A noninvasive method to estimate cardiac output Q without special patient cooperation was developed by modifying a previous acetylene-helium (C2H2-He) rebreathing technique (ART). Estimation of Q using ART is based on a single-compartment model that is valid only under prescribed breathing; e.g., fast, deep breathing, and emptying of the rebreathing bag on each breath. To make the ART less dependent on subject cooperation, a more sophisticated mathematical model and estimation method are needed. For this purpose, we modeled the C2H2 and He concentration dynamics at the mouth over successive breaths using a multi-compartment model. This model takes into account the effects of breathing pattern, compartmental volumes, and gas solubility. From computer simulations and sensitivity analysis, we found that Q could be estimated from the available data with adequate precision. Our model and estimation method were tested on a group of six normal adult subjects, at rest and during submaximal exercise (75 watts). Estimates of Q from our new method (6.5 +/- 0.4 L/min at rest, 12.5 +/- 0.4 L/min at 75 watts) were in agreement with those obtained using a previous ART (7.0 +/- 0.3 L/min at rest, 12.6 +/- 0.5 L/min at 75 watts). We conclude that this approach promises to provide reliable estimates of Q in patients (e.g., children and elderly), at rest and during exercise, without the need of prescribed breathing patterns or changes in rebreathing bag volume.