by Paul Kay, Brent Berlin, and William Merrifield. 1991. Journal of Linguistic Anthropology Volume 1, Issue 1, pp. 12-25.
Addresses the nature of "composite categories" of color and notes that of all the possible combinations of the Hering primaries, only a few composite categories are found (e.g. purple = red + blue, but no yellow + blue or red + green). Composite colors are like orange (red + yellow), brown (yellow + black); they perceived to be composites of these at least, whatever their true nature.
"Categories that are gradient and overlapping, as color categories often are, permit a possibility for the formation of new categories which nongradient, mutually exclusive categories do not."  [the point is that color categories frequently overlap]
Claims that Latin viridis is a green-blue category (without citing any evidence whatsoever!), p. 23; also that Gk. chlo^ros (which he misspells) is a yellow-green word.