Over a 6-month period, 157 patients, 89 of whom had central nervous system tumors, were examined on a prototype 0.12 T resistive nuclear magnetic resonance (NMR) imaging unit. All of the patients had computed tomography (CT), which was used as a standard to which the NMR findings were compared. Studies were done primarily by saturation-recovery technique with short repetition times. The signal intensity with saturation-recovery technique did not allow differentiation among most tumor types. Location, extent, and morphology helped to some extent in attempts at differentiation. In the multiplanar mode, NMR compared favorably to CT with regard to lesion detection. Limited early experience suggests that NMR also may detect some lesions when the CT is negative and may detect additional lesions when one or more are present. The NMR examination was well tolerated by selected patients.