Four different measures (PETCO2, PACO2, PADCO2, and PJCO2) for indirectly estimating arterial PCO2 (PaCO2) from respired gas at the mouth have been investigated. PETCO2 was the end-tidal PCO2. PACO2 was calculated using a reconstruction of the alveolar oscillation of PCO2 obtained from the end-tidal "plateau" in PCO2. PADCO2 was calculated as for PACO2 except that the effects of dead space were incorporated. PJCO2 was calculated from an empirical relationship involving PETCO2 and tidal volume. Six subjects were studied at rest and during cycle ergometry at 50 and 100 W while breathing a variety of gas mixtures. Arterial samples were drawn for determination of true PaCO2. The differences for each method between estimated and true PaCO2 at rest and at 50 and 100 W were as follows: PETCO2, -1.35 +/- 2.64, 1.67 +/- 2.31, and 2.67 +/- 2.02 (SD) Torr; PaCO2, -2.15 +/- 2.73, -0.80 +/- 2.18, and -0.35 +/- 2.31 (SD) Torr; PADCO2, -1.55 +/- 2.54, 0.25 +/- 2.16, and 0.63 +/- 2.26 (SD) Torr; and PJCO2, -1.41 +/- 2.30, 0.12 +/- 1.79, and 0.08 +/- 1.96 (SD) Torr. It is concluded that, at rest, all methods significantly underestimate true PaCO2 and during exercise PETCO2 significantly overestimates PaCO2, but no bias was detected for any of the other methods.