Linguistic Society of America

Cognition in Ethnolinguistics
Author(s): Eric H. Lenneberg
Reviewed work(s):
Source: Language, Vol. 29, No. 4 (Oct. - Dec., 1953), pp. 463-471
Published by: Linguistic Society of America
Stable URL: http://www.jstor.org/stable/409956 .
Accessed: 31/12/2012 06:48
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.

http://www.jstor.org

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

COGNITION IN ETHNOLINGUISTICS
ERIC H. LENNEBERG

Massachusetts Institute of Technology

The republication of Benjamin L. Whorf's articles on what Trager calls
metalinguistics has aroused a new interest in this country in the problem of the
relationship that a particular language may have to its speakers' cognitive processes. Does the structure of a given language affect the thoughts (or thought
potential), the memory, the perception, the learning ability of those who speak
that language? These questions have often been asked and many attempts have
been made to answer them.' The present paper is an attempt to lay bare the
logical structure of this type of investigation.
CRITICALRETROSPECT

A basic assumption. Underlying all of Whorf's theoretical work is the fundamental assumption that the individual's conception of the world (including perception, abstraction, rationalization, categorization) is intimately related to the
nature of his native language.2 Throughout his work Whorf illustrates this idea
with examples from American Indian languages, showing how they differ from
English. However, a demonstration that certain languages differ from each
other suggests but does not prove that the speakers of these languages differ from
each other as a group in their psychological potentialities. To prove this, it would
be necessary to show first that certain aspects of language have a direct influence
on or connection with a given psychological mechanism, or at least that speakers
of different languages differ along certain psychological parameters. In addition
to comparative data Whorf adduces occasionally a different type of evidence.
An example is his analysis of many hundreds of reports of circumstances sur1 Bibliographies of the voluminous literature may be found in the following works:
Kurt Goldstein, Language and language disturbances (New York, 1948); Friedrich Kainz,
Psychologie der Sprache (Stuttgart, 1941/43); George A. Miller, Language and communication (New York, 1952); Charles Morris, Signs, language and behavior(New York, 1946);
David L. Olmsted, Ethnolinguistics so far (SIL, Occasional papers, No. 2; 1950); N. H.
Pronko, Language and psycholinguistics: A review, Psych. bull. 43.189-239 (1946).
This paper was stimulated by research carried on under the auspices of the Values Study
in the Laboratory of Social Relation, Harvard University, and the CommunicationsProject
at the Center for International Studies, Massachusetts Institute of Technology. I wish to
express my thanks to both institutions. I am also greatly indebted to Harry Hoijer for
inviting me to participate in the Conference on Ethnolinguistics, held in Chicago during
March 1953, where the discussion of some of the problems raised in this paper helped to
clarify my thoughts. Finally I gratefully acknowledge the many helpful suggestions made to
me by Noam Chomsky, who read two earlier versions of this article.
2 Whorf is not alone in making this assumption. Cf. Dorothy D. Lee, Linguistic reflection
of Wintu thought, IJAL 10.181-7 (1944); Lucien Levy-Bruhl, Les fonctions mentalesdans les
societts infirieurs, Ch. 4 (Paris, 1910); Leo Weisgerber, Adjektivistische und verbale
Auffassung der Gesichtsempfindungen, Wirter und Sachen 12.197-226 (1929). The last of
these is a representative of what H. Basilius has called Neo-Humboldtian ethnolinguistics,
Word8.95-105 (1952); the entire movement is based on the assumption discussed here.
463

