The aim of this study was to evaluate the neuropsychological instruments in predicting Alzheimer dementia after 5 years in the context of a longitudinal population-based cohort study. A total of 585 nondemented 75-year-old individuals completed neuropsychological examination at the baseline investigation; 479 subjects were followed after 30 months and 404 after 60 months. Cognition, depression and memory complaints were evaluated with psychometric instruments. Known risk factors for Alzheimer dementia were included in the analyses. Univariate logistic regression analyses and stepwise multiple models were calculated. A combination of reduced verbal memory, reduced visual motor speed, subjective memory complaints and the APOE epsilon4 carriage was best in predicting incident probable Alzheimer dementia (R(2) = 0.42, ROC curve = 0.91). The model achieved a positive predictive value of 23.3%, a negative predictive value of 98.7%, a sensitivity of 82.8% and a specificity of 82.4%. Alzheimer dementia can be predicted by neuropsychological instruments measuring episodic memory and motor speed. A high percentage of 98.7% subjects at age 75 years could be predicted as remaining non-demented at age 80 years. The prediction of those unlikely to develop AD would be more important in the future to spare further expensive diagnostic testing and protective therapies in individuals at low risk.