Thirty-one patients with severe drug-resistant epilepsy entered the study. Vigabatrin (2 to 3 g/d, stratified according to weight) and placebo were administered orally, as add-on therapy in random order under double-blind conditions, each for three months using a crossover design. Thirty patients completed both periods. Of these, ten patients (33%) showed a decrease in seizure frequency of 50% or more. In the 15 patients presenting with complex partial seizures, "temporal" electroencephalographic abnormalities, and relatively low seizure frequency, there was a significant reduction in seizure frequency during vigabatrin treatment. No significant treatment effect was found for the remaining 15 patients, who presented with mixed seizure types, multifocal electroencephalographic abnormalities, and high seizure frequencies. Tolerability to vigabatrin was good; the most frequently reported unwanted effect was drowsiness. Plasma concentrations of phenytoin showed a significant reduction during the vigabatrin period. The results demonstrate the efficacy and good tolerability of vigabatrin therapy in patients with severe complex partial epilepsy.