The thermal dilution technique (TD) of measuring cardiac output (Q) has been assumed to be inaccurate when the tricuspid valve is regurgitant (TR). The indicator transit time from the right atrium to pulmonary artery (PA) is prolonged and may increase indicator loss to the heart and adjacent tissue. We surgically created TR in four anaesthetized dogs and compared simultaneous cardiac output measurement made with an implanted PA flow probe with intermittent TD cardiac output measurements. We found an excellent correlation (r = 0.98) between the techniques in the normal heart when Q was increased by 100-150 per cent of control values using an intravenous dobutamine infusion. After TR was produced by incorporating the tricuspid valve leaflet within a suture, Q decreased. Dextran 40 was then infused and Q increased significantly. The rate at which Q increased was greater with the TD technique than with the PA flow probe; however, a significant (r = 0.85) linear relationship was still present with TR. The altered thermal waveform detected by the PA thermistor in TR was characterized by a lower peak amplitude and a slower return to baseline. Within the limitations of an animal model, our data suggest that TD cardiac output may be more accurate than previously assumed, particularly in low output states.