A new technique was validated in vivo in reflectance pulse oximetry for measuring low oxygen saturations. Two pairs of light emitter/detector diodes allow for estimation of light attenuation (LA) in tissue, which is assumed to be responsible for the inaccuracy of pulse oximetry at less than 70 % arterial oxygen saturation. For validation, 17 newborn piglets were desaturated stepwise from 21 % to 1.25 % inspiratory oxygen concentration during general anesthesia, and arterial oxygen saturation was measured with the reflectance pulse oximeter adjusted for LA in tissue, with a standard transmission pulse oximeter and a hemoximeter. LA in tissue could be quantified and was different between snout and foreleg (probability level (p) < 0.05). At arterial oxygen saturations above 70 %, the bias between the methods was at 0 %-1 % and the variability 4 %-5 %. From 2 % to 100 % arterial oxygen saturation, the reflectance pulse oximeter estimated oxyhemoglobin saturation more accurately than a conventional transmission pulse oximeter (p < 0.05). At low oxygen saturations below 70 %, the bias and variability of the reflectance pulse oximeter calibration were closer to the hemoximeter measurements than the transmission pulse oximeter (p < 0.05). The variability of the reflectance pulse oximeter was slightly lower than the traditional oximeter by taking into account the LA in tissue (9 % versus 11 % -15 %, ns), and thus, the quality of the individual calibration lines improved (correlation coefficient, p < 0.05).