 08544, USA. E-mail address: sjleslie@princeton.edu (S.-J. Leslie).
0749-596X/$ - see front matter 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.jml.2010.12.005
2009; Khemlani, Leslie, Glucksberg, & Rubio-Fernandez, 2007; Prasada & Dillingham, 2006; Prasada & Dillingham, 2009). This paper presents four studies designed to investigate the relationship between peoples interpretations of generics and their interpretations of the quantiers all and some. From a logical point of view, statements like all tigers are striped, which we will refer to as universally quantied statements, are only true if every single tiger is striped. The existence of a single stripeless tiger is enough for the universal statement all tigers are striped to be false. Unlike universally quantied statements, generics such as tigers are striped can be true even if there are some stripeless tigers (Carlson, 1977; Gelman, 2003; Krifka et al., 1995; Lawler, 1973). Further, some generic statements are judged true even though a large percentage of the kind lack the property in question. For example ducks lay eggs and mosquitoes carry the West Nile virus are true, but male and immature ducks never lay eggs, and over ninety-nine percent of mosquitoes do not carry the West Nile virus (Carlson, 1977; Cohen, 1996; Leslie, 2007; Leslie, 2008; see also Cimpian, Brandone, et al., 2010; Cimpian, Gelman,
16
S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531
et al., 2010). Recent empirical work conrms that people often judge that these generics are true even when they know that a large percentage of the kind lacks the predicated property (Cimpian, Brandone, et al., 2010; Cimpian, Gelman, et al., 2010; Khemlani et al., 2007; Khemlani et al., 2009). Sentences involving the quantier some (henceforth, existentially quantied statements), e.g. some tigers are albinos or some dogs have only three legs, are true so long as there is at least one albino tiger or one three-legged dog. The corresponding generic statements tigers are albinos and dogs have only three legs, however, are rejected (Carlson, 1977; Carlson & Pelletier, 1995; Khemlani et al., 2009). Some generics are also rejected even though most members of the kind have the property for example, people reject books are paperbacks and Canadians are right-handed, despite knowing that more than fty percent of the kind has the property (Carlson, 1977; Carlson & Pelletier, 1995; Khemlani et al., 2009). For these reasons, semanticists distinguish the meanings of generic statements from the meanings of both universally and existentially quantied statements (e.g., Carlson, 1977; Krifka et al., 1995; Lawler, 1973). Unlike quantiers such as all, most, or some, generics are difcult to analyze from the semantic perspective since they cannot be described in set-theoretic terms (Leslie, 2007). Semantic analyses thus suggest that generics should be much more difcult to acquire and process than quantiers, due to their greater logical complexity (Leslie, 2008; see also Carlson, 1977; Cohen, 1996; Krifka et al., 1995; Pelletier & Asher, 1997). However, recent developmental ndings suggest that generics may be as easy as quantiers for young children to acquire and process, and in some cases even easier (Gelman, Coley, Rosengren, Hartman, & Pappas, 1998; Gelman & Tardif, 1998; Gelman, Hollander, Star, & Heyman, 2000; Hollander et al., 2002; Gelman, 2003; Gelman & Raman, 2003; Goldin-Meadow et al., 2005; Gelman et al., 2008; Hollander, Gelman, & Raman, 2009; Graham, Nayer, & Gelman, in press; Pappas & Gelman, 1998; Papafragou & Schwarz, 2005/2006; Tardif, Gelman, Fu, & Zhu, in press). The generics-as-defaults hypothesis In light of these considerations and others, several theorists have proposed that generics may express default generalizations (Csibra & Gergely, 2009; Gelman, 2010; Gelman & Brandone, 2010; Hollander et al., 2009; Leslie, 2007; Leslie, 2008). That is, the cognitive system may have an automatic, early-developing way of generalizing information from individuals to kinds (Baldwin, Markman, & Melartin, 1993; Graham, Kilbreath, & Welder, 2001; Keates & Graham, 2008; Leslie, 2008). These primitive kind-based generalizations are, according to this hypothesis, later articulated in language as generics. If correct, this hypothesis would explain why generics are understood and produced by young children, despite the semantic complexity that linguists have claimed generics exhibit (Carlson & Pelletier, 1995; Cohen, 1996; Gelman, 2003; Leslie, 2007; Leslie, 2008). The generalizations expressed by quantied statements, in contrast, represent more
sophisticated generalizations not the primitive default ones expressed by generics (Leslie, 2008). Leslie (2008) notes that such a hypothesis would explain otherwise puzzling cross-linguistic data: generic interpretations are always associated with less marked syntactic forms than quantied statements (Dahl, 1985; Krifka et al., 1995). For example, in English, one uses the words all or every to mark a universal generalization, and the word some to mark an existential generalizations. However, there is no word gen that is used to mark a generic generalization English speakers do not say gen tigers are striped like they say all tigers are striped. Instead, the generic interpretation is associated with the absence of a quantier word: tigers are striped. Similar patterns are found cross-linguistically no known language contains a word gen that exclusively marks a generic generalization. Rather, like English, generics are signaled in part by the absence of quantier terms (Carlson & Pelletier, 1995; Dahl, 1985). Default interpretations are often associated with less-marked surface forms (Chomsky, 2000). Thus this cross-linguistic pattern can be explained if generics express default generalizations. Quantier words such as all and some signal that the cognitive system must generalize in a non-default manner, whereas the unmarked generic form allows the cognitive system to rely on its default way of generalizing. The hypothesis that generics, unlike quantiers, express default generalizations generates a number of empirical predictions (Gelman, 2010; Leslie, 2008). For example, if understanding quantied statements requires deviating from the default mode of generalization, then both children and adults should sometimes fail to execute this deviation, and so should incorrectly treat quantied statements as generics. This tendency might be more pronounced in young children, but if the generics-as-default hypothesis is correct, adults should also be prone to making these errors, at least under some circumstances. Prior research conducted by Hollander et al. (2002) and Tardif et al. (in press) found evidence that young children may indeed treat quantied assertions as generics. However they did not nd this with their adult participants. In their study, Hollander et al. (2002) investigated the extent to which children and adults differentiated generics from universal and existential claims by asking them a variety of yes/no questions, each of which appeared either in universal form (e.g. do all shoes have laces?), generic form (e.g. do shoes have laces?), or existential form (e.g. do some shoes have laces?). They found that fouryear-olds and adults successfully differentiated between all three types of questions in their answers, but threeyear-olds did not. Instead, the three-year-olds gave the same responses regardless of whether the question was in universal, generic, or existential form. The difference between the age groups was due entirely to differences in their responses to the two quantiers Hollander et al. found no developmental differences in the responses to the generic questions across these three age groups. The three-year-olds responded as the adults did to the generic questions but then also responded in that same way to the universally and existentially quantied questions. The three-year-olds apparently handled the generic questions
17
like the older children and the adults did, but unlike the older age groups, the three-year-olds seemed to treat the quantied questions as though they were in fact generic questions (Hollander et al., 2002). Tardif et al. (in press) conducted a similar study of Mandarin Chinese speakers, and found the same results with three-year-olds. Mandarin speaking four-year-olds also showed the effect; it was not until age ve that Mandarin speaking children differentiated their responses to quantied questions from their responses to generic questions. Such ndings are predicted by the generics-as-default hypothesis: when confronted with a quantied statement that applies to an entire kind, the young preschoolers failed to inhibit their default tendency toward generic generalizations, and so they treated the quantied assertions as though they were generic. Thus instead of considering whether, e.g., all shoes have laces, the three-year-olds instead evaluated the generic shoes have laces, and responded accordingly. The English speaking three-yearolds were, however, able to respond accurately to quantied questions in a post-test session where all the questions concerned a small discrete set of items (e.g. are all the crayons in the box?) rather than categorywide generalizations, as was the case in the main experiment (Hollander et al., 2002), and the Mandarin speaking three- and four-year-olds were able to distinguish all from some in this context as well (Tardif et al., in press). The results of the main experiment were thus not due to a basic lack of competence with the quantiers; they seemed instead to do with the difculty of processing categorywide quantied statements. When confronted with a quantied claim about an entire category (as opposed to a specic subset), young preschoolers appear to rely on their interpretation of the corresponding generic. In Hollander et al.s (2002) and Tardif et al.s (in press) experiments, adults did not make the error of treating quantied statements as though they were generics. The generics-as-defaults hypothesis would predict that this would not always be the case, however. We wondered if Hollander et al.s and Tardif et al.s adult ndings would hold up across a broader range of test items. Would adult participants have any inclination to treat category-wide quantied statements as though they were in fact generics? Work by Jnsson and Hampton (2006), Osherson, Smith, Wilkie, Lopez, and Shar (1990) and Sloman (1993), Sloman (1998) suggests that this may occur in the case of the quantier all. For example, Jnsson and Hampton found that, in a wide range of circumstances, adults judged that it was more likely, e.g., that all ravens are black, than that all young jungle ravens are black. Clearly, such judgments are erroneous from the point of view of logic, because all ravens are black entails that all young jungle ravens are black. However, if the participants interpreted the universals as generics, then their responses would be much more reasonable, since the generic ravens are black could be true even if jungle ravens were, say, light brown when they are young (Jnsson & Hampton, 2006; Jnsson, personal communication). Indeed, Connolly, Fodor, Gleitman, and Gleitman (2007) conrmed that people routinely reason this way with generics (see also
Cimpian, Brandone, et al., 2010; Cimpian, Gelman, et al., 2010). Their focus was not on the interpretation of generic statements per se, but rather on prototypical properties (Rosch, 1978) properties that generics are well suited to express (Declerk, 1986; Geurts, 1985). While the generics-asdefaults hypothesis would explain Jnsson and Hamptons ndings, their data constitute only indirect support for the hypothesis. Is it possible to nd more direct evidence of adults interpreting quantied statements as generics? Data from a memory study by Leslie and Gelman (submitted for publication) supports the hypothesis that adults, as well as children, may sometimes treat quantied statements as generics. Leslie and Gelmans data suggest that adults and young children both frequently misremember category-wide statements quantied with all and most as generics. This is consistent with the idea that adults may at times treat universally quantied statements as generics. In light of these ndings, we wondered whether adults would be inclined to agree to some false universal statements when the corresponding generic is true, despite knowing about the falsifying counterinstances. Preliminary data reported by Khemlani et al. (2007) suggested that this was indeed the case participants agreed to statements such as all ducks lay eggs and all horses give live birth approximately half the time. We term this the generic overgeneralization (GOG) effect, since it involves overgeneralizing from the truth of a generic to the truth of the corresponding universal statement, as the generics-as-default hypothesis would predict. In Experiment 1, we replicated Khemlani et al. (2007) with a larger range of items. We predicted that the ndings from Khemlani et al. (2007) would hold up, and that people would often judge statements like all ducks lay eggs to be true. While Hollander et al. (2002) and Tardif et al. (in press) found evidence that preschoolers treat statements quantied with some as generics, we reasoned that it was unlikely that adults would make such errors in this experimental context. This is because generics are rarely accepted if no members of the kind have the property, i.e., if the corresponding existential statement is false (Carlson & Pelletier, 1995). Thus, it would not be feasible to nd cases where an existential statement was false, but the corresponding generic was true. The only possible GOG errors involving some, then, would involve people incorrectly judging a true existential statement to be false because the corresponding generic was false, e.g. judging some Canadians are right-handed to be false because the generic Canadians are right-handed is false. However, it is extremely easy to determine that such an existential is true: one only needs to think of a single right-handed Canadian in order to accept it. In this way, true existential statements are easy to evaluate, and so it is unlikely that adults would incorrectly rely on the default generic in such cases. It may be harder, though, for people to conrm that a universal statement is true, since they are only true if every single member of the kind has the property in question. Adults may thus save cognitive effort by relying on the generic to evaluate the universal. If true existential statements are easier to evaluate than universals, however, it is likely that no GOG effect will be found
18
for statements quantied with some. If the statement is easy to evaluate, participants will be unlikely to rely on the generic default. The scope of the overgeneralization effect The generics-as-default hypothesis predicts that adults will tend to incorrectly endorse false universal statements if the corresponding generic is true. A further question is whether this effect would be found for all types of generics, or whether it is more limited in scope. Generics are often used to express essential properties of a kind (Gelman, 2003; Gelman, 2010; Gelman & Brandone, 2010), or properties that bear a principled connection to the kind (Prasada & Dillingham, 2006; Prasada & Dillingham, 2009); for example tigers are striped, horses have hooves, doctors heal people. Prasada and Dillingham (2006), Prasada and Dillingham (2009) dene the notion of a principled connection so as to include only properties that are expected to be highly prevalent among members of the kind. However, generics such as ducks lay eggs and lions have manes seem to express essential properties in the sense of Medin and Ortony (1989) and Gelman (2003), and otherwise resemble principled properties, despite predicating properties that are only true of mature members of one gender of the kind (Cimpian, Brandone, et al., 2010; Cimpian, Gelman, et al., 2010; Leslie, 2007, 2008; Leslie, Khemlani, Prasada, & Glucksberg, 2009). A discussion of the relationship between principled and essential properties is beyond the scope of this paper (but see Prasada & Dillingham, 2009), so here we adopt the term characteristic property to indicate a property that bears a deep causal and explanatory relation to the kind in question a property that results from the nature of that kind. Characteristic properties can, in some cases, occur in only a minority of the members of the kind; for example, laying eggs would seem to stem from the nature of ducks, even though only the mature fertile female ducks possess the property. Leslie (2007), Leslie (2008) terms generics such as ducks lay eggs minority characteristic generics, since they express properties that are characteristic of the kind, but are only possessed by a minority of its members. Regardless of prevalence, not all generics express characteristic properties: for example cars have radios, pigeons sit on statues, and sharks attack swimmers (Leslie, 2007; Leslie, 2008; Prasada & Dillingham, 2006; Prasada & Dillingham, 2009). Would all these generics support a GOG effect? One hypothesis is that all generics produce GOG effects, regardless of whether the property in question is characteristic of the kind or not. On this hypothesis, people should accept universals like all cars have radios and all sharks attack swimmers as frequently as they accept minority characteristic universals such as all ducks lay eggs and all lions have manes. There are, however, some theoretical and empirical reasons to suspect that the GOG effect might be most pronounced in the case of characteristic properties (Khemlani et al., 2007). Characteristic properties are usually possessed by all or almost all members of a kind; in most cases, the only members of a kind that lack a characteristic (or essential) property are in some respect abnor-
mal, e.g., a stripeless albino tiger or a three-legged dog (Gelman, 2003). Thus, while the corresponding universal statements may be strictly speaking incorrect since there are some stripeless tigers and three-legged dogs assertions such as all tigers are striped and all dogs have four legs may be close enough to being correct for practical purposes. This is not so for non-characteristic properties: there is nothing abnormal about the many sharks that never attack swimmers, and so all sharks attack swimmers is patently incorrect. In this way, the tendency to substitute a judgment of the generic for the universal will be generally more successful when the property in question is a characteristic one, and so adult participants may be less likely to show a GOG effect if the property is not characteristic of the kind. However, there are some characteristic properties that do not occur in all or almost all members of the kind, namely minority characteristic ones, as in ducks lay eggs and lions have manes (Leslie, 2008). If this hypothesis is correct, then the GOG effect should be observed for minority characteristic predications, such as all ducks lay eggs and all lions have manes. Data from Khemlani et al. (2007) favored the hypothesis that the GOG effect was primarily to be found for minority characteristic generics. The limited range of test items prevented any rm conclusions being drawn, and so Experiment 1 used an expanded set of items to test the GOG effect. A full taxonomy of the types of predications used in these studies may be found in Table 1 below. If adults do indeed erroneously accept statements like all ducks lay eggs, does this necessarily support the generics-as-default hypothesis? One alternative explanation is that people may interpret the universal quantier all as applying to subkinds of ducks, rather than to individual ducks. That is, people may understand all ducks lay eggs to mean all kinds of ducks lay eggs e.g., Mallard ducks lay eggs, Muscovy ducks lay eggs, and so on. On this interpretation, it would be correct to accept all ducks lay eggs, and so this would not constitute evidence that participants were defaulting to their judgment of a generic in place of a universal. An alternative explanation is that people may interpret all ducks to apply to only a subset of ducks, namely the fertile female ducks. That is, the context may restrict the scope of the quantier, so that people do not interpret all ducks to apply to every single duck in the world, but rather to a subset of those ducks (Stanley, 2007; Stanley & Szabo, 2000). If this explanation is correct, then peoples acceptance of all ducks lay eggs would again not constitute evidence for the generics-as-defaults hypothesis. Experiment 2 sought to address these issues. In Experiment 2a, participants were given population information (e.g. there are 431 million ducks in the world) so as to prime them to interpret the subsequent statement (all ducks lay eggs) as quantifying over (a) individual ducks, and (b) every single duck in the world. If the GOG effect is driven by quantication over subtypes, or by contextual quantier domain restriction, then it should disappear in the context of population information. Further, in Experiment 2b, participants performed a paraphrase task in which they were asked to paraphrase the sentence they evaluated. The paraphrases were then coded
19
for any indications that participants had generalized over subkinds or interpreted the scope of the quantier to be restricted to a subset of the kind. If the GOG effect was simply due to generalization over subkinds or restricted quantier scope, then there should be evidence of this in participants paraphrases. In particular, they should be more likely to provide a paraphrase that either refers to subkinds or reects quantier domain restriction if they accepted the universal than if they rejected it. Another alternative explanation for the effect would be that participants were simply ignorant of the biological facts. Experiment 3 addressed whether the participants who accepted statements such as all ducks lay eggs did indeed know that male ducks do not lay eggs. It is possible though perhaps unlikely that people accepted these false universal statements simply out of ignorance. The experiment used a blocked design; one block replicated Experiment 1 and the other asked participants to evaluate statements involving counterexamples such as male ducks lay eggs. In addition to controlling for background knowledge, we were interested in whether the GOG effect would persist after participants had been asked to think about these counterexamples. Adults take counterexamples into consideration when evaluating universals, so if participants had processed these statements as universals but were somehow unable to bring counterexamples to mind, then the GOG effect should disappear, since counterexamples should have been readily called to mind. Conversely, if participants processed the universal statements as though they were generics, as we predicted, then they should have accepted all ducks lay eggs to some extent, even after rejecting the statement male ducks lay eggs. After all, the truth of the generic ducks lay eggs is perfectly compatible with male ducks failure to do so. Finally, Experiment 4 tested whether the GOG effect persisted when people were asked to choose between two statements such as all ducks lay eggs and some ducks do not lay eggs. The experiment offered people a
correct alternative, and thus made salient the potential error that would be involved in accepting the universal. If the GOG effect is due to participants relying on their judgment of the generic instead of evaluating the universal, then they may continue to erroneously prefer the universal statement. Experiment 1: the generic overgeneralization effect Preliminary work suggests that the GOG effect may occur most clearly for predications that express gender-specic properties that are characteristic of the kind in question, e.g., how the members of the kind reproduce (Khemlani et al., 2007). Can any GOG effect be found for non-characteristic properties, or is it limited to characteristic properties in the case of adults? Prasada and Dillingham (2006), Prasada and Dillingham (2009) and Leslie (2007), Leslie (2008) have identied various other types of generics, so we included each identied type in our experiments, plus a category of statements that are false in generic form. Examples are given in Table 1. Our rst and last types of predications, quasi-denitional and false generalizations, were included for comparison purposes. False generalizations tend to be rejected in generic form despite predicating prevalent properties of the kind (Khemlani et al., 2009). The inclusion of this category allowed us to determine which, if any, of the other types were more likely to be accepted in universal form. If other types of statements were accepted in universal form more frequently than false generalizations, then that would suggest a GOG effect for them. At the other end of the spectrum are the quasi-denitional predications, which are genuinely true in universal form. Acceptance in universal form of any predication types other than quasi-denitional ones constitutes an error, since there are always some members of the kind that lack the property. Majority characteristic predications involve characteristic properties that are true of all the normal members of the
Table 1 Various types of generic generalizations used in Experiments 1, 2, and 3. Predication type Quasi-denitional Majority characteristic Example Triangles have three sides Tigers have stripes Truth value of the generic True True Description Property must be universally true of all the members of the kind; no exceptions Property must be central, principled or essential (Gelman, 2003; Medin & Ortony, 1989) namely, it must be directly related to the nature of the kind in question. It must also be prevalent though not universally had among members of the kind; while some exceptional members (e.g. albino tigers) fail to possess it, all the normal members of the kind must possess it Property must be central, principled or essential (Gelman, 2003; Medin & Ortony, 1989) namely, it must be directly related to the nature of the kind in question. However, it must only be held by a minority of the kind. For our purposes we restricted these items to methods of gestation, methods of nourishing the very young, and characteristic physical traits had only by one gender Property must be prevalent among members of the kind, and must not be a principled connection (Prasada & Dillingham, 2006, 2009). Property must only be had by a small minority of the kind, and must signify something dangerous and to be avoided Property must be prevalent among members of the kind and there must be a sufciently salient alternative property (e.g. being left-handed), so that the generic form of the predication sounds false or mistaken
Minority characteristic
Lions have manes
True
Majority Striking False generalization
Cars have radios Pit bulls maul children Canadians are right-handed
True True False
20
kind (they map onto Prasada & Dillinghams, 2006; Prasada & Dillinghams, 2009 notion of principled connections), and so it would not be surprising if people accepted those statements in universal form. We were particularly interested in minority characteristic, majority, and striking predications throughout our studies. Recent work suggests that these different types of generic predications can be distinguished along various dimensions (Cimpian, Brandone, et al., 2010; Cimpian, Gelman, et al., 2010; Khemlani et al., 2009; Prasada & Dillingham, 2006; Prasada & Dillingham, 2009; Leslie et al., 2009), and so they may be treated in different ways on our tasks. Thus our inclusion of minority characteristic, majority and striking items allowed us to better test the extent and limits of the GOG effect. Method Participants and procedure 56 volunteers participated in the experiment over the Internet. Participants were chosen from Amazons Mechanical Turk system for human interface tasks, an online forum through which individuals can participate in tasks for monetary compensation. All spoke English as their rst language and none had participated in experiments concerning generics before. They were asked to judge their
agreement to each item, one at a time, and registered their responses by selecting buttons marked yes and no over the Internet through an interface written in Ajax. Materials and design Participants were presented with 72 experimental items randomly embedded in a list of 60 llers. The 72 experimental items consisted of six types of predications, and examples are given in Table 1. A full listing of the materials is provided in the Appendix. Each predicate type appeared in one of three statement types: existential (e.g., some ducks lay eggs), generic (e.g., ducks lay eggs), or universal (e.g., all ducks lay eggs). This generated a 6 (predicate type) 3 (statement type) repeated measures factorial design. Statements were counterbalanced such that each participant saw any given predicate in only one statement type. We also included 30 false and 30 true ller items. Filler items were non-quantied factual assertions, such as Paris is the capital of France (true) and New York City is in England (false). Results and discussion Fig. 1 shows the proportions of yes responses to each of the eighteen types of assertions. Existentially quantied
Experiment 1
80 0 40
Existential
Quasi-definitional Majority characteristic
Generic
Minority characteristic Majority
Universal
Striking False generalization
Experiment 2a
80 0 40
Existential
Generic Experiment 2b
Universal
40
80
Existential
Generic
Universal
Fig. 1. Proportions of agreement for Experiment 1, 2a, and 2b as a function of statement type and predicate type; 95% condence intervals shown.
21
and generic statements were deemed true most of the time, while agreement to universal statements depended on the type of predication expressed. In generic form, quasi-denitional and majority characteristic statements were judged true 90% and 96% of the time, respectively. Minority characteristic predications in generic form were judged true 85% of the time, and majority and striking were judged true in generic form 82% and 77% of the time, respectively. The various assertions in existential form were each judged true over 90% of the time, with the exception of quasi-denitional existentials, which had a 78% acceptance rate. This relatively low acceptance rate can be attributed to the scalar implicature effect, a wellknown phenomenon in the language comprehension literature. Given, say, that all triangles have three sides, it is pragmatically odd to assert that some triangles have three sides; people sometimes interpret some as implying not all, and so are reluctant to agree to universally true assertions quantied by some (Bott & Noveck, 2004). False generalizations were rejected most of the time when they appeared in generic form the agreement rate for them was only 38%. Of more interest are the universally quantied predications. Participants were prone to the GOG effect for minority characteristic and majority characteristic predications: universally quantied minority characteristic predications were judged true 51% of the time and principled predications were judged true 78% of the time when, in fact, all of those statements were false. Further, 40 out of 56 participants judged universal minority characteristic predications true over 30% of the time (binomial test, p < .005), and 10 out of 12 items were judged true over 30% of the time (binomial test, p < .05), where 30% reects a proportion high enough to indicate a GOG effect. In contrast, universally quantied majority and striking predications were incorrectly judged true only 13% and 14% of the time, respectively. The proportions of agreement data were subjected to a 6 3 within-subjects ANOVA. As is conventional, we use F1 and F2 to report participant and item analyses respectively, as well as min F0 (Clark, 1973). There was a main effect of predicate type, F1(5, 275) = 95.21, p < .0001, F2(5, 65) = 30.50, p < .0001, min F0 (5, 111) = 23.10, p < .0001, reecting the higher agreement rate to quasi-denitional, principled, and minority characteristic statements than majority, striking, or false generalization statements. There was also a main effect of statement type, F1(2, 110) = 260.33, p < .0001, F2(2, 130) = 338.03, p < .0001, min F0 (2, 229) = 147.07, p < .0001, reecting the higher agreement for some statements (91%) than for generic statements (78%), with the lowest agreement for all statements (42%). Finally, there was a reliable interaction between predicate type and statement type, F1(10, 550) = 72.19, p < .0001, F2(10, 130) = 42.48, p < .0001, min F0 (10, 301) = 26.55, p < .0001. The interaction can be interpreted in terms of the differences in agreement rate for statement type as a function of predicate type. Agreement rates for existentially quantied assertions were comparable across predicate types. In contrast, quasi-denitional and majority characteristic predicates yielded robust agreement rates in the universal condition, minority characteristic predicates tended towards agreement, and majority, striking, and false generalizations were
usually rejected. Finally, generic statements differed as a function of predicate type, with false generalizations eliciting lower agreement than any of the other predications. Planned comparisons revealed that quasi-denitional, majority characteristic, and minority characteristic assertions were each accepted in universal form signicantly more often than false generalizations, ts(110) > 8.01, ps < .0001, ds > 1.53. Neither majority, t(110) = .57, p = .57, d = .11, nor striking assertions were accepted more often than false generalizations, t(110) = .10, p = .92, d = .02. These results support the hypothesis that the GOG effect only occurs for predications that express characteristic properties of the kind. Experiment 1 found that the GOG effect occurs on approximately half the trials when the property is characteristic of the kind but only occurs in one gender. No evidence was found of the GOG effect occurring when the property in question is not characteristic of the kind. Thus Experiment 1 provides evidence that the GOG effect is limited to characteristic properties, at least for adults. It also provides evidence for the generics-as-default hypothesis more generally, since, as predicted, adults exhibited a tendency to accept false universals with true corresponding generics. However, one alternative explanation of these data is that people interpret the quantier all as applying to subkinds instead of individual members of the kind, in which case Experiment 1 would not in fact constitute evidence for generics being defaults. Another explanation is that people interpret the quantier all as applying only to a subset of the kind in particular, to only one gender. That is, perhaps they contextually restricted the scope of the universal quantier so that it does not apply to the entire kind, but only to one gender of the kind just as if someone asserts every student passed the exam, she would naturally be interpreted as talking about the students in her class, not about every single student in the world (Stanley, 2007; Stanley & Szabo, 2000). Experiment 2 was designed to address these alternative deationary explanations. Experiment 2: is the GOG effect due to quantication over subkinds or domain restriction? To determine whether the GOG effect can be explained by supposing that participants generalize over subkinds or restrict the scope of the universal quantier, Experiment 2a tested whether the GOG effect persisted in a context in which participants were directed to consider every single individual member of the kind. Experiment 2b solicited paraphrases, to test whether they indicate reliance on a subkind interpretation or quantier domain restriction. Experiment 2a: priming individual-based interpretations with population information Method Participants and procedure Twenty-seven volunteers from the same population as Experiment 1 participated over the Internet for pay. All were native English speakers and none had participated
22
in experiments on generics in the past. The study used the same procedure as in Experiment 1, except that on each trial, to prime them to consider each and every individual member of a kind in the world, instead of subkinds or a subset of those individuals, participants were provided with additional information about population estimates for the kind in question. For example, participants were visually presented with the following: Suppose the following is true: there are 431 million ducks in the world. Do you agree with the following: all ducks lay eggs. The numerical information was articial, but was designed to convey plausible population estimates for each kind. Materials and design Experiment 2a used the same set of items used in Experiment 1, except that it dropped three items: rectangles are geometric gures, US Presidents are over 35, and strokes cause paralysis. These items were dropped because it was not possible to construct meaningful population estimates for each item, e.g., it is nonsensical to say that there are 45 million rectangles in the world, because the set of rectangles is impossible to count. Filler statements were dropped from the study. In total, participants evaluated 69 items. Each item appeared in existential, generic, or universal form. The study thus generated a 6 (predicate type) 3 (statement type) repeated measures factorial design. Results and discussion Fig. 1 provides the proportions of yes responses to the different assertions used in Experiment 2a. Participants accepted minority characteristic universal assertions 30% of the time, whereas they accepted majority and striking assertions only 6% and 5% of the time respectively. Despite a reduction of the GOG effect, the overall pattern results of the present study replicated those of Experiment 1. The proportions of agreement data were subjected to a 6 3 within-subjects ANOVA, which revealed a main effect of predicate type, F1(5, 130) = 100.18, p < .0001, F2(5, 62) = 43.02, p < .0001, min F0 (5, 116) = 30.09, p < .0001, a main effect of statement type, F1(2, 52) = 192.50, p < .0001, F2(2, 124) = 315.39, p < .0001, min F0 (2, 117) = 119.54, p < .0001, and a reliable interaction between predicate type and statement type, F1(10, 260) = 54.85, p < .0001, F2(10, 124) = 31.77, p < .0001, min F0 (10, 267) = 20.12, p < .0001. Planned comparisons showed that, as in Experiment 1, minority characteristic assertions were accepted in universal form signicantly more often than false generalizations, t(52) = 5.64, p < .0001, d = 1.54. As before, majority t(52) = .50, p = .48, d = .19, and striking assertions were not accepted reliably more often than false generalizations, t(52) = .57, p = .57, d = .15. Our ndings demonstrate that the GOG effect occurred even in a context in which people were asked to consider a population estimate designed to focus attention on every individual member of the kind in the world, rather than on subkinds or on a subset of individuals. This additional information may have reduced the overgeneralization effect; people accepted minority characteristic assertions
30% of the time in this study as compared to 51% of the time in Experiment 1. Thus some of the GOG effect found in Experiment 1 may have been due to some participants employing a subkinds interpretation or a restricted interpretation. However, the results of Experiment 2a indicate that the effect cannot be entirely explained in this way, since the effect remained in a context that encouraged generalization across each and every individual in the world. It is also possible that the GOG effect was reduced here because the context primed participants to employ a more analytic mode of thinking, due to considering numerical information about populations (see Alter, Oppenheimer, Epley, & Eyre, 2007). As a result, the participants may have answered more carefully, and therefore they may have been less likely substitute a default generic judgment for a universal one. The reduction in the GOG effect may thus be due in part to the elimination of subkind and restricted interpretations. Experiment 2b tested the deationary interpretations further by eliciting participants paraphrases of quantied and generic assertions. Experiment 2b: paraphrases of existential, generic, and universal assertions Method Participants and procedure Sixteen volunteers from the same population as Experiment 1 participated over the Internet for pay. All were native English speakers and none had participated in experiments on generics in the past. The experiment consisted of two phases. Participants rst performed an evaluation phase in which they carried out the same agreement task as in Experiment 1. They then carried out a second phase, in which they were given the same assertions as in the rst phase and were asked to type out a paraphrase of each assertion. Participants were specically asked to produce paraphrases that retained as much of the original meaning of the statement as possible. The paraphrase task examined whether individuals used language that suggested a subkind interpretation or a restricted interpretation, e.g., whether a participant paraphrased the universal all ducks lay eggs as all types (/kinds/sorts/species) of ducks lay eggs, or as all female ducks lay eggs. Materials and design Experiment 2b used the same set of items used in Experiment 1, except that all the ller statements were omitted since we were not interested in how the llers were paraphrased. In total, participants evaluated 72 items. Each item appeared in existential, generic, or universal form. The study thus generated a 6 (predicate type) 3 (statement type) repeated measures factorial design. Results and discussion The results of the rst phase of Experiment 2b replicated those of Experiment 1, and are shown in Fig. 1. Participants accepted minority characteristic universal
23
assertions 48% of the time. The proportions of agreement data were subjected to a 6 3 within-subjects ANOVA. There was a main effect of predicate type, F1(5, 75) = 21.46, p < .0001, F2(5, 65) = 19.83, p < .0001, min F0 (5, 138) = 10.31, p < .0001, a main effect of statement type, F1(2, 30) = 87.23, p < .0001, F2(2, 130) = 125.75, p < .0001, min F0 (2, 76) = 50.8, p < .0001, and a reliable interaction between predicate type and statement type, F1(10, 150) = 24.89, p < .0001, F2(10, 130) = 25.24, p < .0001, min F0 (10, 279) = 12.53, p < .0001. This pattern of results is comparable to those of the previous studies. In the second phase of the experiment, participants produced paraphrases of the statements that they had evaluated in the rst phase. The authors coded the paraphrases on three separate dimensions. First, the paraphrases were coded based on what kind of statement was produced (existential, generic, universal, or other). Coding was based on the semantic signicance of the paraphrase, so paraphrases that used the quantiers none or no, e.g., no moose are without antlers, were coded as universal quantiers. Likewise, assertions that made use of the quantiers always and never were coded as universals. Denite and indenite singular assertions, e.g., a paperback is a book, were coded as generics, as were qualied generics, e.g., female ducks lay eggs. Paraphrases that made use of the quantiers there are and a few were coded as existentials, as were negated universals, e.g., not all cheetahs are fast runners. Finally, everything that did not fall into these categories was coded as other. Paraphrases were also coded on (a) whether they mentioned subtypes or not, e.g., some types of pencils are made of wood, and (b) whether they made reference to a restricted subset of the kind, e.g., female ducks lay eggs. The authors agreed on 79.8% of responses (Cohens kappa = .73, z = 43.7, p < .0001). Differences between the coding were reconciled on a case-by-case basis. Table 2 shows the form of participants paraphrases as a function of the statement that they were asked to paraphrase. As the table shows, they paraphrased existential assertions as existentials 62% of the time, generics as generics 59% of the time, and universal as universals 51% of the time. Across all responses, only 16% of the paraphrases were classied as other. For those paraphrases that were different from the original assertions, participants paraphrased universals as generics on 36% of trials, which happened signicantly more often than when they paraphrased existentials, generics, or universals as any other type of statement (MannWhitney tests, all zs > 3.59, all ps < .0001). (That is, universal-as-generic
Table 2 Percentages of statements that were paraphrased as an existential, as a generic, as a universal, or in some other form as a function of the statement type of the original assertion in Experiment 2b. Paraphrased as Original statement type Existential Existential Generic Universal Other 62 12 3 24 100 Generic 8 59 18 16 100 Universal 4 36 51 9 100
paraphrases were provided reliably more often than universal-as-existential paraphrases, than generic-as-existential paraphrases, than existential-as-generic paraphrases, and so on for all the other possible combinations.) This nding also is consistent with the generics-as-default hypothesis, as it reveals a tendency to paraphrase universals as generics. Only 1% of paraphrases included subtyping language, and 1.6% of them included a subset restriction across the experiment as a whole. Both proportions were signicantly greater than 0 (permutation tests, ps < .001). However, the proportion of paraphrases with subtyping language did not correlate with the proportion of GOG errors (Kendalls tau = .02, z = .58, p = .56), and neither did the proportion of paraphrases that were restricted to a relevant subset of the kind (Kendalls tau = .02, z = .76, p = .45). If the GOG effect was driven by these interpretations, then one would expect to see a higher proportion of such paraphrases when the universal was accepted vs. when it was rejected. That is, if the GOG effect was due to participants taking the minority characteristic universals to quantify over subkinds or a restricted subset, then participants who agreed to the universals should be more likely to produce paraphrases that make reference to subtypes or subsets. No such effect was observed, however. In light of this nding, plus the overall low rate of subtyping and restricting paraphrases, Experiment 2b suggests that the GOG effect was not due to subkind or restricted interpretations. An alternative way of assessing whether participants answers were due to subkind or restricted interpretations could have been to point out participants mistakes to them and then see if and how they attempted to defend their original responses. However, peoples tendency to produce post hoc rationalizations for phenomena that are driven by wholly unrelated causes is well documented (e.g., Nisbett & Wilson, 1977). Thus, participants may claim that in fact they were employing a subkind interpretation in order to justify an otherwise incorrect response that was driven by a default to the generic interpretation just as Nisbett and Wilsons participants claimed to prefer the detergent Tide because, e.g. their mother used Tide, when in fact their preference was driven by priming effects. For these reasons, we elicited our paraphrases in a neutral context (i.e. without confronting participants about their errors), in order to minimize the chances of post hoc rationalizations. Experiments 1, 2a, and 2b reveal a GOG effect that cannot be explained by a subkind interpretation of the assertions, or by a domain-restricted interpretation. However, the results may be trivially attributed to a lack of knowledge of the gender discrepancies in the properties expressed by the minority characteristic predications. Perhaps participants were unaware of the gender discrepancies in the minority characteristic statements. To investigate this issue, Experiment 3 tested participants knowledge of gender discrepancies for the minority characteristic items e.g., whether people know that male horses do not give live birth. Experiment 3 also tested whether the GOG effect would be eliminated if people were forced to consider relevant knowledge.
24
Experiment 3: minimizing the generic overgeneralization effect Participants completed the same evaluation task as in Experiment 1, but were also given a knowledge test in which they judged the truth values of false gendered minority characteristic statements such as male ducks lay eggs or female lions have manes. The study manipulated whether the evaluation test was presented before or after the knowledge task. When the evaluation task was presented rst, we expected to replicate the results of Experiment 1, with the additional manipulation check that participants were expected to reject statements such as male ducks lay eggs in the subsequent knowledge test. Otherwise, we could attribute performance in Experiment 1 to participants lack of knowledge of the gender discrepancies in question. When the knowledge test was presented before the evaluation task, then participants may no longer agree with statements such as all ducks lay eggs because they would have recently rejected statements such as male ducks lay eggs, and so would have counterexamples to the universal statements in mind (Johnson-Laird & Hasson, 2003). If participants demonstrated knowledge of the gender discrepancies but nevertheless showed the GOG effect, then this would suggest that the effect is robust and difcult to eliminate. For the evaluation task, we asked participants from the same population as in Experiment 1 to evaluate the truth values of the eighteen different kinds of assertions. We employed a 6 (predicate type) 3 (statement type) 2 (task order) design, and we presented a total of 72 counterbalanced target stimuli. These same participants were given a knowledge test, in which they were given counterexamples to test their knowledge of the relevant gender discrepancies, either before or after the evaluation task.
each item, one at a time, following the same procedure as in Experiment 1. Materials and design Experiment 3 used the same set of 72 items used in Experiment 1, examples of which are given in Table 1. We used a blocked design. In one block, participants were presented with a knowledge test of 55 statements, 12 of which were gendered versions of the minority characteristic generic statements used in Experiment 1. The statements were false, and participants were expected to reject them. Table 3 lists the gendered minority characteristic statements used in the study. The rest of the 55 statements were designed to ensure that there were approximately equal numbers of true and false items, and more importantly that participants did not just employ the strategy of rejecting all the gendered statements out of hand. For example, these items included true gendered statements such as female zebras have stripes. In the other block, participants evaluated the items as they had in Experiment 1. Half of the participants received the knowledge test before the evaluation task and the other half received it after. The ller statements that were used in Experiment 1 were omitted so as to maximize the likelihood that participants would remember their judgments about the gender specicity of the properties when presented with the knowledge test rst. Dropping the ller items thus maximized the likelihood that participants would remember that male ducks do not lay eggs, and would reject all ducks lay eggs. Results and discussion Participants were aware of the relevant gender discrepancies in minority characteristic predications, except for one item: 73% of participants believed that female cardinals are red. Accordingly, this item was dropped from the analyses of all four experiments. The remaining false gendered minority characteristic items were rejected 84% of the time, and levels of rejection did not vary by whether the knowledge task was presented before or after the evaluation task, t(38) = 1.00, p = .32, d = .32. Proportions of agreement to each of the false gendered minority characteristic statements are given in Table 3. We also included
Method Participants and procedure 40 volunteers from the same population as Experiment 1 participated over the Internet for pay. All were native English speakers and none had participated in experiments on generics in the past. Participants judged the truth of
Table 3 Percentage agreement to false gendered minority characteristic statements used in Experiment 3. Predication type Minority Minority Minority Minority Minority Minority Minority Minority Minority Minority Minority Minority characteristic characteristic characteristic characteristic characteristic characteristic characteristic characteristic characteristic characteristic characteristic characteristic Example Male sheep produce milk Male snakes lay eggs Male horses give live birth Male ducks lay eggs Male pigs suckle their young Male insects lay eggs Female lions have manes Female deer have antlers Female moose have antlers Female goats have horns Male kangaroos have pouches Female cardinals are reda Proportion of agreement 3 3 8 8 8 18 20 23 25 33 33 73 Agreement rate Low Low Low Low Low High High High High High High
a Note: participants erroneously agreed with this item 73% of the time, and so the corresponding ungendered minority characteristic statement, cardinals are red, was dropped from the analyses of Experiments 1, 2a, 2b, 3, and 4.
25
items that participants could have deemed gender specic (e.g., female zebras have stripes), but we found that they were aware of the lack of gender specicity in those items. These data suggest that the results in Experiment 1 cannot be attributed to mere ignorance on the part of our participants. The overall agreement rates for the evaluation task were comparable to those found in Experiment 1. Existentially quantied statements were judged as true most of the time (91%). Just as in Experiment 1, quasi-denitional existentials exhibited a scalar implicature effect (76%), i.e., a relatively low agreement level because of the implication that some is taken to mean not all (Bott & Noveck, 2004). Statements in generic form were judged true 77% of the time, and false generalizations were accepted only 33% of time. Universal statements were accepted 40% of the time. We replicated the GOG effect: minority characteristic universals were judged true 41% of the time. Participants agreed with minority characteristic universals 50% of the time when they were presented before the knowledge test, and this agreement rate dropped to 32% when presented after the knowledge test. Even when participants had recently rejected statements such as males ducks lay eggs, they still showed the overgeneralization effect by agreeing with statements such as all ducks lay eggs almost a third of the time. The effect was reduced but not eliminated. The proportion of agreement data were subjected to a 6 3 2 mixed measures ANOVA. As in Experiment 1, we found a main effect of predicate type, F1(5,190) = 104.02, p < .0001, F2(5, 65) = 60.75, p < .0001, min F0 (5, 146) = 38.35, p < .0001, reecting the higher agreement rate to quasi-denitional (86%), majority characteristic (88%), and minority characteristic statements (76%) than to majority (61%), striking (58%), and false generalization statements (46%). There was also a main effect of statement type, F1(2, 76) = 218.19, p < .0001, F2(2,130) = 367.03, p < .0001, min F0 (2, 160) = 136.84, p < .0001, reecting the higher agreement for some statements (91%) than for generic statements (77%), with the lowest agreement for all statements (40%), and a signicant interaction between predicate type and statement
type, F1(10, 380) = 59.89, p < .0001, F2(10, 130) = 50.75, p < .0001, min F0 (10, 356) = 27.47, p < .0001, again replicating the results in Experiment 1. Despite the reduction of the overgeneralization effect when universal statements were presented after the knowledge test, the ANOVA did not yield a signicant effect of task order. Finally, there was no signicant three-way interaction, F1(10, 380) = 1.71, p = .08, F2(10, 130) = 2.29, p < .05, min F0 (10, 441) = .98, p = .46. Fig. 2 displays performance for universal statements as a function of predicate type and task order. Planned comparisons showed that indeed, minority characteristic universals were accepted less often when evaluated after the knowledge test as compared to before (32% vs. 50%), t(38) = 2.05, p < .05, d = .65. Conversely, majority characteristic universals were accepted more often when evaluated after (90%) vs. before (70%) the knowledge test, t(38) = 2.95, p < .01, d = .93. We interpret this result as a sensitizing effect where the knowledge test, in combination with the subsequent evaluation task, alerted participants to the difference between minority characteristic and majority characteristic items. Majority characteristic universals may have been accepted more often to potentially compensate for the reduction in agreement to minority characteristic universals. No other simple effects comparisons between task orders were reliable. Planned comparisons between predicate types (collapsed over the two test orders) replicated previous ndings, and revealed that minority characteristic assertions were accepted in universal form more than false generalizations, t(78) = 6.31, p < .0001, d = 1.41, whereas neither majority t(78) = .95, p = .34, d = .21, nor striking assertions were accepted reliably more than false generalizations, t(78) = .48, p = .63, d = .11. Importantly, even when the knowledge phase was presented rst, minority characteristic predictions were accepted more frequently in universal form than false generalizations, t(38) = 3.26, p < .005, d = 1.03. As Table 3 shows, false gendered minority characteristic items were rejected most of the time. However, some items were accepted as much as 33% of the time, and so
Experiment 3
100
Quasi-definitional Majority characteristic Minority characteristic Majority Striking False generalization
20
40
60
80
Evaluation First
Evaluation Second
Fig. 2. Proportions of agreement to universals for Experiment 3 as a function of predicate type, separated by whether the evaluation task came rst (before the knowledge test) or second (after the knowledge test); 95% condence intervals shown.
26
we examined whether the GOG effect was reduced as a result of participants lack of knowledge about specic items. Five of the 11 remaining minority characteristic items had ignorance rates of less than 10% (with an average of 6%) i.e. the false gendered items (e.g. male ducks lay eggs) were accepted by fewer than 10% of participants. The other six items had ignorance rates greater than 10%, with an average of 25% of participants failing to reject the false gendered item. If the GOG effect had been driven by ignorance, then the effect would have been much greater for the second set of items, since participants were less likely to know about the gender differentiations. However, the items for which the ignorance rate was less than 10% yielded a substantial GOG effect (44% of trials), as did those with an ignorance rate greater than 10% (38% of trials). The GOG effects for the two subsets of items did not differ, t(78) = .70, p = .49, d = .16. Between the knowledge test and the evaluation of minority characteristic universals, there are four possible patterns of response that a given participant could provide for a given minority characteristic item: (1) A GOG effect response, in which the participant correctly rejected the false gendered minority characteristic assertion, but nonetheless accepted the corresponding minority characteristic universal (e.g. s/he rejected male ducks lay eggs but accepts all ducks lay eggs). (2) A correct response, in which the participant rejected both the false gendered minority characteristic assertion as well as the minority characteristic universal (e.g. s/he rejected both male ducks lay eggs and all ducks lay eggs). (3) An ignorant response, in which the participant accepted the false gendered minority characteristic assertion and also accepted the minority characteristic universal (e.g. s/he accepted both male ducks lay eggs and all ducks lay eggs). (4) A noise response, in which the participant accepted the false gendered minority characteristic but rejected the universal (e.g. s/he accepted male ducks lay eggs but rejected all ducks lay eggs). Table 4 shows the distribution of the four types of responses in Experiment 3 as a function of the task order. When the evaluation task came before the knowledge test, participants were equally likely to provide GOG responses (40% of trials) as they were to produce correct responses (44% of trials), t(38) = .49, p = .63, d = .15. Participants were signicantly more likely to produce GOG responses than ignorant responses (10% of trials), t(38) = 3.36, p < .005, d = 1.09, or noisy responses (6% of trials), t(38) = 4.53,
Table 4 Percentages of the four possible types of responses in Experiment 3 as a function of the task order. Task order Type of response GOG effect Evaluation rst Evaluation second 40 19 Correct 44 65 Ignorance 10 13 Noise 6 3
p < .0001, d = 1.43. Further, 17 out of the 20 participants produced more GOG responses than ignorant or noisy responses (binomial test, p < .0001, assuming a prior probability of 1/3). When the evaluation task came after the knowledge task, participants made more correct responses (65%) than GOG responses (19%), t(38) = 5.61, p < .0001, d = 1.77. These two types of responses (GOG vs. correct) were entered into a Chi-square test by task order, and the test revealed a signicant interaction, v2 = 11.52, p < .001, reecting a higher rate of correct responses when the evaluation phase was proceeded by the knowledge test. However, despite the reduced proportion of GOG responses when the evaluation task came second, participants still made more GOG responses than noisy responses (3% of trials), v2 = 11.63, p < .0001, suggesting that the GOG effect was not due to random responding, even when participants completed the knowledge test rst. Experiment 3 showed that the GOG effect is not entirely due to ignorance. Participants were generally aware of the relevant gender distinctions, which eliminates the possibility that overgeneralization effects arise simply out of ignorance. When the evaluation block came rst, participants produced GOG responses on 40% of trials that is, they accepted all ducks lay eggs despite knowing that male ducks do not lay eggs. Thus, the effect found in Experiment 1 was not due to ignorance. Participants also sometimes committed the error even when primed with information about the members of a kind who do not possess the predicate in question (male ducks and female lions for the examples above). Participants who rejected false gendered statements such as male ducks lay eggs nevertheless subsequently agreed with the minority characteristic universal statement all ducks lay eggs on 19% of trials. The reduced GOG effect reects that participants recalled the information from the rst half of the task. It is likely that participants were made more sensitive to whether information was gender specic, since the knowledge portion highlighted this. They may have proceeded more cautiously in the second half of the task, especially when it came to the minority characteristic statements. It is thus not surprising that the GOG effect was reduced; heightened sensitivity to gender-specic properties likely lessened the extent to which participants relied on their judgments of generics to evaluate universals, since they would have been more alert to the possibility of error. What is more remarkable is that the GOG effect persisted nonetheless on almost one fth of trials. This experiment thus suggests that the GOG effect is difcult to eliminate entirely. The overgeneralization effect persisted in Experiment 3 despite the availability of information that, in principle, should prevent the error. Would the effect still persist even when a correct alternative is presented to participants? People might have agreed with statements such as all ducks lay eggs because such statements were presented in isolation. Experiment 4 investigated whether participants would still agree with the above statement when they were explicitly provided with correct alternatives such as only some ducks lay eggs or some ducks dont lay eggs.
27
Experiment 4: the availability of correct alternatives Would people continue to show the overgeneralization error in spite of being presented with correct alternatives? Experiment 4 addressed this question by allowing participants to judge two competing statements on whether they agreed with one over the other. The pairs of statements differed only in the quantiers used to express them. Half the participants were asked to compare universals (e.g., all Xs are Ys) with afrmative existentials (e.g., only some Xs are Ys), and other half compared universals with negative existentials (e.g., some Xs are not Ys). Both formulations of the existential provided correct alternatives to the false universals, and so at least in the case of majority characteristic, minority characteristic, majority, striking, and false generalization predications, participants could choose a correct answer by selecting the existential. Participants who express a preference for universals over afrmative existentials (e.g., only some ducks lay eggs) or negative existentials (e.g., some ducks dont lay eggs) would show the overgeneralization effect. Given the results of our previous experiments, we expected participants to do so for majority characteristic and minority characteristic predications. Method Participants Forty volunteers from the same population as Experiments 1 and 2 participated over the Internet for pay. All were native English speakers and none had participated in experiments on generics in the past. Materials and design Experiment 4 used the same set of 72 items used in Experiments 1 and 2, examples of which are given in Table 1. On each trial, participants saw two versions of each item: a universal version and an existential version. Half the participants received an afrmative existential form (only some Xs are Ys) and the other half received a negative existential form (some Xs are not Ys). This generated a 6 (predicate type) 2 (existential polarity) design, where the polarity of the existential was a between-participants manipulation. Procedure Participants were asked to evaluate each pair of items over the Internet through an interface written in Ajax. The interface consisted of two locations on a horizontal plane, which were labeled A and B. The pairs of target stimuli appeared in these locations, and the order of their appearance was randomized such that half the time the universals appeared in slot A and half the time they appeared in slot B. Thus, the interface layout appeared as follows: All Xs are Ys A Only some Xs are Ys B
Results and discussion Responses on the Likert scale were coded so that a mean score of 6 indicates that participants agreed with universals over existentials, a score of 3.5 indicates no preference, and a score of 1 indicates that participants agreed with existentials over universals for that particular type of predication. Quasi-denitional statements were given a mean rating of 5.01, majority characteristic 4.61, minority characteristic 3.56, majority 1.94, striking 1.98, and false generalizations 1.74. The means of participants responses, separated by the polarity of the existential, are given in Table 5. These differences were assessed via a 6 2 mixed ANOVA, which yielded a signicant main effect of predication type, F1(5,190) = 141.64, p < .0001, F2(5, 65) = 62.50, p < .0001, min F0 (5, 127) = 43.36, p < .0001, but no main effect of polarity and no interaction. Planned comparisons showed that minority characteristic universals (M = 3.56) were preferred more often than majority (M = 1.94, t(78) = 8.6, p < .0001, d = 1.92), striking (M = 1.98, t(78) = 8.57, p < .0001, d = 1.92), and false generalization universal assertions (M = 1.74, t(78) = 10.18, p < .0001, d = 2.28). These data indicate a denite tendency to prefer minority characteristic universals to minority characteristic existentials, as compared to striking, majority, or false generalization universalexistential pairs. The tendency is a dramatic illustration of the GOG effect, as participants need only recognize that, for instance, the population of male ducks satises the existential statement only some ducks lay eggs and some ducks dont lay eggs and thus falsies the universal statement all ducks lay eggs. Use of a Likert scale in this experiment also allowed us to test how condent participants were in their judgments when they displayed a GOG effect. The previous studies used a forced-choice paradigm, so it is possible that a GOG rate of approximately 50% yes responses indicated that participants were uncertain and responding at chance. However, had they relied on their judgment of the generic to evaluate the universal, as per the generics-as-default hypothesis, they should have been condent in their answers. Although the mean preference for minority characteristic universals was 3.56 in this study, participants condence ratings were signicantly bimodal (Hartigans dip test, p < .001), i.e., their responses fell nearer to the end-points of the scale rather than the mid-point. In other words, participants who preferred the minority characteristic universal did so with condence. Experiment 4 demonstrated that even when correct
Table 5 Mean ratings of agreement (and standard deviations) to the universal (6) or the existential (1) formulations of assertions as a function of predication type and polarity of the existential in Experiment 4. Predication type Polarity of the existential statements Afrmative Quasi-denitional Majority characteristic Minority characteristic Majority Striking False generalizations 4.95 4.70 3.77 2.03 1.93 1.83 (1.92) (1.93) (2.23) (1.62) (1.54) (1.63) Negative 5.16 4.51 3.35 1.85 2.03 1.65 (1.61) (1.87) (2.07) (1.54) (1.59) (2.12)
Participants indicated whether they agreed with one of the items over the other by responding on a 6-point Likert scale ranging from 6 (I agree with A more) to 1 (I agree with B more).
28
alternatives were available at the time of choice, participants showed the GOG effect.
General discussion The generic overgeneralization (GOG) effect is a robust phenomenon. Participants agreed to statements such as all ducks lay eggs a substantial portion of the time (Experiment 1), despite knowing that male ducks do not lay eggs (Experiment 3). Participants endorsed these universals despite judging only minutes earlier that only half the kind has the property (Experiment 3). Participants also frequently failed to reject these false universal statements when they were explicitly provided with true alternatives such as some ducks do not lay eggs (Experiment 4). The effect is not likely to be attributable to participants employing a subtyping interpretation on which all ducks lay eggs would be understood to mean all kinds of ducks lay eggs, nor is it likely attributable to participants contextually restricting the scope of the universal quantier so that it only applies to, e.g., female ducks. Experiment 2a showed a GOG effect after participants were primed with population information (there are 431 million ducks in the world) designed to encourage an individual-based interpretation of the statements, and to encourage participants to understand the quantier to range over every single duck in the world (Stanley & Szabo, 2000). While the GOG effect occurred at a lower rate in this context, it still occurred on a substantial portion of trials. The lower rate may have been due to eliminating any subkind and restricted interpretations, or alternatively the numerical population information may have put participants in a more careful and analytic frame of mind (Alter et al., 2007). Further, Experiment 2b found very little evidence of subtyping interpretations or of domain-restricted interpretations when participants provided paraphrases of the statements they had evaluated. Importantly, there was no reliable correlation between subtyping or restricted paraphrases and acceptance of universal statements, which would be expected if the GOG effect were due to participants employing subtyping interpretations. The ndings reported here are instead better explained by the hypothesis that the GOG effect or at least a substantial portion of the observed effect was due to participants relying on their judgments of the corresponding (true) generic, and so failing to properly evaluate the universal in question. If this is so, then the effect that Hollander et al. (2002) and Tardif et al. (in press) identied in preschoolers persists into adulthood, as the generics-as-default hypothesis would predict. The hypothesis that generics express primitive, default generalizations explains a range of ndings including acquisition data, cross-linguistic syntactic patterns, and errors in reasoning tasks. For example, if generics express childrens default generalizations, then this would explain why generics are acquired early despite their semantic complexity (Carlson & Pelletier, 1995; Gelman, 2003; Gelman, 2010; Gelman & Brandone, 2010; Gelman & Raman, 2003; Graham et al., in press; Leslie, 2007; Leslie, 2008). It would also explain why generic forms are always less syntactically marked than quanticational ones (Chomsky, 2000; Dahl, 1985; Krifka et al., 1995; Leslie, 2008).
Importantly, the generics-as-default hypothesis also predicts that quantied statements will sometimes be incorrectly interpreted as generics. This prediction, if correct, would explain Jnsson and Hamptons (2006) inverse conjunction fallacy, Leslie and Gelmans (submitted for publication) nding that quantied statements are frequently recalled as generics, and Hollander et al.s (2002) and Tardif et al.s (in press) nding concerning preschoolers interpretations of all and some. The studies presented in this paper represent a further test of the predictions of the genericsas-default hypothesis. Consistent with the hypothesis, we found that adults show a robust tendency to evaluate some universal statements as though they were generics. Participants did so across a range of contexts, including one in which they were presented with population information concerning the kind (Experiment 2), one in which they were rst primed to consider the counterexamples (Experiment 3) and one in which they were presented with correct alternatives to the universal (Experiment 4). Interestingly, our ndings appear to be limited to predications that involve properties that are characteristic of the kind. Our focus in these experiments was on minority characteristic universals such as all ducks lay eggs, since these universals offered the clearest illustration of the GOG effect. However, participants also accepted majority characteristic universals such as all tigers have stripes, despite the fact that these predications are not strictly true in universal form. After all, there are stripeless tigers (as Siegfried and Roys performances attest). Our experiments did not control for whether or not people are aware that some tigers lack stripes, so we cannot be sure that these statements were not accepted out of simple ignorance. However, Jnsson and Hamptons (2006) studies suggests that something more interesting than ignorance may be driving these results. Forty-six percent of the time their participants judged that a more inclusive universal such as all ravens are black was more likely to be true than a narrower one such as all young jungle ravens are black even when the two statements were presented side by side for comparison (Jnsson & Hampton, 2006, Study 5). Either this constitutes a clear failure to appreciate the logical features of universal statements, or else their participants were interpreting these universals as generics (Jnsson, personal communication). It should be further noted here than nearly all of Jnsson and Hamptons items would count as majority characteristic predications according to our criteria, and so our participants willingness to accept majority characteristic universals could potentially be due to their interpreting them as generics also. Similarly, Leslie and Gelmans (submitted for publication) memory task in which adults often recalled universal statements as generics exclusively used items that would count as majority characteristic by our criteria. Jnsson and Hamptons results, together with Leslie and Gelmans, suggest that majority characteristic universals may be interpreted as generics, and our results suggest that this is true of minority characteristic universals as well. We did not nd that majority or striking universals were accepted, however in our studies, the GOG effect was limited to minority characteristic and possibly majority characteristic items. These ndings support the
29
hypothesis that the GOG effect is limited to cases where a characteristic property is predicated of the kind. Part of the explanation here may be that characteristic properties are usually held by all the normal members of a kind, and so GOG effects concerning such properties may often constitute relatively minor errors. Thus, adults could save cognitive effort by relying on the default generic interpretation to evaluate the universal, without risking signicant error. The error only becomes signicant in the case of these gender-specic properties. For non-characteristic properties, however, the risk of error is more prominent, and so adults may have learned not to rely on their judgments of generics for such items. One outstanding question is whether there are developmental differences in the extent of the GOG effect e.g. might young children show GOG effects even if the property is not characteristic? Hollander et al.s (2002) results suggest that the GOG effect may potentially be broader in scope for young children. However, they did not use a wide enough range of items to support denite conclusions. Future research is needed to address this question, and thereby shed more light on why adults only exhibit this effect when the property is characteristic. In sum, the generic overgeneralization effect the tendency to judge universally quantied statements to be true when the corresponding generic is true occurs for minority characteristic statements like all ducks lay eggs. These universally quantied statements were accepted by particiAppendix Materials for Experiments 1, 2, and 3 in generic form. Predication type Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Quasi-denitional Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Minority characteristic Minority characteristic Item (in generic form) Ants are insects Bachelors are unmarried Cats are animals Dogs are mammals Elms are trees Mushrooms are fungi Poodles are dogs Preschoolers cannot vote Rectangles are geometric gures Sows are pigs US Presidents are over 35 Vixens are foxes Bees have wings Cats have whiskers Cheetahs run fast Cows eat grass Dogs have tails Horses have four legs Kangaroos hop Leopards have spots Pencils are wooden Ravens are black Sparrows have wings Tigers have stripes Cardinals are reda Deer have antlers
pants about half the time, despite their knowing that only female ducks lay eggs. These ndings are plausibly due to participants having a tendency to treat universal statements as though they were generics, as one would predict on Leslies (2008) and Gelmans (2010) hypothesis that generics express primitive, default generalizations (see also Csibra & Gergely, 2009; Gelman & Brandone, 2010; Hollander et al., 2009). This explanation, if correct, has the potential to apply to a range of disparate phenomena, such as Hollander et al.s (2002) developmental ndings, Jnsson and Hamptons (2006) Inverse Conjunction Fallacy, and Leslie and Gelmans (submitted for publication) memory task. These diverse ndings may have a common underlying explanation: people have a tendency to treat universally quantied statements as though they were generics. This tendency is what one would expect if generics involve more basic generalizations, while universally quantied statements involve more sophisticated, non-default ones. Acknowledgments We are grateful to Paula Rubio Fernandez, Axit Fumero, Susan Gelman, Geoffrey Goodwin, Phil Johnson-Laird, Mark Johnston, Alan M. Leslie, Louis Lee, Ray Nickerson, Sandeep Prasada, and Carlos Santamaria for their helpful comments. These studies were supported in part by a NSF Graduate Research Fellowship awarded to the second author.
Predication type Majority Majority Majority Majority Majority Majority Majority Majority Majority Majority Majority Majority Striking Striking Striking Striking Striking Striking Striking Striking Striking Striking Striking Striking False generalization False generalization
Item (in generic form) Barns are red Bars are noisy Bikers have tattoos Cars have radios Clocks are round Italians eat spaghetti Jackets have zippers Pigeons sit on statues Pigs are kept in pens Shoes have laces Squirrels live in parks Subways are crowded Birds carry avian u Coyotes kill pets Insects carry viruses Lions eat people Mosquitoes carry malaria Pit bulls maul people Rats carry disease Rottweillers maul children Sharks attack swimmers Strokes cause paralysis Ticks carry Lyme disease Tigers attack people Animals are mammals Athletes are students (continued on next page)
30
Appendix (continued) Predication type Minority Minority Minority Minority Minority Minority Minority Minority Minority Minority
a
Item (in generic form) Ducks lay eggs Goats have horns Horses give live birth Insects lay eggs Kangaroos have pouches Lions have manes Moose have antlers Pigs suckle their young Sheep produce milk Snakes lay eggs
Predication type False False False False False False False False False False generalization generalization generalization generalization generalization generalization generalization generalization generalization generalization
Item (in generic form) Bees are worker bees Books are paperbacks Canadians are right-handed Computers are pcs Ducks are female Engineers are male Humans are over three years old Lions are male Teachers are female Trees are deciduous trees
characteristic characteristic characteristic characteristic characteristic characteristic characteristic characteristic characteristic characteristic
This item was dropped from the analyses reported in the paper (see Table 3).
References
Alter, A. L., Oppenheimer, D. M., Epley, N., & Eyre, R. N. (2007). Overcoming intuition: Metacognitive difculty activates analytic reasoning. Journal of Experimental Psychology: General, 136, 569576. Baldwin, D. A., Markman, E., & Melartin, R. (1993). Infants ability to draw inferences about nonobvious object properties: Evidence from exploratory play. Child Development, 64, 711728. Bott, L., & Noveck, I. A. (2004). Some utterances are underinformative: The onset and time course of scalar inferences. Journal of Memory and Language, 51, 437457. Carlson, G. (1977). Reference to kinds in English. PhD dissertation, University of Massachusetts, Amherst. Carlson, G., & Pelletier, F. J. (1995). The generic book. Chicago: University of Chicago Press. Chambers, C. G., Graham, S. A., & Turner, J. N. (2008). When hearsay trumps evidence: How generic language guides preschoolers inferences about unfamiliar things. Language and Cognitive Processes, 23, 749766. Chomsky, N. (2000). New horizons in the study of language and mind. Cambridge: Cambridge University Press. Cimpian, A., Brandone, A. C., & Gelman, S. A. (2010). Generic statements require little evidence for acceptance but have powerful implications. Cognitive Science, 34(8), 14521482. Cimpian, A., Gelman, S. A., & Brandone, A. C. (2010). Theory-based considerations inuence the interpretation of generic sentences. Language and Cognitive Processes, 25(2), 261276. Clark, H. H. (1973). The language-as-xed-effect fallacy: A critique of language statistics in psychological research. Journal of Verbal Learning and Verbal Behavior, 12, 335359. Cohen, A. (1996). Think generic: The meaning and use of generic sentences. PhD dissertation, Carnegie Mellon University. Connolly, A., Fodor, J., Gleitman, L. R., & Gleitman, H. (2007). Why stereotypes dont even make good defaults. Cognition, 103, 122. Csibra, G., & Gergely, G. (2009). Natural pedagogy. Trends in Cognitive Sciences, 13, 148153. Dahl, O. (1985). Tense and aspect systems. Oxford: Blackwell. Declerk, R. (1986). The manifold interpretations of generic sentences. Lingua, 68, 149188. Gelman, S. A. (2003). The essential child. New York: Oxford University Press. Gelman, S. A. (2010). Generics as a window onto young childrens concepts. In F. J. Pelletier (Ed.), Kinds, things, and stuff: The cognitive side of generics and mass terms. New directions in cognitive science (Vol. 12). New York: Oxford University Press. Gelman, S. A., & Bloom, P. (2007). Developmental changes in the understanding of generics. Cognition, 105(1), 166183. Gelman, S. A., & Brandone, A. C. (2010). Fast-mapping placeholders: Using words to talk about kinds. Language Learning and Development, 6, 223240. Gelman, S. A., Coley, J., Rosengren, K., Hartman, E., & Pappas, A. (1998). Beyond labeling: The role of parental input in the acquisition of richly structured categories (Vol. 63, No. 1). Monographs of the Society for Research in Child Development (SRCD), Series no 235. Ann Arbor, MI: SRCD. Gelman, S. A., Goetz, P. J., Sarnecka, B. S., & Flukes, J. (2008). Generic language in parentchild conversations. Language Learning and Development, 4, 131.
Gelman, S. A., Hollander, M., Star, J., & Heyman, G. D. (2000). The role of language in the construction of kinds. In D. Medin (Ed.). Psychology of learning and motivation (Vol. 39, pp. 201263). New York: Academic Press. Gelman, S. A., & Raman, L. (2003). Preschool children use linguistic form class and pragmatic cues to interpret generics. Child Development, 74, 308325. Gelman, S. A., Star, J. R., & Flukes, J. E. (2002). Childrens use of generics in inductive inferences. Journal of Cognition and Development, 3, 179199. Gelman, S. A., & Tardif, T. (1998). A cross-linguistic comparison of generic noun phrases in English and Mandarin. Cognition, 66, 215248. Geurts, B. (1985). Generics. Journal of Semantics, 4. Goldin-Meadow, S., Gelman, S. A., & Mylander, C. (2005). Expressing generic concepts without a language model. Cognition, 96, 109126. Graham, S. A., Nayer, S. L., & Gelman, S. A. (in press). Two-year-olds use the generic/non-generic distinction to guide their inferences about novel kinds. Child Development. Graham, S. A., Kilbreath, C. S., & Welder, A. N. (2001). Words and shape similarity guide 13-month-olds inferences about nonobvious object properties. In J. D. Moore & K. Stenning (Eds.), Proceedings of the twenty third annual conference of the Cognitive Science Society (pp. 352357). Hillsdale, NJ: Erlbaum. Hollander, M. A., Gelman, S. A., & Raman, L. (2009). Generic language and judgments about category membership: Can generics highlight properties as central? Language and Cognitive Processes, 24, 481505. Hollander, M. A., Gelman, S. A., & Star, J. (2002). Childrens interpretation of generic noun phrases. Developmental Psychology, 38, 6. Johnson-Laird, P. N., & Hasson, U. (2003). Counterexamples in sentential reasoning. Memory &Cognition, 31, 11051113. Jnsson, M. L., & Hampton, J. A. (2006). The inverse conjunction fallacy. Journal of Memory and Language, 55, 317334. Keates, J., & Graham, S. A. (2008). Category labels or attributes: Why do labels guide infants inductive inferences? Psychological Science, 19, 12871293. Khemlani, S., Leslie, S. J., & Glucksberg, S. (2009). Generics, prevalence, and default inferences. In N. Taatgen, H. van Rijn, J. Nerbonne, & L. Schomaker (Eds.), Proceedings of the 31st annual conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society. Khemlani, S., Leslie, S. J., Glucksberg, S., & Rubio-Fernandez, P. (2007). Do ducks lay eggs? How humans interpret generic assertions. In D. S. McNamara & J. G. Trafton (Eds.), Proceedings of the 29th annual conference of the Cognitive Science Society. Nashville, TN: Cognitive Science Society. Krifka, M., Pelletier, F. J., Carlson, G., ter Meulen, A., Chierchia, G., & Link, G. (1995). Genericity: An introduction. In G. Carlson & F. J. Pelletier (Eds.), The generic book (pp. 1125). Chicago: Chicago University Press. Lawler, J. (1973). Studies in English generics. University of Michigan Papers in Linguistics, 1. Leslie, S. J. (2007). Generics and the structure of the mind. Philosophical Perspectives, 21, 375405. Leslie, S. J. (2008). Generics: Cognition and acquisition. The Philosophical Review, 117(1), 149. Leslie, S. J. & Gelman, S. A. (submitted for publication). Quantied statements are recalled as generics: Evidence from preschool children and adults.
S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531 Leslie, S. J., Khemlani, S., Prasada, S., & Glucksberg, S. (2009). Conceptual and linguistic distinctions between singular and plural generics. Proceedings of the 31st annual cognitive science society. Amsterdam: Cognitive Science Society. Medin, D. L., & Ortony, A. (1989). Psychological essentialism. In S. Vosniadou & A. Ortony (Eds.), Similarity and analogical reasoning (pp. 179195). New York: Cambridge University Press. Nisbett, R., & Wilson, T. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231259. Osherson, D. N., Smith, E. E., Wilkie, O., Lopez, A., & Shar, E. (1990). Category-based induction. Psychological Review, 97, 185200. Papafragou, A., & Schwarz, N. (2005/2006). Most wanted. Language Acquisition, 13, 207251. Pappas, A., & Gelman, S. A. (1998). Generic noun phrases in motherchild conversations. Journal of Child Language, 25, 1933. Pelletier, F. J., & Asher, N. (1997). Generics and defaults. In J. van Benthem & A. ter Meulen (Eds.), Handbook of logic and language (pp. 11251179). Cambridge, MA: MIT Press.
31
Prasada, S., & Dillingham, E. (2006). Principled and statistical connections in common sense conception. Cognitive Science, 99(1), 73112. Prasada, S., & Dillingham, E. (2009). Representation of principled connections: A window onto the formal aspect of common sense perception. Cognition, 33(3), 401448. Rosch, E. (1978). Principles of categorization. In E. Rosch & B. B. Lloyd (Eds.), Cognition and categorization (pp. 2748). Hillsdale: Lawrence Erlbaum Associates. Sloman, S. A. (1993). Feature-based induction. Cognitive Psychology, 25, 231280. Sloman, S. A. (1998). Categorical inference is not a tree: The myth of inheritance hierarchies. Cognitive Psychology, 35, 133. Stanley, J. (2007). Language in context: Selected essays. Oxford: Oxford University Press. Stanley, J., & Szabo, Z. (2000). On quantier domain restriction. Mind and Language, 15(2 and 3), 219261. Tardif, T., Gelman, S. A., Fu, X., & Zhu, L. (in press). Acquisition of generic noun phrases in Chinese: Learning about lions without an -s. Journal of Child Language.
!"#$%&'()*(+,-(./0%"1(2/#13(4%+,(5-&-#%06 7(
."#",89"&-(:-61%-( ( ;-&)1+%<"+-(=#">+?(;1-"6-(0,-0@(A->/#-(0%+%&'(
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
1
Preparation of this article was supported by NSF grant BCS-1226942.
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
/(%&*+,&+#0"+-(',.6+)".6:3+I0.#+,-4+-(:"+-(',.6+'.#"1(),"-+.)"+.6-(+"--"&#,.6,F"*4+:".&,&1+#0.#+C"(C6"+ B"6,"D"+#0.#+:":B")-+(/+#0"-"+'.#"1(),"-+-0.)"+.+/%&*.:"&#.6+&.#%)"+#0.#+1)(%&*-+.+).&1"+(/+'(::(&+
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
2
An apparent exception to this generalization concerns groups such as homosexuals, and certain categories of mental illness, where believing in a natural/biological basis for group membership tends to correlate negatively with prejudice. That is, people who view homosexuality as a chosen lifestyle tend to exhibit more prejudice than people who understand it to be something biologically based and beyond the individuals control (Haslam & Levy, 2006). However, such examples are clearly complex, since anti-homosexual prejudice generally involves more than viewing it as simply a choice it involves viewing it as a morally and/or socially objectionable choice. That is, if one assumes that something is morally and/or socially objectionable, then believing it to be grounded in biology works to absolve the individual of responsibility, thereby, presumably, somewhat reducing ones prejudice or at least some of its expressions. The existence of these more complex patterns of prejudiced thinking, however, do not undermine the basic point here. 3 A more general question is why we essentialize any categories whatsoever, including animal categories. That is, why do we have tendency from such a young age to view some categories in this way? For discussion of this question see Gelman (2003), and Cimpian & Salomon (submitted).
!"#$%!%&'#(()*+,,'*-,,'*#-*.%/.()*,--,'!%#(%0,$1*,-2,3%#(()*+)*4,4+,"-*&5*!.,*622,"*3#-!,-* 78#.#(%'/.#41*9::;<*8#.#(%'/#4*=*>&$"%/6,01*9::?@A**B.,*C'3%,'!*D",,E-*#",*&5!,'*-#%$*!&*.#F,* +,(%,F,$*!.#!*!.,",*G,",*!G&*56'$#4,'!#(*E%'$-*&5*.64#'*+,%'/-H*D",,E-*#'$*I#"+#"%#'-1*,#3.* ,'$&G,$*G%!.*!.,%"*&G'*$%-!%'3!%F,*'#!6",-A*J,*4%/.!*-622&-,*!.#!*3(#--*%'*K'/(%-.*-&3%,!)*.#-1*#!*(,#-!* 6'!%(*F,")*",3,'!()1*+,,'*,--,'!%#(%0,$1*#'$*3,"!#%'()*4,$%,F#(*K6"&2,#'*'&!%&'-*&5*!.,*D",#!*L.#%'*&5* I,%'/*%'F&(F,$*.%/.()*,--,'!%#(%-!*G#)-*&5*!.%'E%'/*#+&6!*!.,*$%55,",'!*-!"#!#*&5*-&3%,!)A*7M&+(,-* -622&-,$*4#'%5,-!,$*#'*%'',"*-62,"%&"%!)1*!.,*3(,"/)*.#$*#*$%-!%'3!%F,*&'!&(&/%3#(*-!#!6-1*4&'#"3.-*G,",* %'56-,$*G%!.*#*$%F%',*"%/.!*!&*"6(,1*#'$*",+,((%&6-*2,#-#'!-*G,",*&5!,'*-#%$*!&*+,*/&%'/*#/#%'-!*!"#$%&' (#)&*+A@*>,3,'!*G&"E*.#-*#(-&*5&6'$*!.#!*3.%($",'*/"&G%'/*62*%'*4&",*2&(%!%3#(()*3&'-,"F#!%F,* 3&446'%!%,-*%'*!.,*N'%!,$*O!#!,-*.#F,*4&",*,--,'!%#(%-!*+,(%,5-*#+&6!*"#3,*!.#'*3.%($",'*/"&G%'/*62*%'* 4&",*2&(%!%3#(()*(%+,"#(*3&446'%!%,-*7>.&$,-*=*D,(4#'1*9::P@A*Q%E,G%-,1*G%!.%'*R-"#,(1*,--,'!%#(%-!*+,(%,5-* #+&6!*,!.'%3%!)*#",*4&",*3&44&'*%'*",(%/%&6-*3&446'%!%,-*!.#'*%'*-,36(#"*3&446'%!%,-*7S%,-,'$"63E*=* T#+,"1*9::P@A* * R!*G&6($*!.6-*-,,4*!.#!1*G.%(,*#'%4#(*E%'$-*#",*3&'-%-!,'!()*#'$*6'%5&"4()*,--,'!%#(%0,$*
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
!"#$%&%'%()*(+*,&&,)'%!-%&'*.,-%,+&*'/!'*0($-1*/!2,*+(34,1*!)50!56*3!'/,3*'/!)*0(37%)8*'(*%)&'%--*&,-,"'%2,* ,&&,)'%!-%&'*.,-%,+&*%)*'/,*+%3&'*9-!",:** * ;(*',&'*0/,'/,3*8,),3%"*-!)8$!8,*4%8/'*/!2,*1%3,"'*&,-,"'%2,*,++,"'&*%)*'/,*"!&,*(+*&("%!-*83($9&6*
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
"#$!$%&#'(&!)*+),-(%!(.-(!.%-$)*/!/%*%$),!'-*/#-/%!0!)*!%)(.%$!1-$%!2'#$-'!3$!)*+%4)*)(%!&)*/#'-$!
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
''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''
4
As noted above, to the extent that children understand kind-wide quantified statements, those statements could well have the same effect and for the same reason; indeed Cimpian and Erickson (2012) found similar effects when children were presented with facts quantified by most. However, as the authors note, this finding should again be interpreted in light of the fact that young children understand and remember kind-wide quantified statements as generics (Hollander et al., 2002; Leslie & Gelman, 2012; Tardif et al., 2010).
!"#$%&'($)&&)*#+%,"-!%)&.%)!%$-/%0#)12304#%#&&#1!)-5)&!%(#5)#2&%-(3'!%63$#1.%-17%2'0!"#0%43$$'1)4-!#% !"-!%!")&%&"-0#7%#&&#14#%4-'&#&%63$#1%!3%(#%&'($)&&)*#+%%% % ,")&%)&%13!.%32%43'0&#.%!3%&-/%!"-!%6#%!"#$%!$&'()*+,-))1!#080#!%963$#1%-0#%&'($)&&)*#:%)1%!")&%
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
"#0%4")57+%,"#%$#4"-1)&$%32%<#1#0)4%5-1<'-<#%)&%&3%&'(!5#%!"-!%!")&%8-0#1!%$)<"!%131#!"#5#&&%'!!#0% #13'<"%<#1#0)4&%!3%)1&!)55%&'4"%(#5)#2&+%H%8-0#1!%6"3%)&%43$$)!!#7%!3%!#-4")1<%"#0%4")57%!"-!%<#17#0% 7)22#0#14#&%-0#%13!%7##8%30%1-!'0-5%$)<"!%131#!"#5#&&%"-*#%#13'<"%&##$)1<5/%)1134'3'&%431*#0&-!)31&%
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
2*%%,'45&C,'+*&?*&!0*&(*!-,'4&)*0*&?,%)&D8/*!08#-(&+),-(0*'<&%)*&10#21*+%2&#;&!&1!0*'%&2$++**(,'4&!%& )*+(&,&-(#.+#..$',+!%,'4&*22*'%,!-,2%&"*-,*;2&@&,5*5<&"/&21*--,'4&%)*.&#$%&@&!0*&;!,0-/&-#?5&E;%*0&!--<&!'& *F1-,+,%&;#0.$-!%,#'&#;&*22*'%,!-,2.&.!/&'#%&"*&0*!(,-/&+#.10*)*'2,"-*&%#&+),-(0*'&!%&%),2&!4*G&H%)*0*I2& 2#.*%),'4&(**1&!'(&,')*0*'%&,'&J!01,*2&%)!%&.!:*2&%)*.&!--&!-,:*K5&L,',.!--/<&2$+)&2%!%*.*'%2&!0*& *;;*+%,=*-/&'*=*0&;#$'(&,'&1!0*'%8+),-(&+#'=*02!%,#'2<&*=*'&!"#$%&!',.!-&:,'(2&7M*-.!'<&NOOP95&M*'*0,+& -!'4$!4*&.!/&%)$2&10#=,(*&!&=*),+-*&?)*0*"/&"*-,*;2&!0*&+#..$',+!%*(&?,%)#$%&*=*0&"*,'4&*F1-,+,%-/& ;#0.$-!%*(&@&1*0)!12&*=*'&!%&!'&!4*&?)*'&-/).,"0$,&"1$(#.2,,)$$&!().,"0,)+-12(.')$"1',)$.2').0"-.)3)0. 2!().-".)*+(&,&-(#.)0-)'-2&0.-/).,"4410&,2-)5.!)(&)65&Q#?*=*0&,.1-,+,%&!'(&,'+)#!%*&%)*&*22*'%,!-,2%& "*-,*;2&.!/&"*<&%)*&2**.,'4-/&.$'(!'*&4*'*0,+&2*'%*'+*&,2&!"-*&%#&+#..$',+!%*&%)*.&>$,%*&*;;*+%,=*-/5& & !"#"$%&'()#*(+,#&"-.%,#'(,/(01%2%.%"'( R#'2,(*0&%)*&>$*2%,#'<&?)!%&(#*2&,%&%!:*&%#&"*&4##(&!%&!&1!0%,+$-!0&;!,0-/&(*.!'(,'4&!+%,=,%/&@&2!/&!'& ,'%*--*+%$!-&!+%,=,%/<&2$+)&!2&(#,'4&.!%)S&T'*&!'2?*0&.,4)%&"*&%)!%&,%&%!:*2&)!0(&?#0:<&1!%,*'+*<&!'(& (*(,+!%,#'5&U=*'&,;&#'*&,2&,',%,!--/&2%$.1*(&"/&.!%)*.!%,+2<&?,%)&*'#$4)&*'*04/&!'(&+#..,%.*'%<&#'*& +!'&+#.*&%#&*F+*-&!%&,%5&V*%&$2&+!--&2#.*#'*&?)#&4,=*2&2$+)&!'&!'2?*0&!'&&0,')4)0-2(.-/)"'&$-<&%)*0*"/& 0*;-*+%,'4&)*0&"*-,*;&%)!%&!",-,%/&,'&>$*2%,#'&+!'&"*&,'+0*.*'%!--/&!+>$,0*(&!'(&,.10#=*(&%)0#$4)&;#+$2<& (*(,+!%,#'&!'(&10!+%,+*5&A)*&+#'%0!2%,'4&!'2?*0&,2&%)!%&%#&"*&4##(&!&1!0%,+$-!0&;!,0-/&(*.!'(,'4&!+%,=,%/& ,%&%!:*2&'!%$0!-<&,')*0*'%&%!-*'%&@&2#.*%),'4&%)!%&,2&;,F*(&!'(&$'+)!'4*!"-*5&T'*&*,%)*0&H)!2&,%K&#0&#'*& (#*2'I%&@&.!%)*.!%,+!-&!",-,%/&,2&!&4,;%<&!'(&)!0(&?#0:&,'&%)*&!"2*'+*&#;&%),2&4,;%&?,--&"*&-!04*-/&;$%,-*5&W'& !++#0(&?,%)&%)*&12/+)#-#4,+!-&-,%*0!%$0*<&-*%&$2&+!--&2#.*#'*&?)#&"*-,*=*2&%),2&!'&)0-&-#.-/)"'&$-<&%#& 0*;-*+%&)*0&+#'+*1%,#'&#;&!",-,%/&!2&!&;,F*(<&2%!"-*&*'%,%/&@&2#.*%),'4&,')*0*'%&%)!%&,2&*,%)*0&1#22*22*(&#0& -!+:*(5&7A)*2*&%?#&!'2?*02&;#0.&!&21*+%0$.<&#;&+#$02*<&"$%&;#0&2,.1-,+,%/&#;&(,2+$22,#'<&W&?,--&21*!:&!2& %)#$4)&%)*/&;#0.&,'2%*!(&!&(,+)#%#./59&& & E&+#'2,(*0!"-*&"#(/&#;&*=,(*'+*&,'&12/+)#-#4/<&+#--*+%*(&"/&R!0#-&X?*+:&!'(&)*0&+#--*!4$*2<&
2$44*2%2&%)!%&,%&,2&>$,%*&(*%0,.*'%!-&%#&1*0;#0.!'+*&%#&"*&!'&*'%,%/&%)*#0,2%&72**&X?*+:<&YZZZ<&NOO[&;#0& 0*=,*?295&A)!%&,2<&!+0#22&.!'/&(,=*02*&*F1*0,.*'%!-&!'(&0*!-8?#0-(&+#'(,%,#'2<&,'+0*.*'%!-&%)*#0,2%2&
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
"#!$%&'(!)*!('%+,!%(!%+!%-.)/(0#(!()!1#23/+(0#2!4'0(!-%&'(!$302!5'%$2/3#!()!%#%(%0$$6!*)/-!0#!3#(%(6!
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

5
Interestingly, these effects were not found when children were asked to consider social groups about which they did not have pre-existing essentialist beliefs (boy/girls at a different school). However, the results of Rhodes et al. (2012) suggest that this could potentially be due to the fact that children only heard a small number of generics concerning this social group it is possible that, had children heard more extensive generic input about these boys and girls at a different school, they may have again come to think of the ascribed abilities in more entity-theoretic terms.
!"#$!%&'&(&)*$(+$*+,&!'$-.+/0*$1!2$')!#$,3&'#.)"$(+$!#+0($!$1+.)$)"(&(2$(3)+.)(&,$0).*0),(&4)$+"$(3)$ .)')4!"($!%&'&(&)*5$ $ 6"$)4)"$1+.)$#.!1!(&,$&''/*(.!(&+"$+7$(3)$0+8).$+7$-)").&,$'!"-/!-)$,+",)."&"-$!%&'&(&)*$&*$
7+/"#$&"$!$4).2$.),)"($*(/#29$!-!&"$%2$:&10&!"$!"#$3&*$,+'')!-/)*$;:&10&!"9$<&!"9$=$>/(3).'!"#9$ */%1&(()#?5$@"$(3&*$*(/#29$0.)*,3++').*$8).)$!*A)#$!$*).&)*$+7$B/)*(&+"*$!%+/($7!1&'&!.$!%&'&(&)*9$ B/)*(&+"*$(3!($8).)$)&(3).$&"$-)").&,$+.$&"#&4&#/!'$7+.15$@"$0!.(&,/'!.9$(3)2$8).)$)&(3).$!*A)#$B/)*(&+"*$ */,3$!*$C!.)$-&.'*$-++#$!($#.!8&"-DE9$+.$B/)*(&+"*$!%+/($!$0!.(&,/'!.$7.&)"#$+7$(3)&.*9$)5-5$C&*$F!""!3$ -++#$!($#.!8&"-DE5$G3/*9$,3&'#.)"$8).)$")4).$!,(/!''2$!"#$$!"2(3&"-$!%+/($-.+/0$+.$&"#&4&#/!'$!%&'&(&)*H$ (3)2$8).)$*&10'2$!*A)#$(+$,+"*&#).$(3)*)$B/)*(&+"*5$:3&'#.)"$&"$%+(3$,+"#&(&+"*$;&5)5$-)").&,$4*5$ &"#&4&#/!'?$8).)$(3)"$(+'#$(3!($!$0!.(&,/'!.$&"#&4&#/!'$&*$!$-++#$!($!$%"&'#(!,(&4&(2$;)5-5$C(3&*$-&.'$&*$-++#$!($ !$-!1)$,!'')#$C-+.0E?5$G3)$,3&'#.)"$8).)$!*A)#$B/)*(&+"*$(3!($)I!1&")#$83)(3).$(3)2$(3+/-3($(3!($(3&*$ &"#&4&#/!'$"))#)#$(+$0.!,(&,)$(+$%)$-++#$!($-+.09$+.$83)(3).$(3)2$$(3+/-3$3)J*3)$8!*$CK/*($-++#E$!($&($L$ (3!($&*9$(3)$)I()"($(+$83&,3$(3)2$3)'#$)"(&(2$4*5$&",.)1)"(!'$,+",)0(&+"*$+7$83!($&($(!A)*$(+$)I,)'$!($(3&*$ /"7!1&'&!.$-!1)5$>(.&A&"-'29$,3&'#.)"$83+$3!#$0.)4&+/*'2$!"*8).)#$B/)*(&+"*$,+",)."&"-$(3)$!%&'&(&)*$+7$ -.+/0*9$.!(3).$(3!"$&"#&4&#/!'*9$0'!,)#$1/,3$')**$)103!*&*$+"$0.!,(&,)9$!"#$()"#)#$&"*()!#$(+$*/00+*)$ (3)$&"#&4&#/!'$3!#$!"$)77+.(M&"#)0)"#)"($-&7(5$G3/*9$K/*($*&10'2$)"%*+$',+%-$-)").&,$B/)*(&+"*$!%+/($ ./0+#+/,$!%&'&(&)*$')#$,3&'#.)"$(+$!#+0($)"(&(2M(3)+.)(&,$,+",)0(&+"*$+7$!$"+4)'$!%&'&(25$ $ 1!','"!23'(45,'/!$ >/00+*)$2+/$!.)$!$1)1%).$+7$!$-.+/0$(3!($&*$!**+,&!()#$(3.+/-3$,/'(/.!'$*().)+(20)*$8&(3$*/%0!.$ 0).7+.1!",)$!($!$0!.(&,/'!.$!,(&4&(2$L$&"$(3)$8!2$(3!(9$*!29$8+1)"$!.)$+7()"$*().)+(20)#$!*$%)&"-$')**$ -++#$!($1!(3$(3!"$1)"5$6$.+%/*($!"#$)I()"*&4)'2$#+,/1)"()#$03)"+1)"+"9$A"+8"$!*$*!','"!23'( !5,'/!9$&*$(3!($83).)$(3).)$&*$,+11+"$A"+8')#-)$+7$(3)$*().)+(20)9$+")N*$0).7+.1!",)$!($(3!($!,(&4&(2$ 8&''$%)$#)-.!#)#$&7$+")N*$1)1%).*3&0$&"$(3&*$-.+/0$&*$1!#)$*!'&)"(9$)4)"$&"$!$*/%(')$8!25$O+.$)I!10')9$&7$ 8+1)"$!.)$!*A)#$(+$&#)"(&72$(3)&.$-)"#).$0.&+.$(+$(!A&"-$!$1!(3$)I!1$;!*$3!*$(.!#&(&+"!''2$%))"$(3)$,!*)$ 8&(3$(3)$>6G*$!"#$PQR*9$7+.$)I!10')?9$(3)2$8&''$0).7+.1$')**$8)''$+"$(3)$)I!1$(3!"$(3)2$8+/'#$3!4)$3!#$ -)"#).$1)1%).*3&0$"+($%))"$1!#)$*!'&)"($;)5-59$61%!#29$>3&39$S&19$=$T&((&"*A29$UVVWH$6.+"*+"9$O.&)#9$ =$P++#9$UVVUH$P++#9$6.+"*+"9$=$@"X'&,3(9$UVVYH$>())')$=$6.+"*+"9$WZZ[H$>())')9$UVWVH$!"#$1!"2$+(3).*?5$ \)1%).*3&0$&"$(3)$*().)+(20)#$-.+/0$,!"$%)$1!#)$*!'&)"($&"$(3)$1+*($*/%(')$+7$8!2*$L$,3),A&"-$!$%+I$ +"$!$7+.19$+.$)4)"$K/*($%)&"-$&"$!"$)"4&.+"1)"($8&(3$*/%(')$,/)*$(3!($1!A)$-)"#).$+.$+(3).$-.+/0$ 1)1%).*3&0$*!'&)"($;)5-59$(3).)$%)&"-$!$0+*().$&"$(3)$.++1$(3!($*!2*$C\!(3]$P+($^+1)"DE?5$ _"#).*(!"#&"-$*().)+(20)$(3.)!($&*$&10+.(!"($7+.$!$.!"-)$+7$.)!*+"*$L$7+.$)I!10')9$&"$1!"2$!,!#)1&,$
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
!-%/#"%3"0"')(%*103/13"5%&1'!)(/*1'*+%)0%(-02"'#1!)-0%;)!,%(,)*$'"09%B"'!1)0*+5%-0"%#,-/*$%#!')2"%!-% 4-0)!-'%-0"=#%/#"%-<%3"0"')(#5%1!%*"1#!%!,-#"%3-2"'0)03%#-()1*%3'-/&#9%?,"%$)<<)(/*!+%-<%!'+)03%!-%"*)4)01!"% !,"4%<'-4%-/'%'"&"'!-)'"%1*!-3"!,"'%#,-/*$%0-!%7"%/0$"'"#!)41!"$5%,-;"2"'9%U#%0-!"$%1!%!,"%7"3)00)03% -<%!,)#%&1&"'5%)!%#""4#%!,1!%3"0"')(#%1'"%-/'%;1+%-<%1'!)(/*1!)03%-/'%4-#!%71#)(%3"0"'1*)V1!)-0#5%10$%#-%)0% !,)#%;1+%!,"+%'"<*"(!%#-4"!,)03%2"'+%</0$14"0!1*%)0%-/'%-/!*--A9%?-%!'+%!-%"@()#"%!,"4%<'-4%-/'% (-02"'#1!)-0#%;-/*$%!,/#%7"5%1$4)!!"$*+5%1%$1/0!)03%!1#A9%%
"#!$%&'!$%((')*!+#)*',!-(!-*!.#(!%*!(/#01/!#.'!2#034!*-$536!*+-(2/!(#!0*-.1!70%.(-8-2%(-#.%3!
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`Û9^J*( Q>$3%?3='(!*'(P-"@-$3'(!*'(7=4C-='(+*(N*'(0(R=C"$3'(+*(!*(89:;9<*(W?(C-?34(,$G=("$3=4a(O?>(B,-C%>=3'( T=3=>-B4($>=($#?A2(5-3%4(3?2(HA$32-2-=4*(=%)'>+?-@-'&78-.9'(<;D0E'(`9JU`JJ*( Q>?V3'(S*(1*'(0(N?4=@,4'(S*(!*(8;MMM<*(!(#A>%=3(?F(@>??FD(+2=>=?2&@=(>=C=G$3B=($3%(T=3%=>( %-FF=>=3B=4(-3("$2,(@=>F?>"$3B=*(1&23.*'+&4+!-3"&.*')9#+*.>+,&$)*'+!"#$%&'&(#:+CA'(9`^L9I\*( P$>C4?3'(R*(8;M\\<*(F-4-3-.$-+9&+G).>"+).+5.(')"%*(1,*W*(%-44=>2$2-?3'(b3-G=>4-2&(?F(.$44$B,A4=224'( !",=>42*( P$>C4?3'(R*)*'(0(1=CC=2-=>'(O*N*(8;MMI<*(H%-+I-.-3)$+J&&K*(P,-B$T?D(P,-B$T?(b3-G=>4-2&(1>=44*( P,=>&$3'(+*'(0(Q?%=3,$A4=3'(R*(c*(89:::<*(d,=3(@?4-2-G=(42=>=?2&@=4(2,>=$2=3(-32=CC=B2A$C(@=>F?>"$3B=D( 6,=(@4&B,?C?T-B$C(,$_$>%4(?F(e"?%=C("-3?>-2&f(42$2A4*(!"#$%&'&()$*'+,$)-.$-'(//8I<'(JMML`:9*( P-"@-$3'(!*(89:;:<*(6,=(-"@$B2(?F(T=3=>-B(C$3TA$T=($#?A2($#-C-2&(?3(B,-C%>=3g4($B,-=G="=32("?2-G$2-?3*( ?-@-'&78-.9*'+!"#$%&'&(#:+BA8I<'(;JJJU;J`:*( P-"@-$3'(!*'(!>B='(]*(P*'(.$>5"$3'(E*(.*'(0(WV=B5'(P*(+*(89::\<*(+A#2C=(C-3TA-42-B(BA=4($FF=B2(B,-C%>=3g4( "?2-G$2-?3*(!"#$%&'&()$*'+,$)-.$-'+/<8`<'(J;ÙJ;^*( P-"@-$3'(!*'(Q-$3'(7*'(0(+A2,=>C$3%'(+*(84A#"-22=%<*(6,=(?>-T-34(?F(B,-C%>=3Y4(#=C-=F4($#?A2($B,-=G="=32D( 6,-35-3T($#?A2(2,=($#-C-2-=4(?F(T>?A@4(B$A4=4(`L&=$>L?C%4(2?(%=G$CA=(=FF?>2*( P-"@-$3'(!*'(0(E>-B54?3'(7*(P*(89:;9<*(6,=(=FF=B2(?F(T=3=>-B(42$2="=324(?3(B,-C%>=3Y4(B$A4$C($22>-#A2-?34D( hA=42-?34(?F("=B,$3-4"*(?-@-'&78-.9*'+!"#$%&'&(#:+B<8;<'(;IMU;\:*( P-"@-$3'(!*'(0(.$>5"$3'(E*(.*(89::M<*(X3F?>"$2-?3(C=$>3=%(F>?"(T=3=>-B(C$3TA$T=(#=B?"=4(B=32>$C(2?( B,-C%>=3Y4(#-?C?T-B$C(B?3B=@24D(EG-%=3B=(F>?"(2,=->(?@=3L=3%=%(=[@C$3$2-?34*(=&(.)9)&.:+//;D/E:(;Ù9I*(
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`*'(Q%-&79.%'(a*(Q*'(+(b@A.7='(`*(0122c4(W7&7-"9(@%&8A%87("&($%-7&>Y96"@<( 9;&L7-=%>";&=*((D9-8E98(&D(9?-#-8&9-%&'()($*+,(-./&I/(3YP3*( W7@#%&'(Q*()*'(J%-7'(/*()*'(+(d@7"&C7-8(b*(012324*(/??79>=(;?(87&7-"9(@%&8A%87(;&(9%>78;-M(9;&>7&>(%&<( =>-A9>A-7*(!*8-#.#)(&4567"*$*86'(JF2;3/(1EPYP23( W;;<'(!*'()-;&=;&'(`*'(+(e&_@"96>'(,*(0122P4*(e#$-;L"&8(%<;@7=97&>=N(=>%&<%-<"_7<(>7=>($7-?;-#%&97R()&( "&>7-L7&>";&(>;(-7<A97(>67(7??79>=(;?(=>7-7;>M$7(>6-7%>*(K++$#(%&'()($*+,(-.9$&4567"*$*86/&1I'(SDOFSS1*( W-77&C7-8'(H*'(0122P4*(A9-#C(5.9.#*-5&*C&8(-(?#7#.6*(X;A>@7<87*(\7B(H;-.*( \*(Z%=@%#'(X;>6=96"@<'(I*'(+(/-&=>'(U*'(012224*(/==7&>"%@"=>(C7@"7?=(%C;A>(=;9"%@(9%>78;-"7=*(L?#.#5"&M*E?-9$& *C&:*7#9$&4567"*$*86/&PGR(33PY31Ef(( \*(Z%=@%#'(X;>6=96"@<'(I*'(+(/-&=>'(U*'(012214*()-7(7==7&>"%@"=>(C7@"7?=(%==;9"%>7<(B">6($-7gA<"97K(L?#.#5"& M*E?-9$&*C&:*7#9$&4567"*$*86'&IF/(cEF322( Z%=@%#'(\*'(+(I7LM'(Q*(X*(0122S4*(/==7&>"%@"=>(C7@"7?=(%C;A>(6;#;=7Â%@">MR(Q>-A9>A-7(%&<("#$@"9%>";&=(?;-( $-7gA<"97*(4(?5*-9$#.6&9-%&:*7#9$(4567"*$*86&LE$$(.#-/&;1/&DE3FDcO*(
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`674G')*"%7*F'%'("$4]"549%,*V3.(L9D:0$8(9=( I3&8919#3",* M'#$4')*+,*R,)*QF'$H"%)*+,*B,*-./C.1,*a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`&#E)@&DD%F%#2%1)O%3T%%#)TB'%#)"#@)'%#+) 789;<.-.$);"-':;)&#;&6'426)><LR>IM+) 7(&#1);&()U+)/+()8)I)--&(()U+)^+);>JJ2?+)9172$B-BE&2"-)%11%#3&"-&1')BD)$6'"#)2"3%EBF&%1+)B%((&#1'+)(&;1).#8' )#'789;<.-.$);"-':;)&#;&6'4K3L56'<=<H<=_+) :$B@%1()!+()8),%-'"#().+)/+);<==L?+)/)@%A%-BC'%#3"-)%P"'&#"3&B#)BD)3$%)2B#2%C36"-)13F6236F%)BD)"#&'"-() "F3&D"23()"#@)$6'"#)1B2&"-)2"3%EBF&%1)"2FB11)3TB)26-36F"-)2B#3%P3+)B.$#)1),&'789;<.-.$96'EM6'<QQH<SQ+) :$B@%1()!+()J%1-&%().+X+()8)^TBF%2K()4+);<=><?+)46-36F"-)3F"#1'&11&B#)BD)1B2&"-)%11%#3&"-&1'+)7(.;&&*)#$8'.?' 1<&'N"1).#"-'A;"*&09'.?':;)&#;&8'37NA:56'4JM()>IM<_H>IMI>+) :B3$O"F3()!+()8)^"7-BF()!+);>LL<?+)4"3%EBF7)-"O%-1)"#@)1B2&"-)F%"-&37V)UB)T%)A&%T)1B2&"-) 2"3%EBF&%1)"1)#"36F"-)K&#@1a)W#),+).%'&#)8)b+)N&%@-%F);0@1+?()!"#$%"$&6')#1&(";1).#'"#*'8.;)"-' ;.$#)1).#';CC+)>>HI_?+)JB#@B#V)."E%+) ."6-()X+);&#)CF%11?+)cY#2B#12&B61)W#D-6%#2%1)"#@)dB'%#)&#)9$&-B1BC$7!)W#)N&B#")X%#K&#1)"#@) b"3F&#")5632$&1B#);%@1+?)O.0&#')#'7<)-.8./<9P'O<"1'N&&*8'1.'B<"#$&Q'' .$&$()!+()/'O"@7()[+():&2$%1B#()X+)/+()N6e&3"()b+()8),F"7()5+)!+);<==<?+).3%F%B37C%)C%FDBF'"#2%) OBB131V)^$%)&'C"23)BD)1%-DHF%-%A"#2%)"#@)3$%)'"##%F)BD)13%F%B37C%)"23&A"3&B#+)D.%(#"-'.?' 7&(8.#"-)19'"#*':.;)"-'789;<.-.$96'R@;I?()_IZH_QS+)
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`#C"B[#&#99#6"B@$B9F&@6A& #6"B"@"B[B"=&B6&"?#&E23F@"B26&2E&9"#3#2"=>#9(&;12$0',&#%6),'-)<03#,&);$63"0&0(6)G1+#15=)A%&-a-H -<<(& & & &
1 Running head: GENERICS AND CONCEPTS
Conceptual Distinctions Amongst Generics Sandeep Prasada, Sangeet Khemlani, Sarah-Jane Leslie, Sam Glucksberg
Please address correspondence to: Sandeep Prasada Department of Psychology Hunter College, City Universty of New York New York, NY 10065 Email: sprasada@hunter.cuny.edu Fax: 212 772 5677 Keywords: generics; concepts; generalizations; semantics Word count (from Abstract to References): 10,759 words
Acknowledgments This work was supported in part by a NSF Graduate Research Fellowship awarded to the second author, and NSF grant BCS-1226942 awarded to the third author. We are grateful to Jeremy Boyd, Adele Goldberg, Phil Johnson-Laird and Mark Johnston for their many helpful discussions and suggestions.
GENERICS AND CONCEPTS
Abstract Generic sentences (e.g., bare plural sentences such as dogs have four legs and mosquitoes carry malaria) are used to talk about kinds of things. Three experiments investigated the conceptual foundations of generics as well as claims within the forrmal semantic approaches to generics concerning the roles of prevalence, cue validity and normalcy in licensing generics. Two classes of generic sentences that pose challenges to both the conceptually based and formal semantic approaches to generics were investigated. Striking property generics (e.g. sharks bite swimmers) are true even though only a tiny minority of instances have the property and thus pose obvious problems for quantificational approaches, and they also do not seem to characterize kinds in terms of the principled or statistical connections investigated in previous research (Prasada & Dillingham, 2006, 2009). The second class minority characteristic generics (e.g. ducks lay eggs) also poses serious problems for quantificational accounts, and appears to involve principled connections even though fewer than half of its instances have the relevant property. The experiments revealed three principal discoveries: first, striking generics involve neither pincipled nor statistical connections. Instead, they involve a causal connection between a kind and a property. Second, minority characteristic generics exhibit the characteristics of principled connections, which suggests that principled connections license the expectation that most instances will have the property, but do not require it. Finally, the experiments also provided evidence that prevalence and the acceptability of generics may be dissociated and provided data that are problematic for normalcy approaches to generics, and for the idea that cue validity licenses low prevalence generics. As such, the studies provided evidence in favor of a conceptually based approach to the semantics of generics (Leslie, 2007, 2008; see also Carlson, 2010).
Conceptual Distinctions Amongst Generics Much of our conceptual knowledge consists of generic knowledge -- knowledge about kinds of things and their properties -- and is expressed via generic sentences such as tigers are striped, ravens are black, and cars have radios. The seemingly mundane nature of such sentences and the knowledge they express disguise the serious challenges generic sentences and generic knowledge pose. Despite significant efforts and some progress, linguists and philosophers have been unable to develop an adequate formal semantic analysis of generic sentences (Krifka et al, 1995; Leslie, 2008). Furthermore, though knowledge about kinds of things is at the core of our conceptual knowledge, research on concept acquisition and representation has generally proceeded without investigating the relationship between kinds and their characteristics that is revealed by our understanding of generic sentences. As a consequence, current theories of conceptual representation do not provide an account of how the knowledge expressed in generic sentences is represented (Prasada, 2012). Recently, however, the idea that generic sentences provide a window into our conceptual system has gained considerable traction (Carlson, 2010; Cimpian & Markman, 2009; Gelman, 2003, 2004; Gelman & Bloom, 2007; Hampton, 2010; Leslie, 2007, 2008, 2012; Pelletier, 2010; Prasada, 1999, 2000, 2010, 2012; Prasada & Dillingham, 2006, 2009). Specifically, generic sentences reveal the distinct ways in which our conceptual systems can represent connections between kinds and properties. On this view, which may be dubbed the conceptually based approach to generics, generic sentences give voice to our cognitive systems default modes of conceiving of kinds (Leslie, 2007, 2008, 2012), so that an adequate account of the meanings of generic
sentences requires an understanding of the ways in which we represent connections between kinds and properties. The conceptually based approach contrasts with the formal, quantificational approaches to generics that have dominated the semantics literature. These formal approaches seek to analyze generics using the set-theoretic tools of formal semantics, and usually involve positing some kind of covert quantification for example, generics may be analyzed as containing an implicit quantifier (e.g. all), which is often taken to range over the normal instances of the kind. Thus tigers are striped would be analyzed as meaning something like all normal tigers have stripes (see Carlson & Pelletier, 1995, for examples of such accounts, especially in the introductory chapter). Such accounts aim to reduce generics to set-theoretic, quantificational claims. The conceptually based approach, in contrast, aims to understand generics in the context of the various ways in which we represent connections between kinds and properties. The primary goal of the experiments in the present paper is to investigate the types of connections our conceptual systems represent between kinds and properties. In so doing, they reveal domain-general ways in which our concepts of kinds are structured (Prasada & Dillingham, 2009). In addition, they provide a characterization of the cognitively fundamental modes of generalization that are hypothesized to underlie generic sentences . (Leslie, 2007, 2008, 2012; see also Carlson, 2010). The experiments also address a number of claims that have been made concerning the semantics of generics within the linguistics literature.
Representing generic knowledge via principled and statistical connections
Our conceptual systems distinguish at least two types of connections, principled and statistical, between kinds and their properties, and thus provide at least two ways to characterize kinds and represent generic knowledge (Prasada & Dillingham, 2006; 2009). Principled connections involve properties that instances of a kind have by virtue of their being the kinds of things they are (e.g., having four legs for a dog) and they involve three dimensions: an explanatory dimension, a normative dimension, and a statistical dimension. As such, principled connections support formal explanations whereby a property of an instance of a kind is accounted for by reference to the kind of thing it is (e.g., Fido has four legs because he is a dog). They also license normative expectations: we expect that instances of the kind should have the properties to which the kind has a principled connection (e.g. Fido, by virtue of being a dog, should have four legs). Instances that lack the property are judged to be defective or incomplete rather than merely atypical. Finally, principled connections ground the expectation that the property will generally be highly prevalent (e.g., most dogs are expected to have four legs). Properties that have a principled connection to a kind are represented as aspects of being that kind of thing (Prasada & Dillingham, 2009). Generic knowledge that involves principled connections may be expressed in generic sentences that either have the bare plural form (e.g., dogs have four legs) or the indefinite singular form (e.g., a dog has four legs) in English (Leslie, Khemlani, Prasada & Glucksberg, 2009; Prasada & Dillingham, 2009). Statistical connections, on the other hand, involve properties that are simply prevalent amongst instances of a kind (e.g., being red for a barn). Statistical connections do not support formal explanations or normative expectations. Furthermore, properties
that have a mere statistical connection to a kind are not represented as aspects of the relevant kind of thing. Generic knowledge involving statistical connections may be expressed via generic sentences that have the bare plural form (e.g., barns are red), but not the indefinite singular form (e.g., #a barn is red) in English (Leslie et al., 2009; Prasada & Dillingham, 2009). Thus, principled and statistical connections provide two different means for characterizing kinds, and reflect at least two different ways in which our generic knowledge can be represented. Both principled and statistical connections provide a means for characterizing kinds generally and are available for characterizing kinds irrespective of content domains. In particular, Prasada & Dillingham (2006, 2009) found that the distinction between principled and statistical connections applies to natural, artificial and social kinds. Furthermore, there is an asymmetry between the two connections: when there is a principled connection between a kind and a property, there is also a statistical connection between the two. However, the presence of even a strong statistical connection does not in itself implicate a principled connection between the kind and the property (Prasada & Dillingham, 2006).
Are principled and statistical connections sufficient for characterizing our generic knowledge? A natural question is whether principled and statistical connections are sufficient for characterizing all of our generic knowledge. There are two classes of generic knowledge that seem to involve connections that are neither principled nor statistical, at least as currently conceived. The first class consists of generalizations such as pit-bulls
maul children or sharks attack swimmers. i In these cases, it appears that the property can hold for just a very small percentage of the instances of the kind and yet the generic will be licensed (Cimpian, Brandone, & Gelman, 2010; Leslie, 2007, 2008, in press). Such low-prevalence generics tend to involve striking (good or bad) properties and thus have been dubbed striking property generics (Leslie, 2007, 2008, in press). The existence of striking property generics shows that high prevalence is not necessary for generic beliefs. Not only does it seem that striking property generics can be true even when a small percentage of instances have the property in question, but they also do not seem to involve an expectation that most instances would or should have the property. As such, striking property generics seem to involve a type of connection that is neither principled nor statistical. Leslie (2007, 2008, in press) and Prasada (2010) suggest that striking property generics involve causal connections between the kind and the property, in the sense that the shared nature of the members of the kind causes them to be disposed to have the property in question, whether or not they actually do. For example, even though most ticks do not carry Lyme disease, their common biological makeup causes them to be susceptible to carrying the disease. According to Leslie and Prasada, such a causal connection between the kind and the property is required for a striking property generic to be accepted. Furthermore, though principled connections and causal connections often co-occur, they can also be decoupled (Prasada & Dillingham, 2006, 2009). For example, though there is a principled connection between being a triangle and being three-sided, there is no causal connection, since triangles are abstract entities. Conversely, though there is likely a causal connection between being a deer tick and carrying Lyme disease,
there is no principled connection between the two because carrying Lyme disease is unlikely to be thought of as an aspect of being a deer tick (see Experiment 1 below). While the conceptually based approach can readily accommodate striking property generics, they have proved intractable for the standard formal semantic approaches. Previous discussions of striking property generics have almost exclusively relied on theorists intuitions rather that on data from nave participants (but cf. Cimpian, Brandone & Gelman, 2010). Since striking property generics provide an important piece of evidence in adjudicating between different theories of generics, Experiment 1 sought to confirm that people do indeed accept them despite knowing that very few members of the kind have the property. Experiment 3 investigated whether striking property generics involve a causal connection between the kind and the property. A second class of generics that does not straightforwardly involve principled nor statistical connections concerns what Leslie (2007, 2008) calls minority characteristic generics such as ducks lay eggs and lions have manes. Minority characteristic generics are similar to generics that involve principled connections in that they involve properties that seem to be part of the nature of the kind. However, they differ from principled generics in that they involve properties that are possessed by less than half of the instances of the kind (e.g., only fertile, mature female ducks lay eggs). As such, they may differ from both principled and statistical generics and potentially constitute a fourth type of generic. Alternatively, minority characteristic generics may involve principled connections, since principled connections ground only the expectation that most or all instances of the kind will have the property, but do not require this to be the case. Prasada & Dillinghams (2006, 2009) stimuli only included items for which most or all
instances had the critical property. Thus, studies have not investigated whether principled connections are implicated when only a minority of the instances have the property. Experiments 1 and 2 investigated whether minority characteristic generics display the non-statistical characteristics of principled connections. The experiments also investigated whether the connections that underlie striking generics differ from principled connections with respect to their explanatory and normative dimensions.
Formal semantic approaches to generics While the primary goal of our studies was to examine the different ways in which we represent generic knowledge, we also addressed a number of claims in the formal semantics literature (e.g., Carlson, 1977; Carlson & Pelletier, 1995; Cohen, 1996; Lawler, 1973; Leslie, 2007, 2008; Pelletier & Asher, 1997). One claim is that some generics are deemed true even though very few members of the kind in question have the property (e.g., Leslies 2007, 2008, in press notion of striking property generics), and a complementary claim is that some generics are false even though the property is possessed by most members of the kind (e.g., Cohen, 1996; Leslie, 2007, 2008). Examples of the latter sort, which we called majority false generalizations, include statements such as books are paperbacks and Canadians are right-handed. Since much weight has been placed on these claims, it is important to confirm that peoples judgments are accurately represented. Further, the semantics literature tends to focus on a very limited number of examples often just one or two examples of generics are provided to make a point. It is thus an open question whether these claims hold reliably
GENERICS AND CONCEPTS 10
over a larger range of example sentences. We expected that the consensus in semantics would be supported and the judgments would generalize to a larger set of items. It is sometimes suggested that low prevalence generics such as ticks carry Lyme disease are only accepted because of the high associated cue validity i.e., because of the high conditional probability of something being a tick given that it carries Lyme disease. If this is correct, then one would expect that the low prevalence striking property generics would only be accepted if they had high associated cue validities, and further, that the higher the cue validity, the more acceptable the generic would be. However, there is no empirical evidence to bear on this issue, and so Experiment 1 addresses it directly. Another highly influential family of formal semantic accounts which we refer to as the normalcy approach relies on the idea that generics express claims about what is normal for the members of the kind, or, relatedly, about which properties they are supposed to have (e.g., Asher & Morreau, 1995; Carlson & Pellier, 1995; Dahl, 1975; Krifka et al., 1995; Nickel, 2008; Pelletier & Asher, 1997). These formal semantic accounts do not simply claim that some generics (e.g., the ones that involve principled connections) communicate information about normalcy; rather, they aim to provide a uniform analysis of the conditions under which any generic statement will be true or false. To test the claims made by these normalcy accounts, we asked people to evaluate whether a) all normal members of the kind have the property, and b) whether members of the kind are supposed to have the property in question. Support for the normalcy approach would be found if people only accepted generics when they also accepted these other related claims. We predicted, however, that striking generics and statistical generics would be accepted, despite having low scores on these measures.
In what follows, we describe three experiments that sought to clarify the conceptual foundations of striking and minority characteristic generics and test the diverging claims between the conceptually based approach and the formal semantics approach to generic statements.
Experiment 1: Which types of generics involve principled connections? The main goal of Experiment 1 was to investigate which types of generic knowledge involve representing a principled connection between a kind and a property. Of primary interest is whether striking property generics and minority characteristic generics involve principled connections. In addition to including striking property and minority characteristic generics, Experiment 1 included majority characteristic generics (e.g., dogs are four legged) which clearly involve principled connections, as well as majority statistical generics (e.g., barns are red) which clearly involve statistical connections. These items provided points of comparison for the striking property and minority characteristic generics. Cue-validities and prevalence estimates were also collected to test claims concerning the role of these factors in licensing generics. The various types of generic generalizations that we considered are provided in Table 1. --------------------------------------------Table 1 About here -------------------------------------------Experiment 1 investigated whether principled connections underlie generics for four predication types (e.g., majority characteristic, minority characteristic, majority
statistical, and striking property; see table 1) via truth-value judgments for five assertion types (see Table 2). ------------------------------------------Table 2 About Here ------------------------------------------Four of the assertion types reflected principled connections (in virtue of, is an aspect of, are supposed to, and all normal, see Table 2), while the fifth assertion type (bare plural generic) is not a measure of whether a connection is principled or not. Participants judged the truth of each combination of assertion and predication. The conceptually based approach makes the following predictions of their performance:
Predictions of the conceptually based approach (1) Majority characteristic predictations (e.g., tigers have stripes) should be highly rated in all five assertion types, since they involve principled connections and support generic statements. (2) Majority statistical predications (e.g., cars have radios) should receive low ratings in all four assertion types that measure principled connections because they do not involve principled connections. However, they should nonetheless receive high ratings in the bare plural form. (3) Minority characteristic predications (e.g., lions have manes) should be rated much more highly than majority statistical predictations (e.g. cars have radios) on the principled connection measures, with the exception of all normal assertions.
(4) Striking property predications (e.g., pitbulls maul children), like majority statistical predications (e.g., cars have radios), should receive low ratings in all assertion types except for the bare plural form. (5) Generics and prevalence: striking property predications should be accepted in bare plural form despite having low prevalence estimates, whereas majority false generalizations (e.g. books are paperbacks) should be rejected in bare plural form despite having high prevalence estimates. (6) Normalcy and generics: Contrary to the predictions of the normalcy accounts of generics, all normal assertions should be rated highly only for the majority characteristic predications, and they should receive much lower ratings for all the other types of predications. We also predict are supposed to assertions, (e.g., pitbulls are supposed to maul children) should be rejected for.striking property and majority statistical predications (7) Cue validity and striking property generics: acceptability ratings of (low prevalence) striking property generics should not be affected by cue validity.
Method Participants. 137 volunteers participated in the experiment over the Internet. Participants were chosen from Amazons Mechanical Turk system for human intelligence tasks (for an evaluation of the validity of this platform for psychological experimentation, see Paolacci, Chandler, & Ipeirotis, 2010). All spoke English as their first language and none had participated in experiments concerning generics before.
Materials and Design. Twelve items of each of the four types of predications discussed above (i.e., majority characteristic, minority characteristic, majority statistical, and striking predications) were included in the experiment. The items were drawn from a larger set of items for which we collected cue validity and prevalence norms (see Appendix A). The criteria for determining predication type are summarized in Table 1. Half of the items for each type of predication involved a natural kind and the other half involved an artifact kind, except for the minority characteristic predications, which involved only natural kinds. ii The four types of predications were matched with respect to the mean cue-validity of the properties they mentioned. In the norming study, participants estimated a mean prevalence of 58.99% for the minority characteristic items. Given this surprisingly high estimate, we sought to determine if the estimate was due to a bimodal response pattern in which most participants estimated the prevalence to be less than 50%, but a small number of participants estimated that most instances had the property, possibly due to their interpreting the task in a different manner. The distribution of responses was indeed bimodal (Hartigan's dip test =.059, p<.0001) with a large mode in the 40-50% range and a smaller mode in the 90-100% range. We will continue to use the term minority characteristic to refer to these predications for the sake of consistency. Though we will not be able to determine if principled connections can have estimated prevalences of less than 50%, the more important question is whether minority characteristic generics (e.g., ducks lay eggs) involve principled conections or not. For this purpose, the items are ideally suited as their prevalences are significantly lower than those of the majority statistical items and thus provide a clear way to distinguish between the effects of principled connections and prevalence. iii
We also included three other types of predications (majority false generalizations, minority false generalizations, and quasi-definitions) to encourage full use of the rating scale. The majority false generalizations (e.g., Canadians are right-handed) were of interest because they provide an opportunity to confirm that bare plural generics can be rejected despite high prevalence estimates. Minority false generalizations were statements that are deemed false as generics, such as cats are white, in which the property is true of a minority of instances. Quasi-definitions were statements such as ants are insects in which the property was virtually definitional and often referred to a superordinate category. The full list of items is given in Appendix A. For each predication, five assertions were generated. Four of the assertions specified phenomena reflecting principled connections (see Table 2). A bare plural form expressing generic knowledge was also generated. The materials generated a 7 (predication type) x 5 (assertion type) mixed factor design, with assertion type a betweenparticipants factor and predication type a within-participants factor. Each participant received all 12 items of each type of the 7 types of predications in one of 5 assertion types, i.e., 84 items in total. We chose to make assertion type a between-participants variable to allow comparisons of how each type of predication was rated for a given assertion type without interference from factors that may be relevant to judging other assertion types. iv Each participant rated all the predications in one of the five assertion types.
Procedure. The experiment was run on the Internet using an experiment interface written in Ajax (Javascript, HTML, and PHP). For each assertion, participants were asked to
evaluate the extent to which the assertion struck them as being true on a 7-point Likert scale, whose end points were labeled definitely true (+3) and definitely false (-3). After familiarization with the response scale, each participant received the items for a given assertion type in a different random order.
Results Given the bi-modal prevalence estimates for the minority characteristic items, a Hartigans dip test for bi-modality was performed on the data from the trials with the minority characteristic predications to determine if they received bi-modal responses in the present experiment as well which would suggest that participants interpreted these items in more than one way. The test revealed no tendency for bi-modality (p > .5) so the data can be interpreted straighforwardly. ----------------------------------------Insert Table 3 about here ----------------------------------------The means for each predication type as a function of assertion type are presented in Table 3. A 4 x 5 omnibus ANOVA with the four non-filler classes of predications and the five assertion types revealed significant effects of predication type, assertion type, and a significant interaction between the two (see Table 4, Test 1). v ----------------------------------------Insert Table 4 about here -----------------------------------------
(1) Do majority characteristic predications, unlike majority statistical predications, involve principled connections? We predicted a large difference between majority characteristic predications (e.g., tigers have stripes) and majority statistical predications (e.g., cars have radios) on the four assertions that reflect principled connections, but a smaller or no difference between the two predications for the truth-value ratings of the bare plural generics. A 2 x 5 ANOVA yielded the predicted interaction as well as significant main effects of predication type and assertion type (see Figure 1; Table 4, Test 2). Planned ttests with Bonferroni corrections for multiple comparisons revealed that majority characteristic predications receiving significantly higher ratings for all five assertion types, but, as predicted, the difference was smallest for the bare plurals. The details of these pairwise tests are reported in Appendix B (Test 1). -----------------Insert Figure 1 about here -----------------These data confirm that items that involve principled connections (majority characteristic predications) receive high scores on our principled connections measures, but items which clearly do not involve principled connections (majority statistical predications) receive lower scores on these measures.
(2) Are minority characteristic predications represented via principled connections? Based on the hypothesis that minority characteristic predications (e.g., lions have manes) involve principled connections, we predicted a substantial difference between
minority characteristic predications and majority statistical predications (e.g., cars have radios) on three of the four principled connection measures, and smaller differences in judgments concerning whether all normal instances of the kind have the property and the ratings of the bare plural generics. This prediction was confirmed. A 2 x 5 ANOVA of predication type versus assertion type with minority characteristic predications and majority statistical predications yielded the predicted interaction as well as significant main effects of predication type and assertion type (see Figure 2; Table 4, Test 3). Pairwise t-tests with Bonferroni correction (Appendix B, Test 2) revealed that, as predicted, the minority characteristic predications received significantly higher ratings than the majority statistical predications on all the principled connection measures except the all normal measure. The minority characteristic predications also received a significantly higher rating on the bare plural assertion. Note that the minority characteristic items received higher ratings than the majority statistical items on all the principled connection measures except the all normal measure, despite the fact that the minority characteristic items had a significantly lower prevalence for the critical properties (58.99%) than the majority statistical items (67.97%; t(11) = 2.38, p < .05, d = 1.28; see Appendix A). This provides clear evidence that the distinction between principled and statistical connections cannot be reduced to a difference in prevalence. Furthermore, it rules out the possibility that minority characteristic predications reflect only statistical connections. -----------------Insert Figure 2 about here ------------------
Because the minority characteristic items all involved natural kinds, it is important to see if the differences on the principled connections measures remain when only the natural kind majority statistical items are considered. This analysis found that the minority characteristic items received significantly higher scores on all the principled connections measures.
(3) Comparing minority characteristic and majority characteristic predications: We expected an interaction between predication type and assertion type for minority characteristic (e.g., lions have manes) and majority characteristic predications (e.g., tigers have stripes). There should be small or no differences for all measures except the one concerning whether all normal instances of the kind have the property. A 2 x 5 ANOVA of predication type versus assertion type with minority characteristic generics yielded the predicted interaction as well as significant main effects of predication type and assertion type (Table 4, Test 4). Pairwise t-tests with Bonferroni connection revealed that the majority characteristic predications were given higher ratings than the minority characteristic predications on all measures (Appendix B, Test 3), with the largest difference on the all normal assertion. The differences in ratings between majority and minority characteristic predications on measures other than the all normal measure are likely due to differences in the prevalence of the properties, which were significantly higher for the majority characteristic items (90.19%) than the minority characteristic items (58.99%, t(11) = 16.87, p < .0001, d = 8.06). This would be consistent with Prasada and Dillinghams (2006, 2009) finding that while principled and statistical connections cannot be distinguished on the basis of prevalence, prevalence
nonetheless has a small but significant effect on participants judgments (see also Khemlani, Leslie & Glucksberg, 2012).
(4) Are striking property generics represented via principled connections? Striking property predications (e.g., pitbulls maul children) should not involve principled connections and so we expected a large difference between striking property predications and majority characteristic predications (e.g., tigers have stripes), but a smaller or no difference in the ratings of the bare plural generics. These expectations were confirmed. A 2 x 5 ANOVA found the predicted interaction as well as significant main effects of predication type and assertion type (Figure 3; Table 4, Test 5). Pairwise t-tests with Bonferroni corrections for multiple comparisons revealed that the majority characteristic items received higher ratings for all five assertion types (Appendix B, Test 3), but, as predicted, the difference was smallest for the bare plurals. These data suggest that striking property generics are not represented via principled connections between a kind and a property. This conclusion is bolstered by the results of the next comparison. -----------------Insert Figure 3 about here ------------------
(5) Comparing striking property and majority statistical items: Since neither striking property (e.g., pitbulls maul children) nor majority statistical predications (e.g., cars have radios) are hypothesized to involve principled connections,
we expected, and found, no main effect of predication type or interaction between predication type and assertion type. A 2 x 5 ANOVA with striking and majority statistical items revealed no significant effect of predication type in the item analysis, but surprisingly, there was a significant effect in the participants analysis (see Figure 4; Table 4, Test 6). There was also a significant effect of assertion type, and a significant interaction. Planned t-tests with Bonferroni corrections for multiple comparisons (reported in Test 5 of Appendix B) revealed that the striking property and majority statistical predications did not differ on any of the assertions except the one pertaining to whether all normal instances of the kind have the property in question. Participants denied this to be the case to a greater extent for the striking property predications than the majority statistical predications. Thus, as predicted, striking property predications mostly patterned like majority statistical predications on phenomena that reveal principled connections. The one assertion type for which the two types of predications differed was the extent to which they supported the expectation that all normal instances should have the property, with striking predications receiving a lower rating. This indicates that they are less like predications that involve principled connections on this measure than the majority statistical predications. These data strongly suggest that striking property predications are not represented via principled connections between the kind and the property. -----------------Insert Figure 4 about here -----------------(6) Can generic acceptance be dissociated from prevalence?
In keeping with the claims made by linguists and philosophers we found that generic acceptance can be dissociated from prevalence. Striking property bare plural generics (e.g., pitbulls maul children) were accepted to the same extent as majority statistical bare plural generics (e.g., cars have radios) (1.27 and 1.23 respectively, t(27) = .34, p = .74, d = .05; see Table 3). However, the prevalence estimates associated with the former were significantly lower than the latter (33.13% and 67.97% respectively; t(11) = 10.34, p < .0001, d = 4.37). Furthermore, we also found that both majority statistical bare plural generics and striking property bare plural generics were accepted significantly more than were majority false generalizations in bare plural form (e.g., books are paperbacks) (1.23, 1.27 and -.06 respectively, ts(27) > 7.73, ps < .0001, ds > 1.52). However, majority statistical and majority false generalizations did not differ in their prevalence estimates (67.97% and 64.39% respectively), and striking property predications had significantly lower prevalence estimates (33.13% vs. 64.39%, t(11) = 8.25, p < .0001, d = 3.11).
(7) Are generics understood as making claims about normal instances? Our resul ts suggest that this is not the case. The assertion All normal xs are p received a relatively high rating for the majority characteristic predications (e.g., all normal tigers have stripes). Nevertheless, this rating was lower than the corresponding bare plural generic (e.g. tigers have stripes) (2.09 vs. 2.48, t(54) = 2.03, p < .05, d = .54). The all normal assertion received much lower ratings and was significantly worse than the corresponding bare plural for the majority statistical (e.g., all normal cars have radios vs. cars have radios), minority characteristic (e.g., all normal lions have
manes vs. lions have manes), and striking property items (e.g. all normal pitbulls maul children vs. pitbulls maul children), ts(54) > 4.59, ps < .0001, ds > 1.23), with especially low ratings for the striking property predications in all normal form. Further, the ratings for an x is supposed to be p also differed significantly from the ratings of the bare plural statements for striking property (e.g., a pitbull is supposed to maul children), t(53) = 5.07, p < .0001, d = 1.36, and majority statistical predications (e.g., a car is supposed to have a radio), t(53) = 3.56, p < .001, d = .95. These data provide evidence against the normalcy approaches to generics.
(8) Cue validity and striking property generics: Is high cue validity needed for licensing low prevalence generics? The answer to this question is no. High cue validity is not a necssariy condition for licensing low-prevalence generics. There are two ways in which Experiment 1 provides data concerning the viability of the cue validity proposal. First, the fact that a stimulus set could be constructed wherein the cue validity of the low prevalence striking property items (e.g., pitbulls maul children) and the majority statistical items (e.g., cars have radios) does not differ, where both sets of generics equally acceptable despite the large and significant difference in prevalence, speaks against the idea that low prevalence generics are only possible if their cue validities are much higher than those of generics that are licensed by high prevalence (majority statistical generics). The cue validities for both conditions were not particularly high (.62 and .97 on a scale that ranges from -3 to +3), so the results are unlikely to be due to a threshold effect wherein the cue validities are high enough to license any generic. More dramatically, the majority false
generalizations (e.g., Canadians are right-handed) also matched to the striking property items on cue validity (.68 vs. .97, t(11) = .85, p = .41, d = .31), but had significantly higher prevalence estimates associated with them (64.39% vs. 33.13%, t(11) = 8.25, p < .0001, d = 3.11). Striking property generics were nonetheless accepted significantly more than the majority false generalizations, as reported above. Because the items were not randomly selected, these data can only be taken to show that it is empirically possible for cue validity and prevalence to be dissociated. A second way to address the cue validity proposal is to perform a median split on the cue validities of the (low prevalence) striking property items to see if the acceptability of the generic differs for the low and high cue validity items. The split yielded a set of items with mean cue validities of .52 and 1.4, which are significantly different from one another, t(10) = 4.39, p < .005, d = 2.54, and mean prevalences of 34.2 and 32.0, which do not differ statistically, t(10) = .54, p = .60, d = .31. Importantly, there was no hint of a difference in the acceptability of the generics (1.49 vs. 1.51) corresponding to each group, t(10) = .29, p = .78, d = .17. These data suggest that low prevalence generics are licensed by something other than high cue validities.
Discussion The results of Experiment 1 confirmed that majority characteristic and majority statistical generics involve principled and statistical connections, respectively. It also provided evidence that striking property generics such as pitbulls maul children are not represented via principled connections. Instead, they pattern like majority statistical predications (e.g., cars have radios), and not like majority characteristic generics (e.g.,
tigers have stripes). In additon, Experiment 1 found that minority characteristic generics (e.g., lions have manes) received higher ratings on the relevant principled connection measures than the majority statistical generics did, even though the minority characteristic items had lower prevalence ratings than the majority statistical items. This result provides strong evidence that ratings on the principled connection measures are not simply a function of prevalence. It also suggests that minority characteristic generics do not involve statistical connections. Minority characteristic items also differed from majority characteristic items: minority characteristic predications displayed a pattern of responses that suggests they are represented via principled connections, but not to the same extent as majority characteristic predications. This result can be understood in two ways. First, minority characteristic generics may involve yet another kind of connection between kinds and properties. Alternatively, minority characteristic generics may well involve principled connections, but our dependent measures were affected by factors other than connection type. We favor the latter interpretation. The obvious additional factor that likely contributed to the difference in ratings between majority and minority characteristic predications is the prevalence of the property. Participants clearly know that roughly half the instances do not have the property for the minority characteristic predications, and that knowledge may have influenced their judgments. Prasada and Dillingham (2006, 2009) found that while principled and statistical connections cannot be distinguished on the basis of prevalence, prevalence nonetheless has a small but significant effect on participants judgments (see also Khemlani et al., 2012). Experiment 1 appears to have yielded a similar result.
Overall, our results show that principled connections underlie minority characteristic generics. This suggests that the statistical dimension of principled connections (at most) grounds the expectation that most instances will have the property, but does not require that most instances have the property. One might note that our examples of minority characteristic generics all involved properties that are only had by one sex. How important is this sex-specificity in shaping peoples judgments? That is, suppose that half the members of an animal species, regardless of sex, had, e.g., manes while the others did not. Would people understand the connection between this kind and the propery of having a mane differently than they understand the connection between lions and the property of having a mane? Real-world examples of this sort are not readily available, and since our items drew on real-world knowledge, we could not directly examine this question here. However, a recent study using novel animal kinds suggests that people do not differentiate between these two cases, at least when it comes to accepting bare plural generics (Cimpian, Gelman, & Brandone, 2010). Thus we cautiously speculate that sex-specificity is not especially important in shaping peoples judgments rather it just so happens that, as a matter of biology, real-world examples have this character. Experiment 1 also provided evidence for the claim in the semantics literature that generics can be rejected despite high prevalence, and conversely, that they can be accepted despite low prevalence. However, we found that people accepted the claim that all normal members of a category have a property (or are supposed to have a property) only for the majority characteristic generics. This suggests that normalcy approaches to generics cannot account for the full range of generic beliefs. We also found that the (low
prevalence) striking property generics were accepted even though they did not have particularly high cue validity estimates associated with them. Further, the striking property generics with the highest cue validity ratings were not more likely to be accepted than the ones with the lowest cue validity ratings. These findings contravene the idea that cue validity licenses low prevalence generics.
Experiment 2: Which types of generics support formal explanations? Experiment 1 provided an indirect measure of the explanatory aspect of principled connections via judgments about whether different predication types are supported a by virtue of assertion. Experiment 2 sought to investigate whether the connections that underlie the kinds and properties in striking and minority characteristic generics explicitly support formal explanations. In this experiment, participants were presented with a question (Q) that asked why an instance of a kind had a given property. They then rated the naturalness of an answer (A) that explained the propertys presence via reference to the kind of thing it was along with the basis for the explanation (that the property is one aspect of being that kind of thing). By specifying the basis for why reference to the kind is explanatory, the experiment ensured that participants rated the acceptability of formal explanations rather than some other type of explanation that may be implicitly deployed by reference to the kind of thing something is. Formal explanations for items that involve principled connections, such as example (1) below, should yield high ratings, however, when a statistical connection is involved, as in (2), formal explanations should receive low ratings (Prasada & Dillingham, 2006). (1) Q: Why does that (pointing to a tiger) have stripes?
A: Because it is a tiger, and having stripes is one aspect of being a tiger. However, (2) Q: Why does that (pointing to a car) have a radio? A: Because is is a car and having a radio is one aspect of being a car. Consequently, we expect that formal explanations involving majority characteristic items (e.g., tigers have stripes) would receive higher ratings than formal explanations involving majority statistical items (e.g., cars have radios). Based on the results of Experiment 1, we further expected that formal explanations involving minority characteristic predications (e.g., lions have manes) would be rated more highly than those involving majority statistical predications. We also expected formal explanations for majority characteristic items to receive higher ratings than those for minority characteristic items. Finally, we expected that formal explanations involving striking property predications (e.g., pitbulls maul children) would not be rated more highly than those involving majority statistical items (e.g., cars have radios).
Method Participants. 25 volunteers from the same participant pool as in Experiment 1. All spoke English as their first language, and none had participated in experiments concerning generics before.
Materials and Design. We used the experimental items from Experiment 1 to generate question-answer pairs such as (1). The false generalization items from Experiment 1 were
included as fillers. The materials generated a one-way repeated measures design. Each participant received the items in a different random order.
Procedure. Participants evaluated the extent to which the responses to the questions sounded natural on a 7-point Likert scale.
Results and Discussion Participants responses on the Likert scale were given numerical values from -3 (very bad) to +3 (very good). The mean responses are given in Table 5. As with Experiment 1, a Hartigans dip test on the data from the trials with the minority characteristic predications revealed no tendency for the data to be bi-modal (p > .5).
----------------------------------------Insert Table 5 about here ----------------------------------------We found a a significant effect of predication type, F1(3,72) = 56.52, p < .0001, F2(3,63) = 58.79, p < .0001, min F(3,110) = 28.82, p < .0001. Planned t-tests revealed that, as predicted, formal explanations involving majority characteristic predications (e.g., tigers are striped) received higher ratings than formal explanations involving majority statistical items (e.g., cars have radios) t1(24) = 10.03, p < .0001, t2(11) = 7.94, p < .0001. This result essentially replicates the findings of Prasada and Dillingham (2006). More importantly, formal explanations involving minority characteristic predications
(e.g., lions have manes) also received higher ratings than formal explanations involving majority statistical predications, t1(24) = 7.28, p < .0001, t2(11) = 6.81, p < .0001. As with Experiment 1, we ran an analysis comparing the minority characteristic items to only the natural kind majority statistical items. We found the same pattern of results. These results add to the evidence from Experiment 1 that minority characteristic generics are represented via principled connections. On the other hand, formal explanations involving striking property predications (e.g., pitbulls maul children) did not receive significantly higher ratings than formal explanations involving majority statistical items (e.g., cars have radios). In fact, they were marginally lower in the participant analysis, t1(24) = 1.81, p = .08, t2(11) = 1.38, p = .19. Finally, the ratings for the formal explanations involving minority characteristic predications (e.g., lions have manes) received lower ratings than formal explanations involving majority characteristic predications (e.g., tigers have stripes), t1(24) = 3.55, p < .005, t2(11) = 4.23, p < .005. This was also the case when only the natural kind majority characteristic items were included in the analysis. This result suggests that factors other than connection type (most likely prevalence) can contribute to the dependent measure. Experiments 1 and 2 suggest that majority characteristic and minority characteristic generics both involve principled connections and that neither striking property nor majority statistical generics do so. This raises the question of whether striking property and majority statistical generics may both involve statistical connections, or whether striking property generics involve a different type of connection between kinds and properties. The fact that striking property generics are accepted as true
despite the low prevalence of their properties strongly suggests that striking property generics do not involve mere statistical connections. Leslie (2007; 2008; in press) and Prasada (2010) suggest that striking property generics involve a causal connection between a kind of thing and the property in question, in the sense that, e.g., the biological makeup of ticks causes them to be disposed to carry Lyme disease. It is not enough that a few members of the kind happen to have the striking property rather there must be something about the nature of the kind that causally grounds the presence of the property. Where there is a mere statistical connection, on the other hand, the attribution of a causal connection sounds infelicitous (e.g., even though dogs wear collars is judged as true, there is nothing about being a dog that causes it to wear a collar).
Experiment 3a: Do striking property generics involve causal connections? To assess whether striking property generics, but not majority statistical generics, involve a causal connection between the kind and property, we asked participants to judge assertions of the form, There is something about Ks that causes them to P. If striking property generics involve causal connections, then participants will accept the assertion of a causal connection for striking predications but not for majority statistical predications.
Method Participants. 20 volunteers from the same participant pool as in the previous experiments. All spoke English as their first language and nNone had participated in experiments concerning generics before.
Materials and Design. Experiment 3a used the striking and majority statistical items used in the previous experiments. Each item was used to generate a statement involving a causal connection (e.g., There is something about cars that causes them to have radios). The materials generated a one-way repeated measures design.
Procedure. Participants evaluated the extent to which the statement struck them as being definitely true (+3) or definitely false (-3) on a 7-point Likert scale.
Results and Discussion Participants rated striking items significantly higher (M = 1.77) than the majority statistical items (M = .48), t1(19) = 5.77, p < .0001, d = 1.16, t2(11) = 4.10, p < .005, d = 1.66. Furthermore, 18 out of 20 participants accepted the striking property predications more often than majority statistical predications in the causal formulation (Binomial test, p < .0005). This finding supports the idea, put forward in Leslie (2007, 2008, in press) and Prasada (2010), that striking property generics are understood as involving a causal connection between the kind and the property. Even though very few mosquitoes actually carry malaria, people still believe that there is a causal connection between mosquitoes and malaria i.e. that there is something about mosquitoes that causes them to carry malaria. No such connection need exist between a kind and a merely statistically prevalent property. vi
Experiment 3b: What mediates causal connections of striking generics?
There are many sorts of causal chains, and so we sought to specify the nature of the causal connection that is thought to hold between a kind and a striking property. Striking properties appear to come about as a result of the underlying physical, material, and functional organization of the kind in question. To test this idea, we asked participants to judge assertions of the form The parts and functional organization of Ks cause them to be P. While stilted, this statement provides a relatively straightforward way of capturing the relevant intuition in a manner that can be applied across a broad range of items.
Method Participants. 18 volunteers from the same participant pool as in the previous experiments. All spoke English as their first language and none had participated in experiments concerning generics before.
Materials and Design. Same as Experiment 3a except that the statement rated by participants was The parts and functional organization of Xs cause them to be Y.
Procedure. Same as Experiment 3a.
Results and Discussion Participants rated striking items significantly higher (M = 1.01) than the majority statistical items (M = .08), t1(17) = 4.15, p < .001, d = .69, t2 (11) = 3.67, p < .005, d =
1.35. Furthermore, 16 out of 18 participants accepted the striking predications more often than majority statistical predications in the causal formulation (Binomial test, p < .005). The results of Experiments 3a and 3b suggest that striking property generics involve representing a causal connection between the kind and the property, and that striking properties are understood to be causally connected to the parts and functional organization of instances of the kind. This causal connection grounds the intuition that the striking properties are systematically related to the nature of the kind. As with all causal connections, how often an effect (the striking property) is actually realized depends on how often the relevant enabling conditions are present and preventative conditions are absent. As such, the causal connection grounds the truth of striking property generics even though the striking property may be rarely realized because the relevant enabling conditions may be quite rare. Generics involving principled connections (majority and minority characteristic generics) would likely receive high ratings on the measures used in Experiments 3a and 3b, indicating that these generics also involve causal connections. However, principled connections cannot simply be causal connections because, as Experiments 1 and 2 show, striking property generics, which involve causal connections, do not support (i) formal explanations, (ii) normative expectations, (iii) by virtue assertions, or (iv) understanding striking properties as aspects of the relevant kind of thing. General Discussion Experiments 1 and 2 provided evidence that minority characteristic generics (e.g., lions have manes) and majority characteristic generics (e.g., tigers have stripes) involve principled connections, while striking property generics (e.g., pitbulls maul
children) and majority statistical generics (e.g., cars have radios) do not. Minority characteristic generics behave more like majority characteristic generics on measures designed to test the presence of principled connections than do majority statistical generics (even though the minority characteristic generics have a lower prevalence than the majority statistical generics). This result strongly suggests that an account based on prevalence alone cannot distinguish between principled and statistical connections. Experiment 1 also found that striking property predications were accepted as true in the bare plural form even though the prevalence of the properties in question was estimated to be very low among members of the kind. Conversely, participants rejected generics such as books are paperbacks and Canadians are right-handed (i.e., majority false generalizations) even though they involve properties that are possessed by the majority of the members of the kind. These findings confirm some claims that have been made in the semantics literature, and provide further evidence that generics cannot be understood to be simply making vague quantificational assertions (Carlson, 1977; Lawler, 1973; Leslie, 2007; 2008). Instead, different types of generics involve distinct ways in which we connect kinds and properties (Prasada & Dillingham, 2006, 2009). Experiment 3 provided evidence that striking property generics, but not majority statistical generics, involve a causal connection between the kind and property. Together, the experiments suggest that our conceptual systems provide at least three ways (principled, causal, and statistical) of representing the connection between a kind and a property. They thus provide evidence for Leslies (2007, 2008) claim that there are three fundamantal ways of generalizating information within our conceptual systems: one pertaining to the generalizations that underlie generics that involve majority and minority
characteristic predications; one pertaining to the generalizations that underlie generics that involve striking predications; and one pertaining to the generalizations that underlie generics that involve majority statistical predications. These three modes of generalization are proposed by Leslie (2007, 2008) to be our default modes of generalization. They may, as a consequence, appear earlier in cognitive development than explicitly quantificational generalizations which involve set-theoretic notions (see Leslie, 2012, for a review). In what follows, we (i) discuss each connection type and its features, (ii) consider the implications of our findings for formal semantic approaches to generics, (iii) consider what these three connection types reveal about our conceptual systems, and (iv) identify some questions for future research.
Three types of connections Principled connections. Prasada & Dillingham (2006, 2009) proposed that principled connections have an explanatory, a normative, and a statistical dimension, and provided evidence that representing a property as having a principled connection to a kind requires representing it as being an aspect of being that kind of thing. Data from Experiments 1 and 2 help to elaborate our understanding of principled connections by showing that both minority characteristic generics and majority characteristic generics support formal explanations and display a normative dimension. These findings extend the scope of principled connections. Previous research had considered only majority characteristic properties. That minority characteristic predications involve principled connections suggests that principled connections do not require high statistical prevalence, but we would argue nonetheless that, if a principled connection is
represented, then people expect the property to be statistically prevalent. This expectation, while sometimes violated, helps explain some otherwise puzzling findings. For example, Khemlani, Leslie and Glucksbergs ( 2009, 2012) finding that 85% of people who accepted the generic ducks lay eggs also agreed to Quacky lays eggs when told only that Quacky is a duck. Further, the prevalence estimates that we collected reflect a tendency to overestimate the prevalence of minority characteristic predications (see also Khemlani et al., 2007, 2009). These findings would all be explained by the hypothesis that principled connections support the expectation of statistical prevalence, even though this expectation is sometimes violated.
Causal connections. Experiment 3 provided evidence that striking property generics involve a causal connection between the kind and the property. Experiments 1 and 2 showed that striking property generics do not support formal explanations or normative expectations, and thus suggest that causal connections need not support formal explanations or normative expectations. vii We hypothesize that causal connections also do not support expectations concerning prevalence. The prevalence of the property is determined by the prevalence of relevant enabling conditions and the absence of relevant prohibiting conditions. How often such circumstances occur is not determined by the kind in question, nor by the causal connection between the kind and the property.
Statistical connections. Finally, there is a class of generics that involve merely statistical connections between the kind and property. These connections support neither formal explanations, normative expectations, nor causal statements concerning the
property in question. Though the connection is justified on a statistical basis, we do not currently know the statistical principle(s) by which our conceptual systems decide that the connection observed in a limited sample of instances may be generalized to the indefinitely many instances that constitute the kind (Prasada, 2010). As Leslie (2007, 2008; following Carlson, 1977; Carlson & Pelletier, 1995) notes, statistical prevalence alone is not enough for a generic to be accepted. Data from Experiment 1 confirmed this. The prevalence estimates for false generalizations (e.g. books are paperbacks, Canadians are right-handed) were high, yet they were consistently rejected in bare plural generic form. Leslie (2007, 2008) suggests that these generics are rejected because the exceptions to the generics have salient, positive alternative properties e.g., the hardcover books, and the left-handed Canadians. Our experiments here did not test this hypothesis, however, since these false generalizations were used only as fillers.
Formal semantic approaches to generics Consistent with a number of claims in the semantics literature, we found that generics can be rejected despite having high associated prevalence estimates, and conversely, they can be accepted despite having low associated prevalence estimates. However, some of our other findings are at odds with many of the semantic accounts of generics that have been proposed. In particular, many accounts cluster together under what we have called normalcy approaches to generics (e.g., Asher & Morreau, 1995; Dahl, 1975; Krifka et al., 1995; Nickel, 2008; Pelletier & Asher, 1997). These accounts differ in their details, but share a common commitment to the idea that generics are only
true if it is normal for the members of the kind to have the property in question. There have been several theoretically-based critiques of this approach (e.g., Cohen, 1996; Leslie, 2007, 2008). However no prior empirical work had tested the proposal directly. We found that people consistently judge that all normal members of the kind have the property and also that members of the kind are supposed to have the property only in the case of majority characteristic generics such as tigers are striped which are often the only examples discussed by proponents of the normalcy approach. Majority statistical and striking property items were accepted in bare plural generic form, but not in the all normal or supposed to formulations. This suggests that normalcy approaches cannot account for the full range of generic beliefs. We also found that striking property generics are accepted despite having low associated prevalence estimates. However, the striking property generics did not have particularly high cue validity estimates associated with them, and further, the striking property generics with the highest cue validity ratings were not at all more likely to be accepted than the ones with the lowest cue validity ratings. This suggests that cue validity does not explain why low prevalence generics are accepted. Instead, the results of the present experiments suggest that low prevalence striking property generics may involve a causal connection between a kind and a property. The present research thus identifies a number of limitations of formal semantic approaches to generics that seek to provide a general account of generics in terms of notions such as prevalence, normalcy, and cue validity.
Generics and conceptual representation
Our findings have important implications for the manner in which conceptual knowledge is represented. Research inspired by prototype theories of conceptual representation has focused attention on the probabilistic nature of much of our conceptual knowledge, and thus emphasizesd the role of statistical connections in conceptual representation (e.g. McRae, de Sa, & Seidenberg, 1997; Rogers & McClelland, 2004; Tyler & Moss, 2001; Yoshida & Smith, 2003). On the other hand, research within the theory-based approach to conceptual representation has focused attention on the explanatory nature of much of our conceptual knowledge, and has emphasized the role of causal knowledge and causal connections in conceptual representation (e.g. Carey, 1985; Gelman, 1990; Gelman, 2003; Keil, 1989; Rehder, 2009; Sloman, Love & Ahn, 1998). Though these two approaches are often discussed as if they are mutually exclusive, a number of researchers have argued for hybrid theories of conceptual representation whereby conceptual knowledge is constituted by causal/explanatory as well as statistical knowledge (Gelman, 1990; Keil, 1989; McNorgan, Kotack, Meehan, & McRae, 2007; Sloman, Love & Ahn, 1998). The present work suggests the need for another dimension to conceptual knowledge namely the formal dimension that involves principled connections. Prasada & Dillingham (2009) provided evidence that representing a principled connection between a kind and a property involves representing the property as an aspect of being that kind of thing. As such, principled connections involve representing a specific type of part-whole relation between the property and the kind of which it is an aspect. Thus, our knowledge of kinds and their properties seems to have a formal dimension which complements the causal and statistical dimensions investigated in
previous research on conceptual representation (see Prasada & Dillingham, 2009 for a more detailed characterization of aspect and kind representations). Generics reflect the three perspectives we can take when thinking and reasoning about kinds of things. We can focus on the formal dimension by attending to principled connections. In doing so, we notice certain formal explanatory relations as well as the basis for certain normative and statistical expectations. Alternatively, we can focus on the material dimension by attending to the material constitution of the instances and thus their causal dispositions to behave in one or another manner in appropriate circumstances. Finally, we may adopt a perspective where we do not attend to either the formal or material characteristics of instances of a kind, but simply note the statistical connections that exist between instances of the kind and various properties. Although we have distinguished the different connection types using different kinds and properties, we do not mean to imply that only a single connection type is represented between a given kind and a given property. There may be cases in which we represent multiple connections between a specific kind and property. For example, it is likely that we represent all three types of connections between the kind dog and the property of having four legs. As previous research and the present experiments show, each connection type grounds different kinds of linguistic and nonlinguistic phenomena, and thus it does not seem that connection types are reducible or replaceable by one another. Finally, the types of connection properties have to kinds need to be distinguished from the centrality or mutability of the property (Sloman, Love & Ahn, 1998). For example, there is a principled connection between being a canary and being yellow even
though being yellow is not a causally central property of canaries. Similarly, it is easy to change the number of legs of a dog, or the color of a banana, or the crunchiness of a carrot, and leave their identity and most of their properties intact even though these properties have a principled connection to the kinds in question. More generally, though there may be tendencies for properties with statistical connections to be less causally central and more mutable than properties that have principled or causal connections to a kind, we do not think it is possible to replace the notion of connection types with information about feature centrality or mutability. In our opinion, conceptual representations of kinds will represent information about the types of connections that exist between the kind and the property as well as information concerning the centrality or mutability of the property for that kind (see Prasada & Dillingham, 2006, 2009 for further discussion).
Summary and directions for future research The experiments in this paper flesh out the conceptually based approach to generics by providing evidence that minority characteristic predications involve principled connections, that striking property predications involve causal connections between kinds and properties, and that principled connections ground an expectation rather than a requirement of high prevalence. The experiments also provide data that are problematic for normalcy approaches to generics and for the idea that cue validity licenses low prevalence generics. Thus they provide further reasons for pursuing the conceptually based approach to the semantics of generics (Leslie, 2007, 2008; see also Carlson, 2010).
The present research also suggests a number of directions for future investigation. By identifying the distinct types of connections we represent between kinds and properties, the present research helps clarify the nature of the acquisition problems faced by children. In acquiring generic knowledge, children must not only figure out which properties to generalize to the kind, they must also figure out what type of generalization (connection type) is involved. An interesting question for future research concerns the developmental time course of distinct forms of generic knowledgedoes generic knowledge involving one of the connection types develop earlier than the other connection types? Another important goal for future research will be to determine how linguistic cues might help children with this problem. Though it is well known that principled and statistical generics show somewhat different linguistic profiles (specifically, only principled connections support indefinite singulars such as a tiger is striped; Lawler, 1973; Leslie et al., 2009), the question of how generics involving causal connections may be expressed linguistically has not been systematically studied. It will also be important to investigate how the distinct ways in which properties are connected with kinds impact inductive and default reasoning (Khemlani et al., 2009; 2012). Finally, the present research raises the question of whether multiplicities that are not construed as kinds (e.g. white bears) can also be characterized via principled, statistical and causal connections (Prasada & Dillingham, 2006; Prasada, Hennefield & Otap, 2012; Prasada, 2012).
References Carey, S., (1985). Conceptual change in childhood. Cambridge, MA: MIT Press. Carlson, G., & Pelletier, F.J. (1995). The generic book. Chicago, IL: University of Chicago Press. Carlson, G. (1977). Reference to Kinds in English. Ph.D. dissertation, University of Massachusetts, Amherst. Carlson, G. (2009). Generics and concepts. In F. J. Pelletier (Ed.) Kinds, things, and stuff: Mass terms and generics. New Directions in Cognitive Science, v. 12. New York: Oxford University Press. Cimpian, A., Brandone, A. C., & Gelman, S. A. (2010). Generic statements require little evidence for acceptance but have powerful implications. Cognitive Science, 34(8), 1452-1482. Cimpian, A., & Markman, E. M. (2009). Information learned from generic language becomes central to childrens biological concepts: Evidence from their openended explanations. Cognition, 113 (1), 14-25. Gelman, R. (1990). First principles organise attention to and learning about relevant data: Number and animate-inanimate distinction as examples. Cognitive Science, 14, 79-106. Gelman, S. A. (1988). The development of induction within natural kind and artifact categories. Cognitive Psychology, 20, 65-95. Gelman, S. A. (2003). The essential child: Origins of essentialism in everyday thought. New York: Oxford University Press.
Gelman, S. A. (2004). Psychological essentialism in children. Trends in Cognitive Sciences, 8, 404409. Gelman, S. A., & Bloom, P. (2007). Developmental changes in the understanding of generics. Cognition, 105. Gelman, S. A. & Markman, E. M. (1986). Categories and induction in young children. Cognition, 23 (3), 183-209. Hampton, J. (2009). Stability in concepts and evaluating the truth of generic statements. In F. J. Pelletier (Ed.) Kinds, things, and stuff: Mass terms and generics. New Directions in Cognitive Science, v. 12. New York: Oxford University Press. Heit, E. & Rubinstein, J. (1994). Similarity and property effects in inductive reasoning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 411422. Keil, F. C. (1989). Concepts, kinds, and cognitive development. Cambridge, MA: MIT Press. Khemlani, S., & Johnson-Laird, P. N. ( in press). Theories of the syllogism and related instances: A meta-analysis. Pschological Bulletin. : Khemlani, S., Leslie, S. J., & Glucksberg, S. (2009). Generics, prevalence, and default inferences. In Proceedings of the 31st Annual Cognitive Science Society. Austin, TX: Cognitive Science Society. Khemlani, S., Leslie, S. J., & Glucksberg, S. (2012). Inferences about members of kinds: The generics hypothesis. Journal of Language and Cognitive Processes, 27(6), 887-900.
Khemlani, S., Leslie, S.J., Glucksberg, S., & Rubio-Fernandez, P. (2007). Do ducks lay eggs? How people interpret generic assertions. In Proceedings of the 29th Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society. Lawler, J. (1973). Studies in English generics. University of Michigan Papers in Linguistics, 1 (1). Leslie, S.J. (2007). Generics and the structure of the mind. Philosophical Perspectives, 21, 375-403. Leslie, S.J. (2008). Generics: Cognition and acquisition. Philosophical Review, 117 (1), 1-47. Leslie, S.J. (2012). Generics articulate default generalizations. Recherches Linguistiques de Vincennes. Leslie, S.J. (in press). The original sin of cognition: Fear, prejudice, and generalization. The Journal of Philosophy. Leslie, S-J., Khemlani, S., & Glucksberg, S. (2011). Do all ducks lay eggs? The Generic Overgeneralization Effect. Journal of Memory and Language. 65, pp. 15-31. Lopez, A., Atran, S., Coley, J.D., & Medin, D.L. (1997). The tree of life: universal and cultural features of folkbiological taxonomies and inductions. Cognitive Psychology, 32, 251-295. Markman, E. M. (1989). Categorization and naming in children: Problems of induction. Cambridge, MA: MIT Press.
McRae, K., de Sa, V., & Seidenberg, M. S. (1997). On the nature and scope of featural representations of word meaning. Journal of Experimental Psychology: General, 126, 99-130. McNorgan, C., Kotack, R.A., Meehan, D.C., & McRae, K. (2007). Feature-feature causal relations and statistical co-occurrences in object concepts. Memory & Cognition, 35, 418431. Osherson, D. N., Smith, E. E., Wilkie, O., Lopez, A., & Shafir, E. (1990). Categorybased induction. Psychological Review, 97, 185-200. Pelletier, F.J. (2009). Kinds, things, and stuff: Mass terms and generics. New Directions in Cognitive Science, v. 12. New York: Oxford University Press. Prasada, S. (1999). Names of things and stuff: An Aristotelian perspective. In: R. Jackendoff, P. Bloom and K. Wynn, (Eds.), Language, logic, and concepts: Essays in memory of John Macnamara. Cambridge, MA: MIT Press. Prasada, (2000). Acquiring generic knowledge. Trends in Cognitive Sciences, 4, 6672. Prasada, S. (2010). Conceptual representations and some forms of genericity. In F. J. Pelletier (Ed.) Kinds, things, and stuff: Mass terms and generics. New Directions in Cognitive Science, v. 12. New York: Oxford University Press. Prasada, S. (2012). Mechanisms for characterizing kinds and classes. Recherches Linguistiques de Vincennes Prasada, S., & Dillingham, E. (2006). Principled and statistical connections in common sense conception. Cognition, 99 (1).
Prasada, S., & Dillingham, E. (2009). Representation of principled connections: A window onto the formal aspect of common sense conception. Cognitive Science, 33, 401-448. Prasada, S., Hennefield, L., & Otap, D. (2012). Conceptual and linguistic representations of kinds and classes. Cognitive Science, 36, 1224-1250. Rehder, B. (2009). Causal-based generalization. Cognitive Science, 33, 310-344. Rehder, B., & Hastie, R. (2004). Category coherence and category-based property induction. Cognition, 91, 113-153. Rips, L. J. (1975). Inductive judgments about natural categories. Journal of Verbal Learning and Verbal Behavior, 14, 665-681. Rogers, T. T. & McClelland, J. L. (2004). Semantic cognition: A parallel distributed processing approach. Cambridge, MA: MIT Press. Sloman, S. (1993). Feature-based induction. Cognitive Psychology, 25, 231280. Tyler, L.K. & Moss, H. E. (2001). Towards a distributed account of conceptual knowledge. Trends in Cognitive Sciences, 5, 244-252. Yoshida, H. & Smith, L. B. (2003). Shifting Ontological Boundaries: How Japanese- and English-speaking children generalize names for animals and artifacts. Developmental Science, 6 (1), 1-36.
Table 1 : Types of predication used to generate assertions for Experiment 1. Truth value of Description the generic
Property must be prevalent though not universally had among members of the kind; some exceptional members (e.g., albino tigers) fail to possess it. Relevant by virtue paraphrase must sound acceptable (Prasada & Dillingham, 2006). Property must only be held by a minority of the kind, and must be central, principled or essential (Medin & Ortony, 1989; Gelman, 2003). For our purposes we restricted these items to methods of gestation, methods of nourishing the very young, and characteristic physical traits had only by one gender. Property must be prevalent among members of the kind, and must not be a majority characteristic connection (i.e. the by virtue paraphrase must sound unacceptable). Property must only be had by a small minority of the kind, and must signify something dangerous and to be avoided. Property must be universally true of all the members of the kind; no exceptions. Property must be prevalent among members of the kind and there must be a sufficiently salient alternative property (e.g. being a lefthanded), so that the generic form of the predication sounds false or mistaken. Property must be held by very few members of the kind but must not signify something dangerous. The generic form of the predication must sound false or mistaken.
Predication type
Example
Majority characteristic
Tigers are striped
True
Lions have manes Cars have radios Pit bulls maul children Ants are insects Canadians are right-handed Rooms are round
True
Majority statistical Striking Quasi-definition Majority false generalization Minority false generalization
True True True
False
False
Table 2 . Examples of the seven types of predication and five types of assertions used in Experiment 1
Predication type Majority characteristic Minority characteristic Majority statistical Striking Quasi-definition Majority false generalization Minority false generalization Aspect Being striped is one aspect of being a tiger Having a mane is one aspect of being a lion Having a radio is one aspect of being a car Mauling children is one aspect of being a pit bull Being an insect is one aspect of being an ant Being right-handed is one aspect of being a Canadian Being round is one aspect of being a room By Virtue Tigers, by virtue of being tigers, are striped Lions, by virtue of being lions, have manes Cars, by virtue of being cars, have radios Pit bulls, by virtue of being pit bulls, maul children Ants, by virtue of being ants, are insects Canadians, by virtue of being Canadians, are right-handed Rooms, by virtue of being rooms, are round Normativity Tigers are supposed to be striped Lions are supposed to have a mane Cars are supposed to have a radio Pit bulls are supposed to maul children Ants are supposed to be insects Canadian are supposed to be right-handed Rooms are supposed to be round All Normal All normal tigers are striped All normal lions have manes All normal cars have radios All normal pit bulls maul children All normal ants are insects All normal Canadians are right-handed All normal rooms are round Bare Plural Tigers are striped Lions have manes Cars have radios Pit bulls maul children Ants are insects Canadians are right-handed Rooms are round
Table 3. Mean turth-value ratings for predication types as a function of assertion type in Experiment 1 Predication type Majority characteristic Minority characteristic Majority statistical Striking Filler items Quasi-definition Majority false generalization Minority false generalization Aspect 2.40 1.48 0.56 0.75 2.44 -0.26 -0.50 By Virtue 2.25 1.28 0.02 0.28 2.71 -1.14 -1.77 Normativity 2.40 1.40 0.19 -0.25 2.57 -1.17 -1.78 All Normal 2.09 0.34 -0.10 -1.27 2.13 -1.67 -2.45 Bare Plural 2.48 1.83 1.23 1.27 2.62 -0.06 -0.76
Table 4. Summary of the ANOVA results for Experiment 1 (N = 137)

Participant analysis ANOVA Test 1. Omnibus Predication type F(3, 396) = 217.45 Assertion type F(4, 132) = 13.00 Interaction F(12, 396) = 6.85 2. Majority characteristic vs. majority statistical Predication type F(1, 132) = 358.56 Assertion type F(4, 132) = 5.25 Interaction F(4, 132) = 3.43 3. Minority characteristic vs. majority statistical Predication type F(1, 132) = 82.84 Assertion type F(4, 132) = 8.45 Interaction F(4, 132) = 2.86 4. Majority characteristic vs. minority characteristic Predication type F(1, 132) = 190.48 Assertion type F(4, 132) = 6.27 Interaction F(4, 132) = 5.79 5. Striking vs. majority characteristic Predication type F(1, 132) = 436.72 Assertion type F(4, 132) = 15.39 Interaction F(4, 132) = 13.66 6. Striking vs. majority statistical Predication type F(1, 132) = 7.86 Assertion type F(4, 132) = 12.99 Interaction F(4, 132) = 10.80 .006 F(1, 22) = 2.07 < .0001 F(4, 88) = 74.25 < .0001 F(4, 88) = 12.30 .164 min F'(1, 35) = 1.64 < .0001 min F'(4, 222) = 14.0 < .0001 min F'(4, 214) = 1.70 .208 < .0001 .0002 < .0001 F(1, 22) = 17.35 < .0001 F(4, 88) = 42.22 .026 F(4, 88) = 5.26 < .0001 F(1, 22) = 29.69 < .0001 F(4, 88) = 35.69 < .0001 F(4, 88) = 12.11 < .0001 F(1, 22) = 262.20 < .0001 F(4, 88) = 70.48 < .0001 F(4, 88) = 41.38 .0006 min F'(1, 32) = 14.34 < .0001 min F'(4, 179) = 7.04 .0007 min F'(4, 218) = 1.85 < .0001 min F'(1, 29) = 25.69 < .0001 min F'(4, 174) = 5.34 < .0001 min F'(4, 215) = 3.92 < .0001 min F'(1, 53) = 163.84 < .0001 min F'(4, 183) = 12.63 < .0001 min F'(4, 201) = 10.27 .0006 < .0001 .120 < .0001 .0004 .004 < .0001 < .0001 < .0001 < .0001 F(1, 22) = 135.12 .0006 F(4, 88) = 23.91 .01 F(4, 88) = 9.51 < .0001 min F'(1, 41) = 98.14 < .0001 min F'(4, 183) = 4.31 < .0001 min F'(4, 205) = 2.52 < .0001 .002 .042 < .0001 F(3, 44) = 61.91 < .0001 F(4, 176) = 105.44 < .0001 F(12, 176) = 14.16 < .0001 min F'(3, 72) = 48.19 < .0001 min F'(4, 165) = 11.57 < .0001 min F'(12, 571) = 4.62 < .0001 < .0001 < .0001 F1 p Item analysis F2 P Combined min F' p
Table 5. Mean judgments of naturalness (-3 to +3) for the question-answer pairs provided in Experiment 2 as a function of predication type. Predication type Majority characteristic Minority characteristic Majority statistical Striking Filler items Majority false generalization Naturalness judgment to question-answer pair 1.69 1.12 -0.31 -0.65 -1.44
Figure 1. Mean truth-value ratings (-3 to +3) as a function of assertion type for majority characteristic and majority statistical items in Experiment 1.
Figure 2. Mean truth value ratings (-3 to +3) as a function assertion of type for majority statistical and minority characteristic items in Experiment 1.
Figure 3. Mean truth value ratings (-3 to +3) as a function of assertion type for majority characteristic and striking items in Experiment 1.
Figure 4. Mean truth-value ratings (-3 to +3) as a function of assertion type for majority statistical and striking items in Experiment 1.
Appendix A: Norming study and materials for Experiments 1, 2, and 3

