The overall accuracy of cardiac output measurements made by impedance cardiography during maximum exercise was studied in man. Initially, the systematic error of the technique was assessed over the range 3.5 to 18 litre . min-1 by comparing with simultaneous measurements of cardiac output made using the direct Fick method. No systemic error was demonstrated in 40 estimations made in 20 subjects. The random error was assessed in 4 subjects in a steady state at rest and during exercise at 80 and 130 W and found to be less than 5% in each subject. The reproducibility of maximum exercise response was assessed in six healthy male subjects (age 26.2 +/- 4.4 years, +/- SEM) who underwent maximum exercise tests twice, 1 week apart, on a bicycle ergometer. Simultaneous recordings of cardiac output and oxygen uptake (VO2) at rest and during each 3 min stage of exercise were made. Highly significant correlations were obtained in the stroke volume (r = 0.84, p less than 0.001), cardiac output (r = 0.98, p less than 0.001) and VO2 (r = 0.98, p less than 0.001) between the two tests. Average maximum cardiac output was 27.0 +/- 1.2 litre . min-1 (+/- SEM) and maximum VO2 was 4.4 +/- 0.2 litre . min-1 (+/- SEM). These results show that measurements of cardiac output were reproducible over one week. Impedance cardiography is non-invasive technique which is as accurate as invasive methods and can be used for maximal exercise testing.