One hundred randomly selected inpatients were examined with both double-contrast barium meal and endoscopy in a blinded prospective fashion. All studies were done by staff personnel, with equal clinical information available to both the radiologist and endoscopist. The final diagnosis was made by a review committee of participating radiologists and endoscopists. Endoscopy was more sensitive (92% versus 54%, p less than 0.001) and specific (100% versus 91%, p less than 0.05) than the double-contrast barium meal. Both procedures significantly affected the clinical outcome of the patient, the effect of endoscopy being significantly greater than that of the double-contrast barium meal. Although errors with the barium study related predominantly to an inability to show subtle lesions, poor patient cooperation and perceptual and technical failures were additional significant factors. Endoscopy is recommended for certain groups of patients.