Introduction. The aim of the study was to investigate whether the preclinical stage of Alzheimer's disease (AD) can be diagnosed in a clinical setting. To this end we investigated whether subjects with preclinical AD could be differentiated from subjects with nonprogressive mild cognitive impairment and from subjects with very mild AD-type dementia. Methods. Twenty-three subjects with preclinical AD, 44 subjects with nonprogressive mild cognitive impairment, and 25 subjects with very mild AD-type dementia were selected from a memory clinic population. Variables that were used to differentiate the groups were demographic variables, the Mini-Mental State Examination score, performance on cognitive tests, measures of functional impairment, and measures of noncognitive symptomatology. Results. Age and the scores for the delayed recall task could best discriminate between subjects with preclinical AD and subjects with nonprogressive mild cognitive impairment. The overall accuracy was 87% The score on the Global Deterioration Scale and a measure of intelligence could best discriminate between subjects with preclinical AD and subjects with very mild AD-type dementia. The overall accuracy was 85% Conclusions. Subjects with preclinical AD can be distinguished from subjects with nonprogressive mild cognitive impairment and from subjects with very mild AD-type dementia. This means that preclinical AD is a diagnostic entity for which clinical criteria should be developed.