When studying respiratory impedance by forced oscillations, part of the flow measured at the mouth is lost in upper airway wall motion and does not enter the trachea. The corresponding error was studied in 10 normal subjects and 8 patients with chronic obstructive pulmonary disease (COPD) by measuring respiratory impedance with the cheeks unsupported, with the cheeks supported, and when upper airway wall motion was simultaneously measured with a head plethysmograph, and corrected for. In normal subjects, wall motion had little influence on respiratory resistance but, whether the cheeks were supported or not, increased the resonant frequency (p less than 0.05) and respiratory compliance (p less than 0.001) and decreased respiratory inertance (p less than 0.001). In patients with COPD, average resistance from 4 to 30 Hz was significantly lower when the cheeks were not supported (3.32 +/- 0.57 cm/H2O X L-1 X s; m +/- SD) than when they were (4.59 +/- 0.73, p less than 0.01) and when the data were corrected (5.41 +/- 1.14, p less than 0.001). Moreover, resistance increased with increasing frequency when wall motion was corrected for and decreased when it was not. Upper airway wall motion also tended to increase resonant frequency and decrease inertance in patients. The data show that supporting the cheeks does not prevent large errors on respiratory impedance and derived parameters, especially in obstructive patients; accurate measurements require that airway wall motion be evaluated and corrected for.