1. Endurance athletes (E) undergo a marked reduction of arterial O2 saturation (Sa,O2) at maximal exercise in normoxia, which disappears when they breathe hyperoxic mixtures. In addition, at a given level of hypoxia, the drop in maximal O2 consumption (VO2,max) is positively related to the individual normoxic VO2,max. 2. These data suggest that the curve relating VO2,max to PI,O2 may be steeper and perhaps less curved in E than in sedentary subjects (S) with low VO2,max values because of the greater hypoxaemia in the latter, whence the hypotheses that (i) the relationship between VO2,max and PI,O2 may be set by the shape of the oxygen equilibrium curve; and (ii) the differences between E and S may be due to the different position on the oxygen equilibrium curve on which these subjects operate. These hypotheses have been tested by performing a systematic comparison of the VO2,max or Sa,O2 vs. PI,O2 relationships in E and S. 3. On ten subjects (five S and five E), VO2,max was measured by standard procedure during cycloergometric exercise. Sa,O2 was measured by finger-tip infrared oximetry. Arterialized blood PO2 (Pa,O2) and PCO2 (Pa,CO2) were determined in 80 microliters blood samples from an ear lobe. The subjects breathed ambient air or a N2-O2 mixture with an inspired O2 fraction (FI,O2) of 0.30, 0.18, 0.16, 0.13 and 0.11, respectively, VO2,max was normalized with respect to that obtained at the highest FI,O2. 4. The relationships between Sa,O2 or normalized VO2,max and FI,O2 (or PI,O2) had similar shapes, the data for E being systematically below and significantly different from those for S. Linear relationships between Sa,O2 and normalized VO2,max, statistically equal between E and S, were found. 5. We conclude that the relationships between either VO2,max or Sa,O2 and FI,O2 (or Pa,O2) may indeed be the mirror images of one another, implying a strict link between the decrease of VO2,max in hypoxia and the shape of the oxygen equilibrium curve, as hypothesized.