A small landmark trial in 1952 showed that excess oxygen use might well be causing a major epidemic of retinal blindness in preterm babies. That a single study of just 65 babies was enough to throw doubt on a long-standing treatment strategy revealed just how powerful a tool the randomized controlled trial could be. Confirmatory evidence from a cooperative trial involving a further 212 babies banished all residual doubt just 4 years later, and it remains a major reproach that we have still not learnt, after 50 years, how to optimize the delivery of oxygen to the preterm baby with further help from this powerful tool. Two well-conducted trials have recently shown that avoiding subclinical hypoxaemia (fractional SaO(2)<92%) in babies more than a month old does nothing to improve later growth or development. It is now time the same question was asked of babies less than a month old, because we might reduce their need for ventilatory support. This is particularly important in babies of less than 28 weeks' gestation, who remain, currently, at serious risk of chronic lung disease and permanent retinal scarring.