Purpose: Magnetic resonance imaging (MRI) accurately identifies mesial temporal sclerosis (MTS), but prediction of successful surgical outcome ranges from 62% to 96% in published studies. Prior investigations only used patients who had received anterior temporal lobectomy (ATL), potentially overestimating the predictive value of MRI-identified MTS (MRI-MTS). Methods: The authors performed an intent-to-treat analysis of 90 consecutive patients assessed for possible ATL, including 13 who did not undergo ATL because of inconclusive intracranial ictal EEG. Four (31%) of these 13 patients had unilateral mesial temporal abnormalities on their MRIs. Results: The positive predictive value of MRI-MTS for seizure cessation decreased from 0.69 to 0.63 after adjustment for these additional false positive results. Four previous studies had revealed a positive predictive value of 0.75 (0.72 after similar adjustment). Conclusions: The authors conclude that the predictive value of MRI-MTS for outcome from ATL may be overestimated by small retrospective studies of highly selected postoperative patients.