An experimental test of some qualitative predictions of the Kreps-Wilson (1982) model of reputation building is conducted in a version of a borrower-lender game first used experimentally by Colin Camerer and Keith Weigelt (1988). A systematic response to changes in the payoff function of borrowers is observed. However, the response is opposite to the direction predicted by the theory. Furthermore, the observed behavior cannot be reconciled with the theory by an appeal to "homemade" priors of the type specified by Camerer and Weigelt. Copyright 1992 by The Econometric Society.