The aim of this study was to assess whether the severity of the infant's lung disease determined the most appropriate change in mean airway pressure (MAP) level to use on transfer from conventional ventilation to high frequency oscillation (HFO). In addition, we wished to assess whether the oscillatory frequency employed affected gas exchange. Ten premature infants with respiratory distress syndrome (RDS) were studied at a mean postnatal age of 1.5 days. During HFO, the infants were studied at a MAP equivalent of that used during conventional ventilation (baseline MAP), then at 2 and 5 cmH2O above baseline at 10 Hz. At the MAP identified as optimum, that is, the one associated with the best oxygenation, the infants were then studied at 10, 15 and 20 Hz. Each oscillatory setting was maintained for 20 minutes after which time arterial blood gases were measured. Prior to transfer to the oscillator, the peak inspiratory pressure was recorded, the P(A-a)O2 calculated and compliance of the respiratory system (Crs) measured. In nine infants, the optimum baseline MAP was +5 cmH2O. Oxygenation at that level was better than on conventional ventilation (P < 0.05), but there was no significant change in CO2 elimination. The optimum MAP was related to the peak pressure during conventional ventilation (P < 0.01) and inversely related to Crs (P < 0.01). There was no significant relationship with the P(A-a)O2. At the optimum MAP, the only significant effect of frequency was an impairment of oxygenation at 20 Hz.(ABSTRACT TRUNCATED AT 250 WORDS)