We study finitely repeated 2 / 2 normal form games, where playersnhave incomplete information about their opponents’ payoffs. In a laboratory experiment we investigate whether players (a) learn the game they are playing, (b) learn to predict the behavior of their opponent, and (c)nlearn to play according to a Nash equilibrium of the repeated game. Ournresults show that the success in learning the opponent’s type depends on the characteristics of the true game. The learning success is much higher for games with pure strategy Nash equilibria than for games with a unique mixed strategy Nash equilibrium, and it is higher for games with symmetricnpure strategy Nash equilibria than for games with asymmetric equilibria. Moreover, subjects learn to predict the opponents’ behavior very well. However, they rarely play according to a Nash equilibrium and we observenno correlation between equilibrium play and learning about the game.