658

Science in China Ser. C Life Sciences 2005 Vol.48 No.6 658 668

The diversity effect of inductive reasoning under segment manipulation of complex cognition
CHEN Antao, LI Hong, FENG Tingyong, GAO Xuemei, ZHANG Zhongming, LI Fuhong & YANG Dong
School of Psychology, Southwest University, Key Laboratory of Basic Psychology of Chongqing City, Chongqing 400715, China Correspondence should be addressed to Li Hong (email: lihong1@swnu.edu.cn)

Received November 9, 2004; accepted May 12, 2005

Abstract The present study proposed the idea of segment manipulation of complex cognition (SMCC), and technically made it possible the quantitative treatment and systematical manipulation on the premise diversity. The segment manipulation of complex cognition divides the previous inductive strengths judgment task into three distinct steps, attempting to particularly distinguish the psychological processes and their rules. The results in Experiment 1 showed that compared with the traditional method, the quantitative treatment and systematical manipulation of SMCC on the diversity did not change the tasks nature, and remain rational and a good measurement of inductive strength judgment. The results in Experiment 2 showed that the participants response rules in the triple-step task were expected from our proposal, and that in Step 2 the feeling of surprise (FOS), which seems implausible but predicted from the diversity premises, was measured, and its component might be the critical part that produced the diversity effect. The feeling of surprise may reflect the impact of emotion on cognition, representing a strong revision to premise probability principle of pure rational hypothesis proposed by Lo et al., and its roles in the diversity effect are worthy of further research. In this regards were discussed the mistakes that the premise probability principle makes when it takes posterity probability as prior probability.
Keywords: inductive reasoning, segment manipulation of complex cognition (SMCC), diversity effect, feeling of surprise (FOS), Bayes theorem. DOI: 10.1360/062004-92

Forming a general reasonable conclusion from a limited number of examples is one of the most marvelous psychological activities of human beings. This kind of activities is inductive reasoning which is a key issue rising from the cross fields of psychology, philosophy, cognitive science, and so on. It is a thinking process of generalizing from concrete cases. The diversity effect of inductive reasoning refers to the phenomenon that the inductive argument of more diverse premise can obtain stronger mental strength than that
Copyright by Science in China Press 2005

of less diverse one. The diversity effect is regarded as the core process of inductive reasoning[1]. Carey firstly explored the diversity effect. She used concrete items (e.g. dogs and bees) to construct the premises[2]. The task was to judge the possibility that all animals had specific property when such different animals as dogs and bees had that property in common. The result was that adults showed diversity effects, while 6-year-old children did not, which implied that diversity effect was learned. Carey also demonstrated

Diversity effect of inductive reasoning under segment manipulation of complex cognition

659

the diversity effects existed among super-ordinate concepts (e.g. mammals and reptiles). More recently, more research has been done on the diversity effect, but the results are full of disputes. Osherson et al.[3] found that adults showed a diversity effect in the arguments of general conclusions as well as in those of specific ones. Heit and Hahn[4] even found this effect on 5-year-olds. However, Lpez et al.[5] found that 9-year-old children showed a diversity effect in the arguments of general conclusions, while they did not show this effect in the arguments of specific conclusions. Therefore they suggested that the evaluation of specific arguments needs more cognitive processes than that of general arguments. The evaluation involves many comparisons among items, which requires more resources, while children may lack sufficient resources to accomplish that. Choi et al.[6] investigated the influence of culture on the diversity effect. Using animals and human beings as premises, they discovered the diversity effect in Korean undergraduates. Nevertheless, Lpez et al.[7] found no diversity effects in the research of Guatemalan Itzaj adults, using various biological concepts familiar to local people. Under various experimental conditions, the participants insisted that the arguments with the homogeneous premises were more likely to be true (typicality effect). The Concept Boundary Effect found by Gelman and Lpez et al. also suggests that the diversity effect may not be true, and they found that the influence of items category boundary on the inductive strength exceeded that of item diversity[4,8]. The experimental tasks of the diversity effect are usually administrated in this way: natural categories (e.g. cows) are used as the premise subjects, and blank properties are used as the premise predicates (e.g. property P/X); the level of conclusion subjects (e.g. mammals) provided is higher than that of premise subjects; the conclusion predicates provided are the same as premise predicates. Two arguments are then provided for comparison, and the participants are required to judge which argument is more likely to be true. The following schema illustrates a typical inductive reasoning task.

