AIMS: The associations between smoking and various socioeconomic indicators may have different implications and causes, which may also vary according to sex and birth cohort. This study analyses how two dimensions of socioeconomic position, an individual (education) and a structural (occupation) indicator, are associated with ever, current and ex-smoking. METHODS: Data on smoking behaviour were collected in five cross-sectional surveys of random samples of the general Danish population aged 20 years or more at intervals between 1982 and 1994. In total, 8,054 men and 8,281 women participated. Logistic regression was used to analyse the influence of education and occupation on smoking behaviour controlling for sex and birth cohort. RESULTS: In cohorts born after 1930 ever and current smoking were related to years of school education and current occupation. The prevalences of ever and current smoking were highest among the least educated, unskilled workers, unemployed persons and persons who received welfare benefits. A significant interaction between birth cohort and education indicated that the educational difference in ever and current smoking increased significantly with increasing year of birth. In multivariate analysis controlling for sex and birth cohort, ex-smoking seemed to be more strongly associated with education than occupation. Those with 12 or more years of school education had twice as high a chance of being ex-smokers as those with 7 years of school or less. CONCLUSION: Smoking behaviour is strongly associated with both individual and structural indicators of socioeconomic position in Danish adults in all cohorts except for those born before 1930.