Aim. To investigate the relationship between early conduct problems and early onset of cannabis use, with special emphasis on possible gender differences. Design. A prospective longitudinal study of a national sample of 2436 adolescents. The sample was followed up over a year and a half, when the adolescents were in their early teens. Setting. Norway. Measurements. On the basis of an earlier study, conduct problems (CP) closely related to the criteria for conduct disorder (CD) in DSM-III-R were decomposed into three dimensions, labelled serious, aggressive and covert. Further, information was collected about alcohol intoxication, daily smoking and use of cannabis. A number of questions were posed about sexual interactions and perceived puberty development. Parental socio-economic status was measured according to the ISCO-88. Separate information was collected as to whether the parents were on social welfare or unemployed. A parental bonding measure (PBI) was used to measure the emotional relationship between respondents and parents. Further, a measure of parental monitoring was used, and information was also collected on other aspects of the family milieu, and on the adolescents' peers. Statistical models. Logistic regression analysis was employed. As the sample consisted of pupils clustered within classes within schools, a three-level error structure for the logistic regression model was estimated. Findings. There was a strong association between early conduct problems and subsequent cannabis initiation. Also conduct problems at a potential subclinical level seemed to have great impact. The effect was significantly stronger in girls than in boys. Serious CP was found to have a moderate effect upon cannabis initiation in boys, whereas aggressive and covert CP had strong effects in girls. Early onset of puberty and early sexual involvement had no impact, whereas early use of cigarettes proved an important precursor to cannabis use. Conclusions. Conduct problems are important precursors of early onset cannabis use, but probably represent gender-specific aetiologies. There might be an important potential for prevention of early onset drug use in the prevention of early conduct problems, in particular for girls.