Since arm exercise affects the respiratory muscles the CO2 rebreathing method for determining cardiac output (Q) has to be evaluated during arm exercise. The purpose of this study was (1) to compare three different methods of determining arterial CO2 tension (PaCO2) during arm exercise, (2) to verify the linearity of the relationship between Q and oxygen uptake (VO2) during arm exercise, and (3) to investigate whether the CO2 rebreathing method according to Collier can determine accurately Q during arm exercise. Sixty male subjects performed arm-cranking exercise at 20%, 40% and 60% of their individual maximal load. Carbon dioxide output (VCO2) was measured by gas exchange measurement, and mixed venous CO2 tension (PvCO2) was determined from the CO2 rebreathing plateau at each exercise level. PaCO2 was estimated in three different ways: (A) by the modified Bohr formula for dead space, (B) by an arterialized blood sample from the hyperaemic ear-lobe, and (C) by the end-expiratory CO2 tension. A, B, and C were used to calculate Qa, Qb and Qc, respectively. The Pearson's correlation coefficient was high (P < 0.01) among the three different ways of estimating PaCO2. The Q-VO2 relationship proved to be linear (P < 0.01). The Q-values showed a good agreement with the direct Fick measurements, and were in the same range compared to other results obtained by dye dilution, electrical impedance cardiography and the exponential CO2 rebreathing method during arm exercise. In conclusion, the CO2 rebreathing method appeared to be accurate to determine Q during submaximal arm exercise.