Estimation of cardiac output by impedance cardiography (QZ) in exercise during normal breathing (NB) has been limited by motion artifact. Our objective was to obtain readable impedance cardiograms on five subjects during upright cycle exercise at 0, 50, 100, 150, and 200 W to permit comparisons of QZ during NB, expiratory breath hold (EXP) and inspiratory breath hold (INSP). Q was also determined using an equilibration CO2 rebreathing method [Q(RB)]. QZ during NB exceeded EXP QZ at 100, 150, and 200 W, and exceeded INSP QZ at 100 W (P less than 0.05). The low EXP QZ values were due to a significantly lower stroke volume at 100, 150, and 200 W (P less than 0.05). For the INSP QZ at 100 W, heart rate was lower than during EXP (P less than 0.05). Regression of QZ (NB) against Q(RB) resulted in a linear relationship (r = 0.93) over the range of Q = 7-26 1/min. The slope of the regression differed significantly from 1.0 (P less than 0.05). We conclude that QZ values obtained during EXP or INSP should not be assumed to represent QZ during NB, at least at work rates greater than 50 W. A consequence of the linear relationship between QZ(NB) and Q(RB) over the range of 0-200 W is that estimates of CO2 rebreathing cardiac output can be obtained by impedance cardiography if QZ is adjusted using an appropriate empirical factor.