Data from three studies on two captive groups of rhesus monkeys, Macaca mulatta, were used to test a hierarchical multiple regression model of adult social organization. The model followed the literature in assuming that matrilineal kinship affects rank distance, i.e. that the rank positions of related adults are generally closer to one another than the positions of unrelated adults. Rank distance, in turn, was assumed to affect affiliative behaviour, tolerance at feeding and drinking sites, and aggression. The model was tested with matrix permutation techniques that allow dyadic variation to be taken completely into account. The results confirmed that rank distance correlates with a range of social behaviour independently of the effects of kinship. Several alternative functional explanations of the role of rank distance were compared; attraction among individuals with similar ranks appeared the most plausible explanation. The study also addressed the general applicability of a previously reported 'class' division in one of the two groups. This division could not be replicated, and may have been a temporary structural phenomenon. © 1991.