OBJECTIVE The authors conducted a study to define the long-term outcomes and risks of stereotactic radiosurgery (SRS) for arteriovenous malformations (AVMs) of the basal ganglia and thalamus. METHODS Between 1987 and 2006, the authors performed Gamma Knife surgery in 996 patients with brain AVMs; 56 patients had AVMs of the basal ganglia and 77 had AVMs of the thalamus. In this series, 113 (85%) of 133 patients had a prior hemorrhage. The median target volume was 2.7 cm(3) (range 0.1-20.7 cm(3)) and the median margin dose was 20 Gy (range 15-25 Gy). RESULTS Obliteration of the AVM eventually was documented on MR imaging in 78 patients and on angiography in 63 patients in a median follow-up period of 61 months (range 2-265 months). The actuarial rates documenting total obliteration after radiosurgery were 57%, 70%, 72%, and 72% at 3, 4, 5, and 10 years, respectively. Factors associated with a higher rate of AVM obliteration included AVMs located in the basal ganglia, a smaller target volume, a smaller maximum diameter, and a higher margin dose. Fifteen (11%) of 133 patients suffered a hemorrhage during the latency period and 7 patients died. The rate of post-SRS AVM hemorrhage was 4.5%, 6.2%, 9.0%, 11.2%, and 15.4% at 1, 2, 3, 5, and 10 years, respectively. The overall annual hemorrhage rate was 4.7%. When 5 patients with 7 hemorrhages occurring earlier than 6 months after SRS were removed from this analysis, the annual hemorrhage rate decreased to 2.7%. Larger volume AVMs had a higher risk of hemorrhage after SRS. Permanent neurological deficits due to adverse radiation effects (AREs) developed in 6 patients (4.5%), and in 1 patient a delayed cyst developed 56 months after SRS. No patient died of AREs. Factors associated with a higher risk of symptomatic AREs were larger target volume, larger maximum diameter, lower margin dose, and a higher Pollock-Flickinger score. CONCLUSIONS Stereotactic radiosurgery is a gradually effective and relatively safe management option for deep-seated AVMs in the basal ganglia and thalamus. Although hemorrhage after obliteration did not occur in the present series, patients remain at risk during the latency interval between SRS and obliteration. The best candidates for SRS are patients with smaller volume AVMs located in the basal ganglia.