Abstract: Breeding groups with multiple stallions occur sympatrically with single-stallion breeding groups in feral horse, Equus caballus, populations. Mutualism and reciprocal altruism between stallions have been proposed to explain the origin and functioning of multistallion bands. However, empirical support for these hypotheses is contradictory and incomplete. Furthermore, there are no explicit tests of the predictions that each hypothesis makes about stallion behaviour and social structure. We compared nine multistallion and 18 single-stallion bands in the Kaimanawa Ranges, New Zealand. Compared with agonistic behaviours, affiliative behaviours were relatively unimportant in the relationships between stallions within bands. The number of stallions in the band did not have a positive influence on mare group size, stability, home range quality or reproductive success in bands. Furthermore, there was a positive relationship between aggression ('intolerance') by the dominant towards subordinate stallions and the subordinates' effort in mare group defence ('helping') but a negative relationship between helping effort by subordinates and their proximity to, and mating with, the bands' mares. Therefore, the predictions of the mutualism and reciprocal altruism hypotheses were not supported. Indeed, for some of the predictions we found the opposite outcomes to be true. Multistallion bands had significantly poorer reproductive success, and dominant stallions were less tolerant of subordinates that helped most and reduced their access to mares. Nevertheless, in all other respects Kaimanawa stallions in multistallion bands behaved like those described elsewhere. Thus, we reject cooperative hypotheses for multimale breeding groups in horses and discuss the mate parasitism and consort hypotheses as better alternatives. Copyright 2000 The Association for the Study of Animal Behaviour.