At a given steady O2 consumption (VO2) in normoxia, cardiac output (Q) is inversely proportional to arterial O2 concentration (CaO2), so that O2 delivery (QaO2=QCaO2) is kept constant and adapted to VO2. The matching between QaO2 and VO2 keeps O2 return (QvO2=QaO2-VO2) constant and independent of VO2 and haemoglobin concentration ([Hb]). This may not be so in hypoxia: in order for QvO2 to be independent of the inspired O2 fractions (FIO2), the slopes of the Q versus VO2 lines should be greater the lower the CaO2, which may not be the case. Thus, we tested the hypothesis of constant QvO2 by determining QaO2 and QvO2 in acute hypoxia. Thirteen subjects performed steady-state submaximal exercise on the cycle ergometer at 30, 60, 90 and 120 W breathing FIO2 of 0.21, 0.16, 0.13, 0.11 and 0.09. VO2 was measured by a metabolic cart, Q by CO2 rebreathing, [Hb] by a photometric technique and arterial O2, saturation (SaO2) by infrared oximetry. CaO2 was calculated from [Hb], SaO2 and the O2 binding coefficient of haemoglobin. The VO2 versus power relation was independent of FIO2. The relations between Q and VO2 were displaced upward and had higher slopes in hypoxia than in normoxia. However, the Q changes did not compensate for those in CaO2. The slopes of the QaO2 versus VO2, lines tended to decrease in hypoxia. QVO2 was lower the lower the FIO2. A significant relationship was found between QvO2 and SaO2 (QvO2= 1.442 SaO2+0.107, r=0.871, n=24, P<10(-7)), which confutes the hypothesis of constant QvO2 in hypoxia.