The objectives of the present study were to evaluate the relationship between the fractional amplitudes of the EEG derived from power spectral analysis (PSA) of the electroencephalogram (EEG) and depth of coma measured clinically with the Glasgow Coma Score, and to assess the accuracy of PSA in predicting long-term outcome. Thirty-two patients rendered unconscious by blunt head injury (mean (GCS = 7) had intermittent EEG recordings daily from 1-10 days post injury. There was a significant correlation between fractional amplitude of the EEG and the GCS. The rate and magnitude of change in the EEG and GCS were also correlated. There were significant differences in PSA parameters between improved and deteriorated patient groups at the termination of monitoring (p = .02) and in the change of PSA parameters over time (p = .02). Using linear discriminant analysis of PSA parameters, the accuracy of outcome prognostication based on the six month outcome was approximately 75%. Accurate classification of outcome was possible in a number of patients in whom there was little or no change in the GCS during the period of monitoring.