Linguistic Relativities

Language diversity and modern thought

John Leavitt
Universite de Montreal

CAMBRIDGE
UNIVERSITY PRESS

164

Linguistic Relativities

series of criticisms of Whorf's Hopi data and his analysis of it, I reserve its discussion for the next chapter.

9
Conclusion
The Neohumboldtian school has sometimes (e.g., Miller 1968; Steiner 1975)
been presented as a European twin of Boasian ethnolinguistics - or rather,
it has been assumed that the latter represented an American version of
Humboldtian essentialism. While this reading is understandable, it fails to recognize that the two schools took Humboldt's legacy in opposite directions.
The Neohumboldtians started with the idea of a necessary and normal unity
of language, thought, and culture, and pushed it to its limits: valorization of a
single language and virtual abandonment of the equally Humboldtian project
of large-scale comparison. While the Boasian school hearkens back to Geists
and Volks, it draws the opposite moral. Its commitment to contrastive comparison and to the critique of modern Western values forced the abandonment of
the unity of language-thought-culture.
The Neohumboldtians show what real twentieth-century linguistic essentialism looks like; by contrast, they show how different the Boasian project is.
The comparison, as we will see, suggests that many of the attacks against the
"Sapir-Whorfhypothesis" would in fact be directed more appropriately against
the Neohumboldtians precisely for the ways in which they differ from Sapir
and Whorf. We might even be justified in speaking of a Right-Humboldtianism
and a Left-Humboldtianism (cf. the Right-Hegelianism and Left-Hegelianism
of the 1840s), except that the Boasian problematic represents a serious enough
shift away from some of Humboldt's positions to put this into question.
While these kinds of essentialist models, like anything else, can be abused,
in some cases a great variety of phenomena do seem to present themselves as
diverse expressions of a single essence. At the same time, it is unlikely that
most human languages or cultures or societies can be grasped adequately in
this way. It is equally unlikely that a human language or culture or society
can be adequately understood using the great alternative to essentialism, linear
causal arguments.
In the face of the overwhelming dominance of universalist modes of
explanation, Boas and company are often drawing on the evident alternative,
essentialist phraseology, to express what are in fact new perceived patterns.

The rise of cognition and the repression

of languages

Sapir died in 1939, Whorf in 1941, Boas in 1942. Dorothy Lee published some
papers on language and world view after this time, as did Harry Hoijer (19041976), who worked primarily on Athabaskan languages, notably Navajo (discussion of his work in Lucy 1992a: 75-82). A collection of Sapir's papers was
published in book form in 1949, some ofWhorf's in 1952, with the definitive
volume of his papers appearing in 1956. But that was pretty much it. From the
1950s on, the scholarly world would be going in a direction hostile to relativist
ideas.
The cognitive revolution
Much of American academic thinking from the 1920s on accepted the behaviorist assumption that the mind was an inaccessible black box, and therefore
that the only thing one could study scientifically was externally observable
behavior. In linguistics, the shift in emphasis from Sapir's culturally connected
and open linguistics to the much more limited descriptive analysis of his Yale
colleague Leonard Bloomfield (1887-1949) took place the 1930s. Bloomfield
had experienced a conversion to behaviorism: language was to be redefined as
verbal behavior (Bloomfield 1933). He and the several generations of American
linguists who followed him sought to create a science of pure description. One
result was that linguistics as a separate field grew closer to psychology and
farther away from anthropology, which maintained an interest in questions of
meaning and context.
With the rise of computing from the 1950s there appeared a new possibility: if we could successfully program a computer to do what humans do, in
any domain, then we can use our very programming as a hypothesis for what
is going on in the human mind. At least hypothetically, we could get a glimpse
inside that black box. This led to the phenomenal rise of new fields of cognitive
psychology and, more broadly, of the new family of disciplines that came to be
known as the cognitive sciences.
The analogy one has chosen to clarify an idea tends to develop a life of its
own: the servant can easily become the master. What are computers good at?
165

166

Linguistic Relativities

Computing and calculating. Not terribly surprisingly, the computer model reinforced a tendency to define human thought as a form of computing or "processing." Steven Pinker puts this well: "'the physical symbol system hypothesis' or
the 'computational' or 'representational' theory of mind ... is as fundamental
to cognitive science as the cell doctrine is to biology and plate tectonics is to
geology" (1994: 78).
Cognitive science is one of the great success stories of recent times and
has led to major discoveries about human thinking, particularly in the area
of child development. But the very universalist program that has proven so
effective also peripheralizes diversity. Whatever the discipline, cognitive science approaches have massively presumed that "cognition" or "thinking" is
a universal process, inherent, perhaps innate, in each individual human, and
that linguistic and cultural specificities are mere surface effects, foam on the
ocean of mind. Cognitive science "combines tools from psychology, computer
science, linguistics, philosophy, and neurobiology to explain the workings of
human intelligence" (Pinker 1994: 17) in a universalist toolbag that characteristically leaves anthropology out.
With his usual clarity, Pinker states the attitude of classical cognitive science
toward the question of language specifics.
A famous hypothesis, outlined by Whorf ... asserts that the categories and relations that
we use to understand the world come from our particular language, so that speakers of
different languages quite literally conceptualize the world in different ways. This is an
intriguing hypothesis, but virtually all modem cognitive scientists believe it is false.
(1990: 200)

On the next page, however, Pinker states the following, apparently seeing no
contradiction with what he has just said:
[L]earning different languages does require one at least to pay attention to different
aspects of the world ... [L]eaming a language does involve learning to categorize
the world in particular ways, even if only for the purpose of using language itself: an
English speaker must think of a table as being more like a board than a forest when
referring to it in a sentence. (1990: 201)
This is characteristic: one defines linguistic relativity in such an extreme way
as to make it seem obviously untrue; one is then free to acknowledge the reality
of the data at the heart of the idea of linguistic relativity - without, until quite
recently, proposing to do any serious research on these data.
Taking linguistic and cultural differences seriously would seem to put the
whole cognitive project in jeopardy, at least as this project was understood
by most practitioners, at least well into the 1990s. Some cognitive scientists
seem really haunted by the specter of relativism. Here is the philosopher
and psycholinguist Jerry Fodor, one of the major figures in conceptualizing
cognition:

Fl.".,.;I

The rise of cognition and the repression of languages

167

The thing is: I hate relativism. I hate relativism more than I hate anything else, excepting, maybe, fiberglass powerboats. More to the point, I think that relativism is very
probably false. What it overlooks, to put it briefly and crudely, is the fixed structure of
human nature. (Fodor 1985: 5)
There is something startling about hearing this late-twentieth-century American
scholar declaring that he hates relativism worse than, say, racism or Nazism
or mass murder or any other hateables. His hatred is, however, understandable: real relativism implies a total absence of rules and potentially a war of all
against all with might making right. Because of this, everyone, except perhaps
for some nihilists, poetes maudits, and sociopaths, has always avoided real
relativism. The only way to have a relative relativism has been to presume a
hierarchy of values, with some kinds of worlds superior to others - the essentialist option; or to seek ways to calibrate among worlds, the Boasian option,
as it is the Einsteinian one: relativity rather than relativism.
As we will see in this chapter, the rise of cognitive science was accompanied
by a restating of what came to be called the "Sapir-Whorf hypothesis" in the
most extreme terms. Three arguments came to the fore repeatedly:
Determinism. The Sapir-Whorf hypothesis says that the language you
speak, and nothing else, determines how you think and perceive. We have
already seen how false a characterization this is: the model the Boasians were
working from was only deterministic in cases of no effort, of habitual thought
or speaking. With enough effort, it is always possible to change your accent
or your ideas.
Hermeticism. The Sapir-Whorf hypothesis maintains that each language
is a sealed universe, expressing things that are inexpressible in another language. In such a view, translation would be impossible and Whorf's attempt
to render Hopi concepts in English an absurdity. In fact, the Boasians presumed, rather, that languages were not sealed worlds, but that they were to
some degree comparable to worlds, and that passing between them required
effort and alertness.
Both of these characterizations are used to set up a now classic article on
linguistic relativity by the psychologist Eleanor Rosch (1974):
Are we "trapped" by our language into holding a particular "world view"? Can we
never really understand or communicate with speakers of a language quite different
from our own because each language has molded the thought of its people into mutually
incomprehensible world views? Can we never get "beyond" language to experience the
world "directly"? Such issues develop from an extreme form of a position sometimes
known as "the Whorfian hypothesis" ... and called, more generally, the hypothesis of
"linguistic relativity." (Rosch 1974: 95)
Rosch begins the article noting how intuitively right the importance of language differences first seemed to her, then spends much of the rest of it attacking this initial intuition.

