We provide experimental evidence for the hypothesis that bounded rationality is an important element of the theory of the firm. We implement a simplified version of a mechanism that was designed in order to perfectly solve the holdup problem under conditions of perfect rationality (Maskin 2002). We test whether this mechanism either is able to perfectly solve our experimental holdup problem or may at least improve economic performance. We show that neither is the case: the implementation of the mechanism worsens economic performance. We reconstruct the main features of participants’ behavior by applying the logit agent quantal response equilibrium (McKelvey and Palfrey 1998) as an equilibrium concept that takes players’ potential mistakes into account.