1. Genioglossus (GG) activation in response to upper airway negative pressure may be an important mechanism in the maintenance of airway patency. This reflex occurs during wakefulness but is diminished during stable non-rapid eye movement (NREM) sleep. Since obstructive events occur more commonly at wake-sleep transitions and during rapid eye movement (REM) sleep than during stable NREM sleep, we assessed the GG reflex during these two vulnerable states. 2. Seventeen healthy adults were studied throughout one evening and overnight. Electroencephalograms (EEGs), electro-oculograms (EOGs), submental electromyogram (EMG), GG EMG (intramuscular electrodes), and choanal plus epiglottic pressures were recorded. The GG reflex response to pulses of -8 cmH2O choanal pressure applied via nose mask during early inspiration was quantified repeatedly during relaxed wakefulness, within five breaths of wake-sleep transition (EEG alpha-theta transition) and during REM sleep. Only trials without EEG arousal were analysed, resulting in data from 14 subjects during sleep onset and 10 subjects during REM sleep (overall, 174-491 trials per state). 3. During wakefulness there was brisk GG reflex activation in response to negative pressure (amplitude: +78.5 +/- 28.3 % baseline (mean +/- s.e.m.); latency to maximal response: 177 +/- 16 ms). 4. At sleep onset, although there was marked variability among individuals, there was no significant reduction in the magnitude of the GG reflex for the group as a whole (amplitude: +33.2 +/- 8.2 % baseline; latency: 159 +/- 15 ms). 5. In contrast, during REM sleep there was a reduction of GG reflex (amplitude: -12.6 +/- 8.3 % baseline (P = 0.017 vs. awake); latency: 160 +/- 10 ms (n.s. vs. awake)) and greater airway collapsibility during the applied pressures (P = 0.043 vs. awake). 6. We conclude that there was no systematic reduction in the GG reflex to negative pressure at sleep onset. Nonetheless, it remains possible that sleep-deprived normal subjects and patients with sleep apnoea could react differently. 7. The apparent inhibition of the GG reflex during REM sleep may help explain why the upper airway is vulnerable to collapse during this state.