168

Linguistic Relativities

f:.
j'

Infinite variability. A third common characterization is that Boasian linguistics holds that, in Martin Joos's words, "languages can differ from each other
without limit and in unpredictable ways" (Joos 1966: 96). This would mean
that the identification of any language universal would disprove the approach.
In fact, the Boasians worked with the universals that were available to them these were mainly derived from psychology - but opposed what they saw as
the unfounded imposition of false universals that in fact reflected only modern
Western prejudices. Joos's hostile formulation has been cited repeatedly as if it
were official Boasian doctrine (see Hymes and Fought 1981: 57).
For over fifty years, these three assertions have largely defined the received
understanding of linguistic relativity. Anyone who has participated in discussions and/or arguments about the "Whorfian hypothesis" has heard them over
and over again.

New universalisms in linguistics

The late 1950s and early 1960s saw the rise of major new orientations in linguistics. The best known of these is Noam Chomsky's rethinking of syntax.
For Chomsky, the miracle that requires investigation is that any speaking
human can produce an infinite number of sentences out of the limited means
of a given language. This infinite productivity is a matter of form, of syntax,
not of semantic content, and it holds for speakers of any language. What is of
interest, then, is the language faculty in itself, not the specifics of any particular
language. Chomsky himself sees the deep structures of language and thought
as innate in the human mind, and so, appropriately, finds precursors in the
Cartesian linguistics of the seventeenth century (Chomsky 1966).
Equally important for the development oflinguistics, however, was a rethinking of the question of language universals, initially in the work of Joseph
Greenberg (1915-2001). Greenberg compared surface structures in a vast range
of languages, and found a hitherto unnoted series of implicational universals: if
a language has a given feature (say object-verb sentence structure) it is statistically more than likely to possess some others as well (in this case, for instance,
postpositions rather than prepositions). Greenberg's findings led to a whole new
kind of linguistic typology, now closely bound with the study of language universals (see, for instance, Comrie 1989; Haspelmath et al. 2001), which has been
one of the most important growth areas in linguistics since the 1960s. Effort
has concentrated on finding reasons for the co-occurrence of features in many
languages. The overall goal is the universalist one of identifying "general principles governing variation among languages" (Comrie 2001: 25). Explanations
have largely been psychological or experiential- general aspects of human cognition or experience are looked to for explanations of commonly occurring correlations among features. In John Lucy's (1992a: 88) characterization,

The rise of cognition and the repression of languages

169

[t]he basic argument has been that the observed commonalities among languages
demonstrates the existence of common cognitive and semiotic processes at work in
all known languages. A marked relativity in conceptual world view simply cannot be
sustained, the argument goes, in the face of this evidence.

If Chomsky's transformational grammar represents a return to a locus of

explanation in the universal human mind, which is to say a new rationalism,
much of the new typology represents a return of a locus of explanation in an
equally universal human experience. In this "implicitly reductionist view,"
[g]rammar is derived from (a conventionalization of) semantics, as semantics is from
pragmatics (or discourse), as pragmatics is from "reality," as - for some people - such
reality is from physical reality. Such an epistemic hierarchization is predicated on a
radical empiricism. (Friedrich 1985: 186)
The implications of these new linguistic models for language diversity are
clear, whether you take an innatist or evolutionary view (Kenneally 2007) or
the view that language reflects general human experience. For all of these,
compared to the importance of language universals, the differences are so
minor as to be of little interest in themselves.

Psychologists and linguistic anthropologists kill Whorf again

In the 1950s, anthropologists and psychologists were interested in experimentation and the testing of hypotheses on what was taken to be the model of the
natural sciences. At a conference on language in culture, Harry Hoijer (1954)
first named a Sapir-Whorf hypothesis that language influences thought.
To call something a hypothesis is to propose to test it, presumably using
experimental methods. This task was taken on primarily by psychologists.
A number of attempts were made to prove or disprove experimentally that
language influences thought (see Lucy 1992a: 127-78; P. Brown 2006). Both
"language" and "thought" were narrowed down to make them more amenable
to experiment: the aspect of language chosen was usually the lexicon, presumably the easiest aspect to control in an experimental setting; thought was interpreted to mean perceptual discrimination and cognitive processing, aspects
of thinking that psychologists were comfortable testing for. Eric Lenneberg
defined the problem posed by the "Sapir-Whorf hypothesis" as that 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?" (1953: 463). Need I recall that Boas, Sapir, and Whorf went
out of their way to deny that different languages were likely to be correlated
with strengths and weaknesses in cognitive processes, i.e., in what someone is
capable of thinking, as opposed to the contents of habitual cognition?

170

Linguistic Relativities

If language influences thought, Lenneberg's argument went, speakers of

languages with many terms for a given domain should be able to make finer
perceptual distinctions in that domain than speakers of languages with few
terms. In the most influential studies, the domain chosen was that of color, one
that had interested an essentialist tradition for centuries, but that had not been
central to the Boasians. Some post-Boasian linguists had turned to colors as
a domain illustrative of the great variation in construal of the world through
vocabulary. The view that color classifications are unlimited and arbitrary is
clearly stated in Bloomfield's Language:
Physicists view the color-spectrum as a continuous scale ... but languages mark off different parts ofthis scale quite arbitrarily and without precise limits. (Bloomfield 1933: 140)

This view was carried on as part of American linguistic lore for the next
several decades. As the anthropologist Verne Ray put it, "[t]here is no such
thing as a 'natural' division of the spectrum. Each culture has taken the spectral continuum and has divided it into units on a quite arbitrary basis" (1952:
258). In what came to be a standard manual of the descriptivist school, H. A.
Gleason uses the diversity of divisions of a continuous spectrum of colors as
the initial example of how language works (1961: 4-5).
Initial experimental findings suggested that speakers of English and some
other languages found it easier to remember colors for which they had words
(e.g., Brown and Lenneberg 1954 [1970]). This was taken to mean that
"codability" enhanced memory, an apparent effect of language on cognitive
processing. Yet by the 1960s there had been little follow-up on actual psychological testing of the implications of vocabulary differences and a notable
lack of breakthrough in the search for "Whorf effects" on perceptual or cognitive processes. Further studies on color seemed, on the contrary, to deny the
importance of language differences. In the late 1960s the linguistic anthropologists Brent Berlin and Paul Kay carried out a survey of languages, shifting
the focus from color language in general to what they called basic color terms,
that is, terms that refer primarily to what we call colors rather than to anything
else ("the terms in each language which have specialized reference to what
English-speaking researchers would call colors," Lucy 1992a: 177). "Red" is
a basic color term, "salmon" is not. If we restrict the acceptable data to such
terms, then some languages have more than others; some have as few as two.
Berlin and Kay started by rephrasing Sapir and Whorf as saying that the
search for semantic universals was "fruitless in principle" because "each language is semantically arbitrary relative to every other language" (1969: 2; cf.
Lucy 1992a: 177-81). If this is what we are calling linguistic relativity, then if
any domain of experience, such as color, is identified in recognizably the same
way in different languages, linguistic relativity must be wrong. As we have seen,
this fits the arguments of Weisgerber and Bloomfield, but not of Sapir or Whorf.

