Previous investigators have suggested that women display lower self-confidence than men across almost all achievement situations. The empirical validity of this suggestion is assessed in an experiment testing the hypothesis that performance evaluation guidelines moderate sex difference in self-confidence. Undergraduates read the guidelines by which their performance on an impending test would be evaluated. Guidelines were: ambiguous (A); clear—specifying the dimensions of performance to be examined, but not providing any examples of others' performance against which subjects' work would be compared (C); or clear with performance examples (CE). Subjects then completed the test and estimated how will they had done. Results showed the predicted pattern in conditions A and C: While women underestimated their actual performanće much more than men when guidelines were ambiguous, they did not do so when guidelines were clear. Further, the rise in self-confidence for women from conditions A to C was greater than that for men. Also as predicted, both sexes' self-confidence and performance were higher in condition C than in condition A. Finally, although condition CE was expected to depress only women's self-confidence and performance, both sexes showed this effect. It is concluded that sex differences in self-confidence are moderated by situation variables and that programs designed to reduce such differences might be improved by a greater focus upon women's response to clearly specifiable factors in achievement settings.