We tested whether the kinetics of systemic O₂ delivery (QaO₂) at exercise start was faster than that of lung O₂ uptake (VO₂), being dictated by that of cardiac output (Q), and whether changes in Q would explain the postulated rapid phase of the VO₂ increase. Simultaneous determinations of beat-by-beat (BBB) Q and QaO₂, and breath-by-breath VO₂ at the onset of constant load exercises at 50 and 100 W were obtained on six men (age 24.2 + 3.2 years, maximal aerobic power 333 + 61 W). VO₂ was determined using Gronlund's algorithm. Q was computed from BBB stroke volume (Qst, from arterial pulse pressure profiles) and heart rate (fH, electrocardiograpy) and calibrated against a steady-state method. This, along with the time course of hemoglobin concentration and arterial O₂ saturation (infrared oximetry) allowed computation of BBB QaO₂. The Q, QaO₂ and VO₂ kinetics were analysed with single and double exponential models. fH, Qst, Q and VO₂ increased upon exercise onset to reach a new steady state. The kinetics of QaO₂ had the same time constants (N.S.) as that of Q. The latter was twofold faster than that of VO₂. The VO₂ kinetics was faster than previously reported for muscle phosphocreatine decrease. Within a two-phase model, because of the Fick equation, the amplitude of phase I Q changes fully explained the phase I VO₂ increase. We conclude that in unsteady states lung VO₂ is dissociated from muscle O₂ consumption. The two components of Q and QaO₂ kinetics may reflect vagal withdrawal and sympathetic activation.