The rise of cognition and the repression of languages

171

What Berlin and Kay found was anything but arbitrary. First, while boundaries between examples associated with a given term could be very fluid, there
was a tendency to agree cross-culturally on focal "centers," i.e., on which red
was the re~dest red. These were also the quickest hues to be recognized and
the most easily remembered. Second, they discovered a uniform attribution of
focal colors depending on the total number of color terms: if there were two
basic color terms the best representatives of them were what we call black and
white, when there was a third term it was best represented by what we call red,
and so forth, allowing the construction of a universal color hierarchy. Beyond
this, they found unexpected support for models of cultural and linguistic evolution: languages spoken by people in small-scale, low-tech societies had few
basic color terms, which, in an apparent correlation with cultural evolution,
increased to the eleven found in English. So not only was there no relativity
effect; on the contrary, color terminology seemed to confirm our own feelings
of cultural advancement.
A few years later Eleanor Rosch Heider (1972) performed a set of tests to
see whether focality, rather than codability, could explain the fact that English
speakers had an easier time with colors for which they had names. She compared English-speaking subjects with speakers of a New Guinea language with
four basic color terms. The result: virtually no difference in color recognition
and memory, regardless of the number of basic terms available. In other words,
the important factor was the universal fact of focality rather than the diverse
facts of vocabulary. "Rosch's finding were seized upon by advocates of universality, who said terminology doesn't affect cognition: color transcends culture"
(Olin 2(03).
The work on color was one contribution to a movement that gained increasing momentum from the 1960s through the 1980s. While new discoveries were
being made on language universals, even work directly bearing on language
diversity seemed to show that it simply didn't matter. Summing up generations
of breakthroughs in the study of what all languages have in common, Pinker
(1994: 59-67) concludes that while language differences can lead to "interesting anecdotes," these are ultimately "canards" or "bunk," of which the supreme
example is the Whorfian "hypothesis of linguistic determinism."
Some apparent support for linguistic relativity came from a consideration
of grammar rather than vocabulary. In 1981, the psycholinguist Alfred Bloom
published a comparison of the comprehension of counterfactual st~tements
among English- and Chinese-speakers ambitiously entitled The Linguistic
Shaping of Thought. Where English has inflections that mark counterfactuals,
Chinese does not, and Bloom felt that this was correlated with the lower degree
of success in the task among the speakers of Chinese. Let us note first of all that
Bloom's experiment deals with "thought processing," not the construal of the
world central to Boasians (DeBernardi 1997). And his findings were quickly

172

Linguistic Relativities

challenged: in a debate in the journal Cognition, the psychologists Terry KitFong Au and Lisa Garbern Liu were able to show that a rephrasing of the test
statements into more idiomatic Chinese virtually erased the difference between
the two groups. Thus this sequence of arguments, too, came to be taken as yet
more evidence against linguistic relativity (e.g., Pinker 1994: 66-7).
A characteristic study was reported recently in my own university's in-house
newspaper under the title "Language and Perception Are Not Connected"
(Baril 2(04). The article starts by saying that according to the "Whorf-Sapir
hypothesis ... language determines perception," and therefore that "we should
not be able to distinguish differences among similar tastes if we do not possess words for expressing their nuances, since it is language that constructs the
mode of thought and its concepts ... According to this hypothesis, every language projects onto its speakers a system of categories through which they see
and interpret the world." The hypothesis, we are told, has been "disconfirmed
since the 1970s" by research on color. The article reports on the research of
Dominic Charbonneau, a graduate student in psychology. Intrigued by recent
French tests in which professional sommeliers, with their elaborate vocabulary, did no better than regular ignoramuses in distinguishing among wines,
Charbonneau carried out his own experiment on coffee - this is, after all, a
French-speaking university, and we take coffee seriously. Francophone students
were asked to distinguish among different coffees; like most of us, they had a
minimal vocabulary for distinguishing them (words like "strong," "smooth,"
"dishwater"). The participants made quite fine distinctions among the eighteen
coffees served, well above the possible results of chance, showing that taste
discrimination does not depend on vocabulary. Conclusion: "Concepts must
be independent of language, which once again disconfirms the Sapir-Whorf
hypothesis" (my italics). And this of course would be true if there were such a
hypothesis, if it was primarily about vocabulary, and if it said that vocabulary
determines perception.

The Boasians and color

We have seen that Bloomfield and his successors in linguistics maintained the
unlimited arbitrariness of color classifications, and so could have served as
easy straw men for the cognitivist return to universals. But what did Boas,
Sapir, Whorf, or Lee actually have to say about color? Did they in fact claim
that color perception or recognition or memory was determined by vocabulary? Sapir and Lee are easy: as far as I have been able to ascertain, neither
one of them talked about color at all. Steven Pinker attributes a relativist and
determinist view of color classifications to Whorf:
Among Whorf's "kaleidoscopic flux of impressions," color is surely the most eyecatching. He noted that we see objects in different hues, depending on the wavelengths

The rise of cognition and the repression of languages

173

of the light they reflect, but that the wavelength is a continuous dimension with nothing
delineating red, yellow, green, blue, and so on. Languages differ in their inventory of
color words ... You can fill in the rest of the argument. It is language that puts the frets
in the spectrum. (Pinker 1994: 61-2)

No he didn't. Whorf never noted anything like this in any of his published
work, and Pinker gives no indication of having gone through Whorf's unpublished papers. As far as I can ascertain, Whorf talks about color in two places;
in both he is saying the opposite of what Pinker says he is saying. In one
instance (1956: 209) Whorf makes the not very startling point that if a group
of people could see only the color blue, they wouldn't develop much of a color
vocabulary. In other words, the existence of a lexical domain depends on that
of an enabling physiology. In the other (1956: 163), he is talking about "color
blindness and unequal sensitivity to colors" as exceptions to the universal perceptual realities discovered by Gestalt psychology. This passage is to be found
on the page on which Whorf looks to psychology to offer a "canon of reference
for all observers."
The only mention of the word "hue" in Whorf's published work is a discussion of how the pronunciation hyuw illustrates his analysis of English phonology (1956: 227).
In fact, whenever Whorf or Boas talked about color perception or discrimination, they did so with due regard to non-linguistic factors. Boas's doctoral dissertation of 1881 ("Contributions to the Understanding of the Color of Water")
had been a study of the absorption of light in water, and in doing his experiments Boas was struck by "the difficulty of judging the relative intensities of
two lights that differed slightly in color" (Stocking 1968: 142). Boas dealt with
the problem with a psychophysical definition of thresholds below which discrimination is impossible. One effect of this for Boas, as he reminisced almost
sixty years later, was to convince him "that there are domains of our experience
in which the concepts of quantity ... are not applicable" (1939: 20). We can see
how this early work was leading toward greater interest in the mind as a factor
in perception; but there is no hint here of linguistic or cultural determinism.
Boas also discusses color in the article on alternating sounds. And it's a
highly instructive passage. Boas has clearly distinguished perception from
apperception and categorization. He is talking about psychological experiments in distinguishing between sensations in different modalities. "As this is
the most important part of our considerations," he writes,
we will illustrate it by a few examples. It is well known that many languages lack a
term for green. If we show an individual speaking such a language a series of green
worsteds, he will call part of them yellow, another part blue, the limit of both divisions
being doubtful. Certain colors he will classify to-day as yellow, to-morrow as blue. He
apperceives green by means of yellow and blue. We apperceive odors in the same way,
and classify new odors with those to which they are similar. (1974: 74-5)

