This study sought to experimentally manipulate the core constructs of the Health Belief Model (HBM) and the Theory of Planned Behavior (TPB) in order to compare the success of interventions to increase preparatory condom use behavior (i.e., purchasing condoms, talking to a boyfriend or girlfriend about using condoms, and carrying condoms) based on these theories. A total of 258 participants were randomly assigned to one of three computer-based interventions (HBM, TPB, or information-only control). A total of 204 (79.1%) completed follow-up assessments 1 month later. Regression analyses were conducted to determine which set of theoretical constructs accounted for the most variance in behavior at baseline. A series of structural equation models were estimated to determine which constructs were the "active ingredients" of change. The TPB accounted for 32.8% of the variance in risky sexual behavior at baseline, while the HBM only explained 1.6% of the variance. Mediational analyses revealed differential intervention effects on perceived susceptibility, perceived benefits, and attitudes toward condom use. However, it was attitudes toward condom use and condom use self-efficacy that were associated with intentions, which then predicted preparatory condom use behavior at follow-up. Except for attitudes, the mediators that were successfully manipulated by the interventions (i.e., perceived susceptibility, perceived severity, and attitudes) were not the same constructs that predicted intentions (i.e., attitudes and condom use self-efficacy), and subsequently predicted behavior. This suggests that the constructs that explain behavior are not the same as those that produce behavior change.