Abstract
A strict universalist theory of color naming predicts that all six Hering primaries should be present in all speakers of all languages; a strict linguistic relativity theory predicts that color categories should vary freely across cultures, but that individuals who come from the same culture and speak the same language should categorize colors similarly. To address this issue, we report two analyses of color naming in the World Color Survey (WCS, Kay, et al., 2003, http://www.icsi.berkeley.edu/wcs/). We scored the color-naming data of all 2616 speakers of the 110 tropical languages of Kay's WCS as “blue”, “dark”, “grue” (green-or-blue), or “other”, according to how they named the colors that English speakers call “blue”. The results were striking: only rarely did all speakers of a given WCS language exhibit the same color-naming pattern for “blue”. This result is good evidence against strict versions of both theories. However, both theories could accommodate this result if color were insufficiently important to all speakers of a language for them to give distinct names to all the colors they can distinguish. We then looked at the consensus for basic color terms used by the 2616 WCS speakers. Five of the six classic Hering primaries showed excellent consensus across the data set: black, white, red, yellow, and green were all assigned with high frequency to chips given those focal color names in English. That result is consistent with the relaxed version of the universalist view, and difficult to explain under the linguistic relativity hypothesis. In contrast to the other five, “blue” is a statistically significantly lower-consensus color term. Inasmuch as the languages in Kay's WCS data set are mostly tropical, this result is straightforwardly predicted by our phototoxicity hypothesis, which predicts that “blue” should be rare in the tropics because of damage to the eye from UV sunlight.