This investigation contrasted path analysis models for 40 males and 40 females based on the predictions of Feltz's (1982) respecification model of Ban-dura' s (1977) self-efficacy theory in the approach/avoidance of two trials of a modified back dive. The hypothesized (respecified) model proposed that previous related experiences, self-efficacy, and heart rate predicted initial back-diving performance and that previous performance and self-efficacy predicted subsequent performance. The hypothesized model also proposed that self-efficacy mediates the influence of autonomic perception of arousal on performance. Results indicated that males had lower state anxiety and autonomic perception scores than females on the first trial. No differences occurred for back-diving performance, self-efficacy, or heart rate. Path analysis results indicated that the hypothesized model fit the data better for females than for males, though it left much unexplained variance for both males and females. Females showed a reciprocal relationship between self-efficacy and performance, whereas males showed a reciprocal relationship between autonomic perception and heart rate. Previous performance and self-efficacy were strong predictors of subsequent performance for both males and females.