[1]

Caws have property P. Horses have property P. All mammals have property P.

[2]

Caws have property P. Hedgehogs have property P. All mammals have property P. (1)

This kind of tasks has the following features: (i) there are at least two premises for each argument, and there must be a difference between the two premise subjects. For example, in argument [1], cows and horses are different, and in argument [2], and cows and hedgehogs are also different. Because this kind of difference exists in one argument, it could be called difference within argument (DWA). (ii) Meanwhile, the DWAs between two arguments must be different. For example, the difference between cows and horses is smaller than that between cows and hedgehogs. The difference between two DWAs could be called difference between arguments (DBA), which is the result of subtraction between two DWAs. The advantage of this kind of tasks is that they are close to the real problem situation. However, it has some limitations. (i) As for the stimuli material, most previous studies use biological categories at different concept levels as the premise subjects. These materials cannot avoid the influence of participants knowledge experience on inductive reasoning. Therefore, participants may deductively evaluate this kind of tasks simply on the basis of their knowledge experience. Related developmental and cross-cultural studies both demonstrate this point. (ii) The traditional researches use natural categories as materials. As a result, it is hard to change the diversity of premises systematically, and it is difficult to quantify the diversity. Therefore, it is impossible to analyze the influences of diversity changes on the inductive reasoning systematically and quantitatively. (iii) In respect of the paradigm, most traditional researches use the whole presentation-comparison paradigm: two complete arguments are presented

660

Science in China Ser. C Life Sciences

simultaneously; the participants are asked to compare their strengths directly. In this case, only one evaluation response can be obtained. However, because inductive reasoning is a complex activity involving several sub-processes[9], it is impossible to analyze the operation laws and interactions of these sub-processes simply with the holistic strength evaluation. Furthermore, these laws and interactions are the basis to understand the psychological mechanisms of inductive reasoning, so the traditional paradigm is not beneficial to the discussion of mechanisms of inductive reasoning. Based on the above analysis, this research attempts to reform in the followings: (i) new artificial novel animal drawings were used as stimuli to avoid the influence of participants knowledge and experience; (ii) quantitative and systematical treatments of the diversity were fulfilled by manipulating local features of these drawings; (iii) participants response on each of these sub-processes of inductive reasoning was recorded through applying segmental presentation-judgment paradigm. 1 Experiment 1

Put differently, the research material was new, and the diversity of premises was quantified in order to test the rationality of the treatment. 1.1 Participants

Seventy undergraduates of Southwest China Normal University took part in the experiment, with the mean age of 20 and SD of 0.9. They ranged from 19.3 to 21.1 years old. Male and female students constituted each half of the participants, respectively. All participants were right-handed, with normal or corrected-to-normal vision. 1.2 Method

The present experiment was conducted by the traditional method where an entire set of question context was presented all at once. The participants were then required to judge the strength of the arguments.

To avoid the limitations of traditional researches mentioned above, we used artifactitious animals drawing and gave them novel names that were used as the premise subjects of inductive argument in order to avoid the influence of knowledge and experience of participants. In addition, we intentionally systematically changed some local features of the animal drawings in order to obtain various premise subjects. Quantification and systematical change of the premise diversity were achieved because the change of difference among premise subjects represents the change of diversity. The artifactitious drawings in Figs. 1 3 were

Fig. 1. Illustration when DBA is 9 d. The changes of Group M in palpus, face, and tail; the changes of Group N in palpus, face, and wing.

Diversity effect of inductive reasoning under segment manipulation of complex cognition

661

Fig. 2.

Illustration when DBA is 5 d.

Fig. 3.

Illustration when DBA is 1 d.

designed according to the above idea (Group M was revised from Yamauchi and Markman[10]). M1, M2, M3, N1, N2, and N3 were simply substitute names for animals. Group M was the sample of insects with eight legs, and group N was the sample of insects with wings. On the whole, these two groups of insects looked obviously different. The experiment instructions told the participants that these objects were certain animals, but did not give any clue about what kind of animals they were. Because there were three premise subjects in each argument, three animals figures were designed in a picture, namely, each picture represents one argument.

And the differences in appearance among three figures that represented three subjects could be changed when needed, thus the combination of different DWA was established. Group M should change in palpus, face, and tail dimensions, and Group N should change in palpus, face, and wing dimensions, and each dimension had three levels (see Fig. 1). Theoretically, 27(333) different combinations of animal figure representing 27 DWAs should be distributed either in Group M or in Group N. 27 DBAs could be obtained corresponding to 27 DBAs of group M and Group N, thus the comparative judgments of 27 DBAS could be obtained.

662

Science in China Ser. C Life Sciences

The DWA of a picture depended on the following principles: if the three figures were different from each other on a certain dimension, the DWA of the picture on this dimension was 3(1 d+1 d+1 d, the d represents the unit of difference between premises. Howerer, the d will be omitted in the following text); if two of the three figures were different on a certain dimension, the DWA of the picture on this dimension was 2(0+1+1); if the three figures was identical on a certain dimension, the DWA of the picture on this dimension was 0(0+0+0). Because the change was in three dimensions, the maximum DWA of a picture was 9(3+3+3), and the minimum DWA was 0(0+0+0). For example, the three figures in Group M of Fig. 1 were different with Group N in three dimensions, their DWAs were all 9. Because DWAs ranged from 0 to 9, the DBAs also ranged from 0 to 9, whereby the premises were quantified and systematically changed. In the present study, three of the ten possible DBA values from 0 to 9, that is, 1(6 5), 5(7 2) and 9(9 0), were chosen. The number out of the parentheses is DBA, and the number in the parentheses is DWA. 1.3 Materials

with the following arguments: Animal M1, M2, and M3 have GAMA in their bodies/All animals have GAMA in their bodies. Animal N1, N2, and N3 have GAMA in their bodies/All animals have GAMA in their bodies. Participants were asked to judge which argument was more likely to be true. Recorded were the participants choices with 0 or 1, that is, if a participant made a judgment in favor of diversity argument, it was recorded as 1, but 0 otherwise. 1.5 Results and discussion

The results of 2 test on the 70 participants choices are shown in Table 1.

Table 1 The effect of quantitative treatment on the inductive strength under mono-step condition 1d 0.057 0.811 5d 1.429 0.232 9d 8.229 0.004**
2

DBA p ** p<0.01.

The experimental stimuli are hand-drawn sketches against white backgrounds (see Fig. 1), with the picture size of 5 cm7 cm. Each picture contained 3 animal figures arrayed like a triangle similar to Chinese character , representing a premise constituents of an inductive argument. In each trial, two pictures were presented simultaneously, which represented two comparative arguments. The DWA of each picture and DBA of each pair of pictures was controlled. According to the quantitative method mentioned above, 3 DBAs were designed in the study, expressed by 6 DWAs respectively, that is, 1(6 5), 5(7 2), and 9(9 0). Each combination of a pair of arguments was measured twice, with once the picture of Group M on the right and the other time on the left. A PC, with a 14 liquid crystal display which was 57 cm away from participant eyes, presented stimuli and recorded the responses of participants. 1.4 Procedure

