The contribution of electroencephalogram (EEG) findings early in the course of neonatal bacterial meningitis to the prediction of severe adverse outcome was assessed in a retrospective cohort study. Infants had known outcomes to 1 year of age and an EEG performed during the first week of illness. EEGs were subclassified as follows: overall EEG description, background activity, presence of positive rolandic sharp waves, presence of seizure activity, and presence of focal abnormal activity. EEG patterns predictive of severe adverse outcome were identified by univariate and multivariate analyses. Of 101 infants admitted with bacterial meningitis, 37 had an EEG performed. Of the 37 infants, 21 had adverse outcomes; nine infants died, and 12 infants had moderate or severe disability. EEG background activity and overall EEG description were identified as predictors of adverse outcome; multivariate analysis indicated that the latter was a stronger predictor (sensitivity 88%, specificity 90%). Infants with normal or mildly abnormal EEGs had good outcomes whereas those with moderate to markedly abnormal EEGs died or survived with adverse outcome. The accuracy of predictions increased when EEGs were repeated. In a high-risk population of infants with bacterial meningitis, moderate-to-markedly abnormal EEG reliably predicts adverse outcome.