The relation between aluminum, fluorine, calcium, and pH in drinking water and the risk for cognitive impairment was studied using data collected in 1988-1989 in a population-based survey of 3,777 French men and women aged 65 years and older (the Paquid study). Cognitive impairment was defined as a score lower than 24 on the Mini-Mental State Examination. Data related to drinking water came from two prospective measurement surveys whose reliability was checked by a variability study. A mixed effects logistic regression was performed, adjusting for the effects of the age, sex, educational level, and occupation of the participants. A significant protective effect was found for high calcium concentration (odds ratio = 0.8 for a concentration > or = 75 mg vs. < 75 mg, p = 0.015). For aluminum, no significant effect was found when pH was not included in the model, but the results showed a weak interaction between aluminum and pH. When the four variables (calcium, aluminum, pH, and aluminum x pH interaction) were considered in the statistical analysis as a group, this group was significantly related to cognitive impairment (p = 0.01).