A double-blind randomized crossover design trial of carbamazepine (CBZ) and oxcarbazepine (OCBZ) was performed with 48 in-patients with epilepsy. All were stabilized on polytherapy including CBZ and had at least two seizures per week. CBZ was replaced by the trial medication. Each trial period started with a titration, followed by a 12-week steady state. Concomitant medications were kept constant during the trial. The criteria for assessment were seizure fit frequency and severity; tolerability; hematology and blood chemistry; plasma levels of antiepileptic drugs; EEG; cardiovascular parameters; and treatment preference. The following differences regarding OCBZ were detected: 9% reduction of the total number of seizures, with a significant reduction of tonic-clonic (20%) and tonic (31%) seizures; increased alertness and concentration ability in five patients; an allergic skin reaction with CBZ that completely disappeared in two patients while receiving OCBZ; an increase of valproate and phenytoin plasma levels in a number of patients, probably caused by reduced enzyme induction; a slight but significant reduction of serum Na, not causing clinical symptoms; less seizures than in the CBZ period in 25 patients (52%); and a preference for OCBZ in 23 patients (48%). We consider OCBZ at least as effective as CBZ with a slightly better tolerability. In severe cases, the wider therapeutic window might improve seizure control.