I experimentally investigate how diversification decisions are affected by information about historical outcomes. In the experiment, subjects have the opportunity to diversify between two simple binary gambles. When subjects lack information about previous outcomes, a vast majority diversify. By contrast, only a minority diversify after learning that one of the gambles has experienced better outcomes than the other in the past. Subjects’ posterior beliefs about winning probabilities predict the propensity to diversify. However, most of the subjects who do not diversify tend to chase the gamble with better historical outcomes, regardless of their beliefs. This behavior is consistent with subjects following a reinforcement heuristic.