School factors and not solely pupil composition probably cause variation in smoking prevalence amongst schools, but there are no theoretical models to explain why. In this paper we propose a hypothesis to explain schools' influence on pupils' smoking and test this using an existing cross-sectional survey of 23,282 pupils from 166 secondary schools in the West Midlands, UK. We hypothesise that school-level educational achievement scores would not be associated with smoking prevalence, but schools providing value-added education given the social background of pupils (authoritative schools) would provide effective support and control, have a relatively strong influence on pupils' lives and be associated with lower than average smoking prevalence. Schools providing value-denuded education (laissez-faire schools) would have a relatively weak influence on pupils' lives and be associated with higher than average smoking prevalence. The school achievement measures were the proportion of pupils achieving 5A-C General Certificates of Secondary Education (5A-Cs) grades and the proportion of half days lost to truancy. Value-added/denuded terms were created by regressing 5A-Cs and truancy on five markers of the social profile of pupils at the school. Authoritative schools achieved better than expected rates on both measures. Laissez-faire schools achieved worse than expected rates on both measures. All other schools were classed as indeterminate. Multilevel logistic regression was used to relate the risk of regular smoking to school culture in both achievement and authoritative/laissez-faire terms, both with and without adjustment for pupil-level risk factors for smoking. As predicted, schools' achievement measures were unrelated to pupils' smoking. The odds ratios (95% confidence intervals) for smoking in authoritative and laissez-faire schools relative to indeterminate schools were 0.80 (0.70-0.91) and 1.16 (1.07-1.27), respectively. Adjustment for pupil-level smoking risk factors had little effect. School culture is an independent risk factor for adolescent smoking. Schools providing effective support and control might protect pupils from smoking.