Philosophy Department, University of Victoria,
Victoria, British Columbia, Canada V8W 3P4
Electronic mail: ddedrick@uvic.ca

We need to reconsider and reconceive the path that will take us from innate perceptual saliencies to basic (and perhaps other) colour language. There is a space between the perceptual and the linguistic levels that needs to be filled by an account of the rules that people use to generate relatively stable reference classes in a social context.

Here are two claims important to the universalist tradition in colour naming: (1) Some colours are psychologically salient, and that salience is not a function of language or culture; (2) Some colour categories are psychologically salient, and that salience is not a function of language or culture. I take it that Saunders and van Brakel (S &vB) accept these two claims as being empirically supported. As they note with respect to (1), certain of Eleanor Rosch’s (1972a) experiments with the Dani of New Guinea "did show that the Dani remembered focal colours better than non-focal, as did Americans" (sect. 2.2, par. 5). With respect to (2) Saunders and van Brakel devote section 2.3 to the relevant work. Though they suggest that "caution" (sect. 2.3, para. 1) is required when it comes to Bornstein et al’s (1976) claim that four month old infants naturally partition the spectrum into 4 colour categories—in much the same way as Western adults do in hue naming tasks—they present no evidence that there is any problem with Bornstein’s hue-categorization claims.

These results specify innate sensitivities to prototypical colours and to hue categories. The Dani had no language for many of the focal colours. Bornstein’s infants had neither language nor culture. So there is something right about Berlin & Kay’s (1969) original claims. There are psychologically salient basic colour categories with an internal structure dominated by focal (prototypical) colour samples. Berlin & Kay may have been wrong about many things (e.g. the number and nature of basic categories) and their original work did not establish these claims. But the claims are pretty much established and the proper foundation for one kind of investigation of human colour language.

Where to go from this minimal toehold? S & vB strive (rightly) to show that there is no determined path from the perceptual saliencies mentioned above to the colour vocabularies found in real languages. They are (rightly) critical of a scientific practice which would like to see —or imagines there to be— tight correlations percolating up from the neurophysiological to the psychophysical to the perceptual and, ultimately, the linguistic. This is currently a no-go. The "Hering code" of psychophysics cannot, at the present time, be represented at the physiological level (Abramov & Gordon 1994). While this may be a merely epistemic gap, that is not likely to be the case when we look to the relations between perceptually salient colours/categories and colour language. It is a standard assumption of the universalist tradition that languages with restricted basic colour vocabularies have composite terms (sect. 2.1, para. 2). Within their reference classes, these terms include at least two of the original basic colour categories and may include two or even three of the unique hues. There is no evidence that these categories —historically primary by the universalist’s own account— have any correlates at the perceptual/psychophysical/physiological levels. Vision science has nothing to say of them and would, I suspect, have to be very different from what it is in order to do so. Composite basic colour categorizations are "salient" only in the linguistic domain. (These points can all be made from within the universalist framework. One does not need to reject it all—and much of vision science!—to recognize these problems.)

We need to reconsider and reconceive the path that will take us from innate perceptual saliencies to basic (and perhaps other) colour language. The universalist tradition has proposed no cognitive account of how people come to have and use basic colour terms. It has limited itself to a correlation game—an attempt to show that colour categories and colour prototypes which have a non-linguistic salience are represented in the languages of the world. When the universalist does not get the fit —between the perceptual and the linguistic— there is very little to say; there is nothing but (supposed) correlations to fall back upon. S & vB make much of putative failures of fit, and their evaluation of the universalist tradition is almost wholly negative. What more positive proposal can we make that is not simply a recapitualation of the standing universalist view? I think we should consider there to be a space between the perceptual and the linguistic which needs to be filled by an account of the rules that people use to generate relatively stable reference classes in a social context. These rules must be stated with some precision and yet be flexible enough to account for the kinds of variation in colour languages that we find. This is a tricky (possibly circular) project but it is not improbable (Harrison 1973). It takes the idea of non-linguistic saliencies seriously and asks how such saliencies may be exploited by colour language users for essentially social purposes.

As S & vB point out (sect. 2.3, para. 2 ), children have some difficulty in mastering colour language. This suggests that the rules that they must learn are relatively complex and difficult to internalize even in Western basic-colour-using contexts — contexts where there is explicit, rote learning of abstract colour categories. I think we ought to concentrate on these languages first. How do their speakers conceptualize colour? This may give us some clue as to how the speakers of rather different languages operate. We may find that categorization based on hue is common to Western languages, but it is brightness that is crucial to others. We may find languages that do not fully integrate hue and brightness, and we may find languages that do not fit a hue-saturation-brightness (HSB) model of colour ordering. It may be possible, therefore, to classify languages in terms that are different from Berlin and Kay’s evolutionary order. Such classification can only be "nontrivial" if there is more to colour categorization than social idiosyncrasy. But even Saunders and van Brakel say (sect. 1, para. 6) that they are not arguing for that view.

Abramov, I., and Gordon, J. (1994) Color appearance: on seeing red—or yellow, or green, or blue. In: Annual Review of Psychology, ed. L. W. Porter and M. R. Rosenzweig. Annual Reviews Inc.

Bornstein, M. H., Kessen, W. & Weiskopf, S. (1976) Color vision and hue categorization in young human infants. Journal of Experimental Psychology 2: 115-29.

Harrison, B. (1973). Form and Content. Basil Blackwell.

Rosch Heider, E. (1972a) Universals in color naming and memory. Journal of Experimental Psychology 93: 10-20.