Ascitic fluid total protein, albumin, cholesterol, their ascites/serum ratios, serum-ascites albumin and cholesterol gradients were measured for their ability to differentiate cirrhotic, malignant and tuberculous ascites in 76 patients. The mean +/- s.d. ascitic fluid total protein, albumin, cholesterol, their respective ascitic fluid/serum ratios in cirrhotic ascites were lower than malignant and tuberculous groups (P < 0.001 for each). The difference between malignant and tuberculous groups was significant for ascitic fluid/serum total protein (P < 0.05) and ascitic fluid/serum albumin (P < 0.01) only. Mean serum-ascites albumin gradient in cirrhotics was higher than in the malignant and tuberculous groups (P < 0.001 for each). The difference between malignant and tuberculous groups was significant (P < 0.01). Mean +/- s.d. serum-ascites cholesterol gradient in cirrhotics was higher than that in malignant and tuberculous groups (P < 0.001 for each). The difference between malignant and tuberculous groups was also significant (P < 0.01). Both serum/ascitic fluid total protein less than 0.5 and ascitic fluid cholesterol less than 55 mg/dL had 94% diagnostic accuracy for differentiating cirrhotic from malignant and tuberculous differentiating cirrhotic from malignant and tuberculous ascites. Serum ascitic fluid albumin gradient greater than 1.1 g/dL, ascitic fluid/serum albumin less than 0.65 and ascitic fluid albumin less than 2 g/dL had diagnostic accuracy of 92, 92 and 91%, respectively. Ascitic fluid total protein had diagnostic accuracy of 88%. None of the tests was able to differentiate between malignant and tuberculous ascites. Measurement of ascitic fluid cholesterol concentration is a simple method of differentiating cirrhotic from non-cirrhotic ascites.