The present investigation was designed to test the predictions of Bandura's (1977) theory of self-efficacy in a competitive, motor-performance situation. Subjects (30 males and 30 females) were randomly assigned to either a high or low self-efficacy condition in a 2 × 2 × 2 (sex × self-efficacy × trials) factorial design. Self-efficacy was manipulated by having subjects compete against a confederate on a muscular leg-endurance task where the confederate was said to be either a varsity ;rack athlete who exhibited higher performance on a related task (low self-efficacy) or an individual who had a knee injury and exhibited poorer performance on a related task (high self-efficacy). Because self-efficacy theory predicts that expectation-performance differences are maximized in the face of obstacles and aversive consequences, the experiment was rigged so that subjects lost in competition to the confederate on both trials. The results supported self-efficacy predictions with the high self-efficacy subjects extending their legs significantly longer than low self-efficacy subjects. Moreover, after failing on the first trial, high self-efficacy subjects extended their legs for a longer time than low self-efficacy subjects on the second trial. A postexperimental questionnaire revealed significant differences in cognitive states (e.g., expectations, attributions, self-talk) between high and low self-efficacy subjects, as well as between males and females. Results are discussed in terms of learned helplessness and differing patterns of sex-role socialization.