It is widely agreed that after two or more seizures patients should be given antiepileptic treatment, but there is still controversy about the treatment of patients after a first unprovoked seizure. In a multicenter, randomized, open trial, patients with a first tonic-clonic seizure were randomized to immediate treatment (carbamazepine, phenytoin, phenobarbital, or sodium valproate) or to treatment only after another seizure. Fifty-two (24%) of the 215 patients randomized to immediate treatment and 85 (42%) of the 204 randomized to delayed treatment experienced seizure recurrence during follow-up. Age, acute treatment of the seizure with benzodiazepines, remote etiologic factors, and EEG abnormalities were significant predictors of relapse. Of the immediately treated patients, 87% had no seizures for a year and 68% had no seizures for 2 years, whereas only slightly fewer initially untreated patients (83% and 60%) achieved these endpoints. Patients treated after the first seizure and those treated after seizure relapse had the same time-dependent probability of achieving 1 and 2 seizure-free years. None of the variables that were prognostic predictors of relapse was significantly associated with the probability of having 1 or 2 years of seizure control. Anticonvulsants in patients presenting a first tonic-clonic seizure reduce the risk of relapse; however, 50% of patients who are not treated will never experience a second seizure. Moreover, the probability of long-term remission is not influenced by treatment of the first seizure.