Oxygen haemoglobin saturations and plethysmograph signal amplitudes were recorded from two oximeters placed on the fingers and toes of 10 patients undergoing oesophagectomy, to assess a method of removing motion artefact from saturation recordings. By examining changes in the plethysmograph amplitude that preceded changes in saturation, episodes of desaturation caused by movement artefacts were removed from the data. The reliability of the method was then determined by scrutinizing two concurrent oximetric profiles from each patient. A total of 1600 h of data were evaluated. Desaturations occurring contemporaneously in both oxygen saturation profiles were presumed genuine, whereas a desaturation occurring in only one of the profiles was classified as artefactual. Our method had a sensitivity of 96%, a positive predictive power of 98% and a specificity of 60%. We modified the method to increase specificity and re-evaluated our data. We found that a useful increase in specificity was associated with a considerable decline in sensitivity.