To review the indications and accuracy of diagnostic thoracoscopy for pleural effusions of unknown origin.Retrospective review of consecutive patients referred for diagnostic thoracoscopy over a 5-year period from 1 January 1989 to 31 December 1993.Tertiary referral cardiothoracic unit.Thirty-four patients referred from either medical or oncology services within a university-affiliated academic complex.All patients had diagnostic thoracoscopy performed under general anaesthesia. Retrospective data were collected in respect of presenting symptoms, gross findings, final pathological findings, amount of drainage, length of hospital stay and complications of the procedure. In 7 patients (21%), iodised talc was insufflated at the same time to create pleurodesis.Final diagnoses were: 17 (50%) malignant disease, 6 (18%) tuberculosis and 9 (26%) 'negative' pathology. In 2 (6%), further intervention was required to make a conclusive diagnosis. The diagnostic sensitivity for malignant disease was 89% and the specificity 100%. For pleural tuberculosis both the sensitivity and specificity were 100%. For 'negative' diagnoses the negative predictive value was 82%. A history of fever and sweats had a marked association (P = 0.002) with the final diagnosis of tuberculosis. No association could be identified between the gross observations at the time of thoracoscopy and the final diagnosis. The average length of hospital stay was 6.7 (range 1-25) days. There was 1 in-hospital death (3%), and 9 patients (26%) had major complications related to the procedure.Diagnostic thoracoscopy is a useful modality for obtaining a diagnosis in effusions of unknown origin where other methods have failed. The presence of symptoms such as fever and sweats is highly associated with a final diagnosis of tuberculosis.