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

464

LANGUAGE, VOL. 29, NO. 4

rounding the start of fires, for instance the empty-gas-drum case.3 An explosion
had been caused by an individual who had carelessly flung a burning cigarette
stub into a gas drum which this person in his insurance report called empty.
Whorf argues that the individual's carelessness was caused by the fact that the
word empty has two different meanings in English: (1) null and void, negative,
inert, and (2) a space which may contain nothing but a vapor, liquid vestiges, or
stray rubbish. The English language forced the individual to call the gas drum
empty, and think of it in terms of that word. Since this word could mean null
and void, Whorf argues that the presence of explosive vapors and inflammable
liquid vestiges could be disregarded by the speaker, who then behaved towards
the drum as if it were absolutely empty. I cannot accept this as evidence for
the assumption that behavior is influenced by language. Clearly, English is
capable of distinguishing between a drum filled with an explosive vapor, one
that contains only air, and one which is void of any matter. This very sentence
is my evidence. The person who caused the fire could have replaced the word
empty by filled with explosive vapor. His failing to do so (as well as his careless
behavior) points to a lack of experience with explosive vapors, perhaps complete
ignorance of their existence. The linguistic-or rather stylistic-fact of the
occurrence of the word empty in the individual's insurance report would indeed
be interesting if Whorf could have shown at the same time that this man had had
plenty of contact with and knowledge of the explosive vapors which form in
emptied gas drums. This Whorf did not try to do. In short, the basic assumption that language affects non-linguistic behavior derives from an inspection of
linguistic facts. Therefore nothing is added to such an hypothesis by referring
back to the same or similar linguistic facts.
Translation. (a) Translation, while useful for the formulation of working
hypotheses of the most exploratory nature, is in itself an inadequate way towards
the finding of objective facts. Obvious as this may seem, it is necessary to spell
out in detail the shortcomings of the translation method in ethnolinguistics.
(b) I illustrate my point with another example taken from Whorf. After posing
the question: 'What do different languages do ... with the flowing face
of nature ...?', Whorf answers: 'Here we find differences in segmentation and
selection of basic terms. We might isolate something in nature by saying, "It is
a dripping spring." Apache erects the statement on a verb ga: "be white (including clear, uncolored, and so on)." With the prefix no-, the meaning of downward
motion enters: "whiteness moves downward." Then to, meaning both "water"
and "spring" is prefixed. The result corresponds to our "dripping spring," but
synthetically it is: "as water, or springs, whiteness moves downward." How
utterly unlike our way of thinking! [NB !] The same verb, ga, with a prefix that
means "a place manifests the condition" becomes gohlga: "the place is white,
clear; a clearing, a plain." These examples show that some languages have means
of expression ... in which the separate terms are not as separate as in English
but flow together into plastic synthetic creations.' Whorf analyzes the Apachean
3 Whorf, The relation of habitual thought and behavior to language, Language, culture,
and personality 75-93 (Menasha, Wis., 1941).
4Whorf, Languages and logic, The technologyreview,Vol. 43 (1941).

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

COGNITIONIN ETHNOLINGUISTICS

465

statement by giving the English equivalent for the general meaning of each
Apachean element, and then compares the resulting sequence of meanings to
the phrase, 'it is a dripping spring'. The sequence of meanings (i.e. the glosses)
and the English phrase are not, however, quite comparable. Whorf does not
give the general meaning of the English morphemes. If he had, something like
this would have resulted:
it, any object or organism which is not an adult human being; is, particle
which denotes that what follows is a predicate of what precedes; a, particle
which denotes that what follows is to be understood generically, not specifically; drip(p), process in which any liquid falls in small natural segments;
-ing, particle which denotes that the preceding process has not come to an
end; spring, something that is not static (eruption of water, device to make
mattresses elastic, and so on).
To abstract a general meaning of a morpheme or lexeme may occasionally be of
some methodological use; but we must not confuse such an abstraction with an
isolable segment of an utterance. General meanings lack reality, so to speak.
It makes no sense to equate the global meaning of an utterance with the sequence
of abstracted, general meanings of the morphemes that occur in that utterance.
To translate the Apachean statement it is a dripping spring appears no less
reasonable than to translate it as water or springs, whiteness moves downward
at a place (or, the place is white, clear; a clearing; a plain--which, I gather from
Whorf, is the synthesis of the elements); for what we translate are equivalent verbal responses to particular stimulus situations, and the Apachean response to the
natural phenomenon in question corresponds to our response it is a dripping
spring. This type of linguistic evidence, therefore, stands or falls with our philosophy of translation. It might be objected here that Whorf's evidence is not the
translation itself but the fact that the Apachean's verbal response to this natural
phenomenon is the same as his verbal response to a different phenomenon,
namely one to which we respond the place is white, clear, etc. and that
the Apachean therefore makes a single response to stimuli to which we make
distinct responses. This objection, while touching upon an important problem,
does not justify the translation method. For what we really want to know is
how the Apachean structure of syntactic categories differs from the English one.
Translation cannot answer this problem. Through it-and that is its value-we
merely know that the problem is not a spurious one.
(c) A further objection to translation as a sufficient method in this type of
research is that it actually vitiates the attempt to demonstrate cognitive difference as evidenced in two or more languages. For, if a language were actually an
aspect of a particular psychological make-up or state of mind (or more precisely,
an aspect of a cognitive process, which is not to be confused with the thought
content), then, in the process of translation, we would be substituting the psychological elements characteristic of one make-up for those of another, so that we
would finally compare two sets of elements of one and the same psychological
structure.
(d) There is a metaphorical element in language per se. The literal meaning of
many metaphors, especially the most frequent ones, never penetrates conscious-

