Females in some mountain gorilla groups can be ranked on the basis of decided, non-aggressive approach-retreat interactions. However, data on four gorilla groups show that females do not form clear agonistic dominance hierarchies. Most aggressive interactions between females were undecided, and most dyads had undecided dominance relationships (Fig. 2). Females ignored aggression or responded to it aggressively more often than they responded submissively (Fig. 3). Moreover, directional inconsistency in aggression (including escalated contests) was high, and in some groups females showed significant bidirectionality in initiation of aggression (i.e., the more often they received aggression from partners, the more often they directed aggression at those partners; Fig. 4). Assigning ranks on the basis of aggressive interactions has limited power to explain variation in rates of initiating aggression and in responses and outcomes to aggression. Aggression was most common during feeding, but usually did not interrupt feeding bouts. It was proportionately more common in contexts that seem to involve competition for access to males (Figs 5, 6). This is consistent with the argument that females depend crucially on male services. As predicted by current models of female primate social relationships, agonistic relationships between gorilla females are unlike those typical in cercopithecine primates. They resemble those of some other primates and some equids in which contest competition for food is relatively unimportant and ecological costs of female transfer are low.