In this paper we conduct a laboratory experiment to test the extent to which Moore and Repullo's subgame perfect implementation mechanism induces truth-telling in practice, both in a setting with perfect information and in a setting where buyers and sellers face a small amount of uncertainty regarding the good's value. We find that Moore-Repullo mechanisms fail to implement truth-telling in a substantial number of cases even under perfect information about the valuation of the good. This failure to implement truthtelling is due to beliefs about the irrationality of one's trading partner. Therefore, although the mechanism should - in theory - provide incentives for truth-telling, many buyers in fact believe that they can increase their expected monetary payoff by lying. The deviations from truth-telling become significantly more frequent and more persistent when agents face small amounts of uncertainty regarding the good's value. Our results thus suggest that both beliefs about irrational play and small amounts of uncertainty about valuations may constitute important reasons for the absence of Moore-Repullo mechanisms in practice.