174

Linguistic Relativities

But is this or is it not an argument for the linguistic determination of perception

or discrimination? It is not. He continues:
It will be understood that I do not mean to say that such sensations are not recognized

in their individuality, but they are classified according to their similarity, and the classification is made according to known sensations.
In other words, vocabulary may playa role in categorization but should not be
expected to affect perceptual discrimination.
Color comes up again in Boas's address at the 1909 Clark University conference on psychology. Speaking to an audience that included Freud, C. G. Jung,
and some of the founders of modern experimental psychology, Boas presented
his view of a specifically cultural and linguistic unconscious, using color as an
example of the diversity of categorization and conceptualization:
Differences [in categorization] ... appear very clearly in the domain of certain simple
sense-perceptions. For instance, it has been observed that colors are classified according
to their similarities in quite distinct groups without any accompanying difference in the
ability to differentiate shades of color. What we call green and blue are often combined
under some such term as "gall-like color," or yellow and green are combined into one
concept, which may be named "young-leaves color." The importance of the fact that
in thought and in speech these color-names convey the impression of quite different
groups of sensations can hardly be over-rated. (1974: 249)
Here the differences in classifying colors are seen as important precisely
because they do not have a direct correlation with "the ability to differentiate
shades of color," something that is presumably part of the human heritage as
such.
Finally, in a comment parallel to his argument about numbers, Boas noted
that American languages have no trouble distinguishing colors - but that they
do so by naming objects of the given color, without fixing a single standard
(cited in Woodworth 1910: 327-8). As all of these examples show, the Boasians
always accepted, indeed insisted on, the autonomy of perception. They were
never linguistic determinists. 1
There is, however, a school of thought that did claim that vocabulary determines perception: not surprisingly, the Neohumboldtians. It is Weisgerber who
says that vocabulary determines the perception of color (e.g., 1950: 141): we
cannot translate color experience from one language to another. "The vocabularies of taste and of smell in various languages reveal the same kind of translation difficulty and can be studied and explained intelligently only by the
neo-Humboldtian assumption" (Basilius 1952: 102).

Reseeing colors
By the early 1960s, there was plenty of ethnographic data on a variety of
systems of naming what we call colors from different parts of the world.

The rise of cognition and the repression of languages

175

What stands out is the variety of dimensions used to categorize visual experience. An exemplary study was that of color language among the Hanun60
of the Philippines, carried out by Harold Conklin (1955 [1964]), one of the
eminent practitioners of ethnosemantics (see Chapter 10 below). By starting
with language and its uses in real situations, rather than with a presumed
universal domain of "color," Conklin shows that in Hanun60, the four "basic"
categories can best be glossed as dark versus light and dry versus wet. For
the Hanun60, in other words, dryness and wetness are essential qualities in
defining a lexical domain that overlaps largely with what we call "color."
This kind of situation, in which a language defines semantic domains in a
way that only partially overlaps with ours, is not rare; there is nothing surprising about it unless we assert that our own semantic domains must be
universal and necessary ones.
This opens up the fundamental criticism of most of the work on color
terms, articulated forcefully by John Lucy (1997a). "Color" itself is a Western
category. The words that many languages use to refer to what we call colors
also, and sometimes primarily, index what we consider non-color qualities
such as shininess or wetness or smoothness. An experiment that offers people
an array of samples distinguished solely by our idea of "color" is only testing how easy it is for them to reconstruct what from their point of view is
a perfectly arbitrary set. If general patterns emerge, this is an artifact of the
proposed task.
[W]hat about the success of the approach? After all ... it works! These color systems
are there! ... I would argue that what is there is a view of the world's languages through
the lens of our own category, namely, a systematic sorting of each language's vocabulary by reference to how, and how well, it matches our own. This approach might well
be called the radical universalist position2 since it not only seeks universals, but sets up
a procedure which guarantees both their discovery and their form ... No matter how
much we pretend that this procedure is neutral or objective, it is not. The procedure
strictly limits each speaker by rigidly defining what will be labeled, which labels will
count, and how they will be interpreted ... Is it any wonder, really, that all the world's
languages look remarkably similar in their treatment of color and that our system represents the telos of evolution? (1997a: 331-2)

If a language takes shininess as a defining quality along with hue, this fact
will simply disappear:
[A] whole level of analysis is missing from the basic color term tradition, namely, no
attention whatsoever is paid to what the various terms actually mean in the sense of
what they typically refer to, their characteristic referential range. Yet somehow a tradition that ignores these issues is supposed to provide a way of discovering semantic
universals. (p. 335)
The obviousness and assumed universality of what we call color is an example of what Bachelard called an epistemological obstacle (see Chapter 10). The
idea of "pure" color is reinforced for us, in our society, by the pervasiveness of

176

Linguistic Relativities

paint and crayon technology. Steven Pinker gives us the Crayola analogy for the
role of focal colors:
[H]umans the world over (and babies and monkeys, for that matter [Query: Aren't
babies human?]) color their perceptual worlds using the same palette, and this constrains the color vocabularies they develop. Although languages may disagree about
the wrappers in the sixty-four crayon box ... they agree much more on the wrappers in
the eight-crayon box - the fire-engine reds, grass greens, lemon yellows. Speakers of
different languages unanimously pick these shades as the best examples of their color
words, as long as the language has a color word in that general part of the spectrum ...
Languages are organized a bit like the Crayola product line, the fancier ones adding
colors to the more basic ones. (Pinker 1994: 62-3)

The history of humankind in a couple of crayon boxes. What could be more

obvious to us, and at the same time more homey and normal and comforting,
than a box of crayons, the childhood tools of pure color, which are an easy
pleasure to use precisely because they have already been designed to exclude
all the hard-to-draw complexities - wetness, ripeness, roughness, shine - of
real visual experience? You could hardly find a more seductive or misleading
analogy for the actual visual categories coded in the languages of the world.
Replacing whole language systems with basic color terms is comparable to
the evolutionists' replacement of whole languages with numerals, which we
could just as easily call basic number terms: "like the fingers on one hand" is
not a basic number term, while "five" is. With both numbers and colors we find
an increase in the number of terms in societies whose members are required to
handle more elaborate technologies. This says nothing about the sophistication
of thought processes in general, but a lot about the necessity in some circumstances of having an array of easily transposable terms that can be abstracted
from actual situations. To paraphrase Boas's reply to the evolutionists on numbers: the way people live in some societies means that they do not need very
many basic color terms. If their way of living changes they pick up or invent
new ones; nor, as Boas actually said about color, should we expect the number
of terms to affect perception as opposed to common habitual classification.
None of this tells us very much about cognition or about how different peoples
construe the world.
It may be answered that the real discovery made by Berlin and Kay was not
simply that basic color terms increase as people require them to, but that they
always do so following the same pattern, one presumably based not on any
language, but on the nature of the human eye and brain. I would not dispute
this. But I would point out that here, too, the parallel to number vocabulary is
very telling. Number terms, too, increase as people require them to, and they,
too, always do so following a single pattern: always from one, two, many, to
one, two, three, many, to one, two, three, four, many, and so forth; nowhere do
we find a counting system of four, seven, one, three, nine, many, or even of one,

