Of 2,545 in-patients with primary lung cancer, pathological proof of the diagnosis was obtained by sputum cytology in 48%, by bronchoscopy in 32%, by biopsy from miscellaneous sites in 12%, and at thoracotomy or necropsy only in 19%. The diagnosis was made solely on clinical and radiological evidence in 8%.One or more satisfactory specimens of sputum from 2,035 patients gave a positivity rate of 59%. When three specimens of sputum were tested the positivity rate was 69% and when four or more were tested, 85%. The maximum false positivity rate was 0-7%. Agreement between the types of malignant cells found in the sputum and in resected or necropsy specimens occurred in 84%. The pathological diagnosis was based solely on sputum cytology in 40% of the patients in whom malignant cells were found in the sputum.The total work of the sputum laboratories was analysed for one year (1967); sputum was examined from 1,598 patients, of whom 266 (17%) had positive results.Of 1,518 patients who had a bronchoscopy, the positivity rate was 53%, and of 524 patients with biopsies from miscellaneous sites the rate was 64%.Sputum cytology, as practised at present, is a very valuable method of diagnosis, but its demands on time and expertise make it unsuitable for general application outside large medical centres. Bartholomew's Hospital and 29 from Brompton Hospital could not be traced.Notes from a total of 2,545 patients were available for study, 913 from St. Bartholomew's and 1,632 from Brompton. There were 2,176 males and 369 (14%) females. One or more specimens of sputum were submitted for cytological investigation from 2,035 (80%) of the patients, bronchoscopy was performed in 1,518 (60%), and biopsies were taken from other sites in 524 (21%). Lung resection was performed in 1,002 (39%) and a further 199 (8%) (1-1%)The miscellaneous group included 10 bronchiolo-623 on 9 May 2018 by guest. Protected by copyright.