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

466

LANGUAGE,VOL. 29, NO. 4

ness, e.g. everybody,in the face of, beforehand,breakfast, inside, already. The
translation method, however, distorts the significance of such forms of speech
and often induces investigators to draw rather ludicrous conclusions. To illustrate the mentality of certain African tribes, Cassirer writes:' 'The languages of
the Sudan usually express the circumstance that a subject is in process of action
by means of a locution which really means [NB!] that the subject is inside that
action. But since, moreover, this inside is usually expressed very concretely,
phrases result such as I am on the inside of walking, I am the belly of walking, for
"I am in the process of walking".'
(e) The process of taking stock of general meanings, which underlies translations, engenders the belief that languages can convey no more and no less than
the general meanings of morphemes. It seems more fruitful to assume6 that much
more is cognized than is expressed by individual morphemes. Morphemes and
their meaning are regarded more appropriately as mnemotechnical pegs of a
whole situation which is brought into consciousness by the statement as a whole.
The general meaning of morphemes is probably of lesser importance in cognition
than the SUM OF ASSOCIATIONS bound up with the complete utterance, or even
with individual morphemes or groups of morphemes.
(f) When the translation involves a juxtaposition of totally different cultures
(say Chukchee and English) we are not only faced with a semantic problem. No
matter what precautions we take in glossing a word, almost no correspondences
can be established between many denotata. For instance, the cultural and physical contexts of Chukchee utterances are, with a few exceptions, incomparable
with the contexts within which English is spoken. Chukchee weapons, food,
manners, standards of any sort, landscape, fauna, and flora are mostly unfamiliar
to English-speaking cultures. Thus, practically no common frame of reference,
no basis for a segmental, one-by-one comparison exists between these two languages. Translation here can be only a very rough approximation of what has
been said and intended originally.
Ad-hoc theories. It is a commonplace in scientific methodology to avoid
etiological theories which are incapable of satisfying more than one single and
specific occurrence of events; yet by necessity working hypotheses often have to
be of this nature. We see a picture fall off a wall directly after hearing a dog bark
in the neighborhood. As a working hypothesis the two events might be causally
related. Upon verification of the hypothesis we note, however, that in general
barking is not followed by things dropping to the ground, nor is the falling of
pictures from the wall usually preceded by barking or similar noises. We are
unable in this instance to formulate a theory because the working hypothesis
cannot be generalized. Turning to ethnolinguistic literature we find an abundance
of working hypotheses where it is difficult to see how they might contribute to
a universally valid and useful theory of language (such that language is related
5 Ernst

Cassirer, Philosophie

der symbolischen Formen: Die Sprache 1.168-9 (Berlin,

1923). Cassirer's source is D. Westermann, Sudansprachen.

6 This assertion and the following are based on evidence from experimental psychology.
See George Humphrey, Thinking: An introduction to experimental psychology, Chs. 4 and 8
(London, 1951); Miller, Language and communication passim.

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

COGNITIONIN ETHNOLINGUISTICS

467

to non-linguistic behavior), because the facts underlying such working hypotheses

cannot be generalized so as to fit more than a single language.' I am not saying
that such hypotheses are right or wrong; many have been proposed by experts
on specific cultures, by scientists of undisputed merit. I am merely pointing to
the difficulty, if not impossibility, of deducing from these hypotheses, if they
are sound, general and verifiable laws. A common means of validating hypotheses
has been barred from the beginning in these cases, namely cross-cultural verification. This, however, does not exclude the possibility that the investigators may

have intra-culturalevidence for each individualhypothesisproposed.

