Computerized spectral and visual EEG analyses were performed in 35 patients with Alzheimer's disease (AD) and compared to 23 patients with major depression and to 61 healthy elderly controls. In particular, we were interested in the diagnostic efficacy of these two techniques in the identification of cases of AD with only mild cognitive impairment (as measured by the Folstein Mini-Mental State score). For the computer analyzed data, in differentiating AD patients from controls, the spectral pooled parasagittal mean frequency was used. In comparing AD patients to depressed subjects, a combined parasagittal delta and theta spectral score was employed. Visual analysis criteria were based on the severity of generalized EEG abnormalities (with or without focal features). We found that spectral analysis afforded only modest advantages over visual EEG analysis in differentiating AD patients from elderly controls as well as from those with major depression. Since the degree of spectral and visual EEG abnormalities correlated with the severity of dementia, both tests more often correctly classified those AD patients with lower Folstein scores. Also, both tests identified primarily the same patients. We did not find the computer to be more sensitive than the eye in the identification of AD patients with mild impairment. However, computerized spectral data was derived from only 4 channels, while 16 channels and a longer recording time were used for visual analysis. In addition, some areas which have been reported to show EEG abnormalities in AD were not included in the computerized data.