We determined the incidence of exercise-induced arterial hypoxaemia and its determinants in sixteen highly trained, healthy runners who were capable of achieving and sustaining very high metabolic rates (maximal O2 uptake = 72 +/- 2 ml kg-1 min-1 or 4.81 +/- 0.13 l min-1). Arterial blood gases and acid-base status were determined at each load of a progressive short-term exercise test and repeatedly during constant-load treadmill running while breathing air and during inhalation of mildly hypoxic, hyperoxic, and helium-enriched gases. Three types of responses to heavy and maximum exercise were evident and highly reproducible within subjects. Four runners maintained arterial PO2 (Pa, O2) within 10 mmHg of resting values, another four showed 10-15 mmHg reductions in Pa, O2, and the remaining eight showed reductions of 21-35 mmHg, i.e. in all cases to a Pa, O2 of less than 75 mmHg and to less than 60 mmHg in two cases. During constant-load exercise, Pa, O2 was often maintained during the initial 30 s when hyperventilation was greatest; then hypoxaemia occurred and in most cases was either maintained or worsened over the ensuing 3-4 min. The most severe hypoxaemia during heavy exercise was associated with no or little alveolar hyperventilation (Pa, CO2 greater than 35 mmHg and PA, O2 less than 110 mmHg) and an alveolar to arterial PO2 difference [(A-a)DO2] in excess of 40 mmHg. During 3-4 min of heavy exercise alveolar PO2 (PA, O2) decreased by 20 mmHg in mild hypoxia (0.17 FI, O2; inspired % O2) and increased by 20 mmHg during mild hyperoxia (0.24 FI, O2) and 10 mmHg during the hyperventilation which accompanied normoxic helium breathing. In all cases Pa, O2 changed in proportion to PA, O2 with no consistent change in the alveolar to arterial PO2 difference [(A-a)DO2]. In view of the correction of hypoxaemia with mild hyperoxia and the very high ratios of alveolar ventilation to pulmonary blood flow (VA/QC = 4-6) achieved during heavy exercise, we think it unlikely that non-uniformity of the VA/QC distribution or veno-arterial shunt could explain the hypoxaemia observed in most of our subjects. By exclusion, we suggest that hypoxaemia may be attributed to a diffusion limitation secondary to very short red cell transit times in at least a portion of the pulmonary circulation.(ABSTRACT TRUNCATED AT 400 WORDS)