To evaluate the efficacy and adverse effects of the type B monoamine oxidase inhibitor selegiline (also known as I-deprenyl) in the treatment of Alzheimer's disease. Long-term, double-blind, placebo-controlled trial. Seven cities (1 or 2 nursing homes in each city) in the Czech and Slovak Republics. A total of 173 nursing-home residents fulfilling the DSM-III criteria for mild to moderate Alzheimer's disease. Selegiline (10 mg per day) or placebo (both including 50 mg ascorbic acid) administered for 24 weeks. Clinical Global Impressions scale and Nurses Observation Scale for Inpatient Evaluation at baseline and at weeks 6, 12 and 24; Clock Drawing Test at baseline and 24 weeks, results of which were evaluated as normal or pathologic, and quantitatively on a modified 6-point scale; Sternberg's Memory Scanning test at baseline and at weeks 6, 12 and 24; Mini Mental State Examination, and electroencephalogram at baseline and 24 weeks; Structured Adverse Effects Rating Scale; physical, laboratory, hematological and electrocardiographic examinations at baseline and weeks 12 and 24. A total of 143 subjects completed enough of the trial to be entered in the analysis. Subjects were analyzed by 2 subgroups depending on whether they had a normal or pathologic result of the Clock Drawing Test. Analysis of variance showed significant improvement with selegiline versus placebo among those with a normal result of the Clock Drawing Test on the Mini Mental Status Examination (total score and orientation-place subscale) and among those with a pathologic result of the Clock Drawing Test of Sternberg's Memory Scanning test (for both speed and accuracy), on the Clinical Global Impressions scale as well as in terms of the dominant frequency on electroencephalograms. Selegiline has a long-term beneficial effect in Alzheimer's disease on memory modalities that reflect the function of the prefrontal areas of the brain, which are rich in dopamine receptors. The delayed appearance of differences between selegiline and placebo supports the notion that the mechanism of action is through neuronal rescue or neuroprotection. The differential response of patients with normal and pathologic results of the Clock Drawing Test may reflect the fact that the evaluation methods' sensitivity to change depends on the severity of dementia.