Purpose: To evaluate the role of stereotactic radiosurgery in the management of recurrent malignant gliomas. Patients and Methods: We treated 35 patients with large (median treatment volume, 28 cm3) recurrent tumors that had failed to respond to conventional treatment. Twenty-six patients (74%) had glioblastomas multiforme (GBM) and nine (26%) had anaplastic astrocytomas (AA). Results: The mean time from diagnosis to radiosurgery was 10 months (range, 1 to 36), from radiosurgery to death, 8.0 months (range, 1 to 23). Twenty-one GBM (81%) and six AA (67%) patients have died. The actuarial survival time for all patients was 21 months from diagnosis and 8 months from radiosurgery. Twenty-two of 26 patients (85%) died of local or marginal failure, three (12%) of noncontiguous failure, and one (4%) of CSF dissemination. Age (P = .0405) was associated with improved survival on multivariate analysis, and age (P = .0110) and Karnofsky performance status (KPS) (P = .0285) on univariate analysis. Histology, treatment volume, and treatment dose were not significant variables by univariate analysis. Seven patients required surgical resection for increasing mass effect a mean of 4.0 months after radiosurgery, for an actuarial reoperation rate of 31%. Surgery did not significantly influence survival. At surgery, four patients had recurrent tumor, two had radiation necrosis, and one had both tumor and necrosis. The actuarial necrosis rate was 14% and the pathologic findings could have been predicted by the integrated logistic formula for developing symptomatic brain injury. Conclusion: Stereotactic radiosurgery appears to proofing survival for recurrent malignant gliomas and has a lower reoperative rate for symptomatic necrosis than does brachytherapy. Patterns of failure are similar for both of these techniques.