Universal Foci and Varying Boundaries in Linguistic Color Categories

Recent research has questioned the universal basis of color categorization and has instead emphasized cross-linguistic variation in boundaries of color categories. We propose that these cross-linguistically varying boundaries can be predicted from near-universal focal colors within the categories. In support of this proposal, we show that: (1) best example choices for color terms in 110 unwritten languages cluster near the prototypes for English white, black, red, green, yellow, and blue – we take these 6 points in color space to approximate universal foci; (2) best example choices cluster more tightly across languages than do category centroids; and (3) a computational model can predict color term boundaries from labelings of best examples reasonably well, for several languages, including one that has been taken to counterexemplify universal tendencies in color naming. Overview It has long been held that there are universal tendencies in color naming, in that linguistic color categories are organized around universally-shared focal points, or prototypes, in color space. Berlin and Kay (1969; B&K for short) showed that the best examples of color terms across a sample of 20 languages seemed to cluster in color space. That study and subsequent work (Kay & McDaniel, 1978; Kay & Maffi, 1999) showed that the most reliable and widespread of these clusters correspond to the six Hering primaries: white, black, red, green, yellow, and blue – suggesting that these points in color space may constitute a universal foundation for color naming. These foci in color space have also appeared to be cognitively privileged, in non-linguistic tasks with speakers of languages that hav