The inference style of clinical depressives was examined using a proverb interpretation task in which self-generated and other-generated proverb interpretations were judged for their normative adequacy. The correspondence between subjects' judgments of the interpretations and independent ratings based on objective criteria was determined using a profile similarity index. Subject-rater correspondence for depressed and nondepressed subjects was examined to determine whether depressed subjects' judgments could be considered accurate (the depressive realism position; Alloy & Abramson, 1979, 1988) or inaccurate and biased with respect to the self (Beck, 1967, 1976). Results indicated that all subjects were less accurate when judging the self than when judging others. However, this discrepancy was significantly larger for depressed subjects who were more inaccurate in the self condition than the nondepressed subjects. No group differences emerged in the other condition, indicating that depressives were capable of judging the normative adequacy of interpretations. In contrast to Beck's theory, an analysis of error type did not indicate that depressive self-inaccuracy was due to a systematic tendency to distort self-relevant information in a negative direction. However, depressed subjects were less likely than nondepressed subjects to distort other-relevant information in a negative direction. Findings are discussed with respect to issues in the nature and assessment of accuracy and bias, as well as current status of the depressive realism hypothesis.