Shape and space in grammar, revisitedShapeCompound color termsLearning about shape by studying colorConclusion and prospectusHaviland->TK, p. 1 “White-blossomed on bended knee”: LINGUISTIC MEDIATIONS OF NATURE AND CULTURE1 John B. Haviland UCSD-Anthropology draft Sept 2003 If, as I will try to illustrate in this paper, much of the conceptual raw material for our understanding of the world comes from language, then what we know as “nature” (what, that is, natural scientists, including folk scientists, invest their energies in investigating) turns out to be—or at least starts out as—a disguise for “culture.” Specifically, important aspects of the natural world are best understood as, in the fashionable phrase, “linguistically mediated culture.” Greg Urban has recently argued (1998) that a central vehicle for naturalizing culture is discourse, something that is simultaneously in and about the world. If so, then a primary culprit in disguising this relationship is the transparency of language. Perhaps the paradigm case of finding “nature” IN “language” (as a particularly tractable subdomain of culture) is the “basic color term” research of Berlin & Kay (1969) and their associates. The method and its underlying logic are familiar. Speakers of different languages are presented with a series of “color stimuli”: Munsell color chips whose sole point of contrast is “color” as we might normally conceive of it. Faced with these chips they are asked to produce appropriate linguistic descriptions. From the resulting expressions, investigators identify elements dubbed “basic color terms” according to a series of interlinked semantic and morphological criteria (eliminating polysemous terms of various sorts, as well as morphologically transparent complex lexemes, and so on). Once identified (and they always are identified), these “basic terms” are passed through yet another filter: investigators locate their prototypical or best exemplars, once again on the color chart. The results of the exercise are compared with the results of similar exercises performed on other languages. We are familiar with the general conclusion: all languages turn out to have very similar sets of color terms, depending on exactly how many “basic” terms there are; systems can be arranged in a putative universal evolutionary sequence in which human languages progressively partition the “color space” into finer and finer discriminations. The finding is linked to 1 This paper started life as an oral presentation called “Chromatapoetics: Tzotzil color terms and spatial imagery,” Cognitive Anthropology Research Group, MPI for Psycholinguistics, Nijmegen, Sept. 7, 1991. Inspired by ongoing work by Victoria Bricker, it resurfaced in the guise of another paper presented in the session “Verbal Practice: Construing the ‘Cultural’ by Constructing the ‘Natural’”, organized by Michael Silverstein at the Annual Meetings of the American Ethnological Society, Portland, Or., March 26, 1999. I am grateful for comments by Bill Hanks and Bruce Mannheim on that latter occasion.Haviland->TK, p. 2 the intractable hard-wiring of our retinas and brains, a biological fact which on this account naturally seeks cultural expression in language. More than thirty years after Berlin and Kay (hereafter B&K) published their original little monograph on the subject, there continues to be a major industry in “basic color research,” ramified across disciplines. There is also a minor intellectual industry which criticizes the same research, spawning in turn ripostes and counter critique. (See Hardin and Maffi 1997 for examples.) One issue is what so-called “color terms” as identified by the B&K procedures “really mean.” How much of their denotational ranges is actually captured by reference to the color chart, or by colorimetric definitions? Are they, indeed, monosemic in the first place, and what licenses an adequate gloss? (These questions are common to all lexicography and by no means easy to answer.) Moreover, what confidence do we have that the color terms identified form a legitimate linguistic set, in terms of the formal/functional criteria that we normally apply to linguistic entities—similar kinds of syntagmatic behavior, participation in paradigmatic links, and so forth? If they do not exhibit such links, we need special arguments to show their semantic inter-relationship; that is, without such formal/functional coherence, the apparent “domain” to which they refer is arguably an extralinguistic imposition. Furthermore, what evidence do we have to exclude wider formal/functional links between “basic color terms” and other linguistic entities, except to appeal from the very start to an essentially circular semantic criterion (making reference to “color”)? One may also question the method of the B&K project, for example its central reliance as stimuli on standardized color chips (which in the heyday of this research were exported to remote corners of the world by enthusiastic color researchers, sometimes with a truly imperial stance). How do we know that we are dealing with “color terms” when color (as defined by the Munsell chips) is the only available “denotation” in the elicitation task, and other possible “meanings” of the resulting expressions are systematically excluded? Except that we impose such a task on informants, why should we think “color” is a domain at all, or that it matters in any given case, when no attention is necessarily paid to the “color naming” practices of the language users involved? Finally, what legitimizes the central methodological move: to exclude from consideration all but the “basic color terms”—defined on a mixture of semantic, morpho-syntactic, and a priori grounds? Tzotzil, a close sister of Tzeltal which was one of B&K’s principal languages, has a fairly standard system of “basic color terms,” following this received account. (See example 1.) Indeed, the facts of Tzotzil “basic color naming” lend little support to the sorts of criticisms, just sketched, that are often launched against this program of research. There is a clearly recognizable set of simple CVC roots in Tzotzil that centrally denote colors, despite some non-color glosses. (For example, sak means not only ‘white’ butHaviland->TK, p. 3 also ‘clean’ and ‘transparent’; yAx2 denotes