When deciding an appropriate upper limit for pulse oxygen saturation (SpO2) in preterm infants the usefulness of current data is limited by the fact that previous studies have examined a population of more mature infants and children or have applied various exclusion criteria which produce results unrepresentative of clinical practice. We tested the hypothesis of previous workers that maintaining the SpO2 below 98% would ensure an arterial oxygen tension (PaO2) less than 12kPa. A total of 477 simultaneous measurements of PaO2 and SpO2 were made using Ohmeda Biox oximeters on 43 infants who were less than 33 weeks gestation and receiving supplementary oxygen. Of 435 measurements performed when the SpO2 was 97% or less, 26 (6%) had a PaO2 greater than 12kPa. Further examination of the data showed that of 108 estimations performed when the SpO2 was less than 94%, none had a PaO2 greater than 12kPa. CONCLUSIONS When using Ohmeda Biox pulse oximeters an upper limit of 97% for SaO2 is not effective in preventing hyperoxaemia; however, a limit of 93% is likely to maintain the PaO2 below 12kPa.