We examined the relationship between tobacco consumption and Alzheimer's disease (AD), cognitive deficit, and memory performances in elderly community residents of southwestern France. Subjects were considered to have a cognitive deficit if they scored under 24 on the French version of the Mini-Mental State Examination, and poor memory performances if they scored under 8 on the Benton Visual Retention Test. Among the 3770 subjects who gave information about their smoking habits, only 356 (9.5%) were current smokers, 998 (26.5%) were past smokers, and 2416 (64%) were never smokers. Current smokers (odds ratio (OR) = 0.52, P < 0.001) and past smokers (OR = 0.54, P < 0.001) had a lower risk of cognitive deficit than did never smokers. However, this significant relationship disappeared after adjustment for potential confounding factors such as occupational category (OR = 0.91 (not significant) for current smokers and OR = 0.87 (not significant) for past smokers). Similar results were obtained for the risk of AD and of poor memory performances. The apparent protective effect of smoking habits on cognitive abilities could be due to a confounding effect of occupational category.