The rise of cognition and the repression of languages

177

three, six, many. The development of number terms, then, also seems based
on an extralinguistic reality. The difference is that in the case of numbers this
fact is so glaringly obvious as to be invisible, while in the case of colors it had
to be discovered. But the principle is exactly the same. In both cases, people
develop more extensive terminological arrays when the need arises, and in both
cases the pattern of development depends on non-linguistic factors: the facts of
human physiology and perceived salience in the case of color terms, the facts
of ordinal sequence in the case of number terms.
The philosophers kill Whorf again

During the period of the rise of cognitive science some philosophers were discovering "the Whorfian hypothesis," and they did not like it. The philosophically minded sociologist Lewis Feuer (1953) says that Whorf's theory is based
on the comparison of vocabularies between languages: he gives the example
of Inuktitut words for snow. Feuer points out, as Boas had, but without citing
him, that we must expect people to develop an elaborate vocabulary on a topic
that is of great interest to them. A distinctive vocabulary would not, then, be a
feature that defines a language, but a mere result of circumstance. Since Feuer
has already presented linguistic relativity as being entirely about vocabulary,
this view would logically eliminate the possibility of linguistic relativity. Feuer
goes on to say that "[t]he 'principle of linguistic relativity' argues that there
are incommensurable cultural universes. An incommensurable cultural universe would be an unknown one. The fact of linguistic communication, the
fact of translation, belies the doctrine of relativity" (p. 95). It's as simple as
that: the fact that we can translate at all, that is, that we can use one language to
denote referential information expressed in another - exactly as the Boasians
said we should be able to do - belies the Boasians' own doctrine. How could
the Boasians have missed this for forty years?
Feuer's arguments were picked up by Max Black (1959 [1962]), who presents
Whorf as believing that the "real world" is totally unstructured, and that any
apparent structure it has is imposed on it by language. If this were true, translation
would be impossible, and Whorf's own effort to render the Hopi world in English
would be a non-sense (arguments repeated in at least one general introduction to
semantics: Palmer 1976: 57). Black (p. 255) consequently cites the "amateurish
crudity" of Whorf's project (on Feuer and Black, see de ForneI2(02).
In a 1970 article, John McNamara again attributes a kind of linguistic solipsism to what he calls "a strong Whorfian hypothesis," proceeds to repeat
the translation argument, and extrapolates that according to Whorf a bilingual
person would be unable to communicate with him- or herself: "in switching to
language B, he would never be able to understand or explain what he had just
communicated in language A" (paraphrased in House 2000: 78).

178

Linguistic Relativities

Donald Davidson (1974) presents the Whorf hypothesis as representing

"conceptual relativism" and linguistic determinism, and as claiming that different languages cannot be calibrated: "Whorf, wanting to demonstrate that Hopi
incorporates a metaphysics so alien to ours that Hopi and English cannot, as he
puts it, 'be calibrated,' uses English to convey the contents of sample Hopi sentences" (p. 6). Davidson gives no page reference for his apparent citation, just
the name of one of Whorf's papers - not, as it happens, one in which the word
"calibrated" appears. In the place where Whorf does talk about calibration,
the passage that we cited on a new principle of relativity (Whorf 1956: 214),
he says the opposite of this: that "speakers of different languages will not be
led by the same physical evidence to the same picture of the universe unless
their linguistic backgrounds are similar or can somehow be calibrated" (my
emphasis). Far from contradicting himself by writing about Hopi in English to
prove that the two cannot be calibrated, Whorf's efforts are attempts to calibrate them.
Davidson goes on to equate difference in world view or conceptual scheme
with untranslatability: "We may identify conceptual schemes with languages,
then, or better, allowing for the possibility that more than one language may
express the same scheme, sets of intertranslatable languages" (p. 185). If it is
possible to translate from one language to another, then the two must form a
"set of intertranslatable languages," and their conceptual schemes must be the
same. Given the universal possibility of translation, Davidson concludes that
all languages must fit in a single set, and therefore by definition the very idea
of conceptual scheme is of no use. It is worth setting this oft-repeated claim
that the possibility of translation proves the irrelevance of language differences against the reverse claim, drawn by Theodora Bynon from the writings
of Weisgerber, that any difficulty in translation proves that languages represent
different worlds. "If in fact words in different languages were simply referring to 'the same objective reality' there would be no translation problems, no
so-called untranslatable words, and the distribution of 'words and objects' ...
would form neat patterns" (Bynon 1966: 472, based on Weisgerber 1963). But
both complete translatability and radical non-translatability are fantasies. In
reality, as translators have found for millennia and translation theorists have
said for centuries, translation from any human language to any other is always
possible but always problematic; it depends on what aspects of the original you
are trying to convey and how much you are willing to burden your translation
in order to do so. A constant shift in point of view, a dual or multiple perspective implying a kind of practical linguistic relativity, seems to be a prerequisite
for the very act of translation (Becker 1995).
It is certainly unfair to present these philosophers solely in terms of their
misreadings of the texts. The point they want to make is fair enough: it is that
thought and experience derive from more than language alone; that the human

The rise of cognition and the repression of languages

179

brain, the rest of the human body, the world, and social life all influence
thought in a way not directly dependent on language, and indeed also influence language.
These critiques - the questions of vocabulary, of sealed language-worlds, of
translation, and of a social influence on thought - would probably hit the mark
if they were aimed at the Neohumboldtians, but it is not at all clear that they
touch the Boasians.

The Neohumboldtians visit the Hopi and kill Whorl again

One ofWhorf's most startling claims is that the grammar of Hopi suggests a
different construction of time and space than do those of Western European
languages. He accepts that there is a universal human experience of time as
duration and succession of events, but asserts that this in itself is no warrant for the modern Western conceptualization of time as divided into three
"spaces" of past, present, and future, with the subject constantly moving at
an unchanging pace from the past through the present into the future. Whorf
correlates this conceptual model, which is Kant and Newton's model, with
some pervasive traits of modern Western languages, notably the importance
of tripartite tense categorization in the verb. As we have seen, Hopi has two
"forms" that cover the experiential material Western languages divide into
tenses: experiences are divided into the "manifested" (the unmarked form of
the verb), which includes both what we would put into the past and the objective, already-there aspects of what we would call the present; and the "manifesting" (marked with the suffix -ni), which includes the transformative and
volitional aspects of the present and all of what we put in the future. Whorf
concludes that this distinction implies a very different conceptualization of
the world and its processes from the three modern Western temporal "spaces."
While this presentation does not take account of other aspects of Hopi grammar which Whorf saw as integral to his total argument (Whorf 1956; P. Lee
1996), this point would be the center of controversy for decades.
In an ironic twist, Neohumboldtian linguistics itself intervened in a way
hostile to the idea of linguistic relativity precisely on this issue of time in
Hopi.
Helmut Gipper was happy with Whorf's overall characterization of the relationship between language and thought but disturbed by what he took to be
some of his extreme formulations. Most shocking was the claim that Hopi had
no linguistic expressions for "time":
After long and careful study and analysis, the Hopi language is seen to contain no
words, grammatical forms, constructions or expressions that refer directly to what we
call "time" ... Hence, the Hopi language contains no reference to "time," either explicit
or implicit. (Whorf 1956: 57-8)

