The performance of three commercially available pulse oximeters was assessed in five anesthetized dogs in which increasing levels of methemoglobin were induced. Hemoglobin oxygen saturation in each dog was monitored with three pulse oximeters (Nellcor N-100, Ohmeda 3700, and Novametrix 500) and a mixed venous saturation pulmonary artery catheter (Oximetrix Opticath). Arterial and mixed venous blood specimens were analyzed for PaO2, PaCO2, and pHa using standard electrodes. An IL-282 Co-oximeter was used on the same specimens to determine oxyhemoglobin and methemoglobin as percentages of total hemoglobin. Methemoglobin levels of up to 60% were induced by intratracheal benzocaine. As MetHb gradually increased while the dogs were breathing 100% inspired oxygen, the pulse oximeter saturation (SpO2) overestimated the fractional oxygen saturation (SaO2) by an amount proportional to the concentration of methemoglobin until the latter reached approximately 35%. At this level the SpO2 values reached a plateau of 84-86% and did not decrease further. When, at fixed methemoglobin levels, additional hemoglobin desaturation was induced by reducing inspired oxygen fraction, SpO2 changed by much less than did SaO2 (regression slopes from 0.16 to 0.32). Thus, at high methemoglobin levels SpO2 tends to overestimate SaO2 by larger amounts at low hemoglobin saturations. Plots of SpO2 versus functional saturation (oxyhemoglobin/reduced hemoglobin plus oxyhemoglobin) show an improved but still poor relationship (regression slopes from 0.32 to 0.46). The Oximetrix Opticath pulmonary artery catheter behaves similarly but provides somewhat better agreement with functional saturation than do the pulse oximeters in the presence of methemoglobinemia. Pulse oximetry data (SpO2) should be used with caution in patients with methemoglobinemia.