Nineteen subjects with the obstructive sleep apnea syndrome (10 with daytime arterial CO2 tension 44 mm Hg or higher) were treated with long-term nocturnal continuous positive airway pressure. The ventilatory response to CO2 (Read's method) was measured in triplicate prior to treatment and after 1, 2, 3, 7, and 14 or more nights of therapy. Seven subjects were tested on at least 4 occasions. For each test, slope of the response line and position of the response line (ventilation at a PCO2 of 60 mm Hg) were calculated. The subjects with initial high daytime CO2 showed no change in slope of response with treatment but showed a progressive increase in ventilation at any given degree of PCO2. Ventilation at a PCO2 of 60 mm Hg increased from a mean of 20.0 +/- 1.3 SEM L/min by 8.0 +/- 2.5 SEM L/min after 2 nights of therapy (p less than 0.05, two-way analysis of variance), and by 16.2 +/- 1.9 L/min after 2 wk or more (p less than 0.01). On average, there was no significant change in either slope or position of response in the subjects with initially normal daytime PCO2. We conclude that airway obstruction in sleep (in obstructive sleep apnea syndrome) leads in some subjects to respiratory failure in the daytime, with a left shift in the ventilatory response to CO2, and that this changes is usually reversible during the next several days.