Functional magnetic resonance imaging (fMRI) was performed in seven patients harboring intracerebral gliomas proven by histological analysis using a noninvasive blood oxygen level-dependent technique based on the documented discrepancy between regional increases in blood flow and oxygen use in response to regional brain activation. We combined fMRI with conventional magnetic resonance imaging (MRI) during motor or language task activation experiments to investigate the potential usefulness of mapping regional brain activity as part of treatment planning in patients with intracerebral gliomas, in whom preservation of areas of functioning brain tissue is critical. Statistical fMRI maps were generated and directly mapped onto conventional MRI scans obtained at the same session. Of the five patients cooperative enough to remain motionless for the study and perform the task, the location of activation in the primary sensorimotor cortex on the side of the tumor was clearly displaced compared with that in the normal contralateral hemisphere in four patients. Four of the five tumors in these patients showed fMRI activation within the periphery of (or immediately adjacent to) areas of presumed tumor based on spin-echo MRI. In some patients with neurological deficit, the extent of activation was reduced on the side of the tumor as compared with the normal hemisphere. The supplemental motor area and the ipsilateral primary motor cortex were also reproducibly activated during motor tasks. We conclude that blood oxygen level-dependent fMRI can localize areas of cortical function in patients undergoing treatment planning for gliomas so that therapy can be directed away from regions of residual function. Our preliminary data suggest that functioning cortex within or adjacent to tumor margins can be demonstrated, which may correspond to partial preservation of clinical function. Our preliminary data also suggest that there may be a quantifiable difference on fMRI between activation in tumor-bearing cortex and activation in corresponding normal cortex in the contralateral hemisphere. We postulate that the magnitude of this difference may relate to the severity of patient deficit.