OBJECTIVES: We wished to identify which potential mediators of the Seropositive Urban Men's Intervention Trial (SUMIT) intervention were in fact changed by the intervention, and further to identify which among these factors distinguished men who decreased their risk behavior relative to those who increased it, irrespective of the intervention arm. METHODS: We examined social cognitive theory and other psychosocial variables that the intervention was designed to affect (potential mediators) in both sets of analyses. These were assessed at baseline, 3-month follow-up, and 6-month follow-up. We tested which potential mediators were changed by the intervention relative to the comparison arm, and which of these factors distinguished men discontinuing risk behavior [unprotected insertive anal intercourse (UIAI) or UIAI with HIV-negative or status-unknown partners] compared with those initiating it. RESULTS: Factors changed by the intervention included partner serostatus assumption making, hedonistic condom outcome expectancies, anxiety and depression. Factors associated with behavioral risk reduction included personal responsibility to protect others from infection and self-evaluative outcome expectancies regarding transmission risk behavior. These constructs are similar and involve the engagement of moral processes and altruism in sexual behavior with others. DISCUSSION: The present results suggest that, although we designed the intervention to enhance personal responsibility to protect others from HIV, we were not successful in this goal. However, changes in this factor did prove to be an important correlate of behavior change. Possible ways to design and deliver more successful interventions are discussed.