We assessed the linearity of oxygen uptake (VO2) kinetics for several work intensities in four trained cyclists. VO2 was measured breath by breath during transitions from 33 W (baseline) to work rates requiring 38, 54, 85, and 100% of maximal aerobic capacity (VO2max). Each subject repeated each work rate four times over 8 test days. In every case, three phases (phases 1,2, and 3) of the VO2 response could be identified. VO2 during phase 2 was fit by one of two models: model 1, a double exponential where both terms begin together close to the start of phase 2, and model 2, a double expontential where each of the exponential terms begins independently with separate time delays. VO2 rose linearly for the two lower work rates (slope 11 ml.min-1 W-1) but increased to a greater asymptote for the two heavier work rates. In all four subjects, for the two lighter work rates the double-exponential regression reduced to a single value for the time constant (average across subjects 16.1 +/- 7.7 s), indicating a truly monoexponential response. In addition, one of the responses to the heaviest work rate was monoexponential. For the remaining seven biexponential responses to the two heaviest work rates, model 2 produced a significantly better fit to the responses (P < 0.05), with a mean time delay for the slow component of 105 +/- 46 s. The amplitude of the primary component (A1) increased linearly across all four work rates (slope 9.1 ml.min-1 W-1, r = 0.957). In contrast, the time constant for the fast component (tau-1) exhibited no significantly change as work rate increased. Thus, although the overall oxygen cost increases nonlinearly with work intensity, the primary component of the VO2 response exhibits approximately linear behavior: linear increase in gain (A1) and invariant time constant (tau-1).