We analyze the evolution of behavioral rules for learning how to play a two-armed bandit. Individuals have no information about the underlying pay-off distributions and have limited memory about their own past experience. Instead they must rely on information obtained trough observing the...