Unlike neurological deficits resulting from intracranial hemorrhage in patients with cerebral arteriovenous malformation (AVM), which have well defined etiology, the pathogenesis and treatment of progressive and/or fluctuating non-hemorrhagic neurological and psychological deficits require clarification. Values for local cerebral blood flow (LCBF) and local partition coefficients (L lambda) were measured by the method using stable xenon (Xe)-computerized tomography (CT-CBF) scanning, and were compared to 133Xe inhalation values using external probes in 16 patients with cerebral AVM's. Values were measured by both methods before and after total excision of AVM's in five cases. Neurological and mental status assessments were compared with LCBF results. Clinical improvement was most evident after total excision of AVM's. Other procedures, such as clipping of vessels, partial excision, and ventriculoperitoneal shunting for hydrocephalus, were associated with frequent complications. Embolization carried risks of cerebral infarction and was not efficacious unless combined with excision. Medical treatment resulted in poor or unsuccessful seizure control, with neurological deterioration despite anticonvulsant therapy. Compared with age-matched normal individuals, LCBF values in patients with AVM's were significantly reduced, particularly adjacent to the AVM's. Mean L lambda values for gray and white matter were normal. After excision, LCBF values in gray and white matter increased significantly up to normal. Due to overestimation of CBF by shunt flow with the 133Xe method of measurement, no correlation was found with results of the CT-CBF method before AVM excision, but significant correlation of the two methods resulted after excision. If accurate LCBF values are obtained by high resolution, direct measurement of tracer clearance from brain tissue, progressive and/or fluctuating neuropsychological deficits correlate with the degree of cerebral steal.