The Mini-Mental State Examination score is often used as a screening test for dementia. We studied its properties for a population of French community residents, using a sample of 2,792 randomly selected subjects in Gironde, France. When the traditional cutoff point 23-24 was used, we observed a sensitivity of 1.0 and a specificity of 0.77. We sought to increase the specificity of the test, while keeping sensitivity at 1.0, with the use of logistic regression analysis. We found (1) a specificity of 0.81 when "recall three objects" and "orientation to time" were added to the Mini-Mental State Examination score; (2) a specificity of 0.90 when Benton's Visual Retention Tests and Isaacs' Set Test of Verbal Fluency were added to the Mini-Mental State Examination score and subscores; and (3) no improvement in specificity when age and educational level were included in the analyses.