Abstract: Dominance has been assumed to be a quality of overwhelming social importance but satisfactory definitions and measures have not been devised. As an indication of predictability of outcome of interaction between animals, it can be explained in terms of ordinary learning processes previous to and during a specific relationship. Agonistic interactions are usually determined and often initiated by the subordinate's behavior, and subordinate behavior is correlated with physiological changes, so that a subordination hierarchy is probably a more useful concept than a dominance hierarchy. Hierarchies develop in stressful conditions, especially in captivity where animals with overresponsive adrenal cortices are at a selective disadvantage. In wild groups hierarchies are tenuous or absent and stress-responsive members are probably advantageous to a group. Group defense and leadership roles are not correlated with rank, but policing is characteristic of high-ranking animals in species where it occurs. There is no evidence that formation of a hierarchy reduces aggression--hierarchies are actually associated with high rates of aggression in primate groups. There is no conclusive evidence that high ranking males have greater overall reproductive success, and an alternative hypothesis that adult males are sexually active for a relatively short stage of their lives fits existing data equally well.