To prospectively assess the value of an algorithm in differentiating spontaneous from secondary bacterial peritonitis, we performed serial paracenteses in 43 episodes of ascitic fluid infection (28 spontaneous and 15 secondary) in 40 patients. The algorithm involved identification of (a) secondary peritonitis associated with gut perforation, based on previously proposed criteria in patients with neutrocytic ascites (ascitic fluid total protein greater than 1 g/dl, glucose less than 50 mg/dl, and lactate dehydrogenase greater than the upper limit of normal for serum) and (b) separation of spontaneous from secondary peritonitis (unassociated with perforation) based on the response of the ascitic fluid cell count to antibiotic therapy. The perforation criteria had 100% sensitivity in detecting episodes of actual gut perforation; their specificity, however, was low (45%). After 48 h of treatment the concentration of ascitic fluid neutrophils was below the baseline pretreatment value in all episodes of spontaneous peritonitis but in only two thirds of the patients with secondary peritonitis. This algorithm is useful in (a) identifying patients who have infected ascites associated with perforation of an intraabdominal viscus, and (b) differentiating spontaneous from nonperforation secondary peritonitis on the basis of the response of the ascitic fluid cell count to appropriate antibiotic therapy. The optimal time for repeat paracentesis in patients with infected ascites appears to be 48 h after initiation of treatment.