Purpose: Most studies on adolescent smoking focus either on the probability of smoking onset or frequency of smoking. We assume the existence of two different qualitatively distinct processes in smoking acquisition. Therefore our objective was to test a two-part latent growth model, which assumes that psychosocial variables associated with the probability of smoking onset are different from, or differently related to variables associated with frequency of smoking given smoking onset. Methods: The predictive associations of blocks of variables of (1) intrapersonal factors, (2) cognitions, (3) role models, and (4) family variables, on both smoking onset, and frequency of smoking given smoking onset, were tested in a nationwide sample of Dutch adolescents by using a two-part model. Summary: Smoking onset was instigated by a variety of factors, while similar and other factors predicted frequency of smoking given smoking onset itself. Self-esteem, attitudes, and proportion of friends smoking, were identified as factors that affected both absolute smoking and frequency of smoking. Overall conclusions: This study illustrates that it makes sense to differentiate between smoking onset and frequency of smoking and that few factors are active in both processes. © 2010 Elsevier Ltd.