In 4 experiments, college students received bogus personality-test feedback that they possessed an actual-self, ideal-self (desired but unpossessed), or rejected-self (unwanted and unpossessed) trait. When the test had low credibility, rejected-self feedback produced positive mood (PM), whereas ideal-self feedback produced negative mood (NM). Self-comparison of the feedback with self-representations apparently revealed the falsity of the feedback, making salient Ss' virtues or shortcomings. The pattern reversed when test credibility was high: Rejected-self feedback led to NM and ideal-self feedback to PM. These effects were not evident, however, when the feedback trait had high personal importance or when Ss generated counterexamples before feedback about an unimportant trait. This suggests that reactions to self-discrepant feedback depend on whether convincing counterexamples are readily accessed during self-comparison.