To explore the association between smoking and breastfeeding, we obtained data from a retrospective questionnaire-based national survey comprising a random sample (n = 34799) of all mothers giving birth in Norway 1970-91. Variables studied were postpartum smoking habits for both parents, duration of breastfeeding, infant's year of birth and parental age. The response rate was 70% (n = 24438). During the study period, the maternal postpartum smoking prevalence decreased from 38% to 26%. The proportion breastfeeding at 6 months increased from 15% to 44% among smokers, and from 30% to 72% among non-smokers. In spite of a considerable increase in breastfeeding both among smokers and non-smokers, the proportion of breastfeeding, non-smoking women at 6 months was twice that of smoking women during the whole period. Furthermore, the duration of breastfeeding was shorter among young mothers and when the fathers were smoking. There was epidemiological evidence that the effect on breastfeeding of smoking might represent both biological and social mechanisms.