Forced-oscillation total respiratory resistance (Rrs) has been shown to underestimate spontaneous breathing lung resistance (RL,sb) in patients with airway obstruction, probably owing to upper airway shunting. The present study reinvestigates that relationship in seven severely obstructed chronic obstructive pulmonary disease patients using a technique that minimizes that artefact. Rrs at 8 and 16 Hz was computed for each successive forced oscillation cycle. Inspiratory and expiratory RL,sb were obtained by analysing transpulmonary pressure (Ptp) with a four-coefficient model, and compared to Rrs over the same periods. "Instantaneous" values of RL,sb were also obtained by computing the dynamic component of Ptp, and compared to simultaneous values of Rrs. In both respiratory phases, good agreement between Rrs and RL,sb was observed up to RL,sb values of approximately 15 hPa x s(-1) x L(-1) at 8 Hz and 10 hPa x s(-1) x L(-1) at 16 Hz. Instantaneous Rrs and RL,sb varied systematically during the respiratory cycle, exhibiting various amounts of flow- or volume-dependence in the seven patients; the amplitudes of their variations were significantly correlated, but Rrs was much more flow-dependent than RL,sb in three patients. Also, Rrs exceeded RL,sb at end-expiration in three instances, which could be related to expiratory flow limitation. In conclusion, total respiratory resistance is reliable up to much higher levels of airway obstruction than previously thought, provided upper airway shunting is avoided.