BACKGROUND Knowledge of pleural malignancy can lead to immediate pleurodesis during thoracoscopy. However, the accuracy of pleural frozen sections is largely unknown. OBJECTIVE To investigate the accuracy of frozen sections in pleural tumor pathology. METHODS A total of 156 frozen pleural sections performed with the question of malignancy were retrospectively reviewed. The original frozen sections were compared to the permanent section slides which were considered as the gold standard. The influence of the following parameters on the frozen section response was evaluated: specimen size, clinical information, as well as the processing by a specialized pulmonary pathologist or not. The reasons of discrepancies were categorized as sampling errors or interpretation errors. RESULTS Frozen sections made up 16.4% of 951 pleural biopsies performed in the same time period. Accurate diagnosis was feasible in 92.3% of the cases. There were 7 (4.5%) deferred (inconclusive) cases and 5 (3.2%) discrepant cases. Sensitivity of the method was 96.26%, specificity 97.87%, the positive predictive value was 99.04%, and the negative predictive value was 92%. There was no association between the responses given during frozen section and specimen size, clinical information, or the evaluation by a specialized pulmonary pathologist. Four of the 12 cases were sampling errors, while 8 cases were interpretation errors mostly made in the absence of fat tissue invasion. Thus, paucicellular lesions without prominent invasion - fat invasion or haphazardly invading cellular proliferation - were those posing most of the difficulties during frozen section. CONCLUSIONS Frozen sections are a highly accurate tool in pleural pathology. Thus, they can be used when an immediate pleurodesis is requested.