The continuity between early smoking experimentation and smoking at age 16 years was analysed for a birth cohort of New Zealand children. This analysis suggested that when due allowance was made for errors in reports of smoking behaviours, there was evidence of relatively strong continuity (r = 0.60) between early smoking experimentation and cigarette smoking at age 16 years. Further analysis suggested that the continuities between early smoking experimentation and later smoking arose from three pathways that linked early smoking experimentation to later smoking. First there was evidence to suggest that children who engaged in early smoking experimentation tended to affiliate with adolescent peer groups whose members smoked. In turn, these peer group affiliations reinforced pre-existing tendencies to cigarette smoking. Secondly, a small component of the apparent continuity between early smoking experimentation and later smoking arose because of common social, individual and contextual factors that were associated with both smoking experimentation and later smoking. Finally, there was evidence of moderate direct continuity in cigarette smoking behaviour over time. The implications of these findings for the development of smoking prevention programmes are discussed and it is concluded that effective programmes need to be embedded in a developmental approach which attempts to reduce both early smoking experimentation and the effects of peer pressure in adolescence on the development of cigarette smoking.