A total of 84 kind-predicate pairs (tigers-striped) were needed for the studies, 12 for each of the seven types of predication used in the experiments as experimental and filler items. We sought to hold cue validity, the probability that an item is part of a category given a particular cue, constant for the five types of predication. We further sought to balance levels of agreement and estimates of prevalence for several of the predicate types. To do this, we conducted a norming study (N = 62) in which participants evaluated 210 different items of varying predication types. Participants were randomly assigned to perform one of three tasks. Participants who performed Task 1 (the agreement task) were asked to judge the truth of a statement in bare plural form (e.g., Tigers are striped). They registered their answer on a 7-point scale ranging from +3 (definitely true) to -3 (definitely false). Participants who performed Task 2 provided prevalence estimates, i.e., they provided an integer from 1-100 to the question, What percentage of tigers are striped? The instructions emphasized that the estimates should reveal the percentage of individuals of the kind that have the property (individual tigers that have stripes) rather than the percentage of kinds of tigers (e.g. Bengal tigers, White tigers) that have the property. Participants were also told that they should interpret the kind terms as referring to all instances of the kind, rather than just a subset if a term (e.g. lion) could be used in either way. Task 3 asked participants to judge, i.e., they were told, Suppose x has stripes. How likely is it that x is a tiger? Participants registered their answer on a 7-point scale ranging from +3 (very likely) to -3 (very unlikely). Of the 210 items that were evaluated, we chose 84 items based on the factors outlined above. Those items are provided in Table A, along with their corresponding agreement ratings, cue validity estimates, and prevalence estimates.
59
Table A. Truth judgments (+3 to -3), cue validity estimates (+3 to -3), and prevalence estimates (0-100%) for materials used in Experiments 1, 2, and 3 as a function of predicate type Predication type Majority characteristic Item airplanes have wings scissors cut ambulances have sirens needles are sharp tables are flat diapers are absorbent horses have four legs birds have wings lemons are sour leopards have spots dogs have tails feathers are light sheep produce milk lions have manes deer have antlers ducks lay eggs elk have antlers cows have udders moose have antlers pigs suckle their young goats have horns snakes lay eggs kangaroos have pouches peacocks have beautiful tails Kind artifact artifact artifact artifact artifact artifact natural natural natural natural natural natural Mean natural natural natural natural natural natural natural natural natural natural natural natural Truth judgment 2.91 2.67 2.50 2.42 2.29 2.65 2.54 2.63 2.25 2.17 2.42 2.29 2.48 2.00 1.46 1.09 2.04 1.67 2.04 1.79 1.54 1.42 2.00 2.33 1.91 Cue validity estimate 0.11 1.00 1.47 0.42 0.05 0.42 -0.11 1.58 1.06 0.68 -0.21 0.44 0.58 -0.06 0.89 1.05 0.58 0.42 1.33 0.67 -0.53 0.11 0.32 1.26 1.32 Prevalence estimate 95.32 91.89 89.11 88.74 87.95 87.58 94.58 94.11 91.42 88.32 87.16 86.16 90.19 51.89 53.63 54.21 56.53 57.00 57.74 60.21 60.95 62.53 63.68 64.89 64.63
Majority statistical
trumpets are loud fire trucks are red cars have radios dolls wear dresses diapers are white rocking chairs are wooden cats like milk dogs bark at strangers eggshells are white summers are humid winters are snowy raccoons eat garbage lead toys poison children knives cut people rusty nails cause tetanus car crashes kill people guns kill people plastic bags suffocate small children rip tides drown swimmers mosquitoes carry malaria loud noises deafen people ticks carry Lyme disease snow storms shut down schools strokes kill people cats are animals mushrooms are fungi rectangles are geometric figures even numbers are divisible by 2
Striking
Quasi-definition
Mean artifact artifact artifact artifact artifact artifact natural natural natural natural natural natural Mean artifact artifact artifact artifact artifact artifact natural natural natural natural natural natural Mean natural natural natural natural
1.77 1.38 1.92 1.46 1.04 0.83 0.83 2.21 1.75 1.75 1.21 1.29 1.13 1.40 1.79 1.63 1.50 1.48 1.38 1.38 1.67 1.63 1.46 1.42 1.39 1.25 1.50 2.42 2.46 2.54 2.21
0.61 -0.11 0.84 0.79 0.47 -0.79 -0.26 0.67 2.37 0.21 1.16 1.58 0.53 0.62 0.89 0.78 1.26 0.21 0.37 0.89 1.26 2.05 1.37 1.37 1.22 -0.05 0.97 0.17 1.47 0.72 2.58
58.99 79.47 75.47 74.74 68.74 78.79 71.21 74.21 62.05 59.79 58.89 57.58 54.74 67.97 43.42 34.47 33.32 34.00 33.32 25.89 21.21 25.63 32.32 46.00 33.63 34.32 33.13 91.74 90.11 89.84 89.05
kangaroos are marsupials ants are insects dogs are mammals sows are pigs elms are trees bachelors are unmarried preschoolers cannot vote US Presidents are over 35 Majority false generalization books are paperbacks computers are PCs bees are worker bees ducks are female humans are over three years old lions are male mammals are placental mammals trees are deciduous trees Canadians are right-handed engineers are male teachers are female Americans are brunettes cars are yellow houses are mansions novels are mystery novels restaurants are Chinese restaurants rooms are round tables are 10 feet long cats are white dogs are beagles mammals are hamsters
Minority false generalization
natural natural natural natural natural social social social Mean artifact artifact natural natural natural natural Natural Natural Social Social Social Social Mean Artifact Artifact Artifact Artifact Artifact Artifact Natural Natural Natural
2.33 2.83 2.00 1.74 2.54 2.63 2.13 2.74 2.38 0.21 1.54 0.96 -0.58 0.00 -0.04 0.92 0.50 -0.50 0.04 -0.26 -0.21 0.21 -0.50 -0.21 0.04 -0.91 -0.58 -0.67 -0.67 -0.88 -0.54
1.37 0.72 -0.06 0.58 -0.05 1.32 0.26 -0.68 0.70 1.68 2.58 1.72 -0.74 0.32 -0.16 1.89 1.84 -0.95 -0.42 0.16 0.21 0.68 -0.21 2.05 1.74 2.42 -1.21 0.05 -0.26 2.00 1.47
86.26 85.63 80.79 80.37 78.42 84.11 91.58 87.26 86.26 59.84 64.37 81.84 49.26 82.53 57.11 69.21 45.79 74.89 72.79 65.21 49.84 64.39 22.84 15.63 29.95 27.42 19.16 19.63 20.63 21.37 15.47
plants are ferns tigers are albino trees are palm trees
Natural Natural Natural Mean
0.04 -0.88 -0.48 -0.52
2.00 -0.37 1.94 0.97
17.37 16.84 13.79 20.01
Appendix B: Summary of the planned t-tests for Experiment 1 (N = 137)

Participant analysis Assertion type 1. Majority characteristic vs. majority statistical Bare plural Aspect All normal By virtue Normative 2. Minority characteristic vs. majority statistical Bare plural Aspect All normal By virtue Normative 3. Majority characteristic vs. minority characteristic Bare plural Aspect All normal By virtue Normative 4. Striking vs. majority characteristic Bare plural Mean 1
Item analysis p t2 p
Mean 2
t1
2.48 2.40 2.09 2.25 2.40

1.23 .57 -.10 .02 .19

t(54) = 7.25 t(50) = 6.44 t(54) = 7.26 t(54) = 8.46 t(52) = 7.87
< .0001 < .0001 < .0001 < .0001 < .0001
t(22) = 9.82 t(22) = 9.04 t(22) = 8.12 t(22) = 12.00 t(22) = 9.92
< .0001 < .0001 < .0001 < .0001 < .0001
1.83 1.48 .34 1.28 1.40

1.23 .57 -.10 .02 .19

t(54) = 3.02 t(50) = 2.83 t(54) = 1.29 t(54) = 4.60 t(52) = 3.72
.019 .033 > .50 .0001 .003
t(22) = 4.62 t(22) = 3.54 t(22) = 1.28 t(22) = 5.85 t(22) = 4.32
.0007 .009 > .50 < .0001 .001
2.48 2.40 2.09 2.25 2.40

Striking
1.83 1.48 .34 1.28 1.40

t(54) = 3.95 t(50) = 3.99 t(54) = 6.16 t(54) = 3.86 t(52) = 4.52 t(54) = 6.39
.001 .001 < .0001 < .002 .0002 < .0001
t(22) = 4.74 t(22) = 4.24 t(22) = 6.40 t(22) = 4.80 t(22) = 4.02 t(22) = 9.84
.0005 .002 < .0001 .0004 .003 < .0001
1.27
2.48
Aspect All normal By virtue Normative 5. Striking vs. majority statistical Bare plural Aspect All normal By virtue Normative
.76 -1.27 .28 -.25

Striking
2.40 2.09 2.25 2.40

t(50) = 5.63 t(54) = 12.65 t(54) = 7.25 t(52) = 9.56 t(54) = .19 t(50) = .53 t(54) = 3.59 t(54) = .86 t(52) = 1.20
< .0001 < .0001 < .0001 < .0001 > .50 > .50 .004 > .50 > .50
t(22) = 9.88 t(22) = 15.18 t(22) = 11.87 t(22) = 13.25 t(22) = .37 t(22) = .85 t(22) = 3.88 t(22) = 1.39 t(22) = 1.85
< .0001 < .0001 < .0001 < .0001 > .50 > .50 .004 > .50 .39
1.27 .76 -1.27 .28 -.25
1.23 .57 -.10 .02 .19
65
Endnotes
i
Whether and how such generics constitute a class of knowledge or of sentences is a
theory-dependent question. In calling them classes, we are pointing to sets of generics that share some salient properties and that may plausibly constitute a class, but what is important here is that they do not obviously involve principled or statistical connections.
ii
Only natural kind items were included as these are the types of examples which have
been used in the literature on generics and it is unclear whether it is possible to have minority characteristic generics in the domain of artifacts. Though this is an important question for future research, we leave it aside, as for current purposes, we wanted to include only those items which clearly appear to involve principled connections despite being true of only a minority of instances.
iii
An important question for future research is to determine exactly why participants had
a tendency to overestimate the prevalence of the minority characteristic items. One intriguing possibility is that the over-estimations are a consequence of the knowledge being represented via principled connections.
iv
We ran a fully within-participants replication and found similar results. However we
found some evidence that judgments for a given assertion type showed interference effects from judgments of other assertion types.
v
The low ratingsthat minority characteristic items received when judging the All normal
xs are p assertions rule out the possibility that participants were interpreting the assertions involving minority characteristic items as quantifying over kinds of kinds (e.g. interpreting All normal ducks lay eggs as All normal kinds of ducks lay eggs) or limiting their interpretation of the noun to the relevant subset of the kind (e.g. interpreting lion in
All normal lions have manes to male lions). Had participants used either of these interpretive strategies, they would have rated the items highly on the all normal measure; however, this is not what we found.
vi
It should be noted that, although we have often used inanimate examples for majority
statistical generics and animate examples for striking generics, our stimuli for each category included roughly even numbers of animate and inanimate examples (see Appendix A).
vii
We should emphasize that we are not proposing that the existence of a causal
connection between a kind and a property is itself sufficient for the generic to be accepted, rather that this class of generics namely striking generics involve such a causal connection.
Cultural transmission of social essentialism

Marjorie Rhodesa,1, Sarah-Jane Leslieb, and Christina M. Tworeka
a
Department of Psychology, New York University, New York, NY 10003; and bDepartment of Philosophy, Princeton University, Princeton, NJ 08544
Edited* by Douglas L. Medin, Northwestern University, Evanston, IL, and approved July 2, 2012 (received for review May 25, 2012)
cognitive development generic language social cognition social categorization conceptual development
n his seminal work on the psychological foundations of social prejudice, Gordon Allport noted that . . . a belief in essence develops. There is an inherent Jewishness in every Jew. The soul of the Oriental, Negro blood, . . . the passionate Latin all represent a belief in essence. A mysterious mana (for good or ill) resides in a group, all of its members partaking thereof (1, p 174). Psychological essentialism is a pervasive cognitive bias that leads people to view members of a category as sharing a deep, underlying, inherent nature (a category essence), which causes them to be fundamentally similar to one another in both obvious and nonobvious ways (2, 3). Numerous previous studies have documented essentialist beliefs about social categories (e.g., gender, race) from the preschool years through adulthood (48); however, to date no research has examined the processes underlying the development of these beliefs. The question of how a belief in essence develops was the focus of our present studies. As a pervasive cognitive bias, psychological essentialism shapes how people think about many types of categories. Psychological essentialist beliefs have been studied most often in the context of biological categories (e.g., tigers) (2, 911) rather than social categories, such as those described by Allport. For biological categories, psychological essentialism facilitates learning and knowledge acquisition. For example, viewing category members as fundamentally alike allows a child to infer that if one tiger is ferocious, then other tigers will be too, even if the individual tigers look different from each other (e.g., orange and white tigers) (12). Similarly, viewing category-linked properties as arising from an underlying nature allows children to infer that a baby tiger will inevitably grow up to be ferocious, even if it does not appear ferocious at birth (9, 10). When applied to social categories, psychological essentialism can have pernicious consequences, however. As suggested by Allports observations, essentialist beliefs about social categories (hereinafter referred to as social essentialism) facilitate social stereotyping and prejudice (1318). For example, social essentialism facilitates the belief that because one girl is bad at math, girls in general will be bad at math, or that because one member of a racial group commits a criminal act, the group must share a criminal nature (13, 16). Two aspects of when social essentialism arises provide insight into the processes possibly underlying its development. First,
www.pnas.org/cgi/doi/10.1073/pnas.1208951109
Author contributions: M.R. and S.J.L. designed research; C.M.T. performed research; M.R. analyzed data; and M.R. and S.J.L. wrote the paper. The authors declare no conict of interest. *This Direct Submission article had a prearranged editor.
1
To whom correspondence should be addressed. E-mail: marjorie.rhodes@nyu.edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1208951109/-/DCSupplemental.
PNAS Early Edition | 1 of 6
PSYCHOLOGICAL AND COGNITIVE SCIENCES
Social essentialism entails the belief that certain social categories (e.g., gender, race) mark fundamentally distinct kinds of people. Essentialist beliefs have pernicious consequences, supporting social stereotyping and contributing to prejudice. How does social essentialism develop? In the studies reported here, we tested the hypothesis that generic language facilitates the cultural transmission of social essentialism. Two studies found that hearing generic language about a novel social category diverse for race, ethnicity, age, and sex led 4-y-olds and adults to develop essentialist beliefs about that social category. A third study documented that experimentally inducing parents to hold essentialist beliefs about a novel social category led them to produce more generic language when discussing the category with their children. Thus, generic language facilitates the transmission of essentialist beliefs about social categories from parents to children.
essentialist beliefs about both biological and social categories have been found early in development (by age 4 y) in every cultural context studied to date, including in both rural and urban communities within the United States (4, 5, 7, 19), among Jewish and Arab children in Israel (8, 2022), in a small shing community in Madagascar (6), and among children in Brazil (23). However, whereas young children show essentialist beliefs about all basic animal categories (e.g., tigers, robins, lizards), they hold essentialist beliefs about only a small subset of the social categories with which they are familiar. For example, European American 5-y-olds from several cultural communities within the United States have been found to have essentialist beliefs about gender categories, but not racial categories (19). Thus, psychological essentialism emerges early in development, but young children view only a small subset of social categories in an essentialist manner (24). Second, across development, substantial cultural variation emerges in which social categories invoke essentialist beliefs. For example, by age 710 y, children growing up in more politically conservative communities in the United States have more essentialist beliefs about race compared with children growing up in more politically liberal communities (19). In India, adults from upper social classes view class-based groups in essentialist terms, but adults from lower classes do not (25, 26). Also, within Israel, essentialist beliefs about ethnicity are more common among older children in religious communities than among older children in secular communities (21). Thus, rudimentary social essentialist beliefs emerge early in development in every cultural context studied to date, yet there is substantial cultural variation in which social categories are viewed in this manner. This pattern suggests that the development of social essentialism results from an interplay between early emerging cognitive biases and cultural inputin particular, that cultural input guides how children map general essentialist biases onto particular categories in their environment. Such a process would explain why some form of social essentialism emerges early in development, but also why children hold essentialist beliefs about only some of the social categories with which they are familiar (those categories for which they receive the requisite form of input), why there is cultural variation in these categories (because input is given about different categories in different communities), and why there is often a lengthy developmental trajectory for social essentialist beliefs (if it takes time to receive a sufcient amount of the input). In these studies, we examined the nature of the cultural input that shapes the development of social essentialism. We tested the role of a particular form of cultural input, namely, generic language (e.g., boys play with trucks, a girl wears pink) (27). Generic statements describe a kind or category in general, rather than some particular members. For example, boys play with trucks (a bare plural generic sentence)
does not refer to a particular boy and need not be true of all boys, but instead reects a belief about the category in general (14). Similarly, a girl wears pink (an indenite singular generic) does not refer to any particular girl, but rather expresses something about girls in general. These statements can be contrasted with nongeneric sentences, such as some boys play with trucks or this girl wears pink, which refer to a subset and an individual, respectively. Previous research has suggested a causal link between hearing generic language and forming essentialist beliefs about animal categories (28, 29). However, children rapidly develop essentialist beliefs about animal categories even in the absence of generic language (2, 30); thus, generic language may have only a weak facilitative effect on the development of essentialism. In contrast, social essentialism develops more slowly, more selectively, and with cultural variation, suggesting that cultural input plays a much more important role. We tested whether generic language plays a powerful role in shaping the development of social essentialism by guiding children to develop essentialist beliefs about social categories that they would not otherwise view in an essentialist manner. Why might generic language elicit social essentialism? Generic language is usually understood as communicating nonaccidental generalizations. Thus, when children hear generic language describing a new property of a familiar category, they assume that there is a kind-based, causal explanation of why the members have the property. For example, on hearing that butteries have dust on their wings, children offer kind-based explanations (e.g., butteries need the dust so they can y) (31, 32). Alternately, on hearing the nongeneric description this buttery has dust on her wings, children assume that the cause is incidental (e.g., she ew through a dusty room). We hypothesized that hearing generic language about a novel category would lead children to infer that it is the sort of cohesive, natural kind that supports such explanations. Accordingly, we tested whether hearing generic language about a novel social category leads children to develop essentialist beliefs about the category. For generic language to facilitate the cultural transmission of social essentialism, children also must hear more generic language for categories for which adults in their communities hold essentialist beliefs. Indeed, we further hypothesized that parents would be more likely to produce generic language to describe categories that they themselves view as supporting the kind of category-based explanations described above (33). Thus, we also tested whether holding essentialist beliefs about a social category leads parents to produce more generic language describing the category when talking to their children. Indeed, parents do produce generic language for at least some social categories (gender) (34); yet whether this language causes children to develop essentialist beliefs as they learn about those categories remains unknown. Further, whether parents selectively produce generic language for categories for which they themselves hold essentialist beliefs has not yet been examined. [There is evidence suggesting that parents and children produce more generic language for animals than for artifacts (33, 43, 44), which could be due to domain differences in essentialism. However, because there are many differences in the structure of animal and artifact categories, these studies cannot provide denitive evidence of the role of essentialism in the production of generics. Furthermore, they do not speak to the role of essentialism in the production of generics for social kinds.] Thus, whether generic language transmits selective, social essentialist beliefs across generations remains an open question. The present studies tested whether two complementary processes underlie the cultural transmission of social essentialism: (i ) that parents produce more generic language when they hold essentialist beliefs about a social category, and (ii ) that hearing
2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1208951109
more generic language leads children to develop essentialist beliefs about the category. We examined these questions over the course of three studies. The rst two studies tested the effects of generic language on the formation of social essentialist beliefs, and the third study tested the effects of essentialist beliefs on parents production of generic language in parentchild conversation. The rst two studies examined both adult and 4-y-old participants. The 4-y-olds were selected because this is the age at which social essentialism begins to emerge and thus the age at which it is most crucial to examine the processes underlying its formation. Adults were included to test for a developmental windowwhether sensitivity to generic language is found only in young children or continues into adulthood. Given that learning about the social world continues over a lifetimeand new social categories may be encountered for the rst time in adulthoodwe predicted that adults would also be sensitive to generic language. The third study examined parentchild pairs, focusing on the parents of preschool-age children. Together, these three studies provide a strong test of whether generic language can serve as a mechanism by which social essentialist beliefs are culturally transmitted. Results essentialism, we introduced children and adults to a novel category of peopleZarpiesvia an illustrated storybook, as in previous work on animal categories (28). Each page presented a picture of a single person displaying a unique physical or behavioral property. The characters were diverse with respect to sex, race, and age; thus, the novel category cut across groupings for which people might already have essentialist beliefs. For example, if all of the Zarpies were Asian, subjects might apply essentialist beliefs to the group because they generally have essentialist beliefs about race. Because the novel group is so diverse, it would initially appear arbitrary (35); thus, levels of essentialism in the absence of generic language should be low (as conrmed by comparison conditions). A single line of text describing the depicted property accompanied each page using the language specied by the participants condition: generic (e.g., Look at this Zarpie! Zarpies are scared of ladybugs) or one of two comparison conditions (specic, e.g., Look at this Zarpie! This Zarpie is scared of ladybugs! or no label, e.g., Look at this one! This one is scared of ladybugs). Participants were randomly assigned to one of these three conditions. In study 1a, adult participants read the storybook twice before completing the test questions. In study 1b, a trained experimenter read 4-y-old participants the storybook four times over the course of two research sessions approximately 3 d apart. Children completed the test questions in a third research session approximately 3 d later. The test questions comprised multiple measures of essentialist beliefs, which assessed the extent to which participants (i ) expect properties associated with the new category to be innate and inevitable (inheritance items), (ii ) expect properties attributed to a single category member to extend to other category members (induction items), and (iii ) view category membership as causing/explaining the development of typical properties (explanation items). In study 1a, adults gave more essentialist responses in the generic condition compared with the specic and no-label conditions (P < 0.001), and more essentialist responses in the specic condition compared with the no-label condition (P = 0.001) [main effect condition; Wald 2 (2) = 169.90, P < 0.001] (Fig. 1). The generic condition increased the odds of an essentialist response by 6.07 [Wald 95% condence interval (CI) = 4.56, 8.08] relative to the no-label condition. In study 1b, children gave more essentialist responses in the generic condition than in the specic condition (P = 0.002) or no-label condition (P < 0.001), but responses to the specic and no-label conditions did not
Rhodes et al.
Study 1. To test whether hearing generic language induces social
conditions did not differ from one another [condition, Wald 2 (2) = 15.55, P < 0.001] (Fig. 2). The generic bare plural condition increased the odds of an essentialist response by 1.94 (95% CI = 1.35, 2.79) relative to the specic condition; the generic indenite singular condition increased the odds of an essentialist response by 1.85 (95% CI = 1.29, 2.65). Thus, study 2 conrms that it is generic language per senot simply the use of plural noun phrasesthat underlies the formation of essentialist beliefs. Further, whereas study 1 provided children with extended exposure to generic language over time, study 2 conrms that hearing generic language leads to the rapid formation of essentialist beliefs among 4-y-olds.
Study 3. Studies 1 and 2 show that hearing generic language elicits social essentialism. For generic language to serve as a mechanism by which communities transmit essentialist beliefs, then people must also selectively produce generic language when they have essentialist beliefs about a social category. Study 3 tested this prediction with an experimental study of parentchild interactions. First, parents were introduced to the category Zarpies via a paragraph that led them to hold essentialist beliefs about Zarpies (by describing Zarpies as a distinct kind of people with many biological and cultural differences from other social groups) or nonessentialist beliefs about Zarpies (by describing Zarpies as a nondistinct kind of people, with many biological and cultural similarities to other populations). A pilot study with a separate sample of adult participants (n = 20) conrmed that the essentialist paragraph (mean = 0.76, 95% CI = 0.68, 0.82) elicited more essentialist beliefs about Zarpies (as indicated by scores on the inheritance and induction items used in studies 1 and 2) than the nonessentialist paragraph [mean = 0.57, 95% CI = 0.49, 0.65, Wald 2 (1) = 10.30, P = 0.001]. After reading the introductory paragraph, parents received a picture book containing the illustrations used in studies 1 and 2, with no accompanying text. They were asked to talk through the picture book with their child and describe the people and events depicted, just as they would a picture book at home. No other instructions were provided. The entire parentchild conversation was videotaped and transcribed. References to the characters in the story were coded as generics (e.g., Zarpies are scared of ladybugs!), Zarpie category labels (e.g., This Zarpie is scared of the ladybug!), pronouns (e.g., She is scared of the ladybug!), social category labels other than Zarpie (e.g., This woman is scared of the ladybug!), or as universally quantied (e.g., All Zarpies are scared of ladybugs!). The total number of parent utterances did not differ by condition [essentialist, mean = 117.67 68.42; nonessentialist, mean = 119.30 37.91; t (17) = 0.06, P = not signicant], and neither did the number of utterances referencing the characters
Fig. 1. Probabilities of essentialist responses by condition for study 1a (adults) and study 1b (children). Error bars represent Wald 95% CIs.
Study 2. Study 2 tested another form of generic language, indenite singular generic sentences (e.g., A Zarpie sleeps in tall trees). Like bare plural generic sentences (the form of generic language used in study 1), indenite singulars are about kinds and categories, not individuals; for example, a cow says moo asserts something about cows in general, not about one specic cow (27, 36). However, unlike bare plural noun phrases, indenite singular noun phrases are grammatically singular. In the two comparison conditions of study 1 (specic and no-label), the subjects of the sentences were singular noun phrases, whereas in the generic condition, they were plural noun phrases. Thus, based on the ndings of study 1, it is possible that it is not generic language per se, but simply the use of plural noun phrases, that leads to essentialism. Testing indenite singulars allows us to control for singularity/plurality across the conditions, to provide a more stringent test of whether it is generic language per se that induces essentialism. This important issue has not been addressed in any previous work on the role of generic language in the development of essentialism, including work on the case of animal categories. In study 2a, adults were randomly assigned to either the generic condition (indenite singular; e.g., A Zarpie sleeps in tall tress) or the specic condition (e.g., This Zarpie sleeps in tall trees). For study 2b, childrenlike adultscompleted the entire study in a single research session. An experimenter read the assigned condition two times and then immediately asked the test questions. Because of this procedural change, we also included a bare plural generic condition, to replicate the ndings from study 1b. In this way, study 2b also allowed us to assess whether children need extended generic input (multiple sessions over the course of a week) to essentialize, or whether these effects emerge more rapidly. Thus, children were randomly assigned to one of three conditions: generic bare plural (e.g., Zarpies sleep in tall tress), generic indenite singular (e.g., A Zarpie sleeps in tall trees) or specic (e.g., This Zarpie sleeps in tall trees). Adults in study 2a gave more essentialist responses in the generic indenite singular condition than in the specic condition [Wald 2 (1) = 41.17, P < 0.01] (Fig. 2). The generic condition increased the odds of an essentialist response by 2.45 (95% CI = 1.86, 3.26). Children in study 2b gave more essentialist responses in both of the generic conditions than in the specic condition (P < 0.001), whereas the two generic
Rhodes et al.
Fig. 2. Probabilities of essentialist responses by condition for study 2a (adults) and study 2b (children). Error bars represent Wald 95% CIs.
differ [condition, Wald 2 (2) = 25.69, P < 0.001] (Fig. 1). The generic condition increased the odds of an essentialist response by 2.21 (95% CI = 1.62, 3.03) relative to the no-label condition. Thus, study 1 demonstrates that hearing generic language about a novel, diverse social category led to the formation of essentialist beliefs about that category among both 4-y-olds and adults.
Table 1. Percentage of character references tting each code by condition, study 3

Essentialist Generic Label Pronoun Other category Quantied 14.32 10.93 61.99 10.93 1.84 (2.56) (3.97) (5.14) (1.91) (0.89) Nonessentialist 6.04 8.56 68.99 15.30 1.11 (1.78)* (4.70) (5.72) (4.16) (0.62)
Numbers in parentheses are means +/ SE. *Different from essentialist condition, P < 0.05.
in the pictures [essentialist, mean 67.67 30.97; nonessentialist, mean = 66.80 22.32; t (17) = 0.07, P = not signicant]. However, a higher percentage of the character references were generic in the essentialist condition compared with the nonessentialist condition [t (17) = 2.15, P < 0.05] (Table 1). No other type of character reference differed by condition. Thus, as predicted, parents produced more generic language when they were induced to hold essentialist beliefs. We also coded the content of the conversation for evaluative comments (i.e., statements that a character was doing something positive, e.g., theyre cool, right? or negative, e.g., thats yucky). Parents produced more negative evaluations in the essentialist condition than in the nonessentialist condition [mean, 2.55 0.67 vs. 0.70 0.34; t (17) = 2.56, P = 0.02]. Positive evaluations did not differ by condition (essentialist, mean = 0.22 0.15; nonessentialist, mean = 0.10 0.10). Discussion These three studies provide strong evidence that generic language is a mechanism by which social essentialist beliefs can be transmitted from parents to children. As shown in studies 1 and 2, hearing generic language about a novel social category led both preschool-age children and adults in our samples to develop essentialist beliefs about the category. The effect of generic language was not simply a function of the plurality of the subject noun phrase, because both bare plural generics (e.g., Zarpies are scared of ladybugs) and indenite singular generics (e.g., A Zarpie is scared of ladybugs) induced essentialism. Furthermore, in the children, we found both immediate effects (when the input and testing occurred within a single session in study 2) and effects that persisted over time (when children were exposed to generic input over the course of several days and tested 3 d later in study 1). Thus, the effects of generic language appear to be powerful and robust. These ndings are particularly striking because the novel category was diverse with respect to race, ethnicity, age, and sex; did not map onto any preexisting social category for which people might already hold essentialist beliefs (24); and thus initially appeared arbitrary. Furthermore, the properties used in the test questions were very unusual ones that people are not normally expected to possess (e.g., disliking ice cream, eating owers). However, essentialist responses involved projecting these properties. For example, the inheritance items asked participants to consider a baby born to a Zarpie mom who likes to eat owers but raised by a non-Zarpie mom who likes to eat crackers. To evidence essentialist beliefs on this question, participants had to overcome the general expectation that people prefer to eat crackers rather than owers. As would be expected, then, given both the diversity of the category and the unusualness of the properties, baseline levels of essentialism (as shown in the comparison conditions of studies 1 and 2) were very low. However, fairly minimal exposure to generic language (several readings of a storybook) approximately doubled the essentialist responses among children, and had even larger effects in adults. Thus, accumulating exposure to generic
language over time, as children learn about familiar social categories in their everyday environment (34), could perhaps lead to even more dramatic increases in social essentialism. Indeed, the effect of generic language appeared to be stronger among children in study 1 (where children received more input across several sessions) than in study 2 (where children received less input in a single session), suggesting that the effects of generic language may increase over time with increasing input. In study 3, the parents were more likely to produce generic language when speaking with their children about a social category for which they themselves held essentialist beliefs. To experimentally test the causal inuence of essentialism on the production of generics, we induced parents essentialist beliefs by describing Zarpies as having different biological and cultural properties from other social groups; nothing was said about whether Zarpies were similar to one another. Furthermore, parents were never shown Zarpies sharing any properties; like the participants in studies 1 and 2, they saw only one Zarpie at a time in the picture book, each of which displayed a unique property (e.g., one Zarpie was shown ipping in the air, the next sleeping in a tree, and so on). However, despite the fact that no within-category similarities were described in the essentialist prime or shown in the picture book, inducing essentialism led parents to produce generic sentences when talking about Zarpies with their children (e.g., generating such sentences as Zarpies sleep in tall trees). These data do not show that generic language creates essentialist thought. Essentialist beliefs (e.g., that categories are innate, coherent, and causally powerful) go far beyond any content that is explicitly communicated by generic language, and essentialism can emerge in the absence of generic language, such as in the case of animal kinds (30). Rather, social essentialism appears to result from the interplay of cognitive biases and cultural input; childrens cognitive biases lead them to assume that some or other social categories reect essential kinds, and generic language signals to them to which categories they should apply these beliefs. These studies indicate that generic language can facilitate the transmission of social essentialism from one generation to another. To determine whether this mechanism accounts for crosscultural variation in social essentialism, future work will need to test whether the effects obtained in the populations studied here (which consisted primarily of educated middle-class families in a diverse, urban environment) extend to other cultural communities. Certainly other cultural factorsfor example, the degree to which a category serves as the basis of segregation or differential treatment (20, 22, 37)could also play roles in shaping the development of social essentialism. Given the pernicious consequences of social essentialism, elucidating the processes underlying its development is a critical goal. Indeed, in study 3, parents who were induced to hold essentialist beliefs about Zarpies were more likely to produce negative evaluative statements about them. Thus, these data suggest that inducing essentialism may contribute to negative social attitudes (13, 16, 3840). Understanding the mechanisms that underlie the development of social essentialism could provide guidance on how to disrupt these processes, and thus perhaps on how to
Fig. 3.
Sample Illustrations.
Rhodes et al.
reduce stereotyping and prejudice. As a society, we often change the way we speak about a given social group (e.g., when a label comes politically incorrect ); grounding these changes in mechanisms that have been empirically shown to inuence the formation of essentialist beliefs could lead to more effective efforts to reduce societal prejudice. Materials and Methods
Participants. Adults in study 1a (53 females, 8 males) and study 2a (22 females, 18 males) were undergraduate students at New York University who volunteered to participate. Children in study 1b (24 males, 22 females; median age, 4.70 y; age range, 4.05.5 y) were recruited from private preschools across New York City. Children in study 2b (21 males, 21 females; mean age, 4.81 y; age range, 4.505.36 y) and families in study 3 (19 parents, 9 males and 10 females; 19 preschool-age children, 7 males and 12 females; mean age, 4.2 y; age range, 3.05.44 y) were recruited from the Childrens Museum of Manhattan. The sample was 64% European American, 9% African American, 21% Asian American, and 6% Hispanic. All participants spoke English as their rst language. All study procedures were approved by the Institutional Review Board of New York University. Written informed consent was obtained from adult participants and parents of participating children; children provided oral assent. Materials. The picture book consisted of 16 illustrated pages (identical across conditions and studies; Fig. 3 presents sample items). Each page presented a picture of a single person displaying a unique physical or behavioral property (e.g., This Zarpie. . .chases shadows, loves to eat owers, has striped hair, sleeps in tall trees). The characters were diverse with respect to sex (half male, half female), race/ethnicity (four white, four black, four Latino, and four Asian), and age (four young children, four older children, four adults, and four older adults). To allow for a visual identier of category membership, the characters had a category-typical clothing style, but no single clothing feature was a necessary or dening feature of category membership. In studies 1 and 2, a statement describing the characters action, using the form of language specied by the condition, accompanied each illustration (Tables S1 and S2). In study 3, no text accompanied the illustrations. For study 3, materials also included introductory paragraphs to induce essentialist or nonessentialist beliefs about Zarpies (SI Text). Procedure. In studies 1a and 2a, adults were given a printed copy of the picture book and asked to read it twice. They were then given a booklet containing all of the test questions (with accompanying illustrations) and asked to complete the test items on their own. The test items included three
1. Allport GW (1954) The Nature of Prejudice (Addison-Wesley, Oxford), p 174. 2. Gelman SA (2003) The Essential Child: Origins of Essentialism in Everyday Thought (Oxford Univ Press, New York). 3. Medin DL, Ortony A (1989) Similarity and Analogical Reasoning, eds Vosniadou S, Ortony A (Cambridge Univ Press, New York), pp 179195. 4. Hirschfeld LA (1995) Do children have a theory of race? Cognition 54:209252. 5. Taylor MG, Rhodes M, Gelman SA (2009) Boys will be boys; cows will be cows: Childrens essentialist reasoning about gender categories and animal species. Child Dev 80:461481. 6. Astuti R, Solomon GEA, Carey S (2004) Constraints on conceptual development: A case study of the acquisition of folkbiological and folksociological knowledge in Madagascar. Monogr Soc Res Child Dev 69:1135, viiviii, discussion 136161. 7. Waxman SR (2010) Names will never hurt me? Naming and the development of racial and gender categories in preschool-aged children. Eur J Soc Psychol 40:593610. 8. Diesendruck G, HaLevi H (2006) The role of language, appearance, and culture in childrens social category-based induction. Child Dev 77:539553. 9. Waxman S, Medin D, Ross N (2007) Folkbiological reasoning from a cross-cultural developmental perspective: Early essentialist notions are shaped by cultural beliefs. Dev Psychol 43:294308. 10. Gelman SA, Wellman HM (1991) Insides and essences: Early understandings of the non-obvious. Cognition 38:213244. 11. Medin DL, Atran S (2004) The native mind: Biological categorization and reasoning in development and across cultures. Psychol Rev 111:960983. 12. Gelman SA, Markman EM (1987) Young childrens inductions from natural kinds: The role of categories and appearances. Child Dev 58:15321541. 13. Leslie SJ The original sin of cognition: Fear, prejudice, and generalization. J Philos, in press. 14. Leslie SJ (2008) Generics: Cognition and acquisition. Philos Rev 117:147. 15. Prentice DA, Miller DT (2007) Psychological essentialism of human categories. Curr Dir Psychol Sci 16:202206. 16. Haslam N, Rothschild L, Ernst D (2002) Are essentialist beliefs associated with prejudice? Br J Soc Psychol 41:87100.
measures of essentialist beliefs (SI Text), all modeled on previous work (28): explanations (four items) (28, 31, 32), inheritance (three items) (4, 5, 9, 10, 41), and induction (six items) (12). Identical materials were used in studies 1b and 2b, but the stories and test questions were presented verbally to children in individual research sessions. The experimenter recorded the childrens responses on an answer form. Child sessions were also videotaped. A secondary coder coded all videos for childrens responses. The percent agreement between the live coder and video coder was 95%, with disagreements resolved by the rst author. In studies 1a and 1b, after the test items, participants completed an assessment of their memory for the stories; analyses of these data conrmed that memory could not account for the obtained pattern of condition differences in essentialism (SI Text). For study 3, sessions were videotaped, and parentchild conversations were transcribed verbatim. The rst and second authors, blinded to condition, coded each utterance according to the scheme provided in Table S3. Initial interrater reliability was 94%, with differences resolved by discussion. Analyses For studies 1 and 2, data were composed of a series of binary responses. Data were analyzed with binomial regression models, testing for an effect of condition separately for each age group, followed by post hoc contrasts with sequential Bonferroni corrections. The dependent variables were entered as the number of times that participants gave essentialist responses out of the total test items. These analyses yielded Wald 2 values as indicators of signicant effects. For ease of interpretation, data are presented as probabilities of essentialist responses, accompanied by Wald 95% CIs. Our main analyses examined a composite measure summing all essentialist responses across the three measures (Chronbachs = 0.85). Similar patterns were found across each measure of essentialism examined separately; descriptive statistics for each measure of essentialism are available in Tables S4 and S5. For study 3, the percentage of category references tting each code was compared across condition via a series of independent-samples t tests, as were the number of negative and positive evaluations. ACKNOWLEDGMENTS. We thank Dr. Susan Gelman for providing the materials from previous work on which the present materials were modeled; Kelli Grobe and Greg Kirschen for assistance with data collection; the staff and families at Temple Emanu-El Nursery School, Park Avenue Christian Church Day School, and the Childrens Museum of Manhattan for participating in this research; Karl Edwards for study illustrations; and Dr. Todd Gureckis for assistance with gure preparation. Funding was provided by a seed grant from the New York University Institute for Human Development and Social Change (to M.R. and S.J.L.) and by National Science Foundation grant BCS-1147543 (to M.R.).
17. Hirschfeld LA (1996) Race in the Making: Cognition, Culture, and the Childs Construction of Human Kinds (MIT Press, Cambridge, MA). 18. Keller J (2005) In genes we trust: The biological component of psychological essentialism and its relationship to mechanisms of motivated social cognition. J Pers Soc Psychol 88:686702. 19. Rhodes M, Gelman SA (2009) A developmental examination of the conceptual structure of animal, artifact, and human social categories across two cultural contexts. Cognit Psychol 59:244274. 20. Deeb I, Segall G, Birnbaum D, Ben-Eliyahu A, Diesendruck G (2011) Seeing isnt believing: The effect of intergroup exposure on childrens essentialist beliefs about ethnic categories. J Pers Soc Psychol 101:11391156. 21. Diesendruck G, Haber L (2009) Gods categories: The effect of religiosity on childrens teleological and essentialist beliefs about categories. Cognition 110:100114. 22. Birnbaum D, Deeb I, Segall G, Ben-Eliyahu A, Diesendruck G (2010) The development of social essentialism: The case of Israeli childrens inferences about Jews and Arabs. Child Dev 81:757777. 23. Sousa P, Atran S, Medin D (2002) Essentialism and folkbiology: Evidence from Brazil. J Cogn Cult 2:195223. 24. Haslam N, Rothschild L, Ernst D (2000) Essentialist beliefs about social categories. Br J Soc Psychol 39:113127. 25. Mahalingam R (2007) Essentialism, power, and the representation of social categories: A folk sociology perspective. Hum Dev 50:300319. 26. Mahalingam R, Rodriguez J (2006) Culture, brain transplants and implicit theories of identity. J Cogn Cult 6:453462. 27. Carlson GN, Pelletier FJ (1995) The Generic Book (Chicago Univ Press, Chicago). 28. Gelman SA, Ware EA, Kleinberg F (2010) Effects of generic language on category content and structure. Cognit Psychol 61:273301. 29. Waxman SR, Lynch EB, Casey KL, Baer L (1997) Setters and Samoyeds: The emergence of subordinate level categories as a basis for inductive inference in preschool-age children. Dev Psychol 33:10741090. 30. Diesendruck G (2003) Categories for names or names for categories? The interplay between domain-specic conceptual structure and language. Lang Cogn Process 18: 759787.
Rhodes et al.
31. Cimpian A, Markman EM (2009) Information learned from generic language becomes central to childrens biological concepts: Evidence from their open-ended explanations. Cognition 113:1425. 32. Cimpian A, Markman EM (2011) The generic/nongeneric distinction inuences how children interpret new information about social others. Child Dev 82:471492. 33. Gelman SA, Coley JD, Rosengren KS, Hartman E, Pappas A (1998) Beyond labeling: The role of maternal input in the acquisition of richly structured categories. Monogr Soc Res Child Dev 63:IV, 1148. 34. Gelman SA, Taylor MG, Nguyen SP (2004) Mother-child conversations about gender: Understanding the acquisition of essentialist beliefs. Monogr Soc Res Child Dev 69:1145. 35. Dunham Y, Baron AS, Carey S (2011) Consequences of minimal group afliations in children. Child Dev 82:793811. 36. Cimpian A, Meltzer TJ, Markman EM (2011) Preschoolers use of morphosyntactic cues to identify generic sentences: Indenite singular noun phrases, tense, and aspect. Child Dev 82:15611578. 37. Bigler RS, Liben LS (2007) Developmental intergroup theory: Explaining and reducing childrens social stereotyping and prejudice. Curr Dir Psychol Sci 16:162166.
38. Dweck CS (2009) Prejudice: How it develops and how it can be undone. Hum Dev 52: 371276. 39. Karafantis DM, Levy SR (2004) The role of childrens lay theories about the malleability of human attributes in beliefs about and volunteering for disadvantaged groups. Child Dev 75:236250. 40. Levy SR, Dweck CS (1999) The impact of childrens static versus dynamic conceptions of people on stereotype formation. Child Dev 70:11631180. 41. Atran S, et al. (2001) Folkbiology doesnt come from folkpsychology: Evidence from Yukatek Maya in cross-cultural perspective. J Cogn Cult 1:342. 42. Haslam N, Levy SR (2006) Essentialist beliefs about homosexuality: Structure and implications for prejudice. Pers Soc Psychol Bull 32:471485. 43. Brandone AC, Gelman SA (2009) Differences in preschoolers and adults use of generics about novel animals and artifacts: A window onto a conceptual divide. Cognition 110:122. 44. Gelman SA, Goetz PJ, Sarnecka BW, Flukes J (2008) Generic language in parent-child conversations. Lang Learn Dev 4:131.
Rhodes et al.
Essence and Natural Kinds: When Science Meets Preschooler Intuition 1

Sarah-Jane Leslie Introduction It is common practice in philosophy to rely on intuitions in the course of an argument, or sometimes simply to establish a conclusion. One question that is therefore important to settle is: what is the source of these intuitions? Correspondingly: what is their epistemological status? Philosophical discussion often proceeds as though these intuitions stem from insight into the nature of things as though they are born of rational reflection and judicious discernment. If these intuitions do not have some such status, then their role in philosophical theorizing rapidly becomes suspect. We would not, for example, wish to place philosophical weight on intuitions that are in effect the unreflective articulation of inchoate cognitive biases. Developmental psychology has discovered a range of belief sets that emerge in the first few years of life, and which plausibly go beyond the evidence to which the child has had access in that time period. In such cases, it is reasonable to suppose that the belief sets do not derive solely from the childs rational reflection on her evidence, but rather show something about the way human beings are fundamentally disposed to see the world. (In some cases, the deep-seated dispositions are also shared with non-human animals.) There are many explanations of why we may be fundamentally disposed to see the world in a particular way, only one of which is that metaphysically or scientifically speaking, the world actually is that way. Another explanation may be that it is simply useful and practical to see the world that way and this may be so even if it misleads us with respect to the metaphysical and scientific structure of reality. One particular way of carving up the world may be, say, efficient from the information processing point of view, without reflecting much about the fundamental nature of reality. Suppose, then, we find that a particular set of philosophical intuitions closely resembles such an early-emerging implicit belief set. This certainly does not establish the falsity of the intuitions, but it should give us reason to subject them to further scrutiny. This is because such a set of intuitions may strike us as completely compelling, and yet in fact reflect no more than a useful but ultimately misleading cognitive bias. Of course it may turn out that this is not the case; there is no argument from early developing to false. The point is purely epistemological: such a discovery should lead us to scrutinize the intuitions, and look for independent and converging evidence for the conclusions they urge. 2 The present paper focuses on essentialism about natural kinds as a case study in order to illustrate this more general point. Saul Kripke and Hilary Putnam famously argued that natural kinds have essences, which are discovered by science, and which determine the extensions of our natural kind terms and concepts. This line of thought has been enormously influential in philosophy, and is often taken to have been established beyond doubt. The argument for the
1
I am indebted to Andrei Cimpian, Michael Devitt, Susan Gelman, Sally Haslanger, Elizabeth Harman, Mark Johnston, Paul Needham, Michael Weisberg, and an anonymous reviewer for helpful discussion, guidance, and invaluable feedback on earlier drafts. All the remaining mistakes are my own. 2 The most general question here is the epistemological status of aspects of philosophical methodology, in light of recent psychological findings. Several theorists have recently begun to consider this question; for intriguing further discussion see, e.g., Gendler (2007), Stich (2009), and Weinberg, Nichols and Stich (2001).
conclusion, however, makes critical use of intuitions, and I note that the intuitions are of the sort had by preschool children, and that they are traceable to a deep-seated cognitive outlook, which is often called psychological essentialism. Further, if we did not have such a cognitive outlook or implicit belief set a belief set which is in fact in place by at least age 4 then we would not have the relevant philosophical intuitions. In light of this, I consider the question of whether natural kinds actually have scientifically discovered or discoverable essences, and whether these putative essences could determine the extensions of our terms and concepts as Kripke, Putnam, and many others have supposed. In fact, a number of philosophers of biology and chemistry have argued that biological and chemical kinds do not have such essences, yet these arguments particularly in the case of chemistry have not been assimilated by philosophers more generally. The reason for this poor assimilation is, I suggest, that the Kripke/Putnam view is just so intuitive. But this fact, I argue, is due to a deep-seated cognitive bias, rather than to any special insight into the nature of reality. The first half of the paper lays out the theory of psychological essentialism (which I rename quintessentialism for ease of exposition) in comprehensive detail, and summarizes some of the major experimental results that support it. I then consider further experimental evidence that suggests that lay people frequently misunderstand or misconstrue scientific findings as confirming their (quint)essentialist beliefs, when in fact the science does the opposite. Some of the more vivid and readily accessible cases involve quintessentialist beliefs concerning social kinds such as race and gender, and so such examples are also considered throughout the paper. The second half of the paper takes up philosophical essentialism about natural kinds, and argues that quintessentialist beliefs are required for the crucial Kripke/Putnam intuitions. I then review extensive findings in biology and chemistry and argue along with a number of philosophers of biology and chemistry that the findings conflict with Kripke/Putnam-style essentialism. To suppose otherwise is to misconstrue science through the lens of quintessentialism. The intuition that such essences must exist is no more than an expression of a belief set that is firmly in place by the middle of the preschool years, but this belief set does not comport with the complexities of actual science. To drive this last point home, we must first understand the hypothesis of psychological essentialism and its detailed predictions, then recall the Kripke/Putnam view, and then investigate the actual facts as they stand in biology and chemistry. The psychology explains where our kind-essentialist intuitions come from, and the biology and chemistry explains why we should suspect and/or suspend those intuitions.
I. Psychological (Quint)essentialism
The Quintessentialists Consider an intelligent species let us dub them the Quintessentialists whose members implicitly believe that a wide range of entities have within them a substance-like quintessence, which causally grounds their most important, stable and enduring properties. Not all entities are believed to have quintessences: for example, artificial and manufactured items are rarely believed to have quintessences. However, biological beings, along with certain non-biological substances, are believed to be bearers of quintessence. It must be emphasized that the Quintessentialists rarely, if ever, explicitly entertain thoughts about quintessence as such; rather
their quintessentialist beliefs are tacit or implicit, though they are frequently manifested in a number of explicit ways. Though they are not the only bearers of quintessence, biological individuals are paradigmatic bearers of quintessence. According to the quintessentialist belief set, each such individual has its own particular quintessence, yet there can be considerable similarities (and in the extreme case, potentially qualitative identity) between the quintessences of distinct individuals. 3 For example, offspring have quintessences that are highly similar to those of their parents. Further, two individuals are generally only able (or disposed) to produce offspring if their quintessences are sufficiently alike. Like begets like, as far as quintessence is concerned, and Quintessentialists are convinced that ones parents quintessence is the main determiner of ones own quintessence. These sorts of similarities in quintessence are taken to form the basis of real or natural kinds, as opposed to merely conventional groupings. (In the case of nonbiological individuals, such as certain substances, origins may also be considered important for example, if two substances are formed in very different ways, then it is reasonable to suppose they have distinct quintessences.) Thus the Quintessentialists believe that there are ways of dividing individuals up into kinds that carve nature at its joints namely the ones that group according to objective likeness in quintessence. Conversely, they also acknowledges that some groupings are merely nominal, and do not reflect genuine similarity classes of quintessences. However, Quintessentialists firmly believe that many of their own words and concepts map directly on to the real, objective kinds. The Quintessentialists also believe that there are a number of levels or degrees of similarity in quintessence, all of which are real and objective, in effect constituting a taxonomic hierarchy of kinds. At the lowest levels of this taxonomy, there are considerable similarities between the quintessences of members of some distinct kinds, while at the higher levels, there is considerable variation between the quintessences of members of the same kind. Importantly, there is a privileged level of this subjective taxonomy that occupies a sweet spot in this tradeoff between within-kind variation in quintessence, and cross-kind quintessential distinctness. At this level, individual members of the same kind have only minimal differences in their quintessences, and these quintessences are quite dramatically different from the quintessences had by members of other kinds. (The Quintessentialists cognitive psychologists call this taxonomic level the basic-level.) The Quintessentialists believe that this privileged taxonomic level is objectively determined, and so there is a privileged way of answering the question of whether a given individual is the same kind of animal (or same kind of plant) as another: namely whether they
3
Sometimes quintessentialism (= psychological essentialism) is characterized as not allowing for individual variation in quintessence that is, sometimes it is assumed that members of the same natural kind will have the same quintessence rather than highly similar quintessence. I think in the general case, quintessentialism is best articulated as allowing at least for the possibility of individual variation. This is most vivid in the case of individual people; since a wide range of social kinds are quintessentialized, there must be as much variation between people as there are different quintessentialized social kinds for them to belong to. Further, the literature on organ transplantation (discussed in more detail in the section on causal minimalism) reflects the belief that a recipient can take on individual characteristics of the donor -- for example, heart recipient Claire Sylvia (1997) believed that she came to enjoy fried chicken and beer as the result of a transplant from a young man named Tim (see also Inspector, Kutz & David, 2004). In order to accommodate these sorts of considerations, the possibility should exist for individual variation in quintessence. It should further be emphasized that the model I am introducing here would allow also for the possibility that, in some cases, the nature of the individuals quintessence would be exhausted by the nature of the quintessence of the (most specific) natural kind to which they belong. (That is, my model is compatible with the limit case where similarity becomes identity.) For example, one might believe that, e.g., mosquitoes do not display variation in their individual quintessences.
belong to the same basic-level kind. The notion of a basic-level kind extends beyond the biological realm and into the chemical realm too; samples of substances can also share quintessences, and there is again a privileged level of the taxonomy of substances where the kinds in play maximize intra-kind similarity (in the limit case, intra-kind qualitative identity) and inter-kind dissimilarity in quintessence. Thus the Quintessentialists are strongly inclined to believe there is a privileged way of answering the question of whether one thing is the same substance as another. It must be emphasized that the Quintessentialists believe that these taxonomic categories are determined by objective facts about quintessence. This has several consequences; firstly that the Quintessentialists strongly believe that membership in these kinds is a purely intrinsic matter a given individuals quintessence is the sole determiner of its membership in a real kind. Further, they believe that quintessence lends itself to being carved at its joints that is, quintessence does not vary continuously between individuals of different kinds, but rather is objectively distributed in such a way that, especially at the basic-level, members of the same kind have considerable sameness of quintessence, while non-members have distinctly different quintessences. Thus, membership in these kinds ought to be close to an all-or-nothing matter; that is, Quintessentialists believe that real kinds should have sharp boundaries. Sometimes they themselves are not able to tell for certain the kind to which an individual belongs, but they believe there should be a correct answer to the question an answer that is determined by the individuals quintessence. (However, other beliefs about quintessence namely that it is transmittable and mixable, as discussed below compete with this one, creating an inevitable tension.) The single most important feature of quintessences is that they have causal powers. An individuals quintessence is understood to be the causal root of many/most of that individuals stable and enduring intrinsic properties. The respects in which members of a real kind are outwardly similar to each other is understood by the Quintessentialists to be the direct upshot of the similarities between their quintessences. Thus, the Quintessentialists consider outward signs of kind-membership to be important, but only in as much as they are indicators of underlying similarities in quintessence. It would be a mistake, though, to think that quintessences only have kind-related causal powers. For example, the Quintessentialists believe that they themselves each have a distinct quintessence, which grounds many of their individually distinguishing properties (in addition, that is, to grounding kind-general properties). For example, individual personality traits are considered to be the upshot of the individuals quintessence. 4 The Quintessentialists further believe that an items quintessence causes its properties in a defeasible way that is, adventitious factors can prevent an individual from manifesting all the properties that the quintessence would otherwise cause it to manifest. This is particularly clear to them when they consider kind-wide quintessential properties that have been altered by clearly external circumstances for example, some Quintessentialists have lost limbs to amputation or accident. It can also happen that an individuals quintessence is somehow thwarted from ever expressing itself in a particular feature for example, some Quintessentialists are simply born without limbs. Since quintessence plays such a central role in beliefs about kind-membership, and since quintessence causally grounds a range of properties, Quintessentialists are ready and able to exploit kind-membership in their reasoning practices. Since natural kinds especially basic4
Again, not all articulations of quintessentialism (= psychological essentialism) take this line, however it seems to me the best way to account for a number of findings; see footnote 2, and the section below on causal minimalism.
level kinds are supposed to be groups with highly similar quintessences, Quintessentialists treat membership in a natural kind as a cornerstone of their inductive practices. If one member of a basic-level kind has a plausibly quintessential property, then other members of that kind may well have the property too. (Of course, these inferences are defeasible in a number of ways, given the nature of quintessential properties. They are, nonetheless, a good default starting point.) In this way, knowledge of kind-membership is more important to the Quintessentialists inductive inferences than anything else: more important than perceptual similarity between individuals, and more important than their sharing any other properties. When it comes to beliefs about their fellows, the same framework is employed. Quintessentialists manifest a tendency to believe that certain social categories most prominently race and gender categories -- group according to real similarities and distinctions in quintessence, and so reflect objective, non-constructed differences in nature. Just as in the nonsocial case, quintessentialist beliefs about these categories involve thinking that membership in them is determined by natural, inborn, intrinsic properties of the individual; that members of these groups are highly similar to one another and distinctly different from non-members; and that membership in these groups is more important and informative about the person than his or her appearance, demeanor, and other such readily observable properties. This quintessentialist belief set about the social domain has a number of unfortunate consequences; in particular, social groups whose members are seen to share very similar quintessences often experience significant degrees of social prejudice. When it comes to the nature of quintessence itself, the Quintessentialists do not have very specific beliefs at all. Instead, they have what might be termed a placeholder notion of quintessence they are sure that something fills this role, but they know not what. They are confident that it has a nature and independent existence, however, and they see no reason why they shouldnt discover more about its nature, especially as their scientific practices mature. However, if one were to push for an elaboration of the metaphysics of quintessence, one might arrive at something like the following: quintessences are substance-like entities in that they occupy or better, pervade space-time regions. They can also mix with each other, albeit only under unusual circumstances. Quintessences can permeate objects, including inanimate objects, and thus sometimes be transmitted. Notably, such transmissions almost always require a physical bearer. In the normal case, an individuals quintessence pervades its insides; internal parts are regarded as more important for quintessence than external surface parts. In this way, Quintessentialists believe that altering an individuals insides is more likely to produce a change in the individuals quintessence than altering an individuals surface properties. In keeping with their beliefs that real, basic-level kinds reflect highly similar quintessences, they also expect that members of the same basic-level kind will have highly similar insides even if their external appearances are quite different. Since ones insides are suffused with ones quintessence, the possibility of physically transplanting a part of an individuals insides into another raises the possibility of transferring some of an individuals quintessence into another. In recent years, the Quintessentialists medical technology has evolved to a point where this is more than just a possibility: internal organs can now be transplanted from one Quintessentialist to another. This raises particular anxieties for the Quintessentialists, since they believe that this practice transfers some of the donors quintessence to the recipient. The donors quintessence retains its causal powers, and as a result, the Quintessentialists believe that the recipient may be irreversibly changed. (The Quintessentialists
are often reminded by their medical professionals that there is no scientific evidence to support this claim, but the lay belief persists nonetheless.) Recently, some Quintessentialist scientists have made considerable technological progress on the possibility of transplanting organs from other species. This is most disquieting to the Quintessentialist public. They fear that receiving transplants from other species may alter them, making them more animal-like. They feel that mixing the quintessences of different basiclevel kinds is profoundly unnatural. While internal parts are the main bearers of quintessence, it strikes the Quintessentialists that a small amount of an individuals quintessence rubs off on things that he or she handles and uses. This belief manifests itself in several ways. For example, Quintessentialists are willing to pay large sums of money for otherwise unremarkable objects, just because they were once owned and used by a celebrated person. They also shrink from items that were handled or used by an infamous and hated person. (This disposition increases as the contact becomes more intimate e.g. a worn item of clothing is more potent in this respect than a touched pen.) At least to some extent, the Quintessentialists believe that they can be causally affected by quintessence transmitted through an object in this way. To the extent that they hold this belief, it may be weaker than their belief in the potency of organ transplantation which is, of course, reasonable given their beliefs about the relationship between internal parts and quintessence. There is, nonetheless, some degree of belief in the causal potency of quintessence transmitted even in such an indirect way. It must be emphasized again that these beliefs in quintessence are rarely explicitly formulated, articulated or entertained by the Quintessentialists. Rather, the belief structure sketched out here is a tacit framework that guides their thinking. Thus, to determine whether an individual is a Quintessentialist, it does not do to ask them directly rather, one must use a number of indirect measures. Further, the Quintessentialists are not taught to have the beliefs they do. These beliefs are not a mere cultural artifact; Quintessentialists from a wide variety of culturally independent groups exhibit the belief structure sketched above. And tellingly, Quintessentialist children manifest this way of thinking from a very young age. The role of culture is often to fill out more specific forms of the implicit beliefs that are characteristic of the early-developing quintessentialist syndrome. Psychological (Quint)Essentialism Psychological essentialism (henceforth psychological quintessentialism) holds that we are the Quintessentialists described above. This well-confirmed psychological hypothesis attributes to us an implicit and early-developing belief in quintessence in a substance-like entity possessed by some individuals and some forms of stuff, a substance which pervades their insides and causally grounds many of their stable and persisting properties. Membership in a natural kind is taken to be determined by objective features of ones quintessence namely the features that are shared by other members of the kind. It is crucial to understand the hypothesis of psychological quintessentialism correctly. It does not involve the claim that to be a Quintessentialist is to subscribe to all the beliefs listed above; rather quintessentialism is better thought of as a syndrome, which can be manifested in a variety of default implicit beliefs and ways of interpreting ones world. Thus, one might be a Quintessentialist without holding each belief mentioned above. Compare, for example, the symptoms of the syndrome that is depression. Such symptoms include having suicidal thoughts;
however it would be a mistake to suppose that one simply could not be depressed unless one has suicidal thoughts. Moreover, default implicit beliefs can be suppressed by explicit learning of contrary facts. For example, a well-informed adult may be less likely to explicitly hold the relevant beliefs about, say, organ transplantation. But this is fully compatible with being a Quintessentialist; that is, exhibiting the psychological syndrome that is manifested in the default beliefs mentioned above. It would also be a mistake to suppose that introspection alone could deliver the result that one is not a Quintessentialist. Introspection at best provides access to explicitly held beliefs, however it is unhelpful in assessing beliefs that are largely implicit. Sometimes, one belief may be held implicitly, while a contradictory belief may be held explicitly. A dramatic illustration of this phenomenon concerns prejudiced attitudes as revealed by tests such as the Implicit Association Test (IAT) (e.g., Greenwald & Banaji, 1995). Many individuals who honestly explicitly disavow sexist and racist (and other prejudiced) beliefs nonetheless may implicitly hold such attitudes even dedicated feminists may implicitly associate, e.g., women with nurturance but men with intelligence. Further, implicit attitudes have been found to better predict a range of behaviors as compared to explicit/introspection-based measures of attitudes (Greenwald, Poehlman, Uhlmann & Banaji, 2009; for excellent further philosophical discussion of these points and related ones, see Gendler, 2008a, 2008b, in press). Implicit attitudes and beliefs are not somehow of lesser significance than explicit ones only a limited range of the beliefs and attitudes that shape our behavior are available to introspection. And again, the explicit rejection of a given belief does not entail that that belief is not held implicitly. So quintessentialism is a syndrome, which encourages one or another cluster or range of default implicit beliefs drawn from the inventory provided in the previous section. These beliefs are not readily available to introspection, and they can persist in the face of known evidence to the contrary. The combination of these three observations means that a certain tempting response to the arguments given in this paper would be mistaken. I have in mind the following protestation, crudely characterized: I know that it is scientifically impossible for a heart transplant to alter ones personality, therefore I am not a Quintessentialist; however I still have the Kripke/Putnam intuitions, therefore it cannot be the case that Kripke/Putnam intuitions are due to quintessentialism! This line of objection, along with its more subtle variants, turns on a failure to understand the hypothesis of psychological quintessentialism, and the significance of implicit beliefs more generally. Some Crucial Findings Despite the fact that the beliefs in question are largely implicit, and so are somewhat more difficult to discern than our explicit beliefs, the last two decades or so have produced a wealth of empirical evidence suggesting we are, indeed, Quintessentialists. In the interests of space, I will not attempt a complete review here, but will rather highlight some of the main findings. (For an extremely thorough review of the data available through 2003, see Gelman, 2003.) There is a considerable amount of evidence that supports the claim that there is a privileged level in our psychological taxonomies namely the so-called basic-level (Rosch, 1978; see also Coley, Medin & Atran, 1997; Coley, Medin, Proffitt, Lynch & Atran, 1999; Gelman, 2003; and many others). Members of the same basic-level kind are perceived to have quintessences that are highly similar, and also highly distinctive, in that members of other basiclevel kinds have notably different quintessences (Gelman, 2003). Accordingly, in inductive
reasoning tasks, both children and adults very often generalize properties to other members of the same basic-level kind (e.g. tigers), but not to more inclusive kinds (e.g. mammals); nor do they limit their generalizations to less inclusive kinds (e.g. Bengal tigers) (Coley et al., 1997; Gelman & OReilly, 1988; Waxman, Lynch, Casey & Baer, 1997). Information about membership in a basic-level kind has a powerful impact on peoples inductive inferences from a very early age. For example, even two-year-olds prefer to base their inductive inferences on information about kind membership (as communicated via language) than on perceptual similarity (Gelman & Coley, 1990) despite the general importance of perceptual similarity (especially shape) in guiding young childrens inferences. Further, preschool children distinguish between the sorts of properties that can be generalized along kind boundaries from those that cannot: children do not generalize transient or accidental properties such having fallen on the floor that morning (Gelman & Coley, 1990; Gelman, Collman & Maccoby, 1986; Gelman & Markman, 1986, 1987). The latter are paradigmatically the sorts of properties that are independent of ones quintessence. It is also worth noting that the results of many of the studies cited above depend on even very young children being willing to accept that an individual can appear for all the world like a member of one kind, yet really belong to another kind. Further, it is kind membership that is understood to be most important when making inferences about deep, shared properties. Notably, these findings extend to substance kinds, with preschoolers understanding that something can look more like coal than gold yet still be gold, and also basing their inferences about projectible, non-obvious properties on this shared kind membership (Gelman & Markman, 1986). Young children also have definite views on the causal roles of nature vs. nurture. For example, preschoolers believe that an infant creature with kangaroo parents will grow up have a pouch and hop, even if it is raised exclusively with goats (Gelman & Wellman, 1991; see also Heyman & Gelman, 2000; Solomon & Johnson, 2000; Springer & Keil, 1989). Young children evidence the same set of beliefs when it comes to gender: for example, that a male baby raised on an island populated only by females will grow up to display boy-typical behavior and preferences (e.g. preferring to play with trucks rather than dolls; Taylor, 1996; Taylor, Rhodes & Gelman, 2009). Preschoolers also display at least some nature-over-nurture beliefs in the case of race, expecting that a child born to black parents will grow up to be black (Hirschfeld, 1996), though in general quintessentialist beliefs about race develop later than quintessentialist beliefs about species and gender (e.g. Rhodes & Gelman, 2009a). More generally, a number of studies have documented beliefs in children from very different cultural backgrounds about the relative power of nature vs. nurture, for example Menominee Indian children, Yukatek-Mayan children, and urban Brazilian children (Atran, Medin, Lynch, Vapnarsky, Ek & Sousa, 2001; Sousa, Atran & Medin, 2002; Waxman, Medin & Ross, 2007). These findings are exactly what one would expect if ones subjects were Quintessentialists. Children and adults also tend to intensify the boundaries of natural kinds that is, they judge that there should be sharp, exclusive boundaries between natural kinds (at the same taxonomic level). For example, Diesendruck and Gelman (1999) found that adults almost always judged that a given animal was definitely a member of a kind or definitely not a member of a kind; their judgments of how typical an exemplar of the kind the individual was, however, was more graded. The same pattern did not hold for artifact kinds, however, reflecting that this pattern is specific to natural kinds. Rhodes and Gelman (2009b) documented similar beliefs in young children. Further, even if adults are themselves uncertain as to how to classify an animal,
they judge that there is a single correct answer which an expert could determine (Luhman & Coley, 2000). Young children also judge that membership in both animal and gender kinds reflect correct, objective, non-conventional facts (Kalish, 1998; Rhodes & Gelman, 2009a). Children at least as young as four understand that animate individuals have very different insides from inanimate individuals (e.g. Gelman, 1987; Gelman & OReilly, 1988; Gelman & Wellman, 1991; Gottfried & Gelman, 2005; Simons & Keil, 1995). They also understand that the internal parts of animate creatures have special importance when it comes to determining both an individuals behavior and its kind-membership (Gelman & Wellman, 1991). Some recent work on infants suggests that the importance of insides is recognized from a very young age. Newman, Herrmann, Wynn, and Keil (2008) found that 14-month-olds expected that animals with the same internal features (e.g. visible red stomachs), rather than external features (e.g. blue hats), would display the same sorts of self-generated motion. However, when the motion of the target did not appear to be self-generated, these expectations did not arise. Thus it appears that, when it comes to apparently animate, self-generated patterns of motion, even infants expect that internal parts will better predictors than external parts. Additional studies have found that both children and adults understand that insides are of considerable importance in determining membership in both biological and chemical kinds (e.g. Keil, 1989; Lizotte & Gelman, 1999; Newman & Keil, 2008). For example, Keil (1989) reports a classic series of experiments that show childrens increasingly sophisticated understanding of the importance of insides in determining kind-membership, with even very young children judging that, e.g., a raccoon that has had its fur shaved and dyed so as to make it look just like a skunk is nonetheless still a raccoon. Interestingly, recent evidence suggests that rhesus monkeys also privilege insides over outward appearance in determining kind-membership. (Phillips, Shankar & Santos, 2010; see also Phillips & Santos, 2007). The majority of studies mentioned so far were conducted with participants primarily from North American communities. However, recent work has discovered similar dispositions among other cultural groups. As of now, essentialist beliefs have been documented in at least the following communities: Indians (Mahalingam, 2003; Meyer, Leslie, Gelman & Stilwell, under review), Brazilians (Diesendruck, 2001; Sousa et al.,2002), the Menominee (Waxman et al., 2007), the Vezo in Madagascar (Astuti, Solomon, & Carey, 2004), Yucatec Mayans (Atran et al., 2001), the Yoruba in Nigeria (Walker, 1999), and the Torguud of Mongolia (Gil-White, 2001). Thus, while of course more work is needed, the available evidence suggests that quintessentialist thinking is not a local phenomenon, but rather is a pervasive aspect of human psychology. Quintessentialism or Simply Causal Minimalism?: The Transmissibility of Quintessence Strevens (2000) argues that the data just described do not demand an explanation in terms of our quintessentialist impulses: instead, they are better explained by positing that people believe that membership in a (natural) kind has a causal influence on many of an individuals properties. According to Strevens, the evidence so far cited supports the view that we have the default belief that it is being a member of the relevant kind that causally confers stripes on tigers; the evidence does not support the view that we have the default belief that there is something, a quintessence, in individual tigers that individually confers stripes on them. Strevens terms his view causal minimalism. As it happens, I do not agree with Strevens that his view better accounts for the data cited above, but it is true that the evidence adduced so far is more or less compatible with a deflationary kind-based account, such as the one he offers.
However, it turns out that there is more evidence. Causal minimalism cannot account for the default belief that quintessences can be transmitted and mixed, causing changes in the recipient without necessarily altering which kind the recipient belongs to. This conceptual possibility is illustrated by Sir Arthur Conan Doyles Adventure of the Creeping Man. In this story, Sherlock Holmes is called to investigate some alarming and mysterious goings-on at the house of an aging professor. Through a characteristically brilliant string of deductions, Holmes figures out what is happening: in an effort to recapture his youthful vitality, the professor has been injecting himself with a serum extracted from the glands of monkeys. This has caused the professor to take on several monkey-like characteristics: he has become increasingly aggressive and his knuckles have become thickened and hairy; he has developed superhuman climbing abilities, and at times he adopts the gait of a monkey. He remains, however, clearly a man as the title of the short story reflects. His kind membership remains unchanged at the time of Holmes intervention. This story however much at odds with modern scientific understanding reflects a basic quintessentialist way of thinking. A serum extracted from the glands of a monkey confers a portion of the monkeys quintessence, complete with its causal powers, thus making its recipient take on the characteristics of the monkey. The tale reflects some of the most specific characteristics of quintessentialist thinking: the notion that the quintessence is causally potent, lies in the internal parts of the individual, and is substance-like in that it can be physically removed, relocated, and mixed with the recipients own quintessence to produce a hybrid mix of outward properties. (Does this explain the demand for bulls testicles among those who desire a certain outward property?) Can such beliefs be systematically studied and documented? In an early study, Johnson (1990) found evidence that young children do indeed believe that such hybrid mixtures of quintessence are possible; for example, they believe that a heart transplant from a mean person would make the recipient become meaner. Meredith Meyer, Susan Gelman, Sarah Stilwell and I have recently conducted a series of experiments to test for such beliefs in adults. In particular, we wished to assess whether people would endorse the possibility of a transplant or transfusion recipients personality changing to become more like the donors. We asked participants to consider donors from a range of social categories and two animal categories. We studied adult participants in both the United States and India, and have found that a sizeable portion of participants from both communities judge that the recipients personality may change to become more like the donors as a result of receiving a transplant or transfusion. 5 In the case of organs taken from animals, we also asked participants whether they believed the recipient would
5
Such beliefs are also attested to in a number of (non-experimental) surveys of attitudes towards transplants (Basch, 1973; Belk, 1990;Hayward & Madill, 2003; Inspector et al., 2004; Sanner, 2001a, 2001b). For example, Inspector et al (2004) found that a third of heart-transplant recipients believed they had in fact taken on characteristics of the donor. Sanner (2001a, b) interviewed people about their attitudes towards various organ transplants, and she reports participants explicit formulations of Creeping Man concerns: I would perhaps look more piggish with a pigs kidney. Would I become half a pig, if I got an organ from a pig? What if I would start grunting? (2001a, p. 22) A number of participants also expressed more general discomfort at the idea of mixing species in this way: I feel instinctively that its wrong to mix different species, it would go wrong. My body would let me know that an animal organ didnt fit. Its contrary to nature. Its unnatural to move body parts between species. (2001b, p. 1495) And quite decisively: The whole pig nature just feels like a big no. (2001b, p.1496)
become a member of the animal kind as a result, and participants systematically rejected this possibility. Participants thus believed that an individual could take on anothers characteristics as a result of a transplant, without changing their kind membership. Causal minimalism does not have the resources to account for this sort of data. Are transplants and transfusions the only means by which quintessence may be transmitted? A quick perusal of auction catalogs will show that people are willing to pay large sums of money for rather ordinary objects, if those objects once had contact with a famous or admired person. Hood and Bloom (2008) and Frazier and Gelman (2009) found that even young children place higher value on such objects, as compared to qualitatively identical duplicate objects. The converse of valuing objects that have been in contact with admired persons is shrinking from items that have been in contact with reviled individuals; this phenomenon has been studied and documented at length by Carol Nemeroff, Paul Rozin, and their colleagues (see Rozin and Nemeroff (2002) for a review). For example, in one famous study, Nemeroff and Rozin (1994) asked participants to rate how they would feel about wearing a sweater that had come into contact with various people, desirable and undesirable, including an enemy of theirs, and a person that the participants judged to be evil. (Unsurprisingly, Adolf Hitler was the favored choice for the latter category.) Participants were strongly opposed to wearing a sweater worn by either a personal enemy or by Hitler; in fact they rated a sweater that had fallen in dog feces as more desirable to wear than Hitlers sweater. Notice that such results are not accounted for by the simple observation that we happen to value or care about historical properties. There are many historical properties of paintings and sweaters and keepsakes and the like that we do not care about, such as having once been located in Europe or twice having been scanned by an x-ray machine. The crucial thing to be explained is the quasi-infection-based character of the authenticity and contamination findings, and a natural hypothesis is that it is an implicit belief in transmissible quintessence which accounts for this. As with transplants and transfusions, the clearest way to determine whether the authenticity and contamination findings are evidence of a belief in a transmissible quintessence would be to see whether people believe that wearing an item of clothing might cause a person to become more like the original wearer, even if the person in question does not know about the items history. For example, suppose one is about to stand trial, and one learns that the judge is unwittingly wearing an item of Stalins clothing, worn by him on and off over several years. Might one become anxious that the judge will be less just as a result? This phenomenon has not been extensively studied; however one experiment suggests that such beliefs can be found. In particular, Johnson and Jacobs (2001) found that elementary school children and adults alike believed that wearing Mr. Rogers sweater would make a person friendlier, even if the wearer did not know about the prior owner. This finding suggests that, just as with transplants and transfusions, the notion of transmissible quintessence is at play here. The history of close-range adoration of relics purportedly from sainted persons may also be worthy of examination. Quintessentialism and Science Some quintessentialist beliefs such as the beliefs that receiving a blood transfusion or that wearing Mr. Rogers sweater can alter ones personality are quite clearly at odds with the deliverances of science. Many other quintessentialist beliefs are, I would argue, similarly at odds with science, although on the surface they do not seem so. In particular, scientific findings,
particularly in biology, are often misunderstood by the general population in a specifically quintessentialist manner. That is, people inappropriately map scientific concepts onto their preexisting quintessentialist beliefs, and then consider those beliefs to be scientifically underwritten. (This may be part of a more general phenomenon, whereby people adopt culturally available ways of elaborating and articulating the same core, quintessentialist beliefs (e.g. Waxman et al., 2007).) While most educated adults have some familiarity with the concepts of genes and of DNA, misunderstandings abound, many of which at least indirectly suggest that people may simply have mapped the concepts onto their implicit notion of quintessence. One illustrative pair of anecdotal examples can be drawn from the geopolitical sphere: on March 13th, 2000, Vladimir Putin reportedly asserted that central power is in Russias genes. 6 Apparently President G. W. Bush concurred; at a White House news conference held on October 17th, 2007, he wondered whether or not it's possible to reprogram the kind of basic Russian DNA, which is a centralized authority. 7 There is, of course, no such thing as specifically Russian DNA; however a quintessentialist view of Russians as a social kind entails belief in a particular Russian quintessence. There is also, of course, no scientific evidence to suggest that genetics per se would specially predispose a person or a group of people to accept centralized authority, yet quintessentialist thinking is highly compatible with such a belief. 8 As the above example illustrates, DNA and genes are often invoked in peoples conceptions of quintessentialized social groups. A number of social groups, such as groups demarcated by race and gender, are often highly quintessentialized; that is, people believe that the members of those social groups have highly similar quintessences to one another, and further that non-members have quite distinct quintessences in effect, these groups are treated as social analogs of basic-level kinds. Lay beliefs about race and genetics often show precisely this pattern. For example, the majority of adults in the United States agree with the following statement: Two people from the same race will always be more genetically similar to each other than two people from different races (Jayaratne, 2001). Such a belief is wholly at odds with contemporary scientific thinking about genetic variability. Instead, the genetic variability within a racial group is just as high as the degree of variability across racial groups (e.g., Graves 2001; Lewontin, Rose, & Kamin, 1984; Templeton, 1999). In general, contemporary genetics is far more concerned with understanding how genetic differences among individuals can explain the phenotypic differences among those individuals, rather than attempting to find genetic explanations for putative group differences. This point has been poorly assimilated, however. As an illustration, when asked to consider possession of the trait being nurturing, Cole, Jayaratne, Cecchi, Feldbaum, and Petty (2007) found that people thought that genetics could better explain (perceived) gender differences than individual differences within genders. That is, contrary to the actual deliverances of genetics, they thought that group-level genetic explanations were more applicable than individual-level ones. 9 Social groups aside, there are numerous other examples of quintessentialist misinterpretations of genetics. For example, Lanie, Jayaratne, Sheldon, Kardia, and Petty (2004)
6
Oleg Shchedrov, "Central Power in Russia's Genes, Putin Says," Reuters (Moscow), 13 March 2000, from Johnson's Russia List, 13 March 2000 7 Russias DNA, Washington Post editorial, October 19th, 2007. 8 I am extremely grateful to Peter Godrey-Smith for drawing my attention to these examples. 9 Interestingly, Keller (2005) found that belief in genetic determinism that ones genes are a powerful determiner of ones traits and character correlated with social prejudice, and that priming people to think about genetics further increased their level of prejudice.
note that people with a family history of a heritable disease sometimes think that they cant get the disease if they dont look like the family members who have it. This is a natural belief to have against the backdrop of quintessentialism: one is susceptible to a heritable disease only in as much as one shares quintessence with the relevant relative, and similarity of appearance is an excellent guide to the extent to which quintessence is shared. People also confidently attribute traits, abilities and dispositions to their genes in the absence of any scientific evidence supporting the idea that those characteristics are genetically based. Lanie et al. report that, even though few respondents could give a remotely adequate characterization of what genes even are, it is interesting that elsewhere in the interview about three quarters of respondents had no trouble giving an example of at least one nonphysical, nonmedical genetic characteristic that ran in their families, even if there was no scientific research in support of their beliefs (2004, p. 311). Such examples included characteristics like being good at home repairs (and perhaps also could have included embracing centralized authority). As another illustration, consider that on the quintessentialist outlook, a quintessence has the power to cause its bearer to have certain characteristics, more or less independently of the bearers environment. And so relatedly, the outlook encourages the belief that if a certain characteristic does not completely spring from the bearers quintessence, then it must be determined by the environment. Thus, quintessence and environment are seen as largely mutually exclusive influences when it comes to determining the causal source of a trait. 10 In contrast, modern genetics holds that phenotypic traits arise from complex interactions between an individuals genes and its environment. Genetics and environment are not in any sense mutually exclusive as causes, rather they operate in tandem. Nevertheless, lay understanding of genetics often implements the quintessentialist model, in which genetic causes and environmental causes are seen as exclusive of one another. For example, in an extensive study, Jayaratne, Gelman, Feldbaum, Sheldon, Petty, and Kardia (2009) found numerous negative correlations, and no positive correlations, between participants endorsements of genetic and environmental factors in explaining the source of a range of traits. In recent years, a number of researchers have argued that quintessentialism is responsible for both resistance to and misunderstanding of evolutionary theory (e.g. Gelman & Rhodes, in press; Hull, 1965; Mayr, 1982, 1988, 1991; Samarapungavan & Wiers, 1997; Shtulman, 2006; Shtulman & Schulz, 2008). The main culprit is the quintessentialist belief that within-species variability is a limited and uninteresting phenomenon, in comparison to within-species similarity. That is, quintessentialism places considerable emphasis on the similarities between members of a given species (or more accurately, a given basic-level kind, since these are not always one and the same), and de-emphasizes the differences between them. (This is, of course, why quintessentialism encourages such rich inductive inferences over these basic-level kinds.) To correctly understand evolutionary theory, however, one must appreciate that there is a considerable degree of variability within a species, since this variability is precisely what allows the evolutionary process to occur. Not only does quintessentialism make people resistant to accepting evolutionary theory for example, Samarapungavan and Wiers (1997) found a sizeable portion of Dutch third- and fifth-graders believe the species to be eternal and unchanging but it further means that even educated adults often wholly misunderstand what the theory claims. Shtulman (2006) found that
10
Of course this is not intended as an absolute and universal quintessentialist principle, but rather a general guiding one for example, there could still be cases in which the environment prevents a quintessential property from being had, an obvious example being a tiger that has lost its tail.
the majority of the talented group of students enrolled in the Harvard Summer School held transformational rather than variational conceptions of evolutionary theory. The correct variational conception of evolution holds that random mutations occur all the time, and some of these confer an adaptive benefit on their possessors, and so render them more likely to successfully reproduce and pass along the mutation to the next generation. The transformational conception of evolution, in contrast, holds that, if a trait is beneficial to the species, then over generations the species as a whole is likely to acquire the trait. Thus, on the transformational view, the quintessentialist emphasis on within-species commonalities is preserved what happens is that (somehow) over time the common quintessence of the species alters so as to produce more beneficial traits. Perhaps unsurprisingly, people who hold this transformational view of evolution are less likely to believe in evolution as scientific fact. After all, what they understand it to be saying is not in fact true. As emerges at some length below, biologys understanding of species and other taxa is deeply at odds with the quintessentialist mind-set. The relationship between ones genotype and ones species membership is complex, probabilistic, and highly dependent on extrinstic factors; there is no such thing as the species genotype. Just as the systematic misconstruals of evolutionary theory illustrate, educated adults may appear to have a working understanding of biology, while in fact their understanding is rife with quintessentialist confusion. For example, consider the following remark by the brilliant semanticists Hans Kamp and Barbara Partee: the vast majority of natural kind terms are sharp in the strict sense of being determinately true or false of everything that is found in the real world. For instance, to belong to a particular biological species an individual must have the DNA of that species; and almost without exception this is a property which an individual organism either definitely has or else definitely lacks (1995, p. 175). 11 How far does this kind of quintesssentialist thinking extend, and has it left its mark on philosophy?
II. Philosophical Essentialism and Psychological Quintessentialism

Natural Kinds and Philosophical Essentialism Hilary Putnam and Saul Kripke famously argued for a version of philosophical essentialism as applied to natural kinds, according to which there are scientifically discoverable necessary and sufficient conditions for belonging to a natural kind. In particular, these necessary and sufficient conditions consist in a specification of the kinds hidden underlying structure. This underlying structure is discoverable by science, and plays a causal/explanatory role in the determination of the kinds perceptible, or manifest, properties. Both Kripke and Putnam further argue that the extensions of natural kind terms are determined by sameness of underlying essence rather than by, e.g., sameness of manifest properties. 12
11
I am grateful to Susan Gelman for directing my attention to this passage. She and Marjorie Rhodes use this passage to illustrate a very similar point in Gelman and Rhodes (in press). 12 It must be acknowledged that Hilary Putnam came to revise his view on the matter, and allow that, e.g., our interests also figure in the determination of the extension of natural kind terms (e.g., Putnam, 1992). This revised view has not, however, been nearly so influential as his original view, and in fact comparatively few philosophers are even aware that Putnam changed his view (Hacking, 2007a). The discussion in this half of the paper is directed towards the view articulated in The Meaning of Meaning, at least as it is widely interpreted that has been so influential, and which is so often characterized as Kripke/Putnam essentialism. For further discussion, see Hacking
The Twin-Earth Thought Experiment Perhaps the best-known illustration of the point is Putnams Twin-Earth thought experiment. We are asked to imagine a planet that is qualitatively identical to Earth, except that where Earth contains H2O, Twin-Earth contains a superficially indistinguishable substance whose complex chemical formula is abbreviated as XYZ. Thus, on Twin-Earth, XYZ fills the oceans and lakes, and falls from the sky as rain; Twin-Earth people drink XYZ to quench their thirst, bathe in it, use to make soups, and so forth. Putnam then asks us to imagine that, in 1950, some inhabitants of Earth set out on a spaceship and reach Twin-Earth. When they arrive there, they are astonished at the similarities, including what they initially believe to be the abundance of water on the planet. Crucially, though, we are encouraged to have the intuition that when the Earths astronauts say there is water in the lakes on Twin Earth, what they say is false. If they go on to perform chemical tests on the liquid that fills the oceans and lakes on Twin Earth, they will realize they were just wrong to have called the substance water. This is because the extension of our term water picks out all and only that which has the underlying chemical composition H2O. Alternatively, imagine an inhabitant of Earth, Oscar, and his Twin-Earthian doppelganger, Twin-Oscar. Putnam argues that the word water as used by each has different extensions for Oscar water applies to H2O, for Twin-Oscar water applies to XYZ. Of course, Oscar and Twin-Oscar may have different beliefs about water and twin-water respectively: namely, Oscar may believe that water is composed of H2O, while Twin-Oscar may believe that twin-water is composed of XYZ. However, Putnam then asks us to roll the clock back to 1750, before anything was known about the chemical composition of water. Putnam maintains that water still has a different extension in the mouths of Oscar and Twin-Oscar even though we may stipulate now that neither Oscar nor Twin-Oscar (nor any local experts) have any beliefs about the nature of the chemical composition of these substances. If we agree with these intuitions, we must suppose that two people can be in duplicate inner psychological states, while their terms have different extensions, due to the nature of the local natural kinds. Putnam famously concludes that meanings just aint in the head! (1975, p.227). In what follows below, I will argue that the intuitions associated with Putnams highly influential thought experiment are quintessentialist intuitions. That is, only beings who are (at least in some respects) Quintessentialists would speak a language whose terms behave as Putnam suggests (or at least share Putnams intuitions that the terms would behave in the relevant ways). If our psychology was relevantly different, then we would not have Twin-Earth intuitions even if the world does indeed conform metaphysically to the Kripke/Putnam essentialist view of it. I will then review evidence that the world does not in fact conform to Kripke/Putnam essentialism, and argue that the relevant intuitions derive solely from our quintessentialist outlook. As one might put it, Twin-Earth intuitions are driven by whats in the head rather than by whats in the world.
(2007a); nothing in this paper disagrees with Hacking on the historical observations, however I will continue to speak of the Kripke/Putnam view, since, as Hacking himself observes, this is how the view is widely understood. There is a crucial further point. It is one thing to qualify the view in the light of isolated counterexamples and quite another to hope that this will deal with the full systematic range of examples of natural kind terms drawn from biology and chemistry. What follows is intended to strongly suggest that we much more than the concession that our interests play a role.
Kripke/Putnam Essentialism and Quintessentialism In the wake of this compelling thought-experiment, Putnam proposes an account of natural kind terms. He suggests that natural kind terms can be given ostensive definitions that is, one may provide a definition for a natural kind term by ostending an instance of the kind, and indicating that the term applies to anything that is the same as the instance in important respects. For example, we could define the term water by pointing to a glass of it and noting that water applies to anything that bears the same liquid relation to the ostended sample: The logic of natural-kind terms like water is a complicated matter, but the following is a sketch of an answer. Suppose I point to a glass of water and say this liquid is water My ostensive definition of water has the following empirical presupposition: that the body of liquid I am pointing to bears a certain sameness relation (say, x is the same liquid as y ) to most of the stuff I and other speakers in my linguistic community have on other occasions called water (1975, p.224-5). And more generally: One can give [someone seeking to learn a natural kind term] a so-called ostensive definition this (liquid) is water; this (animal) is a tiger; this (fruit) is a lemon (1975, p.229). Kripke makes similar remarks about ostension in Naming and Necessity (e.g., p. 135). The crucial thought common to Kripke and Putnam is that underlying essential features determine the extension of same liquid and same substance. Nathan Salmon (1979) convincingly argues that one would be mistaken to suppose that the essentialist metaphysical doctrine is entailed by the above semantic picture. Rather, Salmon argues, the advocates of the semantic picture make essentialist assumptions from the outset. Indeed, Kripke himself explicitly denies that he ever thought to derive conclusions about the metaphysical status of essentialism from his theory of reference (1980, preface). I am in complete agreement with Salmon that the essentialist metaphysics is presupposed by the semantic analysis, rather than entailed by it. (I argue in the final sections of the paper, however, that the detailed presuppositions are ill-founded.) Here, I would note another implicit assumption in Putnams argument: namely that the communities of speakers that he describes are, like his readers, Quintessentialists, at least to some extent. Absent such an assumption, the argument would fail. (Unlike his metaphysical assumptions, however, I believe that this assumption is completely correct.) Imagine a community of speakers let us call them the Phenomenalists whose psychology is notably different from the Quintessentialists, especially when it comes to beliefs about the natures of certain kinds and individuals. The Phenomenalists staunchly deny that anything of interest lies below the surface it would never occur to them to consider two individuals to be essentially similar if they differed in their obvious surface properties. The Phenomenalists thus form their categories and concepts on the basis of perceptual appearance and other such readily-accessible qualities. (In this way, the Phenomenalists take Oscar Wildes remark only the most superficial people do not judge by appearances to be words to live by.) It should be clear that an ostensive definition of the sort Putnam discusses will yield a term with a very different extension for the Phenomenalists. Putnam writes, Suppose I point to a glass of water and say this is water, in order to teach someone the word water (1975, p. 230). If one attempts to teach a Phenomenalist a term in such as way, the Phenomenalist will acquire a term that he will take to have a very different extension than our term water does, at least on
Putnams account of how that term functions. For example, it will strike the Phenomenalist as undeniable that XYZ falls in the extension of this term: it looks like it, smells like it, tastes like it, and so on and what else matters to the Phenomenalist? Of course, if we are simply imagining an isolated Phenomenalist attempting to learn English, we might just dismiss his reaction as a semantic error, especially if we believe in public languages whose meanings are determined by community wide dispositions to use terms in certain ways. The response to this is obvious: let us imagine an entire community of Phenomenalists, whose numbers and practices are sufficient to determine a public language of their own. Even if one believes that languages are public and are thus not determined by individual psychological dispositions, there can be no denying that the psychological dispositions of the community of speakers determine which public language they are collectively speaking. If we do indeed speak a language whose term water does not include XYZ in its extension, this is because we are psychologically very different from the Phenomenalists. It might be objected that the Phenomenalists mindset is at odds with science. That is, havent we discovered that water has an essence (namely H2O), and so wouldnt it be somehow unscientific of the Phenomenalists not to revise the extension of the term after making such a discovery? Even if we grant the (dubious) scientific claim about the essence of water, it should be clear that this objection has no bite. First of all, no speaker is under an obligation to revise her use of a term in face of such scientific discoveries unless her community already had a (quint)essentialist view of the terms extension. If a term in a language has its extension determined by manifest appearance, then scientific discoveries are not relevant to determining its extension (unless of course they bear on facts about manifest appearance). Secondly, Putnam himself is quite explicit that his claims about meaning do not depend on what is known at a time about the relevant underlying essences. Oscar and Twin-Oscars words have different extensions in 1750; similarly, (gold) as used by Archimedes had its extension restricted to substances composed of the element with atomic number 79 even if Archimedes himself (let us suppose) was unable to tell this substance apart from, say, iron pyrite. It is not the scientific discoveries per se that fix the extension of the terms; it is the dispositions of the community of speakers to intend to use their natural kind terms to carve nature at its (quint)essential joints, whatever those may be. Science is here no more than a post-hoc guide as to the determinate semantic consequences of such dispositions. If we consider a Quintessentialist community instead of a Phenomenalist one, it is evident that the speakers are likely to speak (or at least believe that they speak) a language of the sort that Putnam describes. If one teaches the word water to a Quintessentialist via ostention, then she will readily suppose the words extension is determined by similarities in underlying quintessence. That is, she will be prepared to limit the use of the term water to just those quantities of substance whose quintessences are relevantly similar to that of the ostended substance and she will be happy to admit that there may be cases where a given quantity of stuff looks just like the original, and yet differs in its quintessence enough that the term water will not apply to it. Similarly, if she is taught tiger in the same way, she will suppose that the term applies to just those individuals whose quintessences are appropriately similar to the demonstrated individual. And importantly, since their quintessentialist view of the world preceded modern scientific discoveries, Quintessentialists would have understood these terms in this way throughout their entire history, just as Putnam supposes that we do. The Quintessentialists are even able to resolve (or at least appear to resolve) a lingering ambiguity that haunts these ostensive definitions (often referred to as the qua problem; e.g.,
Devitt & Sterelny, 1987; Dupr, 1981, 1993). There are, for example, a host of non-tigers whose quintessences are in many respects similar to the quintessence of our demonstrated tiger: namely lions, panthers, cougars, etc. There are also sub-groups of tigers whose quintessences differ in some important and systematic respects: Bengal tigers, Sumatran tigers, Siberian tigers. Let us suppose the demonstrated tiger was a Bengal tiger. How does the Quintessentialist determine that Siberian tigers but not lions have quintessences that are relevantly similar to the originally labeled animal? The answer is that she employs a helpful assumption: that novel terms name basic-level categories. 13 This assumption fits very well with the quintessentialist outlook: members of basic-level kinds are taken to have maximally similar quintessences, modulo their also having highly distinctive quintessences relative to other kinds. By settling on the basic-level kind, the Quintessentialist thereby adopts a strategy that maximizes the information an application of a natural kind term to an individual conveys about how the individuals quintessence is similar to and different from other individuals quintessences. As a consequence, the Quintessentialists favor using this strategy when learning terms via ostension. Since the Quintessentialists are so accustomed to (unconsciously) applying this strategy, they sometimes leave it wholly implicit in their philosophical works, as they assume that their readers will be Quintessentialists themselves. Put another way, the qua problem can seem not to be that deep precisely because we are Quintessentialists, who privilege basic-level kinds. If we are indeed Quintessentialists, it is easy to see why our intuitions concerning TwinEarth and the like accord with Putnams. Were we, say, Phenomenalists instead, the thoughtexperiment and its conclusions would gain no traction with us. It is thus at least a necessary condition for Putnams account to be successful that we think like Quintessentialists in the relevant respects. Is it also sufficient, or does the world have to cooperate? Do we have TwinEarth intuitions even when the kind in question has no essence? More generally: might we have the intuition that members of a particular kind must share an essence even if in fact they do not? If so, perhaps not only are the specifically Twin-Earth intuitions driven by quintessentialism, but so too is the bedrock idea that the kinds that correspond to our natural kind terms group individuals by their essences. The Tragic Mulatto Before considering whether the kinds that Putnam and Kripke discuss actually do group their members by essence, it is worth recalling a nineteenth and twentieth century North American literary and cinematic trope, the Tragic Mulatto (or Tragic Mulatta, as the trope is sometimes styled). The trope typically involves a young woman who looks and acts like a white person, and in many cases believes herself to be white. Everyone around her unquestioningly believes that she is white, and she rises to an enviable position in society and is sought after by eligible young white men. Then the terrible news is discovered: the woman is not, in fact, white, but rather has mixed racial ancestry. The discovery of one black ancestor spells her ruination. As the literary trope has it, a single drop of black blood suffices to make one black; she loses
13
A separate question is how a Quintessentialist figures out which categories are basic level. There is a wealth of interesting work on this issues (e.g. Murphy, 2002; Rosch, 1973, 1975, 1978; Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976). Shape is an important (though defeasible) guide. It is important to separate this empirical question from the point that a Quintessentialist, who can somehow distinguish basic-level from non-basic-level kinds, even when the kinds in question are novel, can use this ability to fix on a meaning for a term introduced via Kripke/Putnam style ostension.
everything overnight, is rejected by friends and lovers, and cast out of white society. In some instances, the tale ends with her being sold into slavery. This depressing trope was very popular, and clearly was not in any way difficult for its readers to comprehend. Let us consider its structure then: it requires the reader to suppose than a person can appear in every observable respect to be white, and yet in fact not be white, but rather black. This is something that would be incomprehensible to the Phenomenalist about race, yet is immediately comprehensible to the racial Quintessentialist. Since quintessence only defeasibly causes its bearers to have their characteristic observable properties, the possibility exists for a member of a kind to have the relevant quintessence, and yet not share any of the characteristic properties of the kind. Conversely, an individual may have the characteristic superficial qualities associated with a kind, and yet in fact have the quintessence of another kind. (Compare Gelman and Markmans (1986) study, where preschoolers readily accepted that an individual could be a member of chemical or biological kind, despite better resembling members of a different kind.) It is this quintessentialist understanding of race that the Tragic Mulatto trope exploits. The Tragic Mulatto trope has much in common with the Twin-Earth thought experiment; someone whose mind-set allowed them to accept the Tragic Mulatto tale hook, line and sinker would surely also have Twin-Earth intuitions about race. If there can be one tragic mulatto, why not a whole planet? Why not a Twin-Earth populated by individuals who look and act like white people, but are in fact, like the Tragic Mulatto, really black? Just as the Tragic Mulattos suitors incorrectly apply the term white to her, a visitor to this Twin-Earth would incorrectly apply the term white to its inhabitants, and so on, so forth. The same conceptual structure allows for both the Twin-Earth example and the Tragic Mulatto trope. Both hinge on the idea that something can appear in every respect to belong to a kind, but in fact fail to, since the individual lacks the relevant essence. There is, of course, no such thing as black or white essence. Racial groups are social constructs; they are not based on any biologically real essences. Genetic variation within racial groups is just as high as it is across racial groups there is no such thing as black DNA, and there is most certainly nothing intrinsic that would set apart a young woman as truly being black, despite having pale skin and other phenotypic features associated with white people. 14 In this case the world does not vindicate the belief in hidden essence, but racial groups have historically been and continue to be highly quintessentialized. These quintessentialist beliefs suffice for the analog of the Twin-Earth intuitions in the case of race. This suggests that quintessentialism may be not only necessary, but also sufficient for the relevant intuitions.
Some researchers have recently claimed that race does, in fact, have a biological reality, in particular that racial divisions can be at least approximately cashed out in terms of continent-based ancestral breeding populations, which affect to some extent the nature and frequencies of some alleles (see, e.g., Kitcher (2007) for sympathetic discussion). Such findings, even generously interpreted, do not amount to racial essences or black DNA or the like, as Kitcher makes very clear. For example, the claims in question are statistical/probabilistic in nature, and quite weak ones at that: only .0005% of human genetic variation is even putatively explained by membership in these groups (Maglo, 2011). It is also far from clear that these findings hold up to scrutiny (as opposed, perhaps, to simply being expressions of quintessentialism); there are a number of reasons to be skeptical regarding them reasons which range from concerns about the statistical methods employed to objections to the theoretical interpretations bestowed (for an excellent review, see Maglo (2011), and sources cited therein). Further, the Tragic Mulatto is specifically characterized by having largely white ancestry but a single drop of black blood, which suffices to make her black. Nothing in these statistically-based genomic analyses would deliver this result (in fact one criticism of them is that they tend to abstract away from cases of mixed ancestry, thus potentially making the findings appear more clean-cut than they in fact are).
14
Do Biological Kinds have Essences? One outstanding question at this point is whether the kinds that Putnam and Kripke actually discuss turn out to have essences, or whether -- as in the race case -- both our Twin-Earth style intuitions and our confident belief in kind essence are simply driven by unjustified quintessentialist beliefs. For a kind to have an essence in the Kripke/Putnam sense, there must be hidden underlying features that are necessary and sufficient for kind membership -- features of the sort that science investigates and discovers. Moreover, for a natural kind term to have the sort of extension that Kripke and Putnam suppose, it is necessary that any two members of the relevant kind (including any two quantities of the relevant substance) be similar in some such scientifically discoverable and essential respect. Putnam is very explicit that having chemical composition H2O is the relevant essential respect in the case of water, and also that having the appropriate genetic code is the relevant essential respect in the case of lemons. He writes: [A] critic has maintained that the predominant sense of, say, lemon is the one in which anything with the superficial characteristics of a lemon is a lemon. The same critic has suggested that having the hidden structure the genetic code of a lemon is necessary to being a lemon only when lemon is used as a term of science. Both of these contentions seem to me to rest on a misunderstanding The sense in which literally anything with the superficial characteristics of a lemon is necessarily a lemon is extremely deviant At the same time the sense in which to be a lemon something has to have the genetic code of a lemon is not the same as the technical sense (if there is one, which I doubt). The technical sense, I take it, would be one in which lemon was synonymous with a description which specified the genetic code. But when we said that to be water something has to be H2O we did not mean that the speaker has to know this similarly, even though the predominant sense of lemon is one in which to be a lemon something has to have the genetic code of a lemon (I believe), it does not follow that lemon is synonymous with a description which specifies the genetic code explicitly or otherwise (1975, pp.239-240, original emphasis). Kripke makes similar remarks concerning the internal structure of tigers (1980, pp.120-1), and later speaks of scientific discoveries of species essence (p. 138). Does the world cooperate with these claims, or are these quintessentialist misinterpretations of science? That is, are Kripke and Putnam simply reporting scientific facts in their discussions, or are the discussions fueled instead by unfounded quintessentialist convictions that natural kinds must have essences? Might the very claim that natural kinds have scientifically discovered essences be no more than an articulation of inchoate quintessentialist intuition? Species and Sex; Genes and Essence On the question of whether species have these sorts of essences, there is a degree of consensus among philosophers of biology (and indeed biologists) that is almost unprecedented in philosophy at large (e.g., Dupr, 1981, 1993; Ghiselin, 1987; Hull, 1965; Laporte, 1997, 2004; Mayr, 1982, 1988, 1991; Okasha, 2002; Sterelny & Griffiths, 1999, and many others.). There is
no such thing as lemon DNA, no common genetic code that makes for membership in the kind Panthera tigris. 15 To a first approximation, most biologists subscribe to the notion that a species is delineated by the boundaries of an ecological niche, or by the boundaries of a reproductive community. 16 Within such bounds, considerable genetic variation is possible, and conversely, it is possible for there to be less genetic variation across such boundaries. As a concrete illustration, consider the cutthroat trout (Salmo clarki). This species has a number of subspecies, and there is considerable genetic divergence between these subspecies. In a study of seven subspecies of cutthroat trout, zoologists Fred Allendorf and Robb Leary report that: A highly variable pattern of genetic divergence exists among the seven subspecies Very little genetic divergence exists among Colorado, Snake River, greenback, and Yellowstone cutthroat trout. Nei's genetic distances between these subspecies are typical of those reported for conspecific populations in a diversity of freshwater and anadromous fishes (Avise 1974; Avise & Smith 1977; Buth & Burr 1978; Loudenslager & Gall 1980; Buth et al. 1984). In contrast, substantial biochemical genetic divergence exists between coastal, Lahontan, and westslope cutthroat trout and between these fishes and the other four subspecies. These genetic distances are truly exceptional for conspecific populations; they are similar to or larger than values observed between many species of fish (Johnson 1975; Avise & Ayala 1976; Buth & Burr 1978; Phelps & Allendorf 1983; Yates, Lewis, & Hatch 1984). Surprisingly, the coastal, Lahontan, and westslope cutthroat trout are as similar or more similar to the rainbow trout than they are to the other subspecies (Allendorf & Leary, 1988, pp.172-3, emphasis added). Thus a member of one species (e.g. a member of the Lahontan subspecies) may have more genetically in common with a member of another species (namely rainbow trout) than with a member of its own species (e.g. a member of the Snake River subspecies). While such findings may not be the norm in biology, this is not a wholly isolated occurrence; for example, high degrees of intra-species genetic divergence can be found among various other Salmonids as well (Pennell & Barton, 1996). As we shall see, these are not just weird cases; biology systematically confounds our quintessentialist convictions. As against this, recently Michael Devitt (2008) has criticized the consensus in the philosophy of biology, and indeed in biology and zoology, and argued for the existence of intrinsic, microstructral species essences of the Kripke/Putnam variety. It is instructive to consider Devitts arguments, since he articulates a view that is, I think, widely held in philosophy (outside of philosophy of biology). He argues that biologists and zoologists are committed to positing such essences even if they assert otherwise on pain of explanatory inadequacy. Devitt writes: There has to be something about the very nature of the groupa group that appears to be a species or taxon of some other sortthat, given its environment, determines the truth of the generalization [e.g. that Indian rhinos have one horn and African rhinos have two horns]. That something is an intrinsic underlying, probably largely genetic, property that
In what follows, I will limit my discussion to the species-level in the taxonomy, since this is, I think, in keeping with Kripkes and Putnams intentions. 16 These distinct ways of characterizing species are considered in more detail below.
15
is part of the essence of the group. Indeed, what else could it be? Some intrinsic underlying property of each Indian rhino causes it, in its environment, to grow just one horn. A different such property of each African rhino causes it, in its environment, to grow two horns. The intrinsic difference explains the physiological difference. If we put together each intrinsic underlying property that similarly explains a similar generalization about a species, then we have the intrinsic part of its essence (2008, p. 355). And summarizing: I have presented a positive argument for Intrinsic Biological Essentialism. We might sum it up: structural explanations in biology demand that kinds have essential intrinsic properties (2008, p. 365, original emphasis). Devitts argument is intuitively appealing, but not, I think, ultimately successful (see Barker (2010) and Ereshefsky (2010) for further discussion). First of all, the argument is undermined by the fact that members of the same species often do not share the same phenotypic traits; conspecificity is compatible with a great deal of variation in phenotype at a time, and even more dramatically over time. That is, it is possible for a species to radically change its phenotypic features over time without a speciation event occurring. In the extreme case, members of a species at t1 might not have any non-trivial phenotypic properties in common with the members of the species at t2. However, in what follows I will set aside this particular concern, and consider only cases in which a particular phenotypic trait actually is shared by members of a species. Devitt is, of course, indisputably correct that each particular African rhino has some intrinsic features that, in combination with the environment, are causally responsible for that individuals having horns. This does not entail, however, that those very same intrinsic features are also responsible for other African rhinos having horns. Whether this is so is a substantive empirical hypothesis, not one whose truth can be intuited in advance. For a simple illustration of the general point, let us revert to a well-known chemical case, that of jade. There are two very different chemical compounds that both fall under our term jade: jadeite (NaAl(SiO3)2) and nephrite (Ca2(MgFe)5Si8O22(OH)2). Substances composed of these two respective chemical compounds have very similar observable properties (in fact they are far more similar than many conspecific plants and animals). Thus a sample of jadeite and a sample of nephrite will have many properties in common (in fact they will be indistinguishable to all but the most experienced artisans). It is also true that the observable properties of the sample of jadeite are determined by its intrinsic chemical structure in conjunction with the environment, and similarly for the sample of nephrite. Yet there will be no common intrinsic chemical structure that explains the shared features of the two samples of jade. (There will, of course, be the non-explanatory disjunctive property of being composed of either NaAl(SiO3)2 or Ca2(MgFe)5Si8O22(OH)2. However, it is important to see that disjunctive properties cannot play the explanatory role that Devitt has in mind, or else the whole enterprise is trivialized. For example, let us suppose with Devitt that there is a common intrinsic property had by tigers that explains why they are striped. Let us also suppose that there is a different common property that explains why canna lilies are striped. If disjunctive properties are allowed to figure as common intrinsic explanatory properties in Devitts sense, then there will be a further shared intrinsic property that explains why this tiger and this lily both have stripes. If disjunctive
properties are countenanced in this endeavor, then shared properties become far too cheap to be of interest. Certainly, it would not then be a biological hypothesis that a common property explains why Indian rhinos have one horn it would simply be a familiar point about the logic of disjunction.) In the case of jade, we have selected these two chemically different but manifestly similar substances to figure in some of our statues and jewelry; but natural selection can do something similar when it comes to phenotypic features. Consider, for example, the property of having typical external human female genitalia the property that is overwhelmingly used to classify humans, particularly newborn infants, as female. 17 What is the genetic basis for having such a property what intrinsic genetic basis explains the possession of such a property (as Devitt might put it)? This is a question that many educated adults believe they know how to answer: it is having two X chromosomes, they say! Possession of two X chromosomes (or more accurately, possession of at least one X chromosome and no Y chromosome), it is thought, is both necessary and sufficient for being a human female, and this further explains the possession of typical female characteristics. The distinctive essential feature is thus not only thought to be necessary and sufficient for having typical external female genitalia, but it is also explanatory in the sense that Devitt seeks. This proposal does not in fact withstand scrutiny. While the majority of females do not have Y chromosomes, there are some that do; a small but substantial portion of the population have 46, XY karyotypes, but have typical external female genitalia. 18 There are numerous ways in which this situation can arise. For example, various mutations can lead to androgen insensitivity, meaning that that cell receptors do not respond to androgenic hormones, so cells are unaffected by the presence of these hormones. Alternatively, there can be any number of changes to the biochemical processes involved in producing these androgens: for example, there may be a range of alterations affecting the synthesis of testosterone from cholesterol, or there may be changes in enzyme levels that prevent testosterone from being broken down into dihydratestosterone. For each of the determinate ways in which this developmental pathway can be altered, there are a range of different genetic mutations, or combinations of mutations, that may be responsible. Further, this is far from an exhaustive list of the ways in which a 46, XY karyotype may be associated with typical external female genitalia (Kolon, 2008). 19 To return to Devitts argument, suppose we have a number of newborn females, all with typical external genitalia, and suppose we seek to explain why their genitals are formed as they
17
This discussion in this section is concerned with the biological categories of sex (e.g. male/female) not with the social categories of gender (e.g. man/woman). The two are, of course, often confused, in accordance with our tendency to quintessentialize gender to a very high degree. While it is often pointed out that the latter has no intrinsic essential bases, the corresponding point is less frequently made in the case of the former. The discussion here should illustrate that even the seemingly simple categories of sex are in fact highly complex and resistant to delineation in any simple essentialist terms. 18 This discussion abstracts away from cases of individuals with ambiguous genitalia, which can (though need not) result from many of the conditions described here, and from other conditions besides. 19 As a further level of complication, consider the phenomena of mosaicism and chimerism. These are conditions in which the cells within a given individual have different genetic makeups. For example an individual with mosaicism/chimerism might have some cells with karyotype 46, XX and others with 46, XY, or some with 47, XXY and others with 46, XY, or some with SRY mutations and others without, and so on. The effect on the phenotype depends on a range of factors, for example the relative proportion of the different cells, their location and degree of distribution around the body, and so on so forth. For example, if someone has a mix of 46, XX and 46, XY cells (which is rare but possible), they may have typical male genitalia, typical female genitalia, or ambiguous genitalia, depending among other factors on the proportion and location of the two cell types.
are. For each individual infant, there will be a genetic component to the explanation (as well as an environmental explanation), but this genetic component need not be the same for each infant. There may be considerable overlap in explanation (the majority of individuals with female genitalia have 46, XX karyotypes, and --- crucially --- lack the factors that lead such a karyotype to be associated with typical male genitalia), but in the case of some individuals, the explanation will go by a rather different route. One cannot infer a universal and essential intrinsic genetic component on the basis of shared morphological characteristics, even within a given sex of a given species. Of course, Quintessentialists that we are, we find such inferences appealing to the point of being almost irresistible. As a result, one might be tempted to rescue the idea by setting aside cases of 46, XY females, so as to preserve the idea that there is a genetic essence shared by all and only females that explains their typical morphological features. (I want to be absolutely clear that I am in no way attributing such a response to Michael Devitt, however.) There is no question that the individuals in question have the relevant morphological feature typical external female genitalia so if the explanation of why females have this feature is to make reference to a genetic essence shared by all females, the only option is to insist that these individuals are not, in fact, female. That is, if we insist that only 46, XX individuals are female, then we can rescue the idea that the kind human female has an essence, and further that this essence explains the typical morphological features of the category. The real females are the 46, XX females, the others are aberrations, freaks, not really female. I say this is no more than quintessentialism in its most unjustified and pernicious form. Biology does not provide neat, essentialized categories of this sort, however much quintessentialism demands them. 20 There are a number of different biological routes to the phenotype, and the natural world does not label one or another of these as the genuine, true, real, or normatively right one. 21 Sadly, a number of female athletes have had their careers and even their lives ruined by the quintessentialist way of thinking, which is blind to this natural fact. Further, the sort of phenomenon illustrated here is widespread in the biological world. In general, within a given species, individuals who share a common phenotypic feature need not share intrinsic microstructural bases that gave rise to the feature. One simple illustration of this phenomenon involves what are known as phenocopies individuals who display environmentally induced, non-hereditary phenotypic features which are identical to genetically induced, hereditary features displayed by other individuals. This can occur within a particular species for example, the Common Rabbit (Oryctolagus cuniculus) frequently sports black fur. This trait black fur can arise relatively straightforwardly from a given rabbits genetic makeup -- that is, as an inherited trait that manifests itself across various environments. The trait can also occur, however, among the Himalayan breed under certain environmental conditions. Himalayan rabbits, when raised in moderate temperatures, have white body fur with black tails, noses, and ears; if they are raised in cold temperatures, however, they develop wholly black fur. The genetic, developmental and biochemical pathways to black fur differ significantly between the two cases (Baum et al., 2010; Sawin, 1932); there is no common intrinsic explanation for the
20
For a number of years, 46, XY females were referred to in the clinical literature as male pseudohermaphrodites, a label which is no longer in use. Even in the context of medicine, quintessentialism can leave its traces, but it is hard to maintain in the face of a real investigation of nature. 21 Ah, but the variations described in this section tend to be incompatible with reproduction is this not biologys way of signaling that these individuals are not really female? Not more so here than with any genetically based condition that results in infertility.
possession of black fur by the members of O. cuniculus. (And conversely, there need be no genetic difference between a Himalayan rabbit with black fur and one with mostly white fur.) Perhaps one might be tempted to think that this example is misplaced because having black fur is not a species-wide characteristic of rabbits, or perhaps because appeal was made to a particular breed of rabbit (which, one might be tempted to think, would be such that its members share an intrinsic microstructural basis for the trait). This is not so. Consider, for example, that having three toes on the hind feet is a characteristic property of guinea pigs (Cavia porcellus). Possession of this phenotypic property is due to a flexible interaction-effect between a number of factors, both genetic and non-genetic that is, there is a generous range of pathways, all of which lead to having three hind toes. Individual guinea pigs can differ significantly from each other with respect to these factors and yet each have three toes, as was demonstrated in a series of classic experiments conducted by Sewall Wright (1934). Alternatively, consider the salamander Ambystoma talpoideum. A. talpoideum can undergo either metamorphosis from its larval stage to its adult stage, or else can undergo pedomorphosis, that is, it becomes sexually mature while retaining its larval characteristics. A general characteristic property of A. talpoideum is that it tends to undergo pedomorphosis if it is raised in a favorable aquatic environment. However, the genetic basis for this property has been found to differ between two populations of A. talpoideum that are separated by as little as 15 km (Harris, Semtlisch, Wilbur & Fauth, 1990). The developmental pathway leading to metamorphosis can be disrupted in any number of ways (e.g., by increased prolactin secretion, by blockage of thyrotropin-releasing hormone secretion, by reduced sensitivity to thyroxine, or by other means), so the differences in the genetic basis of pedomorphism likely translate into significant differences in the biochemical processes undergone by individuals from the two populations (Harris et al., 1990; West-Eberhard, 2003). Further, the two populations of salamanders are such that the females typically produce approximately 18 eggs each; however, once more the genetic basis for this property differs between the two populations (Harris et al., 1990). Such examples abound. Phenotypic traits are the upshot of complex biochemical processes controlled in most cases by a number of genes. Differences in the genetic level need not translate into differences in the biochemical processes, and even differences in the biochemical processes need not translate into differences in phenotypic traits (see, e.g., Zuckerkandl & Villet (1998) for detailed discussion of the latter point; for example, in some cases there may be an important tradeoff between chemical affinity and chemical concentration). More generally, biologists and geneticists often make use of the notion of canalization the idea that the development of a phenotypic trait, particularly one that is important to survival, can arise despite variation in genes, environment, and the resulting developmental pathways. Canalization of a trait insures that the trait is stably expressed in the face of underlying genetic variation (e.g. Gibson & Wagner, 2000; Waddington, 1942). 22 To put the point in what is perhaps more familiar terminology, we might say that phenotypic traits often exploit a certain multiple realizability at the microstructural level. Notice that it would not be an adequate defense of Devitts idea to say that the human female, the trout, the rabbit, the guinea pig, and the salamander may be the less common sort of case, and that phenocopying and canalization may not be universal phenomena. Whether or not this is true, it is not the sheer number of such cases, but the way in which they subvert the idea
22
Consider, for example, the wide range of developmental pathways that can all lead to typical external female genitalia, as discussed above.
that species as such must meet the relevant (quint?)essentialist standards for being a respectable natural kind. The general theoretical point is this: Every macroscopic phenotypic property depends on a massive number of biochemical reactions, originating with the genes themselves but continuing along the entire developmental pathway; at each point potentially subject to environmental influences, influences from other genes, and so on. If there was a one-to-one correspondence between sameness of macroscopic phenotypic properties and sameness of biochemical pathway, then simply put there would be considerably more phenotypic variation than there in fact is. The idea that one may generally infer shared underlying features from a shared phenotypic feature just does not pass empirical muster. Nature is, fortunately, far more robust. Extrinsic Species-Essence and Plenitude If species do not have intrinsic essences, if there are no general microstructural genetic facts of the sort that Devitt, Putnam and (more offhandedly) Kripke posit as making for species membership then does this mean that species have no essences whatsoever? Recently, several philosophers of biology have argued that species can indeed be said to have essences, albeit extrinsic ones. For example, Samir Okasha (2002) and Joseph LaPorte (1997, 2004) and others have argued that species do indeed have essences, in the sense that there are necessary and sufficient conditions for belonging to a given species. These necessary and sufficient conditions are relational, however, and do not yield the result that species-membership is an essential property of the individual (e.g. it will not be an essential or even necessary property of Socrates that he be a member of Homo sapiens, contra Wiggins (1980)). Nonetheless, if such species essences are to be found, then there would at least be a determinate same animal relation, of the sort that Kripke and Putnam appeal to. (This result would be arrived at only on the basis of some charitable reconstruction in particular we would need to supplement the discussion of ostensive definition with the notion that the same animal relation in fact amounts to the same species relation, which is in effect to grant a solution to the qua problem (Devitt & Sterelny, 1987). Since basic-level kinds do not map neatly onto species, or in some cases onto any taxon (Dupr, 1993), this is already quite a generous amendment to the view.) Broadly speaking, on most of the popular accounts, to be a member of a given species, e.g. Panthera tigris, is to be part of a particular chunk of the genealogical nexus, which begins with a particular speciation event, and ends with either another speciation event, or with extinction. A more specific version of the species concept is the phylogenetic (or cladistic) species concept. LaPorte offers the following as an elaboration of the phylogenetic species concept: a species name like Panthera tigris is to be defined something like as follows: Panthera tigris =df the lineage descending from ancestral population P and terminating in speciation or extinction, P being an appropriate population in the lineage that gave rise to todays tigers (2004, p. 54). Of course we are still owed an account of what a speciation event consists in, as many philosophers of biology have noted (e.g. Dupr, 1993; Kitcher, 1984; LaPorte, 2004; Okasha, 2002; Sterelny & Griffiths, 1999). There are two dominant lines of thought here: that speciation events occur when populations become reproductively isolated from each other, and that such events occur when populations come to occupy distinct ecological niches. Both lines of thought are frequently appealed to by biologists and zoologists, and each plays an important theoretical and explanatory role in at least some contexts of biological inquiry.
Before considering the significance of having multiple scientifically useful ways of characterizing speciation events, it is worth briefly noting that, whatever details one embraces, species membership is not an intrinsic matter (see, e.g., Dupr, 1981, 1993; Kitcher, 1984; LaPorte, 1997, 2004; Okasha, 2002; Sober, 1980; Sterelny & Griffiths, 1999). For example, the dominant accounts of the species concepts deliver the result that, if two populations are geographically separated in such a way as to prevent interbreeding, and result in the two populations occupying distinct ecological niches, then these populations constitute distinct species (at least on many tenable construals of these speciation conditions). No reference is here made to the intrinsic properties of the members of the species; thus the possibility remains that two individuals with identical intrinsic properties could belong to distinct species (as a result of belonging to such geographically separated populations). As another (somewhat simplified) illustration, consider the case of an individual with significant genetic mutations that render the individual incapable of either interbreeding with his parents population or occupying their ecological niche. If there are other similar individuals around to form a distinct interbreeding population or co-occupy a distinct ecological niche, then a speciation event may be underway, and the individual sometimes termed a hopeful monster -- may count among the first population of a new species. If no such fellows are to be found, however, the individual will simply count as an atypical and infertile member of his parents species. 23 Thus, even if we grant that we may be able to specify necessary and sufficient conditions for membership in a species, we are very far indeed from the sort of essentialism Kripke and Putnam had in mind (LaPorte, 2004; Okasha, 2002). By now it should occasion no surprise to note that the pursuit of extrinsic essential conditions for species membership quickly runs into its own complications. As the above discussion indicates, the phylogenetic species concept can be supplemented in at least two useful and explanatory ways that is, one can characterize speciation events as arising when two populations cease to interbreed, or when two populations come to occupy distinct ecological roles. These two characterizations of speciation events can lead to competing results: for example, it is possible for two populations to occupy different ecological niches and yet still interbreed, and so on. Further, the phylogenetic species concept (with its premium on shared ancestry) is not even the only concept on the table; there is, for example, the biological species concept (BSC), which takes interbreeding to be the sole determiner of species membership. These two different species concepts may deliver different results, even if the phylogenetic species concept is supplemented by the interbreeding approach to the speciation question. As an illustration, suppose a species splits into two (i.e. non-interbreeding) species S1 and S2 at time t1. Then later at time t2, S1 and S2 cease to be reproductively isolated from each other, and form an interbreeding population. The phylogenetic species concept will still entail that there are two distinct species at t2 (distinguished by their different ancestry going back to time t1), whereas the BSC in its most straightforward form would posit a single species at t2. A further species concept takes occupancy of an ecological niche to be the sole determiner of species membership, rather than simply a determiner of speciation, and so on so forth. 24
23
Devitt (2008) writes as though the standard species concepts are not in conflict with his Intrinsic Biological Essentialism. The fact that the standard species concepts allow for such cases suggests otherwise, however (Barker, 2010; Ereshefsky, 2010). Thus Devitts proposal is highly revisionary. 24 Devitt (2008) correctly points out that these species concepts fall short of telling us what it is to be a tiger that is, while they may tell us what it is for a population to be a species, or what it is for two organisms to be conspecific, they do not tell us what it is for an organism to be a tiger (or a lion or a zebra, and so on). Some qualification is in order here: we cannot produce any such purely qualitative specification of the essence of a kind like the tiger or the
Notably, despite the fact that these competing concepts deliver different results about whether two individuals are conspecific or not, and also about how many species should be recognized, and so on, they all have a claim to being scientifically useful. This has led a number of prominent philosophers of biology, beginning with Philip Kitcher (1984) and John Dupr (1981, 1993) to argue for pluralism concerning species concepts (see also Ereshefsky, 1998; Holter, 2009; Rosenberg, 1994; Stanford, 1995, and others). 25 The extent of the pluralism is not limited to the range of species concepts laid out above either; for each of the seemingly particular species concepts, there are any number of ways of spelling out the details. Kitcher notes this in a recent paper: I proposed [in 1984] that there were many different species concepts, appropriate for different purposes of inquiry Both Dupr and I, however, tended to think in terms of manageable pluralism, or limited promiscuity; for my part, I took the Biological Species Concept to be one among a number of contenders. The real trouble, however, is that the Biological Species Concept itself allows for indefinitely many ways of development, depending on how one approaches the notions of population and of reproductive isolation (2007, pp. 300-301). The arguments of Kitcher and his colleagues on this point seem to me completely convincing: if we seek to characterize species to give necessary and sufficient conditions for membership in a species then we will need to countenance of plenitude of sets of such conditions, each of which is scientifically useful and appropriate. Thus even the most charitable attempts to reconstruct essentialism about species face the problem of there being too many candidate species essences for us to be able to say: here is the essence of the kind Panthera tigris the set of properties that all and only actual and possible members of the kind possess. Biological science has not uncovered such hidden conditions that properly govern our use of the term tiger; biology does not issue in a privileged same animal relation. At best it has uncovered a multitude of distinct sets of conditions and relations, each with equal claim. 26
lion. However, if we are allowed to directly refer to particular individuals e.g. a particular founding population then we can provide such necessary and sufficient conditions (e.g. to be a tiger is to be descended from this ancestral population prior to any further speciation events occurring among the populations descendents). Devitts point here is correct, though I disagree with his interpretation to the effect that this indicates a significant hole in biology/philosophy of biology. An alternative interpretation would simply be that, in light of the best science and philosophy, we have found that full-blooded, purely qualitative necessary and sufficient conditions are only to be found at the level of what it is to be a species, or what it is to be conspecific, not at the level of what it is to be a tiger. Of course, since tigers constitute a species, this places some necessary conditions on being a tiger, though they will fall short of being also sufficient conditions. This state of affairs parallels a relatively common view in the discussion of philosophical essentialism; many philosophers hold that we can provide qualitative necessary and sufficient conditions for, e.g., being a person (where this may be distinct from being a member of the species Homo sapiens; compare the literature on personal identity), and further that Socrates is essentially a person. However, they do not hold that there are qualitative necessary and sufficient conditions for being Socrates (though there may be non-qualitative ones, e.g. coming from this sperm and this egg). The situation in philosophy of biology does not seem worryingly different. 25 There are a number of points of disagreement and discussion in the literature, for example, whether pluralism is compatible with realism about species, and I will not attempt to reconstruct all the subtleties here. 26 I believe that this is anyway a widespread consequence of philosophical essentialism in its most general form i.e. the thesis that there are two ways in which an item may have a property, namely essentially or accidentally. Put very crudely, where one might suppose there is a single item with n properties, essentialism has difficulty avoiding the consequence that there is something on the order of the cardinality of the power of set of n items instead,
Chemical Kinds and Essence Kripke/Putnam essentialism about biological kinds looks to be untenable, even if one attempts to reconstruct it in the most charitable of ways. But what of chemical kinds, such as water and gold? Surely here the world cooperates with our quintessentialist intuitions. After all, dont we know the essences of water and gold to be H2O and the element with atomic number 79 respectively? More generally, has not chemistry actually discovered a privileged same substance relation (namely being either the same element or the same compound) one which could be pressed into duty in the following sort of manner: Suppose I point to a glass of water and say this liquid is called water My ostensive definition has the following empirical presupposition: that the body of liquid I am pointing to bears a certain sameness relation (say, x is the same liquid as y, or x is the sameL as y) to most of the stuff I and other speakers in my linguistic community have on other occasions called water The key point is that the relation sameL is a theoretical relation: whether something is or is not the same liquid as this may take an indeterminate amount of scientific investigation to determine. Moreover, even if a definite answer has been obtained either through scientific investigation or through the application of some common sense test, the answer is defeasible: future investigation might reverse even the most certain example (Putnam, 1975, pp. 224-5). 27 The idea is elaborated in the following manner: [A] liquid with the superficial properties of water but a different microstructure isnt really water Suppose, now, that I discover the microstructure of water that water is H2O. At this point I will be able to say that the stuff on Twin Earth that I earlier mistook for water isnt really water (Putnam, 1975, pp. 232-3). As this passage makes clear, Putnam believes that the same substance relation is determined by microstructural properties, and further that H2O specifies such a microstructure. Also emphasized in these last two passages is the notion that these microstructural properties these determiners of the same substance relation are discovered. The same theme is also found in Kripke, who makes it clear that we may not only discover, contrary to what we believed, that two substances are distinct, but also that two substances are in fact one and the same:
differing only in which of the n properties are had essentially vs. accidentally. While a number of refinements are in order, essentialism without plenitude on a vast scale is, I think, an untenable position. For the details and the arguments, see Leslie (2011, submitted). 27 Here and elsewhere, Putnam speaks of a same liquid relation. The appeal to chemistry is more plausible, however, if this is characterized as a same substance relation instead. For example, the dominant use of the term water in chemistry is a phase-neutral one, so that water can also occur in solid and gas phases. Nothing in the criticisms of essentialism that follow hang on this adjustment; if anything, this adjustment makes the case for essentialism more plausible. For example, the oft-made claim that water is identical to H2O is clearly false as stated if water is a phase-specific term. For then, by parity of reasoning and the transitivity of identity, the absurdity that water is identical to ice is derivable. See Johnston (1997). In what follows, I will use the term water in a phase-neutral sense, unless otherwise specified.
[I]f this substance [H2O] can take another form such as the polywater allegedly discovered in the Soviet Union with very different identifying marks from that of what we now call water it is a form of water because it is the same substance, even though it doesnt have the appearance by which we originally identified water (1980, pp.128-9). And more generally: [S]uppose some items (let the set of them be I) are discovered and are believed to belong to a new kind K. Suppose later it is discovered that the items in I are indeed of a single kind; however, they belong to a previously known kind, L. Observational error led to the false initial belief that the items in I possessed some characteristic C excluding them from L (1980, p. 136, emphasis added). The basic idea behind this essentialist model for chemical kinds is clear: there is a privileged relation same substance as that science discovers, and which determines the extension of our natural kind terms. Further, this scientific discovery has already been made, at least in the case of water and gold, which have been found to be identical to H2O and the element with atomic number 79 respectively. It may now seem reasonable to infer from these examples and others that, for chemical kinds, the same substance relation can be characterized as a) same atomic number in the case of elements, b) same chemical formula in the case of compounds. Such microstructural essences are the determiners of the extension of chemical kind terms, and macroscopic manifest features are at best a guide to these true essential features. We can now pose the question: does science actually deliver such a privileged same substance relation, or do we simply have once again the quintessentialist intuition that science must do so? That is, does the Kripke/Putnam discussion simply retell the scientific facts, or is the science once again distorted, perhaps by quintessentialist convictions? In this section I will argue that the relevant sciences deliver no such privileged same substance relations. The idea that same atomic number/same chemical formula captures this relation simply will not do, especially in the case of compounds; nor is there is a unique privileged substitute relation. At best, science delivers a number of candidate same substance relations, many of which rely on the macroscopic and the manifest rather than the microstructural. Finally, I discuss some examples that suggest that the chemical kinds of thought and talk may not line up especially neatly with the kinds of chemistry, suggesting once again that the extensions and truth conditions of our thought and talk are governed more by our false quintessentialist mind-set than by any scientifically discoverable essentialism, contrary to what philosophers often suppose. 28 Essentialism and Compounds Let us begin with the case of water, which we are told, is identical to H2O. One initial observation that has not been properly assimilated is that H2O is not in fact a microstructural description (Needham, 2000, 2002, 2011; van Brakel, 1986). It is simply a compositional formula, specifying the proportion of hydrogen to oxygen that is to be found in the substance. It
28
This next section draws on the excellent work of a number of philosophers of chemistry, including Michael Weisberg, Joseph LaPorte, Jaap van Brakel, and especially on the exemplary work of Paul Needham, whose papers should be mandatory reading for philosophers working on natural kinds. The presentation and many of the illustrations are my own, however.
is not a specification of the molecular structure of water; in fact, it can apply to phases of water that have no molecular structure such as ice X, as noted below. Moreover, historically speaking compositional formulas were introduced and widely used at a time before atomism in chemistry was generally accepted; in this sense, compositional formulas do not even involve a commitment to the existence of atoms or molecules, the actual constituents of microstructure (Needham, 2000). Further, since compositional formulas only specify the proportion of elements in a compound, they do not even distinguish between structural isomers, which are chemically distinct compounds composed of the same proportions of elements. For example, the compositional formula C2H6O applies equally to ethyl alcohol and dimethyl ether, but these two compounds are distinct along any number of chemically important dimensions. As it happens, water has no structural isomers, however many, many compounds do. (And many compounds have many, many structural isomers. For example, in the case of large organic molecules even ones that consist only of carbon and hydrogen the number of distinct compounds corresponding to a given compositional formula can number in the millions and greater (Smith, 2011).) A somewhat better candidate for capturing the microstructure of water would be a structural formula -- a formula which would specify to some approximation how the relevant molecule is structured, e.g., by indicating that the two hydrogen atoms are each bonded to the central oxygen atom. There are numerous ways of representing structural formulas, many of them graphic in nature, so as to illustrate the geometry of the bonding patterns. For simplicity, I will use H-O-H as an abbreviation for a more informative structural description of the molecule. The real complications arise when we learn that only in the gas phase does water primarily consist of such H-O-H molecules. That is, the gas phase of water can be reasonably modeled (though with some idealization) by supposing that there are a large number of discrete, separate H-O-H molecules, but this is not true for the liquid phase, nor for the many ice phases. For example, there are at least 15 different forms of ice that have been observed experimentally, each formed under different combinations of temperature and pressure. (And more are theoretically projected, including a metallic phase of ice!) Interestingly, one form of ice, ice X, has no molecular structure at all; its microstructural arrangement is such that there is no distinction between the intra-molecular bonds and the inter-molecular bonds. Rather ice X is an atomic solid, composed of hydrogen and oxygen atoms arranged in a particular lattice structure. It does not contain any molecules to be described, so the question of how to describe its molecules does not arise. It still contains hydrogen and oxygen in a 2:1 ratio, however, and so falls under the compositional formula H2O, despite lacking any molecular structure. In the liquid phase, there is molecular structure, but unlike the gas phase, only a subportion of the molecules are H-O-H molecules. For example, some of the H-O-H molecules dissociate into H+ and OH- ions. Some of these ions attach to H-O-H molecules to form complex ions, e.g. H3O+ and (H2O)OH- (since the complexity of the molecules is rapidly increasing, I will use condensed structural formulas such as these). Other molecules bond together to form polymers (i.e. chains of repeating structural units) of arbitrary length: two H-O-H molecules will combine to form a (H2O)2 molecule, which can combine with another H-O-H molecule to form (H2O)3, and so on and so forth. The polymers and the ions can further combine, so in a given sample of liquid water there may be molecules of the form (H2O)nH+ and (H2O)nOH- for any reasonable n. The patterns of disassociation and bonding in liquid water happen continuously what is a polymer one moment may disassociate into smaller parts the next, and so on (see Needham, 2000, 2011 for more discussion).
This is the beginning (and just the beginning!) of a serious description of the microstructure of water. The idea that H2O constitutes a description of waters microstructure may have its roots in the incorrect idea that the liquid and solid phases of water (and other substances) are basically like the (still somewhat idealized conception of the) gas phase which we may say is composed by H-O-H molecules. That is, the idea may persist that the liquid and solid phases have the same microstructural composition, only with the molecules staying closer together in these phases. In fact the important manifest properties of water -- e.g. its relatively high boiling point, which ensures it is liquid at standard temperatures and pressures and therefore drinkable; the fact that ice is less dense than liquid water, which means fish can survive the freezing of lakes by swimming lower; and so on are due specifically to that fact that liquid and solid water do not simply consist of H-O-H molecules, but rather a variety of dynamically related complex molecules. In light of this complexity, what are we to make of the very idea of a microstructural essence for water and other chemical kinds? As a general rule, a compositional formula like H2O does not provide a good candidate for capturing the essence of a substance. For example, recall that the compositional formula C2H6O applies equally to ethyl alcohol (condensed structural formula C2H5OH) and dimethyl ether (condensed structural formula (CH3)2O), but these two compounds are quite dissimilar (and certainly considered distinct by chemists), and we should not want the result that they are somehow to be said to share a microstrucural essence, and so be the same substance. In general, a compositional formula does not specify a unique chemical kind; in fact, in some cases the same compositional formula applies to millions of distinct chemical kinds (Smith, 2011). But on the other hand, relying on a structural formula to specify the composition of the constituent molecules will also be problematic as general strategy, as the case of water illustrates if we attempt to insist on something to the effect that water is essentially composed of H-O-H molecules, then we will derive the result that water exists only in the gaseous phase. And certainly, no description of molecules will suffice to characterize ice X, since it does not contain molecules. The Same Substance Relation Perhaps the solution will be to examine the microstructure of water and other chemical kinds in even greater detail, and hope that such examination will uncover the elusive essence of chemical kinds. Returning to the case of liquid water (which, remember, was supposed to be the easy case, the poster child for Kripke/Putnam essentialism!), we can begin by noting that the rates and patterns of polymer formation and ionic disassociation are sensitive to the temperature and pressure. Thus, if we take seriously the idea of deriving detailed microstructural descriptions, water at 4 degrees C will have a considerably different microstructure than water at 98 degrees C. Any microstructural specification of the essence of water would have to take account of this variation without delivering the result that these are different substances. As Needham puts it, In view of the sensitivity of the distribution of particular determinate forms of the many determinable features to prevailing conditions, the actual description of the relevant microscopic structure which must be sufficiently relaxed to accommodate all these variations without including too much would be a truly daunting task But first and foremost, it should be asked why this vast range of microscopic structures is associated with one and the same substance kind, rather than a genus of related substances. (2000, p.20).
Given the extensive variation at the microscopic level, is water even a single substance? And more generally: what criteria, if any, does chemistry give us for determining how many different substances are present in a given sample or put differently, does chemistry give us a way of determinately answering Putnams question is x the same substance as y? We have seen that neither sameness of compositional formula nor sameness of molecular structure provides a satisfying answer to the question (because of the possibility of isomers and the possibility of massive variation on the microstructural level respectively). As I understand Needhams work, the real answer is two-fold: 1) such criteria for sameness are far more readily found at the macroscopic level than at the microscopic level, and 2) there is nonetheless no unique and privileged way of answering the question at either level. Put in a slightly simplified way, the point is this: When it comes to manifest kinds, chemists typically find unexpected complexity and variability in the constituents of the examples of the kinds in question, and they are fully prepared to note this complexity and variation without any surprise, because they treat the manifest and macroscopic features as partly determinative of what it is to be, for example, water or gold -- contrary to what Kripke and Putnam envisaged. Chemistrys partial reliance on and respect for the macroscopic and the manifest is at odds with the Kripke/Putnam project. 29 Consider their repeated reference to common microstructure as wholly determinative of the extension of our natural kind terms. Consider, for example, Kripke in the following passage: what I would have wanted to do would have been to discuss the part about gold being a metal. This, however, is complicated because first, I dont know too much about chemistry. Investigating this a few days ago in just a couple of references, I found in a more phenomenological account of metals the statement that its very difficult to say what a metal is. (It talks about malleability, ductility, and the like, but none of these exactly work.) On the other hand, something about the periodic table gave a description of elements as metals in terms of their valency properties. This may make some people think right away that there are really two concepts of metal operating here, a phenomenological one and a scientific one which then replaces it. This I reject (1980, pp.117-8). The thought is that the valency properties or something like that is what makes things deserve the name of metals, rather than the macroscopic features of malleability, ductility and the like. In fact, there are at least two scientific concepts of metal in play in chemistry one denoting a set of elements in the periodic table, the other applying when a solid of any atomic constitution displays certain macroscopic properties as is illustrated by the hypothesis that there exists a metallic phase of ice. This latter is characterized in macroscopic terms, but is nonetheless an important scientific notion, despite Kripkes suggestion to the contrary. This dichotomy between the scientific and the phenomenological especially if the latter includes notions such as
Putnam, especially in his later work (but also in The Meaning of Meaning p. 239), does not seem to take such a consistently hard line on this, and allows instead that our interests also matter when determining whether one thing counts as the same substance as another. See Hacking (2007a) for a helpful discussion. Again, here and throughout the paper, use the Kripke/Putnam project to denote the one that is rooted in Naming and Necessity and The Meaning of Meaning as it is widely interpreted that is, the view that has been so historically influential in philosophy.
29
malleability and ductility is artificial and cannot be sustained. Indeed, there is an interesting further question as to whether the elements that are called metals are so-called because they typically form metallic solids at normal temperatures and pressures, where what counts as a metallic solid is phenomenologically determined, i.e. determined by such properties as malleability, ductility and the like. Returning to the question of the same substance relation, one promising place to look is what is known as Gibbs phase rule, which relates 1) the number of independent substances in a closed system (thus offering a way to count substances), 2) the number of phases in the system (e.g. solid, liquid, etc), and 3) the number of intensive variables (e.g. temperature, pressure), which can be varied independently. The phase rule does deliver the result that water is a single substance. 30 However, it must be understood that Gibbs rule counts independent substances. If multiple substances are related to each other within the system by a reversible chemical reaction, the phase rule will count fewer substances than seems correct. Consider, for example, what happens to a system that contains H2O heated to around 2000K (Here I follow Needham, m.s.). The H2O begins to dissociate into H2 and O2 molecules, which in turn recombine into H2O molecules. The phase rule, along with empirical observations about the behavior of the system, delivers the result that there is only one substance in the system, despite the fact that in general H2, O2, and H2O are not the same chemical kind. What constitutes a distinct independent substance in a system is not in general the same as what constitutes a distinct substance simpliciter. The fact that the phase rule deems water to be a single substance could thus be
The phase rule is as follows: cf+2=V If V that is, that we have but a single substance in the system then the phase rule allows us to make empirical predictions about how that substance will behave with respect to its different phases, as a function of temperature and pressure. For example, if we have but a single substance in the closed system, Gibbs phase rule lets us derive that there will be a unique combination of temperature and pressure at which three phases of the substance will be at equilibrium that is, there will be a specific temperature and a specific pressure at which, e.g., solid, liquid, and gas phases can all coexist in equilibrium (V = 0, so no intensive variables can be varied; this is sometimes known as the triple point). The phase rule also lets us predict that, if c = 1, there will be a range of temperature and pressure pairs at which two phases (e.g. solid and liquid) can co-exist in equilibrium, but also that temperature and pressure cannot be varied independently. However, if again c = 1, but only one phase is present, the temperature and pressure can each vary independently. To return to the question of whether water is a single substance or not, the answer is that, according to Gibbs phase rule, it is. The behavior of water fits this thermodynamic profile of a single substance. As Needham illustrates: [S]uppose the two-phase quantity of water [liquid and gas] is confined at fixed temperature to a closed container fitted with a piston. Any attempt to decrease the pressure by raising the piston and increasing the volume available to the water will fail (as long as two phases remain), because the liquid phase will decrease in volume as matter goes into the gas phase to maintain the pressure. Similarly, attempting to increase the pressure by decreasing the volume will be thwarted by the volume of the gas decreasing as matter goes into the liquid phase to maintain the pressure. Continuing the processes of increasing or decreasing the volume will eventually lead to a single phase being formed (gas in the first case, liquid in the second), which is bivariant, so that pressure and temperature can vary independently. This behaviour is in accordance with the phase rule for a system for which the number of independent substances, c, is one (m.s., p.6). This macroscopic criterion for individuating substances thus delivers the result that water is a single substance.
30
interpreted in the following way: water is a conglomeration of distinct substances, but because they dissociate and recombine as they do, the distinct substances are not independent. An alternative way of distinguishing substances is to consider the entropy of mixing. Suppose we have two separate containers of gas sitting side-by-side, and we then remove the divider between them. If the two gases truly consist of the same substance, there will be no increase in entropy as a result of allowing them to combine. If, however, they are distinct substances then even if they are highly similar the entropy of the system (a measurable quantity) will increase (Needham, 2000). This turns out to be an exceedingly fine-grained test. For example, in addition to structural isomers (e.g. ethyl alcohol and dimethyl ether), there are also spin isomers, for example ortho- and para-hydrogen. A molecule of ortho-hydrogen is one in which the nuclear spins of the two H atoms are aligned parallel; a molecule of para-hydrogen is one in which the nuclear spins are anti-parallel. What is termed hydrogen gas and denoted by the formula H2 normally consists of a mixture of these two spin isomers (at a ratio of about 1:3 of para-hydrogen to ortho-hydrogen). The two spin isomers of hydrogen can be separated in a laboratory situation, however. Further, if a container of ortho-hydrogen is allowed to mix with a container of para-hydrogen, then the entropy of the system will increase; by this test, there are thus (at least) two distinct substances in a normal sample of hydrogen. There are also spin isomers of water, ortho- and para-water, which are distinguished in the same way, depending on whether the spins of the hydrogen nuclei are aligned or not in the HO-H molecule (and in H-O-H links in the polymer chains, and so on). Just as with hydrogen, normal samples of water contain both spin isomers, in a ratio of 1:3 (para to ortho). For various reasons, ortho- and para-water have proved more difficult to separate than ortho- and parahydrogen, however several groups of researchers have recently succeeded in doing so (Kravchuk, Reznikov, Tichonov, Avidor, Meir, Bekkerman & Alexandrovich, 2011; Tikhonov & Volkov, 2002). By the entropy of mixing test, ortho- and para-water count as distinct substances, suggesting that normal samples of water are in fact mixtures of two distinct substances. This disagrees with the results of the phase rule test, however, which suggests that water comprises a single substance. So, are ortho- and para-water really distinct substances? The answer, in some sense, is clearly yes. For example, efforts to separate them are clearly not fools errands attempts to separate a single thing into identical components. On the other hand, it is a distinction that is not generally made water is normally treated as a single substance in chemistry. Tikhonov and Volkov, the first researchers to separate these spin isomers, found that the separated ortho- and para-water took 55 minutes and 26 minutes respectively to settle back into the normal 3:1 ratio when kept in a liquid state (far longer than they had expected), and estimated that ortho- and para-ice frozen at -18 degrees C would take several months to return to the normal ratio. They also note that [t]he OP [ortho/para] separation procedure realized is quite straightforward and may occur in naturein soil, atmosphere, living organisms, and cosmic objects. The scope and the role of this phenomenon are yet to be studied (p. 2363). It would thus not be totally beyond the realm of scientific possibility (though it may stretch the bounds) to imagine a version of the Twin-Earth thought experiment that compared ortho- with para-water. Would the visitors from Ortho-Earth be making an error when they first applied the word water to the substance on Para-Earth? Do these spin isomers bear the same substance relation to each other, or not? By some tests, yes (certainly both are H2O); by other tests, no. There are scientifically useful ways of dividing up substances according to which they are the same substance, but there are other scientifically useful ways of dividing up substances
according to which they are distinct. Nor are we awaiting the results of further investigation to settle which way of dividing up substances is the scientifically real one. A similar example which is often discussed in the literature concerns isotopic variation (LaPorte, 2004; Needham, 2008; Weisberg, 2005). Two atoms are isotopes if they share the same atomic number (number of protons in the nucleus) but differ in mass due to the number of neutrons in their nuclei. Elements are standardly defined in chemistry by their atomic numbers, thus two distinct isotopes are nonetheless examples of the same element. Hydrogen atoms, for example, usually have a single proton and no neutron in their nuclei (protium), however there are two naturally occurring stable hydrogen isotopes, one with a single neutron (deuterium), the other with two neutrons (tritium). Thus, protium, deuterium, tritium all fall under the same element kind (though of course they can obviously be distinguished from each other). Isotopic variation does not tend to affect the qualitative nature of the chemical interactions if the most common variant of an element enters into a given chemical reaction, so will its isotopes. Sometimes isotopic variation can affect the speed of a chemical reaction, but much of the time the effect is negligible. Thus the addition, as it were, of a neutron in contrast to the addition, say, of a proton has little effect on an atoms chemical behavior. In this way, identifying elements with their atomic numbers regardless of isotopic variation is a very reasonable and useful way to classify elements and the compounds containing them. In the case of hydrogen, however, isotopic variation has non-trivial effects due in part to the fact that hydrogen atoms are so small (e.g., deuterium has roughly twice the atomic weight of protium). Consider, then, heavy water, which has deuterium in place of protium. Is heavy water the same as regular water? By the most standard ways of categorizing compounds which is insensitive to isotopic variation they would count as the same substance. (Though of course by other ways of categorizing, they would be distinct for example, there is an entropy of mixing associated with combining isotopic variants. Nothing in the observation that elements group isotopic variants together implies this is somehow a distinction without a difference.) Imagine, then, a miniature version of Twin-Earth, in which a lake that contains heavy water is newly discovered, and let us pose the question of whether the substance in it is the same substance as the water encountered thus far. It should be clear from the foregoing that chemistry does not deliver a univocal answer to the question. Let us consider instead, then, how the members of population at large would be inclined to answer the question. While this paper in general cautions against reliance on intuition in philosophy, this would seem one occasion where we might nonetheless confidently intuit how this would go: the substance in this newly discovered lake would most certainly not be judged to be the same substance that comes out of our taps, etc (LaPorte, 2004; Needham, 2008; Weisberg, 2005). This is for the straightforward reason that heavy water, when drunk in suitable quantities, causes sterility, neurological problems, and death. If approximately 25% of ones daily intake of protium oxide is replaced by deuterium oxide, one rapidly becomes sterile. If higher proportions of deuterium oxide are consumed, this soon leads to acute neurological symptoms, liver hyperplasia, anemia, other symptoms, and eventually death (Kushner, Baker & Tunstall, 1998, p. 81). These effects, broadly speaking, are due to the fact that deuterium in heavy water exchanges with protium in organic molecules in the body. The bond between carbon and deuterium is considerably stronger than the bond between carbon and protium, which slows and otherwise disrupts a number of biochemical processes (Kushner et al., 1998). I submit that we would not consult with chemists before concluding this substance was most certainly not the same substance as our water. Suppose an overenthusiastic philosopher
should try to argue as follows: in general isotopic variation has very little impact on chemical properties, and so elements are identified by their atomic numbers, and compounds by the elements in them, and so it has been discovered that the substance in this new lake is in fact the very same substance as the one we are used to drinking! An apt response to the philosopher here might be: well then, drink up! The moral of this particular tale is that there are occasions on which chemistry does not deliver a univocal answer to the same substance question, yet the significance of the manifest properties of the substance nonetheless do provide such an answer. We have seen thus far that chemistry does not provide a privileged same substance relation. Further, in the case of compounds, we also saw that there is no generally effective way of specifying the essence of the compound (even abstracting away from isotopism and spin isomerism) simply specifying the composition of the compound in terms of its constituent elements is too broad (since it does not distinguish structural isomers), yet specifying the structure of the component molecules cuts too finely (since, e.g., water in its liquid phase is composed of many, many different types of molecules, while in its ice X phase it is not composed of molecules at all). Essentialism and the Elements What, though, of the elements? Certainly, the elements are more hospitable to essentialism than the compounds: in the context of chemistry, elements are quite clearly identified by their atomic numbers. (Again, though, as the case of isotopes illustrates, this does not clearly constitute a privileged same substance relation, as opposed to a same element relation.) Of course, for the Kripke/Putnam thesis to be correct, it is not enough that science provide the essences of kinds, it also has to be the case that these essences determine the extension of the relevant terms/concepts. Thus, an interesting question to explore is the extent to which the natural language terms/concepts for elements have their extensions determined simply by the relevant atomic number. Consider, for example, Kripkes assertion that a material object is (pure) gold if and only if the only element contained therein is that with atomic number 79 (1980, p. 138). Since the word gold is used in scientific contexts as the English name for the element Au, there is a sense in which something like this biconditional is unassailable as a necessary truth this is the sense in which something is copernicum if and only if it has atomic number 112. The term copernicum is not a term of ordinary non-technical language; copernicum is not a manifest kind rather it specifically names a certain element: the one with atomic number 112. That gold can have such a usage also is not something I would dispute. But Kripke intends to make a far more substantive claim: in fact he refers not to elements, but to material objects made of gold, and thus intends his statement to apply to the manifest kind gold -- the kind which has played such an important role in our economic history. Golds historical and economic significance depends, of course, on its manifest properties e.g., its lack of chemical reactivity make it very stable and free from corrosion, its luster and general appearance make it desirable as jewelry and a store of wealth, while its malleability/ductility make it relatively easy for artisans and currency minters to work with. But the manifest properties of an elemental substance depend on more than simply its atomic constituents; they also depend on how those constituents are arranged. When the atoms of a single element bond in different structural patterns, this is known as allotropy. As it happens, Au does not exhibit allotropy at normal temperatures and pressures, but rather forms only what is known as a face-centered cubic lattice, which accounts for many of the substances important
manifest properties, such as its ductility and malleability. It is not, however, necessary that Au atoms bond this way; for example, at very high pressures, Au atoms form instead a hexagonal close-packed lattice. The resulting substance is far less malleable and ductile than the facecentered cubic lattice of normal temperatures and pressures. If such allotropes of Au existed at normal temperatures and pressures, would we be so quick to assert that being composed of Au atoms is necessary and sufficient for falling under the manifest kind gold, despite lacking these economically significant properties? Or consider the element tin (Sn), which has in addition to its familiar shiny, metallic, malleable form, also has a non-metallic allotrope (stable below 13.2 degrees C), which is dull, brittle, and powdery. What if Au displayed such allotropic variation? Would such a dull, brittle and powdery substance fall under the manifest kind gold? One version of this question is a linguistic question: whether the word gold would be applied to such a counterfactual substance. A further question is whether this stuff would be considered the same substance as regular gold for example, whether it would be assimilated into the same economic role as regular gold. Whatever the linguistic outcome would be in such circumstances, the economic outcome seems quite easy to predict: this counterfactual nonmetallic Au would not play the same economic role as regular gold. (Of course if there were a cheap and easy way to convert the one to the other, that would change the economic role of the former, but only in a way that is parasitic on the role of the latter.) What of the linguistic question? My view here is that the linguistic question is less probative than it is often taken to be, in large part because like most linguists who study lexical semantics I believe that polysemy is a widespread and important phenomenon. That is, a single phonological form, e.g. /gold/, may be associated with a range of related (sometime very closely related) but nonetheless distinct meanings. (Polysemy differs from lexical ambiguity, since in the latter case, the meanings are typically not at all related.) Thus, in the counterfactual circumstances we are considering, the possibility that the phonological form /gold/ might come to apply to the non-metallic allotrope would not show that the two were taken to belong to the same manifest kind. In this respect, economic behavior is more telling since it requires one to to put your money where your mouth is, as it were. A real-world example of this form concerns the allotropes of carbon. Carbon has many allotropes, including graphite and diamond. It is an undeniable point that we do not consider graphite and diamond to belong to the same manifest kind, and the discovery that both consist of the 6th element of the periodic table does not alter this one bit (Johnston, 1997; LaPorte, 2004). It was a surprising chemical discovery that diamond contained only the humble carbon atom, but this did not somehow undermine diamonds economic role. In the case of the 6th element, distinct phonological forms (at least in English) are used for each of the allotropes, and a further one (/carbon/) for the element itself. Had history gone differently, so that this 6th element was first discovered in the context of analyzing diamond, it may have been that this 6th element would have been named diamond also, paralleling the situation with gold. The surprising discovery would then have been that lowly graphite contained diamond atoms. This is a counterfactual history in which diamond is more polysemous than it actually is, not one in which the manifest kind diamond was discovered to include just anything composed of the 6th element. For another illustration of polysemy as it applies to terms which have a use in which they pick out elements, let us consider the phonological form /sodium/. As background information, this phonological form applies to the 11th element of the periodic table. Pure macroscopic quantities of this 11th element can take the form of a metal, which is so reactive that it almost
never occurs naturally, and must be covered with paraffin when stored in a laboratory. The phonological form /sodium/ also applies to this dangerous manifest kind. More commonly, however, the 11th element is found as a component of an ionic compound. Table salt is such a compound: it is composed of ions of sodium and chlorine. Sodium ions are typically ingested in the form of table salt, and play a number of extremely important roles in human physiology; for example, sodium ions in conjunction with potassium ions -- help maintain blood pressure. The phonological form /sodium/ is again used without qualification in everyday discourse to pick out these ions. Consider, then, a doctors injunction to increase ones sodium intake, as a means of combating low blood pressure. The doctor here is not encouraging her patient to consume a highly reactive and toxic metal, though of course there is a perfectly standard use of /sodium/ that applies exactly to such a substance. Rather, the doctor is suggesting that her patient consume more sodium ions, probably in the form of table salt. (Note that for the halogens the two senses are distinguished phonologically: e.g. chlorine vs. chloride, where the latter refers only to ions.) In fact there are at least three related but distinct uses of /sodium/ in play here: the use that concerns ions as in increase your intake of sodium, the use that concerns a metallic manifest kind as in sodium is an explosive metal, and the use that concerns an element as in sodium is the element with atomic number 11 . It would however be a mistake and in this case, a deadly one to suppose that philosophical musings on chemistry establish that one of the latter two uses employs the real sense of sodium. (Suppose someone makes this mistake and interprets his doctors advice on increasing his sodium intake as the injunction to ingest a lump of sodium metal. One would not fancy the chances of a malpractice lawsuit pursued on the grounds that sodium is univocal and therefore applies to the lump of metal.) Similar remarks apply to oxygen. Oxygen can be used to pick out the 8th atomic element, and in this sense both O2 gas and O3 gas are composed of oxygen. The manifest kind oxygen the one we must continually inhale to stay alive includes only the former gas. The latter gas, ozone, is in an excellent sense simply not oxygen. (Inhaling ozone in sufficient concentrations rapidly leads to respiratory distress and eventually death.) When a paramedic declares that a patient needs oxygen, it is clear what she means, and what she means does not include ozone. Beyond Chemistry: Quintessence and Chemical Kinds When two manifest kinds are distinguished on the basis of their observable, macroscopic properties, could chemistry (or any other science) ever actually discover that they are one and the same? Of course we might discover they are closely related in some respect, and we may introduce a term that applies to both. Since polysemy is a widespread feature of the lexicon, it is possible that this umbrella term will be phonologically the same as the term for one of the two kinds. But if the two kinds genuinely differ in their important manifest properties, what discovery could lead us to discount this difference? For one thing, even if we allow a distinction between manifest properties and scientifically significant ones, the former do not float free of the latter if there is a difference at the manifest level, there will be a difference at the scientifically significant level, and who is to say that difference is somehow unimportant? Consider, for example, the fact that chemists generally use the term water is a phaseneutral sense. This does not mean that chemistry somehow obliges us to discount differences in phase on the contrary, chemists are at times very interested in phase differences. What then of the term water? It is clearly used in a phase-neutral sense in most scientific contexts (such as
when a chemist writes a paper), but in other contexts in a phase-specific sense (such as when the same chemist requests a glass of water). The term water is polysemous in this way. Moreover, some of the best reasons for supposing that water in any sense comprises a single substance depend on its manifest and macroscopic properties, such as its behavior with respect to Gibbs phase rule. In contrast to the Kripke/Putnam view, which supposes that when it comes to the determination of reference, the manifest and macroscopic is held wholly hostage to the microstructural and microscopic, the picture that emerges in chemistry itself is one on which macroscopic and manifest properties figure prominently. The majority of examples considered thus far have been cases in which (a nave and misleading conception of) chemistry might have led us to consider distinct manifest kinds to be one and the same. What then of the converse? Are there examples in which chemistry suggests that there is more than one kind present where there is arguably but one manifest kind? If we take seriously the Kripke/Putnam emphasis on microstructure, then ortho-water and para-water provide such an example. Moreover, Joseph LaPorte (2004) elegantly and persuasively argues that the case of jade also fits this profile, and that the case of jade in China constitutes a historical Twin-Earth scenario that breaks in a way that is the opposite of what the Kripke/Putnam theory would predict. Recall that there are in fact two chemically distinct compounds with extremely similar manifest properties nephrite (Ca2(MgFe)5Si8O22(OH)2) and jadeite (NaAl(SiO3)2). For approximately 5,000 years, nephrite played an extremely important economic and cultural role in Chinese society, and was more highly prized than even gold. In 1784, the Qianlong Emperor annexed Northern Burma. The region contained a number of mines that were rich in jadeite, a substance that the Chinese had not previously encountered. Jadeite and nephrite are exceedingly similar in their manifest properties, though not strictly identical; highly skilled artisans could tell the difference, since jadeite is slightly harder and heavier than nephrite. However, both substances share their most significant and prized manifest properties; for example, their ability to sustain intricate carving. The Qianlong Emperor loved jade, and welcomed the newfound resources as more of the same substance. Jadeite was seamlessly assimilated into the same economic, artistic and cultural role as nephrite, and the term y was readily applied to jadeite and nephrite alike (Hacking, 2007b, LaPorte, 2004). 31 This example again illustrates the significance of manifest properties in determining what counts as the same substance as something else. One should not, however, be mislead into supposing that manifest properties are all that matters. On this point, I believe that Kripke and Putnam were completely correct; however, I would interpret their insight in terms of our being, at heart, Quintessentialists. We do not believe that looking like a K, feeling like a K, and so on suffices for being a K. To truly be a K requires that one have the quintessence of Ks. On this point, it is instructive to note that, while both jadeite and nephrite have been embraced as true jade, other substances have been deemed false jade, despite having very similar appearances. That is, the manifest kind jade is not simply characterized in terms of superficial descriptive properties the possibility exists for there to be exceedingly convincing cases of fake jade. (The
31
Might this also be a case of polysemy? To best assess this question, it would be important to examine whether there were any notable economic or cultural distinctions drawn between the two. If none were drawn, then it would be reasonable to suppose that this is not a case of polysemy (at least not for most speakers of the language; for the artisans the situation may be different). I cannot rule out this possibility, but I do not think that chemical differences necessitate it either.
internet is replete with advice to buyers on how to avoid being conned in this way by unscrupulous jewelers.) The situation may, I think, be characterized in the following way: we believe that there is a quintessence of jade, which is shared by both jadeite and nephrite, but not by other stones, no matter their appearance. (This leaves open whether we may believe that there are still minor quintessential differences between jadeite and nephrite perhaps in the way that animal kinds below the basic-level are believed to differ in minor ways that are unimportant for general purposes. Compare Bengal tigers and Siberian tigers.) As the case of jade suggests, quintessence does not map well onto chemical properties though of course we may be inclined to suppose it does, just as we do in the biological case. A large difference in chemical properties does not entail a difference in supposed quintessence. I would further propose that the converse holds as, e.g., the case of water and heavy water would illustrate: regardless of how chemically similar two substances may be, if one sustains life while the other ends it, these will be believed to have decidedly different quintessences. As a final illustration of quintessentialist thinking, consider the distinction between natural and synthetic diamonds. Current technological advances have made it possible to produce diamonds in the laboratory which are chemically and physically identical to naturally occurring ones. In some cases it is possible for a trained gemologist with enough time and equipment to tell them apart, but even this is not failsafe. Further, the ways in which they are distinguished often involve detecting imperfections that are more likely to be present in the natural diamonds than in the synthetic ones. It is certainly not possible for a consumer or even for the average jeweler to be able to tell them apart. Nonetheless, consumers are willing to pay far more for natural diamonds, and further the overwhelming majority of consumers choose the natural ones. These two kinds of diamonds are clearly distinguished in peoples thinking: there are websites which offer advice on how to get your fiance to accept a synthetic diamond, and conversely websites in which people (including, we may suppose, the target fiances) declare that they will never settle for a synthetic diamond. Further, the qualifier synthetic is often replaced with a more derogatory term, such as imitation, simulated, or even fake, which strongly suggests that these diamonds are not the same thing as the natural ones, but rather merely mimic their surface and chemical properties while being deeply and underlyingly different. As one jewelry website succinctly puts it, Some customers may feel content to buy simulated diamonds, knowing that they could never afford a near flawless diamond. Others though, insist upon buying only genuine diamonds, as these are a more natural treasure and very rare, especially when compared to laboratory-produced items. 32 The quintessentialist intuition here is that man-made diamonds just arent the same things as natural ones. They are no more than forgeries, simulacra of the real things, whose natures are shaped deep in the ground under high pressure over many years. No lab can confer this nature on anything, because the nature does not simply reside in the chemical structure. The man-made stones simply do not have the quintessence of diamonds. Reflecting on chemical composition and bonding structure does not silence this complaint because our quintessentialist dispositions project differences even when there is no difference in the underlying chemistry. Conclusion
32
http://www.abazias.com/diamondblog/diamond-education/the-difference-between-real-and-synthetic-diamonds
We are Quintessentialists, and we have a tendency to interpret science through the lens of our quintessentialist intuitions. Thus, at first glance, it may seem to us that science confirms our quintessentialist intuitions, when in fact the opposite is true. Closer inspection of the scientific findings shows that against the very natural and highly intuitive expectations of Kripke and Putnam -- there are not good prospects for the discovery of microstructural biological or chemical essences that determine the references of our natural kind terms. There are cases in which members of the same species have quite different DNA and conversely ones in which members of different species have quite similar DNA (recall for example the cutthroat and the rainbow trout). There are cases of sameness of manifest kind with different microstructures (as with jade) and cases of different manifest kinds with the same microstructure (as with natural and synthetic diamonds). Further, neither biology nor chemistry delivers a privileged same species or same substance relation. The highly intuitive character of the Kripke/Putnam picture derives from our quintessentialist mind set; it does not reflect scientific or metaphysical facts about the world. If we did not have something like quintessentialist beliefs then we would not have the relevant Kripke/Putnam intuitions. We do, however, have such beliefs, and we have them despite the fact that the world does not conform to them. Our intuitions here reflect only facts about us, not facts about the deep nature of reality. More generally, this discussion of one of the most intuitive views in philosophy illustrates the urgent need to examine the sources of our philosophical intuitions. In particular, it is important to understand whether a set of intuitions may be due to an early-developing and deep-rooted bias to see the world in a particular way. Such a finding does not entail that the intuitions are misleading, but it does suggest that they will be extremely compelling even if they are misleading. Thus, such a finding gives us reason to subject the intuitions and any conclusions derived from them to further scrutiny. If a conclusion is supported solely by such intuitions, this may be good reason to remain skeptical concerning the conclusion. A question remains: what if quintessentialism not only implicitly shapes our basic beliefs but also our semantic intentions; what if in using natural kind terms we often do in fact intend to refer to kinds united by a common quintessence? Science shows there are not in general good candidates to be the common quintessences, so what happens to reference under this assumption? It is either much less determinate than the work of Kripke and Putnam led us to suppose or, if this is different, it is determinate only relative to a quintessentialist fiction that we all tend to share. A full exploration of this question must be left for another occasion.
References Allendorf, F. W., and R. F. Leary. 1988. Conservation and distribution of genetic variation in a polytypic species: the cutthroat trout. Conservation Biology, 2, 170184. Astuti, R., Solomon, G. E. A., & Carey, S. (2004). Constraints on conceptual development. Monographs of the Society for Research in Child Development, 69. Atran, S., Medin, D.L, Lynch, E., Vapnarsky, V., Ucan Ek' & Sousa, P. (2001) Folkbiology doesn't come from Folkpsychology: Evidence from Yukatec Maya in Cross-cultural Perspective. Journal of Cognition and Culture, 1, 4-42. Barker, M. (2010). Specious Intrinsicalism. Philosophy of Science, 77, 73-91. Basch, S. H. (1973). The intrapsychic integration of a new organ: A clinical study of kidney transplantation. Psychoanalytic Quarterly, 42, 364-384. Baum, P., Schmid, R., Ittrich, C., Rust, W., Fundel-Clemens, K., et al. (2010). Phenocopy A strategy to qualify chemical compounds during hit-to-lead and/or lead optimization. PLoS ONE, 5, e14272. Belk, R. W. (1990). Me and thee versus mine and thine: How perceptions of the body influence organ donation and transplantation. In J. Shanteau & R. J. Harris (eds.), Organ donation and transplantation: Psychological and behavioral factors. American Psychological Association: Washington, DC, 139-149.
Cole, E. R., Jayaratne, T. E., Cecchi, L. A., Feldbaum, M., & Petty, E. M. (2007). Vive la difference? Genetic explanations for perceived gender differences in nurturance. Sex Roles, 57, 211-222. Coley, J.D., Medin, D. L., & Atran, S. (1997). Does rank have its privilege? Inductive inferences within folkbiological taxonomies. Cognition, 64, 73-112. Coley, J. D., Medin, D. L., Proffitt, J. B., Lynch, E., & Atran, S. (1999). Inductive reasoning in folkbiological thought. In D. L. Medin & S. Atran (eds.), Folkbiology.. Cambridge , MA : Bradford, 205-232. Devitt, M. (2008). Resurrecting Biological Essentialism, Philosophy of Science, 75, 344-382. Devitt, Michael & Kim Sterelny (1987) Language and Reality, Basil Blackwell, Oxford. Diesendruck, G. (2001). Essentialism in Brazilian children's extensions of animal names. Developmental Psychology, 37, 49-60. Diesendruck, G., & Gelman, S. A. (1999). Domain differences in absolute judgments of category membership: Evidence for an essentialist account of categories. Psychonomic Bulletin and Review, 6, 338-346. Dupr, John (1981), Natural Kinds and Biological Taxa. Philosophical Review, 90, 6690. Dupr, J. (1993). The disorder of things: Metaphysical foundations of the disunity of science. Cambridge, MA: Harvard University Press. Ereshefsky, M. (1998). Species Pluralism and Anti-realism. Philosophy of Science, 65, 103120. Ereshefsky, M. (2010). Whats wrong with the new biological essentialism. Philosophy of Science, 77, 674-685. Frazier, B. N., & Gelman, S. A. (2009). Developmental changes in judgments of authentic objects. Cognitive Development, 24, 284-292. Gelman, R. (1987). Cognitive Development: Principles guide learning and contribute to conceptual coherence. Paper presented at the Meeting of the American Psychological Association, Division I, New York. Gelman, S. A. (2003). The essential child: Origins of essentialism in everyday thought. New York: Oxford University Press. Gelman, S. A., & Coley, J. D. (1990). The importance of knowing a dodo is a bird: Categories and inferences in 2year-old children. Developmental Psychology, 26, 796-804. Gelman, S.A., Collman, P., & Maccoby, E. E. (1986). Inferring properties from categories versus inferring categories from properties: The case of gender. Child Development, 57, 396-404. Gelman, S. A., & Markman, E. M. (1986). Categories and induction in young children. Cognition, 23, 183-209. Gelman, S. A. & Markman, E. M. (1987). Young childrens inductions from natural kinds: The role of categories and appearances. Child Development, 58, 1532-1541. Gelman, S. A., & Wellman, H. M. (1991). Insides and essences: Early understandings of the nonobvious. Cognition, 38, 213-244. Gelman, S. A., & OReilly, A. W. (1988). Childrens inductive inferences within superordinate categories: The role of language and category structure. Child Development, 59, 876-887. Gelman, S. A., & Rhodes, M. (in press). Two-thousand years of stasis: How psychological essentialism impedes evolutionary understanding. In K. S. Rosengren, S. Brem, E. M. Evans, & G. Sinatra (Eds.), Evolution challenges: Integrating research and practice in teaching and learning about evolution. Cambridge: Oxford University Press. Gendler, T. S. (2007). Philosophical thought experiments, intuitions, and cognitive equilibrium. Midwest Studies in Philosophy, 31, 68-89. Gendler, T. S. (2008a). Alief and Belief. Journal of Philosophy,105, 634-663. Gendler, T. S. (2008b). Alief in action (and reaction). Mind and Language, 23, 552-585. Gendler, T. S. (In press). On the epistemic costs of implicit racism. Philosophical Studies. Ghiselin, M. (1987). Species Concepts, Individuality, and Objectivity. Biology and Philosophy, 2, 127143. Gibson, G., & Wagner, G. (2000). Canalization in evolutionary genetics: A stabilizing theory? BioEssays, 22, 372380. Gil-White, F. J. (2001) Are ethnic groups biological 'species' to the human brain?: Essentialism in our cognition of some social categories. Current Anthropology, 42, 515-554. Gottfried, G. M., & Gelman, S. A. (2005). Developing domain-specific causal-explanatory frameworks: The role of insides and immanence. Cognitive Development, 20, 137-158. Graves, J. L. (2001). The emperors new clothes: Biological theories of race at the millennium. New Brunswick, NJ: Rutgers University Press. Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 8.
Greenwald, A. G., Poehlman, T. A., Uhlmann, E., & Banaji, M. R. (2009). Understanding and using the Implicit Association Test: III. Meta-analysis of predictive validity. Journal of Personality and Social Psychology, 97, 1741. Hacking, I. (2007a). Putnams theory of natural kinds and their names is not the same as Kripkes. Principia, 11, 124. Hacking, I. (2007b). The contingencies of ambiguity. Analysis, 67, 269-277. Harris, R. N., Semtlisch, R. D., Wilbur, H. M., & Fauth, J. E. (1990). Local variation in the genetic basis of paedomorphosis in the salamander Ambystoma talpoideum. Evolution, 44, 1588-1603. Hayward, C., & Madill, A. (2003). The meanings of organ donation: Muslims of Pakistani origin and white English nationals living in North England. Social Science and Medicine, 57, 389-401. Heyman, G., & Gelman, S. A. (2000). Beliefs about the origins of human psychological traits. Developmental Psychology, 36, 663-678. Hirschfeld, L. A. (1996). Race in the Making: Cognition, culture, and the childs construction of human kinds. Cambridge, MA: MIT Press. Holter, B. D. (2009). The Ontology of Species: A Radically Pluralistic Perspective. Ph.D. thesis, Washington State University. Hood, B. M. & Bloom, P. (2008). Children prefer certain individuals over perfect duplicates. Cognition, 106, 455462. Hood, B.M., Donnelly, K. & Byers, A. (2011). Moral contagion and the negativity bias in attitudes towards potential organ transplantation. Journal of Culture and Cognition. Hull, D. L. (1965). The effect of essentialism on taxonomy: Two thousand years of stasis. British Journal for the Philosophy of Science, 15, 314-326 & 16, 1-18. Inspector, I., Kutz, Y., & David, D. (2004). Another person's heart: Magical and rational thinking in the psychological adaptation to heart transplantation. The Israeli Journal of Psychiatry and Related Sciences, 41, 161-173. Jayaratne, T. (2001). National sample of adults beliefs about genetic bases to race and gender. Unpublished raw data. Jayaratne, T. E., Gelman, S. A., Feldbaum, M., Sheldon, J. P., Petty, E. M., & Kardia, S. L. R. (2009). The perennial debate: Nature, nurture, or choice? Black and White Americans' explanations for individual differences. Review of General Psychology, 13, 24-33. Johnson, C. N. (1990). If you had my brain, where would I be? Children's understanding of the brain and identity. Child Development, 61, 962-972. Johnson, C. N., & Jacobs, M. G. (2001, April). Enchanted objects: How positive connections transform thinking about the very nature of things. Poster presented at the Society for Research in Child Development, Minneapolis, MN. Johnston, M. (1997). Manifest Kinds. Journal of Philosophy, 94, 564-583. Kalish, C. W. (1998). Natural and artificial kinds: Are children realists or relativists about categories? Developmental Psychology, 34, 376-391. Kamp, H., and Partee, B. (1995). Prototype theory and compositionality. Cognition, 57, 129-191. Keil, F. C. (1989). Concepts, kinds, and cognitive development. Cambridge, MA: MIT Press. Keller, J. (2005). In genes we trust: The biological component of psychological essentialism and its relationship to mechanisms of motivated social cognition. Journal of Personality and Social Psychology, 88, 686702. Kitcher, Philip (1984), Species, Philosophy of Science, 51, 308333. Kitcher, P. (2007). Does 'Race' Have a Future? Philosophy and Public Affairs, 35, 293317. Kolon, T. F. (2008). Disorders of sexual development. Current Urology Reports, 9, 172-177. Kravchuk, T., Reznikov, M., Tichonov, P., Avidor, N., Meir, Y., Bekkerman, A., & Alexandrovich, G. (2011). A magnetically focused beam of ortho-water. Science, 331, 319-321. Kripke, S. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press. Kushner, D. J., Baker, A., & Dunstall, T. G. (1999). Pharmacological uses and perspectives of heavy water and deuterated compounds. Canadian Journal of Physiological Pharmacology, 77, 7988. Lanie, A. D., Jayaratne, T. E., Sheldon, J. P., Kardia, S. L. R., & Petty, E. M. (2004). Exploring the public understanding of basic genetic concepts. Journal of Genetic Counseling, 13, 305-320. LaPorte, Joseph (2004), Natural Kinds and Conceptual Change. Cambridge: Cambridge University Press. LaPorte, J. (1997). Essential Membership. Philosophy of Science, 64, 96-112. Leslie, S. J. (2011). Essence, Plenitude, and Paradox. Philosophical Perspectives, 25, 277-296. Leslie, S. J. (submitted). Essence and Accident: An Empty Distinction.
Lewontin, R. C., Rose, S., & Kamin, L. (1984). Not in our genes: Biology, ideology, and human nature. New York: Pantheon Books. Lizotte, D. J., & Gelman, S. A. (1999). Essentialism in childrens categories. Poster presented at the Cognitive Development Society, Chapel Hill, NC. Luhmann, C. & Coley, J.D. (2000). Domain Specific Relations between Typicality and Absolute Category Membership. Poster presented at Northeastern University College of Arts & Sciences Experiential Education Expo, May 2000. Maglo, K. N. (2011). The case again biological realism about race: From Darwin to the post-genomic era. Perspectives on Science, 19, 361-391. Mahalingam, R. (2003). Essentialism, culture, and power: Representations of social class. Journal of Social Issues, 59, 733-749. Malt, B. (1994). Water is not H O. Cognitive Psychology, 27, 41-70. Mayr, E. (1982). The growth of biological thought. Cambridge, MA: Harvard University Press. Mayr, E. (1988). Toward a new philosophy of biology: observations of an evolutionist. Cambridge, MA: Belknap Press of Harvard University Press. Mayr, E. (1991). One long argument: Charles Darwin and the genesis of modern evolutionary thought. Cambridge, MA: Harvard University Press. Meyer, M., Leslie, S. J., Gelman, S. A, & Stilwell, S. M. (Under review). Essentialist beliefs about bodily transplants in the United States and India. Murphy, G. (2002). The Big Book of Concepts. Cambridge, MA: MIT Press. Needham, P. (2000). What is Water? Analysis, 60, 13-21. Needham, P. (2002). The discovery that water is H20. International Studies in the Philosophy of Science, 16, 205226. Needham, P. (2008). Is Water a Mixture?Bridging the Distinction Between Physical and Chemical Properties. Studies in History and Philosophy of Science, 39, 66-77. Needham, P. (2011). Microessentialism: What is the Argument?, Nous, 45, 1-21. Needham, P. (Manuscript). The Phase Rule and the Notion of Substance. Stockholm University. Nemeroff, C., & Rozin, P. (1994). The contagion concept in adult thinking in the United States: Transmission of germs and interpersonal in Ethos, 22, 158186. Newman, G., Herrmann, P., Wynn, K., & Keil, F.C. (2008). Biases towards internal features in infants' reasoning about objects. Cognition, 107, 420432. Newman, G. & Keil, F.C. (2008). Wheres the Essence?: Developmental Shifts in Childrens Beliefs About the Nature of Essential Features. Child Development, 79, 1344-1356. Okasha, Samir (2002), Darwinian Metaphysics: Species and the Question of Essentialism. Synthese 131, 191213. Pennell, W., & Barton, B. A. (1996). Principles of Salmonid Culture. Netherlands: Elsevier. Phillips, W., Shankar, M., & Santos, L. R. (2010). Essentialism in the absence of language? Evidence from rhesus monkeys (Macaca mulatta). Developmental Science, 13, F1-F7. Phillips, W. & Santos, L. R. (2007). Evidence for kind representations in the absence of language: Experiments with rhesus monkeys (Macaca mulatta). Cognition, 102, 455-463. Putnam, H. (1975). The Meaning of Meaning In H. Putnam, Mind, Language, and Reality. Cambridge: Cambridge University Press, 215-271. Rhodes, M., & Gelman, S. A. (2009a). A developmental examination of the conceptual structure of animal, artifact, and human social categories across two cultural context. Cognitive Psychology, 59, 244-274. Rhodes, M., & Gelman, S. A. (2009). Five-year-olds' beliefs about the discreetness of category boundaries for animals and artifacts. Psychonomic Bulletin and Review, 16, 920-924. Rosch, E. H. (1973). Natural categories. Cognitive Psychology, 4, 32850. Rosch, E. H. (1975). Cognitive representation of semantic categories. Journal of Experimental Psychology, 104, 192-233. Rosch, E. H. (1978). Principles of Categorization. In E. Rosch and B.B. Lloyd (eds.), Cognition and Categorization. Hillsdale, NJ: Erlbaum, 27-48. Rosch, E. H., Mervis, C. B., Gray, W. D., Johnson, D. M., Boyes-Braem, P. (1976). Basic objects in natural categories. Cognitive Psychology, 8, 382439. Rosenberg, A. (1994). Instrumental Biology or the Disunity of Science. Chicago: University of Chicago Press. Rozin, P., & Nemeroff, C. (2002). Sympathetic magical thinking: the contagion and similarity "heuristics". In Gilovich, T., Griffin, D., & Kahneman, D. (eds.) Heuristics and biases: The psychology of intuitive judgment. Cambridge: Cambridge University Press, 201-216.
Salmon, N. U. (1979). How Not to Derive Essentialism From the Theory of Reference. Journal of Philosophy, 76, 703-725. Samarapungavan, A., & Wiers, R. (1997). Children's thoughts on the origin of species: A study of explanatory coherence. Cognitive Science, 21, 147-177. Sanner, M. A. (2001a). Exchanging spare parts or becoming a new person? People's attitudes toward receiving and donating organs. Social Science and Medicine, 52, 1491-1499. Sanner, M. A. (2001b). People's feelings and beliefs about receiving transplants of different origins--questions of life and death, identity, and nature's border. Clinical Transplantation, 15, 19-27. Sawin, P. B. (1932). Hereditary variation of the chin-chilla rabbit: In coat and eye color. Journal of Heredity, 23, 3946. Shtulman, A. (2006). Qualitative differences between nave and scientific theories of evolution. Cognitive Psychology, 52, 170-194. Shtulman, A., & Schulz, L. (2008). The relation between essentialist beliefs and evolutionary reasoning. Cognitive Science, 32, 1049-1062. Simons, D. J., & Keil, F. C. (1995). An abstract to concrete shift in the development of biological thought: The insides story. Cognition, 56, 129-163. Smith, M. B. (2011). Organic chemistry: An acid-based approach. Boca Raton: Taylor & Francis. Sober, Elliott (1980), Evolution, Population Thinking and Essentialism, Philosophy of Science, 47, 350383. Solomon, G. E. A., & Johnson, S. C. (2000). Conceptual change in the classroom: Teaching young children to understand biological inheritance. British Journal of Developmental Psychology, 18, 81-96. Sousa, P., Atran, S., & Medin, D.L. (2002). Essentialism and Folkbiology: Evidence from Brazil. Journal of Cognition and Culture, 2, 195-223. Springer, K., & Keil, F. C. (1989). On the development of biologically specific beliefs: The case of inheritance. Child Development, 60, 687-648. Stanford, P. Kyle (1995), For Pluralism and Against Realism about Species, Philosophy of Science, 62, 7091. Sterelny, Kim, and Paul Griffiths (1999), Sex and Death. Chicago: University of Chicago Press. Stich, S. (2009). Reply to Sosa. In M. Bishop and D. Murphy (eds.), Stich and His Critics. Oxford: Blackwell, 228236. Strevens, M. (2000). The essentialist aspect of naive theories. Cognition, 74, 149-175. Taylor, M. G. (1996). The development of childrens beliefs about social and biological aspects of gender differences. Child Development, 67, 1555-1571. Taylor, M. G., Rhodes, M., & Gelman, S. A. (2009). Boys will be boys; cows will be cows: Children's essentialist reasoning about gender categories and animal species. Child Development, 79, 1270-1287. Templeton, A. (1999). Human Races: A Genetic and Evolutionary Perspective. American Anthropologist, 100, 632 50. Tikhonov, V. I., & Volkov, A. A. (2002). Separation of water into its ortho and para isomers. Science, 296, 2363. van Brakel, Jaak (1986), The Chemistry of Substances and the Philosophy of Mass Terms. Synthese, 69, 291324. Waddington, C. H. (1942). Canalization of development and the inheritance of acquired characteristics. Nature, 3811, 563-565. Walker, S. (1999). Culture, domain-specificity, and conceptual change: Natural kind and artifact concepts. British Journal of Developmental Psychology, 17, 203-219. Waxman, S. R., Lynch, E. B., Casey, K. L., & Baer, L. (1997). Setters and samoyeds: The emergence of subordinate level categories as a basis for inductive inference. Developmental Psychology, 33(6), 1074-1090. Waxman, S. R., Medin, D. L., & Ross, N. (2007). Folkbiological reasoning from a cross-cultural developmental perspective: Early essentialist notions are shaped by cultural beliefs. Developmental Psychology. 43(2), 294-308. Weinberg, J., Nichols, S. & Stich, S. (2001). Normativity and epistemic intuitions. Philosophical Topics, 29, 42960. Weisberg, M (2005). Water is Not H2O. In D. Baird et al. (eds.) Philosophy of Chemistry: Synthesis of a New Discipline. New York: Springer, 337-345. West-Eberhard, M. J. (2005). Developmental Plasticity and Evolution. New York: Oxford University Press. Wiggins, David (1980), Sameness and Substance. Oxford: Blackwell. Wright, S. (1934). The results of crosses between inbred strains of guinea pigs, differing in their number of digits. Genetics, 19, 537-551. Zuckerkandl, E., & Villet, R. (1988). Concentration-affinity equivalence in gene regulation: Convergence of genetic and environmental effects. Proceedings of the National Academy of Science, 85, 4784-4788.
PHILOSOPHICAL PERSPECTIVES
Philosophical Perspectives, 25, Metaphysics, 2011
ESSENCE, PLENITUDE, AND PARADOX
Sarah-Jane Leslie Princeton University

