In this study, we describe a quantitative EEG method for localization of seizure onset in 29 patients with temporal lobe epilepsy. We used right-left differences in relative power and a logistic regression to predict the side of seizure onset. The optimum bipolar channels were T4-T6/T3-T5 and the optimum frequency band was 4-10.5 Hz. Using a threshold probability of 0.75 for classification, the best quantitative EEG method classified 23 patients (79%) correctly, 2 patients (7%) incorrectly, and 4 patients (14%) as undetermined localization. In comparison, 3 electroencephalographers visually interpreted these same EEGs and classified 20-25 patients (69-86%) correctly, 1 patient (3%) incorrectly and 3-8 patients (10-28%) as undetermined localization. In the 8 patients classified as undetermined localization by at least one interpreter, the quantitative EEG method classified 5 patients correctly, 1 patient incorrectly and 2 patients as undetermined localization. These results suggest that quantitative EEG might improve the reliability of ictal EEG localization and potentially reduce the need for invasive intracranial EEG monitoring.