180

Linguistic Relativities

This quote is from the paper "An American Indian Model of the Universe,"
which Whorf left unpublished during his lifetime, although it is one that has
given rise, as Gipper notes (1994: 283), to a popular idea of a "timeless Hopi
language." But to some degree, at least, this idea is based on a misreading. We
have to recognize (as does Lucy 1992a: 286) that Whorf does not write time,
but "what we call 'time'," doubly qualifying the word with scare quotes and
with a "what we call." How important are quotation marks? Absolutely essential here, since a few lines above and a few lines down Whorf defines what he
means by "what we call 'time''': it is the constructed, spatialized model of time
typical of the modern West, with the past conceived as a "space" that is behind
us, the future as an analogous space "in front" of us. Whorf makes it clear in
the same text that this construction is not to be confused with the universal
experience of temporal change, of it "always getting later."
In Gipper's case, however, there is a legitimate philosophical difference
here as well. It is crucial for Whorf's argument that one be able to distinguish general and immediate human experience that is largely bodily-based something that fits with Whorf's fondness for Gestalt psychology - and the
varying construals that languages and cultures can put upon this experience.
For a Kantian such as Gipper, diversity can only go so far "down": unidimensional flowing time and three-dimensional static space are necessary aspects
of human understanding. Elsewhere, in comparisons with ethology, Gipper
relates them to human bodiliness in the world (1963). If you share this view,
then to claim that the Hopi language implies a very different kind of time and
space from the Newtonian ones is to say that the Hopi are not really human,
or at best that the Hopi language lacks means to express some of the fundamentals of human experience. On the contrary, since Whorf does not share
this view, he is saying, to Kantians and Newtonians in particular (1956: 153),
that being human is more profoundly various than they allow, that, to abuse
Hamlet a bit, it includes "more heavens and earths than are dreamt of in your
philosophy."
Gipper was sufficiently curious and dedicated first to publish a detailed study
ofWhorf's terminology (1963), then to undertake several periods oflinguistic
field research in Hopi territory in Arizona. He found (1972: 213-35) that Hopi
possesses the means to refer to what we call past, present, and future, something he showed through German translation equivalents of Hopi sentences. He
also pointed out cases in which days and other temporal units can be counted,
and that months can be treated using the same kinds of numerotation as objects.
On what one would assume is the crucial point of verb forms, Gipper agrees
with Whorf that there are two, not three (1972: 223-4), but argues that these
are in fact based on the same tripartite division of time that everyone intuitively
possesses, except that instead of having separate forms for past, present, and
future, the Hopi have a present-past opposed to a future.

The rise of cognition and the repression of languages

181

Some of Gipper's students, most extensively Ekkehart Malotki, continued to

work on Hopi. Malotki conducted long-term field research and published on
the Hopi language of space (1979), then on time. In his 1983 book Hopi Time,
he documents temporality in Hopi and says that Whorf's claims for the timelessness of Hopi show that he did not have an adequate understanding of the
language. In nearly 700 pages, Malotki massively demonstrates Hopi usages,
in all parts of speech, that he says show the omnipresence of concepts of time.
He does this sometimes by challenging Whorf's characterizations directly,
more often by showing that the best English translation equivalent of a given
phrase involves past-pre sent-future distinctions.
The first page of Malotki's book carries only two quotes, one from Whorf
and one from Malotki's Hopi field notes. The Whorf quote is the one we have
already considered, saying that Hopi has no words for "time"; the field note,
a Hopi text with interlinear glosses and an English translation, appears to be
about nothing but time.
The only way to judge this juxtaposition is to take it philologically. I have
already pointed out that the distinction between the word time with and without quotation marks, between the word time used without qualification and
what Whorf calls "what we call 'time'," is crucial for understanding what
Whorf meant to say in this manuscript. But here, floating on the page, we have
Whorf apparently saying that Hopi lacks the means of expressing any kind of
temporality.
Here is the sentence that follows on the same page - the English translation
of a sentence from Hopi: "Then indeed, the following day, quite early in the
morning at the hour when people pray to the sun, around that time then he
woke up the girl again." See all the time words? How could Whorf have possibly missed all this?
In fact, this begs the question entirely. Whorf never said that the Hopi can't
or don't talk about time; he said that they don't conceptualize time in the same
way we do, and that language is a source of conceptualization, both theirs and
ours. No one would deny that the most appropriate translation of a given Hopi
sentence into normal English might involve English time words. The question
is whether the forms that convey given referential information in Hopi work the
same way as do English forms, conveying the same background assumptions
and, presumably, the same "metaphysics" of a spatialized past, present, and
future. These are empirical questions which I cannot answer. But Gipper's lists
of German translation glosses, Malotki's lists of English ones, do not answer
them either (cf. P. Lee 1996: 140-1).
While Gipper's book got relatively little attention among English-speaking
linguists, Malotki's was greeted with jubilation as putting the final nail in
Whorf's coffin, the proof that his upsetting claims were based on his incompetence as a linguist. In his review, Bernard Comrie (1984: 132), one of the

182

Linguistic Relativities

major figures in linguistic typology, wrote that "Malotki's presentation and

argumentation are devastating. At the end, one is left only with the question ...
how Whorf could have been led to make the generalisations about Hopi time
expression which he did make." Steven Pinker ( 1994: 63) cites Malotki to attest
to Whorf's "limited, badly analyzed sample of Hopi speech." Even anthropologists of a symbolic persuasion accepted that Malotki had demonstrated that
Whorf's presentation was fundamentally flawed or perverse (see Dinwoodie
2006: 329).
Not everyone has taken this line, however. Penny Lee (1991) offered a
defense of Whorf's analysis of what he calls tensors, which, she argued,
is consistently misrepresented by Malotki. David Dinwoodie (2006: 332)
points out that in many cases we do not know in what contexts Malotki
gathered his data, nor what came from whom in what situations, in terms
of age, activity, or knowledge of English. He then gives a detailed analysis
of Malotki's chapter on the ceremonial calendar (Malotki 1983: 451-80),
arguing that the examples given are either not explained or in fact fit into
Whorf's interpretation.
I am not a specialist of Hopi, of Uto-Aztecan, or of Amerindian languages.
But besides the questions Dinwoodie has raised, some other parts of Malotki's
book seem problematic simply based on the evidence he presents.
One of Whorf's claims is that while modern Western languages spatialize
time, talking, for instance, about durations as being "long" or "short," Hopi
does not (1956: 146). Malotki's counterclaim that Hopi does use spatial metaphors for time (1983: 13-27) seems to boil down to the fact that the same
word is used respectively for "here" and "now," for "there" and "then," and for
"yonder" and "a long time ago," plus a medial term. The terms, in other words,
seem to denote what we would call relative "proximity" in both space and time,
without distinguishing between the two. Simply on the evidence of meaning
and usage, one doesn't know whether these are spatial metaphors for time,
temporal metaphors for space, or, as Whorf himself suggests in the "Model"
manuscript (cf. P. Lee 1996: 100-1), space-time deictics.
Malotki, for his part, simply asserts the priority of space over time as a fact
of human psychology. When the actual linguistic evidence is not clear, he will
impose a solution based on a putative universal psychology - that is reflected
in, or perhaps reflects, modern Western languages. Whorf's approach is the
opposite: he relies on what the language seems to be saying and tries to reconcile his construction of the world to that. This is an imaginative construction,
but one that stays close to the data.
Another point in Whorf's presentation is that in Hopi time units are not
counted as if they were objects. Malotki's Chapter two, on units oftime, seems
to support Whorf more than to contradict him: while Malotki indicates some
units of time that can be counted, and that Whorf may have missed, these are

