What Is The Sapir-Whorf Hypothesis.pdf

WHAT IS THE SAPIR-WHORF HYPOTHESIS?

Paul Kay

University of California, Berkeley

and

Willett Kempton

Michigan State University

Cognitive Science Program

Institute of Cognitive Studies

University of California at Berkeley

April 1983

Acknowledgement: The authors have received help on this material from a number of colleagues too

great to list. We would like to single out Brent Berlin and Eleanor Rosch for special thanks. This

research was supported by grant BNS 78-15900 from the National Science Foundation. The assistance

of Don Burgess with the Tarahumara ﬁeld trials is gratefully acknowledged.

The production of the Berkeley Cognitive Science Report Series is supported by a grant in cognitive

science to the University of California at Berkeley from the Alfred P. Sloan Foundation.

-2-

Historical Background

The doctrine of radical linguistic relativity is to be understood historically as a reaction to the

denigrating attitude toward unwritten languages that was fostered by the evolutionary view prevalent

in anthropology in the nineteenth century. Subsequently, the research of Boas and his students

showed these languages to be as systematic and as logically rich as any European language, and it

was perhaps inevitable that the latter ﬁnding should spawn a doctrine on non-European languages

and cultures antithetical to the evolutionary view. If this doctrine of radical relativity has led to cer-

tain excesses of its own--in which the valid insistence on the integrity of each linguistic system has

led to an underestimation of their common structural features--we should not forget that it nonethe-

less supplied a needed corrective to the ethnocentric evolutionism it replaced. Indeed, outside of cer-

tain rariﬁed academic milieux, the early relativists’ battle for a rational and unprejudiced view of our

nonliterate contemporaries is not yet won.

Sapir was Boas’ student and Whorf Sapir’s. The following two passages are among the most

frequently cited from Sapir and Whorf, respectively, and the reader’s indulgence is begged that we

print them yet again. In the ﬁrst, Sapir expresses, in terms of no less lucid for being poetic, the basic

empirical ﬁnding of the Boasians on the formal completeness and intellectual adequacy of unwritten

languages.

Both simple and complex types of language of an indeﬁnite number of varieties may by

found spoken at any desired level of cultural advance. When it comes to linguistic form,

Plato walks with the Macedonian swineherd, Confucius with the headhunting savage of

Assam (Sapir 1921:219).

In the second passage, Whorf takes the further step, foreshadowed in other writings of Sapir’s

( e.g. , ‘The "real world" is to a large extent unconsciously built up on the language habits of the

group’ 1951 [1929]:160), of claiming that an intellectual system embodied in each language shapes

the thought of its speakers in a quite general way.

The categories and types that we isolate from the world of phenomena we do not ﬁnd

there because they stare every observer in the face. On the contrary the world is

presented in a kaleidoscopic ﬂux of impressions which have to be organized in our

minds. This means, largely, by the linguistic system in our minds (Whorf 1956

[1940]:212f).

-3-

During the past forty years, there have been many attempts to recast the fundamental insights

of Sapir and Whorf, originally expressed in a number of evocative and sometimes metaphorical pas-

sages similar to those just cited, in terms of sufﬁciently prosaic that the doctrine may be subjected to

empirical test. We will not attempt to review that literature but rather endorse Roger Brown’s con-

viction that Eric Lenneberg in 1953 really said all that was necessary (Brown 1976:128). In Brown’s

summary, ‘Whorf appeared to put forward two hypotheses:

(1) Structural differences between language systems will, in general, be paralleled by nonlinguistic

cognitive differences, of an unspeciﬁed sort, in the native speakers of the two languages.

(2) The structure of anyone’s native language strongly inﬂuences or fully determines the world-

view he will acquire as he learns the language’ (Brown 1976:128).

It seems that Whorf’s interpreters have imbued his doctrine with an additional tacit premise. If

the differences in world-view mentioned in (2) and the non-linguistic cognitive differences men-

tioned in (1) are to be interesting, they must be sizeable. Miniscule differences are dull. But if the

differences in the dependent variable (non-linguistic cognition) are big, then probably the differences

in the independent variable (language) are also big. (This last is not a logical consequence, but

appears to be a piece of tacit, plausible reasoning that has in fact occurred historically.) Hence (1)

and (2) appear to have induced the tacit postulation of (3) on the part of Whorf’s followers.

(3) The semantic systems of different languages vary without constraint.

Traditions of Empirical Research

Empirical research inspired by the Sapir-Whorf hypotheses may be divided into two, largely

independent traditions, one devoted to the evaluation of (1) and the other to the evaluation of (3).