We can see from the results of Table 1, when the DBA was up to 9, participants showed very obvious orientation; while when the DBA was 5 or 1, the participants orientation was not obvious. It is obvious that the change of DBA can affect the emergence of diversity effect, which indicates that DBA is the accurate portrait of diversity effect. When the DBA was very small (e.g. 1), there was hardly any orientation in the participants responses; with the increase of DBA, the participants responses began to show orientation: when the DBA was 5, although it was not statistically significant, the participants responses more apparently tended to be in favor of the arguments with larger DBA; when the DBA is 9, the participants very significantly reported that the more the DBA was, the more stronger the argument was. The variation tendency suggests that the control of premise DBA can affect the diversity effect of inductive reasoning; therefore, it is acceptable to use DBA to express the diversity effect of premise. More important, the results also indicate that the diversity effect of premise can be quantified. Based on that, the systematical change of diversity can be realized, which provides a new approach to the further study of

The figure stimuli were presented to the participants,

Diversity effect of inductive reasoning under segment manipulation of complex cognition

663

the diversity effect of inductive reasoning. 2 Experiment 2

Based on the quantitative treatment and systematical change of the diversity effect, the triple-step task was designed in the experiment according to the segmentation manipulation of complex cognition (SMCC), aiming at exploring the mental mechanisms of the diversity effect of inductive reasoning. 2.1 Participants

mate; (iii) when this property is expanded to a larger extent, this kind of FOS can affect the participants judgment based on the idea that the argument with a larger DWA is stronger. Based on the hypothesis, the judgment of the inductive strength was divided into three sub-tasks in the present study: (i) Participants were asked to make an initial estimate of the possibility that three premises of a certain argument were true at the same time. Consequently, the participants initial possibility judgment of each premise combination was obtained. (ii) The participants were explicitly told that all the three premises were true (participants estimates were usually different from it, consequently they were made surprised.), and asked to choose the most surprising argument. (iii) The participants were asked to judge which of the two conclusions was more possible than the other. The three sub-tasks were administrated to the participants one after another: Participants could not see the next task until they finished a specific task. And the answer to the current task would constitute the basis for solving the next one. We call this method the SMCC, which is illustrated in Table 2. 2.3 Stimuli

70 undergraduates of Southwest China Normal University took part in the experiment, with the mean age of 20 and SD of 0.8. They ranged from 19.0 to 21.4 years old, with male and female constituting each half of the participants respectively. All participants were right-handed, had normal or corrected-to-normal vision. 2.2 Method

According to the analysis of relative research reports and theoretical models [11,12], we proposed that the diversity effect involves the following mental processes: (i) Individual makes an initial estimate of the possibilities of each premise proposition on the basis of his or her knowledge and experience; (ii) when knowing unexpectedly that several premises are true simultaneously, the participant may have a feeling of surprise (FOS) when he or she notices that the new information acquired is different from his initial esti
Table 2 Argument 1

The stimuli in this experiment were the same as that of Experiment 1. But the display procedure was different. The stimuli were presented in three steps instead of one (see Table 2).

An example of SMCC Argument 2

Step 1

Tell participants: Proposition1: Animals M1 have GAMA in their bodies. Proposition1: Animals N1 have GAMA in their bodies. Ask participants to infer: Animals M2 and M3 both have Ask participants to infer: Animals M2 and M3 both have GAMA in their bodies. GAMA in their bodies. Ask participants to compare: As for the two arguments inferred from M1 and M2 respectively, which is more likely to be true? Make statement: It is proved by researches that, the above arguments are both true. Thus Fact 1: Actually, animals M2 and M3 both have GAMA in Fact 1: Actually, animals N2 and N3 both have GAMA in their their bodies. bodies. Ask the participants, Which of the two facts above makes you feel more surprised or incredible? Assure the participants that the facts are proved by scientific researches however surprised they are. Ask the participants to make one more judgment: Argument 1: M1, M2, and M3 have GAMA in their bodies; All Argument 1: M1, M2, and M3 have GAMA in their bodies; animals have GAMA in their bodies. All animals have GAMA in their bodies. Ask the participants to judge which of the above arguments is more likely.

