1. Rapid steady-state CO2 responses were determined in six normal subjects at rest and five subjects at four different work loads up to 125 W, by injecting pure CO2 at constant flow into a small mixing chamber in the inspiratory limb of a breathing circuit.
2. The time course of the response of ventilation (V) and mean alveolar Pco2 (Paco2) was checked in separate experiments, where the flow rate of injected CO2 was changed abruptly and the effects were followed for 10 min.
3. V and Paco2 were measured every breath, and the results ensemble-averaged for each subject (two or three runs per subject) and then for the groups as a whole, in 30 s or 60 s time bins.
4. Paco2 during exercise was estimated by graphical reconstruction from the sloping alveolar plateau, and separately by the empirical equation of Jones, Robertson & Kane . At rest, Paco2 was assumed equal to end-tidal Pco2 (Petco2).
5. With the constant inflow technique, 4 min was required to reach steady-state V and Paco2 during exercise, and 6 min at rest.
6. At rest, with 4 min steps (doubtful steady state) the averaged CO2 response was concave up. With 6 min steps the response was almost linear. In neither case was the deviation from linearity statistically significant.
7. During exercise, the averaged CO2 responses were essentially isocapnic at work loads greater than 75 W with either method of deriving Paco2.