The effects of cold-induced vasoconstriction and venous occlusion on the detection of induced hypoxaemia by four pulse oximeters were examined in 10 volunteers. In three further subjects vasoconstriction was maintained until at least one instrument failed to detect the induced hypoxaemia. Time taken to detect hypoxaemia was increased for all instruments to between two and three times the instrument's own control value for both vasoconstriction and venous engorgement (P less than 0.01). There was highly significant variation in detected minimum saturation between the instruments (P less than 0.001). One instrument failed to detect the full extent of desaturation under the experimental conditions and was more likely to fail completely to detect desaturation than the other test instruments when influenced by vasoconstriction (P less than 0.05). Significant impairment in the performance of all the instruments tested occurred in the presence of normal pulse signals. The duration of detected reductions in oxygen saturation was not significantly affected.