Summary Pulse oximetry probes placed peripherally may fail to give accurate values of blood oxygen saturation when the peripheral circulation is poor. Because central blood flow may be preferentially preserved, we investigated the oesophagus as an alternative monitoring site. A reflectance blood oxygen saturation probe was developed and evaluated in 49 patients undergoing cardiothoracic surgery. The oesophageal pulse oximeter results were in good agreement with oxygen saturation measurements obtained by a blood gas analyser, a COβoximeter and a commercial finger pulse oximeter. The median (IQR [range]) difference between the oesophageal oxygen saturation results and those from blood gas analysis were 0.00 (β0.30 to 0.30 [β4.47 to 2.60]), and between the oesophageal oxygen saturation results and those from COβoximetry were 0.75 (0.30 to 1.20 [β1.80 to 1.80]). BlandβAltman analysis showed that the bias and the limits of agreement between the oesophageal and finger pulse oximeters were β0.3% and β3.3 to 2.7%, respectively. In five (10.2%) patients, the finger pulse oximeter failed for at least 10βmin, whereas the oesophageal readings remained reliable. The results suggest that the oesophagus may be used as an alternative monitoring site for pulse oximetry even in patients with compromised peripheral perfusion.