The causes of exercise-induced hypoxemia (EIH) remain unclear. We studied the mechanisms of EIH in highly trained cyclists. Five subjects had no significant change from resting arterial PO(2) (Pa(O(2)); 92.1 +/- 2.6 Torr) during maximal exercise (C), and seven subjects (E) had a >10-Torr reduction in Pa(O(2)) (81.7 +/- 4.5 Torr). Later, they were studied at rest and during various exercise intensities by using the multiple inert gas elimination technique in normoxia and hypoxia (13.2% O(2)). During normoxia at 90% peak O(2) consumption, Pa(O(2)) was lower in E compared with C (87 +/- 4 vs. 97 +/- 6 Torr, P < 0.001) and alveolar-to-arterial O(2) tension difference (A-aDO(2)) was greater (33 +/- 4 vs. 23 +/- 1 Torr, P < 0. 001). Diffusion limitation accounted for 23 (E) and 13 Torr (C) of the A-aDO(2) (P < 0.01). There were no significant differences between groups in arterial PCO(2) (Pa(CO(2))) or ventilation-perfusion (VA/Q) inequality as measured by the log SD of the perfusion distribution (logSD(Q)). Stepwise multiple linear regression revealed that lung O(2) diffusing capacity (DL(O(2))), logSD(Q), and Pa(CO(2)) each accounted for approximately 30% of the variance in Pa(O(2)) (r = 0.95, P < 0.001). These data suggest that EIH has a multifactorial etiology related to DL(O(2)), VA/Q inequality, and ventilation.