By repeated treatment with phenytoin, subgroups of animals with different response to phenytoin were selected from a large group of amygdala-kindled rats. In one subgroup ('phenytoin responders') phenytoin induced reproducible increases in focal seizure threshold, while in another subgroup ('phenytoin nonresponders') animals were resistant to phenytoin. These phenytoin nonresponders, which comprised about 12% of the kindled animals tested, did not differ from phenytoin responders in drug levels, drug adverse effects or location of the stimulation electrode in the amygdala. Treatment of phenytoin responders and nonresponders with other primary antiepileptic drugs showed that valproate and phenobarbital induced much smaller increases in focal seizure threshold in phenytoin nonresponders than in responders, whereas carbamazepine induced about the same threshold increase in both groups. The novel antiepileptic drug vigabatrin, which acts by increasing GABA levels, exerted anticonvulsant effects in phenytoin responders but was inactive in nonresponders. Determination of plasma amino acids before and after vigabatrin treatment demonstrated marked differences in biochemical responses to vigabatrin although drug levels were about the same in both groups. The data demonstrate that amygdala-kindled rats with phenytoin resistance offer a basic approach to the investigation of mechanisms of drug resistance in epilepsy. Furthermore, these animals may be used in the evaluation of new anticonvulsant drugs for treatment of partial seizures which do not respond to the currently available therapies.