Essentialism in philosophy is the view that there are in principle two ways of having properties and including parts: the essential way and the accidental way (e.g, Fine, 1994, 1995; Johnston 1984, 2006; Kripke, 1980; Salmon 1986, 2005; Yablo 1987). As I understand the view, even an extreme essentialist who takes every property and every part to be had essentially should agree with the in-principle distinction while arguing that certain metaphysical considerations mean that the second way of having properties is not, and cannot be, exemplified. Such a view would contrast with the view of an anti-essentialist like Willard van Orman Quine who repudiates the very distinction between the two ways of having properties. Quines view is not that all properties are accidental, for that would be to use what he calls the invidious distinction (1953), rather than repudiate it. His view is that there is not even in principle these two ways of having properties (or satisfying predicates, as he would put it), about which we can ask Which of these two ways do things have properties? What then ostensibly distinguishes the essential way of having properties from the accidental way? An objects essential properties are conditions on what it is to be that object, and this set of conditions fixes just which possibilities or possible worlds the object exists in, namely just those in which it satisfies those conditions. An objects essential properties also fix how long it exists, namely only at those times at which it satisfies those conditions. An objects accidental properties are those of its properties that it can be found without at some times or at some worlds. Likewise, an objects accidental parts are those it can be found without at some times or worlds. It is difficult, and perhaps impossible, to be an essentialist without embracing a plenitude of entities far more entities than common sense explicitly recognizes. This is sometimes noted in the literature, and usually in the context of entities in different categories or in different manifest kinds, such as a statue and the lump of clay which constitutes it (e.g. Fine, 2003, 2008; Johnston, 1992). However the scope of the resulting plenitude of entities and its consequences for metaphysics has not been fully understood. In this paper I argue that a
278 / Sarah-Jane Leslie
long-standing paradox of essentialism is easily resolved once the connection between essentialism and plenitude is appreciated. This paradox has several formulations, versions of which are known as Chisholms Paradox Chandlers Paradox and The Four Worlds Paradox. The paradoxical element, I argue, is illusory, and arises only because two or more distinct entities are mistakenly thought to be identical. Because of the connection between essentialism and plenitude, the essentialist assumptions of the paradox guarantee that there multiple entities in play, even if common sense is taken to suggest otherwise. The apparent paradox is generated by holding to essentialism while taking common senses suggestion too seriously, and insisting there must be a single entity where there are, in fact, many. (This insistence may rest on a misinterpretation of commonsense counting arguably to count the number of Ks is in effect to count the number of discrete, i.e. not-entirely-overlapping, Ks so that what the essentialist is committed to is not really at odds with commonsense, but only with a na ve philosophical construal of it.)
Essentialism and Plenitude

Since essentialism holds that there is not one but rather two ways that an item can have a property or include a part, it opens the door to a proliferation of entities. This has been observed since the early history of essentialism, as is illustrated by the view often attributed to Aristotle as his theory of kooky objects (e.g., F. Lewis, 1982; Matthews, 1982; White, 1971). These kooky objects are items such as sitting-Socrates and musical-Corsicus items that share the essential properties of Socrates and Corsicus, except they are also respectively essentially sitting and essentially musical. When Socrates is seated, why does this further entity sitting-Socrates not come into existence only to be destroyed when Socrates stands? Of course, common sense does not recognize such entities, but common sense need not be a good guide to the whole extent of ontology. If the essentialist does not wish to countenance such items, she owes us an explanation of why the combination of Socrates essential properties plus being seated is somehow unacceptable or inadmissible as an essence: why is this not an essence an entity can have? The scope of this challenge can be illustrated by considering an abstract (and simplified) general model. Suppose simplifying massively for the purposes of illustration; see Yablo (1987) that we have an item with five properties. Suppose further that those properties are strongly modally independent, so that each of the five can be possessed either essentially or accidentally without requiring that the other four be possessed essentially, accidentally or even at all, and likewise for any combination of the properties.1 Finally, suppose the essentialist claims to have discovered that the item in question has two of the five properties essentially, and the other three merely accidentally. Now we can ask: why is there not also another entity present that has only one of the properties essentially?
Essence, Plenitude, and Paradox / 279
Why is there not one that has four of the properties essentially? Why not items that have other combinations of the properties essentially? Since we stipulated that these properties are modally independent from each other no independent metaphysical principles will rule out these possibilities. If we wish to rule out this plenitude of entities, we will need a new metaphysical principle that will rule that all but one of these combinations specifies an allowable essence an essence that an entity can have. Let us call such a principle a principle of limited variety of essences. If we do not insist that the properties in the model be strongly modally independent, then some of the combinations of properties will fail to specify essences. For example, consider a blue square, and suppose we decide that being spatially extended is an essential property of the square, but being blue is accidental. If we attempt to specify an essence by swapping these properties we will fail: since being blue necessitates being spatially extended , nothing is essentially blue but accidentally spatially extended. No principle of limited variety is needed to eliminate this alleged essence general metaphysical principles suffice to rule it out. Interestingly, the example of sitting-Socrates may also be ruled out on general grounds, unless some adjustments are made. Suppose, as is reasonable, that Socrates is essentially a self-maintaining living thing that is, Socrates performs certain life functions, as a usually reliable means of preserving his continued existence. It would seem that sitting-Socrates cannot be essentially self-maintaining, since sitting-Socrates does not undertake any particular actions so as to preserve its continued sedentary condition. Thus, the alleged essence specified by simply adding being seated to the list of Socrates essential properties does not in fact correspond to a possible essence; it is internally problematic, and so no independent principle of limited variety is needed to rule it out. This is not, of course, to say that a similar entity cannot be specified, however one that shares many of Socrates essential properties (though not being self-maintaining), and is essentially seated. General metaphysical principles may narrow the list, but it would be wildly optimistic to suppose that they will narrow the list to just the one item recognized by common sense namely, Socrates. Let us call the combinations of essential properties that are not ruled out on general grounds candidate essences, and let us call essences that entities can have admissible essences. (For extended discussion of how to arrive at candidate essences, see Yablo (1987) and Leslie (submitted).) It should be clear upon reflection that there will be far more candidate essences than there are entities recognized by common sense. Thus, if essentialism is to avoid positing a plenitude of entities, it needs to present a principle of limited variety to deliver the result that only a few candidate essences are in fact admissible essences. Perhaps such a principle can be formulated so as to deliver the result that Socratess essence is the only admissible essence to be found among the relevant candidate essences; for example, perhaps one could argue that to be an admissible
essence, begin a self-maintaining substance must be among the essential properties (and then also argue that no variant on Socrates essence meets this criteria). I am skeptical that any such principle can be defended even in the case of living beings (Leslie, submitted), but I will not argue for this here. For the purposes of this paper we need only consider whether such a principle could be remotely plausible the case of artifacts. As an illustration, imagine we take a ship made of one hundred planks and remove each plank one by one. (For simplicity, let us suppose we burn each plank immediately after removing it from the ship.) Intuitively, there will come a point at which our ship no longer exists certainly this will be the case by the time the last plank is burned, and is naturally thought to come a good bit before that. On the essentialist view, this point is determined by the ships essence being constituted by some minimum number of planks will be an essential property of the ship, and so once we pass this point, the ship ceases to exist. If the essentialist believes there is a unique ship here, she faces the question: what is the number n such that the ship exists if n+1 of its planks still remain, but ceases to exist when only n of its planks still remain? It might seem that this question is an unfair one because of vagueness, but one needs to be cautious on this point. It is one thing to think that there is linguistic vagueness, another to think that there is metaphysical vagueness, especially when it comes to questions about existence. Recall that, for the essentialist, an items essence determines its conditions for existence; if its essence is to be vague, then there will be instances in which the item hovers indeterminately between existence and non-existence. Some philosophers have defended such views, but most reject this notion. So while the essentialist might agree the word/concept ship does not have a definite n associated with it, she still faces the question of what this n might be for the (allegedly) unique entity in question, lest she find herself committed to the idea that there is some range of numbers for which the entity hovers strangely between being and not-being. Suppose, then, that the essentialist determines that n is 20 planks; that is, the essentialist truth about the ship is that when all that remains of it are 20 planks arranged however they might be, the original ship has ceased to be, and cannot be reconstructed by augmenting the 20 remaining planks. This property being such that it has ceased to be and cannot be brought back after its constitution has been reduced in this way to 20 planks is an essential property of the ship. This property thus figures in an admissible essence an essence that is countenanced as one than an entity has. However, consider the following candidate essence: the essence that is just like that of the ship, but which includes instead the property of being such that anything that has it will forever cease to be after its constitution has been reduced in this way to 25 planks. This is quite clearly a candidate essence no general metaphysical principles will rule this out as an impossible essence. But why is this not also an admissible essence? What plausible principle of limited variety could be invoked to allow the n = 20 essence, but not the n = 25 essence? Once
we allow that the n = 20 condition figures in an admissible essence, it will be utterly arbitrary not to allow there are many such essences in the offing: as well as the n = 25 essence there is the n = 28 essence, the n = 17 essence, and so on. And if there are such essences then it follows that many ships (or ship-like entities) were entirely coincident with our ship during the period before it began to be destroyed by disassembly. They are distinct entities because they cease to be at distinct points in the process of disassembly. Once we see how indefensible such a principle of limited variety would be, it is easy to model the vagueness of the term ship without recourse to metaphysical vagueness. The vagueness here is wholly linguistic the phrase this ship simply does not pick out a unique entity in the plenitude. Each entity has fully determinate conditions of existence, but our language and concepts do not single out a particular entity. Could one avoid plenitude by embracing metaphysical vagueness? Suppose we pick out a range of n such that the ship neither determinately exists nor determinately doesnt exist in that range. Then we have deemed that the range figures in an admissible essence (we can even allow that the range have fuzzy boundaries, so as not to rule out higher-order vagueness). Then consider another clearly distinct again so as to be clear that we are not trading on higherorder vagueness range of values of n. If the former figures in a candidate essence, so does the latter. Now we ask again: what plausible principle of limited variety will rule out the latter while allowing the former? Thus, even if we accept metaphysical vagueness, with all its concomitant difficulties, plenitude would still only be avoided by a highly implausible principle of limited variety. In the case of artifacts (and I think even more generally) it is very hard to see how a sufficiently strong principle of limited variety could be defended so as to vindicate the prejudice against plenitude while holding on to essentialism. Essentialism, it might be said, invites plenitude simply by holding that there are not one but two ways to have properties and parts. While this is sometimes noted in the literature (e.g. Fine, 1999; Hawthorne, 2006; Johnston, 2006; Yablo, 1987), its unavoidability and more importantly its implications have not been fully appreciated. In what follows, I discuss a long-standing cluster of apparent paradoxes, which have been taken seriously by Roderick Chisholm (1967, 1973), Willard van Orman Quine (1976), Hugh Chandler (1976), Saul Kripke (1980, fn. 18), David Lewis (1986), Nathan Salmon (1986, 2005), Graeme Forbes (1984), Penelope Mackie (2006) and others, and argue they are readily resolved once we see the connection between essentialism and plenitude.
The Paradox of Tolerant Essence: Chisholm and Chandler

On one formulation, Chisholms (1967, 1973) paradox asks us to consider a ship (call it Ship1) made from 100 planks, and supposes that while the ships (original) constitution is essential to the ship, the ships essence is nonetheless
ever so slightly tolerant along this dimension: we could have made that very ship if we had substituted one different plank. This assumption has intuitive plausibility it seems unnecessarily strict to insist that even one plank out of a hundred could not have been different, else that very ship would not have existed. The apparent paradox then unfolds by asking us to consider a chain of possible worlds w1, . . . , w101, each containing a ship made from 100 planks, such that in w1 there is a ship made of the original 100 planks, in w2 there is a ship that is made of 99 of those planks plus one new one, in w3 there is a ship made of 99 of the same planks as in w2, plus one more, and so on. That is, in each wn+1, there is a ship made of 99 of the same planks as the ship in wn, plus one new plank. Thus, in the first world w1, made of the original 100 planks, while in the 101st possible world, w101, made of none of the original planks.2 The paradoxical reasoning is then as follows: by hypothesis, Ship1s original constitution is essential to it, yet its essence is tolerant, so that it could be made with one different plank. W1 uncontroversially contains Ship1, and so does w2, since w2 represents an alternative original constitution of Ship1. But if w2 contains Ship1, so does w3, since the constitution of the ship in w3 differs from that of the ship in w2 only by one plank. Ship1s essence is tolerant, and Ship1 exists at w2, and the ship at w3 shares the same original constitution as the ship at w2, save only for a small variation that is allowed for by the tolerance in Ship1s essence. Generalizing, for each pair of successive worlds wn and wn+1, it would seem that we have the same ship in each since by hypothesis the ships essence tolerates one-plank changes and so by transitivity of identity, the ship in w1 is identical to the ship in w101, despite their not sharing any of the same planks when they are first constructed. Ship1 exists in w101, but is made from entirely different planks than in w1 despite the fact that we initially stipulated that its original constitution was essential to it, albeit with a small amount of leeway (i.e. one plank). Herein lies the paradox: it would seem that essences cannot be at all tolerant, and yet we surely have the intuition that we could have made this very ship had we changed out just one of the planks.3 Perhaps it might not seem too extreme to avoid all this by supposing that a ships original planks are strictly essential to it. But what, though, of the planks themselves? We might naturally suppose that their original constituting matter is essential to them, but now we can ask: could it have been the same plank had one single proton been replaced by another? If we answer yes, then the paradox reemerges for the planks themselves, and so by extension for the ship. One might suspect that this paradox is driven by vagueness that it is somehow no more than a sorites paradox in disguise, rather than being a sui generis paradox about essence. However, this is not so, as is made clear by another version of the paradox, which focuses on a three-part artifact, whose essence is stipulated in a precise way (compare Chandler (1976)). Suppose in w1 I build an axe Axe1 out of blade Blade1, shaft Shaft1, and handle Handle1. This axe, let us stipulate, is essentially (originally) composed of
Figure 1. The Paradox.
those parts, but its essence is tolerant to this extent: as long as two out of the three parts are involved, Axe1 could have been made with one part different. So we have w2 in which Axe1 is made with a different blade but the same shaft and handle: Blade2, Shaft1, and Handle1. Now surely Axe1s essence is still tolerant in w2. But then there will be w3 in which we build Axe1 with a different shaft than we used in w2: Blade2, Shaft2, and Handle1. Now, however, we have a possible world w3 in which Axe1 is built with only one of the three parts with which it was built in w1 but as we originally stipulated, Axe1s essence allows only one of its parts to be different. As we originally described it, no axe made from Blade2, Shaft2, and Handle1 could be identical to Axe1, since Axe1 must essentially be built from at least two of the following parts: Blade1, Shaft1, and Handle1. Thus the axe in w3 cannot be Axe1; yet by transitivity of identity, it must be Axe1. While these three worlds suffice to illustrate the paradox, it can be made more vivid by considering a fourth world w4, in which the axe in w3 is built with one part different from the axe in w3: Blade2, Shaft2, and Handle2. Since the axe in w3 has a tolerant essence, this is a permissible way for it to be built. But now by transitivity of identity we have the result that the axe in w4 is identical to Axe1, even though it has no parts in common with Axe1. Thus, even though we stipulated at the outset that its original constitution was essential to Axe1, albeit with some degree of tolerance, we obtain the result that Axe1 could in fact have been built with an entirely different original constitution. (See figure 1.) The invocation of vagueness is obviously not to the point here, for we have stipulated it away; instead the problem arises on a certain conception of tolerance in the original constitution of the axe.
Salmons Solution to the Paradoxes

Nathan Salmon (1986; 2005 (originally 1981)), following Hugh Chandler (1976), has argued for an influential solution to the paradoxes. His solution hinges upon rejecting the idea that what is possibly possible is possible simpliciter. That is, Salmon argues only some possible worlds are accessible from the actual world, and that different worlds may be accessible from those immediately accessible worlds; this is coherent, Salmon argues, because the accessibility
relation is not transitive, so worlds that are accessible from accessible possible worlds need not be themselves accessible from the world under consideration. Salmons solution to the paradoxes thus runs as follows: If we make Axe1 out of Blade1, Shaft1, and Handle1 in w1, then since Axe1s essence is tolerant, w2 in which we make Axe1 from Blade2, Shaft1, and Handle1 is accessible from w1. But again, since Axe1s essence is tolerant, there is a world w3, which is accessible from w2, in which Axe1 is made from Blade2, Shaft2, and Handle1, and similarly, mutatis mutandis, for w4. However, w3 and w4 are not accessible from w1; the possibilities that w3 and w4 represent are merely possibly possible, and possibly, possibly possible respectively, relative to w1 not straightforwardly possible. If we suppose that w1 is the actual world, then we might say that had things been different i.e. had we made Axe1 from Blade2, Shaft1 and Handle1 then it would have been possible to make Axe1 from Blade2, Shaft2, and Handle1. However, since in fact we made Axe1 from Blade1, Shaft1, and Handle1, it is not possible Axe1 have been made from Blade2, Shaft2, and Handle1. As Salmon sees it, the paradox is driven by the false assumption that what is possible relative to a world that is possible simpliciter, is itself possible simpliciter. Thus he claims the paradox turns on what he calls the fallacy of possibility deletion. Salmons solution is ingenious, and has been influential.4 I do not think that it is ultimately satisfactory, for a reason that has not been noted. Salmons treatment of the paradox faces a destructive dilemma: either the paradoxical argument stops at the second world, in which case there is no paradox to be explained away by Salmons appeal to the deletion fallacy or he is committed to the view that an items essence could have been different than it is, even if we restrict our interpretation of the relevant could to the accessible worlds i.e. the worlds that are possible simpliciter. The world w2 is accessible from w1 and vice versa; each represents straightforward possibilities for the items that exist in the other. But on Salmons description of the case Axe1 in w1 has a different essence from Axe1 in w2. To see this, we need to be very clear about the notion of a tolerant essence.
What is a Tolerant Essence?

Let us begin by noting that there may be something a little suspicious about the description of Axe1s essence: we began by supposing that its essence was tolerant in that the axe could have been made with one part different relative to its original composition, and also that its original composition consisted of Blade1, Shaft1, and Handle1. Now, it would seem that we might have described this same essence differently. Isnt the tolerant essence just specified equivalent to the following at-one-level intolerant essence? Axe1 is essentially constituted by at least two out of the following three parts: Blade1, Shaft1, and Handle1, plus the appropriate kind of third part if needed. This would seem to pick out the
same essence, only with greater precision, since now the parts are specified in the description of the essence. It is easy to see that any putative tolerant essence will be describable as an essence that is in a certain way intolerant, and surely we should prefer to speak in terms of these more precisely specified intolerant essences. That is, it would seem that once we are more precise about what a tolerant essence could actually come to, it will in fact be equivalent to a certain kind of intolerant essence. One way to emphasize what is going on here is to make a distinction between variably realized essences, i.e. essences whose fixed fulfillment conditions can admit of varied realization from world to world, and invariant essences, i.e. essences whose fixed fulfillment conditions admit of no such variation across worlds. So it may be part of my essence to be conceived somewhere in the universe, but not in any given specific place. (The earth was spinning through space when it happened, and it could have happened slightly earlier or later I assume.) This part of my essence is fixed but subject to various different realizations. It is intolerant at the level of being conceived somewhere, but flexible at the level of exactly where I was conceived. This seems a possible essence for something to have, at least if there are essences. What is not possible not possible simpliciter, since it conflicts with the very notion of essence is an object having an unfixed or thoroughly variable essence, an essence that varies from possible world to possible world, not just in its realization but in its higher-order conditions of fulfillment. An objects essence is its essence in every possible world; any item with a different essence simply cannot be identical to the original object. Unfortunately, the very idea of a tolerant essence, with which the Chisholm/Chandler paradoxes begin, is first introduced plausibly by way of claims about variably realized essences and then gets parlayed into necessarily false claims about unfixed or variable essences, essences whose higher-order fulfillment conditions themselves vary from world to world. Notice for example that Salmons treatment of the paradoxes implies that Axe1s essence could have been different than it is. We in w1 build Axe1 with Blade1, Shaft1, and Handle1, and agree that Axe1s essence is tolerant in that it could have been made with one part different. What this means is that Axe1 could have been made with two out of those three parts, plus a new part of the relevant sort as needed; it does not mean anything more obscure or elusive that this (even if the tolerant essence locution invites us to think otherwise). If we accept Salmons description of the case, then at w2 where Axe1 is made from Blade2, Shaft1, and Handle1 Axe1s essence is there such that it could have been made from two out of those three parts (plus a new part of the relevant sort as needed). But then Axe1 has an essence at w2 which is different from its essence at w1. Since w2 is accessible from w1, we have it that Axe1s essence could have been different than it is. This is just not consistent with the notion of essence. A things essence could not have been different than it is. All that is possible, all that could be
sensibly meant by a tolerant essence, is that a things fixed essential conditions of existence may be open to variable realization. But in accepting the terms of the so-called paradox, this is parlayed into what is in effect the notion of a variable essence. In his reply to Forbes, Salmon writes:
The key feature of Chisholms paradox the feature of it that makes it a peculiarly modal paradox is its essential use of nested modalities. It proceeds from the observation that the truth of the modal principle (II) is no accident but is a necessary truth. (Salmon, 1986, p. 88)
and here is the modal principle:

(II) If a wooden table x is the only table originally formed from a hunk of matter y according to a certain plan P, y is any (possibly scattered) hunk of matter that sufficiently substantially overlaps y and has exactly same mass, volume, and chemical composition as y, then x is such that it might have been the only table originally formed according to the same plan P from y, instead of from y. (Salmon, 1986, p. 77)
Clearly however, this principle cannot be necessary and true, for consider a pair of mutually accessible worlds in each of which x exists, but which are such that the difference between xs constitution in the two worlds approaches but does not quite meet the allowable limits imposed by the requirement of sufficient substantial overlap. If (II) is necessary it follows that xs essential origins are tolerant in the second world in a way that they are not in the first world. That is, it follows that there are possibilities of variable realization of xs essence in the second world that are not found in the first world. This is just what cannot happen, for this implies that x has a variable essence an essence that changes from world to world not just a variably realizable essence. More generally, none of these paradoxes arise if we distinguish variably realizable intolerant essences and variable essences. If Axe1 has the relevant intolerant but variably realizable essence, it can still be made with one part different, as is the case at w2. However, it cannot be made with two parts different, as is supposed to be the case at w3, nor can it be made with all three parts different, as is supposed to be the case at w4. Thus once we are precise in this way once we clarify the shifty notion of a tolerant essence there is no paradox: Axe1 exists in w1 and w2, but not w3 or w4. Problem solved? Not so fast: what could have led Chisholm, Chandler, Kripke, D. Lewis, Salmon, Forbes, Mackie, and others who have taken these paradoxes seriously to have slid from variably realizable essences to variable essences? Indeed there is a natural response they could give at this stage, one which is invariably part of the informal set up of the paradox. After all, so the response goes, there is nothing special about w1 in our construction; we could have started with
any other world in which Axe1 uncontroversially exists. So consider Axe1 as it is made in w2 out of Blade2, Shaft1, and Handle1. We might just as well have begun our discussion at w2 instead of w1, and agreed that Axe1s allegedly tolerant essence really should clarified as a certain kind of variable realized intolerant essence as follows: Axe1 is essentially made from at least two out of the following three parts: Blade2, Shaft1, and Handle1, plus the appropriate kind of third part if needed. There is no magical halo surrounding w1 that compelled us to begin our description of the case with that world w2 might have done just as well. This response helps explain why all these philosophers are so readily led to accept the initial description of w1 and w2; we accepted or stipulated that Axe1 need only have two out of three of its original parts, and there is nothing special about starting with w1 or w2, so Axe1 need only have two out of three of Blade1, Shaft1, and Handle1 as its original parts (as in w1), and it need only have two out of three of Blade2, Shaft1, and Handle1 as its original parts (as in w2). True, once we indulge in this kind of thinking the familiar paradoxes get underway. But the thinking is itself already paradoxical, indeed it is genuinely inconsistent, for it entails that Axe1 has a different variably realizable essence depending on whether we start with w1 or w2. This, once again, is the incoherent idea of a variable essence.
Plenitude and the Paradoxes

What has gone wrong here? I believe that these paradoxes take us in because of an illusion of singularity: Surely, we say, there is only one axe in w2 and it must be taken to be our original Axe1, since w2 was set up precisely to represent the very possibility that lies in the essence of Axe1, namely this: although in w1 it was made of Blade1, Shaft1, and Handle1, it could have been made from just two of these three with one other part, and so in particular it could have been made from Blade2, Shaft1, and Handle1. But since there is nothing special about starting with w1 rather than w2, we can now see that Axe1 could also be made from Blade2, Shaft2, and Handle1. But now we have the paradox. For we stipulated that this was beyond the limit of sufficiently substantial overlap in parts, namely overlap of two out of three parts. We actually dont have a paradox, however. What needs to be shown to begin the paradox is that there is only one axe in w2 and this is far from obvious once we set out the details of the case. Here are two distinct essences; nothing could have both, and if x and y have the first and the second respectively, then x and y are necessarily distinct: Essence1: properties P, Q, R. . .and the property of being originally made of 2 out of 3 of Blade1, Shaft1, Handle1 plus an appropriate third part as needed.
Essence2: properties P,Q,R. . .and the property of being originally made out of 2 out of 3 of Blade2, Shaft1, Handle1 plus an appropriate third part as needed. Both essences are certainly on a par with one another, in the sense that both seem to be equally admissible essences: they differ only in whether it is Blade1 or Blade2 that is specified. What could make it the case that the axe in question has the one essence rather than the other? Speaking loosely of tolerant essences let us avoid these questions, but only at the cost of living with paradox including a two worlds paradox (which has been, to my knowledge, entirely unnoticed in the literature): a paradox consisting of Axe1 having Essence1 in w1 and Essence2 in w2. That is simply impossible, but the more familiar paradoxes depend on this being in place. The underlying issue here is that there is more than one axe in each of the worlds under consideration. One such entity has the essence that we specified for Axe1 when we began our story in w1: an entity that is essentially constituted by at least two out of the following three parts: Blade1, Shaft1, and Handle1, plus the appropriate kind of third part if needed. Another such entity has the essence we (thought we) specified for Axe1 when we began at w2: an entity that is essentially constituted by at least two out of the following three parts: Blade2, Shaft1, and Handle1, plus the appropriate kind of third part if needed; let us call the entity with this essence Axe2. These are distinct entities, with distinct essences. Both entities would seem to be axes, and are composed of the same parts in w2, and are in all categorical respects alike. Both of those axes also exist at w1, however only one of them exists at w3, and neither exists at w4. The transitivity of identity does not lead us into paradox here; it merely appears to because we failed to recognize that with essentialism comes plenitude. There is something at w2 that is identical to w1s Axe1, and there is something at w2 that is identical to something in w3, namely Axe2. This does not mean that Axe1 is identical to anything in w3 this thought is only tempting if we think there is only one axe at w2. (See figure 2 for an illustration.) W4 contains neither Axe1 nor Axe2, however it nonetheless contains an entity that also exists at w2 and w3, namely Axe3. Axe3 is an entity that is essentially constituted by at least two out of the following three parts: Blade2, Shaft2, and Handle1, plus the appropriate kind of third part as needed. As figure 2 illustrates, our intuitions of local identity (i.e. that each world in the sequence contains an axe that is identical to an axe in the previous world) are vindicated. However, the identities are only local, and do not carry across all the worlds, so no paradox arises. Once we recognize that essentialism invites plenitude, the paradox is resolved. At this juncture, it is important to note that plenitude is not simply something that is being wheeled out to resolve the paradox. If that were so, one might reasonably respond to the foregoing with a shrug: so we resolve the paradox, but only at the cost of populating the worlds with multiple co-located
Note: Black lines indicate that an entity exists at a given world. Only the entities relevant to resolving the paradox are shown. Plenitude, however, means there are far more entities present in the worlds. Figure 2. Plenitude as Solution to the Paradox.
axes. This would be, I think, a misunderstanding of the real dialectic, namely that the apparent paradox depends on essentialism, but essentialism itself gives rise to plenitude. Plenitude is not introduced in response to the seeming paradox; rather both have a common ground in the core notion of essentialism, the idea that there are two ways of having properties and two ways of including parts, namely the essential way and the accidental way. To see this, let us return to the set-up of the paradox. We begin by considering an axe made from Blade1, Shaft1, and Handle1. Further, we suppose this original constitution to be essential to the axe, though we allow that it could have been made with one part different. Once we are precise about what we mean and dispense with the incoherent notion of a variable essence, this is means that we are considering an axe, Axe1, that is essentially made from two of the following three parts (plus one more as needed): Blade1, Shaft1, and Handle1. In order for the paradox to be formulated, we must allow that this is sort of essence that an entity can have. But now let us ask the question: what conceivable metaphysical principle could rule that this is an admissible essence while disallowing the following candidate essence: being essentially originally constituted by two of the following three parts (plus one more as needed): Blade2, Shaft1, and Handle1. What metaphysical good fortune could attach to Blade1 but not to Blade2 so that the former but not the latter enters into an admissible essence? What defensible principle of limited variety of essence could deliver this result? The obvious answer is surely that if the essence we specified for Axe1 is an admissible essence, then so must be the essence we specified for Axe2, and similarly for Axe3 (two out of three of Blade2, Shaft2, and Handle1). But now consider a world (w2) in which Blade2, Shaft1, and Handle1 are appropriately assembled: this is a world in which the essences of Axe1, Axe2, and Axe3 are all
instantiated. These three axes are clearly distinct entities, as they have different essences. They can, however, be constituted by the same appropriately arranged parts, and this is what happens at w2. If we believe that entities have essences and so can be distinguished by their essences (as an essentialist must), then it is hard to see how this conclusion is to be avoided: for a given piece of matter with a set of properties at a world, there will be many distinct essences that are satisfied, and so many different entities constituted by this matter, all sharing the same properties at that world. To avoid this conclusion, we need a principle that allows only one of those candidate essences to be admissible to be the sort of essence that an entity can have. And what sort of metaphysical principle would yield the result that Blade1 and Shaft1 but not Blade2 and Shaft2 can figure in the essences of things? Such a principle would, to say the least, be most implausible. (Note further that Blade1 and Blade2, and Shaft1 and Shaft2, could be exactly alike: how could it then be that only one of the candidate essences is admissible?) Once we see (i) that an objects essence does not vary from world to world, even if the realization of that essence varies from world to world, and also recognize (ii) that there is multiple co-location of axes (or axe-like entities) in the worlds under discussion, the original paradox disappears. We are left with a hard-to-avoid consequence of essentialisms central claim that there are two ways of having properties and original parts, namely accidentally and essentially. If one doesnt like co-location as between Axe1 and Axe2, one should either stop being an essentialist or else set about making a case for an invidious distinction between these two candidate essences, a distinction which explains why one but not the other is admissible: Essence1: properties P, Q, R. . .and the property of being originally made of 2 out of 3 of Blade1, Shaft1, Handle1 plus an appropriate third part as needed. Essence2: properties P,Q,R. . .and the property of being originally made out of 2 out of 3 of Blade2, Shaft1, Handle 1 plus an appropriate third part as needed. Good luck with that!
Total Intolerance and Counterpart Theory

It may seem that the essentialist has another way out of the Chisholm/Chandler paradox, namely to cut them off at the start by allowing no scope for variable realization of essence. So, if our essentialist is taken with the essentiality of origins, he had better say things like this: It is not just that I had to be conceived of somewhere, I had to be conceived at the exact spatio-temporal
location where I was actually conceived. It is not just that I had to have some original constitution; I had to have the exact original constitution I actually had. And so on and so forth for everything that has come into being. Chisholm (1973) himself seemed tempted by this kind of response, for he presents strict mereological essentialism as a solution to the specific kind of paradox of variation in parts across worlds which he considers. (More exactly, he advocates mereological essentialism for primary objects and treats so-called vulgar objects as successions of these.) Chisholms thought was that tolerance about constitution, i.e. allowing for variable realization of that part of a things essence, leads to trouble, so why not conclude that essences are not tolerant in this way, so that a thing cannot survive variation in any of its constituent parts? The more general thought is that tolerance, i.e. variable realization of essence, leads to trouble, so why not conclude there are no such essences? This amounts to using the alleged paradox as a way to establish a principle of limited variety of essences namely one that rules out variably realized essence. If the alleged paradox was indeed a paradox, this would be a promising line of argument. However, since the supposed paradox, once clarified, depends on admitting essences no such argument remains. A structurally similar response applies to Graeme Forbes (1984) solution to the paradoxes, namely the invocation of counterpart theory. Here the argument is that it is the assumption that a transitive relation of identity holds across items in different possible worlds which makes for trouble, so we should replace that assumption with the idea that items in different possible worlds can at most be counterparts of each other, where the relation of counterparthood is understood as the relation of similarity in central respects. Forbes (1984) motivation for adopting counterpart theory comes from these alleged paradoxes; since the paradoxes are easily dissolved, depending as they do on admitting variable essences, this motivation does not hold up. But what if one is motivated not so much to resolve the paradoxes, but to avoid plenitude? Here, some versions of counterpart theory are extremely helpful, in particular ones that allow for a single item to have multiple counterparts in a single world, and which then rely on context (or some other additional factor) to select from among these many potential counterparts. Such a version of counterpart theory was most prominently defended by David Lewis (1986), and is an example of an Abelardian account of modality (named after Peter Abelard, of Abelard and Eloise fame). Abelardian accounts of modality hold that the subject position of de re modal claims is referentially opaque. On an Abelardian treatment of de re modal or essentialist claims, such claims hold of entities only under a description or a conception, perhaps made salient by the context of utterance or the particular way the entity is characterized. Abelardianism dispenses with the central essentialist claim that there is a fixed set of essential properties that characterize an object. Instead an object may be essentially F relative to one context, but accidentally F relative to another. Essentialism invites plenitude because it supposes that there are two ways of
having a property, and that if one object has a property accidentally while the other has it essentially, then the two objects cannot be identical. Abelardianism dispenses with the genuine de re modalities that essentialism embraces, and so the argument to plenitude does not apply. If one wishes to avoid plenitude or at least this source of plenitude then embracing Abelardianism is a natural move. (In fact, there are a great deal of similarities between Abelardianism and essentialism with plenitude. In particular, where essentialism with plenitude posits multiple items each with distinct modal profiles, the Abelardian posits multiple counterpart relations, thus obtaining the same range of modal profiles. For a detailed discussion of how this kind of Abelardianism about essence might be taken to avoid the problem of plurality of genuine essence with subsequent co-location, see Leslie (submitted)). Interestingly, however, this is not the move that Forbes makes. Rather, he adopts absolute counterpart theory, which holds that there is a single privileged counterpart relation for each object, so that each object has at most one potential counterpart in each world. As his later work (1986) suggests, Forbes wants to remain a genuine essentialist; he does not believe that there are many counterpart relations which can be made salient by context, and so many equally good accounts of the so-called essences of things. Thus Forbes does not embrace Abelardianism, but rather adopts counterpart theory primarily in response to the paradox as a way of replacing the transitive notion of cross-world identity with the potentially intransitive notion of the counterpart relation. But as Forbes (1986) still accepts that an item has some of its properties essentially and other accidentally and that these two ways of having a property are not mediated by any additional hidden arguments he still faces the question of why there is but one axe here made essentially of, e.g., two out of three of these parts, but not another made essentially of two out of three of these other (overlapping) parts. Again, a principle of limited variety would be needed to block the proliferation of entities.
Haecceitism and Plenitude

A set-up similar to the Chisholm/Chandler paradox, called the Four Worlds paradox (Salmon, 2005 (originally 1981)) is often used to argue for haecceitistic differences between worlds i.e. the thesis that there are possible worlds that are identical in all qualitative respects, yet nonetheless represent distinct possibilities, because their inhabitants, though indiscernible, are numerically distinct. This set-up is as follows: consider, as before, a world w1 in which an axe Axe1 is made of Blade1, Shaft1, and Handle1. Further, Axe1 is again essentially constituted by at least two out of the following three parts: Blade1, Shaft1, and Handle1, plus the appropriate kind of third part if needed. Thus Axe1 can be made (again, as before) at w2 out of the following parts: Blade2, Shaft1, and
Note: The axes in World 2 and World 4 are made of the same parts, but cannot be identical according to the argument. Figure 3. An Argument for Haecceitism.
Handle1. Now we are asked to consider world w3, in which a different axe, Axe2 is made from the following parts: Blade2, Shaft2, and Handle1. Axe2 is essentially constituted by at least two out of the following three parts: Blade2, Shaft2, and Handle1, plus the appropriate kind of third part if needed. Clearly, Axe1 is not identical to Axe2. But now consider world w4 in which Axe2 is made from the following three parts: Blade2, Shaft1, and Handle1. (See figure 3.) Let us further assume that all four worlds contain nothing except these axes. The argument now goes as follows: w2 and w4 represent distinct possibilities, because w2 contains Axe1 and w4 contains Axe2, and Axe1 is not identical to Axe2. However, the two axes in these two worlds have exactly the same parts, and since the worlds contain nothing else, the two worlds are themselves qualitatively identical. Thus, possible worlds can differ haecceitistically, i.e. solely with respect to the identities of the individuals within them. I take no stand here on whether there are haecceitistic differences between worlds or not, however I think that this set-up does not establish this conclusion, even though it is often taken to do so. David Lewis (1986), for example, argues that the set-up pushes essentialists who accept transworld identity and a plausibly tolerant essentialism to an implausible haecceitism. I believe that this is mistaken; in particular, once the connection between essentialism and plentitude has been made, the set-up gives us no additional reason to accept haecceitism at the level of possible worlds. To see this, note that in order for the above argument to be valid, a hidden assumption is required: w2 contains Axe1 and not Axe2 and w4 contains Axe2 and not Axe1. Once we allow co-located entities of the sort discussed above as the essentialist should this assumption is unmotivated. The above set-up can be equally well modeled with w2 = w4, without violating any of the articulated premises; we only need to suppose that this one world contains both Axe1 and Axe2. (See figure 4.) Again, this does not show that there cannot be haecceitistic differences between worlds, but it does show that the essentialist can easily handle this case without recourse to haecceitism about worlds.
Note: In light of plenitude, there is no need to posit a fourth world. Figure 4. A Response from Plenitude.
Conclusion
The Chisholm/Chandler/Four Worlds paradoxes have been misdiagnosed, partly because of (i) a confusion between the coherent notion of a variably realizable essence and the incoherent notion of variable essence and (ii) a blind spot common to most (but not all) essentialists and anti-essentialists. The very distinction between two ways of having properties and including parts the essential way and the accidental way makes for many candidate essence/accident profiles that differ not in respect of what properties and parts they cite, but only in respect of which of these properties and parts are had accidentally and which essentially. The majority of these candidate essence/accident profiles would seem to be admissible in that there seems no metaphysical basis for an invidious distinction between one such profile and another. But then whenever one profile is instantiated so also are the others, and we have multiple co-located objects, differing only in their essences. Once this is recognized, these paradoxes do not motivate restricting the accessibility relation between worlds, resorting to counterpart theory, or even adopting haecceitism. However, proper consideration of the paradoxes serves as a reminder of a surprising consequence of essentialisms core distinction between two ways of having properties and including parts namely massive collocation of things of the same manifest kind, axes, ships.
Notes
1. A precise formulation of what this requirement comes to is tricky, though the oft-cited idea that utterly natural properties be ontologically on a par and
Essence, Plenitude, and Paradox / 295 independent from each other is a potential model. See Yablo (1987) and Leslie (submitted) for discussion. 2. Chisholms original formulation of the paradox featured two entities rather than one, and originally focused on two men, Adam and Noah. If we retain ships as our example, the set-up would be as follows: consider two ships, and suppose that at each world they swap a plank with each other, so that at the final world in sequence the first ship is made from the same planks that the second ship was made from in the first world, and vice versa. Thus we have two worlds that are qualitatively identical (i.e. the first and last worlds), yet differ with respect to the identity of the two ships. I discuss this aspect of the paradox (though with a simpler set-up), in the final section of this paper entitled Haecceitism and Plenitude. 3. It is important to notice that, while these two examples involve original constitution, the paradox can be generated for any cluster of essential properties. (In fact one of Chisholms formulations involves Adam and Noah.) One simply needs to construct a list of an items essential properties, and then decide that a degree of toleration is in order. The same paradox will ensue. 4. See Forbes (1984) for extended discussion, and Salmon (1986) for an extended reply.
References
Chandler, H. S. (1976). Plantinga and the Contingently Possible. Analysis, 36(2), 106109. Chisholm, R. (1967). Identity through Possible Worlds: Some Questions, No us, 1, 18. Chisholm, R. (1973). Parts as Essential to Their Wholes, Review of Metaphysics, 26, 581603. Fine, K, (1994). Essence and Modality, Philosophical Perspectives, 8, 116. Fine, K. (1995). Senses of essence. In W. Sinnott-Armstrong (ed.) Modality, Morality and Belief: Essays in honor of Ruth Barcan Marcus. Cambridge: Cambridge University Press, 5373. Fine, K, (1999). Things and Their Parts. Midwest Studies in Philosophy, 23(1), 6174. Fine, K. (2003). The Non-Identity of a Material Thing and its Matter, Mind , 112(46), 195234. Fine, K. (2008). Coincidence and Form, Aristotelian Society Supplementary Volume, 82(1), 101118. Forbes, G. (1984). Two Solutions to Chisholms Paradox. Philosophical Studies, 46(2), 171187. Forbes, G. (1986). In Defense of Absolute Essentialism. Midwest Studies in Philosophy, 11(1), 331. Hawthorne, J. (2006). Metaphysical Essays, Oxford: Oxford University Press. Johnston, M. (1984). Particulars and Persistence. Ph.D. Dissertation, Princeton University. Johnston, M. (1992). Constitution is Not Identity, Mind , 101(401):89106. Johnston, M. (2006). Hylomorphism , Journal of Philosophy, 103(12), 652699. Kripke, S. (1980). Naming and Necessity, Cambridge, MA: Harvard University Press. Leslie, S. J. (Submitted). Essence and Accident: Against Limited Variety. Lewis, D. (1986). On the Plurality of Worlds. Oxford: Blackwell. Lewis, F. (1982). Accidental Sameness in Aristotle, Philosophical Studies, 42(1), 136. Mackie, P. J. (2006). How Things Might Have Been: Individuals, Kinds, and Essential Properties. Oxford: Clarendon Press. Matthews, G. (1982). Accidental Unities. In M. Schofield & M.C. Nussbaum (eds.), Language and Logos. Cambridge: Cambridge University Press: 22340.

Quine, W. V. O. (1953). Three Grades of Modal Involvement. Proceedings of the XIth International Congress of Philosophy, 14, 6581. Quine, W. V. O. (1976). Worlds Away. Journal of Philosophy, 73(22), 859863. Salmon, N. (1986). Modal Paradox: Parts and Counterparts, Points and Counterpoints. Midwest Studies in Philosophy, 11(1), 75120. Salmon, N. (2005). Reference and Essence (Second Edition). Amherst, NY: Prometheus Books. White, N. (1971). Aristotle on Sameness and Oneness, Philosophical Review, 80, 177197. Yablo, S. (1987). Identity, Essence, and Indiscernibility, Journal of Philosophy, 84, 293314.
Philosophical Perspectives, 26, Philosophy of Mind, 2012
CONCEPTS, ANALYSIS, GENERICS AND THE CANBERRA PLAN1
Mark Johnston Princeton University Sarah-Jane Leslie2 Princeton University
My objection to meanings in the theory of meaning is not that they are abstract or that their identity conditions are obscure, but that they have no demonstrated use.3 Donald Davidson Truth and Meaning
From time to time it is said that defenders of conceptual analysis would do well to peruse the best empirically supported psychological theories of concepts, and then tailor their notions of conceptual analysis to those theories of what concepts are.4 As against this, there is an observation traceable at least as far back to Gottlob Freges attack on psychologism in The Thought that might well discourage philosophers from spending a week or two with the empirical psychological literature. The psychological literature is fundamentally concerned with mental representations, with the mental processes of using these in classification, characterization and inference, and with the sub-personal bases of these processes. The problem is that for many philosophers, concepts could not be mental items. (Jerry Fodor is a notable exception, we discuss him below.) We would like to set out this difference of focus in some detail and then propose a sort of translation manual, or at least a crucial translational hint, one which helps in moving between philosophical and psychological treatments of concepts. Then we will consider just how, given the translation, the relevant psychology, particularly including recent work on the generic character of many of our platitudes or early developing central beliefs (Gelman, 2010; Hollander, Gelman, & Star, 2002; Leslie, 2007, 2008, in press a; Leslie & Gelman, 2012; Leslie, Khemlani, & Glucksberg, 2011; Mannheim, Gelman, Escalante, Huayhua, & Puma, 2011; Tardif, Gelman, Fu, & Zhu, 2011), bears upon
114 / Mark Johnston and Sarah-Jane Leslie
philosophical theories of concepts, and especially on the new style of conceptual analysis that now goes under the heading of the Canberra plan.
A Philosophical Theory of Concepts

What follows in this section is a short account of the general form of a substantial philosophical theory of concepts. We do not endorse it. We simply take it to be a widely held view, at least in its significant sub-parts; we also take it to be a view that animates the common practice of conceptual analysis within philosophy, along with the so-called conceptual theory of the a priori the idea that the source of the a priori lies in the application conditions of concepts. On the substantial theory, concepts are abstract objects individuated by their conditions of application to entities, and it is possession or grasp of these concepts which guides the use of terms. Thinkers who grasp or possess a given concept C will often come to associate it with a term or representation T and then use T to classify or characterize entities in accord with the specific conditions of application of the concept C. For such thinkers T will then express the concept C. Other thinkers, who do not even implicitly know the application conditions associated with C might nonetheless defer to those who use T to express C, and intend to do likewise even though they themselves do not use T to classify or characterize entities in accord with the specific conditions of application of C. They can thus make conceptual mistakes, such as supposing that arthritis is a more general infirmity than inflammation of the joints. They can then be said to have an imperfect or incomplete grasp of the concept expressed by arthritis. Thus far, concepts are abstract objects encoding conditions of application of words, phrases and the like; in effect, they are rules of reference determination across possible situations, rules that we grasp and employ as guides in our use of terms. They are what give spoken and written words life; that is, it is by being associated with this or that concept that a word or phrase comes to have conditions of application. They would also explain how it is that words and phrases like si that appear in distinct languages can have double lives. The speakers of the distinct languages associate different concepts with the same words. Likewise, within a language, the association of different concepts with the same words and phrases is what accounts for ambiguous words and phrases such as bank and the bank he built up. The postulation of concepts, and the hypothesis that it is the grasp of these concepts if not by us then by the experts and lexicographers to whom we defer that properly guides our use of words, makes for concept publicity; that is, the possibility of two speakers with very different beliefs about a subject matter sharing the same concept of that subject matter. Someone who knows that arthritis is a disease of the joints can genuinely disagree with someone who believes it can spread to the bones. The second is in error about the application conditions of a concept that both speakers share. As a result
Concepts, Analysis, Generics and the Canberra Plan / 115
he can be meaningfully corrected by the experts; for example they are not changing the subject on him when they tell him that one cannot have arthritis in the thigh. Since words can combine compositionally to produce phrases of any arbitrary length, the life-givers that are concepts had also better combine compositionally to give life to phrases of arbitrary length (see, e.g., Fodor, 1998, Fodor & Lepore, 2002). So the concept female lion had better be some computable construction out of the concept female and the concept lion. Very likely, it had better amount to an intersection of these concepts; that is, a concept whose conditions of application are given by the conjunction of the application conditions of the constituent concepts. A natural extension of this whole way of thinking involves associating a certain sort of concept with a whole sentence. The concept associated with a whole sentence is a computable construction out of the concepts associated with the sentences constituent words and phrases. The special name for such a concept is a thought, where a thought is not a sentential or representational vehicle but rather what is expressed by such vehicles. On this view, it is precisely because subsentential concepts have application conditions that sentences express thoughts, and so can be true or false. The special name for the application conditions of these structured concepts, the ones that are expressed by whole sentences, is truth conditions. On the substantial theory, if there are concepts, and if our use of terms is explained by our grasp of the associated concepts, then there will be certain immediate inferences that will look like good candidates to be underwritten simply by our grasp of the relevant concepts, since those inferences seem utterly reliable and do not seem to depend on any intermediate empirical premise. So it is with the inference from
Bonnie is a mare
to
Bonnie is a horse.
Furthermore, there is a route to recognizing the truth of the corresponding conditionals of such inferences by appealing simply to our grasp of the relevant concepts. So it is said that it is a priori that
If x is a mare then x is a horse
where this means that there is a route to recognizing its truth that does not depend for its justifying force on any empirical information about how the world works. This also explains why the conditional could not be false. The conditional gives no hostages to fortune; it involves no bet on any contingent way the world
is. Accordingly, the conditional has the defining feature of a so-called conceptual truth: it is not only a priori but necessary. Finally, if our use of terms is explained by our grasp of the associated concepts, and if concepts have necessary and sufficient conditions of application, then we might well hope to make our implicit knowledge of the application conditions of our terms explicit, by reflecting on what seems to be guiding us when we apply them in the vast range of imaginary cases that we can conjure up in the philosophical armchair. We can then form various hypotheses about just which sets of application conditions lie behind which terms, and we can test these hypotheses further by means of the method of consulting real and imaginary cases. The aim is, for each significant philosophical concept that is not simple that is, for each concept that is compounded out of other concepts to arrive at an a priori and necessary biconditional of the form
x is T if and only if x is . . .
where T is a term expressing the concept in question. Not just any such biconditional will do. It has to be non-trivial. And it has to be rich enough to enable the derivation of all the a priori and necessary truths that arise from the concept associated with T. Finding a rich enough a priori and necessary biconditional is not easy; but if we succeeded we would have an analysis of the relevant concept. Generalizing, if we were to provide such analyses for all of the concepts of philosophical importance person, cause, essence, knowledge, mental state, matter, good and so on and so forth we would have articulated the bases for all of our distinctively philosophical a priori knowledge. Something like this substantial picture has a good claim to be at the motivating core of what was once called analytical philosophy (from Gottlob Frege, Kurt Godel, A.J. Ayer, H.P. Grice and Roderick Chisholm through to George Bealer and Frank Jackson). Arguably, with the simple addition that the abstract objects that encode the application conditions of terms are the rules we grasp and which then determine our sense of the rightness of our applications of the term in particular cases, the picture is the real target of the skepticism about meaning that Saul Kripke (1982) saw in certain passages in Wittgenstein. Whether or not one accepts the picture, it goes some way toward illuminating just why a philosophical theory of concepts might take little notice of mental entities or mental processes. Of course, the picture raises many questions: What are the specific application conditions of this or that particular concept? What it is to grasp a concept? What is it for a concept to be associated with a word or phrase? Are concepts really life-givers or is talk of them simply a way of encoding, as it were after the fact, the details concerning the extension of this or that word- or phrase-in-use? Are concepts then idle wheels? Is a priori knowledge
really knowledge deriving from knowledge of the application conditions of our concepts? If so, is there any that goes beyond knowledge of merely stipulative definitions? Is the method of imagining a case and calling on our intuitions about whether a given concept applies in such a case a good way of making explicit our implicit grasp of the application conditions of the concept? Are some concepts, for example the familiar concept plus, more natural than others, in the sense of being privileged attractors of the terms that we use, attractors which we can only avoid by explicitly intending to use our terms in certain specific ways? Is this why our term + expresses the concept plus rather than the concept of some finite function that is like plus for a huge finite stretch but gives no answer just where our capacity to add actually gives out? It may well turn out that these questions have disappointing answers. (For some of best arguments that this is so, see Willard van Orman Quine (1951), Hilary Putnam (1962, 1975), Gilbert Harman (1974, 1975), Jerry Fodor (1998), Timothy Williamson (2003) and Michael Devitt (2005).) Concepts may not be guides in any psychologically real sense. There may be no non-trivial biconditionals that neatly capture the application conditions of our concepts. There may be no a priori knowledge, period, or, at least, almost none beyond that delivered by logic and mathematics. Furthermore, the now popular idea of certain concepts, or the properties they demarcate, being privileged attractors may just be hopeful mysticism. At the end of the day, the frame provided by philosophical talk of concepts, analysis, and the distinctively philosophical a priori may have nothing very interesting left in it. (Obviously that pessimistic view also has its distinguished opponents; among them David Lewis (1970, 1994, 1997), Christopher Peacocke (1992) George Bealer (1987), Frank Jackson (1994, 1998), David Chalmers (Chalmers & Jackson, 2001) and Paul Boghossian (2003).) What has so far been missing from the debate over these central philosophical issues is any systematic account of just how the psychological literature might bear upon them. By way of beginning to fill that gap, we will revisit these issues at the end of the paper, once the interrelations between philosophical and psychological theories of concepts have been set out in appropriate detail. Our ultimate view is that the psychology does resolve the philosophical standoff in a fairly convincing way. In any case, it should be clear that on what we have been calling the substantial theory, concepts are abstracta that encode conditions of application; they are not mental processes or mental representations. One obvious route to the conclusion that concepts are not mental is to note that on the substantial view of concepts, the content of what is thought is a kind of concept, and only someone in the grip of the vehicle/content confusion would think that what is thought is a mental representation. Mental representations might be vehicles that express thoughts, but a mental representation is not the content of what is thought. Representations have characteristic effects, but no one is in danger of coming under the causal influence of a thought content, as opposed to a
representation of it, or a proponent of it. So the concepts that are the variety of things that are thought are none of them mental representations, and neither are their constituent concepts. Furthermore, as Christopher Peacocke (1992) emphasizes, once the substantial theory of concepts is up and running, it seems obviously coherent to suppose that many or most concepts are beyond our grasp, so that we will never bear any interesting mental relation to them. However, psychologists who theorize about concepts are directly concerned with putatively mental entities, namely mental representations (including mental images and the like) that play a central role in our activities of classifying and characterizing and making immediate inferences about objects. A psychological account of our concepts in this sense is an account of just how it is that we, or our cognitive systems, systematically exploit certain sorts of general information to use the relevant mental representations to classify and categorize objects, and make inferences about them. In fact, when psychologists do focus on mental representations they show little or no interest in the rules which determine their possible world extensions. What then is the relation of the psychological literature to the philosophers interest in concepts? Is it simply a confusion of tongues to suppose that psychologists and philosophers are concerned with the same subject matter? We believe that once the import of the psychology is made clear, a different philosophical view of concepts, one which attributes less knowledge to conceptusers as such, becomes much more plausible. On this concepts as terms-in-use view, a concept is a term, and a term is just an interpreted string in a language; a string which has, thanks to its conventional use, an extension, i.e. a range of items to which the term applies. Accordingly, it is only use, perhaps only conventionally governed use, which gives terms the life that arbitrary strings lack. There need be no more to specifying a term than characterizing its syntactic form and giving a rule that determines its extension. There need be nothing behind terms at least nothing like concepts understood as meanings or senses or extension-determining rules that speakers grasp and which then guides their use of the relevant terms. Mutual correction while participating in a convention for using a term selects for a variety of effective heuristics or criteria that then guide individuals in their use of that term. To know how to use a term is just to have some such effective criterion, and this will often fall far short of even a tacit understanding of the conditions of application of the term. Hence, on the concepts as termsin-use view, there is no reason to suppose that speakers can render explicit the conditions of application of their terms simply by reflecting on how they would use them in a variety of circumstances. For such actual and counterfactual usages reflect only the speakers empirical criteria, which may fall short of any grasp of application conditions. Thus armchair philosophical analysis, as traditionally conceived, can at best produce fragmentary and inconclusive results. It can be no more than the articulation of the criteria we presently use in deciding whether to apply a term.
Problems with Psychological Theories of Concepts?

By starting with the philosophers substantial notion of a concept, and adopting a translation of the following sort
Concept in the psychological literature means pretty much what philosophers who have accepted significant parts of the substantive theory have meant by it
one can quickly make mincemeat out of the psychological literature. Jerry Fodor has already done much of this work for us in his Concepts (1998), where he points out that so construed, most, if not all, psychological theories of concepts fail the straightforward requirements of compositionality and publicity. To put it in the broadest possible terms, psychologists seem most interested in discovering our criteria or ways of telling when a given individual has a property or falls within a specified kind, and our criteria or ways of telling which inferences about members of a kind are the ones to make. But it will not be in general true that our ways of telling that something falls under a complex concept F&G is a joint application of our ways of telling whether something falls under the concept F and our ways of telling whether something falls under the concept G. Moreover, there is massive individual variation in our ways of telling whether something falls under the concept F ; we may only be able to look it up in a book, you may have written the book and conducted the complex experiments needed to determine that it is F, and yet so the substantial theory has it all of us can share the same concept F . Someone more enamored than Fodor of the conceptual theory of the a priori say the Christopher Peacocke of A Study of Concepts (1992) and more recently The A Priori (2004) might continue in the same general vein, by noting that, if we simply focus on our ways of telling which inferences about members of a kind are the ones to make, we will fail to distinguish between those inferences underwritten by our grasp or possession of the concept of the kind in question, and those inferences that appeal to collateral knowledge about members of the kind. Yet psychologists studying concepts happen to show no interest in this distinction. Instead, they focus on what generalizations and inferences are cognitively fundamental, in the sense of being (i) early developing and (ii) central to our a posteriori conceptions of the kinds in question. We believe that the perception that these are in any sense defects in the psychological literature derives from the naive translation principle, and not from the psychological theories themselves. We will briefly review the relevant literature, emphasizing as we go just how recent discoveries concerning generics bear on that literature, then propose a different translation scheme, and finally examine the ways in which the psychological theories might put pressure on philosophical theories of concepts.
Some Psychological Theories of Concepts

The focus and the questions that drive psychological theorizing differ significantly from those which organize philosophical discussions of concepts. This is partly obscured by the fact that almost all reviews of the psychological literature on concepts begin with the so-called classical view, which is easy to misconstrue as simply a version of the substantial philosophical theory of concepts. On the classical view, when applying most lexical concepts (a lexical concept is a concept that is expressed by a single word) subjects actually exploit represented necessary and sufficient conditions for falling under the concept. The standard illustration of the classical view is the concept bachelor, which, it is claimed, is composed of the concepts unmarried and man, such that anything is a bachelor just in case it is an unmarried man. Of course, not all concepts can be decomposable in this way; there must be some or other stock of primitive concepts, out of which all other concepts are ultimately composed. Nevertheless, concept learning, on this view, involves combining such primitive concepts to form complex ones, and classifying an item under a decomposable concept really is supposed to be a matter of checking whether the item satisfies the necessary and sufficient conditions specified by the decomposition. Thus, someone who possesses the concept bachelor will classify items as bachelors based on their gender and marital status, in such a way that she will classify something as a bachelor just in case it is a man and unmarried. Mutatis mutandis for concepts such as lion, dog, and table. Notice that, despite the initial appearances, the classical view of the concept bachelor is not just the (false but) banal remark that someone is a bachelor if and only if he is an unmarried man; it is instead the rather heady empirical thesis that we actually use the concepts unmarried and man in deciding whether to count something as a bachelor. It is distinctively a thesis about the criteria we actually use, not, or not just, a thesis about the application conditions of our concepts. Since the 1970s, the classical view, understood as a thesis about the criteria we actually use, has been quite roundly rejected. Much of the reason for its rejection has to do with the discovery by Eleanor Rosch and her colleagues of so-called typicality effects (e.g., Rosch, 1973, 1978; Rosch & Mervis, 1975). For many categories, some members of a category are perceived as being more typical examples of the category than others, and it turns out that how typical a category member is actually predicts a very wide range of experimental results. For example, people are quicker to categorize typical members, and are more confident and consistent in their categorization of typical members. When learning a novel concept, people learn to categorize the typical members first, and they learn the concept faster when presented with typical members in the learning phase. There are also myriad effects of typicality on language learning and use, on reasoning, and so on so forth. (For some very helpful reviews, see Laurence & Margolis, 1999; Murphy, 2002; Smith and Medin, 1981).
Positing that people represent and exploit necessary and sufficient conditions does not explain typicality effects. Knowing that something is a bachelor just in case it is unmarried and a man does not give any information about what makes for the typical versus the atypical bachelors; for they are all alike in respect of being unmarried men. The pope is an unmarried man, but quite an untypical bachelor. Something else has to be posited to explain typicality effects; but once we have posited this something else there may be no empirical reason to also posit the representation of necessary and sufficient conditions. That is, suppose we posit some sort of mental representation that explains the powerful effect of typicality on peoples categorization practice, and suppose we satisfy ourselves that this posited representation explains the experimental findings concerning peoples dispositions to categorize items and generalize and reason about them. Then we will have accounted for all the data without positing operations on subjects representation of necessary and sufficient conditions. From the perspective of explaining and predicting the target empirical results, knowledge of necessary and sufficient conditions then looks like an idle wheel there are no results whose explanation demand positing the representation of necessary and sufficient conditions. The classical view has fallen into disrepute because many investigators believe that something like this has turned out to be true. There may be some confusion concerning the rejection of the classical view: the evidence does not, it is sometimes said, prove that for most of our lexical concepts subjects do not represent and exploit necessary and sufficient conditions. That is correct; the findings of Rosch and her colleagues do not prove that there cannot be necessary and sufficient conditions lurking somewhere in our mental representations. However this form of resistance misses the real thrust of the psychological data; the issue is not that certain empirical results prove the absence of represented necessary and sufficient conditions, but rather that there is simply no evidence for their representation in the case of most lexical concepts.5 Consider, for example, an experiment conducted by Jerry Fodor and his collaborators, in which they asked whether one could find any differences in processing time that would indicate that one concept is composed in part of another. If, e.g., the concept murder is composed in part by the concept kill (as has been claimed), then it should take longer to process murder than kill , since processing the former involves processing the latter as a proper part. However, this prediction of the classical view is not borne out (Fodor, Garrett, Walker, & Parkes, 1980). Again, no evidence in favor of the classical view emerged from that experiment. It is for this sort of reason that psychologists have moved beyond the classical view of concepts. Driven by the need to explain typicality effects, many psychologists have embraced the prototype theory of concepts, where prototypes are statistical functions over properties, which assign weights to features based on how likely a category member is to have that feature, or conversely, based on how likely something with that feature is to be a category member. There are a number
of different proposals that fall under the heading of the prototype theory (see Murphy, 2002, for an extensive review), but they generally appeal to features that are in some way statistically related to category membership. For example, the prototype for dog might include features such as barks, has four legs, has a tail , wears a collar and so on. These features are not candidates to figure in universal necessary and sufficient conditions since not all dogs have these features; an unfortunate creature can still be a dog even if it has three legs, no tail, no collar and no voice. However, the basic idea behind prototype theory as applied to dogs is that if one is confronted with an animal and wishes to determine whether or not it is a dog, one will use this animals features, or lack thereof, in a complex sub-personal calculation based on the weights of the various features in the prototype of dog. (The details of this calculation differ a great deal depending on the particular version of the theory.) The weight that a feature receives in the prototype is generally taken to be determined by two sorts of statistical facts, namely the prevalence of the property among dogs (so barks, has a tail , etc would receive high weights since most dogs have these features), and/or the cue validity of the property; that is, how likely it is that something with that feature is a dog. Thus even though, perhaps, most dogs dont wear collars, the probability of something being a dog if it wears a collar is high, so wears a collar might receive a significant weight in the prototype. The more highly weighted features an individual has, the more typical an exemplar of the kind it will be; prototype theory thus places typicality effects first-and-foremost among the data it aims to explain. Of course, one could try to express the prototype view in terms of some set of necessary and sufficient conditions; for example
x is a bachelor if and only if his lifestyle and behavior resembles an appropriate paradigm, e.g. the one presented by Sean Connery in the early James Bond movies.
Clearly, when expressed this way, the prototype theory for bachelor fails miserably. This is just one sign that prototype theory is just not intended as an account of the application conditions as, in effect, an analysis of the term bachelor. Another sign is that, as this very dated example suggests, the paradigms can change over time without thereby producing any change in the application conditions of the term. Prototype theory has many adherents, and many well-motivated critics. While we may often rely on statistically weighted features in categorization, particularly in rapid, perceptually-based categorization, it seems that this cannot be the whole story. Imagine, for example, that you are presented with a raccoon. A perverse scientist then comes along and alters the creature, dying its fur so that it takes on the markings that are typical of a skunk, and even goes so far as to implant a sac of smelly liquid that the creature can use to spray smells when it is under stress. How would you categorize this creature? It now has all the typical features of a skunk, yet overwhelmingly, from elementary school on up,
people say this is still a raccoon (Keil, 1989). This finding is hard for standard prototype theories to accommodate. Furthermore, it seems increasingly clear that typicality ratings are not solely driven by statistical facts; crucially, the causal status of features also matters. Imagine that two features are equally prevalent among members of a kind, but that one is understood as generally being the cause of the other. Suppose then that an instance of the kind has one feature but not the other. Since the statistical facts are the same in both cases, prototype theory would seem to predict that typicality ratings of the individual would not be affected by which feature is lacking. However, individuals exhibiting the effect but not the cause are rated as less typical than those exhibiting the cause but not the effect (Ahn, Kim, Lassaline, & Dennis, 2000). These results and many others suggest that our ways of categorizing things, and reasoning about things in categories, involve a richly structured knowledge base that is responsive to causal-explanatory factors as well as statistical factors (e.g., Carey, 1985, 2009; Gelman, 2003; Gopnik & Meltzoff, 1997; Keil, 1989). As we will use the term here, this is the outlook typical of the so-called theorytheory of concepts.6 Since theory-theory, so construed, posits that our concepts or criteria for categorization and generalization are sensitive to theory and causalexplanatory structure, there is not too much to be said beyond that about the general features of our concepts. Rather, it may be useful to go on to consider concepts within each broad domain, e.g., natural kind concepts, artifact concepts, social concepts, mental state concepts, mathematical concepts and so on so forth. For example, a view known as psychological essentialism seems to provide a great deal of insight into how our natural kind concepts are structured from a very young age (e.g., Gelman, 2003; Leslie, in press b). However, this view is very likely not applicable to artifact concepts or mental state concepts, and certainly not to mathematical concepts. That sort of domain sensitivity should not be seen as a failing of theory-theory, but as an appropriate response to the complex and myriad ways we have of categorizing things, and of generalizing on the basis of those categories. Probably the view that fits best with the mass of empirical material on concepts is a hybrid of theory-theory and some elements drawn from prototype theory. There are terms like red or dog which we can apply rapidly and without reliance on theory, at least in some circumstances. It is natural to think that this goes by way of sub-personal processing of sensory and perceptual information with subsequent comparison with paradigmatic or prototypical sensory and perceptual profiles. The characterization of such prototypical profiles in particular whether and to what extent they use prevalence and cue validity is a complex piece of empirical psychological theorizing, yet to be completed. Obviously, subjects who use such prototypical profiles are not thereby using the yet-to-be-completed psychological theory. When it comes to the immediate sensory or perceptual application of terms like red or dog, subjects exploit a range of prototypical sensory and perceptual profiles, but they do not exploit a theory of those prototypical profiles.
In what follows, we will take this qualification to theory-theory as read, and emphasize instead the explicitly theoretical elements in our criteria for applying terms.
Concepts and Generalizations

One feature that the foregoing psychological theories of concepts have in common is that they all make some reference to properties that are possessed by members plural of the target category; they all involve focus on forms of generalization concerning the category and its properties. According to the classical theory, the relevant generalizations are (modalized) universal generalizations; the prototype view treats them as probabilistic generalizations; and on the theory-theory, they are complex and theory-laden general beliefs. These observations suggest a possible alternative route to studying the nature of our classificatory and inferential heuristics: we should look to the empirical investigation of our earliest and most fundamental generalizations. Suppose, for example, that it was possible to identify and describe our most basic way of forming general judgments about kinds or categories of moving from information concerning individual members of a category to judgments concerning the category in general. It would be quite surprising if this fundamental manner of generalization was not centrally connected with the heuristics for categorization and inference concerning kinds or categories. Thus a natural and conservative empirical hypothesis would be that our conceptual heuristics in large part consist of such fundamental types of generalizations. Recent interdisciplinary research suggests an intriguing possibility along these lines; namely that our basic way of generalizing information issues in generic generalizations, which are articulated in language via generic sentences such as tigers have stripes, lions have manes, and mosquitoes carry malaria (e.g., Gelman, 2010; Hollander et al., 2002; Leslie, 2007, 2008, in press a; Leslie & Gelman, 2012; Leslie et al., 2011; Mannheim et al., 2011; Tardif et al., 2011). Such generic sentences exhibit a puzzling truth-conditional profile, as a few familiar examples quickly illustrate. Consider, for example, lions have manes this strikes most people as obviously true, yet only mature male lions have manes. Thus, there are perfectly normal lions (i.e. female ones) who lack manes, and yet the generic seems true. Further, there are more male lions than there are maned lions (since some males are immature or lack manes for genetic or environmental reasons), yet the generic lions are male is widely rejected. Perhaps more puzzling are generics such as mosquitoes carry malaria, which are accepted despite the fact that only about one percent of mosquitoes carry the virus. Yet, generics such as books are paperbacks are robustly rejected, even though over eighty percent of books are paperbacks (for more discussion of generics, see Carlson & Pelletier, 1995; Cohen, 1996; Leslie, 2007, 2008; for
empirical investigation of peoples judgments of these sorts of generics and others, see Prasada, Khemlani, Leslie, & Glucksberg, in press). Most importantly for our purposes here, generic generalizations are obviously not equivalent to universal generalizations, as is illustrated by the truth of lions have manes and mosquitoes carry malaria. Even generics such as tigers are striped and dogs have four legs tolerate exceptions in a way that their universal counterparts do not. All tigers are striped is falsified by a single stripe-free albino tiger, and similarly for all dogs have four legs; the generics are more robust, however, and remain true in the face of exceptions. Thus, if generic generalizations constitute our most fundamental way of making general judgments about categories, this would seem to raise a further empirical challenge for the classical view. A proponent of the classical view would have to argue that the general information employed in our classificatory heuristics does not originate from our most basic way of forming general judgments. The information that we use to identify members of a category would not come by way of our basic means of forming general judgments about the category. This is not incoherent, but given the overwhelming absence of empirical evidence in favor of the classical view, it amounts to positing another unmotivated defensive epicycle. But why should we think that generic generalizations are more cognitively fundamental than universal ones? Some of the data in favor of this hypothesis comes from the study of language acquisition. As noted, generics have a very complex truth conditional profile; providing an account of when generic sentences are true or false is a quite demanding task (see e.g., Carlson & Pelletier, 1995; Cohen, 1996; Leslie, 2008). In contrast, it is very easy to provide an account of when universally quantified statements are true (all Ks are F is true iff the set of Ks is a subset of the set of Fs). In light of this, one might expect that universals would be easier for young children to acquire and process than generics; however, this is precisely the opposite of what we find. Generics are produced and understood by preschool-aged children, and the data collected to date suggest that these young children have a remarkably adult-like understanding of generics. For example, preschoolers who know that only boy lions have manes will accept lions have manes but reject lions are boys despite implicitly understanding that there are at least as many boy lions as there are maned lions (Brandone, Cimpian, Leslie, & Gelman, 2012; see also Gelman & Raman, 2003; Gelman, Star, and Flukes, 2002; Graham, Nayer, & Gelman, 2011; for a summary of available evidence on generic acquisition, see Leslie, in press a). Preschoolers are generally competent with the quantifier all when it is applied to a specific set of individuals (e.g., Barner, Chow, & Yang, 2009). For example, if preschoolers are shown six crayons and asked are all of these crayons in the box? they are usually able to answer the question correctly. Most of the work on quantifier acquisition has focused on such
situations; however it should be clear that these sorts of limited, non-projectable pseudo-generalizations are not the sort that are involved in conceptual heuristics. The question, then, is how young children fare with open-ended, category-wide universals not all of these crayons, but all crayons. Several studies indicate that they have considerable difficulty processing universal quantifiers in such kind-wide generalizations. Most intriguing, though, is that when preschoolers are confronted with such kind-wide universals, they do not simply provide random, incorrect answers, instead they treat the universals as though they were generics. That is, preschool children not only consistently evaluate generics just as adults do, they also evaluate kind-wide universals as generics (Hollander et al., 2002; Leslie & Gelman, 2012; Tardif et al., 2011; for a detailed review, see Leslie, in press a). In addition to English-speaking children, such findings have also been documented among Mandarin Chinese- and Quechua-speaking children; similar results have also been found with other quantifiers (Brandone, Gelman, & Hedglen, submitted; Hollander et al., 2002; Mannheim et al., 2011; Tardif et al., 2011). Importantly, these findings are just what one would expect on the hypothesis (Leslie 2007, 2008, in press a) that generics, unlike universals (or someor most-quantified statements), articulate cognitively fundamental generalizations. If the cognitive system has a basic, default way of forming general, open-ended judgments then it may sometimes fall back on this means of generalizing when asked to process a more taxing and sophisticated generalization. This tendency would be most pronounced in young children, who would be expected to struggle with the more taxing generalizations. Not only do young children apparently not struggle with generic generalizations, they substitute their understanding of the generic when asked to consider category-wide quantified generalizations. If generics truly do article cognitively fundamental, default generalizations then one would expect that these effects might not be limited to young children. Adults might also be susceptible to the error of treating quantified statements as generics. Indeed, under a variety of circumstances, adults do show a robust tendency to accept universally quantified statements such as all ducks lay eggs, despite knowing that male ducks do not lay eggs (where the tendency to accept the universal was not due to participants interpreting the universal as quantifying over only females, or over sub-kinds of ducks; Leslie et al., 2011; see also Meyer, Gelman, & Stilwell, 2010). This finding would be explicable if adults were not always evaluating the universal claim, but were instead, like preschoolers, sometimes substituting their evaluation of the corresponding generic. Further confirming evidence can also be found in the study of adult reasoning errors. For example, Steven Sloman (1993, 1998) investigated adults evaluations of arguments that involve the quantifier all, and found that their evaluations did not conform to the logic of universal quantification. His participants judged that arguments such as (A) are strictly stronger than arguments such as (B), despite judging that reptiles are indeed animals:
(A) All animals use norepinephrine as a neurotransmitter; therefore all mammals use norepinephrine as a neurotransmitter (B) All animals use norepinephrine as a neurotransmitter; therefore all reptiles use norepinephrine as a neurotransmitter This pattern of judgment is simply mistaken given the logic of the universal quantifier; however, if we replace the universals in the arguments with generics, then the judgments of the participants would be very reasonable. Since generics tolerate exceptions, the claim animals use norepinephrine as a neurotransmitter can be true even if some animals are exceptions to the claim. If one judges then that reptiles are more likely than mammals to be exceptions to the generic, then argument (A) is indeed stronger than argument (B). Hence these results are as one would expect if adults have a tendency to evaluate universals as generics. Note that adults also judge that universals such as all ravens are black are more likely to be true than universals such as all young jungle ravens are black, despite understanding that the latter are a subset of the former (Jonsson & Hampton, 2006). Again, this is incoherent if one is really dealing with universally quantified statements; however, if one were instead evaluating these universals as generics, this would be a reasonable judgment, since for all one knows young jungle ravens may be exceptions to the generic ravens are black. These results are thus naturally read as lending support to the hypothesis that adults are treating these universals as generics. As a further piece of converging evidence from another experimental paradigm, it has been found that both preschoolers and adults recall previously presented quantified statements as generics (Leslie & Gelman, 2012). The hypothesis that generics, unlike quantified statements, articulate cognitively fundamental, default generalizations thus has a fair amount of empirical support at this time. As a final observation in favor of the hypothesis, we might note that quantified statements require a phonologically articulated element, namely the quantifier itself. That is, we say all tigers are striped or most tigers are striped; however, in the case of the generic, there is no corresponding articulated element (e.g., gen tigers are striped). This is not an isolated fact about English, but rather it would appear that few, if any, natural languages have a dedicated, articulated generic operator (Carlson & Pelletier, 1995; Dahl, 1985). The generics-as-default-generalizations hypothesis offers an explanation for this otherwise puzzling fact: if one wishes to interact efficiently with a system, and the system has a basic, default way of proceeding or performing a task, then one need only issue an explicit instruction to the system if one wishes it to deviate from this default way of proceeding. To convey the idea in more intuitive terms, if one is dealing with a child who, say, by default does not pick up her toys, one only needs to say something if one wishes the child to deviate from her default and actually pick up her toys. If one does not wish the child to pick up her toys on a given occasion, it would be a waste of breath to say dont pick up your toys! since this is what will happen even if one remains silent. Thus,
quantifiers may be articulated in language because one needs to tell the cognitive system, as it were, to deviate from its default, generic mode of generalizing, and instead generalize in the universal manner or the existential manner, and so on and so forth (for more details, see Leslie, 2008, in press a). Generics, by virtue of expressing the default mode of generalization, require no such phonological marking.
The Nature of Generic Generalizations

Suppose that, in the light of the foregoing, we were to go further and adopt the generic encoding hypothesis; namely that the heuristics which most fundamentally guide our use of terms are properly formulated in generic terms.7 How would this impact the extant psychological theories of concepts? That depends on just how we as theorists should understand generic generalizations. There would be minimum impact on the classical view if generics were to be understood either as ceteris paribus generalizations as in other things being equal, all ravens are black or as universals over what is normal as in all normal ravens are black. Likewise, the prototype theory would find support if generics could be explained either in terms of the conditional probability of the kind in question having the feature in question or in terms of the cue validity of the feature; that is, its predictive value as an indicator of the kind. Universally quantified generalizations are easily analyzed, since all Ks are F is true just in case the Ks form a subset of the Fs. Yet, as we have seen, generic generalizations of the form Ks are F are not universals, since they tolerate exceptions. Can such generics be analyzed as equivalent to claims that are quantified with most, usually, or generally? This might seem promising for generics such as tigers are striped or ravens are black. There are, however, generics that can be true even though most members of the kind lack the property: consider mosquitoes carry West Nile virus, lions have manes, sharks attack swimmers. Paraphrasing these true generics with most, usually or generally results in false claims, suggesting that this is not the correct analysis. Conversely, a range of false generics become true when paraphrased with most, usually or generally: most school teachers are female, usually, books are paperbacks and generally, humans are righthanded are all correct claims, yet the corresponding generics (school teachers are female, books are paperbacks, humans are right-handed) would seem to be false. Clearly this latter point also rules out the idea that generics are equivalent to any logically weaker claims, such as statements quantified with some. In the semantics literature, it is often proposed that generics are in some sense or other equivalent to claims about what is normal for members of the kind. There are subtle variants on these accounts (see, e.g., Pelletier & Asher, 1997), but they share the core notion that generics tell us something about the
properties of all of the normal members of the kind. Again, this seems promising for generics such as tigers are striped, since albino tigers may be counted as in some way abnormal. More challenging are generics such as lions have manes and ducks lay eggs, since there is nothing abnormal about maneless female lions, or eggless male ducks. The standard response to such examples is to suggest that these generics involve covert restriction to the relevant sex (or other natural sub-kind) that is, lions have manes is just elliptical for the claims male lions have manes, which is arguably amenable to an all normal analysis. However, if it sufficed for the truth of a generic that all the normal members of one sex have a property, we would expect that a range of clearly false generics would instead be true. For example, it is surely true that all normal male lions are male; why then does the generic lions are male strike us as clearly false? Further, it is normal for male ducks to not lay eggs, yet ducks dont lay eggs would seem to be false (for further discussion of the point, see Leslie, 2008). Recent empirical evidence also raises some additional difficulties for any approach that appeals to implicit restrictions. Andrei Cimpian and his colleagues investigated whether people think that a kind in which a property is had only by one sex (that is, where an effective domain restriction to a single sex is possible) is a better satisfier of a generic such as Xs have manes than a kind in which half the members, regardless of sex, have the property (that is, where no such effective restriction to single sex is possible). If generics of the form Xs have manes are only accepted because people are implicitly restricting the domain to a sub-kind whose members normally possess the property, then they should think that a kind in which the property is had by one sex is a better satisfier of the generic that is, a better satisfier than a kind in which half the males and half the females have the property. However, no such preference was found (Cimpian, Gelman and Brandone 2010; see also Khemlani, Leslie, & Glucksberg, 2009, for distinct empirical considerations against the notion of implicit domain restriction). If generics such as lions have manes do not implicitly mean male lions have manes, then they would seem to count against the analysis of such generics in terms of what is normal for members of the kind. There is after all nothing abnormal about the maneless female lions. Further challenges are also posed by generics such as mosquitoes carry West Nile virus and sharks attack swimmers. Very few members of the kinds in question have the relevant properties, and it is hard to maintain that this tiny minority represents what is normal, while, e.g., the virus-free mosquitoes are somehow abnormal. Examples such as these suggest that generics need not always involve properties that are normal for the kind, or even for any of its sub-kinds. Nor does it seem plausible to suppose that these generics are, respectively, equivalent to other things being equal, all mosquitoes carry the West Nile virus and other things being equal, all sharks attack swimmers. There is another possible approach to the analysis of generics, one which might be favored by prototype theorists trying to accommodate the generic encoding hypothesis. Perhaps some generics such as tigers have stripes are accepted because the property is highly prevalent among
members of the kind (i.e., because most tigers have stripes), while other generics such as mosquitoes carry West Nile virus are accepted because if something carries West Nile virus, it is very likely to be a mosquito that is, because the cue validity associated with the generic is high. If such an account of generics were successful, then this would constitute evidence in favor of the prototype theory of our classificatory heuristics. Since prototype theory standardly involves features that are either highly prevalent or highly diagnostic (or some combination of the two) it would then fit well with the generic encoding hypothesis, and it could naturally assimilate the considerable evidence in favor of that hypothesis. Problems arise for such an account along several dimensions. First and most importantly, even the combination of high prevalence and high cue validity does not suffice for the truth of a generic. Consider examples such as books are paperbacks or tigers are Bengal tigers, (As it happens, there are six remaining subspecies of tigers in the world, with Bengal tigers being more numerous than all the other subspecies combined). In both cases, the majority of the members of the kind have the property in question, and furthermore the property is highly diagnostic of being a member of the kind; still these generics seem to be false. Second, recent empirical work suggests that peoples judgments of low-prevalence generics (such as mosquitoes carry malaria) are not especially sensitive to their estimates of cue validity. If low-prevalence generics were accepted only because the associated cue validity was high, then one would expect that peoples judgments of these generics would scale to some extent with their estimates of cue validity. However, this was found not to be the case (Prasada et al., in press). If the notions of normalcy, prevalence and cue validity do not allow us to give an account of generics, where does that leave us? As noted above, it would be reasonable to suppose that the best account of our classificatory heuristics should mesh with our best account of generics, since the former involves features that are generalized to kinds and categories, while the latter are languages way of letting us articulate our most cognitively basic generalizations. That is the basis of our advocacy of the generic encoding hypothesis. We have also just seen that generics are simply not equivalent to universals (which means the generic encoding hypothesis does not mesh with the classical view), nor do they seem especially amenable to treatment in terms of probabilistic notions such as prevalence and cue validity (which means that the generic encoding hypothesis does not mesh with the prototype theory). Perhaps, then, in accord with the theory-theory, we need to take into account theoretically rich, content-based factors in giving an account of generics. Consider, for example, the low-prevalence generics that have been discussed throughout this section: mosquitoes carry West Nile virus, sharks attack swimmers, ticks carry Lyme disease. We might add to the list as follows: pit bulls maul children, tigers eat people, lead toys poison children, and so on so forth. Once we are on the lookout for content-based factors as opposed to merely formal statistical relations like prevalence or cue validity we can make something of the fact that these generics all share a common feature: the
predicates in question all express properties that make their possessors dangerous or threatening; they involve the sort of property about which we would wish to be forewarned. Indeed, it is natural to suppose that it would be beneficial for us if our most basic way of generalizing was sensitive to such a higherorder feature of the properties predicated of the relevant kinds. Perhaps, then, generics are accepted even at low prevalence levels if the property in question has this type of higher-order feature (see Leslie, 2007, 2008, in press c for detailed discussion; for empirical support see Cimpian, Brandone, & Gelman, 2010). If generic generalizations are indeed sensitive to the nature of the property being generalized, might that account for other troublesome generics? Suppose that, as part of our theoretical knowledge as this is understood by the theory-theory concerning animal kinds, we register that animals kinds tend to be very similar to each other at a certain level of abstraction: by and large, animal kinds have characteristic noises, characteristic modes of locomotion, characteristic diets, characteristic salient physical features, characteristic methods of reproduction/gestation, and of nurturing the resultant young. Outside the domain of animal kinds, artifact kinds have characteristic functions, kinds of professionals have characteristic social roles, and so on. Suppose that this general structural knowledge is exploited by our cognitive system, so that for each of the relevant characteristic dimensions for each type or kind with which we are presented, we first seek to fill in the relevant values. We might call these values the characteristic properties of the kind. Examples of characteristic properties of kinds would include salient, distinctive physical features (e.g., lions have manes), methods of reproduction and nurturing the young for animal kinds (e.g., ducks lay eggs, pigs suckle their young), and for artifact kinds, functions (e.g., Orange-Crusher-2000s crush oranges which could be a true generic even if every Orange-Crusher-2000 is destroyed before it is used). There is thus a sense in which these generics are answering an implicit question about, e.g., how ducks reproduce, or what the function of OrangeCrusher-2000s is. They do not reflect merely statistical claims about how prevalent the property is among members of the kind. Indeed, if a generic attributes a characteristic property to the kind, then the generic may be accepted even if few members of the kind have the property in question (see Leslie, 2008 for more details). Given that generics are indeed sensitive to rich, content-based factors, it may be possible to explain why some generics are accepted even though the property is not prevalent among members of the kind. What, though, of generics such as books are paperbacks, school teachers are female, or Canadians are right-handed? People tend to reject such generics, despite judging that the property in question is highly prevalent (Prasada et al., in press). One proposal is that generics may be sensitive to the nature of the exceptions to the generic claim (Leslie, 2007, 2008). That is, amongst the members of the kind that fail to have the
predicated property, it may matter how they fail to have the property; in particular whether they simply lack the property, or whether they have an equally salient, concrete, positive property instead . Ticks carry Lyme disease is accepted even when the prevalence estimate is low, but the non-infected ticks are known to simply not carry Lyme disease; they do not have a specific alternative property instead. However, the books that are not paperbacks are instead hardcover. Similarly, people who are not right-handed are instead left-handed ; elementary school teachers who are not female are male. Intuitively, one might feel as though one would be overlooking these hardcover books, these left-handed individuals, and these male school teachers if one accepted the generics in question. Thus, a promising proposal is that generics may be also sensitive to the nature of the exceptions to the generalization. Of course, the study of generics does not provide decisive evidence regarding the nature of our heuristics and criteria. However, if the best account of generic generalizations meshes with theory-theory but not other accounts, this could be naturally construed as providing support for the theory-theory approach to categorization and inference. In our terms, the theory-theory fits best with the generic encoding hypothesis; the empirical claim that the heuristics or, more generally, the criteria we use in applying terms frequently take a generic rather than a universal form. Therefore, in what follows, we will assume that the theory-theory offers the best account of our heuristics of categorization, and that the generalizations involved are articulated in language via generic sentences, whose interpretations are governed by rich, content-based factors. This tentative vindication of the theory-theory, because of its capacity to incorporate the generic encoding hypothesis, may initially seem to raise the stocks of the Canberra plan. To the casual eye the Canberra plan, focusing as it does on the theory, or alternatively the set of platitudes, that guides us in the use of a term may seem to best comport with the theory-theory of concepts. As we shall see, this first impression is badly wrong; the theory-theory is plausible only as an account of the heuristics or criteria we employ in using a term, while the Canberra plan is intended as a general account of how we should specify the application conditions of a term. Moreover, as we will see, the generic character of many of our platitudes raises quite systematic problems for the Canberra plan, at least when it is applied to arenas where the relevant generic platitudes are not backed up by the corresponding universal truths (as they typically are in logic and mathematics).
Some Suggestions about Translation

As even this brief review of the psychological literature makes clear, when relating philosophical and psychological theories of concepts, we need a better translation scheme than the following
Concepts, Analysis, Generics and the Canberra Plan / 133 Concept in the psychological literature means pretty much what philosophers who have accepted significant parts of the substantial theory have meant by it.
Otherwise, philosophers might all too easily follow Jerry Fodors lead in despairing of psychological theories of concepts, while psychologists will be immediately led to think that the philosophical discussion is derisory. The na ve translation principle implies that psychologists should be interested in the application conditions of concepts, but of course they are not. They show no interest in studying those who would be best placed to know the application conditions of concepts, namely clear-headed taxonomists and experts in philosophical analysis. Indeed, given the na ve translation manual, it would have be said that psychologists exhibit what can only be seen as a curious and indefensible fixation; they study the cognitive performance of untutored adults and even of young children! So they might discover, for example, that children and untutored adults withhold the term fruit from tomatoes, olives and peppers. But that just entails that such people are not full masters of the concept fruit, which does indeed apply to these ovaries of flowering plants. That is, these subjects simply do not know the application conditions of the term fruit. Why then is their performance at all relevant in the study of concepts? Is this not an appalling waste of time and money? That line of questioning, which is on its face rather silly, would be quite sensible if the na ve translation were adequate. So it is not adequate. We should do better. Often psychologists simply mean by concept of Ks what ordinary speakers mean, namely a contextually relevant part of someones conception or total theory of Ks. In Doing Without Concepts (2009), Edouard Machery has recently argued that in so far psychologists mean just that, and in so far as they appeal to different parts of a subjects conception of Ks when they explain her perception of Ks, memory of Ks, classification of Ks and inferences concerning Ks there is then nothing uniform to go under the heading of a psychological theory of concepts. However, there seems to us to be more relevant uniformity in the psychological literature on concepts. Mostly psychologists mean by concept of Ks psychologically real clusters of heuristics, or more generally criteria, which guide us in classification, characterization and inference concerning Ks. The crucial interpretive point is that most psychologists working on concepts are not concerned with what an analysis would deliver, namely application conditions of concepts, but with heuristic criteria for classification, characterization and inference. If this is right then we should also avoid the opposite error that could be driven by the na ve translation principle, namely the idea that we could use the experimental methods of psychology, or the methods of opinion surveys, to directly study not just the criteria or heuristics we use, but the application conditions of our concepts, thereby letting the empirical work do our analyses for us, as it were.
Three Main Theses

This overview of the relation between philosophical and psychological theories of concepts naturally suggests the following thesis:
Philosophers who believe in concepts/conceptual knowledge/the distinctively philosophical a priori should treat the psychological theories outlined above, along with other psychological theories in the same line of country, as in significant part accounts of the heuristics or more generally the criteria we actually use to apply terms to things.
This, in its turn, suggests a second thesis, concerning the crucial relevance of the psychology to philosophical theories of concepts:
Philosophers should then ask whether and to what extent their theories of concepts, when combined with the best psychological accounts of the heuristics we actually use to apply terms, will entail massive underdetermination in respect of just which concepts we are using those terms to express.
Thirdly,
To the extent that there is such massive underdetermination of concepts and their associated application conditions by the criteria we use in applying terms, philosophers should lose confidence in the usefulness of conceptual analysis, understood as the making explicit of the conditions of application of concepts by means of the method of cases. To that same extent, we should be pessimistic about the range of the distinctively philosophical a priori.
Dogs, Wolves and German Shepherds

In order to illustrate the last two theses, suppose that in actual fact our heuristics for applying the term dog are exhausted by the following down-anddirty criteria:
Is it an animal? Does it have one of the characteristic looks, smells, coat textures, etc, of one of the familiar kinds of dogs? Is it the offspring of an animal with one of the characteristic looks, smells, coat textures, etc, of those one of those familiar kinds of dogs?
As things actually go in suburban environments these three criteria may be good heuristics for collecting together observed instances of the kind dog, the kind we now know to be the species Canis lupus familiaris. However, this is due
to the contingent fact that the canines around us are almost all of them from that species. In fact Canis lupus familiaris, as the name suggests, is a subspecies of the kind grey wolf, Canis lupus, which includes wolves that are not dogs, but which happen to look very like German Shepherds. Wolves typically dont roam in suburban neighborhoods. So, what we in fact track by means of these heuristics are dogs, but in another environment the same heuristics might lead us to track a larger group, namely the grey wolves. What concept in the philosophical sense should we then assign to our word dog: the concept that comports with membership in Canis lupus familiaris or the concept that comports with membership in Canis lupus? Notice that some background intention to use dog to refer to the same sort of animal as the ones around us wont really help here. First, we may not in fact have that intention, and yet presumably we would still have a concept associated with dog. Second, the intention itself does not eliminate Canis lupus from contention, for there is no good reason to suppose that all the grey wolves considered together are not the same sort of animal . In general there are too many sorts of animals for underdetermination to be reduced in this way. Moreover, even on the assumption that these heuristics enable us in our environment to use dog to pick out members of Canis lupus familiaris, the heuristics themselves will be poor candidates to figure in the analysis of the concept dog. For it is not a priori and necessary that
If x is an animal which has one of the characteristic doggish looks, smells, coat textures, etc., and is the offspring of that ilk then x is a dog.
Other animals can be made to look like dogs. A coiffed squirrel can be made to look like a chihuahua, and so can the squirrels immediate forebears. The lesson is that the heuristics are simply indicators of when something around here is a dog. It is a bad verificationist error to mistake such heuristics for the application conditions of the relevant concept. A similar argument goes through on the assumption that these heuristics in fact enable us, at least in our typical environment, to pick out the kind Canis lupus. Animals that are not grey wolves can be made to look, smell and feel like grey wolves. So can their immediate forebears. Satisfying the criteria for applying a concept or better, for applying a mental representation or a term is not satisfying the application conditions (in the philosophers sense) for the concept associated with the representation or term. Indeed, once the criteria/conditions distinction is made, why should we think that application conditions rather than our criteria are what is articulated by our judgments about cases? Once that distinction is made clear, the method of appealing to our judgments as to whether we should apply or withhold a term in a variety of imaginary cases is obviously a way of manifesting our criteria or ways of telling whether the term applies. It is not obviously a way of manifesting
our implicit grasp of the application conditions of terms. Indeed, the empirical explanation of our use of terms, by way of one or another psychological theory of concepts, seems to drive out the supposed explanation by way of implicit grasp of application conditions. But the supposed explanation was doing the crucial work of justifying the method of cases as the royal road to the analysis of our concepts.
A Temptation to be Resisted
Some may still be tempted by the thought that if our heuristics for applying the term dog are in fact just exhausted by the down-and-dirty criteria
Is it an animal? Does it have one of the characteristic familiar doggish looks, smells, coat textures, etc.? Is it the offspring of an animal with one of those characteristic looks, smells, coat textures, etc?
then it just follows that the concept we express by dog is the concept of an animal which has one of the characteristic domestic doggish looks, smells, coat textures, etc., and which is the offspring of an animal with one such perceptual profile. Accordingly, they will say that given how we use the word dog it is a priori and necessary that
x is a dog iff it is an animal with one of the characteristic patterns of domestic doggish looks, smells, coat textures, etc., and it is the offspring of that ilk.
Here we no longer have underdetermination, understood in terms of there being multiple kinds Canis lupus familiairis, Canis lupus that are equally good candidates to fall under the concept dog. Instead we have lack of specificity in the concept expressed by dog; it turns out to be a concept under which the more general kind Canis lupus falls. The problem then is to explain how it could have been rational on our part to accept the scientific discovery that
All dogs are members of the kind Canis lupus familiaris.
At the very least, accepting this will involve a conceptual change, by changing from the topic we had in mind when we use dog to a much more specific topic. That is odd, for it suggests that we could reasonably reject such scientific discoveries out of hand because they confuse a kind with one of its sub-kinds in precisely the same way as the following claim does:
Concepts, Analysis, Generics and the Canberra Plan / 137 All human beings are men.
Obviously, such a rejection of the scientific discovery that all dogs are members of the kind Canis lupus familiaris is wildly out of place, therefore the temptation that prompts it is not to be followed. What then is wrong with the temptation to make the conditions of application of a concept exactly as unspecific as the heuristics which govern its use? Once again, it confuses useful heuristics or criteria for the employment of a term with a priori and necessary conditions for the terms correct application. The usefulness of a heuristic is invariably situation-dependent. In the suburbs of New York, the heuristics are a fair guide to the presence of a member of Canis lupus familiaris. In the Arctic hinterlands, where there are lots of wolves, they are not. The conditions of application of the concept dog are not in this way situation-dependent.
Are There Any Concepts in the Sense of Things to be Analyzed?

Philosophical reflection on the psychological literature yields the following straightforward challenge:
Premise 1: The concept that subjects express in their use of a term T is not fixed by the heuristics that subjects employ in applying T. Premise 2: All that subjects could have introspective or armchair access to, even in the best case of articulating what they know about Ts by means of an large inventory of real and imagined possible cases in which T is applied and withheld are (i) the various deliverances of their heuristics in the real and imagined possible cases, and (ii) by way of inferences to the best explanation of such deliverances, the details of their own heuristics. Conclusion: The method of trying to extract a conceptual analysis, a statement of the necessary and sufficient conditions of application of a concept, from armchair intuitions about cases typically will not work. For if there is a concept that subjects express in their use of a term T, the relevant heuristics will typically underdetermine which concept that is.
This challenge has only been illustrated, not vindicated, by the toy example of our heuristics for dog. For it is open to someone to maintain that when we really take all the heuristics or criteria that we employ in applying T into account then the conditions of application of the concept expressed by T are determined or fixed (or fixed modulo the further facts about which concepts, properties or extensions are privileged attractors). That response will be taken up in detail later in the paper, in the context of discussing the general program of philosophical analysis known as the Canberra plan. Obviously, the actual facts about the kind of heuristics or criteria we really
are employing are highly relevant to the claim that the totality of the heuristics or criteria we use determines the conditions of application of the relevant term or concept. The overall challenge from psychology can only be met if it can be met at all by looking at what psychologists have discovered about these matters. As we noted above, recent psychological results are now converging on the suggestion that many of the heuristics or criteria we use in applying terms involve beliefs that are best expressed in generic form. If this is so then we believe that it can be shown that it is very unlikely that the totality of our heuristics or criteria fix the application conditions of the relevant terms.
Isnt Reference Fixing Analysis Still Viable?

Once upon a time, there was a privileged way of telling whether something was Neptune; all we had to determine was whether it was the planet which caused perturbations in the orbit of Uranus. And, arguably, this way of telling was semantically tied to the name Neptune in the sense that this name was in fact introduced to denote whatever planet was causing the already observed perturbations in Uranus. According to the well-known tale, Alexis Bouvard observed odd perturbations in the orbit of Uranus and hypothesized that these were caused by the gravitational influence of an unknown planet, which was then dubbed Neptune. Then, in 1846, Johann Galle discovered the planet in question. Some would then say that it is a priori that (1) Neptune is the cause of perturbations in Uranus and hence that it is a priori and necessary that (2) Neptune is the actual cause of perturbations in Uranus Finally, someone might say, we now have just what we wanted from an analysis. We have identified a privileged way of telling whether something is Neptune, a way of telling that is semantically tied to the term Neptune, and one which can be used to guide us in determining the application conditions of the term across all possible situations. All we have to do is look to see whether the thing in the possible situation is identical with the actual cause of perturbations in Uranus. The idea can be generalized if we indulge ourselves in various as-if stories about the introduction of our terms. Suppose the term water was introduced to denote the same stuff as the potable liquid found in rivers, lakes and streams, and that falls from the sky when it rains. Then, according to the line of thought in question, it is a priori and necessary that (3) Water is the stuff that is the actual potable liquid found in rivers, lakes and streams, and that falls from the sky when it rains.
Once again, we have something that can play the role of an analysis of the concept water, for if (3) is a priori and necessary then it specifies the application conditions of water across all possible situations. Why then is there any problem about analysis? Analysis is just the articulation and rigidification of the reference fixers of our terms! However, as our brief survey of the psychological facts suggests, ordinary users of terms do not themselves privilege such reference fixers among their ways of telling. Indeed, as in the case of water, they often apply terms on the basis of prototypical perceptual profiles, profiles which may resist any discursive characterization that could appear on the right hand side of an analysis, even a reference fixing analysis. Do they then have different concepts from those who explicitly treat (2) and (3) as a priori? The friend of reference-fixing analyses need not say this. He or she can say that, as Kripke and Putnam have shown, ordinary users of the terms Neptune and water respond to possible cases as if they did treat (2) and (3) as a priori and necessary. So although they have anticipated one important upshot of the psychological literature, namely that ordinary users of terms employ at most fallible heuristics or criteria in their use of those terms, Kripke and Putnam have nonetheless shown us the way to provide thoroughly modern conceptual analyses: namely, find the reference-fixer for the term in question and rigidify the reference-fixing condition. It is Frank Jackson, more than anyone else, who has done the most to advance this route to refurbishing the credentials of analysis. The main difficulty for the thought that reference fixers can provide analyses is that, although in some cases reference fixing descriptions may have played a semantic role precisely in serving, at a particular time, to fix the reference of a term, it is a notable fact that even in these very cases the relevant rigidified reference fixers need not in fact fix the subsequent application conditions of terms across all possible situations. That is, (2) and (3) and their various analogs are not a priori and necessary. To see that (2) and its ilk are not a priori imagine that after Johann Galles observation of Neptune, Alexis Bouvard subsequently noticed odd perturbations in the orbit of Neptune, though of a less dramatic sort than those he found in Uranus. Bouvard then hypothesizes that these too are the gravitational effects of still another planet, one that remains to be observed. The question arises as to how to name the planet that is doing this to Neptune, and it comes to be called Pluto. Galle then subsequently observes Pluto. If history had gone that way, then if (2) is a priori and necessary, the same would hold of (4) Pluto is the actual planet that is causing perturbations in Neptune. However, (4) is certainly not a priori and necessary, even in the scenario imagined. For our little story about Galles second success is compatible with what actually happened in the early 21st century, when shockingly, Pluto was demoted from the status of a planet, because it turned out to be more similar to large
asteroids than to planets. If (4) were indeed a priori and necessary that could not have happened. Instead the only option would to have been to conclude that Pluto never existed. Likewise, we could discover that Neptune is a huge, well-disguised, alien spaceship, so (2) is not a priori and necessary. The crucial lesson is this: an established pattern of usage can override original reference fixers. So long as we are not dealing with what Putnam called one-criterion terms, as, perhaps, in the artificial example of the name Julius introduced just to denote the inventor of the zipper, we can find examples where the term survives the falsity of its initial reference fixing description. Reference fixing descriptions for terms do not in general hold a priori of the items that the terms pick out. This is even more obvious in the case of so-called natural kind terms such as water. (For an extensive discussion of crucial, but often suppressed, complexities surrounding natural kind terms see Leslie (in press b).) We can discover that the water cycle is much more complicated than we supposed, so that the stuff that falls from the sky changes its chemical structure as it approaches the surface of the earth, with the consequence that there is no natural unity in the stuff that falls from the sky and the stuff in lakes and streams. This is not the discovery that water does not exist, as it would have to be if (3) is a priori. Again, an established pattern of usage can override almost any original reference fixer. There is an appealing response that Frank Jackson and David Chalmers offer to these kinds of counter-examples (Chalmers & Jackson, 2001; Jackson, 1994, 1998). They say that, in order for us to have these very intuitions about the possible cases in question, we must be using some other criterion for being Pluto, or Neptune or water. Otherwise, how can we identify the possible situation before us as one in which we have Pluto, though (4) is false, or as one in which we have Neptune, though (2) is false, or as one in which we have water, though (3) is false? That may seem like a compelling question, at least if we take seriously what Kripke (1980) called the vernoscope picture of our relation to possible worlds. On this picture, a possible world is presented to us neutrally, leaving open whether it contains Pluto, or Neptune, water or whatever target phenomenon or feature. It is then up to us to bring to bear our a priori and necessary criteria for telling whether it contains Pluto, or Neptune or water or whatever other target we have in mind. An articulation of our ultimate a priori and necessary criteria for telling will then be the proper analysis of the term for the target. So, say Jackson and Chalmers, counterexamples to analyses are themselves indications that we implicitly possess a priori and necessary conditions for telling, i.e. that we know the true analysis. However, the vernoscope picture, as Kripke (1980) emphasized, systematically misrepresents the nature of our thoughts about possibilities. We happen to know that certain things are possible, and this knowledge of possibilities already comes laden with subject matters, e.g. Pluto, or Neptune, or water. Possible
worlds neutrally described are not our basic epistemic starting points in arriving at knowledge of possibilities; they are simply useful devices for defining validity in modal languages. In particular, in urging the argument above, did we actually find ourselves in such a situation of surveying a possible world neutrally described so that we then had to bring to bear a priori and necessary criteria to determine whether that possible world contains Pluto, or Neptune, or water? No, for the purposes of the argument above, we simply found ourselves in this epistemic situation: the following individual claims are very plausible
Pluto could exist even if (4) were false. Neptune could exist even if (2) were false. Water could exist even if (3) were false.
Clearly we do not need to be relying upon even an implicit analysis of Pluto, Neptune and water to be in that, fairly minimal, epistemic situation. That should be clear because the possession of an implicit analysis is a very strong condition; an implicit analysis would decide every case. For the purposes of the argument above, we needed only to decide three cases. Obviously, something much more fragmentary and thus much less committal than an implicitly possessed analysis could suffice for that. After all, our grounds for the three claims above were simply that established usage can simply override the reference fixers of old. In saying that we need simply be relying on some reasonable criteria for counting as Neptune, Pluto and water, criteria that do not themselves amount to an analysis, and which may not themselves figure in any plausible analysis of the application conditions of the terms Neptune, Pluto and water. Laura Schroeter has offered a very useful way of understanding the general approach of Jackson and Chalmers when it comes to the analysis of our terms. She writes
However exactly we choose to cash it out, it seems that Jacksons and Chalmerss account is ultimately grounded in the conceptual [or term-using] dispositions the subject would form after ideal reflection on hypothetical cases. The rigidified definite descriptions Chalmers and Jackson offer as an analysis of water are correct just in case they accurately summarize the ideal conceptual dispositions the subject would converge on, irrespective of which world she considers as actual (2004, p. 449).
The thought is that something is guiding our use of terms in thinking about merely possible situations. Imagine a huge list of the positive and negative verdicts about whether the term under consideration applies in a fully representative range of possible cases verdicts we would give under ideal conditions; that is, conditions such as no distractions, plenty of computational ability, no straightforward irrationality. If we could codify the pattern found in such a list, we would have
an analysis of the concept expressed by the term; equivalently, we would have an account of the application conditions of the term. If not, we will not have an analysis, but there is no more to know about the application conditions of the term in question than the list of verdicts. What else, Jackson and Chalmers might ask, could determine the application conditions of the term? Having worked through the toy example of dogs, we hope that the reader can now recognize this kind of idea. Once again, it amounts to the conviction that what we rely upon in making judgments about cases, namely our criteria, themselves determine the application conditions of terms. What else could determine the application of our terms? Recall the concepts as terms-in-use view. Mutual correction while participating in a convention for using a term selects for a variety of effective heuristics or criteria that then guide individuals in their use of that term. To know how to use a term is just to have some such effective criterion, and this will often fall well short of even a tacit understanding of the conditions of application of the term. Accordingly, on the concepts as terms-in-use view, there is no reason to suppose that speakers can render explicit the conditions of application of their terms simply by reflecting on how they would use them in a variety of circumstances. For such actual and counterfactual usages reflect only the speakers criteria, which may fall short of any significant grasp of application conditions. (Remember how Sean Connery was once more central to the criteria for counting as a bachelor than were the properties of being male and being unmarried.) Thus philosophical analysis, as traditionally conceived, can at best produce fragmentary and inconclusive results, even under conditions of ideal reflection. One way to press this point home is to characterize just what knowledge is required to fix the application conditions of our terms, and how far that knowledge transcends anything that could be fixed even by full knowledge of the criteria we actually use in applying our terms. As the examples of Neptune, Pluto and water themselves show, scientific discoveries can entail surprising facts about the application conditions of our terms. Neptune may not be the cause of the very perturbations which were cited in the fixing of the reference of the term Neptune; Pluto need not refer to a planet, and water may not apply to rain. In fact, as we shall now see, in order to have the knowledge required to fix the application conditions of our terms, we would either have to know (i) the true total theory of the world (encompassing all the potential sources of surprises about the application conditions of our terms) or (ii) conditionals which specify for each epistemically possible total theory of the world just what the application conditions of our terms would be relative to that total theory. Surely both (i) and (ii) massively transcend what can be extracted from our criteria, even under ideal conditions, at least as those conditions have so far been defined. To see just why the Jackson and Chalmers view that our dispositions to classify possible cases are rich enough to determine the application conditions of our terms leads to one or another of these positions, consider the
following example, which simply recalls the traditional Quinean worries about the inter-animation of application conditions and empirical theory. Suppose that a subject has never understood or even heard of special relativity, and suppose that its important claims do in fact figure in the total true theory of the world. When asked to clearheadedly consider a world in which there is no relevant two-place temporal relation between events that can count as a candidate to be simultaneity, our subject will wrongly conclude that there is no simultaneity in that world, because he has not had the benefit of the relevant scientific understanding of the widespread explanatory advantages associated with special relativity the very ones that justify thinking of simultaneity as frame-dependent. The general lesson is this; if actual science as we know it can reveal surprising facts about the extensions of our terms, for example that simultaneity applies in a world in which there is no relevant two place temporal relation between events that can count as simultaneity, then absent the knowledge of the total true science of our world, we will be liable to make systematic errors about the extensions of our terms when presented with possible cases. Of course, our subject can be presented with an explanation of special relativity, and then asked What is the extension of our term simultaneity under those circumstances? Then he may get it right. But suppose he is given a description of a possible case in a possible world, without knowing just how the ultimately correct scientific theory of our world reveals, as it likely will reveal, surprising facts about the extensions of our terms. Suppose, as is likely, that this forever unknown theory would be even more revisionary and surprising than special relativity. As the history of sciences impact on our use of terms already reveals, if we are in ignorance of the true theory of the world, then we should treat our reactions to vernoscope presentations of possible worlds presentations in which it is left open whether the target things in question are in them as hypotheses, not as data. They could well be overturned by subsequent scientific developments. On the Jackson and Chalmers view that our dispositions to classify possible cases are rich enough to determine the application conditions of our terms, there are two ways of assimilating the familiar point about the unknown surprises that the true theory of the world might deliver. On the first, in order to be sure that our ideal reactions to possible cases get the extensions of our terms right, we need to be in ideal conditions that include knowledge of the surprising implications of the extensions of our terms due to the ultimately correct theory of our world. Only then can we properly evaluate the extensions of our terms in the possible situations presented to us. Let us suppose that for each term T there is such a thing as the list of those reactions we would have to the full range of possible cases, under the following conditions: we have no distractions, plenty of extra computational ability, no straightforward irrationality and we know the surprising implications of the extensions of our terms due to the ultimately correct theory of our world. What are we to make of that list?
First, it looks unlikely that for many terms there will be any useful finite codification of the infinite number of reactions on the relevant list. So, we may have here a final account of what would make an analysis true, without there ever being any analysis, i.e. any actual statement of an analysis, which is true. Second, given the nature of the idealization required, it would be utterly forced to say that for each term we even tacitly know its list, let alone the codification of the list if there is one. Third, and most importantly, we have arrived once more at a central theme of this paper; the real problem with concepts, and with the conceptual a priori and with analyses, is that they have no demonstrated use. Certainly, given the idealization in question, it is not our knowledge of them (explicit or tacit) which guides our use of terms in ordinary life, where we manifestly do not know the surprising implications of the extensions of our terms due to the ultimately correct theory of our world. It is rather our heuristics, our criteria, which guide our use of terms. Whereas the psychological approach to concepts as clusters of heuristics has an explanatory relation to a subject-matter, the philosophical treatment of concepts that goes by way of such highly idealized outputs does not. The same conclusion emerges on the alternative way of assimilating the point about the surprises concerning the extensions of our terms that the true theory of our world might force. Imagine the following response to the worries just raised.
We agree that the comprehensive scientific theories we accept can involve surprising consequences for the use of our terms. What is true for users of a term T is that under ideal conditions of the more restrictive sort, involving only no distractions, plenty of extra computational ability, and no straightforward irrationality, the users will be disposed to give a verdict for each epistemically possible true total theory of the actual world whether, under that theory, T applies in this or that possible case.
This massively multiplies the computational complexity required to generate the relevant list, and so intensifies the worry that there will be no codification of the list. But it also means that the items on the list will themselves be conditional in form; for example, when it comes to the concept of simultaneity, the relevant conditional will look like
If special relativity holds up in the true total theory of our world then. . .
So now the obvious problem is this. Thus articulated, our conceptual knowledge leaves us in the following situation: we should make no judgment about any cases with respect to whether they fall in the extension of the term in question or not until we know the total true theory of our world. Hence, on this alternative way of going, our main point holds a fortiori; if this is the content of our conceptual knowledge, it not only has no demonstrated use, its lack of usefulness is itself demonstrable.
Jackson and Chalmers might insist that there is indeed one remaining use for talk of a priori or conceptual knowledge here, however idealizing it may be. For they each identify physicalist reductionism with a very distinctive thesis, namely with so-called type A physicalism. This is the thesis that the deliverances of ideal reflection about all relevant cases will in principle support rich enough intermediate a priori premises to provide a derivation of the truths about all that there is, from the truths about what there fundamentally is at the physical level. Chalmers denies this thesis and Jackson now defends it, but they both understand physicalist reductionism in this same way. The fact that the deliverances of ideal reflection may not be codifiable is not crucial here, for that only means that we cannot carry out the derivation. As against this highly theoretical use of the conceptually based a priori, we have only the following plausibility argument. Consider what ideal reflection, on this way of going, is. It is no more than having no distractions, plenty of extra computational ability, and being subject to no straightforward irrationality. Consider the input that ideal reflection takes in, namely the dispositions of use associated with the heuristics and criteria we employ in deciding how to use terms. Now consider the massive output needed to meet the standards of type A physicalism: for each term in our language which is not part of the logical or basic physical vocabulary we need a conditional which specifies, for each epistemically possible total theory of our world, just what the possible world application conditions of the term would be relative to that epistemically possible total theory. Is it really credible that ideal reflection using the input yields anything like this output? Is this not a massive, and extremely adventurous, empirical psychological speculation? In any case, it was an important line of inquiry to ask whether the externalist semantics of Kripke and Putnam reintroduced something like analysis by the back door. We take ourselves to have argued for a negative judgment on this more restricted point; reference fixing descriptions do not (except in very artificial cases, where there is one, simple and theory-free, criterion) hold a priori, and in arguing for that conclusion we ourselves need not be evidencing an implicit grasp of an analysis.
Total Theory as a Criterion

A theory of Ts, that is, a theory of the things to which the term T applies, can be readily re-interpreted as a heuristic for applying the term T. Simply proceed as follows: first put a variable X for every occurrence of T in the statement of the theory. You then have a vast open sentence. You now look around the world for the things which satisfy the open sentence. Your heuristic is: apply T to those things and those things only. Furthermore, if we abstract away from the actual psychological limitations which make it reasonable to use only smallish subparts of what one believes about
Ts when determining whether we have a T before us, then ones total theory of Ts at some given moment should be the theory from which one derives ones T-sensitive heuristic. For suppose you omit something you do in fact believe about Ts; then you are not abiding by the requirement of total evidence, the requirement that you bring all relevant beliefs to bear on the question of whether you have a T before you. Of course, your total theory of Ts is likely to evolve over time, and so your ideal heuristic will also change over time. What changes here are not the application conditions of the term T, but your ways of telling whether to apply it. Imagine a child who starts out with the following theory of dogs
If x is an animal which has one of the characteristic doggish looks, smells, coat textures, etc., and is the offspring of that ilk, then x is a dog.
and then becomes a biology major, adding to her earlier theory of dogs the thesis that
Dogs are members of the subspecies Canis lupus familiaris, distinct from the members of the wider species Canis lupus in the following ways: A, B, C.
Obviously, she now has better criteria for telling when something is a dog; she is less likely to confuse dogs and grey wolves. Has her concept dog changed as a result? Has she moved from a concept with the application conditions determined by the smaller childhood theory to a concept whose application conditions are determined by the larger adult theory? As we noted earlier, there is an everyday use of the term concept, on which ones concept of Ks is just some contextually relevant part of ones total theory of Ks. In this sense, in certain contexts, it will be right to say that her concept has changed. However, this is manifestly not the philosophers use of concept, which ties concepts essentially to conditions of application. Once we have that use in mind, the thing to say is that she has just learnt more about the things the concept dog applies to; the extension of the term dog could have remained the same for her, even as she acquires more and more knowledge about the nature of the items in that extension. Suppose instead that there were, earlier and later, two concepts in play; not in the trivial everyday sense of there being two conceptions in play, but in the philosophical sense of two extension-determiners in play. Then the biology major would have made a conceptual error in simply conjoining her later knowledge with her earlier knowledge and treating them as knowledge concerning the same kind of animal. To assume that our total theory determines the application conditions of its concepts, makes what we call learning typically count as conceptual confusion. This is obviously not an objection to the total theory-theory of concepts, so long as we understand that as merely an empirical account of theory-based criteria we use in applying concepts. But if we instead treat the total theory-theory
of concepts as an account of the application conditions of concepts, then the urgent objection will be How is significant learning specifically learning that narrows or broadens our conception of the range of entities a term is to be applied to possible? Moreover, if the total theory-theory account of application conditions were true, someone could not learn a theory in the obvious way; that is, by having it explained to them using words he or she already understands. For the total theory-theory account of application conditions implies that in order to know what the words that occur in the total theory mean, one must already know the total theory. The total theory-theory account of the application conditions of our terms faces a second objection: it wrongly represents cases of genuine disagreement as cases in which the disputants are talking past each other. Suppose, as David Lewis (1970, 1994) argues, that our total folk theory of mental states represents them as inner sates that occupy certain characteristic causal roles, in particular with respect to other such inner states, sensory input and behavioral output. On the face of it, the folk theory is at odds with behaviorism, which represented being in a mental state merely as exhibiting certain patterned connections between sensory input and behavioral output. The folk theory is even more obviously at odds with epiphenomenalism, which denies that mental states are causes, and with the parallelism of Leibnitz, which keeps the mental and the physical realms causally isolated, but running in harmony with each other, so that they behave as if they were interacting. But now suppose that we adopt the total theory-theory view of the application conditions of our terms, and accordingly consider the four theories just discussed folk theory, behaviorism, epiphenomenalism, parallelism as different total theories that thereby establish different application conditions for mental state terms such as belief, desire, pain and so on. Now the manifest disagreement about the nature of the mental among the defenders of such theories is lost. There are simply four different term defining theories, and their proponents are simply talking past each other. The familiar response to both the objection from learning and the objection from disagreement involves a restriction of the theory of Ts to a sub-theory that involves just those criteria that are analytic as opposed to empirical. (Analytic in this context can be taken to mean both a priori and necessary.) The thought is that, quite generally, part of the theory ordinary speakers use in applying a term T will involve a priori and necessary criteria, where this entails, crucially, that these criteria are immune to empirical revision, and so should persist throughout any amount of learning. Likewise, disputants with differing total theories of a domain can still genuinely be in dispute over a common subject matter. For that subject matter can be defined by a common shared core of a priori and necessary criteria. Even setting aside general philosophical skepticism about the analytic/synthetic distinction, the response seems inadequate. It faces a dilemma: are the analytic criteria for a term T themselves sufficient to fix the application
conditions of T, or not? The supposition that they typically are appears to have been empirically embarrassed; that is, the accumulated evidence against what Rosch called the classical view is precisely evidence to the effect that speakers in using terms do not know, and so do not employ, analytic criteria a priori necessary and sufficient conditions that fix the application conditions of those terms. So at best the analytic criteria, in so far as speakers possess them and associate them with terms, will somewhat constrain, but not fix, the application conditions of terms. A number of difficulties now arise on this horn of the dilemma. First, in the case of many of our terms, there is reason to doubt that we actually possess any analytic criteria. Consider the biology major who comes to hold the consolidated theory about dogs:
If x is an animal which has one of the characteristic doggish looks, smells, coat textures, etc. and is the offspring of that ilk then x is a dog. Dogs are members of the subspecies Canis lupus familiaris, distinct from the members of the wider species Canis lupus in the following ways: A, B, C.
None of this is analytic, i.e. utterly insulated from being overturned by surprising empirical discoveries. As Hilary Putnam made clear long ago, we could discover that the things we call dogs are not members of the subspecies Canis lupus familiaris, and indeed are not animals, but instead are cleverly disguised bionic robots collecting information about our domestic habits. Moreover, recall the coiffed squirrels that look like Chihuahuas. What actually is analytic for speakers when it comes to the term dog? Second, consider the cases where we might suppose that some criterion possessed by speakers is at least a pretty good candidate to be analytic. What are we then to make of subsequent learning of empirical matters of fact that intuitively work to restrict the extension of the term? How is that really possible? Suppose for example that for some user of the term arthritis the criterion arthritis is a disorder is an analytic criterion, and so one that constrains, but does not fully specify, the application conditions of his term arthritis. Suppose he now learns the synthetic medical fact that arthritis is inflammation of the joints. How exactly can he have learned this, given the present theory of the application conditions of terms? Prior to the supposed learning, he associated one analytic criterion with arthritis, namely its being a disorder. So, on the hypothesis that only ones analytic criteria constrain the application conditions of ones terms, he is using a term which is no more specific in its application conditions than the term disorder or the phrase some disorder or other. On the hypothesis that only the analytic criteria one associates with a term can constrain its application conditions, the most he can learn is that some disorder or other is inflammation of the joints. But clearly he is in a position to learn more than this, namely that arthritis is inflammation of the joints.
Have we crucially under-estimated what is analytic for our subject? Maybe it is something like arthritis is the disorder that my speech community refers to by the term arthritis. Now perhaps we can explain, consistent with the hypothesis that only ones analytic criteria constrain the application conditions of ones terms, how our subject can learn that specifically arthritis, and not just some disorder or other, is inflammation of the joints. One familiar worry for this line of response is that many speakers who use terms do not have views about how their speech community uses those terms. That is not very telling, since a defender of the view that we have analytic criteria, and that only they constrain the application conditions, could say that rational speakers would come to recognize the truth that arthritis is the disorder that my speech community refers to by the term arthritis , just by understanding that truth. A more telling objection is this: upon reflection it is obviously a synthetic or empirical matter of fact that ones speech community refers to arthritis by arthritis. For just how far ones speech community extends is an empirical matter, and, crucially, it is an empirical matter just how good or bad the members of ones speech community are at spelling and pronunciation. It is manifestly not a priori that many of them are not severely impaired when it comes to spelling and speech production. They may, many or most of them, systematically fail to produce the right graphemes and phonemes, and so not refer to arthritis by arthritis. These points aside, we should say that our target is not the idea that we have some analytic or a priori criteria for applying some concepts. Our point is that the empirical literature makes this look like a special kind of case. This is especially so if the generic encoding hypothesis is true. Our earliest forms of generalization are generic in nature, so the little theories we use as criteria for applying terms are from the beginning generic in form. Even as adults, our thought is rife with generics, and we can relatively easily be gotten to evaluate universal generalizations as generics (Leslie et al., 2011). Thus it is very likely that our general criteria for the use of the concepts are generic in form. Here is the crucial observation in this context: the only generics that are plausibly taken to be a priori are those for which the corresponding universal generalization is a priori. These are the sorts of cases we have in mind:
Yellows are brighter than browns. Primes are divisible by themselves. Irrational numbers are not expressible in the form m/n where m and n are rational. Conjunctions entail their separate conjuncts.
These are the special cases; and our point is that even in these special cases it remains a question whether the application conditions of the constituent concepts
can be defined by way of the backing universal generalizations. (For what it is worth, our best guess is yes, for some simple mathematical and logical concepts, like prime and conjunction, and no, for almost everything else.) For all these reasons, we should reject any simple identification of the application conditions of the term T with the criteria even the analytic criteria derived from our best theory of Ts. The theory-theory of criteria is tenable; it is the favored psychological theory of concepts, and as we have seen, it meshes with the generic encoding hypothesis better than pure prototype theory; but the theory-theory of application conditions, at least in the simple form just considered, is not tenable. The question remains as to whether there is some sophisticated tweaking of the theory-theory of application conditions that is tenable. Hence the relevance of the so-called Canberra plan, which we take to be just such a theory.
Lewis on Defining Terms

John Hawthorne and Huw Price coined the term the Canberra plan in 1996 to denote what was then a growing tendency among Australian philosophers, several of them then located in Canberra, to apply David Lewiss method of defining terms quite generally as a part of a revival of the program of philosophical analysis. Lewis own analyses of the concepts of mind (1970), value (1989) and color (1997) are taken to be paradigms of the Canberra plan. Michael Tooley (1987) provided an analysis of causation which can be easily assimilated to the plan. Johnstons (1992) discussion of color is sometimes cited as an example of the plan, since it begins with central beliefs about color that we would find hard to give up, but unlike Lewiss (1997) somewhat similar discussion of color, the general apparatus of the plan is in fact absent. David Braddon-Mitchell and Robert Nola have recently edited an important collection of papers, Conceptual Analysis & Philosophical Naturalism (2009), a collection which is on the whole extremely sympathetic to the plan. David Lewis is rightly viewed as the father of the Canberra plan, and his Psychophysical and Theoretical Identifications (1972) and How to Define Theoretical Terms (1970) are taken to be its founding documents, with an important addition arising from Lewiss emphasis in New Work for a Theory of Universals (1983) and Putnams Paradox (1984) on natural properties as default referents. One way to see Lewiss proposal, and the Canberra plan more generally, is as a very sophisticated revision one developed over fifteen years of the general idea that the criteria derived from our theory of Ts are indeed the application conditions of the term T. Lewis makes three crucial revisions to the simple version of this idea, the version we have already found to be unworkable. The Canberra plan is the idea that, given these revisions and some associated qualifications, the criteria derived from our common theory of Ts are the application conditions of the term T.
Precisely what revisions to the simple criteria/conditions identification are required, according to Lewis? First, instead of focusing on analytic criteria, which, even if they exist, are in fact very sparse on the ground and obviously insufficient to determine the application conditions of terms, Lewis resorts to what in various places he calls platitudes. Thus in Psychophysical and Theoretical Identifications he writes
Collect all the platitudes that you can think of regarding the causal relations of mental states, sensory stimuli, and motor responses . . . Add also all the platitudes to the effect that one mental state falls under anothertoothache is a kind of pain, and the like. Perhaps there are platitudes of other forms as well. Include only platitudes which are common knowledge among us everyone knows them, everyone knows that everyone else knows then, and so on. (1972, p. 256)
Notice that the items of common knowledge considered individually need not be a priori; they can be highly informative. As an example, Lewis (1997) has the platitudes concerning red include such things as that British postboxes are red.8 It is clear also that Lewis allows that platitudes can be false, for he immediately adds
Form the conjunction of these platitudes; or better, form a cluster of thema disjunction of all conjunctions of most of them. (That way it will not matter if a few are wrong.) This is the postulate of our term-introducing theory (1972, p. 256).
The platitudes embedding a term T used by our speech community make up the knowledge or supposed knowledge common to the members of our speech community. They are the relevant things that are known in the sense that we take ourselves to know them, and we take them to be widely known, at least implicitly, throughout our speech community. Since these platitudes are common (supposed) knowledge, a certain kind of informed and reflective thinker can access them from the armchair, i.e., without any further empirical investigation.9 Though some have tried to place further constraints on the Canberra plan, most notably that the platitudes be genuinely a priori or immune from empirical revision, Lewiss own more liberal account is more realistic, given the actual psychological literature. For each of our terms, we have a rich set of relevant things we take to be known; but outside of logic and mathematics there is little that looks a priori. Second, as the last quotation makes clear, in Psychophysical and Theoretical identifications Lewis adopts a cluster theory of application conditions. Instead of requiring that all of the platitudes taken together determine the application conditions of our terms, he assigns that work to a disjunction of conjunctions, each of which contains collections of most of the platitudes. As he notes,
this has the attractive consequence that the term T can have a reference even when many (though not most) of the platitudes involving it turn out to be false. Third, in New Work for a Theory of Universals and Putnams Paradox Lewis crucially departs from the idea that the reference of our terms is fixed solely by our total set of platitudes involving them. He recognizes that such an account, especially when it takes an appropriately clustered form, may well underdetermine the references of our terms in recognizable ways. Moreover, as Putnam (1980) pointed out, there will be intuitively wrong interpretations of T that are nonetheless compatible with making the T-involving platitudes true. Thus Lewis is led to endorse an external constraint on the reference of terms one that turns on a notion that he finds indispensible for ontology, namely the notion of one property or class of individuals being more natural than another property or class of individuals. A property or class is perfectly natural when it is suited to figure in a fundamental ontology; when things share such a property they exhibit a respect of perfect similarity. The more natural a property or class, the more the sharing of a property, or membership in the class, makes for genuine similarity. For example, the property of being green makes for a genuine similarity among the things that have it, and so is more natural than the property of being grue, i.e. the property of being green and observed before the year 2000 or blue and observed after the year 2000, which does not make for a genuine similarity among the things that have it. Moreover, the class of electrons is more natural than the class of particles; the electrons collectively exhibit a higher degree of genuine similarity than does the wider class of particles. Lewiss external constraint on reference fixing may now be put as follows; the more natural a property or class, the more it works as a default referenceattractor; that is, where there are two interpretations of T that do as well in making our T-involving platitudes true, or mostly true, we should prefer that interpretation which assigns T the more natural property or class. Notice that this way of putting it ranks the satisfaction of most of our T- involving platitudes lexically above the appeal to naturalness. Alternatively, but still in the spirit of Lewis, one could adopt a principle which says that the preferred assignment of an extension to T is one that optimizes both the making true of platitudes and the degree of naturalness of the extension of T. Depending on which way one goes, there will be a difference in what will count as the best analysis of a term or concept. For example, it is now a platitude, among philosophers at least, that justified true belief must somehow be de-Gettierized in order to count as knowledge. If knowledge is a conjunctive property, then the identification of it with the tripartite conjunctive property true and justified and believed will hew to the naturalness constraint better than the identification of it with the four part conjunctive property true and justified and believed and de-Gettierized. Of course, the second identification will make one more platitude true, namely the very platitude that amounts to the standard philosophical intuition in Gettier cases. On the lexical ranking version of the naturalness constraint, this latter fact means that the second,
and philosophically favored, interpretation wins the day. On the other hand, the optimizing form of the naturalness constraint would treat this as a trade-off situation, and might deliver the result that our term knowledge can equally well be associated with either the three-part or the four-part conjunction. (Hereafter, the difference between the lexical and the optimizing forms of the naturalness constraint will not make much difference.)
General Empirical Problems with the Plan

By the Canberra plan we here specifically mean the idea that Lewis has thus outlined a quite general method of analysis. (Variants on this straightforward characterization will be considered as we go.) Lewis himself seemed more cautious than the planners so defined; after all, his own counterfactual analysis of causation does not go by way of the Ramsey/Lewis method. In part, this was because Lewis (1973, 2000) believed that there are two very different forms of causation, causation by biff, roughly, by the transference of energy, and causation by absence, which are not in fact platitudinously identified as two forms of causation. (As it turns out, some friends of the Canberra plan such as David Liebesman (2011) have chided Lewis on just this point.) If we follow the planners in viewing Lewiss account as a quite general method of analysis, we then have an account which (modulo considerations of naturalness) tightly connects the application conditions of a term to a certain privileged criterion that users of the term implicitly possess, namely the criterion which is determined by the disjunction of conjunctions of most of the platitudes involving the term. For Lewis, and for the planners more generally, implicitly possess now has a definite empirical content. Many or most of the users of the term will recognize the relevant platitudes as such upon appropriate reflection. One immediate empirical problem with any specific application of the plan is deciding whether there are in fact enough platitudes on offer in order to fix the application conditions of the term in question. Being platitudinous within a speech community is not like being a priori or being analytic. Those latter properties are at least according to theories which give them any significant role properties a member of the speech community can in principle recognize claims to have simply by considering the claims themselves. In contrast, being platitudinous within a speech community is a substantial psycho-social property; I am only in a position to treat some claim as a platitude in this sense, and so incorporate it into a Lewis-style analysis, if I both take myself to know it and take it to be widely known, even if only implicitly, throughout my speech community. This latter condition is on its face an empirical belief about the claims general uptake in my speech community, and there are reasons to think we have a tendency to overestimate how many of our deep convictions are obvious to others. The first worry then is this: for any given term there do seem to be more platitudes connected with it than there are genuine a priori or analytic claims
involving the term, so that a platitude analysis initially seems more promising as an account of what determines application conditions; however, it turns out to be an empirical question just which claims are platitudes, and there are reasons to believe we may often be mistaken about such questions. So we need to get out of the armchair and do some empirical social psychology in order to perform our analysis by platitudes. So the analysis is not in any real sense a priori. That point could be given an extra twist: since we need to get out of the armchair anyway, why not instead investigate the worldly phenomenon associated with the term? Is that not invariably of more philosophical interest than the investigation of the adventitious social status of claims? Moreover, a claims being platitudinous within a speech community can be a temporary property of that claim; many of the platitudes in which a term figures can suddenly be given up. Suppose investigators seem to discover that Neptune is not a planet but a massive space ship, a space ship that does not orbit the Sun, but simply occupies positions consistent with that, during just the periods when the space ship is visible from the Earth. The discovery gets widely put about. Should we then have any confidence that there are claims about Neptune which still are platitudinous in the relevant sense? (Recall that it need not be known by the members of my speech community that Neptune denotes Neptune; they need not be readers or writers to be speakers, and they may badly mispronounce the names of planets.) Does Neptune cease to have a reference in such conditions? It seems not. Reference seems a more persistent feature than the adventitious social status of claims. A third problem arises from the residual truth in prototype theory, at least when it is restricted to what some philosophers have called partly recognitional terms. These are terms like red or dog, which we apply at least in part on the basis of paradigmatic or prototypical sensory and perceptual profiles, profiles whose linguistic characterization is a complex empirical psychological matter. We are not saying that an extensionally adequate linguistic characterization of the prototypical look of red or of the prototypical looks of dogs can never be given. The point is that even if we had such characterizations they would not be platitudinous in the defined sense. They would be complex bits of empirical psychological theorizing. The same is true of the very claim that the term dog is partly recognitional; philosophers can plausibly speculate that this is so, but it is not a platitude in the relevant sense. It is a bit of tentative, if plausible, psychological theorizing that goes beyond what is known in our speech community. However, if the term dog is in fact partly recognitional then it is fairly likely that its extension is at least partly determined by our dispositions to treat certain looks as prototypical dog-looks. The upshot will then be that a crucial determiner of the extension of our term dog will not be incorporated into an analysis via platitudes. That method of analysis will thus be susceptible to being systematically thrown off for partly recognitional terms like dog or red. (Notice that this point would hold a fortiori if we departed from Lewis,
as Michael Smith does, and restricted the platitudes to those claims that are ostensibly a priori. As observed in footnote 8, Smith agrees.) Some may say that this just shows that philosophers have no business analyzing partly recognitional terms. However, many terms of peculiarly philosophical interest may well turn out to be partly recognitional. Color terms are an obvious example. The attributive adjectives good and bad may be partly recognitional. You need to have visual recognitional capacities to identify a good tennis swing as such. You need to have visual recognitional capacities to identify any badlooking thing as such, including morally-badlooking things such as the torturing of a cat. And as Albert Michottes (1963) work reflects, there are a range of recognizable perceptual prototypes for causation, at least of the biff variety. Even six-month-old infants are sensitive to this profile, as Alan Leslie and Stephanie Keeble (1987) demonstrated.
Generics and the Plan

The Canberra plan, inspired by Lewiss account of the definition of terms, proposes a general method for the analysis of concepts: identify the term or terms which express the target concept, collect the platitudes involving the term that expresses the target concept, and then find the most natural satisfier of at least most of those platitudes. (If there is no satisfier of most of the platitudes, the term will then be empty.) Although the planners confine their attention to philosophically interesting concepts, there is, in fact, no such in-principle limitation built into the plan. If the plan works, then it should work quite generally. If it fails quite generally, then absent special pleading for the concepts of philosophical interest, we should expect it to fail in the philosophical cases as well. One kind of special pleading we regard as very unpersuasive, given the psychological evidence, is that in philosophy our platitudes are invariably universals and not generics. That would be remarkable if true. Events have causes, true justified belief is knowledge, lying is wrong, people choose those acts that seem to them, given their beliefs, to advance the satisfaction of their desires these generics are platitudes, but the various attempts to find universal truths backing each of them has not resulted in anything that is likewise platitudinous. Only if the plan works quite generally for concepts in whatever subject area we choose can it be seriously put forward as a method of conceptual analysis. Concept and conceptual analysis are topic neutral or quasi-formal notions that apply across all subject matters. The moral is that if the plan fails for many subject matters then whatever the plan is achieving in the philosophical cases it is not the analysis of concepts. The plan does fail quite generally when interpreted as a philosophical analysis of concepts in any interesting sense of concepts. First, lets set aside
an uninteresting sense. As noted earlier, a term of the form concept of Ts is often used to pick out contextually relevant part of someones or some groups conception of Ts, as in their concept of men is so old-school. A conception of Ts can be articulated or explicitly set out, but this is not analysis in any interesting sense, nor is it a distinctively philosophical enterprise, even when the term T is widely employed in philosophy. The articulation of conceptions is a matter for history, psychology and sociology; which is not to say that philosophers are to be prevented from pursuing such matters. Suppose instead that we understand concepts as terms-in-use. If we want to know the extension of a concept understood as a term-in-use, we should consult the lexicographer, who has studied the relevant empirical questions about the extensions of our terms. Still, the plan might be put forward as a general account of what makes it the case that any given term in use has the (possible worlds) extension that it does, namely that the extension is determined by the most natural satisfier of the platitudes which involve the term. But now we face problems. Terms in use combine compositionally to give more complex terms in use. However, many of our platitudes are generic in form, and the structure of generics systematically interferes with compositionality. Furthermore, the requirement of naturalness and the structure of platitudinous generics can work together to assign manifestly incorrect extensions to terms. The same points apply when we understand concepts by way of the substantial conception; that is, as items speakers grasp and which themselves determine (modulo considerations of naturalness) the extensions of our terms. Concepts, so understood, combine compositionally. Yet many of our platitudes are generic in form, and the structure of generics systematically interferes with compositionality. Moreover, the requirement of naturalness and the structure of platitudinous generics can work together to assign manifestly incorrect extensions to terms. Finally, when the plan relies on disjunctions of conjunctions of platitudes in order to insulate itself from the possibility of platitudes turning out to be false, it thereby breaks with another requirement on conceptual analysis, namely that good conceptual analyses should underwrite inferences like
Bonnie is a mare
to
Bonnie is a horse.
that is, the rare inferences where we do seem to possess (what looks like) an a priori justification for proceeding from the first belief to the second.
Generic Platitudes that Hold Only in the Minority of Cases

There can be generic platitudes that are true even though they hold of a minority of the very kind those generics concern. Lying is wrong holds true,
and may even be platitudinous, in a world in which people are quite honest and mostly lie righty, say to crazed axe-men at the door. Whenever we have generic platitudes that hold of a minority of cases, there is the likelihood that a platitude analysis will produce failures of compositionality. This is most easily seen in the case of sex-typed generics, but the point is quite general. The possible world extension of the concept female lion should be a specific compositional function of the possible world extension of the concept female and the possible world extension of the concept lion. In particular, the concept female lion is intersective; that is, its extension in a possible world is the intersection of the extension of the concept female in that world with the extension of the concept lion in that world. (Contrast attributively qualified concepts like small lion, which though compositional are not intersective in this sense.) The Canberra plan applied to these three concepts female, lion and female lion will generate the required result that the concept female lion is intersective only if the platitudes concerning female lion are a union of the platitudes concerning lion and the platitudes concerning female. Since being a platitude is a substantial psycho-social property there is no general guarantee that this is so. If it is not so, then there will be a failure to find the right kind of compositionality for the concept female lion. And the point will generalize for every intersective concept for which the platitudes do not align in the proper way. This kind of problem, if and when it arises, has always been taken to be a disabling objection to a philosophical theory of concepts. Recall that Fodor (1998) used this kind of objection against psychological theories; although we think it was misplaced there, the objection is disabling if it applies to a proposed theory of the application conditions of our concepts. We want concepts to combine compositionally, since the (possible worlds) extensions of our terms combine compositionally and concepts are, minimally, extensiondeterminers. The generic encoding hypothesis suggests one way in which the right alignment of platitudes might not be in place. From a very young age, we are interested in what psychologists call basic-level kinds, kinds such as dogs, lions, tigers, tables, and chairs. Basic-level kinds can be contrasted with superordinate kinds (e.g., mammals, furniture), and subordinate kinds (e.g., Bengal tigers, formal dining tables). The first count nouns we learn usually denote basic-level kinds, and even into adulthood, they are the first terms we supply to answer the question, what is this? (e.g., if shown a picture of Princeton Universitys mascot, and asked what it is, one would naturally reply a tiger, rather than a mammal or a Bengal tiger). Our knowledge concerning such kinds knowledge passed on to us in childhood in generic form is likewise encoded in as generic generalizations. When taken collectively for each such kind, this mostly generic knowledge forms perhaps the best candidate to be the platitudes applicable to the kind. After all, we not only know such things but we know that our parents and teachers know them. It would be very implausible to suggest that we only have a concept of a basic-level kind like the kind lion when and if we arrive at
exceptionless generalizations concerning it, and then come to know that these exceptionless generalizations are widely known in our speech community. For one thing, the exceptionless generalizations will not be widely known in our speech community. So our platitudes for such basic-level kinds are, many of them, generic in form. Moreover, in the case of animal kinds, some of these generic platitudes involve properties that are only had by one sex. These include generics that describe methods of reproduction, such as ducks lay eggs and horses give live birth, but also characteristic salient physical traits that happen to be had only by members of one sex, such as lions have manes, deer have antlers or peacocks have fabulous blue-green tails. As emphasized earlier, these generics are just not plausibly interpreted as contextually restricted universals respectively applying to female ducks and horses, or male lions, deer or peacocks. Nor, as noted above, are they encoded as sex-restricted generics such as male lions have manes (e.g., Cimpian, Gelman et al., 2010; Leslie, 2007, 2008, Khemlani et al., 2009). Rather, these are generics that are true despite there being a substantial number of exceptions to the generalization. These sex-typed generics are understood and accepted by preschool children; and plausibly are accepted as platitudes even at this age (Brandone et al., 2012). Surely, if anything is a platitude, lions have manes is a platitude. We now have the following concrete situation. The platitudes for the concept female lion are not a union of the platitudes for the concept female and the platitudes for the concept lion. Even though it is a platitude that lions have manes, it is not a platitude that female lions have manes. Some of our fellow speakers of the language do not regard this as true because they know that female lions do not have manes; others do not regard it as true because they do not know whether it is the male lions or the female lions that have manes. The platitudes associated with an intersective concept such as female lion thus are not given by the union of the platitudes associated with female and the platitudes associated with lion. Note that if platitudes took universal rather than generic form, this particular issue would not arise, for if the only platitudes associated with lion were ones that were satisfied by all lions, and similarly for female, then the platitudes associated with female lion could consist of the union of these platitudes. However, it is simply not empirically plausible that our platitudes here take universal rather than generic form. What if only male lions have manes was also a platitude? It is not clear how this could help. The point is simply that lions have manes is a platitude associated with lions but not with female lions, no matter what other platitudes may be also associated with the relevant concepts; therefore the platitudes associated with an intersective concept are not the union of the platitudes associated with both concepts. Further, it is much less plausible that only male lions have manes is a platitude, certainly it need not be. (In fact, a recent study found that some twenty percent of adult participants were ignorant of this fact, and a full third did not know that only male goats have horns and only female kangaroos have pouches
(Leslie et al. 2011).) We cant plausibly maintain that these adults who do not know that it is the male lions that have manes therefore do not have the concept lion, and hence do not have the concept female lion.10 What then can the friend of the Canberra plan say in order to defend the idea that it is a general method of conceptual analysis?11
Clustering to the Rescue?

There is one feature of the Canberra plan that might be thought to be helpful here. Lewis proposed that we cluster our platitudes; that is, we should take the extension-determiner for a term not to be the totality of platitudes which govern it, since this would make the term empty if even one platitude was false, but rather to be a disjunction of conjunctions, each disjunct of which contains most of the platitudes, with each platitude appearing in at least one disjunct. So if there are three platitudes, P1, P2 and P3, governing a term T we have in effect the claim that the possible worlds extension of T is fixed in the following way
A as it is in W satisfies T if and only if either
(i) the open sentence formed from P1 and P2 by substituting the variable X for T throughout is true of A as it is in W, or (ii) the open sentence formed from P1 and P3 by substituting the variable X for T throughout is true of A as it is in W, or (iii) the open sentence formed from P2 and P3 by substituting the variable X for T throughout is true of A as it is in W. Thus Lewis says that all his account requires is that the platitudes be mostly true of the target extension, which on the famous Lewisian account of adverbs (1975) amounts to the claim that all that is required is that most of the platitudes be true of the target extension. (We confess that we cannot make out anything helpful here in the alternative suggestion, sometimes heard, to the effect that each platitude should be true of most of the items in the extension. This fails for the same reason that accounts that assimilate generics to most-statements fail. For example, the platitude that lions have manes is not mostly true of lions; most lions lack manes, since only typically developing adult male lions have manes.) The relevance of Lewiss clustering proposal is this: clustering allows us to effectively shed or neutralize platitudes that turn out to be problematic, and thus avoid having them constrain the extensions of the relevant terms. Platitudes can be problematic if they are false, and this is what motivated the clustering account in the first place; for not every false platitude makes for an empty concept. Perhaps platitudes can also be problematic if they interfere with the compositional derivation of intersective concepts from their components. Only
one of the disjuncts itself a conjunction needs to hold for the disjunction to hold. So perhaps we can effectively neutralize the impact of the problematic platitude that lions have manes. A moments reflection will show that to resort to clustering here is to jump from the frying pan into the fire. First, we need not change our concept female lion by getting very interested in female lions, thereby coming to have more platitudinous knowledge of them than we do of lions in general. But when we do this, the disjunction of conjunctions of platitudes associated with female lions will change. Is this not a change in concept, according to the cluster view (just as it is a change in concept on the non-cluster view)? Certainly the extension determining cluster itself has changed. Second, a single platitude or two may be all that is holding the possible world extensions of two distinct concepts apart, keeping them disjoint as it were. Such is the case, perhaps, with the concepts Siberian tiger and Bengal tiger, where the relevant true platitudes might be as follows
For Siberian Tigers P4: Xs are predominantly located in or derive from the territory surrounding Siberia P5 : Xs are striped P6 : Xs are large cats
For Bengal Tigers

P7: Xs are predominantly located in or derive from the territory surrounding Bengal. P8: Xs are striped P9: Xs are large cats
The clustering approach now produces chaos. Bengal tigers in any possible world will satisfy a disjunction of three distinct conjuncts each containing two of P4, P5, and P6, since they satisfy P5 and P6. Yet no Bengal tigers are Siberian tigers. Siberian tigers in any possible world will satisfy a disjunction of three distinct conjuncts each containing two of P7, P8 and P9, since they satisfy P8 and P9. Yet no Siberian tigers are Bengal tigers. What if we add this last observation to the list of platitudes; what if we add Siberian tigers are not Bengal tigers to the lists? Even this will not help. It is, in effect, a numbers game; for suppose we also add one more platitude that Siberian and Bengal tigers agree on, say Xs are ferocious. Now there will be five platitudes associated with each concept, and so there will be a disjunct that contains only the three platitudes that both Siberian and Bengal tigers both
satisfy. This disjunct will suffice to give the two concepts the same extension, despite it being a platitude that they do not share the same extension. As an elaboration of this point, let us return for a moment to the case of female lions. Suppose we were able to deal with the earlier difficulty about compositionality by insisting (implausibly) that platitudes must be universals, not generics. The set of platitudes associated with female lion could then be the union of the platitudes associated with female and with lion. But consider the possibility that there are more platitudes associated with lion than with female. Then there will be a disjunct that contains only platitudes associated with lion, and none with female. Clearly, male lions will satisfy this disjunct, and so we have the result that male lions fall in the extension of female lion.12 It is important to notice that this problem does not lie specifically with the idea that a given disjunct should contain most of the platitudes. Even if we raise the requirement to, say, 90% of the platitudes, we need only consider a context in which there are nine times as many platitudes associated with lion than with female. (This could be the result of a case of extreme female sequestration, across the entire animal kingdom.) A final difficulty with clustering may now be noted. Suppose we persuade ourselves that there is an a priori and necessary connection between two concepts; given clustering, how can we be sure that the a priori and necessary status of this connection will be preserved? Perhaps a plausible case is the concept mare and the concept horse, so that it is a priori and necessary that if x is a mare then x is a horse. Intuitively, if anything is due to concepts, this is due to the concept mare and the concept horse, and so the relevant platitude if x is a mare then x is a horse should be part of the roster of platitudes for both concepts. But depending on how many platitudes there are for mares, something, say a perfect inorganic simulacrum of a mare, can fall under most of the platitudes on the roster for the concept mare, even though it is not a horse. How then can it be a priori and necessary that if x is a mare then x is a horse? It should be noted that that many of these objections would still apply even if we limited the platitudes to prima facie a priori claims, as Smith (1994) suggests. Clustering would thus appear to be an inherently problematic strategy. But then how do we avoid the unwanted result that a single false platitude makes the target term empty?
Striking Property Generics and the Requirement of Naturalness

Earlier, we indicated that we wanted to consider some variants of Lewiss view that might be considered further tweaks to which a particular planner might be attracted. Consider the following variant on Lewiss stated theory, one which does not in general associate application conditions with concepts, but only gives a criterion for deciding when one competing account is to be preferred to another: where we have two or more competitors we should prefer, other things
being equal, the account that vindicates more of the relevant platitudes. This criterion would generate a preference even in a case in which almost all of the relevant platitudes are false. In such cases, we believe Lewis would have said that there is no good account of the extension determiner of the concept. So it is a significant weakening of Lewiss actual view. Consider also the following significant strengthening of Lewiss view: the extension-determiner expressed by a term encompasses almost all of the platitudes associated with the term. Both the weakening and the strengthening of Lewiss view run into difficulties when we see that many of our platitudes are generic in form, and then come to understand the import of generic platitudes. Recall that some generics, such as sharks attack bathers, manipulators are evil and mosquitoes carry West Nile virus, can be true even though very few members of the kind in question have the property.13 Such cases arise if the property in question makes its bearers dangerous. One tempting first response is to suppose that, if a property has this feature, then a generic is true if just some members of the kind have the property in question. However this is too simple: consider fish attack bathers and mammals maul children. Some great white, oceanic white-tipped, bull and tiger sharks attack bathers, and some pit-bulls maul children, yet the more general generics concerning fish and mammals do not seem right. Leslie (2007, 2008, in press c) proposes that these sorts of generics are only true if the kind in question is a good predictor of the striking property, where a kind is a good predictor if its members are typically disposed to have the property in question, even if they do not manifest it. Fish attack bathers is false because trout and salmon and sea bass have no such disposition. However, we do not often have access to detailed information about unmanifested dispositions; so the question then arises, how do we in fact select a kind to be the locus of a dangerous-making property generic in the absence of this sort of information? A plausible hypothesis is that, by default, we generalize the property to the basic-level kind. These kinds make up the psychologically privileged level of the subjective taxonomy, which includes kinds such as sharks, tigers, and lions. This pattern of generalization allows us, by default, to generalize these properties to kinds that can be easily and efficiently identified (see Leslie, 2008, in press c for more details). However, there is no guarantee that such kinds will not be overly inclusive; their members may not, in fact, be typically disposed to have the property in question. For example, consider sharks attack bathers. It is plausible to suppose that this generic is, in fact, false, since arguably only great white, oceanic white-tip, bull and tiger sharks have this disposition. This is plausibly a case in which our default practice of generalizing to the basic-level kind leads us to over-generalize the property. It would, however, seem to be a platitude that sharks attack bathers. It certainly satisfies the intuitive test, namely that it is natural to say, in an authoritative tone, it is widely known that sharks attack bathers! In fact, of course, this is one of those cases that Lewis deliberately allowed for, a case in
which a platitude, something supposedly widely known, is false. It is only the great white sharks, the oceanic white-tip sharks, the tiger sharks and the bull sharks that ever attack bathers; for example, basking sharks, whale sharks and megamouth sharks feed only on plankton. But this is not widely known, and so not a platitude. Suppose then, for purposes of argument, that great white sharks possess all the properties platitudinously predicated of sharks. After all they are sharks, and clustering aside, that is how you get to count as a shark on the Lewis view. Thus the great white sharks satisfy more of the platitudes on the roster for the concept shark, since they do attack bathers. So we now have an unintuitive result, namely that the extension of sharks is great white sharks. (By now it should be clear that unintuitive in such contexts does not mean: at odds with the proper analysis of the concept shark. It just means hard to believe, even given the circumstances imagined.) Alternatively, suppose that there is one platitude on the roster of the concept shark that great white sharks do not satisfy. Then we indeed might have a tie, where both sharks and great white sharks each satisfy all but one of the platitudes on the roster of the concept shark, a different one in each case. Then the third distinctive feature of Lewiss account of how to define terms will kick in; when we have such a tie we should take the extension of the term to be the more natural set of things that satisfies the platitudes. As we noted earlier, for Lewis a property or class is perfectly natural when it is suited to figure in a fundamental ontology; when things share such a property or are in such a class they exhibit at least one respect of perfect similarity. When we depart from the perfectly natural, the controlling rule is that the more natural a property or class, the more the sharing of a property or membership in the class makes for genuine similarity. So the class of electrons is more natural than the class of particles; the electrons collectively exhibit a higher degree of genuine similarity than does the wider class of fundamental particles. Likewise, the great white sharks exhibit a higher degree of genuine similarity than the sharks. Thus, if we apply Lewiss criterion of naturalness to break the tie between sharks and great white sharks in the case where each satisfy all but one of the platitudes on the roster of the concept shark, we then get the unwanted result that the extension of sharks is the great white sharks. That unwanted result is the upshot of an unavoidable fact about platitudes, namely that just what the available platitudes are at a time is a contingent psycho-social matter, and a typical feature of naturalness, namely that a so-called basic-level kind like shark is less natural than its sub-kinds such as megamouth shark and great white shark. It may even be possible to generate this sort of unwanted result simply from the facts about naturalness alone. Suppose we have a kind, namely Panthera tigris, more typically known as the kind tiger, and suppose that this kind satisfies all the platitudes governing the term tiger. Depending on how the facts of genetics stand there may be a more natural sub-kind or sub-class of the kind tiger, which also satisfies all the platitudes in question. Here is a way of looking for it; identify some mutant tigers of a very distinctive sort, and consider the kind or class that
includes just the tigers that do not have that mutation. It will be very likely that this sub-kind or sub-class exhibits more genetic and phenotypic similarity among its members than does the kind tiger. To take a concrete example; consider the albinos, mutant tigers with a homozygous occurrence of a recessive gene that controls coat color, which has the effect of producing a stripe-free white coat. The albinos lie among the tigers, but thanks to the generic character of platitudes like tigers have stripes this does not prevent the kind tiger or the class of tigers from satisfying all the platitudes. But now consider the non-albino tigers, they may also satisfy all the platitudes, for the fact that some tigers are albinos is a relatively recherch e fact, and so need not be a platitude. The non-albino tigers are not only more phenotypically similar among themselves than are the tigers; they are genetically more similar among themselves than the tigers. The non-albino tigers are, accordingly, the more natural class. Applying the Canberra plan, we get the unwanted result that our term tiger applies just to the non-albino tigers.14 We could go on in much more detail, but perhaps enough has been said to show that the Canberra plan, understood as a generalization or natural variant on Lewis on defining terms, is not really viable as a method of conceptual analysis. There is another, much more deflationary way of construing the interesting work the advocates of the plan have done on color, mind and value; on this construal the work has been very useful, but it is not the implementation of a recognizable successor to conceptual analysis, let alone a newfangled method of conceptual analysis. It is just the humdrum, old-fashioned method of starting our inquiry with what we happen to take ourselves to know or firmly believe. That method is more or less compulsory anyway, and its credentials do not improve if it is further adorned with implausible aspirations to thereby provide an analysis, or explain the appearance of the a priori.
Conclusion
There are strong grounds for doubting that even when one takes into account all the heuristics and guiding theoretical commitments that we employ in applying a term T they will determine the conditions of application of the concept we express by T. Indeed, there is good reason to doubt that when one takes all the heuristics that we employ in applying T into account, and all the facts about the relative naturalness of the extensions that could be captured by those, then all of this together will determine the conditions of application of the concept we express by T. Just this emerged in the discussion of the Canberra plan. Furthermore, once we assimilate the relevant psychology and philosophy, the following picture presents itself as plausible overview of the subject area that goes under the title of concepts.
1. There are terms, both linguistic and mental. Terms somehow respectively get to be about individuals, classes, kinds, properties, etc. They somehow come to have these worldy items as their extensions. The unsolved mystery of reference or intentionality remains; we have shed no positive light upon it, but have only tried to rule out certain popular conceptions of how it works. Perhaps one crucial determiner of reference is the acceptance correction by others whom we regard as speaking our common tongue, so that we can then cull hopeless hypotheses about the reference of our common terms, and refine useful ones. 2. As a matter of psychological fact, subjects associate criteria with a term; these are heuristics, including perceptual prototypes, and guiding theoretical commitments, which put us in a good position to recognize many items in the extension of the term. These heuristics may be effective only in certain restricted situations, and the guiding theoretical commitments may be, even in large part, false. 3. Even when they are true, the guiding theoretical commitments are likely to be generic in form, allowing for many exceptions which are not counterexamples. (The generic encoding hypothesis.) 4. The psychologically real analog of Freges notion of sense, namely the set of criteria a subject associates with a term, does not determine the reference of terms. It does not guarantee the compositionality of our concepts. It does not make for publicity, for different speakers using the same term with the same extension can employ different theories and different heuristics as guides. 5. The psychologically real analog of sense may serve to explain how it is that terms with the same possible word extensions can have different cognitive values for a given speaker. It can be informative to learn that vixens are female foxes because it unites the criteria we use for vixen with those we use for female foxes. 6. The method of cases, the method of trying to articulate what we know in using a term by considering how we are inclined to apply it to real and merely possible cases, is at best a way of articulating our criteria. It is not a method for analyzing concepts, as is shown by the fact that our reaction to cases is typically explained by our criteria, and not by some supposed grasp of the application conditions of our concepts. 7. There is no empirically explanatory need to postulate concepts, understood as abstract entities (i) which encode the conditions of application of our terms and (ii) grasp of which guides in the use of our terms. Talk of grasping a concept is either an unhelpful metaphor or a description of a supernatural theory of language use, one rendered otiose by the actual empirical psychology of language use. 8. There is accordingly no good reason to think that there are concepts in any substantial philosophical sense. There are terms, public languages,
the extensions of terms, and criteria for applying terms. There is the something or other that gives terms their particular extensions, and hence their conditions of application. But this something or other is never a concept in the substantial philosophical sense. Concepts in the substantial sense have no demonstrated use.
Notes
1. Special thanks to Shamik Dasgupta, Frank Jackson, Joshua ORourke, Nathaniel Tabris, and Gideon Rosen for helpful comments on an earlier draft. 2. Preparation of this article was supported by NSF grant BCS-1226942, awarded to Leslie. 3. 1967, p 22. 4. See, for example, Ramsey (1992), Stich and Weinberg (2001), Laurence and Margolis (2003), Sytsma (2010), and Banicki (2012). 5. It is interesting here to consider the findings of Armstrong, Gleitman, and Gleitman (1983), who found some of the typicality effects even for concepts such as odd number. This shows, of course, that typicality effects are compatible with representing necessary and sufficient conditions. However, the crucial point is that in the case of concepts such as odd number we have independent reason for supposing that we represent necessary and sufficient conditions namely that we (i.e., typical competent adult users of the term odd number) can articulate what they are. There is no corresponding case to be made for the majority of concepts. Thus while the in-principle compatibility point concerning typicality effects and necessary and sufficient conditions is illustrated by Armstrong et al.s findings, it does not address the fact that, for most concepts, we simply have no reason to suppose that typical, competent adults represent and exploit necessary and sufficient conditions. 6. Sometimes theory-theory is reserved for the specific view that our concepts, including young childrens concepts, are exactly like scientific theories. We follow many in the field by using the term more inclusively. 7. It is, perhaps, rather telling that psychological experiments concerned with concepts almost invariably use the generic form to articulate conceptual knowledge. 8. Some friends of the Canberra plan, such as Michael Smith (1994), reject this liberal attitude to platitudes. For Smith, platitudes must be prima facie a priori, i.e. such that if they are true then their truth is not an empirical matter. However this means that the plan will often fail, since, as we have seen, in general there are just not enough prima facie a priori truths around to determine the application conditions of our terms. Indeed, as in the case of space not in fact being Euclidean makes clear, empirical inquiry is itself highly relevant to the question of whether it is the case that if S is true then its truth is an a priori or instead an empirical matter. We often simply cannot imagine what would undermine our central beliefs. So, the category of what we can legitimately take to be prima facie a priori will shrink as we become more and more aware of the inter-animation of the putatively a priori and the empirical. In fact, Smiths own conclusion is that the plan does fail in both of the cases he is concerned with, namely the analysis
Concepts, Analysis, Generics and the Canberra Plan / 167 of red and in the analysis of right. And he provides a plausible diagnosis of this failure; in such cases competence with the term consists in part in a disposition to use it to classify cases correctly on the basis of certain patterns of features. There is no reason to believe that there is any prima facie a priori encapsulation of the relevant patterns. Although Lewis drops the term platitudes in Reduction of Mind (1994) he crucially still talks in terms of putative common knowledge. He writes We have a very extensive shared understanding of how we work mentally. Think of it as a theory: FOLK PSYCHOLOGY. It is common knowledge among us; but it is tacit, as our grammatical knowledge is. We can tell which particular predictions and explanations conform to its principles, but we cannot expound those principles systematically (p. 56). Our use of platitudes throughout is meant to conform simply to this idea that we can articulate the relevant putative common knowledge in which our terms are entangled, and then use it to define our terms. In fact, Lewis (1997) seems to persist with a platitude-based analysis of color terms. We have argued that the platitudes associated with intersective concepts do not always include the union of the platitudes associated with the component concepts. A related point that Jerry Fodor (e.g., 1998) has often made is that the platitudes associated with the complex concept will also not usually be exhausted by this union of platitudes. That is, it could be a platitude that black bulls are dangerous without it is being a platitude that black things are dangerous or its being a platitude that bulls are dangerous. One response to failures of compositionality due to the widespread application of ones favored technique of analysis is instead to identify a base of uncompounded terms, apply ones favored technique of analysis only to them and then simply let the syntactical methods of combination by which the language generates compound terms determine the extensions of compound terms. This, however, would be an unhappy way to patch up the failures of compositionality due to widespread application of the Canberra plan. First, it is likely that many terms in any plausible base will be partly recognitional terms like red and dog and cause, for which the platitude-based analysis is ill-suited because of the reasons noted earlier. Second, many of the target terms of philosophical analysis are unlike physical cause and more like free will in being themselves complex though not simply intersective combinations (or any other simple combination) of their respective components (here free and will). Note that it will not help to insist, implausibly, that each concept have the same number of platitudes associated with it. For consider then the complex concept adult female lion. By assumption there is the same number of platitudes associated with all three concepts, so each concept contributes one third of the total number of platitudes. But then again there will be a disjunction that contains only platitudes associated with adult and lion, and none with female. Adult male lions will again satisfy this disjunction. Far and away, most cases of manipulation involve parents and children; you get to count as a manipulator by manipulating a bit now and then, but the overwhelming majority of parents and children who manipulate a bit now and then are not evil. The same holds for lawyers, politicians and the police.
9.
10.
11.
12.
13.
168 / Mark Johnston and Sarah-Jane Leslie 14. For an analog of this sort of point in the case of Lewiss analysis of the concept of personal identity, see Johnston (2010), pp. 7980 On the Unhelpfulness of Reference Magnetism.
References
Ahn, W., Kim, N. S., Lassaline, M. E., & Dennis, M. J. (2000). Causal Status As A Determinant of Feature Centrality. Cognitive Psychology, 41, 361416. Armstrong, S.L., Gleitman, L.R., & Gleitman, H. (1983). What Some Concepts Might Not Be. Cognition, 13, 263308. Banicki, K. (2012). Connective Conceptual Analysis and Psychology. Theory and Psychology, 22 (3), 310323. Barner, D., Chow, K., & Yang, S. (2009). Finding Ones Meaning: A Test of the Relation Between Quantifiers and Integers in Language Development.Cognitive Psychology, 58, 195219. Bealer, G. (1987). The Philosophical Limits of Scientific Essentialism. Philosophical Perspectives, 1, 289365. Boghossian, P. (2003). Blind Reasoning. Proceedings of the Aristotelian Society, Supplementary volume, 77, 225248. Braddon-Mitchell, D., & Nola, R. (2009). Conceptual Analysis and Philosophical Naturalism. Cambridge, MA: MIT Press. Brandone, A., Cimpian, A., Leslie, S. J., & Gelman, S. A. (2012). Do Lions Have Manes? For Children, Generics are About Kinds, Not Quantities. Child Development, 83(2), 423433. Brandone, A., Gelman, S. A., & Hedglen, J. (submitted). Young Childrens Intuitions about the Truth Conditions and Implications of Novel Generics. Carey, S. (1985). Conceptual Change in Childhood . Cambridge, MA: MIT Press. Carey, S. (2009). The Origin of Concepts. New York: Oxford University Press. Carlson, G.N., & Pelletier, F.J. (1995). The Generic Book. Chicago: Chicago University Press. Chalmers, D. (1996). The Conscious Mind . New York: Oxford University Press. Chalmers, D., & Jackson, F. (2001). Conceptual Analysis and Reductive Explanation. The Philosophical Review, 110, 315361. Cimpian, A., Brandone, A. C., & Gelman, S. A. (2010). Generic Statements Require Little Evidence for Acceptance But Have Powerful Implications. Cognitive Science, 34(8), 14521482. Cimpian, A., Gelman, S. A., & Brandone, A. C. (2010). Theory-Based Considerations Influence The Interpretation of Generic Sentences. Language and Cognitive Processes, 25(2), 261276. Cohen, A. (1996). Think Generic: The Meaning and Use of Generic Sentences. Ph.D. dissertation, Carnegie Mellon University. Dahl, O. (1985). Tense and Aspect Systems. Oxford: Blackwell. Davidson, D. (1967). Truth and Meaning. Synthese, 17, 304323. Devitt, M. (2005). There is No A Priori. In E. Sosa and M. Steup (eds.), Contemporary Debates in Epistemology, pp. 105115. Cambridge, MA: Blackwell Publishers. Fodor, J. A. (1998). Concepts: Where Cognitive Science Went Wrong. Oxford: Oxford University Press. Fodor, J. A., Garrett, M. F., Walker, E., & Parkes, C. (1980). Against definitions. Cognition, 8, 263367.

