Between 1982 and 1986, 27 patients with seizure disorders due to cerebral arteriovenous malformation (AVM) were surgically treated by the authors. These patients had no history or clinical manifestation of intracranial hemorrhage. All were treated with anticonvulsant agents by their neurologists but became disabled due to inadequate control of seizures by medication, side effects of the anticonvulsant drugs, or the effects on their professional lives of even infrequent seizures. The age of the patients ranged from 13 to 61 years. There were 13 males and 14 females. The AVM's were smaller than 2 cm in four patients, between 2 and 4 cm in five, and larger than 4 cm in 18. The most frequent location of the AVM's was in the temporal lobe, followed by the frontal, parietal, and occipital lobes. All patients had preoperative electroencephalography (EEG) and intraoperative electrocorticography. Intraoperative recording of the amygdala and the hippocampus by depth electrodes was performed if the AVM's were located in the temporal lobe. Superficial or posterior temporal lobe AVM's often have remote seizure foci that involve the amygdala and hippocampus. All patients underwent craniotomy and total excision of their AVM's. Surgery was carried out under local anesthesia to allow localization by electrical stimulation if the AVM involved the speech area or the sensorimotor cortex. Based on the EEG findings, excision of the epileptogenic lesion in addition to the AVM was performed in 18 patients. In seven patients with AVM's located in the temporal lobe, remote seizure foci were identified and excised. The remote epileptic activity was particularly prominent in AVM's in the temporal lobe and usually involved mesial temporal structures. Microscopic study of excised seizure foci showed gliosis in 26 cases, hemosiderin deposits in 10, and focal hemorrhage in four. There were no operative deaths. Two patients developed a hemiparesis and three suffered temporary dysphasia after surgery. Two patients had visual field deficits. The results of postoperative seizure control during the average follow-up period of 3 years 11 months were excellent in 21 patients, good in three, fair in two, and poor in one. The latter patient, whose epileptic lesion was not completely excised because of its location in the motor cortex, had poor seizure control postoperatively. Another patient required a second operation to remove a remote seizure focus. In this series, proposed mechanisms of seizure associated with cerebral AVM include focal cerebral ischemia secondary to arteriovenous shunting, gliosis of the surrounding brain, and a secondary epileptogenesis in the temporal lobe.(ABSTRACT TRUNCATED AT 400 WORDS)