In a prospective study for the diagnosis of pancreatic cancer, ultrasonography, radionuclide scanning, selective arteriography, and endoscopic retrograde cholangiopancreatography (ERCP) were compared. Eighty-nine consecutive patients were investigated; 58 underwent laparotomy, and 36 were found to have periampullary cancer; seven had other malignant tumors within the abdomen, and nine had pancreatitis on biopsy. Five had other benign disease, and there was one negative laparotomy. Thirty-one patients who did not have laparotomy have shown no evidence of cancer at one year follow-up. Ultrasonography was found to be more reliable than scan or arteriography in the detection and diagnosis of a mass in the pancreas. ERCP achieved the highest rate of correct definitive diagnosis.