Sister-to-Sister: The Black Women's Health Project is a skill-building HIV/STD risk-reduction intervention for African American women that had significant effects in reducing self-reported sexual risk behavior and biologically confirmed sexually transmitted disease (STD) incidence. The present analyses were conducted to identify which theory-based factors that were addressed in the intervention accounted for its success. The data were collected in the context of a randomized, 5-group intervention trial with assessments at baseline and at 3, 6, and 12 months following the intervention. A mediation analysis was conducted with condom use at last sex, self-reported 12 months after the intervention, as the outcome variable. Mediators were also measured at the 12-month follow-up. Mediators were derived from social cognitive theory: condom use knowledge, hedonistic beliefs regarding effects of condom use, expected sex partner reactions to condom requests, sex partner approval of condom use, self-efficacy for impulse control, self-efficacy for carrying condoms, and self-efficacy to achieve consistent condom use with partner. The intervention significantly improved all potential mediators except condom use knowledge (p = .15), hedonistic beliefs (p = .08), and self-efficacy for impulse control (p = .20). Analyses testing each mediator separately revealed that expected partner reaction, partner approval of condom use, self-efficacy for condom carrying, and self-efficacy for condom use were significant mediators. When they were entered into a multivariate mediation analysis, however, only self-efficacy for condom use was significant (p = .001). These results highlight the importance of self-efficacy in explaining the effects of skill-building sexual risk-reduction interventions on women's use of condoms. Self-efficacy was more important than characteristics of male partners.