The reliability of resting electroencephalogram (EEG) power spectra was assessed for 49 normal and 44 schizophrenic subjects. We specifically examined internal consistency reliability to determine how much EEG data are necessary to ensure small measurement error. Twenty-one 8-s epochs of resting EEG were collected from each subject from site Cz. Epochs containing artifacts or blinks were eliminated. Power was computed in the bands delta (0.125-3 Hz), theta (3.125-8 Hz), alpha (8.125-13 Hz), beta1 (13.125-20 Hz) beta 2 (20.125-25 Hz) and beta 3 (25.125-30 Hz). Internal consistency was computed using coefficient alpha (Cronbach, 1951). Results for both groups indicated that eight artifact-free epochs of data were sufficient to give a coefficient alpha value of around .9. The schizophrenic and normal groups did not differ with respect to coefficient alpha. The proportion of artifacts in the data from the schizophrenics indicated that to obtain eight artifact-free epochs from members of each group, 40% more data would be required from the schizophrenics.