To test the hypothesis that variable hyperoxia potentiates preretinal neovascularization in newborn rats, and to establish a more reliable animal model of ROP in which therapies designed to inhibit abnormal angiogenesis can be tested. Immediately after birth, litters of Sprague Dawley albino rats and mothers were placed in an incubator containing 40% oxygen. After 12 hours, the oxygen was increased to 80% with a transition time of less than 1 min. For the ensuing 7, 10, or 14 days, the oxygen was altered between 40% and 80% every 12 hr in a stepwise fashion. Other litters were kept in constant 80% oxygen or in room air for the same three time periods. After exposure, rats were either killed or placed in room air for an additional 2, 4, or 7 days before being killed. When rats were killed immediately after oxygen exposure, the resulting vessel loss in rats exposed to 40%/80% oxygen was identical to that of animals exposed to 80%(vessels constituted 12.2+/-2.2% of total retinal area in cyclic oxygen vs 12.0+/-1.2% in constant oxygen). However, preretinal neovascularization subsequently occurred in 66%(63/96) of all rats exposed to cyclic oxygen followed by a room air period but in no rats (0/50) exposed to constant oxygen followed by room air. Preretinal vascular proliferation consisted of glomerular tufts of endothelial cells, or mature, lumenized vessels containing red blood cells. Consistency of oxygen therapy is more important than overall oxygen level in inducing retinopathy. Consideration should be given to tighter control of intended oxygen therapy in premature infants, regardless of the target saturation level.