The authors tested the hypothesis that smoking exerts a protective effect on Alzheimer's disease or dementia in a population-based cohort of 668 people aged 75-101 years (Sweden). Smoking was negatively associated with prevalent Alzheimer's disease (adjusted odds ratio = 0.6, 95% confidence interval 0.4-1.1) and dementia (adjusted odds ratio = 0.6, 95% confidence interval 0.4-1.0). Over 3-year follow-up (1989-1992), the hazard ratios of incident Alzheimer's disease and dementia due to smoking were 1.1 (95% confidence interval 0.5-2.4) and 1.4 (95% confidence interval 0.8-2.7). Mortality over 5-year follow-up was greater among smokers in demented (hazard ratio = 3.4) than nondemented (hazard ratio = 0.8) subjects. Smoking does not seem protective against Alzheimer's disease or dementia, and the cross-sectional association might be due to differential mortality.