We bench-tested a commercially available instrument for measuring respiratory impedance (Siregnost FD5: Siemens) and found that resistance (R) and phase changes were accurately recorded in models. In a single human subject, total respiratory resistance (R(l)) was closely comparable to resistance measured by the Mead-Whittenberger technique. The derived continuous variable (R(os)) was similar to R at less than 4 cmH2O . litres-1 . sec, but underestimated R at higher values. Ros was highly correlated with airways resistance by body plethysmography (R(aw)), but with a low slope and high intercept (R(os) = 1.38 + 0.59 R(aw): r = 0.89). Because of turbulence, both in model larynxes and in normal subjects, R(os) tends to rise with increases in flow in either direction. R(os) also tends to fall as lung volume rises, and vice versa, reflecting cyclic changes in airway calibre. We devised indices of expiratory narrowing of airways from the maximum flow-volume loop, and the plethysmographic alveolar pressure-flow loop, and compared them with the slope of the relation between R(os) and lung volume during tidal breathing, in nine normal subjects and 16 patients with airflow obstruction. Twelve of the 16 patients, all with abnormal flow-volume loops, had high R(os)-volume slopes, demonstrating excessive expiratory narrowing even during tidal breathing. We found no patients with normal inspiratory R(os) together with an abnormal R(os)-volume slope. Thus unstressed inspiratory calibre was never dissociated from airways narrowing on expiration.