The processes that underlie the formation of the dominance hierarchy in a group are since long under debate. Models of self-organisation suggest that dominance hierarchies develop by the self-reinforcing effects of winning and losing fights (the so-called winner-loser effect), but according to ‘the prior attribute hypothesis’, dominance hierarchies develop from pre-existing individual differences, such as in body mass. In the present paper, we investigate the relevance of each of these two theories for the degree of female dominance over males. We investigate this in a correlative study in which we compare female dominance between groups of 22 species throughout the primate order. In our study female dominance may range from 0 (no female dominance) to 1 (complete female dominance). As regards ‘the prior attribute hypothesis’, we expected a negative correlation between female dominance over males and species-specific sexual dimorphism in body mass. However, to our surprise we found none (we use the method of independent contrasts). Instead, we confirm the self- organisation hypothesis: our model based on the winner-loser effect predicts that female dominance over males increases with the percentage of males in the group. We confirm this pattern at several levels in empirical data (among groups of a single species and between species of the same genus and of different ones). Since the winner-loser effect has been shown to work in many taxa including humans, these results may have broad implications.