OBJECTIVE To determine the usefulness or otherwise of the awake electroencephalogram (EEG) in the diagnosis of Alzheimer's disease (AD). DESIGN Prospective collection of one or more awake EEGs in patients diagnosed as having AD or mixed AD and multi-infarct dementia according to current systematic criteria with analysis of those cases confirmed by postmortem examination. Systematized blind interpretation of EEGs. SETTING Tertiary care practice with both ambulatory and hospitalized patients, ie, neurological department in general hospital and psychogeriatric unit in psychiatric hospital. PATIENTS A series of 86 subjects with AD and 17 with mixed AD and multi-infarct dementia being those members of a consecutive series on whom postmortem information was available. Awake EEGs in 56 age- and sex-matched control subjects. RESULTS Seventy-five patients with AD (87.2%) and 13 of the mixed group (76.5%) had abnormal EEGs on first testing, giving a sensitivity of 87.2% for uncomplicated AD. Ultimately, 79 (92%) of 86 patients with AD had abnormal EEGs. Twenty (35%) of 56 EEGs for matching control subjects were abnormal. Moderately abnormal or severely abnormal EEGs were found in 10 (50%) of 20 of the patients with AD of less than 4 year's duration compared with two (4.1%) of 49 of the control subjects, giving a specificity of 95.9% for EEGs with this degree of abnormality. The normal EEG had a negative predictive value of 0.825 with respect to the diagnosis of AD in these populations. CONCLUSIONS Widespread availability, low cost, and high sensitivity support the use of the awake EEG in the diagnosis of AD.