The purpose of this study was to discover whether selfefficacy beliefs play the mediational role ascribed to them by Bandura (1986) and social cognitive theory, and whether these beliefs are stronger predictors of performance than are other presumed determinants and common mechanisms. We focused on the influence of self-efficacy on mathematics because our interest in self-efficacy was founded on a broader interest in education and the academic performance of students. In this context, the solving of math problems afforded a clearer and more reliable assessment than was possible in other academic contexts, but results would nevertheless inform social cognitive theory and its claims about self-efficacy in general.