Counting the Color Words in a Language

Some languages divide the visible spectrum into eleven named regions and some into three, and the temptation is to conclude that their speakers see differently. The evidence does not support that conclusion, and the reason it does not is instructive about how such claims should be tested at all.

Speakers of a language with a single term covering green and blue can sort colored chips as finely as anyone else, and they notice a mismatch between two shades as quickly. What differs is not perception but the effort of description. Asked to name a chip in isolation, they use a modifier, in much the same way that an English speaker reaches for a comparison when naming a shade between pink and orange. The category is coarse; the eye is not.

There is a genuine effect, but it is smaller and stranger than the strong claim. When a speaker has to hold a color in memory for a few seconds before matching it, the name given to it at the moment of seeing shifts the memory slightly toward the center of that name's range. The effect disappears when the language faculty is occupied with something else during the delay, which is the experiment that makes the mechanism clear: the interference is at the stage of storing a description, not at the stage of seeing.