Recent studies have shown that most newly diagnosed epileptic patients can be satisfactorily treated with a single antiepileptic drug. We therefore undertook a prospective randomised pragmatic trial of the comparative efficacy and toxicity of four major antiepileptic drugs, utilised as monotherapy in newly diagnosed epileptic patients. Between 1981 and 1987 243 adult patients aged 16 years or over, newly referred to two district general hospitals with a minimum of two previously untreated tonic-clonic or partial with or without secondary generalised seizures were randomly allocated to treatment with phenobarbitone, phenytoin, carbamazepine, or sodium valproate. The protocol was designed to conform with standard clinical practice. Efficacy was assessed by time to first seizure after the start of treatment and time to enter one year remission. The overall outcome with all of the four drugs was good with 27% remaining seizure free and 75% entering one year of remission by three years of follow up. No significant differences between the four drugs were found for either measure of efficacy at one, two, or three years of follow up. The overall incidence of unacceptable side effects, necessitating withdrawal of the randomised drug, was 10%. For the individual drugs phenobarbitone (22%) was more likely to be withdrawn than phenytoin (3%), carbamazepine (11%), and sodium valproate (5%). In patients with newly diagnosed tonic-clonic or partial with or without secondary generalised seizures, the choice of drug will be more influenced by considerations of toxicity and costs.