Three different constructs for measuring social influence were utilized in the present study to explain adolescents' present and future smoking behavior at 6 (T2), 12 (T3), and 18 months (T4) after the first test. Social influence was assessed by measuring the social norms, perceived smoking behavior, and direct pressure. The impact of the social influence constructs was also assessed in the context of broader models, including attitudes and self-efficacy expectations, intention, and previous behavior. The three social influence measures correlated significantly with intention and behavior. Stepwise regression analyses showed that perceived behavior and pressure made significant contributions, after entering social norms, in explaining actual and future adolescent smoking behavior. Adding attitudes and self-efficacy increased the predictive power of the model significantly. In agreement with the theory of Fishbein & Ajzen (1975), intention was the most powerful predictor in explaining present and future smoking behavior. Attitudes, self-efficacy, and the social influences also made small unique contributions improving the explanatory power by approximately 5%. Previous behavior, however, had a substantial unique contribution in predicting future behavior after attitudes, social influences, self-efficacy, and intention were entered in the equations. Since social influences may exert their impact via different routes, it is recommended that smoking prevention programs discuss not only overt pressures such as direct pressure from peers, parents, and media, but also address the more covert social pressures such as modeling and the adolescents' ability to cope with these covert influences. Furthermore, norms on nonsmoking should be made explicit.