In order to evaluate a new thoracic electrical bioimpedance (TEB) system for measurement of stroke volume based on the Sramek-Bernstein equation, 391 paired values of cardiac output were measured simultaneously with the standard thermodilution method. These values were obtained from 16 patients selected for having the most severe illness during a 6-month period; the intent was to evaluate the bioimpedance method in the worst possible situations. The correlation coefficient (r) was 0.83, slope was 0.87, intercept was 1.53, and the mean difference between the two methods was 16.2 +/- 11.8 (SD)% in the total series. In 285 paired samples where satisfactory conditions were met, r was 0.90, slope was 0.98, intercept was 0.34, and the mean difference was 11.8 +/- 8.9%. The data indicate satisfactory correlations between these two methods. When the TEB waveform is satisfactory, the agreement between TEB and thermodilution is as good as the agreement between serial thermodilution methods. Difficulties may arise with dysrhythmias, tachycardia (heart rate greater than 150 beat/min), metal in the chest or chest wall, sepsis, hypertension, and extremely oily skin. Mechanical ventilation did not appear to be a problem.