The ventilatory response of the newborn to CO2 was studied using a rebreathing method that minimized changes in arterial PO2 during the test. The aim was to study the variability of the ventilatory response to CO2 and take this into account to assess the relative magnitude of the response to CO2 during rapid-eye-movement (REM) sleep and quiet sleep (QS). Five full-term babies aged 4-6 days were given 5% CO2 in air to rebreathe for 1.5-3 min. O2 was added to the rebreathing circuit to maintain arterial O2 saturation and transcutaneous PO2 (Ptco2) at prerebreathing levels. Tests were repeated four to five times in REM sleep and QS. Mean Ptco2 levels varied between individuals but were similar during REM sleep and QS tests for each subject. The mean coefficient of variability of the ventilatory response was 35% (range 15-77%) during QS and 120% (range 32-220%) during REM sleep. PtcO2 fluctuations during tests [6.0 +/- 3.0 (SD) Torr, range 1-13 Torr] were not correlated with ventilatory response. Overall the ventilatory response was significantly lower in REM sleep than in QS (12.2 +/- 3.0 vs. 38.7 +/- 3.0 ml.min-1.Torr-1.kg-1, P less than 0.001; 2-way analysis of variance) due to a small (nonsignificant) fall in the tidal volume response and a significant fall in breathing rate. In 12 REM sleep tests there was no significant ventilatory response; mean inspiratory flow increased significantly during 8 of these 12 tests. We conclude that there is a significant decrease in the ventilatory response of the newborn to CO2 rebreathing during REM sleep compared with QS.