The relative roles of ventilation-perfusion (VA/Q) inequality, alveolar-capillary diffusion resistance, postpulmonary shunt, and gas phase diffusion limitation in determining arterial PO2 (PaO2) were assessed in nine normal unacclimatized men at rest and during bicycle exercise at sea level and three simulated altitudes (5,000, 10,000, and 15,000 ft; barometric pressures = 632, 523, and 429 Torr). We measured mixed expired and arterial inert and respiratory gases, minute ventilation, and cardiac output. Using the multiple inert gas elimination technique, PaO2 and the arterial O2 concentration expected from VA/Q inequality alone were compared with actual values, lower measured PaO2 indicating alveolar-capillary diffusion disequilibrium for O2. At sea level, alveolar-arterial PO2 differences were approximately 10 Torr at rest, increasing to approximately 20 Torr at a metabolic consumption of O2 (VO2) of 3 l/min. There was no evidence for diffusion disequilibrium, similar results being obtained at 5,000 ft. At 10 and 15,000 ft, resting alveolar-arterial PO2 difference was less than at sea level with no diffusion disequilibrium. During exercise, alveolar-arterial PO2 difference increased considerably more than expected from VA/Q mismatch alone. For example, at VO2 of 2.5 l/min at 10,000 ft, total alveolar-arterial PO2 difference was 30 Torr and that due to VA/Q mismatch alone was 15 Torr. At 15,000 ft and VO2 of 1.5 l/min, these values were 25 and 10 Torr, respectively. Expected and actual PaO2 agreed during 100% O2 breathing at 15,000 ft, excluding postpulmonary shunt as a cause of the larger alveolar-arterial O2 difference than accountable by inert gas exchange.