We investigated the interaction between respiration and sleep in ten male outpatients with severe, stable, maximally treated congestive heart failure (CHF). Cheyne-Stokes respiration (CSR), defined as periodic breathing with apnea or hypopnea, was found in all patients with a mean duration of 120 +/- 87 minutes [50.2 +/- 34.4 percent total sleep time (TST)]. The CSR was found predominantly during stage 1 (20.6 +/- 6.7 percent TST) and stage 2 (25.8 +/- 6 percent TST) NREM sleep and occurred rarely during slow wave sleep (SWS) (1.6 +/- 1 percent TST) and REM sleep (1.6 +/- 0.5 percent TST). All apneas and hypopneas were central. Despite normal awake arterial oxygenation (SaO2) (96.1 +/- 1.6 percent), significant, severe hypoxemia was found during sleep in seven patients with SaO2 less than 90 percent for 9 to 59 percent TST (mean +/- SD, 23 +/- 23 percent TST), and this was significantly related to the duration of CSR (r = 0.66, p less than 0.05). The mean minimum SaO2 for sleep stage was lowest during stage 1 (82.1 percent +/- 2.6 percent) and stage 2 (78.9 percent +/- 2.8 percent) NREM sleep, intermediate during REM sleep (84.5 percent +/- 1.8 percent) and highest during SWS (87.6 percent +/- 2.7 percent). Sleep was disrupted to a variable extent in all patients with a short mean TST (287 +/- 106 minutes), a high proportion of stage 1 sleep (26 +/- 19 percent TST), virtual absence of SWS (5 +/- 7 percent TST) which was found in only four patients, and a high number of sleep stage changes (30 +/- 27/hour) and arousals (28 +/- 25/hour). Arousals occurred predominantly during stage 1 (17 +/- 20/hour) and stage 2 (10 +/- 7/hour) NREM sleep and the majority immediately followed the hyperpneic phase of CSR. The amount of CSR (percent TST) was inversely related to the length of TST (r = -0.73, p less than 0.05), and directly related to the number of sleep stage changes (r = 0.79, p less than 0.01) and the number of arousals (r = 0.66, p less than 0.05). We conclude that in severe, stable CHF, CSR occurs predominantly during light sleep, that despite normal awake arterial oxygen saturation, significant hypoxemia may develop during sleep due to CSR, and that sleep is unstable and disrupted due to frequent arousals caused by the hyperpneic phase of CSR. These sequelae of CSR may be important determinants of the clinical status and outcome of patients with severe CHF.