Recorded values (SpO2) of oxygen saturation from six noninvasive pulse oximeters were compared to each other and to the value from a cooximeter. Simultaneous measurements were obtained from each instrument in eight healthy nonsmoking volunteers rendered hypoxic greater than 70% SpO2. Functional arterial oxygen saturation (SaO2), fractional SaO2 (%HbO2), and percent fraction of carboxyhemoglobin (HbCO) and methemoglobin (MetHb) were determined by a cooximeter. The difference between SaO2 and %HbO2 was 2.3 +/- 0.3 (SD) % (p less than .001) with 1.4 +/- 0.1% HbCO and 0.9 +/- 0.1% MetHb. The SpO2 values in two instruments were close to %HbO2, while the other instruments correlated approximately with SaO2. The greatest SpO2 difference among the instruments was 2.7 +/- 1.9% (p less than .001). These results may indicate that, under normal dyshemoglobin levels, some pulse oximeters are calibrated to estimate SaO2 and others to estimate %HbO2. Since the pulse oximeter using two wavelengths cannot measure accurately %HbO2 or SaO2 in the presence of dyshemoglobin, SpO2 values would be independent from %HbO2 and SaO2. A standard calibration method for pulse oximeters should be established by the manufacturers.