In two closely related experiments, college student subjects were instructed to choose between a drug that allegedly interfered with performance and a drug that allegedly enhanced performance. This choice was the main dependent measure of the experiment. The drug choice intervened between work on soluble or insoluble problems and a promised retest on similar problems. In Experiment 1, all subjects received success feedback after their initial problem-solving attempts, thus creating one condition in which the success appeared to be accidental (noncontingent on performance) and one in which the success appeared to be contingent on appropriate knowledge. Males in the noncontingent-success condition were alone in preferring the performance-inhibiting drug, presumably because they wished to externalize probable failure on the retest. The predicted effect, however, did not hold for female subjects. Experiment 2 replicated the unique preference shown by males after noncontingent success and showed the critical importance of success feedback.