BACKGROUND--Despite extensive investigations some patients with suspected lung cancer may undergo thoracotomy without preoperative histological proof of malignancy. A questionnaire on the use of histological examination of peroperative frozen sections in such patients was sent to 50 thoracic surgeons. Replies were received from 41 surgeons and indicated an absence of consensus on the usefulness of histological examination of frozen sections in this context, confirming the need for this study. METHODS--During one year 60 consecutive patients undergoing thoracotomy for suspected lung cancer without a prior histological diagnosis were studied prospectively. At thoracotomy the surgeon assessed the lesion macroscopically and a verdict on whether it was malignant was recorded. A biopsy specimen was then taken for examination of a frozen section and the result recorded. The appropriate operation was performed and the surgeon's verdict and the report on the frozen section were compared with the definitive histological diagnosis based on a paraffin section. RESULTS--Of 50 malignant lesions, 43 were identified by the surgeon and 47 by examination of the frozen section (sensitivity 86% and 94% respectively). Of 10 benign lesions, four were identified by the surgeon and nine by examination of the frozen section (specificity 40% and 90% respectively). CONCLUSIONS--Clinical and macroscopic assessment at thoracotomy are inferior to examination of frozen sections in suspected lung cancer, particularly where the lesion is benign. Lung resection should not be performed without examination of peroperative frozen sections when thoracotomy is performed for suspected but unproved lung cancer.