In previous studies of human and experimental acute pancreatitis, three main assumptions have been made. First, that the disease is due to activation of pancreatic proteolytic enzymes in the pancreas with resulting "autodigestion" of the gland. Second, that interstitial pancreatitis is a mild form of hemorrhagic pancreatitis into which it may progress, and third, that bacteria play little part, if any, in the initiation of the disease. These assumptions are now questioned. In the present study in dogs, levels of proteolytic enzymes in blood, thoracicduct lymph and peritoneal fluid were measured using benzoylarginine amide. Raised levels of amidase were found in hemorrhagic, but not with interstitial, pancreatitis, and biochemical examination of amidase suggested it was not a pancreatic protease, but with its broad specificity and stability derived from bacteria. Addition of antibiotic to the blind duodenal loop in hemorrhagic pancreatitis reduced the level of blood amidase, but Trasylol given intravenously did not, nor did it inhibit amidase in vitro. In all animals, histological examination was made of the pancreas at time of death. On bacteriology, it is concluded that experimental interstitial pancreatitis results from damage to the pancreatic duct system without infection, and haemorrhagic pancreatitis mainly from reflux of bacteria into the pancreatic ducts from the duodenum. Only bacteria such as Escherichia coli and Clostridium welchii that produce proteolytic enzymes and cytotoxins appear to be able to cause haemorrhagic pancreatitis, and these bacteria may explain the release of vasoactive polypeptides and the vascular effects. In hemorrhagic pancreatitis such bacteria were found in the pancreas, but none in interstitial pancreatitis. Evidence is given to suggest that pancreatic proteolytic enzymes are unlikely to cause the cell necrosis which is a pathological feature of hemorrhagic pancreatitis, and that "autodigestion" is likewise unlikely to be a cause of this condition. An extrapancreatic source of proteolytic enzymes from bacteria is now suggested in haemorrhagic pancreatitis, and more attention to bacteriology in human acute pancreatitis is urgently needed. Amidase levels were highest in peritoneal fluid, suggesting a rationale for peritoneal lavage in the treatment of acute pancreatitis, and it is unlikely that Trasylol can give any benefit. The assessment of treatment of acute pancreatitis will be unsatisfactory as long as the proportion of haemorrhagic to interstitial pancreatitis in any series is not known accurately.