Complex problems occur concerning the diagnosis and treatment of GI bleeding. How often can the exact cause and site of the bleeding be determined? What are the advantages and complications of various forms of treatment? There are no universal answers, applicable to all situations. Nevertheless, certain assumptions seem to be justified. Diagnosis of the cause of severe bleeding, often suggested by the patient's history, is confirmed most effectively by selective angiography, which should give an answer in over 95% of cases. The causes of slow, persistent bleeding, manifested either by gross blood or by positive tests for blood in the stools, are diagnosed in about the same percentage of cases by endoscopy. It is when the bleeding is intermittent and of minimum or moderate severity that determination of the site of origin is the most difficult; angiography cannot help in the majority of cases, and endoscopy can miss tiny lesions that are not bleeding at the time of examination. However, laparotomy, especially when it is combined with intraoperative enteroscopy or angiography, should reduce the number of unsolved diagnoses to under 5% of the total. Surgery is the primary method of therapy in many cases. Primary control of bleeding also can be obtained in approximately 80% of cases of severe bleeding by endoscopy, by the use of coagulation or laser, or by selective arterial infusion of pitressin. Recurrences must be treated by surgery, except in some instances where selective embolization by the angiographer may result in a cure. The surgeon must attempt to cure the hemorrhage and to prevent recurrence at a later date by such measures as gastric resection and vagotomy for bleeding duodenal ulcers to prevent stomal ulcers, or by subtotal rather than segmental colectomy for widespread colonic diverticulosis. There will be an irreducible percentage of lesions that continue to develop in other sections of the GI tract after successful primary treatment of the original cause of bleeding; AVMs, angiodysplasia, and vasculitis are examples. Fortunately, such diseases are relatively rare. It is not unreasonable today to expect that the measures discussed herein will cure bleeding in 95% of cases; if death occurs it rarely is due to hemorrhage, but to some type of organ failure.