Values of effective pulmonary blood flow (Q(EP)) and cardiac output, determined by a non-invasive foreign gas rebreathing method (CO(RB)) using a new infrared photoacoustic gas analysing system, were compared with measurements of cardiac output obtained by the direct Fick (CO(FICK)) and thermodilution (CO(TD)) methods in patients with heart failure or pulmonary hypertension. In 11 patients, of which three had shunt flow through areas without significant gas exchange, the mean difference (bias) and limits of agreement (+/- 2 S.D.) were 0.6 +/- 1.2 litre x min(-1) when comparing CO(FICK) and Q(EP), and -0.8 +/- 1.3 litre x min(-1) when comparing CO(FICK) and CO(TD). When correction for intrapulmonary shunt flow was applied (i.e. calculation of CO(RB)) in all 11 patients, the bias between CO(FICK) and CO(RB) was 0.1 +/- 0.9 litre x min(-1), primarily because agreement improved in the three patients with significant shunt flow. In the eight patients without significant shunt flow, the agreement between Q(EP) and CO(FICK) was 0.3 +/- 0.9 litre x min(-1). In conclusion, a foreign gas rebreathing method with a new infrared photoacoustic gas analyser provided at least as reliable a measure of cardiac output as did thermodilution. In the absence of significant shunt flow, measurement of Q(EP) itself provides a reliable estimate of cardiac output in heart failure patients. The infrared photoacoustic gas analyser markedly facilitates clinical use of the rebreathing method in general, which makes the method available to a larger group of clinicians working with patients with cardiovascular diseases.