We noticed that arterial oxygen saturation by pulse oximetry (SpO2) was generally lower when determined by the Ohmeda Biox 3700 pulse oximeter than when determined by the Nellcor N-100 pulse oximeter, and we investigated whether this finding was consistent and the reason for the discrepancy. We placed both oximeters simultaneously on 30 infants with indwelling arterial catheters and measured arterial partial pressure of oxygen (PaO2), percentage of fetal hemoglobin, and complete cooximetry, including arterial oxygen saturation (SaO2) with a Radiometer OSM-3 cooximeter, with and without correction for fetal hemoglobin levels, in four samples of blood from each infant during a 12-hour period for a total of 120 samples. The Nellcor SpO2 was consistently higher than the Ohmeda SpO2 by a mean (+/- SD) of 1.61% +/- 2.69% (p < 0.001). The Nellcor SpO2 correlated best with functional SaO2 (oxyhemoglobin (HbO2)/(HbO2 + reduced hemoglobin)) x 100); Ohmeda SpO2 correlated best with fractional SaO2 (HbO2/(HbO2 + reduced hemoglobin + carboxyhemoglobin + methemoglobin)) x 100), reflecting a fundamental difference in the calibration algorithms used in the two instruments. A desired PaO2 of 50 to 100 mm Hg, is maintained when the range of SpO2 is 90% to 96% for Ohmeda SpO2 and 92% to 98% for Nellcor SpO2 in the neonate, giving a positive predictive value in this study of 94% to 95%. We conclude that SpO2 determined by pulse oximeters of different brands is not interchangeable, and this may be of clinical importance in predicting PaO2 on the basis of SpO2.