Pulse oximetry is a useful technique for noninvasive oxygen monitoring in sick infants. We simultaneously measured oxygen saturation by pulse oximetry and on arterial blood samples by co-oximetry as well as PaO2 and the relative content of fetal (F) and adult hemoglobin in order to evaluate the reliability of pulse oximetry. Comparisons were made in triplicate in ten infants with acute cardiorespiratory disease less than 7 days of age and in 11 infants with chronic lung disease greater than 28 days of age. Oxygen saturation pulse oximetry and arterial saturation were well correlated over a wide range of saturation values. In infants with chronic lung disease, PO2 derived from pulse oximetry was within 10 torr of measured PaO2 in 73% of comparisons. In contrast, calculated PaO2 was within 10 torr of measured PaO2 in only 50% of comparisons in patients with acute disorders. The chronic infants all had less than 10% hemoglobin F, but in the acute infants, hemoglobin F ranged from 26% to 83%. Nonetheless, correction of oxygen dissociation curves for type of hemoglobin in these acute infants failed to improve the correlation between calculated and measured PaO2. We conclude that pulse oximetry saturations and their derived PaO2 values correlated well with measured arterial saturation and PaO2 obtained from arterial blood samples in neonates with chronic lung disease and prolonged oxygen dependence. In infants with acute cardiorespiratory problems, pulse oximetry unreliably reflects PaO2, but may be useful in detecting clinical deterioration.