Step 2

Step 3

664

Science in China Ser. C Life Sciences

2.4

Procedure

In the first step, only premise pictures and the corresponding properties were presented (see Table 2). The question was how it was possible that all the three animals had a property. The second step was based on the first one. Participants were informed that the two groups of animals were equally likely to have a specific property. They were asked to tell which statement was more surprising. Based on the first two steps, in the third step, the participants were asked to judge which of the two arguments was stronger in the arguments strength. In Step 1, if a participant chose the one with smaller DWA, it was recorded as 1, but 0 otherwise. In Step 2, if a participant chose the one with larger DWA, it was recorded as 1; but 0 otherwise. In Step 3, if a participant chose the one with larger DWA, it was recorded as 1, but 0 otherwise. 2.5 Results and discussion

participants responses in Step 3 need comparing with their strength judgments under the mono-step condition to test the difference. Because the paradigm was within subjects, we conducted a non-parametric test for two independent samples with SPSS 10.0 to compare the difference. The result was as follows (Table 3):
Table 3 Non-parametric test of diversity effects between two paradigms DBA 1d 5d 9d Mono-step 51.4% 57.1% 67.1% Triple-step 54.2% 66.7% 66.9% Z 0.326 1.165 0.60 p 0.745 0.244 0.952

(i) Test the rationality of SMCC. When the material was presented all in one step participants immediately made judgment of inductive strength. By contrast, when material was presented in triple steps participants made judgment only in the last step. In the triple-step condition, was there any confusion produced during Steps 1 and 2, thereby producing an unexpected influence on Step 3? If it was, it implied that the triple-step condition probably had changed the nature of the task, and made it significantly different from the nature of the mono-step condition, which might be a challenge to the rationality of SMCC. If it was not, it suggested that the triple-step presentation had not changed the nature of the task, and SMCC might be rational. Therefore, in order to test the rationality of SMCC,
Table 4 DBA Step Percentage
2

The result indicates unanimously that there was no significant difference between the strength judgments under the mono-step condition and the triple-step condition, and it happened for each of all the three levels of DBA. The result suggested that there was no significant difference between the participants response inclination under these two paradigms, no matter what the DBA was. It was obvious that the triple-step condition did not change the nature of the inductive reasoning task. And we could indicate that the SMCC is rational. (ii) The response rule of the triple-step tasks. Having testified the rationality of SMCC, we further discussed the response rule of triple-step tasks. The response in Step 1 was actually a judgment of prior probability. The response in Step 2 had two aims: one was to inform that two premise combinations were equally likely to be true; the other was to record the participants feeling of surprise. The response in Step 3 was to judge the extent to which the premise combinations supported the general conclusion. The following table (Table 4) showed the participants responses of three steps at three levels of DBA.
5d S2 52.8% 0.222 0.637 9d S2 68.1% 9.389 0.002**

The response inclinations in each step under triple-step conditiona) S3 54.2% 0.5 0.480 S1 63.9% 5.556 0.018* S3 66.7% 8.000 0.005** S1 70.8% 12.5 0.001** S3 66.7% 8.000 0.005**

S1 56.9% 1.389 0.239

1d S2 52.8% 0.222 0.637

a) S1, namely Step 1, recorded the percentage of the participants who thought the more difference between premises the more likely these premises were true simultaneously. S2, namely Step 2, recorded the percentage of participants who thought it surprising that the premises with a larger DWA should have the same property. S3, namely Step 3, recorded the percentage of participants who believed the conclusion based on the arguments with larger DWA was more likely to be true. * p < 0.05; ** p < 0.01.

