This study used a prospective design to test a model of the relation between social cognitive variables and physical activity in a sample of 277 university students. Social support, self-efficacy, outcome expectations, and self-regulation were measured at baseline and used to predict physical activity 8 weeks later. Results of structural equation modeling indicated a good fit of the social cognitive model to the data. Within the model, self-efficacy had the greatest total effect on physical activity, mediated largely by self-regulation, which directly predicted physical activity. Social support indirectly predicted physical activity through its effect on self-efficacy. Outcome expectations had a small total effect on physical activity, which did not reach significance. The social cognitive model explained 55% of the variance observed in physical activity.