Purpose: This paper aims to examine how influences at home and school interact to predict smoking among adolescents. Design/methodology/approach: Data were collected from 15-year-old pupils from Norway (n=1,404 in 73 Grade 10 school classes). Multilevel logistic regression analysis was used to determine how family and school influences interact to predict adolescent smoking behaviour. Findings: A total of 14 schools strictly enforced the ban on smoking among pupils, and 60 schools strictly enforced existing restrictions on teachers' smoking. Pupils from all schools were exposed to smoking by both their families and teachers. Although the main effects of school enforcement and family smoking were not associated with adolescent smoking behaviour, a cross-level interaction term between these variables was significantly associated with school level variation in regular smoking (odds ratio [OR] 5.57, 95 percent confidence interval 1.47-21.12). Individual rather than school's perception of exposure to teacher's smoking, and parental norms were associated with adolescent smoking behaviour, irrespective of the school's level of enforcement. Practical implications: The effect of school smoking restrictions seems to be dependent on smoking norms at home. Thus, in addition to school-based anti-smoking policies, those directed at parents are needed as well. Originality/value: The present study used a multilevel statistical approach to study the association between the independent factors and adolescent smoking behaviour, an approach that takes into account the cluster effect in hierarchically structured data. The use of this approach is important in studies that involve data with hierarchies as a means to avoid false conclusions. (Contains 2 tables and 1 figure.)