Fodor, J. A., & Lepore, E. (2002). The Compositionality Papers. New York: Oxford University Press. Frege, G. (1956). The Thought: A Logical Inquiry. Mind , 65, 289311 Geach, P.T. (1967). Identity. Review of Metaphysics, 21, 312. Gelman, S. A. (2003). The Essential Child: Origins of Essentialism in Everyday Thought. New York: Oxford University Press. Gelman, S. A. (2010). Generics as A Window onto Young Childrens Concepts. In F. J. Pelletier (ed.), Kinds, Things, and Stuff: The Cognitive Side of Generics and Mass Terms (New Directions in Cognitive Science v. 12.), pp. 100123. New York: Oxford University Press. Gelman, S.A. & Raman, L. (2003). Preschool Children use Linguistic form Class and Pragmatic Cues to Interpret Generics. Child Development, 74, 308325. Gelman, S. A., Star, J., & Flukes, J. (2002). Childrens Use of Generics in Inductive Inferences. Journal of Cognition and Development, 3, 179199. Gopnik, A., & Meltzoff, A.N. (1997). Words, thoughts, and theories. Cambridge, MA: MIT Press. Graham, S. A., Nayer, S. L., & Gelman, S. A. (2011). Two-Year-Olds use the Generic/NonGeneric Distinction to Guide Their Inferences About Novel Kinds. Child Development, 82, 493507. Harman, G. (1974). Meaning and semantics. In M. K. Munitz and P. K. Unger (eds.), Semantics and Philosophy, pp. 116. New York: NYU Press. Harman, G. (1975). Language, Thought, and Communication. In K. Gunderson (ed.), Language, Mind, and Knowledge: Minnesota Studies in the Philosophy of Science VII, pp. 27998. Minneapolis, MN: University of Minnesota Press. Hawthorne, J., & Price, H. (1996). How to Stand Up For Non-Cognitivists. Australasian Journal of Philosophy, 74 (2), 275292. Hollander, M. A., Gelman, S. A., & Star, J. (2002). Childrens Interpretation of Generic Noun Phrases. Developmental Psychology, 38, 883894. Jackson, F. (1994). Armchair Metaphysics. In M. Michael and J. OLeary-Hawthorne (eds.), Meaning in Mind , pp. 2342. Boston: Kluwer Academic Publishers. Jackson, F. (1998). From Ethics to Metaphysics: A Defense of Conceptual Analysis. Oxford: Clarendon Press. Johnston, M. (1992). How to Speak of the Colors. Philosophical Studies, 68(3), 221263. Johnston, M. (2010). Surviving Death. Princeton, NJ: Princeton University Press. Jonsson, M. L. & Hampton, J. A. (2006). The Inverse Conjunction Fallacy. Journal of Memory and Language, 55, 317334. Keil, F. (1989). Concepts, Kinds, and Cognitive Development. Cambridge, MA: MIT Press. Khemlani, S., Leslie, S. J., & Glucksberg, S. (2009). Generics, Prevalence, and Default Inferences. Proceedings of the 31st Annual Cognitive Science Society. Amsterdam: Cognitive Science Society. Kripke, S. (1980). Naming and Necessity. Cambridge MA: Harvard University Press. Kripke, S. (1982). Wittgenstein on Rules and Private Language. Cambridge, MA: Harvard University Press. Laurence, S., & Margolis, E. (2003). Concepts and Conceptual Analysis. Philosophy and Phenomenological Research, 67 (2), 253282. Leslie, A.M., & Keeble, S. (1987). Do Six-Month-Old Infants Perceive Causality? Cognition, 25, 265288. Leslie, S. J. (2007). Generics and the Structure of the Mind. Philosophical Perspectives, 21, 375403. Leslie, S. J. (2008). Generics: Cognition and Acquisition. Philosophical Review, 117(1), 147. Leslie, S. J. (in press a). Generics Articulate Default Generalizations. Recherches Linguistiques de Vincennes.

S. J. (in press b). Essence and Natural Kinds: When Science Meets Preschooler Intuition. In T. Gendler and J. Hawthorne (eds.), Oxford Studies in Epistemology, 4. Leslie, S. J. (in press c). The Original Sin of Cognition: Fear, Prejudice, and Generalization. The Journal of Philosophy. Leslie, S. J., & Gelman, S. A. (2012). Quantified Statements are Recalled as Generics: Evidence from Preschool Children and Adults. Cognitive Psychology, 64, 186214. Leslie, S. J., Khemlani, S., & Glucksberg, S. (2011). All Ducks Lay Eggs: The Generic Overgeneralization Effect. Journal of Memory and Language, 65, 1531. Lewis, D. (1970). How to Define Theoretical Terms. Reprinted in his Philosophical Papers, pp. 7895, vol. 1, (1983), New York: Oxford University Press. Lewis, D. (1972). Psychophysical and Theoretical Identifications. Australasian Journal of Philosophy, 50, 249259. Lewis, D. (1973). Causation. Journal of Philosophy, 70, 556567. Lewis, D. (1975). Adverbs of Quantification. In E. L. Keenan (ed.), Formal Semantics of Natural Language, pp. 315. Cambridge: Cambridge University Press. Lewis, D. (1983). New Work for a Theory of Universals. Australasian Journal of Philosophy, 61, 343377. Lewis, D. (1984). Putnams Paradox. Australasian Journal of Philosophy, 62, 221236. Lewis, D. (1989). Dispositional Theories of Value. Proceedings of the Aristotelian Society, Supplementary volume 63, 113137. Lewis, D. (1994). Reduction of Mind. In S. Guttenplan (ed.) A Companion to the Philosophy of Mind , pp. 41231. Oxford: Basil Blackwell. Lewis, D. (1997). Naming the Colours. Australasian Journal of Philosophy, 75(3), 325342. Lewis, D. (2000). Causation as Influence, Journal of Philosophy, 97, 209212. Liebesman, D. (2011). Causation and the Canberra Plan. Pacific Philosophical Quarterly, 92(2), 232242. Machery, E. (2009). Doing Without Concepts. New York: Oxford University Press. Mannheim, B., Gelman, S. A., Escalante, C., Huayhua, M., & Puma, R. (2011). A developmental analysis of generic nouns in Southern Peruvian Quechua. Language Learning and Development, 7(1), 123. Margolis, E., & Laurence, S. (1999). Concepts: Core Readings. Cambridge, MA: MIT Press. Meyer, M., Gelman, S. A., & Stilwell, S. M. (2011). Generics are a Cognitive Default: Evidence from Sentence Processing. In Proceedings of the 33rd Annual Conference of the Cognitive Science Society. Boston, MA: Cognitive Science Society. Michotte, A. (1963). The Perception of Causality. Andover: Methuen. Murphy, G. (2002). The Big Book of Concepts. Cambridge, MA: MIT Press. Peacocke, C. (1992). A Study of Concepts. Cambridge, MA: MIT Press. Peacocke, C. (1993). How Are A Priori Truths Possible? European Journal of Philosophy, 1, 17599. Peacocke, C. (1997). Metaphysical Necessity: Understanding, Truth and Epistemology. Mind, 106, 52174. Peacocke, C. (1998). Implicit Conceptions, Understanding and Rationality. in E. Villanueva (ed.), Philosophical Issues 9: Concepts, pp. 12148. Peacocke, C. (2005). The A Priori. In F. Jackson and M. Smith (eds.), The Oxford Handbook of Contemporary Philosophy, pp. 739767. Oxford: Oxford University Press. Pelletier, F. J. & Asher, N. (1997). Generics and Defaults. In J. van Benthem and A. ter Meulen (eds.), Handbook of Logic and Language, pp. 11251179. Cambridge, MA: MIT Press. Prasada, S., Khemlani, S., Leslie, S. J., & Glucksberg, S. (in press). Conceptual Distinctions Amongst Generics. Cognition. Putnam, H. (1962). It Aint Necessarily So. The Journal of Philosophy, 59, 658671. Leslie,

Putnam, H. (1973). Explanation and Reference. Reprinted in his Mind, Language and Reality: Philosophical Papers, vol. 2, (1975), pp. 196215. New York: Cambridge University Press. Putnam, H. (1975). The Meaning of Meaning. Reprinted in his Mind, Language and Reality: Philosophical Papers, vol. 2, (1975), pp. 215272. New York: Cambridge University Press. Putnam, H. (1980). Models and Reality, Journal of Symbolic Logic, 45(3), 464482. Quine, W.V.O. (1951). Two Dogmas of Empiricism. Reprinted in his From a Logical Point of View, (1953), pp. 2047. Cambridge MA: Harvard University Press. Ramsey, W. (1992). Prototypes and Conceptual Analysis. Topoi, 11, 5970. Rosch, E. (1973). Natural Categories. Cognitive Psychology, 4, 328350. Rosch, E. (1978). Principles of Categorization. In E. Rosch & B.B. Lloyd (eds.), Cognition and Categorization, pp. 2748. Hillsdale: Lawrence Erlbaum Associates. Rosch, E. & Mervis, C.B. (1975). Family Resemblances: Studies in the Internal Structure of Categories. Cognitive Psychology, 7(4), 573605. Schroeter, L. (2004).The Limits of Conceptual Analysis. Pacific Philosophical Quarterly, 85, 425453. Sloman, S. A. (1993). Feature-based induction. Cognitive Psychology, 25, 231280. Sloman, S. A. (1998). Categorical inference is not a tree: The myth of inheritance hierarchies. Cognitive Psychology, 35, 133. Smith, E. E., & Medin, D. L. (1981). Concepts and Categories. Cambridge, MA: Harvard University Press. Smith, M. (1994). The Moral Problem. Oxford: Wiley-Blackwell. Stich, S.P., & Weinberg, J.M. (2001). Jacksons Empirical Assumptions. Philosophy and Phenomenological Research, 62 (3), 637643. Sytsma, J. (2010). The Proper Province of Philosophy. Review of Philosophy and Psychology, 1(3), 427445. Tardif, T., Gelman, S. A., Fu, X., & Zhu, L. (2011). Acquisition of Generic Noun Phrases in Chinese: Learning about Lions without An -S. Journal of Child Language, 30, 132. Tooley, M. (1987). Causation: A Realist Approach. New York: Oxford University Press. Williamson, T. (2003). Understanding and Inference. Proceedings of the Aristotelian Society, 77, 249293.
Sarah-Jane Leslie/Generics articulate cognitively default generalizations
Sarah-Jane Leslie Princeton University
Generics Articulate Default Generalizations
Introduction: Our Judgments of Generics Generics are sentences that express generalizations about kinds, such as tigers are striped, ducks lay eggs, and ticks carry Lyme disease. 1 Quantified sentences such as all tigers are striped, most barns are red, and some ticks carry Lyme disease can also express generalizations about kinds, but unlike generics, they convey information about how many members of the kind have the property in question. Quantified statements have proved highly amenable to semantic analysis for example, all tigers are striped is true just in case the set of tigers is included in the set of striped things. More generally, quantified statements depend solely on straightforward content-neutral information of the sort that can be represented in set-theoretic terms and as a result can be readily analyzed using the standard tools of formal semantics (e.g. Barwise & Cooper, 1981). In contrast to quantified statements, generics have proved difficult for semanticists to analyze. As a simple illustration of why this is so, consider
the generics ducks lay eggs and ducks are female. There are more female ducks than there are egg-laying ducks (since some female ducks are infertile, immature, and so on), yet the former statement seems intuitively true, while the latter does not. Similarly, consider ticks carry Lyme disease; this would seem to be clearly correct, yet only one percent of ticks carry the disease. Conversely, however, books are paperbacks does not seem to be correct, yet over eighty percent of books are paperbacks. In Leslie (2007, 2008), I suggested that our judgments of generics may be sensitive to non-quantitative, content-based factors. In particular, I proposed that our judgments are sensitive to whether the property in question is characteristic of the kind e.g., does the property specify how the kind reproduces? How it nurtures its young? Its identifying salient physical features? and also to whether the property in question is striking e.g., dangerous, threatening, appalling; the sort of property about which one would wish to be forewarned. If a property is either characteristic or striking, then the generic may well be accepted at low prevalence levels hence the disposition to accept ducks lay eggs and ticks carry Lyme disease respectively. I also proposed that our judgments may be sensitive to the nature of the exceptions to the generic claim. That is, amongst the members of the kind that fail to have the predicated property, it matters how they fail to have the property in particular whether they simply lack the property, or whether they have an equally salient, concrete, positive property instead. The non-infected ticks simply do not carry Lyme disease they do not have an alternative property instead. However, the books that are not paperbacks are instead hardcover. Consider also people are right-handed, elementary school teachers are female in both cases the predicated property is highly prevalent, yet the generic is intuitively false. But again there are positive counterinstances people who are not right-handed are instead left-handed; elementary school teachers who are not female are male. Intuitively, it feels as though we would be overlooking these hardcover books, these lefthanded individuals, and these male school teachers if we accepted the generics in question. Thus, I proposed that generics are likely to be rejected even if the property is highly prevalent, if there are these positive counterinstances. (For a more detailed presentation of these aspects of the proposal, see Leslie, 2007, 2008, in press a; for a discussion of social prejudice in light of this account, see Leslie, in press b.) In this earlier work, I relied on my own evaluations of the generics in question, and on evaluations that have been made in the literature by linguists and philosophers for a number of years now (e.g. Carlson, 1977; Carlson & Pelletier, 1995; Lawler, 1973). That is, I took for granted that ticks carry Lyme disease can be true even if few ticks carry the disease; this was one key point I sought to explain. Recently, several experiments have been
conducted to test whether nave participants also share the theorists intuitions that, e.g., a generic such as ticks carry Lyme disease can be true even though very few ticks carry the disease. Yet one might wonder what the value of conducting such empirical tests could possibly be: havent the theorists already established the conclusion that the empirical work will surely support? That is, dont we already know that ticks carry Lyme disease can be true despite the low percentage of infected ticks? Empirical work in this regard is valuable, even if it is unlikely to overturn orthodoxy. For one, even if it is unlikely to overturn orthodoxy, one can never be certain. Suppose, for example, that false beliefs about the percentage of ticks that carry Lyme disease abound in the population: most people believe that approximately 80% of ticks are infected. Suppose further that people only reliably accept generics when they believe the prevalence of the property to be high. In this scenario, the generic sentence ticks carry Lyme disease is reliably asserted, but only because of false beliefs about prevalence. A theorist may grow up hearing ticks carry Lyme disease may read the sentence repeatedly in the media and so hear the sentence as true. Later, while conducting work on generics, the theorist may learn that in fact very few ticks carry Lyme disease, but since her belief in the truth of ticks carry Lyme disease has been built up over many years, she may still remain confident that the sentence is true, and so write that ticks carry Lyme disease can be true despite few ticks carrying the disease. However, all this scenario would reflect is that our beliefs in the truth of generics are not easily overturned, despite being in the general case driven solely by underlying beliefs about prevalence. If it is only the theorists who accept generics at low prevalence levels, this should give us pause for thought and whether such a situation obtains cannot be determined on the basis of theorist intuition. In this case, however, results from recent empirical studies comport with the observations made by linguists and philosophers. For example, Amanda Brandone, Andrei Cimpian, Susan Gelman and I found that adults are significantly more likely to accept sentences such as birds lay eggs (or lions have manes, or pigs give milk their young) than they are to accept sentences such as birds are female (or lions are male, or pigs are female), despite appreciating that there are comparable numbers of female birds (or male lions, or female pigs) and egg-laying birds (or maned lions, or nursing pigs) (Brandone, Cimpian, Leslie & Gelman, in press). That is, for a range of examples, participants understand that only one sex of the kind has a particular characteristic property, yet they are far more likely to accept a generic that predicates the characteristic property than they are to accept a generic that predicates belonging to that sex. Sandeep Prasada, Sangeet Khemlani, Sam Glucksberg and I further examined participants judgments of a range of generics, including generics
that predicate non-prevalent but strikingly dangerous properties (e.g. ticks carry Lyme disease, sharks attack swimmers, mosquitoes carry malaria), and generics that predicate prevalent properties, but have positive counterinstances (e.g., books are paperbacks, Canadians are righthanded, and teachers are female). We also elicited prevalence estimates from participants that is, participants were asked to judge what percentage of the kind has the property in question. As predicted, participants gave much higher prevalence estimates for the positive counterinstance items than for the strikingly dangerous items, yet were much more likely to endorse the strikingly dangerous generics than the positive counterinstance generics (Prasada, Khemlani, Leslie & Glucksberg, submitted). It is sometimes suggested that generics such as ticks carry Lyme disease are accepted because, even though few ticks carry it, ticks are the only creatures that do so. That is, even though few ticks carry Lyme disease, Lyme disease carriers are overwhelmingly ticks. Thus, perhaps we do not need to appeal to content-based factors such as whether the property is dangerous or not; perhaps it is simply that if something carries Lyme disease, then it is almost certainly a tick. Such a measure would be the inverse of prevalence: prevalence can be modeled as the conditional probability of carrying Lyme disease given that one is a tick, while this measure would be the conditional probability of being a tick given that one carries Lyme disease. In the psychology literature, this measure is known as cue validity. While there are some examples, actual and hypothetical, that tell against this possibility (see, e.g., Leslie, 2008), it is difficult to rule out definitively while relying solely on our intuitions as pure theorists of generics. However, it is easy to evaluate this hypothesis in the laboratory. To this end, Prasada et al. also elicited cue validity ratings in our study, and so we were able to control for cue validity across their different categories of generics. That is, the items in the strikingly dangerous category had the same average cue validity ratings as the items in the positive counterinstance category, yet the above results obtained nonetheless. If the strikingly dangerous generics were accepted because they had high cue validity ratings, this would not have been possible; the positive counterinstance generics had the same cue validity ratings, yet they were not accepted despite being associated with much higher prevalence estimates. Cue validity cannot explain why these striking generics were accepted at low prevalence levels. The category of strikingly dangerous generics in the Prasada et al. study was constructed so as to contain generics that predicate dangerous yet non-prevalent properties. That it was possible to do so that is, to construct a range of generics that predicate strikingly dangerous properties and are accepted despite having low associated prevalence estimates and mid-range cue validity estimates lends indirect support to the hypothesis in Leslie
(2007, 2008). However, the support is only indirect, since it does not involve directly testing the impact of a propertys being dangerous on peoples judgments. Recent work conducted by Andrei Cimpian, Amanda Brandone, and Susan Gelman, however, examined precisely this question (Cimpian, Brandone & Gelman, 2010). In their experiment, participants were told about a novel animal kind, e.g. lorches. They were then told that a certain percentage of lorches had a particular property, e.g., purple feathers, then they were given some further information about the feathers. Crucially, in some cases, this further information described a dangerous property e.g., the feathers were poisonous to the touch whereas in other cases it was anodyne. The participants were then asked to evaluate a generic attributing the property to the kind. The experimenters found that, at low prevalence levels, participants were significantly more likely to accept the generic if the property in question was dangerous, thus lending support to the hypothesis that our judgments of generics are sensitive to factors such as whether the property being predicated is strikingly dangerous (Leslie, 2007, 2008). One point to note is that in the Cimpian et al. study, the nondangerous properties were still inherent, biologically-based properties (e.g., purple feathers), and so may have been understood to be characteristic properties of the kind. 2 That is, the rate of acceptance of these non-dangerous but inherent properties may already have been high relative to more accidental properties. To examine this, Cimpian et al. adapted their paradigm to include properties that would likely be interpreted as accidental, non-characteristic properties e.g. muddy feathers, broken legs. In the case of such properties, participants almost never accepted the generic at low prevalence levels; instead they only accepted the generic when a strong majority of the kind possessed the property, which is consistent again with the predictions made in Leslie (2007, 2008). Susan Gelman and Paul Bloom (2007) also found converging evidence of the impact of the inherent and biological vs. the adopted and accidental on adults judgments of generics. Recent empirical work thus suggests that adults judgments of generics are sensitive to rich, content-based factors such as whether the property in question is strikingly dangerous, whether it is an inherent characteristic property of the kind, or whether it is merely accidentally possessed by members of the kind. The impact of these factors does not appear to be reducible to abstract, quantitative measures such as prevalence and cue validity. The Psychological Significance of Generics The question one faces at this juncture is whether the foregoing observations and findings simply amount to the following: there is a class of sentences, evaluations of which are sensitive to a number of surprising factors. This
would be of some interest to some theorists, but may not be terribly gripping at the end of the day. Is this as far as the study of generics can take us, or might they hold interest from the perspective of the nature of cognition more generally? Taking a step back for a moment, let us note that one of the most important tasks a learner faces is how to arrive at general information that applies to situations and instances that they have not yet encountered how to move from the particular to the general. If one is unable to generalize information in this sense, ones knowledge will be forever bound to the hereand-now every new situation must be encountered without the support of prior knowledge. As it happens, if one touches a hot stove and burns ones hand, one will be careful about touching that stove in the future, and one will also be careful about touching other stoves. This sort of learning involves generalization in the sense I am using it here in the first instance, generalization from one time to future times; and in the second, from one instance of a kind to other instances of a kind, namely other stoves. There is a sense in which the capacity to generalize is the backbone of all learning without it, one could only catalog discrete and seemingly unrelated events. It is perhaps unsurprising, then, that infants in the first year of life are able to form general judgments (e.g. Baldwin, Markman & Melartin, 1993; Graham, Kilbreath & Welder, 2001). They are, for example, able to form expectations concerning the properties of novel instances of a kind or category on the basis of experience with other instances. Thus they expect that, if one item makes a rattling noise when it is shaken, perceptually similar items will do so too. Infants are not simply trapped in the here-and-now they can extract general information from a situation and apply it to a novel one. There must therefore be something about the infants cognitive capacities that allows them to do this, something in their cognitive makeup that permits them to form and exploit general judgments a cognitive mechanism that forms general judgments. We might then call the generalizations that are formed by this mechanism cognitively fundamental generalizations. This mechanism would represent our most natural and default way of forming general judgments, and would produce the sorts of general judgments that we make from our earliest days. An interesting question, then, is whether language later provides us with a way of giving voice to these generalizations that we have been making all along. It is not guaranteed that natural language must provide us with such a vehicle, but it is reasonable to suppose it might (and would be quite surprising if it did not). We can thus ask the question: of the many ways we have of linguistically expressing generalizations, which form articulates our cognitively fundamental generalizations? That is, if an adult forms a general
judgment using this fundamental, default mechanism of generalization and wishes to communicate this belief using natural language, what linguistic form will be used? The answer I propose here and elsewhere is: the generic form. One observation that can be made in support of this hypothesis is that, while there is a word in English all that signals a universal generalization, a word some that signals an existential one, and so on for all the quantifier words both determiners and adverbs there is no word gen in English that signals a generic generalization. That is, we do not say gen tigers are striped in the way we say all tigers are striped or most tigers are striped. 3 This is not an isolated fact about English either: there is no known language that has a dedicated, articulated generic operator no word gen that uniquely signals that a generalization is a generic one. While languages differ dramatically in how generic claims are syntactically manifested, one does not find phonological forms that are dedicated to the expression of generic claims generics are always unmarked in this way (Dahl, 1985). This would seem a puzzling fact: why should this be the case? And even if a language with a gen in it were to be found, we would still be left with the fact that the overwhelming majority of languages do not contain such a linguistic item a linguistic quasi-universal that would still cry out for explanation. Such an explanation is forthcoming if generics are languages way of letting us give voice to cognitively fundamental generalizations (Leslie, 2007, 2008). If one wishes to interact efficiently with a system, and the system has a basic, default way of proceeding or performing a task, then one need only issue an explicit instruction to the system if one wishes it to deviate from this default way of proceeding. As an intuitive illustration, suppose that one is dealing with a recalcitrant teenager who by default never tidies his room. If on a particular occasion one does not wish for said teenager to tidy his room, one need not say anything it would be a waste of breath, as it were, to say please do not tidy your room today, since if one says nothing, this is precisely what will happen. One need only speak up if one wishes the teenager to deviate from his default; then one must ask that the room be cleaned. Quantifiers, I propose, are articulated precisely because quantified generalizations are non-default generalizations. For the cognitive system to form a universal generalization it must be told to form a universal generalization, as it were and this requires that there be a word in natural language like all. Such non-default generalizations must be explicitly marked they are like the instruction to the teenage to clean his room. However, generic sentences contain no such marking there is no item in them that specifies what sort of generalization is here being expressed. As a result, the cognitive system supplies its own default means of generalizing.
To explicitly instruct a system to operate according to its default is inefficient and unnecessary akin to instructing the teenager not to tidy his room hence the universal or near universal absence of a word gen. The hypothesis that generics articulate cognitively default generalizations, then, allows us to explain this otherwise puzzling crosslinguistic fact. Further, it issues in a number of empirical predictions. For example, if processing quantified statements requires the cognitive system to deviate from its generic default, we would expect that sometimes this default would be inadequately suppressed, and quantified statements would be processed as though they were generics. This might show up in a number of ways in evaluation tasks, in memory tasks, in reasoning tasks. We might further predict that this tendency would be found in adults, but that it would be most pronounced in young children since in general, children are less likely than adults to be able to successfully inhibit default processes. We would also correspondingly predict that generics would prove quite easy for young children to acquire easier to acquire than quantifiers. In case these predictions seem in any way obvious, consider the following contrasting point of view. Quantified statements depend solely on clean, quantitative information, of the sort that is theoretically highly tractable. Generics, on the other hand, appear to be sensitive to all sorts of messy, complicated, content-based factors, as the discussion in the previous section indicates. Surely it would be easier to acquire and process the former than the latter! Further, the fact that there is no word gen should make the learners task even harder she is not even presented with an articulated word whose meaning she can try to figure out. So we have something elegant, tractable, and articulated like all on the one hand, and the opaque, elusive and mysterious generic on the other. It would be quite natural to suppose that a child would find the former easier to acquire than the latter, and that the latter would impose greater processing demands so that if anything children and adults alike should default from the generic to the quantified, rather than vice-versa. Generics in Acquisition Toddlers begin to produce generics at approximately 30 months of age, which is the first time that they reliably have the requisite background syntactic capabilities (e.g. sufficiently many words per utterance, plurality, etc; Gelman, 2010; Gelman, Goetz, Sarnecka & Flukes, 2008). By the time children are between 3 to 4 years of age, they produce generics as frequently as adults do which is not something that is true for all parts of language by a long shot (Gelman et al., 2008). Thirty-month-old toddlers are further able to draw different inferences from generics than from specific statements. That is, if they are shown some members of a novel animal kind performing
an action such as drinking milk and told either blicks drink milk or these blicks drink milk, they are more likely to generalize the action of drinking milk to novel blicks if they hear generic language (Graham, Nayer & Gelman, 2010). Thus, 30-month-olds understand that generics express generalizations, in contrast to specific statements, and further are able to deploy this fact in drawing inferences. Two-year-olds also understand that generics tolerate exceptions. For example, Susan Gelman and Lakshmi Raman (2003) showed young children pictures of familiar animals that were exceptions to a generic generalization the children might have been shown, e.g., a picture of penguins. The children were then either asked do these birds fly? or do birds fly? If young children understand that generics tolerate exceptions, then they should answer no to the specific question but yes to the generic one, and the experimenters found that toddlers as young as 30 months were able to do just this. Several other studies have also found that preschoolers understand that generics tolerate exceptions in this way (e.g., Chambers, Graham & Turner, 2008; Gelman & Bloom, 2007; Gelman, Star & Flukes, 2002). What of preschoolers comprehension of troublesome generics the sorts of generics that have caused difficulty for semanticists studying these sentences? Amanda Brandone, Andrei Cimpian, Susan Gelman and I have begun to research this issue. The question we posed was: if a preschooler knows that, e.g., only girl birds lay eggs, how will she evaluate birds lay eggs and birds are girls? If a child has the relevant background knowledge, then she will at least implicitly believe that there are as many girl birds as there are egg-laying birds. Thus, if preschoolers judgments, unlike adults, are driven by considerations of prevalence, then such a child should be equally likely to endorse the two generics. If, however, preschool children are sensitive to the same factors as adults, then they should be more likely to accept birds lay eggs than birds are girls. We used a range of items that included methods of reproduction, salient identifying physical features, and means of nurturing the young, and we only considered date from trials in which the preschooler in question knew that only one gender of the animal kind in question had the property. We found that preschoolers, like adults, were far more likely to endorse generics that attribute a characteristic property to the kind than they were to endorse generics that predicated belonging to a given gender. 4 This suggests that even for preschoolers, generics are not simple expressions of beliefs about prevalence (Brandone et al., in press). While a more complete investigation into preschoolers understanding of generics is needed (and indeed is underway), the available evidence suggests that, like adults, they understand that generics a) tolerate exceptions, and b) do not depend solely on considerations of prevalence. The
10
data so far also indicate that generics are acquired easily and early. We might ask, then, how the acquisition of generics compares to the acquisition of quantifiers. Consider, for example, the quantifier most, which one might suppose to be the closest cousin of generics, since it too could be said to tolerate exceptions. Is most acquired in a comparable time frame? The answer to this question appears to be a resounding no. Preschoolers appear to find most very difficult to process, even on simple tasks. For example, suppose there are six crayons, some of which are in a box, others of which are not. The child is then asked are most of the crayons in the box? Papafragou and Schwarz (2005/2006) found that preschool children found the task exceedingly difficult; there was not even a consistent pattern to their responses. Further, even 6- to 8-year-olds who are quite old in terms of language acquisition differed from adults on the task. Papafragou and Schwarzs finding has subsequently been replicated (e.g., Barner, Chow & Yang, 2009; Leslie & Gelman, 2012, experiment 2). (Some researchers have suggested that preschoolers have some understanding of most, which manifests itself if the task is made even easier (Haldberda, Taing & Lidz, 2008) however, even if preschoolers have some competence with most, it is clear that it is minimally very difficult for them to process.) In the case of the quantifiers all and some, however, even young preschoolers are able to pass a comparable task that is, they can say whether, e.g., some/all of a small set of crayons before them are in a box or not (e.g., Barner et al., 2009). There is a sense, though, in which such scenarios do not require children to process full-blooded generalizations. They are not asked to make generalizations about crayons in general -- as a kind or class -- but only to form some beliefs about a set of crayons directly before them. The judgment that all of these crayons are in the box does not involve any expectation as to whether a newly introduced crayon will be in the box or not. It simply does not apply to any crayon that is not one of these crayons. The judgment here is not an open-ended, kind-wide, projectable generalization. (Note that generic generalizations can only be of this openended sort there is no way of even formulating a generic that pertains only to these 6 crayons.) Developmental studies of the acquisition and processing of quantifiers, however, have mostly considered only these sorts of scenario: ones in which children are presented with a small set of items and then asked to evaluate quantified statements pertaining to those particular items. Let us then pose the question: How do preschoolers fare when asked to consider open-ended, kind-wide, projectable quantified generalizations? Defaulting to the Generic In a relatively early study, Michelle Hollander, Susan Gelman, and Jon Star (2002) tested preschoolers comprehension of kind-wide quantified
11
statements. In their study, 3- and 4-year-olds and adults were asked a number of yes/no questions. The questions were either existentially quantified (e.g. are some fires hot?), universally quantified (e.g., are all fires hot?), or generic (e.g. are fires hot?). Importantly, for all three question types, their participants were asked to evaluate open-ended, kind-wide generalizations as opposed to being asked questions about a specific and limited set of items before them. The questions covered a range of properties some were had by all or almost all members of the kind, others by some but not all members of the kind, and others by no members of the kind. In the case of responses to the generic questions, the experimenters found no developmental differences in the responses; that is, 3-year-olds, 4year-olds and adults gave statistically indistinguishable responses to the questions when they occurred in generic form. In the case of the quantified questions, however, they found clear developmental changes, with all three age groups giving different patterns of responses. What was most intriguing about the data, though, were the response patterns provided by the 3-yearolds: this youngest group of participants responded in the same way regardless of whether the question was universal, existential, or generic. This pattern of responses was the same as the pattern of responses provided by the 4-year-olds and the adults to the generic questions. Thus an intriguing possibility emerged: might the 3-year-olds be interpreting the quantified statements as though they were generics? Such a possibility is precisely what one would expect to find if generics articulate cognitively default generalizations. Processing and evaluating kind-wide quantified statements would require inhibiting the generic default and engaging in a more taxing cognitive process people, especially young children, may not always do so successfully, and so would be susceptible to treating quantified statements as though they were generics. It should be noted that, even if the 3-year-olds found the quantified statements difficult to process, there are many, many possible alternative response patterns that they might have produced. They could, for example, simply have said yes to every quantified question, or no, or given responses with no clear pattern to them. (Such responses are frequently encountered in developmental work if children do not understand a task or a question, they often respond at random.) But this is not what happened here; instead they gave the same responses to the quantified questions as they gave to the generic questions which in turn were indistinguishable from the responses adults gave to those generic questions. Further, the 4-year-old children also differed from adults in their responses to the quantified but not the generic questions. In particular, their response patterns were intermediate between the 3-year-olds generic pattern, and the adults pattern. Thus, it is possible that, in some trials, the 4-year-olds were also interpreting
12
quantified statements as generics. After they had asked the children these kind-wide questions, the experimenters asked the children universal and existential questions that pertained to specific sets of items before them (e.g. are all of these crayons in the box?). Here, even the youngest children were able to answer correctly thus, they did not lack a basic competence with all and some, but rather appeared to be unable to properly process them when confronted with an open-ended, kind-wide generalization. Hollander et al.s findings were recently replicated and extended to Mandarin-speaking preschoolers (Tardif, Gelman, Fu & Zhu, 2010). The results from the Mandarin-speaking population were even more dramatic: even Mandarin-speaking 4-year-olds gave indistinguishable responses to universals, existentials, and generics. Again, the researchers found no developmental differences in answers to the generic questions the Mandarin-speaking preschoolers responded just like adults to the generic questions, but also gave this generic response pattern to the quantified questions. The results from Hollander et al. (2002) and Tardif et al. (2010) suggest that, even if young children are able to understand all and some when used in the context of a small set of items (e.g., all of these crayons), they nonetheless have difficulty processing kind-wide quantified statements, and may fall back on their interpretation of the relevant generic instead. However, in the Hollander et al. and Tardif et al. studies, children had to rely on their background knowledge of the kind to evaluate the kind-wide statements, whereas they had the all the relevant instances before them (e.g. the 6 crayons) when asked to evaluate quantified statements concerning a small set of items. To what extent is this asymmetry important? Certainly, relying on memory increases the cognitive demands associated with performing a task, and so perhaps the children only defaulted to the generic because of the difficulty of the task. However, it would constitute an even more dramatic illustration of the power of the generic default if they treated quantified statements as generics even when they do not have to rely on memory to complete the task. Susan Gelman and I (Leslie & Gelman, 2012, experiment 4) presented 4-year-olds and adults with pictures of a novel animal kind, introduced with a novel label e.g., these are gorps. In each case, there were 8 individual members of the novel kind, and we ensured that the child understood that each of the 8 was indeed a member of the kind. The 8 depicted instances were identical, except for a specific target feature e.g. curly hair that was had by only 6 of the 8. Each child was then asked a question about the novel animals: the question was either a generic (do gorps have curly hair), a specific universal (do all of these gorps have
13
curly hair?), or a kind-wide universal (do all gorps have curly hair?). The 4-year-olds, like the adult, tended to agree to the generic question, saying yes to it on 83% of trials. They also performed well with the specific universal, correctly denying that all of these gorps have curly hair on 79% of trials. However, when asked the kind-wide universal question, the 4-yearolds responded yes on 51% of trials, which was significantly higher than the rate at which they accepted the specific universal (21% of trials). The children had the pictures of the gorps in front of them during questioning, and so had the falsifying counterexamples the hairless gorps before them when evaluating the kind-wide universal. They clearly understood that not all of these gorps have curly hair, yet they were not consistently able to correctly reject the kind-wide universal, even under these circumstances. Again, these otherwise puzzling data are readily explained by the hypothesis that preschoolers sometimes substitute their evaluation of a generic for their evaluation of a kind-wide universal. The preschoolers replied that all gorps have curly hair because they did not consistently evaluate the universal as a universal, but rather on some trials assimilated it the generic gorps have curly hair. We are currently investigating this tendency at other prevalence levels, with different age groups, and with other quantifiers (in particular, some). Thus, several experiments have turned up findings that are precisely what one would expect if generics do indeed express cognitively default generalizations, in contrast to quantifiers. The data would be otherwise quite puzzling and surprising e.g., why are 4-year-olds able to use hairless gorps to consistently reject all of these gorps have curly hair but not to consistently reject all gorps have curly hair? However, the data reviewed thus far all concern preschool children. On the generics-as-defaults hypothesis, one would expect this tendency to be most pronounced in young children, but one would also expect to find evidence of it in adults. Thus, we might pose the question: do adults ever interpret quantified statements as generics? Sangeet Khemlani, Sam Gluckberg and I investigated precisely this question (Leslie, Khemlani & Glucksberg, 2011). In particular, we were interested to learn whether adults would ever accept universals such as all ducks lay eggs, despite knowing that, e.g., male ducks do not lay eggs. We found that adults have a robust tendency to do precisely this. For example, if asked to first evaluate universals such as all ducks lay eggs and then later to evaluate statements such as male ducks lay eggs, participants who correctly rejected the latter were just as likely to have accepted the universal as they are to have rejected it. One might wonder what would happen if the ordering was reversed if participants first rejected male ducks lay eggs (as the vast majority of them do), and were then subsequently asked to evaluate the universal all ducks lay eggs. Reversing the order of the blocks
14
in this way reduced the tendency to accept the universal, but it did not eliminate it: on almost 20% of trials, adults still went on to accept all ducks lay eggs despite having just rejected male ducks lay eggs (Leslie et al., 2011, experiment 3). (This percentage of trials was significantly greater than zero, and significantly greater than the percentage of trials on which participants gave other incorrect response patterns.) Again, these data are difficult to explain, unless one supposes that the adult participants were not really evaluating the universals at all but were instead evaluating the corresponding generic ducks lay eggs. 5 Meredith Meyer, Susan Gelman, and Sarah Stilwell tested adults evaluations of universally quantified statements and also found evidence that they were interpreting them as though they were generics. Importantly, they also found that the tendency to do so becomes stronger when adults are under time pressure (Meyer, Gelman & Stilwell, 2010). Converging evidence is also to be found in studies on adult reasoning. Several studies have found that adults make specific reasoning errors when dealing with kind-wide universal statements errors that would be readily explained if they were treating those universals as generics. For example, Steven Sloman (1993, 1998) investigated adults evaluations of arguments that involve the quantifier all, and found that their evaluations did not conform to the logic of universal quantification. His participants judged that, e.g., argument A is stronger than argument B, despite judging that reptiles are indeed animals:
(A) All animals use norepinephrine as a neurotransmitter; therefore all mammals use norepinephrine as a neurotransmitter (B) All animals use norepinephrine as a neurotransmitter; therefore all reptiles use norepinephrine as a neurotransmitter
Such an assessment would be readily understandable if Slomans participants were in fact interpreting these universals as generics, since reptiles maybe be naturally thought to be more likely than mammals to be exceptions to the generic animals use norepinephrine as a neurotransmitter. Similarly, Martin Jnsson and James Hampton found that adults consistently judged universals such as all ravens are black to be more likely to be true than universals such as all young jungle ravens are black despite the fact that young jungle ravens are a subset of ravens, and so if the latter universal is false then so is the former (Jnsson & Hampton, 2006). If one replaces the universals with generics, however, then this judgment is very reasonable: since generics tolerate exceptions, it is possible for ravens are black to be true even if young jungle ravens are not. Since the latter are an unknown quantity, one could reasonably be more confident in the more inclusive statement. Thus, again, if Jnsson and Hamptons participants were interpreting the universals as generics, the findings would be readily
15
explicable. Further evidence for the generics-as-default hypothesis also emerged from a recent memory study. Since much of the foregoing evidence derives from evaluation tasks, Susan Gelman and I wished to use a measure that would tap into a distinct set of behavioral responses. To do so, we tested preschoolers and adults recall of quantified statements and generics (Leslie & Gelman, 2012). We predicted that 3-year-olds and 4-year-olds and adults would all tend to recall quantified statements as generics. We indeed found a marked tendency in all three age groups to recall statements quantified with all and most as generics. Adults and 4-year-olds generally recalled some statements correctly, however 3-year-olds tended to recall even these existentials as generics. We then examined 3-year-olds recall of no statements for example no butterflies taste with their mouths. One possibility was that 3-year-olds may have simply been dropping the first word namely the quantifier and so employing a dumb strategy which would not lend any particular support to the generics-as-defaults hypothesis. If this was the case, then the 3-year-olds should recall no butterflies taste with their mouths as butterfies taste with their mouths i.e. no statements would be recalled as positive generics. However, 3-year-olds almost never recalled no statements as positive generics instead, no statements were frequently recalled as negated generics, e.g. butterflies dont taste with their mouths. Thus the youngest children recalled quantified statements as generics even when doing so required introducing a negation into the sentence, thereby adding to the growing body of evidence in favor of the hypothesis that generics express cognitively default generalizations. Conclusion The available empirical evidence would seem to support the hypothesis that generics articulate cognitively fundamental, default generalizations, while quantifiers articulate more taxing and sophisticated ones. The hypothesis pulls together and explains a range of otherwise puzzling results from a variety of experimental paradigms across diverse age groups. If we combine this idea with the observations concerning the non-quantitative nature of generic generalizations, we are led to the following conclusion: our most basic way of forming general judgments is not fundamentally driven by considerations of prevalence or cue validity as many psychologists have supposed (e.g. Rosch, 1978), but rather is sensitive to rich, content-based factors, such as whether the property being generalized is dangerous, or characteristic of the kind in question.
References Baldwin, D. A.; Markman, E.; Melartin, R. (1993). Infants ability to draw inferences
16
about nonobvious object properties: Evidence from exploratory play. Child Development 64: 71128. Barner, D.; Chow, K.; Yang, S. (2009). Finding one's meaning: A test of the relation between quantifiers and integers in language development. Cognitive Psychology 58: 195-219. Barwise, J.; Cooper, R. (1981). Generalized quantifiers and natural language. Linguistics and Philosophy 4: 159-219. Brandone, A.; Cimpian, A.; Leslie, S. J.; Gelman, S. A. (in press). Do lions have manes? For children, generics are about kinds not quantities. Child Development. Carlson, G. (1977). Reference to kinds in English. Doctoral dissertation, University of Massachusetts, Amherst. Carlson, G.; Pelletier, F.J. (1995). The Generic Book. Chicago: University of Chicago Press. Chambers, C.G.; Graham, S.A.; Turner, J.N. (2008). When hearsay trumps evidence: How generic language guides preschoolers inferences about unfamiliar things. Language and Cognitive Processes 23: 749-766. Cimpian, A.; Brandone, A. C.; Gelman, S. A. (2010). Generic statements require little evidence for acceptance but have powerful implications. Cognitive Science 34: 1452 1482. Dahl, O. (1985). Tense and aspect systems. Oxford: Blackwell. Gelman, S. A. (2010). Generics as a window onto young childrens concepts. In F. J. Pelletier (ed.). Kinds, things, and stuff: The cognitive side of generics and mass terms. (New Directions in Cognitive Science v. 12.) New York: Oxford University Press. Gelman, S. A.; Bloom, P. (2007). Developmental changes in the understanding of generics. Cognition 105: 166-183. Gelman, S.A.; Goetz, P.J.; Sarnecka, B.S.; Flukes, J. (2008). Generic language in parent-child conversations. Language Learning and Development 4: 1-31. Gelman, S. A.; Star, J.; Flukes, J. (2002). Childrens use of generics in inductive inferences. Journal of Cognition and Development 3: 179-199. Graham, S. A.; Kilbreath, C. S.; Welder, A. N. (2001). Words and shape similarity guide 13month-olds inferences about nonobvious object properties. In Proceedings of the Twenty Third Annual Conference of the Cognitive Science Society: 35257. Hillsdale, NJ: Erlbaum. Graham, S. A.; Nayer, S. L.; Gelman, S. A. (2011). Two-year-olds use the generic/non-generic distinction to guide their inferences about novel kinds. Child Development 82: 493-507. Gelman, S.A.; Raman, L. (2003). Preschool children use linguistic form class and pragmatic cues to interpret generics. Child Development 74: 308-325. Halberda, J.; Taing, L.; Lidz, J. (2008). The development of most comprehension and its potential dependence on counting-ability in preschoolers. Language Learning and Development, 4: 99-121.
17
Hollander, M. A.; Gelman, S. A.; Star, J. (2002). Childrens interpretation of generic noun phrases. Developmental Psychology 38: 883-894. Jnsson, M. L.; Hampton, J. A. (2006). The inverse conjunction fallacy. Journal of Memory and Language 55: 317-334. Lawler, J. (1973). Studies in English generics. University of Michigan Papers in Linguistics 1. Leslie, S. J. (2007). Generics and the structure of the mind. Philosophical Perspectives 21 : 375-405. Leslie, S. J. (2008). Generics: Cognition and acquisition. The Philosophical Review 117: 1-49. Leslie, S. J. (in press a) Generics and Generalization. Oxford: Oxford University Press. Leslie, S. J. (in press b). The original sin of cognition: Fear, prejudice, and generalization. Journal of Philosophy. Leslie, S. J.; Gelman, S. A. (2012). Quantified statements are recalled as generics: Evidence from preschool children and adults. Cognitive Psychology 64: 186-214. Leslie, S. J.; Khemlani, S.; Glucksberg, S. (2011). All ducks lay eggs? The generic overgeneralization effect. Journal of Memory and Language 65: 15-31. Meyer, M.; Gelman, S. A.; Stilwell, S. M. (2011). Generics are a cognitive default: Evidence from sentence processing. In Proceedings of the 33rd Annual Conference of the Cognitive Science Society. Boston, MA: Cognitive Science Society. Papafragou, A.; Schwarz, N. (2005/6). Most wanted. Language Acquisition 13: 207251. Prasada, S.; Khemlani, S.; Leslie, S. J.; Glucksberg, S. (submitted). Conceptual distinctions amongst generics. Rosch, E. (1978). Principles of categorization. In E. Rosch & B. B. Lloyd (Eds.) Cognition and categorization: 27-48. Hillsdale, NJ: Lawrence Earlbaum Associates. Sloman, S. A. (1993). Feature-based induction. Cognitive Psychology 25: 231280. Sloman, S. A. (1998). Categorical inference is not a tree: The myth of inheritance hierarchies. Cognitive Psychology 35: 133. Tardif, T.; Gelman, S. A.; Fu, X.; Zhu, L. (2011). Acquisition of generic noun phrases in Chinese: Learning about lions without an -s. Journal of Child Language 30: 1-32.
This paper will focus on bare plural generics, however English also allows for indefinite singular (e.g., a tiger is striped) and definite singular generics (e.g., the tiger is striped). 2 Cimpian et al. also found that participants were more likely to accept the generic at low prevalence levels if they were told that the property in question was distinctive i.e. that no other animals on the island had the property. One possibility is that this finding reflects an effect of cue validity on peoples judgments. (Note that the possibility that cue validity affects judgments is different from the hypothesis that cue validity is the explanation to the exclusion of the impact of danger -- of judgments of low-prevalence generics such as ticks carry Lyme disease the latter is
18
inconsistent with the evidence that is currently available, while the former remains an empirical possibility.) That is, all else being equal, people may be more likely to accept a generic if cue validity is high. Another possibility is that this effect is limited to salient physical characteristics of the kind the sorts of features that are extremely helpful to us in perceptually identifying whether something is a member of a given kind or not, even if the features are only had by some members and that the effect of distinctiveness operates by way of making it more likely that the property in question is seen as characteristic of the kind. That is, rather than there being a general effect of cue validity on peoples judgments across the board, there may be a more limited effect of distinctiveness on identifying physical characteristics. This is an open empirical question at this point in time. (Again, this is a question that would be exceedingly difficult to decide by considering theorists intuitions about a small number of examples, but that can be readily resolved in the laboratory.) 3 It is sometimes proposed in conversation that we do have such a word, namely the adverb generally. However, inserting generally into a generic produces a change in meaning. For example, books are generally paperbacks, Canadians are generally right-handed, and elementary school teachers are generally female would all appear to be perfectly acceptable statements, while their generic counterparts are reliably rejected. 4 Male and female are not words that preschoolers are generally familiar with, hence the use of boys and girls. However, we wanted to be sure that the children did not think we were asking whether birds were human girls, so we conducted a follow-up study where we asked them to evaluate generics such as mommy bears are girls. Our participants readily endorsed such items. We were also concerned that there was a miss-match between the types of main verbs in the two sentences the gender attribution items were all of the form are boys/girls, where as the characteristic items either contained action verbs or the verb to have (e.g. lions have manes). In a follow-up study, we converted all of the predicates to the form are animals that e.g. lions are animals that have manes vs. lions are animals that are boys. The same results were obtained as in the original study. 5 One possible alternative explanation of these data could be that participants were interpreting the universal quantifier as ranging over subkinds of ducks, rather than individual ducks. That is, perhaps participants understood all ducks lay eggs to mean that mallard ducks lay eggs, and mandarin ducks lay eggs, and so on for each subkind of duck. There are several reasons to believe this was not the case. As a direct measure, we asked participants to paraphrase the statements they had been asked to evaluate; if a subkind interpretation could account for the high (close to 50%) rate of acceptance of the universals, one would expect that a reasonable percentage of the paraphrases would contain some indication of this. Crucially, participants who accepted a given universal should be more likely to use subkinds in their paraphrases than participants who rejected that universal. Instead, vanishingly few paraphrases had any reference to subkinds whatsoever, and there was no difference to be found along this dimension between participants who accepted vs. rejected a given universal (Leslie et al., 2011, experiment 2B). (The same applied to another possible explanation: namely that participants were understanding the universal to be domain restricted, e.g. to only be about female ducks. The paraphrase task found no evidence of this whatsoever.) Secondly, we ran a version of the study where participants were presented with population information designed to make a subkind interpretation very
19
unnatural. For example, participants were told: There are 4.16 million ducks in the world. Evaluate the following statement: All ducks lay eggs. The use of population information should make a subkind interpretation (or any kind of domain restricted interpretation) very odd, and yet we still found a robust tendency to accept the universal statements (Leslie et al., 2011, experiment 2A). Thirdly, Susan Gelman and I tested whether adults would be more likely to accept universal statements about novel kinds if a) not all individuals have the target property, but b) the kinds were presented to them as having salient subkinds which were such that each subkind could be said to generically have the property. That is, we manipulated the availability of a subkind interpretation of a universal, but did not find any increase in the (extremely low) rate at which adults accepted the universal as a result (Leslie & Gelman, 2012, experiment 4). Thus, there does not appear to be any evidence that participants are disposed to employ a subkind interpretation of these universals, and so this is very unlikely to be the explanation of the findings. Instead, we propose that, like young children, adults have some tendency to default to the generic when evaluating quantified statements.
!"#"$%&'()*"$'%+,-%.%"/ 0( 12$23452#"(6"'-%"(
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
!Y$%&#$#.+7*!7=!./+-!#$.+,'%!?#-!-(&&7$.%1!>5!ZIX!)$#*.!"SI<MFF[T\F9! !X7$!%@.%*-+;%!1+-,(--+7*!7=!./%-%!8(%-.+7*-!-%%!P-/%$!#*1!R7$$%#(!KMTTVN3!S7/%*!KMTT[3!FLL\N3!H#?'%$!KMTU]N3! H%-'+%!KFLLU3!FLL^N3!_$+=4#!%.!#'93!KMTTVN3!Y%''%.+%$!#*1!P-/%$!KMTTUN3!#67*)!7./%$-9! !
F M
!"# $%&'()*+,*-*.+*'-/*-*'01*23)45461*4-51710)8*.591*23&10*+-*:'-*;1*:3+2<51/*;(*)<::1))461* +<%542%4:'543-8* * =94)*>4-/*3?*01201)1-5'543-*4)*:1-50'%*53*@1&4)A)*!BCDE#*2'210,*'-/*4)*)'4/*53*;1*'*5042'05451*)50<:5<01,* :3-)4)54-7*3?*'*F<'-54?410G3210'530*H,*'*I1)504:530*I*!24:>4-7*3<5*591*/3+'4-*3?*H#,*'-/*591*J:321*J,*'%)3* :'%%1/*591*+'504K*!24:>4-7*3<5*591*203210541)*'5504;<51/*53*H*1%1+1-5)*3?*I#L*=042'05451*)50<:5<01)*5'>1* 591*?30+*MH*.I8*.J8AL** * =91*)5'-/'0/*641&*'))4+4%'51)*71-104:)*53*594)*+3/1%,*&459*591*<-203-3<-:1/*!"#*5'>4-7*591* 2%':1*3?*N'%&'()OL*P5*4)*<)<'%%(*91%/*59'5*!"#*3::<0)*&91-1610*59101*'01*35910&4)1*<-;3<-/*6'04';%1)*4-* 591*I1)504:530*Q*59'5*4),*6'04';%1)*59'5*'01*-35*591+)1%61)*;3<-/*;(*'*203-3<-:1/*F<'-54?4:'543-'%* '/610;*30*F<'-54?(4-7*/1510+4-10L*=91*+3)5*)50'4795?30&'0/*610)43-*3?*591*)5'-/'0/*641&*93%/)*59'5* 4-/1?4-451)*)<:9*')*;'01*2%<0'%)*!'-/*4-/1?4-451*)4-7<%'0),*)<:9*')*N'*54710O#*:3-504;<51*201/4:'51)*'-/* <-;3<-/*6'04';%1)*53*%374:'%*?30+)*:3-5'4-4-7*591+*!1L7L,*R14+,*BCS"T*U'+2,*BCSBT*U'+2*V*I(%1,*BCCW#L* X3-)4/10,*591-,*!W#Y* * !W# =4710)*'01*)50421/* * N=4710)O*:3-504;<51)*'*201/4:'51*'-/*6'04';%1,*&94:9*4)*+'221/*53*591*I1)504:530,*'-/*591*201/4:'51*N4)* )50421/O*4)*+'221/*53*591*J:321,*01)<%54-7*4-Y* * !Z# .54710)!K#8*.)50421/!K#8* * =94)*4-510+1/4'51*%374:'%*?30+*9')*'-*<-;3<-/*6'04';%1*4-*45)*I1)504:530,*)3*591*M/1?'<%5A*6'04';%1*;4-/10* !"#*4)*4-503/<:1/*53*:3+2%151*591*%374:'%*?30+Y* * !E# [1-*K*.54710)!K#8*.)50421/!K#8* * =91*)5'-/'0/*641&*9')*'*)50'4795?30&'0/*+3546'543-Y*N54710)*'01*)50421/O*'-/*591*%4>1*&3<%/*)11+*53*;1* 71-10'%4\'543-)*3?*)3+1*)305,*'-/*591(*2'5510-*0'5910*%4>1*'/610;4'%%(*F<'-54?41/*5042'05451*)50<:5<01)L* R3&1610,*591*)5'-/'0/*641&*9')*'%)3*%3-7*01:37-4\1/*591*1K4)51-:1*3?*)3]:'%%1/*^]71-104:)T*59'5*4),* 71-104:)*)<:9*')Y* * !_# ^4-3)'<0)*'01*1K54-:5* * &9101*591*201/4:'51*4)*N>4-/*)1%1:54-7OL** `-*591*?':1*3?*45,*594)*&3<%/*-35*)11+*53*?45*&459**591*)5'-/'0/*641&*]]*'-*4-/464/<'%*:'--35*;1* 1K54-:5,*3-%(*'*)21:41)*30*>4-/*:'-*;1T*)3*45*&3<%/*;1*21:<%4'0*53*501'5*!_#*')*9'64-7*591*%374:'%*?30+*N[1-* K*./4-3)'<0)!K#8*.1K54-:5!K#8OL*I'5910,*!_#*&3<%/*)11+*53*;1*+3)5*-'5<0'%%(*501'51/*')*)4+2%(*'))1054-7* 59'5*591*>4-/*$%#&'()**45)1%?*9')*591*2032105(*3?*;14-7*1K54-:5L*=94)*+1'-)*59'5,*'5*%1')5*)3+154+1),*;'01* 2%<0'%)*:'-*01?10*/401:5%(*53*>4-/)L*P?*3-1*594->)*59'5*;'01*2%<0'%)*:3-504;<51*201/4:'51)*'-/*6'04';%1)*53* 59140*%374:'%*?30+),*594)*+1'-)*59'5*4-*)1-51-:1)*%4>1*!_#,*591)1*201/4:'51)*+<)5*;1*M5(21])94?51/*<2A*)3*')* 53*24:>*3<5*>4-/)*Q*-'+1%(*591*>4-/*&93)1*+1+;10)*)'54)?(*591*201/4:'51*4-*F<1)543-L*!J1+'-54:*5(21]
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`%(#!()&(#&%0-0!%#&'(Z30!%#,&(/"#4"(P0-!#,&(,$(%"0(!%)&.)-.(P#0/(#!(%,(H0(+-0$0--0.5(H3%($,-(%"0(+3-+,!0!( ,$(%"#!(.#!43!!#,&5(%"02(4)&(H0(%-0)%0.(0Z3#P)10&%125(!#&40(%"0(1,'#4)1($,-*!(,$('0&0-#4!(!34"()!(M%#'0-!()-0( !%-#+0.N(0&.(3+(H0#&'(+)-)1101(#&(%")%(H,%"(#&P,1P0()&(,443--0&40(,$(%"0(P)-#)H10(H#&.#&'(,+0-)%,-()*#6( + ,-*+."/01*+2"*3( a#0H0!*)&()-'30!(%")%(/0(!",31.(-0+1)40(%"0(!%)&.)-.(P#0/(/#%"(%"0(!#*+10(P#0/5()44,-.#&'(%,(/"#4"()11( '0&0-#4!5(#&413.#&'(,&0!(!34"()!(M%#'0-!()-0(!%-#+0.N5(")P0(1,'#4)1($,-*!(1#90(F>DI( ( F>D JK%#&4%F.#&,!)3-!D( F<<D;%-#+0.F%#'0-!D(
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
N
/W2:;'*,3/,7>73/19.;2.9592-'&/'-'&C.,./24/<6'5/95/9./52/B,/*,1/A/,7>73//,%)-')-'.-)*-7"-$"4-'.-)*-7"-8"$2"'9"4-%.-$"4-7:.)%#4%$4-*7."$9"$.-'#-.)%#4%$4-2*#4')'*#.7/X5/<2E&1/B,/KE95,/9:;&'E.9B&,/52/.E;;2.,/56'5/569./'-'&C.9./49>E*,./9-/56,/ ."5%#)'2/'-'&C.9./24/=@/9./*,1?/A/2-,/('-/*,'19&C/62&1/56'5/569./9./56,/:,5';6C.9(./24/7"'#!-$"4/<9562E5/569-)9->/56'5/ 569./49>E*,./9-3/.'C3/56,/&,@9('&/,-5*C/42*/=9./*,1?7/R,,/J,.&9,/!#YYZ$/42*/:2*,/19.(E..92-/24/569./;29-57/ //
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
L
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
!"#$%"&'(")*""+'),"'!"+)"+#"'!")!-'($)'./"("!01+'2"%!$1!/3"45'1%6$"!'),1)'),/!'/!'0"%"45'&$"')7'47*8 4"3"4'07%2,78!5+)1#)/#'16%""0"+)'9:;<<-'2='>?:'@'),7$6,'7A'#7$%!"-'&/AA"%"+)'"B1024"!'1%"'$+&"%' &/!#$!!/7+'),"%"C= D'' E,/!'!""0!')7'("'1'!5!)"01)/#'2,"+70"+7+='F7%'"B1024"-'/)'!""0!'/+&/!2$)1(45')%$"'),1)'),"' G/+&'!"#$%&'%(',1!'0"0("%!-'($)'' ' 9::CE/6"%!',13"'0"0("%!' ' ,1!'1'3"%5'+1)$%14'%"1&/+6'7+'*,/#,'/)'/!'A14!"-'1+&'!7'7+'!7'A7%),='E,"!"'"B1024"!'H$/)"'!)%7+645'!$66"!)' ),1)'),"')%$),'#7+&/)/7+!'7A'6"+"%/#!'1%"'+7)'I$!)'1'01))"%'7A'),"'G/+&',13/+6'),"'2%"&/#1)"&'2%72"%)5-'1!' ),"'!/024"'3/"*'%"H$/%"!='' J7)/#"'),1)'),"'!)1+&1%&'3/"*'#1+'%"1&/45'1##7007&1)"'),"!"'&1)1'@'"=6=-'9<KC'*7$4&'("'07!)' +1)$%1445'1+145L"&'1!'9:?CM' ' 9:?C'N"+'B'O)/6"%!9BCP'OB'/!'1'A/%!)87%&"%'/+&/3/&$14P' ' E,/!'/!'!)%1/6,)A7%*1%&45')%$"-'!/+#"'"1#,')/6"%'/!'/)!"4A'1'A/%!)87%&"%'/+&/3/&$14='' Q+'),"'#7$%!"'7A'2%"!"+)/+6',/!'7*+'3/"*-'./"("!01+',/0!"4A'2%73/&"!'!70"'1&&/)/7+14'"B1024"!' 7A'),"'%"4"31+)'!7%)'7A'!)1)"0"+)M' ' R!'Q'!)%"!!"&'*,"+'6/3/+6'1'%$&/0"+)1%5'0")12,5!/#!'7A'G/+&!-'G/+&!-'$+4/G"'7%&/+1%5'7(I"#)!-'1%"' 6"+"%14M'),"5',13"'0"0("%!'9:;<<-'2=>:<C=' ' S7+!/&"%'),"'7##$%%"+#"'7A'),"'(1%"'24$%14'T7%&/+1%5'7(I"#)!U',"%"='V+'),"'!/024"'3/"*-'),/!'&"+7)"!'1' G/+&'@'+10"45'),"'G/+&'*,7!"'0"0("%!'#702%/!"'144'1+&'7+45'7%&/+1%5'7(I"#)!='./"("!01+',"%"'1/0!')7' &"+5'),1)'1+5),/+6'),1)'/!'1+'7%&/+1%5'7(I"#)'/!'6"+"%14-'/+'),"'!"+!"'7A',13/+6'0"0("%!='W$)',"' "B2%"!!"!'),/!'),7$6,)'$!/+6'),"'(1%"'24$%14'#7+!)%$#)/7+'T7%&/+1%5'7(I"#)!U='X7'7+'./"("!01+Y!'7*+' !"01+)/#!-',/!'!)1)"0"+)'/!'A14!"Z'/)'&"+/"!'),"'2%72"%)5'7A'("/+6'6"+"%14')7'$)*#+%,-'.'-%,/'0#.12*3$(=' J7%'/!'),/!'1+'/!741)"&'"B1024"'/+'./"("!01+Y!')"B) KZ'),"'*7%%5'/!'+7)'I$!)'),1)'),"'!)1)"0"+)-'3"%5' !$%2%/!/+645-'#70"!'7$)'A14!"'7+'./"("!01+Y!'7*+'!"01+)/#!-'($)'14!7'),1)'),/!'!$66"!)!'I$!)',7*' 4,.15%.4('),"')%"1)0"+)'7A'(1%"'24$%14!'1!'14*15!'G/+&8&"+7)/+6'1#)$1445'/!'88'1+&'I$!)',7*',1%&'/)'/!')7' 274/#"'),/!'!"01+)/#'/+)"%2%")1)/7+-'"3"+'/+'7+"Y!'7*+'),7$6,)=' ' E,"!"'#7+!/&"%1)/7+!'!$66"!)'),1)'),"'!/024"'3/"*'015'+7)'("'),"'A/+14'*7%&'/+'),"'!"01+)/#!'7A' 6"+"%/#!='[7*"3"%-'),"'!/024"'3/"*'/!'2%/01%/45'1'3/"*'1(7$)'),"'476/#14'A7%0'7A'6"+"%/#!-'!7'),"'01/+' ''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''
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
D
!"#!$%&$!'"$()"*$*)++$,%-"$&.%-$)!#$/0)+)!1$!%$/,,%23!$&%.$!'"$#"-/3!),/++1$."+"(/3!$#!.2,!2."$%&$4"3".),#5$ 6"!$2#$3%*$!2.3$!%$!'/!$)##2"5$$ $ !"#$#%$&$'$()*'"+',-./0+$(+$1) ) 73$')#$#"-)3/+$89::$;)##".!/!)%3$%3$4"3".),#<$=."4$>/.+#%3$?.%?%#";$/3$/,,%23!$!'/!$*/#$#%-"*'/!$)3$ !'"$#/-"$#?).)!$/#$6)"0"#-/3@#$()"*<$02!$*)!'$/$-/A%.$;)&&"."3,"5$>/.+#%3<$+)B"$6)"0"#-/3<$!%%B$0/."$ ?+2./+$3%23$?'./#"#$!%$."&".$;).",!+1$!%$B)3;#$C$)3$!')#$."#?",!$!'"$!*%$()"*#$/."$)3$/4.""-"3!5$D%*"(".<$ >/.+#%3$/+#%$?%#)!";$/$-%3/;),$?.";),/!"$%?"./!%.<$!"<$*'),'$!%%B$/#$)3?2!$/$?.";),/!"$%&$)3;)();2/+#<$ /3;$4/("$/#$%2!?2!$/$?.";),/!"$%&$B)3;#5$E!"$;)&&".#$&.%-$!'"$3%*F#!/3;/.;$!#$$%?"./!%.$)3$!'/!$!'"$+/!".$ )#$/$(/.)/0+"$0)3;)34$%?"./!%.$!'/!$."+/!"#$G"#!.),!%.$!%$H,%?"<$*'"."/#$!'"$&%.-".$)#$/$#)34+"F?+/,"$ ?.";),/!"$%?"./!%.5$>/.+#%3$')-#"+&$E89I9J$,/-"$!%$."A",!$!'"$!"$%?"./!%.$)3$&/(%.$%&$!#$5J$$ >%3#);".<$&%.$"K/-?+"L$ $ EMNJO)4".#$/."$#!.)?";$ $ PO)4".#Q<$&%.$>/.+#%3<$)#$/$."&"..)34$!".-<$*'%#"$;"3%!/!)%3$)#$!'"$B)3;$%&'()*+),5$D%*"(".<$!'"$?.";),/!"$ P)#$#!.)?";Q$)#$/$?.";),/!"$!'/!$/??+)"#$!%$)3;)();2/+#<$3%!$B)3;#$C$23+)B"$?.";),/!"#$#2,'$/#$P)#$"K!)3,!Q$C$ /3;$#%$!'"$=R$%?"./!%.$'/#$!%$&).#!$0"$/??+)";$!%$!'"$?.";),/!"5$O'2#$EMNJ$*%2+;$0"$/3/+1S";$/#$EMTJL$ $ EMTJ=RE K$EK$)#$#!.)?";JJE!)4".#J$ $ *'"."$P!)4".#Q$."&".#$!%$%&'()*+),<$/#$)!$;%"#$%3$6)"0"#-/3@#$()"*5 :$O'"$2#"$%&$!'"$!"$%?"./!%.<$'%*"(".<$ -/B"#$>/.+#%3@#$?.%?%#/+$/$(".1$;)&&"."3!$/3)-/+$&.%-$!'"$#)-?+"$()"*5$O'"$"K)#!"3,"$%&$#2,'$/3$ %?"./!%.$)3$!'"$+%4),/+$&%.-$-"/3#$!'/!$X2"#!)%3#$,%3,".3)34$)!#$)3!".?."!/!)%3$/.)#"$C$)&$!'"."$)#$/$!"$ %?"./!%.<$*"$,/33%!$#2??%#"$!'/!$!'"$X2"#!)%3$%&$-./$/$B)3;$)3'".)!#$)!#$?.%?".!)"#$&.%-$)!#$-"-0".#$)#$ /$?2."+1$"K!./F#"-/3!),$&/,!5$Y%.$"K/-?+"<$*"$,/3$./)#"$!'"$X2"#!)%3$%&$*'"!'".$!"$)#$#"3#)!)("$!%$&/,!%.#$ #2,'$/#$*'"!'".$!'"$?.";),/!"$%3$*'),'$)!$%?"./!"#$"K?."##"#$/$;/34".%2#F-/B)34$?.%?".!1<$/3;$#%$%3$ #%$&%.!'5$>/.+#%3$')-#"+&$;%"#$3%!$/)-$!%$?.%();"$/3$/3/+1#)#$%&$!"<$02!$4+%##"#$!'"$!.2!'$,%3;)!)%3#$%&$ 4"3".),#$/#$."X2).)34$!'/!$!'"$?.%?".!1$)3$!'"$#,%?"$%&$!'"$!"$%?"./!%.$)#$'/;$*)!'$#2&&),)"3!$."42+/.)!1$01$ !'"$-"-0".#$%&$!'"$B)3;5$EO'%24'<$%&$,%2.#"<$!')#$)#$)3!"3;";$/#$3%$-%."$!'/3$/$.%24'F/3;F."/;1$4+%##5J$ O'"$().!2"#$%&$6)"0"#-/3@#$#)-?+"$/,,%23!$/."$+%#!$)&$*"$?%#)!$!"$FF$)3$!')#$."#?",!<$!"$)#$3%$;)&&"."3!$&.%-$ !#$5$ $ U'1<$!'%24'<$;);$>/.+#%3$?%#)!$!"$)3$!'"$&).#!$?+/,"V$W/.!$%&$!'"$/3#*".$!%$!')#$X2"#!)%3$-/1$ '/("$0""3$-"!/?'1#),/++1$-%!)(/!";$."+2,!/3,"$!%$/??+1$P)#$#!.)?";Q$/3;$!'"$+)B"$;).",!+1$!%$/$B)3;$C #%-"!')34$!'/!$6)"0"#-/3$)#$.)4'!$!%$X2"#!)%35$D%*"(".<$!')#$*/#$3%!$!'"$%3+1$-%!)(/!)%35$>/.+#%3$ $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
:
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
O
%I)'%$(5$!&,3#$%&'3&'$'6#%#)"#%#)'%3&262(6$%)"/'%G32+,#,!,"6$H%"$%#)',&%"6#'!'4'6#0%I),$%,$%#2%,64,!"#'%#)"#%"%4',!#,!% &'"4,69%2<%#)'%3&262(6$%,$%62#%&'+'/"6#%)'&'%Q%'090%"%&'"4,69%26%:),!)%"%9&2(3%2<7%'0907%42!#2&$%,$%$"+,'6#7%"64%#)'% $'6#'6!'%,$%($'4%#2%$"*%#)"#%32+,#,!,"6$%#),6-%%&'(")*'+%',(%!"6%2(#$."&#%#)',&%23326'6#$0%
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
J
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
OO
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
!B41!=(1!'?!/!-/01!>*=0/*!?'0<1(!>*=0/*.)+!'9!=(,!-=)!9')1!)4/)!C1!</9!01C0.)1!"#A%K"#R%!/-'21!=(.9:!>*=0/*(O!E)41.0! 5')410(.,Y!*'21!L'49.!/93!H.**YF;!B4.(!015/.9(!01/('9/-*+!/<<1>)/-*1,!C4.*1!E)41.0.!5')410(!*'21!/**!-'+(.F!().**!3'1(! 9')!4/21!)41!01*12/9)!01/3.9:;!
@A
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`HFHB!_!.C-<N!.C0.! -<=2G-<-.2!*-<F,501*!/0<!;2!*-(E59!.120.2=!0*!2DE12**-<F!0!*,;*2.!)G!.C2!F2<2105-a0.-)<*!.C0.!;012!E5,105*! /0<!2DE12**B!7>!;,.!05.21<0.-6259B!)<2!(-FC.!120*)<0;59!E)*-.!.A)!)&*!)E210.)1*B!)&*7!0<=!)&*:BA-.C! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
7>
!89!6-2A!V!AC-/C!_!A-55!<).!01F,2!G)1!C212B!<)1!*E255!),.!-<!(,/C!=2.0-5!V!-*!.C0.!*)(2!F2<21-/*!E12=-/0.2!AC0.!A2! (-FC.!/055!!"#$#!%&$'(%'!!E1)E21.-2*!)G!.C2!N-<=B!0<=!.C2*2!F2<21-/*!012!0//2E.0;52!0*!;).C!;012!E5,105*!0<=! -<=2G-<-.2!*-<F,501*H!b.C21!F2<21-/*!E12=-/0.2!E1)E21.-2*!)G!.C2!N-<=!;2/0,*2!.C)*2!E1)E21.-2*!(0N2!.C2-1!;20121*! *.1-N-<Fc=0<F21),*!"2HFHB!"#7%%B!)1!;2/0,*2!()*.!(2(;21*!)G!.C2!N-<=!C062!.C2!E1)E21.9!"2HFHB!"#>%%H!MC2*2!F2<21-/*! 012!*)(2AC0.!-<G25-/-.),*!0*!-<=2G-<-.2!*-<F,501*H!_G!.C-*!-*!/)112/.B!.C2<!-<=2G-<-.2!*-<F,501!F2<21-/*!-<6)N2!0! 12*.1-/.-)<!)<!.C2!10<F2!)G!F2<21-/!F2<2105-a0.-)<*!.C0.!A2!/0<!2DE12**!6-0!;012!E5,105!F2<21-/*d!C)A2621B!AC2<!0! F-62<!F2<21-/!-*!G25-/-.),*59!2DE12**2=!0*!0<!-<=2G-<-.2!*-<F,501B!A2!=)!<).!<22=!0!<2A!0<059*-*!)G!.C0.!F2<21-/B!;,.! /0<!10.C21!-<6)N2!.C2!0<059*-*!,*2=!G)1!-.*!;012!E5,105!/),<.21E01.H! e)(2!A),5=!01F,2!.C0.!-<=2G-<-.2!*-<F,501*!-<6)562!0!=-GG212<.!N-<=!)G!G)1/2!.C0<!;012!E5,105*!V!.C0.!.C29! 2DE12**!.C0.!.C2!E1)E21.9!-*!*)(2C)A!()12!-(E)1.0<.!)1!T2**2<.-05U!)1!T()=0559!/2<.105U!.)!.C2!N-<=H!_!.C-<N!.C0.! .C-*!/0<!;2!2DE50-<2=!;9!.C2!);*2160.-)<!.C0.!-<=2G-<-.2!*-<F,501*!&*+,$!&!.C2!/C010/.21-*.-/!120=-<F!"AC2<! 060-50;52%B!AC2120*!;012!E5,105*!=)!<).H!MC-*!);*2160.-)<B!/)(;-<2=!A-.C!-<=2E2<=2<.59!().-60.2=!);*2160.-)<*! /)</21<-<F!E)59*2(9!-<!0!<,(;21!)G!<),<!EC10*2*B!0//),<.*!G)1!.C2!*2<*2!.C0.!-<=2G-<-.2!*-<F,501*!0<=!;012!E5,105*! =-GG21!-<!G)1/2!"*22!Y2*5-2!-<!E12**!;B!G)1!()12!=-*/,**-)<d!*22!05*)!f015*)<!:$7$d!Y2*5-2!2.!05H!:$$Rd![10*0=0!\! S-55-<FC0(!:$$#B!:$$R%H!
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
WT
'@>/&/'"!'%'B%1'*,'>/%&'*>/!/'%!'=/$"0"*,)!:'"?/?:'(1'*%F"+4'*>/'<&/."0%*/'*,'&%+4/',D/&'!)(F"+.!',=':'/?4?:'."+,!%)&!' %+.'*>/+'>/%&'*>/'!/+*/+0/'%!'!%1"+4'*>%*'"&.)#/&$4&/)#$,"5+'"!'/;*"+0*?'\,*/'*>%*'*>"!'"!'<,!!"($/'/D/+'"='*>/'+,)+' <>&%!/'"!'M)%+*"="/.'5/?4?'I],!*'."+,!%)&!'%&/'/;*"+0*J9?'@>"!'"!'+,*:',='0,)&!/:'*>/'&/$/D%+*'&/%."+4:'%!'"*'*/$$!')!' +,*>"+4'%(,)*'B>/*>/&'"+./="+"*/'!"+4)$%&!'61+&,1'0%+'(/'F"+.-&/=/&&"+4'"+'*>/'B%1'*>%*'(%&/'<$)&%$!'0%+'(/?'
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`>1-3$">!-<-23$1"&1$>=)-$1#;-R,"0510=.$-=1-2,$1"-$4&)*.2>(='$,-6&=10),$A$&,$01$'>-,$0=$6>,1$+0=.(0,10)$1"->20-,$A$ 1"-=$1"0,$+0=-$>5$&2.(6-=13$-<-=$05$)>22-)13$!>(+'=F1$-=,(2-$1"&1$4>1"$-?;2-,,0>=,$!-2-=F1$1#;-R,"051-'$'>!=$1>$ 4-)>6-$;2-'0)&1-,$>5$0='0<0'(&+,$&1$+>.0)&+$5>267$$
89
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
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!U#*%!,1:#!*+,*!$%!-,)!:,2!?*+0:!;0)<!#3!,)0>,1!10-;:!0*:%13A!4C%:*B'0)C!,*!,!-,*7!,)<!-1%,'12!-#)&%2!*+%!-1,0>!*+,*! *+%!0)<0&0<B,1!>%>/%':!10-;!,-%5)%'3%)I!9+%!1#C0-,1!3#'>!>B:*!*+%'%3#'%!.'#&0<%!*+%!:*'B-*B'%!)%%<%<!*#!&0)<0-,*%! *+0:I!
E6
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
XY
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`M:*;3"%*,$P#+.",(*6,87-*$#*+,22#+.*"4*".*6#2#*%,&#3,6*,8.2a*.,*%811,%#*.3'.*7'(08'0#%* '2#*4'"279*#44"+"#(.79*-#%"0(#-*A*$8.*.3"%*"%*2'1"-79*$#+,&"(0*'(*"&1,2.'(.*08"-"(0*12"(+"17#*'+2,%%* 7"(08"%."+%:*N,2*#F'&17#@*.3#*#(."2#*Y"("&'7"%.*12,02'&*"(*%9(.'F*"%*$'%#-*,(*.3"%*"-#'*G#:0:@*b3,&%59*
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Â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`()%/16# 77=#Ba]\BCS=# I.1(*$&6#G=#ASPBPE=#G/&/1+'*#.&0#I$&'/7,*=#H&#J=K=#L/((/,+/1#A/0=E6#<(#/=A&$%(#?=A&C#/&I4J;;=##?/%#b$18M# _c3$10#N&+2/1*+,"#L1/**6#77=#BaQOa=# I-+/1'-+.6#G=#ABCCÊ=#</3/1/&'/#,$#`+&0*#>'1$**#U.&4).4/*F&KC4J'C.&LC#?JC?"&I"3C#4()=A&M7N9A#OOCQ[PD=# I-$5*8"6#?=#ABCCDE=#$%"&D(#(3C.(=4&@'+?'C3=#I.591+04/6#!>M#!HT#L1/**=# I-$5*8"6#?=#ASPPPE=#K"O&%+'(P+#=&(#&4%"&=4J/B&+;&.C#?JC?"&C#/&3(#/F#I.591+04/6#N`M#I.591+04/# N&+2/1*+,"#L1/**=# I+57+.&6#>=#ASPBPE=#T-/#+57.',#$3#4/&/1+'#(.&4).4/#.9$),#.9+(+,"#$&#'-+(01/&d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c7(.&.,+$&*=#-+?#(4(+#A&QQ67Q9A#B[\SD=# I+57+.&6#>=6#@#!.185.&6#R=#!=#ASPBBE=#T-/#4/&/1+'e&$&4/&/1+'#0+*,+&',+$&#+&3()/&'/*#-$%#'-+(01/&# +&,/171/,#&/%#+&3$15.,+$&#.9$),#*$'+.(#$,-/1*=#-%(./&0"1".+23"#4A&587896#[]B\[CS=#
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`@Y(-402[((abU40>( N&"B@03"?C(L0@33*( O@D#%&'(V*()*(67898:*(O@&@0"13(%3(%(Y"&>4Y(4&?4(C4,&=(1<"D>0@&E3(14&1@$?3*((T&(c*(\*(L@DD@?"@0(6@>*:*((J).8"/+ 2%).("/+*.8+,;=::K+(`@Y(-402[((abU40>(N&"B@03"?C(L0@33*'($$*(988Q97G*( O@D#%&'(V*()*'(.(WD44#'(L*(6788R:*(M@B@D4$#@&?%D(1<%&=@3("&(?<@(,&>@03?%&>"&=(4U(=@&@0"13*(G&(.);)&./+ LMA'(9KKQ9dG*( O0@@&Z@0='(-*'(6788G:*(7*.):-";*;)&."+&:+(-.-5)$);#*((`@Y(-402[(P4,?D@>=@*( O04@&@&>"]2'(\*'(.(V?42<4U'(+*(69Jdd:*(SC$@Q3<"U?"&=(P,D@3(%&>(?<@(V@#%&?"13(4U(T&?@004=%?"B@3*(T&(O*( !<"@01<"%'(W*(L%0?@@'(%&>(P*(S,0&@0(6@>3*:'(!5&>-5;)-"/+2#>-"/+*.8+7-*.).("/+I&'+DDK(M40>0@1<?[(XD,Y@0( )1%>@#"1(L,ZD"3<@03'($$*(79QKJ*( e%3D%&=@0'(V*(67899:*(T>@4D4=C'(O@&@0"13'(%&>(!4##4&(O04,&>*(T&(!<%0D4??@(;"??'(@>*'(F-E).)";+ 7-;*>%#")$"6+NO>'&5*;)&."+).+;%-+C.;&'&(#+&:+,-O/+4-.8-5/+*.8+;%-+,-':*(M40>0@1<?[(V$0"&=@0'($$*(9RJQ78J*( e@"#'(T*'(69Jd7:*(2%-+,-E*.;)$"+&:+P-:).);-+*.8+D.8-:).);-+Q&=.+!%5*"-"*(L<*M*(>"33@0?%?"4&'(N&"B@03"?C(4U( +%33%1<,3@??3'()#<@03?*( e@"#'(T*'(69JJ8:*(/Q?C$@(L04&4,&3(%&>(M4&2@C()&%$<40%*(<).(=)";)$"+*.8+!%)'&"&>%#'(L1'(9GRQ9RR*( e4DD%&>@0'(+*()*'(O@D#%&'(V*()*'(.(V?%0'(\*(67887:(!<"D>0@&E3("&?@0$0@?%?"4&(4U(=@&@0"1(&4,&($<0%3@3*( P-I-'&>E-.;*'+!"#$%&'&(#/+1R/(ddGQdJI*((( \%@=@0'(S*c*'(.(S"DC'(e*(67899:*(a&(D%&=,%=@(f,?"D"?CE[(L041@33"&=(14#$D@b"?C(%&>(14##,&"1%?"B@(@UU"1"@&1C*( S)'-#+D.;-58)"$)>').*5#+T-I)-U"6+G&(.);)I-+,$)-.$-/+0@1B/(G7GHGGF*( \4<&3?4&'(+*'(.(5@3D"@'(V*\*(6"&($0@33:*(!4&1@$?3'()&%DC3"3'(O@&@0"13'(%&>(?<@(!%&Z@00%(LD%&*(!%)'&"&>%)$*'+ !-5">-$;)I-"*(
!"#$%&'(%&)*+,*-(&.&/01234&25&/36/0&"78&91#"7/:;&:7/13$31/"/:27(&<7&=(>3217178:?@%&=(="79917&"78& A(B/2@025&)189(-%&!"#$%&'()*+",-'./'*+)'0*1,2'"3'4%/51%5)(&C23831;0/D&E23/9&F6GH:;"/:279%&$$(&IJJKLII(& !"#$%&'(&"78&M14H1%&N(%&*++L%&!#"$'6.-7"1#-)'*"'4"5.7(&C23831;/D&!H6O13&.;"81#:;&F6GH:90139(& !:7P%&=(&)IQQR-(&.7"$023"(&0*%/3"#,'8/727&"9),.%'"3':+.&"-"9+2(& !3:5@"%&A(%&F1HH1/:13%&E(=(%&S"3H927%&>(T(%&/13&A16H17%&.(%&S0:13;0:"%&>(%&U&V:7@%&>(&)*++R-(&>1713:;:/4D&.7& <7/3286;/:27(&<7&>(T(&S"3H927&"78&E(=(&F1HH1/:13&)189(-&;+)'<)/)#.7'="">(&S0:;"P2D&S0:;"P2&N7:W139:/4&F3199(& V"9139207%&F(&)*++J-(&X"31&FH63"H9&"78&C27@14&.7"$023"(&?%*1#%&'4%/51%5)'0)$%/*.7-@'ABCD%&J+K,Y(& V"OH13%&=(%&)*+JL-(&B/68:19&:7&Z7PH:90&P1713:;9(&E/.F)#-.*2'"3'(.7+.5%/':%9)#-'./'4./51.-*.7-&*)*-(& V19H:1%&B(&=(&)IQQJ-(&>1713:;9&"78&/01&B/36;/631&25&/01&A:78(&:+.&"-"9+.7%&':)#-9)7*.F)-%&GCBCD@&LJRK[QL(& V19H:1%&B(&=(&)IQQ,-&>1713:;9D&S2P7:/:27&"78&.;\6:9:/:27(&:+.&"-"9+.7%&'H)F.)I%&CCJBCD%&*K[J(& V19H:1%&B(&=(&):7&$3199&"-(&>1713:;9&.3/:;6H"/1&C15"6H/&>1713"H:]"/:279(&H)7+)#7+)-'4./51.-*.K1)-',)' L./7)//)-(&& V19H:1%&B(&=(&):7&$3199&G-(&':HH"34&SH:7/27&:9&/01&^7H4&A"7&:7&/01&^G"#"&.8#:7:9/3"/:27D&C6"H&S0"3";/13& S27;1$/9%&>1713:;9%&"78&>17813(&M/%&2*.7':+.&"-"9+2(& V19H:1%&B(&=(%&):7&$3199&;-(&_01&^3:P:7"H&B:7&25&S2P7:/:27D&E1"3%&F31?68:;1%&"78&>1713"H:]"/:27(&;+)'N"1#/%&'"3' :+.&"-"9+2(& V19H:1%&B(&=(&):7&$3199&8-(&S"3W:7P&N$&/01&B2;:"H&`23H8&O:/0&>1713:;9(&OP3"#,'0*1,.)-'./'8P9)#.$)/*%&' :+.&"-"9+2(& V19H:1%&B(&=(%&U&>1H#"7%&B(&.(&)IQ*I-(&a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a6"7/:5:;"/:27(&<7&Z(&V(&!117"7&)18(-%&!"#$%&'0)$%/*.7-'"3'?%*1#%&'4%/51%5)(& S"#G3:8P1D&S"#G3:8P1&N7:W139:/4&F3199%&$$(&LK*R(&
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`H1a+)D.%(#"-'.;'?<)-*'!"#$%"$&6)=S()>HI<+) 3%F))!%6-%#()/+);>JJL?+).%'"#3&2)4B#13F"&#31)B#)P7C%H.$&D3&#E)/#"C$BF"+)O#),+T+)4"F-1B#)"#@)Y+)U+) 9%--%3&%F);%@1+?)C<&'H&#&()7'T..MB)4$&2"EBN)b#&A%F1&37)BD)4$&2"EB)9F%11()CC+)IIJHILK+) [&-W&#1B#()c+);>JJ>?+)@1%*)&:')#'1<&'@&0"#1)7:'.;'H&#&()7'R.%#'9<(":&:+)9$+X+)3$%1&1()b#&A%F1&37)BD) !"11"2$61%331()/'$%F13+) [&-W&#1B#()c+);>JJL?+)P$%).%'"#3&21)BD)3$%)4B''B#)TB6#)U)#*+)O#),+T+)4"F-1B#)"#@)Y+)U+)9%--%3&%F);%@1+?) C<&'H&#&()7'T..M+)4$&2"EBN)4$&2"EB)b#&A%F1&37)9F%11()CC+IKIHIJK+) ) )
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition

Sarah-Jane Leslie
Princeton University
Introduction: The Puzzle of the Generic Philosophers of language sometimes theorize about obscure sentences, rarely found in everyday discourse, or about the applicability of sentences to strange, unrealized possible circumstances. Ordinary speakers of English often lack clear intuitions about the truth or falsity of these sentences, and so the theorizing is based on the judgments of the select few whose intuitions are sufficiently honed. Other times, however, one need not venture into obscurity to find a phenomenon worthy of attention. If one so chooses, one may ask any English speaker whatsoever whether sentences such as tigers have stripes, birds lay eggs, or mosquitoes carry the West Nile virus are true, or whether mosquitoes dont carry the West Nile virus and birds are female are false. The response is invariably confident and immediate. These simple examples present a puzzle with no simple solution. That tigers have stripes is true might not seem particularly troubling; even if some tigers lack stripes, most tigers sport them. Examples such as birds lay eggs, though, indicate that it is not necessary for the truth of these sentences that the majority of the kind in question satisfy the predicate; most birds do not lay eggs, yet birds lay eggs is true. Closer
A special thanks is owed to the following people for their invaluable comments on this essay: Mark Johnston, Paul Pietroski, Gideon Rosen, and two anonymous reviewers. I am also very thankful to the following people for their help with the essay: Paul Benacerraf, Noam Chomsky, Delia Graff Fara, Michael Fara, Gilbert Harman, Susan Gelman, Lila Gleitman, John Hawthorne, Thomas Kelly, Philipp Koralus, Barry Lam, Richard Larson, Ernie Lepore, Alan Leslie, Sandeep Prasada, Daniel Rothschild, and Brett Sherman.
Philosophical Review, Vol. 117, No. 1, 2008 DOI 10.1215/00318108-2007-001 2007 by Cornell University
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
reflection also indicates that majority satisfaction is not sufficient either; most mosquitoes99 percent, in factdo not carry the West Nile virus, yet mosquitoes dont carry the West Nile virus strikes us as obviously false. Why should mosquitoes carry the West Nile virus seem so patently true, and mosquitoes dont carry the West Nile virus seem so clearly false? There are ninety-nine times as many mosquitoes without the virus than there are with it. Similarly, why should birds lay eggs strike us as true, while birds are female does not? After all, the egglaying birds are, perhaps even necessarily, a subset of the female birds. These everyday sentences present the philosopher of language with a challenge: what is the basis for this odd distribution of truth-values? The truth conditions of this class of sentences, known as generics, have been the subject of much debate but little successful theorizing. The puzzle of the generic does not end with its truth conditions, however. While generics have proved nigh intractable for theorists, they turn out to be a cakewalk for language learners. Young children engaged in learning their native language grasp and produce generics far more quickly and readily than they do explicitly quantified sentences. Explicit quantifiers such as all, every, and even some are comparatively more difficult for young children to master. These quantifiers for which we can provide precise and concise truth-theoretic definitions pose more of a challenge to language learners than do these strange and elusive generics. What is more, there is no generic operator signifying that a particular sentence is a generic, and therefore possessed of these odd truth conditions. We do not say gen tigers have stripes in the way we might say most tigers have stripes; there is no pronounced operator associated with generics. This is no quirk of English either; no known language has a dedicated generic operator. That this is a feature shared so widely is another fact that cries out for explanation. Could this really be a mere coincidence? The absence of a dedicated generic operator ought to make the acquisition of generics more difficult. Children are not even provided with an explicit object of study; there is no particular part of the sentence with which they could learn to associate the meaning of generics. Quantifiers such as all and some are not only truththeoretically simpler; they are actually articulated! If the challenge to the philosopher of language is to explain our judgments of the truth-values of generics, then the challenge to the cognitive scientist
2
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition is to explain how generics could be easier to master than the explicit quantifiers. This essay aims to meet both challenges. In doing so, it develops an approach to semantics that places more emphasis on cognitive considerations than do more traditional approaches. My approach aims first and foremost to answer the question: how do human beings actually understand, produce, and acquire their languages? What are the cognitive processes that underwrite these abilities? It is with these questions in mind that I approach generics, and so the account I offer is an example of this way of doing semantics. To the extent that my account of generics is successful, I take it to provide support for the larger project of cognitively oriented semantics.1 I will begin by discussing the logical form of generics. My main concern in this essay will not be with the nature of their logical forms, so the discussion will be rather brief, comprising only those issues that are directly relevant to the ensuing argument. I then go on to consider prominent existing accounts of generics and argue that each faces serious difficulties. My positive account begins with the section entitled Troublesome Generics and is the topic of the remainder of the article. The Logical Form of Bare Plural Generics The examples cited above were all bare plural generics. That is, the item in the sentence that received a generic interpretation was a plural common noun with no determiner or quantifier preceding it: mosquitoes, birds, and tigers. Genericity manifests itself in other constructions too; indefinite singulars (e.g., a tiger is striped), and definite singulars (e.g., the tiger is striped) can both receive generic interpretations in English. To simplify matters, for the remainder of this essay I will focus solely on bare plural generics. There are interesting questions surrounding the other two genericsthe indefinite singular and
1. We could make a preliminary distinction between two tasks, which semanticists and cognitive scientists might usefully divide up between them. The first is the familiar one of articulating semantic structure in sentences of a given language so as to enable a recursive assignment of truth conditions, or alternatively propositions, to all the sentences of that language. The assignments are understood as meaning giving, and the language in question can thus be taken to be a vast paring of sentences and meanings. The second task is to explore the actual cognitive abilities of speakers in order to explain just how it is that they are able to speak the language so understood. It is this second task, as it applies to the case of generics, that I am mostly concerned with here.
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
the definite singularbut a discussion of them will be beyond the scope of this essay. In the interest of concise terminology, I will henceforth use the term generic to mean sentence containing a generically interpreted bare plural. Thus I am to be understood as theorizing about this more restricted class of sentences. Not all bare plurals receive generic interpretations in English. Consider, for example: Tigers are striped. Tigers are on the front lawn. The interpretation of the first sentence is intuitively quite different from the interpretation of the second. Whereas the first expresses a claim about tigers in general, the second concerns only some particular group of tigers. In cases such as this, we say that the bare plural receives an existential interpretation, in contrast to a generic interpretation. This essay will be concerned only with bare plurals that are interpreted generically. For extensive discussion of existential interpretations of bare plurals, see Carlson 1977 and Kratzer 1995.2 While the semantic interpretation of generics is a matter of controversy, the logical form of bare plural generics is quite well understood. It is generally agreed amongst linguists that that there is a twoplace operator Gen that functions as an adverb of quantification in the sense of Lewis 1975. That is, the operator Gen is unselective, binding any variables that are free in the sentence. Adverbs of quantification include usually, generally, typically, always, sometimes, and so on. The logical forms of sentences containing adverbs of quantification
2. Two tests are useful in distinguishing existential from generic interpretations. The first concerns entailment: the existentially interpreted common noun can be replaced by a common noun that picks out a superset of the former, without change in truth-value; we can replace tiger with animal salva veritate in tigers are on the lawn. This is not so when the interpretation is generic: the truth of tigers are striped does not entail that animals are striped, despite the truth of tigers are on the front lawn entailing that animals are on the front lawn. The second test concerns interactions with adverbs of quantification: we cannot insert an adverb of quantification into a sentence receiving the existential interpretation without inducing a significant change in meaning. If a sentence receives the generic interpretation, however, then we can. Usually is a good candidate to be inserted into sentences such as tigers are striped, though some troublesome generics, such as mosquitoes carry the West Nile virus, become false with the insertion of usually. However, in those cases we can insert sometimes with a minimal change of super ficial meaning. Tigers are sometimes on the front lawn, however, has a distinctly different meaning from (2).
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition are very often given by a tripartite structure (Heim 1982; D. Lewis 1975; Kamp 1981). A tripartite structure consists of an operatorusually a quantifiera Restrictor, and a Scope: Quantifier x, . . . ,z [Restrictor x, . . . ,z] [Scope x, . . . ,z] The Restrictor specifies the domain over which the variables range, and the Scope specifies the property that is attributed to the relevant members of the domain. For example, we can represent generally tigers are striped as: Generally x [Tiger(x)] [Striped(x)] If a sentence contains an adverb of quantification, then any variables in the Restrictor that would be otherwise free are bound by that adverb. But how do those variables come to be bound if there is no explicit quantificational adverb in a sentence? I will adopt the now commonplace proposal that the generic operator Gen occurs in a sentences logical form just in case there is no explicit quantificational adverb in that sentence (Heim 1982; Kratzer 1995). Gen binds free variables in the Restrictor when no articulated adverb of quantification, such as generally, occurs in the sentencethat is how Gen occurs in the logical form of sentences such as tigers are striped, ravens are black, mosquitoes carry the West Nile virus, and so on. Gen can thus be thought of as a default variable-binding operator, arising only when the sentence does not contain an explicit quantificational operator. Since the sentence tigers are striped contains no explicit quantificational operator, its logical form will involve the default operator Gen. The logical form of tigers are striped will be represented as: Gen x [Tiger(x)] [Striped(x)] More generally, the logical forms of the generics we are concerned with in this essay will belong to the following schema:3
3. The astute reader may have noticed that not all nonexistentially interpreted bare plural subjects can be readily assimilated into a view that posits a generic operatorhow are we to treat sentences like dodos are extinct? This is not an existential statement concerning just some dodosit is more general than thatbut it is also not readily understood as containing a quantifier ranging over individual dodos. Here, Krifkas (1987) distinction between I-generics (indefinite generics) and D-generics (definite generics) is particularly useful. D-generics predicate properties of
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
Gen x1 . . . xn [Restrictor(x . . . xn )] y1 . . . yn [Scope(x1 . . . xn ,y1 . . . yn )]4 I know of no contemporary theorists that do not take this schema to underwrite the logical form of generics. There is, however, much dispute about how Gen is to be semantically interpreted, as we will see in the next section. Thus far I have been speaking of Gen appearing in the logical form of sentences, without being explicit about how we should understand this notion of logical form. From here on in, I will understand the relevant notion of logical form to be the linguists notion of LFa psychologically real level of linguistic representation. Thus I will assume that the introduction of Gen is one of the operations the language faculty performs when formulating a sentences LF. If one is unfamiliar with the notion of LF, the important point is simply that the emergence of Gen will be considered to be a psychologically real process.5
their subjects that do not hold of particular members of a kind but only of a kind itself. I-generics predicate properties of their subjects that can hold of particular members of the kind. Examples of D-generics include the Dodo is extinct, dogs are widespread, and dinosaurs died out many years ago, and examples of I-generics include all the generics we have encountered in the essay until now. D-generics have received comparatively little attention in the literature. Their subjects are taken to be referring terms that denote kinds, and little more is said than that. The LF of dodos are extinct is simply Extinct(d), where d picks out the kind Dodo. I will not pursue the matter of D-generics further. 4. If a variable is free in the Scope but not in the Restrictor, that variable will be bound by an existential quantifier. Thus for Cats have a tail we first obtain: [Cats(x)] [Tail(y) & Have(x, y)] Since y is free only in the Scope, the Scope must undergo existential closure. x, however, is free in both locations and so will be bound by Gen: Gen x [Cats(x)] y[Tail(y) & Have(x, y)] For an extensive discussion of this phenomenon, see Kratzer 1995. 5. If one is uncomfortable with the claim that these tripartite structures arise at LF, but is nonetheless content with the claim that they arise at some psychologically real level of linguistic representation or other, then one may reinterpret my references to LF as references to this level of representation. Nothing I say depends on this level being LF; that the level in question is psychologically real will be sufficient for the following discussion. For expository simplicity, I will follow Heim (1982) and describe the level as LF. I must emphasize that the important features are just that this level should be linguisticrather than conceptualand be psychologically realized. Nothing more specific to LF needs to be assumed.
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition Existing Accounts of Generics Although there is much consensus in the literature about the general outline of the logical form of generics, there is little or no consensus about the meaning of Gen. Gen is clearly not equivalent to any of the classical quantifiers such as all, most, or some (Carlson 1977). All is too strong; generics can remain true in the face of exceptions, while universals cannot. Most is both too weak and too strong; most chicken are female, yet chickens are female is false, while on the other hand mosquitoes carry the West Nile virus is true, though most mosquitoes do not carry the virus. Some is evidently too weak since certainly some chickens are female, but again the corresponding generic is false. Some adverbs of quantification may seem more promising as glosses: generally, usually, and typically, are all successful glosses of many generics: Tigers have stripes. Generally, tigers have stripes. Dogs bark. Typically, dogs bark. Cats have whiskers. Usually, cats have whiskers. In each case, the second sentence is a good paraphrase of the first. Since Gen appears to share properties with adverbs of quantification (such as being unselective), it would be appealing to locate it semantically among these other adverbs. However, there are many generics that cannot be paraphrased using generally, usually, or typically: Bees are sterile (false). Generally, bees are sterile (true).6 Books are paperback (false). Typically, books are paperback (true).7 Sharks attack bathers (true). Usually, sharks attack bathers (false).
6. Generally, bees are sterile; only the queen and a few males are capable of reproducing. 7. Typically, books are paperbacks; only a small percentage of books are hardcover.
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
Similar difficulties undermine the suggestion that Gen functions as a restricted universal quantifier meaning roughly all normal. The most sophisticated versions of this approach are due to Asher and Morreau (1995) and Pelletier and Asher (1997), who maintain that Ks are F is true if and only if for each individual K, the most normal worlds for that K, according to the ordering base, are such that that K is F. (The ordering base of worlds contains a contextually determined similarity metric on worlds.) Put in more intuitive terms, this amounts to the claim that the generic is true if and only if each K would be F under the circumstances that are most normal for that K. This has some appeal, but quickly runs into difficulty with gender-specific generics such as ducks lay eggs, and lions have manes. There is nothing abnormal about being a male duck or a female lion, yet none of these creatures would have the properties needed to make the universal true except under exceptionally abnormal circumstances. Pelletier and Asher suggest restricting the quantifier further in these cases, claiming that the only Ks that are relevant to the truth of ducks lay eggs are the female ducks, and so the truth of that generic is determined by whether female ducks lay eggs under the most normal conditions. They suggest that this sort of domain restriction is acceptable as long as the restricted domain is a subkind of the kind in question; female ducks constitute a subkind of ducks, hence the truth of the above generic. This should leave us wondering why ducks are female is not true; why are we blocked from the appropriate subkind restriction in that case, but not the other? Alternatively, why is ducks dont lay eggs not also a true generic? If female ducks constitute an appropriate restriction, surely male ducks do likewise; why then is ducks dont lay eggs not also rendered true via restriction to the male ducks? We should also wonder what prevents bees are sterile from being a true generic; do worker bees not constitute a subkind of bees? Or for that matter, why do we not accept marsupials are kangaroos? Surely kangaroos count as a natural subkind of marsupials. These accounts also encounter difficulties with generics such as mosquitoes carry the West Nile virus. As we have noted, less than 1 percent of mosquitoes actually carry the virus, so it is quite an unusual and perhaps abnormal property for a mosquito to have. Certainly one should be reluctant to say that all normal mosquitoes carry the virus. Similarly, one would need quite an unusual ordering relation of worlds as more or less normal to have it turn out that the most normal state of affairs for a mosquito is one in which it carries the West Nile virus. We should also note that there is no natural subkind with which to restrict
8
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition the domain here. I would suggest that no noncircular domain restriction is forthcoming for this sentence and its like. Ariel Cohen (1996) proposes a rather different account, in which he asks us to consider the set Alt(F) for a property F, where Alt(F) is the contextually supplied set of alternatives to F. The notion of a set of alternatives to a predicate is familiar from theories of focus semantics (see Krifka 1995). The notion is an intuitive one: the alternative to, for example, being male is being female ; the alternatives to walking to the department are driving to the department, taking the shuttle to the department, and so on. The alternatives to F need not be incompatible with one another, though in many cases they will be. As Cohen understands the set of alternatives, F itself is included among Alt(F). Cohen also makes use of the alternatives to being a K; for example, the relevant alternatives to being a cat may be being another kind of mammal, and the relevant alternatives to being a mosquito may be being another kind of insect. With the notion of alternatives in place, Cohen suggests a distinction between absolute and relative generics:
Absolute generics: Ks are F is true iff the probability that an arbitrary K that satisfies some predicate in Alt(F) satisfies is F is greater than .5 Relative Generics: Ks are F is true iff the probability that an arbitrary K that satisfies some predicate in Alt(F) satisfies is F is greater than the probability that an arbitrary member of Alt(K) that satisfies some predicate in Alt(F) satisfies is F.
We have the restriction satisfies some predicate in Alt(F) in order to deal with, for example, reproduction-related generics. The alternatives to give live birth will be lays eggs, engages in mitosis, and so on. Since the males of a species will not engage in any of these alternatives, they are not considered in the evaluation of such generics as cats give live birth. This account avoids the prediction that cats are female will also be true since the alternative to is female is is male, and so a broader range of Ks are considered in its evaluation. I will refer to the Ks that satisfy some predicate in Alt(F) simply as the relevant Ks. Absolute generics, then, are true if more than half of the relevant Ks satisfy the predicate in question; tigers have stripes would be an example of a true absolute generic since most tigers are striped. Relative generics are true if the rate of F-ness among the relevant Ks is greater
9
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
than the rate of F-ness among the relevant non-Ks.8 Mosquitoes carry the West Nile virus is thus a true relative generic; mosquitoes are far more likely than other insects to carry the virus. It is important to note that there is no explicit marking of the distinction between absolute and relative generics; there is no independent way to tell whether a generic should be interpreted as absolute or relative. Generics are not ambiguous; Cohen does not claim this, and it would not be right to claim it. For example, Dogs have four legs does not have two readings, one of which is false and the other of which is true. (It is true as an absolute generic since most dogs have four legs. It is not true as a relative generic, though, since the incidence of four-leggedness among dogs is lower than among the likely alternatives to dogs, which are, presumably, other mid-sized mammals. There are almost no three-legged animals in the wildthey tend to meet quickly with sticky endsbut many pet dogs live to old age despite being short one leg.) Similarly mosquitoes carry the West Nile virus is not intuitively ambiguous between a false absolute reading and a true relative reading. If generics are not ambiguous, then each generic must either be absolute or relative. Cohen gives us no procedure to determine whether a generic is absolute or relative, but there is an implicit assumption in his discussion: namely, that if a generic is true on one of the classifications, then it belongs to that classification, and is thereby true simpliciter. A generic belongs to whichever class makes it true; if neither class makes it true, then the generic is false. Dividing generics up in this way is ingenious and captures a much broader range of data than offered by the other accounts above. As it stands, however, it is too permissive and renders too broad a range of generics true. For example: Books are paperbacks. Chickens are female.9 Bees are sterile. Dogs have three legs. These are all intuitively false, but are rendered true by Cohens accountthe first three being true absolute generics, the fourth being
8. Here I am conflating frequency and probability and will continue to do so. Nothing interesting hangs on the distinction here. 9. I am told that approximately 80 percent of chickens are hens.
10
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition a true relative generic since a three-legged animal is very likely to be a dog. Cohen recognizes that his account is too permissive and introduces his homogeneity constraint to rein in the account. The homogeneity constraint requires that the above probability conditions (exceeding .5, or exceeding that of the arbitrary alternative to the kind) should hold in all salient partitions of the kind. We should, of course, wonder just what constitutes a salient partition of a kind. In some cases it is rather intuitive: books are paperbacks is false because the kind, Book, is naturally partitioned into paperback books and hardcover books. Being a paperback book occurs with the required probability in only one of the partitions (namely among the paperbacks), and not in the other (i.e., among the hardcovers).10 Thus books are paperbacks is false because it does not meet the homogeneity constraint. Similarly Chickens are female is false because there is a salient partition into the female chickens and the male chickens; the generic thus violates homogeneity. The homogeneity constraint, however, comes dangerously close to ruling out all generics that have any exceptions whatsoever. Surely any predicate is F makes salient a simple two-part partition into Fs and nonFs; why does the occurrence of the predicate itself not suffice to evoke this partition? The rate of F-ness among the non-Fs will trivially be zero; if such a partitioning is salient and permissible, generics would collapse into universals. Cohen clearly cannot rule out simple two-way partitions of the kind into the Fs and the non-Fs since that is exactly what occurs in the two generics discussed above. Presumably, the explanation of why chickens are female violates homogeneity in virtue of the partition of the kind into female chickens and nonfemale chickens (i.e., male chickens) must rely on the idea that gender is an inherently salient partition. It is not that it is made salient by the use of the predicate is female; if this was the explanation, then any predicate might make salient a
10. In Cohen (2004) he claims that the relevant partitions of the kind Book are into mystery books, encyclopedias, biographies, and so on. Homogeneity is violated because almost no encyclopedias are paperbacks, and so it is not the case that all salient partitions of the kind Book consist of more than 50 percent paperbacks. I must admit thatwhile I see the appeal of the homogeneity constraintI cannot understand why Cohen thinks this is the correct explanation of the falsity of books are paperbacks. Even if encyclopedias (and all other types of books) were uniformly, say, 80 percent paperbacks, surely books are paperbacks would still be false. It is the existence of the nontrivial number of hardcover books that matters for the falsity of books are paperbacks, not how publishing houses allocate that number across genres. Homogeneity does yield a plausible explanation of why the generic is false, though, as long as we understand the partition as into hardbacks and paperbacks.
11
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
corresponding partition into the Fs and the non-Fs. But if gender is an inherently salient partition, then we would expect that lions have manes, peacocks have big blue tails and cardinals are red should all be false generics. Cohen also argues that bees are sterile is false because there is a salient partition of bees into workers, drones, and queens, and almost none of the queens are sterile. Unfortunately, the same constraint will make the true generic bees reproduce false, as none of the workers reproduce. Homogeneity also does not offer a way to rule out dogs have three legs as a true relative generic without also ruling out dogs have four legs as a true absolute generic; any partition that is not homogeneous with respect to the first will also be nonhomogeneous with respect to the second. The homogeneity constraint faces difficulties. The notion of a relative generic is also troublesome. The category of relative generics was introduced to deal with sentences like mosquitoes carry the West Nile virus. The incidence of West Nile virus-carrying is higher among mosquitoes than among arbitrary insects, so it is a true relative generic. We should worry, though, that the possibility of relative generics opens the door too wide. Cohens accountif it is able to capture the truth of dogs have four legsmust treat dogs have three legs as a true relative generic. Similarly, any time, say, a handicap or birth defect D is more frequent among one kind than all others, then Ks have D will be a true relative generic. Thus, for example, humans are one legged, or autistic, or blind, would be true relative generics by Cohens lights. Cohens theory also erroneously predicts that certain generics containing negation would be true: none of us think that mosquitoes dont carry the West Nile virus is true, nor do we think that sharks dont attack bathers is true, yet these should be true absolute generics on his view. It is not even clear that we ought to treat troublesome generics as relative. Let us consider mosquitoes carry the West Nile virus. We can imagine a world in which fleas also carry the West Nile virus and at a significantly higher rate than that of mosquitoes. At this point, Cohen still predicts the truth of mosquitoes carry the West Nile virus, since mosquitoes may still carry it at a rate higher than that of the arbitrary insect. Imagine, though, that fleas came to drastically outnumber all other insects; the rate with which the arbitrary insect carried the West Nile virus would likewise increase. (The rate at which the arbitrary insect carries the virus is determined by dividing the number of insects with the virus by the total number of insects. Increasing the flea population while
12
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition keeping its high rate of virus carrying constant will, of course, increase this average.) If the flea population swelled to sufficiently immense proportions, then the rate at which the arbitrary insect carries the virus would exceed the rate at which mosquitoes carry the virus. Thus whether mosquitoes carry the virus at a rate higher than that of the arbitrary insect depends on these facts about relative population size. It seems, however, that whether it is true that mosquitoes carry the West Nile virus does not depend on how many fleas there are relative to other insects. One does not need to know how many fleas there are to know whether mosquitoes carry the West Nile virus. Troublesome Generics Let us consider the counterexamples that plague existing accounts. Many generics of the form Ks are F11 hold of a large majority of the Ks, and their exceptions are often in some way defective. Take, for example, tigers have four legs: it is only tigers with birth defects, or that have met with unfortunate accidents, that do not have four legs. These generics are readily handled by existing accounts. As we saw, however, there are a number of examples that spell difficulty for some or all of these accounts. These troublesome generics are either true, despite a majority of the Ks failing to satisfy the predicate, or false, despite the majoritys satisfying it. It would be desirable to understand the circumstances under which these troublesome generics arise. Is there a reasonably narrow range of features that is sufficient to identify potentially troublesome generics? I claim that the bulk of these troublesome cases fall into one of two categories: Type A Bees reproduce. Bees are sterile (false).
11. Of course there are many generics that are not of this form because, for example, a generically interpreted bare plural occurs in object position. To achieve full generality, we should speak in terms of generically interpreted material in the restrictor satisfying the scope, but it is cumbersome to speak in this way. In the interests of simplicity, the discussion here will be restricted to syntactically simple generics of the form Ks are F, where Ks is the only generically interpreted material in the sentence. A successful treatment of this restricted class can, I believe, be readily extended to the broader class of sentences, however such a discussion is beyond the scope of this essay.
13
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
Birds lay eggs. Chickens are female (false). OrangeCrusher 2000s crush oranges (even if no OrangeCrusher 2000 has ever been used). Firemen fight fires (even if there are no fires). Type B Sharks attack bathers. Mosquitoes carry the West Nile virus. Tigers eat people. Rottweilers maul children. The patterns in the A group are best explained by first noticing that for certain types of kinds, including biological, artifact, and institutional kinds, our background knowledge leads us to have certain strong expectations concerning them. For example, we expect that biological kinds will exhibit certain characteristics, or else face extinction. The most obvious expectation of biological kinds is that reproduction will be possible. It is common knowledge that for an animal kind to survive, certain conditions must be met. We generally suppose that: There must be both male and female members of the species. There must be a manner in which this reproduction and subsequent gestation occurs. There must be adult members of the species. The young must be nourished in some way. I suggest that generic statements that express determinate versions of these claims are true even if there happen to be a large number of exceptions. Our background assumption is that these claims are true in virtue of the kind under consideration being a successful biological kind, so it takes a very large proportion (almost 100 percent) of exceptions for us to give up these claims. Hence the apparent truth of the following, despite a large number of exceptions: Bees reproduce. Chickens lay eggs. Cats suckle their young.
14
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition Conversely, generics that conflict with these presuppositions are very difficult to render true: Bees are sterile. Turtles die in infancy. If, however, the number of exceptions approaches 100 percent, the statement will be taken to be true; mules are sterile is such an example. Similar effects can be seen with artifact and institutional kinds. We presume that artifacts will have a function, so generics characterizing that use are tolerant of exceptions. OrangeCrusher 2000s crush oranges is such an example; the sentence strikes us as true, even if none have ever been plugged in and useddue, perhaps, to lack of demand for crushed oranges. Similarly, firemen fight fires would remain true even if the climate had become so humid that there were never any fires to fight. Let us turn now to Type B generics. These generics are true even though the vast majority of Ks are exceptions to the generalization. The examples in the B group have something in common: in all of them, the sentence attributes harmful, dangerous, or appalling properties to the kind. More generally, if the property in question is the sort of property of which one would be well served to be forewarned, even if there were only a small chance of encountering it, then generic attributions of the property are intuitively true. We see a similar phenomenon elsewhere in our judgments: compare the number and regularity of times one must worry to be a worrier versus the number of murders one must commit to be a murderer.12 It should be evident upon reflection that the criteria than govern troublesome generics reflect our psychology. This is especially evident in the B group: the more striking, appalling, or otherwise gripping we find the property predicated in the generic, the more tolerant the generic is to exceptions. For the sentences in the B group, it appears that it is sufficient for the truth of the generic that just some of the members of the kind have the property in question. Our judgments about the A group too, I claim, depend on our background assumptions and expectations. If a generic pertains to certain kind-relative expectationsreproduction
12. Interestingly, Larson (1998) suggests that there is a generic operator present in the logical form of attributions like John is a worrier and John is a murderer. If this is correct, then an account of the disparities in the required regularities would follow from the account of generics that I am proposing.
15
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
for animal kinds, function for artifact kinds, and so onthen it too is very tolerant of exceptions. As a first-pass hypothesis, we might say that if a generic expresses the manner in which an animal reproduces and nurtures its young, then it is again sufficient for its truth that some members of the kind satisfy the predicate. If a generic expresses the function of an artifact, then it is enough that this be the intended function of the artifact. I will call these specific kind-relative expectations characteristic dimensions of a kind, and will return to the relation between characteristic dimensions and generics in more detail below. If the above classification of troublesome generics is appropriate, then this suggests that our goal of understanding the quirky behavior of generics might be furthered by looking carefully into some of the relevant psychological literature. If the relation between objective statistical frequencies and the truth-values of generics is in some way mediated by psychological considerations, it would be useful to determine exactly what role these considerations are playing. As it turns out, the above observations are not the only reasons to consider the psychological literature when studying generics. The suggestion about the importance of our psychology for the understanding of generics is reinforced when we look at the near-paradoxical nature of the acquisition of generics. Before considering generic acquisition, however, let us take a moment to consider another potential class of troublesome generics. Dutch Sailors and Bulgarian Weight Lifters: An Aside concerning Logical Forms One might worry that my classification of troublesome generics leaves out an important group. What of generics such as Dutchmen are good sailors and Bulgarians are good weight lifters, which are supposed to be true, even though few Dutchmen are sailors, and so a fortiori few Dutchmen are good sailors (and similarly for Bulgarian weight lifters). Historically, these observations form the basis of the Port-Royal Puzzle, which notes the invalidity of the following argument: Dutchmen are good sailors. So, Dutchmen are sailors. This puzzle has traditionally been thought to be connected with genericsthat it is one of many strange features of generics that only a few Dutchmen need to be good sailors for the premise to be true, but far
16
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition more would need to be sailors (good or bad ones) for the conclusion to be true. On closer investigation, however, it is less than obvious that this puzzle has to do with the truth conditions of generics rather than the interaction of comparative adjectives and agentive nominals. Consider the following nongeneric version of the Port-Royal Puzzle: Barry is a good dancer. So, Barry is a dancer. This argument is also invalid. For the premise to be true, it is sufficient that Barry exhibit talent when he dances. For the conclusion to be true, however, Barry would need to have organized his life so that dancing is his chosen vocation, or at least one of his avocations (Larson 1998). The premise may well be true even if Barry is a philosopher who dances once a year, but the conclusion cannot be true under those circumstances. This suggests that the puzzle has more to do with adjectives and agentive nominals than with generics. (An agentive nominal, for our purposes, is a nominal derived from a verb, usually by adding er to the stem.) Richard Larson (1998) considers this nongeneric version of the Port-Royal Puzzle and offers a solution to it. Following Condoravdi (1989), he proposes that, when agentive nominals are modified by adjectives, the agentive nominal occurs as a verb in the Restrictor of the sentences logical form. It does not occur as a noun in the Scope. The logical forms of sentences such as Barry is a good dancer are generalizations over events in which Barry dances: Q e [C(e, Barry) & dancing(e, Barry)] [good(e, C)], where C(e, Barry) denotes a contextual restriction to relevant events involving Barry. Barry is a good dancer means, on this account, something to the effect of generally, events of Barry dancing are good qua events of dancing. In uttering Barry is a good dancer, we are predicating something of events of Barry dancing, namely, that they are good qua events of dancing. We are not directly predicating of Barry the property of being a good dancer. Since the agentive nominal is not part of what is being predicated in Barry is a good dancer, we would not expect the sentence Barry is a dancerin which the agentive nominal is part of whats predicated to be entailed. Thus the nongeneric version of the Port-Royal Puzzle is explained.
17
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
Larson does not discuss the generic version of the puzzle, but his work extends naturally to explain it. Consider (3), for example: (3) Bulgarians are good weight lifters. Weight lifter is an agentive nominal (as is sailor). Following Larson and Condoravdi, I suggest we treat weight lifter as part of the Restrictor rather than the predicate. This massively reduces the number of exceptions there are to the generalization. In uttering (3), we are not generalizing over Bulgarians and claiming they satisfy is a good weight lifter, but rather we are generalizing over Bulgarian weight lifters and claiming that they are good at weightlifting. The Port-Royal Puzzle rests on an error about logical form. Thus I claim that once we properly understand the logical form of these claims, they are not troublesome generics. The Paradox of Generic Acquisition Recall from our earlier discussion that we are taking Gen, the generic operator, to be a default operator that arises to bind any variables that remain free in a sentences restrictor once all the articulated quantifiers in the sentence have bound their variables. Gen is thus a peculiar operator, occurring only when the sentence does not provide articulated quantifiers to bind all the restrictors variables. As we have noted, this is not an idiosyncratic feature of English: Gen is not articulated in any known language (Krifka et al. 1995; Dayal 1999), and so exists only as a shadowy default operator that arises in the absence of explicit ones. As we have seen in previous sections, the interpretation of this default operator is complex. Gen does not track straightforward statistical regularities, and if my above classification of troublesome generics is correct, it is sensitive to whether the predicated property is striking, or among characteristic features of the kind. Moreover, it appears to be a linguistic universal that the operator with these strange qualities is the default one; Gen is never articulated, so it is natural to conclude that Gen is a default operator across languages, as it is in English. A puzzling question now arises: how does a language learner ever come to master generics? Not only is the interpretation of Gen rather complicated, the operator is not even phonologically realized. How does a language learner come to know that it is this particular operator with this peculiar meaning that binds unbound variables in the absence of other operators? The task of learning generics appears quite daunting.
18
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition Surely, one would think that explicit quantifiers such as every and most would be more easily acquired. For one, they are phonologically realized, and thereby present the child with an object of study, in contrast to our elusive Gen. Secondly, they have precise statistical meanings, of the sort that linguists and philosophers of language can readily seize on as semantic interpretations, and which one might imagine the language learner could quickly grasp. For example, one might fondly imagine the child noticing that every A is B is said only when the As are a subset of the Bs, while most As are Bs is said as long as more than half of the As are Bs. Fair enough, but then all of a sudden our language learner encounters sentences like the previous ones, but with a blank in place of the quantifier. The child then has to learn that Mommy and Daddy refrain from uttering every, or all, or most, or usually in circumstances where As are generally Bs, unless B is really striking, in which case it is enough that just some As are Bs, or B is a characteristic feature of As, in which case again just some is enough, except in the case of artifacts and their design, in which case no As need to be Bsand so on. And again, the child must learn to associate this complex range of conditions with the absence of a quantifier or determiner.13 The task seems practically impossible: how do we ever master this part of language? To make matters all the more puzzling, it happens that generics are acquired quite early on. Children start using generics by two years of age, which is significantly earlier than explicit quantifiers (Gelman 2003; Roeper, Strauss, and Pearson 2006). That children ever master generics is perplexing; that children master them more readily than explicit quantifiers borders on the paradoxical. This is a phenomenon that demands explanation. Semantics from a Cognitive Perspective But why, one might ask, should we concern ourselves with the facts of language acquisition? Traditional approaches to semantics do not
13. Associations with absence are notoriously difficult for children to master. If children as old as four or five are presented with two stimuli that differ only in that one stimulus has an additional feature that the other lacks, and are reinforced only for responding to the stimulus lacking the feature, they fail, by and large, to learn the discrimination. If they are reinforced for the stimulus with the distinctive feature, they learn the discrimination quite rapidly; there is something especially challenging about learning associations with absences. See Sainsbury (1973) for discussion of this Feature Positive Effect.
19
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
usually take such considerations into account. Often, semantic interpretations consist of assigning things like objects, properties, senses, or truth conditions to sentences and their parts. These more traditional approaches to semantics might be viewed as dealing with the question of how language is related to the world. This is an interesting question and has received much attention. There are, however, different questions related to meaning that we might wish to address: how do human beings actually understand their natural languages?14 What are the mental processes that underlie our ability to comprehend and produce the sentences of our languages? Notice that this question might remain unanswered even if one had a successful finite and recursive way of matching sentences with meanings. There is no guarantee that such a model would be the correct model of human language comprehension. A broad range of empirical data from acquisition, disorders, cognitive development, and so forth would bear on the adequacy of a given model from the psychological point of view. These latter sorts of considerations are rarely taken into account by semanticists. If ones interest lies solely with characterizing the language-world relation, then these sorts of consideration may be irrelevant. If, however, one is interested in how human language comprehension and production actually proceeds, then one would be well served to consider such data. Exactly what one should do with the data, or what a cognitively oriented semantics should look like, may not be immediately obvious. I hope that the remainder of this essay will give the reader a sense of how one might proceed. For now, I will thus set aside the tools of formal semantics and will not directly concern myself with the relation of generics to the world. I will instead develop an account of generics that is couched in the framework of cognitive science. Infant Generalizations The contrast between the difficulties we would expect children to face in acquiring generic language and the apparent ease with which they
14. That this question exerts an influence on semantic research is evidenced by the pursuit of finitely axiomatizable language-to-world semantic theories. The motivation for requiring that semantic theories be finitely axiomatizable is primarily to explain how finite beings can understand a potentially infinite number of natural language sentences. This constraint figures in just about every contemporary semantic theory.
20
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition master it suggests that they do not learn the meaning of Gen in the manner discussed abovethat is, by isolating and matching a complex set of truth conditions with the absence of an explicit quantifier. A different proposal is needed. Let us begin by making the following obvious claim: Generics express generalizations. This is intended as a weak and untendentious claim; generic sentences are not about some specific instances of the category mentioned in the Restrictor, but rather about the category in general. Now, inductive or ampliative learninglearning that goes beyond the instances with which one is presentedis central to the acquisition of knowledge throughout life. Inductive learning is what allows the child to avoid a hot stove after a single burn, the adult to avoid repeating errors indefinitely, and perhaps even the conditioned rat to flee a shock box after hearing a warning tone. Without the ability to make generalizations that go beyond particular, experienced instances and so respond accordingly to novel events and items, an animal would probably not last very long. No theorist would deny that the capacity to generalizethe capacity for inductive learningis innately given.15 The capacity to generalize manifests itself early in life. Graham, Kilbreath, and Welder (2001) found that twelve- to fourteen-month-old infants are willing to generalize nonobvious properties, such as rattling when shaken, both to objects that were perceptually similar to the original item and to objects that were introduced with the same label, (e.g., these are blicketts!), but not to any other objects. Graham, Kilbreath, and Welders experiment was intended to demonstrate the impact of language on infant inductive categorization tasks; the impact is striking considering that infants are only just beginning to use language at that age. It is also interesting to notice that while labeling aided the infants in generalizing to a broader range of items, they were already inclined to generalize to the class of perceptually similar items. Baldwin, Markman, and Melartin (1993) further found that infants as young as nine months were willing to make similar generalizations. Thus the inclination to generalize, though aided by language, does not depend on language but is, rather, an early developing, presumably innate, cognitive disposition. This innately given cognitive disposition is surely available to infants before they have acquired any language, as the study concerning nine-month-olds suggests. The disposition constitutes the human minds default manner of generalizing information from a few instances
15. Empiricist-nativist controversies often concern whether anything is innately given apart from this ability.
21
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
to many. Thus an interesting hypothesis presents itself: just as, at the level of logical form, Gen is a default operator, at the level of language comprehension, its contribution is to invoke the minds default manner of generalizing the material in the Scope over the items that satisfy the Restrictor. Recall the notion of a tripartite structure discussed earlier. These tripartite structures can be schematically represented as follows: Quantifier [Restrictor] [Scope]. The hypothesis I am suggesting is that, in the case of generics, the relation between the Restrictor and Scope is computed by means of the minds default method of generalizing. This would stand in contrast to, for example, the computation of set-theoretic relations between the extensions of the Restrictor and the Scope. If Generalized Quantifier Theory were understood as a view about psychological processeswhich it need not bethen these set-theoretic computations would be central to our comprehension of explicit quantifiers. On the view of generics here advocated, it is the innately given default mechanism of generalization that underwrites our comprehension of generics. This mechanism is invoked in understanding, evaluating, and producing generics. If a speakers knowledge and experiences with members of a kind K leads her default mechanism to generalize the property of being F to that kind, then she will express this with the generic Ks are F. Similarly, her hearer would judge the utterance to be true if, given his knowledge and experience, his default mechanism would generalize the property of being F to the Ks. If this is how generics are grasped, the paradox of acquisition is dissolved. At the time of acquisition, the child would already be in possession of this default method of generalizing, complete with all its quirks. Thus the task of learning how generics should be semantically interpreted would be simplified. Once a learner is able to partition material into a Restrictor and a Scope, and understand that material (the nouns, verbs, and so on), she is equipped to understand genericsmodulo the implicit knowledge that the default Gen must bind the Restrictors free variables. But it is also quite natural to imagine that whatever operations of the language faculty are responsible for the emergence of Gen at Logical Form (LF) are part of the language facultys innate endowment; it is hard to imagine what evidence could push a learner to adopt the practice of thus binding the restrictors free variables. Minimally, the childs innate linguistic knowledgeknowledge of Universal Grammar, if one likesmust entail that an LF cannot contain free variables if it is to be interpretable. The child must then find some operator if the sentence
22
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition does not provide one, and there is some elegance to the idea that this default operator has as its interpretation the childs default mode of arriving at generalizations. As generic acquisition requires only that the child learn how her language partitions material into Restrictor and Scope, and the meaning of the nonquantificational material therein, this view predicts that the unarticulated generic would be acquired before the explicit quantifiers. The truth-conditionally tractable and phonologically articulated quantifiers are so because they need to be learned, in a way that generics do not.16 The seeming paradox of generic acquisition is dissolved. Defaults and Inhibitions We are investigating the following hypothesis: the default operator Gen invokes the cognitive systems default mode of generalizing. I argued that there must be such a mode and that it operates prior to the acquisition of generics, even though the latter occurs quite early on. If generics are understood by way of this default, we might further conjecture that the comprehension of nondefault quantifiers requires the conceptual system to override or inhibit its default operations. When the conceptual system is confronted with an LF of a tripartite nature, it must compute the relation between the Scope and the Restrictor in a way that is appropriate to the operator. In the case of generics, where the only operator is the default Gen, the least marked of operators, the relation between Restrictor and Scope is simply the relation of the latters being generalized by the default mechanism to the former, so the evaluating conceptual system need only invoke its default method of operation. In the case of explicit, marked quantifiers such as all, however, the evaluating conceptual system must use a computational process other than its default; it must use one that will ensure the material in the Scope is understood as universally generalized to the material in the Restrictor. We may think of the quantifiers, articulated and marked, as supplying explicit instructions to the evaluating conceptual system. For a conceptual system to perform a nondefault generalization, such as a
16. Some contemporary theorists hold that quantifiers are innate, even though they emerge later in development. I have no quarrel with this view, though I take it as self-evident that if quantifiers are innate, so are generics. See S. J. Leslie 2007, forthcoming for more discussion of the relationship between generics and quantifiers.
23
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
universal one, it must be explicitly commanded to do so. Absent such instruction, the conceptual system simply employs its default. Chomsky (2000) suggests that similar phenomena occur elsewhere in language. Consider, for example, the sentence John climbed the mountain. This is understood as meaning that John climbed up the mountain; to obtain the interpretation that John climbed down the mountain, we must explicitly use the preposition down. The unmarked case climbed the mountain is never interpreted as climbed down the mountain. This may reflect a fact about our conception of climbing, to the effect that by default we conceive of climbing as climbing up. To deviate from this default interpretation, we must use the more marked form climbed down the mountain, which makes use of an explicit preposition. Similarly, in the case of generalizations, the unmarked generic invokes the cognitive systems default mode of generalizing. To invoke a nondefault generalization, an explicit and marked quantifier must be used. The connection, then, between generic generalizations and unmarked surface forms is not an accidental one, but rather reflects a deep fact about human cognition. It is no coincidence that no human language has a dedicated, articulated generic operator. We might hypothesize that part of what makes the mechanism involved in the comprehension of generics the default mechanism of generalization is that other methods of generalizingsuch as that corresponding to universal generalizationsrequire some inhibitory processing.17 If generics alone reflect this default mode of generalizing,whereas other methods of generalizing require inhibitory processing, then we would expect certain empirical consequences. First, we would expect that children would handle generics with greater ease than the explicit quantifiers, which would be manifested by their acquiring generics first. We might also expect that generics would show up more frequently in
17. The notion of some processes being defaults and others requiring these defaults to be inhibited or overridden is ubiquitous in cognitive science, neuroscience, and biology. The spinal column of a chicken implements processes that, by default, cause the chicken to run around aimlessly. These processes must be inhibited by the chickens brain for the chicken to refrain from such behavior. If the chicken loses its head, then the spinal columns default mode of operation kicks in; this is the science behind the description running around like a chicken without a head. Less vivid examples of inhibition and defaults can be found anywhere from cellular processes, neural networks, glandular regulatory systems, visual cognition, action-planning systems, and even at the conscious level of temptation resistance.
24
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition maternal speech.18 Maternal speech is indeed rife with generics, suggesting that generics constitute a ready means of imparting valuable information to children. In contrast, explicit quantifiers are rarely found in maternal speech (Gelman et al. 1998). We might also expect the more surprising result that if faced with a sufficiently demanding cognitive task, young children might even interpret sentences containing explicit quantifiers as though they were generics. That is, if the processing demands of explicit quantifiers are greater, then under some circumstances, a young child may fail to implement the more demanding processes and resort to the default ones associated with generics. Thus we might find some experimental situations in which young children erroneously act as though explicitly quantified statements were in fact generics. We might also expect that adult nongeneric generalizations would, from time to time, inappropriately exhibit some characteristics of generics, especially if the information-processing demands were made great enough. These predictions are all supported by empirical results. I will delay discussion of adult errors until the end of the essay, as it will be helpful to first characterize generic generalizations in more detail. The prediction concerning childrens mistakes, however, is strongly supported by data from a recent experiment by Hollander, Gelman, and Star (2002), to which we now turn. Hollander, Gelman, and Star recently conducted an interesting experiment that investigated young childrens understanding of sentences containing all and some and generics. Their subjects included a group of three-year-olds, a group of four-year-olds, and a control group of adults. The participants were asked yes/no questions that were either generic, or contained all or some; for example, a subject might have been asked are fires hot? (generic), are all fires hot? (all), or are some fires hot? (some). (The questions were selected so that threeyear-olds could reasonably be expected to know the answer; the subject matter was no more esoteric than the above example concerning the hotness of fires.) Within each category of question, there were
18. Maternal speech is the characteristic manner that mothers adopt when addressing young children and babies. Maternal speech is regarded as a specifically evolved means of communication that aids children in learning their native language and conveys information to them in ways that they most readily grasp. It is not, for example, an accidental cultural artifact.
25
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
questions concerning wide-scope properties that were had by all or almost all the members of the category, questions concerning narrowscope properties that were had by only some members, and questions concerning irrelevant properties that were had by none (for example, do fish have branches?). This last class of questions was included to ensure that the subjects were paying attention and providing sensible answers. The experimenters found that, whereas the adults and four-yearolds distinguished their answers depending on the type of question they were asked, the three-year-olds did not. There was no statistically significant difference between the three-year-olds answers to questions containing all or some or generics; the three-year-olds responded as though they had been asked only one type of question rather than three very different sorts of questions. This alone is remarkable, but the experimenters further found no significant difference across the age groups answers to the generic question; the three-year-olds responded as though every question they were asked was a genericeven though many questions contained all or some. The older children and the adults differentiated between the three question types by answering the all and some questions differently from the three-year-olds, but the response patterns to the generic questions remained unchanged with age. These results suggest that while three-year-olds have difficulty with questions containing all and some, their competence with the generic questions is already close to that of an adult. Since generic questions elicit the same response patterns regardless of age, we may conclude the three-year-olds correctly understand and process generics. This alone is noteworthy; the asymmetry between the young childrens competence with generics and their difficulty with all and some reinforces the thesis that generics are mastered much earlier in development. The data invite an even stronger conclusion, however; the threeyear-olds difficulty with all and some appears to take the form of assimilating these explicit quantifiers to generics. The three-year-olds did not simply lack competence with questions containing all and some; they answered them as though they were generics. Not only were these young children proficient with generics yet uncomfortable with explicit quantifiers, they actually resorted to treating the explicit quantifiers as generics. The experimenters went on to test the three-year-olds understanding of all and some under different circumstances. They showed
26
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition their subjects four crayons (for example) some, all, or none of which were in a box. They then asked the children are all the crayons in the box? or are some crayons in the box? One might reasonably think that these circumstances would make it easier for the children to answer appropriately; they were no longer being asked about categories in the abstract but rather about concrete items before them. This second experiment thus placed the children in lower-demand circumstances than those of the first experiment; they were no longer expected to give answers concerning abstract generalities, and so the task carried fewer cognitive demands with it. In this lower-demand situation, the children by and large gave correct answers to the questions; they were able to treat all and some appropriately. Thus three-year-olds are not incapable of processing all and some, but it seems that they find them difficult to processdifficult enough that it matters for their success how demanding the task is. This is most consistent with all and some requiring inhibitory processing. Inhibitory processes, by their very nature, are more taxing for the conceptual system to implement than noninhibitory ones. As the demands on the conceptual system are increased, it becomes more likely that it will be overtaxed, and so fail to implement the inhibitory process, and revert to the default. The conceptual system nonetheless has the ability or competence to perform the inhibitory process; it simply fails to do so because it is overwhelmed. Psychologists have held that there is a significant increase in the ability to implement inhibitory processes between ages three and four (Friedman and A. M.Leslie 2005). If I am right and explicit quantifiers require inhibitory processing, then it is no surprise that the fouryear-olds succeed where the three-year-olds fail. That the three-year-olds were able to give the correct answers to a simplified task also suggests that it is not lack of competence or knowledge about the interpretation of all and some but rather difficulty using that knowledge in the more demanding circumstances. One should bear in mind in evaluating these data that three-year-olds are generally quite linguistically competent; these are no eighteen-month-olds, stringing together two-word sentences. The surprising data is not that they can correctly handle all and some under some circumstances, but that they fail to do so under others. That this difficulty consists of sometimes treating sentences with all and some like generics lends much support to the claim that the comprehension of generics reflects a more basic process than the comprehension of these explicit quantifiers.
27
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
The Hollander, Gelman, and Star results are exactly what one would expect if the theory I am proposing is correct.19 While the explicit quantifiers all and some can be successfully processed by young children under low-demand conditions, in high-demand conditions children erroneously fall back into their default mode and act as though sentences containing all and some were generics. If generics do indeed express the conceptual systems default mode of generalizing, then we would expect exactly this pattern; deviations from the default, such as are required by all and some require inhibitory processing, whereas the default does not. Children therefore handle generics with far greater ease than explicit quantifiers.20 Sesame Street provides children with a handy little jingle to aid them in understanding all, some, and none. No such jingle exists for generics. How could it? Top linguists have fallen short of adequately articulating even an extensionally adequate semantics of generics could the Sesame Street writers be expected to do any better? Fortunately, this is no great loss to the shows preschool audience; they already understand generics.
Information Gathering Thus far I have been proposing the following solution to the paradox of generic acquisition: generics express the conceptual systems default method of generalizing. This method is innately givenquirks and all and is in operation long before even the earliest manifestations of language acquisition are seen. The childs innate language endowment
19. They themselves do not consider the hypothesis that generics invoke the default mode of generalization; to the best of my knowledge this hypothesis is original to me. Hollander, Gelman, and Star do note that the children are falling back on the generic in the more difficult task, but simply state that the reason for this should be considered in future research. 20. One might wonder whether young children grasp all the troublesome aspects of generics. Perhaps these features are acquired more slowlyperhaps more slowly than explicit quantifiers are acquired. This is an empirical question, and I do not know of data that bear directly on the question at this time. However, it would not, I think, be terribly surprising to find that young children readily accept generics such as ducks lay eggs and lions have manes, and even tigers eat people, while rejecting generics such as ducks are girls and tigers dont eat people. (Maternal speech, of course, contains troublesome true generics and is devoid of troublesome false ones.)
28
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition would then only need to provide the learner with three principles for the acquisition of generics to proceed: 1. 2. 3. All variables must be bound for an LF to be interpretable. Variables that are free in the Restrictor are bound by a default operator. This default operator invokes the conceptual systems default means of generalizing.
If the acquisition of generics proceeded even roughly along these lines, we can begin to see why generics are so easy to acquire; they correspond most closely to what the child already knows how to do. This might make the life of the learner a little easier, but it still remains for us theorists to characterize this default method. If an account of generics as reflecting this default method is to have independent plausibility as an account of the conceptual systems basic means of generalizing, it would be desirable to characterize this mechanism of generalization on the basis of psychological data. Thus while I will be inevitably guided by the desire to give an account of generics, I will endeavor to approach the problem from a broader psychological perspective. This mechanism is, by hypothesis, a basic information-gathering mechanism that underwrites much of our early knowledge of the world. Young children need information, and they need it fast; this mechanism should be as efficient as possible. An evolved mechanism such as this one could increase its efficiency by off-loading as much work as possible onto its environment. This particular mechanism is concerned with drawing inferences from particular instances of a category to novel and unobserved ones; such a mechanism could increase its efficiency by taking advantage of regularities in the world. There are, after all, many useful regularities out there in our environment. For example, individual cats bear many similarities to one another. Cats, on the whole, eat the same things, lick themselves in the same manner, make the same noises to get our attention, move around in the same fashion, and so on. There are regularities at other levels of abstraction too; just as individual cats resemble each other in a multitude of ways, cats, dogs, and other animal kinds are also alike in many respects. They each have a characteristic diet, characteristic noises, modes of locomotion, and so on. These regularities are provided to us by our environment; an efficient information-gathering system should
29
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
take advantage of them, and thereby minimize the extent to which the learner must relearn pieces of information (Shipley 1993). If one were able to use information from related kinds to predict dimensions of regularity for novel ones, then one would be able to fill in the value on this dimension on the basis of limited experience with the new kind. For example, experience with meowing cats, barking dogs, and clucking chickens might indicate to a learner that animal kinds have characteristic noises associated with them. The learner, upon encountering a single croaking frog, could at least tentatively conclude that croaking is the frogs characteristic noise. Let us call these dimensions of regularity characteristic dimensions. Characteristic dimensions of animal kinds include perhaps noises, modes of locomotion, means of reproduction and gestation, and diet; characteristic dimensions of artifact and social kinds may include function and role respectively. An efficient information-gathering system could use information from a related category to predict characteristic dimensions of a novel one, then use even a single instance of the novel category to obtain the value of that characteristic dimension for the new category. Macario, Shipley, and Billman (1990) found that children use precisely this sort of information in experimental tasks. In particular, they found that children use information from related categories to infer regularities about novel ones. If children are presented with a category whose members are similar along a given dimension, for example, color, they expect that novel but related categories will also have members that are similar along that dimension.21 We might hypothesize then that part of the operation of the mechanism we are characterizing consists in inferring the existence of the characteristic dimensions of novel kinds from related kinds, then using very fewsometimes only oneinstance of the novel kind to supply a value for the dimension (cf. Shipley 1993).
21. In Macario, Shipley and Billmans experiment, children were told that Smurfette was very picky about which blocks she liked; in fact she liked only these blocks, where a collection of blocks that were uniform along some dimensions but not others was ostended. The ostended collection might, perhaps, consist of blocks that were all red, but were of many different shapes. The children were then shown a single block, and told that Papa Smurf liked it, and were then asked to point out what other blocks they though Papa Smurf might like. If Smurfettes blocks were uniformly red, and Papa Smurfs single block was blue, children picked out other blue blocks for Papa Smurf. If Smurfettes blocks were uniform in shape but varied in color, then the children used the shape of Papa Smurfs block to determine which others he might like.
30
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition This hypothesis gains support from an examination of maternal speech about kinds. Novel kinds are often located as members of a superordinate kind; for example, mothers are often heard saying things like look, its a prairie dog. Thats a kind of animal (Shipley, Kuhn, and Madden 1983). If the child is disposed to reason in the manner described above, then she may conclude on the basis of her knowledge of other animal kinds that prairie dogs will have, for example, diet, noise, and locomotion as characteristic dimensions. She can then observe the single prairie dog at hand and draw (at least tentative) general conclusions about other prairie dogs. It should be noted that this is a very powerful mechanism for induction; it draws robust inferences from very few examples. (Of course such a mechanism is vulnerable to error, as will be discussed in the next section.) Learning is facilitated in that minimal experience is needed for valuable information to be gathered. If one has a prior notion of what the characteristic dimensions of a kind are, then ones attention may be focused on finding values for those dimensions, thereby filtering out less important information, and of course one may use as few as one instance of the kind to supply values for those characteristic dimensions. Thus, providing one has already located appropriate characteristic dimensions for a kind, thenmodulo the remarks in the next sectionone is well placed for learning inductively about that kind. If this is a fair characterization of the relevant cognitive dispositions, then we can begin to shed some light on the troublesome generics. Recall that one of the two main classes of troublesome generics pertained to features that were intimately linked to background expectations we have of the kinds in question. For animal kinds, these features were often related to reproduction, and for artifact and social kinds, they were related to role and function (e.g., OrangeCrusher2000s crush oranges). This is perhaps because these features lie along characteristic dimensions for these kinds. We might conjecture that generalizations concerning values along characteristic dimensions are made rapidly, in the face of but a few examples, and are not readily underminedthus it only takes a few egg-laying bees for us to judge that bees lay eggs is true. In her seminal article, Shipley (1993) argues that the attribution of characteristic dimensionsor over-hypotheses, as she terms themplays an integral part in our inductive practices. She goes on to pose the question: what is the source of these characteristic dimensions? Shipley (ibid., 296) wonders if they emerge as knowledge becomes more
31
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
organized, or if they are the outgrowth of innate dispositions, but does not suggest an answer to the question. I suspect that they are not simply by-products of organized worldly knowledge; rather, I believe they are the guiding principles behind not only the organization of our worldly knowledge, but even its acquisition. Shipley does not dwell long on inductive generalizations that have more exceptions than satisfiers (e.g., birds lay eggs), though she does wish to subsume them under her account. She suggests in passing that birds lay eggs is best understood as birds are a kind of animal such that the mature female lays eggs. If the availability of such paraphrases is sufficient for the truth of a generic, then we should wonder why the truth of birds are a kind of animal such that the egg-layers are mature females does not imply the truth of birds are mature females. (Hers is just a passing comment, though, and not intended as an account of generics, of course.) The above asymmetry, though, highlights the fact that, however characteristic dimensions arise, they do not arise from simple statistical observation of the kinds. Thus Shipleys suggestion that characteristic dimensions may arise from the organization of our worldly knowledge seems to me to fall short of a full explanation. Objective statistical facts cannot account for why our world knowledge is organized as it is; if positing previously establishedthrough either nature or early nurturecharacteristic dimensions can account for it, then this is a hypothesis that should be taken seriously as the source and explanation of this aspect of our knowledge. I do not pretend to know the extent to which these principles are innately given and to what extent they are provided by early nurture. Perhaps maternal speech suggests characteristic dimensions to very young children by drawing implicit parallels through iteration. (We are all familiar with that particular singsong tone of voice informing the child of such parallelisms as cats say meow, dogs say woof, and chickens say cluck!) Minimally, though, I imagine that children are innately disposed to seek out characteristic dimensions for kinds. These dimensionsthough of course not their valueswill be uniform within domains of kinds, such as artifacts, animals, and social roles. Whatever the origin of characteristic dimensions, I claim that they play a central role in our primitive generalizations, and so in our judgments of generics. They provide an outline of information to be gathered about a new kind; characteristic dimensions provide a learner with an informational template. When a value is found for a characteristic dimension of a kind, it is thereby generalized to the kind by the basic
32
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition generalization mechanism, and so the generic that predicates that property of the kind is accepted. Ducks, being an animal kind, have reproduction22 as a characteristic dimension, so the inductive learner looks for a value to fill in this dimension; even limited experience will deliver laying eggs as the appropriate value, and so the property is generalized to the kind, and ducks lay eggs is accepted as true. Revision and Counterinstances As it stands, the account I have offered so far falls short of explaining our judgments of troublesome generics. If laying eggs is a value along the characteristic dimension of reproduction, why isnt not laying eggs also a possible value on that dimension? Why do the non-egg-laying ducks not provide us with equally good candidate values? Thus far we have no explanation of why this might be so. Naturally, generalizations that are based on limited experience should be open to revision in the face of counterinstances, at least under some circumstances. It is one of the most widely noted features of generics that they tolerate exceptions; that is, Ks that are not F. A moments reflection should convince any theorist that far more needs to be said about the toleration of counterinstances; why, for example, are male birds tolerable counterinstances to birds lay eggs but not to birds are female? Why does the latter sentence have an almost universal flavor to itit would seem to require for its truth that all birds be femalewhile the former in no way demands that all birds be egglayers? Why is mosquitoes carry the West Nile virus intuitively true, yet books are paperbacks false, when the percentage of mosquitoes with the virus is about one-eightieth the percentage of books with paperback covers? I propose that a powerful factor here is whether the counterinstances are positive rather than negative. The distinction I have in mind is as follows: a positive counterinstance to Ks are F occurs when an instance of the kind K has a concrete alternative property, that is, when it has a positive alternative23 to the property F, whereas a negative
22. Of course young children might not understand the notion of reproduction as such, but they are aware from an early age that adult animals have babies that are members of the same kind as their parents (see Gelman 2003). 23. The notion of alternative that I have in mind here is similar but not identical to Cohens notion. For my purposes, a property G is an alternative to a property F only if the two properties are incompatible. Thus being male is an alternative to being female,
33
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
counterinstance occurs when an instance simply fails to be F. Whether a counterinstance counts as positive or negative is highly dependent on the property being predicated. For example, if we are considering birds are female, then its counterinstances are positive ones; those birds that fail to be female do so in virtue of possessing the positive alternative property of being male. If instead we are considering birds lay eggs, the birds that fail to do so simply fail to do so. They do not lay eggs. They are merely negative counterinstances, failing to possess the property in question, without possessing a positive alternative property. If some birds were so constituted as to bear live young, these birds would be positive counterinstances; males and barren females are not. There may be reasons to be skeptical of attempts to draw a metaphysical distinction between positive and negative properties. The distinction at hand, however, is not intended as a metaphysical distinction, but rather a psychological one. What matters is whether we take the counterinstances as negative or positive. The metaphysical basis for this distinction may be uninteresting; human perception is the important factor here. As my reader may have guessed, this positive/negative distinction is relevant to generics in that negative counterinstances are more likely to constitute tolerable exceptions than positive ones. For a generic, such as birds lay eggs, we are pitting the positive feature of giving birth in a certain manner against the males and barren females negative feature of not doing so. This should be contrasted with birds are female, which is false, despite the egg layers being a subset of the females. The counterinstances to birds are female are positive counterinstances; they possess the positive alternative property of being male. Thus birds lay eggs is accepted as true, while birds are female is not, at least in part because birds are female would require us to overlook positive counterinstances, whereas birds lay eggs does not. That the positive/negative distinction is relevant here is supported by the following thought experiment: the generalization expressed by birds lay eggs is tenable in the face of male birds, but what if we were to learn of some female birds that bore live young? It seems natural to think that under those circumstances the generalization would be weakened to something like birds lay eggs or give live birth.
but carrying malaria is not an alternative to carrying the West Nile virus since a single mosquito could carry both. Obviously, I do not consider being F to be an alternative to itself; here, I differ from Cohen.
34
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition Similarly, in evaluating generics such as lions have manes and peacocks have fabulous blue tails, we are pitting perceptually striking positive features in the males against the effective absence of those features in the females. The females of those species are drab and unremarkable compared with the malesthey effectively lack the features that the males possess, and so essentially are only negative counterinstances.24 It is worth considering another intuitive test here. As things stand, we are inclined to assent to peacocks have fabulous blue tails. Would we be so quick to do so if the females had fabulous pink tails rather than brown stumps? I am inclined to think that we would then assent only to peacocks have fabulous pink or blue tails and not to peacocks have fabulous blue tails since such an equally striking feature had by the females is not so easily overlooked. Generics such as peacocks have blue tails and lions have manes are readily judged true because the females of the species lack these properties; they do not exhibit competing properties, or at least the competing properties they do exhibit are quite boring and unremarkable.25 This contrastive feature of judgments of generics has not been noted in the literature so far. That our judgment would change if the way in which female peacocks lack blue tails changed suggests that in making our judgments we do not consider only the members of the category that possess the feature in question, but also the ones that lack it. In particular, it is relevant how they lack the feature. There is an intuitive difference between simply lacking a feature and lacking it in virtue of having another, equally memorable, feature instead. The peacock example illustrates this: as things actually stand, the unfortunate female peacocks simply lack those fabulous blue tails. Had nature been kinder to them, however, they might have had fabulous pink tails instead and would then have lacked blue tails in virtue of
24. It should be clear here why the distinction is only psychological. A stumpy brown tail is no less existent than an ornate blue one, but we pass so easily over the former it might as well not be there. As an autobiographical note, I believed that female peacocks had no tails whatsoever, before this project led me to learn otherwise. 25. I am not sure whether characteristic dimensions are playing a role here, or whether the negative/positive contrast alone is sufficient, given that the proportion of peacocks with the tails is close to half. (That is, we are not required to overlook the overwhelming majority in accepting this generic.) It is plausible to think that we expect animal kinds to enjoy some uniformity in appearance, though the dimension of appearance would have to be rather more complicated than, say, gestation, since there is much more potential for variability with appearance.
35
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
having pink tails. I suggest the following hypothesis: in making a generic judgment that Ks are F, it matters how the non-F Ks fail to be F. If they fail to be F in virtue of having an equally salient, memorable, and striking feature, the generic is unlikely to be judged true. If, however, they fail to be F in a nonstriking, uninteresting way (such as merely lacking F), then we are far more likely to judge the generic to be true. This contrastive approach that I am advocating points toward an understanding of why some generics are judged to be false, despite the majority of the Ks satisfying the predicate: Books are paperbacks (false). Greyhounds are English Greyhounds (false). Chickens are female (false). Here, the alternatives are not merely the absence of the given property; there are concrete alternatives to the properties cited. When we pit one concrete property against another, the number of Ks that must be F is considerably greater. The alternatives in question here are being hardcover, being an Italian Greyhound, and being male. The alternatives in these cases are particularly hard to ignore. The first two properties pick out subkinds of the kind in question, and the third picks out a property such that if no chickens have it, the kind Chicken will fail to be able to reproduce. Thus we would need approximately universal satisfaction of the predicate for these to be true, which squares with intuition about these examples. Let us combine the diagnosis that this positive/negative distinction plays an important role in our judgments of generics with the claim that generics reflect primitive facts about the cognitive system. This implies that negative facts have less psychological impact than positive ones. There is some support for this claim. In their seminal book, Nisbett and Ross (1980) argue that people are affected by information in proportion to the vividness and concreteness of that information and note that this predicts that the absence of an event or property should have disproportionately little impact, in virtue of such information being more easily overlooked. The authors illustrate this nicely with an example from Sir Arthur Conan Doyles Memoirs of Sherlock Holmes. In The Silver Blaze, Holmes calls Inspector Gregorys attention to the curious incident of the dog in the night-time. Inspector Gregory (still trapped in an infantile way of thinking) replies in puzzlement, The dog did nothing in the night-time, to which Holmes triumphantly replies, That
36
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition was the curious incident. The nonoccurrence of the dogs barking signaled to Holmes that the intruder must have been someone the dog knew. This conclusion had not readily presented itself to the less gifted Inspector Gregory, however, because it was inferred from a nonoccurrence rather than from an actual event. Nonoccurrences, being by their very nature nonconcrete and nonvivid, are easily overlooked. Nisbett and Ross cite experimental evidence in support of this intuitive claim. Both humans and animals have great difficulty learning associations that involve the absence rather than the presence of features. Animals and young children have difficulty in even the simplest learning tasks if they are asked to distinguish between two stimuli that are identical save for one feature and to respond only to the one that lacks the feature. For example, they may be presented with two circular stimuli, one with a smaller circle in the middle, and the other without. Animals and young children can be readily taught to respond only to the stimulus containing the inner circle, but it is very difficult to train them to respond only to the stimulus that lacks the circle (Sainsbury 1973; Jenkins and Sainsbury 1970). (This is known as the Feature Positive Effect.) Adults are, of course, capable of mastering this simple task, but if the scenario is made complicated enough to be taxing, adults encounter far more difficulty with negative associations than with positive ones (Newman, Wolff, and Hearst 1980). That there is such a strong asymmetry between our abilities to grasp the two types of associationidentical save for whether the stimulus to which one must respond is the positive or negative onesurely reflects an asymmetry in how we process presences versus absences. If we are not presented with a positive object for association, we have difficulty learning the association at all. Mere absences are not on a cognitive par with concrete presences. Characteristic Dimensions Revisited Thus far I have discussed two factors that influence our judgments of generics: the background expectations associated with kinds and the nature of the counterinstances to the generalization. Generics that pertain to background expectations are very tolerant of exceptions, at least when those exceptions simply fail to have the property in question when they are negative counterinstances. I suggested earlier that a feature of our primitive inductive learning mechanism is that it operates by identifying characteristic dimensions of a category, then locating a value on that dimension on the basis of what is often very limited information.
37
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
I noted, though, that this observation alone cannot explain the nature of troublesome generics. Now that we have the negative/positive distinction in hand, however, we can frame the following hypothesis: the inductive learning mechanism is strongly biased towards filling in only positive values for characteristic dimensions, and once it has done so, it readily retains the value in the face of negative counterinstances. Thus despite the large number of bees that do not lay eggs (all bees, save the queen), these bees do not possess positive properties that could supply values for this characteristic dimension. Only queen bees supply such a positive value, but since bees are an animal kind, we are eager to find a value for their method of gestation, so we seize on the only positive instances available, and conclude that bees lay eggs. This hypothesis is both formed and retained in the face of the counterinstances since they are all negative. One might wonder whether we need to appeal to characteristic dimensions at all now that we have the negative/positive distinction in hand. Could we perhaps explain our judgments of generics on that basis alone? It would make for an elegantthough perhaps somewhat odd account of generics if we could simply observe that when the counterinstances possess positive alternative properties, the generic is false, otherwise it is true. I do not believe such an account would be successful; it would classify some intuitively false generics as true. For example, there are some dogs with oddly shaped lumps on their backs, while the large majority of dogs simply lack such lumps. We do not, however, assent to dogs have oddly shaped lumps on their backs. The only counterinstances to the generic are negativethe counterinstances merely lack the strange lumpsyet this does not suffice to make the generic true. Thus it cannot be enough for the truth of a generic that all its counterinstances be negative; if there are a great many of those counterinstances and the generic does not pertain to a characteristic dimension,26 then it will be false. Of course there are some generics that feature properties that are neither striking (see below) nor found along characteristic dimensions of the kind. In these cases, though, we need more than just a few instances to support the generalization. Examples of such generalizations perhaps include cars have radios and barns are red. Here it seems quite important to the truth of these generics that a large portion of the kind have the property in question; these generics will therefore
26. Or predicate a striking propertysee below.
38
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition be nontroublesome. Even these generalizations, though, are subject to a negative/positive asymmetry among their counterinstances. Compare, for example, cars have radios to buildings are less than one thousand feet tall. The latter generalization, whose counterinstances are positiveand rather salientstrikes us as false. Generics that do not pertain to characteristic dimensions or predicate striking properties, then, require that the majority of the kind possess the relevant property. This alone does not suffice for their truth, however; their counterinstances must also be negative. Generics that pertain to characteristic dimensions can tolerate any number of exceptions as long as they are negative. Providing that there is some available positive value for the characteristic dimension, negative counterinstances will not undermine the truth of the generic in question. Negation and Generics The negative/positive distinction also lets us understand the function of negations embedded in generics, a topic that has been overlooked in the literature. I know of no account that handles negation in a satisfactory way. Consider the following generics: Fair coins come up heads (false). Fair coins dont come up heads (false). Mosquitoes dont carry the West Nile virus (false). It is evident from our judgments of the first two sentences in the group that we do not understand negations here as taking wide scope over the bare plural,27 else we would judge the second sentence to be true in virtue of the first being false. It is not immediately obvious then why mosquitoes dont carry the West Nile virus is not taken to be true, once we recognize that negation does not take wide scope. We have no account so far as to why Mosquitoes carry the West Nile virus, and Mosquitoes dont carry the West Nile virus
27. Carlson (1977) suggests that negations take wide scope over existentially interpreted bare plurals, but he does not mention the interaction between generically interpreted bare plurals. I do not know of anyone who has proposed a substantive theory of the interaction of negation and generically interpreted bare plurals.
39
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
could not both be true. The first could be true because it is a striking property (see below), the second because it holds of the preponderance of mosquitoes. Cohen (1996), for example, would predict that both are true: the first is a true relative generic and the second is a true absolute generic. Understanding generics as inherently contrastive, though, points the way toward an answer. In evaluating the second of the pair, we are pitting the mere absence of a property against a very striking positive property. One way to lack the property expressed by a negated predicate will normally be to possess its positive counterpart. Since in many, perhaps all, cases we think of negated predicates as expressing mere absences and their unnegated counterparts as expressing concrete positive properties, we would require there to be basically no instances of Ks that have the positive property. This explains the apparent strength of negations embedded in generics. Ks are not F carries some sense of Ks are never F; this is because just a few Ks being F is enough to make us judge the generic to be false when the contrast is between the absence of a property and its instantiation. Striking Properties The other main class of troublesome generics consists of generics in which the predicate expresses a property that is striking, often in virtue of being dangerous or appalling. These are the sorts of properties that one would like to know aboutif there is a nontrivial chance that one will encounter something with these traits, one would be well served to have some prior warning. Such generics are extremely tolerant of exceptions; it does not appear to count against them if the vast majority of the kind does not have the property in question; that some of the kind does is sufficient. One might be tempted to think that striking property generics are simply existentials in disguiseafter all, we saw in the introduction to this article that some bare plurals are perfectly able to receive existential interpretations. Perhaps we have simply misclassified these generics. I do not believe this is the right response. Sharks attack bathers seems to belong with dogs eat meat rather than dogs are in my yard. One might, for example, imagine a shark lover objecting to the crude generalization expressed by sharks attack bathers, while not disagreeing with the fact that there have been some incidents of sharks attacking bathers. This suggests that sharks attack bathers does indeed receive a generic interpretation.
40
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition The entailment pattern of striking-property attributions also resembles that of generics rather than existentials. Whereas the truth of dogs are on my lawn entails the truth of animals are on my lawn, mosquitoes carry the West Nile virus does not suggest to most speakers the truth of animals carry the West Nile virus. We can also insert a quantificational adverb into striking-property attributions with only a small change in meaning, but cannot do so for existentials; compare mosquitoes sometimes carry the West Nile virus with dogs are sometimes on the lawn. The latter, but not the former, means something quite different from the original adverb-free sentence. Thus these sentences with striking-property predicates have an entailment pattern like that of generics and unlike that of existentials.28 For these reasons, I will continue to classify them as such. As these sentences are to count as generics, a factor in our judgments must be how striking the predicated property is. It is not hard to imagine why a primitive mechanism of generalization might be biased toward emphasizing such properties. If we allow ourselves to indulge in evolutionary explanations, we might imagine that our ancestors who rushed to generalize upon seeing just one lion eat their companions fared better than those who were more cautious in their generalizing. Of course, the mechanism does not generalize such information to every kind that has members possessed of these properties. As noted above, we do not accept animals carry the West Nile virus or even insects carry the West Nile virus, even though both kinds have some members that carry it (namely those few mosquitoes). We also dont decide that accountants are murderers is true even though some accountants have been known to kill. The mechanism, I suggest, looks for a good predictor of the property in question; it avoids generalizing to overly broad kinds or to irrelevant kinds. In particular, for a kind to be the locus of a striking-property generalization, it seems that the members of the kind that lack the property must at least be disposed to have it. It is important, for example, that the virus-free mosquitoes be capable of carrying the virus. If there is no such shared disposition, the
28. Other features of the predicate here suggest that they should be considered generics rather than existentials; the predicates are individual level rather than stage level. It is widely agreed that individual-level predicatespredicates that express properties normally had by an individual for an extended part of its existencegive rise to generic interpretations, whereas stage-level predicatespredicates that express properties normally had only fleetinglygive rise to existential interpretations. See Carlson 1977 and Kratzer 1995 among others for discussion.
41
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
generalization is not made. (After watching Dr. Evil attach lasers to the heads of sea bass and set them on Austin Powers, we do not conclude that sea bass attack people. We understand that the behavior of these sea bass is traceable to Dr. Evils intervention and is not a reflection of their nature.) And it is on the grounds that there is no disposition shared among sharks that leads them to attack us that the friends of sharks often challenge the generalization that sharks attack bathers. It is only the odd demented great white shark that ever does so, and only then because it mistakes a person for a seal. (I leave to the reader to decide for herself whether she is convinced by this.) As before, the claim that something is a factor in our judgments of generics is lent plausibility by evidence of the factors more general psychological impact. There is experimental evidence that indicates our disposition to generalize information in proportion to how striking it is. Rothbart et al. (1978) conducted an experiment to test the effects of striking information on our estimates of statistical frequencies. Their subjects heard brief descriptions of the behavior of fifty people, forty of whom were well behaved and ten of whom engaged in criminal activities. For half the subjects, the described criminal acts were quite horrific, consisting of crimes such as rape and murder, while for the other half, the crimes were much less disturbingshoplifting, forgery, and the like. The subjects were then asked to estimate the frequency of criminals in the groups they had just heard about; they consistently provided higher statistical estimates of criminality for the more extreme group, even though the actual percentage of criminals was the same for both groups. Statistical generalizations, like the ones made by the subjects in the Rothbart et al. experiment, are of course a type of generalization. I have argued that generics express the cognitive systems default generalizations and that other generalizations involve inhibiting or overriding this default. Since inhibitory processes can fail in high-demand situations, we should not be surprised by the Rothbart et al. findings. Their subjects were given a large amount of information to handle fifty individual descriptions is a lot to try to recall and assess for statistical patternsso it is reasonable to consider this a high-demand task. When faced with such a task, the cognitive system may fail to fully inhibit its default processesin this case, the processes associated with generics. Thus if, by default, we are disposed to generalize striking information more readily than nonstriking information, then we would predict that in high-demand situations we might make overly high statistical generalizations concerning striking information.
42
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition Semantic Truth Conditions and Worldly Truth Makers We have described in some detail the cognitive mechanism involved in understanding generics and have begun to see how it is related to other mechanisms via, for example, inhibitory processing. Since this mechanism is responsible for our understanding of generics, providing an account of this mechanism has also allowed us to understand the circumstances in which generics are true or false. Though there may be a further refinement or two needed, we can describe the circumstances under which a generic of the form Ks are F is true as follows: The counterinstances are negative, and: If F lies along a characteristic dimension for the Ks, then some Ks are F, unless K is an artifact or social kind, in which case F is the function or purpose of the kind K; If F is striking, then some Ks are F and the others are disposed to be F; Otherwise, almost all Ks are F. I would suggest that these worldly truth specificationsthese descriptions of how the world must be for the sentence to be trueshould not be mistaken for semantically derived truth conditions, however. To illustrate the distinction, let us assume that a dispositionalist theory of color is correct: what it is to be red is to be experienced as red by standard observers in standard conditions. We can then specify the worldly truth makers for the claim Bob is red; this claim is true if and only if Bob is experienced as red by standard observers in standard conditions. This is a specification of the circumstances in the world that must obtain for Bob is red to be true. Such a specification does not tell us anything about the semantically derived, compositionally determined truth conditions for Bob is red however; they describe only how the world must be so as to conform to the claim. It may well be that for Bob to be experienced as red by standard observers in standard conditions, there must exist standard observers to experience him as such. Then, metaphysically speaking, the truth of Bob is red entails the existence of standard observers. It is in no way part of semantic competence to recognize that the truth of Bob is red entails that there exist standard observers, however. This is not plausibly a semantic entailment, but merely a metaphysical one.
43
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
The semantic truth conditions for Bob is red may well be no more than Red(Bob). This respects the compositional structure of the sentence; Bob is red is a singular statementthe application of a monadic predicate to a singular term. To substitute the regimented version of the worldly truth specifications results in the following monstrosity: Bob is red is true if and only if Gen x, y [StandardObserver(x) & StandardCondition(y)] [ExperiencesAs In(x, Bob, red, y)] But Bob is red is not a generic; it is a singular statement, and its semantic truth conditions should respect that. For this reason, and others, it is very often desirable to simply disquote individual expressions when giving semantic truth conditions. Any further analysis of individual expressions very often belongs to metaphysics rather than to semantics (see King 2002 for a related, though somewhat different view). Similarly, I am inclined to view the above truth specifications as worldly descriptions rather than semantically derived truth conditions. I propose instead that the best semantic truth conditions we can supply for generics will expose their logical forms by means of a tripartite structure, but will simply use the generic operator Gen (S. J. Leslie 2007, forthcoming). The worldly truth specifications provided at the beginning of this section are exactly that; they are specifications of how the world must be for a generic to be true. But just as Bob is experienced as red by standard observers in standard conditions was a poor candidate for the semantic truth conditions of Bob is red, the complex and disjunctive conditions listed above are poor candidates for the semantic truth conditions of generics. Just as we are best off using is red in the truth conditions of Bob is red, we are best off using Gen in the semantic truth conditions of generics. When it comes to providing semantic truth conditions for generics, we may well be best off disquoting the generic operator Gen.
Conclusion The central thesis of this essay is that our understanding of generics reflects our default mechanism of generalization. If this claim is correct, then we have an explanation of how children can grasp these apparently complex sentences before they grasp more theoretically tractable ones, such as those containing all. On this view, the odd features of
44
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition generics are not linguistically based pieces of knowledge that the child must string together, but rather are features of the childs innately given faculty of generalizationthe oddness of generics derives from the cognitive system itself. We isolated and identified these quirky features of the default mechanism of generalization and showed how they manifest themselves in psychological behavior that is independent of generics. The notion of characteristic dimensions, for example, may be central to our capacity to gather information efficiently. The effect of how striking a property is was seen to influence adult judgments, at least in relatively high-demand situations. If generics, which are influenced by strikingness, reflect our default method of generalizing, then we would predict these effects in adults. Just as children sometimes fall into error by failing to fully inhibit this default process, and thereby treat all and some like generics, adults fall into error by also failing to inhibit this default process. When adults are asked to make statistical generalizations under sufficiently difficult circumstances, they do not manage to implement the inhibitory processes necessary to keep the generalizations purely statistical. Their statistical generalizations thus bear the mark of their default generalizations, where striking information plays a central role. This does not, of course, mean that judgments concerning generics are erroneous. In making a generic generalization, the default process is what is requiredgeneric generalizations are not statistical generalizations. With them, there is no normative demand to inhibit the default; rather, the default is to be embraced.
References
Asher, N., and M. Morreau. 1995. What Some Generic Sentences Mean. In The Generic Book, ed. G. Carlson and F. J. Pelletier, 300339. Chicago: Chicago University Press. Baldwin, D. A., E. Markman, and R. Melartin. 1993. Infants Ability to Draw Inferences about Nonobvious Object Properties: Evidence from Exploratory Play. Child Development 64: 71128. Carlson, G. 1977. Reference to Kinds in English. PhD diss., University of Massachusetts, Amherst. Chomsky, N. 2000. New Horizons in the Study of Language and Mind. Cambridge: Cambridge University Press. Cohen, A. 1996. Think Generic: The Meaning and Use of Generic Sentences. PhD diss., Carnegie Mellon University.
45
xxx
pr0001
October 1, 2007
10:43
SARAH-JANE LESLIE
. 2004. Generics and Mental Representation. Linguistics and Philosophy 27: 52956. Condoravdi, C. 2005. The Middle: Where Semantics and Morphology Meet, MIT Working Papers in Linguistics, no. 11. Cambridge, MA: MIT Press. Dayal, V. 1999. Bare NPs, Reference to Kinds, and Incorporation. In Proceedings of Semantics and Linguistic Theory (SALT) 9, ed. T. Matthews and D. Strolovich, 3451. Ithaca, NY: Cornell University. Friedman, O., and A. M. Leslie. 2005. Processing Demands in Belief-Desire Reasoning: Inhibition or General Difficulty? Developmental Science 8: 218 25. Gelman, S. A. 2003. The Essential Child. New York: Oxford University Press. Gelman, S. A., J. Coley, K. Rosengren, E. Hartman, and A. Pappas. 1998. Beyond Labeling: The Role of Parental Input in the Acquisition of Richly Structured Categories. Monographs of the Society for Research in Child Development (SRCD), series no 235, vol. 63, no. 1. Ann Arbor, MI: SRCD. Graham, S. A., C. S. Kilbreath, and A. N. Welder. 2001. Words and Shape Similarity Guide 13Month-Olds Inferences about Nonobvious Object Properties. In Proceedings of the Twenty Third Annual Conference of the Cognitive Science Society, ed. J. D. Moore and K. Stenning, 35257. Hillsdale, NJ: Erlbaum. Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD diss., University of Massachusetts, Amherst. Hollander, M. A., S. A. Gelman, and J. Star. 2002. Childrens Interpretation of Generic Noun Phrases. Developmental Psychology 36: 88394. Jenkins, H. M., and R. Sainsbury. 1970. Discrimination Learning with the Distinctive Feature on Positive or Negative Trials. In Attention: Contemporary Theory and Analysis, ed. D. Mostofsky, 23975. New York: Appleton-CenturyCrofts. Kamp, H., 1981. A Theory of Truth and Semantic Interpretation. In Formal Methods in the Study of Language, ed. J. Groenendijk, J. Janssen, and M. Stokhof, 277322. Dordrecht: Forts Publications. King, J. 2002. Two Sorts of Claims about Logical Form. In Logical Form and Language, ed. Georg Peter and Gerhard Preyer, 11832. Oxford: Oxford University Press. Kratzer, A. 1995. Stage-Level and Individual-Level Predicates. In The Generic Book, ed. G. Carlson and F. J. Pelletier, 12576. Chicago: Chicago University Press. Krifka, M. 1987. An Outline of Genericity. Technical Report SNS-Bericht 8725. T ubingen: Seminar fur naturlich-sprachliche Systeme, T ubingen University. . 1995. Focus and the Interpretation of Generic Sentences. In The Generic Book, ed. G. Carlson and F. J. Pelletier, 23865. Chicago: Chicago University Press.
46
xxx
pr0001
October 1, 2007
10:43
Generics: Cognition and Acquisition
Krifka, M., F. Pelletier, G. Carlson, A. ter Meulen, G. Chierchia, and G. Link. 1995. Genericity: An Introduction. In The Generic Book, ed. G. Carlson and F. J. Pelletier, 1125. Chicago: Chicago University Press. Larson, R. 1998. Events and Modification in Nominals. In Proceedings from Semantics and Linguistic Theory (SALT) 8, ed. D. Strolovitch and A. Lawson, 14568. Ithaca, NY: Cornell University. Leslie, S. J. 2007. Generics, Cognition, and Comprehension. PhD diss., Princeton University. . Forthcoming. Generics and the Structure of the Mind. Philosophical Perspectives. Lewis, D. 1975. Adverbs of Quantification. In Formal Semantics of Natural Language, ed. E. L. Keenan, 315. Cambridge: Cambridge University Press. Macario, J. F., E. Shipley, and D. Billman. 1990. Induction from a Single Instance: Formation of a Novel Category. Journal of Experimental Child Psychology 50: 17999. Newman, J., W. Wolff, and E. Hearst. 1980. The Feature-Positive Effect in Adult Human Subjects. Journal of Experimental Psychology: Human Learning and Memory 6 (5): 63050. Nisbett, R. E., and L. Ross. 1980. Human Inference: Strategies and Shortcomings of Social Judgment. Englewood Cliffs, NJ: Prentice-Hall. Pelletier, F., and N. Asher. 1997. Generics and Defaults. In Handbook of Logic and Language, ed. J. van Benthem and A. ter Meulen, 112579. Cambridge, MA: MIT Press. Roeper, T., U. Strauss, and B. Z. Pearson. 2006. The Acquisition Path of the Determiner Quantifier Every: Two Kinds of Spreading. In Current Issues in First Language Acquisition, University of Massachusetts Occasional Papers 34, ed. T. Heizmann, 97129. Amherst, MA: Graduate Linguistic Student Association. Rothbart, M., S. Fulero, C. Jenson, J. Howard, and P. Birrell. 1978. From Individual to Group Impressions: Availability Heuristics in Stereotype Formation. Journal of Experimental Social Psychology 14: 23755. Sainsbury, R. 1973. Discrimination Learning Utilizing Positive or Negative Cues. Canadian Journal of Psychology 27: 4657. Shipley, E. 1993. Categories, Hierarchies, and Induction. In The Psychology of Learning and Motivation, 30, ed. D. Medin, 265301. New York: Academic Press. Shipley, E., I. Kuhn, and E. C. Madden. 1983. Mothers Use of Superordinate Category Terms. Journal of Child Language 10: 57188.
47
xxx
pr0001
October 1, 2007
10:43
48
Philosophical Perspectives, 21, Philosophy of Mind, 2007
GENERICS AND THE STRUCTURE OF THE MIND
Sarah-Jane Leslie Princeton University 1. Introduction: The Semantics of Generics

