Background. Vigabatrin is a newly licensed drug for use in patients with epilepsy. We investigated whether this drug was comparable to standard first-line monotherapy in efficacy and incidence of adverse events. Methods. We enrolled 459 patients with newly diagnosed, untreated partial epileptic seizures from 44 centres and randomly assigned them carbamazepine 600 mg daily (n = 230) or vigabatrin 2 g daily (n = 229). After initial maintenance doses were reached, doses were adjusted downwards (in the case of adverse events) or upwards (in the case of seizures) by the clinician. The primary outcome was time to withdrawal because of lack of efficacy or adverse events. Secondary outcomes included efficacy (time to 6-month remission of seizures, time to first seizure after initial dose stabilisation), and adverse events (incidence and severity). Analysis was by intention to treat. Findings. Time to withdrawal for lack of efficacy or adverse events did not differ between groups (p = 0.318). Vigabatrin was better tolerated than carbamazepine with fewer withdrawals, but was more frequently associated with psychiatric symptoms (58 [25%] vs 34 [15%]) and weight gain (25 [11%] vs 12 [5%]). Carbamazepine was associated with rash (22 [10%] vs seven [3%]). All efficacy outcomes favoured carbamazepine and failed to show equivalence between the two drugs. No significant difference was found for time to achieve 6 months of remission from seizures (p = 0.058), but the most powerful outcome, time to first seizure after the first 6 weeks from randomisation, showed carbamazepine to be significantly more effective than vigabatrin (p = 0.0001). Interpretation. Vigabatrin seems less effective but better tolerated than carbamazepine, which is the first-choice drug for the treatment of partial epilepsies. Vigabatrin cannot therefore be recommended as a first-line drug for monotherapy in this group of patients.