We used actuarial methods to study outcome after temporal lobectomy in 135 consecutive patients classified into subgroups according to preoperative MRI findings. Sixty months after surgery, 69% of patients with foreign tissue lesions, 50% with hippocampal sclerosis, and 21% with normal MRIs had no postoperative seizures. An eventual seizure-free state of 2 years or more, whether the patient was seizure-free since surgery or not, was achieved by 80% of patients with foreign tissue lesions, 62% of those with hippocampal sclerosis, and 36% of those with normal MRIs. Outcome was worse in those with normal MRIs than in the other two groups. Early postoperative seizures with later remission (the "running down" phenomenon) occurred in all groups. Late seizure recurrence was present only in the hippocampal sclerosis group. These data show that preoperative MRI is a useful predictor of outcome and that actuarial analysis provides insight into different longitudinal patterns of outcome in MRI subgroups. This information can now be used in preoperative counseling.