Recent evaluations of smoking prevention programs have suggested considerable promise for curricula emphasizing resistance of social influences. The present study extends these evaluations by addressing key methodological limitations in previous work. Twenty-two matched schools were randomized to experimental and control conditions. Grade 6 students received a 6-week core curriculum, plus additional sessions through Grades 7 and 8. Questionnaires, and saliva samples to validate self-reported smoking behavior, were collected at five times over the 2-year study period. Cross-sectional and longitudinal analyses examined program impact for five levels of initial smoking experience, ranging from "never smoker" through regular, weekly smoker. Significant program effects were documented, most clearly for those having some experience with smoking before the program began and for those with smoking peer and family models. This study provides the methodologically most rigorous test to date of social influence programs for smoking prevention and documents for the first time significant effects for those at high risk for smoking.