Pleural fluid analysis is often the initial diagnostic test used to determine the cause of a pleural effusion. We prospectively studied 33 consecutive patients with pleural effusions to determine whether the fluid arose from a transudative or an exudative process. Clinical judgment by an internist before thoracentesis and both serum and pleural fluid protein and lactic dehydrogenase levels (commonly referred to as "Light's criteria") were compared to the patient's final diagnosis. The internist correctly classified 15 of 17 exudative processes and all 16 transudative processes; the presence of any one of Light's three criteria correctly classified 15 of 17 exudative processes, whereas the absence of all three criteria correctly classified 14 of 16 transudative processes. Clinical judgment and Light's criteria are comparable in their ability to predict whether an exudative or transudative process was responsible for the effusion. Both methods are associated with errors, though of different kinds; these errors occurred infrequently. Recognizing the limitations of these methods will permit the most accurate effusion categorization.