In the past, preoperative pulmonary function abnormalities have identified a group of patients in danger of postoperative cardiorespiratory morbidity and mortality. We selected a group of 56 patients, each of whom had a lung mass and had demonstrated significant abnormalities in screening pulmonary function. By using temporary unilateral pulmonary artery occlusion and quantitative macroaggregate lung scanning, we then studied these patients for split pulmonary function. Those patients whose noncancerous lung had a calculated forced expiratory volume in 1 sec greater than 800 ml and a circulation that could accommodate all of the cardiac output without producing hypertension or arterial hypoxemia were offered thoracotomy. Of the 56 patients, we judged 6 to be physiologically inoperable and did not offer surgery. Another 4 patients were not offered surgery, and 4 refused surgery. Forty-two patients underwent surgical exploration-of these, 17 then had a pneumonectomy and 13, a lobectomy. Of the 30 patients resected, 6 died in surgery (4 from respiratory insufficiency). These cardiorespiratory mortality rates (neumonectomy, 17.6 per cent; lobectomy, 7.7 per cent) are lower than those reported previously when patients had equivalent pulmonary function abnormality. A follow-up of 49 of 56 patients revealed that 59 per cent of the patients undergoing either pneumonectomy or lobectomy were still living 1 to 3 years after the resection. Our results suggested that the preoperative testing of split pulmonary function permitted an attempt at surgery in patients who might otherwise be considered inoperable by history, physical examination, screening pulmonary function tests alone.