Cross-Species Differences in Color Categorization

Article excerpt

Berlin and Kay (1969) found systematic restrictions in the color terms of the world's languages and were inclined to look to the primate visual system for their origin. Because the visual system does not provide adequate neurophysiological discontinuities to supply natural color category boundaries, and because recent evidence points to a linguistic origin (Davidoff, Davies, & Roberson, 1999), a new approach was used to investigate the controversial issue of the origin of color categories. Baboons and humans were given the same task of matching-to-sample colors that crossed the blue/green boundary. The data and consequent modeling were remarkably clear-cut. All human subjects matched our generalization probe stimuli as if to a sharp boundary close to the midpoint between their training items. Despite good color discrimination, none of the baboons showed any inclination to match to a single boundary but rather responded with two boundaries close to the training stimuli. The data give no support to the claim that color categories are explicitly instantiated in the primate color vision system.

(ProQuest Information and Learning: ... denotes formula omitted.)

The discontinuities that humans see in the color spectrum are so obvious that there is an inclination for cognitive psychologists to believe them inevitable and, hence, to propose that color categories are intrinsic to early levels of the neurophysiology that underpins color vision (Berlin & Kay, 1969; Franklin, Clifford, Williamson, & Davies, 2005; Kay & McDaniel, 1978; MacLaury, 1992; Pitchford & Mullen, 2002; Ratliff, 1976). However, even though color categories require discontinuities in perception, not even primary color categories (red, yellow, green, and blue) could be derived from any discontinuities in the spectral sensitivity of the three cone types by Sperling and Harwerth (1971) or, as Kay and McDaniel (1978) proposed, from the output of opponent process cells (Abramov & Gordon, 1994; Webster & Mollon, 1991). So even though some current research still stresses the role of precortical color processing in the formation of color categories, their appearance must be the result of organization at higher levels (Abramov & Gordon, 1994; Gegenfurtner & Kiper, 2003).

With respect to how color categories might be implemented in cortical visual areas, it is known that cells in V1 may be responsive to quite narrow ranges of wavelength and brightness (Yoshioka, Dow, & Vautin, 1996) with fairly much the same selectivity higher up at V4 (Schein & Desimone, 1990). Some color vision scientists (e.g., Hanazawa, Komatsu, & Murakami, 2000; Okajima, Robertson, & Fielder, 2002; Zeki, 1983) have proposed that such cells are the origin of color categories. There would seem to be even more reason to site color categories in the inferotemporal cortex, for lesions to that area produce achromatopsia (Cowey, Heywood, & Irving-Bell, 2001) and other (noncolor) disorders of perceptual categorical processing (Wilson & DeBauche, 1981). Yet there is no evidence for any cells, even in the inferotemporal cortex, that have the properties necessary for categorical perception proposed by Hamad (1987). The argument from categorical perception should predict, for example, "green" cells as being least sensitive to color change in the middle of the category and most sensitive to boundary colors (e.g., chartreuse or turquoise).

Despite this lack of a neurophysiological underpinning, what we shall call the nativist argument gains support from the observation, extended by Franklin and Davies (2004), that 4-month-old babies show a preference for looking at blue stimuli after habituating to green (Bornstein, Kessen, & Weiskopf, 1976). There are also a few studies with nonhuman primates that point to the nativist position with respect to color categories (Matsuzawa, 1985; Sandell, Gross, & Bornstein, 1979). There are, however, concerns about Sandell et al. …