Thirty-five of 68 patients had arteriograms demonstrating gastrointestinal tract bleeding. Seventy-three percent were older than 40 years. Significant overlap existed between patients whose arteriograms showed bleeding and those whose did not, in the amount of blood administered before arteriography. However, 85% of patients receiving 5 units or more demonstrated bleeding. Seventeen of 22 patients who manifested signs of shock demonstrated bleeding, but 34% of cases which demonstrated bleeding manifested no shock. Preoperative localization of bleeding assisted the surgeon in cases where no abnormality was externally visible or palpable. The arteriogram showing no bleeding was not helpful. The clinical course, not the lack of demonstrable bleeding, determined management. A history of gastrointestinal tract bleeding or disease did not correlate with study results showing bleeding. Complication rate was low. The data show that emergency arteriography is helpful and accurate. Guidelines for patient selection are discussed.