The present investigation was designed to replicate and extend the Feltz (1982) study of the causal elements in Bandura's (1977) theory of self-efficacy. Path analysis techniques were employed to investigate the predictions based on Bandura's model of self-efficacy, along with the additional influence of autonomic perception on the approach/avoidance behavior of female college students (N = 80) attempting a modified-back dive. The Bandura model predicted a reciprocal relationship between self-efficacy and back-diving performance, and between self-efficacy and physiological arousal (heart rate). It was also predicted that autonomic perception was a better predictor of self-efficacy than was physiological arousal, but not better than previous back-diving performance. Additionally, self-efficacy was hypothesized to be the mediator of past performance accomplishments, physiological arousal, and autonomic perception on back-diving performance. Bandura's model was tested against a "full" model that included performance, autonomic perception, and actual physiological arousal, along with self-efficacy as direct causal influences of back-diving performance. Results provided greater support for the full model. Although one's self-efficacy was the major predictor of performance on Trial 1, subjects' heart rates also significantly predicted performance on Trial 1. After Trial 1, back-diving performance on a previous trial was the major predictor of performance on the next trial. Furthermore, one's perception of autonomic arousal was a significant influence on self-efficacy but not on performance. Previous back-diving performance, however, was a better predictor of self-efficacy than autonomic perception. No reciprocal relationship was found between self-efficacy and physiological arousal. Moreover, the full model explained more performance variance than did the Bandura model.