The management of duodenal perforation associated with endoscopic sphincterotomy is controversial. Despite the fact that many patients recover without surgery, surgical opinion tends to favor immediate operation upon diagnosis since the mortality is high when sepsis is advanced. To refine the criteria for operative management, all duodenal perforations after endoscopic sphincterotomy over a 5-year period were studied. In a series of 464 consecutive endoscopic sphincterotomies, 8 duodenal perforations occurred; additionally, 4 patients with duodenal perforation were referred from elsewhere for management. Six patients were managed initially with nonoperative treatment (group I), and six underwent exploratory surgery upon diagnosis or hospital transfer (group II). One patient in group I was operated on 4 days after diagnosis. Of the seven surgically treated patients, three had repair of the duodenal perforation and drainage of the abscess or phlegmon, but four had no gross inflammation or visible duodenal perforation requiring repair at exploration. The clinical features of abdominal pain with physical signs significantly correlated with operative findings of pus or phlegmon (p < 0.05). Improvement in symptoms within 24 hours is correlated with spontaneous recovery (p < 0.01). Neither the presence of retroperitoneal air nor contrast leak is predictive of the need for surgery, and neither correlated with the size of the perforation. It is concluded that duodenal perforation may be treated successfully without surgery when the symptoms are mild and improve rapidly with medical treatment, but surgery should be undertaken if pain and abdominal signs are prominent, if suppuration is suspected, or if symptoms do not improve after a brief period of nonoperative management.