Generics are sentences such as ravens are black, tigers are striped, and mosquitoes carry the West Nile Virus. Such sentences occur frequently in everyday speech; we continually rely on these constructions to convey information to one another. It is far from clear, however, how we are to assign truth conditions to these sentences. If we focus our attention on generics such as tigers are striped, the situation does not seem too vexing. Tigers are striped cannot be understood as expressing a universal claim to the effect that all tigers have stripes; it is compatible with the existence of stripeless albino tigers in the way that the universal claim is not. Perhaps tigers are striped should be understood as equivalent to most tigers are striped, or generally tigers are striped. Alternatively, perhaps it means that all normal tigers are striped. Accounts to this effect can be found in a variety of guises throughout the literature (for example, Krifka, M., F. Pelletier, G. Carlson, A. ter Meulen, G. Chierchia, and G. Link. (1995); Greenberg (2003); Asher and Morreau (1995); Pelletier and Asher (1997)). Let us consider Asher and Morreau (1995), and Pelletier and Asher (1997) in more detail. These authors do not make the simple claim that a generic of the form Ks are F is true iff all normal Ks are F, but rather claim that such a generic is true iff for each individual K, the most normal worlds for that K according to a certain ordering of worlds, are such that that K is F. (The ordering of worlds is given by a contextually determined similarity metric on worlds.) Or less formally, we have that the generic is true iff each K would be F under the circumstances that are most normal for that particular K. Thus, while an albino tiger might in actuality lack stripes, this, we might suppose, is only because the actual circumstances are abnormal for that tiger, in as much as she is albino. A world in which that particular tiger was not albino would be a more normal world for that tiger than the actual world.
One might find such an account convincing in the case of such generics as tigers are striped. Consider, however, generics such as ducks lay eggs, lions have manes or peacocks have blue-green tails. No male ducks lay eggs, and no female lions have manes, nor do female peacocks sport blue-green tails. And unlike the case of an albino tiger, these creatures do not lack the relevant properties because actual circumstances are in some way abnormal for them. The most normal of circumstances will be ones in which any given male duck never lays an eggit would only be in the most abnormal or worlds that such a duck would lay eggs. These generics are intuitively true, yet it is not the case that for each given member of the kind, the most normal circumstances for that individual will be one in which they possess the relevant property. Pelletier and Asher (1997) attempt to rescue their account from the difficulties posed by such generics through quantifier domain restriction. Perhaps, when we say ducks lay eggs, we are not talking about ducks simpliciter, but rather are talking only about the female ducks. They suggest that this sort of domain restriction is acceptable as long as the restricted domain is a sub-kind of the kind in question; female ducks constitute a natural sub-kind of ducks, hence the truth of the above generic. Numerous puzzles arise if we adopt this proposal. For example, it is odd that ducks lay eggs and are female is not a true generic. If the predicate lays eggs requires the domain to consist only of female members, why can we not truly predicate is female here? Alternatively, why can we not restrict the domain to the male ducks, and thereby render true ducks dont lay eggs? The males surely constitute just as natural a sub-kind of ducks as do the females. Similarly, if domain restriction is permitted so long as the restriction is to a natural sub-kind why are generics such as Dogs are poodles and mammals are cows not true? It is hard to see why poodles and cows are not natural sub-kinds of dogs and mammals respectively. One might think that these are paradigmatic examples of natural sub-kinds. Difficulties continue as we consider generics such as mosquitoes carry the West Nile virus. This is a true generic, even though less than 1% of mosquitoes carry the virus. It is certainly not true that all normal mosquitoes carry the West Nile virus or that circumstances are somehow more normal for a given mosquito if it is infected. It is more plausible that all normal mosquitoes are, in fact, virus-freethat any circumstances in which a given mosquito is infected are quite abnormal for that mosquito. (If anything, the account under discussion here would seem to predict the truth of mosquitoes dont carry the West Nile virus.) The truth of mosquitoes carry the West Nile virus does not appear to have anything to do with what is normal for a mosquito. Since, if anything, it is quite abnormal for a mosquito to carry the virus, no further restrictions on the domain of normal mosquitoes will predict that the generic is true. No means emerge by which we may invoke domain restriction here. It seems that generics cannot be assimilated to all normal, or any such variant. We might wonder if some generics are made true by the predicated property being uniquely possessed by the kind in question. The truth of mosquitoes carry
Generics and the Structure of the Mind / 377
the West Nile virus, for example, might depend on mosquitoes being the viruss sole bearers. There is, I believe, a reading of the sentence to that effect, but it is not the only reading. The unique-possession reading is obtained by emphasizing mosquitoes, or paraphrasing as it is mosquitoes that carry the West Nile virus. Notice that one could disagree with such an assertion by claiming that deer ticks also carry the virus: if it is mosquitoes that carry the West Nile virus, then it cannot be that deer ticks carry it also. There is, however, another very natural reading of mosquitoes carry the West Nile virus on which there is no incompatibility between its truth and the truth of deer ticks also carry the virus. There is nothing contradictory in an assertion of mosquitoes carry the West Nile virus, and deer ticks do too. There is thus a reading of mosquitoes carry the West Nile virus whose truth does not require that mosquitoes be the only insects that carry the virus. This is the more natural reading of the statement, and is obtained so long as mosquitoes is not pronounced with stress. We readily judge the sentence to be true on this reading, and an account of generics must explain why this is so. It is also not generally sufficient for the truth of a generic that the kind in question be the unique possessor of the predicated property. Consider, for example, humans are autistic. Humans are the only species whose members are afflicted with this condition, but we do not judge the generic to be true. Similarly, humans are, roughly speaking, the only species to sport one-legged members, yet humans are one-legged is quite obviously false. Ariel Cohen (1996) offers an intriguing and sophisticated account of generics, which is perhaps the most subtle and promising account to be found in the literature. I consider his account at considerable length in Leslie (2007, 2008), however, and find it to be vulnerable to counterexample. Cohen begins by asking us to consider the notion of the contextually supplied alternatives to a predicate is F, which he denotes as Alt(F). A contextually supplied alternative is a natural, intuitive notion: the alternatives to drives to the department include walks to the department, bikes to the department, takes the bus to the department, and so on. The alternative to is female is is male, and the alternatives to lays eggs include bears live young, and so on. The predicate is F is always included in the set Alt(F), and the alternatives need not be incompatible with each other. Cohen also introduces the notion of the contextually supplied alternatives to a kind K, which he denotes by Alt(K). Again, the notion is highly intuitive: the alternatives to the kind Cat include other midsized mammals, and the alternatives to the kind Ant include other insects, and so on. With these notions in place, Cohen distinguishes between two different classes of generics: absolute and relative generics. The truth conditions for the two classes are given as follows:
Absolute generics: Ks are F is true iff the probability that an arbitrary K that satisfies some predicate in Alt(F) satisfies is F is greater than .5
378 / Sarah-Jane Leslie Relative Generics: Ks are F is true iff the probability that an arbitrary K that satisfies some predicate in Alt(F) satisfies is F is greater than the probability that an arbitrary member of Alt(K) that satisfies some predicate in Alt(F) satisfies is F.
An absolute generic, then, is true so long as the probability that the arbitrary relevant K (i.e. K that satisfies some predicate in Alt(F)) satisfies is F is greater than .5. For example, tigers have stripes is a true absolute generic, since the probability that the arbitrary tiger has stripes is greater than .5. Similarly, ducks lay eggs is a true absolute generic, because, of the ducks that perform some or other alternative to laying eggs (i.e. some or other means of gestating), the probability that the arbitrary such duck lays eggs exceeds .5. A relative generic is true so long as the probability that the arbitrary relevant K satisfies is F is greater than the probability that the arbitrary relevant member of the relevant alternative kinds satisfies is F. Mosquitoes carry the West Nile virus is a true absolute generic, since the arbitrary mosquito is far more likely to carry the virus than the arbitrary insect. Cohen adds a further condition to his account, which all generics must meet if they are to be true: the homogeneity condition:
Homogeneity: These conditions must hold for all salient partitions of the kind K.
That is, the probabilistic conditions specified above must hold in all the salient partitions of the kind. The notion of a salient partition is left somewhat vague in Cohens work, but it is fairly intuitive. For example, gender is a salient partition in animal kinds, and so if a generic is to be true of an animal kind, it must meet the conditions specified above for both the male and the female members of the kind. Thus, despite the fact that over 80% of chickens are female, chickens are female is a false generic. Even though the arbitrary chicken has a .8 chance of being female, this is not so amongst the male chickens. The probability of an arbitrary male chicken being female is, of course, zero, and so the generic is false. The homogeneity condition is quite strict, however. Cohen is able to classify generics such as chickens are female as false, only at the price of also classifying generics such as lions have manes, peacocks have fabulous blue-green tails, and cardinals are red as likewise false. The predicated properties in the latter three generics are possessed only by the males of the species, and so the homogeneity condition is violated. These generics are wrongly predicted to be false. It is also difficult to derive the result that dogs are four-legged is true, while dogs are three-legged is false, on Cohens account. If number of legs constitutes
a salient partition of the kind Dog, then dogs are four-legged will be false, since homogeneity will be violated. If, however, number of legs does not constitute a salient partition, then dogs are three-legged will be predicted to be true also! The arbitrary dog has a much greater chance of having three legs than your arbitrary mid-sized mammalthree-legged dogs live on easily with human help, while three-legged wolves, foxes and tigers are rarely so fortunate. Dogs are three-legged would thus be a true relative generic on Cohens view. There are quite a few other generics that are intuitively false, and yet are predicted to be true relative generics. Humans, for example, are far more likely than the arbitrary mammal to be blind, one-legged, or paralyzed, which should suffice to render true humans are blind, humans are one-legged, and humans are paralyzed. The truth conditions associated with relative generics, then, appear to be overly weak. They are also too strong, though, under some circumstances. As things stand, mosquitoes carry the West Nile virus is correctly predicted to be a true relative generic, since the arbitrary mosquito is far more likely than the arbitrary insect to carry the virus. Let us suppose, though, that fleas also carry the virus, and do so at a greater rate than mosquitoesmosquitoes carry the West Nile virus is still correctly predicted to be true, since the arbitrary mosquito is still more likely to carry the virus than the arbitrary insect. If, however, the flea population grew so rapidly as to vastly outnumber all other insects, at some point it would cease to be the case that the arbitrary mosquito would be more likely than the arbitrary insect to the carry the virus. At this point, mosquitoes carry the West Nile virus would fail to be a true relative generic on Cohens account. Intuitively, however, it would remain true. The size of the flea population has no bearing on the truth of mosquitoes carry the West Nile virus. Cohens account, then, is subject to counterexample, despite its ingenuity and promise. The foregoing discussion should indicate that the truth conditions of generics is no simple matter. The extant accounts of generics are fittingly complex; they invoke everything from iterated modalities to non-standard logics to comparative probabilities. I argue in detail in Leslie (2007, 2008) that each account is subject to systematic counterexample. There is no successful account of the truth conditions of generics to be found in the current literature. In the interest of brevity, I will not rehearse these criticisms here, but will rather describe my own positive account of generics. In light of this account, I will consider the question of whether generics are quantificational. Certainly, their logical forms closely resemble those of quantified statements. I will argue that, despite appearances, generics are in no sense quantificational. The generic operator Gen is a variable-binding operator that is used to express generalizations, and yet is not a quantifier. In the final portion of the paper, I argue that, while this is a surprising result from the point of view of philosophy of language, it is to be expected given the role of generics in our psychology.
2. Generics and Cognition

As I argue here and in Leslie (2007, 2008), no existing account of generics is successful. Counterexamples aside, however, it is instructive to consider just how complicated the accounts turn out to be. Consider Cohens account once again:
Absolute generics: Ks are F is true iff the probability that an arbitrary K that satisfies some predicate in Alt(F) also satisfies is F is greater than .5
Relative Generics: Ks are F is true iff the probability that an arbitrary K that satisfies some predicate in Alt(F) satisfies is F is greater that the probability that an arbitrary member of Alt(K) that satisfies some predicate in Alt(F) satisfies is F
Homogeneity: These conditions hold for all salient partitions of the kind K.
We may compare the complexity of Cohens account to the simple and concise definition we can provide for the quantifier all:
All Ks are F is true iff {x: x is a K} {x: x is F}
Other accounts of generics are no less complex; they may feature iterated modalities, or even invoke non-monotonic logics. This asymmetry in complexity between accounts of generics and accounts of explicit quantifiers such as all becomes even more remarkable when we consider data from language acquisition. Strangely enough, young children find generics easier to acquire and master than explicit quantifiers. Generics appear in childrens speech very early in development, significantly before explicit quantifiers do (Gelman 2003; Roeper, Strauss, and Pearson 2006). And under some circumstances, it has been found that three-year-old children will even interpret explicitly quantified statements as though they were generics (Hollander, Gelman, and Star 2002). It seems that they find generics so much easier to comprehend than quantified statements, they rely on this easier interpretation at times, rather than attempting to process the more taxing quantificational claim. Explicit quantifiers, whose semantics have proved quite tractable for the theorist, are more challenging for the young child than generics, which have remained baffling to linguists and philosophers throughout forty years of research.
That generics would be easier for children to acquire than explicit quantifiers is all the more puzzling in light of the fact that there is no articulated operator associated with generics. We do not say Gen tigers are striped in the way we say most tigers are striped. This is no quirk of English either; there is no known language that has a dedicated, articulated generic operator. Children, then, would not only have to master an immensely complex set of truth conditions, they would have to associate that set with the absence of an operator. Associations with absence are notoriously difficult for children to master,1 so it seems almost paradoxical that they ever master generics, let alone before they master quantifiers. Even if we set aside the acquisition patterns, the fact that no known language has a dedicated articulated generic operator is a fact that cries out for explanation. In every language that has been studied to date, generics have the least marked surface forms; it is natural to conjecture that this is a linguistic universal. Why should these strange, elusive generalizations always correspond to unmarked surface forms? Or to frame the question once more in terms of acquisition: how do all language learners come to know that it is these peculiar, quirky generalizations that are to be associated with the absence of operators? It is not, I think, an accident that natural languages least marked generalizations are also its most complex and unlearnable. Children do not ever learn truth conditions for generic claims. Rather, the generalizations that generic sentences express correspond to the cognitive systems most primitive, default generalizations. The ability to generalize pre-dates the acquisition of language; infants as young as 12 months readily form category-wide generalizations on the basis of experience with a few instances of the category (Graham, Kilbreath, and Welder 2001). There must, then, be an early-developing cognitive mechanism responsible for these most basic generalizations. This mechanism must be in place before the child begins to learn language, since pre-verbal infants have the capacity to generalize. Generics, I claim, give voice to the generalizations this primitive mechanism produces. The child, already possessed of the mechanism at the time of language acquisition, does not need to learn something new and complex; she need only continue to form the generalizations she has been making since her earliest days. I will argue in the next section that the strange truth conditional features of generics can be traced to what are, quite plausibly, some quirks and biases in this mechanism. (For more details concerning this section and the next, see Leslie (2007, 2008).) If the comprehension of generics invokes a mechanism that the child has already been using for some time, it is not surprising that the acquisition of generics occurs so earlywe would predict that this would be so. And if this mechanism is the cognitive systems most basic, default means of generalizing, we can begin to understand why generics have the least marked surface forms across languages. Chomsky (2000) suggests unmarked surface forms invoke cognitive defaults. For example, he notes that a sentence such as John climbed the mountain can only be understood to mean that John climbed up the mountain.
The unmarked form climbed the mountain is never interpreted to mean climbed down the mountain. Chomsky (2000) suggests that this may reflect a fact about our concept of climbingby default, we think of the act of climbing as involving climbing up an item, not down it. To obtain the non-default reading, we must explicitly include the preposition down, thereby employing a more marked form. We might think of such marked information as supplying an instruction to the conceptual system to the effect that it must deviate from its default means of proceeding. I suggest that explicit quantifiers serve to direct the cognitive system to generalize in specific, non-default mannersto form, say, universal or existential generalizations. Generics, in virtue of being unmarked, do not supply any such instructions. The cognitive system proceeds with its default, automatic means of generalizing. Thus it is no coincidental fact that generics always correspond to the least marked surface forms. If all this is correct, we would expect some further predictions to be borne out. If generics give voice to our most primitive, default generalizations, while explicit quantifiers, in contrast, require our conceptual system to actively diverge from this default, then we would expect that explicit quantifiers would be more difficult to process. We would expect that, if sufficiently heavy cognitive demands are placed on people, they might mistakenly treat quantified statements as though they were generics. I am currently collaborating with Sam Glucksberg and Sangeet Khemlani to investigate whether adults are susceptible to such errors. It has already been found, however, that young children are indeed susceptible. Hollander, Gelman and Star (2002) asked three-year-olds, four-year-olds, and adults a series of yes/no questions involving either generics, or the explicit quantifiers all and some. (For example, they asked their subjects are all fires hot?, are some fires hot? or are fires hot?) They found that the threeyear-olds gave the same pattern of responses, regardless of whether the questions involved quantified statements or generics. This pattern of responses, remarkably, was identical to the pattern of responses that four-year-olds and adults gave to the generic versions of the questions. (Naturally, the older subjects altered their responses when they were confronted with questions containing quantifiers.) The three-year-olds, however, responded as though they were faced with generics in all cases. The experimenters further found that the three-year-olds were able to successfully handle questions containing quantifiers, so long as they were only asked about a limited set of items that were immediately before them, rather than about categories in general. Thus, if the three-year-olds were presented with four crayons and asked are all the crayons in the box? or are some of the crayons in the box?, they responded appropriately. In Leslie (2007, 2008), I suggest that this is because the cognitive demands are lower when the children are asked to consider a limited number of items, and higher when they are asked to consider abstract categories. Scenarios of the latter type impose sufficiently high cognitive demands that the youngest children were unable to process explicit quantifiers.
Instead, they fell back on the less taxing default operations associated with generics. Thus the rather surprising data of Hollander, Gelman and Star (2002) is to be expected if the account of generics I am proposing is correct. There is a significant amount of empirical work that remains to be done, but the data that is currently available points to the theory I am advocating. GoldinMeadow, Gelman and Mylander (2005) studied congenitally deaf children who have never been exposed to either spoken language or sign language, and so go on to develop their own communicative gesture system called home signs, and found that these children routinely employ gestures that are most naturally understood as generics. The experimenters were extremely conservative in their classification of gestures as genericas opposed to episodic or otherwise specificbut they nonetheless found that congenitally deaf children produce generic gestures at (or at least extremely close to) the rate at which hearing children produce generic utterances.2 These findings are remarkable since the deaf children have never been exposed to any form of language, and so a fortiori have never encountered generics. Generic gestures are produced even in the absence of any adult model of their production. A further piece of tantalizing data is to be found in Everetts (2005) work on the Amazonian language Piraha. Everett claims that Piraha is devoid of explicit quantifiers, yet nonetheless features generics. Pirahas generics have truth conditions that correspond to the truth conditions of English generics. For example, panthers eat people is a true generic in Piraha, even though very few panthers actually eat people, as is birds lay eggs (Everett, pers. comm.). Everetts work on Piraha is in its early stages, and his claims stand in need of independent confirmation. It is fascinating to think, though, that there may be a language that is so minimal as to lack explicit quantifiers, but which nonetheless contains generics. There is, then, a strong case to be made for the idea that generics give voice to our most primitive generalizations. This claim alone, of course, does not shed much light on the strange truth-conditional nature of these statements. I will argue in the next section, however, that the otherwise puzzling truth conditions of generics become far more tractable against the backdrop of this theory. In particular, I will argue that the strange truth conditional features of generics can be traced to what might reasonably be quirks and biases in our most primitive mechanism of generalization.
3. The Fundamental Mechanism of Generalization

We can, I believe, identify four main features of our most primitive mechanism of generalization, which go a long way towards explaining our puzzling truth-conditional judgments concerning generics. It is clear that this mechanism ought to be an efficient information gathering mechanism, since it is our most basic and immediate means of
obtaining information about categories. One way such a mechanism might be efficient is for it to take advantage of regularities out there in the world. For example, animal kinds are quite similar to one another at a level of abstraction: Animal kinds, by and large, have characteristic noises, characteristic modes of locomotion, characteristic diets, characteristic methods of reproduction, and so on. I claim that this mechanism operates in part by identifying these dimensions of regularitycharacteristic dimensions, as I call themwithin a domain of kinds, such as the domain of animal kinds, the domain of artifact kinds, and so on. Once the mechanism has identified a characteristic dimension for a domain of kinds, it will then seek to fill in a value along the dimension for a given kind within the domain.3 If the mechanism is seeking only to find a value along a characteristic dimension, it can do so on the basis of very limited experience with members of the kind. If noise is determined to be a characteristic dimension for animal kinds, for example, then if one is faced with even a single instance of a novel animal kind, one is poised to take whatever noise that particular instance makes and project it, at least tentatively, across the kind. No statistical surveying is needed for this generalization to be made. The generalization can be made on as little evidence as an encounter with a single instance of the kind. (See Macario, Shipley and Billman (1990), and Nisbett, Krantz, Jepson and Kunda (1983) for experimental evidence that we do, in fact, employ some such means of arriving at generalizations.) The generalization is, of course, potentially subject to revision, though only under limited circumstances, as I discuss below in connection with the fourth and final feature of the mechanism. In the case of animal kinds, it is natural to suppose that reproduction is a characteristic dimension. Having identified reproduction as a characteristic dimension of animal kinds, we can fill in the value for ducks upon learning of the egg-laying ducks. The existence of even just a few egg-laying ducks will suffice to provide a value for this dimension. Since all we are looking to do is locate a value along a dimension, no statistical information about the percentage of ducks that lay eggs, etc, is needed. The fact that less than half of ducks lay eggs has no effect on this generalization. The second main feature of this mechanism is how it deals with information that is particularly striking, often horrific or appalling. Consider the generics:
Mosquitoes carry the West Nile Virus. Sharks attack bathers. Pittbulls maul children. Tigers eat people.
These generics are intuitively true, yet very few members of the kind in question actually possess the predicated property. When dealing with the sort of information that one would be well-served to be forewarned about, the mechanism
generalizes in the face of even just a few members of the kind possessing the property. Of course, it would be inefficient to generalize such information to every kind that has a member with the property. We do not judge animals carry the West Nile Virus or even insects carry the West Nile Virus to be true, even though both kinds have some members that carry the virus, since those viruscarrying mosquitoes belong to both kinds. The mechanism, I suggest, looks for a good predictor of the property in question, and thereby avoids generalizing too broadly. For a kind to be a good predictor of a striking property, I propose, the members of the kind that lack the property must at least be disposed to have it. It is important, for example, that the virus-free mosquitoes be capable of carrying the virus. If there is no such shared disposition, the generalization is not made. If, for example, an accountant or two are convicted of murder, we do not judge that accountants are murderers, because we do not think that accountants are in any way generally disposed to be murderers. The odd murdering accountant is a bad apple, in no way indicative of the nature of accountants in general. Interestingly enough, defenders of such claims as Muslims are terrorists routinely argue that there is something about Islamits very doctrinethat instills in its followers the disposition to perform terrorist acts. (See, e.g., B. Lewis 1990.) I discuss these issues further in Leslie (2007). The third feature of the fundamental mechanism of generalization deals with rather more neutral information. When faced with information that is neither striking nor found along a characteristic dimension for the given kind, it appears that the mechanism in question requires the majority of the kind to possess the property for it to be generalized. Consider, for example, barns are red, or cars have radios. These are true generics, but only because a large majority of the kind in question satisfies the predicate. Were only a small percentage of barns to be red, we would not consider barns are red to be true. The final main feature of this mechanism involves how it deals with the members of the kind that are exceptions to the generalizations. It matters very much whether they simply fail to have the property in question, or whether they have an equally vivid, concrete positive property instead. I claim that we are far more prepared to maintain our generalizations when faced with the former kind of negative counterinstances than when faced with the latter kind of positive ones. For example, let us say that we determine that reproduction is a characteristic dimension of animal kinds, and understand that birds, being animals, must therefore have reproduction as a characteristic dimension. We observe some egg-laying birds, and fill in lays eggs as the appropriate value. The male birds simply fail to lay eggs. They constitute merely negative counterinstances, since they do not possess some equally positive alternative property. Our generalization is easily retained in face of these male birds. Were we to learn of some birds that bear live young, however, the generalization would need to be weakened to birds lay eggs or bear live young. It cannot be maintained as is in the face of such positive counterinstances.
Alternatively, consider the true generic peacocks have fabulous blue-green tails. Female peacocks are unfortunate creatures that drably lack tails. Yet were they to have fabulous pink tails instead of the stumps they actually have, then peacocks have fabulous blue-green tails would not be true. The female peacocks would constitute positive counterinstances, and the generalizations could not be maintained. In these circumstances, the generalization would have to be weakened to peacocks have fabulous blue-green or pink tails.
4. Semantic Truth Conditions and Worldly Truth-Makers

We can now describe how the world must be for a generic of the form Ks are F to be true:
The counterinstances, if any, are negative, and: If F lies along a characteristic dimension for the Ks, then some Ks are F If F is striking, then some Ks are F and the others are disposed to be F Otherwise, the majority of Ks are F
These worldly truth specifications should not be mistaken for semantically derived truth conditions, however. The distinction is quite intuitive, though it is rarely drawn. Let us suppose, for example, that a dispositionalist theory of color is correct: what it is to be red is to be experienced as red by standard observers in standard conditions. The dispositionalist theory of colorwhich is clearly a metaphysical theory, not a semantic oneprovides us with the worldly truth makers for Bob is red; this claim is true iff Bob is experienced as red by standard observers in standard conditions. These are the only circumstances in which Bob is red will be a true statement. An analysis of these circumstances, such as the dispositionalist theory provides, is not a semantic analysis of Bob is red. It tells us nothing about semantically derived truth conditions for Bob is red. It tells us nothing about the semantic entailments of Bob is red, and it tells us nothing about the compositional structure of the sentence. The dispositionalist theory is a metaphysical theoryno more, and no less. For example, it may well be that for Bob to be experienced as red by standard observers in standard conditions, there must exist standard observers to experience him as such. If this is so, as is quite plausible, the truth of Bob is red metaphysically entails the existence of standard observers. The entailment is surely not semantic, however. It would be bizarre to think that it is part of being semantically competent with Bob is red that one recognize that the truth of the claim entails the existence of standard observers. If this was so, no one save the dispositionalists themselves would count as semantically competent with Bob is red! Considerations of compositionality, and a desire to keep semantics at least in touch with syntax suggest that the semantic truth conditions for Bob is red
may be no more complex than Red(Bob)that is, the application of a monadic predicate to a singular term. These semantic truth conditions line up well with the sentences syntax, and respect its compositional structure. To substitute the regimented version of the worldly truth specifications results in the following monstrosity:
Bob is red is true iff Gen x, y [StandardObserver(x) & StandardCondition(y)] [ExperiencesAs_In(x, Bob, red, y)]
where Gen is the generic operator. But Bob is red is not a generic; it is a singular statement, and to claim otherwise leads one far from the project of arriving at a semantics that mirrors sentences compositional structures. Such considerations, among others, often lead semanticists to simply disquote individual expressions when giving semantic truth conditions. (Cf. King (2002).) In giving semantic truth conditions for generics, we can make use of a tripartite structure to lay bare their logical forms (Heim 1982; D. Lewis 1975; Kamp 1981). A tripartite structure consists of an operatorusually a quantifiera Restrictor, and a Scope:
Quantifier x, . . . , z [Restrictor x, . . . , z] [Scope x, . . . , z]
The Restrictor specifies the domain over which the variables range, and the Scope specifies the property that is attributed to the relevant members of the domain. For example, we can represent most tigers are striped as:
Most x [Tiger(x)] [Striped(x)]
In the case of generics, we posit a generic operator Gen that occurs to bind variables and to relate the Restrictor to the Scope. The logical form of tigers are striped is represented as:
Gen x [Tiger(x)] [Striped(x)]
In general, generics are regimented as:

Gen x. . .z [Restrictor(x. . .z)] [Scope(x. . .z)]
I propose that the semantic truth conditions of generics will use the generic operator Gen, as above. The tripartite structure serves to lay bare the compositional structure of generics by dividing the material into Restrictor and Scope, and so is an essential part of the semantic truth conditions. If we pursue our analysis further, though, I suggest that we leave semantics behind us. The worldly truth specifications at the beginning of this section are specifications of how the world must be for a generic to be true. They are
not semantic truth conditions. Those disjunctive, non-compositional conditions are ill-suited to serve as semantic truth conditions of generics, just as Bob is experienced as red by standard observers in standard conditions is ill-suited as the semantic truth conditions of Bob is red. We ought to handle the generic operator disquotationally, just as we ought to handle the predicate is red disquotationally.
4.1. Inferences Involving Generics

One might object that it is all well and good to simply disquote the generic operator, but no satisfying account of generics can do without a metalanguage analysis of the operator. If we do not provide a metalanguage analysis, we do not predict any semantically valid inferences will be specifically licensed by the operator. Nicholas Asher made this point at length in his commentary on my work at a recent semantics workshop (Asher 2006). Asher views this as a shortcoming of my proposal, but I must admit I see it as an advantage. With one exception, which I consider below, I claim there simply are no inferences concerning Gen that hold in virtue of form.4 That is, while we endorse many inferences concerning a given generic, we invariably find other generics for which the corresponding inference patterns dont hold. Whatever inferences we may be disposed to make with a given generic, these inferences are not due to the semantics of generics alone. Other than one which I explain below, there is no schematic inference that holds regardless of the particular generic that figures in it Theorists who study inferences concerning generics do not tend to insist that there are deductively valid inferences licensed by them, but rather consider defeasibly valid, or default, inferences. The notion of defeasible validity is the central notion of non-monotonic logics. Defeasibly valid inferences are warranted inferences that may be retracted in light of additional information. They are inferences that hold given what is currently known, but may be given up in light of new information. A prominent example from this literature is as follows:
Birds fly. Tweety is a bird So, Tweety flies.
This inference is defeasibly valid, because the conclusion is warranted relative to the premises, but the conclusion should be rejected were we to discover additional information such as that Tweety is a penguin. Theorists in the default reasoning tradition such as Asher and Morreau (1995) and Pelletier and Asher (1997) have claimed that generics as such support
these inferences. They claim that the following schema is a defeasibly valid one:
Ks are F a is a K So, a is F
It is undeniable that we are inclined to endorse various instances of this schema, such as the above example concerning Tweetys flight. Asher and his collaborators take it that any semantic account of generics must explain why we endorse these instances.5 Recall from earlier discussion that these theorists propose accounts of generics that we may roughly gloss as the claim that Gen means all normal . This analysis, they claim, explains why people find the above inferences attractive: if we know that all normal Ks are F, then on assumption that a is a normal K (a reasonable assumption absent information to the contrary), we know that a is F. The above inference schema is only attractive, however, for a limited range of generics. Given that less than one percent of mosquitoes carry the virus, the following inference is not an attractive one:
Mosquitoes carry the West Nile Virus. Buzzy is a mosquito. So, Buzzy carries the West Nile Virus.
Nor is the following inference one that we should be inclined to endorse:

Ducks lay eggs. Beaky is a duck. So, Beaky lays eggs.
These examples are telling: the default inference schema does not hold for all generics. It is not warranted in virtue of the form of these sentences. It works well for some generics, but fails for others. In particular, the default reasoning inferences hold well for those generics for which the corresponding claim all normal Ks are F is also true, and it fails for those that dont. I would suggest that it is the often concomitant belief that all normal Ks are F that is underwriting the inferences here. We feel the inferences to be correct or incorrect depending on whether the corresponding all normal claim is true. Many true generics have true corresponding all normal claims; for such generics, the above inferences are attractive. However, as I discuss in the first section of this paper, a true all normal claim is neither necessary nor sufficient
for the truth of a generic in general, and those generics whose corresponding all normal claims are false do not participate in the default inferences. To conclude on the basis of considering this restricted range of generics ones in which the corresponding all normal claim is truethat the above inference schema is valid is an objectionable move. It is akin to studying the inferential profile of an inclusive disjunction by considering only the cases in which both disjuncts are true, and then claiming that inclusive disjunctions entail their two disjuncts. As noted above, I do believe that there is one inference involving generics that does appear to be universally acceptable:
Ks are F. Ks are G. So, Ks are F and G.
I do not know of any particular instances where we are reluctant to endorse this inference. It should be noted that this fact alone does not entail that the inference must be semantic, and entailed by a metalanguage analysis of Gen the data to be explained here is that whenever we accept the two premises, we also accept the conclusion. To capture this data, I suggest the following thesis concerning our fundamental mechanism of generalization: it generalizes a conjunction of properties to a kind iff it generalizes each conjunct to the kind. This is a thesis about the workings of this mechanism. If it is a true thesis, however, it must be noted that it explains the relevant dataour inclination to endorse all instances of the above schemaas well as any account that located the basis of the inference in a metalanguage analysis. Further, the account I am offering of the generic conjunction-introduction judgments here fares better than any extant account of generics when confronted with certain instances of the schema. Asher and his colleagues treat generics as restricted universals, and so explain the conjunction introduction schema as an instance of conjunction introduction in the scope of a universal quantifier. Consider, however, the following, intuitively valid argument:
Peacocks lay eggs. Peacocks have fabulous blue-green tails. So, Peacocks lay eggs and have fabulous blue-green tails.
An informal poll suggests that people are happy with the above inference. Yet no restricted universal account can explain why this is so, because there are no peacocks that satisfy the conjunctive predicate. No individual peacock lays eggs and has a fabulous blue-green tail, since only the females do the former, and only the males possess the latter. I know of no account of generics that explains
how a generic could be true even though no member of the kind satisfies the predicate, or is even disposed to satisfy it. If, however, generics with conjunctive predicates are judged true just in case each conjunct is generalized to the kind by the mechanism of generalization I have characterized, then we would expect that this statement would be judged true, even though no peacock satisfies the predicate.
5. Quantificational Behavior
Thus I do not take it as a failing of my account that I do not provide a metalanguage analysis of Gen. This does not mean that there are not semantic questions to be asked here. A further semantic question that we might ask is: Is Gen a quantifier? It is certainly a variable binding operator, and it is used to express generalizations, in the sense that generics involve commitments as to the properties of unencountered instances of the kind in question. The other examples of such items are all quantifiers: all, every, most, and so on. Should Gen be considered a member of this class? In his 1977 dissertation, Greg Carlson informally classifies generics as nonquantificational statements. They are not, he notes, about how many, or how much, in the way that quantified statements are. A quantified statement such as most Ks are F can be given in response to the question how many Ks are F? Upon being asked how many tigers have stripes? for example, one could respond by saying most tigers have stripes, or all tigers have stripes, and so on. One cannot, however, respond by saying tigers have stripes. This is simply not an appropriate response to the question. Of course, this is no more than a loose and informal characterization of what it is to be quantificational, however. Carlsons original account did not treat Gen as a variable binding operator relating a Restrictor and a Scope. Rather, he took the logical forms of generics to include a one-place predicate operator G that takes a predicate of individuals to a predicate of kinds, and treated the bare plural in sentence as a singular term referring to a kind. The logical form of tigers have stripes, for example, was G(Striped)(tigers). The predicate G(Striped) is a monadic predicate of kinds, and tigers is a singular term referring to the kind Tiger. Thus at the time of Carlsons observation, generics were assigned very different logical forms than those of quantified sentences. In light of the more recent views of the logical forms of generics, which posit an operator Gen that relates the material in the Restrictor to the material in the Scope, let us revisit the question of whether Gen could be considered a quantifier. The informal observation that quantifiers tell us how much or how many can be given a formal analog in terms of isomorphism invariance. If we restrict out attention to operators that bind a single variable, this requirement can be simplified. For such an operator this requirement reduces to the idea that, for it to be quantificational, it must be definable as a function from the
Figure 1.
cardinalities of sets to truth values (Westerstahl 2005; Larson and Segal 1995; Higginbotham and May 1981; Sher 1991; Glanzberg forthcoming). I will argue that there is no such definition of Gen.6 How exactly are we to understand the claim that quantifiers are definable as functions from set cardinalities to truth values? We begin by observing that any two monadic predicates is A and is B partition the universe of discourse into four disjoint subsets. If we let M be the universe of discourse, and let A and B be the sets {x: x is A} and {x: x is B} respectively, then we have the four-way partition represented in Figure 1. We can further define k, l, m, and n as follows:
j = |M-(AB)| k = |A-B| m = |AB| n = |B-A|
(Where |A| is the cardinality of A.) Consider now a regimented sentence of the form Qx[Ax][Bx]. For such an operator Q, we can formulate the following condition: If Q is a quantifier, then Q is definable as a function from k, l, m, and n to true or false. For example, we can give the truth conditions of the following sentence schemas as follows:
All x [Ax] [Bx] is true iff k = 0 Most x [Ax][Bx] is true iff m > k Some x [Ax][Bx] is true iff m > 0
A little reflection should suffice to show that no such truth conditions will be forthcoming for sentences of the form Gen x [Ax][Bx]. An example considered earlier provides a nice illustration of this. As things stand, peacocks have fabulous blue-green tails is true. Were female peacocks to have fabulous pink tails in place of the brown stumps they in fact have, peacocks have fabulous blue-green tails would not be true; only the weaker peacocks have fabulous blue-green or pink tails would be true in those circumstances. This thought experiment does not involve altering any of the relevant cardinalities, however. The numbers of peacocks, blue-green-tailed individuals, blue-greentailed peacocks, and non-blue-green-tailed non-peacocks all remained unaltered. That our intuitions about the truth of a generic can be altered by such facts that do not pertain to cardinalities reflects the fact that generics depend on far more than cardinalities for their truth and falsity. Other examples are forthcoming. Consider the true generic Mosquitoes carry the West Nile virus. We can regiment this as Gen x [x is a mosquito] [x carries the West Nile virus]. Let us assume that mosquitoes are the only things that carry the virus, and further imagine that all and only those mosquitoes that carry the virus are blind mosquitoes; every mosquito that carries the virus is blind, and every mosquito that does not carry it has excellent vision. Then the set {x: x carries the West Nile virus} is the same as the set {x: x is a blind mosquito}. Thus an operator that is sensitive only to the cardinalities of sets will not distinguish between them. If the truth of generics was definable in terms of set cardinalities, then mosquitoes are blind mosquitoes would be true in the circumstances described if mosquitoes carry the West Nile virus was true. Of course, we think the latter is true but the former false, which shows that Gen is not a function from cardinalities to truth values. The latter example suggests a stronger conclusion than that Gen cannot be defined as a function from cardinalities to truth values. It suggests that we cannot give adequate truth conditions for generics if we identify the semantic values of predicates with their extensions. In the example described, the set of things that carry the West Nile virus is the same as the set of things that are blind mosquitoes, and yet the two sentences have different truth values. Clearly, generics operate on semantic values more fine-grained than extensions. My reader may have noticed that there is a more liberal use of the term quantifier than I am countenancing hereone according to which even proper names are quantifiers (Barwise and Cooper 1981). (There is, however, some resistance to such a liberal usage, and many theorists reserve the term for operators that behave in the way I am characterizing here (Larson and Segal 1995; Higginbotham and May 1981; Sher 1991; Glanzberg forthcoming; May 1991).) This more liberal use of the term understands quantifiers to be functions from sets (i.e. predicate extensions) to truth values (rather than the more strict criterion that they be functions from set cardinalities to truth values). It is important to notice here that even on the liberal view, generics are not quantifiers. They are
not functions from extensions to truth values. Thus even a theory that counts proper names as quantifiers cannot count generics as such. It appears that even intensionsextensions across possible worldsare also too coarse-grained to provide an adequate treatment for generics. Consider, for example, the predicates in the last example rigidified with actually or in actuality. Such predicates are commonly understood to have the same extension in every possible world that they have the actual world. Nonetheless, mosquitoes in actuality carry the West Nile Virus is true, but mosquitoes in actuality are blind mosquitoes is false. These predicates have the same intensions, but the sentences have different truth values. In case one does not care for appeals to predicates rigidified by using actually, other examples are forthcoming. Consider a virus, say, the M-virus, that is necessarily only mosquito bornenothing else can carry it. (The metaphysical details behind the case are unimportant; it is certainly not semantically incoherent to suppose there could be such a virus.) So, necessarily, everything that carries the M-virus is a mosquito. The sets {x: x carries the M-virus} and {x: x is a mosquito that carries the M-virus} are the same sets, even across possible worlds. Correspondingly, the predicates carries the M-virus and is a mosquito that carries the M-virus have the same intensions. Nonetheless, intuitions are quite strong that mosquitoes carry the M-virus is true, while mosquitoes are mosquitoes that carry the M-virus is false.7 (Similarly, we might elaborate on the tale and say that the M-virus necessarily co-occurs with blindness in its carriers; intuitions about mosquitoes are blind mosquitoes remain unchanged.) This leaves us with the tantalizing suggestion that generics join propositional attitudes in requiring their semantics to operate on entities that are even more fine-grained than intensions. That aside, it is of course evident that Gen is in no sense a quantifier. Evidently, even a modalized version of Barwise and Coopers liberal sense of quantification, as described above, will still not count generics as quantifiers. Gen is not even a function from intensions to truth values. Generics may express generalizations in the sense that they involve commitments as to the nature of previously unencountered instances, but they are not quantificational. They are generalizations that are not fundamentally about how much or how many.8
6. Primitive Generalizations
Let us combine the observation that generics are not quantificational with the claim that they express the conceptual systems default generalizations. This entails that our most basic means of forming general judgments is not primarily concerned with how many. These cognitively primitive generalizations do not operate on set extensions, or any such abstraction. They are not grounded in such extensional or statistical information, but rather depend on factors such as how striking and important the information in question happens to be.
As surprising as this may initially seem, the claim fits well with the Two Systems view of cognition set forth by Daniel Kahneman and his colleagues. They argue that we can identity two systems of cognition. On the one hand, we have System 1, a fast, automatic, effortless lower-level system; on the other, System 2, a slower, more effortful higher-level system, whose workings are rulegoverned. These two systems can arrive at conflicting judgments, which is how we are able to identify the presence of two systems, rather than one. As in intuitive illustration, consider the following problem (Kahneman 2002):
A bat and a ball cost $1.10 in total. The bat costs $1 more than the ball. How much does the ball cost?
Most people report an initial inclination to answer 10 cents. This answer springs to mind quickly and automatically. It is quite opaque why this response comes to mind; it just seems like the right answer, at least at first. To arrive at the correct response of 5 cents, we need to invoke algebraic reasoning. We know that Bat + Ball = 1.10, and Bat = Ball + 1. Solving for Ball, we obtain 5 cents. The reasoning that leads to this response is perfectly transparent to us; we could explain our thought process to others, for example. It invokes conscious, deliberate reasoning that follows the rules of algebra. The initial response of 10 cents is supplied here by System 1; the response is fast and automatic, and it is quite opaque how we ever arrived at it. The response of 5 cents is supplied by System 2; it is arrived at by effortful, conscious, rule-governed reasoning. Kahneman and his colleagues argue that it is System 1 that underlies the results of their famous feminist bankteller experiment. In this experiment, subjects are given the following description:
Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations
They are then asked, under a variety of circumstances, to assess the probabilities of the proposition that
Linda is a bankteller
And the proposition that

Linda is a bankteller who is active in the feminist movement
Time and again, people assign a higher probability to the latter proposition than to the former. Of course, this amounts to assigning a higher probability to a conjunction than to one of its conjuncts, in violation of the rules of probabilities. There is a certain intuitive pull associated with such a response; it just seems that
Linda is more likely to be a feminist bankteller than a bankteller. It takes the stern application of probabilistic norms to appreciate that this is an error. Again, System 1 is responsible for the intuitive responsethe response that experimental subjects overwhelmingly provide. It takes System 2 to intervene and impose the laws of probabilities for us to resist the response given by System 1. System 1 is here neglecting the most basic set-theoretic aspect of probability. If the extension of an event-type A is included in the extension of an event-type B, then the probability that an event in A will obtain cannot exceed the probability that an event in B will obtainthis is why no conjunction can be more probable than its conjuncts. System 1 is not sensitive to these set-theoretic considerations. System 1 neglects information about quantity too. Our intuitive evaluations of painful procedures are often insensitive to the duration of the procedure, and depend instead on the peak amount of pain, and the end amount of pain, both of which are non-extensional attributes (Kahneman and Frederick 2002). Thus people will claim to prefer a painful procedure extended by a period of less intense pain to one that is not thus extended. People are also willing to pay the same amount to save endangered birds, whether they would be saving 2,000, 20,000, or 200,000 birds, and Toronto residents are willing to pay the same to clean up one or two polluted lakes in the region as they are to clean up all polluted lakes in Ontario. Various studies have also documented the neglect of base-rate information in making statistical estimates (Kahneman and Frederick 2002). In this vein, people are very often moved far more by concrete, vivid examples than by dry, though well-founded, statistical facts. Nisbett, Borgida, Crandall, and Reed (1976) offer a nice illustration of this. They ask us to imagine that we are deciding whether to buy a Volvo or a Saab. We read Consumer Reports and find the consensus there reported to be significantly in favor of Volvos. These reports are based on the experiences of several hundred consumers. We then encounter someone at cocktail party who exclaims: A Volvo! Youve got to be kidding. My brother-in-law had a Volvo. First, that fancy fuel injection computer thing went out. 250 bucks. Next he started having trouble with the rear end. Had to replace it. Then the transmission and the clutch. Finally, he sold it in three years for junk. The logically proper response is to add this one negative case to the overwhelmingly positive statistics based on several hundred casesi.e. to essentially discount it. It must be admitted, though, that most of us would think twice about getting a Volvo after hearing this testimony. The power of such a personal and vivid anecdote is remarkable. At the intuitive level, is it far more convincing than abstract statistical facts. There are various experimental analogs of the scenario described above; that people place more weight on a single anecdote than on statistically robust information has been well-documented. (See, e.g., Hamill, Wilson, and Nisbett 1979, Tversky and Kahneman 1972, Nisbett, Borgida, Crandall, and Reed (1976) Nisbett and Borgida 1975). Of course, if this is correct, my reader will be far more convinced by the Volvo anecdote than by a description of the experimental data, so I will say no more on this point.
The evidence surveyed so far suggests that System 1the more primitive systemis not particularly sensitive to information about how much or how many. I suggest that generics are judgments issued by System 1. They are thus non-quantificational; they do not depend on considerations of quantity, or any such information easily captured by set-theory. They are, however, automatic, effortless, and cognitively basic. Quantifiers, in contrast, express judgments issued by System 2, the rule-governed, extension-sensitive, higher-level system. Quantifiers do depend on considerations such as how much and how many. They are thus easily describable in the terms of set-theory. Generics share other characteristics with System 1 also. The judgments issued by System 1 are sensitive to the affective nature of information, such as how striking it happens to be. Rothbart, Fulero, Jenson, Howard, and Birrell (1978) conducted an experiment to test the effects of striking information on our estimates of statistical frequencies. Their subjects heard brief descriptions of the behavior of fifty people, forty of whom were well-behaved, and ten of whom engaged in criminal activities. For half the subjects, the described criminal acts were quite horrific, consisting of crimes such as rape and murder, while for the other half, the crimes were much less disturbingshoplifting, forgery, and the like. The subjects were then asked to estimate the frequency of criminals in the groups they had just heard about; they consistently provided higher statistical estimates of criminality for the more extreme group, even though the actual percentage of criminals was the same for both groups. These estimates, provided by System 1, were thus influenced by the striking nature of the information at hand. System 1 is also frame dependent; its judgments depend on how information is presented or described. For example, we are more likely to favor a medical procedure if it is described as having a 90% survival rate, than if it is described as having a 10% mortality rate. I wonder if generics might perhaps be rightly thought of as frame dependent. For example, people have the firm intuition that mosquitoes carry the West Nile Virus is true, but mosquitoes are mosquitoes that carry the West Nile Virus is false. In evaluating the second sentence, there is a feeling that we are inappropriately overlooking the mosquitoes that dont carry the virus, but this feeling is not present when evaluating the first, even though, of course, the two claims have the same exceptions.
7. System 1 and Errors of Judgment

Much of the work that has been done on understanding the nature of System 1 has focused on errors of judgment. It must be emphasized that this is an artifact of research methods, however; there is nothing inherently erroneous about System 1. The situations in which one can most confidently conclude that two distinct systems are operating are ones in which conflicts arise between the responses that people arrive at intuitively, and ones that, upon reflection, they recognize as
correct. Such a situation offers the most direct evidence for the existence of two systems. These circumstances obtain when people are given a task that calls for a rule-governed, quantity-sensitive response of the sort that System 2 provides, but nonetheless they rely on the easier, quicker, automatic judgment delivered by System 1, rather than invoking System 2s more arduous operations. This simply means that System 1 may be erroneously overused, not that it is in some way inherently erroneous. Some questions and tasks appropriately involve the working of System 1. For example, instead of asking subjects to rate the probabilities of Lindas being a bankteller rather than a feminist bankteller, we might have asked them to judge how much Linda resembles a bankteller rather than a feminist banktelller. No error would have been involved in judging that Linda more closely resembles a feminist bankteller than a bankteller. The task at hand called for a System 1 judgment, so no error is associated with providing one.9 Generics, I suggest, express System 1 judgments. There is no erroneous overuse of System 1 involved in uttering and understanding generics; the use of System 1 is appropriate to the production and comprehension of generics. Errors only arise when we interpret quantified statements as though they were generics. As mentioned earlier, there are situations in which three-year-olds do exactly this; they interpret quantified statements as though they were generics. These three-year-olds ought to employ System 2 to understand the quantified statements, but they incorrectly rely on the easier, more available judgments of System 1namely those associated with generics. One only makes such an error when one treats a quantified statement as a generic; treating generics as generics constitutes no error whatsoever.
8. Conclusion
Generics express our most primitive and fundamental generalizations. These generalizations are not about how much or how many in the way that quantificational statements are. Rather, they stem from System 1, a lower-level cognitive system that is not particularly sensitive to such considerations. Quantifiers, in contrast, belong to System 2; they invoke the more sophisticated workings of this higher-level system. Thus quantifiers are harder to process and understand than generics, and so are acquired later, and prove more difficult for children to master. The non-extension-based, non-quantificational generalizations expressed by generics may be more puzzling for the theorist than these more transparent quantifiers, but they are more basic and fundamental to human cognition.
Acknowledgments In January and February of 2006, an earlier version of this paper was presented to the Philosophy Departments at UC San Diego, UC Berkeley, Columbia University, the University of Michigan, New York University,
and to the Department of Linguistics and Philosophy at MIT. An earlier version was also presented at the Rutgers University Semantics Conference in September 2006, and at the Eastern Division of the American Philosophical Society in December 2006. I was fortunate enough to have Nicholas Asher and John Hawthorne respectively as commentators at those events, and I benefited enormously from their comments. I am further indebted to the audiences at those talks for their helpful comments and questions. I would also like to thank the following people for invaluable discussions of the ideas in this paper: Noam Chomsky, Shane Frederick, Susan Gelman, Lila Gleitman, Gil Harman, Irene Heim, Mark Johnston, Daniel Kahneman, Richard Larson, Alan Leslie, Sandeep Prasada, Gideon Rosen, and Kai von Fintel. Notes
1. See, for example, Sainsbury (1973). The difficulty of forming associations with absences is known in the literature as the Feature Positive Effect, and has been well documented over the last thirty years. 2. A gesture was classified as generic if it represented a prototypical property of the kind in question that was not present in context, and did not contain any further indications (such as being accompanied by a negation gesture) that it was intended to refer to a particular instance of the kind in that context. The children studied here have two means of invoking their subjects of predicationthey can either employ deictic gestures, or mimetic gestures that function as common nouns (Goldin-Meadow, Butcher, Mylander and Dodge 1994). If a particular instance of a kind was present in the context (e.g. a toy elephant was at hand), a gesture was classified as generic only if it reflected a stable, prototypical property of elephants that was not possessed by that particular elephant. For example, a pointing gesture followed by a trunk gesture would not be classified as generic if that particular elephant had a trunk. If the elephant in question lacked a trunk, then for the gesture to count as generic, it would have to lack any indication that it was intended to point out a particular feature of that elephantfor example, if it was accompanied by a negating gesture, then it would be understood to mean that that particular elephant does not have a trunk. The classification procedure was thus very conservative. Of course there is nonetheless a certain interpretative gap, and it is impossible to be certain that all the relevant gestures were in fact generics, as opposed to, say, normative claims to the effect that that particular elephant ought to have a trunk. However, even such a normative claim draws on generic knowledgethe child would only be disposed to make such a normative claim if she already believed the corresponding generic. See Prasada and Dillingham (2006a, b) for an intriguing discussion of the role of generic knowledge in licensing such normative beliefs. 3. Elizabeth Shipley (1993) suggests that much early learning proceeds in this manner. (She calls characteristic dimensions overhypotheses, but the idea is much the same.) Macario, Shipley and Billman (1990) provide experimental evidence that supports this idea.
400 / Sarah-Jane Leslie 4. Obviously there are semantically valid inferences involving generics in a loose sense, for example birds fly and tigers have stripes, therefore tigers have stripes. The validity of this inference is obviously not licensed by the fact that the conclusion is a generic, but is rather an instance of conjunction elimination. The fact that generics, on my account, are truth evaluable is enough to preserve the validity of these inferences. The putative inferences Asher and I are discussing here are ones that would hold in virtue of the fact that the premises or conclusion are generics. 5. The fact that we are inclined to endorse some particular inferences does not, of course, establish that the inference is semantic. I imagine that we are all inclined to defeasibly endorseabsent information to the contraryinferences such as Marys boyfriend has cheated on her three times, so hell cheat on her again, and Janes boyfriend has stolen money from her three times, so hell steal from her again, and so on. We should not conclude that there is some semantic property of the schema As boyfriend has done X three time, so hell do X again that underwrites these inferences. It is not for the semanticist to explain our defeasible endorsement of these inferences. 6. In order to make use of the simplified version of isomorphism invariance, I will focus on cases where Gen binds a single variable. Although Gen is unselective, and therefore capable of binding more than one variable, it will suffice to demonstrate that Gen is not isomorphically invariant in this simplest case. An operator fails to satisfy isomorphic invariance if there are some instances in which it is not isomorphically invariant. I will thus focus on generics of the form Gen x [Kx] [Fx], and show that the truth of such generics are not functions of the cardinalities of sets. 7. The explanation of these differential intuitions lies, I believe, in the nature of the counterinstances to the claims. Whether a counterinstance is positive or negative depends specifically on the generalization at handmale birds are intolerable, positive counterinstances to birds are female, since they have the positive alternative property of being male, but are permissible, negative counterinstances to birds lay eggs, as they simply fail to lay eggs, and do not have a positive alternative property in its place. We are, of course, considering the same birds in both caseswhether they are positive or negative counterinstances is relative to the specific property (understood (hyper-)intensionally) being predicated. Now, the mosquitoes that do not carry the M-virus do not possess a positive property that is an alternative to carrying the M-virus, but they do possess a positive alternative property to being a mosquito that carries the M-virus namely, being a mosquito that does not carry the M-virus. This subtle shifts turns merely negative counterinstances into positive ones, because they possess the positive, concrete property of being a mosquito that is virus-free. Intuitively speaking, it feels somehow that we would be overlooking the virus-free mosquitoes in judging true mosquitoes are mosquitoes that carry the M-virus, while this is not so in judging true mosquitoes carry the M-virus. The phenomenology is quite similar when we consider ducks are female and ducks lay eggs. I would suggest that this feeling of overlooking is a hallmark of the phenomenon of positive counterinstances. 8. (Technical note) Some NPs that are plausibly quantificational are built up out of an n-place quantifier (in the strict sense), and some predicate-denotations
Generics and the Structure of the Mind / 401 that are frozen in some of its argument positions (Westerstahl 2005). A simple example of this is the NP all red balls. This quantificational NP is built up out of the 2-place quantifier all, with its first argument place filled in by red balls, to create the 1-place quantifier all red balls. All red balls does not pass the cardinality test for quantifierhood, but since it is thus derivable from a strict quantifier, we may decide to count it as a quantifier. More complicated examples include NPs such as everyone except John and more women than men. If generic NPs could be understood as being built up out of an n-place cardinality operator with frozen argument places, then this would undermine the argument that they are not quantificational. It is vital to note, however, that the NPs above are all quantifiers in the liberal sense of Barwise and Cooper (1981). That is, they are all functions from predicate denotations to truth values. The ones mentioned above all partition the powerset of the domain of discourse into those sets of individuals that are mapped to True, and those that are mapped to False. If we begin with a cardinality operator and freeze some of its argument places with predicate denotations, we will not construct an operator that does otherwise. Certainly, we will never construct one that operates on more fine-grained entities than extensions or intensions. Thus hopes of arguing that generics are quantificational on these grounds will surely be dashed. 9. There may even be many situations in which System 1 is less prone to error than System 2. For example, in our interactions with other people, and our judgments concerning those interactions, we are often better off trusting our instincts than approaching the situation dryly and analytically. A recurrent theme in the television series Star Trek is the plight of the unfortunate Data, who lacks human intuitions, particularly those concerning social situations. His attempts to compensate for this lack of intuition with careful and considered analytic reasoning are all too often met with failure. When it comes to reading and responding to social cues, the careful, conscious reasoning of the sort we associate with System 2 is a poor substitute for the automatic, intuitive deliverances of System 1.
References
Asher, N. (2006). Comments on Leslies Generics and the Structure of the Mind. Presented at the Rutgers University Semantics Conference, September 2006. Asher, N. and Morreau M. (1995), What Some Generic Sentences Mean. In G. Carlson and F. J. Pelletier (eds.) The Generic Book. Chicago. Chicago University Press. 300339. Barwise, J., and Cooper, R. (1981), Generalized Quantifiers and Natural Language. Linguistics and Philosophy 4:159219. Carlson, G. (1977), Reference to Kinds in English. Ph.D. dissertation. Department of Linguisitics. University of Massachusetts, Amherst. Amherst, MA. Chomsky, N. (2000), New Horizons in the Study of Language and Mind . Cambridge University Press. Cambridge. Cohen, A. (1996), Think Generic: The Meaning and Use of Generic Sentences. Ph.D. dissertation. Department of Linguistics. Carnegie Mellon University. Pittsburg, PA. Everett, D. (2005), Cultural Constraints on Grammar and Cognition in Pirah a: Another Look at the Design Features of Human Language. Current Anthropology, August-October 2005.

Gelman, S. A. (2003), The Essential Child . Oxford University Press. New York. Glanzberg, M. (forthcoming), Quantifiers. In E. Lepore and B. Smith (eds.), Handbook of Philosophy of Language. Oxford University Press. Goldin-Meadow, S., Butcher, C., Mylander, C., & Dodge, M. (1994), Nouns and verbs in a self-styled gesture system: Whats in a name? Cognitive Psychology 27: 259319. Goldin-Meadow, S., Gelman, S. A. and Mylander, C. (2005), Expressing generic concepts with and without a language model Cognition 96(2): 109126. Graham, S. A., Kilbreath, C. S. and Welder, A. N. (2001). Words and Shape Similarity Guide 13-month-olds Inferences about Nonobvious Object Properties. In J. D. Moore and K. Stenning (eds.), Proceedings of the Twenty Third Annual Conference of the Cognitive Science Society. Erlbaum. Hillsdale, NJ. 352357. Greenberg, Y. (2003), Manifestations of genericity. Routledge. New York. Hamil, R., Wilson, T., and Nisbett, R., (1979), Insensitivity to sample bias: Generalizing from atypical cases. Journal of Personality and Social Psychology, 39: 578589. Heim, I. (1982). The Semantics of Definite and Indefinite Noun Phrases. Ph.D. dissertation. Department of Linguistics. University of Massachusetts, Amherst. Amherst, MA. Higginbotham, J. and May, R. (1981), Questions, quantifiers and crossing. The Linguistic Review 1: 4180. Hollander, M. A., Gelman, S. A., and Star, J. (2002), Childrens Interpretation of Generic Noun Phrases. Developmental Psychology 36(6): 883894. Kahneman, D., (2002), Maps of Bounded Rationality: A Perspective on Intuitive Judgments and Choice, Nobel Prize Lecture. Kahneman, D. and Frederick, S. (2002), Representativeness revisited: Attribute substitution in intuitive judgment. In T. Gilovich, D. Griffin, and D. Kahneman (eds.) Heuristics & Biases: The Psychology of Intuitive Judgment. Cambridge University Press. New York. 4981. Kamp, H. (1981), A theory of truth and semantic interpretation. In J. Groenendijk, J. Janssen and M. Stokhof (eds.), Formal Methods in the Study of Language. Forts Publications. Dordrecht. 277322. King, J. (2002). Two Sorts of Claims about Logical Form. In Georg Peter and Gerhard Preyer (eds.) Logical Form and Language. Oxford University Press. Oxford. Krifka, M., F. Pelletier, G. Carlson, A. ter Meulen, G. Chierchia, and G. Link. (1995). Genericity: An Introduction. In The Generic Book, edited by G. Carlson and F. J. Pelletier, 1125. Chicago: Chicago University Press. Larson, R. and G. Segal (1995), Knowledge of Meaning. MIT Press. Cambridge, MA. Leslie, S. J. (2007), Generics, Cognition, and Comprehension. Ph.D. dissertation. Department of Philosophy. Princeton University. Princeton, NJ. Leslie, S. J. (2008), Generics: Cognition and Acquisition. Forthcoming in The Philosophical Review. Lewis, B. (1990), The Roots of Muslim Rage, The Atlantic Monthly. Lewis, D. (1975), Adverbs of Quantification. In E. L. Keenan. (ed.), Formal Semantics of Natural Language. Cambridge University Press. Cambridge. 315. Macario, J. F., Shipley, E., and Billman, D. (1990). Induction from a Single Instance: Formation of a Novel Category. Journal of Experimental Child Psychology, 50: 179199. May, R. (1991), Syntax, Semantics, and Logical Form. In Asa Kasher. (ed.) The Chomskyan Turn. Blackwell. Oxford. 334359. Nisbett, R. E., & Borgida, E. (1975), Attribution and the psychology of prediction. Journal of Personality and Social Psychology 32: 932943 Nisbett, R. E., Borgida, E., Crandall, R., & Reed, H. (1976), Popular induction: Information is not necessarily informative. In J. S. Carroll & J. W. Payne. (eds.), Cognition and social behavior. Erlbaum. Hillsdale, N J. 227236.

Nisbett, R. E., Krantz, D. H., Jepson, C., and Kunda, Z. (1983), The Use of Statistical Heuristics in Everyday Inductive Reasoning. Psychological Review, 90: 339363. Pelletier, F. and Asher, N. (1997), Generics and Defaults. In J. van Benthem and A. ter Meulen (eds.) Handbook of Logic and Language. MIT Press. Cambridge, MA. 11251179. Prasada, S. and Dillingham, E. (2006a), Principled and statistical connections in common sense conception, Cognition 99(1): 7311. Prasada, S. and Dillingham, E. (2006b), Representation of Principled Connections: A Window onto the Formal Aspect of Common Sense Conceptions. MS under review. Roeper, T., Strauss, U. & Pearson, B. Z. (2006). The acquisition path of the determiner quantifier every: Two kinds of spreading. In Current Issues in First Language Acquisition, University of Massachusetts Occasional Papers, 34: 97129. edited by T. Heizmann,. Amherst, MA: GLSA. Rothbart, M., Fulero, S., Jenson, C., Howard, J., and Birrell, P. (1978), From Individual to Group Impressions: Availability Heuristics in Stereotype Formation. Journal of Experimental Social Psychology, 14: 237255. Sainsbury, R. (1973), Discrimination Learning Utilizing Positive or Negative Cues. Canadian Journal of Psychology, 27(1): 4657. Sher, G. (1991), The Bounds of Logic. MIT Press, Cambridge, MA. Shipley, E. (1993). Categories, Hierarchies, and Induction. In D. Medin. (ed.) The Psychology of Learning and Motivation, 30: 265301. Tversky, A. and Kahneman, D. (1972). Availability: A Heuristic for Judging Frequency and Probability. Cognitive Psychology, 4: 207232. Westerst ahl, D., (2005), Generalized Quantifiers, The Stanford Encyclopedia of Philosophy (Winter 2005 Edition), Edward N. Zalta (ed.), http://plato.stanford.edu/archives/ win2005/entries/generalized-quantifiers.
May 3, 2008
Time: 04:14pm
t1-v1.3
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45
If, Unless, and Quantification

Sarah-Jane Leslie
Keywords Conditionals quantification compositionality modality unless
1 Introduction: Quantified Conditionals and Compositionality

A language is semantically compositional if the meanings of its complex expressions are wholly determined by the meanings of their parts, and the manner in which those parts are combined. The belief that natural languages are semantically compositional has played a central role in contemporary semantics. The belief is not stipulative, but is an empirical claim. It thus is conceivable that we might discover a counterexample to the thesis that natural languages are semantically compositional. We might, for example, discover that there are complex natural language constructions whose meanings do not depend solely on their parts, and the way in which those parts combine. A few such putative counterexamples have been discussed over the last thirty years, and one highly influential example was discussed by James Higginbotham in 1986. Higginbotham argued
S.-J. Leslie Department of Philosophy, 1879 Hall, Princeton, NJ 08544, USA
C. Viger and R. J. Stainton (eds.), Compositionality, Context and Semantic Values, xxxx DOI: 10.1007/978-1-4020-8310-5 1, C Springer Science+Business Media B.V. 2008
UN
CO RR
EC
Abstract Higginbotham (1986) argues that conditionals embedded under quantifiers (as in no student will succeed if they goof off) constitute a counterexample to the thesis that natural language is semantically compositional. More recently, Higginbotham (2003) and von Fintel and Iatridou (2002) have suggested that compositionality can be upheld, but only if we assume the validity of the principle of Conditional Excluded Middle. I argue that these authors proposals deliver unsatisfactory results for conditionals that, at least intuitively, do not appear to obey Conditional Excluded Middle. Further, there is no natural way to extend their accounts to conditionals containing unless. I propose instead an account that takes both if and unless statements to restrict the quantifiers in whose scope they occur, while also contributing a covert modal element to the semantics. In providing this account, I also offer a semantics for unquantified statements containing unless.
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90
that, when a conditional containing either if or unless is embedded under a quantifier, as in no student will succeed if they goof off, the meaning of the conditional varies depending on the nature of the quantifier in whose scope it occurs. Much discussion has come in the wake of Higginbothams 1986 article, such as Pelletier (1994a, b), Janssen (1997), von Fintel (1998), von Fintel and Iatridou (2002), and Higginbotham (2003). That quantified conditionals pose a challenge to the idea that natural languages are semantically compositional has acquired an almost folkloric status, and is frequently discussed in surveys and encyclopedia entries on compositionality. Pelletier discusses the possibility in his 1994 survey article on compositionality (1994b); in his Handbook of Logic and Language paper on compositionality, Janssen discusses the possibility that quantified conditionals may constitute a counterexample to the thesis that natural language are compositional, and in a Stanford Encyclopedia of Philosophy entry on compositionality, Zoltan Szabo discusses quantified conditionals as a possible counterexample to this thesis. I do not believe that quantified conditionals behave in a non-compositional manner. I will begin by considering conditionals that contain if, and consider a very simple account of their compositional structure. This simple account, which treats embedded conditionals as predicates of their quantified subjects, delivers satisfactory truth conditions for the most part, but runs into difficulties with conditionals that do not obey the principle of Conditional Excluded Middle. For that reason, I reject this simple account, and instead argue that an account that takes if-clauses to restrict quantifiers delivers the desired results, so long as we recognize that there is a covert modal element in the semantics of quantified if-statements. I then consider quantified unless-statements, and propose a parallel account. We should understand quantified unless-statements as restricting the quantifying determiners in whose scope they occur, while also contributing a covert modal element to their semantics. In order to provide such an account, however, we need to understand the semantics of the unquantified versions of these statements, and so I develop a semantics for unquantified unless-statements. The account of quantified if and unless-statements I propose here provides a uniform meaning of if and unless; their semantics do not vary depending on the nature of the quantifier in whose scope they occur. We need not ascribe any sort of chameleon-like semantics to if and unless, which would have their meaning depend on the nature of the quantifier under which they are embedded. Some of the discussion of Higginbothams claim has centered on the question of whether a chameleon-like semantics for conditionals would constitute a genuine counterexample to compositionality, or whether the principle of compositionality is sufficiently vague as to absorb the possibility (Pelletier, 1994a ,b; Janssen, 1997). The principle of compositionality is sufficiently vague so as to encompass a variety of precisifications. Some of the more liberal formulations of the principle are arguably compatible with an items possessing a chameleon-like semantics, though the stricter formulations are not. I will not take up the question of whether compositionality is compatible with a chameleon-like semantics for an item, but will rather
UN
CO RR
EC
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3

91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
2 The Puzzle of Quantified Conditionals
Higginbotham (1986) claims that if makes a different semantic contribution in (1) and (2) below, as does unless in (3) and (4): (1) (2) (3) (4) Every student will succeed if they work hard. No student will succeed if they goof off. Every student will succeed unless they goof off. No student will succeed unless they work hard.
Higginbotham claims that (1) can be understood to contain a material conditional, and (3) an inclusive logical disjunction, and so no puzzle arises for those sentences. But if (2) were to contain a material conditional, then (2) would be true if and only if every student goofed off and didnt succeed. Similarly, if (4) were to contain a disjunction, it would be true if and only every student both failed to work hard and failed to succeed.1 Those truth conditions are not appropriate to the English sentence, however: (2) does not seem to entail that every students goofs off, and
1 No student will succeed if he goofs off is equivalent to: for every student, its false that he will succeed if he goofs off, which in turn is equivalent to: for every student, he will goof off and he wont succeed. Similarly, No student will succeed unless he works hard is equivalent to: for every student, its false that he will succeed unless he works hard, which, on the assumption that unless means or, is equivalent to: for every student, he will not succeed and he will not work hard. Here and for the rest of the paper I will make occasional reference to the truth functional equivalence of no x (A) and every x (not A) when both quantifiers have wide scope over the sentence, as do Higginbotham (2003) and von Fintel and Iatridou (2002). This is not intended as a claim about the semantics of no, nor as a claim that the two constructions are everywhere intersubstitutable, but merely as the observation that they are truth functionally equivalent when they have wide scope over the sentence in which they occur.
UN
CO RR
Elementary inferences involving these [subordinating conjunctions] proceed very well when they are understood as truth functional connectives, the material conditional [for if] and the non-exclusive or [for unless] . . . The puzzle that I wish to discuss is independent of the issues of most prominent concern in that literature [on the semantics of conditionals], and it will be just as well to state it initially with the understanding that these classical terms of logical theory are truth functional. The puzzle is this: the words if and unless seem to have different interpretations, depending on the quantificational context in which they are embedded.
EC
TE
He claims that, while the if and unless in (1) and (3) have the semantic values they would have were they not embedded under quantifiers, the if and unless in (2) and (4) have different semantic values altogether. It is important to notice here that Higginbotham (1986) is assuming that indicative conditionals have the semantics of material conditionals:
DP
RO O
argue that the proposed chameleon-like semantics does not even accurately capture the truth conditions of the relevant English sentences, and will offer a uniform semantics in its place.
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
(2) No student goofs off and succeeds. (4) No student succeeds and doesnt work hard.
(1) (1) (2) (2)
Every student will succeed if they work hard. Every student will either succeed or not work hard. No student will succeed if they goof off. No student goofs off and succeeds.
(For clarity, I have formulated the material conditional in (1) as a disjunction.) These putative paraphrases do not adequately capture the truth conditions of the English sentences (1) and (2).2 To see the intuitive non-equivalence of (1) and (1), consider poor Bill, who, no matter how hard he works, will never succeed at calculus. Bill knows this, and does not in fact try hard in his calculus class since he knows it is futile. Bill will then satisfy the material conditional in (1), since he does not satisfy its antecedent the equivalent disjunction will either succeed or not work hard is satisfied by Bill in virtue of his failing to work hard. Thus (1) may be true of a class containing Bill, since Bill presents no obstacle to its truth. But is (1) true if Bill is among the relevant students? The answer is quite clearly no. Bill is a student in that class, and so it is simply not true that every student will succeed if they work hard. Bill is a clear counterexample to this; no matter how hard he works, he wont succeed in this class. Counterexamples to the paraphrasing of (2) by (2) also exist. Imagine a student in a New Jersey high school lets call her Meadow whose father has managed to scare the life out of her teacher. This teacher has no intention of giving Meadow anything less than an A in his class, no matter what she does. So it is simply not true that no student in the class will succeed if he goofs off, for Meadow will succeed no matter what she does. It so happens, though, that Meadow is quite interested in the
2 Higginbotham (2003) and von Fintel and Iatridou (2002) discuss counterexamples of this nature, though they use them to object to restrictive analyses, which I will consider below. I am indebted to them for the structure of the counterexamples presented in this section of the paper.
UN
CO RR
EC
TE
DP
These truth conditions contrast with the truth conditions of (1) and (3), where if and unless contribute a material conditional and an inclusive or respectively. Higginbotham concludes then that if and unless make different contributions depending on the nature of quantifier they are embedded under. This, he claims, is a counterexample to compositionality. We should wonder whether Higginbothams (1986) analysis adequately captures the truth conditions of (1)(4). He proposes that (1) and (2) can be analyzed as (1) and (2):
RO O
(4) does not entail that every student will fail to work hard. In fact both (2) and (4) are intuitively compatible with every students recognizing them to be true and working hard as a result. Higginbotham (1986) notes this, and suggests that the truth conditions of (2) and (4) are rather given by (2) and (4):
May 3, 2008
Time: 04:14pm
t1-v1.3

181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
3 The Semantics of Conditionals Containing If

Let us set aside conditionals that contain unless for now, and focus on ones that contain if. Unless-statements are considerably more complex than ifstatements, so it will be helpful to first formulate an account of if-statements. I will take up unless-statements in Part III of this paper.
3.1 A Simple Solution

There is a tempting solution to Higginbothams puzzle of quantified conditionals, which would seem to let us deliver a fully compositional account in a most straightforward manner. To see this Simple Solution, let us put aside worries specific to conditionals for a moment, and consider the truth conditions of quantified sentences in general. Standard accounts of quantified sentences of the form Q Ns VP assign
UN
CO RR
Intuitively, (3) cannot be true of any class that contains Bill, who will fail no matter what he does. But if Bill is again aware of his predicament, and so resolves not to waste his time trying in vain, then (3) may be true of a class containing Bill. Bill satisfies the disjunction succeeds or goofs off, and so poses no obstacle to the truth of (3). Thus we might have a class that contains Bill, of which (3) is false but (3) is true. Similarly, if Meadow is amongst the relevant students, (4) cannot be true, since she will succeed no matter what. It is intuitively false that no student will succeed unless she works hard, if Meadow is one of the students. If Meadow is once again interested in the subject matter and so elects to work hard, however, (4) may be true of a class containing her. If Meadow works hard, then she will not satisfy the conjunction succeeds and doesnt work hard, and so (4) might still be true of Meadows class. Thus neither (3) and (3), nor (4) and (4) are equivalent. Thus Higginbothams (1986) non-compositional account does not even adequately capture the truth conditions of quantified conditionals, and so is untenable.
EC
TE
DP
(3) (3) (4) (4)
Every student will succeed unless they goof off. Every student either succeeds or goofs off. No student will succeed unless they work hard. No student succeeds and doesnt work hard.
RO O
subject matter, and does not in fact goof off. Meadow is no obstacle to the truth of no student goofs off and succeeds, then, since she does not goof off, and so does not satisfy the conjunction goofs off and succeeds. Thus it can be true of a domain containing her that no student in it goofs off and succeeds. While (2) cannot be true of a class that includes Meadow, (2) can be, so we must reject (2) as an analysis of (2). These same counterexamples tell against Higginbothams (1986) analysis of (3) and (4) as (3) and (4):
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
to them truth conditions that depend on how many of the Ns possess the property denoted by the VP in particular on whether the number or portion of Ns required by the quantifying determiner Q possess the property denoted by the VP. A relevant question to ask, then, is whether the truth of quantified conditionals depends on how many of the relevant items possess the conditional property, or to put it in terms that do not make reference to conditional propertieshow many of the relevant items satisfy the open conditional in question? Von Fintel and Iatridou (2002) argue that we can indeed provide a fully compositional account of quantified conditionals in this manner, and I argued as much myself in Leslie (2003a, b). While the Simple Solution offers an elegant, appealing and uniform treatment for the majority of cases, I will argue in the next section that, if we pursue the Simple Solution, we will be forced once again to adopt a chameleon-like semantics for if in a limited number of cases. I will take this to be good reason to look for an alternative account. Let us consider in more detail how the Simple Solution would proceed. We saw above that the unfortunate Bill raised difficulties for Higginbothams noncompositional account of conditionals, since his presence is enough to render false every student will succeed if they work hard, even if Bill does not in fact work hard. On the Simple Solution, we would predict that Bill would falsify every student will succeed if they work hard iff Bill fails to satisfy x will succeed if x works hard. Intuitively, Bill does not satisfy this conditional: it is false that Bill will succeed if he works hard. Thus we would predict that Bills presence would be incompatible with the truth of every student will succeed if they work hard. Similarly, we would predict that Meadow would indeed be a counterinstance to the claim no student will succeed if they goof off. For the quantified statement to be true of a domain containing Meadow, Meadow would have to fail to satisfy x will succeed if x goofs off. However, on any natural interpretation of the conditional, it is true that Meadow will succeed if she goofs off. It is clear, then, why no student will succeed if they goof off cannot be true of a class that includes Meadow. This treatment is completely compositional with respect to the contribution of the conditional to the truth conditions of the entire sentence. It is also completely independent of any particular semantic treatment of conditionals themselves. We have offered no explanation of when an object satisfies the embedded conditional; this account of how conditionals compose appears to be independent of whatever the ultimate account of semantics for conditionals turns out to be. Just as it is not necessary to provide an account of when an item satisfies the predicate is F in order to highlight the compositional structure of Q Ns are F, if the Simple Solution was to succeed, it would not be necessary to provide an account of when an item satisfies an open conditional in order to see that a compositional analysis of quantified conditionals is possible. The Simple Solution, then, is an appealing option, and thus far it appears to handle our data correctly. There is, however, a class of if-statements that are not well handled by the Simple Solution, namely those if-statements that do not obey the Law of Conditional Excluded Middle. I will also argue that unless-statements are simply not amenable to anything like the Simple Solution, but first let us consider those quantified if-statements that resist the Simple Solution.
UN
CO RR
EC
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3

271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
3.2 Conditional Excluded Middle
Higginbotham (2003) and von Fintel and Iatridou (2002) each claim that we can give compositional interpretations to the troublesome quantified conditionals, but only if we assume that if-statements obey the law of Conditional Excluded Middle (CEM). Higginbotham describes this principle as follows: Writing the Stalnaker conditional as , we have the validity of (CEM), or Conditional Excluded Middle: (CEM) ( ) ( ) (2003, p. 186)
Then by assumption its false that a is Q, if it is P is not equivalent to if a is P, then its false that a is Q, though its false that: a is Q, if it is P is nonetheless interpretable and acceptable. Then (6) should also be interpretable and acceptable: (6) No x is Q if x is P. (6) should be true just in case none of the relevant items satisfy the open conditional x is Q, if x is P. That an item can fail to satisfy the open conditional without satisfying if x is P then its false that x is Q should not affect our analysis. There is
UN
(5) a is Q, if it is P.
CO RR
Higginbotham is reluctant to endorse CEM, but takes it to be the only means of giving a compositional account of quantified conditionals. He writes, Compositionality can be restored under certain assumptions [namely CEM] about the meaning, or the presuppositions, of conditionals. However, I am not aware at present of any way of grounding these presuppositions that is not stipulative (p. 182). Von Fintel and Iatridou (2002) are more enthusiastic in their endorsement of CEM, since it is part of a theory of conditionals to which von Fintel is antecedently committed. Neither von Fintel and Iatridou, nor Higginbotham provide much explanation of why they believe CEM is a necessary assumption when analyzing quantified conditionals in particular, however. That von Fintel and Iatridou would assume CEM is perfectly understandable, since one of the authors has defended such an analysis of conditionals elsewhere. It is less than clear from his 2003 paper, though, why Higginbotham feels obliged to accept CEM. The Simple Solution is just a way of dealing with quantified conditionals, and so should be neutral on the truth of CEM. If CEM is a true principle governing unquantified conditionals, then it should also govern quantified ones, but if certain unquantified conditionals do not obey CEM, we have no explanation of why these conditionals should suddenly obey it when they appear under a quantifier. The Simple Solution made no assumptions whatsoever about the semantics of unquantified conditionals we gave the truth conditions of quantified conditionals solely in terms of how many items satisfied or failed to satisfy the embedded conditional. If we encounter a conditional that does not obey CEM, then, we should be able nonetheless to analyze quantified versions of that conditional compositionally. Suppose, for example, (5) is a conditional that does not obey CEM:
EC
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360
(7) This fair coin will come up heads if flipped.
(8) This fair coin will not come up heads if flipped
(9) No fair coin will come up heads if flipped.
One might deny that (7) and (8) really are false, and claim instead, for example, that they are simply indeterminate, or lack a truth value. Certainly the defender of CEM as a general principle should argue for some such claim. I will not discuss such a possible defense here, but rather the discussion will proceed on the highly intuitive assumption that this is a genuine counterexample to CEM. It is worth noting, though, that it is far easier to convince oneself that (7) and (8) are indeterminate, than it is to convince oneself that that their quantified counterparts (9) and (10) are: (9) No fair coin will come up heads if flipped. (10) Every fair coin will come heads if flipped. (9) and (10) strike most people as quite clearly false. Thus even if one is inclined to reject (7) and (8) as counterexamples to CEM on the grounds that they are indeterminate rather than false, one still needs an explanation of why (9) and (10) seem quite clearly false and not at all indeterminate. Any natural extension of the simple solution to cases of indeterminacy would predict that the quantified statements should be indeterminate if their embedded conditionals are indeterminate.
4
I am indebted to Jim Higginbotham and David Chalmers for pointing this out to me.
UN
CO RR
But (9) strikes us as false under these circumstances!4 (9) expresses a much stronger claim: (9) would be true only if each coin was sure not to come up heads if flipped. That is, (9) is true iff each coin satisfies the open conditional x will not come up heads if x is flipped. A friend of the Simple Solution might respond by invoking CEM here. The conditional x will not come up heads if x is flipped is related to (9)s embedded conditional via CEM: if CEM holds, then an item can fail to satisfy x will come up heads if flipped if and only if it satisfies x will not come up heads if flipped.
EC
However, (8) seems to be false also; it seems that we have a conditional that does not obey CEM.3 Let us now consider how the Simple Solution handles quantified conditionals whose embedded conditionals do not satisfy CEM. The above discussion suggests that the arbitrary fair coin fails to satisfy x will come up heads if x is flipped. The Simple Solution would then predict that (9) would be true of any given collection of fair coins, since each fair coin will fail to satisfy the embedded conditional:
TE
DP
Suppose that we have a fair coin before us, and we are contemplating what will happen if we decide to flip it. On the assumption that the coin in question really is fair, (7) is intuitively false. Since (7) is a false conditional, if it obeyed CEM, then (8) would be true:
RO O
nothing in the account presented here that even suggests that CEM is an assumption required to provide a semantics for quantified conditionals. It is a controversial matter whether CEM is a principle governing all conditionals, or whether there are some that do not obey it. A good candidate for a conditional that does not obey CEM is (7):
May 3, 2008
Time: 04:14pm
t1-v1.3

361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
(10) Every fair coin will come up heads if flipped. In our explanation of (9)s falsity, we stipulated that to fail to satisfy x will come up heads if flipped just is to satisfy x will not come up heads if flipped, and used that equivalence to arrive at the conclusion that each of the relevant coins must, in fact, satisfy x will come up heads if flipped. These conditions, though, are exactly ones in which (10) ought to be true; thus our analysis predicts the truth of (10), despite its obvious falsity. We have purchased our explanation of (9)s falsity only at the price of predicting (10)s truth. Out of the frying pan and into the fire. It is clear that (10) is false as long as (at least some of) the coins fail to satisfy the open embedded conditional, even though they also fail to satisfy the CEMequivalent conditional. (9), however, is only true if the coins satisfy this CEMequivalent conditional; it is not enough that they simply fail to satisfy the open embedded conditional. The proposed defense of the Simple Solution has led to the awkward position of requiring that our quantified conditionals both obey and fail to obey CEM. It appears that CEM is a necessary stipulation when we are providing a semantic analysis of conditionals under quantifiers such as no, but not if the quantifier is one such as every, CEM applies only if it applies to the unquantified version of the conditional. Thus if conditionals such as (7) and (8) do not obey CEM, we are forced to alter their semantics so as to conform to CEM when they occur under quantifiers like no, but not when they occur under quantifiers like
UN
CO RR
EC
Intuitively, it appears that (9) requires something like this for its truth: the coins in question must all fail to come up heads if flipped for (9) to be true. Thus it seems that the Simple Solution will be adequate only if we assume that an item fails to satisfy an open conditional if P(x) then Q(x) if and only if it satisfies the conditional If P(x) then not Q(x) i.e. if we do assume that all conditionals obey CEM. Higginbotham (2003) notes that this assumption is strange and stipulative; we have no explanation of why we would need to assume CEM for our analysis. On the account sketched here, we would in fact predict, on the face of it, that we would not need to assume CEM. If our difficulties were resolved by assuming that quantifiers demand that conditionals in their scope obey CEM, though, perhaps this would justify our adopting the stipulation. The situation, however, is not quite so straightforward. To provide an adequate analysis of (9), we were forced to assume that the coins in question failed to satisfy x will come up heads if x is flipped if and only if they satisfied x will not come up heads if flipped. Fair coins do not intuitively satisfy x will come up heads if flipped, but no fair coin will come up heads if flipped is clearly false. We explained this by assuming that, in order to fail to satisfy x will come up heads if flipped, an item must satisfy x will not come up heads if flipped. Fair coins clearly do not satisfy this latter conditional, so we concluded that, despite appearances, fair coins must satisfy x will come up heads if x is flipped after all. We were then able to explain the falsity of (9), which is true if and only if none of the coins satisfy x will come up heads if flipped. But this explanation of why (9) is false unfortunately predicts that (10) will be true:
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450
3.3 A Modalized Restrictive Account
5 A quantificational adverb is an adverb such as always, sometimes, often, never, and so on. Lewis (1975) argued that these adverbs quantify over cases or situations. Thus, for example, the sentence John always wins is to be analyzed to mean that all relevant situations involving John are ones in which he wins. 6
I.e. parts of possible worlds; see Kratzer (1989). In our discussion, nothing will hang on the use of situations rather than worlds. (An account that uses situations rather than worlds is useful in dealing with so-called donkey sentences, such as if a farmer owns a donkey, he beats it (Heim, 1990). We will not be concerned with such sentences here.)
UN
A popular account of quantified conditionals emerges from the tradition that began with David Lewis (1975), which takes if-statements to restrict quantifiers and quantificational adverbs.5 Lewis argued that if-statements that occur in the scope of quantificational adverbs restrict the domain of quantification of that adverb. For example, we would analyze always, if m and n are positive integers, the power m n can be computed by successive multiplication as involving quantification over pairs of positive integers. The sentence is analyzed to mean that, for all pairs of positive integers m and n, the power m n can be computed by successive multiplication. Thus the if-clause if m and n are positive integers provides the domain of quantification for the adverb always. Most contemporary theorists in this tradition assume that, if no explicit adverb of quantification is present in a conditional statement, then a covert universal quantifier over possible situations6 occurs in the sentences logical form. On this view, conditionals serve to restrict the domain of possible situations over which the quantifier ranges be it an explicit quantificational adverb or a covert universal quantifier. It is almost always assumed that, if an explicit adverb of quantification occurs in the sentence, then the conditional will restrict that adverb, and no covert universal will occur in the sentences analysis. On such an account, an if-statement of the form If R, then M (i.e., in which no explicit quantificational adverb occurs) would be analyzed as:
CO RR
EC
TE
DP
every. In this way we find ourselves back at square one; one semantic analysis applies to conditionals under every, and another to conditionals under no. This suggests, I think, that we have not properly understood the logical form of conditionals embedded under quantifiers. The Simple Solution so far fares considerably better than Higginbothams original account it provides adequate truth conditions in the vast majority of cases, and the violations are localized to those marginal and controversial conditionals that fail to obey CEM. Nonetheless, we have no explanation of why CEM is a necessary assumption for providing the semantics of conditionals embedded under no. We have even less of an explanation of why this assumption does not apply to conditionals embedded under every. The Simple Solution, though initially most appealing, is not ultimately adequate. Another approach is called for.
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3

451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495
No [C R] [M]
Or more formally:
Kratzers treatment of if-statements as restricting quantificational operators has been very influential. As it stands, though, it does not accurately capture the truth conditions of (1) and (2), since it is susceptible to the same counterexamples as Higginbothams (1986) account. Let us consider Bill once again doomed to failure regardless of how hard he works whose presence suffices to falsify (1). Should Bill decide not to work hard, though, then he poses no obstacle to the truth of (1*): he is not among the students who work hard, and so is irrelevant to (1*)s truth or falsity. Thus (1) will be false while (1*) may yet be true. Similarly, the inclusion of the fortunate Meadow who will succeed no matter what among the relevant students is enough to render (2) false. Should Meadow decide not to goof off, though, then (2*) may well still be true, since only those students that actually goof off are relevant to the truth of (2*). This analysis, then, does not fare any better than Higginbothams original (1986) account. It should be clear, though, exactly what the root of the difficulty is for this version of the restrictive account the analysis is ignoring possible circumstances that are relevant for the truth of the quantified conditional because they are merely possible, and not actual. This difficulty does not arise for the restrictive analysis when the quantificational element is an adverb of quantification or a covert universal, because we are taking those quantifiers to range over possible situations. The truth conditions
UN
CO RR
(1*LF) Every x [x is a student and x works hard] [x will succeed] (2*LF) No x [x is a student and x goofs off] [x will succeed]
EC
(1) (1*) (2) (2*)
Every student will succeed if they work hard. Every student who works hard will succeed. No student will succeed if they goof off. No student who goofs off will succeed.
TE
Thus never, if R, then M is interpreted to mean that no relevant situations in which R is true are situations in which M is true. Lewis confined his original discussion to adverbs of quantification, but it is a natural further step to treat if-statements as restricting quantificational NPs such as no students, if the if-statement occurs in the scope of such an NP (see, e.g. Kratzer 1991; von Fintel 1998). On this view, we would construe (1) and (2) as (1*) and (2*) below:
DP
RO O
where C denotes the set of contextually relevant situations, and R and M are the interpretations of the antecedent and consequent respectively. Thus if R, then M is true iff all of the contextually relevant situations in which R is true are ones in which M is true. If an explicit adverb of quantification occurs in the sentence, then that adverb will take the place of the covert universal quantifier. For example, Never, if R, then M would be analyzed as:
All [C R] [M]
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540
(11) If Beryl is in Paris, she often visits the Louvre. Geurts points out that (11) can be read as saying that on many of the occasions in which Beryl is in Paris, she visits the Louvre, or as saying that whenever Beryl is in Paris, she pays many visits to the Louvre. The first reading is obtained by taking the if clause to restrict the overt quantificational adverb often, while the second is obtained by taking the if clause to restrict a covert universal, of the sort that is standardly taken to occur in the absence of a quantificational adverb. Geurts account differs from some more conventional views in that he claims that a conditional may have a covert universal associated with it, even when the sentence contains an explicit quantifier or quantificational adverb. Thus Geurts does not take such explicit items to block the emergence of a covert universal. Geurts, though, only discusses this covert operator in contexts where it is the operator that the conditional is restricting. However, there is no reason that I know of that would prevent this covert universal quantifier over possible worlds from occurring in the logical form of a conditional statement, even though the conditional is itself restricting an explicit quantifying determiner. I propose that a conditional may contribute a covert universal quantifier to the semantics, even though the conditional itself serves to restrict an explicit quantifier. Further, I suggest that, when a conditional restricts an
UN
CO RR
of conditionals such as Bill will succeed if he works hard do not simply depend on the happenings of the actual world, because the covert universal is taken to range over possible situations. If we restricted the domain of the universal to actual situations, then we would predict inappropriate truth conditions if, as it happens, Bill never actually works hard. A modal element is needed to deliver the correct truth conditions for conditionals. This suggests that our objection to treating quantified conditionals as restricted quantifiers, then, would be defeated were we able to include such a modal element in their truth conditions. Meadow falsifies no student will succeed if they goof off even if she does not actually goof off, because were she to goof off, she would succeed nonetheless. This modal fact is enough to guarantee that Meadow falsifies the quantified conditional, regardless of how events in the actual world unfold. We need to take these possible events into account when giving the truth conditions of quantified conditionals, just as we must when we are giving the truth conditions of conditionals that contain quantificational adverbs. Indeed, it would be rather surprising were the two types of constructions not to require such parallel treatment. There are a variety of ways, it would seem, in which this idea might be implemented. We might take the quantifier to range over possible individuals, for example. Here, I will pursue a particular means of implementing the idea, which fits rather well with some recent work by Bart Geurts (m.s.), though there are other ways that one might implement the idea. Geurts (m.s.) argues that, even when a conditional statement contains an explicit quantificational NP or quantificational adverb, the conditional may still serve to restrict a covert universal, in the same way that it does when no explicit quantifier or quantificational adverb is present. Geurts asks us to consider the following sentence:
EC
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3

541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
7 There is a fair amount of contextual variability associated with the restricting nominal student here. I have been eliding the details of this restriction, other than including a parenthetical relevant in my representation of the logical form of these statements. There is far more that needs to be said here. In particular, it seems that some contextual restrictions allow the extension of the restricted nominal to change across the possible situations, while others do not. For example, if I say every student will succeed if they work hard with my introductory logic class in mind, there is a reading of the sentence on which it applies to any students who might possibly take my class. The utterance would then be a commentary on how I run my course. On this reading, the statement is false if the likes of Bill is even a possible member of my class. There is another reading of the sentence, though, on which it only applies to the students that have actually enrolled in my class, and thus understood is a commentary on the intellectual abilities of these students. On this reading, it does not matter whether Bill might have enrolled that he has not in fact enrolled is enough to discount him from the evaluation of the statement. We should, I think, understand this variability as part of the general phenomenon of contextual variability in nominals the property picked out by is a student might be such that its extension does not vary across the relevant possible situations, or it might be less rigid. (We could also locate difference between the readings in the set of relevant possible worlds we are considering. The proposal presented here is neutral between the two implementations, however, I am inclined to locate the restriction in the restricted nominal.) It should be noted, though, that it is less clear how these two readings would be generated, if we understood the statement to be quantifying over actual individual students, and attributing conditional properties to them, as we would under the Simple Solution. Unless we take the quantificational NP to range over possible individuals, it may be hard to avoid the consequence that the only available readings of the statement should be ones that pertain to the students that are, in fact, members of my class.
UN
CO RR
Cw, w0 picks out a contextually determined restriction on the possible worlds over which we are quantifying. These truth conditions correctly predict that every student will succeed if they work hard will be false if Bill is among the relevant students. Since there are relevant worlds in which Bill works hard but does not succeed, the statement is false. Similar remarks apply to the quantified conditional No student will succeed if they goof off, which is falsified by Meadows presence, regardless of how hard she actually works. Since there are relevant possible worlds in which Meadow goofs off and still gets an A, the quantified conditional cannot be true.7
EC
TE
w Cw, w0 : Every x [x is a (relevant) student in w & x works hard in w] [x will succeed in w]
DP
explicit quantifier, this covert universal takes wide scope over the entire statement. Thus, when a conditional restricts a quantifier, every relevant possible world must be such that the quantified statement holds in it, for the entire statement to be true. This account delivers the correct results for the quantified conditionals we have been considering. The unfortunate Bill doomed to failure regardless of how hard he may work posed a problem for a straightforward account of the conditional as restricting the quantifier. If we understood the conditional as restricting the quantified NP, with no modal element present, we would predict that the quantified conditional would be true, so long as Bill did not in fact work hard. The quantified conditional is not, however, intuitively true under those circumstances. The fact that, had Bill worked hard, he still would not have succeeded is enough to falsify the quantified conditional. I propose that we amend the above analysis, so as to include wide-scope quantification over contextually relevant possible worlds:
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
w Cw, w0 : No x [x is a fair coin in w & x is flipped in w] [x will come up heads in w]
w Cw, w0 : Every x [x is a fair coin in w & x is flipped in w] [x will come up heads in w] It is clear that this analysis correctly predicts that every coin will come up heads if flipped will be false. The Modalized Restrictive Account is able to capture the strong truth conditions of both the quantified conditionals. The Simple Solution issued in overly weak conditions for the conditional under no, unless CEM was assumed to apply. However, once CEM was assumed, the truth conditions for the conditional under every were predicted to be overly weak. Only a differential application of CEM captured the strong truth conditions of both statements. Our restrictive analysis allows us to avoid any such differential assumptions. This consideration constitutes good reason to prefer a restrictive analysis of conditionals to the Simple Solution. Furthermore, we will see in the next section that no version of the Simple Solution is applicable to quantified unless-statements, while a Modalized Restrictive
UN
CO RR
On this analysis, the statement is true iff in all relevant possible circumstances, none of the coins that are flipped will come up heads. These are the truth conditions we have been seeking, and we are able to arrive at them without making questionable assumptions about the plausibility of CEM in such a case. Similarly, we have at hand a straightforward, parallel analysis for every coin will come up heads if flipped:
EC
This Modalized Restrictive Account is thus able to deliver the correct truth conditions for (1) and (2). The Simple Solution, of course, was also able to handle these sentences correctly. However, our Modalized Restrictive Account, unlike the Simple Solution, delivers the intuitively correct results when faced with conditionals that do not obey Conditional Excluded Middle, without employing ad hoc assumptions. Higginbotham (2003) claimed that a compositional account of quantified conditionals is forthcoming only if we assume that conditionals under quantifiers obey CEM. He was rightly uncomfortable with this result, feeling it to be little more than stipulation. As we saw above, the troubles run deeper than unexplained stipulation; the stipulation only applies to conditionals embedded under quantifiers such as no. If a conditional occurs under every, it obeys CEM only if its unquantified counterpart obeys CEM. Thus, in our above example, every coin will come up heads if flipped is a false claim, even though there is no coin in the domain that satisfies x will not come up heads if flipped. If CEM held here, the universally quantified claim would only be predicted to be false if there were such a coin. Thus CEM has to be imposed differentially on conditionals, depending on the nature of the quantifier they are embedded under. It was just this sort of chameleon-like semantics, though, that we set out to avoid. Our Modalized Restrictive Account yields the right predictions without recourse to such uncomfortable assumptions and chameleon-like analyses. The Modalized Restrictive Account would render no fair coin will come up heads if flipped as:
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3

631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675
4 The Semantics of Conditionals Containing Unless
Q [N] [P if R]
(3 LF) handles Bills case adequately: Bill does not satisfy x will succeed unless x goofs off, since its false that Bill will succeed unless he goofs off. Thus if (3 LF) is the logical form of (3), we would predict that (3) would not be true if Bill is among the relevant students. But what of (4 LF)? We wish to predict that a sentence whose logical form is given by (4 LF) will not be true if Meadow is among the relevant students. (4 LF) is true if none of the relevant students satisfy the open unlessstatement, or alternatively if all of the relevant students fail to satisfy it. Meadow will present an obstacle to the truth of (4 LF) iff she satisfies x will succeed unless x works hard . . . and here we encounter a difficulty. Meadow will succeed unless she works hard is intuitively false. This is not a true sentence in the scenario we have described. Meadow will succeed no matter what she does, so its false that Meadow will succeed unless she works hard. Thus, if the logical form of (4) was given by (4 LF), Meadow would pose no obstacle to the truth of (4). She fails to satisfy the open unless-statement, and so it is quite possible that no student in a class containing her would satisfy it. The Simple Solution, then, does not even begin to accommodate unlessstatements. It appears that the truth conditions of quantified unless-statements do not depend on how many members of the domain satisfy the open unlessstatement. Statements of the form No Ns P unless t hey R cannot be understood to mean that No Ns satisfy P unless they R. In the case of unless-statements, we do not need to invoke conditionals that fail to obey CEM to raise difficulties for the Simple Solution. It cannot handle these rather basic examples of quantified unless-statements.
UN
CO RR
EC
(3) (3 LF) (4) (4 LF)
Every student will succeed unless he goofs off. Every x [x is a (relevant) student] [x will succeed unless x goofs off] No student will succeed unless he works hard. No x [x is a (relevant) student] [x will succeed unless x works hard]
TE
A parallel account for unless would render the logical forms of (3) and (4) as follows:
DP
The Simple Solution to the puzzle of quantified conditionals treated embedded if-statements as predicating a conditional property of the quantified NP subject. The truth value of a sentence of the form Q Ns P, if R would then depend on how many of the relevant Ns possess the conditional property. The logical form of such a sentence, we have claimed, might be given as follows:
RO O
Account delivers the desired results. Treating quantified if and unless-statements in a uniform manner constitutes further reason to prefer the restrictive account to the Simple Solution in the case of if-statements.
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720
Treating unless as meaning if . . . not is the most obvious way to fill out the claim that Meadow really does satisfy the relevant unless-statement: Its true that Meadow will succeed if she doesnt work hard. Higginbotham (2003) proposes that we handle unless in this manner, and claims that a compositional treatment of quantified unless-statements is possible so long as unless is assimilated to if . . . not. (Higginbotham provides few details, so it is not clear whether he proposes this to deal with situations such as Meadows, or for some other reason.) Besides a general desire not to simply dismiss as pragmatic any phenomenon that threatens semantic simplicity, there are other considerations that weigh against treating unless as if . . . not. Geis (1973) produces a battery of reasons not to equate unless with if . . .not, and I refer my reader to his excellent article for more detailed discussion than I can provide here. Geis notes that unless and if . . . not behave differently with respect to the possibility of coordinate structures. There is no obstacle to conjoining clauses containing if . . . not, but we cannot do the same with clauses containing unless. Compare, for example: John will succeed if he doesnt goof off and if he doesnt sleep through the final. *John will succeed unless he goofs off and unless he sleeps through the final. Unless and if . . . not also interact differently with negative polarity items. Naturally, negative polarity items can occur in the scope of if . . . not. They cannot, however, occur in the scope of unless: John wont succeed if he doesnt ever attend class. *John wont succeed unless he ever attends class. As a final point against the identification of if . . . not and unless, we should note that clauses containing if . . . not can be modified by only, even, except, while clauses containing unless cannot: John will succeed only if he doesnt goof off. *John will succeed only unless he goofs off. John will succeed even if he doesnt work hard. *John will succeed even unless he works hard. John will succeed except if he doesnt work hard. *John will succeed except unless he works hard. I will take these considerations and others in Geis (1973) to tell strongly against the identification of unless with if . . . not that Higginbotham (2003) suggests, and so this particular means of deriving the falsity of Meadow will succeed unless she works hard is untenable. Perhaps other means might be proposed, but I do not know of any other such proposals.
UN
CO RR
EC
TE
DP
One might, of course, just deny that Meadow will succeed unless she works hard really is false. One might argue that it is merely pragmatically unacceptable for one reason or another. I find such a solution deeply unsatisfying. To my ear, and the ears of my informants, it is simply false that Meadow will succeed unless she works hard. Any account that validates that intuition should be preferred to one that dismisses it. In what follows I will propose such an account, and to the extent that it is successful, it provides us with a far more satisfactory account than one that chalks up the appearance of falsity here to mere pragmatic factors.8 Let us then accept at face value the intuition that its false that Meadow will succeed unless she works hard. One way that we might frame our puzzle is as follows: for unquantified unless-statements, there appears to be a uniqueness requirement. This uniqueness requirement has it that, for the unless-statement to be true, it would have to be the case that working hard is the only relevant way in which Meadow will fail to succeed. Since this is false in the case described, the unless-statement is predicted to be false. This uniqueness requirement, however,
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3

721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765
4.1 Von Fintels Account of Unless
All [C R] [M]
It is thus part of the truth conditions of M unless R that all relevant situations in which R is false are ones in which M is true. It should be clear that this is extensionally equivalent to the analysis we would give for If not R, then M, and so the common intuition (see, e.g., Higginbotham 2003) that unless is akin to if . . . not is captured by this part of von Fintels treatment. Unless does not simply mean if . . . not (Geis, 1973, see also fn 9). Von Fintel recognizes this, and so includes a so-called uniqueness condition, which he formulates as follows:
UN
To make progress here, we will need to understand the semantics of unquantified unless-statements in more detail. There has been relatively little contemporary discussion of unless, but fortunately von Fintel (1992, 1994) offers an excellent discussion that will be extremely helpful to us here. Von Fintels account extends and formalizes Geis (1973), and includes a uniqueness condition that explains why Meadow will get an A unless she works hard is false. Von Fintels account of unless-statements follows in the Lewis-Kratzer tradition of treating conditionals as restrictions on quantificational adverbs, and he assumes, along with most theorists, that a covert universal quantifier occurs in the absence of an explicit quantificational adverb. Let us begin by considering cases in which no adverb of quantification is present in the sentence, and so the quantifier in question is a covert universal. Von Fintels account of unless-statements has two parts. The first part treats unlessstatements as having as part of their meaning something akin to if . . . not. Thus M unless R has its interpretation given in part by:
CO RR
EC
TE
DP
seems to disappear when unless-statements are embedded under some quantifiers such as no students. A will succeed unless A works hard is false if A can succeed while working hard. However, the mere fact that working hard and succeeding are compatible for each student does not suffice to make true No student will succeed unless they work hard. The quantified statement is much stronger. It is not made true by the mere compatibility of working hard and succeeding. For Meadow, hard work and success are certainly compatible, but this does not mean that No student will succeed unless they work hard can be true of a class containing her. It cannot. No student will succeed unless they work hard specifically rules out the possibility of students like Meadow, who may succeed without hard work. I will argue that a Modalized Restrictive Account of quantified unless-statements will deliver the results we are seeking. Once we have a satisfactory account of unless-statements that occur in the presence of adverbs of quantification, it will be a simple matter to extend this account to handle unless-statements that are embedded under quantifiers.
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810
S (All [C S] [M] R S)
All [C R] [M] & S (All [C S] [M] R S) It should be obvious by now that we will not be able to use this analysis to give an account of quantified unless-statements in any straightforward manner. If we try to treat No students are M unless they are R as No x [x is a student] [x is M unless x is R] and use von Fintels analysis of the unless-statement, we will obtain: No x [x is a student] [All [C {s: x is R in s}] [{s: x is M in s}] & S (All [C S] [{s: x is M in s}] {s: x is R in s} S)] (where C is the set of relevant situations.) But as long as all the students fail to satisfy at least one of the conjuncts of the analysis, the statement will be true. As before, this predicts that Meadow will pose no obstacle to the truth of no student will get an A unless they work hard, since she will not satisfy the uniqueness condition of the unless-statement. Once again, the uniqueness condition essential for an account of unless-statements that do not occur under quantifiers creates difficulties once we try to embed the statement under no.
UN
CO RR
Thus, for any set of situations S, if all relevant situations that are not S situations are also M situations, then S includes R as a subset. It is this condition that explains the falsity of, e.g., Meadow will get an A unless she goofs off. It is certainly true that all relevant situations in which Meadow does not goof off are situations in which she will get an A, thus the first condition of the analysis is satisfied. But the uniqueness condition will not be satisfied. Consider a proper subset of the (possible) situations in which she goofs off say, situations in which she both goofs off and chews gum in class. Clearly, the set of situations in which she goofs off is not a subset of the situations in which she both goofs off and chews gum in class, at least on the very natural assumption that there are some relevant, possible situations in which she goofs off but does not chew gum. However, since Meadow will get an A in all relevant situations, she will a fortiori get an A in situations in which she either doesnt goof off, or doesnt chew gum in class. But this disjunctive set of situations just consists of the situations denoted by [C S], where S is the set of situations in which she both goofs off and chews gum. Thus, we have found a set S of situations such that all the relevant non-S situations are situations in which Meadow gets an A, but the set of situations in which Meadow goofs off is not a subset of this set S. Thus the uniqueness condition is not satisfied. The uniqueness condition will only be satisfied if all the situations in which R holds are situations in which M does not hold. If there are any R-situations that are also M-situations, then if we subtract these situations from R, we will obtain a set S that falsifies the uniqueness condition. Von Fintels account of statements of the form M unless R thus contains two conjuncts:
EC
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3

811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
4.2 A Modified Account of Unless
Before we return to our main project of giving an account of unless-statements under quantifiers, let us consider how von Fintels account fares when there is an explicit adverb of quantification present in the unless-statement. Von Fintel (1992) formulates his account in general terms as follows: Q [C R] [M] & S (Q [C S] [M] R S) where Q is the relevant quantifier either a covert universal as before, or an adverb of quantification that explicitly occurs in the sentence. Let us see how his account handles a statement such as (12): (12) John never succeeds unless he works hard (Or to make the scope of the adverb more apparent, we may substitute the more awkward Never, unless he works hard, does John succeed.) Clearly, (12) cannot be true if there are any possible, contextually relevant situations in which John succeeds without working hard. Von Fintel (1994) claims that (12) also requires for its truth that any time John works hard, he succeeds, but this seems to me too strict a requirement for the truth of (12). (12) may be true, yet there be some relevant situations in which even hard work does not suffice for Johns success. My intuitions, and those of my informants, have it that it should not be part of the truth conditions of (12) that every situation in which John works hard is one in which he succeeds. If John is someone who finds his coursework extremely difficult, and so never succeeds without hard work, (12) will be true, even if John sometimes finds the work so difficult, that he fails despite working hard. Von Fintels account, however, predicts that the truth conditions of (12) would include such a strict requirement. His above analysis, applied to (12), would be as follows: No [C {s: John works hard in s}] [{s: John succeeds in s}] & S (No [C S] [{s: John succeeds in s}] {s: John works hard in s} S)
The first conjunct above is perfectly correct it states that no relevant situation in which John does not work hard is a situation in which John succeeds. The second conjunct the uniqueness condition imposes an overly demanding condition, however. The second conjunct is not satisfied as long as there is some set of situations S such that none of the relevant non-S situations are situations in which John succeeds, and yet S does not contain the situations in which John works hard. Suppose, for example, that amongst the contextually relevant situations are ones in which the subject matter is just too difficult for John to master. No matter how hard he works, he wont succeed in those situations. Intuitively, (12) can be true despite the possibility of such situations, but the uniqueness clause in von Fintels account is violated under these circumstances. To see that this is so, let us take S to be the set of situations in which the subject matter is not too difficult for John. Let us further suppose that there are some relevant possible situations in which John works hard, even though the subject matter,
UN
CO RR
EC
TE
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900
{s: John works hard in s} S
However, since S is the set of situations in which the subject matter is not too difficult for John, [C S] is the set of relevant situations in which the subject matter is too difficult for John. In the scenario we are describing, none of these situations are situations in which John succeeds. Thus it is true that: Thus von Fintels uniqueness clause is violated, and so (12) is predicted to be false, so long as there are some situations in which John works hard but still doesnt succeed. Intuitively, however, it may be true that John never succeeds unless he works hard, even though sometimes his hard work isnt enough to secure his success. Sometimes the subject matter is simply beyond him. Thus von Fintels account does not correctly handle unless-statements that contain the quantificational adverb never. If von Fintels account included only its first part the requirement that no situations in which John does not work hard be ones in which he succeeds we would have the intuitively correct truth conditions for John never succeeds unless he works hard. As we saw above, however, the uniqueness clause is needed to provide adequate truth conditions for unless-statements that contain universals, be they covert or overt. How can we accommodate this data in a compositional manner? I believe that the problem lies in the formulation of von Fintels uniqueness clause. Further evidence that it is not properly formulated emerges when we consider unless-statements that contain adverbs of quantification such as usually or rarely, as in (13) and (14). Von Fintel (1994) claims that statements such as (13) and (14) are ill-formed and semantically deviant. I must admit that I simply do not share this intuition, nor do my informants. Since (13) and (14) are perfectly fine to my ear, I will aim to provide an account of unless that captures their truth conditions adequately.
(Or to make it absolutely clear that the quantificational adverbs have scope over the whole statements: (13) Usually, unless John goofs off, he succeeds. (14) Rarely, unless John works hard, does he succeed. I cannot find anything objectionable about these sentences.) Von Fintels account cannot be successfully applied to (13) and (14); this is natural since von Fintel does not intend that it should apply to them. The account applied to (13) would yield the following:
UN
(13) John usually succeeds unless he goofs off. (14) John rarely succeeds unless he works hard.
CO RR
EC
TE
DP
No [C S] [{s: John succeeds in s}]
RO O
regrettably, is just too difficult him. (This supposition is just the one described the preceding paragraph.) Then the set of situations in which John works hard will not be a subset of S, and so for this S it is false that:
May 3, 2008
Time: 04:14pm
t1-v1.3

901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945
Most [C S] [{s: John succeeds in s}]
Once again, the uniqueness clause is not satisfied, and so we would predict that John usually succeeds unless he goofs off would be false as described. It is enough to falsify von Fintels analysis that John very occasionally bribes his teacher and, having done so, usually succeeds as a result. Intuitively, however, Usually, John succeeds unless he goofs off is not so strong a claim as to be incompatible with such circumstances. In the above example, it was important that we stipulated that John only occasionally bribes his teacher to succeed. If this was common practice for him, then (13) would not be true. It does not seem correct to say, for example, that Meadow usually succeeds unless she goofs off. Thus in the case of most or usually, some second conjunct is needed, for it is not enough for the truth of the claim that most of the situations in which Meadow does not goof off be ones in which she succeeds. The first conjunct of von Fintels analysis alone would not suffice here, though it seemed that it would suffice when the adverb of quantification was never. We can mount a similar argument against the appropriateness of the uniqueness clause in the case of (14), which is an unless-statement that contains the adverb rarely. Suppose, for example, that it only occasionally happens that John succeeds without working hard for the most part, he only succeeds when he works hard. Then (14) is intuitively true. If we suppose further that, sometimes, the subject
9
I am assuming here that usually can be understood as most, and so am setting aside any additional normative or otherwise modal import usually may possess; nothing will hang on this simplifying assumption.
UN
CO RR
EC
TE
As before, I cannot find fault with the first conjunct of the account it is certainly necessary for the truth of (13) that most situations in which John does not goof off are ones in which he succeeds. It is the uniqueness condition that gives us cause for concern. Suppose, for example, that for the most part, when John doesnt goof off, he succeeds, and again for the most part, when John does goof off, he doesnt succeed. Let us say, though, that once in a while John bribes his teacher, in which case he usually succeeds, no matter what he does. It seems that (13) is true under these circumstances, as long as John very rarely bribes his teacher, but the uniqueness clause is violated. We may take our S to be the set of situations in which John does not bribe his teacher. John almost never bribes his teacher if he is planning to work hard what would be the point? so the situations in which he does bribe his teacher are generally ones in which he goofs off. Thus the set of situations in which John goofs off are not a subset of S, i.e. of the situations in which he refrains from bribing his teacher. However, as we have described the example, most of the situations in which John does decide to bribe his teacher are ones in which John succeeds, so it is true that:
DP
RO O
Most [C {s: John goofs off in s}] [{s: John succeeds in s}] & S (Most [C S] [{s: John succeeds in s}] {s: John goofs off in s} S)9
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990
Q [C R] [M]
Q [C R] [M] & Q [M C] [C R]
All [C {s: John goofs off in s}][{s: John succeeds in s}] & All [{s: John succeeds in s} C] [C {s: John goofs off in s}] The sentence is true just in case all relevant situations in which John doesnt goof off are ones in which he succeeds, and all relevant situations in which John succeeds are ones in which he doesnt goof off. These truth conditions are equivalent to the ones that von Fintel provides for unless-statements that contain universal quantifiers.10
10 The formulation of von Fintels uniqueness clause needs to be amended in order for these to be strictly equivalent, but it is a minor adjustment, and is independently motivated. As it stands, von Fintel has the following as his uniqueness clause:
However, the clause, as it stands, is violated if there are irrelevant R-situations (i.e. situations that are in R, but not in C). That is, statements such as John will succeed unless he doesnt work hard would be predicted to be false if there are possible situations outside of the contextually relevant ones in which John doesnt work hard situations in which, e.g., John dies in a freak accident.
S (Q [C S] [M] R S)
UN
CO RR
Q M unless R is true, then, if and only if Q of the relevant non-R situations are M situations, and Q of the relevant M situations are non-R situations. If there is no explicit adverb of quantification in the sentence, then I assume that a covert universal is present. Thus John will succeed unless he goofs off is analyzed as:
EC
It would be surprising if this part of the analysis was not correct, given the strong intuitions that unless is semantically similar to if . . . not. The uniqueness clause so far has proved tricky, however. We would like it to be equivalent to von Fintels uniqueness clause when the quantifier is a universal, but we would like it to effectively evaporate when the quantificational adverb is never. We would like to have some version of a uniqueness clause when the adverb is usually, but we would like it to amount to the requirement that most of the situations in which M holds are ones in which R doesnt hold, so as to allow the truth of John usually succeeds unless he goofs off if John very occasionally bribes the teacher, but not if he does so as a matter of course. I propose that we analyze statements of the form Q M unless R as follows:
TE
DP
RO O
matter is simply too hard for John, and despite his best efforts, he does not succeed, then (14) remains intuitively true. Once again, however, von Fintels semantics predicts that the statement will be false. I shall not go through the details again, but my reader may convince herself that this is so by taking the set S to be the set of situations in which John works hard, and the subject is not too difficult for him. Let us then summarize our desiderata for an account of unless-statements of the form Q M unless R. The first conjunct of von Fintels analysis was absolutely correct in all cases:
May 3, 2008
Time: 04:14pm
t1-v1.3

991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
No [C {s: John goofs off in s}] [{s: John gets an A in s}]
Strictly speaking, this analysis predicts that John never gets an A if he goofs off is true if there are simply no relevant possible situations in which John goofs off. Intuitively, though, the English conditional suggests that it is a live possibility that John will goof off, and, of course, that in such possible situations, John will fail to get an A. I do not think that the situation is much different in the case of John never gets an A unless he works hard, which we would represent as:
To see that the uniqueness clause is violated, take S to be the contextually relevant situations in which John doesnt work hard (i.e. S = CR). This difficulty is easily remedied by rendering the uniqueness clause as: The adjustment is minor, and surely reflects von Fintels original intentions. Once we have made this adjustment, the two clauses are provably equivalent when the quantifier in question is a universal. S (Q [C S] [M] CR S)
UN
No [ C {s: John works hard in s}] [{s: John gets an A in s}] & No [C {s: John gets an A in s}] [C {s: John works hard in s}]
CO RR
If Q is symmetric, then the two conjuncts are equivalent. We are thereby able to capture the intuition that there is no real uniqueness clause when the quantifier is no it is enough for the truth of the unless-statement that no relevant non-R situations be M situations. The uniqueness clause does not in fact evaporate in a noncompositional manner, but simply becomes redundant if the quantifier in question is no. There is, I think, a suggestion of sorts to the effect that there are some situations in which M and R both hold, and the uniqueness clause provides an explanation of this suggestion, so it is not completely vacuous. For example, John never gets an A unless he works hard suggests that there are some possible situations in which John gets an A by working hard. This would seem to be related to the implication or presupposition carried by if-statements that contain never, such as John never gets an A if he goofs off. We would analyze the if-statement as
Q [C R] [M C] & Q [M C] [C R]
EC
TE
DP
Our first desideratum, then, is satisfied: If the quantifier is a universal, our account is equivalent to von Fintels. What, though, of the quantifier no, or never? We wished that the uniqueness clause would evaporate in such cases, and this is indeed what we obtain. The determiner no is a symmetric determiner (Barwise & Cooper, 1981); no As are Bs is true iff no Bs are As is also true. We may swap the material in the restrictor with that in the scope, without changing the truth value of the claim, if the quantifier is no or its equivalent. But the uniqueness clause we are considering amounts to just this exchange! We might just as easily have written M C in place of M in the first conjunct, and so formulated our analysis as follows:
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
Most [C {s: John goofs off in s}] [{s: John succeeds in s}] & Most [{s: John succeeds in s} C] [C {s: John goofs off in s}]
4.3 Aside: Uniqueness Clauses and Coordinate Structures

One might worry that, in allowing the uniqueness clause to evaporate in unlessstatements containing never, we lose an explanation of why unless clauses cannot be conjoined. Geis (1973) points out that coordinate structures with unless are not permissible, for example: (15) *John will get an A unless he goofs off and unless he sleeps through the final. Von Fintel (1991, 1994) proposes that his uniqueness clause explains the impermissibility of this statement since the unless clause expresses the unique minimal restriction that makes the conditional true, there cannot be another such clause. If both restrictions made the conditional true, then the uniqueness clause would not be satisfied in either case. Von Fintels account of unless-statements containing never also features a uniqueness clause, and so he claims that it also predicts the unacceptability of conditionals such as
UN
CO RR
The unless-statement will be false if most of the relevant situations in which John succeeds are just ones in which he bribes the teacher, then kicks back, since the second conjunct will not be true under those conditions. As long as we are only considering the occasional bribe, however, the unless-statement will be true. We thus arrive at a plausible and appealing account of unless by adopting this formulation of the uniqueness clause. Our account now yields the right results even when the sentence in question contains a quantificational adverb that is not a universal.
EC
TE
The second conjunct here is a perfect parallel to the analysis of the if-statement above, and it carries with it a similar suggestion (implication or presupposition, depending on the details of ones account) that there are some live possibilities in which John gets an A. Those possibilities cannot be ones in which John doesnt work hard, so we derive the suggestion that its possible for John to work hard and get an A. We do not need to make any assumptions specific to unless-statements here however we account for the parallel suggestion with if-statements should carry over here. (Von Fintels account of if-statements is an example of an account that treats this suggestion as a presupposition.) The account set out so far also provides an appealing analysis of unlessstatements that contain usually, as in usually M, unless R. We wanted our account to require that most of the situations in which M holds be situations in which R holds. For example, we wished to explain why John usually succeeds unless he goofs off was compatible with Johns occasionally slipping the teacher a bribe, but not with his doing so as a matter of course. We are now able to do so; John usually succeeds unless he goofs off will be analyzed as:
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3

1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125
(16) *John never gets an A unless he works hard and unless he bribes the teacher. I have argued that his uniqueness clause issues in overly strong truth conditions for unless-statements containing never, and so is not a desirable component of an account of such statements. The uniqueness clause also did violence to unlessstatements containing usually and rarely, which, pace von Fintel, are quite acceptable. But do we achieve our intuitively correct truth conditions at the cost of losing our explanation of the impermissibility of coordinate structures like the ones above? On closer inspection, it is far from clear that von Fintels uniqueness clause does in fact explain the unacceptability of the coordinate structures. His semantics predict only that the two sets of situations denoted by the two unless-clauses are coextensive. If they are not coextensive, the statement is false, not defective. And if they are coextensive, then his semantics predicts that the statement will be true! Consider, however, statements such as:
We should also be hesitant to offer a straightforwardly semantic explanation for the impermissibility of these constructions, since conjoined unless clauses are far more acceptable when they occur at the left periphery of the sentence. Consider, for example: (19) Unless he goofs off, and unless he sleeps through the final, John will succeed. Rearranging the sentence in this way significantly increases its acceptability, but von Fintels account, or any obvious variation on it, would predict that such rearrangement would not impact the sentences acceptability. Unless we assume that moving the clauses to the left periphery alters the truth conditions of the statement, a truth conditional explanation of the permissibility of coordination will not be forthcoming.11 We should also not lose sight of the lingering phenomenon that von Fintel (1991, 1994) points to namely that unless clauses can be disjoined, as in: (20) I wont go to the party unless Bill comes, or unless there is free beer.
11
I am indebted to John Hawthorne for bringing this phenomenon to my attention.
UN
CO RR
(18) *John will get an A unless he goofs off and unless he sleeps through the final are impermissible, as opposed to simply entailing that John will goof off if and only if he sleeps through the final.
EC
This statement is just as unacceptable as the two above, but it is far from clear why this should be so on von Fintels account. Since the two unless-clauses denote states of affairs that are coextensive in the possible worlds that are likely to be relevant, there is no obstacle to the uniqueness clause being satisfied for both unless clauses. Relatedly, it is not clear why statements such as:
TE
(17) *I will respect the list of endangered species unless it contains renates and unless it contains cordates.
DP
RO O
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170
4.4 Unless-Statements Embedded under Quantifiers
w Cw, w0 : Qx [Nx Rx] [Mx] & Qx [Nx & Mx] [Nx Rx] Or more perspicuously:
We will thus treat, e.g., no student will succeed unless they work hard as: w Cw, w0 : No x [x is a (relevant) student in w & x does not work hard in w] [x succeeds in w] & No x [x is a (relevant) student in w & x succeeds in w] [x is a (relevant) student in w & x does not work hard in w] We correctly predict that Meadows presence is enough to falsify the claim, no matter how hard she in fact decides to work. Since there is a relevant possible world in which Meadow succeeds without working hard, the statement is false. Let us recall the intuition we had earlier: that unless-statements embedded under no seem to be in some sense equivalent to if . . .not-statements. We had no sense that there was a uniqueness clause making a contribution to the truth conditions of the statement. On this analysis we can understand why this is so. Just as unlessstatements that contain the quantificational adverb never seem ed to be equivalent to if . . . not statements, the same is true for ones embedded under no, since in both cases the quantifiers are symmetric, and so the uniqueness clause does not add any additional demands to the truth conditions. We are also able to predict at last that both the over-protected Meadow and the unfortunate Bill suffice to falsify every student will get an A unless they goof off.
UN
CO RR
w Cw, w0 : Qx [Nx & not Rx] [Mx] & Qx [Nx & Mx] [Nx & not Rx]
EC
We have formulated a promising account of unless-statements that occur with quantificational adverbs. The unless-statements serve to restrict the range of possibilities that fall under the domain of the adverbial quantifiers. It is not difficult, then, to extend this account so that unless-statements embedded under quantifying determiners serve to restrict those quantifiers. As in the case of if-statements, a modal element must be introduced into the semantics, and I will continue to do so by means of a wide-scope covert universal quantifier over possible worlds. I propose that we analyze quantified unless-statements of the form Q Ns M, unless they R by letting the antecedent R restrict the quantifier in the same way that it restricts an adverb of quantification in the account provided above. The logical form of such a statement would then be:
TE
DP
RO O
These data together suggest that the behavior of unless clauses in coordinate structures needs considerably more investigation before it will be understood. Positing a uniqueness clause does not provide us with the explanation we seek. While more work is certainly in order, the phenomenon of coordination does not provide us with a reason to prefer von Fintels account to mine.
May 3, 2008
Time: 04:14pm
t1-v1.3

1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215
w Cw, w0 : Most x [x is a (relevant) student in w & x does not goof off in w] [x succeeds in w] & Most x [x is (relevant) student in w & x succeeds in w] [x is a student in w & x does not goof off in w] Similarly, few students will succeed unless they work hard is to be analyzed as: w Cw, w0 : Few x [x is a (relevant) student in w & x does not work hard in w] [x succeeds in w] & Few x [x is a (relevant) student in w & x succeeds in w] [x is a student in w & x does not work hard in w] We have thus managed to provide an account of quantified unless-statements that adequately captures their truth conditions, without attributing a chameleonlike semantics to unless. Quantified statements containing unless could not be understood as attributing conditional properties to a particular number or proportion of items in a domain restricted by the relevant nominal, as the Simple Solution would have it. They can, however, be analyzed by way of taking the unlessstatement to restrict the quantifier, albeit in a somewhat complex manner. We have seen, though, that an adequate account of unquantified unless-statements extends naturally to accommodate quantified unless-statements.
5 Conclusion
A uniform semantic analysis of if and unless embedded under quantifiers is possible. These constructions thus do not pose a threat to the thesis that natural language is semantically compositional. The semantics of these statements is not, however, a straightforward matter. The Simple Solution to Higginbothams puzzle according to which their truth and falsity depend on the number or proportion of the relevant items that satisfy the open conditional ran into difficulty when we considered if-statements that do not obey Conditional Excluded Middle, and was wholly unable to deal with unless-statements. Ultimately, we found that both types of conditional ought to be treated as restricting their quantified NP subjects, while also contributing a covert modal element to the semantics. Given the complexity of
UN
CO RR
EC
TE
DP
No matter how hard Bill and Meadow actually work, the statement is false if either is among the relevant students. The statement is false in Bills case because there is a relevant possible world in which he does not goof off and yet does not succeed, and so the first conjunct of the analysis is false in that world. It is false in Meadows case because of the relevant possibility of her succeeding without working hard, and so falsifying the second conjunct in that world. Unless-statements containing most and few can be given a parallel analysis. Most students will succeed unless they goof off will be analyzed as:
RO O
w Cw, w0 : Every x [x is a (relevant) student in w & x does not goof off in w] [x succeeds in w] & Every x [x is (relevant) student in w & x succeeds in w] [x is a (relevant) student in w & x does not goof off in w]
May 3, 2008
Time: 04:14pm
t1-v1.3
S.-J. Leslie
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
References
Barwise, J., & Cooper, R. (1981). Generalized quantifiers and natural language. Linguistics and Philosophy, 4, 159219. Geis, M. (1973). If and Unless. In B. Kachru, R. Lees, Y.Malkiel, A. Petrangeli, & S. Saporta (Eds.), Issues in linguistics: Papers in Honor of Henry and Renee Kahane (pp. 231253). Urbana: University of Illinois Press. Geurts, B. (m.s.). On an ambiguity in quantified conditionals. University of Nijmegen. Heim, I. (1990). E-type pronouns and donkey anaphora. Linguistics and Philosophy, 13, 137177. Higginbotham, J. (1986). Linguistic theory and Davidsons program in semantics. In E. Lepore (Ed.), Truth and interpretation: Perspectives on the philosophy of Donald Davidson (pp. 2948). Oxford: Basil Blackwell. Higginbothan, J. (2003). Conditionals and compositionality. In J. Hawthorne (Ed.), Philosophical perspectives (Vol. 17, pp. 181195). Janssen, Theo M. V. (1997). Compositionality. In J. van Benthem & A. ter Meulen(Eds.), Handbook of logic and language (pp. 417475). Amsterdam: Elsevier. Kratzer, A. (1989). An investigation of the lumps of thought. Linguistics and Philosophy, 12, 607653. Kratzer, A. (1991). Conditionals. In A. von Stechow & D. Wunderlich (Eds.), Semantics: An international handbook of contemporary research (pp. 651657). New York: de Gruyter. Leslie, S.-J. (2003a). Compositional conditionals. Paper presented to the Australian National University Philosophical Society. Leslie, S.-J. (2003b). Compositional conditionals. Paper for the Conference on Recent Research in Semantics, Carleton University. Lewis, D. (1975). Adverbs of quantification. In E. Keenan (Ed.), Formal semantics of natural language. Cambridge: Cambridge University Press. Pelletier, F. J. (1994a). On an argument against semantic compositionality. In D. Prawiz & D. Westerstahl (Eds.), Logic and philosophy of science in uppsala (pp. 599610). Kluwer Dordrecht. Pelletier, F. J. (1994b). The principle of semantic compositionality. Topoi, 13,1124. Von Fintel, K. (1992). Exceptive conditionals: The meaning of Unless. Proceedings of the North East Linguistics Society, 22, 135151. Von Fintel, K. (1994). Restrictions on Quantifier Domains. Ph.D. Thesis, University of Massachusetts, Amherst. Von Fintel, K. (1998). Quantifiers and if-clauses. The Philosophical Quarterly, 48(191), 209214. Von Fintel, K. & Iatridou, S. (2002).If and when if-clauses can restrict quantifiers. Paper for the Workshop in Philosophy and Linguistics, University of Michigan. Szab o, Z. G., Compositionality. In Edward N. Zalta (Ed.), The stanford encyclopedia of philosophy (spring 2005 edition), http://plato.stanford.edu/archives/spr2005/entries/compositionality.
UN
CO RR
EC
TE
DP
Acknowledgment Many thanks to John Hawthorne, Richard Larson and Paul Pietroski, Christopher Viger, and an anonymous reviewer for extremely helpful discussion and comments on an earlier draft of this paper. A much earlier version of the paper was presented in 2003 to the Australian National University Philosophical Society, and at the Carleton University Conference on Recent Research in Semantics, under the title of Compositional Conditionals. I am indebted to both audiences for their questions and comments.
RO O
the analysis required to give adequate truth conditions for these constructions, it is hardly surprising that theorists have doubted that a successful, uniform account of them would be possible. I hope to have shown that, despite these doubts, we can indeed provide such an analysis.
This article was downloaded by: [Princeton University] On: 10 November 2012, At: 14:28 Publisher: Psychology Press Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Language and Cognitive Processes

Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/plcp20
Inferences about members of kinds: The generics hypothesis

Sangeet Khemlani , Sarah-Jane Leslie & Sam Glucksberg
a a b a
Department of Psychology, Princeton University, Princeton, NJ, USA

b
Department of Philosophy, Princeton University, Princeton, NJ, USA Version of record first published: 25 Oct 2011. To cite this article: Sangeet Khemlani, Sarah-Jane Leslie & Sam Glucksberg (2012): Inferences about members of kinds: The generics hypothesis, Language and Cognitive Processes, 27:6, 887-900 To link to this article: http://dx.doi.org/10.1080/01690965.2011.601900
PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-andconditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.
LANGUAGE AND COGNITIVE PROCESSES 2012, 27 (6), 887 900
Inferences about members of kinds: The generics hypothesis

Sangeet Khemlani1, Sarah-Jane Leslie2, and Sam Glucksberg1
1 2
Department of Psychology, Princeton University, Princeton, NJ, USA Department of Philosophy, Princeton University, Princeton, NJ, USA
Downloaded by [Princeton University] at 14:28 10 November 2012
People routinely make inferences based on kind membership. For example, if you were told that a particular kind of animal is a tiger, then you would likely infer that it has stripes. Under what conditions are people willing to infer that a member of a given kind has a property? Two hypotheses were examined. The base rate or prevalence hypothesis holds that people rely only on their knowledge of the statistical frequency of a property among its kind to infer whether a member has that property. An alternative is the generics hypothesis, which states that people are influenced by their belief that the relevant generic generalization is true. In other words, if people agree to the generalization, ducks lay eggs, then they should be willing to make the inference that an arbitrary individual duck lays eggs, despite their knowledge that the majority of ducks do not lay eggs (i.e., juveniles, males, and infertile females). We present data that support the second hypothesis. Rather than being driven solely by beliefs about prevalence, agreement to the relevant generic predicted performance on an inference task beyond estimated prevalence or cue validity. These findings suggest that models of categorization that are based solely on statistical or simple probabilistic principles are incomplete. They also provide support for the idea that generics articulate core conceptual beliefs that guide our interactions with the world.
Keywords: Generics; Generics hypothesis; Default inferences; Non-monotonic reasoning; Concepts.
Consider an arbitrary dog. How confident are you that the dog has four legs? The inference is compelling, and it is a type of inference known as a default inference because, absent any information to the contrary, it leads one to conclude that an arbitrary dog has four legs. Now consider an arbitrary Canadian. How confident are you that he or she is right-handed? We might be less inclined to draw the default conclusion that the Canadian is right-handed than that the dog has four legs. What might account for the difference between the two inferences? It may be that a very high percentage of dogs have four legs, while a somewhat lower percentage of
Correspondence should be addressed to Sangeet Khemlani, Department of Psychology, Princeton University, 3-C-6 Green Hall, Princeton, NJ 08540, USA. E-mail: skhemlani@gmail.com This work was supported in part by a National Science Foundation Graduate Research Fellowship awarded to Sangeet Khemlani. We thank Jeremy Boyd, Adele Goldberg, Phil Johnson-Laird, Mark Johnston, Olivia Kang, Niklas Kunze, Max Lotstein, Adam Moore, and Lance Rips for their many helpful discussions and suggestions. # 2012 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business http://www.psypress.com/lcp http://dx.doi.org/10.1080/01690965.2011.601900
888
KHEMLANI ET AL.
Canadians are right-handed. Alternatively, if our conceptual knowledge concerning kinds contains structure richer than bare information about base rates or prevalence, then that structure might be reflected in our inferences about members of kinds. This paper presents an experiment designed to test whether such conceptual structure can be detected in our default reasoning. If so, then the view that categories and categorical induction can be adequately characterised in terms of the encoding and representation of statistical relationships, including prototype and exemplar theories, must, at best, be incomplete (e.g., Maddox & Ashby, 1993; Rosch, 1978; for a recent and comprehensive review see Murphy, 2002). Similar concerns would also apply to much of the recent work on Bayesian approaches to categorization however, sophisticated Bayesian approaches would likely be able to model this conceptual structure (Griffiths, Kemp, & Tenenbaum, 2008; Tenenbaum & Griffiths, 2001). How might generics affect categorization and default inferences beyond consideration of statistical relationships such as prevalence? People often use generic statements, such as tigers are striped, barns are red, ducks lay eggs, or ticks carry Lyme disease to express generalizations about kinds (Brandone, Cimpian, Leslie, & Gelman, in press; Carlson & Pelletier, 1995; Gelman, 2003; Krifka et al., 1995). Unlike quantified statements*for example, all tigers are striped, most barns are red*generics do not communicate information about how many members of the kind have the property in question. For example, one cannot answer the question, How many tigers are striped? by replying that tigers are striped (Brandone et al., in press; Carlson, 1977; Leslie, 2007). Correspondingly, there is no direct relation between the prevalence of a property among members of a kind and the acceptability of the relevant generic. Indeed, ducks lay eggs is accepted even though only mature fertile females lay eggs, but ducks are female is rejected (Brandone et al., in press; Leslie, Khemlani, & Glucksberg, 2011). Further, ticks carry Lyme disease is accepted even though only 2% of ticks carry the disease, but books are paperbacks is rejected, despite the fact that over 80% of books are paperbacks (Khemlani, Leslie, Glucksberg & Rubio-Fernandez, 2007; Leslie, 2007, 2008). Leslie (2007, 2008) hypothesised that these acceptance patterns arise because generics are sensitive to four main factors: 1. If a generic predicates a characteristic property of a kind, it is accepted, regardless of prevalence (e.g., ducks lay eggs, and lions have manes). Examples of a kinds characteristic properties include (for animal kinds) its method of locomotion, diet, salient physical characteristics, method of reproduction, method of nurturing its young; function (for artifact kinds); and societal role (for professional kinds). 2. If a generic predicates a striking or dangerous property of a kind, it is accepted if just some members of the kind have the property (e.g., ticks carry Lyme disease, sharks attack bathers; see Cimpian, Brandone, & Gelman, 2010). 3. If the property is neither dangerous nor characteristic, then the majority of the kind must have the property for the generic to be accepted (e.g., barns are red, cars have radios; see also Prasada & Dillingham, 2006, 2009), with one caveat. 4. Generics are rejected*even if they predicate highly prevalent properties*if the members of the kind that lack the property have an equally salient alternative property instead, for example, being a hardcover book instead of a paperback (books are paperbacks); being male instead of female (ducks are female); being left-handed instead of right-handed (e.g., Canadians are right-handed).
GENERICS AND DEFAULT INFERENCES
889
Recent findings suggest that generics may articulate our core conceptual knowledge about kinds (e.g., Gelman, 2010; Hollander, Gelman, & Star, 2002; Leslie, 2007, 2008; Leslie et al., 2011; Leslie & Gelman, 2011; Prasada & Dillingham, 2006, 2009; Prasada, Khemlani, Leslie, & Glucksberg, 2011). If this hypothesis is correct, then our conceptual knowledge is not based solely on quantitative information such as prevalence but rather is sensitive to the semantic factors outlined above. To date, most studies that have tested this hypothesis have been developmental (e.g., Brandone et al., in press). In this paper, we take a different approach and address the question of how conceptual knowledge is structured by examining systematic patterns of default inference among adults. In particular, we ask: what factors promote default inferences? If generics express core conceptual knowledge, then one would predict that default inferences would be influenced by the same factors that determine our agreement to a generic, and not be based solely on beliefs about prevalence. How do adults and children reason with generics? A fair number of studies have addressed this question (Chambers, Graham, & Turner, 2008; Cimpian, Brandone, & Gelman, 2010; Connolly, Fodor, Gleitman, & Gleitman, 2007; Elio & Pelletier, 1993, 1996; Gelman & Bloom, 2007; Graham, Nayer, & Gelman, in press; Gelman & Raman, 2003; Gelman, Star, & Flukes, 2002; Hampton, 2009; Pelletier & Elio, 2003). For both adults and children, belief in the generic generalization seems to dispose people to judge that a given member of the kind will have the relevant property, a view we call the generics hypothesis. However, previous work has not systematically considered low prevalence accepted generics (e.g., ticks carry Lyme disease; for one exception see Cimpian, Brandone, & Gelman, 2010) or high prevalence rejected generics (e.g., books are paperbacks). As a result, we do not know whether the tendency to make strong inferences from generics is a result of generic belief per se, or whether these inferences are driven by beliefs about the relative frequency with which the property occurs. For example, if people learn that Tweety is a bird, they may use their knowledge that most birds fly to infer by default that Tweety flies (Gerla, 1994). We refer to this account as the base rate or prevalence hypothesis. The experiment reported here examined whether default inferences are guided by the relevant generic generalization (the generics hypothesis) or by beliefs about the prevalence of the relevant property for the kind (the prevalence hypothesis). To test this, we chose a variety of items for which generic agreement is dissociated from estimates of prevalence. Table 1 includes the taxonomy of these types of generics along with descriptions of the operationalisations of the categories to which they belong. Participants were told that an arbitrary individual was a member of a particular kind, and they were then asked how confident they were that the individual had a given property. For example, they were told that Jumpy is a tick, and then asked to report their confidence in the claim that Jumpy carries Lyme disease. No generics were presented here*participants evaluated Jumpy carries Lyme disease without seeing the generic ticks carry Lyme disease. That is because the presentation of an explicit generic might lead participants to rely more on the generic than they otherwise would. Instead, participants could draw on whatever knowledge or information they deemed relevant. The generics hypothesis predicts that participants inferences would be sensitive to the same factors as generic generalizations, and thus agreement to a generic would be correlated with the tendency to make a default inference, over and above beliefs about prevalence.
890
TABLE 1 Various types of generic generalizations and operational denitions of the categories to which they belong
KHEMLANI ET AL.
Item type Definitional
Example Triangles have three sides
Truth value of the generic True True
Definition Property must be universally true of all the members of the kind; no exceptions. Property must be prevalent though not universal among members of the kind; some exceptional members (e.g., albino tigers) fail to possess it. The property must be central, principled or essential (Gelman, 2003; Medin & Ortony, 1989; Prasada & Dillingham, 2006, 2009). In our study, we included only items that passed linguistic tests as outlined by Prasada and Dillingham (2006, 2009). Property must be held by only a minority of the kind, but must be central, principled or essential (Gelman, 2003; Medin & Ortony, 1989), and so pass Prasada and Dillinghams tests (Prasada & Dillingham, 2006, 2009). For our purposes we restricted these items to methods of gestation, methods of nourishing the very young, and characteristic physical traits held by only one gender. Property must be prevalent among members of the kind, and must not be an essential or principled property. Property must hold for only a small minority of the kind, and must signify something dangerous and to be avoided. Property must be prevalent among members of the kind, and there must exist a sufficiently salient alternative property (e.g., being left-handed), such that the generic form of the predication sounds false or mistaken. Property must be held by very few members of the kind, and must not signify something dangerous. The generic form of the predication must appear false or mistaken. No members of the kind may hold the property.
Majority characteristic Tigers are striped
Minority characteristic Lions have manes
True
Majority statistical Striking Majority FGs
Cars have radios Pit bulls maul children Canadians are right-handed
True True False
Minority FGs False
Rooms are round Sharks have wings
False False
Note: FGs False generalizations.
891
METHOD Norming study

A norming study was conducted to provide data on prevalence estimates, cue validity ratings, and generic agreement for the items used in the experiment (see Table 2). Sixty-four online volunteers were randomly assigned to perform one of three tasks: 1. Prevalence estimation. Fifteen participants estimated the prevalence of a property within a category, for example, they were asked What percentage of mosquitoes carry malaria? and responded on a 0 100 scale. 2. Cue validity rating. Twenty-seven participants evaluated the cue validity (Beach, 1964) of each item, as cue validity could potentially mediate prevalence estimation and generic agreement. That is, people may determine that a given mosquito has malaria because the probability of having malaria and not being a mosquito is quite low. Thus, participants in the norming study were told, Suppose x carries malaria and were asked, How likely is it that x is a mosquito? They responded on a 7-point Likert scale in which 3 very likely, 0 not sure, and 3 very unlikely. 3. Generic agreement. Twenty-two participants agreed or disagreed with generic statements, for example, mosquitoes carry malaria, by responding on a 7point Likert scale in which 3 definitely agree, 0 cant tell, and 3 definitely disagree.
Participants
Twenty-nine volunteers completed the study for monetary compensation through Mechanical Turk, an online platform hosted through Amazon.com (for an analysis of the validity of results obtained through this platform, see Paolacci, Chandler, & Ipeirotis, 2010). None of the participants reported that they had received any training
TABLE 2 Mean ratings of condence to the given conclusion (3 Im condent its true and 3 Im condent its false and prevalence estimates (0 100), cue validities (3 very likely, 0 not sure, 3 very unlikely), and generic agreement (3 denitely agree, 0 cant tell, 3 denitely disagree) for the corresponding generic generalization as a function of item type Experimental data Mean confidence rating 2.5 2.6 1.7 1.5 0.7 0.6 0.3 2.5 Norming data
Item type Definitional Majority characteristic Minority characteristic Majority statistical Striking Majority FG Minority FG False
Prevalence estimate 89 89 65 67 24 60 18
Cue validity 1.0 1.1 0.6 0.5 0.9 0.6 0.5
Generic agreement 2.6 2.7 2.1 1.2 1.3 0.0 0.9
Note: Norming data were not obtained for the false items. FG False generalization.
892
KHEMLANI ET AL.
in logic, and none had participated in experiments involving generics before. The participants completed the experiment online using an interface written in Javascript, HTML, and PHP.
Materials
Five types of items that are generally accepted in generic form were used (see Table 1): definitional, e.g., triangles have three sides; majority characteristic, e.g., tigers are striped; minority characteristic, e.g., ducks lay eggs; majority statistical, e.g., cars have radios; and striking item types, e.g., mosquitoes carry malaria. Three item types that are generally rejected as generics were also used: majority false generalizations (FGs), e.g., Canadians are right-handed; minority FGs, e.g., Rooms are round; and false items, e.g., sharks have wings. The participants were never presented with generic statements, but rather the generics were used to construct two statements: one was a premise concerning the category membership of an individual (e.g., Buzzy is a mosquito) and a second was a conclusion about a property that the individual might possess (e.g., Buzzy carries malaria). The names of the individuals were chosen so as not to imply anything relevant about the truth of the property statement. For example, for items for which gender was relevant, only gender-neutral names were used. The materials are provided in Appendix 1.
Design and procedure

Participants carried out a default inference task in which they provided their level of confidence that a conclusion (e.g., Buzzy carries malaria) was true by selecting from a 7-point Likert scale where 3 meant Im confident its true and 3 meant Im confident its false. The eight different item types were distributed such that participants received 20 false items and 10 each of the other item types. Thus, they were given 90 items, in which 50 of the corresponding generics were accepted and 40 were rejected. The majority FGs (e.g., books are paperbacks, Canadians are righthanded) served to provide a comparison for the majority statistical items (e.g., cars have radios). The generics hypothesis predicted that people would be more willing to make default inferences from the majority statistical items (e.g., cars have radios) than from the majority FG items (e.g., Canadians are right-handed). The items that corresponded to rejected generic types*especially minority FGs and false statements*also served to provide opportunities for low-confidence ratings, and thereby balanced the Likert scale responses. Each participant received a different randomised order of the experimental items, and the participants were familiarised with the response scales before performing the default inference task.1
RESULTS AND DISCUSSION

The mean confidence ratings for default inferences and the norming data for each generic type are provided in Table 2.
We also ran a separate, counterbalanced block of the generic agreement task on the participants in the main experiment. For brevity, these data were not used as predictors in any of the analyses in the paper, though they produced effects comparable to the generic agreement data collected in the norming study (as analyzed in Khemlani, Leslie, & Glucksberg, 2009). The participants default inferences were not affected by whether the generic agreement task or the inference task came first, and so the two samples were collapsed in all subsequent analyses.
1
GENERICS AND DEFAULT INFERENCES TABLE 3 Correlation matrix among the three predictors taken from the norming study (N 64) Variable 1. Prevalence estimates 2. Cue validity ratings 3. Generic agreements 1 0.67 0.77 2 0.83 3
893
To test whether the strength of default inferences is accounted for by prevalence or cue validity (but not generic agreement), the prevalence estimates, cue validity ratings, and generic agreement data (collapsing over item type) were regressed against mean confidence ratings as implemented in the lme4 package (linear mixed-effects models using S4 classes; Bates, Maechler, & Dai, 2008) in the statistical software R (version 2.10; R Development Core Team, 2008). The data were subjected to a hierarchical linear regression analysis that compared three models: Model 1 used prevalence alone as a predictor, Model 2 added cue validity as a predictor, and Model 3 added generic agreement data as a predictor. Table 3 provides a correlation matrix for the three predictors used in the models, and Table 4 gives the results of the analysis. In all three models, prevalence estimates were significant positive predictors of confidence ratings. The model that considered generic agreement as a predictor accounted for significantly more variance than either of the other two models 2 (R2 Model 3 .65 vs. RModel 2 .58, p B .0001). Likewise, prevalence and cue validity (Model 2) accounted for less variance than the model that considered generic agreement (Model 3). The analysis suggests that a strictly prevalence-based account of default reasoning (Model 1) is incomplete. Generic agreement and prevalence accounted for more variance than the account based on prevalence alone.
Analysis of item types by default inference confidence ratings and prevalence estimates
For an analysis of participants performance for specific item types, mean confidence ratings and prevalence estimates were subjected to Friedman nonparametric analyses of variance, which revealed that confidence ratings varied significantly as a function of
TABLE 4 Hierarchical regression analysis for three predictors of condence ratings: prevalence estimates, cue validity ratings, and generic agreement data Model 1 2 Predictor Intercept Prevalence estimates Intercept Prevalence estimates Cue validity ratings Intercept Prevalence estimates Cue validity ratings Generic agreement B 0.66 0.03 0.57 0.03 0.56 0.22 0.01 0.19 0.54 SE (b) 0.14 0.00 0.13 0.00 0.14 0.13 0.00 0.18 0.08 b 0.74 *** 0.58 *** 0.23 *** 0.36 *** 0.07 ns 0.55 ***
Note: R2 .55 for Model 1; R2 .58 for Model 2; R2 .65 for Model 3 (all ps B .001). ***p B .001.
894
KHEMLANI ET AL.
item type, x2(7, 28) 170.29, p B .0001, as did prevalence estimates, x2(6, 28) 79.49, p B .0001. Additional planned comparisons tested whether confidence ratings and prevalence estimates differed significantly across the various item types and are provided in Table 5. For all comparisons, we computed the Bonferroni correction with a family-wise alpha rate of .05, and we include values of Cliff s d, a nonparametric effect size indicator whose value ranges from 1 to 1 (see Cliff, 1993). We found that default inferences and prevalence estimates often coincide. For instance, definitional (e.g., triangles have three sides) and majority characteristic (e.g., tigers have stripes) items did not differ reliably on their mean confidence ratings or their corresponding prevalence estimates of 89% and 89%, respectively. Similarly, the mean confidence ratings and prevalence estimates for minority characteristic (e.g., ducks lay eggs) and majority statistical (e.g., cars have radios) were not significantly different. These comparisons suggest a strong relationship between mean confidence ratings and estimated prevalence in line with the prevalence hypothesis. However, the theoretically instructive comparisons are those in which both estimated prevalence and agreement to the generic differed. Striking (e.g., ticks carry Lyme disease) and majority FG items (e.g., books are paperbacks) provide clear evidence that prevalence was not the only factor driving default inference. Consider first the striking item types. If participants relied primarily on prevalence information, the mean confidence ratings for striking items should have been lower than the ratings for majority FGs because the respective prevalence estimates for these two item types are reliably different (24% vs. 60%). Despite this difference, the mean confidence rating for striking generics is comparable to those for majority FGs (0.70 vs. 0.60). The opposite effect holds for the mean confidence ratings of majority FG items (e.g., books are paperbacks), which were lower than one would expect if participants relied exclusively on prevalence. The estimated prevalence of 60% for majority FGs is comparable to the estimated prevalence of 65% for minority characteristic items, yet the mean confidence ratings for these two item types are
TABLE 5 Planned comparisons between different item types used in the experiment and the norming study. For each comparison, we provide a pairwise analysis of mean condence ratings obtained for the main experiment as well as mean prevalence estimates obtained from the norming study Wilcoxon test, z-value 0.88 0.51 0.83 0.97 1.17 3.35 4.12 1.65
Planned comparison Definitional vs. majority characteristic Mean confidence rating: 2.5 vs. 2.6 Mean prevalence estimate: 89% vs. 89% Minority characteristic vs. majority statistical Mean confidence rating: 1.7 vs. 1.5 Mean prevalence estimate: 65% vs. 67% Striking vs. majority FG Mean confidence rating: 0.70 vs. 0.60 Mean prevalence estimate: 24% vs. 60% Minority characteristic vs. majority FG Mean confidence rating: 1.7 vs. 0.60 Mean prevalence estimate: 65% vs. 60%
p-value .30 .30 .30 .30 .30 B .005 B .0005 .30
Cliffs d .10 .17 .12 .09 .16 .82 .75 .27
895
reliably different (0.6 vs. 1.7, respectively). Also note that agreement for majority FG and minority characteristic items is 0.0 and 2.1 respectively, that is, generic agreement corresponds with the confidence ratings while prevalence estimates do not. Since the only difference between these two item types is whether or not they are agreed to as generics, it is strong evidence that participants relied on information other than prevalence, in this case, whether or not the item is agreed to as a generic. These two cases can be construed as instances of base rate neglect (Kahneman & Tversky, 1973; Tversky & Kahneman, 1982). The above analyses show that striking, majority statistical, and majority FG items yield default inferences that a prevalence-based model cannot account for. These results corroborate the generics hypothesis, namely that default inferences are guided in part by belief in the underlying generic generalization, and thus that default inferences are sensitive to the same factors as generics are (e.g., whether the property is strikingly dangerous or not).
GENERAL DISCUSSION
What factors guide default reasoning about arbitrary members of kinds? We examined two alternative hypotheses. The prevalence hypothesis posits that confidence in a default inference is proportional to a reasoners belief about the statistical properties of a kind, for example, that most dogs have four legs. The generics hypothesis proposes that inferences concerning the properties of individual members of a kind are driven by belief in the corresponding generic generalization, for example, the belief that dogs have four legs. If it is indeed the case that generic generalizations are conceptually fundamental (e.g., Gelman, 2010; Leslie, 2007, 2008), then one would expect that the generics hypothesis would be correct, and that default inferences would be sensitive to the same factors that drive acceptance of generic generalizations. Our findings suggest that this is indeed the case. We tested these hypotheses by providing participants with category information such as, Speedy is a car, and then asking them to rate their confidence in a default inference, for example, Speedy has a radio. We compared confidence ratings with prevalence estimates, cue validity ratings, and generic agreement from a separate group of participants, and found that generic agreement and estimated prevalence is a better predictor of performance than estimated prevalence alone. For instance, prevalence estimates for striking (e.g., mosquitoes carry malaria) and majority FG items (e.g., Canadians are right-handed) differed reliably (24% and 60%, respectively), yet ratings of confidence about default inferences were comparable (0.7 and 0.6, respectively). We propose that the explanation for this is that, despite the fact that striking items were associated with lower prevalence estimates than majority FG items, striking items were more likely to be accepted in generic form than majority FGs. Since peoples default inferences are guided by both their beliefs about prevalence as well as their belief in the relevant generic generalization, this led to the striking items being rated comparably to the majority FGs in the default inferences task, despite their disparate associated prevalence estimates. In other words, the fact that people accept striking generic generalizations led them to be more confident of the default inference than their beliefs about prevalence alone would license and conversely for majority FGs. Likewise, the prevalence estimates for majority FG items like Canadians are right-handed were comparable to those for majority statistical items (e.g., cars have radios) and minority characteristic items (e.g., ducks lay
896
KHEMLANI ET AL.
eggs), yet people were much less likely to agree that the arbitrary Canadian is righthanded than that the arbitrary car has a radio or that the arbitrary duck lays eggs. People know, of course, that only female ducks lay eggs (Leslie et al., 2011), and so the tendency to agree that an arbitrary duck, Quacky, lays eggs is high even though people know that immature ducks and male ducks cannot lay eggs. Cimpian, Brandone, and Gelman (2010) report that people may also discount information about prevalence when considering generic assertions that concern novel kinds. Their participants were willing to agree to striking and characteristic generics even when a propertys prevalence was very low (e.g., accepting lorches have dangerous feathers even if only 30% of lorches have dangerous feathers). However, their participants expected that, if a striking or characteristic generic is true, the property would be prevalent (e.g., upon being told that lorches have dangerous feathers, they estimated that over 90% of lorches would have dangerous feathers). Thus they found, as we did, that inferences concerning the properties of individuals were influenced by agreement to the generic. The study presented here indicates that default inferences are sensitive to the same factors that generics are sensitive to, and thus may well be guided by belief in the generic generalization. The findings presented here provide additional evidence for the hypothesis that our core conceptual information is richly structured and sensitive to the nature of the property being generalised, for example, whether it is characteristic of the kind or strikingly dangerous, rather than being driven solely by considerations such as how many members of the kind have the property. Any model of concepts that is built solely on statistical measures such as prevalence and cue validity (e.g., Rosch, 1978 and successors) will not be adequate to model human cognition. In recent years there has been a surge of interest in Bayesian models of cognition, with theorists arguing that generalization, categorization, and induction can be unified under a Bayesian framework (Griffiths et al., 2008; Tenenbaum & Griffiths, 2001). Despite successful applications of these models, critics have countered that human cognition is far too complex to be explained by such a framework (Boroditsky & Ramscar, 2001; Jones & Love, in press; Love, 2001). Though the global adequacy of Bayesian models is beyond the scope of our analysis, we note that our findings suggest that inferences about particular members of kinds cannot be modelled by a simple or unstructured application of the probabilistic calculus. That is, peoples judgments about whether a given member of a kind K has a property P does not simply reduce to their beliefs about the conditional probability of having P given that something is a K (nor to their beliefs about the conditional probability that something is a K given that it has P; nor to the conjunction of the two sets of beliefs). Our results indicate that a far more complex and sophisticated model*whether Bayesian or otherwise*will be needed to account for how we reason about members of kinds. The model must reflect the rich conceptual structure we find in generic generalizations.
Manuscript received 15 November 2010 Revised manuscript received 22 June 2011 First published online 19 October 2011
REFERENCES
Bates, D. M., Maechler, M., & Dai, B. (2008). lme4: Linear mixed-effects models using S4 classes [R package Version 0.999375-24]. Retrieved from http://cran.r-project.org/web/packages/lme4/index.html
897
Beach, L. R. (1964). Cue probabilism and inference behavior. Psychological Monographs: General and Applied, 78, 1 20. Boroditsky, L., & Ramscar, M. (2001). First, we assume a spherical cow. . .. Behavioral and Brain Sciences, 24, 656 657. Brandone, A., Cimpian, A., Leslie, S. J., & Gelman, S. A. (in press). Do lions have manes? For children, generics are about kinds rather than quantities. Child Development. Carlson, G. (1997). Reference to kinds in English. Doctoral dissertation, University of Massachusetts, Amherst, MA, USA. Carlson, G., & Pelletier, F. J. (1995). The generic book. Chicago, University of Chicago Press. Chambers, C. G., Graham, S. A., & Turner, J. N. (2008). When hearsay trumps evidence: How generic language guides preschoolers inferences about unfamiliar things. Language and Cognitive Processes, 23, 749 766. Cimpian, A., Brandone, A., & Gelman, S. A. (2010). Generic statements require little evidence for acceptance but have powerful implications. Cognitive Science, 34, 1452 1482. Cliff, N. (1993). Dominance statistics: Ordinal analyses to answer ordinal questions. Psychological Bulletin, 114, 494 509. Connolly, A., Fodor, J., Gleitman, L. R., & Gleitman, H. (2007). Why stereotypes dont even make good defaults. Cognition, 103, 1 22. Elio, R., & Pelletier, F. J. (1993). Human benchmarks on AIs benchmark problems. In Proceedings of the 15th Annual Conference of the Cognitive Science Society (pp. 406 411). Hillsdale, NJ: Lawerence. Elio, R., & Pelletier, F. J. (1996). On reasoning with default rules and exceptions. Proceedings of the 18th Annual Conference of the Cognitive Science Society (pp. 131 136). Hillsdale, NJ: Lawrence Erlbaum. Gelman, S. A. (2003). The essential child. New York: Oxford University Press. Gelman, S. A. (2010). Generics as a window onto young childrens concepts. In F. J. Pelletier (Ed.), Kinds, Things, and Stuff (pp. 100 123). New York: Oxford University Press. Gelman, S. A., & Bloom, P. (2007). Developmental changes in the understanding of generics. Cognition, 105(1), 166 183. Gelman, S. A., & Raman, L. (2003). Preschool children use linguistic form class and pragmatic cues to interpret generics. Child Development, 74(1), 308 325. Gelman, S. A., Star, J. R., & Flukes, J. E. (2002). Childrens use of generics in inductive inferences. Journal of Cognition and Development, 3(2), 179 199. Gerla, G. (1994). The probability that Tweety is able to y. International Journal of Intelligent Systems, 9, 403 409. Graham, S. A., Nayer, S. L., & Gelman, S. A. (2011). Two-year-olds use the generic/nongeneric distinction to guide their inferences about novel kinds. Child Development, 82(2), 493 507. Grifths, T., Kemp, C., & Tenenbaum, J. (2008). Bayesian models of cognition. In R. Sun (Ed.), Cambridge handbook of computational psychology. Cambridge: UK Cambridge University Press. Hampton, J. (2009). Stability in concepts and evaluating the truth of generic statements. In F. J. Pelletier (Ed.), Kinds, things, and stuff (pp. 80 100). New York: Oxford University Press. Hollander, M. A., Gelman, S. A., & Star, J. (2002). Childrens interpretation of generic noun phrases. Developmental Psychology, 38(6), 883 894. Jones, M., & Love, B. (in press). Bayesian fundamentalism or enlightenment? On the explanatory status and theoretical contributions of Bayesian models of cognition. Behavioral and Brain Sciences. Kahneman, D., & Tversky, A. (1973). On the psychology of prediction. Psychological Review, 80, 237 251. Khemlani, S., Leslie, S. J., & Glucksberg, S. (2009). Generics, prevalence, and default inferences. In N. Taatgen & H. van Rijn (Eds.), Proceedings of the 31st Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society. Khemlani, S., Leslie, S. J., Glucksberg, S., & Rubio-Fernandez, P. (2007). Do ducks lay eggs? How humans interpret generic assertions. In D. S. McNamara & J. G. Trafton (Eds.), Proceedings of the 29th Annual Conference of the Cognitive Science Society (pp. 395 400). Nashville, TN: Cognitive Science Society. Krifka, M., Pelletier, F. J., Carlson, G., ter Meulen, A., Chierchia, G., & Link, G. (1995). Genericity: An introduction. In G. Carlson & F. J. Pelletier (Eds.), The generic book (pp. 1 125). Chicago, IL: Chicago University Press. Leslie, S. J. (2007). Generics and the structure of the mind. Philosophical Perspectives, 21(1), 375 405. Leslie, S. J. (2008). Generics: Cognition and acquisition. The Philosophical Review, 117(1), 1 49. Leslie, S. J., & Gelman, S. A. (2011). Quantied statements are recalled as generics: Evidence from preschool children and adults. Manuscript under review. Leslie, S. J., Khemlani, S., & Glucksberg, S. (2011). Do all ducks lay eggs? The generic overgeneralization effect. Journal of Memory and Language, 65, 15 31. Love, B. C. (2001). Three deadly sins of category learning modelers. Behavioral and Brain Sciences, 24, 687 688.
898
KHEMLANI ET AL.
Maddox, W. T., & Ashby, F. G. (1993). Comparing decision bound and exemplar models of categorization. Perception and Psychophysics, 53, 49 70. Medin, D. L., & Ortony, A. (1989). Psychological essentialism. In S. Vosniadou & A. Ortony (Eds.), Similarity and analogical reasoning (pp. 179 195). New York, NY: Cambridge University Press. Murphy, G. L. (2002). The big book of concepts. Cambridge, MA: MIT Press. Paolacci, G., Chandler, J., & Ipeirotis, P. G. (2010). Running experiments on Amazon Mechanical Turk. Judgment and Decision Making, 5, 411 419. Pelletier, F. J., & Elio, R. (2003). Logic and cognition: human performance in default reasoning. In P. Ga rdenfors, J. Wolenski, & K. Kijania-Placet (Eds.), In the scope of logic, methodology, and philosophy of science, Vol. I (pp. 137 154). Dordrecht: Kluwer Academic. Prasada, S., & Dillingham, E. (2006). Principled and statistical connections in common sense conception. Cognition, 99, 73 112. Prasada, S., & Dillingham, E. (2009). Representation of principled connections: A window onto the formal aspect of common sense perception. Cognitive Science, 33, 401 448. Prasada, S., Khemlani, S., Leslie, S. J., & Glucksberg, S. (2011). Conceptual distinctions amongst generics. Manuscript under review. R Development Core Team (2008). R: A language and environment for statistical computing (Version 2.10). Vienna: R Foundation for Statistical Computing. Retrieved from http://www.R-project.org Rosch, E. (1978). Principles of categorization. In E. Rosch & B. B. Lloyd (Eds.), Cognition and categorization (pp. 27 48). Hillsdale, NJ: Lawrence Earlbaum Associates. Tenenbaum, J., & Grifths, T. (2001). Generalization, similarity, and Bayesian inference. Behavioral and Brain Sciences, 24, 629 640. Tversky, A., & Kahneman, D. (1982). Evidential impact of base rates. In D. Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncertainty: heuristics and biases (pp. 153 160). Cambridge, UK: Cambridge University Press.
APPENDIX 1
Materials used in the experiment. Participants were given a category membership premise and were asked to rate their condence in the predication
Item type Definitional Definitional Definitional Definitional Definitional Definitional Definitional Definitional Definitional Definitional Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Majority characteristic Category membership Barky is a dog Breezy is an elm Fluffy is a poodle John is a bachelor Righty is a rectangle Snoozy is a cat Snuffly is a sow Triggy is a triangle Vivi is a vixen x is an even number Tweety is a bird Cheety is a cheetah Grazer is a cow Fido is a dog Aussie is a kangaroo Regal is a lion Hempel is a raven Squiggy is a squirrel Prowly is a tiger Spotty is a cat Predication Barky is a mammal Breezy is a tree Fluffy is a dog John is unmarried Righty has four sides Snoozy is an animal Snuffly is a female pig Triggy has three angles Vivi is a female fox x is divisible by 2 Tweety has wings Cheety runs fast Grazer moos Fido has a tail Aussie hops Regal roars Hempel is black Squiggy eats nuts Prowly has stripes Spotty has four legs

