To assess the effect of smoking on bone mineral density (BMD) at different life stages, to examine whether the effect of smoking differs between men and women, and to discover whether its effect in women differs according to history of estrogen use, a cohort study was carried out with single cross-section measurement of BMD by single and dual photon absorptiometry. The setting was the Framingham Study, a population-based cohort study with over 40 years prospectively collected data on smoking. Subjects (n = 1164) consisted of cohort members participating in the 20th biennial Framingham examination (1988-1989). The measurements included in the study were BMD measured at the hip, spine and radius, smoking history ascertained at all Framingham Study examinations since 1948, and other factors affecting BMD (age, weight, estrogen use, caffeine use, alcohol use and physical activity). Neither current smoking, recent (last 10 years) smoking, nor early adulthood smoking resulted in significantly lower BMD at any skeletal site among women who had not taken estrogen. Among women who had taken estrogen, BMD at most sites was lower among current or recent smokers, although the small numbers of smokers made it difficult to find significant differences at all skeletal sites. Among men, a consistently lower BMD at all skeletal sites was observed for smokers regardless of when in their life they smoked (4-15.3% lower), although the effect of smoking during early adulthood was of a lesser magnitude (4-8% lower). Former male smokers who had quit < 10 years ago had lower BMD than men who had quit > or = 10 years ago. In conclusion, in women who had used estrogen, BMD was lower in current or recent smokers than it was in non-smokers. In men, smoking at any stage of life had adverse effects on the skeleton that was independent of weight, alcohol or caffeine use, implying other mechanisms for smoking's effect on bone.