Twenty-five patients with tumors of the central nervous system received bromodeoxyuridine (BUdR), 200 mg/sq m, by intravenous infusion every 8 hours for 3 days before surgery. Excised tumor specimens were fixed in chilled 70% ethanol, embedded in paraffin, and cut into 6-micron sections. Each section was reacted with monoclonal antibodies against BUdR and stained with immunoperoxidase to identify nuclei that had incorporated BUdR. The growth fraction of each tumor was estimated by calculating the ratio of BUdR-positive nuclei to the total number of tumor cells in three to six microscopic fields in viable areas of the tumor. In seven cases, the tumor doubling time was measured from the serial computerized tomography scans and an attempt was made to estimate the cell cycle time. The growth fractions ranged from 9.1% to 46.5% in malignant gliomas, 2.0% to 6.7% in low-grade gliomas, 11.2% to 43.2% in metastatic brain tumors, 0.8% to 1.9% in pituitary adenomas, 3.9% to 4.6% in acoustic neurinomas, and 6.2% to 8.2% in meningiomas and cerebellar hemangioblastomas. The estimated cell cycle time was 5 to 12 days in most malignant gliomas and brain metastases; however, the actual cell cycle time should be substantially shorter because cell loss was not considered in the calculation. Although the growth fraction appeared to correlate with the biological malignancy of each tumor, the tumor doubling time did not reflect growth potential. It is possible that unpredictable cell loss plays an important role in tumor growth at certain sizes. Therefore, the cell cycle times calculated in this study are considerably overestimated and should be interpreted with caution.