The aim of this study was to define the number of pleural biopsy samples necessary for optimum diagnostic performance and determine to what extent they are complementary. Eighty-four closed pleural biopsies were performed in our department between June 1996 and January 1998 on 55 males and 29 females with an average age of 64.4 +/- 16.7 years. The study of the pleural fluid included: pH, biochemical testing of pleura/serum (proteins, lactate dehydrogenase, glucose, cholesterol, triglycerides, albumin and adenosine deaminase), haemogram, cytology and microbiological testing (Gram-staining, aerobes, anaerobes and mycobacteriae cultures). The biopsies were performed using a Cope needle, with a total of five biopsies for each patient: four for pathological examination (taken numerically in the order in which they were performed: D1, D2, D3 and D4) and one for microbiological testing. In those cases in which the diagnosis was uncertain or effusion persisted, a thoracoscopy or thoracotomy was performed. There were no significant differences in the diagnostic yield of each individual sample (D1, D2, D3 and D4), but there were differences in the sum of the samples, depending on the number of biopsies performed.This was true for total group and the group with carcinomas, but not for the group with tuberculosis. The increase in diagnostic yield with the number of biopsies was more remarkable in the carcinoma cases, where it increased by 35% when four biopsies were performed (54% with one biopsy versus 89% with four biopsies, P < 0.002). In conclusion, the diagnostic yield increased with the number of biopsy samples in the total group and the group with malignancy but not in the group with tuberculous effusions. The best diagnostic performance for malignant pathology was obtained with four samples. In pleural tuberculosis, the diagnostic yield did not increase with more biopsy samples. One high quality sample should be enough to obtain a diagnosis.