Arousal and ventilatory responses to rebreathing from a small bag, initially .apprx. 7% CO2 in 40% O2, were measured via a nose mask in 13 normal human adults. With deepening non-rapid-eye-movement sleep (NREM), males aroused at increasing alveolar PCO2 (mean .+-. SE: stage II 58.6 .+-. 1.7, stage III 61.2 .+-. 1.0, stage IV 63.8 .+-. 0.8 Torr), whereas in rapid-eye-movement sleep (REM), arousal alveolar PCO2 was 57.7 .+-. 0.7 Torr, i.e., much lower than in stage III and IV NREM. Females showed no significant change in arousal alveolar PCO2, (II 57.6 .+-. 0.9, III 57.3 .+-. 1.3, IV 59.4 .+-. 0.9, REM 56.3 .+-. 1.0 Torr). Male ventilatory response was 2.5 .+-. 0.1 (SE) 1 .cntdot. min-1 .cntdot. Torr-1 and fell by 49% in NREM (1.29 .+-. 0.13) and by 69% in REM (0.78 .+-. 0.18). Female response was little affected by state, being similar to male NREM response (wake 1.39 .+-. 0.14, NREM 1.40 .+-. 0.13, REM 1.11 .+-. 0.26 l .cntdot. min-1 .cntdot. Torr-1). In NREM tests, there was no change in sleep state until arousal, whereas in REM, subjects awoke abruptly with the onset of rebreathing (11 cases), showed a transient arousal with onset but continued in REM until final arousal (21 cases) or changed to NREM at onset (2 cases). These arousal results contrast sharply with findings in tracheostomized dogs and in obstructive sleep apnea syndrome, where asphyxic arousal is later in REM than in NREM, suggesting that the events at test onset in REM may be related to upper airway sensitivity to CO2 specific to REM.