We compared the relative sensitivity of two interictal PET techniques, bolus injection of [15O] labeled water for estimation of cerebral blood flow (H2(15)O CBF-PET), and 18F 2-deoxyglucose (18FDG-PET) for cerebral glucose metabolism (CMRglc), and T2-weighted magnetic resonance imaging, in 28 patients with medically intractable complex partial seizures undergoing evaluation for surgery. There were statistically significant associations between lateralization by 18FDG-PET, and MRI, but not H2(15)O CBF-PET, and lateralization of the epileptic focus as defined by scalp-sphenoidal ictal EEG. Fifteen patients had surgery or subdural electrodes. 18FDG-PET was more closely associated with a good outcome than H2(15)O CBF-PET, which, in addition, showed hypoperfusion contralateral to the epileptic temporal lobe in several cases. H2(15)O sensitivity may have been reduced by technical factors, but 18FDG-PET appears to be more specific for localization of epileptic zones.