OBJECTIVE: To compare the methods for continuous and bolus thermodilution cardiac output measurements. DESIGN: In vivo and in vitro experimental studies. SETTING: Surgical research division in a university hospital. SUBJECTS: Eight calves and flow bench model. INTERVENTIONS: Data were collected in vivo from eight calves instrumented with pulmonary artery catheters, which allowed both continuous and bolus thermodilution measurements. The pulmonary artery catheter was placed through the external jugular vein. All in vitro measurements were performed using a flow bench model. MEASUREMENTS AND MAIN RESULTS: A total of 232 bolus and continuous thermodilution measurements were analysed in vivo to determine the degree of agreement between the two methods. The absolute measurement bias was 0.14 L/min with 95% confidence limits ranging from -0.83 to 1.15 L/min. In vitro analysis of 576 measurements at six different temperature points (range 31 degrees to 41 degrees C), using clinically relevant flows (2 to 9 L/min), showed overestimation of flow values using continuous and bolus thermodilution methods. However, the continuous method showed better accuracy by a lower degree of overestimation. Systematic error was 9.7 +/- 8.4 (SD) % for continuous and 11.1 +/- 6.3% for the bolus method (p < .001). This effect was especially evident at lower flow rates. The influence of various temperatures on the accuracy and reproducibility of both methods of measurement was statistically significant but not clinically relevant. The infusion of lactated Ringer's lactate solution (infusion rates 100 to 1000 mL/hr) affects both methods at a low flow rate of 2 L/min, without causing a significant effect on continuous measurement at a higher flow rate (4 L/min). Shunting of 50% of circulating volume to the distal part of the thermal filament of the pulmonary catheter impaired the accuracy of continuous measurement without affecting results from bolus measurements (systematic error -26.8 +/- 8.2% for continuous and -5.2 +/- 4.1% for bolus thermodilution). CONCLUSIONS: Continuous thermodilution cardiac output measurement provided higher accuracy and greater resistance to thermal noise than standard bolus measurements. The correct placement of the catheter is essential for precise measurements.