The purpose of the study described here was to investigate the reliability of impedance cardiography (IC) in measuring cardiac output (CO) and central blood volume. Absolute values and changes in these variables obtained by impedance cardiography and by isotope- or thermodilution techniques were compared. The reproducibility of IC within the same day was compared with that of isotope dilution and the reproducibility in IC from day to day was derived. Finally, the effects of the readings of impedance tracings by different observers were quantified. The results are based on 270 measurements in 37 healthy subjects and in 25 unmedicated patients with ischaemic heart disease. We obtained significant correlations between absolute values (y = 0.68x + 1.48) and changes (y = 1.00x + 0.0003) in CO measured by IC and isotope- or thermodilution. IC significantly overestimated absolute values of CO (P less than 0.001). We found a qualitative but no quantitative correlation between thoracic fluid volume measured by IC and central blood volume measured by isotope dilution. IC was highly reproducible both when studies were repeated within the same day (SD on differences in CO = 0.36 1 min-1 for IC; SD on differences in CO = 0.30 1 min-1 for isotope dilution) and on different days (SD on differences in CO = 0.45 1 min-1). A low intra-observer variability was found (SD on differences in CO = 0.12 1 min-1). We conclude that impedance cardiography is reliable in measuring changes in cardiac output and thus suitable for repeated measurements in studies on the haemodynamic effects of physiological or pharmacological intervention. Impedance cardiography is sufficiently reliable for comparison of absolute values of CO between different groups of patients. We cannot recommend impedance cardiography for quantitative studies of central blood volume.