We investigated statistical parametric mapping (SPM) use for positron emission tomography (PET) with [(18)F]fluorodeoxyglucose (FDG) data analysis in mesial temporal lobe epilepsy. The study involved 14 patients with temporal lobe epilepsy ultimately treated by anterior temporal lobectomy. Surgical outcome in terms of seizure control was favorable in 12 patients. Two different SPM approaches were designed to analyze each FDG-PET scan: a direct comparison with a control group (n = 27) and a search for significant interhemispheric asymmetry considering the asymmetry existing in the control group. Statistical inference was performed, first, without correction for multiple comparisons (making the hypothesis of temporal hypometabolism) and, second, after correction for multiple comparisons. Search for temporal interhemispheric asymmetry under the hypothesis of temporal hypometabolism was the most reliable SPM approach: hypometabolism was identified on the side chosen for resection in most cases (sensitivity, 71%; specificity, 100%) and was predictive of favorable postsurgical outcome in 90% of the patients. There was no false-positive result within the control group using this approach. After correction for multiple comparisons, SPM also identified in some patients temporal hypermetabolic areas as well as extratemporal cortical and subcortical hypometabolic areas on the side of resection but also on the contralateral side. In a further step, SPM was used for a group analysis of patients with favorable outcome after reversing scans when needed to set an identical lateralization in all patients. This analysis identified multiple ipsilateral temporal and extratemporal hypometabolic regions; when temporal metabolic changes were specifically assessed, the contralateral mesiotemporal region was found hypermetabolic, possibly as a manifestation of compensatory mechanisms in the presence of a unilateral epileptogenic lesion.