To determine whether sleep quality is influenced by the mode of mechanical ventilation, we performed polysomnography on 11 critically ill patients. Because pressure support predisposes to central apneas in healthy subjects, we examined whether the presence of a backup rate on assist-control ventilation would decrease apnea-related arousals and improve sleep quality. Sleep fragmentation, measured as the number of arousals and awakenings, was greater during pressure support than during assist-control ventilation: 79 +/- 7 versus 54 +/- 7 events per hour (p = 0.02). Central apneas occurred during pressure support in six patients; heart failure was more common in these six patients than in the five patients without apneas: 83 versus 20% (p = 0.04). Among patients with central apneas, adding dead space decreased sleep fragmentation: 44 +/- 6 versus 83 +/- 12 arousals and awakenings per hour (p = 0.02). Changes in sleep-wakefulness state caused greater changes in breath components and end-tidal CO2 during pressure support than during assist-control ventilation. In conclusion, inspiratory assistance from pressure support causes hypocapnia, which combined with the lack of a backup rate and wakefulness drive can lead to central apneas and sleep fragmentation, especially in patients with heart failure.