Diversity effect of inductive reasoning under segment manipulation of complex cognition

665

We noticed from Table 4 that when the DBA was 1 the participants performances were all at chance level. These results were similar to those of Experiment 1. Two reasons might account for the phenomenon: one was that the DBA was too small to be detected by the participants; the other was that the participants noticed the existence of DBA but they thought it too small to support the diversity effect. Apparently in the former case the participants only made random choice, while in the later case, emergence of the diversity effect might be conditioned. When the DBA was 5, the participants responses in Step 1 and Step 3 were all significantly consistent with experimental expectation, whereas in Step 2 there was no significant inclination. By comparing this result with that of Experiment 1, we can see that when the DBA was 5, the participants did not show any significant response inclination, while in Experiment 2, there was a significant inclination. We will discuss the possible reasons in General discussion. When the DBA was 9, the result showed that participants showed very significant inclination in each step. Participants were clearly aware of the DBA in Step 1, and they felt surprised when they were informed that the premise combinations with larger and smaller DWAs had the same property. In Step 3 they tended to report that the conclusion based on the premise combination with larger a DWA was stronger than that with a smaller DWA. The responses in the three steps accorded with the experiment expectation very well. 3 General discussion

in that it realized the study of the diversity effect, and excluded irrelevant information. SMCC was a treatment similar to slowing down the motion of movies, whose main feature was that it divided inductive reasoning into several continual stages. Accordingly we could design particular task for each stage. Because the participants thinking could only go forward with the progressive presentations of tasks, the thinking processes could be analyzed like slow movies. It was believed that this design was helpful for observing the responses in the three stages, and the mental mechanisms of the diversity effect could be analyzed based on the synthesis of these responses. Both the responses in Step 3 of SMCC and the holistic responses of the traditional paradigm were the judgments of inductive strength. If participants responses to those two questions were similar, it could be believed that the responses in Step 3 of SMCC were equivalent to those in the traditional paradigm, namely, the additional first two steps in SMCC did not affect the nature of the task, which can prove the rationality of the new paradigm. The result of Experiment 2 indicated that there was no significant difference between the responses in Step 3 and those in traditional paradigm, that two tasks are equivalent, and that the new paradigm does not change the nature of inductive reasoning. The three sub-tasks in SMCC examined respectively three different stages of inductive reasoning: initial probability judgment (initial though), probability modification (belief modification) and the feeling of surprise, and inductive strength judgment. The responses of three steps were obtained in the experiment. Only through the analysis of participants responses in the tasks showing the diversity effect can we discover the rules of this effect. Since participants showed the effect only when the DBA is 5 and 9, let us examine the participants responses of every stage under the two conditions. It can be inferred from the results of Table 4 that when the DBA was 5 or 9 participants responses in Step 1 were significantly inclined to a premises combination with smaller DWA, which suggested that par-

The result in Experiment 1 showed that when the DBA was 9d, the participants significantly showed diversity effect. However, when the DBA was at a lower level (1 or 5) they did not show any diversity effect. This result means: (i) the emergence of the diversity effect depended on the DBA; (ii) the emergence of the diversity effect was conditioned; (iii) the DBA could rationally represent the premise diversity, which suggested that the aim of quantitative treatment and systematic manipulation of premise diversity could be achieved. The advantage of this solution lied

666

Science in China Ser. C Life Sciences

ticipants thought that the premise combination with smaller DWA was more likely to be true. It accorded with the expectation from sub-task hypothesis. It reflected the initial belief that the things which were more similar to each other in appearance were more likely to have the same property, and vice versa. However, the participants responses in Step 2 were complicated. They had no significant inclination when the DBA was 5. By contrast, when the DBA was 9, they showed significant inclination, which affords food for thought. Step 2 actually included two activities: one was to inform the participants that the two premise combinations were equally likely to be true (which was out of participants expectation); the other was to record the participants surprise responses. For this result, there are two possible interpretations. One is that the participants may instinctively think the DBA 5 is not large enough to surprise them, however when the DBA is large enough (9) the participants feel surprised. In this case, there must be a threshold of DBA for FOS. When the DBA is equal to or above the threshold, participants would feel surprised; when the DBA is below the threshold, participants would not feel surprised. The other interpretation is that the DBA (5) is big enough, but there may be some intervening factors in the experiment. At present, we cannot explain why participants were not surprised in Step 2 when the DBA was 5. Whatever case it may be, further study is needed. When it comes to Step 3, participants showed a very significant inclination under the two DBA conditions, which was a manifestation of the diversity effect. The results showed that participants believed that the premise with larger DBA could strongly support a more general conclusion. Synthesizing the holistic responses of three steps, the following conclusions could be obtained: participants initial conception holds that the premise combination with smaller DWA is more likely to be true; the conceptual modification makes the participants feel more surprised that the premise combination with larger DWA should also be true; when judging the strength of argument conclusion, participants significantly believe that the prem-

