1. The influence of the upper airway on the ventilatory and arousal responses to hypercapnia in wakefulness and sleep was investigated using a chronic animal model. 2. Experiments were performed in five unrestrained dogs trained to sleep naturally in the laboratory. The animal rebreathed through a chronic tracheostoma (thus excluding the upper airway from the breathing circuit), or through the snout (intact upper airway). Resistance to breathing and volume of dead space during quiet tracheal breathing were matched to those in quiet nasal breathing during wakefulness and sleep. CO2 rebreathing tests were performed during wakefulness, rapid eye movement (REM) and non-REM (NREM) sleep, during nasal and tracheal breathing. 3. The ventilatory response to hypercapnia was significantly lower in nasal breathing compared with tracheal breathing, in all behavioural states. This was due to a smaller tidal volume and lower breathing frequency. 4. The ventilatory response to CO2 was lowest during REM sleep, irrespective of route used for breathing. 5. Alveolar partial pressure of CO2 (PA,CO2) level at arousal was identical in NREM nasal and tracheal rebreathing tests. Differences in PA,CO2 levels at arousal between NREM and REM sleep were not significant in nasal tests and only marginally different during tracheal breathing. 6. We conclude that nasal breathing influences the hypercapnic ventilatory response in wakefulness and sleep, and that the presence of CO2 in the upper airway does not affect arousal in NREM and REM sleep.