The mechanisms responsible for the various sleep apnea patterns have not been elucidated. In 5 hypersomnolent patients, we analyzed 613 events that would have been labeled mixed and central apneas had usual recording sensitivity been employed. Recording air flow at high sensitivity revealed that in most of these events expiratory flow was present throughout most of the prolonged period during which there was no inspiratory effort. In 378 events, resumption of inspiratory effort was unaccompanied by air flow (occluded events), whereas in 144 events resumption of inspiratory effort was accompanied by inspiratory air flow (not-occluded events). In another 91 events, expiratory air flow ceased during the interval between inspiratory efforts and then recommenced unaccompanied by expiratory effort suggesting airway closure and reopening. Because expiratory air flow is present during a substantial portion of the prolonged interval between inspiratory efforts seen in our sleeping patients, we conclude that "apneas" should be defined relative to absence of inspiratory air flow. In addition, it appears that the initial portion of "mixed," and at least some "central apneas," are associated with increased expiratory upper airway resistance.