We determined both (1) the reliability (reproducibility) of seizure onset location in patients with temporal lobe recorded with extracranial EEG and (2) if interictal spike (IIS) location influences that reliability. EEGs were recorded with the 10-20 system and sphenoidal electrodes. Between 1985 and 1993, 166 patients with suspected temporal lobe epilepsy were eligible for inclusion with a total of 734 seizures recorded. Each seizure onset was classified as either localized, lateralized, or nonlocalized. Individual patients were then grouped according to interictal spike location (i.e., unilateral, bilateral, or none) as well as by ictal onset reproducibility patterns (i.e., concordant, semiconcordant, conflicting, or nonlocalized) based on each patient's monitoring session. Seizure onset location was reproducible in 68% of the cases and variable patterns of seizure onset were seen in 32% of patients. Patients with unilateral IIS (68%) were more likely to have consistent ictal onset patterns than patients with bilateral IIS (27%). The first seizure more often predicted subsequent seizure onset patterns when it was localized than when it was nonlocalized. Of 18 patients with conflicting temporal lobe seizures, up to four seizures were necessary to detect independent bilateral temporal lobe seizures.