Pulse oximetry is a major improvement in the assessment of oxygenation. The device uses plethysmography and light absorbance measurements at two wavelengths to estimate oxygen saturation. It is inaccurate, however, when more than two types of hemoglobin are present. This article describes two infants with methemoglobinemia in whom pulse oximetry overestimated oxygen saturation. We discuss the mechanism of this systematic error and emphasize that pulse oximetry should not be used to estimate true oxygen saturation in the presence of methemoglobin. However, a disparity between oxygen saturation estimates by pulse oximetry and by calculations based on the arterial partial pressure of oxygen and the oxygen-hemoglobin dissociation curve can provide an important clue to the presence of such abnormal types of hemoglobins. Therapy should be based on direct measurements of oxyhemoglobin by cooximetry and not on measurements of oxygen saturation by pulse oximetry or on saturations calculated from the Pao2 and the oxyhemoglobin dissociation curve.