This study used mediation analyses, implemented in a longitudinal structural equation modeling framework, to examine the mechanisms by which a social-influence-based school drug use prevention program (Project ALERT) achieved its effects on past month cigarette use and alcohol misuse. Participants were 4277 South Dakotan middle-school students (2554 treatment and 1723 control) measured at baseline and 1 year later on past month cigarette use and alcohol misuse, as well as cigarette- and alcohol-related mediating variables targeted by the Project ALERT curriculum (i.e., resistance self-efficacy, positive and negative beliefs about use, and peer influence). Results for cigarettes showed that all hypothesized mediating variables were significant mediators of ALERT's effect on intentions to smoke and past month cigarette use, with peer influence being the strongest. Results for alcohol point to positive beliefs about the consequences of drinking as an important mediator for alcohol misuse. Taken together, the findings highlight an avenue for program improvement through increased impact on peer influence to use alcohol and drugs.