In a series of experiments involving measurement of left ventricular blood flow by thermodilution in the coronary sinus, a consistent, but unexpected, variation in calculated coronary sinus flow, related to respiration, was seen. In some patients the variation was small, but in others it was as much as 300%. To discover its cause, five patients were investigated prospectively with continuous right atrial pressure monitoring during coronary flow measurement and respiratory manoeuvres. In three, ice-cold saline was injected into the right atrium during respiratory manoeuvres while coronary sinus temperature was monitored continuously and the position of the catheter was monitored fluoroscopically. The cause was found to be movement of the catheter such that thermistor approached the right atrium, and the variation was therefore spurious. We report these findings and discuss other known causes of inaccurate coronary flow measurement using the thermodilution technique.