AIM To investigate effects of cigarette consumption level and socio-economic circumstances during adolescence on adult smoking. METHODS 1958 British birth cohort (all births 3-9 March 1958). Logistic regression used to predict (i) smoking at 41 years and (ii) persistent smoking (at 23, 33 and 41 years) from cigarette consumption and socio-economic circumstances at 16 years, indicated by social class and educational qualifications. RESULTS Of 6537 subjects with full smoking history, 30% smoked at 16 years, 23% smoked at 41 years and 19% smoked at 23, 33 and 41 years (persistent smokers). Heavier smokers at 16, 23 and 33 years were more likely to smoke at 41 years than lighter smokers. The odds ratio (OR) of smoking at 41 years was 2.5 for men and 3.0 for women who smoked >/=60 cigarettes/week at age 16, relative to <20 cigarettes/week. Subjects from manual social backgrounds and those with no qualifications had elevated risks of being a smoker at 41 years or a persistent smoker. These effects were robust to adjustment for adolescent consumption level (e.g. adjusted OR for no qualifications was 3.8). However, adolescent consumption level modified the effect of educational achievements. Among lighter adolescent smokers, those gaining higher qualifications had lower prevalence of smoking at 41 years (16%) than men with no qualifications (83%); among heavier adolescent smokers, prevalence was more similar for subjects with higher (56%) and no qualifications (69%). CONCLUSIONS Socio-economic background appears to influence adult smoking behaviour separately from adolescent cigarette consumption which is a recognized measure of nicotine dependence. There was some evidence that effects of early nicotine dependence are modified by educational achievements.