The mechanisms responsible for hypoventilation during rapid-eye-movement (REM) sleep in patients with diaphragmatic weakness are not clear. Therefore, we studied the effects of unilateral (UDP) and bilateral (BDP) diaphragmatic paralysis on ventilation during wakefulness (W), slow-wave sleep (SWS), and REM sleep. Studies were performed in 3 trained dogs in which small silicone cuffs had been implanted surgically around the phrenic nerves. Reversible diaphragmatic paralysis was induced during wakefulness or sleep by bathing the phrenic nerves in local anesthetic injected through a catheter attached to the phrenic cuffs. The UDP reduced abdominal expansion and increased rib cage expansion, but had no effect on minute volume of ventilation (VI) or end-tidal PCO2 (PACO2). The BDP produced marked abdominal paradox, but did not reduce VI during W or SWS and had no effect on tidal volume or respiratory frequency. In contrast, during REM sleep, VI was decreased by an average of 21% mainly because of a fall in tidal volume, and PACO2 increased by 2.4 mm Hg compared with that during the intact state. Furthermore, addition of extra dead space to the breathing circuit, which was well tolerated in intact dogs, led to a progressive increase in PACO2 during BDP and to arousal from sleep. The findings indicate that acute BDP does not impair ventilation during quiet W or SWS, but reduces ventilation during REM sleep, supporting the concept that during REM sleep maintenance of ventilation is critically dependent on diaphragmatic compensation for intercostal and accessory muscle inhibition.