1. Sramek and Bernstein's method of impedance cardiography is a simple, non-invasive and inexpensive computerised way of measuring stroke volume and systolic time intervals. In this study measurements made using the method were compared with those found simultaneously using established reference techniques. 2. In healthy volunteers there was no significant bias (d) and narrow 95% limits of agreement (d +/- 2s) when impedance and mechanophonocardiographic measurements of pre-ejection period (PEP, d = 0.3, d + 2s = 7.3, d-2s = -6.6 ms), ventricular ejection time (VET, d = 1.5, d + 2s = 17.7, d-2s = 14.6 ms) and PEP/VET ratio were compared. 3. In critically ill patients there was moderate agreement between impedance and thermodilution measurements of stroke volume (d = 8.1 (P < 0.05), d + 2s = 35.5, d-2s = -19.4 ml) and drug-induced changes in stroke volume were accurately detected. 4. In healthy volunteers agreement between impedance and dye dilution measurements of stroke volume was moderate, and similar at rest and during exercise (d = 3.4, d-2s = -31.1, d + 2s = 37.9 ml), however impedance underestimated exercise-induced increases in stroke volume (P < 0.05). 5. In patients with coronary heart disease impedance measurements correlated with angiographic left ventricular ejection fraction included the PEP/VET ratio (r = -0.81), stroke volume index (r = 0.65) and Heather index (r = 0.58, all P < 0.001), however the PEP/VET ratio could not be used to estimate the left ventricular ejection fraction with sufficient accuracy. 6. This impedance method provides reproducible semi-quantitative measurements of cardiac performance and blood flow. Its use for making pharmacodynamic measurements can be justified when invasive methods are considered inappropriate.