During the authors' initial 4-year experience with radiosurgery using the Leksell cobalt-60 gamma unit, they treated 121 patients with cerebral arteriovenous malformations (AVMs). The radiosurgical dose to the margin of the nidus was 20 Gy for lesions less than 2.0 cm in diameter (volume < or = 4.2 cm3); 18 Gy for malformations 2.1 to 3.0 cm in diameter (volume 4.2-14.1 cm3); and 16 Gy for malformations greater than 3.0 cm (volume > 14.1 cm3). Fifty-one patients underwent follow-up angiography between 1 and 3 years after treatment, and complete obliteration of the nidus was confirmed in 38 (74.5%) of these patients. Thirty-two (74.4%) of 43 AVMs with volumes of 10 cm3 or less and six (75%) of eight larger AVMs (volume 11-30 cm3) showed complete obliteration. Analysis of the time course of AVM nidus shrinkage and obliteration showed that most of the radiosurgically induced effect had occurred by 36 months after treatment. Retrospective analysis of the dose plans for 10 AVMs that were not obliterated by 36 months after gamma knife radiosurgery at the authors' institution (eight cases) or elsewhere (two cases) revealed that six AVMs had not been covered completely by the prescribed isodose. Six (5%) of the 121 patients developed neurological deficits as a direct result of radiosurgical treatment. The authors infer from these data that malformations up to 30 cm3 in volume (approximately 4.0 cm in average diameter) can be treated effectively with an acceptably low complication rate using a radiosurgical dose of 16 Gy to the margin of the nidus. The obliteration rate for the larger malformations that were treated with a dose of 16 to 18 Gy appears to be similar to that for smaller ones treated with 18 to 20 Gy. As more experience accrues using radiosurgery to treat AVMs, patient selection criteria and the variables associated with successful obliteration of the nidus should become more clearly defined.