Until a technique is developed for assessing the world-view of a people independently of the

language they speak, no direct test of (2) is possible. It seems probably that in attempting to ﬁnd

empirical support for (3), anthropologists have sought to provide indirect evidence for (2).

The bulk of the research in both traditions has concerned the domain of color. Empirical work

on color devoted to evaluation of the Sapir-Whorf hypothesis by anthropologists and linguists

belongs mainly to the tradition concerned with (3). Studies in this tradition have sought to describe

-4-

and compare the semantics of the color lexicons of different languages. Some of the better known,

early studies are those of Gleason (1951), Ray (1952), Conklin (1955), Lenneberg and Roberts

(1956), Bohannon (1963), and Berlin and Kay (1969). (From the early seventies on, there have been

a large number of studies in this tradition.) The studies prior to 1969 tended to support (3); those

since 1969 have tended to discredit (3). Kay and McDaniel (1978) summarize the ﬁndings of the

later studies regarding constraints operating in color classiﬁcation across languages and attempt to

explain these constraints as a consequence of the neurophysiology of human color vision. In this

paper we accept the main lines of Kay and McDaniel’s conclusions and so take (3) to be discredited.

There appear to be strong constraints on possible interlinguistic variation in the encoding of color.

The experiments reported in this paper belong to the tradition of research, primarily conducted

by psychologists, concerned with evaluating (1). We should note, however, before closing this sec-

tion, that since empirical work on the Sapir-Whorf hypothesis has been restricted essentially to the

domain of color, the above conclusions are correspondingly restricted. There are other areas of

human thought and belief--religion is an obvious example--in which constraints like those imposed

by peripheral neural mechanisms on possible color classiﬁcations seem a priori unlikely to operate.

Such domains therefore offer greater potential scope for application of notions like (2) and (3).

The basic research paradigm in the evaluation of hypothesis (1) has involved establishing a

correlation between a linguistic and a non-linguistic, cognitive variable within a single language.

Among the deﬁning studies in this tradition of research are those of Brown and Lenneberg (1954),

Burnham and Clark (1955), Lenneberg (1961), Lantz and Stefﬂre (1964), and Stefﬂre, Castillo and

Morely (1966). The linguistic variable has come in two forms, ‘codability’ and ‘communication

accuracy.’ The non-linguistic variable has been ‘memorability.’ Support for the Sapir-Whorf

hypothesis was claimed when a correlation was found between the memorability of a color and its

value on one of the linguistic variables. Rosch (Heider 1972; Heider and Olivier 1972) has shown,

however, that focality or perceptual salience universally determines both memorability and

codability/communication accuracy. Prior to Rosch’s work, codability and communication accuracy

were assumed to differ across languages and once this is shown to be false the correlation obtaining

-5-

between memorability and a linguistic variable no longer gives support to the Whorﬁan thesis. (See

Brown 1976 for a lucid exposition of this and related points, especially pp. 113f, 149.)

Recently, Lucy and Shweder (1979) have performed experiments that they interpret as reinstat-

ing the Whorﬁan claims made by the early studies. Lucy and Shweder succeed in developing a

stimulus array that exhibits a communication-accuracy/memorability correlation independent of

focality. However, their way of doing this is to begin with a standard stimulus array and selectively

substitute chips into this array until the result has the desired property. Thus, Lucy and Shweder

claim to have shown that a stimulus array can be constructed in which there is a communication-

accuracy/memorability correlation independent of focality.

What such a correlation may mean is, however, unclear. ‘Communication accuracy’ is esta-

blished with two groups of subjects. Members of the ﬁrst group are shown an array of colors and

instructed to name each color so that another person could pick the color out. The second group is

given the array of colors and the set of names produced by the ﬁrst group and asked to pick out the

colors intended by the various names. Colors that are more often picked out correctly are deﬁned as

having higher communication accuracy. Communication accuracy is therefore a function not only of

the code embodied in the language used but also of the particular array of contrasts comprised by the

stimulus set. For example, a focal green stimulus will be highly codable as green in any array.

However, its communication accuracy score will be high in an array lacking other very green chips

but low in an array that contains other very green chips. Since the Lucy-Shweder stimulus array is

purposely constructed by successive approximations until it yields a (focality-independent)

communication-accuracy/memorability correlation, it is unclear what meaning the Lucy-Shweder

correlation may have outside of the particular array of colors that was deliberately constructed to

produce it.

However one judges the claims and counterclaims in this research tradition, it seems clear that

any within-language correlational study, matching a linguistic variable purporting to measure the

degree to which the language ‘favors’ a color to a nonlinguistic cognitive variable like memorability,

is at best an indirect test of (1). Any such study depends on quite tenuous inference to connect the

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: