The objective of the study was to compare the serum-effusion albumin gradient (serum albumin level minus pleural effusion albumin level) to Light's traditional criteria (pleural fluid/serum total protein ratio greater than 0.5, pleural fluid/serum LDH ratio greater than 0.6, and pleural fluid LDH greater than 200 U/L) for identifying exudative pleural effusions. The design included prospective measurement of the serum-effusion albumin gradient and Light's criteria in patients with pleural effusions in an inpatient ward in a military teaching hospital. Fifty-nine consecutive patients with pleural effusions who were undergoing diagnostic or therapeutic thoracentesis in whom the etiology of the effusion could be determined were studied. Serum and pleural effusion fluid chemistries were measured in order to determine both the serum-effusion albumin gradient and Light's criteria. Using an albumin gradient of 1.2 g/dl or less to indicate exudates and greater than 1.2 g/dl to indicate transudates, 57 of the 59 patients (41 exudates; 18 transudates) were correctly classified. Two patients with malignant effusions were misclassified as having transudates. Although Light's criteria correctly identified all of the exudates, five patients with congestive heart failure were misclassified as exudates. Four of these patients had had previous diuretic therapy, and all had a clinical response to further diuretic therapy. We conclude that although Light's criteria for exudates are very sensitive, an albumin gradient of 1.2 g/dl or less tends to be more specific, especially in cases of chronic congestive heart failure.