Appendix (Continued )
899
Item type Majority characteristic Minority characteristic Minority characteristic Minority characteristic Minority characteristic Minority characteristic Minority characteristic Minority characteristic Minority characteristic Minority characteristic Majority statistical Majority statistical Majority statistical Majority statistical Majority statistical Majority statistical Majority statistical Majority statistical Majority statistical Majority statistical Striking Striking Striking Striking Striking Striking Striking Striking Striking Striking Majority FG Majority FG Majority FG Majority FG Majority FG Majority FG Majority FG Majority FG Majority FG Majority FG Minority FG Minority FG Minority FG Minority FG Minority FG Minority FG Minority FG Minority FG Minority FG Minority FG False False False False False False False
Category membership Hammy is a hamster Bleaty is a sheep Browny is a moose Fuzzy is a goat Galloper is a horse Oinky is a pig Quacky is a duck Roary is a lion Sparky is a mammal Tweety is a cardinal Boozey is a bar Brassy is a trumpet Cozy is a jacket Footsy is a shoe Luigi is an Italian Metroline is a subway Nike Relax is a shirt Old Rickety is a barn Speedy is a car Ticky is a clock Bob had a stroke Buzzy is a mosquito Fido is a Rottweiler Fins is a shark Jumpy is a tick Leo is a lion Marsha is a hurricane Squeaky is a rat Stripey is a tiger Wingy is a bird Happy Times is a book Joe is a Canadian X-Rig is a computer Beaky is a duck Dr. Jones is an engineer Jane is an American Pat Brown is a school teacher Plucky is a lion Southern cross is a tree Viv is an athlete Fluffy is a dog Junes is a restaurant Kate is an American Kitty is a cat Nurse Jones is a nurse Pablo is a Spaniard Pouncer is a tiger Table 371 is a table The Thornberry Suite is a room Waggy is a dog Ally is an alligator Bushy is a fox Cheeky is a hamster Father OReilly is a Catholic priest Fawny is a deer Feathers is a bird Freddy is a frog
Predication Hammy has nipples Bleaty has an udder Browny has antlers Fuzzy has horns Galloper gives live birth Oinky suckles its young Quacky lays eggs Roary has a mane Sparky produces milk Tweety is red Boozey is noisy Brassy is loud Cozy has a zipper Footsy has laces Luigi eats spaghetti Metroline is crowded Nike Relax has a collar Old Rickety is red Speedy has a radio Ticky is round Bobs stroke caused paralysis Buzzy carries malaria Fido mauls children Fins attacks swimmers Jumpy carries Lyme disease Leo eats people Marsha damaged buildings Squeaky carries disease Stripey attacks people Wingy carries avian flu Happy Times is a paperback Joe is right-handed X-Rig is a PC Beaky is female Dr. Jones is male Jane is brunette Pat Brown is female Plucky is male Southern cross is a deciduous tree Viv is a student Fluffy is blind Junes is a Chinese restaurant Kate is dyslexic Kitty is white Nurse Jones is a man Pablo is Jewish Pouncer is albino Table 371 is 10ft long The Thornberry Suite is round Waggy has three legs Ally has fur Bushy lays eggs Cheeky has stripes Father OReilly is married Fawny eats meat Feathers suckles its young Freddy can fly
900
KHEMLANI ET AL.
Appendix (Continued )
Item type False False False False False False False False False False False False False
Category membership Gills is a fish Grizzly is a wolf Hexy is a hexagon Hissy is a snake Kanga is a kangaroo Larry is a leaf n is an odd number Robby is a robin Sammy is a shark Savannah is a lion Sharpy is a television Whiskers is a tiger Zebediah is a zebra
Predication Gills is a mammal Grizzly says meow Hexy has four sides Hissy has legs Kanga has fins Larry is blue n is divisible by two Robby is four-legged Sammy has wings Savannah is black Sharpy makes coffee Whiskers has horns Zebediah has spots
Cognitive Psychology 64 (2012) 186214
Contents lists available at SciVerse ScienceDirect
Cognitive Psychology
journal homepage: www.elsevier.com/locate/cogpsych
Quantied statements are recalled as generics: Evidence from preschool children and adults
Sarah-Jane Leslie a,, Susan A. Gelman b
a b
Department of Philosophy, 1879 Hall, Princeton University, Princeton, NJ 08544, United States Department of Psychology, 530 Church Street, University of Michigan Ann Arbor, Ann Arbor, MI 48109, United States
a r t i c l e
i n f o
a b s t r a c t
Generics are sentences such as ravens are black and tigers are striped, which express generalizations concerning kinds. Quantied statements such as all tigers are striped or most ravens are black also express generalizations, but unlike generics, they specify how many members of the kind have the property in question. Recently, some theorists have proposed that generics express cognitively fundamental/default generalizations, and that quantied statements in contrast express cognitively more sophisticated generalizations (Gelman, 2010; Leslie, 2008). If this hypothesis is correct, then quantied statements may be remembered as generics. This paper presents four studies with 136 preschool children and 118 adults, demonstrating that adults and preschoolers alike tend to recall quantied statements as generics, thus supporting the hypothesis that generics express cognitively default generalizations. 2011 Elsevier Inc. All rights reserved.
Article history: Accepted 1 December 2011 Available online 4 January 2012 Keywords: Generics Quantiers Memory Semantics Children
1. Introduction Much of cognitive psychology has focused on how our core conceptual knowledge of the world is represented (e.g., Carey, 2009; Gelman, 2003; Holyoak & Morrison, 2005; Keil, 1989; Murphy, 2002; Rosch, 1973; Smith & Medin, 1981; and myriad others). A signicant portion of this core conceptual knowledge consists of generalizations about kinds (e.g., Gelman, 2003; Margolis & Laurence, 1999; Prasada, 2000). For example, the beliefs that connect tigers with stripes, dogs with tails, and doctors with healing people all involve the generalization of properties to kinds. What is the nature of these kind-based generalizations? Are they at heart quantitative and statistical (e.g., Rosch, 1973), or some Corresponding author.
E-mail address: sjleslie@princeton.edu (S.-J. Leslie).
0010-0285/$ - see front matter 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.cogpsych.2011.12.001
S.-J. Leslie, S.A. Gelman / Cognitive Psychology 64 (2012) 186214
187
thing more complex and richly structured (e.g., Carey, 2009; Gelman, 2003; Prasada & Dillingham, 2006, 2009)? Researchers have studied these questions using an enormously wide variety of methods. One example includes the body of empirical work devoted to studying the nature of kind-based generalizations in memory (e.g., McCarthy, 1995; Meyer, 1970; Smith, Shoben, & Rips, 1974). In particular, researchers have for many years studied the nature of semantic memory (or gist memory) as a way of assessing which aspects of a target will be encoded, and which will be forgotten (e.g., Jacoby & Dallas, 1981; Kintsch, 1988; Tulving, 1972). Thus, classic and ongoing studies of semantic memory provide a window onto aspects of conceptual representation. This paper presents a series of studies conducted on both children and adults, designed to assess their semantic memory for different types of kind-based generalizations. Kind-based generalizations may be either quanticational (e.g., All tigers have stripes) or generic (e.g., Tigers have stripes). Which (if either) of these forms corresponds to our core conceptual beliefs about kinds? Answering this question has important consequences for the study of concepts and categorization. If our most basic way of generalizing information about kinds is quanticational, this means that it can be characterized in quantitative, statistical terms that our semantic knowledge is organized in terms of logic sets and set-inclusion relations, as these are the hallmarks of quanticational generalizations (e.g., Barwise & Cooper, 1981). Generic generalizations, in contrast, cannot be characterized in these terms, but rather reect richer and more complex relations between the kind and the property, which cannot be reduced to purely formal, quantitative terms (Carlson, 1977; Leslie, 2007). The precise nature of generic generalizations is not yet fully understood (but see Cimpian, Brandone, & Gelman, 2010; Cimpian, Gelman, & Brandone, 2010; Khemlani, Leslie, & Glucksberg, 2009, in press; Leslie, 2008; Prasada & Dillingham, 2006, 2009 for some important work). The main focus of this work is the nature of the underlying conceptual knowledge that is expressed by the linguistic form, rather than the linguistic form itself. However, understanding how our conceptually fundamental generalizations are expressed in language is itself of interest for several reasons. For example, it provides insight into subtle linguistic differences with implications for implicit messages that are conveyed in ordinary conversation. Further, because language is a primary source of information to children regarding regularities in the world around them (Gelman, 2010), this investigation can shed light on how conceptually important information is communicated to young children by parental speechand how they are interpreted by young children as they hear such speech. Recent theoretical and empirical work has begun to address whether general conceptual knowledge is quanticational or generic. In particular, Leslie (2007, 2008) and Gelman (2010) argue that generics, not quantied statements, articulate conceptually central generalizations. They argue that generics express cognitively fundamental, default generalizations, and that quantied statements, in contrast, express more sophisticated, less accessible generalizations. We briey review the data in favor of this claim, and then report a series of experiments designed to further test the predictions of this hypothesis. Children begin to produce generics around age two-and-a-half, and by the time they reach 3 4 years, they produce generics as frequently as adults (Gelman, Goetz, Sarnecka, & Flukes, 2008). Recent evidence suggests that two-and-a-half-year-olds understand that generics express general claim about kinds, rather than specic claims about individuals (Gelman & Raman, 2003; Graham, Nayer, & Gelman, 2011). Thus, as soon as children master the requisite syntactic skills, they produce and understand generics (Gelman, 2010; Gelman et al., 2008). When it comes to expressing generalizations about kinds, generics are used much more frequently than universal quantiers (e.g., all, every, any) in both childrens speech and in child-directed speech (Gelman, 2003; Gelman, Hollander, Star, & Heyman, 2000). Universal quantiers are used primarily in reference to specic sets of individuals (e.g., pick up all those toys) rather than to kinds as a whole (e.g., all tigers have stripes). Hollander, Gelman, and Star (2002) found that both 3- and 4year-olds could appropriately interpret the quantiers all and some in reference to a specic set of items (e.g., are all the crayons in the box?); however, they also found that 3-year-olds had difculty with all and some when the quantiers were used to make kind-wide generalizations (e.g., are all res hot?). Interestingly, while Hollander et al. found signicant developmental differences in the case of category-wide uses of all and some, they found no developmental differences in the
188
case of generics, suggesting that generics may be easier for young children to understand than quantiers. Tardif, Gelman, Fu, and Zhu (2011) recently replicated these experiments with Mandarin speaking preschoolers, and found that Mandarin speaking 3- and 4-year-olds had difculty with categorywide uses of quantiers. Hollander et al. (2002) and Tardif et al. (2011) further found that the pattern of errors that Englishand Mandarin-speaking preschoolers made with all and some specically involved their responding to quantied statements as though they were generics. That is, English- and Mandarin-speaking preschoolers gave the same pattern of responses regardless of whether they were presented with quantied statements (using all or some) or with generics, and further, this pattern was indistinguishable from the pattern of responses that older children and adults gave when presented with generics. Thus, these results may potentially reect that the preschoolers relied on their interpretations of generics when confronted with category-wide statements quantied with all and some. Might these results reect a more general tendency to assimilate quantied statements to generics? Leslie, Khemlani, and Glucksberg (2011) report evidence suggesting that adults may exhibit this tendency under some circumstances. In particular, adults sometimes accept statements such as all ducks lay eggs, despite knowing that male ducks dont lay eggs, and even despite having explicitly judged minutes before that male ducks dont lay eggs (Experiment 3). This nding would be explicable if English-speaking adults, like English- and Mandarin-speaking preschoolers, were relying on their interpretation of the widely accepted generic ducks lay eggs, instead of evaluating the incorrect universal all ducks lay eggs (Khemlani, Leslie, Glucksberg, & Rubio-Fernandez, 2007; Leslie et al., 2011; see also Jnsson & Hampton, 2006; Sloman, 1993, 1998, for results that are readily explicable on this hypothesis). From a theoretical perspective, the question remains why children and adults have this tendency. Leslie (2008) and Gelman (2010) hypothesize that generics express cognitively default generalizations, whereas quantied statements express more taxing and cognitively sophisticated ones. Such a hypothesis would explain why generics are so easy for children to acquire, and why they are the preferred way of expressing kind-based generalizations both in childrens speech and in child-directed speech: generic generalizations are the easiest sort of generalization for young children to process. It would also explain the tendency to interpret quantied statements as though they were generics. Because quantied statements involve more taxing, non-default generalizations, people sometimes fall back on the default generic interpretation instead of processing the more taxing one. Leslie (2008) also notes that such a generics-as-default hypothesis would explain an otherwise puzzling cross-linguistic fact. To express a quantied generalization in English, we have to use a specic quantier for example, all(/most/some) tigers are striped. To express a generic generalization, however, we just omit the quantier: tigers are striped. There is no special linguistic marker (gen) that specically signals a generic. Interestingly, this is not an isolated fact about English: no known language has a word or linguistic marker that specically signals a generic (Dahl, 1985; Krifka et al., 1995). In this way, generics are less syntactically marked than quantied statements. Chomsky (2000) argues that less marked syntactic forms may be associated with default interpretations. Thus, if generics express cognitively default generalizations, this would explain why they are syntactically less marked than quantied statements. The present paper reports a series of experiments designed to further test the generics-as-default hypothesis. The hypothesis was investigated using a memory task modeled on Gelman and Raman (2007), in which participants were presented with a series of pictures of animals, and told a novel fact about each kind of animal. Following a distractor task, they were asked to recall the fact they had been taught, upon being presented with the picture of the relevant animal. Gelman and Raman (2007) used this paradigm to test childrens and adults memory for generic vs. specic statements (e.g., cats sweat through their paws vs. this cat sweats through its paws). They were especially interested in memory for the type of sentence generic vs. specic and found that 3- and 4-year-olds and adults all tended to successfully recall generics as generics, and specics as specics. Gelman and Raman (2007) were primarily interested in how well young children differentiated information about individuals from information about kinds. We sought to adapt their paradigm to test a different set of predictions: namely that both children and adults would recall quantied statements as generics. If generics express cognitively more default generalizations than quantied state-
189
ments do (Gelman, 2010; Leslie, 2008), then quantied statements may be more likely to be stored in memory as generics than vice versa. Previous tests of the generics-as-default hypothesis have relied on participants evaluations of quantied statements, to see if there is evidence that they sometimes substitute their judgments of generics for their judgments of quantied statements (e.g., Hollander et al., 2002; Khemlani et al., 2007; Leslie et al., 2011; Tardif et al., 2011). (Conversely, some prior research has examined the recall of nouns vs. modiers (e.g., Ehrlich, 1975), but to our knowledge, no one has focused on quantiers and how they compare to generics.) The present experiments provide converging evidence by use of a distinctly different set of methods, which tap into a different set of behavioral responses. There are numerous studies across many areas that examine errors in recall in order to assess cognitive defaults. One early example is found in psycholinguistics, in particular the study of the differential processing of negative vs. afrmative statements. Adults are more likely to incorrectly recall a negative statement (e.g., John did not sing) as an afrmative statement (e.g., John sang) than vice versa, which researchers took to indicate that the afrmative form is more basic than the negative form (e.g., Cornish & Wason, 1970; Fillenbaum, 1966; Mehler, 1963). Other diverse examples of this methodology include work on scripts, indicating that people have default expectations concerning how events should unfold, which give rise to errors in recall and recognition tasks (e.g., Bower, Black, & Turner, 1979), and work on the processing of stereotype consistent/inconsistent information, which indicates that stereotypical beliefs are, by default, applied to individuals, at least under high processing loads (e.g., Macrae, Hewstone, & Grifths, 1993). Thus, the use of a memory paradigm would seem a promising way of testing the generics-as-default hypothesis. In Experiment 1A, adult participants were presented with photos of animals, and told a novel fact about each animal. Half of the facts were presented in generic form, and the other half were presented as containing a particular quantier either all, most, or some. We predicted, in accordance with the generics-as-default hypothesis, that participants would be more likely to recall quantied statements as generics than vice versa. Experiment 1B was designed to provide an even stronger test of the hypothesis that quantied statements will be recalled as generics. In Experiment 1B, adult participants never heard any generics in the learning phase only quantied statements. We predicted that participants would nonetheless recall some of the statements as generics. Experiments 2A and 2B followed the same design as Experiments 1A and 1B respectively, but were conducted with 3- and 4-year-old participants. Experiment 3 was designed in part to rule out a potential alternative explanation of 3-year-olds performance, namely that they are simply forgetting the rst word, i.e. the quantier. That is, if 3-year-olds recalled all three quantied statements as generics, then this may just be due to a dumb strategy of shortening the statement to exclude the rst word. Experiment 3 addressed this by pairing generics with the negative quantier no (e.g., No spiders have ten legs). If 3-year-olds are simply dropping the rst word, then they should recall such statements as positive generics (e.g., no spiders have ten legs would be recalled as spiders have ten legs). Finally, Experiments 4A and 4B were designed to rule out additional alternative explanations for the results of the earlier studies.
2. Experiment 1A Experiment 1A was designed to test adults memory for generics vs. quantied statements. The design was adapted from Gelman and Raman (2007): participants were shown a series of photographs of different animals, and were told a novel fact about each animal. Participants heard half of the facts as generic statements, and the other half as statements that contain a particular quantier (either all, most or some). Each participant heard only one particular quantier throughout. We predicted that quantied statements would be recalled as generic more often than generic statements would be recalled as quantied. One outstanding question is whether all three quantiers would be recalled as generic, or whether they might pattern differently. We predicted, following Leslie et al. (2011), that adults would have a tendency to recall statements quantied with all as generics, but what about most and some? Although several experiments have shown that adults differentiate most from generics (e.g., Cim-
190
Table 1 Items used in Experiments 1 and 2. Category Bear Rabbit Snail Hippo Spider Bee Rhino Camel Seahorse Cat Generic form Bears climb trees Rabbits keep food in their cheeks Snails breathe through their feet Hippos like to swim Spiders shed their skin Bees have ve eyes Rhinos have horns made out of hair Camels store fat in their humps Seahorses come out of their dads tummies Cats sweat through their paws Quantied form All/Most/Some All/Most/Some All/Most/Some All/Most/Some All/Most/Some All/Most/Some All/Most/Some All/Most/Some All/Most/Some All/Most/Some bears climb trees rabbits keep food in their cheeks snails breathe through their feet hippos like to swim spiders shed their skin bees have ve eyes rhinos have horns made out of hair camels store fat in their humps seahorses come out of their dads tummies cats sweat through their paws
pian, Brandone, et al., 2010; Cimpian, Gelman, et al., 2010), no experiments have yet been conducted to test whether most statements are ever interpreted as generics. We predicted that most statements would also tend to be recalled as generics, since most conveys information that applies broadly to the members of the kind. The quantier some may be interpreted somewhat differently by adults, however. Leslie et al. (2011) failed to nd any evidence that adults ever treat some statements as generics. This may be because adults interpret some as conveying information about only a limited number of members of a category (Bott & Noveck, 2004), and so in this way some statements may be interpreted more like specic statements about a particular member or set of members of the kind. Thus we predicted that some statements would be more readily distinguished from generics in memory than would all and most statements. 2.1. Method 2.1.1. Participants Forty-eight undergraduates (32 women; 16 men) were recruited from a large Midwestern public university. One additional participant was tested but dropped from analysis because of insufcient prociency in English. 2.1.2. Materials Ten photographs of animals were used.1 Each animal had a novel fact associated with it. The photographs and the facts were a subset of those used in Gelman and Raman (2007); see Table 1. Gelman and Raman (2007) pretested these items with preschoolers, and found that the facts were unfamiliar to children. Each fact was either presented in generic form or in quantied form (all, most, or some). 2.1.3. Procedure Each participant was tested individually. They were asked to listen carefully, because they would be asked to remember the information provided. They were presented with ve test items, one at a time. For each item, the participant rst saw a photograph depicting the animal, and then heard a novel fact associated with the animal. For example, participants saw a photo of a bear and heard one of the following sentences: Bears climb trees (generic), All bears climb trees (all), Most bears climb trees (most), or Some bears climb trees (some). Following presentation of the rst ve items, participants received a 4-min distractor task in which they were asked to solve a series of multidigit multiplication problems. They were then tested for their recall of the ve sentences. For each, they were shown the same photograph as they had seen during the initial presentation phase, and asked to recall what they had been told about the animal depicted. The entire process was then repeated with the
1 Participants in Studies 1A, 1B, 2A, and 2B each received an additional six sentences (therefore 16 total). However, due to experimenter error, the additional sentences were not counterbalanced across conditions. We therefore report only those results from the 10 sentences that were properly counterbalanced. The major patterns of results presented here are equivalent to those from the full set of 16 items.
191
remaining ve items. Responses were audiotaped and transcribed. The testing process took approximately 12 min to complete. Participants were randomly assigned to one of three groups: generic/all, generic/most, and generic/ some. Each group heard half the facts in generic form, and the other half in quantied form (either all, most, or some, depending on the condition). Assignment of item to wording condition (i.e., generic vs. quantied) was counterbalanced. Order of items within each block of 5 was presented in one of two random orders. 2.1.4. Coding The transcribed responses were independently coded by two researchers. Responses were coded as Generic, Some, Most, All, or Other. The Generic category included indenite singular generics (e.g., a tiger is striped) but excluded denite singular generics (e.g., the tiger is striped), because the latter is ambiguous between a generic and a specic reading. We also adopted the conservative practice of coding any statements involving adverbs of quantication (e.g., always, mostly, usually, sometimes) as belonging to the semantically closest quantier category (e.g., always as All, mostly and usually as Most, sometimes as Some). This is consistent with standard linguistic treatments of these adverbs (e.g., de Swart, 1993; Heim, 1982; Kratzer, 1995; Lewis, 1975). Any other statements, including a failure to recall anything, were coded as Other. Partial recall of a sentence was coded, as were multiple forms in a given response (e.g., cats sweat through their paws . . .. Most cats sweat through their paws was coded as both Generic and Most). Agreement on the different coding categories ranged from 99% to 100%, and Cohens Kappas ranged from .98 to 1.00. All discrepancies in coding were resolved by discussion. 2.2. Results 2.2.1. Percentage correctly recalled We calculated the percentage of trials that were recalled in the correct format (generic, all, some, or most, depending on the wording of the sentence that had been presented). These scores were entered into a 2 (sentence type: generic, quantied) 3 (condition: generic/all, generic/most, generic/some) ANOVA, with condition as the between-subjects factor and sentence type as the within-subjects factor. We obtained a main effect of sentence type, F(1, 45) = 13.55, p = .001, g2 = .23, indicating that generic sentences were recalled correctly more often than quantiers (67.5% vs. 45%, respectively). We also obtained a main effect of condition, F(2, 45) = 3.56, p < .05, g2 = .14, indicating that recall was higher in the generic/some condition than in either of the other two conditions. (All g2 results that we report use the partial g2 formula (SSeffect/(SSeffect + SSerror)). Tabachnick and Fidell (1989) suggest that partial g2 is an appropriate alternate computation of g2.) However, this main effect must be interpreted in light of the sentence type condition interaction, F(2, 45) = 5.74, p < .01, g2 = .20. For the generic/all and generic/most conditions, generic sentences tended to be recalled more accurately than quantied sentences, p = .10 and p < .001, respectively (see Fig. 1). In contrast, for the generic/some condition, generic sentences and quantied sentences were correctly recalled at equal rates. 2.2.2. Percentage recalled as opposite (generic as quantied and vice versa) We calculated the proportion of trials that were recalled in the opposite form to what was presented (generic sentences recalled as quantied; quantied sentences recalled as generic). These scores were entered into a 2 (sentence type: generic, quantied) 3 (condition: generic/all, generic/most, generic/some) ANOVA, with condition as the between-subjects factor and sentence type as the within-subjects factor. We obtained a main effect of sentence type, F(1, 45) = 9.38, p < .01, g2 = .17, indicating that quantied sentences were more often recalled as generic (38% of trials) than generic sentences were recalled as quantied (18%). We also obtained a main effect of condition, F(2, 45) = 4.03, p < .05, g2 = .16. Overall, participants more often recalled sentences in the opposite form in the generic/all and generic/most conditions than in the generic/some condition. However, this nding must be interpreted in light of
192
Fig. 1. (A) Experiment 1A, mean proportion of generic sentences recalled in correct vs. opposite format. (B) Experiment 1A, mean proportion of quantied sentences recalled in correct vs. opposite format.
the signicant sentence type condition interaction, F(2, 45) = 6.10, p = .005, g2 = .21. This interaction shows that, for the generic/most condition, quantiers are more often recalled as generic than vice versa (59% vs. 7%), p < .001. For the generic/all condition, there was a non-signicant trend towards the quantier being recalled more often as generic than vice versa (39% vs. 24%), p = .20. In contrast, for the generic/some condition, there is no difference between how often quantiers are recalled as generic and generics are recalled as quantied (22% vs. 17%), p > .6. 2.2.3. Correct responses vs. opposite responses We next were interested in the relative frequency of correct vs. opposite responses, to assess the strength of the tendency (predicted by the generics-as-default hypothesis) for adults to recall quanti-
193
ers as generic. We therefore conducted a series of paired t-tests, comparing the percentage of correct responses with the percentage of opposite responses within each condition, as a function of sentence type. The results appear in Fig. 1. For the generic/all condition, generic sentences were recalled correctly more often than in opposite (all) form, t(15) = 2.47, p = .026, whereas quantied sentences were recalled equally often as generic (opposite response) and as all (correct response), t(15) = 0.38, p > .7. For the generic/most condition, generic sentences were recalled correctly more often than they were recalled as most (opposite response), t(15) = 6.45, p < .001, whereas quantied sentences were recalled as generic (opposite) more often than they were recalled as most (correct), t(15) = 2.30, p = .036. Finally, for the generic/some condition, sentences were recalled in correct format more often than in opposite format, regardless of whether the sentence being recalled had been presented in generic or quantied form, ts(15) > 3.1, ps < .01. Altogether, these data reveal an asymmetry between generics and the quantiers all and most, but no asymmetry between generics and the quantier some.
2.2.4. Percentage of errors recalled as opposite The analysis of opposite errors could reect overall accuracy rates. In other words, if participants are more accurate in recalling generics, then the absolute rate of opposite errors will necessarily be relatively low. Therefore, we also examined the opposite errors as a percentage of errors. In other words, we ask whether there are systematic differences in the kinds of recall errors, controlling for differential rates of accuracy. To conduct this analysis, we gave each participant scores for the percentage of errors that showed the opposite pattern (generics recalled as quantied; quantiers recalled as generic), separately for each wording type. Whenever a participant had no errors, their scores were not included in the analysis (remaining Ns were 12 for the generic/all condition, 10 for the generic/ most condition, and 8 for the generic/some condition). We then conducted a 2 (sentence type: generic, quantied) 3 (condition: generic/all, generic/most, generic/some) ANOVA, with condition as the between-subjects factor and sentence type as the within-subjects factor. This analysis revealed no signicant effects for sentence type or condition, although in the generic/most condition, quantied sentences were recalled as generic (59%) nearly twice as often as generic sentences were recalled as quantied (30%), p = .08, one-tailed.
2.3. Discussion As predicted, generics tended to be correctly recalled as generics, whereas quantied statements were frequently recalled as generics. Participants showed little tendency to recall generics as quantied statements, but showed a marked tendency to recall quantied statements as generics. This effect was driven by most and to a lesser extent all; some statements were no more likely to be recalled as generics than generics were to be recalled as some statements. This nding is consistent with Leslie et al.s (2011) ndings: whereas adults sometimes interpreted all statements as generics in their tasks, they showed no tendency to interpret some statements as generics. One outstanding question from Leslie et al. is whether this tendency is specic to the quantier all, or whether it is the quantier some that is exceptional in this regard. Our data show that most statements are recalled as generics, which suggests that the general phenomenon is not limited to the quantier all. Indeed, the tendency for quantiers to be recalled as generic was strongest for most, and weaker (though present) for all. A possible explanation for why some is treated differently is that adults understand some statements to express distinctly limited generalizations. That is, adults may interpret some statements as pertaining only to a limited subset of the kind not as a broad generalization over members of the kind like all and most statements and generics may only be default interpretations for generalizations that pertain broadly to the members of the kind. We return to this possibility in the discussion of Experiment 2A. First, however, we investigated whether adults would recall quantied statements as generics even if they did not hear any generics in the learning phase.
194
3. Experiment 1B In Experiment 1A, each participant heard half of their statements in generic form. To what extent does the tendency to recall quantied statements as generics depend on participants hearing generics? Clearly the results of Experiment 1A cannot be explained by simple confusion between the two types of statements heard, because in the case of all and most, the tendency to recall quantied statements as generics outstripped the tendency to recall generics as quantied statements. However, an even stronger test of the predictions of the generics-as-default hypothesis would be to see if participants recall generics even if they only hear quantied statements in the learning phase. In Experiment 1A, not a single participant recalled a statement as quantied with a quantier other than the one they heard in the learning phase. Errors overwhelmingly consisted of participants either making opposite errors, or failing to remember anything at all. Would this still be the case if they heard no generics, however? We predicted that the tendency to recall quantied statements as generics would be so strong that, even if generics were not included in the learning phase, participants would still produce them in the test phase. Experiment 1B tested this prediction by pairing most with some. We selected these quantiers for two reasons. Most was selected because participants had displayed the strongest tendency to recall most statements as generics. One possible alternative explanation for this might be that most statements are for some reason especially hard to remember, and easily confused with other statements. If this explanation was correct, then one would expect participants often to recall most statements as some statements in this task, so that the rate of opposite errors in this task would be comparable to the rate of opposite errors in the most vs. generics task in 1A. If, however, the generics-as-default interpretation is correct, then there should be fewer opposite errors with most in this experiment than in 1A. The quantier some was chosen because the structure of 1B allowed for a more subtle test of whether adults have any tendency to recall some statements as generics. If there is no such tendency, then some statements here should not be recalled as generics the primary errors with some would be opposite errors. 3.1. Method 3.1.1. Participants Eighteen adults (3 women, 15 men) were recruited from a large Midwestern university. None had participated in the previous study. 3.1.2. Materials The same materials as in Experiment 1A were used. 3.1.3. Procedure The procedure was the same as in Experiment 1A, except that the participants all heard half the statements as some statements, and the other half as most statements. 3.1.4. Coding The same coding practice as in 1A was adopted. Agreement was 100%. 3.2. Results We conducted a series of paired t-tests to determine whether there were signicant differences in recall for sentences presented with most vs. some. On all three measures, there were no signicant differences in recall for most vs. some: percent correct (33% for most; 42% for some), percent recalled as opposite (23% for most recalled as some; 18% for some recalled as most), and percent of errors recalled as opposite (31% of most recalled as some; 23% for some recalled as most), ps > .40.
195
We were interested in whether the errors with most found in Experiment 1A were due to a specic tendency to recall most as generics, or whether most just tends to be recalled as whatever other type of sentence it is paired with. To test this, we compared the rate of opposite errors with most in Experiment 1A (59% of trials) with the rate of opposite errors with most in this experiment (23%). More opposite errors occurred with most in Experiment 1A, t(32) = 3.45, p = .002, suggesting that the high error rate in Experiment 1A is not due to most being somehow unmemorable. Opposite errors with some here occurred at the same low rate as in 1A: 17.5% in 1A and 18% in 1B, t(32) = .04, p > .9. The primary purpose of this experiment was to examine whether quantied sentences are ever recalled as generic. In contrast to Experiment 1A, where errors only entailed forms that had been presented in the original sentence block, here we found that participants consistently recalled quantied forms as generic. Over 60% of participants recalled a quantied sentence as generic as least once (N = 11 out of 18 participants). They did so more often for most sentences (21%) than some sentences (12%), t-paired(17) = 2.12, p < .05. Notably, errors in which the quantied sentences were recalled as generic were as frequent as errors in which the quantied sentences were recalled in the opposite quantied form, for both most and some sentences, ps > .30. Finally, as in Experiment 1A, we examined the relative frequency of correct vs. opposite errors. We predicted that the overwhelming tendency in Experiment 1A to produce opposite errors in the most condition would no longer be found, given that the opposite wording in this case is another quantier rather than a generic. And indeed this was the case: correct responses (33%) and opposite errors (23%) occurred at equal rates, p > .3. Furthermore, as in Experiment 1A, in the some condition, correct responses (42%) were more frequent than opposite errors (18%), p = .039.
3.3. Discussion Participants sometimes recalled statements as generics, despite never hearing a generic in the learning phase. As in 1A, participants never recalled any statement as containing a quantier that they did not hear in the learning phase. It is thus remarkable that they recalled generics, despite only hearing quantied statements in the learning phase. Even though participants did not show a bias towards recalling generics in the case of some statements in 1A, in 1B they recalled some statements as generics on 12% of trials. More specically, some statements were recalled as generics just as frequently as they were recalled as most statements i.e. the rate at which some statements were recalled as generics was equivalent to the rate of opposite errors for some. Further, the majority of participants recalled at least one some statement as a generic. These ndings suggest that even for some statements, there may be some underlying bias towards assimilating them to generics. Statements quantied with most were recalled as generics here more frequently than some statements were recalled as generics, which is as one would expect given the strong tendency to recall most as generics in 1A. Further, the rate of opposite errors for most here was signicantly lower than in 1A. If most had been recalled as some here as frequently as most statements were recalled as generic in 1A, this would have suggested that most statements were just easily confused with other statements, and would have undermined the interest of the ndings in 1A. However, the rate of opposite errors in this study was signicantly lower than in 1A. This nding, coupled with the substantial rate at which most statements were recalled as generics here, reects that these results are due to a sizeable underlying tendency to remember most statements as generics. Experiments 1A and 1B thus show that adults have a distinct tendency to recall statements quantied with all and most as generics, and also have a smaller tendency to recall some statements as generics, as predicted by the generics-as-default hypothesis. In Experiments 2A and 2B, we extended our ndings to 3- and 4-year-old children.
4. Experiment 2A Experiment 2A followed the same design as Experiment 1A, but was conducted with 3- and 4-yearold participants. The generics-as-default hypothesis predicts that, like adults, preschool children
196
should recall quantied statements as generics. The tendency to do so should be at least as strong among young children as it is among adults. Furthermore, we examine potential differences among the quantiers all, most, and some. For adult participants, the tendency to recall some statements as generics is weak compared to all and most statements. Why might this be? One explanation could be that adults interpret some as expressing a very bounded and restricted sort of generalization they understand it to mean that only a few members of the kind have the property. Understanding some this way entails showing a scalar implicature effect (Bott & Noveck, 2004) that is, believing that some implies not all. Developmental evidence can speak to this issue, because very young children do not show a scalar implicature effect (Barner, Chow, & Yang, 2009; Huang & Snedeker, 2009; Hurewitz, Papafragou, Gleitman, & Gelman, 2006; Noveck, 2001; Papafragou & Musolino, 2003). So it is possible that for the youngest children, some expresses an open and unbounded generalization which is more easily assimilated to a generic generalization. Later in development, some comes to be understood as expressing a more bounded and restricted generalization, and in this way is more similar to a specic claim than to a generic generalization. Consistent with this interpretation, Hollander et al.s (2002) and Tardif et al.s (2011) data suggest that English speaking 3-year-olds, but not 4-year-olds, may interpret some statements as generics. If this is part of a more general tendency, then one would likely nd that 3-year-olds recall some statements, as well as all and most statements, as generics. 4.1. Method 4.1.1. Participants Twenty-six 3-year-olds (12 girls, 14 boys; mean age = 3.67 years; range = 3.043.99 years) and thirty 4-year-olds (14 girls, 16 boys; mean age = 4.53 years; range = 4.045.21 years) were recruited from preschools in a small Midwestern city. Sixteen children were assigned to the generic/all condition, 16 to the generic/most condition, and 24 to the generic/some condition. More children were assigned to the generic/some condition so that we could examine more closely the effects of age within this condition. An additional 10 participants (four 4-year-olds and six 3-year-olds) were dropped from the study because they did not recall at least one predicate from each wording condition. 4.1.2. Materials The same materials as in Experiment 1 were used. 4.1.3. Procedure The main procedure was the same as in Experiment 1A, except that the children played with Legos as a distractor task. The main phase was then followed by a post-test, in which each child was asked to respond yes or no to three questions. The questions involved whatever quantier the child had heard in the rst phase. For the rst question, two crayons were place in a box, while four were left outside the box, and the child was asked, Are all/most/some crayons in the box? Then two more crayons were moved into the box, and the child was asked the same question again. Finally, the remaining two crayons were moved into the box, and the child was asked the question a third time. 4.1.4. Coding The same coding practice as in Experiment 1 was used. Agreement across the different coding categories ranged from 95% to 100%, and Cohens Kappas ranged from .885 to 1.00. 4.2. Results 4.2.1. Post-test Children performed extremely well on all, correctly denying that 2/6 or 4/6 are all (100%), and correctly reporting that 6/6 is all (93%). Children also performed well on some, correctly reporting that 2/6 is some (73% at age 3, 100% at age 4) and that 4/6 is some (91% at age 3, 100% at age 4).
197
There was also sharp developmental change in how often children treated some as contrastive with all (i.e., the scalar implicature): 9% of 3-year-olds and 60% of 4-year-olds did so, which is consistent with Barner et al. (2009), Hurewitz et al. (2006), and Pouscoulous, Noveck, Politzer, and Bastide (2007), whose results taken jointly suggest that 4-year-olds sometimes show a weak scalar implicature effect, whereas 3-year-olds tend not to. Finally, children performed poorly on most (only 42% correctly reporting that 2/6 is not most, and only 67% correctly reporting that 4/6 is most), which is consistent with previous work on childrens comprehension of most (Barner et al., 2009; Halberda, Taing, & Lidz, 2008; Papafragou & Schwarz, 2005/2006). Thus results for most should be interpreted cautiously. The majority of children (83%) did not consider most to be contrastive with all (replicating Papafragou & Schwarz, 2005/2006), although the children who did so were both 4-year-olds. For the primary analyses, we rst collapsed across age groups (3-year-olds, 4-year-olds), examining the percent correct recall, the percent opposite recall, and the percent of errors that were recalled as opposite. 4.2.2. Analyses collapsing across age groups 4.2.2.1. Percentage correct recall. We calculated the proportion of trials that were recalled in the correct format (generic, all, some, or most, depending on the wording of the sentence that had been presented). These scores were entered into a 2 (sentence type: generic, quantied) 3 (condition: generic/all, generic/most, generic/some) ANOVA, with condition as the between-subjects factor and sentence type as the within-subjects factor. See Fig. 2 for results. We obtained a main effect of sentence type, F(1, 53) = 17.28, p < .001, g2 = .25. Recall was higher for generic sentences (41%) than for quantied sentences (18%). Although there was no signicant interaction with condition, we conducted post-hoc tests to determine whether the effect of sentence type held up within each condition considered separately (generic/all, generic/most, and generic/some), and found that generics were recalled better than quantiers within each, ps < .05 for generic/all and generic/most conditions; p = .059 for generic/some. 4.2.2.2. Percentage recalled as opposite (generic as quantied and vice versa). We calculated the proportion of trials that were recalled in the opposite form to what was presented (generic sentences recalled as quantied; quantied sentences recalled as generic). These scores were entered into a 2 (sentence type: generic, quantied) 3 (condition: generic/all, generic/most, generic/some) ANOVA, with condition as the between-subjects factor and sentence type as the within-subjects factor. We obtained a main effect of sentence type, F(1, 53) = 20.24, p < .001, g2 = .28, indicating that quantiers were more often recalled as generics (39%) than generics were recalled as quantied (12%). This result held up within each condition examined individually, all ps < .05. 4.2.2.3. Correct responses vs. opposite responses. We next were interested in the relative frequency of correct vs. opposite responses, to assess the strength of the tendency (predicted by the generics-as-default hypothesis) to recall quantiers as generic. We therefore conducted a series of paired t-tests, comparing the percentage of correct responses with the percentage of opposite responses within each condition, as a function of sentence type. For generics, correct responses always exceeded opposite (quantied) errors, all ps < .05. In contrast, for quantiers, correct responses never exceeded opposite (quantied) errors. For both all and most, opposite (generic) errors exceeded correct responses (ps = .01 and .015, respectively). For some, opposite errors were equivalent in frequency to correct responses, p > .2. 4.2.2.4. Percentage of errors recalled as opposite. As in Study 1A, we also examined the opposite errors as a proportion of errors, in order to control for differential rates of accuracy across the different cells of the design. To conduct this analysis, we gave each participant scores for the percentage of errors that showed the opposite pattern (generics recalled as quantied; quantiers recalled as generic), separately for each wording type. Three participants in the generic/most condition had no errors; their scores were not included in the analysis. We then conducted a 2 (sentence type: generic, quantied) 3 (condition: generic/all, generic/most, generic/some) ANOVA, with condition as the between-subjects factor and sentence type as the within-subjects factor.
198
Fig. 2. (A) Experiment 2A, mean proportion of generic sentences recalled in correct vs. opposite format. (B) Experiment 2A, mean proportion of quantied sentences recalled in correct vs. opposite format.
We obtained a main effect of sentence type, F(1, 48) = 17.82, p < .001, g2 = .27, indicating that, even controlling for error frequency, children recalled quantiers as generic more frequently (46% of errors) than they recalled generics as quantied (18% of errors). Post-hoc comparisons indicated that this distinction held up within each of the conditions separately (generic/all, generic/most, generic/some), ps < .05. 4.2.3. Analyses separated by age group In order to examine developmental differences between 3- and 4-year-olds, we next conducted a set of analyses that separates the two age groups. We were particularly interested in how performance
199
compared across ages in the generic/some condition, given the age differences in performance on the post-test with some (with greater evidence for scalar implicatures at age 4 than age 3). We therefore collapsed the data from generic/all and generic/most into a single group. We calculated the proportion of trials that were recalled in the correct format, and entered these scores into a 2 (sentence type: generic, quantied) 2 (condition: generic/all or most, generic/ some) 2 (age: 3 years, 4 years) ANOVA, with condition and age as between-subjects factors and sentence type as the within-subjects factor. Here we report only those effects involving age, as that was the only novel factor. There was a main effect for age, F(1, 52) = 5.35, p = .025, g2 = .09, indicating that 4-year-olds showed better recall than 3-year-olds (36% and 24%, respectively). There was also a trend toward a 3-way interaction involving sentence type, condition, and age group F(1, 52) = 2.82, p = .099, g2 = .05. Post-hoc analyses showed that both 3- and 4-year-olds recalled generics more accurately than the quantiers all and most, p = .065 at age 3; p < .01 at age 4. However, for the generic/some condition, there was a strong age effect: 3-year-olds recalled generics better than some (38% vs. 10%), p < .02, whereas 4-year-olds recalled generics and some at equal rates (43% vs. 40%), p > .70. We next looked at the absolute percentage of trials with opposite errors, again reporting only those effects involving age. We obtained a main effect of age group, F(1, 52) = 5.58, p = .022, g2 = .10, indicating that 4-year-olds made more opposite errors than 3-year-olds (Ms = 30% and 20%, respectively). We also performed planned t-tests to determine if the asymmetry in recall (remembering quantied as generic more than the reverse) held up within each age group and wording condition separately. Results indicated that the asymmetry held up for all and most at both age groups, ps < .02, but that only 3-year-olds showed the asymmetry for some (p = .078); 4-year-olds did not show the asymmetry for some (p > .16). We next examined the relative frequency of correct vs. opposite responses. For all and most, both age groups show the same asymmetry between generics and quantiers: for generics, correct responses exceed opposite errors, ps < .05, and for quantiers, opposite errors exceed correct responses, ps < .05. For some, however, we see a developmental pattern: whereas 3-year-olds also show the asymmetry (for generics, correct responses tended to exceed opposite errors, p < .05, one-tailed, and for some, opposite errors tended to exceed correct responses, p < .05, one-tailed), 4-year-olds show a different pattern (for generics, correct responses show a non-signicant tendency to exceed opposite errors, p = .065, one-tailed, but for some, opposite errors are equal to correct responses, p = 1.0). Finally, we examined opposite errors as a percentage of errors, again reporting only those effects involving age. There was a main effect of age group, F(1, 45) = 5.74, p = .021, g2 = .11, indicating that 4-year-olds made more opposite errors than 3-year-olds. We also examined whether the generic/ quantier difference held up within each age group. For both 3-year-olds and 4-year-olds, quantiers were more often recalled as generic than vice versa, ps < .05.
4.3. Discussion As predicted, preschoolers tend to recall quantied statements as generics. Their recall for generics was much better than their recall for quantiers in both age groups, and quantied statements were more likely to be recalled as generics than vice versa, even controlling for the overall error rates. Preschoolers had difculty correctly recalling quantied statements: they correctly recalled them on only 18% of trials, whereas generics were correctly recalled on 41% of trials. These ndings provide further support for the generics-as-default hypothesis. The post-test data indicate that preschoolers have difculty interpreting most, even when it applies to a specic and limited set of items (i.e., crayons in a box). This replicates ndings by Papafragou and Schwarz (2005/2006) and Barner et al. (2009) (though Halberda et al. (2008) suggests that preschoolers may have some limited understanding of most). Clearly, the preschool-aged participants in our study had difculty interpreting most in the post-test, so our ndings concerning recall for most with this age group should be interpreted cautiously. Importantly, however, we obtain the same pattern of results on the memory task with all, where preschoolers display excellent performance on the post-test. Thus, childrens tendency to recall quantied statements as generics, like that of adults, cannot be due to difculty with the quantier in question.
200
As we predicted, we obtained developmental differences in childrens recall of some statements. 4-year-olds, like adults, were more successful in differentiating some statements from generics than they were in differentiating all and most statements from generics. 3-year-olds, however, tended to recall some statements as generics, just as they recalled all and most statements as generics. These results mirror Hollander et al.s (2002) nding that 3-year-olds, but not 4-year-olds or adults, interpret category-wide some statements as generics. Although more research is needed to establish the basis of this developmental difference, it is interesting to note that our post-test data also show a developmental difference between 3- and 4-yearolds in their comprehension of some. Sixty percent of 4-year-olds in our study showed a scalar implicature effect that is, if all the crayons were in the box, they denied that some crayons were in the box. In contrast, only 9% of 3-year-olds showed a scalar implicature effect. This suggests that 4-year-olds understand some as expressing a more bounded generalization, whereas for 3-yearolds, some may express a broader and less limited sort of generalization. Thus, 3-year-olds may interpret some as more clearly expressing a generalization about the kind at large, whereas 4year-olds may interpret some as expressing a more limited claim about a few particular members of the kind. To the extent that some statements are understood to make claims about the kind in general, they may be more likely to be interpreted as generics. 5. Experiment 2B Experiment 2A successfully replicated Experiment 1A using preschool aged participants. We thus sought to replicate Experiment 1B with similarly aged children. As with adults in 1B, we wondered whether preschoolers would recall quantied statements as generics even if they do not hear any generics in the learning phase. Since performance in the post-test was poor for most, we paired all with some for this task. Although we were primarily interested to see whether children would recall generics even if they do not hear them initially, this study also allowed us to check whether the results with all in 2A were due to some general tendency to confuse all statements with other kinds of statements. That is, we wished to check that the high rate of opposite errors with all in Experiment 2A was not simply due to a tendency to recall all as whatever other statement it is paired with. If all statements just get confused with whatever other statement they are paired with, then children should be as likely to recall all as some in this task as they were to recall all statements as generics in 2A. Comparing the rates of opposite errors across these studies would allow us to address this possible explanation. We were also interested to see whether 4-year-olds would show any tendency here to recall some statements as generics. Although they did not appear to do so in Experiment 2A, they might, as did adults in 1B, display a subtle tendency to do so in this study. 5.1. Method 5.1.1. Participants Sixteen preschoolers (8 girls, 8 boys; mean age = 4.44 years; range = 3.425.06 years) were recruited from the same pool as in Experiment 2A. None of them participated in 2A. An additional 5 children (two 4-year-olds and three 3-year-olds) were tested but were dropped from analysis; 4 of them because they didnt remember at least one predicate from each wording condition, and 1 because she provided all her answers in Spanish. 5.1.2. Materials The same materials as in Experiment 1B were used. 5.1.3. Procedure The same procedure as in Experiment 2B was used, except that the quantiers all and some were used (instead of most and some, because most is difcult for young children to understand). The children received the same post-test as in Experiment 2A.
201
5.1.4. Coding The same coding practice as in the previous studies was used. Agreement on the different coding categories ranged from 93% to 99%, and Cohens Kappas ranged from .82 to .98. 5.2. Results 5.2.1. Post-test Children performed extremely well on all, correctly denying that 2/6 or 4/6 are all (both 92%), and correctly reporting that 6/6 is all (100%). Children also performed well on some, when shown 2 or 4 out of 6, correctly reporting that 2/6 is some (75%) and that 4/6 is some (92%). Children showed mixed performance on some when shown 6/6, with 7 out of 15 children responding that 6/6 is some (47%). (One child had missing data on the post-test.) There was no relationship between age and performance on this post-test, though this may have been due to the small samples and restricted age range. 5.2.2. Recall of quantied statements We conducted a series of paired t-tests to determine whether there were signicant differences in recall for sentences presented with all vs. some. On the percentage of trials recalled correctly, children performed better with some (42%) than with all (16%), t-paired(15) = 2.89, p = .011. The percentage of trials recalled in the opposite form was higher for all (25%) than for some (2%), tpaired(15) = 3.31, p < .01. We also examined the relative frequency of correct vs. opposite errors. In the all condition, correct responses and opposite errors occurred at equal rates, t-paired(15) = 0.96, p > .3. Furthermore, in the some condition, correct responses were more frequent than opposite errors, t-paired(15) = 4.57, p < .001. Finally, the percentage of error trials recalled in the opposite form showed a trend to be higher for all (27%) than for some (9.5%), t-paired(13) = 1.92, p = .077. We also conducted post-hoc tests on accuracy and percentage of opposite errors, to determine if the effects reported above held true for both younger and older preschoolers, by splitting the participants into two equal groups, a younger group (3.424.53, mean age 3.96) and an older group (4.61 5.06, mean age 4.87). (No analyses were conducted on the percentage of errors that were recalled as opposite, as the N of 14 was insufcient to split the group in two for that measure.) We predicted, based on the greater evidence for scalar implicatures among the older children, that they would be more likely to distinguish some from all than the younger children. This was indeed the case. Each of the effects reported above was signicant among the older group (all ps < .01) but non-signicant among the younger group (all ps > .2). In order to see whether results concerning all in Experiment 2A might be due to all just being easily confused with whatever it is paired with, we compared the rates of opposite errors with all in Experiment 2A to the rates of opposite errors with all in this study. Rates of opposite errors with all tended to be lower in this study: 40% of trials were opposite errors in Experiment 2A but only 25% of trials in this study, t(30) = 1.60, p = .06, one-tailed. This comparison suggests that the rate at which all was recalled as generic in Experiment 2A is not likely to be explained by a general tendency to confuse all with other statements. The rate of opposite recall for some was also lower here: 37% in Experiment 2A vs. only 2% in this study, t(38) = 4.11, p < .001. This result further suggests that young childrens errors in recalling quantiers in Experiment 2A were largely due to a tendency to recall quantied statements as generics, rather than a base rate of erroneous recall. The other primary purpose of this experiment was to examine whether quantied sentences are ever recalled as generic, even in the strong case when no generics are supplied during the learning phase. In contrast to Experiment 2A, where the only forms produced for all and some trials were those that had been presented in the original sentence block (i.e., the original quantier, or else generic), here we found that participants consistently recalled quantied forms as generic. Over 60% of participants recalled a quantied sentence as generic at least once (N = 10 out of 16 participants). Notably, errors in which the all sentences were recalled as generic were equivalent to errors in which they were recalled in the opposite quantied form (some), p > .6, and errors in which the some sentences were recalled as generic were higher than errors in which they were recalled in the opposite quantied form (all), p < .05. Although this latter pattern was not signicant when
202
examined within each age subgroup individually, due to reduced statistical power with the smaller Ns, the same pattern obtained for both the younger and older subgroup. 5.3. Discussion As found with the adult participants in Experiment 1B, preschoolers sometimes recalled statements as generics, despite not hearing a single generic in the learning phase. More specically, they recalled all statements as generics as often as they recalled all as some, and they recalled some statements as generics more often than they recalled some as all. Even though every statement heard in the learning phase was quantied, participants were at least as likely to recall them as generics as they were to confuse one of the quantiers with the other. This reects a marked tendency to recall quantied statements as generics. This experiment also conrmed that the high rates of recalling all as generic in Experiment 2A were not due to the quantier all simply being easily confused with whatever it is paired with. The rates of opposite errors in this study for all tended to be lower than when all was paired with generics in Experiment 2A, suggesting again a tendency to recall all statements specically as generics, rather than to just confuse them with other statements more generally. Further, while some was remembered more accurately than all in both this study and in Experiment 2A, the rates of opposite recall for some were again lower in this study than in the latter experiment. As with all, this nding suggests that some is more easily recalled as generic than confused with another quantier, as is further supported by the fact that preschoolers were more likely to recall some as generic than as all in this study, despite never hearing a generic in the learning phase. These patterns held for 4-year-olds as well as 3-year-olds, suggesting that 4-year-olds, like adults, have a mild tendency to recall some statements as generics.
6. Experiment 3 Experiments 2A and 2B nd that 3- and 4-year-olds, like adults, recall quantied statements as generics. These ndings would seem to support the generics-as-default hypothesis. However, there is an alternative explanation available at least in the case of 3-year-olds namely that they are simply forgetting the rst word (i.e., the quantier). Since 4-year-olds, like adults, tended to successfully recall some statements, it is unlikely that they were employing a dumb strategy of simply dropping the rst word. Such a strategy would apply equally well to some statements; thus, one would expect that some statements would also be recalled as generics, if the results were simply driven by this strategy. However, because 3-year-olds did recall some statements as generics, the possibility remains that they were simply forgetting the rst word they heard. Experiment 3 was designed to address this possibility. In Experiment 3, 3-year-olds were presented with (positive) generics as before, but this time they were paired with the negative quantier no. We reasoned that if the 3-year-olds were simply forgetting the rst word, then they should recall no statements as positive generics: for example, no spiders have nine legs would become spiders have nine legs.2 Previous work suggests that this experiment would be a particularly strong test, since at least adults show a bias towards recalling negative statements as afrmative statements, but not vice versa (Cornish & Wason, 1970; Fillenbaum, 1966; Mehler, 1963; but see Hrmann, 1971). Thus, if 3-year-olds did not recall no statements as positive generics, this would constitute clear evidence that they were not simply dropping the rst word. In contrast, if 3-year-olds were recalling all, most and some statements as generics because of a cognitive bias towards interpreting quantied statements as generics, then no statements should either be correctly recalled as no statements, or else they should be recalled as negative generics (e.g., spiders dont have nine legs). If no statements were correctly recalled as no statements, then this would suggest that the previous ndings did not just reect a dumb strategy of
2
We are indebted to Adam Bowen for suggesting this way of testing whether 3-year-olds are simply forgetting the rst word.
203
dropping the rst word, but that the effect is limited to positive quantiers such as all, most, and some, and does not extend to negative quantiers like no. However, if 3-year-olds recalled no statements as negative generics that is, if no spiders have nine legs is recalled as spiders dont have nine legs then this would suggest that the effect is not just limited to positive quantiers. Since recalling such a negative generic involves restructuring the sentence to include a negation, it could not be due to a dumb strategy of dropping the rst word. Rather, it would reect that even negative quantiers such as no may be recalled as generics. Previous empirical (e.g., Leslie et al., 2011) and theoretical work (e.g., Leslie, 2007, 2008) only considered generics as defaults for positive quantiers, so this experiment also represents a potential extension of the thesis to negative quantiers. (Or, to use more precise semantic terminology, previous work has only considered monotone increasing quantiers, and this experiment represents an extension of the thesis to monotone decreasing ones (Barwise & Cooper, 1981))
6.1. Method 6.1.1. Participants Sixteen 3-year-olds (8 girls, 8 boys; mean age = 3.56; range = 3.083.97 years) were recruited from the same pool as in Experiment 2. None had participated in any of the prior studies. An additional 4 children were tested but were dropped from analysis; 3 because they failed to remember at least one predicate for each wording condition, and 1 because she only produced barking sounds.
6.1.2. Materials The same pictures were used as in the previous studies; however, new facts were used for the quantier no, so that the novel facts would remain true. Since the facts used previously were true when associated with positive quantiers, they would become false when associated with negative ones, and thus new facts were needed. A number of positive facts were also replaced, so that the negative and positive facts would be similar (e.g., hippos eat plants vs. no hippos eat meat). The new materials are shown in Table 2.
6.1.3. Procedure The same procedure as in Experiment 2B was used, except that half the items were presented as positive generics, and the other half were presented as no statements.
Table 2 Items used in Experiment 3. Category Bear Shark Rabbit Snail Hippo Skunk Spider Bee Chimp Rhino Buttery Frog Camel Snake Seahorse Cat Generic form Bears can bark Sharks breathe underwater Rabbits keep food in their tummies Snails breathe through their side Hippos eat plants Skunks go out at night Spiders have eight legs Bees have ve eyes Chimps can run fast Rhinos have horns made out of hair Butteries taste with their feet Frogs sleep all winter long Camels store fat in their humps Snakes have tongues Seahorses come out of their dads tummies Cats sweat through their paws No form No No No No No No No No No No No No No No No No bears can sing sharks breathe in air rabbits keep food in their cheeks snails breathe through their mouths hippos eat meat skunks sleep at night spiders have nine legs bees have six eyes chimps can swim rhinos have horns made out of bone butteries taste with their mouths frogs stay awake all winter long camels store water in their humps snakes have ears seahorses come out of their moms tummies cats sweat through their tails
204
6.1.4. Coding Responses were coded as Positive Generic, Negative Generic, No, Positive Quantier, or Other. Statements were coded as Positive Generic if they were in generic form (bare plural or indenite singular) and did not contain any negations. If a statement was in generic form but did contain a negation (e.g., spiders dont have nine legs), then it was coded as Negative Generic. Whether a statement counted as a generic or not was determined by its subject noun phrase (NP). Occasionally, children produced responses that contained the quantier no, but in object position, with the subject NP in generic form, e.g. squirrels have no stinking tails. These were coded as Negative Generics, since the quantier no was attached to the object NP stinking tails. If no modied the subject NP, then the statement was coded as No. If the statement contained a negative adverb of quantication such as never, then it was coded as no also, since such an adverb could modify the subject NP at logical form (Heim, 1982; Lewis, 1975). Responses were coded as Positive Quantier if they contained a positive quantier (all, most some) in the subject, or a positive adverb of quantication (always, mostly, sometimes). All other statements were coded as Other. Transcripts were coded by two coders, with agreement across coding categories ranging from 97% to 100%, and Cohens Kappas ranging from .92 to 1.0. 6.2. Results As in the previous studies, children were more successful at recalling generic statements than quantied statements (see Fig. 3). Specically, children correctly recalled generic sentences (52%) at a higher rate than no-quantied sentences (21%), t-paired(15) = 2.58, p = .021. Of primary interest was the rate at which children recalled no statements as positive generics, because a high rate of doing so would suggest that 3-year-olds are simply dropping the rst word of the sentence. This rarely happened, however: no statements were recalled as positive generics on only 8% of trials, which was equal to the rate at which the (positive) generics were recalled as no statements (also 8%). This pattern held up when controlling for the rate of producing errors: the percentage of errors that were opposite for generic sentences (i.e., recalled as no statements) was equivalent to the percentage of errors that were opposite for no statements (i.e., recalled as positive generic), 15% and 11%, respectively, p > .6.
Fig. 3. Experiment 3, mean proportion of sentences recalled as positive generic (e.g., Bears. . .), no-quantied (e.g., No bears. . .), and negative generic (e.g., Bears dont. . .).
205
Although no statements were rarely recalled as positive generics (8%), they were frequently recalled as negative generics (47% of trials), t-paired(15) = 3.93, p = .001. No statements were correctly recalled as no statements on 21% of trials, and there was a non-signicant trend towards no statements being recalled as negative generics more often than they were correctly recalled as no statements, t-paired(15) = 1.66, p = .12. Further, 11 out of 16 participants recalled no statements as negative generics more often than they recalled no as no, despite not hearing any negative generics in the learning phase. Only 4 participants showed the opposite pattern of recalling no statements correctly more often than recalling them as negative generics. 6.3. Discussion No-quantied statements were rarely recalled as positive generics, which strongly suggests that 3-year-olds are not simply dropping the rst word of the sentence in recall. If the previous results were due to such a strategy, then one would expect that no statements would be frequently recalled as positive generics. Instead, they were recalled as positive generics only to the same (low) extent that generics were recalled as no statements. These rates are naturally interpreted as the background rate at which the 3-year-olds genuinely misremembered negative statements as positive statements, and vice versa (which, interestingly, did not differ depending on whether the original statement was positive or negative, in contrast to, e.g., Cornish & Wason, 1970). Thus, the present results demonstrate that 3-year-olds were not simply forgetting the rst word in the sentence, and thus that 3-year-olds tendency to recall quantied statements as generics in Experiments 2A and 2B cannot be explained away by such a dumb strategy. Further, 3-year-olds have a tendency to recall no statements as negative generics, and they did so more than twice as often as they recalled no statements as no (i.e., in the form they had been originally presented, although this difference did not reach signicance). Tellingly, 69% of participants more often recalled no statements as negative generics rather than as no statements, whereas only 25% showed the reverse tendency. This nding is notable because the participants did not hear any negative generics in the learning phase. This suggests that the tendency to recall quantied sentences as generic is not limited to positive (monotone increasing) quantiers, but extends even to negative (monotone decreasing) quantiers such as no. 7. Experiment 4A Experiments 4A and 4B were designed to test another alternative interpretation of the results. It is possible that participants interpret the quantied statements as quantifying not over individual members of the kind, but rather over subkinds. That is, perhaps participants both preschoolers and adults interpret, for example, all bears climb trees to mean all kinds of bears climb trees (e.g., grizzly bears climb trees, black bears climb trees, polar bears climb trees, and so on). If so, this would mean that participants were interpreting the quantied statements as making generic claims about all/most/ some/none of the subkinds. If they interpret the quantied statements this way, then they should endorse statements with the quantiers all or most as long as the predicate is generically true of all or most of the subkinds. Importantly, if this were the case, then the pattern of results presented in Studies 13 could be accounted for without reference to the hypothesis that generics are cognitively default relative to quantiers. It is unlikely that adults employed such interpretations. Leslie et al. (2011, Experiment 2B) solicited paraphrases of similar quantied statements, and found vanishingly little evidence of subkind-based interpretations. However, to our knowledge no previous work has investigated whether preschoolers may be inclined to interpret quantied statements in this way. Moreover, paraphrases may not provide the most sensitive test of adults interpretations. Therefore, Experiment 4A was designed to directly test whether adults and preschoolers show any tendency to interpret universally quantied statements as pertaining to subkinds rather than individuals. Participants (preschoolers and adults) were presented with item sets that each depicted eight members of a novel animal kind (e.g., gorps). Six of the eight instances had a distinctive feature
206
(e.g., spots) whereas the remaining two did not. Thus the stimuli were designed so that the generic statement (e.g., Gorps have spots) would be likely deemed true, whereas the universal claim (interpreted as pertaining to the individual gorps; e.g., All gorps have spots) would be deemed false. In one condition (the one-kind condition), the animals were all the same color and varied only in the distinctive feature, thus rendering a subkind based interpretation of the universal quantier unlikely. Thus, participants should uniformly reject the universal statement. In the other condition (the subkind condition), however, the animals were divided into two groups of four, differing in color, with three out of four in each group having the relevant feature. Thus a subkind based interpretation of the universal quantier would be available, if participants were inclined to employ it. Crucially, if participants interpreted the universal as quantifying over subkinds, then they should agree that All gorps have spots in the subkind condition, because all subtypes of gorps have spots (even though not every individual does). Thus, if participants are inclined to give subkind interpretations to all, they should endorse all statements more often in the condition where the subkinds are available than in the homogeneous condition where they are not. If, however, participants are not inclined to interpret universals as quantifying over subkinds, then they should treat the two conditions equivalently.
7.1. Method 7.1.1. Participants Thirty-two preschoolers (16 girls, 16 boys; mean age = 4.42; range = 3.515.07) were recruited from the same pool as in Experiments 2 and 3. Four additional children were tested but dropped from analysis, due to a yes answer bias (answering yes to all questions, including warm-up questions). 42 adults (31 female, 11 male; mean age = 18.64; range = 1821) were also recruited from the same pool as in Experiment 1. Sixteen preschoolers and 34 adults were assigned to the one-kind condition and 16 preschoolers and 8 adults were assigned to the subkind condition.
7.1.2. Materials Ten item sets were employed, each consisting of drawings of 8 exemplars of a novel animal kind. In the one-kind condition, the 8 exemplars were identical, except for the target property. In the subkind condition, half the exemplars were one color, and the other half were another color, forming two subkinds; the two subkinds were spatially segregated. Four of the 10 sets were used in the warm-up task: for 2 warm-up sets, all 8 animals had the target property; for the other 2 warm-up sets, none of the 8 animals had the target property. The remaining six sets appeared in the test phase. In all of these sets, 6 out of 8 of the animals had the target property. In the subkind condition, the target property was equally distributed among members of the subkinds (i.e., 3 out of 4 animals in each subkind had the target property). See Fig. 4 for a sample stimulus set.
7.1.3. Procedure Participants were each tested individually. They were shown the picture sets one at a time. For each set, the animals were introduced with a novel label (e.g., These are gorps). Participants were then asked to point to all instances (e.g., Point to the gorps) and were corrected if they missed any. They were then asked a question about the target property, in either generic form (Do gorps have humps?) or universally quantied form (Do all gorps have humps?). Question form was manipulated within-subjects, whereas the one-kind vs. subkind distinction was manipulated betweensubjects. The generic and all wording questions were blocked, and order of blocks was counterbalanced across participants. Thus, each block included two warm-up questions (one in which 0/8 instances depicted the target property; one in which 8/8 instances depicted the target property), followed by three test questions (in which 6/8 instances depicted the target property). Assignment of item set to wording block was counterbalanced across participants.
207
Fig. 4. Sample item sets, Experiment 4 (single subset and distinct subsets).
7.2. Results Performance on the warm-up task was excellent across conditions and question wording (82% correct for preschoolers; 97% correct for adults), indicating basic understanding of the task. Of greater interest was the test phase. Responses to the test questions were entered into a three-way ANOVA, with wording (generic vs. all) as a within-subjects factor, and age group and condition (one-kind vs. subkind) as between-subjects factors. There was a main effect of wording, F(1, 70) = 118.19, p < .001, g2 = .63, a main effect of age, F(1, 70) = 16.64, p < .001, g2 = .19. and a wording age-group interaction, F(1, 70) = 46.31, p < .001, g2 = .40. Post-hoc tests reveal that at both age groups, participants said yes more often in response to the generic question than the all question, ps < .005, indicating that at no age was the quantier all considered simply equivalent to the generic. Furthermore, children and adults did not signicantly differ from one another in their responses to generic questions, endorsing them 78% and 86% of the trials, respectively. In contrast, there was a
208
sharp developmental difference in response to the all questions, with children endorsing them 58% of the trials and adults endorsing them <1% of the trials, p < .001. Importantly, there were no signicant effects involving condition (one-kind vs. subkind). If the results from Experiments 13 could be explained by a tendency to interpret universals as applying to subkinds, then the responses within at least one age group to universal questions should have differed by condition, that is, by whether a subkind interpretation was available or not. The results of this experiment do not provide any support for this hypothesis.
7.3. Discussion If either children or adults showed a tendency to interpret all as quantifying over subkinds, they should have been more likely to respond yes to all questions in the subkind condition than in the one-kind condition. Experiment 4A did not nd any evidence of this tendency. However, children did accept universals at a surprisingly high rate when only 6/8 exemplars had the target property. While this tendency was not found to differ between the one-kind and subkind conditions, it is in itself quite puzzling. An interesting comparison point is childrens very good performance on the all comprehension post-tests in Experiments 2A and 2B, in which children rarely (less than 10% of trials) agreed that all the crayons are in the box when only 2/6 or 4/6 of the crayons were in the box. What might account for this difference? One explanation might be that there was something inherently puzzling in the novel materials used in Experiment 4A, or perhaps the difference between having only six crayons as opposed to eight animals was crucial to childrens performance. The generics-as-default hypothesis, however, offers another possible explanation. Perhaps the difference in performance was due to the fact that children had to consider kind-wide universals in Experiment 4A (all gorps), whereas in the post-tests of Experiments 2A and 2B, the universal quantier occurred in a partitive construction (all of the crayons), and thus only applied to a specic and limited set, namely the six crayons present in the available context. The generics-as-default hypothesis predicts that kind-wide quantied statements will sometimes be interpreted as generics; perhaps the results of Experiment 4A are due to children evaluating the generic instead of the kind-wide universal on approximately half the trials. That is, perhaps the results of Experiment 4A are examples of the same phenomenon documented by Hollander et al. (2002), Leslie et al. (2011), and Tardif et al. (2011). If so, this is an especially dramatic illustration, since the children had counter-examples right before them. A full investigation of the phenomenon would constitute a more extensive set of experiments and so is left for future work. However, as an initial foray, we wished to test whether the high rate of accepting universals was due to the use of the kind-wide construction, or whether there was some feature of the stimuli used that was confusing to the children. To test this, we ran an adaptation of Experiment 4A using the universal partitive construction in place of the kind-wide universal.
8. Experiment 4B 8.1. Method 8.1.1. Participants Sixteen preschoolers (8 girls, 8 boys; mean age = 4.39; range = 3.295.79) and 10 adults (5 female, 5 male; mean age = 19.1; range = 1821) were recruited from the same subject pool as in Experiment 4A. An additional three children were tested, but were dropped from analysis due to a yes bias in two cases, and a no bias in the third (i.e., providing the same response [either yes or no] to every question, including the warm-up questions).
8.1.2. Materials The same materials were used here as in the one-kind condition in Experiment 4A.
209
8.1.3. Procedure The procedure was the same as the one-kind condition of Experiment 4A, with two differences: no generic questions were asked (because generics only apply to entire kinds, not to specic sets (Carlson & Pelletier, 1995)), and the universal questions involved partitive constructions (e.g., Do all of these Gorps have humps?) so that they quantied over just the particular individuals depicted in the drawings. 8.2. Results As in Experiment 4A, performance on the warm-up trials was excellent (80% correct among preschoolers; 95% correct among adults). More importantly, participants also performed extremely well on the test trials. Adults rejected the universal (all of these) on 100% of trials in the test conditions. Crucially, childrens responses to the test questions were signicantly improved from Experiment 4A, with children incorrectly agreeing to the universal when only 6/8 have the target property on only 21% of trials, as compared with 58% of trials in Experiment 4A, t(46) = 2.91, p < .01. 8.3. Discussion Children performed much better when the universal applied only to a specic set (e.g., all of these gorps), as opposed to when the universal applied kind-wide (e.g., all gorps). Children correctly rejected the universal on 79% of trials in Experiment 4B, which is comparable to the 80% of trials on which they answered correctly on the warm-up (e.g., agreeing that all the gorps have wings when 8 out of 8 of them do) in contrast to 42% accuracy on all trials in Experiment 4A. This comparison suggests that the tendency to agree to the universal in 4A was not due to some confusing feature of the novel stimuli, but rather was due to all applying to the entire kind, rather than to only the particular depicted instances. More work would need to be done to be fully certain of the explanation of this nding. However, it should be noted that this is exactly the sort of nding that the generics-as-default hypothesis would predict. The generics-as-default hypothesis predicts that both children and adults will sometimes interpret kind-wide quantied statements as generics (Hollander et al., 2002; Leslie et al., 2011; Tardif et al., 2011). If the children in Experiment 4A were doing this, it would explain the high rate of agreement to the kind-wide universals, because the children were often instead evaluating the generic, which was itself robustly accepted. Experiments 4A and 4B did not nd any evidence that either children or adults are inclined to interpret quantied statements as applying to subkinds rather than individuals, which would have been an alternative explanation of the results of Experiments 13. Instead, Experiments 4A and 4B together turned up what is potentially a further piece of support for the generics-as-default hypothesis, which future work will need to explore in more detail. 9. General discussion Several researchers in psychology and philosophy have recently proposed that generic sentences may articulate cognitively default generalizations (Csibra & Gergely, 2009; Gelman, 2010; Gelman & Brandone, 2010; Hollander, Gelman, & Raman, 2009; Leslie, 2007, 2008; Leslie et al., 2011). In contrast, quantied statements may express cognitively more sophisticated generalizations. The results presented here lend support to this generics-as-default hypothesis, as they indicate that both preschoolers and adults have a tendency to recall quantied statements as generics. In addition to characterizing the results presented here, the generics-as-default hypothesis would explain why Hollander et al. (2002) and Tardif et al. (2011) found that English- and Mandarin speaking preschoolers seem to rely on their evaluation of the corresponding generic when asked to evaluate category-wide quantied statements an effect that Experiment 4 here suggests may be quite powerful indeed, at least in the case of all. It is also consistent with ndings by Leslie et al. (2011), which suggest that adults may similarly rely on their judgments of generics to evaluate certain universally quantied statements. Leslie (2008) also argues that such a hypothesis would explain why no known
210
language has a word (e.g., gen) that signals a generic statement in the way that a quantier such as all signals a universal statement. Certain adult reasoning errors would also be explicable on this hypothesis, for example Jnsson and Hamptons (2006) nding that adults judge all ravens are black to be more likely to be true than all young jungle ravens are black. From a logical point of view, this is an error, because the young jungle ravens form a subset of the ravens, and so if all ravens are black is true, then all young jungle ravens are black must also be true. However, if adults were interpreting these universal statements as generics, then their judgments would be appropriate: ravens are black could be true even if young jungle ravens werent black just as tigers are striped is true even though albino tigers arent striped (see Connolly, Fodor, Gleitman, & Gleitman, 2007; see also Cimpian, Gelman, et al., 2010). Thus Jnsson and Hamptons ndings would be easily explained if category-wide universal statements tended to be interpreted as generics by adults. A similar explanation may also apply to Slomans (1993, 1998) ndings concerning inferences about categories. 9.1. Is it just that generics are shorter? One possible alternative explanation for the ndings reported here could be that people simply have a general tendency to recall sentences in shortened form. That is, perhaps generics are favored in recall because they are shorter by one word than quantied sentences. This explanation seems implausible for a number of reasons. First, it is not clear how such an explanation could explain why all and most but not some were recalled as generics by adults and 4-year-olds, because in all three cases the generic is shorter than the quantied statement. Further, Experiment 3 indicates that even 3-year-olds were not simply dropping the quantier, since they tended to recall no statements as negative generics, despite the fact that negative generics are not inherently shorter than no statements (as no is replaced by dont here). Nonetheless, to further examine this question, we looked to see whether participants displayed a general tendency to shorten the lengths of the predicates in recall. If generic recall was simply due to an overall tendency to shorten, then one would expect that predicates would also tend to be shortened in recall. We did not nd consistent evidence of this, however; in most cases, participants produced predicates of equal length to those used in the learning phase. Further, although participants in Experiments 1A and 2A were more likely to shorten rather than lengthen a predicate, participants in Experiments 1B and 2B did not show such a tendency. The breakdown from these four studies is as follows. Study 1A: Adults kept predicates the same length on 73% of trials, shortened predicates on 17% of trials, and lengthened on 10% of trials (shorter > longer, p < .01). Study 1B: Adults kept predicates the same length on 74% of trials, shortened predicates on 14% of trials, and lengthened on 12% (shorter = longer, p > .4). Study 2A: Children kept predicates the same length on 43% of trials, shortened predicates on 35% of trials, and lengthened on 22% (shorter > longer, p < .05). Study 2B: Children kept predicates the same length on 52% of trials, shortened predicates on 22% of trials, and lengthened on 26% (shorter = longer, p > .5). Thus we did not nd any consistent tendency to shorten predicates in recall, as would be expected if our data were to be explained by a general tendency to shorten. A further question is whether the tendency to recall a sentence as a generic is affected by its length. To examine this, we analyzed the likelihood of recalling a sentence as a generic as a function of its length (coded as short (2 or 3 words), medium (4 words), or long (5 or 6 words)). Again, this analysis did not reveal a consistent pattern: although adults were more likely to recall longer sentences as generics, preschoolers were not. (The rates of recall as generic was as follows: for Study 1A (adults), 44% for short sentences, 56% for medium sentences, and 60% for long sentences; for Study 1B (adults), 7% for short sentences, 11% for medium sentences, and 22% for long sentences; for Study 2A (preschoolers), 39% for short sentences, 41% for medium sentences, and 41% for long sentences; for Study 2B (preschoolers), 27% for short sentences, 25% for medium sentences, and 30% for long sentences.) Again, the fact that there was not a consistent pattern of generic recall as a function of sentence length argues against the idea that our results simply depend on a tendency to shorten sentences in recall. Given that preschoolers rate of generic recall was not affected by sentence length, this is unlikely to be the explanation of the results. It is interesting, however, that adults showed a somewhat greater tendency to recall longer sentences as generic. One possible explanation for the nding is con-
211
sistent with the generics-as-defaults hypothesis, namely, that this tendency may be due to longer sentences imposing somewhat higher processing demands. That is, if processing and remembering a longer sentence imposes higher cognitive demands, then this may lead to an increased tendency to default to the generic. Some indirect support for this possibility is found in Meyer, Gelman, and Stilwell (2011), who found that adults were more likely to evaluate universals as generics when time constraints were introduced. Further research is needed to address this question; in particular it will be important to examine whether increased cognitive load inuences adults tendency to recall quantied statements as generics. 9.2. Developmental trajectories in the recall of quantied statements Adults showed a denite overall tendency to recall quantied statements as generics. However, in the case of some and to some extent all, the ndings presented here reect an interesting developmental trajectory: younger children are more inclined than adults to recall these quantied statements as generics. This result is what the generics-as-default hypothesis would predict: the youngest children are the most susceptible to defaulting to the generic interpretation. This was evident for all: adults were equally likely to recall all as generic as they were to recall it correctly, whereas preschoolers were more likely to recall all as generic. In the case of some, the developmental trajectory was quite dramatic, with 3-year-olds showing a strong tendency to recall some as generic, whereas 4-year-olds and adults showed very little tendency to do so. It is possible that this developmental change with some is connected to a developing appreciation for scalar implicature, meaning that 3-year-olds may interpret some as indicating more open-ended and potentially kind-wide generalizations than 4-year-olds and adults do. This is only a tentative interpretation, however, and further investigation into 3- and 4-year-olds interpretations of some is needed before one could be condent of such an explanation. One way to test the hypothesis would be to examine whether individual performance with some on the memory task corresponds to appreciation of scalar implicature. In the current study, scalar implicature appreciation was assessed based on a single trial of the post-test, and so we did not have sufcient power to examine correlations at the individual level. However this is a question that will be important to assess in future work. The fact that developmental changes occurred with recall of all and some suggests that there may also be a more general developmental trajectory toward increasingly suppressing the tendency to default to the generic in recall. In light of this, it is interesting to note how poorly even adults performed in recalling most. A possible explanation of this may be that most is simply a more difcult quantier to process than all and some, as previous researchers have suggested (e.g., Barner et al., 2009; Barwise & Cooper, 1981; Papafragou & Schwarz, 2005/2006). 9.3. Conclusion If the generics-as-default hypothesis is indeed correct, there are a number of implications for understanding human cognition. The way we most naturally generalize information from individuals to kinds is of central important in cognitive psychology (Prasada, 2000; Smith, 1989). The generics-asdefault hypothesis claims that our most basic and automatic way of generalizing such information involves generic generalizations, rather than quanticational ones. Although the studies reported here did not focus on the exact nature of generic generalizations, a number of recent studies suggest that generic generalizations, unlike quanticational generalizations, do not depend solely on statistical information such as prevalence (Cimpian, Brandone, et al., 2010; Cimpian, Gelman, et al., 2010; Khemlani et al., 2007, 2009, in press; Leslie, 2007). Although quantied statements such as most lorches have feathers are accepted or rejected depending on the percentage of lorches that have feathers, adults acceptance/rejection of generics such as lorches have feathers depends on more complex factors, such as whether the feathers are potentially dangerous (Cimpian, Brandone, et al., 2010; Leslie, 2007, 2008). Similarly, adults are more likely to endorse generics such as dondrets have tails if having a tail is a characteristic or essential property of dondrets that emerges with maturation, than if only infant dondrets have tails even if there are more infant dondrets, and therefore more dondrets
212
with tails. Quantiers such as most do not follow such patterns, but rather depend only on whether the majority of dondrets have tails (Cimpian, Gelman, et al., 2010). If generic generalizations are indeed cognitively more basic than quanticational ones, then this would mean that our fundamental way of generalizing information is sensitive to more than just statistical distributions of properties. Rather, our basic general judgments would be sensitive to factors such as whether the property to be generalized is potentially dangerous (Cimpian, Brandone, et al., 2010; Leslie, 2007, 2008) or whether it is an essential property that emerges with development (Cimpian, Gelman, et al., 2010; Gelman, 2003). Such a hypothesis, if correct, would t well with theorybased accounts of cognition (e.g., Carey, 2009; Gelman, 2003; Gopnik & Meltzoff, 1997; Gopnik & Wellman, 1994), according to which general knowledge about kinds is arranged according to causal structures, and is driven by more than simple statistical distributions of properties. Acknowledgments This research was supported in part by research funds from NICHD Grant HD-36043 to Gelman. We are grateful to the parents, teachers, and especially children who participated in this research, including the staff at the following schools: Pooh Corner, Generations Together, the JCC Child Care Center, Morning Star, and Annies Child Care Center-West. We thank Erika Manczak, Sarah Stilwell, Amanda Markowitz, Stacey McGregor, Becca Kanine, and Jaime Bortnick for their assistance in collecting and entering the data, and Adam Bowen for suggesting Experiment 3. References
Barner, D., Chow, K., & Yang, S. (2009). Finding ones meaning: A test of the relation between quantiers and integers in language development. Cognitive Psychology, 58, 195219. Barwise, J., & Cooper, R. (1981). Generalized quantiers and natural language. Linguistics and Philosophy, 4, 159219. Bott, L., & Noveck, I. A. (2004). Some utterances are underinformative: The onset and time course of scalar inferences. Journal of Memory and Language, 51, 437457. Bower, G. H., Black, J. B., & Turner, T. J. (1979). Scripts in memory for text. Cognitive Psychology, 11, 177220. Carey, S. (2009). The origin of concepts. New York: Oxford University Press. Carlson, G. (1977). Reference to kinds in English. Ph.D. dissertation, University of Massachusetts, Amherst. Carlson, G., & Pelletier, F. J. (1995). The generic book. Chicago: University of Chicago Press. Chomsky, N. (2000). New horizons in the study of language and mind. Cambridge: Cambridge University Press. Cimpian, A., Brandone, A. C., & Gelman, S. A. (2010). Generic statements require little evidence for acceptance but have powerful implications. Cognitive Science, 34, 14521482. Cimpian, A., Gelman, S. A., & Brandone, A. C. (2010). Theory-based considerations inuence the interpretation of generic sentences. Language and Cognitive Processes, 25, 261276. Connolly, A., Fodor, J., Gleitman, L. R., & Gleitman, H. (2007). Why stereotypes dont even make good defaults. Cognition, 103, 122. Cornish, E. R., & Wason, P. C. (1970). The recall of afrmative and negative sentences in an incidental learning task. Quarterly Journal of Experimental Psychology, 22, 109114. Csibra, G., & Gergely, G. (2009). Natural pedagogy. Trends in Cognitive Sciences, 13, 148153. Dahl, O. (1985). Tense and aspect systems. Oxford: Blackwell. de Swart, H. (1993). Adverbs of quantication: A generalized quantier approach. New York and London: Garland. Ehrlich, M. F. (1975). Cumulative learning and long term retention of sentences. Acta Psychologica, 39, 241250. Fillenbaum, S. (1966). Memory for gist: Some relevant variables. Language and Speech, 9, 217227. Gelman, S. A. (2003). The essential child. New York: Oxford University Press. Gelman, S. A. (2010). Generics as a window onto young childrens concepts. In F. J. Pelletier (Ed.). Kinds, things, and stuff: The cognitive side of generics and mass terms. New directions in cognitive science (Vol. 12). New York: Oxford University Press. Gelman, S. A., & Brandone, A. C. (2010). Fast-mapping placeholders: Using words to talk about kinds. Language Learning and Development, 6, 223240. Gelman, S. A., Goetz, P. J., Sarnecka, B. S., & Flukes, J. (2008). Generic language in parentchild conversations. Language Learning and Development, 4, 131. Gelman, S. A., Hollander, M., Star, J., & Heyman, G. D. (2000). The role of language in the construction of kinds. In D. Medin (Ed.). Psychology of learning and motivation (Vol. 39, pp. 201263). New York: Academic Press. Gelman, S. A., & Raman, L. (2003). Preschool children use linguistic form class and pragmatic cues to interpret generics. Child Development, 74(1), 308325. Gelman, S. A., & Raman, L. (2007). This cat has nine lives? Childrens memory for genericity in language. Developmental Psychology, 43(5), 12561268. Gopnik, A., & Meltzoff, A. N. (1997). Words, thoughts, and theories. Cambridge, MA: Bradford Books/MIT Press. Gopnik, A., & Wellman, H. (1994). The theory theory. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain specicity in cognition and culture. NY: Cambridge University Press.
213
Graham, S. A., Nayer, S. L., & Gelman, S. A. (2011). Two-year-olds use the generic/non-generic distinction to guide their inferences about novel kinds. Child Development, 82, 493507. Halberda, J., Taing, L., & Lidz, J. (2008). The development of most comprehension and its potential dependence on countingability in preschoolers. Language Learning and Development, 4, 99121. Heim, I. (1982). The semantics of denite and indenite noun phrases. Ph.D. dissertation, University of Massachusetts at Amherst. Holyoak, K. J., & Morrison, R. G. (Eds.). (2005). The Cambridge handbook of thinking and reasoning. Cambridge, UK: Cambridge University Press. Hollander, M. A., Gelman, S. A., & Raman, L. (2009). Generic language and judgments about category membership: Can generics highlight properties as central? Language and Cognitive Processes, 24, 481505. Hollander, M. A., Gelman, S. A., & Star, J. (2002). Childrens interpretation of generic noun phrases. Developmental Psychology, 38, 883894. Hrmann, H. (1971). Semantic factors in negation. Psychologische Forschung, 35, 116. Huang, Y., & Snedeker, J. (2009). Semantic meaning and pragmatic interpretation in 5-year-olds: Evidence from real-time spoken language comprehension. Developmental Psychology, 45, 17231739. Hurewitz, F., Papafragou, A., Gleitman, L., & Gelman, R. (2006). Asymmetries in the acquisition of numbers and quantiers. Language Learning and Development, 2, 7796. Jacoby, L. L., & Dallas, M. (1981). On the relationship between autobiographical memory and perceptual learning. Journal of Experimental Psychology: General, 110, 306340. Jnsson, M. L., & Hampton, J. A. (2006). The inverse conjunction fallacy. Journal of Memory and Language, 55, 317334. Keil, F. C. (1989). Concepts, kinds, and cognitive development. Cambridge, MA: MIT Press. Khemlani, S., Leslie, S. J., Glucksberg, S., & Rubio-Fernandez, P. (2007). Do ducks lay eggs? How humans interpret generic assertions. In Proceedings of the 29th annual conference of the cognitive science society. Nashville, TN: Cognitive Science Society. Khemlani, S., Leslie, S. J., & Glucksberg, S. (2009). Generics, prevalence, and default inferences. In Proceedings of the 31st annual conference of the cognitive science society. Austin, TX: Cognitive Science Society. Khemlani, S., Leslie, S. J., & Glucksberg, S. (in press). Inferences about members of kinds: The generics hypothesis. Language and Cognitive Processes. Kintsch, W. (1988). The role of knowledge in discourse comprehension: A constructionintegration model. Psychological Review, 95, 163182. Kratzer, A. (1995). Stage-level and individual-level predicates. In G. Carlson & J. Pelletier (Eds.), The generic book (pp. 125175). Chicago: Chicago University Press. Krifka, M., Pelletier, F. J., Carlson, G., ter Meulen, A., Chierchia, G., & Link, G. (1995). Genericity: An introduction. In G. Carlson & F. J. Pelletier (Eds.), The generic book (pp. 1125). Chicago: Chicago University Press. Leslie, S. J. (2007). Generics and the structure of the mind. Philosophical Perspectives, 21, 375405. Leslie, S. J. (2008). Generics: Cognition and acquisition. The Philosophical Review, 117, 149. Leslie, S. J., Khemlani, S., & Glucksberg, S. (2011). All ducks lay eggs? The generic overgeneralization effect. Journal of Memory and Language, 65, 1531. Lewis, D. (1975). Adverbs of quantication. In Formal semantics of natural language (pp. 315). Cambridge: Cambridge University Press. Macrae, C. N., Hewstone, M., & Grifths, R. J. (1993). Processing load and memory for stereotype-based information. European Journal of Social Psychology, 23, 7787. Margolis, E., & Laurence, S. (1999). Concepts: Core readings. Cambridge, MA: MIT Press. McCarthy, R. A. (1995). Semantic knowledge and semantic representations. Hove, UK: Taylor & Francis. Mehler, J. (1963). Some effects of grammatical transformations on the recall of English sentences. Journal of Verbal Learning and Verbal Behavior, 2, 346356. Meyer, D. E. (1970). On the representation and retrieval of stored semantic information. Cognitive Psychology, 1, 242299. Meyer, M., Gelman, S. A., & Stilwell, S. M. (2011). Generics are a cognitive default: Evidence from sentence processing. In Proceedings of the 33rd annual conference of the cognitive science society. Boston, MA: Cognitive Science Society. Murphy, G. L. (2002). The big book of concepts. Cambridge, MA: MIT Press. Noveck, I. (2001). When children are more logical than adults: Experimental investigations of scalar implicatures. Cognition, 78, 165188. Papafragou, A., & Schwarz, N. (2005/2006). Most wanted. Language Acquisition, 13, 207251. Papafragou, A., & Musolino, J. (2003). Scalar implicatures: Experiments at the semanticspragmatics interface. Cognition, 86, 253282. Pouscoulous, N., Noveck, I., Politzer, G., & Bastide, A. (2007). Processing costs and implicature development. Language Acquisition, 14, 347375. Prasada, S. (2000). Acquiring generic knowledge. Trends in Cognitive Science, 4, 6672. Prasada, S., & Dillingham, E. (2006). Principled and statistical connections in common sense conception. Cognition, 99, 73112. Prasada, S., & Dillingham, E. (2009). Representation of principled connections: A window onto the formal aspect of common sense conception. Cognitive Science, 33, 401448. Rosch, E. H. (1973). Natural categories. Cognitive Psychology, 4, 328350. Sloman, S. A. (1993). Feature-based induction. Cognitive Psychology, 25, 231280. Sloman, S. A. (1998). Categorical inference is not a tree: The myth of inheritance hierarchies. Cognitive Psychology, 35, 133. Smith, E. E. (1989). Concepts and induction. In M. I. Posner (Ed.), Foundations of cognitive science (pp. 501526). Cambridge, MA: MIT Press. Smith, E. E., & Medin, D. L. (1981). Categories and concepts. Cambridge, MA: Harvard University Press. Smith, E. E., Shoben, E. J., & Rips, L. J. (1974). Structure and process in semantic memory: A featural model for semantic decisions. Psychological Review, 1, 214241. Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariate statistics. Philadelphia, PA: Harper & Row.
214
Tardif, T., Gelman, S. A., Fu, X., & Zhu, L. (2011). Acquisition of generic noun phrases in Chinese: Learning about lions without an -s. Journal of Child Language, 30, 132. Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory (pp. 381402). New York: Academic Press.
Child Development, March April 2012, Volume 83, Number 2, Pages 423433
Do Lions Have Manes? For Children, Generics Are About Kinds Rather Than Quantities
Amanda C. Brandone
Lehigh University
Andrei Cimpian
University of Illinois at Urbana-Champaign
Sarah-Jane Leslie
Princeton University
Susan A. Gelman
University of Michigan
Generic statements (e.g., Lions have manes) make claims about kinds (e.g., lions as a category) and, for adults, are distinct from quanticational statements (e.g., Most lions have manes), which make claims about how many individuals have a given property. This article examined whether young children also understand that generics do not depend purely on quantitative information. Five-year-olds (n = 36) evaluated pairs of questions expressing properties that were matched in prevalence but varied in whether adults accept them as generically true (e.g., Do lions have manes? [true] vs. Are lions boys? [false]). Results demonstrated that children evaluate generics based on more than just quantitative information. Data suggest that even young children recognize that generics make claims about kinds.
A central task of early childhood is to learn broad generalizations about categories in the world. Children must learn that lions have manes, milk builds strong bones, and stoves are hot. Generic sentences such as Lions have manes provide an important means of conveying such knowledge (Gelman, 2003). Generics make claims about kinds (e.g., lions as a category) rather than individuals (e.g., the lions at the zoo) and express generalizations about shared properties of category members (Carlson, 1977; Carlson & Pelletier, 1995). Generics also link to core properties, as they express predicates that are relatively enduring (not transient), timeless (not contextually bound), and inherent (not accidental) (Carlson & Pelletier, 1995; Prasada, 2000; see also Prasada & Dillingham, 2006). Generics appear frequently in natural speech, including conversation with young children (e.g., Gelman, Goetz, Sarnecka,
This research was supported by an NSF Graduate Research Fellowship to Amanda C. Brandone, University of Illinois funds to Andrei Cimpian, Princeton University funds to Sarah-Jane Leslie, and NICHD Grant HD-36043 and NSF Grant BCS0817128 to Susan A. Gelman. We are grateful to the children, parents, and teachers at the UM Childrens Center for participating in this research. We also thank Kristin Lang, Jenna Hedglen, Ben Boldt, Melissa Yako, and the Concepts and Theories in Human Development and UIUC Cognitive Development Labs for assistance in data collection. Correspondence concerning this article should be addressed to Amanda Brandone, Department of Psychology, Lehigh University, Bethlehem, PA 18015. Electronic mail may be sent to acb210@lehigh.edu.
& Flukes, 2008), and they have been hypothesized to play a central role in human reasoning (Gelman, 2003; Prasada, 2000). Nevertheless, an issue that remains unresolved concerns how the meaning of these utterances is represented. Within the linguistic and philosophical literatures, there has been considerable debate over how to characterize the semantics of generics (for a review, see Leslie, 2008). Leslie (2007, 2008) has argued that the truth of a generic statement does not depend on how many kind members possess the property in question; there is no straightforward mapping between statistical facts about the world and judgments about generics (see also Carlson, 1977; Carlson & Pelletier, 1995; cf. Prasada & Dillingham, 2006). Consider these examples. Lions have manes is true, but Lions are male is false, even though only male lions ever possess manes. Mosquitoes carry the West Nile Virus is true, even though less than 1% of mosquitoes are actually infected. Conversely, People are right-handed is false, despite the fact that the vast majority of individuals are right-handed. Examples such as these demonstrate that when deciding on the truth value
2012 The Authors Child Development 2012 Society for Research in Child Development, Inc. All rights reserved. 0009-3920/2012/8302-0004 DOI: 10.1111/j.1467-8624.2011.01708.x
424
Brandone, Cimpian, Leslie, and Gelman
of a generic, people rely on more than just statistical information. Recent empirical work conrms that, for adults, generic meaning does not reduce to quantication that is, to considerations of how many members of the kind have the property. Two main ndings have emerged from this research. First, adults acceptance of generics is not accounted for by the prevalence of the predicated properties alone. (It is only in certain circumscribed cases that statistical frequency is sufcient to make a generic true; Prasada & Dillingham, 2006, 2009.) For example, Khemlani and colleagues (Khemlani, Leslie, & Glucksberg, 2009; Khemlani, Leslie, Glucksberg, & Rubio-Fernandez, 2007) showed that adults judged some generics such as Ducks lay eggs and Mosquitoes carry the West Nile virus to be true despite knowing that a large percentage of the kinds lacks the predicated properties; however, they rejected other generics such as Books are paperbacks and Canadians are right-handed despite knowing that most members of those kinds have the predicated properties. Furthermore, Cimpian, Brandone, and Gelman (2010) found that adults were willing to accept even some novel generics (e.g., Lorches have purple feathers) at low prevalence levels (below 50%). Second, research has shown that adults acceptance of generics reects biases in their conceptual representations. For example, Cimpian, Brandone, et al. (2010) found that generic predications of properties that were described as distinctive or dangerous (e.g., venomous scales) were accepted more often than generic predications of other similar properties (e.g., shiny scales), presumably because properties that are either distinctive or dangerous to humans are privileged in our kind representations (see also Leslie, 2007, 2008). Similarly, Cimpian, Gelman, and Brandone (2010) and Gelman and Bloom (2007) showed that considerations about the nature and origin of the predicated properties (e.g., whether they are innate vs. acquired) are sufcient, independently of prevalence, to determine how adults reason about generic sentences. Together, these data suggest that, at least for adults, generics do not reduce to quanticational statements. An important remaining question is whether this is also the case for children: Do children understand generic meaning as distinct from quantication? This question has important implications for broader theoretical questions regarding the nature of childrens concepts. Some have argued that as young children build their lexical and conceptual repertoires, they are guided by
abstract conceptual knowledge (e.g., ontology, animacy, function), including a conceptual distinction between individuals and kinds (e.g., Carey, 1985, 2009; Gelman, 2003; Gelman & Brandone, 2010; Gopnik & Meltzoff, 1997; Wellman & Gelman, 1998). However, others have proposed a more bottom-up process of lexical and conceptual development in which perceptual features of the environment get linked to concrete linguistic features of the input via associative learning (e.g., Colunga & Smith, 2005; Sloutsky & Fisher, 2004; Sloutsky, Kloos, & Fisher, 2007; Smith, Jones, & Landau, 1996). These theoretical perspectives offer distinct proposals for how children might interpret generics. If childrens concepts go beyond the available evidence and incorporate abstract knowledge, then children may be able to interpret generics as being about abstract kinds and as distinct from statements about how many members of the kind have the relevant property. If, on the other hand, childrens cognition is limited to statistical computations and associative learning, then children may instead interpret generics as synonymous with quantied statements. Much of the recent evidence regarding childrens production and comprehension of generics is suggestive of the possibility that generics are not equivalent to statements about frequency even for young children. Children as young as 2 years of age produce generics (Gelman, Goetz, et al., 2008; Gelman, Waxman, & Kleinberg, 2008) and do so more often when the context encourages a focus on kinds (Brandone & Gelman, 2009; Gelman, Goetz, et al., 2008; Gelman, Waxman, et al., 2008). For example, preschoolers produce more generics when talking about animals than artifacts (Brandone & Gelman, 2009; Gelman, Goetz, et al., 2008; Gelman, Waxman, et al., 2008)consistent with evidence suggesting that animal categories are more richly structured and kind-like than artifact categories (e.g., Keil, 1989). These data suggest that children understand generics to be making claims about kinds. However, production data are ambiguous: When children produce a generic (e.g., Birds y), we cannot be certain whether they are referring to the kind or to a quantied set (e.g., all or most birds). So far, the evidence from comprehension studies also cannot denitively rule out the possibility that children understand generics as quanticational. Preschoolers seem to recognize that generics (a) refer beyond individuals in the immediate context (Cimpian & Markman, 2008; Cimpian, Meltzer, & Markman, 2011; Gelman & Bloom, 2007; Gelman & Raman, 2003), (b) are broad in scope but allow for
Do Lions Have Manes?
425
exceptions (Chambers, Graham, & Turner, 2008; Gelman & Bloom, 2007), and, more specically, (c) imply scope that is intermediate between all and some (Hollander, Gelman, & Star, 2002). Although these studies demonstrate sophisticated understanding of generics in preschoolers, their results are still compatible with the possibility that children interpret generics as equivalent to quanticational statements about individuals (e.g., perhaps as equivalent to most statements). Thus, existing data provide suggestive, but not conclusive, evidence that children do not interpret generics as making purely quanticational claims. In the current experiments, we examine this question directly. Specically, we ask: Do young children interpret generics as reecting only statistical facts about the world? Or, as in the case of adults, do judgments about the truth of generics depend on additional information?
Study 1 We compared childrens judgments about pairs of statements that, although matched in prevalence, vary in their acceptability as generics. Consider, for example, Lions have manes versus Lions are male. Since only male lions have manes, Lions have manes applies to at most as many lions as Lions are male. Yet, despite the similar underlying statistics of these two statements, only Lions have manes seems acceptable, arguably because having manes is a characteristic property of the kind lion, whereas being male is not. The question of what makes a property characteristic of a kind is an exceedingly complex one (see Leslie, 2007, 2008; Prasada & Dillingham, 2006, 2009). For current purposes, a property was deemed characteristic if it involved (a) a salient physical property (e.g., Lions have manes), (b) a means of gestation (e.g., Birds lay eggs), or (c) a means of nurturing the young (e.g., Pigs give milk to their babies). A property was deemed noncharacteristic if it involved attributing a gender to the kind (e.g., Lions are male). Theoretically, this does not exhaust all the possibilities for characteristic versus noncharacteristic properties; nevertheless, these were the guidelines selected to operationalize characteristic and noncharacteristic properties in the current study. The previous example illustrates the research strategy used in this study: We compared the likelihood that participants would accept as true each of two kinds of statements presented in generic form: (a) those for which predicates are characteristic of the
kind yet true of only one gender (e.g., Lions have manes) and (b) those for which predicates are not characteristic of the kind yet also true of only one gender (e.g., Lions are boys). (Note that because of the age of our participants, we substituted the terms boys and girls for males and females. Pilot testing conrmed that children understood these terms as referring to the gender of the animals and did not interpret them to mean human boys and human girls.) Pairs of characteristic and noncharacteristic statements (e.g., Lions have manes vs. Lions are boys) were chosen such that, within each pair, the property predicated in the noncharacteristic statement (e.g., being male) served as a precondition for the property predicated in the characteristic statement (e.g., having a mane) and thus applied to a superset of the animals to which the characteristic property applied. To ensure that participants were aware of these inclusion relations, we tested their knowledge of the gender requirements for the properties predicated in the characteristic statements (e.g., Do boy lions have manes? Do girl lions have manes?). Only items on which participants demonstrated requisite knowledge (e.g., that only boy lions have manes) were analyzed. By equating the prevalence of the information expressed in the two types of statements and varying whether or not the predicated property was characteristic of the kind, we could distinguish responses driven by prevalence information from responses driven by knowledge about kinds. If children interpret generics as reecting purely statistical facts about the world, they should be equally likely to accept the characteristic and noncharacteristic statements because the properties predicated in each are roughly equivalent in prevalence. If, on the other hand, children interpret generics as claims about kinds like adults do, then they should be more likely to accept the characteristic than the noncharacteristic statements. We predicted that childrens understanding of generics would not be driven solely by statistical facts about the world; thus, acceptance of the characteristic items should be greater than acceptance of the noncharacteristic items.
Method Participants Fourteen children (7 males, 7 females; M = 5.36 years, SD = .52) and 14 undergraduates (5 males, 9 females) participated. Two additional children were excluded because they did not know any
426

Table 1 Study 1: Part 1 Items Item type Characteristic Item Do birds lay eggs?a Do pigs give milk to their babies?a Do horses grow their babies in their tummies? Do lions have manes?a Do deer have antlers? Do goats have horns?a Are birds girls?a Are pigs girls?a Are horses girls? Are lions boys?a Are deer boys? Are goats boys?a Do tigers have stripes? Do cats have tails?a Do kangaroos hop? Do cows say moo?a Do dogs have wings? Do snakes have feet?a Do hamsters y Do sh say woof?a
of the gender properties. An additional 28 undergraduates participated in pretesting (see below). Children were recruited from schools in a Midwestern city. Undergraduates were recruited from the subject pools at two large public universities. Participants were predominantly European American and from middle-income homes. Materials and Procedure Part 1. . Participants evaluated the truth or falsity of 20 items presented in question form (e.g., Do lions have manes?) in one of four semirandom orders. Participants responded by answering denitely yes, denitely no, or a little yes, a little no. The a little yes, a little no response was included to increase the sensitivity of the measure by creating a midpoint response, which allowed participants to show doubt about an item without completely rejecting it. Answer choices corresponded to pictures of a thumb pointing up, down, and sideways. To motivate the task, children were asked to help an alien learn about animals on Earth. Children responded by pointing to the thumb pictures. Undergraduates responded to written questions by circling the appropriate thumb images. Items consisted of six characteristic and noncharacteristic item pairs (see Table 1). Within each pair, the characteristic item expressed a property that is characteristic of the target kind yet present in only one gender (e.g., Do lions have manes?). This gender information then served as the predicate for the corresponding noncharacteristic item (e.g., Are lions boys?). Because the property predicated in each noncharacteristic item was a precondition for the property predicated in the corresponding characteristic item, characteristic and noncharacteristic properties were roughly equivalent in prevalence (see the Adult Pretests section below). Eight ller items with clear, obvious answers were also included (four characteristic, four noncharacteristic; see Table 1) to screen out children who did not understand the task. The ller items expressed properties that are also considered characteristic (e.g., Do tigers have stripes?) or noncharacteristic (e.g., Do dogs have wings?) of their target kind; however, because in all cases the properties in these ller items are present or absent in both genders and thus do not allow for prevalence-matched comparison items, these items were not included in our central analyses. Children responded correctly to the vast majority of the ller items (M = 97.3%). Every child answered all or all but one of the ller items correctly.
Noncharacteristic
Filler
Indicates items used in Study 2.
Part 2. . In this part, we tested participants knowledge of the gender requirements for the characteristic properties (e.g., that only male lions have manes). Participants were asked about 10 properties: 6 characteristic properties (from Part 1) that apply to individuals of only one gender, 2 ller properties that apply to individuals of both genders (e.g., both boy and girl tigers have stripes), and 2 ller properties that do not apply to individuals of either gender (e.g., neither boy nor girl dogs have wings). Participants were asked separate, sequential questions about whether these properties apply to the males and females of the species (e.g., Do boy lions have manes? Do girl lions have manes?). Question pairs were presented in one of four semirandom orders. Whether the female or male version of the question appeared rst was counterbalanced across subjects. Participants responded by answering denitely yes, denitely no, or a little yes, a little no. Children pointed to indicate their answer to a second aliens questions about boy animals and girl animals. Undergraduates circled the appropriate thumb image. If participants responded incorrectly to a Part 2 question, that item pair was excluded from their Part 1 data. Responses were considered incorrect if participants answered either of the
427
questions incorrectly (e.g., saying that boy lions do not have manes or that girl lions do). To be conservative, the intermediate answer choice was counted as a negative response. Both children and undergraduates knew the majority of the facts tested in Part 2 (MChildren = 4.07 of 6 possible; MUndergraduates = 4.93). Adult pretests. . To conrm that the predicates of the characteristic and noncharacteristic items are considered equally prevalent, we conducted two pretests with separate groups of undergraduates. In the rst, 16 undergraduates estimated the prevalence of the properties (e.g., having manes, being male) used in the main task. For example, participants were told, It is estimated that there are roughly 32,000 individual lions in the world. Please estimate what percentage of those individual lions have manes. Participants estimated the characteristic and noncharacteristic properties to be roughly equal in prevalence, F(6, 10) = 0.39, p = .87, applying to an average of 54.3% and 49.6% of category members, respectively. In the second pretest, a separate group of 16 undergraduates performed a comparative judgment for each item pair. They were asked, for example, Which do you think there are more of? (a) Boy lions; (b) Lions that have manes; (c) They are equal. Across all item pairs, the dominant response was They are equal. Participants selected this response on 67.7% of trials, signicantly more often than expected by chance (33.3%), t(15) = 3.54, p = .003. Participants selected the characteristic response (e.g., Lions that have manes) on 6.3% of trials (signicantly less likely than expected by chance), t(15) = 12.84, p < .001, and the noncharacteristic response (e.g., Boy lions) on 26.0% of trials (at chance levels), t(15) = 0.72, p = .49. Overall, pretest results conrmed that adults consider the characteristic and noncharacteristic properties to be roughly equal in prevalence.
Results Our central analysis involved the proportion of denitely yes responses for the characteristic and noncharacteristic items. We conducted a 2 (item type: characteristic, noncharacteristic; within subject) 2 (age group: children, undergraduates; between subjects) analysis of variance (ANOVA). The dependent variable was a proportion score based on the number of times participants selected the denitely yes response out of the total number of item pairs on which they demonstrated requisite knowledge in Part 2. Results revealed the predicted main effect of item type, F(1, 26) = 34.22, p < .001, gp2 = .57. As shown in Table 2, both children, F(1, 13) = 11.27, p = .002, gp2 = .30, and undergraduates, F(1, 13) = 24.16, p < .001, gp2 = .48, gave signicantly more denitely yes responses for characteristic than noncharacteristic items. (The proportion of denitely no and a little yes, a little no responses are also displayed in Table 2.) This pattern of results held for 10 of 14 children, p = .012, and 11 of 14 undergraduates, p = .006 (sign tests). Results also revealed a nonsignicant trend toward an effect of age group, F(1, 26) = 3.23, p = .084, gp2 = .11, with children tending to choose denitely yes more often than undergraduates (Ms = 0.61 and 0.43, respectively). Additional analyses were conducted to compare the proportion of denitely yes responses against rates expected by chance (.33, Table 2). As predicted, for the characteristic items both children, t(13) = 6.20, p < .001, and undergraduates, t(13) = 4.37, p = .001, selected the denitely yes response at rates greater than expected by chance. This was not the case for the noncharacteristic items. Children, t(13) = 1.41, p = .18, and undergraduates, t(13) = 1.24, p = .24, selected the denitely yes response for these items at chance levels.
Table 2 Mean Proportion (Standard Deviation) of Responses by Response Type, Age Group, and Item Type Proportion of responses (SD) Study 1 Age group Adult Children 2 Children Item type Characteristic Noncharacteristic Characteristic Noncharacteristic Characteristic Noncharacteristic Denitely yes .65 (.28)* .21 (.36) .76 (.26)* .46 (.35) .86 (.33)* .29 (.45) A little yes, A little no .35 (.28) .78 (.35)* .17 (.24)* .47 (.37) .03 (.10)* .71 (.45)* Denitely no .00 .01 .07 .07 .11 .00 (.00)* (.04)* (.19)* (.20)* (.30)* (.00)*
*Indicates p < .05 for comparisons against chance (.33).
428
Discussion Study 1 addressed the question of whether young children equate generics with quantication by comparing childrens true and false judgments about pairs of statements that, although matched in prevalence, vary in whether or not they are considered acceptable generics by adults. Results showed that children were more likely to accept generics that express characteristic than noncharacteristic properties. These ndings suggest that young children do not interpret generics as statements about the number or proportion of individual category members that possess a particular property. Instead, even young children recognize that generics make claims about kinds. The argument that children do not interpret generics as equivalent to quantication rests on the nding that children were less likely to reply denitely yes to the noncharacteristic items. We argue that children did so because they (a) recognize that generics make claims about kinds and (b) do not consider the noncharacteristic items to be true of their kinds. However, there may be alternative reasons why children were less likely to accept the generics that involved noncharacteristic properties. One possibility is that childrens responses were inuenced by the relative novelty of the noncharacteristic stimuli. Children may be less likely to have heard statements expressing the noncharacteristic properties of the target categories than statements expressing their characteristic properties, which may in turn make children less likely to agree with these less familiar items. Although a stringent test of this possibility would require a study with novel categories, at least one aspect of the current study suggests that novelty effects cannot account for childrens responses. Consider the ller items in Part 2 expressing properties that are characteristic of both genders (e.g., Do boy girl tigers have stripes?). It is unlikely that children have previously heard statements expressing these properties (because the properties do not vary by gender and; thus, parents would be unlikely to talk about the genders separately). Nevertheless, children responded denitely yes on these items 78.6% of the time. The fact that children consistently gave denitely yes responses on questions they have likely not heard before casts doubt on the possibility that the novelty of the noncharacteristic stimuli led to their low acceptance levels. Another possibility is that children provided fewer denitely yes and more a little yes, a little no responses on the noncharacteristic items because
they were confused by the structure of these questions (Are Xs Ys?). To explore this possibility, we tested an additional 10 children (M = 5.19 years) on a similar task in which participants responded to questions that were structurally equivalent to the noncharacteristic items but varied in whether their correct answer was yes or no. This task included six items each for which the correct answer was (a) denitely yes (e.g., Are princesses girls? Are daddy bears boys?), (b) denitely no (e.g., Are ladies boys? Are mommy ducks boys?), and (c) a little yes, a little no (e.g., Are kids girls? Are baby birds boys?), for a total of 18 items. We reasoned that if childrens responses to the noncharacteristic items were driven by confusion with the structure of these questions, then children should encounter the same difculty on this task. This was not the case. Although all the items in the control task were structurally analogous to the noncharacteristic items in the main study, children responded appropriately on an average of 78.0% of trials. Importantly, children had no trouble responding denitely yes where appropriate, doing so on 82.4% of items for which it was the correct response. These data suggest that childrens low level of denitely yes responses to the noncharacteristic items on the central task was not due to the unusual structure of these items. Note that this control task also speaks against two additional concerns. First, the animal items (e.g., Are daddy bears boys?) rule out the possibility that children interpreted questions such as Are lions boys? to mean Are lions human boys? (This possibility was remote anyway, as children would presumably have favored denitely no rather than a little yes, a little no if they had interpreted the questions in this way.) Second, this task rules out the possibility that children had more trouble responding to the noncharacteristic items because they had difculty dealing with relations between sets (e.g., relations between the set of lions, boys, and boy lions; see Markman, 1989; Piaget, 1952). A nal alternative explanation we considered was that participants may have responded differently to the characteristic and noncharacteristic items as a result of semantic differences due to how the items were framed. All of the characteristic items either used the verb to have to attribute a physical feature to the kind (e.g., Do lions have manes?), or else used a verbal predicate to attribute a habitual action to the kind (e.g., Do birds lay eggs?). In contrast, all of the noncharacteristic items used nominal predicates (i.e., the verb to be + count noun) to attribute a gender to the kind
429
(e.g., Are lions boys? Are birds girls?). This difference in syntactic frames is potentially problematic if statements about what a kind is carry different implications than statements about what a kind has or does. There is some evidence to suggest that this might be the case. In particular, several studies have shown that using nominal versus verbal predicates to describe a characteristic of an individual inuences judgments about the permanence of that characteristic. For example, Gelman and Heyman (1999) showed that 5-year-olds judge novel personal characteristics as more stable over time and context when they are expressed using a nominal predicate (e.g., She is a carrot-eater) than when they are expressed by a verbal phrase (e.g., She eats carrots whenever she can; see also Cimpian, Arce, Markman, & Dweck, 2007; Markman, 1989). Likewise, Reynaert and Gelman (2007) showed that adults judge physical and mental illnesses described using nominal predicates (e.g., He is a baxtermic and He is baxtermic) as more permanent and stable than those described using verbal predicates (e.g., He has baxtermia). Thus, when describing characteristics of an individual, nominal and verbal predicates seem to carry different implications. Of particular interest here is the question of whether statements about what a kind is versus what a kind has or does require different amounts of statistical evidence to be accepted as true. Specically, given the permanence and centrality that is implied when properties are described using nominal predicates, it may be that more statistical evidence is required to agree with a statement about what a kind is than to agree with a statement about what a kind has or does (see Gilson & Abelson, 1965). On this view, participants may have been less likely to endorse the noncharacteristic than the characteristic items not because of differences in whether or not the predicated properties are viewed as characteristic of their kinds, but rather because greater statistical evidence is necessary to endorse statements about what a kind is. To address this possibility, in Study 2 we examined childrens responses to characteristic and noncharacteristic item pairs that are both presented with nominal predicates using the verb to be.
well as of items that are not characteristic of the kind yet also true of only one gender. However, to control for the possibility that the item effects observed in Study 1 were driven by semantic differences due to how the characteristic and noncharacteristic items were framed (about what a kind is vs. what a kind has or does), in this study both the characteristic and noncharacteristic items were framed in terms of what the kind is. For example, to match the noncharacteristic item Are lions boys? the characteristic item Do lions have manes? was replaced with Are lions animals with manes? If, as argued, childrens responses in Study 1 were driven by whether or not the predicated properties are viewed as characteristic of their kinds, then children should again be more likely to endorse generics that express characteristic than noncharacteristic properties, even when both types of items are presented with nominal predicates using the verb to be.
Method Participants Twelve children (5 males, 7 females; M = 5.10 years, SD = .47) participated. Four additional children were excluded because they did not know any of the gender properties. The children, all of whom were recruited in a small Midwestern city, were predominantly European American and came from a range of socioeconomic backgrounds. Materials and Procedure Part 1. . Children evaluated the truth or falsity of 12 items. Eight of the items were based on the four characteristic and noncharacteristic item pairs with which children from Study 1 were most familiar (see Table 1 for a full list). In addition, there were 4 ller items (2 characteristic, 2 noncharacteristic). Every child answered all or all but one of the ller items correctly. Importantly, all items used nominal predicates using the verb to be (e.g., characteristic: Are lions animals with manes?; noncharacteristic: Are lions boys?; ller: Are tigers animals with stripes?). All other aspects of the design were identical to Study 1. Part 2. . Part 2 was designed to test participants knowledge of the gender requirements for the characteristic properties in Part 1 (e.g., that only male lions have manes). This section was identical in structure to Part 2 of Study 1; however, only the four characteristic properties from Part 1 and four ller items were examined. If participants
Study 2 As in Study 1, we examined childrens endorsements of items expressing properties that are characteristic of the kind yet true of only one gender, as
430
responded incorrectly to a Part 2 question, that item pair was excluded from their Part 1 data. Children demonstrated knowledge of the gender requirements for roughly half of the properties tested (M = 1.83 of 4 possible). (It is unclear why children did not respond appropriately to more of the Part 2 items. The most common error was choosing denitely yes in response to questions about both male and female category members [M = 1.38 of 4 possible], potentially suggesting that these participants, who were somewhat younger than those in Study 1, had not yet learned how males and females differ on some of the characteristic properties.) Results Our central analysis involved the proportion of denitely yes responses for the characteristic and noncharacteristic items. We conducted an ANOVA with item type (characteristic, noncharacteristic) as a within-subject variable. The dependent variable was a proportion score based on the number of times participants selected the denitely yes response out of the total number of item pairs on which they demonstrated requisite knowledge in Part 2. Results revealed the predicted main effect of item type, F(1, 11) = 17.33, p = .002, gp2 = .61. As shown in Table 2, children gave signicantly more denitely yes responses for characteristic than noncharacteristic items. This pattern held for 8 of 12 children, p = .008 (sign test). Additional analyses were conducted to compare the proportion of denitely yes responses against rates expected by chance (.33; see Table 2). As predicted, for the characteristic items children selected the denitely yes response at rates greater than expected by chance, t(11) = 5.54, p < .001. In contrast, children selected the denitely yes response for the noncharacteristic items at chance levels, t(11) = )0.30, p = .77. Discussion The purpose of this study was to evaluate the possibility that participants were more likely to agree with the characteristic than the noncharacteristic items in Study 1 because of semantic differences due to how these items were framed (i.e., as claims about what a kind has or does versus claims about what a kind is). Thus, in Study 2, children were asked to judge characteristic and noncharacteristic items that were both framed in terms of what the kind is (e.g., Are lions animals with manes? vs. Are lions boys?). We found that
even when the characteristic and noncharacteristic items were framed in the same way, children were still signicantly more likely to endorse generics that express characteristic than noncharacteristic properties. Together with the results of Study 1, these ndings lend further support to our claim that when deciding on the truth value of a generic statement, what matters most is not the statistical prevalence of the property, but rather whether or not that property is viewed as characteristic of the kind.
General Discussion This article explored whether young children understand generic meaning as distinct from quantication. We compared childrens responses to questions about properties that, although roughly equivalent in prevalence (each applying to approximately 50% of kind members), differed in whether or not they are considered by adults to be generically true of the kind (e.g. Do lions have manes? or Are lions animals with manes? vs. Are lions boys?). As predicted, children were more likely to accept generics that express characteristic than noncharacteristic properties. These data expand upon the growing literature concerning childrens comprehension of generics (e.g., Cimpian & Markman, 2008; Gelman & Bloom, 2007; Gelman & Raman, 2003; Hollander et al., 2002) in two ways. First, our ndings suggest that children recognize that generic facts can be considered true even if they are not true of all or even most members of a category (e.g., Lions have manes). Second, our data suggest that children recognize that some generic statements can be considered untrue even if they express properties that are fairly frequent (e.g., Lions are boys). Overall, the data presented here provide the rst direct evidence that, like adults, young children do not interpret generics as statements about the number or proportion of individual category members that possess a particular property. Our data reveal instead that, even for young children, generics make claims about kinds. These ndings also speak to an ongoing debate about the nature of childrens concepts. There is now substantial evidence to suggest that adults concepts cannot be characterized by statistical information alone; they are better characterized as theory-like explanatory structures that incorporate not only knowledge about shared features (including nonobvious ones) but also causalexplanatory links between those features (e.g., Ahn, Flanagan,
431
Marsh, & Sanislow, 2006; Murphy, 2002; Rips & Collins, 1993). The data presented here are consistent with the notion that the concepts of young children also consist of more than just statistical regularities (see also Gelman, 2003; Gopnik & Schulz, 2004; Keil, 2006). Young childrens interpretation of generic sentences reveals that they treat certain categories, including the animal categories examined here, as abstract generic kinds. Of course, both statistical information and abstract conceptual knowledge likely play a role in childrens reasoning about kinds (e.g., Prasada & Dillingham, 2006, 2009). In fact, the current experiments provide some evidence that both children and adults also recognize and consider the role of prevalence information in evaluating the truth of generic statements. Consider participants responses to the noncharacteristic items (e.g., Are lions boys?). Both children and adults avoided the denitely no response on these items in favor of the less denitive a little yes, a little no answer choice. This response may reect participants awareness that the noncharacteristic properties are true of at least some individuals within the target kinds. By selecting the a little yes, a little no response, participants were able to convey their knowledge of the prevalence of the noncharacteristic properties while still denying that these properties characterize the kind. This pattern of results is consistent with the view that although children do not rely on prevalence considerations alone when evaluating statements about kinds, they nonetheless do not ignore the statistical evidence. An important question for future research is how childrens statistical learning mechanisms (e.g., Saffran, Aslin, & Newport, 1996) interact with their abstract conceptual knowledge and their nely tuned causal reasoning abilities (e.g., Gopnik & Schulz, 2007) to support conceptual development (see Sobel & Kirkham, 2007; Tenenbaum, Grifths, & Kemp, 2006). An additional important question this work raises concerns the nature of the differences between properties that can and cannot be truthfully predicated of a kind. The question of what makes a property generically true of a kind does not yet have a precise answer. According to one view (Prasada & Dillingham, 2006, 2009), there are two circumstances under which it may be legitimate to predicate a property of an entire kind: (a) when a property is highly prevalent among kind members and (b) when a property has a principled connection with the kindthat is, when instances possess the feature by virtue of being the kinds of things they are (e.g., dogs bark
because barking is part of what it means to be a dog). Which properties are linked to their kinds in this principled way? A number of candidate property types have been proposed. For example, properties may be predicated generically if they t dimensions specied by overhypotheses (e.g., each kind of animal has a characteristic diet, sound, mode of reproduction; Shipley, 1993; see Leslie, 2007, 2008). Relatedly, properties that t with peoples na ve causal theories about the relevant kinds have been shown to be accepted when predicated in generic form (Cimpian, Gelman, et al., 2010; Gelman & Bloom, 2007). Leslie (2007, 2008) has proposed that properties that are especially striking or dangerous (e.g., Sharks attack swimmers) may also be predicated generically if they are viewed as part of a kinds natural disposition (see also Cimpian, Brandone, et al., 2010). However, as Leslie (2007, 2008) points out, a property cannot be correctly predicated of a kind if there is an equally salient, positive alternative property that applies to other members of the kind. For example, Lions are boys may be considered false because female lions possess an equally salient alternative property (i.e., being female). In contrast, Lions have manes may be considered true because female lions lack this property and do not exhibit a salient competing feature. Overall, these proposals begin to address the important question regarding which sorts of properties get linked to their categories in principled ways, which get linked to their categories via prevalence alone, and why. Further work, including developmental research, on these questions is needed. In particular, research on how children learn which properties are principled versus not and on the potential role of parental input in this learning process (e.g., Gelman, Coley, Rosengren, Hartman, & Pappas, 1998; Gelman, Taylor, & Nguyen, 2004) may be especially informative. To conclude, our ndings demonstrate that for children, as for adults (Cimpian, Brandone, et al. 2010; Cimpian, Gelman, et al., 2010; Khemlani et al., 2007; Khemlani et al., 2009), generic meaning does not reduce to considerations of how many individuals have the relevant property. Even when children know how prevalent the relevant properties are among category members, they do not use this knowledge as the sole basis for evaluating the truth of generic statements. Rather, considerations about whether generically predicated properties are characteristic of their kind play a deciding role in childrens reasoning about generic sentences.
432
References
Ahn, W., Flanagan, E. H., Marsh, J. K., & Sanislow, C. A. (2006). Beliefs about essences and the reality of mental disorders. Psychological Science, 17, 759766. Brandone, A. C., & Gelman, S. A. (2009). Differences in preschoolers and adults use of generics about animals and artifacts: A window onto a conceptual divide. Cognition, 110, 122. Carey, S. (1985). Conceptual change in childhood. Cambridge, MA: MIT Press. Carey, S. (2009). The origin of concepts. New York: Oxford University Press. Carlson, G. N. (1977). A unied analysis of the English bare plural. Linguistics and Philosophy, 1, 412456. Carlson, G. N., & Pelletier, F. J. (Eds.). (1995). The generic book. Chicago: University of Chicago Press. Chambers, C. G., Graham, S. A., & Turner, J. N. (2008). When hearsay trumps evidence: How generic language guides preschoolers inferences about unfamiliar things. Language and Cognitive Processes, 23, 749766. Cimpian, A., Arce, H. C., Markman, E. M., & Dweck, C. S. (2007). Subtle linguistic cues affect childrens motivation. Psychological Science, 18, 314316. Cimpian, A., Brandone, A. C., & Gelman, S. A. (2010). Generic statements require little evidence for acceptance but have powerful implications. Cognitive Science, 34, 14521482. Cimpian, A., Gelman, S. A., & Brandone, A. C. (2010). Theory-based considerations inuence the interpretation of generic sentences. Language and Cognitive Processes, 25, 261276. Cimpian, A., & Markman, E. M. (2008). Preschool childrens use of cues to generic meaning. Cognition, 107, 1953. Cimpian, A., Meltzer, T. J., & Markman, E. M. (2011). Preschoolers use of morphosyntactic cues to identify generic sentences: Indenite singular noun phrases, tense, and aspect. Child Development, 82, 15611578. Colunga, E., & Smith, L. B. (2005). From the lexicon to expectations about kinds: A role for associative learning. Psychological Review, 112, 347382. Gelman, S. A. (2003). The essential child: Origins of essentialism in everyday thought. Oxford, UK: Oxford University Press. Gelman, S. A., & Bloom, P. (2007). Developmental changes in the understanding of generics. Cognition, 105, 166183. Gelman, S. A., & Brandone, A. C. (2010). Fast-mapping placeholders: Using words to talk about kinds. Language Learning and Development, 6, 223240. Gelman, S. A., Coley, J. D., Rosengren, K., Hartman, E., & Pappas, A. (1998). Beyond labeling: The role of maternal input in the acquisition of richly-structured categories. Monographs of the Society for Research in Child Development, 62(1, Serial No. 253). Gelman, S. A., Goetz, P. J., Sarnecka, B. S., & Flukes, J. (2008). Generic language in parentchild conversations. Language Learning and Development, 4, 131.
Gelman, S. A., & Heyman, G. D. (1999). Carrot-eaters and creature-believers: The effects of lexicalization on childrens inferences about social categories. Psychological Science, 10, 489493. Gelman, S. A., & Raman, L. (2003). Preschool children use linguistic form class and pragmatic cues to interpret generics. Child Development, 74, 308325. Gelman, S. A., Taylor, M. G., & Nguyen, S. (2004). Motherchild conversations about gender: Understanding the acquisition of essentialist beliefs. Monographs of the Society for Research in Child Development, 69(1, Serial No. 275). Gelman, S. A., Waxman, S., & Kleinberg, F. (2008). The role of representational status and item complexity in parentchild conversations about pictures and objects. Cognitive Development, 23, 313323. Gilson, C., & Abelson, R. P. (1965). The subjective use of inductive evidence. Journal of Personality and Social Psychology, 2, 301310. Gopnik, A., & Meltzoff, A. N. (1997). Words, thoughts, and theories. Cambridge, MA: Bradford Books MIT Press. Gopnik, A., & Schulz, L. (2004). Mechanisms of theory formation in young children. Trends in Cognitive Sciences, 8, 371377. Gopnik, A., & Schulz, L. (Eds.). (2007). Causal learning: Psychology, philosophy, and computation. New York: Oxford University Press. Hollander, M. A., Gelman, S. A., & Star, J. (2002). Childrens interpretations of generic noun phrases. Developmental Psychology, 38, 883894. Keil, F. (1989). Concepts, kinds, and cognitive development. Cambridge, MA: Bradford Book MIT Press. Keil, F. C. (2006). Explanation and understanding. Annual Review of Psychology, 57, 227254. Khemlani, S., Leslie, S. J., & Glucksberg, S. (2009). Generics, prevalence, and default inferences. Proceedings of the 31st Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society. Khemlani, S., Leslie, S. J., Glucksberg, S., & Rubio-Fernandez, P. (2007). Do ducks lay eggs? How humans interpret generic assertions. Proceedings of the 29th Annual Conference of the Cognitive Science Society. Nashville, TN: Cognitive Science Society. Leslie, S. J. (2007). Generics and the structure of the mind. Philosophical Perspectives, 21, 375405. Leslie, S. J. (2008). Generics: Cognition and acquisition. Philosophical Review, 117, 149. Markman, E. M. (1989). Categorization and naming in children: Problems in induction. Cambridge: Bradford Book MIT Press. Murphy, G. L. (2002). The big book of concepts. Cambridge, MA: MIT Press. Piaget, J. (1952). The origins of intelligence in children. New York: International University Press. Prasada, S. (2000). Acquiring generic knowledge. Trends in Cognitive Sciences, 4, 6672.
Do Lions Have Manes? Prasada, S., & Dillingham, E. M. (2006). Principled and statistical connections in common sense conception. Cognition, 99, 73112. Prasada, S., & Dillingham, E. M. (2009). Representation of principled connections: A window onto the formal aspect of common sense conception. Cognitive Science, 33, 401448. Reynaert, C. C., & Gelman, S. A. (2007). The inuence of language form and conventional wording on judgments of illness. Journal of Psycholinguistic Research, 36, 272295. Rips, L. J., & Collins, A. (1993). Categories and resemblance. Journal of Experimental Psychology: General, 122, 468486. Saffran, J. R., Aslin, R. N., & Newport, E. L. (1996). Statistical learning by 8-month-olds infants. Science, 274, 19261928. Shipley, E. F. (1993). Categories, hierarchies, and induction. Psychology, Learning, and Motivation, 30, 265301. Sloutsky, V. M., & Fisher, A. V. (2004). Induction and categorization in young children: A similarity-based
433
model. Journal of Experimental Psychology: General, 133, 166188. Sloutsky, V. M., Kloos, H., & Fisher, A. V. (2007). When looks are everything: Appearance similarity versus kind information in early induction. Psychological Science, 18, 179185. Smith, L. B., Jones, S. S., & Landau, B. (1996). Naming in young children: A dumb attentional mechanism? Cognition, 60, 143171. Sobel, D. M., & Kirkham, N. Z. (2007). Interactions between causal and statistical learning. In A. Gopnik & L. E. Schulz (Eds.), Causal learning: Psychology, philosophy, and computation (pp. 139153). New York: Oxford University Press. Tenenbaum, J. B., Grifths, T. L., & Kemp, C. (2006). Theory-based Bayesian models of inductive learning and reasoning. Trends in Cognitive Sciences, 10, 311318. Wellman, H. M., & Gelman, S. A. (1998). Knowledge acquisition. In D. Kuhn & R. Siegler (Eds.), Handbook of child psychology: Vol. 2. Cognition, perception, and language (4th ed., pp. 523573). Hoboken, NJ: Wiley.
Running Head: ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS
Essentialist Beliefs About Bodily Transplants in the United States and India Meredith Meyer University of Michigan Sarah-Jane Leslie Princeton University Susan A. Gelman and Sarah M. Stilwell University of Michigan
Author Note Meredith Meyer, Department of Psychology, University of Michigan; Sarah-Jane Leslie, Department of Philosophy, Princeton University; Susan A. Gelman, Department of Psychology, University of Michigan; Sarah M. Stilwell, Department of Psychology, University of Michigan. This research was supported by National Science Foundation Grant 059394 to Gelman. We gratefully acknowledge the research assistance of Elizabeth Anastasia and Jocelyn Kuhn. Correspondence concerning this article should be addressed to Meredith Meyer, Department of Psychology, 530 Church Street, University of Michigan, Ann Arbor, MI 48109. Phone: (734) 615-0575 E-mail: mermeyer@umich.edu Keywords: Causal reasoning, concepts, culture, psychology, psychological essentialism
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Abstract Psychological essentialism is the belief that some internal, unseen essence or force determines the common outward appearances and behaviors of category members. We investigated whether reasoning about transplants of bodily elements showed evidence of essentialist thinking. Both Americans and Indians endorsed the possibility of transplants conferring donors' personality, behavior, and luck on recipients, consistent with essentialism.
!"
Respondents also endorsed essentialist effects even when denying that transplants would change a recipient's category membership (e.g., predicting that a recipient of a pig's heart would act more pig-like, but denying that the recipient would become a pig). This finding runs counter to predictions from the strongest version of the "minimalist" position (Strevens, 2000), an alternative to essentialism. Finally, studies asking about a broader range of donor-to-recipient transfers indicated that Indians essentialized more types of transfers than Americans, but neither sample essentialized monetary transfer. This suggests that results from bodily transplant conditions reflect genuine essentialism rather than broader magical thinking. Keywords: Causal reasoning, concepts, culture, psychology, psychological essentialism
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Essentialist Beliefs About Bodily Transplants in the United States and India 1. Introduction In Sir Arthur Conan Doyle's 1923 story The Adventure of the Creeping Man, an eminent professors aide asks Sherlock Holmes to investigate some peculiar changes in the professor's behavior and appearance. The professor has become increasingly aggressive, and his knuckles
!"
have become thickened and hairy; he has developed superhuman climbing abilities, and at times he adopts a strange slouching gait. Holmes cleverly deduces that the professor has been injecting himself with a serum derived from langur monkeys in an attempt to maintain his youth and energy. Although the serum does indeed fulfill its intended purpose, it also causes the unfortunate side effects that so worried the professor's aide. The idea that a serum extracted from a monkey might produce monkey-like appearance and behavior in a human seems clearly at odds with modern-day scientific thinking. However, psychological essentialism would predict that such an idea might nonetheless strike people as intuitively plausible. Psychological essentialism is the theory that people tacitly believe that some internal unseen essence or force determines the common outward appearances and behaviors of the members of a natural kind (Gelman, 2003; Medin & Ortony, 1989). For instance, a monkey's gait, temperament, and climbing abilities are seen as arising from an internal kind-specific monkey essence. In the case of the "creeping man," the professor's monkey-like behaviors would thus be attributed to the transfer of some of the monkey's essence from inside the monkey--the essence's natural home--to the insides of the professor. Numerous studies indicate that essentialism underlies much of how people reason about natural kinds. Infants, toddlers, and preschoolers often generalize internal features and behaviors based on category membership, even when category membership and perceptual attributes
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS
!"
conflict (e.g., preschoolers expect that a property attributed to a leaf-insect will be found in other dissimilar-looking insects rather than a similar-looking leaf). These results suggest a strong belief in the importance of the non-obvious in determining category structure, which in turn licenses inductive inferences (Dewar & Xu, 2009; Gelman & Coley, 1990; Gelman & Markman, 1986, 1987; Graham, Kilbreath, & Welder, 2004). As well, both children and adults expect animals to maintain their kind membership and species-typical behaviors even if external superficial features of the animal or its environment are changed (e.g., Gelman & Wellman, 1991; Keil, 1989). Notably, the underlying basis is also seen as causal; people privilege internal causal features as bases for categorization over non-causal or outward features (Ahn, Kim, Lassaline, & Dennis, 2000; Gelman & Gottfried, 1996; Gelman & Kremer, 1991; Gottfried & Gelman, 2005). Finally, people often construe certain categories as sharply-bounded, natural, and immutable rather than invented and fluid -- a pattern of thinking that is readily explained if people attribute an inalterable and naturally-occurring essence to these categories (Bastian & Haslam, 2006; Haslam, Bastian, Bain, & Kashima, 2006; Rhodes & Gelman, 2009; Taylor, Rhodes, & Gelman, 2009). One need not have beliefs about the precise nature or location of an essence to believe that it exists and causes common features of category members. Rather, essentializing may in some cases simply involve a belief that there is something deep, internal, and unalterable that is causally responsible for an individual's properties; this is often referred to as an essence placeholder (Gelman, 2003; Medin & Ortony, 1989). Importantly, however, even though people need not have beliefs about the specific location of essence, essences are thought to reside in internal parts (Gelman & Wellman, 1991; Gottfried & Gelman, 2005; Newman & Keil, 2008). If essences are thought to pervade the internal parts of people and animals, and to have causal
!"
powers, then the transfer of internal parts from one individual to another may be thought to cause the recipient to take on some of the donors characteristics. This is, of course, precisely the premise behind Conan Doyles Creeping Man. The current study focuses on this hypothesized pattern of essentialist thinking, examining whether people systematically believe that transfer of various internal bodily elements could cause recipients to become more like their donors. We studied this hypothesized pattern of reasoning about bodily element transfers (hereafter referred to simply as "transplants") by surveying two culturally distinct samples, people living in the United States and in India. Psychological essentialism would be reflected by the belief that transplants might confer the traits or characteristics of a donor onto a recipient. Investigating such a reasoning bias in this domain serves to extend prior work on psychological essentialism in several important ways. First, many of the essentialist predictions that have been documented to date actually have some empirical support or grounding in reality: For instance, it is sensible to expect that members of the same animal kind would share similar internal features, and that changes to the outward appearance of an animal would not interfere with its membership in a species or its innately-determined behaviors. However, there is no scientific model to account for why transplants might lead to transference of features. Thus, evidence of essentialist reasoning about transplants would provide especially strong evidence for the farreaching nature of this style of thinking, because it would reflect a reasoning error. The present studies also test the extent to which essentialist reasoning extends to the social domain. The majority of studies on essentialism to date have focused on beliefs about natural kinds, particularly animals. However, a number of researchers have proposed that people often hold essentialist beliefs about social kinds (Allport, 1954; Astuti, Solomon, & Carey, 2004; Birnbaum, Deeb, Segall, Ben-Eliyahu, & Diesendruck, 2010: Diesendruck & Haber, 2009;
!"
Diesendruck & haLevi, 2006; Gelman & Heyman, 1999; Gelman, Taylor, & Nguyen, 2004; GilWhite, 2001; Haslam, Rothschild, & Ernst, 2000; Hirschfeld, 1996; McIntosh, 2009; Prentice & Miller, 2007; Rhodes & Gelman, 2009; Rothbart & Taylor 1992), and, importantly, that essentialist beliefs may be linked to stereotyping and social prejudice (Allport, 1954; Bastian & Haslam, 2006; Keller, 2005; Leslie, in press a; Rothbart & Taylor, 1992; Smiler & Gelman, 2008). Thus, understanding the nature and extent of social essentialism is of considerable interest. However, previous work on social essentialism has focused primarily on race, gender, and ethnic categories. Are these the only social categories that are essentialized, or might essentialist thinking extend to other social categories? The current studies address this question by asking respondents to consider transplants from members of a wide range of social categories (e.g., a philanthropist, a murderer, a homeless person, a person with a different sexual orientation, as well as a person with a different gender), as well as transplants from animals (a pig and a chimpanzee). These studies also advance our understanding of essentialism by contributing to an ongoing debate regarding the very existence of essentialism. Although many studies suggest the operation of essentialism in people's reasoning about a wide range of categories (Gelman, 2003), one alternative account known as the "minimalist" position denies that people's category-based inferences involve a representation of essence at all (Strevens, 2000). The minimalist position instead suggests that people appeal to non-essentialist causal laws (so-called "K-laws") linking kind membership with observable properties or behaviors. Strevens further claims that belief in these K-laws is enough to explain the data in favor of essentialism. Taking as an example the process by which one infers that tigers will have stripes, Strevens explains, "...you must believe there is something about tigers that causes them to have stripes, but you do not have to believe
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS that this something is essence. For example, you might have no opinion about what does the causing, or you might think that a mechanism that is not an essence does the causing (as in modern biology), or you might think that it is just a brute fact about the world that being a tiger causes an animal to grow stripes" (2000, p. 154). According to minimalism, the causal laws linking kinds with observable features are believed to be in effect for an individual so long as that individual has membership in the relevant kind category; consequently membership in a kind licenses predictions of kind-typical
!"
behavior. On the strongest version of this claim, it is membership in a kind, not an essence or any other further causal intermediary, that people appeal to as the basis for predicting categorytypical properties. We refer to this account as the "unmediated" version of minimalism, in the sense that kind membership and properties are believed to be directly connected by causal laws (e.g., it is just a brute fact that they are causally linked [Strevens, 2000, p. 154; see also Figure 3, p. 155]). Another interpretation of the minimalist position, however, allows that individuals may draw on specific mediating causes linking kind membership and outward properties to draw kind-based inferences, but suggests that there is not sufficient evidence that these beliefs appeal to essences. We refer to this reading as the "mediated" reading, in the sense that specific mediating links may be represented (though they are described as non-essentialist). On both the unmediated and mediated views described above, minimalism claims that people's inferences about natural kinds do not involve representation of a causal essence. In contrast, psychological essentialism posits that people believe that individual members of a kind have essences that pervade their internal parts, and cause their external kind-typical appearance and behavior. A given individual is a member of a kind and displays kind-typical features because it has this sort of essence.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS In many respects, psychological essentialism and minimalism make very similar
!"
predictions, and to date no experiment has presented a direct test of the competing positions (but see Ahn et al., 2001, and Strevens, 2001, for a review of relevant indirect evidence). The current study was designed to directly test predictions from essentialism and the unmediated reading of minimalism (see the General Discussion for issues related to a mediated reading, specifically to what extent it constitutes a distinct theory from placeholder essentialism, once it is appropriately constrained by the empirical evidence). Specifically, it addressed this issue by asking respondents to consider not just whether transplants would confer aspects of the donor on the recipient, but also whether the recipient's category membership would change to that of the donor. If respondents endorsed the possibility that transplants might transfer characteristics of the donor without changing the recipient's category membership, this would provide empirical support against unmediated minimalism because it would indicate that people were not simply basing predictions of kind-typical behavior on category membership, but rather on an intervening causal force capable of being transmitted, and exerting its effects independent of category membership. More generally, it would provide evidence, contra Strevens, that people rely on more than just beliefs in K-laws linking category-membership and typical properties in their reasoning. (Interestingly, in the Conan Doyle story, the professor is not said to have become a monkey and so ceased to be a man rather, he merely takes on some monkey characteristics. Thus the pattern of reasoning suggested by the Conan Doyle tale is not readily explained by the unmediated minimalist approach.) More broadly, investigating people's essentialist beliefs about transplants also potentially provides insight into the factors people may consider when donating and accepting transplants. Transfer of internal body parts is now relatively commonplace across the world; hundreds of
!"
thousands of soft tissue and bone transplants take place every year (Lederer, 2008), and the types of organs that can be transplanted successfully are ever-increasing. Although medical advances in transplant viability have provided more and more people with improvements in quality and length of life, the need for organs far outstrips the supply (Feeley & Moon, 2009). Investigating people's implicit beliefs and feelings about transplants may help contribute to our understanding of how people arrive at the decision to become donors. Further, the shortage of human donors has prompted scientists to investigate the possibility of xenotransplantation, in which organs are transplanted from non-human animals (usually pigs) to humans (McLean & Williamson, 2005). One outstanding question is whether essentialist beliefs will pose a significant obstacle to the acceptance of xenotransplants. To date, little research has focused on essentialist beliefs about transplants directly. A number of anecdotal reports from transplant recipients hint at such essentialist thinking, however. In Claire Sylvia's autobiographical account A Change of Heart (1997), for example, Sylvia attributes a series of personality changes following a heart and lung transplant to her donor, claiming that her new cravings for beer, as well as increased assertiveness and "masculine" energy, can be explained by having received the organs of an eighteen-year-old male motorcycle enthusiast and avid beer drinker. Several studies of attitudes about organ transplants also report evidence consistent with such essentialist reasoning. For example, people often report reluctance to receive organs from animals, and sometimes from humans from different social groups, and consistently explain that reluctance by suggesting that transplants might cause both their appearance and personality to become more similar to those of their donors (Basch, 1973; Belk, 1990; Coffman et al., 1998; Hayward & Madill, 2003; Sanner, 2001a, 2001b). And in a small survey of heart recipients, a full third of respondents endorsed the
!"#
possibility of having taken on aspects of their human donor's personality or other psychological properties (Inspector, Kutz, & David, 2004). The studies described above certainly suggest that people may hold essentialist beliefs concerning organ transplants. However, the conclusions that can be drawn regarding the prevalence of essentialism are limited, because these studies either focused exclusively on small samples of individuals who had already received a transplant, or the study design consisted of minimally-structured interview techniques. To address these limitations, we designed the current studies to target two large culturally distinct samples (American and Indian), and we employed a series of structured survey questions directly addressing essentialist construals of transplants. One systematic psychological study on transplants provides highly relevant evidence regarding our research question. Hood, Gjersoe, Donnelly, Byers, and Itajkura (2011) investigated people's views on receiving transplants from a highly immoral donor (a violent murderer) vs. a morally positive donor (a volunteer worker). When asked to consider receiving either a heart or a kidney transplant, participants ratings of happiness dropped substantially from baseline (i.e., the happiness rating when the identity of the hypothetical donor was unknown) if they were told that the hypothetical donor was a murderer. Further, this decrease was significantly larger than the corresponding increase in happiness that occurred if participants were told that the donor was a volunteer worker, which suggests that individuals were more sensitive to the negativity in a morally objectionable donor than they were to the positivity in a morally upright donor. Hood and colleagues couch their findings in the larger literature concerned with the notion of "moral contagion," which was pioneered by Paul Rozin and his colleagues. Studies of moral contagion have found that people are extremely uncomfortable with the idea of coming
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS into contact with personal items that belonged to individuals deemed morally negative (e.g.,
!!"
Hitlers sweater). In contrast, contact with morally positive individuals (e.g., Mother Teresa) also affects value judgments (Frazier, Gelman, Wilson, & Hood, 2009; Newman, Diesendruck, & Bloom, 2011), but typically does not elicit comparably large gains in positive emotions, paralleling the effect found by Hood and colleagues (2011) (Nemeroff & Rozin, 1994; Rozin, Markwith, & McCauley, 1994; Rozin, Millman, & Nemeroff, 1986). Nemeroff and Rozin (1994) suggest that one of the reasons that individuals are so uncomfortable with physical contact (albeit indirect) with immoral individuals is that they are concerned about a transfer of essence: "The underlying assumption ...either explicitly or implicitly, seems to be that, through contact, some 'essence' or 'soul stuff,' some as yet undefined contagious entity, may be transmitted" (p. 159). In their study of organ transplants, Hood and colleagues (2011) similarly suggest that respondents are concerned that they may become morally contaminated by the murderer's essence, and note that such fears are predicted by psychological essentialism. However, although the results from Hood et al. are indeed consistent with the claim that respondents are employing essentialist reasoning, such reasoning is not necessarily responsible for people's discomfort with the idea of receiving a murderer's organ. To be confident that people are employing essentialist reasoning, one would need evidence that people are concerned about the causal effects of the transplants, in particular that the transplant will cause them to become more like the donor. This belief was not directly examined in Hood and colleagues' fascinating study on organ transplants. Furthermore, we know of only two studies in the broader moral contagion literature that address the causal effects predicted by essentialism. One is an unpublished study demonstrating that both older children and adults reported that wearing Mr. Roger's sweater would result in friendlier behavior, even when the person wearing it was unaware of its prior owner (Johnson & Jacobs,
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS 2001). The second is a finding that adults are more likely to expect members of a culture to possess traits of animals they typically ingest vs. animals they interact with, with one possible interpretation being that ingestion is believed to transfer a causal essence (Nemeroff & Rozin, 1989). Absent direct evidence of causal thinking, then, one cannot necessarily attribute Hood and colleagues' findings on organs to psychological essentialism. There are a number of alternative non-causal reasoning patterns that could have led to unhappiness with a murderer's
!"#
organ. For instance, people might feel "creeped out," or feel that they have been contaminated, or experience other negative emotions when thinking about contact with the organs of a morally negative person. None of these negative reactions necessarily involves fearing that they would become like the donor, and so do not require essentialist thinking. People might also consider the social ramifications of receiving an organ from a criminal, expecting that others might view them with suspicion or disgust. Finally, people might simply assume that a murderer would be more likely to have engaged in unhealthy behaviors that would weaken the physical integrity or function of the organ (e.g., drug and alcohol abuse), thereby rendering it less desirable. We designed the current studies to investigate essentialism and its implications directly. Based on the pervasiveness of essentialist thinking when considering the relation between categories and outward properties in other domains (Gelman, 2003), we predicted that people would also employ essentialist thinking here, and so would report that a transplant would have causal consequences -- namely causing the recipient to take on characteristics of the donor. Study 1 was a survey of American respondents, whereas Study 2 employed the same survey with a sample of Indian respondents to investigate the ways that essentialism about transplants may be manifested in these different countries. We also assessed other potential reactions to transplants
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS that were not grounded in essentialist biases. Study 3 served as a control study, assessing whether respondents (both American and Indian) endorsed causal effects for the transfer of
!"#
something that was non-bodily, namely money. If respondents endorsed causal change for bodily transplants but not for money, this would strengthen the argument that people were displaying genuine essentialism in reasoning about transplants. Finally, Study 4 (also with American and Indian respondents) addressed whether essentializing of bodily transplants depended on whether the bodily element was internal and/or biological. 2. Study 1 To address the complexities involved in the expression of essentialist and non-essentialist reasoning about internal bodily elements, we manipulated donor type, transplant type, and recipient identity. We first provide an overview of the survey and then address each of these manipulations below. Surveys started with a vignette introducing a hypothetical transplant, described as necessary for the recipient's continued health. Respondents next viewed a list of possible donors and rank-ordered them according to how much they would like each one to be the donor. Respondents then considered each donor in turn, and provided a liking rating indicating how much they liked the idea of each one being a donor, and then provided an openended explanation for their liking ratings. Finally, they answered four Likert-scale questions, two of which assessed causal-essentialist beliefs, and the other two of which assessed non-essentialist beliefs 1. Notably, essentialist and non-essentialist inferences were not mutually exclusive, and an individual could simultaneously endorse both (e.g., a respondent could both endorse the idea that a transplant might confer a donor's traits on the recipient [essentialist] and express concerns that ########################################################
1
The term "non-essentialist" does not imply denial of essentialism, but rather is used to describe beliefs that are not
indicative of essentialism.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS others might view the recipient differently after the transplant [non-essentialist]). An Essentialism composite score was derived from responses to the two essentialist
!"#
questions. The first essentialist question, Personality/Behavior, assessed people's beliefs that the transplant would cause the recipient's personality or behavior to become more like the donors. The second essentialist question, Luck, assessed an alternative form of essentialist thinking by asking whether individuals might expect a transplant to cause fortunate or unfortunate events to befall them -- the idea being that one's essence might also be seen as causally involved in bringing about one's circumstances (see Olson, Banaji, Dweck, & Spelke, 2006, for evidence that even young children negatively evaluate individuals who experience bad luck and extend this judgment to others in the same social group, as well as Callan, Ellard, & Nicol, 2006 for evidence of adults' beliefs in "immanent justice" -- the idea that prior misdeeds causally bring about misfortune or punishment). Endorsement of either the personality/behavior change or luck change questions would constitute evidence for essentialism, because it is these endorsements that explicitly reflect beliefs in the causal relationship between essence and outward properties or circumstances. In other words, both cases entail a change in the recipient of the transplant, whereby the recipient takes on characteristics of the donor (personality/behavior or luck). Similarly, a Non-Essentialist composite score was derived from responses to the two nonessentialist questions. The first non-essentialist question, Reputation, asked respondents about the degree to which they believed that others might view them more negatively or positively post-transplant. The second non-essentialist question, Creeped Out/Contaminated, assessed the degree to which respondents anticipated disgust or feelings of dirtiness or contamination upon their receiving the transplant. Based on past surveys suggesting essentialist views of organ transplants (Basch, 1973; Belk, 1990; Coffman et al., 1998; Hayward & Madill, 2003; Sanner,
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS 2001a, 2001b) as well as studies of disgust, contagion, and purity (e.g., Nemeroff & Rozin, 1994), we predicted that respondents would endorse both kinds of reasoning. Donor Types. Respondents were provided with a wide range of potential donors (Table
!"#
1), classified into five groups: similar-to-self, different-from-self, positive, negative, and crossspecies. We provided similar-to-self types to assess baseline ratings. That is, we assumed that donors who were similar to the respondent (either through close family ties or possession of similar characteristics and environmental background) would be construed as contributing the most similar essence (and thus effect the fewest changes). Thus scores for various feelings or beliefs about the effects of these transplants would provide a baseline reference point against which we could compare scores for more dissimilar donor types. Different-from-self types were described as differing from the respondent on a major and culturally salient dimension (gender and sexual orientation). Positive and negative donor types were also included, inspired by past studies of magical contagion and feelings about organ transplants (e.g., Hood et al., 2011; Rozin & Nemeroff, 1994). Positive donor types were described as possessing positive attributes (being highly intelligent) or performing positive acts (philanthropy). Negative donor types were described as being in negative environmental circumstances (homeless) or having performed negative acts (murder). Finally, respondents considered non-human (cross-species) donors, which we included in light of findings that animal species typically are highly essentialized (Atran & Medin, 2008; Gelman, 2003; Taylor et al., 2009) as well as people's reported discomfort with this type of donor as found in earlier studies (Coffman et al., 1998; Sanner, 2001a, 2001b). ----------------------------------------------Insert Table 1 about here -------------------------------------The use of cross-species donor types also provided an opportunity to test the different
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS predictions made by essentialism and unmediated minimalism (Strevens, 2000). Recall that
!"#
according to this view, two claims are central: First, essences are not systematically represented in people's kind-based inferences, and second, individuals appeal instead directly to kind membership as a basis for predicting kind-typical behavior (as kind membership is the necessary condition for "K-laws," or causal laws linking kinds with properties, to be in effect). On this view, respondents should judge that cross-species transplants would bring about changes in personality/behavior or luck only if they also judge that the transplant would bring about a change in category membership, because this interpretation of minimalism holds that it is category membership rather than essence that is believed to be the basis for people's predictions of outward behaviors. In contrast, an essentialist construal of organs would allow for transplants to result in outward changes in the recipient despite the recipient's original category membership remaining the same. To address the competing claims of essentialism and unmediated minimalism, we added a question for the cross-species items, If you received the transplant/transfusion from a (pig/chimpanzee), would you be a (pig/chimpanzee) after the transplant/transfusion? (We did not include this question for within-human transplants, because Strevens does not discuss social categories, and it is unclear how to extend the minimalist view to encompass them. For example, is a murderers negative behavior to be attributed to his/her membership in the social category murderers? This is a problematic explanation since one only comes to belong to this social kind by engaging in negative behavior in the first place namely by murdering someone. So it is unclear how one could make sense of unmediated minimalism for such categories.) We then analyzed responses on the questions assessing essentialist change only for those individuals who did not endorse category change. Based on the wealth of studies indicating the presence of
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS essentialism in other domains of reasoning, we expected our findings to support essentialism
!"#
rather than unmediated minimalism; that is, we expected that people would still endorse transfer of personality, behavior, or luck, even if they denied that the transplant would change their category membership. Transplant Types. To investigate whether participants would respond differentially to different kinds of transplants, we created three versions of the survey, each one asking about a distinct type of internal physical transfer: a blood transfusion, a DNA transplant, and a heart transplant. However, because receiving a heart from a family member would entail the family member's death, the family member as a potential donor for the heart transplant was omitted. Based on findings suggesting that adults believe essence to be widely distributed in the body (Newman & Keil, 2008), we expected that all three types of transfer would lead respondents to expect essence-based effects, i.e., score higher on the Essentialist measure for non-similar-to-self donor types relative to baseline similar-to-self types. Another reason to predict essentialist reasoning for all three transplant types can be found in the numerous expressions and metaphors that bear witness to essentialist beliefs about the heart (e.g., the seat of emotion, kindness, or morality) and blood (e.g., hot- or cold-blooded temperaments). However, given recent advances in biology and genomics, educated adults are well aware that DNA plays a part in causing outward properties and behaviors (though of course peoples understanding of the specifics may be sketchy or inaccurate; Mayr, 1982), and in light of this, we expected that people would be especially likely to expect a DNA transplant to have a causal influence. This pattern of thinking would lead to higher scores on the Essentialist measure for the DNA transplant relative to the heart transplant and blood transfusion scenarios. Recipient Identity. As a final manipulation, we asked respondents to consider transplants
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS given to either themselves or to their (hypothetical) infants. This manipulation was included to control for the possibility that, when considering themselves as the recipient, people might
!"#
believe that knowing they had received a transplant might cause them to act more like the donor, e.g., respondents might imagine that knowing they had received the organ of a murderer might cause them to behave more violently because they thought that they should. This would result in higher scores on the Essentialist measure, but would not constitute essentialist reasoning. The inclusion of infants (who would be unaware of the transplant and thus not prone to this "placebo" effect) allowed us to assess beliefs about transplants independent of this concern; no differences, however, were expected along this dimension in terms of its interaction with effects for donor or transplant type (the two major manipulations for which we expected main effects). Summary of Predictions Liking Ratings: Different-from-self, negative, and cross-species donor types will be liked less than similar-to-self (baseline) donor types; liking for positive donor types will be weakly or not differentiated from similar-to-self types. Such findings would in part replicate those found in previous assessments of moral contagion beliefs (Hood et al., 2011; Nemeroff & Rozin, 1994), and would replicate studies demonstrating distaste for receiving organs from non-human donors (e.g., Coffman et al., 1998; Sanner, 2001a, 2001b). Essentialist Measure: Essentialist beliefs about organ transplants will be revealed by higher scores on the Essentialist measure for other donor types relative to similar-to-self donor types. This will hold true regardless of whether the transplant type is blood, DNA, or heart, demonstrating that essence is construed as pervasive throughout the body, as indicated in past studies of psychological essentialism (e.g., Newman & Keil, 2008);
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS however, especially high scores were predicted for DNA, given its known (though typically unspecified) role in causing outward features. Essentialist Measure and Minimalism: Higher scores on the Essentialist measure for
!"#
cross-species transplant types relative to similar-to-self types will be observed even when respondents deny that receiving a transplant from a non-human animal will lead to the recipient changing category membership. That is, it will be the transferred essence per se that is represented as a causal force, rather than simply category membership being used as a basis for predicting category-typical features. Such findings would run counter to predictions from an unmediated reading of minimalism, which denies that people systematically represent causal essences in explaining how category-typical behaviors arise (Strevens, 2000). Non-Essentialist Measure: Non-essentialist styles of reasoning about transplants will be revealed by higher scores on the Non-Essentialist measure for different-from-self, negative, and cross-species donor types relative to similar-to-self donor types; in contrast, positive donor types may be weakly or not differentiated from similar-to-self for this style of reasoning. These predictions are based on findings from the moral contagion literature (e.g., Hood et al., 2011; Nemeroff & Rozin, 1994). 2.1 Method 2.1.1 Participants Respondents (n =104) were obtained using Amazon's Mechanical Turk, a crowd-sourcing platform that allows people to complete online tasks for compensation. We offered the survey only to individuals in the United States and compensated each respondent $1.85 for successful completion. Demographic information from these respondents (obtained through self-report) is
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS displayed in Table 2. Four additional respondents provided data that were discarded due to
!"#
incorrect answers to attention-check questions embedded in the main survey to ensure data was from actively participating respondents. Additionally, data from two surveys associated with the same IP address were discarded, as this indicated that a single person may have taken the survey twice. ----------------------------------------------Insert Table 2 about here -------------------------------------2.1.2 Survey materials and scoring Text of the survey questions appears in Table 3. Each survey began with a vignette stating that the recipient or the recipient's infant needed a transplant or transfusion for continued health. A list of donors was provided for respondents to evaluate, all of which were described as healthy, disease-free, and possessing compatible organs according to the respondent's doctor. We created three versions of the survey (blood, DNA, and heart), each describing the relevant transplant or transfusion. Infant-as-recipient vs. self-as-recipient was a within-subjects variable, whereas transplant type (blood, DNA, heart) was a between-subjects variable. An approximately equal number of respondents answered questions about each transplant type, and within each transplant type, approximately half answered questions about their infants as recipients first and the other half about themselves as recipients first (cell sizes ranged from n = 16 to n = 19). Text of all six scenarios (Blood to Infant, Blood to Self, DNA to Infant, DNA to Self, Heart to Infant, Heart to Self) appears in Appendix A. Following each vignette, respondents rank-ordered their preferences for all ten donors. This gave respondents the opportunity to consider all donors before responding to the target essentialist and non-essentialist questions. Feelings and beliefs about each donor were assessed individually. Donors were presented in a randomly determined fixed order. Following each donor, respondents provided a Likert-
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS scale rating of how much they liked the idea of receiving a transplant from each donor (Like),
!"#
then an open-ended explanation describing the reason behind their rating, and then responses to four Likert-scale questions addressing the target essentialist and non-essentialist reasons: Personality/Behavior (essentialist), Luck (essentialist), Creeped Out/Contaminated (nonessentialist), and Reputation (non-essentialist). Essentialist and non-essentialist questions appeared in one of four possible orders determined by a Latin-square design in approximately equal frequency within each transplant type (blood, DNA, and heart) and recipient type order (infant as recipient first, self as recipient first). Calculation of the Essentialist composite score was performed in two steps. First, the absolute change in luck from the midpoint value of 4 (= no change in luck) was calculated and then fit to a 1-7 point scale. (This approach allowed comparable values to be generated for both the Personality/Behavior and Luck items, with values of 1 always indicating minimal effects [i.e., either no change in personality/behavior or no change in luck], and 7 indicating maximal effects [i.e., the most change in personality/behavior or change in luck].) Thus, absolute values of 0, 1, 2, and 3 were converted to adjusted values of 1, 3, 5, and 7 respectively. Next, the average of the adjusted Luck score and the Personality/Behavior score was calculated, and this average was used as the Essentialist composite score. Calculation of the Non-Essentialist composite score was performed similarly by (1) calculating the absolute change in reputation from the midpoint value of 4 (= no change in how others viewed the recipient) and then fitting this to a 1-7 point scale and (2) taking the average of the adjusted Reputation score and the Creeped Out/Contaminated score. Cronbach's alpha was very high for both the composite Essentialist measure and the composite Non-Essentialist measure (>.9), thus validating the composite measures.
!!"
For the cross-species donor types only, a Category Change question came last (after the essentialist and non-essentialist questions), assessing whether cross-species transplants were thought to result in the recipient changing his/her category membership. Finally, two attention check items were also included, embedded within the main survey (To indicate that you are paying attention to this survey, please respond by marking 3/6 for this question). Respondents' data were discarded if they did not answer as requested. Demographic information was obtained after respondents provided their judgments about all donors. ----------------------------------------------Insert Table 3 about here -------------------------------------2.2 Results 2.2.1 Overview The data are presented in four main sections, below: Liking data, Essentialist composite, Test of minimalism, and Non-essentialist composite. Throughout, ratings for similar-to-self donor types were considered a baseline rating, with planned uncorrected pairwise comparisons conducted between similar-to-self (baseline) donor types and all other types (different-from-self, positive, negative, and cross-species). As noted earlier, the heart transplant item was not included for the family member donor trials, as we did not wish to ask about the death of a family member. Thus, the baseline for the heart transplant condition differed from the baseline for the other two conditions, in excluding that one donor. Accordingly, we also conducted supplementary analyses without the family member donor type, in order to determine whether the results held up when the same baseline was used for all three transplant conditions (reported in Table 4, footnote c). Finally, we present data from respondents' open-ended explanations. In the main analyses of the Likert-scale measures (Liking ratings, Essentialist composite measure, and Non-Essentialist composite measure), ANOVAs were employed using a mixed-
!"#
factorial 3 (transplant type: blood, DNA, heart) x 5 (donor type: similar-to-self, different-fromself, positive, negative, and cross-species) design, with transplant type as the between-subjects factor and donor type as the within-subjects factor. Preliminary analyses including the withinsubjects factor of recipient identity (infant vs. self) and the between-subjects factor of respondent gender demonstrated that neither participated in any interaction that reversed any main effects of donor type or transplant type. Given the complexity of the design, we thus present analyses collapsed across recipient identity and respondent gender. 2.2.2 Liking data: rank-ordered preferences and Likert-scale ratings Rank-Ordered Preferences. Because ranking data were ordinal and thus not suited for parametric comparisons, a Friedman's test was performed on respondents' rankings for donor types. We collapsed across transplant type (blood, DNA, heart) for simplicity, leaving transplant type to be explored in detail in the main parametric ANOVA analysis of Likert-scale Liking ratings. Rankings significantly differed from each other, Friedman's X2 (4) = 353.42, p < .001. Wilcoxon signed-ranks tests between similar-to-self and all other donor types revealed that similar-to-self types were preferred significantly more than all other donor types (different-fromself: Z = 8.74, positive: Z = 6.02, negative: Z = 8.82, cross-species: Z = 8.85, all ps < .001). Likert-Scale Ratings. The ANOVA (described above in the Results overview) revealed a main effect of donor type on liking ratings, F(4, 404) = 273. 56, p < .001,
p 2
=.73. Paired
comparisons of similar-to-self (baseline) types against all other types revealed that transplants from similar-to-self types were liked significantly more than transplants from all other types (Table 4), a pattern consistent with that revealed in the preference rankings. No other main effects or interactions were observed. 2.2.3 Essentialist composite
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS The ANOVA (see Results overview) revealed a main effect of donor type on the Essentialist composite measure, F(4, 404) = 17.99, p < .001,
p 2
!"#
= .15. Paired comparisons of
similar-to-self (baseline) types against all other types revealed that transplants from similar-toself types had lower scores than all other donor types, ps effect of transplant type, F(2, 101) = 3.18, p = .05,
p 2
= .06, with post-hoc Tukey comparisons
showing that DNA received significantly higher scores (M = 2.09, SD = .93) than blood (M = 1.57, SD = .70), p = .04, with scores for heart intermediate and not different from the other two (M = 1.82, SD = .96), ps > .05. No other main effects or interactions were observed; most importantly, transplant type did not interact with donor type (F(8, 404) = 1.35, p = .21), indicating that the main effect seen for donor type that was indicative of essentialism was not restricted to considerations of DNA transplants. American respondents thus broadly endorsed the possibility that transplants from a range of individuals could cause changes in the recipient's outward features or circumstances. This result provides evidence for essentialist reasoning. It is arguably the case that the Personality/Behavior item (i.e., To what extent do you think your personality or behavior might change to become more like that of this donor after the transplant/transfusion?) is the most direct measure of what is typically considered essentialist thinking. Although our main analysis focused on the composite Essentialist measure including the Luck item, we also conducted a secondary analysis to ensure that patterns remained the same when focusing exclusively on the more canonical Personality/Behavior item. Analyses did indeed replicate results seen for the composite Essentialist measure, namely showing (1) a significant main effect of donor type (F(4, 404) = 11.56, p < .001,
p 2
= .10) with paired
comparisons indicating that scores for all other donor types were significantly higher than for similar-to-self types, (2) a significant main effect of transplant type (F(1, 101) = 5.65, p = .005,

