Background: Discrepancies between pulse oximetry saturation (SpO2) and arterial saturation (SaO2) at low blood oxygenation values have been previously reported with significant variations among instruments and studies. Whether pulse oximeters that attenuate motion artifact are less prone to such discrepancies is not well known. Objective: To prospectively assess the agreement of the Masimo V4 pulse oximeter within the critical 85–95% SpO2 target range. Patients and Methods: For all consecutive babies with gestational age <33 weeks, postnatal age <7 days, and an umbilical arterial line, SpO2 was measured continuously and SaO2 analyzed on an as-needed basis. Bland-Altman techniques provided estimates of the difference (D = SaO2 – SpO2), standard deviation (SD), and 95% limits of agreement (D ± 2*SD). Results: There were 1,032 measurements (114 babies) with SpO2 between 85 and 95%. The 95% limits of agreement were –2.0 ± 5.8, –2.4 ± 9.2, and –1.9 ± 5.3 in the SpO2 categories 85–95, 85–89, and 91–95%, respectively. For the SpO2 categories 85–89% and 91–95%, only 52% (53/101) and 59% (523/886) of SpO2 values, respectively, corresponded to the analogous SaO2 categories. In the 85–89% SpO2 category, SaO2 was lower than 85% in 39 of the 101 (39%) measurements. Conclusion: SaO2 was lower on average than SpO2 with an increased bias at lower saturation. The –2.4 ± 9.2 95% limits of agreement for SaO2 – SpO2 in the 85–89% SpO2 category suggest that SpO2 and SaO2 are not interchangeable and intermittent SaO2 assessments are warranted when the targeted SpO2 is within this range.