To test the short-term physiologic effects (indexes of respiratory effort, ventilation, and gas exchange), leaks, patient-ventilator asynchrony, and comfort of four noninvasive ventilation (NIV) facial, oronasal, or oral interfaces with major differences in internal volume. Prospective, short-term, crossover randomized physiologic study. Medical intensive care unit in a university hospital. Fourteen consecutive patients receiving NIV for either hypoxemic (n = 7) or hypercapnic (n = 7) acute respiratory failure. Four interfaces, tested randomly over consecutive sequences, had very high (977 mL), high (163 mL), moderate (84 mL), or virtually no internal volume (mouthpiece). The pressure level was increased in two patients with the larger mask, and was decreased in all patients when using the mouthpiece. Despite differences in internal volume, no apparent dead space effect was observed on minute ventilation, work of breathing, or arterial CO2 levels. Compared with baseline, NIV was uniformly successful in reducing indexes of respiratory effort: the pressure-time product of the respiratory muscles decreased from a median (25th-75th interquartile range) of 179 (158-285) cm H2O.sec.min to values between 91 and 111 during NIV, with no differences between masks (p = 0.84). Few differences were observed in terms of arterial blood gases and breathing pattern. Leaks and patient-ventilator asynchronies were greater with the mouthpiece, and comfort with this interface was deemed poor for most patients. The internal volume of the masks had no apparent short-term dead space effect on gas exchange, minute ventilation, or patient's effort, suggesting that, with the exception of mouthpiece, interfaces may be interchangeable in clinical practice provided adjustment of the ventilatory device parameters are performed.