The aim was to evaluate the association between maternal smoking in pregnancy and delinquency of the offspring in youth and early adulthood. The study population consisted of 5966 male members of the Northern Finland birth cohort of 1966, of whom 355 (6.0%) had committed a crime which led to a criminal record by early 1989, which was taken as an indicator of delinquency. Data collection was started during pregnancy and the development and health of the children was followed up continuously to the age of 14. The incidence of delinquency by the age of 22 was 4.6% among the sons of the mothers who did not smoke during pregnancy and 10.3% among those of the smokers. No clear dose-response gradient was observed, and those stopping smoking during the first trimester of the pregnancy had only a slightly smaller incidence of delinquency than those continuing smoking through pregnancy. The association of maternal smoking with delinquency was studied by controlling a number of social and demographic variables using an approach based on stratification by a confounder score. The adjusted incidence difference between smoking and non-smoking mothers was 4.1% (95% confidence interval [CI] : 2.5-5.8) and the ratio was 1.73 (95% CI : 1.41-2.12). A parallel analysis by logistic regression yielded an estimated odds ratio of 1.74 (95% CI : 1.37-2.11) for the same comparison. Even though the association between maternal smoking and delinquency of the offspring remained after adjustment for the available social and demographic factors, maternal smoking may be symptomatic of a certain lifestyle and norm-breaking behaviour which is likely to increase delinquency in the children rather than being an agent with a direct causal role. This is supported by the lack of any clear dose-response pattern and the minor importance of stopping smoking. Another possibility is that maternal smoking may cause some kind of brain damage which affects the behaviour of the child.