Errors in self-reports of three academic performance measures are analyzed. Empirical errors are shown to depart radically from both no-error and random-error assumptions. Self-reports by females depart farther from the no-error and random-error models for all three performance measures. (Author/BW)