The cell type of bronchial carcinoma predicted from the results of bronchial biopsy at fibreoptic or rigid bronchoscopy or of percutaneous aspiration lung biopsy was compared with the type determined by histological examination of specimens obtained by thoracotomy, biopsy of an extrapulmonary metastasis, or necropsy in 180 cases. The rates of agreement with the final diagnosis were 95.7% for bronchial biopsy through the fibreoptic bronchoscope and 86.5% through the rigid bronchoscope. For percutaneous biopsy, which was usually carried out on tumours inaccessible to the bronchoscope, the rate of agreement was 61%, significantly lower than by the other methods (p less than 0.001). The diagnosis of oat-cell carcinoma by any technique was very reliable. Bronchial biopsy was more reliable than was percutaneous biopsy in diagnosing squamous-cell carcinoma. With any technique the commonest error was the incorrect diagnosis of squamous-cell carcinoma or adenocarcinoma as large-cell undifferentiated carcinoma.