The quantitative results of Sources of Self‐Efficacy in Science Courses‐Physics (SOSESC‐P) are presented as a logistic regression predicting the passing of students in introductory Physics with Calculus I, overall as well as disaggregated by gender. Self‐efficacy as a theory to explain human behavior change [Bandura [1977] Psychological Review, 84(2), 191–215] has become a focus of education researchers. Zeldin and Pajares [Zeldin & Pajares [2000] American Educational Research Journal, 37(1), 215] and Zeldin, Britner, and Pajares [2008] Journal of Research in Science Teaching, 45(9), 1036–1058] found evidence that men and women draw on different sources for evaluation of their self‐efficacy in science fields. Further, self‐efficacy is one of the primary dimensions of students' overall science identity and contributes to their persistence in physics [Hazari, Sonnert, Sadler, & Shanahan, 2010Journal of Research in Science Teaching 47(8), 978–1003]. At Florida International University we have examined the self‐efficacy of students in the introductory physics classes from the perspective of gender theory, with the intention of understanding the subtleties in how sources of self‐efficacy provide a mechanism for understanding retention in physics. Using a sequential logistic regression analysis we uncover subtle distinctions in the predictive ability of the sources of self‐efficacy. Predicting the probability of passing for women relies primarily on the vicarious learning experiences source, with no significant contribution from the social persuasion experiences, while predicting the probability of passing for men requires only the mastery experiences source. © 2012 Wiley Periodicals, Inc. J Res Sci Teach 49: 1096–1121, 2012