Abstract
Focal choices for basic color terms in different languages are tightly clustered around similar regions of color space, pointing to strong universal tendencies in color naming (Kay and Regier, PNAS 2003). We asked whether the foci within these clusters can also show significant variation across languages, by analyzing data from the World Color Survey (WCS), which includes data from an average of 24 speakers from each of 110 unwritten languages (available at http://www.icsi.berkeley.edu/wcs/data.html). For each language we tested for consensus terms (used by a majority of speakers) whose foci were closest to the foci for one of the 11 basic terms for English. Variations in the mean foci for the cluster of terms corresponding to English “red”, “green”, “blue” or “yellow” were then compared to the variation predicted by randomly sampling the foci for 20 individuals across different languages. Clusters for “blue” and “green” were restricted to the subset of languages that included distinct consensus terms for both color regions. For all four terms the variance in both the mean hue and mean lightness of foci across languages was significantly greater than expected by a single population. This within-cluster variability suggests that the choice of exemplars for color categories can be modulated by unknown factors differing among the languages. However, these variations were small compared to the differences between clusters (e.g. “red” and “yellow” clusters were separated by ∼10 times the standard deviation of foci within each cluster, with only one WCS language with a consensus term nearer to the intermediate English term “orange”) while variability among speakers of a common language was pronounced. This suggests that factors contributing to focal color differences between languages — and in particular language per se - impose only a weak constraint on color naming by the WCS respondents.