Over 150 cases of central nervous system tumors have been studied with positron emission tomography using fluorine-18-labeled fluorodeoxyglucose (18FDG) as a tracer. From this material 100 consecutive cases of cerebral glioma have been reviewed and analyzed. The results show a strong correlation of tumor grade with glycolytic rate, with visual "hot spots" present in all high-grade neoplasms and in only four low-grade tumors. The quantitative accuracy is limited by three basic factors. First, the measurement of tissue uptake, as compared with the parent technique, autoradiography, is more difficult because detection must be done outside the body. Effects such as scattered radiation and self-attenuation introduce errors unless properly corrected. A more serious problem when measuring small structures, such as a rim-shaped high-grade glioma, is the limited spatial resolution. The most advanced scanner, the Neuro-PET, has a resolution of 6 to 7 mm. Second, corrections are needed for backflow, including free tracer at the time of the scan that will return to the blood and "trapped" tracer that will backflow because of the presence of phosphatase. These corrections are calculated from the blood activity using nominal rate constants for 18FDG. Our study found no significant alteration in rate constants between normal and tumoral tissue. Finally, a lumped constant is needed to correct for kinetic differences between 18FDG and glucose. If there is a change in the mechanism of either membrane transport or the hexokinase reaction, the lumped constant may change. However, measurements of actual glucose utilization in tissue culture lines from six patients support the 18FDG results.