Project RESPECT was a multisite randomized trial comparing three clinic-based interventions' ability to increase condom use and prevent infection with HIV and sexually transmitted diseases. Because Project RESPECT had guiding concepts that determined the content of the sessions, the authors investigated how the intervention operated using these theoretical variables. Growth curve analysis and structural equation modeling estimated the correlation between intentions toward condom use and self-reports of condom use and isolated the treatment effects on mediating variables--attitudes, self-efficacy, and social norms--that predict intentions. The correlations between intentions and behavior exceeded .70 for both genders, justifying the emphasis on intentions. Project RESPECT was effective through changing attitudes and self-efficacy for females in both counseling interventions. For males, only enhanced counseling had significant effects on these two mediator variables.