We studied 74 consecutive patients with temporal lobe epilepsy who were treated surgically and in whom the volumes of mesial temporal structures were determined preoperatively by magnetic resonance imaging. We divided the patients into three groups according to the volumetric findings: unilateral (63.5% of the patients), bilateral (23%), or no atrophy (13.5%) of the amygdala-hippocampal formation. Two distinct surgical approaches were used: selective amygdalohippocampectomy (n = 37) or anterior temporal lobe resection (n = 37). Outcome was assessed at least 1 year after surgery, according to Engel's modified classification. Patients with unilateral mesial temporal atrophy had significantly better results compared with the other two groups (p < 0.001): We found excellent results (class I or II outcome) in 93.6% of the patients with unilateral atrophy, in 61.7% of those with bilateral atrophy, and in 50% of the group with no significant atrophy of mesial temporal structures. The two different surgical techniques were equally effective, regardless of the pattern of atrophy. In conclusion, magnetic resonance volumetric studies in temporal lobe epilepsy proved to be an important preoperative prognostic tool for surgical treatment, but they did not provide guidance for selecting one surgical approach compared to the other.