A clinicopathologic staging method for colorectal carcinoma was applied prospectively to 503 patients treated by surgical resection over a period of 7 1/2 years. The method grouped separately those patients known to be incurable at the time of resection and allowed for an anatomically precise definition of the extent of tumor spread. Survival studies showed that prognosis did not significantly deteriorate with spread of tumor beyond the bowel wall unless there were demonstrable metastases, infiltration of a free serosal surface, or if local resection was incomplete. Highly significant decrements in survival occurred when lymph node metastases were demonstrable and when unresected tumor was known to be present. The staging system from which these observations were made formed an improved guide to prognosis when compared with the original Dukes' method. Patients with histologically high-grade tumors had a poorer survival rate than those with low or average tumors with the same extent of spread.