We discontinued anticonvulsant drugs in 92 patients who had been free of seizures during two years of treatment with a single drug. All the patients had epilepsy that had previously been untreated, and had been randomly assigned to receive carbamazepine, phenytoin, or sodium valproate. Thirty-one patients relapsed, and 61 remained free of seizures. The mean duration of the follow-up in the patients remaining free of seizures was 35 months (range, 6 to 62). There was no significant difference between the relapse rate among adults (35 percent) and that among children (31 percent). Our results suggest that the number of seizures a patient had before control was achieved, the number of drugs tried as single-drug therapy, and the type of treatment withdrawn all influenced the outcome. Among the various types of seizures, complex partial seizures with secondary generalization carried the worst prognosis. In comparison, the risk of relapse was 65 percent lower in patients with generalized seizures and 97 percent lower in patients with complex or simple partial seizures in the absence of secondary generalized attacks. Among the four electroencephalographic classes, class 4 (abnormal before treatment and unchanged before withdrawal) carried the worst prognosis. The risk of relapse was 94 to 99 percent lower in patients in the other three electroencephalographic classes. Among the three anticonvulsants, withdrawal of sodium valproate carried the worst prognosis. In comparison, the odds of relapsing were 28 percent lower after withdrawal of phenytoin and 85 percent lower after withdrawal of carbamazepine. We conclude that withdrawal of anticonvulsant medication should be considered in patients free of seizures for two years.