We investigated whether cigarette smoking is negatively associated with Alzheimer's disease (AD) in a population-based, frequency-matched, case-control study of 152 AD patients and 180 controls. Ever having smoked was associated with lower risk of AD (adjusted odds ratio = 0.61; 95% confidence interval: 0.37-0.99). Additional multivariate analyses demonstrated that education and history of hypertension modified this association. The direction of the modification was for higher education level and history of hypertension to further reduce the risk. The "dose-response" pattern showed the greatest risk reduction among those who smoked least and suggests a biologic mechanism of a dose-dependent up-regulation of nicotinic (cholinergic) brain receptors. These data, although consistent with current opinion about pathophysiology of AD, do not suggest smoking should be used as a preventive strategy for AD.