Pulse oximetry is a technique for estimating arterial oxygen saturation continuously and non-invasively. Reflectance pulse oximetry might become useful for monitoring the fetus during labour but it is much more susceptible to all kinds of physiological variations than the well-established transmission pulse oximetry for neonatal or adult monitoring. This review focuses on the accuracy of reflectance pulse oximetry. Results of human, animal, in vitro and theoretical models indicate that factors such as; blood volume fraction differences, haematocrit, and blood flow differences are major sources for inaccurate pulse oximetry readings in the fetal arterial oxygen saturation range of 10-80%. These factors cannot be overcome by systems using two wavelengths sensors with the 660/890 or 940 nm combination. Reported precision values (S.D. of difference between pulse oximeter and blood sample saturation) range between 2.5 and 12.9% for various 660 nm sensors. Most sensors were tested only once with a limited number of animals. A new 735/890 nm sensor (Nellcor Puritan Bennett) demonstrates a promising accuracy (precision around 5%) in two studies. Various other sensors have also been developed, but are not or scarcely evaluated. Without thorough establishment of the reliability of this technique, clinical fetal oxygen saturation data are still of limited value.