In 1976 we reviewed a randomly selected cohort of 227 patients with duodenal ulcer first diagnosed in 1963. The cohort comprised cases diagnosed in both hospitals and general practice. Fifty patients had died, 12 had emigrated, and 154 (93%) of the remaining patients were interviewed. Fifty-seven medically treated patients had no symptoms, 44 had mild symptoms, and 19 had more severe symptoms. The remaining 34 patients had been treated surgically. Cases diagnosed in hospital had a more severe prognosis than those diagnosed in general practice. A random sample of 65 general practitioners and 78 medical and surgical gastroenterologists tried to predict the results of this study. The range of the predictions was very wide showing that individual prognostic estimates were highly unreliable. The mean prediction by all doctors differed little from the actual result, suggesting that the collective experience of the medical profession is more reliable. The predictions of general practitioners, physicians, and surgeons showed small systematic differences, presumably reflecting the different types of patients they treat.