The benefits of radiosurgery for cavernous malformations are difficult to assess because of the unclear natural history of this vascular lesion, the inability to image malformation vessels, and the lack of an imaging technique that defines "cure." The authors selected for radiosurgery 47 patients who harbored a hemorrhagic malformation in a critical intraparenchymal location remote from a pial or ependymal surface. Of these, 44 patients had experienced at least two hemorrhages before radiosurgery. The mean patient age was 39 years; six patients had previously undergone attempted surgical removal. The malformation was located in the pons/midbrain in 24 cases, the medulla in three, the thalamus in nine, the basal ganglia in three, deep in a parietal lobe in four, and deep in a temporal lobe in four. Patients had sustained initial hemorrhages from 0.5 to 12 years prior to radiosurgery (mean 4.12 years). In these patients, who were not typical of the majority of patients with cavernous malformations, there were 109 bleeds before radiosurgery in 193 prior observation-years, for 56.5% annual hemorrhage rate (including the first hemorrhage), or an annual rate of 32% subsequent to the first hemorrhage. The mean follow-up period after radiosurgery was 3.6 years (range 0.33-6.4 years). The proportion of patients with hemorrhage after radiosurgery was significantly reduced (p < 0.0001), as was the mean number of hemorrhages per patient (p = 0.00004). In the first 2 years after radiosurgery, there were seven bleeds in 80 observation-years (8.8% annual hemorrhage rate). In the 2- to 6-year interval after radiosurgery, the annual rate decreased to 1.1% (one bleed). After radiosurgery, 12 patients (26%) sustained neurological worsening that correlated with imaging changes. In eight patients these deficits were temporary; two underwent surgical resection and died. Two patients had new permanent deficits (4%). A significant reduction was observed in the hemorrhage rate after radiosurgery in patients who had deep hemorrhagic cavernous malformations, especially after a 2-year latency interval. This evidence provides further support to the belief that radiosurgery is an effective strategy for cavernous malformations, especially when located within the parenchyma of the brainstem or diencephalon.