In patients with gliomas who were stable or improving, we noted a disparity between clinical status and computed tomography (CT) brain scan results. To elucidate this finding, 29 patients were sequentially scanned with 2.0 mCi of 201Tl (5-30 min), 20 mCi [99mTc]gluceptate (GH) (3-4 hr) and 7-10 mCi 67Ga (48-72 hr). A total of 198 images were obtained. A set of three scans at a midpoint in follow up was selected for analysis. Seven patients who died had neuropathologic data available; brain sections were reconstructed to match radionuclide views without knowledge of image results. In the seven patients with autopsy data, 201Tl offered the most accurate correlation with viable tumor. Gallium-67 gave similar results in patients not receiving steroids. Technetium-99m GH scans could not allow differentiation between tumor, necrosis, and edema. Similarly, the CT scan could not routinely differentiate between fibrotic, nonfibrotic, necrotic, and neoplastic tissue. In the 22 patients without autopsy data, 201Tl scans commonly showed smaller and more focal abnormal uptake when compared with [99mTc]GH and 67Ga scans. Thallium-201 scans more accurately reflect viable tumor burden than other radionuclide studies of primary brain tumors, are minimally affected by concomitant steroid administration, can be performed immediately following tracer administration, and complement the anatomic data obtained from CT scans.