The rise of cognition and the repression of languages

183

isolated and do not seem to form anything like the pervasive pattern that is
found in modern Western languages.
But as we have seen, Whorf's central argument bears on the difference
between verb forms with and without the suffix -ni. This obligatory distinction
between what he calls "manifested" and "manifesting" is the main issue in
the "American Indian Model" manuscript, the paper that provoked the whole
Neohumboldtian attack on Whorf. This is one of the key arguments he uses to
claim a difference between Hopi and SAE constructions of time; it is the one
that immediately justifies his "extreme" position that Hopi has no forms that
correspond to "what we call 'time'." Surely his presentation of this part of the
language would have to be a major target, if not the major target, in an attack on
his interpretation of time in Hopi. Yet when Malotki comes to talk about these
forms, which he calls "the Hopi tense system," something odd happens. One
would think that a book about time in the Hopi language would devote a lot of
space/time to tense. But unlike the long discussions of other topics, with pages
of Hopi examples, here there is a presentation of four and a half pages, with
one of the pages a chart and the half page a citation. Not a single Hopi example
is given (in his review, Comrie insists on the large number of examples Malotki
offers, as against the few given by Whorf). And Malotki does not get to this
crux of all the arguments until page 622 of the book, in the last section of the
last substantive chapter (the next and last one is called "Conclusions"), which
bears the title "Miscellaneous."
Malotki's argument is simple:
The unidirectional flow of time out of a "before-now" or "already" via a "now" into
an "after-now" or "not-yet", often diagrammatically represented by a one-dimensional
time line drawn from left to right and featuring an arrow tip pointing into the future,
may actually mirror time reality as experienced by human beings; however, there is no
compelling or intrinsic reason to find this "natural" state of affairs reflected in the tense
categorization of a language. (1983: 624)
This universality and naturalness of a spatialized three-part conceptualization
is expressed in a diagram on page 626 which contrasts "the time system" of
past-pre sent-future ("the time system," not any particular time system) with
the "Hopi tense system" which Malotki calls nonfuture (unmarked), which
"generally allow[s]" past or present time interpretations, versus future (-ni).
In other words, it is possible, if you try hard enough, to find present, past, or
future in the referents of Hopi sentences, even though these are not the grammatical categories that are being marked.
Malotki begins his discussion by trying to define tense. He quotes from
John Lyons (1968: 305): "Many treatments of tense have been vitiated by
the assumption that the 'natural' division of time into 'past', 'present', and
'future' is necessarily reflected in language." Lyons puts "natural" into quotes;

184

Linguistic Relativities

he, in fact, is very careful not to say whether or not he thinks that "past,
present, future" is natural, only saying that it has often been suggested that
the directionality of time is given in "nature" (his scare quotes) and that "past,
present, future" in tense comes from the traditional grammars of Greek and
Latin (p. 304). What does Malotki himself think about three-part division? He
accepts it as natural, as characteristic of human experience before or beyond
language, adding that languages do not necessarily reflect it directly.
So we are back to arguments about naturalness that we thought we had left
behind at the end of the eighteenth century: some languages directly reflect
the natural state of affairs, others fall into inversion and require extra work.
Languages that codify time in a tripartite way (past-present-future) are natural,
while those that are different on this score are unnatural. Hierarchy has just
been reintroduced among the languages of the world.
Malotki asserts, on the evidence of his translation equivalents, that for -ni
"the temporal function is primary" - it's a future marker - but it also has
"secondary, atemporal functions." This is his way of recasting Whorf's
apparently uncontested observation that -ni indexes a form that includes both
what we call the future and some aspects of the present. The difference in
interpretation is due to a difference in method. Malotki starts off by separating out meanings into primary and secondary ones, with the primary one
corresponding to those we are familiar with. Whorf, on the contrary, starts
with the language itself, its indexed grammatical categories and the range of
their meanings, without defining some as primary and others as secondary
based on Western categories.
For Malotki, the two Hopi forms indicate the difference between a future
and an unmarked "nonfuture." On page 625 he shows how you can find a past
or present force for any unmarked verbs if you try. This is based on producing serviceable English translation equivalents and then asserting their priority
over the actual language forms of Hopi. "Perfective verbs generally admit a
past tense translation, imperfective verbs a present tense rendering, although
a past tense reading for the latter and a present tense reading for the former is
also possible under special syntactic and contextual conditions." Does Malotki
now go on to spell out these conditions and give dozens of examples, as he has
done for all kinds of bits of Hopi language for the last 600 pages? No, he does
not. Instead there is a reference: "For more specifics concerning the Hopi tense
situation ... see Voegelin et al. 1980: 582 [it's really 1979], who concur in the
view that Hopi sentences are not restricted to a minimum division of time into
future and nonfuture tenses."
The Voegelins' discussion does indeed question Whorf's formulations.
Their argument, like his, assumes the universality of time deixis in verbs
and the possibility of finding it in Hopi forms, although in the form of combined tense and modality. On the page cited, they note that the addition of

The rise of cognition and the repression of languages

185

non-obligatory elements "make possible a further distinction of the nonfuture

tense into present time and past time."
Malotki calls this a function of disambiguating present from past. Hopi, in
other words, is ambiguous, but it has tricks that allow disambiguation. What is
being disambiguated? The natural and "universal time system" that just happens to be our three-part division, but which is poorly represented by the Hopi
tense system.
Malotki goes on to refute claims of linguacentrism such as the one I have
just made by saying that in fact Whorf was also wrong in seeing Western
languages as expressing a three-part time system. The last half page of this
discussion is a quote - from whom is not clear - saying that English is really
a two-part system, since what is called the future tense is not morphological
but is built up out of other forms: since forms with "will" "are not strictly
part of the 'basic' tense system," English can only be said to have a past and
a non-past. So our feeling of having a three-part system is not linguistic, but
experiential.
The argument has a respectable history (Lyons 1968: 306; Chomsky, preface
to Schaff 1964 [1973]; see Alford 2002), but here it seems specious: it amounts
to saying that analytical grammatical forms are necessarily less clear semantically than synthetic ones, that "j'irai" is somehow a "real" future where "I will
go" is not, when in fact in both semantics and usage, both forms carry a clear
future sense, opposed both to present and past.
Malotki really does appear to believe, and deeply, in the universality of
our three-part system of imagining time. His idea of proving its pertinence
in this case is to show that it is possible to use contextual and other clues to
determine whether someone speaking Hopi is in fact referring to past, present,
or future, even though the language-forms themselves do not indicate this.
But we could as well argue that it is usually possible to infer whether someone speaking English, a language without grammatical evidentiality, is in fact
referring to something he or she experienced directly, knows by hearsay, or
is concluding based on evidence. Does this mean that English speakers are
secretly orienting themselves constantly in a world of evidentiality, but are
hiding it? Are all the grammatical categories that have been identified in every
language in the world to be treated as universal categories most of which most
languages just hide?
Near the end of his glowing review of Malotki's book, even Comrie finds
that he cannot swallow its argument about tense.
It is possible that individual claims made by Malotki may be subject to revision or
even rejection ... [H]is claim that Hopi has a tense system based on the opposition
between future and non-future ... strikes me as questionable: given the wide range
of modal uses of the so-called future, it is at least plausible that this is a modal rather
than a temporal distinction, with the result that Hopi would have no tense distinctions.

