Each year from 1974, when they entered secondary school, to 1978, when they reached school-leaving age, a cohort of over 6000 schoolchildren from Derbyshire, England, answered a questionnaire about their own and their family's smoking practices and their social activities. Their replies revealed a steady increase in the prevalence of smoking during adolescence. Those children who in 1974 smoked, had friends of the opposite sex, were highly involved in social activities, experienced peer pressure to smoke and rejected the health hazards of smoking were more likely to be regular smokers in 1978 than were other children. Similarly, those children who, when aged 11–12 years, had parents or siblings who smoked, had friends of the opposite sex, and were highly involved in social activities increased their smoking rapidly in subsequent years. Sex and social class differences in the strength of these associations suggest that an understanding of the development of smoking during adolescence requires knowledge of the particular character of the social relationships among different subgroups of that age-group and the various meanings of smoking to them.