Most of the evidence for group differences in color vision between males and females comes from color-naming behavior. We examined color vision in another way, at the level of judgments of inter-color similarity. 67 teenaged subjects (37 males, 30 females) provided triadic (odd-one-out) judgments for 32 colors. These stimuli were pastel shades, being the pigment chips used in the D15 and D15-DS panel tests. Their unsaturated, non-focal nature reduces the possibility of subjects encoding the stimuli verbally and basing their responses on semantic knowledge of color-term proximity.

Multidimensional scaling solutions for the males' and the females' data separately suggest relative differences in the salience of color-space axes, with the males tending to attend more to a lightness axis and less to a red-green axis. This was confirmed using individual-differences MDS to obtain color-axis salience parameters for each subject. Individual indices of goodness-of-fit also showed the males to be less reliable in their judgments.

A second group of 35 adult subjects (11 males, 24 females) showed the same group differences. These differences are substantial (comparable to intra-group variation) as well as robust. Possible explanations include the existence of photopigment heterozygosity among females (males are hemizygous), and gender differences in overall color awareness. The procedure lends itself to investigation of other factors affecting color vision, and we present preliminary results from a comparison of smokers and non-smokers.