Lung perfusion was studied in 10 patients (mean age 58 yr) in the lateral position during enflurane anesthesia. They were ventilated through a double-lumen endotracheal catheter: 1) by one ventilator with free distribution of ventilation between the lungs, with no (zero) end-expiratory pressure (ZEEP); 2) as above but with a general positive end-expiratory pressure (PEEP) of 9 cmH2O; or 3) by two ventilators with equal distribution of ventilation between the lungs and with a selective PEEP of 8 cmH2O to the dependent lung only. Total ventilation was on average 8 l/min (BTPS) throughout the study. During the first method, 34% of ventilation was distributed to the dependent and 66% to the nondependent lung. Cardiac output (thermodilution) was 4.5 l/min, 57% being distributed to the dependent lung as assessed by iv boli of Xenon 133. During the second method, ventilation was assumed to be distributed equally between the lungs. Cardiac output was decreased to 3.8 l/min, and the dependent lung received 81% of lung blood flow. During the third method, cardiac output was significantly greater than during the second method (4.1 l/min), 51% passing to the dependent lung. Peak and end-inspiratory airway pressures were 5–18 cm H2O lower during selective than during general PEEP. Arterial oxygen tension was significantly greater during the third method than during either of the other ventilator settings and the alveolar-arterial oxygen tension difference was almost halved compared with the first method. It is concluded that differential ventilation with selective PEEP improves ventilation-perfusion matching and thus oxygenation.