p 2
!"#
= .10), with post-hoc Tukey comparisons indicating that scores for DNA were higher than
blood (p = .003), with heart intermediate and not different from either, and (3) no interaction between donor type and transplant type (F(4, 404) = 1.29, p = .25). Thus, this secondary analysis is entirely consistent with the primary analysis of the composite item. Individual response patterns. The distribution of individual response patterns on the Essentialist measure is presented in Table 5, focusing on comparisons between similar-to-self and all other donor types. For each non-similar-to-self donor type, we present the proportion of respondents providing the following three response patterns: (1) an average score on the Essentialist measure that was lower for the non-similar-to-self type than the similar-to-self type (a difference inconsistent with essentialism), (2) an average score on the Essentialist measure that was equal for the non-similar-to-self and similar-to-self types (also inconsistent with essentialism), and (3) an average score on the Essentialist measure that was higher for the nonsimilar-to-self type than the similar-to-self type (consistent with essentialism). If individuals did not expect essentialist effects, the predicted distribution of individual responses would consist mostly of Pattern 2 responses (equal scores), with lower and equal numbers of Pattern 1 and Pattern 3 responses. If, on the other hand, individuals were employing essentialism, responses would be skewed toward Pattern 3 (more Pattern 3 responses than Pattern 1 responses). This was indeed the case; binomial tests comparing the number of individuals reporting higher scores on non-similar-to-self types vs. the number of individuals reporting lower scores on non-similar-toself types were significant for all four donor types, ps < .03. Response patterns indicating essentialism ranged from 31% of the time (for different-from-self types) to 52% of the time (for cross-species types). 2.2.4 Test of minimalism: Category change and essentialist reasoning
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Recall that according to an unmediated reading of the minimalist hypothesis (Strevens,
!"#
2000), people explain individuals' behavior based on an appeal to category membership, rather than a causal essence. This view would thus predict that changes in personality, behavior, or luck would be endorsed only in cases when respondents also endorsed change in category membership. Essentialism, however, allows for endorsement of change in personality, behavior, or luck absent endorsement of category change, because the causal agent--essence--can be transferred. To address these competing claims, we re-examined scores on the Essentialist composite measure as they related to scores on the category change questions (i.e., If you received the transplant/transfusion from a (pig/chimpanzee), would you be a (pig/chimpanzee) after the transplant/transfusion?). Scores on the Essentialist measure for cross-species and similar-to-self donor types were again compared, but this time including only scores from individuals who endorsed absolutely no category change for any cross-species transplants (n = 85). Thus, we examined whether people would display essentialist reasoning even when they wholly denied any possibility of the recipient changing category. Scores for similar-to-self were again considered baseline and compared against scores for the cross-species donor types in a 2 (donor type: similar-to-self vs. cross-species) x 3 (transplant type: blood, DNA, heart) ANOVA. A main effect of donor type was observed, F(1, 82) = 18.04, p < .001,
2 p =
.18, with
scores for cross-species types higher (M = 1.97, SD = 1.29) than for similar-to-self types (M = 1.45, SD = .74). There were no other main effects or interactions. Thus, respondents reported that a cross-species transplant could lead to outward changes in the recipient even when denying that that the transplant would change the recipient's category membership. This result is inconsistent with unmediated minimalism, and consistent with the idea that respondents expected transplants to transfer causal essence.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS 2.2.5 Non-essentialist composite The ANOVA (see Results overview) revealed a main effect of donor type on the NonEssentialist measure, F(4, 404) = 198.07, p < .001,
p 2
!"#
= .66. Paired comparisons of similar-to-
self (baseline) types against all other types revealed that transplants from similar-to-self types had significantly lower scores than all other donor types (Table 4), suggesting that respondents engaged in non-essentialist explanations for reasoning about why transplants from certain types are desirable or not. No other main effects or interactions were significant. 2.2.6 Open-ended explanations We identified seven non-mutually-exclusive categories of open-ended explanations: (1) Safety/Health (explanations that appealed to the safety of the procedure or health of the donor or recipient), (2) Morality (explanations that mentioned the moral status of the donor or the overall moral ramifications of the transplant procedure), (3) Creeped Out/Reputation (explanations that described feeling disgusted or creeped out, or beliefs that others might view the recipient differently; these are the same dimensions that contributed to the Non-essentialist composite measure addressed by close-ended questions), (4) Appeal to Category (explanations that offered the donor's category membership as the reason for a decision, apparently as if to convey that further explanation was unnecessary, (5) Essentialist (explanations that described the possibility of a transplant conferring traits or characteristics to the recipient; these correspond to the Essentialist composite measure addressed by close-ended questions), (6) Denial of Essentialism (explanations that contained explicit rejection of essentialist predictions), and (7) Unclear or Other (explanations that did not fall into any of the previous six categories). Prior to formal coding, any mention of specific donor types or transplant types was replaced with neutral "donor X" and "transplant X" text (e.g., "I worry that a heart from a man would be too big" was
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS modified to, "I worry that a transplant X from a donor X would be too big"). This was done so
!"#
that coders would not be biased to identify certain explanation styles preferentially according to donor or transplant type. A primary coder then classified responses into the seven explanation types. A second coder provided classifications for 20% of responses, and disagreements were resolved through discussion. Agreement for the categories ranged from very good to excellent, with Cohen's kappa ranging from .72 to .86 (average Cohen's kappa = .79). Sample explanations and classifications are displayed in Table 6, and proportion of respondents providing each type of explanation for the different donor types are displayed in Table 7. Explanations that appealed to safety and health (as well as unclear or other explanations) tended to predominate. However, explanations corresponding with the close-ended Non-Essentialist composite measures were also observed, and appeals to overall moral aspects and category membership were also fairly common. Finally, although respondents sometimes explicitly rejected essentialist effects in their explanations, essentialist explanations were also observed; in fact, a full 34.61% of respondents gave at least one essentialist explanation across all their responses. ---------------------------------------Insert Tables 4, 5, 6, and 7 about here ------------------------------2.3 Discussion The present study replicated past findings regarding distaste for transplants from negative, different, and cross-species donor types. We also demonstrated that respondents preferred transplants from similar-to-self types significantly more even than positive donor types. This effect was unexpected given that positive individuals are unlikely by themselves to elicit disgust or moral concerns (e.g., Nemeroff & Rozin, 1994) and suggests that, when thinking about bodily transplants, the respondents in this study most preferred donors who are most similar to themselves. Although we did not predict this finding, it is readily explained by
!"#
essentialism: if people think that receiving a transplant may lead to changes in their essence, then they might reasonably dislike the idea of receiving a transplant from anyone who differs from them, even along a positive dimension. That is, people dislike the prospect of any change in their essence positive or negative and so any salient difference between the donor and recipient leads to increased resistance to the transplant. In this way, peoples specifically essentialist reasoning about transplants differs from their more general way of reasoning about moral contagions. Regarding the underlying reasons for people's feelings about various donor types, we found evidence for the role of essentialism. As predicted, essentialism was seen most when reasoning about DNA transplants. However, it was also employed for other transplant types as well, namely for blood transfusions and heart transplants. Further, it was observed when respondents considered transplants both to their infants as well as themselves, suggesting that responses were not driven by expectations of a "transplant placebo" effect, whereby people would predict that knowing about the transplant would be the cause of any changes in the recipient. It is important to note that obtaining evidence for essentialism required that respondents actively reject formal knowledge about the function of bodily elements. Thus, it is very likely that the survey was a highly conservative measure of the pervasiveness and strength of the essentialist phenomenon. Indeed, average scores on the Essentialist measure were typically low (i.e., always under "3"). However, the effect size obtained for donor type (
p 2
= .15) was large
(Kinnear & Gray, 2004) and suggests that people were highly consistent in reporting essentialist effects for non-similar-to-self donors. Furthermore, respondents engaged in essentialist reasoning about all the different categories under consideration, including social categories as well as non-
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS human species. Individual response patterns on the Essentialist composite measure further
!"#
indicated the strength and pervasiveness of the effect; respondents were significantly more likely to provide scores for non-similar-to-self types that were higher than similar-to-self types (versus lower), adding more support to the claim that essentialist reasoning was often used when considering transplants from other individuals. As well, respondents' open-ended explanations also frequently invoked essentialist explanations (particularly for human donors differing from self). Finally, we also found that respondents engaged in non-essentialist styles of reasoning when considering the various donor types, reporting concerns about contamination and/or reputation change post-transplant. As with essentialist reasoning, non-essentialist reasoning was seen for considerations of transplants to both the respondents themselves and their infants, across a range of transplant types, and for donors that were both human and non-human. 3. Study 2 Study 1 demonstrated essentialist thinking about transfer of internal bodily elements in a sample of Americans. Although these data are indicative of systematically-held lay beliefs that transplants can have causal effects, left unanswered is whether these findings would be observed in both countries, or if these patterns of reasoning are particular to a Western sample. In Study 2, we expanded the focus of our study by targeting respondents in India so as to investigate possible cultural similarities and differences in essentialist (and non-essentialist) beliefs about transplants. India is an especially interesting country to examine because it differs from the United States on a number of dimensions that may affect how transplants are conceptualized. For instance, Hindu Indians typically are more concerned with contamination than Westerners -- a difference that has been attributed to a high degree of concern with maintaining purity, and a belief that it can be threatened by indirect as well as direct interpersonal contact (e.g., Hejmadi, Rozin, & Siegal,
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS 2004). Such concerns are prominently featured in dietary laws and customs that regulate what foods may be eaten (Olivelle, 1995), as well as rules regarding the caste system --reflected, for instance, by traditional restrictions of contact or association with the "untouchable" caste (Narula, 1999). In light of such pervasive contamination concerns, it is plausible that when thinking about bodily transplants, contamination-based reasoning (indicated by the NonEssentialist measure) might predominate, or even be employed to the exclusion of causalessentialist reasoning. Further, there are socio-cultural differences in the regulation of organ transplants that might also contribute to Indians showing stronger contamination concerns. Until 1995, paying donors for organs was legal in India, and commercial transplants were characterized by poor
!"#
oversight and improper screening of donors, in numerous cases resulting in the transmission of serious viruses such as HIV and hepatitis (Salahudeen et al., 1990). Although the practice was ultimately banned, paying for organs is still a relatively common practice in some areas of the country, with donations often coming from marginalized or impoverished individuals, reportedly often under duress (Jha, 2004; Kalbag, 2008). In contrast, commercial transplants in the United States have been illegal on the national level since 1984, and laws regulating organ transplants have been more strictly enforced, with infractions occurring very rarely (Panjabi, 2011). In light of this, we predicted that Indians would report higher levels of discomfort with organ transplants, particularly from the "negative" donor type, either because these types are seen as victimized (leading to feelings of guilt), or because they are seen as more unclean (leading to feelings of contamination). It is therefore likely that such feelings would contribute to higher scores on both of the measures of non-essentialist reasoning, namely the "creeped out/contaminated" and "reputation" items.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Thus, given Indians' predicted heightened focus on contamination, as well as specific
!"#
socio-cultural factors surrounding how organ transplants have been regulated and conducted, one might expect an exclusive focus on non-essentialist explanations in an Indian sample. On the other hand, it is also plausible that evidence for essentialism would still be found amongst the Indian respondents, largely replicating the findings from the American sample. As discussed earlier, non-essentialist explanations can be entertained independent of essentialist explanations, meaning that it is entirely possible that respondents would engage in both styles of reasoning simultaneously. As well, ample literature suggests that essentialist thinking is found across varied cultures (Atran, 1990, 1998; Atran & Medin, 2008; Gil-White, 2001; Hirschfeld, 1996, 2001; Waxman, Medin, & Ross, 2007), including Indian cultures (Mahalingam, 2003; Mahalingam, Haritatos, & Jackson, 2007). Given that essentialism in other domains has been found to be so widespread, we predicted that essentialist thinking about organ transplants would also be found cross-culturally, and thus would be detectable amongst Indians as well as Americans. In order to assess this prediction, we employed the same survey and analyses that were used with Americans. Again, we were interested in examining the degree to which essentialist and non-essentialist styles of reasoning were employed when considering different potential donors. 3.1 Method 3.1.1 Participants Respondents (n = 140) were obtained using Amazon's Mechanical Turk. We offered the survey only to individuals in India and compensated each respondent $1.85 for successful completion. Demographic information from these respondents is displayed in Table 2. Sixty-
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS eight additional respondents provided data that were discarded due to incorrect answers to the
!!"
attention-check questions. The proportion of respondents failing attention-check questions was much higher in this sample than in the American sample. This is likely due to the higher proportion of individuals in India using Mechanical Turk as a primary source of income rather than a secondary or tertiary source, as is common in the U.S. (Ross, Irani, Silberman, Zaldivar, & Tomlinson, 2010); this difference likely increases volume but decreases quality across respondents. Whatever the cause for the discrepancy, answering both attention-check questions correctly by chance would be predicted only 2% of the time, and thus answers from retained individuals are highly likely to be from engaged individuals with proficient English. Additionally, we discarded data from five pairs of surveys that had duplicate IP addresses within each pair, as this indicates a likelihood that these surveys were taken by the same individual twice. 3.1.2 Survey materials and scoring The survey and scoring were identical to that given to American respondents in Study 1. Respondents answered questions about each type of transplant or transfusion; the number of respondents answering questions about heart transplants was higher (n = 56) than the other two transplant types (blood n = 44, DNA n = 40). Within each transplant type, approximately half responded to questions about an infant as recipient first, and the other half responded to questions about the self as recipient first (cell sizes ranged from n = 18 to n = 24 in the non-heart conditions; in the heart condition, n = 25 [infant first] and n = 31 [self first]). 3.2 Results Analyses exploring the predicted main effects of donor type and transplant type on rankordered preferences, Likert-scale liking ratings, Essentialist scores, and Non-essentialist scores
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS were identical to those used for the American respondents in Study 1, namely non-parametric analyses of rank-ordered preferences and ANOVAs for the remaining measures featuring a 3
!"#
(transplant type: blood, DNA, heart) x 5 (donor type: similar-to-self, different, positive, negative, cross-species) design. As with American respondents, the infant vs. self as recipient variable and respondent gender did not participate in any interactions that reversed any significant main effects of donor or transplant type. Thus, we collapsed across these variables for the final analyses. 3.2.1 Liking data: Rank-ordered preferences and Likert-scale ratings Rank-Ordered Preferences. Rankings significantly differed from each other, Friedman's X2 (4) = 398.94, p < .001. Wilcoxon signed-ranks tests between similar-to-self and all other donor types revealed that rankings for similar-to-self and positive types did not differ (Z = .68, p > .05), but similar-to-self types were significantly preferred over all other donor types (differentfrom-self: Z = 9.72, negative: Z = 9.17, cross-species: Z = 10.12, all ps < .001). Likert-Scale Ratings. The ANOVA (see Study 1 Results overview) revealed a main effect of donor type on liking ratings, F(4, 548) = 372.16, p < .001,
p 2
= .73. Scores for similar-to-self
types and positive types were not significantly different; however, transplants from similar-toself types were liked significantly more than from the other three types (different-from-self, negative, and cross-species), a pattern consistent with that seen in the preference rankings (Table 4). There was also a main effect of transplant type, F(2, 137) = 4.08, p = .02,
p 2
= .06. Tukey
post-hoc tests indicated that DNA donations were liked less (M = 4.14, SD = 1.67) than heart transplants (M = 4.63, SD = 1.30), p = .02, whereas blood transfusions (M = 4.52, SD = 1.54) were intermediate and not different from the other two, ps > .05. Finally, these main effects were qualified by a donor type x transplant type interaction, F(8, 548) = 5.07, p < .001,
p 2
= .07.
!"#
Because our main prediction centered on the main effect of donor type, to examine the nature of this interaction we again compared all other donor types against similar-to-self types within each transplant type. Importantly, comparisons of similar-to-self against all other types within each transplant type revealed the same overall pattern described for the main effect of donor type, with the single exception that transplants from similar-to-self types were significantly more liked than transplants from positive types in the blood condition (t(43) = 3.62, p = .001) (whereas they were not differentiated in the overall main effect). 3.2.2 Essentialist composite measure The ANOVA (see Study 1 Results overview) revealed a main effect of donor type on the Essentialist composite measure, F(4, 548) = 44.49, p < .001,
p 2
= .25. Paired comparisons of
similar-to-self (baseline) types against all other types revealed that transplants from similar-toself types had significantly lower scores than all other donor types (Table 4), indicating that Indian respondents endorsed essentialist reasoning. There was also a main effect of transplant type, F(2, 137) = 4.25, p = .02,
p 2
= .06, with Tukey post-hoc comparisons indicating that DNA
transplants (M = 3.29, SD = 2.21) received higher scores than blood transfusions (M = 2.52, SD = 2.11), p = .03, and heart transplants (M = 2.72, SD = 1.87), p = .04. There were no other main effects or interactions; most importantly, there was no interaction between donor type and transplant type, indicating that the main effect of donor type indicative of essentialism was not restricted to DNA. Thus, there was clear evidence for essentialist reasoning in Indian respondents, which was especially strongly (but not exclusively) observed for beliefs about DNA transplants. As in Study 1, we also examined scores on the Personality/Behavior item considered separately (a more canonical measure of essentialism) in order to determine if patterns were
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS similar on this item to patterns on the composite Essentialist measure. The results were largely
!"#
replicated, with findings indicating (1) a main effect of donor type (F(4, 548) = 16.40, p < .001,
p 2
= .16) with paired comparisons indicating that scores for similar-to-self types were lower than
2 p
for all other donor types and (2) a main effect of transplant type (F(1, 137) = 9.70, p < .001, = .12) with post-hoc Tukey comparisons indicating that scores for DNA were higher than for both heart and blood (ps < .01), whereas heart and blood did not differ from each other.
However, unlike results for the Essentialist composite score, there was a significant donor type x transplant type interaction, F(4, 548) = 2.63, p = .03,
p 2
= .04. Paired-samples t tests revealed
that scores on similar-to-self were not significantly different from either different-from-self (p = .12) or negative (p = .10) donor types within heart transplants. All other comparisons, however, were significant, with similar-to-self receiving lower scores. Thus, aside from a slight attenuation of the main effect of donor type within heart transplants, results were consistent with the primary analysis of the composite essentialism item; most importantly, patterns consistent with essentialist thinking were shown for a variety of donor types and across a range of transplant types.# Individual response patterns. The distribution of responses on the Essentialist measure are presented in Table 5, again (as with Americans) focusing on comparisons between similar-toself and all other donor types. Responses consistent with essentialism were seen for all nonsimilar-to-self donor types, with individuals' scores consistent with essentialism (i.e., higher relative to similar-to-self) ranging from almost 44% (for different-from-self types) to 70% (for cross-species types). Binomial tests comparing the number of individuals reporting higher scores on non-similar-to-self types vs. the number of individuals reporting lower scores on non-similarto-self types were significant for all four donor types, ps < .001.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS 3.2.3 Test of minimalism: Category change and essentialist reasoning People's responses to the essentialist questions relative to responses on the category change questions were again compared to assess the claims of the unmediated minimalist approach. Scores on the Essentialist measure were compared for cross-species and similar-toself donor types using data only from respondents who had uniformly rejected the idea of a transplant causing a change in category membership (n = 74) (the same analysis used for Americans). There was a main effect of donor type, F(1, 71) = 24.47, p < .001,
p 2
!"#
= .26, with
scores for cross-species types higher (M = 3.07, SD = 1.81) than for similar-to-self types (M = 2.10, SD = 1.29). There were no other main effects or interactions. Indian respondents therefore showed essentialist reasoning regarding cross-species transplants even when they did not expect the transplant to exert any change whatsoever in the recipient's category membership. 3.2.4 Non-essentialist composite measure The ANOVA (see Study 1 Results overview) revealed a main effect of donor type on the Non-Essentialist measure, F(4, 548) = 201.94, p < .001,
p 2
= .60. Paired comparisons of similar-
to-self (baseline) types against all other types showed that similar types received significantly lower scores than all other types (Table 4). There was also a main effect of transplant type, F(2, 137) = 4.21, p = .02,
p 2
= .06. Tukey post-hoc comparisons indicated that DNA transplants (M =
3.76, SD = 2.16) received higher scores than both blood transfusions (M = 3.12, SD = 2.04), p = .04, and heart transplants (M = 3.14, SD = 1.80), p = .02. Finally, there was a significant donor type x transplant type interaction, F(8, 548) = 3.80, p < .001,
p 2
= .05. Because our main
prediction centered on the main effect of donor type, to examine the nature of this interaction we again compared all other donor types against similar-to-self types within each transplant type. The pattern of results was largely the same as the overall main effect, with the exception that for
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS heart transplants, scores for similar-to-self types were no longer different from positive types (t(55) = 1.65, p = .11) or different-from-self types (t(55) = 1.61, p = .11). Thus, on the whole, there was still strong evidence for non-essentialist reasoning in Indian respondents for a wide
!"#
range of donor types, with only slight attenuation of this effect observed within heart transplant scenarios. 3.2.5 Open-ended explanations Coding of open-ended explanations was identical to the procedure used for American responses. Agreement between two coders on 20% of the responses ranged from good to excellent, with Cohen's kappa ranging from .67 to .81 (average kappa = .74). As with Americans, concerns with safety and health (as well as unclear or other types of explanations) predominated, but essentialist explanations were observed as well, and in fact were more commonly seen in Indians' responses (52.14% of Indians gave at least one essentialist explanation across all their responses, compared to about a third of Americans). Appeals to the morality of the donor were also notably prominent (Table 7). 3.3 Discussion Patterns of preferences and liking were remarkably similar to those seen in Americans. In particular, similar-to-self donor types were typically preferred over other types of donors (different-from-self, negative, and cross-species types), though for Indian respondents, this difference was not significant when considering positive types. Evidence for essentialist thinking similar to that of Americans was also revealed in the Indian sample. Respondents endorsed the possibility of outward characteristics (personality/behavior or luck) changing upon receiving the transplant to become more like those of the donor. This pattern of reasoning was seen both for human donor types who differed from the recipient as well as cross-species donor types,
!"#
indicating that, as with Americans, essentialist construals of transplants extended beyond other species to other types of people. Finally, Indians also indicated non-essentialist concerns with transplants, reporting they would feel contaminated and/or fear social disapproval if they were to receive transplants from non-similar-to-self types. Taken together, the data indicate that the Indian sample entertained both causal and non-causal styles of explanations when reasoning about transplants from other individuals. 4. Study 3 In both Study 1 (with Americans) and Study 2 (with Indians), respondents provided higher scores on the Essentialist measure for all other donor types relative to donors similar to themselves. We interpret this pattern as constituting evidence for essentialist reasoning about transplants, because it suggests that people expect transplants to transfer aspects of a donor to a recipient. Alternatively, however, the data may reflect a broader magical thinking style, in which individuals expect that any kind of transfer from one person to another has the capacity to transfer aspects of the original owner to the new owner. Although this would be a noteworthy finding in its own right, this would not suggest that respondents represented essence as exclusively internal, bodily, and transferable, but rather that transplants are one of many ways in which the transfer of anything at all can confer traits of a donor on a recipient. To address this alternative explanation, we created a control study in which respondents (both American and Indian) were asked to consider receiving money rather than a transplant from the various donor types. Money was described as being transferred from the donor's account to the recipient's account, again without any direct personal interaction. In light of our original hypothesis -- namely that Studies 1 and 2 reflected people's tendency to represent
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS essence as a causal force within the body -- we predicted the absence of essentialist thinking in the money transfer scenario. 4.1 Method 4.1.1 Participants
!"#
Respondents were obtained using Amazon's Mechanical Turk (American n = 34; Indian n = 29). The number of respondents was approximately equal to that of each of the bodily transplants in the prior studies. We offered the survey only to individuals in the United States (for the American sample) and India (for the Indian sample) and compensated each respondent $1.85 for successful completion. Data from two additional American respondents and seventeen additional Indian respondents were discarded due to incorrect answers to the attention-check questions. Additionally, we discarded data from the Indian sample from nine surveys that had duplicate IP addresses, as this indicates a likelihood that these surveys were taken by the same individual more than once. 4.1.2 Survey materials and scoring Surveys opened with a vignette describing a money donation. The donation was described as helping to pay for an expensive and necessary medical treatment, an aspect that allowed the necessity and urgency of the donation to be roughly equivalent to that of the transplant scenarios. The amount of money was specified as $2,000 for the American sample and 100,000 rupees for the Indian sample; amounts were approximately equal based on the dollar-torupee exchange rate the month the study was run. The remaining content of the survey was otherwise identical to that of Studies 1 and 2 except that animal donor types were excluded, as animals obviously do not possess money and thus could not serve as potential donors. (As well, because animals were excluded, no category change questions were asked, as these had
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS originally been provided only for the animal donor types.) The eight human donors were again classified into the same similar-to-self, different-from-self, positive, and negative donor types. Approximately half of the respondents within each country sample (American and Indian) responded to questions about an infant as recipient first, and the other half responded to questions about the self as recipient first (cell sizes ranged from n = 14 to n = 17). Scoring and calculation of the Essentialist and Non-Essentialist measures were identical to Studies 1 and 2. Exact wording of money transfer vignettes appears in Appendix A. 4.2 Results 4.2.1 Overview Infant- vs. self-as-recipient and gender of respondent did not participate in any
!"#
interactions that reversed any donor type main effects found in preliminary analyses. They were thus removed from the primary reported analyses. As in Studies 1 and 2, preference rankings were analyzed by a non-parametric Friedman's test, and American and Indian samples were analyzed separately. Likert-scale measures (Liking ratings, Essentialist scores, and NonEssentialist scores) were analyzed in separate one-way within-subjects ANOVAs examining the factor of donor type (similar-to-self, different-from-self, positive, and negative). A final analysis compared essentialist beliefs across Studies 1, 2, and 3, thereby making comparisons based on country (U.S. vs. India) as well as transplant type (bodily vs. non-bodily money control). 4.2.2 Liking data: Rank-ordered preferences and Likert-scale ratings Rank-Ordered Preferences. For Americans, rankings significantly differed from each other, Friedman's X2 (3) = 80.41, p < .001. Wilcoxon signed-ranks tests between similar-to-self and all other donor types revealed that rankings for similar-to-self and positive types did not differ (Z = .41, p > .05), but similar-to-self types were significantly preferred over the other two
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS donor types (different-from-self: Z = 4.46, negative: Z = 5.10, both ps < .001). For Indians, rankings also significantly differed from each other, Friedman's X2 (3) = 67.33, p < .001. Wilcoxon signed-ranks tests between similar-to-self and all other donor types indicated that
!"#
rankings for similar-to-self and positive types again did not differ (Z = .93, p > .05), but similarto-self types were significantly preferred over the other two donor types (different: Z = 4.63, negative: Z = 4.51, both ps < .001). Americans and Indians thus showed the same pattern, significantly preferring similar-to-self donor types over different and negative types, but not differentiating between similar-to-self and positive donor types. Likert-Scale Ratings. For Americans, the ANOVA revealed a main effect of donor type on liking ratings, F(3, 99) = 90.86, p < .001,
p 2
= .73. Scores for positive and similar-to-self
types were not differentiated; however, donations from similar-to-self types were liked significantly more than those from the other two types (different-from-self and negative) (Table 8). An identical pattern was seen in the Indian sample; there was a main effect of donor type on liking ratings, F(3, 84) = 109.50, p < .001,
p 2
= .80, with scores for positive and similar-to-self
types not differentiated, and donations from similar-to-self types liked significantly more than those from the other two types (different-from-self and negative) (Table 8), a pattern consistent with that observed in the preference rankings. 4.2.3 Essentialist composite measure For Americans, there was no significant main effect of donor type on the Essentialist measure, F(3, 99) = 1.71, p = .17. We nevertheless conducted paired comparisons of similar-toself (baseline) types against all other types, which revealed that scores for similar-to-self types were not significantly different from either positive or negative types, and in fact were significantly higher than different-from-self types (the opposite direction of that obtained in
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Study 1) (Table 8). A similar non-significant effect was seen for donor type in Indians as well, F(3,84) = 1.13, p = .34. We similarly conducted paired comparisons of similar-to-self types against the other three types, which showed that scores for similar-to-self types did not differ from any other type (Table 8). Thus, the prediction that money donations would not result in causal-essentialist thinking was supported. Examination of Personality/Behavior scores alone (the more canonical measure of
!"#
essentialism) similarly did not reveal evidence of essentialist reasoning about money. Unlike the analysis of the composite Essentialist measure, there was a main effect for donor type in the American sample, F(3, 99) = 6.35, p = .001,
p 2
= .16. However, t tests comparing similar-to-self
baseline types to all other donor types indicated that similar-to-self types received significantly higher scores than different-from-self and negative donor types, again the opposite direction of that obtained in Study 1, and inconsistent with essentialist reasoning about money. There was no main effect of donor in the Indian sample, F(3, 84) = 1.01, p = .39. Individual response patterns. Examination of the distribution of responses on the Essentialist measure further indicate that money donations were not essentialized. Distributions are presented in Table 9, again focusing on comparisons between similar-to-self and all other donor types (as was done in Studies 1 and 2). Binomial tests comparing the number of individuals reporting higher scores on non-similar-to-self types (a pattern indicative of essentialism) vs. the number of individuals reporting lower scores on non-similar-to-self types (a pattern indicating lack of essentialism) were largely non-significant, ps > .05, with the single exception that significantly more Americans provided lower scores for different-from-self types (p < .01), which is opposite to the pattern obtained in Study 1 and inconsistent with essentialist reasoning on this control task.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS 4.2.4 Non-essentialist composite measure For Americans, there was a main effect of donor type on the Non-Essentialist measure, F(3, 99) = 47.57, p < .001,
p 2
!!"
= .59. Paired comparisons of similar-to-self (baseline) types
against all other types showed that similar-to-self types were not differentiated from positive or different-from-self types, but received significantly lower scores than negative types (Table 8). A similar pattern of results was seen in Indians; there was a main effect of donor type, F(3, 84) = 14.43, p < .001,
p 2
= .34, with paired comparisons indicating that similar-to-self types again
were not differentiated from either positive or different-from-self types, but received significantly lower scores than negative types (Table 8). These patterns are different from those obtained in Studies 1 and 2, in which similar-to-self donor types were differentiated from all other donor types. 4.2.5 Open-ended explanations A primary coder assessed respondents' explanations for essentialist predictions, as the main question motivating this study was whether respondents would expect money donations to be essentialized. An independent coder coded 20% of responses. Agreement was perfect (kappa = 1.0). Only two open-ended responses showed evidence of essentialism (one from negative types, in an Indian respondent, and one from positive types, in an American respondent), accounting for less than 1% of responses. 4.2.6 Essentialism across countries and transplant types Data Treatment. Compared to American respondents, Indian respondents gave overall higher values for all questions contributing to the composite essentialism and non-essentialism measures (ps < .01); furthermore, scores specifically for baseline similar-to-self donors were also higher for Indian respondents (ps < .01), suggesting a non-equivalence in baseline scores
!"#
between countries. This pattern is consistent with prior observations of overall higher agreement (irrespective of question type) on the part of Indian vs. American online participants (Schulze, Seedorf, Geiger, Kaufmann, & Schader, 2011) and prevented meaningful direct comparisons of scores across the two samples. Thus, to correct for this difference, we calculated scores for each participant reflecting the difference between the non-similar-to-self donor essentialism scores relative to the similar-to-self baseline essentialism scores. Note that these scores involve the essentialism composite measure only, and not any of the other measures (Liking or Nonessentialist scores). These difference-from-baseline (DFB) scores were analyzed using a country (U.S. vs. India) x donor type (positive, different, negative) x transplant condition (bodily vs. nonbodily money control) ANOVA. (Species donor types were not included in this analysis, as no species donors were provided for the money control condition.) Analysis. Notably, there was a main effect for transplant condition, F(1, 303) = 9.97, p = .002,
p 2
= .03, with DFB scores for bodily transplants higher (M = .29, SD = .77) than for non-
bodily money donations (M = -.06, SD = .84); this is consistent with the idea that bodily transplants but not money would be essentialized. There was also a significant main effect for donor, F(2, 606) = 13.88, p < .01,
p 2
= .04. Tukey post-hoc comparison indicated that positive
DFB scores (M = .33, SD = .87) and negative DFB scores (M = .26, SD = 1.03) were both significantly higher than different DFB scores (M = .08, SD = .83, ps < .01), but were not different from each other, p = .14. This result suggests that positive and negative donor types were essentialized more than different-from-self types (but recall from analyses performed for each country separately in Studies 1 and 2 that all three types were essentialized to some degree, evidenced by all three types receiving significantly higher scores than the similar-to-self
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS baseline). There were no other main effects or interactions; most notably, country did not participate in any significant effects. Finally, because cross-species donors were not included in the money control and thus were excluded from the prior analysis, we also compared the two countries in essentializing
!"#
cross-species donors within the bodily transplant conditions alone. Again, the analyses involved the essentialism composite measure only. Indian respondents' DFB scores were significantly higher (M = 1.16, SD = 1.57) than American respondents' (M = .68, SD = 1.26), t(242) = 2.56, p = .01, Cohen's d = .34. -----------------------------------------Insert Tables 8 and 9 about here ---------------------------------4.3 Discussion In Studies 1 and 2, we argued that people's endorsement of non-similar-to-self donors transferring characteristics to recipients through transplants reflected psychological essentialism. Study 3 was designed to rule out an alternative explanation for these findings, namely that respondents in Studies 1 and 2 may have been relying on a broader magical belief that any kind of transfer, not just internal bodily elements, might confer donors' traits on recipients. Study 3 served as a control to rule out this possibility by asking respondents to consider a money donation rather than a transplant. Both Americans and Indians did not expect money transfer to result in the recipient taking on characteristics of the donor, as indicated by scores on the Essentialism measure for non-similar-to-self types being largely equivalent to scores for similarto-self types. Respondents instead endorsed causal-essentialist effects (transfer of personality/behavior and/or luck) only when considering transfer of internal body parts (Studies 1 and 2), not merely any kind of transfer (such as money). Overall, Study 3 strengthens our claim
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS that Studies 1 and 2 were indicative of genuine essentialist thinking on the part of both Americans and Indians. Analysis of difference-from-baseline (DFB) scores provides further evidence for the
!"#
claim that bodily transplants are uniquely essentialized. Specifically, DFB scores were higher for bodily transplants than for non-bodily monetary transfer, consistent with the belief that bodily donations, but not monetary donations, may confer donors' characteristics to recipients. Thus, results from this analysis again indicate that in both countries, respondents differentiated the causal implications of receiving bodily vs. non-bodily donations; the former was essentialized, whereas the latter was not. Notably, U.S.-India differences were typically not observed when examining difference-from-baseline scores in the analysis of Studies 1, 2, and 3; the sole exception to this was that Indian respondents essentialized cross-species donors more than Americans. 5. Study 4 Results across the first three studies indicate that internal bodily elements, but not monetary donations, are seen as possessing a causal force that can be transferred via organ transplants. Although this strongly suggests that internal bodily elements are uniquely essentialized, monetary donations differ from bodily transplants in a number of ways that raise questions as to exactly which characteristics are necessary for something to be essentialized. Money is both external and acquired, typically on a temporary basis, and thus differs from a heart, blood, or DNA, each of which is an internal and inherent or biologically-endowed part of someone's body. Must an element be both internal and biological for it to be seen as capable of transferring causal essence? We designed Study 4 to investigate whether the internal vs. external and biological vs.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS acquired dimensions are relevant when making essentialist predictions. We compared individuals' responses to questions regarding three types of transfers: (1) a heart transplant,
!"#
similar to that described in the prior studies, in which the donated element was thus both internal and biological, (2) a pacemaker exchange, in which the pacemaker from one individual was described as being put into the recipient's body and was thus an internal but acquired nonbiological element, and (3) a skin graft, in which a portion of an individual's skin was described as being grafted onto the recipient and was thus an external but biological element. As in Studies 1-3, we examined individuals in both the United States and India to compare patterns of essentializing along this dimension. Given that Studies 1-3 had already established that individuals essentialize the heart using similar-to-self donors as a baseline, we did not assess similar-to-self donors in Study 4. Instead, we expanded the range of possible donors within positive and negative types, with the plan of comparing essentializing across the three donor conditions (heart, transplant, and skin graft). We also modified the survey structure relative to Studies 1-3 in two ways. First, we asked respondents about donations to themselves only, rather than to both themselves and their infants, because the infant vs. self dimension did not participate in any effects of theoretical interest in the prior studies. Second, we shortened the survey by including only a Likert-scale liking item, an open-ended explanation/justification item, a single essentialism question, and a single nonessentialism question. The liking item and open-ended explanation request were identical to those in Studies 1-3 (Table 3). To provide respondents with a broader range of options for the essentialism and non-essentialism questions, we asked respondents for a percentage rather than a score on a 1-7 Likert scale. This change allowed for finer-grained analysis of the presence of these constructs. The question assessing essentialism was modeled after the
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Personality/Behavior item from Studies 1-3 and asked, "How possible is it that your personality/behavior might change to become more like that of this donor because of the graft? Please provide a value ranging from 0% to 100%, where 0% indicates this is
!"#
completely impossible and 100% indicates this is completely possible." The non-essentialism question was modeled after the Creeped Out/Contaminated item from Studies 1-3 and asked, "How 'creeped out' or 'contaminated' would you feel if you received your graft from this donor? Please provide a value ranging from 0% to 100%, where 0% indicates you would not feel at all creeped out or contaminated and 100% indicates you would feel completely creeped out or contaminated." 5.1 Method 5.1.1 Participants Respondents were obtained using Amazon's Mechanical Turk (American n = 136, Indian n = 253). 2 We offered the survey to individuals in the United States (for the American sample) and India (for the Indian sample) and compensated each respondent $1.49 for successful completion. Data from 14 additional American respondents and 91 additional Indian respondents were discarded due to incorrect answers to the attention-check questions. Additionally, we discarded data from eight surveys with duplicate IP addresses from the American sample, and from 31 surveys that had duplicate IP addresses from the Indian sample. Finally, respondents with average essentialism scores greater than three standard deviations outside their respective condition means (within each country) were removed from analysis. This excluded three ######################################################## $#Approximately twice as many Indian respondents were obtained due to higher variability in their responses.
Standard deviations within each transfer type condition were approximately two to three times higher in the Indian vs. American sample.#
!"#
American participants (one each from the heart, pacemaker, and skin graft conditions); no Indian respondents had outlying data. 5.1.2 Survey materials Respondents read descriptions about one of three transfers: heart transplant (American n = 49; Indian n = 82), pacemaker exchange (American n = 47, Indian n = 87), or skin graft (American n = 40, Indian n = 84). Each respondent answered questions about a total of 12 donors, with six positive and six negative donors (Table 1). Text of the vignettes describing the transfers appears in Appendix A. Respondents saw four questions about each donor. The liking item and open-ended explanation request appeared first, followed by the Personality/Behavior and Creeped Out/Contaminated items in counterbalanced order. As well, approximately half the participants saw one randomly-determined order of donors, whereas the other half saw this order reversed. Order of Personality/Behavior and Creeped Out/Contaminated and donor order were crossed, and cell sizes within the four versions created by this crossing were approximately equal (American: heart n = 11-14, pacemaker n = 11-13, skin graft n = 9-11; Indian: heart n = 18-23, pacemaker n = 19-28, skin graft n = 19-23). 5.2 Results For the sake of brevity, we focus solely on analyses of the Personality/Behavior and Creeped Out/Contaminated questions. For each question, a mixed-factorial 3 (transfer type: heart, pacemaker, skin graft) x 2 (donor type: positive vs. negative) ANOVA was used, with transfer type as a between-subjects variable and donor type as a within-subjects variable. As in Studies 1-3, scores from Indian respondents on the essentialism item were substantially higher for the overall average relative to Americans (ps < .01). Although this may genuinely reflect
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS greater endorsement of essentialism, it is also possible that the difference reflects the same general tendency to provide higher values seen in Indian respondents relative to American
!"#
respondents in Studies 1-3. This possibility prevented direct comparison between countries; thus, analyses for the two groups were performed separately. American Sample. For the Personality/Behavior item, there was a main effect of transfer type, indicative of differences in essentializing according to transfer condition, F (2, 133) = 4.83, p = .01,
p 2
= 07. Tukey post-hoc comparisons indicated that a heart transplant was judged more
likely to transfer personality or behavior than either a pacemaker exchange or a skin graft, ps < .03; scores for pacemaker exchange and skin graft did not differ, however (p > .05) (Table 10). Finally, the donor type x transfer condition interaction was not significant, p > .05. There was also a main effect of donor, F(1, 133) = 5.93, p = .02,
p 2
= .04, whereby respondents reported
that positive donors would be more likely to transfer personality or behavior (M = 6.87%, SD = 13.91) than negative donors (M = 3.83%, SD = 7.93). For the Creeped Out/Contaminated item, there was a main effect of donor, F(1,133) = 61.36, p < .001,
p 2
= .32, whereby respondents indicated they would feel more creeped out or
contaminated by donations from negative donors (M = 28.03%, SD = 27.86) than positive donors (M = 14.28%, SD = 26.63). No other main effects or interactions were significant, ps > .05 (see Table 10 for means according to transfer type condition). Indian Sample. For the Personality/Behavior item, there was a main effect of donor, F(1, 250) = 4.18, p =.04,
p 2
= .02, whereby respondents reported that positive donors would be more
likely to transfer personality or behavior (M = 25.62%, SD = 28.88) than negative donors (M = 22.61%, SD = 25.32). No other main effects or interactions were significant, p > .05; most notably, this included no main effect for transfer type condition (see Table 10 for means),
!"#
indicating no difference in endorsement of essentialism depending on the three types of transfers. For the Creeped Out/Contaminated item, there was a main effect of donor, F(1, 250) = 57.20, p < .001,
p 2
= .19, whereby respondents indicated they would feel more creeped out or
contaminated by donations from negative donors (M = 33.38%, SD = 28.43) than positive donors (M = 19.99%, SD = 25.77). No other main effects or interactions were significant, ps > .05 (see Table 10 for means according to transfer type condition). ----------------------------------------------Insert Table 10 about here -------------------------------------6. General Discussion A significant body of work suggests that psychological essentialism contributes powerfully to the ways that we think about category structure, license and constrain inductive inference, and make judgments about the causes of outward behaviors and features of social, natural, and other kinds (Gelman, 2003; Medin, 1989; Medin & Ortony, 1989). The current studies assessed whether essentialist beliefs also affect people's reasoning about bodies and transfers of internal bodily elements, in particular whether people believe that receiving a transplant can cause a recipient to take on characteristics of the donor. Psychological essentialism predicts this pattern of belief, because internal parts are taken to be suffused with essence. Thus, when an internal part is transferred from one individual to another, some of the donors essence is also transferred, and so will have a casual influence on the recipient. These studies constitute the first systematic, direct investigation of these essentialist beliefs in the general public. We found clear evidence for essentialist thinking: Respondents frequently endorsed the possibility of organ transplants or blood transfusions conferring a donor's characteristics, behaviors, or outward circumstances on the recipient.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Further, our data indicate not just the presence of essentialist thinking, but also point to
!"#
other aspects of its prevalence and nature. We found essentialist thinking in two distinct samples, Americans and Indians. The presence of essentialist thinking in two different countries supports the idea that essentialism is a bias that obtains across widely varying environmental inputs. As well, participants endorsed essentialist effects not just following a DNA transplant, but also following a heart transplant or blood transfusion. Additionally, Indians (though not Americans) expected similar effects after transfer of other types of bodily elements that had long-term associations with the donor (pacemaker, skin graft). Notably, however, a control study (Study 3) indicated that causal effects were not deemed plausible when considering money donations, for either Indians or Americans. Taken together, these data indicate two additional important features of essentialism. First, causal essence is not promiscuously located in just anything that is associated with an individual (e.g., money). Second, causal essence is located in parts that biologists would argue are not, in fact, causally implicated in outward behaviors or qualities (such as the heart or blood). This latter finding is especially striking, because such a reasoning error indicates the operation of an essentialist bias independent of knowledge about the potential causal role of DNA, and suggests that respondents often reject formal biological knowledge in favor of essentialism. (It is worth noting that responses were not universally indicative of essentialism; indeed, many individuals provided response patterns and open-ended explanations indicating explicit lack of belief in essentialist effects of transplants, suggesting that some people favored formal biological knowledge over essentialism. This is perhaps unsurprising; the explicit measures used in the current study were almost certainly conservative in their ability to detect essentialism.)
!"#
We suggest that it is people's representation of essence per se that is responsible for the effects. This is an important point, because the very existence of an essentialist bias has been questioned. As described in the Introduction, the minimalist approach argues for an alternative explanation for most of the findings in studies on essentialism thus far (Strevens, 2000). Specifically, Strevens argues that people do not represent an essence as a causal force in drawing inferences about an individual's features. Instead, according to the strongest interpretation of this position, people appeal directly to category membership as the basis on which to predict outward features. Our studies demonstrate, in contrast, that people who did not expect a transplant to result in a change in category membership (e.g., from human to pig) nevertheless endorsed the possibility that the transplant could change the recipient's category-typical behaviors to be more like those of the animal donor. Thus, category membership was not construed as necessary for outward features, because causal effects were deemed possible even in the absence of category change. These findings thus run counter to the predictions of unmediated minimalism. The findings reflect that people appeal to something more than just non-specific kind-property links (K-laws) in drawing these sorts of inferences. Our discussion has focused thus far on unmediated minimalism, but what of mediated minimalism, according to which people appeal to causal intermediaries in their reasoning? In order to account for our data, these causal intermediaries would have to have a range of characteristics: at the very least, they would have to be the sorts of things that can be transferred via organ transplants, and that exert causal influence on the recipient; they would have to be inherent; and they would have to be, at least for American adults, internal and biological. Once the nature of these causal intermediaries is constrained in this way, it becomes unclear how this
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS view differs from psychological essentialism 3. Psychological essentialism is characterized as a
!!"
placeholder notion (Medin & Ortony, 1989), indicating that it does not attribute to people any specific beliefs about the nature of essence. Placeholder essentialism thus claims simply that people believe in an inherent, causal something-or-other, which is intimately connected with kind membership, but which is also separable from it. To suppose that belief in causal intermediaries with these characteristics is a distinct view from psychological essentialism is to saddle essentialism with more theoretical commitments than it in fact carries. If mediated minimalism is thus to be a distinct view, it would have to be far more carefully delineated in the literature than it has been thus far. The current studies also speak to a related literature examining people's avoidance of contact with morally negative individuals and the items with which these individuals have associated, i.e., magical contagion studies. A number of studies in this line of research have found extreme distaste for even indirect contact with morally negative entities (e.g., Nemeroff & Rozin, 1994), findings that were recently extended by work indicating that people also dislike the idea of receiving organs from such individuals (Hood et al., 2011). The present studies add to """""""""""""""""""""""""""""""""""""""""""""""""""""""" #"To clarify, philosophical discussions of essence often invoke something more elaborate and specific than what is implied by psychological essentialism. For example, when Kripke (1980) or Putnam (1975) speak of natural kinds having essences, they mean to claim in part that there is a particular property, possession of which is both necessary and sufficient for being a member of the kind. Thus, being H2O is often claimed to be the essence of water, in the sense that all and only things that are H2O count as water this is just what it is, metaphysically or scientifically speaking, for something to be water. For additional discussion of different kinds of essentialism, see Gelman, 2003, Chapter 1 and Leslie, in press b."
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS this literature by showing the same distaste for organ transplants from negative donor types, those different from oneself, or different species. Importantly, our studies revealed specific reasons for this distaste: People are often alarmed by thoughts of essence-based changes in behavior or characteristics, as well as by feelings of contamination and social disapprobation. Our data also provide insight into cultural differences and similarities in people's thinking. Striking similarities across the two countries were observed. Both Americans and Indians reported liking transplants from animals the least (replicating past studies suggesting particular distaste for animals as donors, e.g., Coffman et al., 1998; Sanner, 2001a, 2001b), followed by negative donor types, then different-from-self types, and finally positive types. Respondents from both countries also gave cross-species and negative donor types high scores on both the essentialist and non-essentialist measures, suggesting that reasons for distaste for dissimilar or negative entities are shared across these two cultural contexts.
!"#
On the other hand, cross-country differences were also revealed. First, Indian respondents appeared to essentialize cross-species donors more than Americans, as indicated by a greater difference between scores for these types and baseline similar-to-self types. Second, Indian respondents also appeared to essentialize more types of transfers than Americans; specifically, they did not differentiate essentialist predictions for heart, pacemaker, or skin graft transfer. It may be that Americans and Indians possess distinct concepts of essence, with Indians treating a broader range of elements as having causal properties. Alternatively, American and Indian representations of essence may be fundamentally similar, but Indians may be more likely to expect internal essences to be transferred to objects or parts as a result of extensive direct contact with the individual (see Nemeroff & Rozin, 1994, for discussion of this latter form of essentialism). Further research is needed to determine which interpretation is valid, as well as the
!"#
factors that contribute to people's tendency to extend essentialist effects to non-internal or nonbiological elements. The present investigation still leaves open a number of other important issues. For example, we found evidence for essentialist thinking about a wide range of social categories. How exactly are these essentialist beliefs to be modeled? Haslam and colleagues have proposed that, in the social domain, essentialism includes two highly dissociable (though not mutually exclusive) dimensions: naturalness and entitativity (Haslam et al., 2000). They argue that some essentialized social groups tend to be conceived of as more natural or biologically based (e.g., race, gender), whereas other essentialized social groups tend to be conceived of as more entitative, with members highly similar to one another (e.g., homosexuals). One might predict that it is only categories deemed biologically-based that would also be considered viable candidates for transferring bodily causal essence, yet our data suggest that essentialism is influential for more than just categories considered to be natural or biologically based such as gender or species. Results from our study may thus be seen as indicating the existence of pseudobiological conceptualizations of a wide range of social categories, a notable finding that warrants further investigation. Questions for future research include: what is the causal route by which people expect essences from various categories to take their effects, and does this differ depending on how natural a category is considered? Our data also point to potential questions regarding the distinction between categorybased essence vs. individual-based essence. Here, we argue that transplants from various social and animal categories may be predicted to transfer category-typical attributes including personality and behavior. Personality and behavior, however, are also arguably in equal measure components of one's individual identity; how might psychological essentialism function when
!"#
considering what causal forces underlie unique individuals? A number of researchers have noted potential parallels as well as areas of disconnect between essentializing individuals vs. categories (Blok, Newman, & Rips, 2005; Gelman & Hirschfeld, 1999; Gutheil, Gelman, Klein, Michos & Kelaita, 2008; Gutheil & Rosengren, 1996; Leslie, in press b; Rips, Blok, & Newman, 2006), yet the precise relationship between the two has not been thoroughly established. Future work within the domain of transplants may contribute to resolving this; for instance, is a transfer of essence via a transplant expected to compromise or alter the identity of either the donor or the recipient? If so, does this draw on the same essentialist intuitions that were revealed in the current study? Also unanswered is how essentialist beliefs about bodies develop. A large body of literature indicates the presence of essentialist beliefs from an early age, and a number of studies indicate that children expect essence to be located internally (Gelman, 2003). Might such essentialist thinking be observed when asking children to consider the effects of transplants (e.g., by asking children to consider the results of trading a heart with a pig)? Studies by Johnson (1990) as well as Gelman, Gottfried, and Schulz (1999) provide intriguing initial evidence that bears on the question of how children reason about transplants. In both sets of studies, children were asked to consider the outcomes of trading brains with various individuals. Children showed developmental progression in the ability to recognize how the brain functions to contain thoughts and memories. Johnson additionally compared children's responses to brain transplant scenarios and other body parts (heart and mouth). Although children by early elementary school were likely to expect the traded brain to be the most important factor in determining the recipient's later behavior and thinking, they also showed some evidence of expecting the other transplants to result in causal changes as well; for instance, the majority of children expected a heart from a mean person would reduce the recipient's kindness. Future work can extend these findings by
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS examining precisely what kinds of donors (e.g., members of a wide range of social groups as well as animals), body parts (e.g., heart vs. blood), and traits (e.g., kindness, intelligence, etc.) might be essentialized in the contexts of transplant scenarios.
!"#
Finally, our studies may shed some light on the concerns that real-life transplant patients may have before and after their procedures. Specifically, our results point to the possibility that patients may entertain non-scientifically-based worries; despite the absence of medical evidence to suggest that organ transplants or blood transfusions can cause changes in personality or life circumstances, our respondents still considered such outcomes possible. Belief in such possibilities, we suggest, is rooted in essentialism a deep and pervasive cognitive bias.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS References
!"#
Ahn, W., Kalish, C., Gelman, S. A., Medin, D. L., Luhmann, C., Atran, S., ... Shafto, P. (2001). Why essences are essential in the psychology of concepts. Cognition, 82, 59-69. Ahn, W., Kim, N. S., Lassaline, M. E., & Dennis, M. J. (2000). Causal status as a determinant of feature centrality. Cognitive Psychology, 41, 361-416. Allport, G. W. (1954). The nature of prejudice. Reading, MA: Addison-Wesley. Atran, S. (1990). Cognitive foundations of natural history. Cambridge, UK: Cambridge University Press. Atran, S. (1998). Folk biology and the anthropology of science: Cognitive universals and cultural particulars. Behavioral and Brain Sciences, 21, 547-569. Atran, S., & Medin, D. (2008). The native mind and the cultural construction of nature. Boston, MA: MIT Press. Astuti, R., Solomon, G., & Carey, S. (2004). Constraints on conceptual development: A case study of the acquisition of folkbiological and folksociological knowledge in Madagascar. Monographs of the Society for Research in Child Development, 69, vii-135. Basch, S. H. (1973). The intrapsychic integration of a new organ: A clinical study of kidney transplantation. Psychoanalytic Quarterly, 42, 364-384. Bastian, B., & Haslam, N. (2006). Psychological essentialism and stereotype endorsement. Journal of Experimental Social Psychology, 42, 228-235. Belk, R. W. (1990). Me and thee versus mine and thine: How perceptions of the body influence organ donation and transplantation. In J. Shanteau & R. J. Harris (Eds.), Organ donation and transplantation: Psychological and behavioral factors (pp. 139-149). Washington, DC: American Psychological Association.
!"#
Birnbaum, D., Deeb, I., Segall, G., Ben-Eliyahu, A., & Diesendruck, G. (2010). The development of social essentialism: The case of Israeli children's inferences about Jews and Arabs. Child Development, 81, 757-777. Blok, S. V., Newman, G. E., & Rips, L. J. (2005). Individuals and their concepts. In W. K. Ahn, R. L. Goldstone, B. C. Love, A. B. Markman, & P. Wolff (Eds.), Categorization inside and outside the lab. Washington D. C.: American Psychological Association. Callan, M. J., Ellard, J. H., & Nicol, J. E. (2006). The belief in a just world and immanent justice reasoning in adults. Personality and Social Psychology Bulletin, 32, 1646-1658. Conan Doyle, A. (1923). The adventure of the creeping man. In Sir Arthur Conan Doyle, The Case- Book of Sherlock Holmes, 1927; republished 2001, Cornwall, UK: House of Stratus. Coffman, K. L., Sher, L., Hoffman, A., Rojter, S., Folk, P., Cramer, D., ... Makowka, L. (1998). Survey results on transplant patients' attitudes on xenografting. Psychosomatics, 39, 379383. Dewar, K., & Xu, F. (2009). Do early nouns refer to kinds or distinct shapes? Psychological Science, 20, 252-257. Diesendruck, G., & Haber, L. (2009). God's categories: The effect of religiosity on children's teleological and essentialist beliefs about categories. Cognition, 110, 100-114. Diesendruck, G., & haLevi, H. (2006). The role of language, appearance, and culture in children's social category-based induction. Child Development, 77, 539-553. Feeley, T. H., & Moon, S. (2009). A meta-analytic review of communication campaigns to promote organ donation. Communication Reports, 22, 63-73.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Frazier, B. N., Gelman, S. A., Wilson, A., & Hood, B. (2009). Picasso paintings, moon rocks, and hand-written Beatles lyrics: Adults evaluations of authentic objects. Journal of Cognition and Culture, 9, 114. Gelman, S. A. (2003). The essential child: Origins of essentialism in everyday thought. New York: Oxford University Press. Gelman, S. A., & Coley, J. D. (1990). The importance of knowing a dodo is a bird: Categories and inferences in 2-year-old children. Developmental Psychology, 26, 796-804.
!"#
Gelman, S. A., & Gottfried, G. D. (1996). Causal explanations of animate and inanimate motion. Child Development, 67, 1970-1987. Gelman, S. A., & Heyman, G. D. (1999). Carrot-eaters and creature-believers: The effects of lexicalization on childrens inferences about social categories. Psychological Science, 10, 489-493. Gelman, S. A., & Hirschfeld, L. A. (1999). How biological is essentialism? In D. L. Medin & S. Atran (Eds.), Folkbiology (pp. 403-446). Cambridge, MA: MIT Press. Gelman, S. A., & Kremer, K. E. (1991). Understanding natural cause: Children's explanations of how objects and their properties originate. Child Development, 62, 396-414. Gelman, S. A., & Markman, E. M. (1986). Categories and induction in young children. Cognition, 23, 183-209. Gelman, S. A., & Markman, E. M. (1987). Young childrens inductions from natural kinds: The role of categories and appearances. Child Development, 58, 1532-1541. Gelman, S. A., Taylor, M. G., & Nguyen, S. (2004). Mother-child conversations about gender: Understanding the acquisition of essentialist beliefs. Monographs of the Society for Research in Child Development, 69, vii-126.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Gelman, S. A., & Wellman, H. M. (1991). Insides and essences: Early understandings of the nonobvious. Cognition, 38, 213-244. Gil-White, F. J. (2001). Are ethnic groups biological "species" to the human brain?
!"#
Essentialism in our cognition of some social categories. Current Anthropology, 42, 515554. Gottfried, G. M., & Gelman, S. A. (2005). Developing domain-specific causal-explanatory frameworks: The role of insides and immanence. Cognitive Development, 20, 137-158. Gottfried, G. M., Gelman, S. A., & Schultz, J. (1999). Children's understanding of the brain: From early essentialism to biological theory. Cognitive Development, 14, 147-174. Graham, S. A., Kilbreath, C. S., & Welder, A. N. (2004). Thirteen-month-olds rely on shared labels and shape similarity for inductive inferences. Child Development, 75, 409-427. Gutheil, G., Gelman, S. A., Klein, E., Michos, K., & Kelaita, K. (2008). Preschoolers' use of spatiotemporal history, appearance, and proper name in determining individual identity. Cognition 107, 366-380. Gutheil, G., & Rosengren, K. S. (1996). A rose by any other name: Preschoolers' understanding of individual identity across name and appearance changes. British Journal of Developmental Psychology, 14, 477-498. Haslam, N., Bastian, B., Bain, P., & Kashima, Y. (2006). Psychological essentialism, implicit theories, and intergroup relations. Group Processes and Intergroup Relations, 9, 63-76. Haslam, N., Rothschild, L., & Ernst, D. (2000). Essentialist beliefs about social categories. British Journal of Social Psychology, 39, 113-127.
!"#
Hayward, C., & Madill, A. (2003). The meanings of organ donation: Muslims of Pakistani origin and white English nationals living in North England. Social Science and Medicine, 57, 389-401. Hejmadi, A., Rozin, P., & Siegal, M. (2004). Once in contact, always in contact: Contagious essence and conceptions of purification in American and Hindu Indian children. Developmental Psychology, 40, 467-476. Hirschfeld, L. A. (1996). Race and making: Cognition, culture, and the child's construction of human kinds. Cambridge, MA: MIT Press. Hirschfeld, L. A. (2001). On a folk theory of society: Children, evolution, and mental representations of social groups. Personality and Social Psychology Review, 5, 107-117. Hood, B. M., Gjersoe, N. L., Donnelly, K., Byers, A., & Itajkura S. (2011). Moral contagion attitudes towards potential organ transplants in British and Japanese adults. Journal of Culture and Cognition, 11, 269-286. Inspector, I., Kutz, Y., & David, D. (2004). Another person's heart: Magical and rational thinking in the psychological adaptation to heart transplantation. The Israeli Journal of Psychiatry and Related Sciences, 41, 161-173. Jha, V. (2004). Paid transplants in India: The grim reality. Nephrology Dialysis Transplantation, 19, 541-543. Johnson, C. N. (1990). If you had my brain, where would I be? Children's understanding of the brain and identity. Child Development, 61, 962-972. Johnson, C. N., & Jacobs, M. G. (2001, April). Enchanted objects: How positive connections transform thinking about the very nature of things. Poster presented at the Society for Research in Child Development, Minneapolis, MN.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Kalbag, S. (2008, February 2). India's kidney racket story goes global. Hindustan Times. retrieved from http://www.hindustantimes.com/India-s-kidney-racket-story-goesglobal/Article1-273439.aspx Keil, F. C. (1989). Concepts, kinds, and cognitive development. Cambridge, MA: MIT Press.
!"#
Keller, J. (2005). In genes we trust: The biological component of psychological essentialism and its relationship to mechanisms of motivated social cognition. Journal of Personality and Social Psychology, 88, 686702. Kinnear, P. R., & Gray, C. D. (2004). SPSS 12 made simple. New York: Psychology Press. Kripke, S. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press. Lederer, S. E. (2008). Flesh and blood: Organ transplantation and blood transfusion in 20th century America. New York: Oxford University Press. Leslie, S.-J. (in press a). The original sin of cognition: Fear, prejudice, and generalization. The Journal of Philosophy. Leslie, S.-J. (in press b). Essence and natural kinds: When science meets preschooler intuition. Oxford Studies in Epistemology, 4. Mahalingam, R. (2003). Essentialism and beliefs about gender among Aravanis, the "not-men" of Tamilnadu. Sex Roles, 49, 489-496. Mahalingam, R., Haritatos, J. , & Jackson, B. (2007). Essentialism and the cultural psychology of gender in extreme son preference communities in India. American Journal of Orthopsychiatry, 77, 598-609. Mayr, E. (1982). The growth of biological thought. Cambridge, MA: Harvard University Press. McIntosh, J. (2009). The edge of Islam: Power, personhood, and ethnoreligious boundaries on the Kenya coast. Durham, NC: Duke University Press.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS McLean, S. A. M., & Williamson, L. (2005). Xenotransplantation: Law and ethics. Aldershot: Ashgate Publishing Ltd.
!!"
Medin, D. L. (1989). Concepts and conceptual structure. American Psychologist, 44, 1469-1481. Medin, D. L., & Ortony, A. (1989). Psychological essentialism. In S. Vosniadou & A. Ortony (Eds.) Similarity and analogical reasoning (pp. 179-196). New York, NY: Cambridge University Press. Narula, S. (1999). Broken people: Caste violence against India's "Untouchables." Human Rights Watch. Nemeroff, C., & Rozin, P. (1989). "You are what you eat": Applying the demand-free "impressions" technique to an unacknowledged belief. Ethos, 17, 50-69. Nemeroff, C., & Rozin, P. (1994). The contagion concept in adult thinking in the United States: Transmission of germs and interpersonal influence. Ethos, 22, 158186. Newman, G. E., Diesendruck, G. & Bloom, P. (2011). Celebrity contagion and the value of objects. Journal of Consumer Research, 38, 215-228. Newman, G. E., & Keil, F. C. (2008). Wheres the essence? Developmental shifts in childrens beliefs about internal features. Child Development, 79, 1344-1356. Olivelle, P. (1995). Food in India. Journal of Indian Philosophy, 23, 367-380. Olson, K. R., Banaji, M. R., Dweck, C. S., & Spelke, E. S. (2006). Children's bias against lucky vs. unlucky people and their social groups. Psychological Science, 17, 845-846. Panjabi, R. K. L. (2011). The sum of a human's parts: Global organ trafficking in the twenty-first century. Pace Environmental Law Review, 28, 1-144. Prentice, D. A., & Miller, D. T. (2007). Psychological essentialism of human categories. Current Directions in Psychological Science, 16, 202-206.
!"#
Putnam, H. (1975). The meaning of meaning. In H. Putnam, Mind, language, and reality (pp. 215-271). Cambridge: Cambridge University Press. Rips, L. J., Blok, S. V., & Newman, G. E. (2006). Tracing the identity of objects. Psychological Review, 113, 1-30. Rhodes, M., & Gelman, S. A. (2009). A developmental examination of the conceptual structure of animal, artifact, and human social categories across two cultural contexts. Cognitive Psychology, 59, 244-274. Ross, J., Irani, L., Silberman, M. S., Zaldivar, A., & Tomlinson, B. (2010). Who are the Crowdworkers? Shifting demographics in Mechanical Turk. In Extended abstracts of CHI 2010 (ACM Conference on Human Factors in Computing Systems), pp. 2863-2872. Rothbart, M., & Taylor, M. (1992). Category labels and social reality: Do we view social categories as natural kinds? In G. Semin and K. Fiedler (Eds.), Language, interaction and social cognition (pp. 11-36). London: Sage. Rozin, P., Markwith, M., & McCauley, C. (1994). Sensitivity to indirect contacts with other persons: AIDS aversion as a composite of aversion to strangers, infection, moral taint, and misfortune. Journal of Abnormal Psychology, 103, 495-505. Rozin, P., Millman, L., & Nemeroff, C. (1986). Operation of the laws of sympathetic magic in disgust and other domains. Journal of Personality and Social Psychology, 50, 703-712. Salahudeen, A. K., Woods, H. F., Pingle, A. Suleyman, M. N., Shakuntala, K., Nandakumar, M., ... Daar, T. M. (1990). High mortality among recipients of bought living-unrelated donor kidneys. Lancet, 336, 725-728. Sanner, M. A. (2001a). Exchanging spare parts or becoming a new person? People's attitudes toward receiving and donating organs. Social Science and Medicine, 52, 1491-1499.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Sanner, M. A. (2001b). People's feelings and beliefs about receiving transplants of different origins--questions of life and death, identity, and nature's border. Clinical Transplantation, 15, 19-27.
!"#
Schulze, T., Seedorf, S., Geiger, D., Kaufmann, N., & Schader, M. (2011, June). Exploring task properties in crowdsourcing: An empirical study on Mechanical Turk. In Proceedings of the 19th European Conference on Information Systems (ECIS), Paper 122. Smiler, A., & Gelman, S. A. (2008). Determinants of gender essentialism in college students. Sex Roles, 58, 864-874. Strevens, M. (2000). The essentialist aspect of naive theories. Cognition, 74, 149-175. Strevens, M. (2001). Only causation matters: Reply to Ahn et al.. Cognition, 82, 71-76. Sylvia, C. (with Novak, W.) (1997). A change of heart: A memoir. New York: Time Warner. Taylor, M. G., Rhodes, M., & Gelman, S. A. (2009). Boys will be boys; cows will be cows: Children's essentialist reasoning about gender categories and animal species. Child Development, 80, 461-481. Waxman, S. R., Medin, D. L., & Ross, N. (2007). Folkbiological reasoning from a cross-cultural developmental perspective: Early essentialist notions are shaped by cultural beliefs. Developmental Psychology, 43, 294-308.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Appendix A: Text of Vignettes for Each Transplant Type and Recipient Studies 1-3 Blood Transplant, Infant Recipient Imagine that you have a baby boy/girl, and doctors have found a problem with his/her
!"#
blood that is resulting in serious health problems. These doctors have determined that your baby needs a blood transfusion. The doctors have identified a variety of possible donors, all of whom are physically disease-free and healthy. The doctors have determined that all the possible donors are compatible and thus are confident the transfusion will be successful. The donations will be anonymous, and neither you nor the donor will ever need to meet, nor will your donor ever see your baby. You and your baby's doctor meet to discuss the upcoming transfusion. The doctor gives you a list of the potential donors for your baby and asks you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your baby's donor. Please fill out this questionnaire as if this situation were real. Blood Transfusion, Self Recipient Imagine that doctors have found a problem with your blood that is resulting in serious health problems. These doctors have determined that you need a blood transfusion. The doctors have identified a variety of possible donors, all of whom are physically disease-free and healthy. The doctors have determined that all the possible donors are compatible and thus are confident the transfusion will be successful. The donations will be anonymous, and neither you nor the donor will ever need to meet. You and your doctor meet to discuss the upcoming transfusion. The doctor gives you a list of the potential donors and asks you to fill out a questionnaire indicating your preferences
!"#
and thoughts about these donors. Your responses will be used to determine the selection of your donor. Please fill out this questionnaire as if this situation were real. DNA Transplant, Infant Recipient Imagine that you have a baby boy/girl 4, and doctors have found a faulty gene in your baby's body that is resulting in serious health problems. These doctors have determined that your baby needs some DNA from another individual transplanted into his/her body. This new DNA will contain a healthy version of the gene. The doctors have identified a variety of possible donors, all of whom are physically disease-free and healthy. The doctors have determined that all the possible donors are compatible and thus are confident the DNA transplant will be successful. The donations will be anonymous, and neither you nor the donor will ever need to meet, nor will your donor ever see your baby. You and your baby's doctor meet to discuss the upcoming transplant. The doctor gives you a list of the potential donors for your baby and asks you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your baby's donor. Please fill out this questionnaire as if this situation were real. DNA Transplant, Self Recipient Imagine that doctors have found a faulty gene in your body that is resulting in serious health problems. These doctors have determined that you need some DNA from another individual transplanted into your body. This new DNA will contain a healthy version of the gene. The doctors have identified a variety of possible donors, all of whom are physically disease-free and healthy. The doctors have determined that all the possible donors are compatible and thus are confident the DNA transplant will be successful. The donations will be anonymous, and ######################################################## 4 Sex of infant recipient matched self-reported sex of respondent
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS neither you nor the donor will ever need to meet.
!"#
You and your doctor meet to discuss the upcoming transplant. The doctor gives you a list of the potential donors and asks you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your donor. Please fill out this questionnaire as if this situation were real. Heart Transplant, Infant Recipient Imagine that you have a baby boy/girl, and doctors have found a problem with his/her heart that is resulting in serious health problems. These doctors have determined that your baby needs a heart from another individual transplanted into his body. Also imagine that medical advances now allow for the transplant of an adult's heart into an infant's body; your infant will have adequate room in the chest cavity to accommodate the heart, and it will function normally in your baby's body. The doctors have identified a variety of possible donors, all of whom are physically disease-free and healthy. The doctors have determined that all the possible donors are compatible and thus are confident the heart transplant will be successful. The donations will be anonymous, and neither you nor the donor will ever need to meet, nor will your donor ever see your baby. You and your baby's doctor meet to discuss the upcoming transplant. The doctor gives you a list of the potential donors and asks you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your baby's donor. Please fill out this questionnaire as if this situation were real. Heart Transplant, Self Recipient Imagine that doctors have found a problem with your heart that is resulting in serious health problems. These doctors have determined that you need a heart from another individual
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS transplanted into your body. The doctors have identified a variety of possible donors, all of
!"#
whom are physically disease-free and healthy. The doctors have determined that all the possible donors are compatible and thus are confident the heart transplant will be successful. The donations will be anonymous, and neither you nor the donor will ever need to meet. You and your doctor meet to discuss the upcoming transplant. The doctor gives you a list of the potential donors and asks you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your donor. Please fill out this questionnaire as if this situation were real. Money Donation, Infant Recipient Imagine that you have a baby boy/girl, and his doctor has diagnosed him with a disease that is resulting in serious health problems. The doctor has determined that your baby needs an expensive treatment in order to get better. To help you pay for the treatment, your doctor's office asks members of the community to donate money. A variety of possible donors are located, all of whom are willing to have $2000.00/100,000 rupees transferred from their bank accounts to an account set up for your treatment. The donations will be anonymous, and neither you nor the donor will ever need to meet, nor will your donor ever see your baby. You meet with a representative of your baby's doctor's office, who asks you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your baby's donor. Please fill out this questionnaire as if this situation were real. Money Donation, Self Recipient Imagine that your doctor has diagnosed you with a disease that is resulting in serious health problems. The doctor has determined that you need an expensive treatment in order to get better.
!"#
To help you pay for the treatment, your doctor's office asks members of the community to donate money. A variety of possible donors are located, all of whom are willing to have $2000.00/100,000 rupees transferred from their bank accounts to an account set up for your treatment. The donations will be anonymous, and neither you nor the donor will ever need to meet. You meet with a representative of your doctor's office, who asks you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your donor. Please fill out this questionnaire as if this situation were real. Study 4 Heart Transplant Imagine that doctors have found a problem with your heart that is resulting in serious health problems. You meet with the doctors, and one of them informs you,"You need a heart transplant. In other words, we will replace the heart you have now with the heart from another individual." The doctors have identified a variety of possible donors, all of whom were in good physical health, but unfortunately recently suffered a fatal car accident. The doctors have determined that all the possible donors are compatible and thus are confident the heart transplant will be successful. The donations will be anonymous. You and the doctors meet to discuss the upcoming transplant. They give you a list of the potential donors and ask you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your donor. Please fill out this questionnaire as if this situation were real. Pacemaker Exchange
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Imagine that doctors have found a problem with your heart that is resulting in serious health problems. You meet with the doctors, and one of them informs you,"You need a
!"#
pacemaker. In other words, you need a device that helps make your heart beat at a regular rate. A pacemaker is put inside the body near the heart in a surgical procedure. Your pacemaker will be taken from another individual who no longer needs it." The doctors have identified a variety of possible donors, all of whom were in good physical health, but unfortunately very recently suffered a fatal car accident. The doctors have determined that the pacemakers from all the possible donors are functioning and compatible with your body and thus are confident the pacemaker will work successfully. The donations will be anonymous. You and the doctors meet to discuss the upcoming surgery. They give you a list of the potential donors and ask you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your donor. Please fill out this questionnaire as if this situation were real. Skin Graft Imagine that doctors have found a problem with the skin around your elbow that is resulting in serious health problems. You meet with the doctors, and one of them informs you, "You need a skin graft. In other words, you need some new skin surgically attached to your elbow. Your skin will be taken from the elbow of another individual who no longer needs it." The doctors have identified a variety of possible donors, all of whom were in good physical health, but unfortunately very recently suffered a fatal car accident. The doctors have determined that the skin from all the possible donors is compatible with your body and thus are confident the skin graft will be successful. The donations will be anonymous.
!"#
You and the doctors meet to discuss the upcoming skin graft. They give you a list of the potential donors and ask you to fill out a questionnaire indicating your preferences and thoughts about these donors. Your responses will be used to determine the selection of your donor. Please fill out this questionnaire as if this situation were real.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 1 Donor types and individual donors in each type for Studies 1-3 and Study 4
Donor Type Studies 1-3 Similar-To-Self (baseline) Different-From-Self Positive Negative Cross-Species Study 4 (described as same gender) Positive Donor
a
!"#
same gender, blood-related family member same gender, sexual orientation, age, and background as respondent different gender, same age and background as respondent different sexual orientation, same gender, age, and background as respondent same gender, high IQ same gender, well known for philanthropy and charity work same gender, convicted of violent murder same gender, homeless pig chimpanzee
Negative
a person with a high IQ charity worker who gave money to good causes kind person liked by many talented artist talented mathematician talented musician homeless person person with a low IQ schizophrenic thief compulsive gambler violent murderer
Actual terminology specifying gender of donors depended on respondents' self-reported gender, obtained prior to survey administration. "Same gender" was termed "man" or "male" for male respondents and "woman" or "female" for female respondents; "Different gender" was termed "woman" or "female" for male respondents and "man" or "male" for female respondents; for instance, text for a male respondent for the murderer donor was "man convicted of violent murder"
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 2 Demographic characteristics of respondents in Studies 1-4
Study 1 Americans (n=104) 44.2 55.8 Study 2 Indians (n=140) 66.3 33.7 Study 3 Americans (n=34) 23.5 76.5 Study 3 Indians (n=29) 75.9 24.1 Study 4 Americans (n=136) 44.9 55.1 Study 4 Indians (n=253) 67.6 32.4
!!"
Demographics Sex (%) Male Female Age (years) Mean (SD) Religion (%) Atheist/Agnostic Christian Jewish Hindu Muslim Other or Unspecified
30.8 (9.8)
36.9 (12.7)
36.2 (12.3)
27.8 (10.1)
37.7 (11.87)
31.4 (10.31)
29.8 59.6 4.8 3.8 1.9 0.0
0.0 16.4 0.0 66.4 10.7 6.4
27.3 65.2 6.1 0.0 0.0 1.4
0.0 6.9 0.0 72.4 3.4 17.3
25.0 66.9 2.2 0.0 0.1 5.9
1.2 16.3 0.0 68.4 9.9 4.2
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 3 Text of Survey Questions

Item Preference Like Open-ended Essentialist: Personality/ Behavior Content / Scale
a
!"#
The following is a list of potential donors. Please rank order the options according to how much you would like to receive your transplant from each individual. Would you like to receive your transplant/transfusion from this potential donor? 1-7: 1=Definitely No, 4=Neutral, 7=Definitely Yes Please provide a brief explanation describing your reasoning. (no scale) To what extent do you think your personality or behavior might change to become more like that of this donor after the transplant/transfusion? 1-7: 1=Definitely Would Not Change, 4=Not Sure, 7=Definitely Would Change To what extent do you think your luck might change if you received your transplant/transfusion from this donor? 1-7: 1=Become Much Less Lucky, 4=Stay the Same, 7=Become Much More Lucky To what extent do you think that other people might view you differently if you received your transplant/transfusion from this donor? 1=7: 1=View Much More Negatively, 4=View the Same, 7=View Much More Positively To what extent would you feel "creeped out" or "contaminated" if you received your transplant/transfusion from this donor?
Essentialist: Luck Non-Essentialist: Reputation
Non-Essentialist: Creeped Out / Contaminated Category Change

a b
1-7: 1=Not At All, 4=Somewhat, 7 = Very If you received the transplant/transfusion from a (pig/chimpanzee), would you be a (pig/chimpanzee) after the transplant/transfusion? 1-7: 1=Definitely No, 4=Not Sure, 7=Definitely Yes
Self-as-recipient wording is shown; respondents also answered an infant-as-recipient version in which their hypothetical infants were described as the recipient of the transplant.
b
Category change questions were asked only for cross-species (pig and chimpanzee) donors.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 4 Liking, Essentialist, and Non-Essentialist Scores by Donor Type for Study 1 (American Respondents) and Study 2 (Indian Respondents)
Measure Liking Donor Type American Respondents similar-to-self (baseline) positive different-from-self negative cross-species Indian Respondents similar-to-self (baseline) positive different-from-self negative cross-species
a
!"#
M
c
SD 0.79 0.91 1.34 1.72 1.62 1.10 1.01 1.40 1.41 1.51
Sig. ** ** ** **
Essentialist M 1.54 1.84 1.67 1.86 2.22 2.44 2.94 2.65 2.78 3.61 SD 0.81 1.10 0.85 0.96 1.39 1.30 1.42 1.23 1.24 1.70 t
Sig.
Non-Essentialist M 1.42 1.74 1.97 3.24 4.45 2.42 2.71 2.83 3.47 5.19 SD 0.68 0.84 0.92 1.42 1.94 1.52 1.38 1.36 1.35 1.54 t
Sig. ** ** ** ** ** ** ** **
6.45 6.02 4.87 3.53 2.48 6.10 5.94 4.54 3.75 1.94
5.96 12.57 17.26 23.82 1.59 14.33 16.53 27.25
4.00 1.96 3.96 5.51 6.09 2.66 3.72 8.77
** 0.06 ** ** ** * ** **
4.26 6.94 14.60 17.56 4.09 4.67 8.82 17.29
0.11 ** ** **
Pairwise comparisons were always against scores for similar-to-self (baseline) donor types; Study 1 df = 103, Study 2 df = 139
b c
* = p < .05 ** = p < .01
Similar-to-self baseline for the DNA and Blood conditions included scores for both the close family member and an individual of same gender, age, sexual orientation, and background. However, as noted in the main text, the baseline for the Heart condition only included scores for the latter (excluding the close family member). We therefore also conducted secondary analyses equating for baseline (i.e., employing only the same gender, age, sexual orientation, and background donors as baseline), which replicated all ANOVA donor type main effects and patterns of significant and non-significant results from paired comparisons for both American and Indian samples. Thus, results from this secondary analysis are entirely consistent with those reported in the primary analysis.
Table 5
!"#
Proportion of Respondents Providing Each of Three Response Patterns on Essentialist Measure in Study 1 (American Respondents) and Study 2 (Indian Respondents)
Donor Type differentfrom-self 0.15 0.53 0.31 0.21 0.35 0.44
Response Pattern American Respondents Essentialism score lower than similar-to-self Essentialism score equal to similar-to-self Essentialism score higher than similar-to-self Indian Respondents Essentialism score lower than similar-to-self Essentialism score equal to similar-to-self Essentialism score higher than similar-to-self
positive 0.10 0.51 0.38 0.14 0.28 0.58
negative 0.08 0.43 0.48 0.24 0.24 0.52
cross-species 0.10 0.38 0.52 0.13 0.17 0.70
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 6 Sample Open-Ended Explanations According to Type
Explanation Type Safety/Health Morality Creeped Out/ Reputation Category Label Essentialist Denial of Essentialism Unclear/ Other Sample Explanation Worried about disease Because philanthropists are kind-hearted people" I will feel very contaminated and creepy "Definitely no, because it's a chimpanzee" The cruel murderer's qualities will come to me The heart does not pass along any of those qualities Sounds fine
!"#
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 7 Proportion of Respondents Providing Target Open-Ended Explanation Types for Study 1 (American Respondents) and Study 2 (Indian Respondents)
Explanation Type Safety/ Health Morality 0.50 0.63 0.60 0.66 0.56 0.22 0.31 0.25 0.38 0.30 0.05 0.09 0.29 0.17 0.05 0.07 0.14 0.49 0.48 0.04
a
!"#
Donor Type
Creeped Out/ Reputation 0.02 0.10 0.04 0.29 0.26 0.04 0.03 0.04 0.16 0.11
Category Label 0.23 0.00 0.05 0.04 0.14 0.29 0.05 0.10 0.09 0.16
Essentialist 0.02 0.13 0.23 0.15 0.04 0.06 0.24 0.45 0.28 0.09
Denial of Essentialism 0.04 0.10 0.17 0.14 0.00 0.04 0.08 0.16 0.05 0.01
Unclear/ Other 0.94 0.80 0.73 0.81 0.79 0.99 0.96 0.83 0.89 0.90
American Respondents similar-to-self different-from-self positive negative cross-species Indian Respondents similar-to-self different-from-self positive negative cross-species
a
Because each donor type included two donors, and because each respondent provided explanations for both selfas-recipient and infant-as-recipient, the proportion of respondents providing the different styles of explanations do not add to 1.0 across rows.
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 8 Liking, Essentialist, and Non-Essentialist Scores by Donor Type for Study 3 (Money Transfer Control)
Measure Liking Donor Type American Respondents similar-to-self (baseline) different-from-self positive negative Indian Respondents similar-to-self (baseline) different-from-self positive negative
a
!"#
M 6.21 5.73 6.28 2.96 6.04 5.07 6.04 2.27
SD 1.12 1.18 0.80 1.54 0.92 1.02 0.72 1.24
Sig.
Essentialist M 1.80 1.54 1.87 1.77 2.51 2.27 2.53 2.59 SD 1.39 1.10 1.34 1.40 1.75 1.42 1.57 1.51 t
Sig. * ns ns ns ns ns
Non-Essentialist M 1.09 1.15 1.18 2.65 2.24 2.34 2.16 3.46 SD 0.88 0.86 0.77 1.51 1.36 1.31 1.32 1.26 t
Sig.
2.96 0.57 10.19 4.55 0.00 13.47
* ns **
2.13 0.52 0.15 1.16 0.16 0.31
0.68 0.80 7.27
ns ns **
** ns **
0.55 0.79 4.22
ns ns **
Pairwise comparisons were always against scores for similar-to-self (baseline) donor types; American df = 33, Indian df = 28
b
* = p < .05 ** = p < .01
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 9 Respondents Providing Each of Three Response Patterns on Essentialist Measure in Study 3
Donor Type differentfrom-self 0.47 0.44 0.09 0.38 0.28 0.34
!"#
Response Pattern American Respondents Essentialism score lower than similar-to-self Essentialism score equal to similar-to-self Essentialism score higher than similar-to-self Indian Respondents Essentialism score lower than similar-to-self Essentialism score equal to similar-to-self Essentialism score higher than similar-to-self
positive 0.24 0.44 0.32 0.28 0.34 0.38
negative 0.35 0.38 0.26 0.38 0.21 0.41
ESSENTIALIST BELIEFS ABOUT BODILY TRANSPLANTS Table 10 Essentialist and Non-Essentialist Scores by Transfer Type for Study 4
Measure Essentialist M (%) Non-Essentialist M (%) SD 18.38 23.16 21.92 22.67 28.19 29.05 19.40 29.61 26.40 23.01 23.48 22.96
!"#
Transfer Type
SD 13.55 7.75 5.66 25.26 26.53 21.80
American Respondents heart 8.44 pacemaker 3.01* graft 2.95* Indian Respondents heart 23.66 pacemaker 24.26 graft 24.42
* significantly lower than heart, p < .03
Preyer and Peter run05.tex V1 - May 4, 2007 5:45pm
Page 133
5
Moderately Sensitive Semantics
Sarah-Jane Leslie
1 . I N T RO D U C T I O N What is context sensitivity? What tests are reliable indicators of this phenomenon? Here I shall take up and develop some themes of Herman Cappelen and Ernie Lepores book Insensitive Semantics, in order to better understand the phenomenon, and the tests that reveal it. If we eschew such tests, and rely on intuitions about what is said, then, Cappelen and Lepore argue, it is hard to resist the conclusion that all of language is contextually sensitive. While most semanticists take themselves to be Moderate Contextualistswho hold that natural language contains a broad but limited stock of context-sensitive itemsCappelen and Lepore claim that Moderate Contextualism is an unstable position. The same intuitions that lead semanticists to espouse Moderate Contextualism should lead them to instead espouse Radical Contextualismwhich is the view that context sensitivity is so rampant, no natural language sentence ever semantically expresses a proposition independent of context. The phenomenon of what is said is so unconstrained that, if we try to capture it semantically, we shall be forced to adopt Radical Contextualism. Since, Cappelen and Lepore claim, the arguments for Moderate Contextualism hinge on the desire to account for what is said, there is a slippery slope from Moderate Contextualism to Radical Contextualism. Cappelen and Lepore level three main objections against Radical Contextualism, and thus, if their slippery slope stands, against Moderate Contextualism. They claim that the contextualist classies as context-sensitive items that fail their three tests, which is an unacceptable result. They then argue that, if Radical Contextualism were true, we would be unable to communicate with each other with the ease and reliability that we, in fact, routinely employ. Finally, they claim that Radical Contextualism is internally inconsistent. In what follows, I shall defend Moderate Contextualism against both Cappelen and Lepores slippery slope argument and their claim that Moderate Contextualism posits context-sensitive items that fail their three tests. Specically, I will
Page 134
134
Sarah-Jane Leslie
argue that the Inter-Contextual Disquotation/Real Context Shifting Argument test classies more items as context-sensitive than Cappelen and Lepore suggest. In addition to the basic set, I shall argue that the ICD/RCSA test itself indicates that there is a range of items that are also contextually sensitive. I will refer to this set of contextually sensitive items that are not included in the basic set as the intermediate set. The intermediate set, which is determined by Cappelen and Lepores ICD/RCSA test, is quite restricted; it excludes many of the items that the Radical Contextualist would count as context-sensitive. For example, the intermediate set includes tall, ready, enough, and every, among others, but does not include weighs 80 kg, is red, and is tall for a pregnant giraffe that is standing up straight. Thus if we take the ICD/RCSA test seriously, as I think we should, it provides the Moderate Contextualist with a way of blocking Cappelen and Lepores slippery slope argument. The ICD/RCSA test gives us a means of investigating contextual sensitivity without relying on intuitions about what is said. We have a sober and restricted means of determining when a given item is contextually sensitive, and this means tells in favor of Moderate Contextualism. I shall then consider the other two tests in light of these considerations. I will argue that ICD/RCSA is not negotiable as a test of context sensitivity, and so we should look to understand how the items in the intermediate set could appear to fail the other two tests, despite their being context-sensitive. I provide an account of how the members of the intermediate set behave in says that reports and collected predications that explains their apparent failure on the Report and Collection tests. My account proceeds by rst considering how the semantic values of items including then, there, local, nearby, left, and right are determined in reports and collected predications, and then simply extending this account to the intermediate items. The investigation sheds some light on the complex phenomenon of non-indexical context sensitivity. Since I am only concerned with defending Moderate Contextualism, I shall not address Cappelen and Lepores charge that Radical Contextualism is internally inconsistent. As Moderate Contextualism is in fact a stable position, this criticism does not apply; the charge is specic to Radical Contextualism, as it rests on the claim that every sentence is contextually sensitive. I will also not address the claim that contextualism cannot account for our ability to communicate with each other across contexts in the main body of the chapter, though I have included an appendix in which I discuss the matter. I argue there that Cappelen and Lepores positive view, Semantic Minimalism, cannot explain our communicative practices any better than Moderate or even Radical Contextualism. Although it may appear that I thoroughly disagree with the view presented in Insensitive Semantics, this is not so. Cappelen and Lepores main point concerns semantic methodology, in particular, our unwarranted reliance on intuitions about what is said. On this point, I am in full agreement with them. Their point is an important one, and should be recognized as such. Unfortunately, this pointtheir main pointhas been overlooked because of the provocatively
Page 135
135
small size of their basic set. If I am right, however, this is no more than an artifact of their misapplication of their own tests, and not itself reective of the quality and plausibility of their position. I hope that this chapter will help to clarify this point, and so allow us to see how valuable Cappelen and Lepores real contribution is to the debate.
2 . M O D E R AT E C O N T E X T UA L I S M A N D R E A L C O N T E X T S H I F T I N G A RG U M E N TS Let us begin with Cappelen and Lepores (henceforth CL) third test, which concerns Inter-Contextual Disquotation and Real Context Shifting Arguments. This test is, I believe, the most important and the most telling. CL note that:
It is a constitutive mark of a context sensitive expression e that it can be used with different extensions (semantic values) in different contexts of utterance . . . it follows from this constitutive fact alone that for any context sensitive expression e our use of e in this context . . . with whatever extension it takes on in this context need not be the same as whatever extension it takes on in another context. There can be no denying that this is so. Based on this constitutive fact about context sensitivity, the following test recommends itself for judging whether e behaves as it should by actually using e in a context of utterance . . . and simultaneously describ[ing] another use of e with a distinct semantic value in another context. Since e is not context sensitive unless its semantic values can shift from context to context, and since the semantic values e takes in, say, this context of utterance . . . can be distinct from the semantic value it takes in some other context, to test whether e is context sensitive or not, simply use e ; in order to use e , put it in a sentence S and then use S. e is context sensitive only if there is a true utterance of an instance of the following schema for Inter-Contextual Disquotation (ICD for short, where S contains e ): (ICD) There are (or can be) false utterances of S even though S. (2005: 1045)
CL claim that the members of their basic set pass this test with ease. Consider, for example, the following obviously true remark: there are false utterances of I am female even though I am female. The ICD test has a sister test; if an item passes ICD, we should be able to construct a Real Context Shifting Argument (RCSA) for the item, and vice versa. An RCSA is essentially an ICD, but with the contextual details explicitly supplied, rather than left to the interpreters imagination. We may think of ICD and RCSA as two sides of the same test. We might, for example, construct the following RCSA for then: Then Lets think back to the year 2000. Barry lived in California then. The other day, Gideon and I were talking about last summer. He said Barry lived in California then, but that was false because Barry was living in Princeton last summer. This is so even though, thinking back to 2000, Barry lived in California then.
Page 136
136
Sarah-Jane Leslie
Such an RCSA helps us ll out the otherwise awkward there are false utterances of Barry lived in California then even though Barry lived in California then. It should be clear that RCSAs simply function to esh out the bare bones of the ICDs; the difference between an RCSA and an ICD is only one of packaging. The ICD/RCSA test is a convincing test of context sensitivity. If an item can be shown to pass it, it is hard to deny that it is context-sensitive. On the other hand, if an item does not pass ICD/RCSA, then this is good evidence that it is not contextually sensitive, and any data or intuitions to the contrary should be reconsidered. CL describe the test as reecting a constitutive fact about what it is to be context-sensitive; I have no quarrel with this claim, but rather embrace it. This test should be taken seriously.
2.1. Do the Items in the Intermediate Set Really Fail This Test?
Let us consider just which items pass ICD/RCSA. CL write as though all items that lie outside their basic set fail ICD/RCSA, though in providing examples of such failures, they focus exclusively on items that only a Radical Contextualist would consider context-sensitivepredicates such as knows, is red, and weighs 80 kg. Many a Moderate Contextualist would deny that these expressions are contextually sensitive, while maintaining that the basic set is too restrictive. At this point in their argument, it is understandable why CL would focus exclusively on these items; they have already argued that there is no such stable position as that which the Moderate Contextualist would endorsethere is only their position, which acknowledges no semantic context sensitivity outside of the basic set, or Radical Contextualism. This argument proceeds by way of claiming that the Moderate Contextualist cannot embrace the intuitions that lead to his moderate case without also embracing those that lead to a radical one. There is alleged to be a slippery slope between Moderate and Radical Contextualism, so that the Moderate Contextualist has no principled way of drawing the line between his items and those of the Radical Contextualist. Since CL take themselves to have established this, they may suppose there is no reason why they cannot object equally well to the Moderates position by focusing on the more radical items. Thus they focus on whether items such as knows and weighs 80kg fail their tests. This transition is too hasty, however. All the Moderate Contextualist needs is some way or other to resist the slide down the slippery slope, and she cannot be held responsible for the Radical Contextualists failings. If there is some limited range of items outside the basic set that pass ICD/RCSA, then the Moderate Contextualist will have found the rm footing she needs along the slope to Radical Contextualism. She will have found a principled way to draw the line between the items she takes to be context-sensitive, and the multitude that the Radical Contextualist classies as such.
Page 137
137
FN:1
I will argue in what follows that there is a set of items that is intermediate between the tiny basic set and the Radical Contextualists teeming horde. These intermediate items easily pass ICD/RCSA, while items like knows, weighs 80 kg, and is tall for a pregnant giraffe that is standing up straight do not. Since passing this test is matter of our having intuitions about the truth or acceptability of various stories or utterances, I have informally polled people who are not involved in this debate, and have found that their intuitions accord with mine. My informal poll is a far cry from a controlled experiment, and such an experiment might prove useful here; it would certainly trump my poll, were the results found to ultimately differ. A poll, however informal, is valuable in that it keeps the debate from reducing to SJs-intutions-versus-CLs. The purpose of the poll is to reect that I am not alone in construing the data this way. (Nor are my opinions only shared by Moderate Contextualists with a stake in the debate. Most of the people I spoke with were either not philosophers of language, or were not philosophers at all.) Consider the following Real Context Shifting Arguments for various intermediate items.
Enough
Ive just moved apartments, and Im hanging a picture in my living room. Its pretty light, so a small picture hook is strong enough. But yesterday, my friend was helping me hang a 25lb mirror, and he said Oh, a small picture hook is strong enough. That was false (and I have a cracked mirror to prove it), even though, given how light this picture is, a small picture hook is plenty strong enough.
Ready
John is woefully unprepared from his APA interviews, and is about to pass out from nerves. He really needs more time before his interviews; he is simply not ready. There is an evangelist hovering around the convention, looking to see whether there are any lost souls for whom the time is right to accept Jesus Christ as their Personal Savior. Spotting John, the evangelist says Ah, hes ready! This is true; as it turned out the evangelist managed to convert John, in large part thanks to his looming interviews, coupled with the fact that he just wasnt ready.
Every
Princeton has really clamped down on grade ination. Im teaching Intro to Logic this semester, and it is denitely not the case that every student will get an A. Im going to make sure of that, or else the administration
See Hawthorne (2006) for more examples of RCSAs for items not found in the basic set.
Page 138
138
Sarah-Jane Leslie will get ticked off with me. My friend Des has a lot more guts about these things than me, though. He was telling me about his Kant class the other day, and said My students are great. I dont care what the administration says. Every student will get an A! That was true, tooI saw his grade sheetbut as for my class, its decidedly false that every student will get an A.
Its raining
Im in New Jersey right now, and the weather is beautiful. Its one of those clear spring days, and its denitely not raining. My poor grandmother called me from Scotland this morning though, and one of the rst things she told me was, its raining. She was rightI checked the weather report for Scotland. Anxious to make sure the day would remain nice, I looked outside. No worries; it wasnt raining.
Tall
Seeing as how he measures 6 3 , Tom is tall. He plays basketball from time to time, and once he called me from the court, because he was feeling nervous before the game. To reassure him, I said well, itll help that youre tall. He replied are you kidding me? You should look at the guys Im up against. Im not tall at all!. He was right (the other guys were approaching 7 !), even though at 6 3 , Tom is denitely tall. These scenarios are perfectly intelligible, and as natural as anything else in this debate. And as far as I am able to tell, they all count as RCSAs. In each of them, the sentence in question is used in the context (twice in factonce at the beginning, and again at the end), and mentioned in another, where it receives a different truth value. By CLs own standards, enough, ready, every, tall, and its raining pass their RCSA test for context sensitivity. There are two objections to the validity of these RCSAs that I can imagine that CL might raise, but they each turn out to be non-starters since they also apply to RCSAs for non-indexical members of the basic set such as then. The rst is that, when we use the sentence a second time at the end of the RCSA, we need a brief remark to bring us back to the original context. For example, in Enough we have: Ive just moved apartments, and Im hanging a picture in my living room. Its pretty light, so a small picture hook is strong enough. But yesterday, my friend was helping me hang a 25lb mirror, and he said Oh, a small picture hook is strong enough. That was false (and I have a cracked mirror to prove it), even though, given how light this picture is, a small picture hook is plenty strong enough.
Page 139
139
The italicized phrase makes a signicant contribution to the naturalness of the RCSA. Now, we do not need to supply such a phrase in giving an RCSA for an indexical. For example, I might simply write I am female. The other day Bill said I am female. This was false, even though I am female. No phrase, however minimal, is needed to bring us back to our context. This, however, is not so with those members of the basic set that are not pure indexicals. Consider the RCSA we gave for then, for example: Lets think back to the year 2000. Barry lived in California then. The other day, Gideon and I were talking about last summer. He said Barry lived in California then, but that was false because Barry was living in Princeton last summer. This is so even though, thinking back to 2000, Barry lived in California then. Removing the italicized phrase here undermines the RCSA for then as much (if not more) than removing the corresponding phrase from Enough. We cannot insist that an RCSA be no more complex than is needed for a pure indexical or we will rule out items such as then. The second challenge to the validity of my RCSAs might hold that, for an RCSA to be successful for a predicate is F, we must be able to remark in the RCSA that the relevant utterance is false because the speaker was not concerned with F-hood. CL end their unsuccessful RCSA for weighs 80 kg by writing If someone were to utter Rupert weighs 80 kg her utterance would be false, even though he weighs 80 kg. The utterance would be false, not because Ruperts weight has changed, but because the speaker is concerned with something other than what Rupert weighs, for example with what a scale registers were he to step on it fully clothed (2005: 111; my emphasis). Now, one can end an RCSA for a pure indexical in such a manner, for example, Bills utterance of I am female is false because Bill was concerned with something other than whether I am female. But we cannot so amend our RCSA for then, on pain of incoherence: Lets think back to the year 2000. Barry lived in California then. The other day, Gideon and I were talking about last summer. He said Barry lived in California then, but that was false because Gideon was concerned with something other than what Barry was doing then. This amended RCSA is clearly unacceptable; it is every bit as bizarre as CLs proffered RCSA for weighs 80kg. Again, unless we are to decide that items such as then are not contextually sensitive, we cannot require that an RCSA for an expression e contain a remark that the false utterance of . . . e . . . was false because its speaker was not concerned with e. This cannot be a constraint on a successful RCSA. I will thus take myself to
Page 140
140
Sarah-Jane Leslie
have met CLs challenge, and have provided RCSAs for the intermediate items above. A question that now arises is whether we can construct successful RCSAs for just about any item, as the Radical Contextualist might hope, or whether only a modest collection of items are amenable to RCSAs. It is not a trivial matter to construct an RCSA for an item, and it seems to me that there are many items that do not pass the RCSA test. For example, the RCSA that CL attempt to provide for knows (i.e. their Known Rupert ) is decidedly awkward, and Im not convinced that mine is much better:
Knows
Right now Im writing a paper on semantics, and am not thinking about epistemology. I know a lot of things right now, including that I have hands. But earlier today, Jim called me, and we talked about epistemologyin particular about the possibility of being a brain in a vat. I said Wow, I guess I dont know that I have hands! That was true, since I cant rule out the possibility of my being a brain in a vat. But now that Im back to writing my semantics paper, I know that I have hands. This RCSA is considerably less natural, to my ear at least, than the ones above. Perhaps a better one could be constructed, or perhaps knows is just not context-sensitive, as various leading theorists have argued (see Stanley 2005; Hawthorne 2004, for example). Insofar as some putatively context-sensitive expressions pass RCSA while others fail it, the Moderate Contextualist has found her way off of CLs slippery slope. I am myself neutral on the question of whether knowledge attributions are contextually sensitive, but I would very much like to be sure that Moderate Contextualism does not led inexorably to Radical Contextualism. CL argue that the intuitions that tell us that is tall is context-sensitive should also tell us that is tall for a giraffe is context-sensitive. They ask: Are we talking about pregnant giraffes? (Pregnancy, they tell us, affects giraffes necks in ways that are relevant to determining their height.) Once we have decided whether or not we have in mind pregnant giraffes, the questions arise: Are the giraffes in question standing up straight? Have they just taken a bath? (CL inform us that bathing makes a difference to a giraffes height.) CL claim that all these further factors are ones that start to seem contextually relevant, once we start down the contextualist road. (Why this is so is not wholly apparent; to me these seem more like cases of vagueness or perhaps mere determinability, but let us put that aside.) The challenge to the Moderate Contextualist is then to provide a principled way of acknowledging the context sensitivity of is tall without allowing that is tall for a pregnant giraffe that is standing up straight is also context-sensitive. Here, CLs RCSA test is most helpful. Consider the following (decidedly odd) RCSA:
Page 141
141
Tall for a pregnant giraffe that is standing up straight

Georgina is simply not tall for a pregnant giraffe that is standing up straight. No one who saw her would claim that she was. But the other day, we were talking about pregnant giraffes that are standing up straight but have just taken a bath. I said Georgina is tall for a pregnant giraffe that is standing up straight! What I said was true, because taking a bath shrinks giraffes by a small amount. Of course, now that I am just looking at a dry Georgina, she is denitely not tall for a pregnant giraffe that is standing up straight. There is a clear difference in acceptability and plausibility between the RCSA above, and Tall, reprinted below: Seeing as how he measures 6 3 , Tom is tall. He plays basketball from time to time, and once he called me from the court, because he was feeling nervous before the game. To reassure him, I said well, itll help that youre tall. He replied are you kidding me? You should look at the guys Im up against. Im not tall at all!. He was right (the other guys were approaching 7 !), even though at 6 3 , Tom is denitely tall. The Moderate Contextualist, it seems, can distinguish between is tall and is tall for a pregnant giraffe that is standing up straight; and she can do so using CLs own test for context sensitivity! Whether one can construct as RSCA for a term then, seems a robust and theoretically well-grounded test. I fully agree with CL that it is an important test for context sensitivity, though of course I disagree with them about which terms pass it. As CL emphasize throughout their book, though, their main concern is with semantic methodology, rather than with defending the boundaries of their basic set. Inasmuch as predicates such as is tall for a pregnant giraffe that is standing up straight are not amenable to RCSAs, the introduction of the RCSA as a benchmark is an extremely valuable contribution to the debate. RCSAs sister test, ICD, is also very helpful in this regard. My intuitions, and the intuitions of the neutral parties polled, are that the following ICDs are easily heard as true: I There are false utterances of Im female even though Im female. Now There are false utterances of Jason is reading an email now even though Jason is reading an email now. Ready There are false utterances of SJ is ready even though SJ is ready.
As I say, I did not conduct a controlled experiment, but the people I spoke to were instructed to hold xed tense/time of evaluation, facts about the world, and the referents of proper names.
FN:2
Page 142
142
Sarah-Jane Leslie Enough There are false utterances of steel is strong enough even though steel is strong enough. Its raining There are false utterances of its raining even though its raining.
No one I spoke to had any difculty hearing the above group (I , Now, Ready, Enough, and Its raining ) as true. One or two people found the following ones unnatural, though they reported that they were able to hear them as true if they thought about them: Then There are false utterances of John went to the store then even though John went to the store then. Tall There are false utterances of Tom is tall even though Tom is tall. Every There are false utterances of every student got an A even though every student got an A. It is important to note that while Tall and Every were less natural than Ready, Enough, and Its raining, so was the basic set member Then. Tall and Every were not more difcult for people to hear than Then. And remarkably enough, Ready, Enough, and Its raining struck people as more natural than even Then! Overwhelmingly, however, people could not hear the following ones as true: Tall for a pregnant giraffe that is standing up straight There are false utterances of Georgina is tall for a pregnant giraffe that is standing up straight even though Georgina is tall for a pregnant giraffe that is standing up straight. Red There are false utterances of Clifford is red even though Clifford is red. Weighs 80kg There are false utterances of Smith weighs 80kg even though Smith weighs 80kg. Of these, Red was the only one that anyone even wondered whether they could hear as true. It turned out, though, that this had do to with the vagueness of rednot with the possibility of our being concerned with the color of, say, Cliffords innards, to take CLs example. Everyone was condent about the falsity of the ICDs containing weighs 80kg and is tall for a pregnant giraffe that is standing up straight. People were condent that these statements could not be true. These intuitions are exactly those that the Moderate Contextualist would predict, and they accord well with naturalness of the corresponding RSCAs. Thus if we grant CLs methodology of testing items for context sensitivity in this manner, and embrace ICD/RCSA as a valid and accurate test, then Moderate
Page 143
143
Contextualism emerges as the leading view. Minimalism wrongly predicts that many items that easily pass will in fact fail, and Radical Contextualism wrongly predicts that items that fail will in fact pass. Moderate Contextualism makes the correct predictions. This test also serves to rescue the Moderate Contextualist from CLs slippery slope. Thus, adopting CLs methodology seems to vindicate Moderate Contextualism.
3 . T H E OT H E R T WO T E S TS The ICD/RCSA test is not the only test that CL offer. They suggest that we adopt two additional tests for context sensitivity, which I will refer to as the Report test, and the Collection test. The Report test concerns inter-contextual disquotation. An item is context-sensitive, they claim, only if it typically blocks inter-contextual disquotational indirect reports (2005: 88). They write:
Suppose you suspect, or at least want to ascertain whether, e is context sensitive. Take an utterance u of a sentence S containing e in context C. Let C be a context relevantly different from C (i.e. different according to those standards signicant according to contextualists about e ). If theres a true disquotational indirect report of u in C , then thats evidence that S is context insensitive. ( To be disquotational just means you can adjust the semantic values of components of S that are generally recognized as context sensitive, i.e., we just test for the controversial components.) (2005: 89)
CL argue that the members of their basic set pass this test with ying colors. If Bill utters I am male, Mary cannot correctly report him by saying Bill said that I am male. In contrast, they claim that items that are context-sensitive according to the contextualist often fail this test. If Nina utters John is ready in the course of discussing an exam he is to take, CL, in the context of writing their book at a caf, can correctly report her by saying Nina said that John is ready. The next test, the Collection test, concerns so-called collected predications. On this test they write:
If a verb phrase v is context sensitive . . . then on the basis of merely knowing that there are two contexts of utterance in which A v-s and B v-s are true respectively, we cannot automatically infer that there is a context in which v can be used to describe what A and B have both done . . . On the other hand, if for a range of true utterances of the form A v-s and B v-s we obviously can describe what they all have in common by using v . . . then thats evidence in favor of the view that v in these different utterances has the same semantic content, and hence, is not context sensitive. (2005: 99; original emphasis)
An indexical such as yesterday clearly passes this test; if there is a true utterance of John left yesterday, and a true utterance in another context of Bill left yesterday, we cannot automatically assume that there is a context in which John and Bill left yesterday is true. In contrast, they suggest, from a true utterance of John
Page 144
144
Sarah-Jane Leslie
is ready, and a true utterance of Bill is ready, then, as they put it, the following collective description is perfectly natural: Both John and Bill are ready. Given that the intermediate items pass ICD/RCSA while apparently failing the other two tests, we appear to have a tension in CLs own theory. Whatever CLs personal convictions may be, it seems that people overwhelmingly have the intuitions that suggest the intermediate items pass ICD/RCSA. On the other hand, it does appear that the intermediate items fail the Report and Collection tests. One option is to reject ICD/RCSA as a test for context sensitivity. This option seems to me to be the least attractive, for reasons independent of my favoring Moderate Contextualism. The ICD/RCSA test captures a constitutive feature of context sensitivity. CL themselves insist on this:
it is a constitutive mark of a context sensitive expression e that it can be used with different extensions (semantic values) in different contexts of utterance . . . It follows from this constitutive fact alone that for any context sensitive expression e our use of e in this context . . . with whatever extensions it takes on in this context need not be the same as whatever extension it takes on in another context. There can be no denying this is so. (2005: 104)
FN:3
They go on to notice that it is the truth of an ICD statement containing e that guarantees the inadequacy of a disquotational T-statement for e. If S contains e and there can be false utterances of S even though S, then clearly the following biconditional cannot hold: S is true iff S. CL point out that ICD is precisely what shows us that the biconditional I am female is true iff I am female is false. There can be false utterances of I am female even though I am female, so this biconditional is false, and therefore unusable in a T-schema, or any similar semantic framework. If an item passes ICD, then we cannot give its truth conditions disquotationally, for doing so would result in a biconditional that is false, precisely because we have not recognized the items context sensitivity. If an item passes ICD, and relatedly, RCSA, it must be recognized as contextually sensitive. I see no way of denying this fact, and I do not believe that CL would care to deny it either. Since it is hard to see how an item could pass the ICD/RCSA test but not be contextually sensitive, let us try to understand how the intermediate items could appear to fail the Report and Collection tests despite their being contextually
CL speak at times as if none of their three tests were necessary, nor jointly sufcient, for establishing that an item in contextually sensitive. It seems to me quite possible, though, that a true ICD is sufcient for an item to be counted as contextually sensitive, provided that the truth of the ICD cannot be traced to vagueness, ambiguity, or such factors. Constructing a corresponding RCSA for the item is important in ruling out those factors; the minimal ICD might be heard as true for such extraneous reasons, but in producing an RCSA we can ensure that we are paying attention only to relevant differences between contexts. If an ICD for an item is true, and its truth cannot be traced to vagueness, ambiguity, etc, then it hard to see how that item could fail to be context-sensitive. Certainly, the truth of the ICD guarantees that we cannot give a disquotational T-schema for the sentence.
Page 145
145
sensitive. I will argue that once we have a better understanding of the nature and behavior of context-sensitive items other than the pure indexicals, it will become clear how some context-sensitive items might seem to fail these two tests. In what follows, I will discuss how the semantic values of context-sensitive items such as then, there, nearby, local, left, and right are determined when they occur in the scope of says that reports, and in collected predications. I offer an account of how their values are determined in such constructions, and argue that the account extends naturally to the intermediate items. The account allows us to see why the intermediate items appear to fail the Report and Collection tests, thereby resolving the theoretical tension generated by their passing ICD/RCSA while failing the other two tests. The account I will offer is not intended as a full-edged theory, but is rather somewhat schematic in nature. I do not doubt that there is more to be said about the items in question. I have also tried to remain neutral between the various moderate theories of context sensitivity, and have not tied my account to any particular one. Let us begin to develop our account by noting that the intermediate items are not always amenable to being disquotationally reported, or collected. It is not difcult to construct a scenario in which such reports and collections are unacceptable. Consider the following scenario, for example (adapted from Leslie 2004):
3.1. Johns APA Nightmare

John is on the job market, and about to be interviewed for his dream job. Unfortunately, he hasnt prepared for the interview at all. He can barely even remember what his thesis is about, and is unbelievably nervous. Now, the night before Johns APA nightmare, he was very hungry. People were going out for dinner, and inquired if John was in a position to eat dinner right then. He was indeed ready to eat dinner, and appropriately responded, Yes, Im ready. Johns classmate Mary, on the other hand, is very well prepared for her interviews. She has her dissertation summary down pat, has her syllabi memorized, and is well equipped to answer any questions thrown at her. Her friend takes one look at her condent demeanor and remarks I gotta hand it to you. You are ready.
Report test
The interviewing committee decides that they will allow John to reschedule, if his thesis advisor agrees that he needs to take more time to prepare. The thesis advisor disquotationally reports Johns utterance by saying Hey, he said himself last night that hes ready. The report is simply false under these circumstances.
Page 146
146
Sarah-Jane Leslie
Collection test
The interviewing committee reacts with surprise when they hear this, since John looks to them to be woefully unprepared, especially in contrast to Mary. The thesis advisor collects the two true utterances that attribute readiness to John and Mary, and remarks in response: Nope, theyre both ready. The collection is false as described. If there is some salient activity to be ready for, disquotational reports and collections only strike us as true if they pertain to that salient activity. Parallel situations are easily produced for the other intermediate items. This is not to say that the intermediate items pass the Report and Collection tests after all. Recanati (2006) discusses examples such as my example above, but CL (2006b) reply that the existence of such scenarios does not count against their view. In particular, they note that the existence of some contexts in which the relevant reports and collections may take place, despite relevant contextual differences, is all that they require. We still need an account of why we are able to ever disquotationally report and collect in the face of relevant differences in context. In their recent article Shared Content, Cappelen and Lepore suggest that our intuitions in scenarios such as the one above are tracking speech act content rather than semantic content; the salience of the APA interview leads us to focus on expressed propositions that are relevant to it, and thereby blind us to the minimal, shared semantic content (2006c ). They encourage us to look to contexts of ignorance and indifference: contexts in which we might, for example, say John said he was ready without knowing or caring exactly what activity was under discussion at the time of Johns utterance. The ignorance and indifference constitute relevant differences between the contexts, yet such differences do not render disquotational reports or collections unacceptable. Of course, it is vital that the reports and collections we consider occur in contexts that are relevantly different from the contexts of the original utterances. Clearly, it is irrelevant to consider reporting and collecting contexts in which the relevant contextual factors are the same as in the original context. As CL note in their reply to Hawthorne, we can disquotationally report uses of I if the right circumstances are in place, i.e. if the reporter is identical to the original speaker (2006a). The question then becomes: under what circumstances are two contexts relevantly different from each other? CL themselves clearly do not think that there is an easy answer to this question; that this is so becomes clear in the course of their response to Hawthorne (CL 2006a). Hawthorne (2006) notes that nearby and left seem to fail the Report test. Hawthorne writes:
Let us try left and nearby. Suppose Ernie is in New York City and I am in Birmingham. Ernie says A nearby restaurant has good Vietnamese food. I can report this by saying
Page 147
147
Ernie said that a nearby restaurant has good Vietnamese food, even though I am far away from him. Suppose that Ernie is facing me. A car goes to Ernies left and my right. Ernie says The car went left. I can say Ernie said that the car went left, even though my orientation is radically different to his. (2006)
In their reply, CL defend the claim that nearby and left are context-sensitive, and argue that Hawthornes example does not show them to fail the Report test. They argue as follows:
For the test to be applicable, it is essential that the contexts of the report and the reportee be relevantly different, i.e. that the relevant contextual features should be different in the contexts of the report and the reportee. So what are the contextually relevant features for, say, nearby? Maybe the answer is something like this: the character of nearby determines that an utterance of nearby in a context C refers to a location salient in C, but that location neednt be the location of C (i.e. it neednt be the location where the speech act occurs). It is whatever location is salient in C, and that could be the location of the reporteeif, for example, that location is made salient in indirectly reporting her. Hawthornes examples can then be understood as examples in which we can disquotationally report utterances containing nearby because the relevant contextual features in the two contexts (i.e. the salient location) are the same. If this is the correct diagnosis, all the examples show is that when the same location is salient in two context, C and C , youll be able to disquotationally report utterances of sentences containing nearby from C to C (or the other way around). But this is no more surprising than being able to disquotationally report utterances of sentences containing the rst person pronoun when the reporter is the same as the reported speaker. (2006a)
CL go on to adduce support for this claim by noting:

According to this suggestion, nearby refers to whatever location is salient in the context of utterance, and if your intuition is that Hawthornes disquotational reports are true, that is because the location of the reportee has become salient in the context of the report. Further support for this diagnosis is provided by imagining examples in which the salient locations differ in the context of the report and the context of the reportee, and hence blocks the disquotational report. Consider this variation in Hawthornes example (i)(ii): i. Ernie says to John, walking on 7th street in NY looking for a restaurant: A nearby restaurant has good Vietnamese food. ii. John, walking around Birmingham looking for a restaurant, reports Ernies utterance to his friends by saying: Ernie says that a nearby restaurant has good Vietnamese food. (ii) is intuitively false, and the above account provides an explanation: in the context of (ii) it is the location of the report (i.e. the location where the indirect report is uttered) that is salientthe search for a restaurant in Birmingham has made Birmingham the salient location. (2006a)
These observations concerning nearby are quite analogous to my observations concerning the intermediate items such as ready. As evidenced by Johns APA Nightmare, disquotational reports of ascriptions of readiness are false if there is a salient activity to be ready for, and the reported utterance did not concern that
Page 148
148
Sarah-Jane Leslie
activity. If Johns thesis adviser says John said himself that he was ready, on the basis of Johns saying in connection with dinner Im ready, this report is false. If there is a salient activity to be ready for in the context of the report, then we cannot disquotationally report readiness ascriptions, unless they pertain to that activity. Similarly, if there is a salient location in the context of the report, then we cannot disquotationally report utterances containing nearby, unless, of course, they pertain to the same salient location. But in the case of ready and other intermediate items, CL direct us to consider only contexts of ignorance and indifference, and thus to set aside such scenarios as Johns APA Nightmare. It is those contexts that are relevant to the Report and Collection tests. Should not the same advice apply to Hawthornes examples? Hawthornes examples seem to be prime cases of reports under ignorance or indifference, but in case we are unconvinced, let us borrow an example directly from CLs reply to Hawthorne. (The example is offered in response to Hawthornes suggestion that the reports contain mixed quotation, but it is well suited to our purposes here.) They write:
Consider an example involving nearby: Imagine answering a ringing payphone on the street, and a womans voice says: Theres a river nearby, then she hangs up. Asked what the caller said, you reply: The woman on the phone said there is a river nearby It seems perfectly possible that the speaker intended to use nearby and not to talk about nearby and that in so doing, she succeeded in saying something true (if our earlier diagnosis (of the character of nearby) is correct, thats what we would expect). (2006a; my emphasis)
FN:4
FN:5
The reporting context described here is clearly one of ignorance and indifference, in that the reporter does not know or care where the woman was calling from. CL claim that ignorance and indifference are factors that lead us to focus on semantic content rather than speech act content. Yet even in these circumstances of ignorance and indifference CL agree that we are able to disquotationally report an utterance containing nearby, a term that they take to be contextually sensitive. Presumably they take it that, as in Hawthornes example, the womans location becomes salient, just in virtue of our reporting her utterance. That is, no precursor is needed to render salient the location; it is enough for the reporter to utter The woman on the phone said there is a river nearby, for the womans location to be made sufciently salient. It is also not in any way a necessary condition for the report that the reporter know what or where the location is. He may be completely ignorant of this. Both the reporter and the audience understand nearby to operate on the womans location, whatever it may be.
Ernie Lepore suggested to me (pers. comm.) that perhaps only predicative uses of nearby are contextually sensitive, and that when it is used as a modifying adjective, it is not. I will return to this issue below, but it is important to note here that the utterance that is disquotationally reported is one in which nearby occurs predicatively. Not everyone nds such disquotational reports as acceptable as CL do. One may well doubt that the reporter could properly report the womans utterance as above. Similarly, in the case of
Page 149
149
It appears then that nearby, which even CL (probably) recognize to be contextually sensitive, behaves like the intermediate items on the Report test. That is, we can disquotationally report utterances containing nearby in contexts of ignorance and indifference, just as we can disquotationally report utterances containing the intermediate items in such contexts. If, however, the context of the report is not one of ignorance and indifference, and there is a salient value for the item in the context of the report, then a disquotational report will be unacceptable (unless, of course, that same value was salient in the context of the original utterance). CL point out that this is the case for nearby; if the context of the report is one in which Birmingham is a salient location, we cannot say Ernie said that a nearby restaurant has good Indian food if Ernie said this while discussing eateries in New York City. Similarly, if a particular activity is salient in a context, we cannot disquotationally report utterances of the form x is ready if they were made in contexts where a different activity was salient. We saw this in Johns APA Nightmare; Johns adviser cannot say to the interviewing committee, John said he was ready on the basis of Johns having said Im ready in connection with the previous nights dinner. It is not hard to see that the same holds for the other intermediate items. (Consider, for example, the Real Context Shifting Arguments given above. If we place ourselves in the context of those stories, unqualied disquotational reports of the utterances mentioned in the stories are false.) The intermediate items behave exactly like nearby in this respect. We can disquotationally report utterances containing them in contexts of ignorance and indifference, but not in contexts in which a different value for the context-sensitive item is salient. There are other similarities between nearby and the intermediate items. Consider what happens if we preface our disquotational report with a brief description of the context in which the original utterance occurred. Including such a preamble allows us to disquotationally report utterances containing nearby and the intermediate items, even if the context of the report includes a contextually salient value for the item in question, which would otherwise prohibit us from disquotationally reporting. In the case of nearby, suppose we are in Birmingham and looking for a good Indian restaurant there, as per CLs example, but instead of uttering out-of-the-blue Ernie said that there was a really good Indian restaurant nearby, I say instead, when we were in Princeton and talking about restaurants in NJ, Ernie said that there was a really good Indian restaurant nearby. This little preamble of mine makes it clear that in the context
ready, one could reasonably be skeptical that, if we had no idea what John was said to be ready for, we could unqualiedly report Bills utterance of John is ready by saying Bill said that John was ready. Many people feel that, in such circumstances, we would be obliged to add but I dont know what for, or some such qualication. Of course if this is correct, this casts doubt on the claim that items such as ready actually pass the Report test. Im sympathetic to this position, but I will not pursue it here. I will rather grant CL their data, and show that, even if it is correct, it does not mean that items such as ready are not contextually sensitive.
Page 150
150
Sarah-Jane Leslie
FN:6
of Ernies utterance, Princeton, not Birmingham, was salient, and so nearby should be interpreted accordingly. With this small preamble, we can override the contextually salient location of Birminghamthe salience of which is enough to render false a disquotational report that does not follow such a preamble. The same pattern is to be found among the intermediate items. If Johns thesis adviser says to the interviewing committee, John said he was ready with no preamble, his report is false. If, however, he begins with a preamble, he can successfully offer a disquotational report: when we were talking about dinner last night, John said that he was ready. As is the case with nearby, a little preamble allows us to disquotationally report the utterance. CL do not discuss preambles, but in keeping with their earlier treatment of nearby, I imagine that they would understand the preamble as making salient another locationin my above case, the preamble makes Princeton salient, and so nearby is interpreted relative to this. I propose that, if we simply extend this model to encompass the intermediate items, we will eliminate the tension between the ICD/RCSA test and the Reporting test. If an item such as nearby occurs in a says that report, its semantic value depends on (a) the location that was salient in the context of the original utterance, as introduced by a preamble or some such means; if no such location is available then it will depend on (b) a location that is salient in the reporting context, and if there is no such location, then (c ) it will be understood as dependent on some location or other that was salient in the context of the original utterance, even though the reporter and his audience are ignorant as to which location this may be. This last case, case (c ) includes contexts of ignorance and indifference, and if nearby operates in the manner here described, we can understand how it can be contextually sensitive, while permitting disquotational reports in contexts of ignorance and indifference. What we, in effect, have is a pattern of cascading defaults, which determine the semantic value of a contextually sensitive item guring in a report. The intermediate items can be assimilated to this schema. As Johns APA Nightmare suggests, the predicate is ready operates in a parallel fashion. If the predicate is ready occurs in a says that report, its extension depends on (a) the salient activity in the context of the original utterance, as introduced by a preamble or some such means; if no such activity is salient, then it will depend on (b) an activity that is salient in the reporting context, and if there is no such activity, then (c ) it is understood as dependent on some activity or other that was salient in the context of the original utterance. The reporter and her audience need not have any clue as to the nature of this activity, any more than they need know which location was salient in the original context when reporting utterances containing nearby. These conditions hold, mutatis mutandis, for the
This is not, of course, intended as a full description of how nearby operates. I do not doubt there is far more to say than has been said here.
Page 151
151
FN:7
other intermediate items, as far as I am able to tell. It is this pattern of cascading defaults that allows us to disquotationally report the intermediate items under conditions of ignorance and indifference. It should be noted that we can nd members of the basic set exhibiting this behavior, at least to some extent. We can, contra CL, sometimes use such items in says that reports under conditions of ignorance and indifference. Suppose, for example, that I am talking to Rachel while she is on her cell phone. I have no idea where she is calling from, and I dont bother to ask. In the course of the conversation, Rachel remarks Its raining here. I can report Rachel by saying Rachel said that its raining there. This report is not, of course, strictly disquotationalthere is a mandatory switch from here to therebut nonetheless a member of the basic set is used in a report made under conditions of ignorance. We have no trouble interpreting the occurrence of there as picking out Rachels location, wherever that may be. When members of the basic set that are not pure indexicals occur in says that reports, they are also sensitive to the occurrence of preambles that describe the context of the original utterance. Such a preamble can easily override whatever values may be salient in the reporting context. Consider, for example: I had a great conversation with John the other day. The sun was shining then, and we were both in a good mood. We got to talking about Platos Greece, and John said that philosophy was done so much better then. The reporting context contains a salient value for thennamely the day that the reporter spoke to Johnand it is this value that the rst occurrence of then picks up. On the basis of the minimal preamble we got to talking about Platos Greece, we easily understand the second occurrence of then as referring to the time of Platos Greece. It is important to notice how easily we obtain the value of the second then on the basis of this minimal information. Nothing in the preamble we got to talking about Platos Greece contains a reference to the time period of Platos Greece, yet we automatically extract such a time as a value for then. Clearly, we can extract values for contextually sensitive items on the basis of minimal and oblique preambles, and such values override other potentially salient values in the reporting context. If the members of the basic set behave in this way, it is not surprising that the intermediate items do too.
Even if one does not think that this is the most natural report possible, it is no less natural than disquotationally reporting an utterance by Rachel of There is a restaurant nearby under the same conditions. For that matter, it is not clear that our intuitions are as rm as CL would have us believe when it comes to disquotational reports of the moderate items under ignorance and indifference. It is not very natural to utter Nina said that John is ready if one has no idea of what Nina took John to be ready for. If one feels a temptation to amend the report Rachel said that its raining there by adding but I dont know where she is, consider whether one does not also feel that it would also be more natural to say Nina said that John is ready, but I dont know what for, or Rachel said there is a restaurant nearby, but I dont know where she is.
Page 152
152
Sarah-Jane Leslie
My reader may be wondering why I have separated out values made salient by a preamble from ones that are otherwise salient in the context. Could not both be treated under the banner of salience? Preambles affect the context by making a particular value salient; they are one among many ways by which a value may become salient in a context. I dont think there is anything wrong with this way of thinking, but it is helpful to separate out values that are salient in the reporting context, and those that become salient only in connection with the context of the original utterance. The preambles I have been discussing thus far function to make values salient in the latter way. The distinction is helpful when we consider multiple, collected reports, to which we now turn. Multiple, collected says that reports are possible with members of the basic set, such as then. We might say, for example, The other day, I spoke to John about Platos Greece, and Barry about Shakespeares England. Both said that intellectual standards were much higher back then. Here the preamble does not serve to render salient one single value for then, but rather makes salient one value for Johns utterance, and another for Barrys. We understand the collected report to attribute one utterance to John in which the time of Platos Greece determines the extension of then, and another to Barry in which the time of Shakespeares England determines the extension of then. These different attributions are possible even though we choose to report the utterances with a collected said that report. The preamble (again, minimal and oblique) is enough to allow us to interpret the report as attributing assertions to John and Barry in which then takes on different respective values. We do not always even require a preamble for such a distributive interpretation. Distributive interpretations, it seems, are possible even under conditions of ignorance and indifference. To see this, let us elaborate CLs scenario in which one answers the phone only to hear a woman mysteriously utter there is a river nearby and then hang up. Let us imagine that, immediately after the woman hangs up, weirdly enough, a man calls, also utters there is a river nearby, and promptly hangs up. If asked what the callers said, one might reply The woman and the man both said that there is a river nearby. This report is as natural as the single report CL describe. One might offer this collected report even if one had no idea where the speakers were calling from, and one certainly need not assume that they are calling from the same location to offer it. We easily understand nearby as tied to each speaker; we understand the report distributively, as equivalent to the woman said there is a river nearby, and the man said there is a river nearby, where the semantic value of the rst occurrence of nearby depends on whatever location is salient in the womans context, while the semantic value of the second depends on whatever location is salient for the man. Given that we can collect says that reports for items that are contextually sensitive, and can even do so under conditions of ignorance, it is no surprise that we can also do this for the intermediate items, as CL point out:
Page 153
153
Were thinking about different utterances of John is ready. Were imagining the following two contexts of utterance of (1): (1) John is ready. Context of Utterance C1 In a conversation about exam preparation, someone raises the question of whether John is well prepared. Nina utters (1). Context of Utterance C2 Three people are about to leave an apartment; they are getting dressed for heavy rain. Nina utters (1). . . . In (1.1) we report on her utterance in C1, in (1.2) her utterance in C2: (1.1) Nina said that John is ready (1.2) Nina said that John is ready. (1.3) In both C1 and C2, Nina said that John is ready. (2005: 901)
It should be clear that the possibility of these reports does not count against the context sensitivity of ready. The behavior of ready is not different from that of nearby or even then in this regard. (To make the parallel with then even more explicit, consider: The other day, I spoke to John about Platos Greece, and then later about Shakespeares England. In both contexts, he said that intellectual standards were much higher back then.) So far it seems that the behavior of the intermediate items in reporting contexts does not differ in unacceptable ways from the behavior of items that are clearly contextually sensitive. Once we understand their behavior on the model I have proposeda model that is required to account for the behavior of contextuals such as nearby, and perhaps even Kaplanian adverbs such as there and thenit becomes clear why the intermediate items appear to fail the Report test, despite their being contextually sensitive. We have seen that, for all these items, a minimal preamble lets us understand the value of the context-sensitive item in a says that report as dependent on what was salient in the context of the original utterance. The value is thus dependent the context of the original utterance, as opposed to the reporting context, and further, this can be so across multiple, collected reportseven if the various reported utterances occurred in very different contexts. We have seen that this is so for then and nearby, as well as for the intermediate items. If we nd ourselves in a context of ignorance and indifferencethat is, if no preamble is given and nothing relevant is known about the original context, and the reporting context does not provide the relevant value needed to determine the items semantic valuethen we can nonetheless disquotationally report utterances containing nearby, and in some circumstances, even there. We simply understand the relevant, required value to be what it was in the original context, whatever that may have been. I see no reason why we cannot understand reports containing intermediate items in contexts of ignorance and indifference in the same way. We understand out-of-the-blue utterances of Bill said that John was ready to mean roughly that Bill said
Page 154
154
Sarah-Jane Leslie
that John was ready for some activity or other that was salient in the context of Bills utterance. Let us bolster this last claim by considering the following report. Imagine that we do not have any particular activity in mind, and we do not know what activities were salient in the original contexts. Then consider the following report: Tom said that John was ready. Later, Barry said that John wasnt ready, but that was in another context, so he and Tom didnt disagree. Such a report may well be true. Tom can say that John was ready, while Barry can say that John wasnt ready, and yet Tom and Barry need not contradict each other. For CL, however, John is ready and John is not ready are contradictoriesis ready is an invariant predicate, and so either John possesses the property it expresses or he does not. Barry and Tom would have to be in disagreement, since they predicated contradictory things of John. CL must claim that our intuition that the above report might be true is based on our tracking speech act content, rather than semantic content, since the semantic content of the complements of the reports are contradictories. It is not clear why we would be tracking speech act content here, though, given that we are tracking semantic content in the scenarios they use for their Report test. The proposal I have outlined, however, offers a natural rendering of the above report. Since this is a context of ignorance and indifference, we understand the initial report Tom said that John was ready to mean roughly that Tom said that John was ready for something or other that was salient in the context of Toms utterance, and we understand the second report Barry said that Tom wasnt ready to mean roughly that Barry said that John wasnt ready for something or other that was salient in the context of Barrys utterance. Since we are told that the two contexts differed, it is easy to see how the two reports could be true, and yet Tom and Barry fail to disagree. The contents of the reports are not contradictories on the view proposed here, which ts well with our intuitions concerning scenarios such as this.
3.2. The Collection Test

We have seen how we might reconcile the results of the ICD/RCSA test with those of the Report test. It is not difcult to extend the model described above to also help us reconcile the results of the ICD/RCSA test with those of the Collection test. Consider nearby and local. Both appear to fail the Collection test, as the following example illustrates: John lives in St Louis, and Bill lives in San Francisco. Both John and Bill buy their food from local farmers. And last night, both John and Bill went to a nearby restaurant for dinner.
Page 155
155
We can collect predications containing local and nearby across relevantly different contexts. That means both items apparently fail the Collection test. The situation here, though, is not too different from the one we faced when dealing with collected says that reports. There, we understood the context-sensitive elements to be indexed to their respective contexts, so that they might have different semantic values in each of the multiple reports that were collected. For example, if we say I spoke to John about Platos Greece, and then later on about Shakespeares England. In both contexts, he said that intellectual standards were much higher back then, we understand that two utterances are being reported, and the semantic value of then in the rst is the time of Platos Greece, and in the second it is the time of Shakespeares England. An analog of this approach is appropriate here: we understand local in the collected predication above as meaning local w/r/t John and Bill respectively, and nearby as meaning nearby John and Bill respectively. But now consider simple, non-collected predications such as John went to a nearby restaurant. Here, I propose that nearby may receive either a subjectbased interpretation, i.e. nearby John, or a context-based interpretation, which is relative to a salient location in the context of utterance. If there is such a salient location in the context of utterance, then the item is most naturally interpreted relative to that location; otherwise it is interpreted relative to the subject, so long as that is appropriate. (I will discuss these conditions of appropriateness below.) Thus, if one utters out-of-the-blue John went to a nearby restaurant last night, nearby can be naturally interpreted as meaning nearby John, and so the utterance may be true even if John is in New York, while the utterer is in Birmingham. If, however, one is trying desperately to nd a restaurant in Birmingham, and ones friend says John went to a nearby restaurant last night, nearby is here interpreted as nearby the location of the speaker, and so is false. Having drawn the distinction between these two ways in which the semantic value of nearby may be determined, we can notice that not all sentences lend themselves to the subject-based interpretation. Consider, for example John is nearby. This sentence can only be interpreted to mean that John is near the location of the utterer (or at least near some location that is salient in the context of the utterance). A subject-based interpretation of nearby here would yield the unacceptable John is nearby John, and so is not available as an interpretation of that sentence. Only the context-based interpretation is acceptable here. If collected predications drawing on different contexts are only acceptable when the context-sensitive item receives a subject-based interpretation, then we would predict that sentences in which the subject-based interpretation is not available could not be collected. This is indeed the case: if Bill utters in San Francisco John is nearby, and Barry utters in Princeton Shanna is nearby, we cannot conclude that both John and Shanna are nearby. The collection is unacceptable.
Page 156
156
Sarah-Jane Leslie
I believe that this treatment explains Hawthornes otherwise puzzling observation that some collected predications of nearby and left are permissible, while other are not. Hawthorne writes:
Suppose in Birmingham, I say, I am going to a nearby restaurant and Ernie, in New York, says I am going to a nearby restaurant. We can certainly collect with John and Ernie are going to nearby restaurants. But suppose Ernie says There is good Vietnamese food nearby and I say, There is good Indian food nearby. We cannot collect with There is good Vietnamese food and Indian food nearby. A similar pattern holds for left. Suppose Ernie turns to his left, saying I am turning left, and I turn to my left, saying, I am turning left. I can collect: We are both turning left. I can also use verb phrase ellipsis: I turned left and Ernie did too. But suppose Ernie and I are facing each other. Noticing a ball rolling to from his right to his left, Ernie says The ball is moving left. Noticing a balloon moving from my right to my left, Ernie says The balloon is moving left. I cannot collect with The ball and the balloon are moving left. Relatedly, I cannot say The ball moved left and so did the balloon. This particular pattern of success and failure cries out for some kind of explanation. (2006)
With there is good Vietnamese food nearby and there is good Indian food nearby, there is no subject-based interpretation available. For these sentences, the dummy subject there does not support a subject-based interpretation. If we consider, say, good Indian food to be the dislocated subject of the sentence, as many syntacticians do, this does not help matters; the sentence would then mean Good Indian food is nearby, which does not allow for a subject-based interpretation, just as John is nearby does not. As my treatment predicts, because there is no subject-based interpretation available, we cannot collect these predications. In the case of left, it matters whether the subject determines a frame of reference appropriate for right and left. People determine such frames, so it is possible to interpret John and Ernie both turned left as John and Ernie both turned left from the point of view of John and Ernie respectively. Balls and balloons, however, do not determine such framesthere is no such thing as left from the point of view of a balloon. Only the context-based interpretation is possible in such cases, and we cannot collect these predications. Ernie Lepore (pers. comm.) has suggested to me that perhaps nearby is ambiguous between a context-sensitive predicative sense, as in John is nearby, and a context-insensitive adjectival sense, as in John went to a nearby restaurant. The suggestion is ingenious, and would capture the data described thus far, without recourse to subject-based interpretations. I think, however, that it is not adequate to draw the distinction between predicative and adjectival uses, for there are some adjectival uses that also block collection. Consider, for example John is in a nearby restaurant. Let us suppose that Bill in St Louis utters John is in a nearby restaurant and Mary in Portland utters James is in a nearby restaurant. We cannot conclude that both John and James are in nearby restaurants. The simple change of verb from went to/is going to to is in sufces to block the collection. Nearby here can only be interpreted as
Page 157
157
FN:8
nearby a location salient in the context of utterance. It is, however, occurring in adjectival form. This shows that the situations in which nearby does not allow for collection are not limited to ones in which nearby occurs as a predicate. Whether we use to be in rather than to go to makes the difference between predications with nearby that can be collected and ones that cannot. It is hard to see how this could be explained by positing ambiguities in nearby. The model here described predicts this outcome; while one can go to a restaurant that is near ones location, one cannot be in a restaurant that is near ones location, since one is located in the restaurant. An item cannot be near another item if the former is located within the latter. In John is in a nearby restaurant, nearby cannot mean nearby John; John is in the restaurant, and so cannot be nearby the restaurant. The subject-based interpretation is unavailable, and so cannot be invoked in interpreting the sentence. Collection, as in John and Bill are both in nearby restaurants, is therefore unacceptable. My account lets us see why the small change in verb can make the difference between collected predications being acceptable or unacceptable. The distinction between subject-based interpretations and context-based ones, then, is an important one. If we posit that context-sensitive items such as nearby and left can receive subject-based interpretations when such interpretations are acceptable, we are able to explain why some collections are permissible, while others are not. I take this to be good reason to assume that some contextsensitive items can receive these subject-based interpretations, and that collected predications across different contexts are permissible only when the items receive subject-based interpretations. There is no reason to suppose that the intermediate items do not behave similarly. I suggest that the intermediate items sometimes receive subject-based interpretations, and that we can collect predications containing them only when they receive these subject-based interpretations. As we have seen, there are circumstances under which we cannot collect intermediate items such as ready. If a particular activity is salient in a context, then we cannot collect predications of readiness, unless of course they pertain to that activity. We saw this in Johns APA Nightmare; in the context of that scenario, we cannot collect the true utterance of John is ready (said in connection with dinner) and the true utterance of Mary is ready (said in connection with APA interviews) to obtain John and Mary are both ready. Such a collection is impermissible in that context. This, I claim, is because the context-based interpretation of ready is favored, since the context of the collection supplies a salient activity. Similarly, if we are hungrily seeking a restaurant in Birmingham, we cannot say both John and Bill went to nearby restaurants last night unless the restaurants in question are near to our current location.
Stanley (2000, 2002) offers an important and detailed treatment of intermediate items such as tall along these lines.
Page 158
158
Sarah-Jane Leslie
FN:9
If, however, the context is one of indifference, where no salient value for the context-sensitive item is provided, then we can collect predications containing intermediate items, such as ready and tall. Since we have established that some context-sensitive items permit subject-based interpretations, I propose that we understand the intermediate items to allow subject-based interpretations where appropriate. Just as with nearby, we can collect predications containing intermediate items across different contexts only when those items receive a subject-based interpretation. So, for example, we understand the collection both John and Bill are ready to mean roughly that both John and Bill are ready for activities that are salient for John and Bill respectively. Notice that we can collect predications containing nearby even in contexts of ignorance, so long as the subject-based interpretation is available. If Tom utters John went to a nearby restaurant last night, and George utters Bill went to a nearby restaurant last night, then I might collect the two utterances and say Both John and Bill went to nearby restaurants last night. I might do this even if I have no idea where either John or Bill happen to be. My collection is understood as meaning Both John and Bill went to restaurants that are nearby their respective locations, whatever those may be. I suggest that we understand collected predications containing intermediate items in contexts of ignorance and indifference in a similar way. If Tom utters John is ready and George utters Bill is ready, then I might collect these and say Both John and Bill are ready, even if I have no idea what they are ready for. We understand the collection to mean that John and Bill are ready for activities that are salient for them both respectively, whatever those activities may be. In general, the intermediate items are no worse off than items such as nearby with respect to the Report and Collection tests. In short, once we recognize the possibility of subject-based interpretations of context-sensitive items, we need not think that the possibility of collection counts against an items being contextually sensitive. We should not assume that, just because we might say A and B are F, this means that we are predicating the very same property of both A and B. This would be akin to arguing that we predicate the same property of every girl when we say every girl loves her mother (Stanley 2005), or arguing that John and Bill must desire the same thing if we are to be able to say both John and Bill want to be department chair. As a nal remark on the possibility of collecting the intermediate items, it should be noted that the predicate is ready does not allow collection in all cases. Consider, for example, the impermissibility of the following collection:
Just as in the Report test, not everyone is as condent of the data as CL are. One might reasonably doubt that such collects really are acceptable. I will once again respond to CL on their own terms, however, and show how, even if these data are correct, items such as ready may nonetheless be contextually sensitive.
Page 159
159
My guests have arrived for dinner, and are very hungry. I ask them if they are ready to eat, and they reply yes, we are ready. I call into the kitchen to check on the status of the beef tenderloin, and the cook replies it is ready. Wonderful, I think to myself. My guests and the tenderloin are both ready. We cannot collect in these circumstances. My guests and the tenderloin are ready is no more acceptable than the zeugma Mary left in a huff and a taxi. As linguists have long noted, ready allows for both tough readings and control readings, hence the ambiguity of the goose is ready to eat. On the control reading, the subject the goose controls the unpronounced subject of to eat; this is the interpretation according to which the goose is ready to eat, in the same sense that ones guests may be ready to eat. On the so-called tough reading, the subject is identied with the unpronounced object of to eat; in this sense, the goose is ready to eat in the way that a beef tenderloin may be ready to eat. Just as we cannot accept my guests and the beef tenderloin are ready to eat (except on an interpretation that calls to mind Hannibal Lecter and the like), we cannot accept my guests and the beef tenderloin are ready. The explanation of the failure of collection when the innitival clause is articulated is straightforward: we cannot collect predications that differ so in their underlying syntax. Since collection fails in exactly the same way when the innitival clause is not articulated, this suggests that we are nonetheless at some level representing the innitival clause and its dependence relations to the subject. This suggests that there is no such predicate as is ready, periodthere is always some completion required, though not always articulated. The syntactic nature of the completion determines whether collection is permissible. This point is, of course, specic to ready. It is telling, though, since these data strongly suggest that we represent the object of the readiness, even if we do not articulate it. If this is so, then the Report and Collection tests are misleading in the case of ready; CL use these tests to argue that there is a simple, invariant predicate is ready, which predicates a single property of its subject, no matter the context. The syntactic data above suggest this cannot be so, thus these tests cannot be trusted. Instead, we should rely on the ICD/RCSA test, which correctly characterizes ready as contextually sensitive.
So-called because sentences such as John is tough to please are only interpreted in this way. John is tough to please can only mean that it is tough for the arbitrary person to please John; it cannot mean that it is tough for John to please the arbitrary person. The point is made most clearly when we try to collect predications of readiness across people and inanimate objects. This is because inanimate objects almost invariably occur with tough interpretations; it is hard to imagine cases in which we might say of an inanimate object that it is ready to itself undertake an action. People, on the other hand, are usually said to be ready to do such-and-such a task. The failure of collection can occur when all the subjects are people, however. Consider: John is ready to run a marathon. Bill is ready to shoot a basket. George is ready to fall down. They are all ready. This collection strikes us as a sort of bad joke, a sign of zeugma.
FN:10
FN:11
Page 160
160
Sarah-Jane Leslie 4 . C O N C LU S I O N
Cappelen and Lepores Insensitive Semantics forces us to think carefully about the phenomenon of semantic context sensitivity. What are our standards for counting an item as contextually sensitive? Are these so lax that all of natural language turns out to be context-sensitive? And how do we explain the behavior of context-sensitive items in a variety of linguistic constructions, across a variety of contexts? I have argued that Cappelen and Lepores ICD/RCSA test should be a benchmark for context sensitivity. It reects basic, constitutive facts about the nature of context sensitivity, and it sets the bar high enough that only a limited range of items are able to pass it. With that test in hand, we are able to draw a principled distinction between those items that are context-sensitive and those that are not. We are no longer in no danger of sliding down the slope into Radical Contextualism. Since the ICD/RCSA test should be taken seriously, a challenge emerges. There are a fair number of items that pass the ICD/RCSA test, and yet can occur in disquotational reports and collected predications, both across different contexts. How is this possible, if these items are contextually sensitive? I have tried to sketch out how this might happen, and in doing so, I hope to have shed some light on the phenomenon of non-indexical context dependence. Cappelen and Lepore are right to point out to us the prima-facie incompatibility of an items being contextually sensitive, yet amenable to disquotation and collected predication. It is no simple matter to understand how this is possible, and we are indebted to Cappelen and Lepore for pointing out that this is a phenomenon in need of explanation. Finally, as a meditation on semantic methodology, Insensitive Semantics reminds us time and again that intuitions about what is said are poor guides to semantic content. This point, so often overlooked, is an important one for semanticists to take on board. I have attempted here to extend Cappelen and Lepores work so as to provide tests and standards for context sensitivity that do not depend intuitions about what is said. On the importance of developing and adhering to such tests, I am in full agreement with Cappelen and Lepore. I differ from them only on exactly which items pass the tests. 5 . A P PE N D I X : M I N I M A L I S M A N D C O M M U N I C AT I O N Cappelen and Lepore level three charges against the contextualist: they argue that she classies items that fail their three tests as context-sensitive, that she cannot account for the ease with which we communicate with each other, and that her
Page 161
161
FN:12
theory is internally inconsistent. In the main body of this paper, I argued that Moderate Contextualism is indeed a stable position, contra CL, and that it does not count any items as context-sensitive if they do not pass their ICD/RCSA test. I also provided an account of non-indexical context sensitivity that explains why some of the Moderate Contextualists context-sensitive items might appear to fail the Report and Collection tests. Thus I take myself to have answered CLs rst objection to contextualism, namely that it counts items that fail their tests as context-sensitive. I will not address their third objectionthat contextualism is internally inconsistentbecause the objection applies only to theories that hold that all sentences are contextually sensitive. That is, it is only an objection to Radical Contextualism, not to Moderate Contextualism, and my concern here is only to defend Moderate Contextualism. In this appendix, I will consider CLs claim that only their view, Semantic Minimalism, is able to explain how we can successfully communicate across contexts, or perhaps even within a given context. I will argue that Semantic Minimalism is hard pressed to account for our communicative practices. Simply put, the minimal proposition that is semantically expressed is far too minimal to be what we care about in communication. If Semantic Minimalism was the correct view of semantics, then it would follow that we would almost never be concerned with semantic content in communication, but rather with speech act content. On CLs view, a huge range of propositions are expressed by every utterance, though only one of those propositions is the semantic content of the utterance, while the others constitute the remainder of its speech act content. If their view was correct, our intuitions about even the truth and falsity of an utterance would often fail to track the truth value of the proposition semantically expressed, and would rather track one of the many propositions that are merely said, not semantically expressed. Since our intuitions about the truth and falsity of an utterance may often fail to track its semantic content, the minimal proposition that is semantically expressed does not explain our ability to communicate with one another in the way that CL claim it does. CL introduce their objection to contextualism by writing:
If RC [Radical Contextualism] were true, it would be miraculous if people ever succeeded in communicating across diverse contexts of utterance. But there are no miracles; people do succeed in communicating across diverse contexts of utterance with boring regularity. So, RC isnt true. Only slightly more elaborated, it goes like this: If RC were true, then whats said by an utterance by a speaker A in context of utterance C depends, at least in part, on very
The minimal proposition expressed by a sentence S is the semantic content of S, and if S does not contain any members of the basic set, then the minimal proposition expressed by S is just the proposition that S. Remember that CL take themselves to have established that Moderate Contextualism collapses into Radical Contextualism; that is why they are addressing their objection only the Radical Contextualist.
FN:13
Page 162
162
Sarah-Jane Leslie
specic features of C . . . In sum, if RC were true, it would be a miracle if speakers in difference contexts were ever able to agree, disagree, or more generally, share contents. (2005: 123; my emphasis)
FN:14
It is not easy to understand exactly what the objection is here, since on their own view what is said by a given utterance depends on specic features of the context of the utterance (93). (In fact, what is said by a given utterance even depends on features outside the context of the original utterance; see their rejection of Original Utterance Centrism.) CL devote the entire last chapter of their book to describing and emphasizing how varied and unconstrained speech act content, or what is said, turns out to be, once we look closely. What is said, on their view, depends heavily on features of the context of utterance, among other things. How, then, do CL envision their objection applying to contextualism, but not to their own view? It is their minimal proposition that they suppose gives their account an advantage over its alternatives. This minimal proposition that S the proposition that is semantically expressed by an utterance of Sis always among the indenitely many propositions expressed by an utterance of S (2005: 205). It is the availability of this context-invariant proposition that allows us to communicate with each other across contexts. They write:
Semantic Minimalism, and no other view, can account for how the same content can be expressed, claimed, asserted, questioned, investigated, etc. in radically different contexts. It is the semantic content that enables audiences who nd themselves in radically different contexts to understand each other, to agree or disagree, to question and debate with each other. It can serve this function simply because it is the sort of content that is largely immune to contextual variations. (2005: 152)
Since it is their minimal proposition that CL believe enables us to communicate across contexts, let us consider the nature of these minimal propositions, so as to see whether they can indeed serve this purpose. In particular, let us consider the proposition that CL believe is semantically expressed by sentences containing the intermediate items, such as John is ready. CL hold that John is ready expresses the proposition that John is ready, which is true iff John is ready (2005: 155). The question that naturally arises is, what is the nature of the property of being ready? When does a person or object posses this property?
They write: Whats crucial to notice here (and in general) is that our intuitions about what speakers say with their utterances are inuenced by, at least, the following sorts of considerations: (1) Facts about the speakers intentions and beliefs . . . (2) Facts about the conversational context of this particular utterance . . . (3) Other facts about the world . . . (4) Logical Relations (193). They also state that theres no reason to think that [these factors] exhaust all the factors that inuence our intuitions about what speakers say (194). Finally, its important to note that they take our non-theoretic beliefs about intuitions about what speakers say, assert, claim, etc, at face value (191).
Page 163
163
CL are adamant that a semanticist should not have to answer such metaphysical questions. Nonetheless, they sketch how they imagine an analysis of this property would proceed:
Think about what metaphysicians do. For at least the last two millennia, metaphysicians have been asking What-Do-they-Have-in-Common-Questions (CQ, for short). Suppose youre curious about what it is to be G. Then you ask (this is at least one of the questions you ask): (CQ) What do all G things have in common? . . . Think about what people who are ready have in common. To make this vivid, imagine As being ready to commit a bank robbery, Bs being ready to eat dinner, and Cs being ready to take an exam. Thinking about A, B and C, youve got two options: a) You might think, as we do: Well, they have a common relation they stand in to their respective projects: Theres something in common between As relation to the bank robbery, Bs relation to the dinner and Cs relation to the exam. What they have in common is that they are all ready. b) Alternatively, you might think that theres nothing these people have in common. The fact that we would describe them as all being ready for their various projects doesnt mean that they have anything whatsoever in common. Theres no state of readiness that they share with respect to their respective tasks. We nd (a) overwhelmingly plausible. Its not just a pun that we feel comfortable describing them all as being ready. They really are all ready. Thats different from their all being done with the tasks, or excited about them, or prepared for them, or good at them, etc. (15967)
So CL conclude it is this shared property that is ready expresses. A has it in virtue of being ready to rob a bank, B has it in virtue of being ready to eat dinner, and so on. Generalizing, we have that being ready is a property that one possesses in virtue of being ready to do some task or other. Any time A is ready to do X, for any X at all, then A is ready. One should wonder that this does not make it very easy to be ready. CL are aware of this, and though they do not specically address this question with respect to readiness, they say of their corresponding claim concerning enough:
The following concern might now be raised: Doesnt that make it very easy to have had enough? If thats all it takes, havent we all had enough all of the time? Suppose the answer is yes (though we have no idea whether this is correct or not; presumably it all depends on doing more serious metaphysics, but suppose its correct). When you think real hard [sic.] about enoughness, maybe thats all it takes. If so, then its not that hard to have had enough. (168 fn.)
Thus while CL hedge slightly in their parenthetic remarks, they are quite open to the above metaphysical analysis of properties such as being ready. Let us for now assume that this is indeed their intended account of the property of being ready, and therefore than the proposition that John is ready is co-extensive with
Page 164
164
FN:15
Sarah-Jane Leslie
FN:16
the proposition that John is ready for something. The question now arises: is this proposition the one that we care about in communication? CL intend this minimal proposition to be one that is communicated by all utterances of John is ready. Now, it is not in dispute that all utterances of John is ready communicate that John is ready for something. (CL sometimes write as though the contextualist would deny this, but I do not see why this would be so.) Even if contextualism is true, the truth of John is ready entails that John is ready for somethingi.e. if John is ready for some contextually salient activity, then he is ready for something. It is quite reasonable to think that a proposition that is at least co-extensive with CLs minimal proposition is communicated by all utterances of John is ready, no matter which view of semantics one adopts. But what are we to make of utterances of Its false that John is ready? Surely the proposition semantically expressed by its false that John is ready is the negation of the proposition expressed by John is ready. But if John is ready semantically expresses a proposition that is co-extensive with John is ready for something, then its false that John is ready must semantically express a proposition that is co-extensive with not: John is ready for something. This proposition is false if there is anything at all that John is ready for. But is it reasonable to think that we communicate such a proposition with every utterance of its false that John is ready? Surely this is not so. We simply do not interpret utterances of its false that John is ready as communicating anything this strong. We do not intend to convey this information by uttering such a sentence, and we do not glean this information from utterances of the sentence. Since CL urge us to take our pre-theoretic intuitions about what is said at face value, it is hard to see how such a proposition could be among those that are said by any utterance of its false that John is ready, let alone one that is said by every such utterance.
These propositions (and the properties that gure in them) are co-extensive, but not identical. I assume that CL adopt a structured approach to propositions so that this is possible. CL repeatedly argue as though the contextualist should want to deny that this is so, and claim that there is something in common between people who are all ready for their respective tasks. I must confess that I do not understand why they anticipate this line of objection. Clearly, there are many, many things that two people who are ready for different tasks have in commonthey are both self-identical, both complex material objects, both human beings, and of course, both ready to do something. No one would deny that there is something in common between two people who are ready to perform different tasks. The question is not whether there are some commonalities between these people, the question is we semantically predicate a common property of them when we say that A and B are both ready. Consider e.g. girls who all love their respective mothers. There are many properties shared in common between girls who love their mothersthey are all self-identical, they are all girls, they all love someone, and so on. This observation has no bearing on the semantics of each girl loves her mother. We do not conclude from the fact that these girls share common properties that we are semantically predicating one of these common properties of all of them when we say each girl loves her mother. That there are many properties shared between the objects of thought and talk is an observation that is completely orthogonal to issues concerning semantic context sensitivity.
Page 165
165
Even if we were to be somehow convinced that this proposition is always among those that are said by an utterance of its false that John is ready, it is clearly not the proposition that we care about in communication, since we are not even aware that we are communicating it to one another. Minimal propositions are poorly suited for the communicative work for which CL intend them. They are rarely, for example, our objects of disagreement. I discussed this point in Leslie (2004), and Hawthorne (2006) also makes this point. Let us return to Johns APA Nightmare to see this more clearly.
5.1. Disagreement
Recall that poor John is on the job market, and is thoroughly unprepared for his interviews. Looking at him, his thesis advisor says: Well, at least Johns ready. The department chair appropriately responds: Are you out of your mind? Hes clearly not ready! In other words, the chair adamantly disagrees with the advisors claim that John is ready. What is the basis for this disagreement? It is surely not that the chair takes John not to be ready for anything. He may well think John is ready to do a variety of things: pass out, bolt from the APA, etc. But: If the chair thinks John is ready to bolt from the APA, And if being ready to do something is sufcient for having the property being ready, And the property of being ready is what is semantically predicated of John by and utterance of John is ready, And if disagreements are over semantically expressed propositions, Then the chair would have no grounds to disagree with the thesis advisor. It is clear that the minimal proposition that CL endorse is not sufcient to explain disagreements. A corollary is that the minimal proposition is also not our object of debate and deliberation, since these notions are intimately tied to disagreement. When the department chair questions the advisors judgment, he is not questioning the truth of a proposition that is co-extensive with John is ready for something. It is simply not true that it is the semantic content that enables audiences who nd themselves in radically different contexts to understand each other, to agree or disagree, to question and debate with each other (152). The semantic content does not even serve this purpose within a context. In their paper, Shared Content, Cappelen and Lepore provide a variety of other purposes that shared content serves. I shall not go through every purpose they cite, but it is clear that their minimal proposition is simply not up to task. Consider, for example, the following two additional purposes:
Page 166
166
Sarah-Jane Leslie
5.2. Responsibility
Responsibility is another phenomenon that CL claim is based on shared content. Again the minimal proposition is not sufcient to bear the burden. If the interviewers ask Johns thesis advisor if John would like to take more time and reschedule the interview and the advisor responds, No, hes ready, then John may hold his advisor responsible for the consequences of this. It is no defense on the part of the thesis advisor that the proposition semantically expressed was true in virtue of Johns being ready to bolt from the APA.
5.3. Reasons for Action

