Abstract: We consider a simple two-agent model of strategic experimentation with payoff externalities and unobservable actions. Each agent decides whether to invest effort (or resources) into an independent risky project, if he invests and succeeds his payoff is greater if the other agent succeeds too. We characterize the unique symmetric Markov Perfect Equilibria of the game. The dynamic structure of the problem allows agents to coordinate their experimentation in such a way that they provide a strictly more effort than they would have provided in the absence of the externality. Nevertheless, they provide too little effort in a dynamically inecient way relative to the rst best. The magnitude of efficiency losses is critically linked to the common prior about projects’ quality. If agents are too pessimistic they do not provide any effort even though it is socially optimal to do so.