We assessed the clinical utility of [11C]flumazenil-PET (FMZ-PET) prospectively in 100 epileptic patients undergoing a pre-surgical evaluation, and defined the specific contribution of this neuro-imaging technique with respect to those of MRI and [18F]fluorodeoxyglucose-PET (FDG-PET). All patients benefited from a long term video-EEG monitoring, whereas an intracranial EEG investigation was performed in 40 cases. Most of our patients (73%) demonstrated a FMZ-PET abnormality; this hit rate was significantly higher in temporal lobe epilepsy (94%) than in other types of epilepsy (50%) (P < 0.001). Most FMZ-PET findings coexisted with a MRI abnormality (81%), including hippocampal atrophy (35%) and focal hypometabolism on FDG-PET (89%). The area of decreased FMZ binding was often smaller than that of glucose hypometabolism (48%) or larger than that of the MRI abnormality (28%). FMZ-PET did not prove superior to FDG-PET in assessing the extent of the ictal onset zone, as defined by intracranial EEG recordings. However, it provided useful data which were complementary to those of MRI and FDG-PET in three situations: (i) in temporal lobe epilepsy associated with MRI signs of hippocampal sclerosis, FMZ-PET abnormalities delineated the site of seizure onset precisely, whenever they were coextensive with FDG-PET abnormalities; (ii) in bi-temporal epilepsy, FMZ-PET helped to confirm the bilateral origin of seizures by showing a specific pattern of decreased FMZ binding in both temporal lobes in 33% of cases; (iii) in patients with a unilateral cryptogenic frontal lobe epilepsy, FMZ-PET provided further evidence of the side and site of seizure onset in 55% of cases. Thus, FMZ-PET deserves to be included in the pre-surgical evaluation of these specific categories of epileptic patients, representing approximately half of the population considered for epilepsy surgery.