CL point out that what others say often provides us with reasons for action. If Johns thesis advisor says no, Johns ready, the interviewing committee takes this as a reason for them to go ahead and interview John right away, rather than rescheduling his interview. The advisors utterance would not provide them with a reason to do this if it was not understood to mean that John is ready to be interviewed. If the advisor could truly utter Johns ready in virtue of Johns being ready to bolt from the APA, then his utterance would not provide the interviewing committee with a reason to go ahead with the interview. The minimal proposition clearly cannot explain the communicative phenomena that CL wish it to explain. The minimalist is no further forward than the Radical Contextualist in explaining communication. We might wonder to what extent the difculties we raised for the minimal proposition depend on the particular metaphysics of being ready that CL seem to adopt. I believe that any invariantist account of being ready will face similar difculties. Let us adopt a different analysis of being ready let us say that x is ready iff x is ready for the activity that is most salient to x. Then the minimal proposition expressed by John is ready will be co-extensive with John is ready for the activity that is most salient to him. In Johns APA Nightmare, this activity is clearly his APA interview, so John is ready here expresses a proposition that is co-extensive with John is ready for his APA interview. Notice that this is an invariant property; in every context, an utterance of John is ready expresses this same property, so John is ready is not contextsensitive on this picture. This account offers a much better rendering of the situation described above; an utterance of John is ready will be false, and any disagreement over whether John is ready amounts to disagreement over whether he is ready for his interview, and so on. It seems that we have the correct result so far.
CL suggest an analogous analysis of being tall on at 2005: 171, so it is not a stretch to imagine that they might adopt such an account of being ready.
FN:17
Page 167
167
FN:18
Unfortunately, it is easy to replicate the difculties we encountered earlier. Imagine, for example, that two evangelists are hovering around the APA, looking for lost souls that are ripe for conversion. They discreetly observe John, and it is manifestly clear to them that he is not ready for his interview. Neither of them doubts this fact. The evangelists come to discuss whether the time is right to approach John to discuss his accepting Jesus Christ as his Personal Savior. Is John ready to be saved, they wonder? The rst evangelist says, Take one look at him. Hes lost and bewildered. This guy is denitely ready. The second expresses doubt, saying I disagree. Lets wait and see how things unfold. If his interview goes badlyas it surely willthen hell be ready. Right now, I just dont think hes quite ready. The evangelists are in disagreement over whether John is ready, though neither of them doubts that John is not at all ready for his interview. Since Johns interview is the most salient activity for him at the time, if the above account of being ready was correct, the evangelists would not be in disagreement over whether John is ready. The minimal proposition would, once again, fail to be the object of disagreement. Similarly, if the second evangelist insists, saying No, John is just not ready, and as a result they miss a perfect opportunity to convert John, the rst evangelist may hold the second one responsible for what he said. It would be no defense for the second evangelist to claim that what he said was semantically true, because John was clearly not ready for the activity that was most salient to him. A minimal, invariant proposition is simply not sufcient to explain our communicative practices. We often do not care about the minimal proposition even within a given context. It is thus also rarely what we care about in communication across contexts. (Years later, John might confront his thesis advisor, demanding to know why he told the interviewing committee that John was ready, when he obviously was not. The thesis advisors defense that John was ready for something, namely to bolt from the APA, is no more convincing in this new context than it was in the original context.) If Semantic Minimalism were the correct view, we would be no further forward in explaining communication than we would be if Radical Contextualism were true. On both theories, communication depends on what is said, and what is said is an elusive, context-dependent phenomenon.
On the Moderate Contextualists view, the proposition that is semantically expressed is quite plausibly what we care about in communication, at least in many cases. Certainly, it is easy to see how it is the object of agreement and disagreement, debate and discussion in the scenarios described above. Of course, if Moderate Contextualism is correct, then the proposition that is semantically expressed is, of course, dependent on features of the context. It depends on them in more constrained ways than CLs speech act content does, however. In the main body of this chapter, I sketched an account of how we make use of preambles and the like to help us interpret reports of utterance that were made in different contexts, and how we interpret these reports if we do not know anything about the context of the original utterance. It need not be wholly mysterious how we could communicate across contexts if Moderate Contextualism were true. Still, cross-contextual communication is a remarkable phenomenon, and it is not obvious on any account how it proceeds.
Page 168
168
Sarah-Jane Leslie
Semantic Minimalism does not explain how we are able to communicate with each other any better than its alternatives. CL therefore ought not to criticize contextualism on these grounds, since their criticism applies equally to their own theory. Nonetheless, CL are absolutely correct to point out that there is a phenomenon here that needs to be explained. Given that what we care about in communication depends so heavily on contextual features, how is that we are able to understand each other so well? How can we share content across contexts, given how variable the content we care about turns out to be? At this time, we have no satisfying answers to these questions, and we are indebted to Cappelen and Lepore for reminding us that this is so.
REFERENCES Cappelen, H., and E. Lepore (2005), Insensitive Semantics. Oxford: Blackwell. and (2006a), Replies to Bach, Hawthorne, Korta/Perry and Stainton, Philosophy and Phenomenological Research. and (2006b), Reply to Critics. Mind and Language, 21/1: 5073. and (2006c ), Shared Content, in E. Lepore and B. Smith (eds.), Oxford Handbook of Philosophy of Language, 102056. Oxford: Oxford University Press. Hawthorne, J. (2004), Knowledge and Lotteries. Oxford: Oxford University Press. (2006), Testing for Context Dependence, Philosophy and Phenomenological Research. Leslie, S. J. (2004), Comments on Cappelen and Lepores Shared Content , Conference on Language and Linguistics, University of Connecticut. Recanati, F. (2006), Crazy Minimalism, Mind and Language, 21/1: 2134. Stanley, J. (2000), Context and Logical Form, Linguistics and Philosophy, 23/4: 391434. (2002), Nominal Restriction, in G. Peters and G. Preyer (eds.), Logical Form and Language, pp. 36588. Oxford: Oxford University Press. (2005). Knowledge and Practical Interests. Oxford: Oxford University Press.
I rather suspect that a satisfying account of the phenomenon will belong to psychology, not to semantics.
"#$%&!'$()*!+,&-.$/-!,0!1/2&34%560$7! 8!
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
N
!.$):#$#0-1(!1<!0%-+!#$0-/,)!?#+!+6::1$0);!E4!Z"[!7$#(0!\D"&NLL]^_L@!P%#(=+!01!*16-+)!`(0%1(4!<1$!#+=-(7!a6+0! ?%4!H!$)a)/0!:$#78#0-/!#//16(0+!1<!(1$8#0-3),4!+%-<0);!:$);-/#0-1(+G!?-0%!#!$)+6,0#(0!B/18:,-/#0-1(!1<!0%)!,)>-/1(F@!! !
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
D
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
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
7
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`)%@7)@"8&88&YZZ[YTa>& & ;)-02,%8&_>&]>8&YTTZ8&9A-7#'&;,%*.#.,%2&,3&_"%"-.6&P"%#"%6"2W&A$,&;,%#-)2#.%@&b."$2=8&.%&_>&]>&;)-02,%&& )%*&V>&c>&X"00"#."-&Q"*2>U8&A'"&_"%"-.6&L,,E>&;'.6)@,W&H%./"-2.#1&,3&;'.6)@,&X-"228&55>&RRd[Ref>& & ;,'"%8&J>8&RSSY8&9<%&#'"&@"%"-.6&72"&,3&.%*"3.%.#"&2.%@70)-2>=&c,7-%)0&,3&P"()%#.628&879:;8&Yfe[RST>& & M%,!"8&c>8&N&X-)2)*)8&P>8&RSYY8&9C7)0&;')-)6#"-&;,%6"5#2=8&?%&X-,6""*.%@2&,3&#'"&ee-*&J%%7)0&;,%3"-"%6"& ,3&#'"&;,@%.#./"&P6."%6"&P,6."#1>&L,2#,%8&gJW&;,@%.#./"&P6."%6"&P,6."#1>& & M%,!"8&c>8&N&X-)2)*)8&P>8&27!(.##"*8&9C7)0&;')-)6#"-&;,%6"5#2&)%*&#'"&],-()#./"&C.("%2.,%&,3& ;,%6"5#7)0&\"5-"2"%#)#.,%=>& & `)E,338&_>&QYTfÛ>&<)+%#=&>"*%=&,#0&?,#@%*)/!&A("#@!B&;'.6)@,W&H%./"-2.#1&,3&;'.6)@,&X-"22>& & `)$0"-8&c>8&YTê8&9P#7*."2&.%&h%@0.2'&@"%"-.62>=&H%./"-2.#1&,3&g.6'.@)%&X)5"-2&.%&`.%@7.2#.62&YQYU>& &
!"#$%&'()*(+,--./*(01$2(%13456$2(76$(8"6$(4"93%74(:172(75;4:<=(>?@$638$2:74(32#$<:397:3"2<("A(<%7476( 38@43%7:;6$<*(B"923:3"2'(!"'(.CDE.FF*( ( G";<%";4";<'(!*'(!"#$%&'()*'(G"43:H$6'(I*'(J(K7<:35$'(L*(+,--M/*(G6"%$<<329(%"<:<(725(38@43%7:;6$( 5$#$4"@8$2:*(N729;79$(L%O;3<3:3"2'(#$'(PQMEPMD*(
!"#$$%&'()$#*+,*(#-(+./(,*$'(0%*(#*(+./(12%0%(%30#*#-+&%+#,*45( 67%$().%&%8+/&(),*8/9+-:(;/*/&#8-:(%*3(;/*3/&(
<%&%.=>%*/(?/-$#/ @(
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àWa`bS0+>0Dc'7)D4*)DA(.45D+:$-$D:*D$D -$#45D-$#K<!! ! d(#$4451!)+!:+..+>!$#!*P$-?4*!,.+-!$!>8+445!0(,,*.*#)!$.*#$1!7*@*.$4!*P%8$#A*7!(#!E8$/*7?*$.*e7! 83!6'+/!8$@*!)8*!7$-*!%8$.$%)*.<!d+.!*P$-?4*1!(#!"%)!f!7%*#*!@((1!F$%:*)8!*P?.*77*7!8(7!.*4'%)$#%*!)+! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
W
!"!7?*%($4!)8$#/7!)+!F$./!^+8#7)+#1!^+78!g#+:*1!M$..5!L$-1!Z$%8*4!L*74(*1!E$445!F%6+##*44DG(#*)1!E$#0**?! N.$7$0$1!$#0!R$@(0!E+7$!,+.!)8*(.!@*.5!8*4?,'4!%+--*#)7!$#0!7'AA*7)(+#7!%+#%*.#(#A!)8(7!?$?*.<!! I !f!$-!A.$)*,'4!)+!R$@(0!E+7$!,+.!)8(7!*P$-?4*<!
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``@&U-#567).&^__a@&U74%).&[\\^@&Z#%%)2%#*.&[\\b@&/-85%#.&^_`b@&1,U7))%55XZ()%3.&()& :#%66M9&c33%#-),%6&7H&34%6%&*%)%#(,6.&347"*4&34%0&87"5$&)73&6%%!&37&2%&!%3-:47#(,-5&()&-)0& 63#-(*43H7#8-#$&6%)6%.&6%%!&37&2%&")#%6:7)6(G%&37&34%&-,3"-5&$(63#(2"3(7)&7H&34%&:#7:%#30&-!7)*&34%& !%!2%#6&7H&34%&'()$9&<I706&$7)T3&,#0B&7)&34%&#%5%G-)3&()3%#:#%3-3(7)&(6&)73&,4-55%)*%$&20&34%&%!:(#(,-5& H-,3&34-3&2706&!763&,%#3-()50&$7&,#0.&7#&%G%)&34-3&2706&()&H-,3&,#0&d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
@
'K)+)'O'0#'"/(B%."/.'"/'2/)'-)+!"2/'28',4)'$4)/2#)/2/',40,',4"!'$0$)+'0,,)#$,!',2'"%%B#"/0,);''
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
8
(P%(+*(*"6$%+6$*(!&)+6$0(%.)%(",&'(+,0$5+,+%$(*+,<-&)#(<$,$#+!*(E$7<7(1)(2"'(0"$*,Y%(!#'3("#(1)(5#+$,0(0"$*,Y%(&$%()( 5#+$,0(0#+>$(0#-,=3F(!),(.)>$(,"#6)%+>$(5"#!$4(."/$>$#()*(%.$*$($:)6;&$*(G(),0("%.$#*(<+>$,(%.#"-<."-%(%.$( ;);$#(*-<<$*%(G(%.+*(+*(,"%(),()!!-#)%$(!.)#)!%$#+Z)%+",("5(%.$(;.$,"6$,",7(E["/$>$#(+%(0"$*(*$$6(%.)%(%.$( !",>$#*$(."&0*(G(%.$*$(+,0$5+,+%$(*+,<-&)#(<$,$#+!*(.)>$(",&'(%.$(,"#6)%+>$4(,"%(%.$(0$*!#+;%+>$4(+,%$#;#$%)%+",7( 9.+*(+*(0+*!-**$0(5-#%.$#(+,(5""%,"%$(\7F%
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
T
%%U,(%)(&*+,%?+)%-"(%S5&0#?#$&-#+,*%#*%-"&-%-"(%10+**%E(.(/'0#?#()%+?%-"(%#2(&0*%&**+$#&-(2%!#-"%3(#,1%&%OF%2+(*%,+-% #-*(0?%(.'0&#,%-"(%2#??#$50-9%!#-"%?#,2#,1%&%,+)/&-#B(%)(&2#,1%?+)%E/+*-%3+9*%2+,L-%$)9F%&,2%E*+/(%3+9*%2+,L-%$)9F6%% U,%-"(%+-"()%"&,2:%9+5%&*(%1(-%-"(%,+)/&-#B(%)(&2#,1%#,%-"(%,(1&-#B(%(.#*-(,-#&0%$&*(:%$+,*#2()%E-"()(%&)(%,+%/(,% 7&,9/+)(=F6%R#D(!#*(%!#-"%E-"(%+,09%/&,%#,%-"(%)++/%*&#2%!(%*"+502%1+%-+%!&)%!#-"%@)&,F%7)(?())#,1%-+%;#00&)9% <0#,-+,=%&,2%E-"(*(%1#)0*%-"&-%)5,%-"(%J-&-(%V('&)-/(,-%-"#,D%!(%*"+502%$&'#-50&-(F:%+)%(B(,%E/+*-%+?%-"(%0&2#(*% !"+%)5,%-"(%*-&-(%2('&)-/(,-%-"#,D%!(%*"+502%$&'#-50&-(F:%!"#$"%&0*+%*5''+)-%-"(%*-)&#1"-?+)!&)2%10+**%#,%-()/*% +?%E(.(/'0#?#()%+?%-"(%#2(&0*%&**+$#&-(2%!#-"%3(#,1%&%OF6%@%(,-()%#,-+%/+)(%2(-&#0%+,%-"(*(%'5)(09%*(/&,-#$%S5(*-#+,*% #,%R(*0#(%7?+)-"$+/#,1%&=6%
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
N
%V-,%&-,$%)*.;<..*-'%-+%2-3(.$&(%*'%!1$%.;*$'!*+*;%;-'!$4!9%.$$%S$.3*$%J+-,!1;-&*'=%6L/%
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
[
#!#;.""4<,+#(.7)*+&+D%2;,+#*.#*-4"#(,%42#24=-*#"++2#*.#<+#$.7)6#4)#%#(%"+#6+"(&4<+6#<8#\.-+)#AOPPQB1#<74,64)=#.)# %)6#%6%;*4)=#%#64"(7""4.)#4)#\%&,".)#AQ]]RBJ#"7;;."+#*-%*#%#"*.&+#2%)%=+&#"+*"#*-+#;&4(+#.$#<%)%)%"#*.#ÔM,<1#<7*# 4)(.2;+*+)*#(%"-4+&"#.),8#(-%&=+#*-+#(7"*.2+&"#^QM,<>#E+#24=-*#(.)"46+&#*-+#;%4&#.$#=+)+&4("#5<%)%)%"#"+,,#$.&# Ô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Ô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ÔM,<>B##
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
L
&N#(,&3,-/*2%&/"#8-&.#1&-#&8+3,(%-/+3&-.,%,&5,+,(*'%&'/+&",&4#8+3&*+&O,%2*,&H4#(-.'#6*+5&/I;&P,($&"(*,42$7&*+& 9(,0*#8%&1#(:&HO,%2*,7&QRRS7&QRRLI7&C&/(58,&-./-&%#6,&5,+,(*'%&*+0#20,&9(,3*'/-*+5&%&'('%)*(+,)+%&9(#9,(-*,%&#4&-.,& :*+3&B&4#(&,D/692,%&-.,&5,+,(*'&!38':%&2/$&,55%)&*%&6/3,&-(8,&"$&-.,&4/'-&-./-&2/$*+5&,55%&*%&-.,&'./(/'-,(*%-*'& 6,/+%&"$&1.*'.&38':%&(,9(#38',;&G.,%,&5,+,(*'%&/(,&+#-&9/(-*'82/(2$&(,%9#+%*0,&-#&&"-$.'!/&6,6",(%&#4&-.,& :*+3&./0,&-.,&9(#9,(-$7&/%&!38':%&2/$&,55%)&*228%-(/-,%&H%*+',&6#%-&38':%&H*;,;7&6/2,7&*66/-8(,7&#(&*+4,(-*2,&38':%I&3#& +#-&2/$&,55%I;&J-.,(&5,+,(*'%7&,;5;&!'/(%&./0,&(/3*#%)7&/(,&6/3,&-(8,&"$&-.,&6/K#(*-$&#4&'/(%&./0*+5&(/3*#%&H4#(&6#(,& 3*%'8%%*#+7&%,,&/2%#&>(/%/3/&T&U*22*+5./67&QRRV7&QRRW@&>(/%/3/7&=.,62/+*7&O,%2*,&T&X28':%",(57&%8"6*--,3I;&HG.*%& 3*%-*+'-*#+&3#,%&+#-&/6#8+-&-#&/+&/6"*58*-$&*+&-.,&5,+,(*'%7&.#1,0,(7&"8-&(/-.,(&%*692$&-#&3*44,(,+-&1/$%&-./-&-.,& 1#(23&'/+&6/:,&/&5,+,(*'&-(8,;I&!Y#$%&'($)&*%&-.,&2/--,(&%#(-&#4&5,+,(*'&B&!"#$%)&.,(,&48+'-*#+%&3,%'(*9-*0,2$7&/+3& -.,&5,+,(*'&*%&6/3,&-(8,&"$&-.,&4/'-&-./-&-.,&6/K#(*-$&#4&"#$%&'($;&!Y#$%&3#+?-&'($)&*%&/&5,+,(*'&#4&-.,&4#(6,(&%#(-&B& !"#$%)&.,(,&./%&*-%&+#(6/-*0,7&*3,/2&%,+%,7&/+3&!")$%(/+!0$*%&/&'./(/'-,(*%-*'&9(#9,(-$&#4&-.*%&*3,/2&+#-*#+&#4&/&"#$;& Z./-&'#+%-*-8-,%&/&'./(/'-,(*%-*'&9(#9,(-$&#4&/+&*3,/2&*%&3*%'8%%,3&/-&2,+5-.&*+&-.,&4#22#1*+5&%,'-*#+%;$ N#(,&5,+,(/22$7&5,+,(*'%&-./-&*+0#20,&'./(/'-,(*%-*'&9(#9,(-*,%&'/+&#''8(&1*-.&,*-.,(&1'(*$23#('3$,#14*%),& H,;5;7&!"#$%&3#+?-&'($)&#(&!38':%&2/$&,55%)I&#(&1*-.&+!5*6+!+)*$,+!0#3'(&,#14*%),&H,;5;&!/&"#$&3#,%+?-&'($)&#(&!/&38':& 2/$%&,55%)I;&[#1,0,(7&5,+,(*'%&-./-&*+0#20,&+#+M'./(/'-,(*%-*'&6/K#(*-$&9(#9,(-*,%&'/+&#+2$&",&4#(682/-,3&1*-.& "/(,&928(/2&%8"K,'-%\&!/&'/(&./%&/&(/3*#)&#(&!/&"#$&'(*,%)&/(,&3,'*3,32$&%-(/+5,&H#(&,2%,&-/:,&#+&/&3*44,(,+-&6,/+*+5& /2-#5,-.,(I;&]*6*2/(&(,6/(:%&/992$&-#&-.,&%,,6*+52$&'#+-(/($&9/*(%&#4&5,+,(*'%&3*%'8%%,3&.,(,&B&,;5;7&'#+%*3,(&-.,& 9/*(&!%'*,+-*%-%&'/(,&/"#8-&-(8-.)&/+3&!%'*,+-*%-%&3#&+#-&'/(,&/"#8-&-(8-.)@&#+2$&-.,&4#(6,(&'/+&",&9(#9,(2$& (,4#(682/-,3&/%&/+&*+3,4*+*-,&%*+582/(\&!/&%'*,+-*%-&'/(,%&/"#8-&-(8-.);&F+/2$^*+5&-.,%,&_+#(6/-*0,?&5,+,(*'%&/%& *+0#20*+5&'./(/'-,(*%-*'&9(#9,(-*,%&#4&/+&*3,/2&:*+37&/+3&-.,*(&3,%'(*9-*0,&'#+-(/(*,%&/%&*+0#20*+5&6/K#(*-$&9(#9,(-*,%& #4&/&3,%'(*9-*0,&:*+37&/22#1%&8%&-#&,D92/*+&-.*%&3*44,(,+',&/%&",*+5&9/(-&#4&-.,&6#(,&5,+,(/2&3*44,(,+',&",-1,,+& './(/'-,(*%-*'&/+3&6/K#(*-$&5,+,(*'%;&`#(&6#(,&3,-/*2%7&%,,&O,%2*,&H4#(-.'#6*+5&/I;&&
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
H
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
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
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
QQ
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
############################################################
9:
#P,/#*!'53!7!$6+#=#)'#@/!$!"4#)*#$%,14%#$%&/&#@!33#.&#)"#!0&"$!-!).3&#7/!$&/!,"#-,/#'&'.&/*%!5#!"#$%&#>!"0+#$%,14%#!"# ',*$#7)*&*#$%!*#!*#',/&#,-#)"#!0&)3!?)$!,"#/)$%&/#)"6$%!"4#&3*&8#
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
!"##$%&'()$*"'+()")',-'*$)(&.""/"%#0'1"'+(%'/(2"')"%)"',-'#3"'4.$/('-(+$"'%,./(#$5"'-,.+"',-'#3")"' &"%".$+)0'&$5"%'#3"'())6/4#$,%)'#3(#'6%*".7$"'#3"/8'93"'+,%%"+#$%&'4.$%+$47"'$)'),/"#3$%&'7$2"'#3"' -,77,1$%&:';-',%"'$)'('/"/<".',-'('),+$(7'2$%*0'(%*'#3(#'),+$(7'2$%*'3()'('4(.#$+67(.'4.$/(.='.,7"',.' -6%+#$,%0'#3"%'#3"."'$)'('4.$/('-(+$"',<7$&(#$,%'#,'-67-$77'#3(#'.,7"',.'-6%+#$,%0'(%*'*,'),'"--"+#$5"7=8';-'(' &$5"%'-"(#6."'$)'$/4,.#(%#'-,.'-67-$77$%&'#3$)'.,7"0'#3"%'"5"%'$-'#3"."'$)'%,'&"%".(7',<7$&(#$,%'-,.'4",47"'#,' 3(5"'#3"'-"(#6."0'#3$)'+."(#")'('4.$/('-(+$"',<7$&(#$,%'-,.'/"/<".)',-'#3"'2$%*'#,'4,))"))'#3"'-"(#6."8' 936)0'"5"%'$-'#3"."'$)'%,#3$%&'$%#.$%)$+(77='&,,*'(<,6#'."-.($%$%&'-.,/'+.=$%&0'><,=)'*,%?#'+.=@'+(..$")' %,./(#$5"'-,.+"'<"+(6)"'$#')6&&")#)'#3(#'#3"',%7='1(='#,'"--"+#$5"7='47(='#3"'/(7"'.,7"'$%5,75")'%,#' +.=$%&8'!$/$7(.7=0'(%'6##".(%+"',-'>('1,/(%'46#)'-(/$7='<"-,."'+(."".@'+(%'+,%5"='('."<62"'<"+(6)"'$#' )6&&")#)'#3(#'('1,.2$%&'/,#3".'$)'%,#'-67-$77$%&'3".'),+$(7'.,7"'4.,4".7=8';#'+,%5"=)'%,')6+3'."<62"'$-' *$."+#"*'#,1(.*)'('/(%0'3,1"5".A')$%+"'3"'$)'%,#'('/"/<".',-'#3"'."7"5(%#'),+$(7'2$%*0'3"'1$77'%,#'<"' 47(=$%&'3$)'.,7"'4,,.7='$-'3"'46#)'+(."".'<"-,."'-(/$7=8'B;%'-(+#0'#3"'."5".)"'/(='"5"%'<"'3"7*'()'#.6"8C'' ' D%'$%#".")#$%&'E6")#$,%'+,%+".%)'13"#3".',%"'/$&3#'-""7'#3"'%,./(#$5"'-,.+"',-'#3")"'&"%".$+)' "5"%'$-',%"'*""/)'#3"'4.,4".#='$%'E6")#$,%'#,'<"'&"%".(77=!"#$%&'()*+%8'F3$7"'$#'$)'$/4,.#(%#'#,'<"'+7"(.' #3(#'#3")"'&"%".$+)',%7='3(5"'('4.$/(!-(+$"',.'*"-"()$<7"'-,.+"'G'#3(#'$)0',%"'/$&3#')#$77'*"+$*"'#3(#'(77' #3$%&)'+,%)$*"."*'$#'$)'<"##".'#,'#,-'-67-$77'#3"'.,7"'$%'E6")#$,%'G';'#3$%2'$#'$)'47(6)$<7"'#3(#',%"'/$&3#'H-""7' #3"'%,./(#$5"'-,.+"?'"5"%'$-',%"'#3$%2)'#3(#'$%'&"%".(7',%"',6&3#'#,-!#,'3(5"'#3"'4.,4".#='$%'E6")#$,%8' 93(#'$)0')644,)"'),/",%"'<"7$"5")'#3(#0'$%'&"%".(70'(&&."))$5"'*.$5$%&'$)'%,#'#3"'),.#',-'(+#$5$#=',%"' ,6&3#'#,'"%&(&"'$%8'!644,)"0'3,1"5".0'#3(#'#3(#'4".),%'$)'('<6)'*.$5".0'(%*'-6.#3".'(++"4#)'#3"'.,7"',-'(' <6)'*.$5".'$)'#,'&"#')#6*"%#)'#,'+7())',%'#$/"'%,'/(##".'13(#0'(%*'#3(#'(&&."))$5"'*.$5$%&'$)'%"+"))(.='-,.' -67-$77$%&'#3$)'.,7"8';#'$)'%,#'-(.-"#+3"*0';'#3$%20'#,')644,)"'#3(#')6+3'('4".),%'/$&3#'-""7'#3(#'$%'),/"' )"%)"'#3(#'3"',.')3"',6&3#'#,'*.$5"'(&&."))$5"7=0'*")4$#"'I6*&$%&'#3$)'#,'&"%".(77='<"'4,,.'+,%*6+#8' BD&($%0'#3$)'$)'%,#'#,')(='#3(#'#3"'4".),%'/(='%,#'67#$/(#"7='*"+$*"'(&($%)#'*.$5$%&'(&&."))$5"7=8C' ' 93"."'(."'"5"%'/,."'*.(/(#$+'$776)#.(#$,%)'#,'<"'-,6%*8'D'-()+$%(#$%&'"J(/47"'$%5,75")'(%' $%#".5$"1'1$#3'('-,./".'4$/4'%(/"*'K"5$%',%'./'&!01%('2)#!3'4%8 LM'K"5$%'."+,6%#)')""$%&',#3".'4$/4)' 43=)$+(77='())(67#'#3"'4.,)#$#6#")'$%'#3"$.'H"/47,=?0'(%*'<"$%&'/,.(77='."467)"*'<='#3"$.'(+#$,%)'(%*' -""7$%&'('7,#',-'+,/4())$,%'-,.'#3"'1,/"%8'93(#'$)0'3"'I6*&"*'#3(#'$#'1()'/,.(77='1.,%&'#,'<"(#'64' 4.,)#$#6#")0'(%*'()'('.")67#'3(*'(%'6%+,%5"%#$,%(77='"&(7$#(.$(%'B<='4$/4N4.,)#$#6#"')#(%*(.*)C' ."7(#$,%)3$4'1$#3'('4.,)#$#6#"'%(/"*'O,$)8'P,1"5".0'K"5$%'."4,.#)'1$)3$%&'#3(#'3$)'/,.(7'E6(7/)'1,67*' >4())@0'(%*'&,")',%'#,'*")+.$<"'#3"'-,77,1$%&'$%+$*"%#:! ' !"#$%&'F"'*$*'3(5"'(%',++()$,%'13"."'O,$)'*$*'),/"#3$%&'#3(#'1()'#,#(77='6%(++"4#(<7"8'D%*'(#' #3(#'#$/"0')3"'3(*'7"-#'#3"'3,6)"8'D%*')3"'3(*'<""%'&,%"'-,.'/(=<"'('1""2',.'),/"#3$%&'7$2"' #3(#8'D%*'#3$)'1()'#,#(77='6%(++"4#(<7"'<"3(5$,.'$%'(%='."7(#$,%)3$40'7"#'(7,%"'('4$/4N3,' ."7(#$,%)3$4'Q' D%*';'."/"/<".'3".'1(7#R$%&'$%'#3"'3,6)"'7$2"'%,#3$%&'3(44"%"*8';#'1()'7$2"0'3"=0'3,1S)'$#' &,$%&T'D%*')3"'+,/")'$%0'(%*')3"S)'3"(*$%&'#,'#3"'2$#+3"%8'D%*';'."/"/<".'#3"'&6=)'1()'#3"."8' D%*'),';'-"7#'7$2"0'#,'),/"'"J#"%#0';'3(*'#,'*,'),/"#3$%&8';'3(*'#,'(+#8';'+,67*%S#'7"#'3".'I6)#' (..,&(%#7='1(72'$%'7$2"'#3(#'1$#3'#3"'&6=)')$##$%&'#3"."8'!,';'."/"/<".';'13""7"*'(.,6%*'(%*';' )7(44"*'3".8';')7(44"*'3".8'D%*';')7(44"*'3".'3(.*8';'/"(%0';')7(44"*'3".'."(7'3(.*'Q' ''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''
LM
';'(/'5".='&.(#"-67'#,'U(..='O(/'-,.'<.$%&$%&'#3$)'"J(/47"'#,'/='(##"%#$,%8'
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`=_ab;!
!"#$%&'()&'*++,&'!"#$%&'($()#*+&!"(&,(-$#$.&-$/&01(&23&'($()#*&4($5($*(1)'-#).)'/011$23430"%&'!42%$50$' 6$77"%'8%09$2103:)' !"#$%&'()&'*+++&';<$%$20=1&'>2$?@$%=:'4/9$2A1'4%/'B2"A4A0703:C)'6#$.7#15#*1&-$/&8"#9212:";&'DDE'DD*FDGH)' !"#$%&'()&'DII*&';J%'3#$'5$%$20='@1$'">'0%/$>0%03$'10%5@7421)C'<27)$-9&23&4(=-$5#*1&'*KLHME'*KHFDI+)' <2$$%A$25&'N)&'DIIH&',-$#3(15-5#2$1&23&.($()#*#5;)'O$P'N"2QE'R"@37$/5$)' S%"A$&'T)&'U'-2414/4&'V)&'DI**&';.@47'!#424=3$2'!"%=$B31C&'W%'8)2*((/#$.1&23&5"(&>>)/&?$$7-9&@2$3()($*(& 23&5"(&@2.$#5#A(&4*#($*(&42*#(5;B&X"13"%&'6(E'!"5%0309$'V=0$%=$'V"=0$3:)' S%"A$&'T)&'U'-2414/4&'V)&'1@AY033$/&';.@47'!#424=3$2'!"%=$B31'4%/'3#$'O"2Y4309$'.0Y$%10"%'">' !"%=$B3@47'R$B2$1$%3430"%C)'' Z4P7$2&'T)&'*+[H&';V3@/0$1'0%'\%5701#'5$%$20=1)C'0$#A()1#5;&23&,#*"#.-$&8-:()1&#$&6#$.7#15#*1'*L*M)' Z$170$&'V)'T)&'DII[&';<$%$20=1'4%/'3#$'V32@=3@2$'">'3#$'60%/C&'8"#9212:"#*-9&8()1:(*5#A(1&'D*L*ME'H[GF]IH)' Z$170$&'V)'T)&'DIIK&';<$%$20=1E'!"5%030"%'4%/'(=?@01030"%C&'8"#9212:"#*-9&C(A#(D&'**[L*ME'*F][)' Z$170$&'V)'T)&'>"23#="Y0%5'4&''($()#*1&-$/&'($()-9#E-5#2$)'J^>"2/E'J^>"2/'8%09$2103:'-2$11)' Z$170$&'V)'T)&'>"23#="Y0%5'A&';\11$%=$'4%/'O43@247'S0%/1E'_#$%'V=0$%=$'6$$31'-2$1=#""7$2'W%3@030"%C&' FG32)/&457/#(1&#$&H:#15(=292.;I&])' 6=!"%%$77F<0%$3&'V)&'DIID&';à@$$20%5`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b'S#$Y74%0&'V)b'Z$170$&'V)'T)b'<7@=Q1A$25&'V)&'1@AY033$/&';!"%=$B3@47'/0130%=30"%1'4Y"%513' 5$%$20=1C)' R#"/$1&'6)&'Z$170$&'V)T)&'4%/'cP"2$Q&'!)&'0%'B2$B42430"%&';<$%$20='Z4%5@45$'4%/'3#$'.$9$7"BY$%3'">' V"=047'\11$%304701YC)' '
The Original Sin of Cognition: Fear, Prejudice and Generalization * Sarah-Jane Leslie Penultimate Draft: Please refer to final version for citation purposes
Introduction Long before we learn to talk, our expectations concerning novel members of a category are shaped by our experience with already encountered members. We expect, for example, that objects that share obvious perceptible qualities will also share dispositional properties. If a given item rattles when shaken, nine-month olds expect that other items which share the same perceptible profile will rattle when shaken. 1 By our first birthday, these inductive inferences are guided by language; we expect that even superficially dissimilar objects will share their hidden properties if they are identified by the same common noun; if, for example, each is introduced as a blickett. 2 From the very beginning, we are inclined to generalize from experience with a given item to other items that we perceive as belonging to a common category. There is, presumably, some innate cognitive mechanism that is responsible for these early inductive generalizations. In earlier papers, I argue that generics sentences such as ravens are black and tigers are striped express the generalizations that are delivered by
I would like to thank the following people for their invaluable help, comments, and conversation: K. Anthony Appiah, Andrei Cimpian, Kai-Yuan Cheng, Susan Gelman, Tamar Szabo Gendler, Sam Glucksberg, Susan Fiske, Elizabeth Harman, Sally Haslanger, Richard Holton, Mark Johnston, Thomas Kelly, Sangeet Khemlani, Rae Langton, Eden Lin, Ron Mallon, Sally McConnell-Ginet, Sarah McGrath, Alexander Nehamas, Joyce Carol Oates, Rachel Parsons, Debbie Prentice, Mark Richard, Gideon Rosen, Sandy Waxman, Helen Yetter, the members of my fall and spring 2009 Philosophical Issues in Language and Cognition class, the audience at the New York Society for Women in Philosophy, and anonymous Journal of Philosophy reviewers. 1 D. A. Baldwin, E. Markman, and R. Melartin, Infants Ability to Draw Inferences about Nonobvious Object Properties: Evidence from Exploratory Play, Child Development, 64 (1993): 711-728. 2 S. A. Graham, C. S. Kilbreath, and A. N. Welder, Words and Shape Similarity Guide 13-month-olds Inferences about Nonobvious Object Properties, Proceedings of the Twenty Third Annual Conference of the Cognitive Science Society, (2001): 352-357.
this basic mechanism of generalization. 3 If this is so, then generics provide us with a window onto the workings of this mechanism. In this paper, I am concerned with a particular aspect of this mechanism, namely the route by which we reach general conclusions regarding dangerous or harmful features. What follows in no way purports to be an exhaustive analysis of all the factors that have formed and sustained prejudiced attitudes. Rather, my aim is to identify and discuss one particular cognitive bias that has given birth to many a prejudice. While I adopt a cognitive perspective here, this is not to imply that economic, political and cultural perspectives are not of equal (or even greater) value and importance. These various perspectives are not in competition with each other; rather they complement each other by providing different levels of explanation. Moreover, even within the domain of cognitive explanations of prejudice, I do not purport to offer anything close to a full psychological account of prejudiced attitudes. The focus is on a particular subset of negative stereotypes: ones that involve generalizing extreme and horrific behavior from a few individuals to a group, for example Muslims are terrorists or Blacks are rapists. Two things, though obvious, are worth noting at the beginning. First, cognitive bias explanations do not excuse racial or cultural prejudice, any more than noting than we are hardwired to seek out and accumulate resources serves as an excuse for extreme covetousness or for theft. Secondly, offering a psychological explanation for prejudice does not entail that prejudice is inevitable. Quite the contrary the closing sections of this paper will discuss some ways in which we might combat prejudice, with a particular focus on how we might prevent the formation of these attitudes in the course of childhood development. These suggestions are based on recent psychological research, and so far from implying
S. J. Leslie, Generics: Cognition and Acquisition, Philosophical Review, 117, 1, (2008): 1-49; S. J. Leslie, Generics and the Structure of the Mind, Philosophical Perspectives, (2007): 375-405.
that prejudice is an inevitable feature of human psychology the cognitive perspective on prejudice may point to some novel means of combating it. 4 Striking Property Generalizations In previous work, I argue that a variety of philosophical, linguistic and psychological considerations suggest that generic sentences may be languages way of letting us give voice to cognitively primitive generalizations. 5 This hypothesis has subsequently received further support by new psychological data. 6 There are now a variety of convergent reasons for supposing that the generalizations we articulate using generics reflect deep-seated aspects of our psychology. In theorizing about generic generalizations it is helpful to identify various sub-classes of these generalizations, one of which I term striking property generalizations. This class includes claims such as: Mosquitoes carry the West Nile virus Sharks attack bathers Deer ticks carry Lime Disease Pit-bulls maul children Tigers eat people These claims are intuitively true, even though very few members of the kind in question possess the predicated property. As it happens, less than one percent of mosquitoes carry the West Nile virus, and yet we are quick to assent to mosquitoes carry the West Nile virus,
For an exemplary defense of the utility of psychological accounts of prejudice against various criticisms and misunderstandings, see E. Machery, L. Faucher, L., and D. Kelly, On the Alleged Inadequacy of Psychological Explanations of Racism, forthcoming in The Monist. 5 Leslie, 2008, op. cit.; Leslie, 2007, op. cit. 6 See, e.g., S. J. Leslie, S. Khemlani & S. Glucksberg, All Ducks Lay Eggs: The Generic Overgeneralization Effect, under review; S. J. Leslie & S. A. Gelman, Quantified Statements are Recalled as Generics: Evidence from Preschool Children and Adults, under review; A. Brandone, A. Cimpian, S. J. Leslie, and S. A. Gelman, Do Lions have Manes? Children Interpret Generics as Making Claims about Kinds rather than Quantified Sets, under review.
even after learning this statistical fact. (Conversely, mosquitoes dont carry the West Nile virus remains patently false, even though 99% of mosquitoes dont carry the virus.) It may appear that these generics require for their truth only that some of the kind possess the property in question. 7 This is not true for generics in general; for example, some cats are female, but cats are female is false, and some (in fact, most) mosquitoes dont carry the West Nile virus, but the corresponding generic is false. Such examples abound. I suggest that the generics above are special in that their predicates express properties that we have a strong interest in avoiding. 8 If even just a few members of a kind possess a property that is harmful or dangerous, then a generic that attributes that property to the kind is likely to be judged true. Since we are working under the hypothesis that generics give voice to psychologically primitive generalizations, this observation implies that our basic way of dealing with dangerous or harmful information involves the rapid generalization of this information to the appropriate kind or category. We do not wait around to see what percentage of tigers eat people before drawing a general conclusion even a single instance may be enough for us to conclude that tigers eat people. It is not hard to see the evolutionary benefits of such a disposition, since the costs of under-generalizing such information are
One might be tempted to think that these generics are true because, e.g., only mosquitoes carry the West Nile virus. There is certainly a reading of mosquitoes carry the West Nile virus to that effect; to see this, try stressing mosquitoes, as in MOSQUITOES carry the West Nile virus, or else paraphrasing the sentence as it is MOSQUITOES that carry the West Nile virus. Consider, however, whether ones intuitions would change upon learning that deer ticks also carry the virus. This would falsify one reading of the sentence; an assertion of it is MOSQUITOES that carry the West Nile virus can be countered by the observation that deer ticks do likewise. However, there is still is a salient reading of the sentence upon which it remains true. There is nothing contradictory, or even infelicitous, about the remark that mosquitoes carry the West Nile virus, and deer ticks do too. Thus these generics do not depend on the propertys being uniquely possessed by the subject. 8 For some empirical confirmation that generics are more likely to be accepted at low prevalence levels if the property in question is dangerous, see A. Cimpian, A. Brandone, and S. Gelman, Generic statements require little evidence for acceptance but have powerful implications, forthcoming in Cognitive Science.
potentially huge. Our ancestors were far better off jumping to conclusions, as it were, rather than taking the time to judiciously determine the precise likelihood of their being eaten. The tendency to rapidly generalize such striking information manifests itself elsewhere in our thinking. Consider, for example, how many murders one must commit to be a murderer, versus how many times one must worry to be a worrier. The latter case requires one to worry with considerable regularity, whereas a single murder suffices to make one a murderer. The disposition to generalize strikingly negative information on the basis of even a single event thus appears to be a pervasive aspect of our thinking. For generalizations concerning neutral or positive information, we require the instances or events to occur with a significant regularity; this is not so with negative information. There is a fundamental asymmetry between the impact of very negative information and the impact of neutral or more positive information on our intuitive generalizations. 9 The introduction conditions, as it were, of striking property generalizations how the world must be for us to form or accept these generalizations are very undemanding when it comes to how prevalent the property has to be in the relevant population. What, though, of the elimination conditions of these generalizations how does acceptance or rejection of these generalizations impact the inferences we are willing to draw? 10 We are content to accept ticks carry Lyme disease despite knowing that very few ticks actually carry the relevant bacterium. One might suppose that an ideally rational agent would be very
9
The same arguably applies to strikingly positive information. One extremely large charitable donation presumably suffices to make one a philanthropist, though if it is a one time occurrence, the donation must be very large indeed. (A single gift of a moderate sum does not a philanthropist make, unfortunately.) Such cases are less clean-cut, however, and examples are far less readily available. 10 I must emphasize that I mean elimination conditions here to be read as wholly psychological, and not at all normative (hence the scare quotes). That is, I mean to highlight the inference that we actually draw from these generalizations, not the inferences that we ought to draw from them. For more discussion, see S. J. Leslie, Generics, (Oxford: Oxford University Press, forthcoming).
hesitant to suppose that an arbitrary tick carries Lyme disease, in light of these statistical facts. However, recent psychological results suggest that acceptance of a generic strongly influences our judgments concerning whether an arbitrary member of a kind has a property over and above our beliefs about the prevalence of the property. This is especially so for striking property generalizations. Sangeet Khemlani, Sam Glucksberg and I found that people were as likely to agree that Jumpy the tick carries Lyme disease as they were to agree that Joe the Canadian is right-handed despite the very large discrepancy between the subjects own (roughly correct) judgments of the prevalence of the respective properties in the respective populations. Sixty-five percent of our participants who accepted the striking property generalizations judged with varying degrees of confidence that an arbitrary member of the kind would have the striking property. 11 Andrei Cimpian, Amanda Brandone, and Susan Gelman found comparable results using a very different experimental design. While their participants frequently accepted novel striking property generics at low prevalence levels, if they were presented with a novel striking property generic and asked to
We showed participants the following information: Suppose you are told: Jumpy is a tick. What do you think of the following statement: Jumpy carries Lyme disease. Participants responded by marking a 7-point scale that ranged from strongly agree to I couldnt possibly tell to strongly disagree. Participants were never shown a generic in this portion of the experiment, so they had to rely solely on background beliefs. A variety of examples were used, with varying degrees of prevalence and predictiveness (cue validity). Acceptance of the generic was found to predict responses above and beyond estimates of prevalence and predictiveness. As an intuitive illustration of this, average prevalence estimates for striking property generalizations were 33% (already inflated relative to actual fact), and average prevalence estimates for items like Canadians are right-handed were 60%, yet the mean responses were the same for both types of items despite the gross differences in average estimated prevalence (and overall similar ratings of cue validity). For more details, see S. Khemlani, S. J. Leslie & S. Glucksberg, Generics, Prevalence, and Default, Proceedings of the 31st Annual Cognitive Science Society (2009).
11
estimate how prevalent the property might be among the kind, they gave extremely high estimates in many cases, 100%. 12 These findings suggest that these generalizations play a powerful role in guiding our inferences concerning property possession, despite their relatively weak acceptance conditions. These primitive generalizations are not psychologically inert rather they play a powerful role in guiding our judgments about members of a kind. In an early paper on generics, Robert Abelson and David Kanouse noted that some generics require very little evidence for acceptance, and yet once accepted psychologically they appear to be commonly taken in a rather strong sense, as though the quantifier always had implicitly crept into their interpretation. 13 Our most primitive method of generalization seems to encourage us in reasoning from some to many or most, or even to all, at least when striking properties are in play. Fear, Prejudice, and Generalization The cognitive disposition to generalize strikingly negative information very widely may serve a useful purpose in the non-social realm. When we turn to generalizations about groups of people, however, it can lead to disastrous consequences. I should reiterate that what follows is in no way intended as an exhaustive account of the cognitive factors underlying racism, and it does not touch on the social and cultural factors that have bred and sustained
Cimpian et al., op. cit. Participants who were told that, e.g., 10%, 30%, or 50% of lorches have feathers that can cause massive bleeding frequently accepted the generic lorches have dangerous feathers. However, participants who were instead told lorches have dangerous feathers, and then asked to estimate the prevalence, gave much higher estimates often as high as 100%. Interestingly, this asymmetry was only found if the property was striking and/or characteristic of the kind. For more accidental properties such as having muddy feathers, no such asymmetry was found. See Leslie, 2008, op. cit. for a discussion of the different types of generics striking, characteristic, and more the accidental majority generics. 13 Interestingly, their clearest examples are of striking property generics, though they do not identify them as such. (They do, however, note that it is the predicate that determines how likely the generic is to be accepted on the basis of weak statistical evidence.) R. P. Abelson & D. E. Kanouse, Subjective acceptance of verbal generalizations. In S. Feldman (Ed.), Cognitive consistency: Motivational antecedents and behavioral consequents (New York: Academic Press, 1966: 171-197), here p. 172; see also D. Schneider, The Psychology of Stereotyping (New York: Guilford Press, 2004).
12
racism. Rather, I seek only to identify a particular and pervasive aspect of our thinking which itself leads us down the dark road to prejudice. Our most primitive method of generalization has the potential to enshrine a pervasive form of bigoted thinking. 14 The basic idea is quite simple: just as it takes but a few instances of sharks attacking bathers, or of mosquitoes carrying the West Nile virus, for us to make the corresponding category-wide generalization, so also a strikingly negative action on behalf of a few members of a racial, ethnic, or religious minority (or out-group) may lead us to a general belief concerning their entire group. I discuss below why it may be that these particular social groups, as opposed to others, tend to be the targets of such generalizations, and I then identify the fundamental error behind these generalizations in the case of social groups. (That is, I explain how it may be true that mosquitoes carry the West Nile virus, and yet decidedly false that, say, Muslims are terrorists.) Once we identify the enabling error behind such generalizations over social groups, the opportunity arises to consider novel ways of combating this way of thinking. As a result of a profound and pervasive cognitive bias built into our most primitive method of generalization, a few appalling acts on behalf of some members of a given group can lead us to draw conclusions about the group in general. As the available experimental evidence suggests, acceptance of a striking property generalization can lead one to draw the
14
It should be clear that the thesis set out here need not conflict with other theories of the psychological roots of racism. I rather take my thesis to be quite compatible with them, allowing that each theory may delineate a different aspect of this complex phenomenon. For example, the sophisticated analysis of group identification provided by Social Identity Theory is extremely important to our understanding of group dynamics, and is not at all challenged by my identification of this particular cognitive bias of ours (e.g. H. Tajfel, Social identity and intergroup behavior (Cambridge, England: Cambridge University Press, 1982); H. Tajfel & J. C. Turner, The Social Identity Theory of Intergroup Behavior, in S. Worchel and W. Austin (eds.) Psychology of Intergroup Relations (Chicago: Nelson Hall, 1986)). Similarly, Realistic Conflict Theorys insights into the role of competition over limited resources fill in an important part of the picture that my account does not touch (e.g. M. Sherif, In common predicament: Social psychology of intergroup conflict and cooperation (Boston: HoughtonMifflin, 1966)). It would be quite surprising if prejudice turned out to have a single and uniform psychological basis, rather than being the result of many disparate factors. To seek the psychological explanation for prejudice is likely a mistaken quest.
corresponding conclusion about the arbitrary member of the group conclusions which go beyond even the perceived statistical facts. Extreme and aberrant actions on behalf of the few can thus lead to conclusions concerning the group at large, and these conclusions will influence our judgments concerning a newly encountered member of the group. A rather pristine version of this pattern of reasoning is presented in Laura Ingalls Wilders Little House on the Prairie. The Ingalls rather prejudiced neighbor, Mrs. Scott, claims that The only good Indian is a dead Indian certainly a very sweeping and inclusive claim. She immediately justifies the claim by citing the Minnesota Massacre, going as far as to say To anyone who disagrees, I say, remember the Minnesota Massacre! Mrs. Scott reasons from the single horrific incident of the Minnesota Massacre to the conclusion that there are no good (living) Indians. Considering that the Scotts and the Ingalls were living in Kansas at the time, it is unlikely that she believed any of the Indians they actually encountered had any personal involvement in the Minnesota Massacre. Nonetheless, she took the single incident to justify the claim that the only good Indians were dead Indians. She was also apparently confident that only someone who had forgotten about the massacre would disagree with her on the point. 15
It is important to distinguish my thesis from the thesis that racist generalizations are the result of illusory correlations (see L. J. Chapman, Illusory Correlation in Observational Report, Journal of Verbal Learning and Verbal Behavior, 6, (1967): 151-155; D. L. Hamilton & R. K. Gifford, Illusory Correlation in Interpersonal Perception: A Cognitive Basis of Stereotypic Judgment, Journal of Experimental Social Psychology, 12 (1976): 392-407; B. Mullen & C. Johnson, Distinctiveness Based Illusory Correlations and Stereotyping: A Meta-Analytic Integration. British Journal of Social Psychology, 29 (1990): 11-28.) Illusory correlations occur when people overestimate the degree to which two rare events co-occur. If event A is rare relative to event B, and event C is rare relative to event D, then people may guess that A and C co-occur far more often than B and D, even though the actual rates of co-occurrence are the same. Hamilton and Gifford propose that racist generalizations like the ones discussed here may be explained by illusory correlations it is more rare to encounter a member of a minority than a member of the majority, and strikingly awful events are, thankfully, more rare than neutral ones. Perhaps we are susceptible to illusory correlations in such cases, and so overestimate how often minority members commit horrific acts. Illusory correlations are a real phenomenon, however I believe that they are insufficient to explain the nature of the generalizations discussed here. Its important to note that, while two rare events may be judged to co-occur more often than two more common events, the magnitude of the difference is not huge. It is certainly
15
Mrs. Scotts reasoning is a perfect illustration of how human beings can move from a horrific particular to a sweepingly prejudiced generalization. If reasoning of this sort really is a pervasive cognitive disposition then we should find many examples of it, in whatever historical period we happen to examine. We should not be surprised if this mechanism of generalization has hovered perpetually in the background wherever human beings were formulating prejudiced attitudes towards social groups. A detailed historical analysis is beyond the scope of this paper, but let us briefly consider a recent and vivid example. Nothing has done more to harm the plight of Muslims in America than 9/11. In the aftermath of 9/11, hate crimes against Muslims rose more than 1,600%, according to FBI statistics. 16 Hate crimes are, by definition, crimes motivated by the mere fact that the victim is a member of a particular group; the hate crimes following 9/11 were motivated by the fact that the victims in question were Muslims. Many of these crimes were committed against Muslim women and children; the perpetrators surely were not under the impression that their victims were themselves involved in or personally responsible for the 9/11 bombings. It was sufficient that the victims were Muslims. We might characterize the reasoning of the hate
not the case that the two rare events are taken to co-occur all the time, or even most of the time. Illusory correlations would predict that the Ingalls neighbor Mrs. Smith might overestimate the number of Indians that have been involved in massacres, but would not predict the sweeping universality of her claim the only good Indian is a dead Indian. Further, while the experimental support for illusory correlations is not at all limited to striking or negative events illusory correlations can be observed between two perfectly neutral or positive events (see, e.g., Hamilton and Giffords experiment 2) theorists tend to only employ them in explaining negative stereotypes, often ones involving strikingly negative behaviors of the sort here discussed. Hamilton and Gifford argue that since for most varieties of behavior the norm is positive in value, undesirable (non-normative) behavior is statistically less frequent than desirable behavior and [so] can also be considered distinctive (op. cit. p. 394) If we take illusory correlations seriously as an explanation of stereotype formation, though, we must predict that stereotypes will form whenever a minority member engages in a rare activity, since it is only the perceived frequency, not the polarity or extremity that is relevant. For example, I suppose that bouncing around on a pogo stick is a very rare activity, and certainly one hears about far fewer instances of pogo-stick jumping than acts of terrorism. We would thus have to predict that we would be at least as inclined to conclude that Muslims bounce around on pogo sticks upon hearing of a few Muslims engaging in this activity as we were to conclude that Muslims are terrorists in the wake of 9/11. 16 Reported in the San Francisco Chronicle, http://sfgate.com/cgibin/article.cgi?f=/c/a/2002/11/26/MN224441.DTL
crime perpetrators as moving from the horrific events of 9/11 events which involved a rather small number of extreme individuals to the conclusion that the arbitrary Muslim deserved to be victimized in virtue of being Muslim. The conclusions drawn from the 9/11 attacks did not concern just the bombers and their supporters, but concerned Muslims in general. Such generalizations were made even by members of Congress. Shortly after 9/11, Representative John Cooksey told a Louisiana radio station, If I see someone [who] comes in that's got a diaper on his head and a fan belt wrapped around the diaper on his head, that guy needs to be pulled over. In Georgia, Representative, now Senator, C. Saxby Chambliss told law enforcement official to just turn [the sheriff] loose and have him arrest every Muslim that crosses the state line. 17 These statements again reflect conclusions pertaining to Muslims quite generally. They do not reflect the more moderate conclusion that only some Muslims had any involvement in 9/11 whatsoever. Obviously the aftermath of 9/11 is not an isolated historical example. Consider, for example, the origins of anti-Algerian prejudice in France. The relationship between the French and the Algerians is a complicated one, as would be expected given their history of war, colonization and occupation, and indeed anti-Algerian racism is still virulent in France today. The cognitive bias under investigation here is, of course, too simple to account for all the subtleties of a racism with such a complex history. However, if we trace anti-Algerian prejudice to its early days, we find again the dead hand of striking property generic reasoning.
17
Reported by Human Rights Watch, http://www.hrw.org/reports/2002/usahate/usa1102-05.htm
Early relations between Algerian workers and Frances indigenous population were quite amicable. 18 In 1923, however, French-Algerian relations took a sharp turn for the worse. A wave of anti-Algerian violence began, in which North Africans were attacked at random. The attacks included a public lynching in the rue Fremicourt in Paris, and it was unsafe for North Africans to venture into the surrounding area. The media denigrated North Africans, and petitions were circulated that called for the undesirables to be driven from the area. 19 According to Neil MacMaster, this rapid swell in hostility towards North Africans can be traced to a single catalytic incident. On November 7 1923, Khemile Ousliman, an unemployed North African man, knifed a woman in the Rue Fondary. Ousliman, who was likely mentally ill, had been obsessed with the woman, and had repeatedly made sexual advances towards her. When she refused, he slit her throat, then turned in a frenzy on some passers by, killing another woman and wounding two others. Immediately following this incident, there began a surge of anti-Arab violence, hatred, and discrimination throughout France. Seven years later, Paul Catrice, a Catholic priest and immigration expert, remarked that If the Sidi, in general, inspires a certain repulsive fear, it is because of the memories of certain sensational crimes from which Parisians have drawn unconsidered generalizations. 20 These unconsidered generalizations are exactly those considered in this paper. There is no a priori reason to think that human beings would be disposed to reason from a single sensational event to a category-wide generalization; certainly there is no logical demand for such thinking. We are, however, possessed of a
18
Beaugency, 1914, as reported in N. MacMaster, "The Rue Fondary Murders of 1923 and the Origins of AntiArab Racism," in Jan Windebank and Renate Gunther (eds.), Violence and Conflict in the Politics and Society of Modern France (Lampeter, UK: Edwin Mellen Press, 1995:149-160). 19 Ibid., p. 150. 20 Ibid., p. 158.
particular cognitive bias a style of generalization that makes such reasoning not only possible, but pervasive. More speculatively, since the veil of years here is thicker, the early origins of AngloAmerican prejudice towards Africans and Native Americans may have been fueled by sensational reports of horrific acts relayed in travel logs, which were extremely popular among the newly literate population of Britain. Very few people could afford to travel abroad themselves, so the reports of a small number of explorers were the source of public knowledge of foreign lands and their inhabitants. The initial impressions of the English population vis--vis Africans and Native Americans derived almost wholly from these travel logs. 21 Winthrop Jordan, in his discussion of travel logs on Africa, writes To judge from the comments of voyagers, Englishmen had an unquenchable thirst for the details of savage life It is scarcely surprising that civilized Englishmen should have taken an interest in reports about cosmetic mutilation, polygamy, infanticide, ritual murder and the like of course English men did not really do any of these things themselves It would be a mistake to slight the importance of the Negros savagery, since it fascinated Englishmen from the very first. English observers in West Africa were sometimes so profoundly impressed by the Negros deviant behavior that they resorted to a powerful metaphor with which to express their own sense of difference from him. They knew perfectly well that Negroes were men, yet they frequently described the Africans as brutish or bestial or beastly. The hideous tortures, the cannibalism, the rapacious warfare, the revolting diet (and so forth page after page) seemed somehow to place the Negro among the beasts. 22 These travel logs, which did so much to shape Englands early image of Africa, contained endless gory accounts of shocking behavior (allegedly) exhibited by the Africans. It should be noted that the travelers themselves often reported only specific incidents of cannibalism, or other specific instances of horrific violence. That is, it would be overly simplistic to place the blame for the formation of early negative stereotypes squarely on the
R. Cole, Sixteenth Century Travel Books as a Source of European Attitudes toward Non-White and NonWestern Culture, Proceedings of the American Philosophical Society, 116, 1 (1972): 59-67. 22 W. Jordan, White over Black: American Attitudes toward the Negro, 15501812 (Chapel Hill: University of North Carolina Press, 1968), pp. 25-28.
21
explorers. Many of them were quite responsible in their reporting, and did not indulge themselves in broad generalizations. Given the nature of our default system of generalization, they did not have to. The reporting of specific instances would suffice to encourage very general beliefs in the mind of the reader. Generalizations, Dispositions and Predictors If the foregoing is correct, then the same pattern of generalization is in play for both claims like mosquitoes carry the West Nile virus and claims like Muslims are terrorists. Surely though, there must be some dissimilarities between them: in particular, is it not the case that the former claim is true, while the latter is false? Even if the same unreflective mechanism is responsible for both judgments, it is surely making an error in judging that Muslims are terrorists an error which is not (necessarily) involved in judging that mosquitoes carry the West Nile virus. Let us then consider striking property generics in more detail. Their truth conditions are not quite as straightforward as the earlier discussion suggests. We have been speaking as if a generic Ks are F is true iff some Ks are F, given that being F is a dangerous or harmful property. But this would suggest that insects carry the West Nile virus, or even animals carry the West Nile virus would also be true certainly there are some insects, and therefore some animals that carry the virus, namely those few unfortunate mosquitoes. Similarly, the truth of tigers eat people would entail the truth of mammals eat people, and from the truth of sharks attack bathers we should conclude that fish attack bathers. People do not tend to find these inferences acceptable, so the truth conditions of these generics must involve some further complexity.
In earlier work, I suggested that the mechanism of generalization in question seeks a good predictor of the property in question. 23 It is easy enough to see an evolutionary rationale behind generalizing striking properties only so far up the taxonomic hierarchy. If our ancestors had undertaken to avoid all mammals after seeing a tiger eating one of their companions, the costs of doing so may well have outweighed the benefits. (One could waste a lot of time running from small harmless creatures.) Someone who avoided all animals, big or small, after witnessing a lion maul his companion would be at a significant disadvantage relative to a more sophisticated competitor who limited his conclusions to lions alone. An efficient generalizing mechanism, we might suppose, should seek a good predictor of the striking property a kind that is inclusive enough to aid us in avoiding the property, but not so inclusive as to needlessly hamper our activities. I further suggest that what makes a kind a good predictor of a striking property is that the members of the kind that do not possess the property are typically disposed to possess it. 24 It matters, then, for the truth of mosquitoes carry the West Nile Virus that the virusfree mosquitoes will carry the virus if circumstances allow. Sharks attack bathers is true only if the sharks that never in fact cause harm to humans would typically do so given half a chance, and so on. Statements such as animals carry the West Nile Virus and sea creatures attack bathers are false, because the members of the kinds in question do not share the relevant dispositions. A generic statement in which a striking property is predicated is, I claim, true if and only if some members of the kind in question possess the relevant property,
23 24
Leslie, 2008, op. cit. Ibid. The metaphysics of dispositions is an intriguing subject, and I shall not delve into it here, but rather rely on our intuitive understanding of the notion. For two recent discussions of dispositions, see M. Fara, "Dispositions and habituals", Nous, 39 (2005): 43-82; D. Lewis, "Finkish dispositions", Philosophical Quarterly, 47 (1997): 143-58.
and the others are typically disposed to possess it. 25 To determine which striking property generics are strictly true and strictly false, then, would require some rather detailed knowledge of dispositions and capacities. Dispositions are not directly observable in the way their manifestations are, and so we do not normally possess such knowledge. We thus often operate under uncertainty when it comes to attributing dispositions, and so must adopt certain heuristics to guide our judgments. To probe this strategy further, let us set aside questions of whether sentences are true or false, and consider how our basic mechanism of generalization must work, if my theses here are correct. Dispositions, Essences, and Basic-Level Kinds If the preceding remarks are correct, then our basic mechanism for generalization, when confronted with the manifestation of a striking property, seeks to generalize that property to a kind whose members are disposed to manifest it. Detailed scientific knowledge of dispositions is not likely to be available for most of these generalizations. However, even in the absence of scientific knowledge, we nonetheless often form (tacit) beliefs about the shared natures and dispositions of members of certain kinds. In the psychological literature, these kinds are said to be essentialized. 26 We essentialize a kind if we form the (tacit) belief
If this is correct, then we must allow for the possibility that some of the striking property generics listed above are, in fact, strictly false. Perhaps it is only Great White Sharks that are disposed to attack bathers (as it is sometimes claimed), or perhaps only mosquitoes with a particular mutation are capable of carrying the virus. If these turn out to be the facts, then my account predicts that the above generics are in fact false, and it is only the weaker claims Great White Sharks attack bathers and mosquitoes with a particular mutation carry the West Nile Virus that are true. This seems to be the intuitively correct conclusion here: under such circumstances, the more inclusive generic claims are, strictly speaking, false. 26 D. L. Medin & A. Ortony, Psychological essentialism, in S. Vosniadou & A. Ortony (eds.) Similarity and Analogical Reasoning (New York: Cambridge University Press, 1989: 179196); S. A. Gelman, The Essential Child (New York: Oxford University Press, 2003).
25
that there is some hidden, non-obvious, and persistent property or underlying nature shared by members of that kind, which causally grounds their common properties and dispositions. 27 For example, one might believe, implicitly or explicitly, that there is something about tigers that causes them to have stripes, to growl, to hunt their prey, and so on. These are not accidental features of tigers; they are grounded in the very nature of tigerhood. What is more, we believe that even a crippled albino tiger possesses this intrinsic, essential nature, even if she does not manifest its outward effects. The essence of tigers causally grounds these dispositions, though does not guarantee their manifestation, since adventitious factors may intervene. 28 Different levels of the subjective taxonomic hierarchy for biological kinds are essentialized to differing degrees, in the sense that some levels are seen as possessing highly distinctive essences which ground a large number of shared features, while others are not. For example, simply being a mammal is not predictive of wide range of properties, since animals as diverse as tigers, whales, mice, and humans all count as mammals. Conversely, while Bengal tigers share many properties in common with each other, this shared nature is not particularly distinctive since many of those properties are also shared by Siberian tigers. There tends to be a privileged level of the subjective taxonomic hierarchy at which the
There is a variety of evidence to suggest that we view kinds as essentialized from a young age. For example, pre-school aged children expect that members of the same basic-level kind will have the same internal organs, even if some of the members look quite different from the others (see Gelman, op. cit. and references therein). They also maintain that a raccoon dressed up as a skunk is nonetheless a raccoon, and possessed of raccoon innards, thus demonstrating a belief that there is more to kind-membership than outward appearance. Children also have strong views about the power of nature over nurture when it comes to cross-species comparisons; they expect that a cow raised from birth by pigs will look like a cow, say moo, and so on. Gelman and her colleagues argue at length that these convictions reflect a belief in the essences of these kinds. From a very young age, we think that there is something intrinsic to the nature of cows that will cause it to resemble other cows, regardless of how it is raised. 28 The relevant notion of essence at work in the psychological literature is obviously not the philosophers stricter notion of that intrinsic aspect of a thing which grounds all and only the intrinsic metaphysical necessities that hold of the thing. It should also be noted that the claim being made here is that our folk theories treat biological kinds in this way, not that this is the correct metaphysics of these kinds. The claim is merely psychological.
27
essence of the kind is taken to ground a wide range of properties that are shared by its members but not collectively shared by members of another kind. This privileged level of the taxonomy is known as the basic-level. The notion of a basic-level kind is due to Eleanor Rosch and her colleagues, who found that various measures of psychological salience converged on a particular taxonomy that is psychologically privileged. 29 The most interesting feature of basic-level kinds from our point of view is that, from a young age, we see them as having rich inductive potential, thanks to their members sharing highly predictive natures. These highly essentialized basic-level kinds appear to be the default starting points for our common inductive generalizations. As members of basic-level kinds are viewed as sharing a nature, we are inclined to treat these kinds as supportive of inductive generalizations and inferences about non-obvious properties. Further, since this nature is taken to be distinctive that is, not shared by other comparable kinds we are reluctant to generalize properties to the more inclusive kinds above the basic-level on the taxonomic hierarchy. The psychologically privileged status of basic-level kinds is the result
E. H. Rosch, Principles of categorization, in E. Rosch & B. Lloyd (eds.), Cognition and Categorization (Hillsdale, N.J.: Erlbaum Associates, 1978:27-48). For example, when asked to identify what is in a picture of the animal that is Princetons mascot, people tend to identify it as a tiger, rather than as a Bengal tiger, or as a mammal, or a vertebrate, and so on. Alternatively, if people are asked to list features that they associate with various kinds, basic-level kinds have the greatest number of features that are both widely shared by members of the kind and not shared by members of comparable kinds. For kinds that are taxonomically below the basic level (so-called subordinate kinds), the features listed tend to be ones that are listed for other subordinate kinds; Bengal tigers share most of their psychologically salient features with other types of tigers. If asked to list features for superordinate kinds such as mammal, people have much greater difficulty coming up with features, and often list features that are not widely shared by members of the kind. In the course of language acquisition, names for basic-level kinds are learned first, and there is a high degree of cross-cultural agreement about basic-level taxonomy, even though cultures may differ significantly on the taxonomy of superordinate and subordinate kinds. The notion of a basic-level kind is an explanatory psychological notion, not an explanatory biological notion. In many cases, the basic-level kind corresponds to genus or a species considered as biological taxa, but this is not always the case. For example, while the basic-level kind tiger corresponds to a species of the genus Panthera, the basic-level kind jellyfish corresponds to the class Scyphozoa, which has many orders, families, genera, and species below it.
29
of a useful trade-off between the extent to which natures or essences are taken to be shared and the extent to which they are taken to be distinctive. I suggest that, in making a striking property generalization, the default is to choose a relevant highly essentialized basic-level kind as the target of the generalization. Since such kinds are the primary targets of our inductive generalizations and inferences, it is perhaps not surprising that they are also the targets of these particular generalizations. Basic-level kinds are fine-grained enough to reduce wasted effort, but still coarse-grained enough to allow one to err on the side of caution. And since they are the most readily recognizable categories, they make for a very practical starting point for striking property generalizations. Most importantly, basic-level kinds because they are highly essentialized are supportive of inferences concerning the dispositions of its members. Upon seeing a tiger eat a companion of ours, we conclude that there is something about tigers that disposes them to eat us it lies in their nature to eat us, given half a chance. A typical tiger is thereby, i.e. thanks to his underlying intrinsic nature, disposed to eat us. 30 I thus propose that when an instance of a particular basic-level kind manifests a striking property, by default we take the manifestation of that property to be grounded in some nature common to the members of that basic-level kind. Unless we learn otherwise, we therefore take the disposition to manifest the property to belong to the typical members of the kind, that is, those that share the common nature.
If one is unconvinced by this example, consider the following. A Savannah cat is a hybrid bred from the domestic cat and an African hunting cat known as the serval. The adult Savannah is quite magnificent, weighing in at 40lbs, and possessed of a long neck and beautiful leopard-esque spots. There have been no documented cases (to my knowledge) of Savannahs attacking either people or other pets, though the breed has only been around for a few years, so its general tendencies are not yet well understood. Suppose, however, that tomorrow we see splashed across the New York Times a report of a Savannah savagely attacking a toddler. Would one suspend judgment as to the typical Savannahs violent tendencies, or would one immediately view the entire breed as dangerous?
30
It must be noted that non-striking properties are not treated in this way. Upon learning that a given tiger is female, we do not conclude that typical tigers, by nature, are disposed to be female a disposition that simply fails to manifest itself in the case of male tigers. If we saw a hamster with an odd growth on its back, it would not occur to us to decide that the disposition to grow such lumps is grounded in hamsterhood, though rarely manifested. The rapid generalizations to typical underlying dispositions are specific to striking properties. Social Kinds and Essence In recent years, a number of social psychologists have argued that we view some social kinds as essentialized in much the way we view animal kinds. This line of thinking was popularized by Rothbart and Taylor, 31 and has since received a significant amount of empirical support. 32 Within psychology, the basic observation originated with Gordon Allport in 1954, who wrote that: a belief in essence develops. There is [a belief in] an inherent Jewishness in every Jew. The soul of the Oriental, Negro blood, the passionate Latin all
31
M. Rothbart & M. Taylor, Category labels and social reality: Do we view social categories as natural kinds? in G. Semin and K. Fiedler (eds.), Language, Interaction and Social Cognition (London: Sage, 1992:11-36). 32 E.g.: L. A. Hirschfeld, Race in the Making (Cambridge, MA: MIT Press, 1996); F. J. Gil-White, Are ethnic groups biological species to the human brain? Current Anthropology, 42, (2001): 515-554; N. Haslam, L. Rothschild, L., & D. Ernst, Essentialist beliefs about social categories, British Journal of Social Psychology, 39, (2000): 113-127; N. Haslam, L. Rothschild, & D. Ernst, Are essentialist beliefs associated with prejudice? British Journal of Social Psychology, 41, (2002): 87100; S. Demoulin, J. Leyens, & V. Yzerbyt, Lay theories of essentialism, Group Processes and Intergroup Relations, 9,(2006): 25-42. A related and fascinating line of empirical research has been pioneered by Leyens and colleagues under the heading of infra-humanization. They present a series of empirical findings that suggest people are more reluctant to attribute uniquely human emotions to members of (at least some) out-groups. While people readily attribute to out-group members emotions that are shared with animals (e.g. fear, anger, surprise), they are less likely to attribute uniquely human sentiments (e.g. shame, resentment, love) to out-groups members vs. in-group members. The researchers argue convincingly that this reflects a tendency to deny fully human essence to certain out-groups. This work thus suggests that, not only are out-group members seen as possessing a distinctive essence, but that essence is fundamentally less than fully human. Conversely, one might interpret their findings as reflecting that, to the extent that the in-group is essentialized at all, its essence consists of a purely human essence. See, e.g., J-Ph. Leyens, A. Rodrguez-Prez, R. Rodrguez-Torres, R. Gaunt, M. P. Paladino, J. Vaes, et al, Psychological essentialism and the differential attribution of uniquely human emotions to ingroups and outgroups, European Journal of Social Psychology, 31, (2001): 395411; J.-Ph. Leyens, B. Cortes, S. Demoulin, J. F. Dovidio, S. T. Fiske, R. Gaunt, et al. Emotional Prejudice, Essentialism, and Nationalism, European Journal of Social Psychology, 33, (2003): 704717.
represent a belief in essence. A mysterious mana (for good or ill) resides in a group, all of its members partaking thereof. 33 Rothbart and Taylor argue that we may view some social kinds such as racial or ethnic kinds not simply as essentialized, but more specifically as natural kinds, and in some cases as biological kinds. This may be true, but it obscures the central notion. In the case of animal kinds, there is no need to separate out these different threads of belief, but in the case of social kinds it is imperative to do so. The notion of essence or at least the notion of essence that is relevant for our purposes here is not limited to biological kinds, or even to natural kinds (though such kinds constitute paradigmatic examples of essentialist thinking). For the remainder of this paper, I will understand a kind or group to be essentialized just in case its members are viewed as sharing a fundamental nature that causally grounds a substantial number of their outwardly observable properties. This nature need not be biologically grounded, nor need it be seen as immutable or strictly necessary for membership in the kind. 34 With this understanding of the notion of essence we can generalize the idea of a basic-level kind to the social arena: these will be social kinds that are perceived to have essences that occupy a sweet spot in trade-offs between distinctiveness (which is compromised as groups become less inclusive) and predictiveness (in the sense of grounding
33 34
G. Allport, The Nature of Prejudice (Reading, MA: Addison Wesley, 1954), pp. 173-174. Recent work in social psychology has begun to unravel the various threads of essence. For example, Haslam and colleagues (2000, op. cit.; 2002, op. cit.) distinguish between the perception of a group as a natural kind (e.g. having sharp boundaries, being determined by nature rather than man, being immutable) and the perception of a group as entitative. A group is highly entitative if its members are perceived as being very similar to one another, and if membership in the group is highly informative about the nature of the individual. Their notion of entitativity most closely corresponds to how I am here understanding essence, though I hesitate to adopt this terminology, because it is put to somewhat different uses by other social psychology researchers. (For example, Demoulin et al., op. cit. understand a group to be entitative if, in addition to Haslam et al.s criteria, the group has common goals and will face a common fate.) Interestingly, Haslam et al. (2000, op. cit.) found that groups that were viewed as more entitative tended to be accorded lower status in society, but the degree to which groups were viewed as natural kinds did not predict their perceived status. B. Bastian and N. Haslam (Psychological essentialism and stereotype endorsement Journal of Experimental Social Psychology, 42, (2006): 228-235) further found that people who were inclined to essentialize social groups were more likely to endorse social stereotypes, and were also more likely to attribute the persistence of stereotypes within a culture to the nature of the group being stereotyped (i.e. as opposed to socio-cultural conditions).
the maximal number of common features a feature which is compromised as groups becomes more inclusive). These kinds will thus be taken to have highly distinctive essences that ground a large number of shared features. Such social kinds, we may suppose, will surely include racial, ethnic, and religious kinds. As with basic-level animal kinds, these social kinds play a privileged role in our inductive practices. The perception that they share highly predictive natures makes us willing to generalize across the group, and the perception that these natures are distinctive means that we may be reluctant to generalize to more inclusive categories to the category human being, for example. As with animal kinds, it is these highly essentialized social kinds that are the typical targets of our striking property generalizations. To recap, then, our most primitive generalizations, voiced in language as generics, are especially sensitive to information that is particularly striking, horrific, or appalling. When we encounter individuals engaging in such an act, we are naturally inclined to seek to generalize this action to a kind to which the individuals belongs. The correctness conditions of these generalizations require that some members of the kind must indeed have the relevant property, and also that the other members must be typically disposed to have the property. We do not, however, normally have good information about unobservable dispositions available to us, so as a proxy, we generalize the property to a kind that we perceive to have a highly predictive and distinctive essence. In the social domain, this means that aberrant actions on behalf of the few such as the 9/11 terrorists, or the Native Americans involved in the Minnesota massacre, or the Algerian Khemil Ousliman, and so on can lead us to draw conclusions about an entire social kind. To the extent that Muslims, say, are perceived as forming a deeply uniform group whose nature is quite different from other comparable
groups (e.g. Christians, Jews), conclusions such as Muslims are terrorists may be drawn on the basis of the actions of the few. It is important, I think, to emphasize that this notion of essence or nature here is not a biological notion. An essence can of course be biologically grounded, but it need not be: a kind can be essentialized without being construed as a biological kind. 35 Much of the philosophical literature on race places considerable emphasis on the faulty conception of race as biologically grounded. Race is not a biological notion; there is no genetic ground for dividing people up along racial lines. 36 This is a tremendously important point that should be made repeatedly; however, we should not overestimate its capacity to alter peoples prejudiced convictions. A belief in essence in a shared nature may be the more important belief to change, but beliefs in essence can survive the loss of belief in a biological essence. 37
Interestingly, N. Haslam and S. Levy (Essentialist beliefs about homosexuality: Structure and implications for prejudice, Personality and Social Psychology Bulletin, 32, (2006): 471-485) found that people who endorsed biological explanations for homosexuality were less likely to evidence anti-gay prejudice. However, Haslam et al. (2002, op. cit.) found that the degree to which people perceived gay men to constitute an entitative group (see footnote 34) did correlate with prejudice. One potential explanation of this phenomenon is that people who believe that sexual orientation is a simple biological phenomenon do not consider gay people to have fundamentally different natures from heterosexual people, but rather view sexual orientation as a more adventitious phenomenon (more akin to differences in taste in movies, or in athletic abilities, perhaps). People who view homosexuality as a choice and more specifically as a morally reprehensible choice may take this decision to act immorally (by their lights) as indicative of a fundamentally different nature. (E.g. I could never choose to do something so despicable, so anyone capable of making that choice must be deeply different from me and mine.) Thus belief in shared nature can come apart from beliefs about biological bases, as this case perfectly illustrates. Homosexual people and perhaps gay men in particular (Haslam and Levy, op. cit.) may still essentialized by others even absent belief in biology. (And of course striking property generalizations concerning gays abound: gays are pedophiles, gays are sexual predators.) 36 For clear and accessible discussion of the biology and its potential philosophical significance, see K. A. Appiah, Race Culture, Identity: Misunderstood Connections, in K. A. Appiah & A. Guttman (eds.) Color Conscious: The Political Morality of Race (Princeton, NJ: Princeton University Press, 1996: 30-106); K. A. Appiah, How to Decide if Races Exist, Proceedings of the Aristotelian Society, 106, 1, (2006): 365-382. 37 A further concern one might have with the emphasis on the non-biological nature of race is that it seems to support the wrong counterfactuals. For example, if geneticists discovered that in fact there were significant genetic differences between races, what would follow? Using the non-biological grounds of race as the basis for discussion of prejudice and as the cornerstone of philosophical thinking about how to combat it seems to hold such efforts unnecessarily hostage to the deliverances of empirical science. Greater genetic variance need not translate into deep difference in essence or nature. The real issue here would seem to be the latter, rather than the former.
35
To illustrate, we may consider attitudes towards Muslims a group that is rarely if ever taken to be biologically defined. The absence of belief in biological essence does not preclude belief in an essence more broadly understood, as is illustrated by persisting beliefs that Muslim are fundamentally different from other groups of people, while also being fundamentally all alike. For example, in an influential article entitled The Roots of Muslim Rage, emeritus Princeton professor Bernard Lewis discusses Islam and its followers. He argues that we are now facing nothing less than a clash of civilizations, since Islamic fundamentalism has given an aim and a form to the otherwise aimless and formless resentment and anger of the Muslim masses (my emphasis note the high level of generality implied by his language). 38 Lewis goes on to describe the followers of Islam in generic, essentialized terms: There is something in the religious culture of Islam which inspired, in even the humblest peasant or peddler, a dignity and a courtesy toward others never exceeded and rarely equaled in other civilizations. And yet, in moments of upheaval and disruption, when the deeper passions are stirred, this dignity and courtesy toward others can give way to an explosive mixture of rage and hatred with impels even the government of an ancient and civilized country even the spokesman of a great spiritual and ethical religion to espouse kidnapping and assassination, and try to find, in the life of their Prophet, approval and indeed precedent for such actions. 39 According to Lewis, in virtue of following Islam, the arbitrary Muslim be they peasant or peddler is disposed in times of calm to be most courteous towards other, but when faced with upheaval, this persons finer dispositions give way to an explosive mixture of rage and hatred. Lewis suggests that there is just something about Islam that affects people in this way. In virtue of being Muslim, people possess the dispositions he describes. This is a way of essentializing the followers of Islam, of attributing a shared essence to them and further it
38 39
B. Lewis, The Roots of Muslim Rage. The Atlantic Monthly, 266, (September 1990). Ibid.
is one that causally grounds their putative violent dispositions. In effect, Lewis is offering a rationalization of anti-Muslim striking property generalizations. Essence and Attribution Errors Highly essentialized social groups are prime targets of striking property generalizations. Extreme actions committed by a very small number of the members of such a group can result in the acceptance of a generalization concerning the whole group. If the group is highly essentialized then we readily suppose that the disposition to such action is widely shared among the members indeed that it is grounded in the very nature of the kind. When presented with a) strikingly awful actions, and b) a highly essentialized group, our primitive cognitive mechanism delivers the corresponding generalization on this basis alone. One interesting observation is that we do not tend to make these generalizations concerning groups to which we belong. Groups of which we are members seem to be less likely to be targets of this particular sort of generalization. Of course, we may well be aware that some members of the familiar group may have committed some or other horrific deed, but we do not generalize this information in the way that we do when dealing with members of our own groups. The difference, I propose, lies in the differential tendency to view the deplorable actions as grounded in the nature or essence of the group in question. If a member of a group to which we belong commits an appalling act and there is no highly essentialized sub-group that includes the individual but not us we do not reach beyond that individual in attributing the disposition to so act. If such an individual belongs to a highly essentialized group that excludes us, however, we may well view the inclination to appalling action as part and parcel of the essence of that group.
If this hypothesis is correct, we have thereby identified another sort of attribution error. Long ago, psychologists identified the so-called Fundamental Attribution Error, which consists in a certain asymmetry between attributions to oneself and to others. People tend to attribute their own less-than-stellar behavior in a given situation to aspects of that situation, and yet attribute identical behavior in others to persistent personality traits. For example, we may attribute anothers involvement in a car accident to her being a bad driver, but would attribute our own involvement to external factors such as poor weather conditions. 40 There has been some discussion of another attribution error, known as the Ultimate Attribution Error (UAE), which is concerned with attributions across groups. 41 To be guilty of the UAE, one must be inclined to attribute negative actions of members of groups to which we belong (in-groups) to situational factors, but attribute the negative actions of members of groups to which we do not belong (out-groups) to persistent traits. There is some empirical evidence that we are susceptible to the UAE, but it is less than overwhelming. 42 The attribution error that I have in mind should not be confused with the Ultimate Attribution Error. The UAE would have us explain away negative actions of in-group members by reference to situational factors in a way that parallels our tendency to explain away our own bad behavior as situationally determined while attributing the negative actions
40
E. E. Jones & R. E. Nisbett, The actor and the observer: Divergent perceptions of the causes of the behavior, in E. E. Jones, D. E. Kanouse, H. H. Kelley, R. E. Nisbett, S. Valins and B. Weiner (eds.), Attribution: Perceiving the causes of behavior (Morristown, NJ: General Learning Press, 1972: 79-94). 41 T. Pettigrew, The Ultimate Attribution Error: Extending Allports Cognitive Analysis of Prejudice, Personality and Social Psychology Bulletin, 5, 4, (1979): 461-476. 42 For a review, see M. Hewstone, The Ultimate Attribution Error? A Review of the Literature on Intergroup Causal Attribution, European Journal of Social Psychology, 20, (1990): 311-335. Some of the best evidence in support of the Ultimate Attribution Error may be found in work done on the language used to describe actions. There is a tendency to describe negative actions committed by in-group members in more situational, less characterizing language than that used to describe negative actions committed by out-group members. See A. Maass, D. Salvi, L. Arcuri, & G. Semin, Language use in intergroup contexts: The linguistic intergroup bias, Journal of Personality and Social Psychology, 57, (1989): 981-993.
of out-group members to more persistent inner factors. The Ultimate Attribution Error thus offers the same broad contrast as the Fundamental Attribution Error. However, the attribution error I have in mind here perhaps we should call it the Supreme Attribution Error, so as not to be outdone is concerned with a different contrast. We may very well explain negative behavior of in-group members by reference to their persistent personality traits, but we will not explain it by reference to the group essence. In contrast, when faced with highly negative behavior on the part of members of unfamiliar essentialized groups, we may take the disposition to such behavior to belong not only to the personality of perpetrator, but to the very nature or essence of the group. 43 To illustrate the difference between the UAE and the attribution error I am discussing here, consider the reactions of Americans to the Oklahoma City bombings, and to the bombings of 9/11. To commit the UAE here, one would have to attribute Timothy McVeighs actions to situational factors, while refraining from offering such excuses for the 9/11 bombers. I imagine that few people are inclined to think that McVeigh was simply a victim of circumstances. By contrast, to commit the attribution error that I am describing, one can believe that McVeighs actions were wholly the consequence of his rather demented personality. The attribution error would lie in understanding McVeighs dispositions to be grounded in his individual personality, while taking the 9/11 bombers to be manifesting in
43
To the best of my knowledge, this particular attribution error has not been tested, nor even discussed. It is a wholly empirical question whether we are in fact disposed to commit this attribution error my aim here is to put it forward as a testable hypothesis. The closest empirical test that I have been able to uncover in the literature is V. Y. Yzerbyt, A. Rogier, & S. Fiske, Group Entitativity and Social Attribution: On translating situational constraints into stereotypes, Personality and Social Psychology Bulletin, 24, (1998): 10901104. They found that for more entitative groups (see footnote 34), we tend to overestimate the extent to which behavior of members is attributable to the groups nature vs. to situational characteristics. When the groups in question were not perceived as entitative, this effect was not found. However, the behavior in question here was mildly positive rather than strongly negative, and importantly no asymmetry was found depending on whether the group being assessed was an in-group or an out-group. It is precisely in this respect that I would predict an asymmetry in the case of strikingly awful behavior, though not in the case of bland or positive behavior.
the words of Bernard Lewis an explosive mixture of rage and hatred inherent in Muslims. 44 It bears emphasizing once again that similar conclusions are not drawn when we are faced with neutral or positive information. This bias is not traceable to the essentialization of these kinds alone. If we look out of our window one morning and see a group of Muslims jumping on pogo sticks, we are not inclined to conclude something very general about Muslims. We do not suppose that the essence of the group grounds the disposition to jump on pogo sticks. When we encounter Muslim doctors, we do not conclude Muslims are doctors, and suppose that the non-doctors among them are nursing a yet-unmanifested disposition to practice medicine. And so on with an open-ended range of examples. The disposition to generalize in this way is specific to the strikingly negative, to the threatening, the vile, the horrific, and the dangerous.
This attribution bias should not be confused with the out-group homogeneity bias (G. A. Quattrone and E. E. Jones, The Perception of Variability Within Ingroups and Outgroups: Implications for the Law of Small Numbers, Journal of Personality and Social Psychology, 38, (1980): 141-152; E. E. Jones, G. C. Wood & G. A. Quattrone, Perceived Variability of Personal Characteristics in In-groups and Out-groups: The Role of Knowledge and Evaluation, Personality and Social Psychology Bulletin, 7, (1981): 523-528; B. Park & M. Rothbart, Perception of Out-group Homogeneity and Levels of Social Categorization: Memory for the Subordinate Attributes of Ingroup and Outgroup Members, Journal of Personality and Social Psychology, 42, (1982): 1051-1068). The out-group homogeneity bias consists in peoples perceiving out-groups to be more homogenous than in-groups. The magnitude of this effect while robust and statistically significant is not large; we are inclined to rate out-groups as only slightly more homogenous than in-groups (B. Mullen & L. Hu, Perceptions of Ingroup and Outgroup Variability: A Meta-Analytic Intergration, Basic and Applied Social Psychology, 10, 3, (1989): 233-252). It does not entail our perceiving an out-group to be such that all its typical members share a given disposition. The out-group homogeneity bias is also concerned with a wide range of traits, including (often primarily) positive and neutral ones. It is not at all limited to strikingly negative ones; in fact, if a study of perceived out-group homogeneity that focuses on strikingly negative characteristics exists, I have not found it. There is also evidence that members of minority groups may exhibit the reverse bias, rating themselves as more homogenous than the majority out-group (A. Guinote, The Perception of Group Variability in a NonMinority and a Minority Context: When Adaptation Leads to Out-group Differentiation, British Journal of Social Psychology, 40, (2001): 117-132: B. Simon & R. Brown, Perceived Intra-group Homogeneity in Minority-Majority Contexts Journal of Personality and Social Psychology, 53, (1987): 703-711; B. Simon & T. Pettigrew, Social Identity and Perceived Group Homogeneity: Evidence for the Ingroup Homogeneity Effect, European Journal of Social Psychology, 20, (1990): 269-286). It would be surprising, though, if minority members were inclined to make these striking property generalizations over their own groups, however (though the question is ultimately an empirical one, of course). The out-group homogeneity bias is a well-documented and important bias, but it does not explain the attribution bias I am concerned with here.
44
Looking for Hope: Familiarity and Identification That we resist making such generalizations concerning groups to which we belong does not, on the face of it, hold much promise for helping to reduce the tendency to generalize in this way. Simply put, one cannot be a member of every essentialized group. However, increasing familiarity, knowledge, and solidarity may serve as a surrogate of sorts: people who are very familiar or identified with a class seem to reject striking property generalizations over that class. Those with considerable experience of Catholic priests tend to reject the generic claim that priests molest children. This claim is far more likely to be made by someone with only the most limited contact with Catholic priests. As a matter of sad fact, the incidence of molestation by fathers is significantly higher than by priests, yet few if any of us are inclined to accept the generalization fathers molest children. A key factor seems to be a significant degree of familiarity or even solidarity with members of the relevant kind. How do we then cope with sensational negative information concerning members of such familiar kinds? The case of dogs is telling. When a Rottweiler mauls a child, we do not impugn dogs in general. We instead seek a more restrictive generalization that cites a better predictor of the tendency to maul children. While a Labrador owner might rest content with the generalization that Rottweilers maul children, the Rottweiler breeder will further restrict the generalization to, say, poorly trained Rottweilers. If the search for less inclusive groups were carried to its ultimate conclusion, we would simply treat each individual wholly on their own terms. (As discussed above, if there are no available essentialized kinds that include the perpetrator but not ourselves, this may be exactly what we do.) Even short of this, however, there is much room for improvement. For
example, we might prefer such generalizations as disturbed fanatical extremists blow up buildings in place of Muslims blow up buildings. This search for increasing specificity may be part of a more general tendency, as familiarity and knowledge grows, to view increasingly restrictive sub-kinds as comprising the basic-level. In the biological domain, for example, recent findings suggest that the Itzaj Mayan people treat individual tree species as highly essentialized basic-level kinds, and the category tree as a less essentialized superordinate kind. American college students, however, treat tree as a basic-level kind. A natural explanation is that the Itzaj Mayans are simply more knowledgeable about trees than the American college students, and so the differences between the sub-kinds of trees are too salient to them to be ignored. 45 (Consider the neophyte wine drinker to whom there seems to be just two kinds of wine: red and white. In contrast, the most discerning oenophile might see each triple of vineyard, grapes and vintage as carving out so unique a kind that each must be considered solely on its own terms, with no higher generalizations permitted.) If the above observations concerning familiarity, knowledge and solidarity are true, this would further serve to expose the fatal, reinforcing logic of segregation. The more separated and distant we are, the more see each other in the terms that invite invidious striking property generalizations. These claims fit nicely with the so-called contact hypothesis, originally advanced by Gordon Allport in 1954. 46 He predicted that contact between members of different groups would reduce prejudices directed towards the other group, but only if a) the groups have equal status, b) they participate in cooperative activity towards common goals, c) the
45
J. D. Coley, D. L. Medin, J. B. Proffitt, E. Lynch, & S. Atran, Inductive Reasoning in Folkbiological Thought, in. D.L. Medin & S. Atran (Eds.), Folkbiology (Cambridge, MA: The MIT Press, 1999: 205-232). 46 Op. cit.
acquaintance is personalized, and d) the contact is sanctioned by authorities and/or social norms. Whether all these conditions must be met, and whether there are additional necessary conditions has been a matter of some controversy over the ensuing decades, but the core idea has received considerable empirical support. 47 Prejudice is reliably reduced when members of different groups interact in cooperative and personal ways. The conditions that the contact hypothesis sets forth may well be conditions that lead us to resist viewing broadly inclusive social kinds as highly essentialized. Looking for Hope: Language and Essentialism Familiarity and knowledge can get us to view increasingly less inclusive kinds as the basiclevel kinds. We might also consider which factors originally lead us to view some social kinds as possessed of widely shared and highly distinctive natures, and whether anything can be done to undermine this tendency. When such a belief in the groups shared nature is lacking, we do not tend to accept these generalizations for example, we do not think that accountants are murderers even though some accountants have been known to commit murder, nor do we think that fish attack bathers even though some fish that is, some sharks do indeed attack bathers. A belief that such a shared essence exists and grounds the disposition to the striking behavior is a precondition for these generalizations. If I am correct to suppose that striking property generalizations have played a role in the formation of some of our social prejudices, then the question arises, how might we most effectively combat and undermine these generalizations? The claim that the essentialization of these groups is a precondition here raises a tantalizing empirical suggestion: if we can
47
T. Pettigrew & L. Tropp, Does Intergroup Contact Reduce Prejudice? Recent Meta-Analytic Findings, in S. Oskamp (ed.) Reducing Prejudice and Discrimination (Mahwah, N.J.: Lawrence Erlbaum Associates, 2000: 93115).
identify factors which lead us to essentialize a given kind, then we may be able to develop novel means of combating prejudice. Recent work by Susan Gelman and her collaborators is, I think, particularly telling. Their focus has been primarily on the development of essentialist thinking about animals and other biological kinds. By age 4, children have essentialist beliefs about such kinds. 48 Since it seems that we may essentialize social kinds a similar manner, this invites the hypothesis that factors lead us to essentialize biological kinds might perhaps also contribute to our essentialization of non-biological social kinds. It is important to notice that, while the disposition to essentialize some kinds or others may be an immutable feature of our cognition, this does not determine which kinds we essentialize. While the available cross-cultural evidence suggests that essentialization may be a universal human phenomenon, there is significant cross-cultural variation in which kinds are seen as essentialized. 49 For example, the different castes in India have been seen as highly essentialized, especially by members of the upper castes. 50 The Ancient Greeks are often said to have believed that there were two fundamental kinds of human beings: Greeks and Barbarians, each endowed with their own distinctive natures. We might suppose that class in English society has, at least until very recently, been essentialized, and certainly medieval European notions of the Great Chain of Being involve highly essentialist ways of
48 49
Gelman, 2003, op. cit. E.g., S. Atran, Itzaj Maya folkbiological taxonomy: Cognitive universals and cultural particulars, in. D.L. Medin & S. Atran (Eds.), Folkbiology (Cambridge, MA: The MIT Press, 1999: 233-284); Coley et al., op. cit.; Gil-White, op. cit.; R. Astuti, G. E. A. Solomon & S. Carey, Constraints on conceptual development. Monographs of the Society for Research in Child Development, 69, 3, 277, (2004); for a review, see G. Diesendruck, Categories for names or names for categories? The interplay between domain-specific conceptual structure and language, Language and Cognitive Processes, 18, (2003): 759-787. 50 R. Mahalingam, Essentialism, Power, and Representation of Caste: A Developmental Study (Ph.D. Dissertation, University of Pittsburgh, 1998).
thinking about the different strata of society. (Rebellious peasants were often said to be going against nature itself.) It is not difficult to come up with many more such examples. Thus, while the capacity to essentialize may be an innate feature of human psychology, it does not fix which kinds are to be essentialized. We are not hard-wired to form the specific social categories that we do. 51 The examples just noted do not involve slicing humanity along the same lineaments as contemporary American thinking on race or ethnicity: the Greek term barbarian encompassed a wide variety of groups which we would distinguish in contemporary American society, while the social classes demarcated in medieval Europe cut more finely than the racial and ethnic categories which are so salient to us. Since it is not a fixed and immutable part of our psychology to essentialize the specific groups that we do, the urgent question arises, what determines which groups we essentialize? As with everything else in this domain, I am highly skeptical of the possibility of providing a single, univocal answer to this question. The issue here as elsewhere is partly historical, partly economic, partly social, and partly who-knows-what-else. There is a variant on the question, though, which may prove more tractable from the cognitive perspective. I propose that we instead ask the question: what are some of the cues that suggest to young children that a particular kind or group is to be essentialized?
51
For a fascinating empirical demonstration of how malleable our social categories (other than gender) may be, see R. Kurzban, J. Tooby & L. Cosmides, Can race be erased?: Coalitional computation and social categorization, Proceedings of the National Academy of Sciences, 98, 26, (2001): 15387-15392. These authors argue that, contra prior claims, racial categories are not automatically encoded when we perceive others. Rather, we primarily encode information about membership in coalitions. The authors argue that the encoding of race is just a special case of this more general tendency one that gets reinforced over a lifetimes participation in our society, thus acquiring an air of inevitability. It can be significantly undermined in certain contexts, however.
Though the question is far from settled, recent work on childrens understanding of race suggests that racial groups come to be essentialized only later in development. 52 This accords well with the observation that there is significant cross-cultural and historical variation in which social kinds are essentialized we do not come into the world hard-wired to essentialize racial kinds, and presumably the same also holds for ethnic and religious kinds. 53 This is something that we have to learn to do. How, then, do children learn to essentialize one kind rather than another? This question is beginning to be investigated, especially in the case of animal kinds. If the same cognitive processes are implicated in the essentialization of both animal and social kinds, then we might form the tentative hypothesis that findings concerning the one will apply to the latter. Current research is far from providing a full account of the phenomenon; however, it does point to some factors which can increase the extent to which children and adults essentialize particular kinds; in particular, recent findings suggest that the language we use to talk about individuals and groups can have a significant impact on the degree to which they are essentialized. 54 Here, as elsewhere in psychology, something that looks on its face to be
52
For example, M. Rhodes & S. A. Gelman, A developmental examination of the conceptual structure of animal, artifact, and human social categories across two cultural contexts, Cognitive Psychology, in press. Interestingly, gender may be essentialized significantly earlier (e.g. Gelman, 2003, op. cit.; M. G. Taylor, M. Rhodes, & S. A. Gelman, Boys will be boys, cows will be cows: Childrens essentialist reasoning about human gender and animal development, Child Development, 79, (2009): 1270-1287). And unlike racial categories, evidence suggests that gender categories are strongly and automatically encoded in person perception (Kurzban et al., op. cit.); see footnote 51. 53 For an evolutionary perspective on this point, see L. Cosmides, J. Tooby & R. Kurzban, Perceptions of Race, Trends in Cognitive Sciences, 7, (2003): 173-179; also Kurzban et al., op. cit. 54 On this point, I am in agreement with Hirschfeld, 1996, op. cit.; D. Sperber, Explaining culture: A naturalistic approach (Cambridge, MA: Blackwell, 1996); and S. Carey, On the origin of causal understanding, in D. Sperber, D. Premack, & A.J. Premack (eds.), Causal cognition: A multi-disciplinary debate (Oxford: Oxford University Press, 1995: 268308. Subsequent empirical findings have, I would argue, supported these theorists insights in this regard. See also Diesendruck, 2003, op. cit.
superficial may have a deeper significance because of the idiosyncratic character of our species cognitive structure. 55 One recent study is quite telling on this point. Susan Gelman, Elizabeth Ware, and Felicia Kleinberg designed three picture books about a novel animal, which differed from each other only in the language they used to talk about the animals depicted therein. 56 The first type of book made frequent use of generic language (e.g. zarpies hate ice cream), while the second contained no generics, but did make use of the common noun zarpie (e.g. this zarpie hates ice cream). The third book did not contain the common noun zarpie at all, and the individuals were simply referred to as this (e.g. this hates ice cream). A group of parents agreed to read one of these picture books frequently to their four year old children over a period of several weeks. The children and their parents were tested (separately) with a battery of tests designed to assess the extent to which they were prepared to treat zarpies as an essentialized kind (e.g. the extent to which they viewed zarpie traits as innate vs. learned, the extent to which they generalized novel properties from one zarpie to another, and a number of other such measures). Children and adults who saw/heard the zarpies described with generic noun phrases showed by far the greatest tendency to essentialize zarpies. This effect held up over a period of several weeks, including weeks in which the children were no longer actively reading the picture books, which suggests the effect may be quite robust. (In some respects, the effect of
In what follows, I wish to be clear that I am not endorsing the so-called Whorfian view that language determines cognition in any non-obvious way. Like many of the researchers on whose work I draw (e.g. Susan Gelman) I believe that the results and implications discussed can be readily accommodated on the rather bland supposition that language conveys information, both semantically and pragmatically, and that the information we grasp has an obviously important impact on how we conceptualize the world. For example, a child that hears a kind repeatedly described with generic language may come to essentialize that kind because the child recognizes that the surrounding adults see this as a kind worthy of discussion in generic terms. This recognition need not be conscious in order to be explicable in pragmatic terms. 56 S. A. Gelman, E. A. Ware & F. Kleinberg, Effects of Generic Language on Category Content and Structure, under review.
55
hearing generics actually grew more pronounced as time went on.) This study thus offers clear evidence that language plays a role in shaping the extent to which we essentialize a given kind. 57 Other studies have compared the use of nouns/labels to verbs or adjectives/descriptions, and found that the former foster essentialism, or at least related phenomena. 58 For example, Susan Gelman and Gail Heyman 59 found that five and nine year olds drew different conclusions depending on whether they were told, e.g., Rosie eats carrots whenever she can vs. Rosie is a carrot-eater. The use of the common noun led children to view the property as more stable, and they judged that Rosie was more likely to persist in eating carrots in the absence of the usual parental encouragement. 60 In an induction task conducted with Israeli preschoolers, Gil Diesendruck and Heidi haLevi found that fiveyear-olds preferred to generalize novel properties based on membership in familiar social categories as conveyed by labels (e.g. a Jew vs. an Arab), rather than based on either shared
Zarpies, as presented in the study, are intended to comprise a novel animal kind. Will the same effects of language be found if the group in question is a social kind? Marjorie Rhodes and I are currently collaborating to test this hypothesis there is no data available at time of writing, however. 58 For more discussion of nouns vs. adjectives, see A. Wierzbicka, Whats in a noun? (Or: How do nouns differ in meaning from adjectives?), Studies in Language, 10, (1986): 353389; see also Carey, 1995, op. cit. 59 S. A. Gelman & G. D. Heyman, Carrot-eaters and Creature-believers: The effects of lexicalization on childrens inferences about social categories, Psychological Science, 10, (1999): 489493. 60 Walton and Banaji found that the same held for adults, and Markman and Smith report a related contrast between nouns and adjectives (G. M. Walton & M. B. Banaji, Being what you say: The effect of essentialist linguistic labels on preferences, Social Cognition, 22, (2004): 193213; Markman and Smiths work is reported in E. M. Markman, Categorization and naming in children: Problems of induction (Cambridge, MA: MIT Press, 1989)). In a fascinating recent study, Carnaghi, Maass, Gresta, Bianchi, Cadinu, and Arcuri (Nomina Sunt Omnes: On the Inductive Potential of Nouns and Adjectives in Person Perception, Journal of Personality and Social Psychology, 94, 5, (2008): 839-859) extend a similar paradigm to social kinds, and again find a parallel contrast between nouns and adjectives in the responses of their adult participants. These studies on adults tend to focus on the elicitation of previously held beliefs (with the exception of Walton and Banajis experiment 1), however, rather than on the formation of novel ones. Since we are primarily concerned with formation rather than elicitation, my focus is on the studies that investigate the impact of language on the formation of new beliefs about essentialism.
57
physical traits or shared personality traits as conveyed by appearance and/or description (e.g. shy vs. friendly). 61 It must be emphasized that these linguistic cues are not sufficient for a kinds being essentialized. 62 We can and do use common nouns and generics to talk about nonessentialized kinds. In Gelman, Ware, and Kleibergs experiment, it is important to notice that zarpies are clearly depicted as a type of biological creature that is, language aside, the depicted zarpies formed a kind that was ripe for essentialization. (An interesting control task would be to look at the effects of language on less essentializable kinds, such as artifacts.) We should therefore not be misled into supposing that these linguistic cues are in and of themselves sufficient for essentialization of the kind. A very interesting question, though, is the extent to which these cues are necessary (or at least close to necessary that is, only rarely would these cues not be needed). Certainly, it is difficult if not impossible to think of an essentialized kind for which we do not have a labeling common noun. Even if such cues are not even strictly necessary, the evidence points to their being at the very least important contributing causes. One might thus reasonably hypothesize that altering these cues would have a powerful effect. The above results and others point, I think, to a novel proposal for combating prejudice. If prejudice and essentialism about social kinds are crucially linked, then we might suppose that undermining the latter will help mitigate the former. No studies to date have looked at possible ways to reduce existing essentialist beliefs, but the ones described above
G, Diesendruck & H. haLevi, The role of language, appearance, and culture in childrens social category based induction, Child Development, 77, (2006): 539-553. 62 Carey, 1995, op. cit. seems to suggest that labeling may be both necessary and sufficient for essentialism: Essentialism, like taxonomic structure, derives from the logical work done by nouns I submit that biological essentialism is the theoretical elaboration of the logical-linguistic concept, substance sortal (p. 267-277). Arguably, however, she has in mind a rather different conception of essentialism than the one under discussion here.
61
have identified some factors which promote essentialist thinking about a given kind. In particular, they shed some light on the cues that young children use to determine which kinds they should essentialize. The empirical evidence suggests that the use of nouns as labels promotes essentialism, and that the use of generics compounds this. The intriguing upshot of this is that our very choice of words to describe racial/ethnic/religious kinds may subtly communicate to our children that these kinds are to be essentialized. We need not say anything negative about these groups the use of generics or even simply labeling nouns may communicate that these are essentializable groups, and so open the door to prejudice. The contrast between nouns on the one hand and verbs and adjectives on the other suggests a possible way of mitigating this effect. In the early days of research on autism, researchers would often speak of autists or autistics using a noun to label this group of people. It was then thought that this promoted a certain undesirable way of thinking about this group, and so researchers were urged to instead speak of autistic people using an adjective instead. However, this sort of adjective-noun compound is all too easily heard as just another common noun unit (compare Bengal tiger). Nowadays, the preferred locution is instead people with autism a locution which emphasizes that they are people first and foremost, and that autism is just one property among many which they possess. The condition does not define them. 63 The role of common nouns as opposed to other linguistic devices such as adjectives in guiding categorization and generalization has been extensively studied in the non-social arena. A range of data suggest that common nouns have an impact on how we categorize and
63
For empirical confirmation of the effects of such rephrasing, see C. J. Cunningham, Illnesses as Labels: The Influence of Linguistic Form Class, (Undergraduate Honors Thesis, University of Michigan, 1999).
generalize from as young as fourteen months of age. 64 Interestingly enough, these effects are only found when the noun is clearly presented in a naming phrase (e.g. A blicket! but not Blicket!), reflecting that the common noun must be clearly used to indicate that the item in question belongs to a kind. 65 Further, for infants at least as young as sixteen months, using a description or an adjective (e.g. this is blickish!) in place of a noun does not produce the same effect. 66 These infants were significantly less likely to generalize a property across perceptually dissimilar items introduced by the same description, as opposed to dissimilar items introduced with the same common noun labels. Recent and innovative work by Sandra Waxman 67 has begun to extend these findings into the social realm. Waxman presented preschoolers with pictures of people of different races, and told them that one of the people depicted had a novel property (e.g. like to play a game called zaggit), then asked the children whether they thought the other people depicted would also have the property in question. In the first study, no labels were used, and Waxman found that the preschoolers did not use race a guide to their attributions rather
C. Fennell & S. R. Waxman, What paradox? Referential cues allow for infant use of phonetic detail in word learning, Child Development, in press; S. R. Waxman, Specifying the scope of 13-month-olds expectations for novel words, Cognition, 70, (1999):B35-B50; S. R. Waxman & A. E. Booth, Seeing pink elephants: Fourteen-month-olds interpretations of novel nouns and adjectives, Cognitive Psychology, 43, 3, (2001): 217242: S. R. Waxman & A. E. Booth, The origins and evolution of links between word learning and conceptual organization: New evidence from 11-month-olds. Developmental Science, 6, 2, (2003): 130-137. 65 Fennell and Waxman, in press, op. cit. 66 J. Keates & S. A. Graham, Category labels or attributes: Why do labels guide infants inductive inferences?, Psychological Science, 19, (2008): 1287-1293. 67 S. R. Waxman, Names will never hurt me? Naming and the Development of Racial and Gender Categories in Preschool Aged Children, European Journal of Social Psychology, in press. As best as I am able to tell, until recently very little work had been done on the social implications of using common nouns and/or generics. In a seminal paper, Semin and Fiedler characterized the differential social impact of some types of language, and argued that the more abstract the language, the more potent the social impact. Adjectives (e.g John is aggressive), argue Semin and Fiedler, are more abstract and therefore more socially potent than, say, descriptions of actions (e.g. John hit Mary). However, Semin and Fiedler do not consider any more abstract forms of language than adjectival descriptions. I would argue that common noun labels are more abstract (in their sense) and thus more potent than adjectives, and further than generics are more abstract and potent forms of speech still. See G. Semin & K. Fiedler, The Cognitive Functions of Linguistic Categories in Describing Persons: Social Cognition and Language, Journal of Personality and Social Psychology, 54, (1988): 558-568.
64
they attributed the property just as readily across racial boundaries as they did within. In her second study, however, the person to whom the novel property was attributed was either introduced with a description (e.g. this one likes to eat big lunches) or with an unfamiliar common noun (e.g. this one is a Wayshan). Waxman found that the children who heard the description were again just as likely to attribute the novel property across racial boundaries as within, but this was not so for the children who heard the common noun. These children tended to not generalize the novel property across racial boundaries, but rather confined their generalizations to members of the same racial group. Of course these were of necessity rather artificial conditions, but this ingenious experiment clearly highlights the impact of labeling as opposed to merely describing on preschoolers social thinking. 68 Another dramatic illustration of the power of labeling in social situations is due to Zijing He and Renee Baillargeon. 69 In their experiment, 35-month-olds watched three adults interact with each other. The three adults introduced themselves using unfamiliar labels: the first adult said Im a mayloo, the second said Im a boga, and the third said Im a boga, too. In one condition, the first adult then left, leaving the two bogas together. The second boga made a friendly gesture towards the first boga, and the gesture was either welcomed or rebuffed. The children were surprised when the gesture was rebuffed in this condition (as indicated by a looking-time measure). In another condition, the second adult was a mayloo rather than a boga. The same interaction ensued (the first mayloo left, and then the boga made a friendly overture toward the mayloo), but here the children were not surprised when the gesture was rebuffed in fact, they looked longer when the overture was reciprocated. These results indicate that as young as 35 months, toddlers have expectations about intra-
68 69
Waxman also conducted the same test for gender categories with similar results. Z. He and R. Baillargeon, Early Expectations about In-group and Out-group Agents, in preparation.
versus inter-group interactions, and that these expectations can be triggered by no more than a difference in common nouns. 70 The foregoing discussion suggests that we might consider altering our ways of speaking about race, ethnicity, religion, and so on. Instead of labeling a person as a Muslim, we might instead describe the person if needed as, say, a person who follows Islam, thus emphasizing that person is the relevant kind sortal, and that following Islam is a particular property that the individual happens to possess. Instead of speaking of Blacks and African Americans we might instead adopt locutions such as people with darker skin. 71 (We might notice that already such a locution undermines the viability of essentialist thinking about race how much sense does it make to, e.g., persist in looking for genetic differences and inherent IQ differences between racial groups, if we think through the lens of this locution?) It might also be advisable to take some care in choosing our adjectival phrases. In particular, we should take care to avoid phrases which in their own way evoke an essentialist way of thinking. An extreme example of this would be, e.g., people who share a genetically determined nature with other darker-skinned people, but there are less dramatic possibilities which nonetheless may still promote essentialism. I know of no empirical data on this, but we might conjecture that reference to ancestry, blood, nature, and other related concepts would make for less-than-ideal adjectival phrases. Thus we might prefer people
70
At time of writing, the experimenters have not explored the contrast between nouns and adjectives using this paradigm; this would be a fascinating further study however. 71 The locution people of color would also seem to be a step in the right direction. We might, however, prefer an even more obviously descriptive locution, since there is perhaps still something conducive to essentialization in the mysterious of color. Of course people of color is most commonly used as a general term for non-White people rather than for Black people specifically. While such generality is not a necessary feature of a descriptive locution, it does highlight an interesting issue: could any genuinely descriptive phrase be strictly extensionally equivalent to a common noun label for an essentialized kind? I think the answer here is no if we take a kind to be essentialized, we expect the common noun label to apply just in case an individual partakes of the relevant essence, even if the individual does not superficially fit the descriptive bill for category membership. (Such a mindset is, of course, a pre-requisite for making sense of notions like passing e.g., looking White despite really belonging to a non-White category.)
with darker skin with its focus on outward, superficial properties to people of African descent, since the latter leaves more conceptual room for thinking in terms of shared, hidden, heritable properties. Adopting such a way of speaking and thinking may have some immediate benefits; the results of Carnagi and colleagues 72 suggest that hearing a member of a familiar social kind described by an adjective rather than a noun can reduce the extent to which adults expect the individual to conform to a stereotype. It is possible, though, that the real benefits would not lie just in the alteration of the attitudes of adults; the really intriguing possibility would be to decrease the extent to which children in our society grow up essentializing social kinds. The empirical results summarized here do not, of course, account for all the factors that lead children to essentialize, nor do they purport to. However, the evidence strongly suggests that the use of labels and generics contribute to essentialization, and so we may expect that the converse will also hold: reducing the use of labels and generics for racial, ethnic, and religious groups may reduce the extent to which children grow up essentializing these groups. Conclusion Depending on ones point of view, this linguistic revisionism may seem more or less appealing. The philosophical literature on race very often centers on the question of whether racial categories a) can be said to exist, and b) ought to be said to exist. Theorists are divided on this latter issue because of a number of complex reasons having to do with broad social, cultural, political, and economic factors. 73 I will not venture a view on these issues here,
Carnaghi et al., 2008, op. cit. Some excellent examples encompassing a variety of viewpoints include, K. A. Appiah, The Uncompleted Argument: DuBois and the Illusion of Race, in L. A. Bell and D. Blumenfeld (eds.) Overcoming Racism and Sexism (Lanham, MD: Rowman & Littlefield, 1995: 79-79); K. A. Appiah, 1996, op. cit.; S. Haslanger,
73
72
though of course my proposal will be far more appealing to theorists like Anthony Appiah and Naomi Zack, who believe that we should eschew racial categories for political, social, and ethical reasons. 74 This work adds one consideration in favor of the stance they advocate: we may be wired in a way that makes having essentialized social kinds without concomitant prejudice extremely difficult. Eschewing reference to social kinds, at least by way of common nouns and generics, may be a direct way to reduce the extent to which we unconsciously teach our children to essentialize. By comparison with Appiah and Zack, my perspective in this paper has been considerably more narrow I have explored these issues solely from the point of view of cognitive psychology. The best remedy for racial injustice from the political point of view may not be the same as, or even wholly consistent with, the best preventative strategy from the cognitive point of view. To raise just one difficult question: can the members of the aggrieved group operate effectively in the political and cultural arenas without their publicly essentializing the group in positive ways, thereby paradoxically taking on the risk of reinforcing the very kind of prejudice I have described? The perspective offered here is not intended as the final word on the matter nothing could be further from my intent. My aim has rather been to take a small slice of a huge phenomenon, and identify some of the contributing cognitive factors. Identifying these
"Gender and Race: (What) Are They? (What) Do We Want Them To Be?" Nos, 34, 1, (2000): 31-55; C. Mills, Blackness Visible: Essays on Philosophy and Race (Ithaca, NY: Cornell, 1998); L. Outlaw, On W.E.B. DuBoiss The Conservation of the Races, in Bell and Blumenfeld, 1995, op. cit., 79-102; L. Outlaw, On Race and Philosophy (New York: Routledge, 1996); R. Sundstrom, Racial Nominalism Journal of Social Philosophy, 33, 2, (2002): 193-210; P. Taylor, Appiah's Uncompleted Argument: Du Bois and the Reality of Race, Social Theory and Practice, 26, 1 (2000): 103-28; N. Zack, Race and Mixed Race (Philadelphia, PA: Temple University Press, 1993); N. Zack, Philosophy of Science and Race. (New York: Routledge, 2002). For a wonderfully illuminating and comprehensive discussion of how some socio-political considerations may interact with cognitive considerations, see D. Kelly, E. Machery & R. Mallon, Racial Cognition and Normative Racial Theory, in J. Doris, S. Nichols and S. Stich (eds.) The Oxford Handbook of Moral Psychology (Oxford: Oxford University Press, forthcoming). 74 K. A. Appiah, 1995, op. cit.; K. A. Appiah, 1996, op. cit.; N. Zack, 1993, op. cit.; N. Zack, 2002, op. cit.
contributing cognitive factors leads us to pose the question of how best to alter those factors, and thereby combat those aspects of prejudice that arise from the original sin of cognition, namely its primitive tendency to generalize striking negative information across the members of highly essentialized kinds.
This file is to be used only for a purpose specified by Palgrave Macmillan, such as checking proofs, preparing an index, reviewing, endorsing or planning coursework/other institutional needs. You may store and print the file and share it with others helping you with the specified purpose, but under no circumstances may the file be distributed or otherwise made accessible to any other third parties without the express prior permission of Palgrave Macmillan. Please contact rights@palgrave.com if you have any queries regarding use of the file.
PROOF
AQ1
Eros and the Redemption of the Gods1

Sarah-Jane Leslie
Roger Scrutons Death-Devoted Heart is probably the most satisfying musical, critical, and philosophical interpretation of Tristan und Isolde in the English language. One of the most illuminating aspects of this book lies in Scrutons original charting of the streams of Kantian philosophical anthropology running through Tristan und Isolde. Of course, it has long been commonplace to understand Wagner as a Schopenhauerian acolyte, who came to the sage of Frankfurt by way of Hegel and Feuerbach, but Scruton is the first to lay the groundwork for the Kantian framing of Tristan und Isolde. He writes: According to Kant, human beings stand in a peculiar metaphysical predicament one not shared by any other entity in the natural world. We see ourselves, he argued, in two contrasting ways both as objects, bound by natural laws; and as subjects, who can lay down laws for themselves. The human object is an organism like any other; the human subject is in some way transcendental, observing the world from a point of view on its perimeter, pursuing not what is but what ought to be ... [A]ll human distinctness and all human dignity reside in our transcendental freedom. To abuse that freedom is to degrade human nature; to respect it, in both self and other, is to enter the ideal Kingdom of Ends that is our abstract and mystical reward. (Death-Devoted Heart, p. 123) Human existence embodies a metaphysical paradox, which Kant resolves by way of the distinction between the phenomenal and the noumenal: on the one hand we are animals bound by the laws of the empirical or phenomenal world, yet in ourselves we are inescapably
99
9780230251687_08_cha07.indd 99
3/22/2012 9:52:31 AM
PROOF
100 Sarah-Jane Leslie
free, rational, and moral subjects of experience. The deepest moral commandment is to recognise the embodied noumenal freedom of human beings, to treat oneself and others as subjects rather than objects, as ends rather than means. Scruton interprets Tristan und Isolde through this Kantian lens, a lens which is further tinted by his phenomenological understanding of erotic love as deeply revelatory of the others embodied freedom, an understanding developed in detail in his well-known work, Sexual Desire. While there is nothing in Wagners intellectual formation which would explicitly mark him as a Kantian, Wagner is unquestionably influenced by Hegel, Feuerbach, and Schopenhauer, and it would not be surprising if he had recovered the central Kantian vision from these thinkers. Here I wish to set aside the issues of actual historical influence and focus instead on the fertility of this Kantian framing of erotic love as it appears in Wagner. This philosophical conception of the erotic is beautifully evoked in Scrutons analysis of Tristan und Isolde, and forms the foundation upon which Scrutons main insights rest. This essay begins by amplifying Scrutons characterisation of the erotic as it figures in Tristan, and then considers Wagners two other mature tragedies Der Ring des Nibelungen and Parsifal in this light. Are Scrutons insights in Death-Devoted Heart limited to Tristan, or do they penetrate something more general and profound in Wagners world view? I argue that Scrutons Kantian erotics provide a backdrop against which all three of Wagners mature tragedies should be interpreted, and that this way of viewing them resolves some long-standing puzzles about the meaning of the Ring and Parsifal. This in itself lends support to Scrutons analysis in Death-Devoted Heart, as it shows his insights not to be restricted to this single opera. In doing so, I present novel interpretations of the Ring and Parsifal interpretations that avoid the pitfalls that plague the standard conceptions of the works. In particular, this interpretative framework allows us to understand Parsifal as continuous with the moral and philosophical spheres of Wagners earlier work, instead of representing a strange and even disturbing aboutface, as critics since Nietzsche have so often maintained. But first let us examine just how Scrutons Kantian erotics the idea of the erotic as the celebration of the incarnation of the noumenal helps illuminate Tristan und Isolde.
Tristan und Isolde
If erotic love is to be love, it must be essentially directed to the embodied subject. Sexual love plays on our dual nature we are not disembodied
9780230251687_08_cha07.indd 100
3/22/2012 9:52:31 AM
PROOF
Eros and the Redemption of the Gods 101
spirits, nor are we simply objects. Erotic engagement between pure transcendental agents is unthinkable, since the bodily aspect of erotic activity is essential. Desire is directed towards the incarnate person, the subject-made-flesh: ... the body exalts and reveals the person, shows the subject in the object, and makes of that subject an object of desire of the desire to be united with this person ... Our sexual emotions are founded on individualising thoughts: it is you who I want and not the type or pattern. This individualising intentionality does not merely stem from the fact that it is persons (in other words, individuals) whom we desire. It stems from the fact that the other is desired as an embodied subject and not as a body. (Death-Devoted Heart, pp. 1412) Erotic love brings the other as an embodied Thou most starkly into view; on pain of the pornographic degrading of the sexual act, the others body cannot be seen as a mere object, but rather must be seen as the flesh-and-blood house of the soul. In the sphere of free erotic love, we come in contact with the sacred: the subject incarnate in the flesh before us. But in the grip of the pornographic attitude we encounter the flesh as an obscene but fascinating means to our own pleasure, and this precisely because the embodied subject is removed from view. The erotic is desecrated by the pornographic. By treating the pornographic attitude as a profanation of the embodied subject, Scruton thereby suggests that our sense of the erotic is deeply tied to our sense of the sacred. The sacred appears when we see an item in this world as housing or incarnating something not wholly of this world; we experience something in the phenomenal realm as pervaded by the transcendental. True religion and true morality consist in the recognition of the transcendental in the phenomenal, and hence of the subject in the flesh a recognition manifested in the accompanying refusal to treat the other as a mere replaceable thing. If erotic desire is one sphere of the sacred, then forced sexual relations constitute a particularly vile profanation or sin. Following this devolution of the erotic to its conclusion, we arrive at the ultimate sacrilege, against which Wagner rages both in Der Ring and Tristan: a person robbed of all vestiges of subjecthood, forced to submit repeatedly to unwanted sexual advances, forever in the bondage of objecthood: the institutionalised rape of forced marriage. We first encounter Isolde en route to such a fate. She is being forcibly dragged from her homeland to marry Tristans uncle, King Marke. She
9780230251687_08_cha07.indd 101
3/22/2012 9:52:31 AM
PROOF
has been chosen by others as a suitable bride because of her looks and station. Had another met these criteria, that other would have done just as well as a bride for the King. It is not Isoldes inherent nature as an individual, free, and irreplaceable subject that is prized here, but rather something cruder and more familiar. Isoldes autonomy is being trampled: she wants no part of this marriage. She is being shipped to Cornwall as a spoil of war, an object among objects, an alluring cargo unable to determine her own fate. The impending desecration infuriates Isolde throughout the opening scene of the opera. Forced marriage, and forced sex more generally, are recurrent themes in Wagners works. In contrast to these dehumanising horrors, Wagner offers us movingly rendered visions of true erotic love, most notably the love of Siegmund and Sieglinde, and that of Tristan and Isolde. In their love Tristan and Isolde are purged of any compromising sense of objectification, and any corresponding tendency to admit of substitutions. The lovers stand subject-to-subject, eye-to-eye, I-to-I. Each to the other is the irreplaceable core of life, the embodied noumenal. Thus, Scruton argues, Tristan und Isolde offers us an image of the sacred in the context of erotic love. What, then, of the shattering ending of the opera? What is to be made of the lovers joint sacrifice? One can easily feel that this is the only possible resolution of the opera: Tristan and Isoldes love is too pure and uncompromising for this world. Tristan and Isolde are transfigured in love and death, and this transfiguration has redemptive power: they thereby redeem love itself from the crimes that their social world has perpetrated against it crimes of objectification, substitution, and subordination. Through their love, and its culmination in death, Tristan and Isolde redeem themselves and their world, as King Markes final peroration makes clear. The sacred is recovered through the quintessentially human. As Scruton puts it: When writing of the redemption achieved by his lovers, Wagner is using this term in its true religious sense, to mean a regaining of the sacred in a world where sacrilege is the prevailing danger ... the message of Wagners greatest works is that man makes himself by sacralizing himself, and that by sacralizing himself he also sacralizes the world ... By setting aside the empirical world and its claims, by scorning death for the sake of a goal that only free beings can embrace or conceive, the act of sacrifice sanctifies the one who performs it. It brings the sacred into being ... The sacred moment, in which death is scorned for the sake of love, casts its light back over the entire life
9780230251687_08_cha07.indd 102
3/22/2012 9:52:31 AM
PROOF
that had led to it. Redemption ... consists in a changed conception of the empirical world a recognition that freedom really does exist in this world and that we too possess it. And this freedom is discovered in the most earthbound of our passions the passion of erotic love. As with the Christian vision, redemption requires incarnation; but in Wagner incarnation is no longer Gods means to redeem us from the world, but our means to redeem ourselves in it. (Death-Devoted Heart, pp. 1823) Read this way Tristan thus embodies a deeply religious sentiment. It is one in which the human being is his or her own redeemer, and in which the key to redemption is seeing another I-to-I, as erotic love enables us to do. Through their awed perception of the other, the lovers glimpse the sacred, and are led to prefer perfect union with the other in death to any of lifes inevitable compromises: Love leads us to sacrifice precisely through that aspect of it that nourishes our sense of the sacred: the individualizing intentionality that is disclosed in the look of love, and which we desecrate through all our compromises and substitutes. This sacrifice offers a kind of proof that we can transcend our mortal condition, that even in this passion that robs us of our freedom we are supremely free ... Death accepted for loves sake is a triumph over the empirical world, a final proof of freedom and personality against the meaningless flow of causes ... through their sacrifice [Tristan and Isolde] restore belief in our human potential and renew in us the will to live. Hence the redemption of the lovers in death is also a renewal of the community in life. And that is the religious meaning of Tristan und Isolde. (DeathDevoted Heart, pp. 1934) By taking love to its extreme conclusion, Tristan and Isolde do more than let us merely glimpse the sacred; we ourselves are sanctified through their sacrifice.
2 Der Ring des Nibelungen

How far can the moral and religious frame of Tristan, as Scruton constructs it, be usefully applied to Wagners other mature works? If the moral and religious vision of Tristan is anticipated in the Ring, we must, I claim, accept Brnnhilde as the cycles central salvific figure. Even as things lie on the surface of the Ring, this should come as no
9780230251687_08_cha07.indd 103
3/22/2012 9:52:31 AM
PROOF
Please provide better quality image?
Example 1 The Glorification of Brnnhilde leitmotiv
great surprise, for it is Brnnhilde who ultimately frees the world from Alberichs curse. Brnnhilde accomplishes what no man or god has been able to do: she relinquishes the ring to the Rhinemaidens, thus returning it to its rightful owners, purified of its curse. She brings to an end the cycle of horrors that had fanned out from Alberichs renunciation of love and theft of the gold. It is Brnnhilde who takes the stage in the final scene of Gtterdmmerung, and frees man and god alike from the scourge of the ring. The Rings final leitmotiv the last heard before the curtain falls also suggests Brnnhildes salvific prominence. This leitmotiv (see Example 1) occurs only once before the final scene: as a blessing for Brnnhilde. As her parting words in Act III, Scene I of Die Walkre, Sieglinde sings O hehrstes Wunder! Herrlichste Maid!, and so introduces the leitmotiv that will end the opera cycle. Since the leitmotiv is not heard again until the final scene of Gtterdmmerung, it does not pick up any other meanings or associations along the way it remains wholly associated with Brnnhilde, and with the blessing bestowed upon her by the hapless Sieglinde. This soaring motif is often referred to as the Redemption Motif (or Redemption by Love) in contemporary musical analysis. This is a perfectly adequate title, however it is not Wagners own: he named this final leitmotiv the Glorification of Brnnhilde. This more explicit designation makes vivid the motifs attachment to our heroine. (Though of course its musical history makes the same point and in a deeper and more telling way.) It is no misnomer to call it the Redemption Motif, but this description suggests that the motif can be detached from the character of Brnnhilde that it belongs to the more abstract notions of fate and salvation rather to this flesh-and-blood woman. The motif is one of redemption only because Brnnhilde herself redeems the world. The final musical word, in the language of the leitmotivs, belongs to Brnnhilde the Redeemer.
9780230251687_08_cha07.indd 104
3/22/2012 9:52:31 AM
PROOF
The only other character that one might reasonably be tempted to regard as the central salvific figure in the Ring is Siegfried, the hoped-for free being. To be sure, Siegfried plays an important enabling role in the worlds salvation: he wrests the ring from Fafner, thus paving the way for Brnnhilde to ultimately return it to the Rhinemaidens. This contribution is indispensable, but we must not mistake that for his actually performing the ultimate redemptive act, for that belongs to Brnnhilde alone. Siegfried also figures essentially in Brnnhildes transformation, first by joining with her in the transfiguring bond of love and by betraying her and mistreating her in a shockingly callous, instrumental way. Ultimately, Siegfrieds transgressions are too great for him to be a world-redemptive figure, and he lacks the self-knowledge required to see his own role in the events which engulf him. If we suppose that Brnnhilde is the Ring s central salvific figure, we no longer need to view the Ring as a work deeply compromised by Siegfrieds banality. That Brnnhilde eclipses him as the redeemer is not merely a matter of musical and dramatic fact; it is a matter of necessity from the point of view of the redemption of the gods. 2 Brnnhilde is the child of the Father-God Wotan his favorite child, his beloved. But the path of redemption requires the god to give up his child to the world of man. Certain pains and humiliations are known only to mortal man, but these pains and humiliations must be borne by the once-divine, if salvation is to be possible. The child of god must be made flesh. In the Christian narrative, the Son of God is made flesh to redeem mankind; the divine does not partake of this salvation. As Scrutons discussion of Tristan makes so vivid though, Wagner has no interest in portraying mans salvation as lying in the hands of the divine. In Tristan, salvation is a purely human affair: as Scruton puts it, Wagner is presenting in dramatic outline the image of man as his own redeemer (p. 185). In the Ring Wagner takes this vision a step further: man or more precisely, woman redeems the divine. Nothing in the text or stage directions suggests that the world of man is particularly affected by Brnnhildes actions. The Gibichung vassals survive the final fire and flood, and presumably continue on their way once the fire dies and the water recedes. They are bequeathed a world free of the ring and its curse, but otherwise we are given no indication that they are particularly affected by the eschatological events of Gtterdmmerung.
9780230251687_08_cha07.indd 105
3/22/2012 9:52:32 AM
PROOF
Rather, it is the gods themselves who are redeemed. It is their sins symbolised by the ring and its curse that Brnnhilde takes upon herself, that she expiates with her own death and suffering: Mein Erbe nun nehm ich zu eigen. Verfluchter Reif! Furchtbarer Ring! Dein Gold fass ich und geb es nun fort ... Das Feuer, das mich verbrennt, reinge vom Fluche den Ring!3 Now I take up my inheritance. Accursed ring, terrible ring, I take your gold and now I give it away ... The fire that consumes me shall cleanse the ring from the curse! (Gtterdmmerung, Act III, Scene III)
As she is devoured by fire, the ring is purified of its curse. The ring here represents the accumulated objectifications it has helped encourage, among them the objectifying misdeeds of the gods. Brnnhilde takes up these sins and dies to atone for them by proxy. As Scruton puts it: Yet, in another way, the gods can achieve redemption, namely, through the agency of a mortal. The great sacrifice of which mortals alone are capable, since it is a sacrifice motivated by mortal love, can spread its radiance over the immortals, provided they too participate. In this way Brnnhildes incarnation as a mortal becomes the instrument for the salvation of the gods, by rendering them mortal too. The significance of her immolation is that it is a willing renunciation of consciousness by the transcendental perspective itself in other words a self-immolation of the divine. (Lectures on Wagners Ring) The Christian narrative is thus tellingly turned on its head. The child of god becomes incarnate in mortal flesh, takes up the sins of others, and then suffers and dies to expiate them but here it is not the sins of man that are atoned for, but rather the sins of the gods. What precisely are these sins of the gods those which require such a redemption? I would suggest that they too, are sins of objectification. Scrutons elucidation of Tristans moral world applies equally to the world of the Ring: both are worlds in which the fundamental moral commandment is to see and treat others as Thous, not things; as means, not ends. These are the counts on which the gods are to be charged these are the sins that cry out for redemption.
9780230251687_08_cha07.indd 106
3/22/2012 9:52:32 AM
PROOF
From his first appearance in the Ring, Wotan is enmeshed in a profoundly instrumental view of those around him.4 When we first meet him in Das Rheingold he has bartered away his sister-in-law in exchange for his grand castle, Valhalla. We are quickly reassured that he does not intend to keep to the bargain; it was merely a manipulative gambit to induce the giants to work for him. He sees the giants as foolish brutes, there to be tricked and manipulated into doing his bidding. He is also quite unconcerned about the anxiety and humiliation this episode causes Freia he blithely uses her as a bargaining chip in his bid to use to giants to build his castle. This original episode of instrumentality has been set in motion before Wotan even appears on stage, but its ramifications endure until the very end of the cycle. It is Wotans original act of using Freia to manipulate the giants that necessitates his first wresting the ring from Alberich (thus precipitating the Nibelungs curse), and then, more significantly, necessitates his giving the ring to the giants, instead of returning it to the Rhinemaidens. Had Wotans false bargaining not boxed him into a corner, all manner of horrors could have been avoided. Here is the Fall that sets the stage for what is to come, but it is the Fall of the Gods; Man has nothing to do with it. This is no isolated incidence of instrumentality either. Wotans description of Valhalla and its heroes confided to Brnnhilde in his Act II monologue in Die Walkre is a candid description of purely instrumental, objectifying relations: Dass stark zum Streit uns fnde der Feind, hiess ich euch Helden mir schaffen: die herrisch wir sonst in Gesetzen hielten, die Mnner, denen den Mut wir gewehrt, die durch trber Vertrge trgende Bande zu blindem Gehorsam wir uns gebunden, die solltet zu Sturm und Streit ihr nun stacheln, ihre Kraft reizen zu rauhem Krieg, So that enemies would find us strong in battle, I told you to fetch heroes to me, such as once we masterfully subjected to our laws; men whose spirits we curbed, and, through shady treaties deceitfully binding, held them to us in blind obedience. You were to spur them to storm and strife, tempt their strength into bitter war
9780230251687_08_cha07.indd 107
3/22/2012 9:52:32 AM
PROOF
dass khner Kmpfer Scharen ich sammle in Walhalls Saal!
so that hosts of bold warriors would gather in Valhallas hall.
The assembled heroes of Valhalla are there not as a reward for deeds done in life, but to defend Wotan against his enemies. These heroes, once curbed by shady treaties, are roused by the Valkyries into mortal conflict with each other so that Wotan may collect the dead for his army. (However stately the Valhalla theme may seem, there remains something hollow in it, not just musically but dramatically.) And what of Wotans treatment of Siegmund and Sieglinde, whom he purports to love dearly? He allows Sieglinde, his daughter, to be abducted, raped, and abused by Hunding and his ilk (let us note in passing that he later desires to inflict a similar fate on another daughter, Brnnhilde herself). He abandons Siegmund to a life of lonely adversity, hunted and despised by all he meets. All this he does to his own children in the hope that Siegmund will be able to recover the Ring from Fafner. But then then! when he realises that Siegmund will never be free enough from him to obtain the ring, he condemns him to death. If the plan fails to work, so much the worse for Siegmund. Better to accede to Frickas demands, and so at least reap the small benefit of having mollified the wife. These are the sins that must be expiated. Wotan of course never goes so far as to renounce love, but he is rightly called Licht-Alberich because his sins are sins against love sins of treating others as objects, instead of as the subjects true love requires us to recognise. (We might note that even Wotans erotic encounters serve a purpose beyond union with the other: Erda was seduced for wisdom, and the Walsungs mother for the sake of begetting a free being.) Along with the gods, love itself must be redeemed redeemed from the abuses and curses it has endured. The ring represents these crimes against love. As Scruton writes, Marxs thoughts in Das Kapital about the fetishism of commodities ultimately derive from the Kantian contrast between treating people as ends and treating them as means. The same contrast animates Wagners vision of love and desire in Tristan and also in the Ring. Indeed the Nibelungs ring epitomizes this contrast: it is the object obtained by foreswearing love, so putting the transactions of the market in the place of personal union. (Death-Devoted Heart, p. 144)
9780230251687_08_cha07.indd 108
3/22/2012 9:52:32 AM
PROOF
The ring is the fruit of Schwartz-Alberichs original crime against love, and so comes to stand for all such crimes, including those committed by Licht-Alberich Wotan. Only with the fires of Brnnhildes pyre and the flood of the Rhine can the ring be purified; only thus can love be redeemed. *** Valhallas world a world of lofty ideals built on crude objectification is the only world Brnnhilde has ever known before she encounters Siegmund in Act II, Scene IV of Die Walkre. This scene is the pivot point in Brnnhildes transformation into a mortal woman, for here she sees something she has never before seen: pure love. She goes to Siegmund reluctantly, compelled by the henpecked Wotan to grant victory to Hunding. She is puzzled by Wotans command she has always believed that he loves Siegmund, yet since Wotan cannot will a free man to life she is told she must fight for Frickas slave, Hunding.5 Siegmund cannot play the required role of recapturing the ring, so Wotan has thrown him over in capitulation to his wife. Such is the way of Wotans Valhalla everyone has their use in the divine chess game, and their use-value trumps love every time. Wotan has cast untold afflictions on Siegmund for the sake of making him free, but Wotans plan of salvation by indirection has failed, so Siegmund is left to die. Brnnhilde leaves Valhalla to seek Siegmund on earth. As she greets him and speaks of the afterlife promised to him, Valhallas lofty theme plays throughout the orchestra. Brnnhilde is stately, serene, otherworldly a goddess descended from above to appear to a lowly mortal. It all sounds splendid, says Siegmund, but will he find his beloved Sieglinde in this other world? When Brnnhilde responds in the negative, the orchestra dies back, and in place of Valhallas pomp, we hear a lonely instrument playing the theme of the Walsungs love. Siegmund considers his reply for a moment, and then sings: So grsse mir Walhall, grsse mir Wotan, grsse mir Wlse und alle Helden, grss auch die holden Wunschesmdchen: zu ihnen folg ich dir nicht. Then greet Valhalla for me, greet Wotan too, greet Volsa for me and all the heroes, greet the lovely wishmaidens too. I will not follow you to them.
9780230251687_08_cha07.indd 109
3/22/2012 9:52:32 AM
PROOF
Siegfried will not leave Sieglinde. In a last attempt, Brnnhilde tries to get Siegfried to see Sieglinde in objectual terms: So wenig achtest du ewige Wonne? Alles wr dir das arme Weib, das md und harmvoll matt von dem Schosse dir hngt? Nichts sonst hieltest du hehr? So little do you value everlasting bliss? Is she everything to you, this poor woman who, tired and sorrowful, lies limp in your lap? Do you think nothing else glorious?
This sort of objectification is what Valhalla is built on, but Siegmund will have none of it. Because of his love for another mortal, Siegmund is above it all; he is better than the god and his hero-vassals. The divine thus finds itself shamed by mortal man. The god has abused Siegmund and since he is of no use now intends to betray and abandon him. But Siegmund himself will not betray and abandon his true love. Sieglinde is without price for him, she admits of no substitutions, precisely because she is for him not an object but a subject. She serves no purpose for him, and so cannot be traded or replaced. The love that has escaped the gods has found root in a lowly mortal. Siegmund sees something the gods have not he sees another as a Thou. Brnnhilde recognises this and is ashamed of herself and her divine kinsfolk. This flesh-and-bone human man has unknowingly raised himself above the gods: he is their moral superior, and he has seen what they cannot see. Brnnhilde realises at that moment perhaps unconsciously that if the gods are to be redeemed, they must be transformed into flesh. They must take on human form and embrace human love, else they will be forever morally below man. Following the similar theme in Carl Jungs Answer to Job, we might call this Brnnhildes Answer to Siegmund. In Jungs vision, Job shames the god when, after being needlessly tormented by Yahweh, he remains steadfast and faithful. Yahweh, Jung supposes, recognises the moral superiority of this decent human being, and so sees that in order to save Himself He must become flesh. Only by taking on human aspect can the divine atone for His misdeeds; only by becoming a man can Yahweh be redeemed. Brnnhilde takes the same path. She sides with the man over the god, and backs Siegmund in battle. This predictably enrages Wotan, who
9780230251687_08_cha07.indd 110
3/22/2012 9:52:32 AM
PROOF
responds by stripping her of her godhead and locking her in sleep. She will awaken as a mortal woman, and she will be roused by a mortal man. With this mortal man, she will have the chance to find love to attain the transformed consciousness that so inspired her in Siegmund. All this must happen if the gods are to be redeemed. Love alone, though, is not enough to effect Brnnhildes transformation into a world-redemptive figure. As there are sins that must be expiated, suffering is demanded. Brnnhildes suffering, however, is only possible because of her love for Siegfried; her trauma is the lovers trauma, the trauma of the spurned and rejected. The role of loves pain is evident in the final words of the so-called Schopenhauerian ending: Alles Ewgen Selges Ende, wisst Ihr wie Ich gewann? Trauernder Liebe tiefstes Leiden Schloss die Augen mir auf: enden sah Ich die Welt.6 The blessed end of all things eternal, do you know how I reached it? Deepest suffering of grieving love opened my eyes: I saw the world end.
Wagner never set these words to music, but this was not because he rejected their import rather he deemed that their meaning was contained in the music itself, and so the words were unnecessary.7 It is Siegfried who inflicts this pain on her when he drinks the memory-sapping potion and abandons her for Gutrune. That, however, is only the beginning. Siegfried then takes it upon himself to drag her from her mountain home, and force her to marry Gunther against her will. The very person whom she loves whom she most wishes would treat her as a subject forces upon her the ultimate objectification. She is distraught, humiliated, enraged. Her situation recalls Isoldes in Act I of Tristan. Of this Scruton writes: We discover Isolde wrestling with the awareness that the man consecrated in her desires is nevertheless treating her as a mere means and at the same time forcing upon her the loathsome vision of herself as object. Her anger at Tristan is thus also a longing to regain possession of herself, to be once again a free subject in her own eyes and in his. (Death-Devoted Heart, p. 152)
9780230251687_08_cha07.indd 111
3/22/2012 9:52:32 AM
PROOF
And again: The perspective on Isolde that is above everything loathsome to her [is] the view of her as sexual object, when her overwhelming and unconfessed desire is to be a sexual subject, freely giving to the other who freely gives in turn. (Death-Devoted Heart, p. 151) Both women are not only to be forced to marry against their will to be forced to submit sexually to men they do not care for but are forced to do so by men they love. These men who most of all have the duty to meet them I-to-I, as Scruton puts it, are instead meeting them I-to-It, Thou-to-Thing binding them as mere objects to be used as others wish. We quickly see that Tristan is as much a victim of the phenomenal world as Isolde, and that he soon rises to meet her squarely I-to-I. The same cannot be said for Siegfried, and he thus forfeits any claim to be the world-redeeming hero we might have hoped he would be. On the surface it seems that he is bound by events beyond his control he did not willingly forget Brnnhilde, but was rather tricked by Hagens Drink of Forgetting. That things are perhaps not so simple, however, is demonstrated by commentator Warren Darcys devastating question: Can we imagine a drink that would make Siegmund forget Sieglinde?8 Such a drink strikes us as impossible; their love is too profound, too transcendent. Love aspires to take us beyond the phenomenal realm, beyond the world of causally determined objects. When love is found in its purest form as exemplified by Siegmund and Sieglinde, and by Tristan and Isolde the phenomenal world cannot defeat it; love endures even the obliteration of its objects. Such is the love between Siegmund and Sieglinde, but the same cannot be said for their son. A drink an empirical force acting on him as object suffices to drive all memory of Brnnhilde from his mind. He accepts Gutrune as an immediate substitute, then after meeting the Rhinemaidens declares, And yet/had I not given Gutrune my word/I would cheerfully have chosen for myself/one of these pretty women!9 As Scruton writes, The crime against love is the admission of substitutes.10 Siegmund would have no substitutes he preferred death to eternal life without Sieglinde. Not so for Siegfried: his love for Brnnhilde could be dissolved with a potion, and a Rhinemaiden would do just as well as Gutrune. (Indeed, one cannot but worry that Siegfried, had the Valkyrie appeared to him as she did to his father, would have
9780230251687_08_cha07.indd 112
3/22/2012 9:52:32 AM
PROOF
wanted to know exactly how appealing Valhallas wishmaidens were before making his decision.) Even if we decide that Siegfried should not be held accountable for the effects of the drink, his subsequent treatment of Brnnhilde would still be inexcusable, since more is owed to even a stranger than Siegfried grants to her. No person ought to be abducted and forced into an unwanted marriage; this is an indefensible act of objectification, of desecration. Siegfrieds father finds himself wounded and weaponless at Hundings door precisely because he fought to spare an unknown girl this fate. He single-handedly fights and kills the kinsmen who are pressing her into a forced marriage. Siegmund willingly risks his life to defend a stranger against this horror. What, then, is to be said of his son, who dons the Tarnhelm to deceive his victim, physically overpowers her to take her ring, and forces her into submission overnight, then drags her off down the Rhine to be married and subsequently raped the next day? The drink here cannot afford Siegfried a defense, for this is appalling and repulsive behavior regardless of ones personal connection to the victim. At the wedding scene he continues to be completely unmoved by her plight he gives no indication of any human sympathy towards this strange woman he has abducted, and who is clearly utterly distressed. Rather, the scene ends with Siegfried remarking, Lasst das Weibergekeif!/Als Zage weichen wir gern/gilt es mit Zungen den Streit 11 (Enough of scolding women!/As cowards we gladly give way/when tongues wage war), then skipping gaily off the stage, singing his happy tune. The moral sphere of Wagners operas is not one in which such acts are treated lightly. This extreme objectification of another person is condemned: Siegfrieds acts necessitate his death. To be sure, he falls as a result of Hagens plotting, but Hagens designs only succeed because of Siegfrieds moral failings. We cannot imagine a drink that would make Siegmund forget Sieglinde, and we certainly cannot imagine a plot, however dastardly, that would lead Siegmund to then unperturbedly drag her back to be raped once more by Hunding. In his final moments, Siegfried manages some degree of awakening to the truth. At the moment of death, he sees Brnnhilde Self-to-Self. Brnnhilde, Heilige Braut, with its haunting, otherworldly quality, shows that Siegfried is not wholly lost. He is still capable of love, and at death his love for Brnnhilde emerges more clearly than ever before. Here, perhaps, Siegfried attains something of the purity of love his father Siegmund felt for Sieglinde. This moment helps to redeem Siegfried at a personal level; we must be careful not to read too much into this.
9780230251687_08_cha07.indd 113
3/22/2012 9:52:32 AM
PROOF
It might go some ways towards redeeming Siegfried, but it is hardly enough to redeem the world. Wotan himself has such a personal redemptive moment when he bids farewell to Brnnhilde. Until this final scene of Walkre, Wotan has never truly seen anyone as a pure subject, as a pure end-in-themselves. His relationships have all been deeply instrumental even his relationship with Brnnhilde, whom he calls the blind instrument of his will. Throughout the final scene he has been intent on punishing her disobedience in a rather vile way, and she has apologetically begged him not to do so. In her final entreaty, however, she casts off her apologetic, submissive aspect and for the first time she commands Wotan: Auf dein Gebot entbrenne ein Feuer; den Felsen umglhe lodernde Glut; es leck ihre Zung, es fresse ihr Zahn den Zagen, der frech sich wagte, dem freislichen Felsen zu nahn! At your demand let fire blaze up; round the rock let it burn with flaring flames; let its tongues flicker, its teeth devour any coward who rashly dares To approach the fearsome rock.
Musically she assumes entire authority, proudly and boldly telling the god what must be done. As her vocal line ends, the orchestra swells with the theme of the Valkyries Ride here majestic, shorn of its previous roughness and then answers with a new motif, triumphant, tender, and transfiguring (Example 2). This is the one and only time Wotan ever sees another as a complete Thou. He finally sees his daughter Self-to-Self, but scant moments later her eyes must close on him forever. Both Wotan and Siegfried open themselves to redemption in these final moments. If the one they both finally see can fulfil her destiny, Wotan and Siegfried can be redeemed, along with love itself. Brnnhilde herself sees both father and husband, complete with their sins and
Please provide better quality image?
Example 2 Sleeping Brnnhilde motif
9780230251687_08_cha07.indd 114
3/22/2012 9:52:32 AM
PROOF
failings, and she forgives them. She takes the gods original sin the sin against love upon herself, and in her death expiates it. In this way, the goddess-made-flesh redeems the divine, and cleanses the gods of their original sin of objectification.
3 Parsifal
What, though, are we to make of Parsifal? At first blush its moral world, its understanding of where salvation lies, seems profoundly at odds with that of Wagners other mature works. In those works, as Scruton suggests, the deepest moral commandment is to view others as Thous, and erotic love is embraced as the best means of accomplishing this. Women and the erotic bond are likewise both exalted, and they play central redemptive roles in the operas. And more profoundly, this insight into another the ability to see another as a wholly free subject belongs to humanity, not to the gods. In this way humanity is set above the divine, and the latter are redeemed only by assuming human form. A natural and widespread reading of Parsifal sees Wagner as abandoning all this in his final opera. Friedrich Nietzsche famously sees Parsifal this way, as Wagners final groveling before the cross. Inflamed by the apparent valorisation of chastity, Nietzsche writes, Parsifal is a work of perfidy, of vindictiveness, of a secret attempt to poison the presuppositions of life a bad work. The preaching of chastity remains an incitement to anti-nature: I despise everyone who does not experience Parsifal as an attempted assassination of basic ethics.12 The moral hierarchy of Parsifal appears to be headed by the Christian God conventionally conceived, to whose perfection humans can only aspire. The human hero Titurel once came close, but now is so aged that he lies in a living grave, his life maintained by the impossibly exalted Holy Grail. However, Titurels son and successor, Amfortas, has sullied himself by having sexual contact with a woman, and so is sunk in filth, sin, and self-loathing. Amfortas and the knights of the Grail are saved only by the pure hero Parsifal, who proves his worth by spurning the sinful temptress that is Woman. It is, of course, a bare biographical possibility that Wagner might have had such an about-face, rejecting his previous tenets and groveling before the cross. It is possible that Wagner, in the years following his composition of the Ring, came to believe that his early intellectual
AQ2
9780230251687_08_cha07.indd 115
3/22/2012 9:52:33 AM
PROOF
hero, Feuerbach, was deeply mistaken: that mortal man could not be his own savior, let alone the savior of the divine. It is possible that Wagner came to see an all-male order of chaste knights idolising disembodied holiness as the highest of ideals, and that he came to regard woman as a desecrating, corrupting influence to be demeaned, subjugated, and, most of all, avoided. It is possible that Wagner came to affirm a world in which Sieglindes deepest fear would be justified: that she defiled Siegmunds purity by sleeping with him, filthy and polluted as she was from Hundings rape. A world in which the sprden Wonnen (frigid joys) of a loveless, sexless gathering of near-dead men is to be prized above all else. A world in which Siegmund would have been wrong to choose Sieglinde over Valhalla. A more promising alternative is that the standard interpretation of Parsifal is misguided, and that a deeper, more revealing analysis will find in it the same moral vision that animates his other mature works. Parsifal, I shall argue, retains the underlying moral structure of the earlier operas: the fundamental commandment is to see others as subjects, not objects, and treat them as ends, not means. While erotic love itself does not play so central a role as it does in the other operas, what gives erotic love its core significance remains central: we are redeemed only when we see and treat others as free and perfect Thous. The original sin in Parsifal is still the sin of objectification, and, as in the Ring, divinity itself is guilty of this sin. The Redeemer, Christ himself, is in sore need of redemption, and this redemption can only come from mortal man. *** Let us linger a moment on the Grail Knights. A sickness fills the order, but it is not a sickness that is solely due to Amfortas it is a deeper and more pervasive sickness. The Grail Knights are not far from Valhallas world, a world of lofty ideals built on crude objectification. The knights idolise the Grail; it fascinates them as the ultimate redeeming means. They demand that it be disclosed, and that the life-giving communion be distributed, so that their strength may be renewed indeed so that they may avoid decline and death. When Amfortas is reluctant to unveil the Grail, the knights clamor around him insisting that he do so. Their demeanor is more like that of addicts desperate for a fix than of holy men tending the sacred for its own sake. There is an unnerving kind of obscenity in their demand for the manna of the Grail. With the notable exception of Gurnemanz, the knights as we first encounter them are idolaters or, more precisely, spiritual materialists. A
9780230251687_08_cha07.indd 116
3/22/2012 9:52:33 AM
PROOF
spiritual materialist is one who sees the promises of religion as primarily a means of fulfilling their ordinary desire for advantage. In Saving God, Mark Johnston characterises the spiritual materialist in these terms: The spiritual materialist is inauthentic in his engagement with religion, and with his spiritual quest of search, precisely because he simply turns his ordinary unredeemed desires toward some supposedly spiritual realm. (p. 16) Spiritual materialism is a way of objectifying the sacred, and thus of desecrating it. This disposition is most clearly exemplified by Titurel. His chilling, vampiric presence in the first act is possibly the most disturbing conceit employed on any operatic stage. Titurel is so desperate to cling to life at all costs that he demands to behold the Grail even though he is too feeble to leave the grave. He is a living corpse, entombed but not yet deceased. He forces Amfortas to reveal the Grail so that he may continue his undead existence despite the agony that the sight of the Grail causes in his son. Titurel is pitiless in his pursuit of endless life. The music of the Grail Knights in the first act has a tense and disquieting feel. Its undertones are sinister, and its apparently otherworldly quality evokes a dark and disturbing realm. The religion itself is poisoned, and one senses that the poison runs far deeper than Amfortass transgression and his fixation on his own impure blood. The poison is the poison of spiritual materialism; the knights have objectified the Grail, and thereby have objectified the Redeemer. It is from this sin that the order must be saved. The relationship of the knights to the Grail dramatises spiritually materialistic religion, including debased Christianity. After all, the endless life granted by Wagners Grail is no more than a literal manifestation of the promise of eternal life inherent in debased understandings of the Christian sacrament of Holy Communion. The latter guarantees eternal life in another world to those who partake of it; the former grants deathlessness in this world. In the Christian faith, our corpses are to be reassembled and reanimated on Judgement Day; here, Titurels corpse is continually animated. But it is fundamentally sacrilegious to engage with the sacrament as a mere means of avoiding death. To do so is ghoulish, disturbing, and ultimately horrific. It is as far removed from the truly salvific as we can imagine. This is spiritual materialism in its essence: the holy is only of interest insofar as it can help us deal with our fears. The sacred is no more than a means to another end.
9780230251687_08_cha07.indd 117
3/22/2012 9:52:33 AM
PROOF
Indeed, it does not seem forced to see in the knights open eroticisation of the Grail a pornographic conceit, this time directed upon the sacred itself.13 But contra Nietzsche, this poisoning of the presuppositions of life is not valorised by Wagner, it is something which Parsifal the innocent one, made wise by compassion has come to redeem and transform. *** The objectification of the Grail is not the only sin of objectification in Parsifal: another sin pervades the opera, and it is for this sin that the greatest atonement is demanded. To see what sin this is, we must understand Wagners attitude in this opera to the erotic. At first glance, it looks as though Nietzsche is right. Wagner who once saw erotic love as a path to salvation now presents it as sinful, as a near-insurmountable obstacle to salvation. But if all erotic encounters in the realm of the Grail are sin, how was Titurel able to father Amfortas? How is Parsifal able to go on to father Lohengrin? A blanket condemnation of all sexual acts does not sit easily with the facts of the opera. But if not all sexual acts are rejected, what then is so shameful about Amfortass dalliance with Kundry? And what is so redemptive about Parsifals abstention? As Wagners other operas keenly reflect, sexual acts are far from being morally equal. The moral gulf between Hundings forcing himself on Sieglinde and her loving union with Siegmund could not be greater. Brnnhildes freely giving herself to Siegfried at the end of the eponymous opera stands in stark contrast to Gunthers unwanted attentions (if one believes that Gunther dared come near her). The same contrast would have arisen in Tristan if Isoldes marriage to King Marke had been consummated. Kundrys relations with Amfortas are no more consensual than the consummations of any of these forced marriages. This is a point on which we must be absolutely clear. She is compelled by Klingsor to sleep with Amfortas against her will, just as he tries to make her sleep with Parsifal. Kundry tries her upmost to resist Klingsor in this, but cannot. Her resistance is manifest throughout the first scene of Act II, for example: Kundry Ich will nicht! Klingsor Jetzt schon erklimmt er [Parsifal] die Burg. Kundry I ... will not! Klingsor He [Parsifal] is already mounting the tower.
9780230251687_08_cha07.indd 118
3/22/2012 9:52:33 AM
PROOF
Kundry Oh! Wehe! Wehe! Erwachte ich darum? Muss ich? Muss? Klingsor Ha! Er ist schn, der Knabe! Kundry Oh! Oh! Wehe mir!
Kundry Oh! Alas! Alas! Did I wake for this? Must I? Must I? Klingsor Ha! The boy is handsome! Kundry Oh! Woe is me!14
Here Kundry is a miserable instrument of Klingsors will. She fights desperately to resist, but Klingsor overpowers her by his dark magical power. (Recall that this is a power he has gained by castrating himself; it is clear from Klingsors musical signatures that Wagner regards this as a perversion, even though it leaves Klingsor technically chaste.) Kundry must do as Klingsor commands, and his command is that she prostitute herself to serve his ends. One of the most disturbing aspects of the scene is Klingsors repeated insinuation that Kundry must enjoy these acts of rape-by-proxy for example, Er ist schn, as quoted above. These comments are disturbing because they are psychologically realistic depictions of the mentality of rape: deep down, the victim must really want it.15 Nor was Amfortass sexual encounter with Kundry a meeting of free subjects. Amfortas was just another in what we may infer was a string of men to whom Klingsor has forcibly pimped Kundry. It was not the free giving of the embodied self to another. They did not meet I-to-I, Thou-toThou. There was no erotic love in their encounter the mere possibility of that was destroyed at the outset by Klingsor, since Kundry does not come to these meetings as a free subject, but rather as a compelled object. Their entanglement was an obscenity forced, impersonal, alienated. This was the nature of the act that led to Amfortass downfall. Erotic love is not maligned in Parsifal; it simply does not figure. To suppose that it is frowned upon in these operas is to confuse a union of free subjects with the obscenity of forced prostitution. This is as gross an error as confounding Hundings rape with Siegmunds love. Klingsor is the primary transgressor here to be sure, but Amfortas bears the guilt of participating in such an act. He engaged in sex as a union of objects, not subjects obscene, pornographic sex. He did not could not see Kundry as a Thou, since she came to him as a wholly determined object. Just as the john degrades himself with his actions, so has Amfortas.
9780230251687_08_cha07.indd 119
3/22/2012 9:52:33 AM
PROOF
There is a profound sense in which Amfortas could not have slept with Kundry under those circumstances had he been antecedently inclined to treat her as a Thou. Here it is instructive to consider which counterfactual emendations to the plot would be coherent, and which would not. In Act I, Kundry is widely objectified, particularly by the young Grail knights, who liken her to a beast. (Even Parsifal physically attacks her in this early act.) Only one person resists objectifying her: Gurnemanz. He protects her from the young knights, and rebukes them for addressing her in such demeaning terms. Now: we can readily imagine the young knights succumbing to her compelled seductions, but can we imagine Gurnemanz doing the same? This seems unthinkable: Gurnemanz would not be susceptible. His resistance to objectifying her would bar him from participating in such an act. (Similarly, if we do not mind crossing the boundaries between operas, is it not clear that Siegmund would be immune to such advances, even before he ever laid eyes on Sieglinde?) Amfortas, had he not been susceptible to the original sin of objectification, would never have fallen into Klingsors trap. This is Amfortass transgression the one for which he must atone. It is not sex that is sin, but rather objectification. This is the deep stain that sullies Amfortas. However and this will lead to the most surprising point in the opera the story of this sin does not end here, but rather reaches all the way to the One who first condemned Kundry to this existence, to the One who first punished her laughter with His curse. Kundrys forced prostitution is depicted as a direct consequence of her curse. For example, in the opening scene of Act II, she and Klingsor have the following exchange: Klingsor Gefiel er dir wohl, Amfortas, der Held, den ich dir zur Wonne dir gesellt? Kundry Oh! Jammer! Jammer! Schwach auch er! Schwach alle! ... Meinem Fluche mit mir alle verfallen! Klingsor How did you like the hero Amfortas whom I ensnared to your charms? Kundry O anguish! Anguish! He too was weak! Weak are they all! All fall victim with me to my curse!
The curse is here presented as the root of the unfortunate episode with Amfortas: it is the curse that lies at the root of their degrading encounter.
9780230251687_08_cha07.indd 120
3/22/2012 9:52:33 AM
PROOF
On one level of understanding, the episode is due to Klingsors plotting, but there is still a deeper level at which Amfortas must bear the blame, since the episode could not have happened if he had not been antecedently inclined to objectify Kundry. Now a third level of understanding emerges: the whole miserable episode was caused by the curse. Kundrys later remarks suggest a similar interpretation: Kenntest du den Fluch, der mich durch Schlaf und Wachen, durch Tod und Leben, Pein und Lachenpa zu neuem Leiden neu gesthlt, endlos durch das Dasein qult! ... da kehrt mir das verfluchte Lachen wieder, If you knew the curse which afflicts me, asleep and awake, in death and life, in and laughter, newly steeled to new affliction, endlessly through this existence! ... The accursed laughter assails me once again: a sinner sinks into my arms! Then I laugh laugh I cannot weep, can only shout, rage, storm, rave in an ever-renewed nightmare from which, though repentant, I scarcely wake. (Emphasis added)
ein Snder sinkt mir in die Arme! Da lach ich lache, kann nicht weinen: nur schreien, wten, toben, rasen in stets erneuter Wahnsinns Nacht, aus der ich bssend kaum erwacht.
These acts of loveless, objectifying sex are part of Kundrys curse. One might distill the curse to its essence: her curse is to be endlessly treated as an object, never as a subject. The curse has another aspect to it: it prevents Kundry from dying, condemning her instead to an endless string of reincarnations. The themes of death, redemption, and subjectification are so interwoven in Wagner that they may be seen as aspects of the same phenomenon. Tristan and Isolde are led to sacrifice themselves for the sake of their I-to-I erotic love, and the same may be said of Brnnhilde. The converse plight
9780230251687_08_cha07.indd 121
3/22/2012 9:52:33 AM
PROOF
is that of the Flying Dutchman and I would suggest Kundry. The Flying Dutchman is condemned to eternal life, with only one hope of finding release: he must find a loving wife who remains faithful until her death. Only a woman who engages with him as a free and perfect subject will be capable of such a redemptive act. To the extent that a woman falls short of loving him for the unique and irreplaceable embodied Thou that he is, she will be vulnerable to the crime of substitution that is, infidelity. Kundrys curse echoes that of the Flying Dutchman. She too cannot die until she is seen and treated as an embodied subject. The doubleedged nature of her curse both guarantees she will be objectified by those she meets, and also binds to endless life as a consequence. Kundrys own original sin was to mock Christ on the cross. A terrible thing, to be sure; we should not be misled into thinking of Kundry as an angel. But what of her punishment, and what of the one who inflicted it? To condemn a person to endless objectification this in itself is surely a profound sin, in itself a failure to treat that person as a subject. This is the Redeemers sin. It is for this that the Redeemer Himself must be redeemed, hence the last line of Parsifal, Erlosung dem Erloser (the Redeemer redeemed), which tamer interpreters of the opera have long found so paradoxical. We might think of Kundrys curse as symbolic of the original religious separation of men and women the original conception of woman as sinful, and so to be avoided and demeaned the root of the idea that only an all-male group of chaste knights could be worthy of the Grail. From the first Fall, woman is condemned to be objectified, and so man must shrink from her if he is to avoid committing the sin of objectification. If Tristan and the Ring show how erotic love can overcome this original rift, Parsifal confronts the rift at its source. Before this was mans sin, it was Gods sin. Kundry, in an otherwise obscure moment, subtly equates the Redeemer with Klingsor. Parsifal asks, Wer durft ihn [Amfortas] verwunden mit der heilgen Wehr? (Who dared to wound him [Amfortas] with the holy weapon?), to which the factually correct answer is indisputably Klingsor. However, the music recedes and darkens at the moment, and Kundry quietly replies: Er Er der einst mein Lachen bestraft (He He who once punished my laughter). This description can only apply to the Redeemer, to Christ on the cross no one else punished Kundrys laughter. The sudden solemnity in the orchestra and in the vocal line further indicate that this is no casual assertion.16 Her remark is not to be interpreted literally,
9780230251687_08_cha07.indd 122
3/22/2012 9:52:33 AM
PROOF
but at a depth-psychological level it leads us to compare Klingsor and the Redeemer. Both are sinners, both are objectifiers; both have sinned against Kundry, both have transgressed against the feminine. *** Amfortass sin recapitulates the Redeemers sin. It is the symbol of this deep, primordial error. Amfortass wound which so perfectly parallels the Redeemers own wound cannot be healed until this ill is addressed. On the exoteric level, it cannot be healed until Parsifal retrieves the spear; on a more esoteric level, it cannot be healed until the wrong done to Kundry is redressed. It cannot be healed until her curse is broken until she is treated as a Thou. Parsifal accomplishes this: the curse is undone when he refuses to bow to its power. Of course, he does not have an easy time doing so. Towards the end of Act II, the curse manifests itself in a particularly brutal way: Kundry objectifies herself, begging Parsifal to do the same. This is a particularly harrowing scene to watch, but Parsifal enlightened through compassion resists and rebukes her thus: Die Labung, die dein Leiden endet, beut nicht der Quell, aus dem es fliesst, das Heil wird nimmer dir gespendet, eh jener Quell sich dir nicht schliesst. The solace to end your sorrows comes not from the source from which they flow: grace shall never be bestowed on you until that source is sealed to you.
In this scene Parsifal is able to do what even Kundry herself cannot do. He sees her and treats her as Thou, and understands that this is what is needed for the salvation for all involved. He indicates this to her in the final line of Act II: Du weisst, wo du mich wiederfinden kannst! You know Where you can find me again!
When she does indeed find him again in Act III, Parsifal baptises her an act symbolic of the Redeemer himself acknowledging her as a subject after untold centuries. ***
9780230251687_08_cha07.indd 123
3/22/2012 9:52:33 AM
PROOF
If the fundamental sin in Parsifal is the sin of objectification then the operas underlying moral structure is continuous with the moral worlds of Tristan and the Ring. Scruton describes this moral sphere beautifully in Death-Devoted Heart, and uses it to lay out a deeply satisfying analysis of Tristan perhaps one of the most satisfying interpretations on offer. The Ring, as this essay argues, is naturally interpreted against this same moral backdrop. Parsifal is at first blush more recalcitrant, yet this is only a superficial seeming. Scrutons original insight into Tristan extends elegantly to cover the totality of Wagners mature tragedies. The deepest commandment in these operas is to see others as Thous and treat them as ends in themselves. As in the Ring, in Parsifal divinity is guilty of transgressions against subjecthood, and can only be redeemed by a human being. The capacity of mortal man to see others I-to-I outstrips that of the gods. Scrutons vision of Wagners moral landscape can encompass all of Wagners mature tragedies, including Parsifal. By refusing to objectify Kundry, Parsifal breaks the cycle of sin, and by baptising her, he redeems the one who originally condemned her. This is how the Redeemer is redeemed. At the superficial level of dramatic action, there is no reason why Parsifal must treat Kundry in any particular way in Act III; he already has the Spear, and that alone at his level of understanding should suffice to redeem Amfortas and the knights. Yet at a more profound level, we sense that Parsifal could not proceed to redeem the order without treating Kundry as he does. If he simply did not encounter her in Act III, or if in doing so he mistreated her, or even simply failed to release her from her torment, we sense that redemption of the order would not be possible. This is because the order must be redeemed from something deeper than Amfortass transgression: it must be saved from the primordial sin of the Saviour, and from the knights woman-reviling collaboration in the Saviours act of objectifying Kundry. The sin of objectification manifests itself throughout the order, infecting even the knights relationship with the Grail itself. They no longer treat it as a sacred item to be revered for its own sake, but rather view it primarily as a means of boosting their vitality and avoiding death. Amfortas has denied them the Grail in the intervening years between the events of Act I and those of Act III, though, and the symbolic head of this objectification of the Grail, Titurel, has died at long last. His ghoulish, undead existence has come to an end, which is perhaps a sign of hope for a new relationship between the knights and the Grail. At the beginning of the final scene of the opera, however, the order is still unredeemed, and this manifests itself in the knights
9780230251687_08_cha07.indd 124
3/22/2012 9:52:33 AM
PROOF
aggressive behavior towards Amfortas they circle him, demanding to see the Grail, and the threat of violence sits heavy in the air. They crave their vitality boost; they want to ensure death is kept at bay. When Parsifal presides over the operas final act of Communion, the Grail music is subtly changed; it no longer has the tense, disturbing, ghoulish quality it did before. Its feel is now more genuinely hopeful, raising the possibility that the knights relationship to the Grail will no longer be as it once was. A further sign of a new future is that, for the first time, someone who beholds the Grail nonetheless dies: Kundry. She slowly sinks lifeless to the ground while Parsifal holds the Grail aloft. While there are many ways to view this event, it raises the possibility that the Grail can no longer be used to stave off death that was something that occurred only in the orders earlier, unredeemed state. The knights will be forced to alter their relationship with the Grail; they will no longer be able to treat it as a means of quieting their fear of death. The more esoteric meaning of this presents a vision of how mans relationship to the divine must evolve if the divine is to be truly salvific. Man cannot simply transfer his earthly desires to the heavenly realm if he is to achieve true salvation. Spiritual materialism must be rejected, and the divine must be approached on its own terms, as a whole and perfect end. Kundrys presence in this final scene also signifies other changes in the order. For the first time, a woman is present when the Grail is disclosed. She enters their inner sanctum and beholds the Grail they are charged to tend. Man and woman are no longer kept artificially apart from each other. The curse on the feminine is now broken man is no longer bound to objectify her, and so man and woman can intermingle without sin. The original rift is healed; the sexes are united in their redemption.
Notes
1. I am extremely grateful to Andy Hamilton, Andrew Huddleston, Mark Johnston, Alexander Nehamas, Jan Perkins, and Nick Zangwill for their helpful comments. The greatest debt of gratitude is, of course, owed to Roger Scruton, whose brilliant work on Wagner forms the foundations of this essay. 2. That Brunnhilde rather than Siegfried is the Ring s central redemptive figure seems to me quite obvious, and it is an interesting question why so many commentators have missed this. Warren Darcy to pick just one example describes Siegfried as the gods redeemer in his The Metaphysics of Annihilation: Wagner, Schopenhauer, and the Ending of the Ring,
9780230251687_08_cha07.indd 125
3/22/2012 9:52:33 AM
PROOF
126 Sarah-Jane Leslie Music Theory Spectrum 16(1), 1994, pp. 140. Scruton himself seems to adhere to a similar view in his Lectures on Wagners Ring (Council of Humanities Lecture Series, Princeton University). It is not that these commentators offer arguments that Siegfried rather than Brunnhilde must be the central salvific figure; they simply assume without argument that it must be so. (That Wagner himself may have been under a similar illusion is no great defense the operas speak for themselves.) One notable exception in this regard is J. Wentzel van Huyssteen (Building Effective Bridges to Culture: God and Redemption in the Work of Richard Wagner, in P. Middleton (ed.), The God of Love and Human Dignity: Essays in Honor of George M. Newlands (London: T&T Clark/Continuum, 2007). Unless otherwise noted, all English translations of the Ring are by Lionel Salter and William Mann. For an exemplary analysis of the character of Wotan, see work by Philip Kitcher and Richard Schacht, Finding an Ending: Reflections on Wagners Ring (New York: Oxford University Press, 2004). Einen Freien kann ich nicht wollen: fr Frickas Knechte kmpfe nun du! (Act II, Scene II). Translation by Deryck Cooke, I Saw the World End (Oxford: Oxford University Press, 1979). Wagner wrote that their meaning is already expressed with the greatest precision in the effect of the musically sounding drama. For further discussion, see Warren Darcys The Pessimism of the Ring, Opera Quarterly 4(2), 1986, pp. 2448. (The translation is Darcys.) The Pessimism of the Ring, Opera Quarterly, p. 47, fn. 41. Und doch, trg ich nicht Gutrun Treu, der zieren Frauen eine htt ich mir frisch gezhmt! (Gotterdammerung, Act III, Scene I). Death-Devoted Heart, p. 145. Gotterdammerung, Act II, Scene IV. Nietzsche Contra Wagner. As is made dramatically plain in Hans-Jurgen Syberbergs film Parsifal, and in the recent Baden Baden production, under the stage direction of Nikolaus Lehnhoff. All English translations of the Parsifal libretto here provided are due to Lionel Salter. Scruton describes such forced prostitution in Sexual Desire: The fantasy [described in a passage from Histoire dO] is that of a supremely achieved prostitution, in which the woman does not have even the liberty of refusal in which the market economy has been replaced by an economy of command. In this achieved prostitution, the womans spirit is wholly overcome by the force of masculine autocracy. At the same time, however, the command destroys the basis of personal relation, and compels her to attach her interests and her joys to the abstract penis, irrespective of who owns it or why he seeks her submission to it. This is the root fantasy of obscenity, and describes the content of the feat which underlies the paroxysm of shame. As it implies, its enactment involves the abrogation of all individualised feeling in the woman; her enslavement is merely, to use the Marxian language, the realised and objective form of this subjective alienation. (p. 159)
3. 4.
5. 6. 7.
8. 9. 10. 11. 12. 13.
14. 15.
9780230251687_08_cha07.indd 126
3/22/2012 9:52:33 AM
PROOF
Eros and the Redemption of the Gods 127 16. In addition to the musical and dramatic clues, the capitalisation pattern in the libretto clearly shows that Kundry is here referring to the Redeemer. The full quote is: Er Er He He der einst mein Lachen bestraft: who once punished my laughter: Sein Fluch ha, mir gibt er Kraft; His curse ha! gives me strength; Both occurrences of Er are capitalised, as is Sein in the third line. (Other words occurring in comparable grammatical positions throughout the libretto are not capitalised.) The capitals are often not preserved in the English translations, but this, I think, obscures an important point.
9780230251687_08_cha07.indd 127
3/22/2012 9:52:34 AM

Philosophy &amp; Psychology Writings of Sarah-Jane Leslie

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Philosophy &amp; Psychology Writings of Sarah-Jane Leslie

Caricato da

Copyright:

Formati disponibili

Journal of Memory and Language 65 (2011) 1531

Contents lists available at ScienceDirect

Journal of Memory and Language

Do all ducks lay eggs? The generic overgeneralization effect

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

Lions have manes

Majority Striking False generalization

True True False

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

Quasi-definitional Majority characteristic Minority characteristic Majority Striking False generalization

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

S.-J. Leslie et al. / Journal of Memory and Language 65 (2011) 1531

Preparation of this article was supported by NSF grant BCS-1226942.

!"#$%&%'%()*(+*,&&,)'%!-%&'*.,-%,+&*'/!'*0($-1*/!2,*+(34,1*!)50!56*3!'/,3*'/!)*0(37%)8*'(*%)&'%--*&,-,"'%2,* ,&&,)'%!-%&'*.,-%,+&*%)*'/,*+%3&'*9-!",:** * ;(*',&'*0/,'/,3*8,),3%"*-!)8$!8,*4%8/'*/!2,*1%3,"'*&,-,"'%2,*,++,"'&*%)*'/,*"!&,*(+*&("%!-*83($9&6*

!"#$%&'($)&&)*#+%,"-!%)&.%)!%$-/%0#)12304#%#&&#1!)-5)&!%(#5)#2&%-(3'!%63$#1.%-17%2'0!"#0%43$$'1)4-!#% !"-!%!")&%&"-0#7%#&&#14#%4-'&#&%63$#1%!3%(#%&'($)&&)*#+%%% % ,")&%)&%13!.%32%43'0&#.%!3%&-/%!"-!%6#%!"#$%!$&'()*+,-))1!#080#!%963$#1%-0#%&'($)&&)*#:%)1%!")&%

!"#$!%&'&(&)*$(+$*+,&!'$-.+/0*$1!2$')!#$,3&'#.)"$(+$!#+0($!$1+.)$)"(&(2$(3)+.)(&,$0).*0),(&4)$+"$(3)$ .)')4!"($!%&'&(&)*5$ $ 6"$)4)"$1+.)$#.!1!(&,$&''/*(.!(&+"$+7$(3)$0+8).$+7$-)").&,$'!"-/!-)$,+",)."&"-$!%&'&(&)*$&*$

1 Running head: GENERICS AND CONCEPTS

GENERICS AND CONCEPTS

GENERICS AND CONCEPTS

GENERICS AND CONCEPTS

Representing generic knowledge via principled and statistical connections

GENERICS AND CONCEPTS

GENERICS AND CONCEPTS

GENERICS AND CONCEPTS

GENERICS AND CONCEPTS

GENERICS AND CONCEPTS

GENERICS AND CONCEPTS 10

GENERICS AND CONCEPTS 11

GENERICS AND CONCEPTS 12

GENERICS AND CONCEPTS 13

GENERICS AND CONCEPTS 14

GENERICS AND CONCEPTS 15

GENERICS AND CONCEPTS 16

GENERICS AND CONCEPTS 17

GENERICS AND CONCEPTS 18

GENERICS AND CONCEPTS 19

GENERICS AND CONCEPTS 20

GENERICS AND CONCEPTS 21

GENERICS AND CONCEPTS 22

GENERICS AND CONCEPTS 23

GENERICS AND CONCEPTS 24

GENERICS AND CONCEPTS 25

GENERICS AND CONCEPTS 26

GENERICS AND CONCEPTS 27

GENERICS AND CONCEPTS 28

GENERICS AND CONCEPTS 29

GENERICS AND CONCEPTS 30

GENERICS AND CONCEPTS 31

GENERICS AND CONCEPTS 32

Experiment 3b: What mediates causal connections of striking generics?

GENERICS AND CONCEPTS 33

Procedure. Same as Experiment 3a.

GENERICS AND CONCEPTS 34

Philosophy & Psychology Writings of Sarah-Jane Leslie

Philosophy & Psychology Writings of Sarah-Jane Leslie

!"#$%&%'%()(+,&&,)'%!-%&'.,-%,+&'/!'0($-1/!2,+(34,1!)50!563!'/,3'/!)0(37%)8'(%)&'%--&,-,"'%2,* ,&&,)'%!-%&'.,-%,+&%)'/,+%3&'*9-!",:** * ;(',&'0/,'/,38,),3%"-!)8$!8,4%8/'/!2,1%3,"'&,-,"'%2,,++,"'&%)'/,"!&,(+&("%!-83($9&6

!"#$%&'($)&&)#+%,"-!%)&.%)!%$-/%0#)12304#%#&&#1!)-5)&!%(#5)#2&%-(3'!%63$#1.%-17%2'0!"#0%43$$'1)4-!#% !"-!%!")&%&"-0#7%#&&#14#%4-'&#&%63$#1%!3%(#%&'($)&&)#+%%% % ,")&%)&%13!.%32%43'0&#.%!3%&-/%!"-!%6#%!"#$%!$&'()+,-))1!#080#!%963$#1%-0#%&'($)&&)#:%)1%!")&%

!"#$!%&'&(&)$(+$+,&!'$-.+/0$1!2$')!#$,3&'#.)"$(+$!#+0($!$1+.)$)"(&(2$(3)+.)(&,$0).0),(&4)$+"$(3)$ .)')4!"($!%&'&(&)5$ $ 6"$)4)"$1+.)$#.!1!(&,$&''/(.!(&+"$+7$(3)$0+8).$+7$-)").&,$'!"-/!-)$,+",)."&"-$!%&'&(&)$&$