The potential of differential ventilation (DV) with selective positive end-expiratory pressure (PEEP) has been tested versus conventional ventilation with and without general PEEP. Gas exchange and central haemodynamics were studied in 15 subjects with no clinical or radiological signs of pulmonary disease. The rationale of the method was to ensure ventilation of the well-perfused dependent lung and to counteract airway closure within that lung. The subjects were intubated with a double-lumen catheter prior to scheduled abdominal surgery. During general anaesthesia in the lateral posture, they were given DV. The mean inspired oxygen fraction was 0.32. Fifty per cent ("even" tidal volume (VT) distribution) or 70% ("inverted" VT distribution) of the inspired volume was administered to the dependent lung. Two synchronized ventilators were used. In eight subjects DV was also combined with PEEP applied solely to the dependent lung (selective PEEP). The major findings were that DV with even VT distribution reduced venous admixture by 26% (P less than 0.05) and the alveolo-arterial oxygen tension gradient (P(A-a)o2) by 30% (P less than 0.05) in comparison with conventional ventilation in the lateral position. The addition of selective PEEP further reduced the P(A-a)o2 by 13%. P(A-a)o2 was consequently 43% lower than during conventional ventilation without PEEP in the lateral posture (P less than 0.01). Selective PEEP also had less impact on cardiac output than general PEEP (P less than 0.05). It is concluded that DV with even distribution of VT and selective PEEP can reduce the P(A-a)o2 in anaesthetized lung-healthy subjects in the lateral position.