A METHODOLOGY
TOWARDS

Codificationand cognition.(a) A basic maxim in linguisticsis that anything

can be expressed in any language.8 There may be differences in the ease and
facility for the expression of certain things among various languages but at
present we do not know whether this difference in ease is attributable to the
properties of a given language qua vehicle of communication9 or to the cultural
development of the speakers. In fact, this is one of the problems to be solved in
ethnolinguistics. Now, if we believe, as we do, that we CANsay anything we wish
in any language, then it would seem as if the content or subject matter of utterances does not characterize or, indeed, give us any clear information on the
communicative properties of a language. Thus we are led to the somewhat banal
conclusion that the only pertinent linguistic data in this type of research is the
HOW of communicationand not the WHAT. This HOWI call the codification;the
WHAT I call the messages. Codification can be studied in three phases: (1) the
process of encoding; (2) the code; (3) the process of decoding.10The study of the
code results, for instance, in statements about the structure of phonemes, morphemes, and syntactic categories; about acoustic characteristics of speech sounds;
about the frequency distributions and the transitional probabilities of given
segments; about the efficiency of the code within stated contexts. In these instances meaning can be excluded entirely from our research, at least theoretically,
and we have therefore an assurance that we are actually studying aspects of
codification. Unfortunately, however, it is not always equally easy to decide
whether a phenomenon is pertinent to codification or not. Many assertions about
language which derive from semantic observations or, at any rate, which include
elements of meaning, nevertheless seem to be relevant to codification. Most
7 Most of Whorf's and Dorthy Lee's working hypotheses are of this nature. Harry
Hoijer's
Cultural implications of some Navaho linguistic categories, Lg. 27.111-20 (1951), and the
tentative connections between various linguistic features and nonlinguistic behavior mentioned by Claude L6vy-Strauss, Language and the analysis of social laws, Amer. anthr.
53.155-63 (1951), also fall into this category of working hypotheses.
8 Cf. Sapir, The grammarian and his language, Selected writings of Edward Sapir 153-4
(Berkeley, 1949). It is assumed here that any vocabulary can be expanded.
9 The use of the term vehicleof communicationdoes not mean that I deny (or even take a
position toward) the epistemological contention that language and knowledge are indistinguishable. I am merely referring to the communicative capacities of language.
10John B. Carroll, Report and recommendationsof the Inter-disciplinary SummerSeminar
in Psychology and Linguistics 8 (Ithaca, N. Y., 1951).

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

468

LANGUAGE,VOL. 29, NO. 4

obvious in this connection is the fact that a language always selects for codification highly specific aspects from the physical and social environment. This raises
two questions: How can we describe objectively the aspects that are being
selected out of a great number of other possible aspects? Why are these aspects
selected and not others? There can be little doubt that these considerations,
though clearly of a semantic character, have a bearing on the problem of codification. Hence, the distinction between codification and messages is not the same
as between syntactics and semantics or between form and meaning. All those
observations about meaning are relevant to codification which refer to an aspect
of speech behavior which is forced upon the individual speaker by the rules of
his language and where infringement of the rules would result in defective communication. For instance, an individual reporting about a given event is forced
to stipulate very definite conditions, aspects, and relationships if he wants to be
understood." However, he is free to report on the event in the first place, and
also to elaborate on circumstances of the event which are not included in the
compulsory stipulations. Whatever information is optional in his communication
is message.
(b) Once we have clearly isolated data on codification, such data may be
related hypothetically to nonlinguistic behavior. If the researcher is interested
in cognition,'2 as I am, he will investigate relations that obtain between codification and such behavior as is indicative of memory, recognition, learning, problem
solving, concept formation, and perception, hoping to show that certain peculiarities in these processes can be explained by--and only by-knowledge of the
speakers' peculiarities of codification.'3
The intra-cultural approach. (a) Ethnolinguistic research based on crosscultural comparison must endeavor to isolate data, both on codification and on
cognition, that are general enough to have comparable equivalents in at least
two different languages and cultures; otherwise comparison would be meaningless. It is not infrequent, however, that a working hypothesis relates a certain
cognitive datum to some phenomenon pertinent to codification which appears to
be unique, lacking entirely a parallel in any other language. There is a simple way
of studying this situation; I call it the intra-cultural approach, because it reduces
cross-cultural comparison to a desirable but not indispensable expansion of investigations. This method is so easy to manipulate that many investigators may
perhaps come to use it even where the cross-cultural approach is applicable
directly.
(b) I begin with a practical demonstration of the method. Problem: Languages
differ in their systems of classifying the ten million odd colors which every normal
individual can discriminate.'4 Under laboratory conditions the power of color
These conditions, aspects, and relationships are primarily but not exclusively expressed by grammatical categories.
12
For a modern definition of this term see Robert Leeper, Cognitive processes, Hand"'

book of experimental psychology 730-57 (ed. S. S. Stevens; New York, 1951).

