Clinical laboratory and endoscopic data were collected prospectively in 268 patients with bleeding gastric ulcer who were admitted between September 1985 and November 1987. There were 22 deaths, giving a hospital mortality rate of 8.2%. Surgery was undertaken in 68 patients (25.4%) with a mortality rate of 17.6% (11.8% at 30 days). There was one fatality in 104 (1.0%) patients less than or equal to 60 years compared with 21 deaths (12.8%) in patients greater than 60 years (P less than 0.001). Cirrhosis (P less than 0.01), malignant disease (P less than 0.03), chronic obstructive airways disease (P less than 0.02), congestive cardiac failure (P less than 0.02) and ischaemic heart disease (P less than 0.08) were each associated with an increased risk of mortality. Outcome in patients greater than 60 years was related to systolic blood pressure at admission (P less than 0.03), haemoglobin (P less than 0.02), serum bilirubin (P less than 0.02), and total transfusion requirements (P less than 0.001). For ulcers less than or equal to 1 cm, 1- less than or equal to 2 cm, greater than 2 cm in size, mortality rates were 1.9%, 11.4% and 18.0%, respectively. Initial endoscopy findings of a visible vessel, fresh blood, or active spurting/oozing haemorrhage were associated with rebleeding rates necessitating emergency surgery of 30.0%, 36.4% and 40.0%, respectively. There was no evidence of rebleeding in 187 patients (79.9%) managed conservatively and only five patients (2.7%) in this group succumbed, whereas rebleeding did occur in 47 patients (20.1%) with 13 subsequent deaths (27.7%; P less than 0.001). In patients greater than 60 years the presence of endoscopic stigmata of recent haemorrhage should lead to early consideration of therapeutic endoscopy and/or early surgery, particularly for ulcers greater than 1 cm in size.