Path analysis was used to test the predictive and mediational role of self-efficacy beliefs in mathematical problem solving. Results revealed that math self-efficacy was more predictive of problem solving than was math self-concept, perceived usefulness of mathematics, prior experience with mathematics, or gender (N = 350). Self-efficacy also mediated the effect of gender and prior experience on self-concept, perceived usefulness, and problem solving. Gender and prior experience influenced self-concept, perceived usefulness, and problem solving largely through the mediational role of self-efficacy. Men had higher performance, self-efficacy, and self-concept and lower anxiety, but these differences were due largely to the influence of self-efficacy, for gender had a direct effect only on self-efficacy and a prior experience variable. Results support the hypothesized role of self-efficacy in A. Bandura's (1986) social cognitive theory.Social cognitive theory suggests that self-efficacy, "people's judgments of their capabilities to organize and execute courses of action required to attain designated types of performances" (Bandura, 1986, p. 391), strongly influences the choices people make, the effort they expend, and how long they persevere in the face of challenge. According to Bandura (1986), how people behave can often be better predicted by their beliefs about their capabilities than by what they are actually capable of accomplishing, for these beliefs help determine what individuals do with the knowledge and skills they have.Although researchers have established that self-efficacy is a strong predictor of behavior (Maddux, Norton, & Stoltenberg, 1986), research on the relationship between self-efficacy and academic performance in areas such as mathematics is still limited (Bouffard-Bouchard, 1989). Studies of math self-efficacy have been largely correlational, and researchers have emphasized the need to construct causal models with which to conceptualize and test hypothesized relationships (Hackett & Betz, 1989;Meece, Wigfield, & Eccles, 1990). When causal modeling has been used, most models have excluded key variables identified as influencing math performance (most notably, self-concept), or the theoretical framework used to hypothesize relationships was not based on social cognitive theory. Thus, results have added little to a better understanding of self-efficacy's influence. Bandura (1986) mance; that is, when these determinants are controlled, self-efficacy judgments are better predictors of performance. Bandura also argued that constructs such as selfconcept, perceived usefulness, and anxiety are "common mechanisms" of personal agency in the sense that they, like self-efficacy beliefs, also influence an outcome. However, these mechanisms are, to a great extent, the result of selfefficacy judgments-their influence is largely due to the confidence with which individuals approach a task. Consequently, although strong correlational relationships are observed between these mechanisms and related outcomes, ...