Abstract: Social, genetic and reproductive consequences of natal dispersal were investigated in female horses,Equus caballus, living in a herd with a natural social structure. Dispersal did not as a rule reduce the level of competition the young mares faced: they did not selectively join groups with fewer resident females than the groups they left, and they did not attain higher ranks; there was also no tendency for females to disperse to groups with the fewest resident females, and they suffered more aggression from the mares in their new groups than in their natal groups. These results therefore do not support the hypothesis that a function of natal dispersal is to reduce intra-sexual competition. The young mares nevertheless dispersed non-randomly, generally joining harems with one stallion and at least two subadult females; and they preferred to move to groups with familiar females but no familiar males. As a result, most were closely related to some females of their new groups, but distantly related or unrelated to the male(s). Since after dispersal the young mares bred only with a male of their new groups, inbreeding coefficients of most (85%) of their offspring were lower than from matings between half siblings. These results are consistent with the hypothesis that a function of natal dispersal is to avoid close inbreeding. Dispersal did not appear to involve reproductive costs: the young mares suffered no delay in age at first reproduction, and the survival rates of their first foals tended to be higher if the females had emigrated, although not significantly so.