Factors associated with changes in the smoking behaviour of approximately 6000 schoolchildren (two cohorts aged between 10 and 12 years in 1979) over 12 months are described. They were measured twice as part of a randomized controlled trial of a smoking prevention programme. Four groups were defined: (a) those who became smokers (adopters); (b) those who remained non-smokers; (c) those who became non-smokers (quitters), and, (d) those who remained smokers. Personal and social variables were ordered using a logistic regression model according to the strength of their association with adopting and quitting smoking. Factors distinguishing adopters from children who remained nonsmokers were, being a member of the older cohort, having friends who smoke, having siblings who smoke, approving of cigarette advertising and having a relatively large amount of money to spend each week. Factors distinguishing quitters from children who continued to smoke were, having siblings who do not smoke, being a member of the younger cohort, disapproving of cigarette advertising and having a relatively small amount of money to spend each week. Initial attitude scores were indicative of future smoking behaviour and where smoking behaviour changed, attitudes also changed so that the two remained congruent. The younger cohort improved their knowledge of smoking hazards over the year irrespective of their smoking behaviour. The older cohort showed significant differences in knowledge which were dependent upon smoking category, with 1980 smokers having lower knowledge scores than non-smokers and showing an apparent decrement in their previous knowledge.