Data from three bargaining games - the Dictator Game, the Ultimatum Game, and the Third-Party Punishment Game - played in 15 societies are presented. The societies range from US undergraduates to Amazonian, Arctic, and African hunter-gatherers. Behaviour within the games varies markedly across societies. The paper investigates whether this behavioural diversity can be explained solely by variations in inequality aversion. Combining a single parameter utility function with the notion of subgame perfection generates a number of testable predictions. While most of these are supported, there are some telling divergences between theory and data: uncertainty and preferences relating to acts of vengeance may have influenced play in the Ultimatum and Third- Party Punishment Games; and a few subjects used the games as an opportunity to engage in costly signalling.