,
186

Linguistic Relativities

(It does, however, have a profuse system of aspectual distinctions, and the absence

of grammaticalisation of time reference as tense says nothing about the ability to

conceptualise location in time - one hesitates to ask what conclusions Whorf would
have drawn from the absence of gender as the grammaticalisation of sex in many
languages.)
The admission is necessary, while the accompanying parenthesis is both characteristically snide about Wharf and characteristically unjustified. Whorf was
talking about Hopi grammar and the forms it makes it easy to conceive, not
about whether Hopis had the "ability" to refer to what Comrie is calling "location in time," i.e., to distinguish events in temporal relation to each other and to
specify the temporal relationship between the event being referred to and the
situation of enunciation. Like any Boasian, Whorf would presume that speakers of any language have the ability to conceive and convey any referential
information, including information about what we call "time" and what we
call "sex."
The difference between Malotki's and Whorf's approaches to the same distinction (presence and absence of -ni in the verb) makes sense in terms of
overall philosophical divisons. Where Gipper looked to Kant for his universals, Malotki looks to psychologists; but their arguments are identical. Both
are believers in the universal, necessary quality of the three-part temporal distinction, which Malotki calls "the time system" without qualification. For a
Kantian, past-present-future "spaces" are simply the form of the human intuition of time. A given language need not, certainly, reflect this directly: those
that have three-part tense are, however, natural reflections of this intuition,
while those that do not must use other means, sometimes rather cumbersome
ones, to convey this universal presupposition. When Malotki asserts that "the
temporal function is primary" for the suffix -ni, even though it has other "secondary, atemparal functions," this is as much as to say that his method for
analyzing Hopi forms is to identify the universal semantic core conveyed by
each form (-ni is primarily temporal), then explain away usages that do not fit
this (-ni also has non-temporal functions).
For Whorf, by contrast, since the basic human experience of time is not
necessarily spatialized, there is no reason to assume that the three-part spatialized model of time separated from space that is typical of modern Western
languages should represent the only possible human conceptualization of
space-time. So when Wharf meets a Hopi form, instead of asserting a "primary function" based on a Western category and then listing "secondary functions," he tries to let the form itself speak directly and proposes a category,
perhaps a surprising one to those used to Western grammars, that covers all
of its functions (on Whorf's descriptive method, see P. Lee 1991; 1996). As
we have seen, this respecting of the specificity of each language in creating
a descriptive metalanguage, and in drawing conclusions from the analysis, is

The rise of cognition and the repression of languages

187

the linguistic method Boas was proposing at the beginning of the twentieth
century.
There is another issue that marks the difference between the two
approaches. The whole point of Sapir and Whorf's analogy to Einstein is
that his relativity, like quantum mechanics, shows that the world that science can glimpse is different from the world as we experience and construe
it through "habitual thought." It is what is sometimes called a discontinuist
position in the philosophy of science (see the discussion of Bachelard in
Chapter 10). Gipper, however, was a convinced continuist: he saw Einstein,
whom he admired to the point of having a large photo of him on display in
his office (RalfThiede on LINGUIST List 2.567, 1991), as having provided
a firmer foundation for Newton's universe (Gipper 1972: 2), which therefore
remained both absolute reality and the necessary lived world of any human
subject.
In reading Malotki telling us what all humans are supposed to presuppose
about time, I am reminded of passages from two linguists, one French, one
American. In his book L'homme de paroles, Claude Hagege writes of rationalist grammarians such as Le Laboureur:
They purely and simply declared that since all men had received the same logical principles as a gift, speakers of Latin, who regularly practiced inversion, therefore had to
be speaking differently from the way they thought, while for the French conception and
expression coincided. (1986: 206)
Some years later, A. L. Becker gave the following warning, which reads like a
retrospective critique of Malotki:
Our general tendency is to "read into" our experience of a distant language the familiar things that are missing, all the silences, and then we claim that all these things are
"understood," "implied," or "part of the underlying logical structure" of those languages
. .. It takes a while to learn that things like tenses, and articles, and the copula are
not "understood" in Burmese, Javanese, or Malay ... In Burmese these things aren't
implied; they just aren't there. (1995: 7-8)
The great resonance of Malotki's odd demonstration, the sense of palpable
relief that he had (finally!) killed off the revenant Whorf, are understandable
in terms of the philosophical divisions we have been discussing. For most
linguists and cognitive scientists of the period, this was a welcome elimination of a troublesome and potentially devastating claim of human conceptual
variability.

The reinforcement of divisions

The 1950s through the 1980s saw the progressive triumph of universalist cognitive science. From the 1980s, one saw the concomitant rise of relativistic

188

Linguistic Relativities

postmodernism. By the end of the 1980s there had been a mas~ive retur~ to t~e
old split between universalizing natural sciences and their ancIllary socIal SCIences on the one hand, particularizing humanities and their ancillary cultural
studies on the other. Some things, in the prevailing view, were universal, others
so particular as to call for treatment as fiction or anecdote. Nothing in ?e~w~en
was of very much interest, and North American anthropology, the dIscIplIne
that had been founded upon and achieved a sort of identity in crossing the
natural-sciencelhumanities divide, faced an identity crisis. Symptomatically,
one noticed many scholarly bookstores disappearing their linguistics sections
into "cognitive science," their anthropology sections into "cultural studies."
In this climate, linguistic relativity was heresy, Whorf, in particular, a kind
of incompetent Antichrist. The "Whorfian hypothesis" of linguistic relativism
or determinism became a topos of any anthropology textbook, almost inevitably to be shown to be silly. Otherwise serious linguists and psychologists
(e.g., Pinker 1994: 59-64) continued to dismiss the idea oflinguistic relativity
with an alacrity suggesting alarm and felt free to heap posthumous personal
vilification on Whorf, the favorite target, for his lack of official credentials,
in some really surprising displays of academic snobbery. Geoffrey Pullum, to
take only one example, calls him a "Connecticut fire prevention inspector and
weekend language-fancier" and "our man from the Hartford Fire Insurance
Company" (Pullum 1989 [1991]: 163). This comes from a book with the subtitle Irreverent Essays on the Study of Language. But how irreverent is it to make
fun of somebody almost everybody has been attacking for thirty years?

'1
10

The return of the repressed

What has collapsed has not necessarily been superseded.

Trabant (1986: 206)

The previous chapter sought to set the record straight on what I take to be misrepresentations of linguistic relativity over several decades from the 1950s. In
this chapter we look at some ways in which key elements of the Boasian project survived and eventually grew again. The chapter has three parts: we look
first at some other schools of thought that sought to account for sociocultural
and linguistic diversity without falling into either universalist or essentialist
patterns; we then look at ways in which the Boasian heritage came to be preserved through the height of cognitive hegemony; and finally we examine how
this heritage has been revived dramatically, notably within cognitive science
itself, especially since the late 1990s.
Once in a while during the period just considered, someone from outside
would look at the controversy and wonder about the heaps of condemnation
of linguistic relativity and particularly of Whorf. A case in point is the literary
theorist George Steiner, who in 1975 called Whorf's work "the crowning statement" of the Leibniz-Humboldt "monadist" case, a theory
of great intellectual fascination. The "metalinguistics" of Whorf have for some time
been under severe attack by both linguists and ethnographers. It looks as if a good
deal of his work cannot be verified. But [his] papers ... constitute a model which has
extraordinary intellectual elegance and philosophic tact. They are a statement of vital
possibility, an exploration of consciousness relevant not only to the linguist but also to
the poet and, decisively, to the translator. (1975: 88)

Parallels: structuralism and symbolic anthropology

Other twentieth-century schools of thought posed questions comparable
to those of the Boasians about the nature and implications of linguistic and
cultural diversity. I Here I will concentrate on European structuralism and its
developments in the Prague School of the 1930s and the Paris-based structuralism of the 1960s and 1970s, the last of which drew equally on a primarily
189