The performance of cytological diagnosis in serous effusions was evaluated through a historical large series study which was carried out at Department of Surgical Pathology of Hospital A.C. Camargo. Files from the period of 1966 to 1990 were reviewed. Out of 4297 serous effusions samples (from 3379 patients), 2520 were pleural, 1763 ascitic and 14 pericardiac, with cytological report of malignancy in 917, 688 and 4 cases, respectively. Cytological diagnoses were confirmed true or false after correlation with the final diagnosis of the patients (clinical follow-up and/or histological evaluation). The results observed were 1982 false-negative, 21 false-positive, 1588 true-positive and 468 true-negative cases. Suspicious (161 or 3.75%) and inconclusive (77 or 1.79%) cases were not considered for statistical evaluation. Sensitivity, specificity, efficiency, as well as positive and negative predictive values were 44.5%, 95.7%, 50.1%, 98.7% and 20%, respectively. These values in pleural and ascitic effusions, separately showed similar performance, which was not observed in pericardiac samples due to small casistics. These data show low sensitivity and negative predictive values which contrast with the high specificity and positive predictive values that can be partially explained by the methodological limitations and invasive features of neoplasias.