In a two-year period, 289 patients with acute upper-gastrointestinal hemorrhage were evaluated. The site of hemorrhage was proven by endoscopy in more than 90% of cases. The frequency of any particular bleeding site was characterized for several subgroups and was found to correlate with: (1) the severity of the bleeding episode; (2) the iatrotropic stimulus of hemorrhage, and (3) the history of ethanol consumption. No difference was found in the frequency of ulcer and acute mucosal lesions as sites of hemorrhage between patients who consumed aspirin and those who did not. A history of symptoms of peptic ulcer correlated with the presence of an ulcer as the bleeding site. It is possible that failure to identify important patient subgroups may be partially responsible for the disagreement between previous reports.