The theoretical basis of Seyfarth's priority of access model of female cercopithecine grooming was critically examined and alternative models suggested. These models, named 'engagement', 'interference' and 'rank difference', generated different predicted grooming distributions by assuming the operation of different constraints. These constraints were, respectively, the time available for grooming, active interference on the part of higher ranked animals and a depression of grooming relations as rank difference between animals grew. These priority of access models were compared with observed grooming patterns in two groups of free-ranging olive baboons, Papio cynocephalus anubis: one large, one small. The fit of these models was poor. An alternative method of examining the effects of rank on grooming behaviour using multiple regression was successful. In the small troop the rank of the groomee explained a significant amount of the variance in grooming whilst the rank distance between groomer and groomee did not. In the large troop the opposite effect was found. In the light of these results the merits of bottom-up modelling versus top-down description are discussed. The question of group Size in primates and its relationship to social complexity are addressed.