In this paper we experimentally test a theory of boundedly rational behavior in a "lemons market." We analyzed two different market designs, for which perfect rationality implies complete and partial market collapse, respectively. Our empirical observations deviate substantially from these predictions of rational choice theory: Even after 20 repetitions, the actual outcome is closer to e±ciency than expected. Our bounded rationality approach to explaining these observations starts with the insight that perfect rationality would require the players to perform an in¯nite number of iterative reasoning steps. Boundedly rational players, however, carry out only a limited number of such iterations. We have determined the iteration type of the players independently from their market behavior. A significant correlation exists between the iteration types and the observed price offers.