ises with larger DWA (more diversity) can support general conclusions better, thus the diversity effect is clearly demonstrated. The responses of the three steps of SMCC obey certain rules. What is the connection between the rules and the mental mechanisms of the diversity effect? What are the mental mechanisms of the diversity effect of inductive reasoning? Heit[1] and Sanjana and Tenenbaum[13] tried to interpret this effect with Bayes Theorem. Based on that study, Lo et al.[11] put forward the premise probability principle (PPP), and suggested that the less probable the premise combination was, the more probable the conclusion supported by it was. It is expressed briefly as follows: The strengths of arguments A, B / C and A ', B '/ C can be expressed respectively as follows: P(C | A B) , P C | A ' B ' . According to Bayes

Theorem, there is a conditional probability as follows: P( A B) P( A | B) , so the strengths of the above P( B) two arguments can be rewritten as: P( A | B) = P( A B) P(C A ' B ') and . P( B) P( A ' B ') Because A C , B C , and A ' C , B ' C , there is P(C A B) = P(C A ' B ') = P(C ) , Then P(C ) P(C ) and . P( A B) P( A ' B ')

Obviously, it was the magnitude of the two denominators that determined the strength judgment, and the denominator was the probability of the premise combination being true. Therefore, the probability of premise combination was in inverse ratio with the strength. In general, people are naive to novel material; they had no prior probability. In this study, participants could establish their prior probability only after the presentation of the pictures in Step 1, that is, the recorded responses in Step 1 of Experiment 2 were the prior probability of premise combination holding true. In the traditional studies that apply Bayes Theorem to

Diversity effect of inductive reasoning under segment manipulation of complex cognition

667

analyze the strength judgment, researchers all agree that there should be a prior probability about the event, but it has never been identified experimentally. This study made use of novel pictures under SMCC and discovered that participants did have a process of forming the prior probability, and recorded the phenomenon. But how is the prior probability formed? It is an issue for further study. In Step 2 of Experiment 2, participants got to know a fact opposite to their initial judgment: the premise combination with larger DBA was equally likely to be true as that with smaller DBA. In this case, participants estimation of the probability of premise combination changed from being unequal to equal. According to the PPP, now, P( A B) equals to P ( A ' B ') , then P(C ) / P( A B) and P(C ) / P( A ' B ') should also be equal. Consequently, when the premise probabilities were equal, the conclusion probabilities should also be equal. But, the present results that participants thought the premises with smaller DWA supported the conclusions more strongly. The basic hypothesis of PPP was that prior probability was modified according to Bayes Theorem and ended up with a posterior probability (argument strength). Nevertheless the results of SMCC showed that when the participants were informed that they were both true, participants prior probabilities were modified (to become equal). Therefore, according to PPP, we can only predict that the two strengths are equal. Because the study of Lo et al. applied the traditional paradigm (mono-step presentation), although they put forward PPP and analyzed the issue of prior and posterior probability based on Bayes Theorem, they could not measure participants prior probability. Hence the so-called posterior probabilitys being modified by prior probability can only be a speculation. In present study, the data in Step 1 of SMCC were rightfully the participants prior probability (the probability in which the premise combination with larger DWA was true was smaller than that with smaller DWA.); in Step 2, the participants were informed a clear fact, thus a modification of prior prob-