13What I am proposing to do here is not in principle different from what Whorf (for
instance) occasionally suggested. The difference between Whorf and me is rather in our
respective attempts to substantiate our hypotheses.

14Cf.

Ralph M. Evans, An introduction to color 230 (New York, 1948).

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

COGNITIONIN ETHNOLINGUISTICS

469

discrimination is probably the same for all human beings, irrespective of their
language background. But we do not know whether the habitual grouping of
colors, according to certain labels provided by every language, might not affect
some other cognitive processes involving color stimuli. To be more specific, in
English obviously not all colors are named with equal ease and unambiguity.
Do English-speaking people therefore recognize easily-named (i.e. highly codeable) colors with greater facility than colors not so easily named?'5
The first step toward solution of this problem is to ascertain the linguistic
facts.'6 A representative sample of English speakers is drawn and a number of
colors are prepared that have comparable perceptual properties. Then the notion
'codeable' is investigated and defined operationally, so that we can divide the
physical color stimuli by means of one or a combination of a few simple criteria,
into two groups: one consisting of 'highly codeable' and one of 'less codeable'
colors. I must omit here the details of this procedure and also the reasoning that
underlies the individual steps leading to the development of such a criterion. Let
me simply state that UNANIMITY
IN RESPONSE proves to be a useful criterion
(among others) in this connection. Some colors are consistently given the same
name by every speaker; others are given a variety of names, sometimes as many
names as there are subjects. Regarding the speakers now as a group giving a
linguistic response to each color, we may say that some colors have the property
of eliciting a homogeneous response from English-speakers, whereas other colors
elicit a heterogeneous response. This is to say that linguistic communication in
English is more efficient when some colors are referred to than when others are.'7
There are cogent reasons to assume that the distinction made here between the
colors is a purely linguistic one, and that there are no physical properties in the
colors or physiological ones in the eye which would elucidate the difference in
response made by English-speakers to these colors.'8
15 This is a
specific question within a problem that has been posed by many other investigators. Sapir said: 'Language is a ... self-contained, creative symbolic organization,
which not only refers to experience largely acquired without its help, but actually defines
experience for us by reason of its formal completeness and because of our unconscious
projection of its implicit expectations into the field of experience.' (Conceptual categories
in primitive languages, quoted by I. J. Lee, The language of wisdom and folly 265 [New
York, 1949].) Sapir makes the same point in The status of linguistics as a science, Selected

writings 162.

'1 The following is an outline of research in progress carried on by Roger Brown of Harvard University and myself. The details of the project will be published as soon as the data
are fully assembled.
17 If there is no well defined name for a color, it is reasonable to assume that linguistic
communication about it is poor.
18 Again space does not permit me to cite all the evidence in support of this assertion.
The interested reader may inspect the colors used; they are produced by the Munsell Color
Co., a scientific research organization. Most of them are published in the two volumes of
the Munsell book of colors (Baltimore, 1921 and 1942). Codeable colors have the notation
2.5 PB/7/6, 5 PB/4/10, 10 P/3/10, 5 RP/6/10, 5 YR/3/4, 3 GY/7.5/11.2, 7.5 GY/3/4, 2.5
G/5/8, 5 Y/8/12, 7.5 G/8/4, 2.5 R/7/8. Non-codeable colors have the notation 10 BG/6/6,
8.5 B/3/6.8, 10 PB/5/10, 2.5 R/5/10, 8 RP/3.4/12.1, 7.5 R/8/4, 2.5 Y/7/10, 7.5 Y/6/8, 7.5
YR/5/8, 5 P/8/4, 5 BG/3/6. Colorimetric and psycho-physical data on these colors are pub-

lished in Journal of the Optical Society of America 30.573-645.

