Cardiac output (CO) determination by thermodilution, which was introduced by Fegler in 1954, has gained wide acceptance in clinical medicine and animal experiments because it has several advantages over other methods with respect to simplicity, accuracy, reproducibility, repeated measurements at short intervals, and because there is no need for blood withdrawal. However, errors in determination of CO by thermodilution may be introduced by technical factors and the patients' pathological conditions. The current review summarizes these issues and provides our recommendations, based on the medical literature published between 1954-1992. To obtain more reproducible and accurate CO values by thermodilution, one should make several determinations (1) by using 10 ml injectate at room temperature for adults and 0.15 ml.kg-1 injectate for infants and children; (2) at evenly spaced intervals of the ventilation cycle; (3) when rapid intravenous fluid administration is discontinued; (4) by observing thermodilution curves so that baseline pulmonary artery temperature drift or the existence of intra- and extracardiac shunts are noticed. Finally, CO determination by thermodilution may be unreliable or impossible in patients with low CO states and tricuspid or pulmonary regurgitation. Since non-invasive CO monitoring has not replaced CO determination by thermodilution, intimate knowledge of this method is crucial for anaesthetists to prevent errors in the management of patients.