An individual's fitness depends not only on their phenotype but also on the phenotypes of their competitors and contemporaries. Sexual attractiveness may be strongly influenced by an individual's familiarity to potential mates or the rarity of the individual's phenotype. Such effects can cause negative frequency-dependent selection, maintaining striking polymorphisms in ornamentation. Here we test whether preferences for women's hair color, which is highly polymorphic between European populations, reflects patterns of positive or negative frequency-dependence. We assigned each participant to one of four experimental treatments in which we manipulated the frequency of hair colors in a set of 18 images presented consecutively. The four treatments were: blond hair rare, brown hair rare, red hair rare and an even distribution of the three hair colors. Following immediately on from this experimental phase, participants rated the same set of 9 faces, 3 with each hair color. Results showed that the experimental manipulation of hair color frequency did not significantly influence hair-color attractiveness. However, there were sex differences in ratings: men rated blond hair as most sexually attractive and women rated both blond and brown hair equally as attractive as red hair. Self-reported natural hair color had weak but significant effects on hair color preferences, but these effects disappeared when we restricted analyses to participants of European descent only. Our findings do not support the hypothesis that men's preferences for women's hair color are negative frequencydependent, favoring rare hair colors.