One of the greatest challenges in exercise physiology is to develop a valid, reliable, non-invasive and affordable measurement of cardiac output (CO). The purpose of this study was to evaluate the reproducibility and accuracy of a new impedance cardiograph device, the Physio Flow, during a 1-min step incremental exercise test from rest to maximal peak effort. A group of 12 subjects was evaluated to determine the reproducibility of the method as follows: (1) each subject performed two comparable tests while their CO was measured by impedance cardiography using the new device (COImp1, COImp2), and (2) in a subgroup of 7 subjects CO was also determined by the direct Fick method (COFick) during the second test. The mean difference between the values obtained by impedance (i.e. COImp1-COImp2) was -0.009 l.min-1 (95% confidence interval: -4.2 l.min-1, 4.2 l.min-1), and CO ranged from 3.55 l.min-1 to 26.75 l.min-1 (n = 146). When expressed as a percentage, the difference (COImp1-COImp2) did not vary with increasing CO. The correlation coefficient between the values of COImp and COFick obtained during the second exercise test was r = 0.94 (P < 0.01, n = 50). The mean difference expressed as percentage was -2.78% (95% confidence interval: -27.44%, 21.78%). We conclude that COImp provides a clinically acceptable evaluation of CO in healthy subjects during an incremental exercise.