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

470

LANGUAGE, VOL. 29, NO. 4

The next step is to determine whether there is a difference in ease of recognition

by English-speakers between the colors constituting one group and those constituting the other. The two groups, I repeat, are perfectly balanced in physical
and perceptual properties; the only difference is that the colors in one group
have well defined names in English, whereas the colors in the other do not.1gIf
we now use, in random order, colors belonging to either group, say ten from
each, in a standard recognition test,20 we can easily discover whether English
speakers do better when they have to recognize colors which are highly codeable
in their language than when they recognize less codeable colors. In the actual
performance of the experiment21 this appears to be the case. Statistically, codeable colors are recognized significantly more often than less codeable ones, and
thus there is good evidence that the particular linguistic fact, codeability, affects
the cognitive process, recognition.
(c) Suppose now that this entire color research were repeated in a different
culture where a different language is spoken. If our predictions about recognition,
based on previously determined facts of codification (which vary of course from
language to language), should not be borne out in this other language, the argument advanced in the first experiment would be seriously weakened. Conversely,
if the results should be confirmed, this would fortify the argument. In either case,
of the basic hypothesis is independent of cross-cultural
however, VALIDATION
comparison. The cross-cultural comparison merely adds or subtracts weight. It
is very important to realize that the validation itself is the result of intra-cultural
correlation of two sets of recognition behavior on the one hand (in the described
context we may say 'good' and 'bad' behavior) with two sets of English speech
behavior on the other hand (efficient and not so efficient linguistic communication). It appears that recognition behavior is inefficient where speech behavior
is inefficient.
(d) Not only is the validity of this experiment independent of cross-cultural
comparison; but if cross-cultural comparison is desired, the method dispenses
with the necessity for translation, or the exact equation of linguistic data between
OF
one language and another. For what will be compared are CORRELATIONS
forms.
BEHAVIOR
WITH
RECOGNITION
not
SPEECH
Superlinguistic
BEHAVIOR,
ficially it may look as if the translation method were implicitly the same as the
intra-cultural method, for both methods seem to be concerned with the meaning
of certain linguistic forms which are being compared. However, the intra-cultural
method resembles the translation method only in its very elementary and primary
step: both methods recognize the existence of a problem on the grounds of intui19 Codeability of colors does not seem to be linked to cultural importance or preference
for these colors. The reader may convince himself of this by trying to name all the colors in
his environment. He will notice that colors for which he has a 'good' name occur much less
frequently than colors which are difficult to label unambiguously.
20Such tests are described in K. Koffka, Principles of Gestalt psychology, Chs. 11-3
(New York, 1935).
21 The test colors were exposed four at a time, for two seconds. After a waiting period of
thirty seconds, subjects had to find the test colors on a color chart of 120 colors. All colors
were identified by numbers. The subjects used in this experiment were not required to use
any color name whatever.

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions

COGNITIONIN ETHNOLINGUISTICS

471

tive knowledge of the meaning of forms. The translation method defines meanings
by trying to equate forms of a language foreign to the investigator to forms of his
native language (where meanings are said to be known). The intra-cultural
method need not rely on this haphazard procedure; instead, it objectifies the
intuited meanings of forms by carefully relating them to stimuli of the environment. Thus it is possible (at least in some instances) to specify meaning by
referral to the physical properties of those stimuli.
(e) Stated in general terms, the intra-cultural approach consists of the following. Some aspect of codification is described in order to correlate it with nonlinguistic behavior. A frame of reference is established in terms of which both
the speech behavior and the non-linguistic behavior can be described or specified;
a particularly convenient frame of reference is the physical environment within
which both types of behavior take place. In the experiment described, the frame
of reference was provided by the stimuli sensed as colors. The speech events
(color terms) and the behavioral events (recognition) were related to these
stimuli. The specifications of the physical properties of the stimuli served as a
metalanguage, so to speak, for the description of both types of events.
The fundamental principle of the intra-cultural method is that the physical
stimuli, whatever they may be, can be classified on the grounds of linguistic
criteria so that the constituents of each class are all characterized by the particular way in which they are codified. It is necessary that the codification criterion
should be the ONLYcriterion by which the stimuli can be grouped in this way. If
now the non-linguistic behavior in response to the stimuli thus classified varies
systematically in accordance with the class to which the individual stimulus has
been assigned, we may attribute such regular variation in non-linguistic behavior
to the regular variation in the speech correlates.

This content downloaded on Mon, 31 Dec 2012 06:48:54 AM

All use subject to JSTOR Terms and Conditions