We tested the hypothesis that hyperoxemia defined as arterial PO2 above 12 kPa can be detected by pulse oximetry using 95% oxygen saturation as the upper limit. Thirty artificially ventilated neonates with an indwelling arterial catheter were studied registrating transcutaneous oxygen saturation (Ohmeda Biox 3700 Pulse Oximeter) and transcutaneous PO2 continuously during a 4-hour period and measuring arterial oxygen saturation and PO2 intermittently. 46 episodes of arterial hyperoxemia were observed. Pulse oximetry had a sensitivity of 30%, detecting 14 of these 46 hyperoxemic episodes, and a specificity of 93%. The accuracy for separating hyperoxemia from normoxemia by pulse oximetry could be improved by shifting the cut-off point from 95% to 92%. With this optimal cut-off point sensitivity was 70% and specificity 62%. We conclude that pulse oximetry is not reliable for detection of hyperoxemia.