Even before the advent of modern game theory, reciprocity and reputational effects have been invoked to help explain the sustained existence of cooperation. Hobbes argued that the Foole, who thinks it rational to renege on an agreement if one’s partner has already done her part, mistakes a repeated game for a one-shot game. Likewise, the Foole ignores reputational effects to his peril. Yet reciprocity and reputation have little bite in one of the more important evolutionary transitions in human history: the move from small hunter-gatherer groups to large cultural communities without formal political institutions. Herbert Gintis, Robert Boyd and Samuel Bowles have argued that so-called strong reciprocators (those inclined to both cooperate and punish norm-violators) allow for the sustained existence of cooperation during this transition. Yet if administering a punishment is costly, as it often is, then the existence of punishment is itself in need of explanation. One popular solution is to invoke group selection.  This approach, however, is not without its detractors and has recently been called into question for its historical implausibility.

I propose an alternative account, inspired by the work of David Gauthier (1986) and Robert Frank (1988). Both Gauthier and Frank assume that agents can, with a certain degree of success, identify the type of their counterpart. The idea is that cheap talk, such as subtle body cues and minute facial expressions, often reveal one’s underlying type. I extend this story to the case of costly punishment, borrowing from the evolution of preferences literature. I develop a model in which agents are randomly paired to play a one-shot two-stage game, consisting of the Prisoner’s Dilemma and a punishment phase.  Cooperation is possible for large swaths of parameter space.  Lastly, I show that strong reciprocators do particularly badly in a donation game with third-party monitoring, suggesting that strong reciprocity is neither necessary nor sufficient to explain cooperation.