Individuals' attributional patterns are important determinants of behavior (Weiner, 1985). Controllability has been identified as an influential dimension of causal attributions. Self-efficacy (Bandura, 1977) is also a powerful predictor of participatory behavior, and the relationship between attributions and efficacy is an important area of study. Researchers have begun to investigate controllability in instructional settings in physical activity. Typically, contingency of feedback has been used to manipulate perceptions of controllability, but there have been methodological limitations that have resulted in inconsistent findings. The purpose of this study was to investigate the effect of controllability on causal attributions, efficacy, and performance in an exercise setting. It was hypothesized that non-contingent (uncontrollable) negative outcomes on an initial task would negatively affect causal attributions, thus decreasing efficacy and performance on a subsequent task. Participants, 150 college-aged women, were randomly assigned to one of three conditions: performance-contingent, negative non-contingent, and positive non-contingent. They completed two experimental tasks: 1) hand grip dynamometer; and 2) a wall sit, where they were asked to maintain a sitting posture without support for as long as they could. They viewed a video demonstration of each task, completed an efficacy questionnaire (Gernigon, Fleurance, & Reine, 2000), executed the task, received normative feedback about their performance, and then completed a causal attribution survey (Gernigon, Thill, & Fleurance, 1999) specific to that task. On the initial task, the hand grip, individuals in the performance contingent group were given their actual scores and told how their performance compared to age appropriate norms. Women in negative non-contingent group were told their performance was in the poor range, while those in the positive non-contingent group were told their performance was in the excellent range, regardless of their actual scores. For the second task, the wall sit, all participants were given accurate normative feedback about their performance. Data were analyzed using analyses of variance to test for group differences. As predicted, contingent feedback yielded a more stable, internal pattern of attributions. Positive non-contingent feedback produced higher levels of efficacy for the second task, but the contingent group outperformed both non-contingent groups on the wall sit. Taken together, these results provide insight into how perceptions of controllability can influence the cognition and motivation as individuals approach physical activities. One clear implication for teachers supported by these findings is the importance of providing feedback that is contingent on performance, rather than simply providing positive feedback. Keyword(s): adult physical activity/fitness, college level issues, exercise/fitness/physical activity