In mechanically ventilated patients, pressure and flow tracings can be used to assess respiratory pump muscle activity or rest. When the ventilation is delivered through an endotracheal tube, the respiratory system can be considered a one-compartment model, and activation of the respiratory muscles results in deformations and variability of the pressure tracings. This is also true when mechanical ventilation is delivered nasally. With intermittent positive-pressure ventilation delivered through a nasal mask (nIPPV), we have recently shown that the glottis can interfere with ventilation even in the absence of diaphragmatic surface electromyographic (EMG) activity. On the basis of our observations, we suggested that when mechanical ventilation is delivered through a nasal means of access, the respiratory system cannot be considered a one-compartment model. To confirm this hypothesis, we submitted one healthy subject to nIPPV while his glottis was continuously monitored through a fiberoptic bronchoscope and his diaphragmatic activity was monitored with a bipolar esophageal electrode. During wakefulness or sleep, we observed irregularities in the nasal mask pressure waveform, in nasal mask peak pressure, and in actual VT despite the absence of respiratory pump muscle activity. These irregularities were related to significant variations in the glottic width, rather than to the reappearance of transient phasic inspiratory muscle activity. We conclude that during nIPPV, deformations in the mask pressure waveform can be induced by variations in the glottic aperture without activation of the diaphragm. Thus, when mechanical ventilation does not bypass the glottis, the respiratory system does not behave like a one-compartment model, and EMG remains the only reliable technique for assessing diaphragmatic muscle activity.