A classical data set is used to predict the effect of selection on sexual dimorphism and on the population means of three characters-stature, span, and cubit-in humans. Given selection of equal intensity, the population means of stature and of cubit should respond more than 60 times as fast as dimorphism in these characters. The population mean of span should also respond far more rapidly than dimorphism, but no numerical estimate of the ratio of these rates was possible. These results imply that sexual dimorphism in these characters can evolve only very slowly. Consequently, hypotheses about the causes of sexual dimorphism cannot be tested by comparing the dimorphism of different human societies. It has been suggested that primate sexual dimorphism may be an allometric response to selection for larger body size. We show that such selection can indeed generate sexual dimorphism, but that this effect is too weak to account for the observed relationship between dimorphism and body size in primates.