To measure fetal cardiac output by thermodilution, we injected 2.5 ml of iced 5% glucose rapidly into the superior vena cava and recorded areas of temperature-time curves in the brachiocephalic artery (AS) and descending aorta (BS). One minute later we repeated the injection into the inferior vena cava and recorded areas under two curves (AI and BI). Cardiac output was calculated as Qco = H (AS + BI - AI - BS) (ASBI - AIBS), where H is the change in caloric content of the blood. In a mechanical model we found thermal results correlated well with timed collections (r = 0.98) with a slope of 0.995. In 7 acutely prepared fetal lambs, electromagnetic flow probe measurements of total cardiac output (descending aorta + brachiocephalic artery) correlated well (0.90) with thermal measurements. The slope (1.16) was significantly greater than 1.0, reflecting the inclusion of coronary and pulmonary flows in thermal, but not in flow probe, measurements. In 11 acutely and chronically prepared fetal lambs, the correlation between microsphere and thermal measurements was 0.81 with a slope of 1.09, not significantly different from 1.0. Hence, thermodilution gives results comparable to other methods and is convenient, indefinitely repeatable, inexpensive, and nonradioactive.