Prior research shows that undergraduates tend to identify more strongly with the field of science after participating in scientific research. However, mediators that might account for this association are not well understood. In the current study, we propose that science self-efficacy may serve this mediational function. Specifically, data from a 2-year longitudinal study were used to test a model in which science self-efficacy was expected to mediate the association between research involvement and identity as a scientist. The ethnically diverse sample included 251 undergraduates who were recruited from colleges and universities across the United States. The hypothesized mediation model was tested with a cross-lagged panel analysis. As expected, greater levels of research experience at Time 1 predicted higher identity as a scientist at Time 3, and this association was mediated by science self-efficacy at Time 2. Exploratory analyses testing for ethnic and gender differences in the model suggested that the associations in the model were similar for undergraduates from diverse backgrounds. From a theoretical standpoint, the current study provides novel insight into how research experience, efficacy, and identity relate to one another over time. Applied implications center on the importance of involving undergraduates in research that has the potential to bolster their science self-efficacy.