Eight anesthetized tracheostomized cats were placed in an 8.2-liter airtight chamber with the trachea connected to the exterior. Thirty-two combinations of high-frequency oscillations (HFO) (0.5-30 Hz; 25-100 ml) were delivered for 10 min each in random order into the chamber. Arterial blood gas tensions during oscillation were compared with control measurements made after 10 min of spontaneous breathing without oscillation when the mean arterial PCO2 (PaCO2) was 30.1 Torr. Ventilation due to spontaneous breathing (Vs) and oscillation (Vo) were derived from the chamber pressure trace and a pneumotachograph, respectively. As the oscillation frequency increased, oscillated tidal volume (Vo) decreased from a mean of 39 (0.5 Hz) to 3.3 ml (30 Hz) when 100 ml was delivered to the chamber. From 6-25 Hz, apnea occurred with Vo less than estimated respiratory dead space (VD); the minimum effective Vo/VD ratio was 0.37 +/- 0.05. Although Vo was maximal at 10 Hz at each oscillation volume, the lowest PaCO2 occurred at 2-6 Hz, and arterial PO2 rose as expected during hypocapnia. Above 10 Hz, PaCO2 was determined by Vo and was independent of frequency, whereas at lower frequencies, PaCO2 was related to Vo; below 6 Hz, PaCO2 varied inversely with the calculated alveolar ventilation. As oscillations became more effective, both PaCO2 and Vs fell progressively and were highly correlated; apnea occurred when PaCO2 was reduced by a mean of 4.5 Torr. Mean chamber pressure remained near zero up to 15 Hz, indicating functional residual capacity did not change. We conclude that externally applied HFO can readily maintain gas exchange in vivo, with Vo less than VD at frequencies over 2 Hz.