ability happened( the probability of premise combination with larger DWA was the same as that with smaller DWA. thus a feeling of surprise appeared), which resulted in a modified posterior probability; In Step 3, a final strength judgment was obtained based on the modified posterior probability. According to PPP, after Step 2 was completed, the participants were informed the fact that the premise combination with larger DWA was as true as that with smaller DWA. Therefore, the judgment should be obtained that the premise combination with larger DWA and one with smaller DWA equally support the conclusion. But, the experimental results turned out not to be the case, suggesting that PPP cannot explain the diversity effect well enough. Why can PPP not explain the diversity effect observed in the study very well? We think the main reason may be that the interaction between thinking and emotion is ignored. According to the results of the study, we guess that the FOS in Step 2 may enlarge the strength estimate in Step 3 that follows. As the similar cases often happen in real life, since the thing which is early regarded not very probable (it is hence unusual and incredible) happens, the relevant thing in a wider range should also be true, so does the diversity effect happen. Therefore the FOS we observed is probably the mental mechanism of the diversity effect. Currently, cognitive event-related potentials (ERPs) technology has provided more accurate scientific evidence for the complex thinking processes such as attention[14], Chinese cognition[15], and thinking[16]. Further research on advanced thinking activities such as the diversity effect of inductive reasoning, especially the function of FOS in diversity effect will be carried out by applying the ERP technology.
Acknowledgements This work was supported by the National Natural Science Foundation of China (Grant No. 30370488) and National Key Subject of Basic Psychology, SWU (Grant No. XGZ04006).

References
1. Heit, E., A Bayes analysis of some forms of inductive reasoning, in Rational Models of Cognition (eds. Oaksford, M., Chater, N.), New York: Oxford University Press, 1998, 248 Press, 1985. 274. 2. Carey, S., Conceptual Change in Childhood, Cambridge: MIT

668
3. Osherson, D. N., Smith, E. E., Wilkie, O. et al., Category-based induction, Psychological Review, 1990, 97: 185 nitive Psychology, 2001, 43: 243 273. 200. 4. Heit, E., Hahn, U., Diversity-Based Reasoning in Children, Cog5. Lpez, A., Gelman, S. A., Gutheil, G. et al., The development of category-based induction, Child Development, 1992, 63: 1070 1090. 6. Choi, I., Nisbett, R. E., Smith, E. E., Culture, category salience, and inductive reasoning, Cognition, 1997, 65: 15 32. 7. L pez, A., Atran, S., Coley, J. D. et al., The tree of life: Universal and cultural features of folkbiological taxonomies and inductions, Cognitive Psychology, 1997, 32: 251 295. 95. 8. Gelman, S. A., The development of induction within natural kind and artifact categories, Cognitive Psychology, 1988, 20: 65 9. Holland, J. H., Holyoak, K. J., Nisbett R. E., Induction: Process of Inference, Learning, and Discovery, 2nd ed., Cambridge: MIT Press, 1989. 10. Yamauchi, T., Markman, A., Inference using categorization, Journal of Experimental Psychology: Learning, Memory, and Cognition, 2000, 26: 776 795.

Science in China Ser. C Life Sciences

11. Lo, Y., Sides, A., Rozelle, J. et al., Evidential diversity and premise probability in young childrens inductive judgment, Cognitive Science, 2002, 181 206. 12. Chen, A. T., Li, H., The study of diversity effect of inductive reasoning, Advances in Psychological Science (in Chinese), 2003, 31: 103 108. 13. Sanjana, N. E., Tenenbaum, J. B., Bayesian models of inductive generalization, in Advances in Neural Processing Systems 15 (eds. Becker, S., Thrun, S., Obermayer, K), Cambridge: MIT Press, 2003. 14. Luo, Y. J., Wei, J. H., Research of cross-modal attention on deviance-related components of ERPs, Science in China, Series C, 1997, 40: 604 612. 15. Luo, Y. J., Wei, J. H., Event-related potential on the cross-modal identification of Chinese language, Science in China, Series C, 1999, 42: 113 121. 16. Mai, X. Q., Luo, J., Wu, J. H. et al., Aha! Effects in guessing riddle task: An ERP study, Human Brain Mapping, 2004, 22: 261 270.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.