Pressure-cycled infant ventilators generate pressure pulses with short rise-times. Gas flow is approximately equal to the derivative of pressure when lung compliance is low, and hence contains high-frequency components. The authors defined bandwidth as that frequency fm, below which 99.9% of the energy of the signal resided. Simulation of the measurement process using measurement systems with frequency response similar to sixth-order Bessel filters and a lung model comprising series resistance, inertance and compliance showed that measurement systems with frequency response flat +or-lO% to f m yield time domain errors less than 3% of the peak value. The authors digitized pressure and flow signals from 10-20 ventilator (Healthdyne 105) breaths in 33 stable mechanically ventilated infants. The transducers' frequency responses had been measured between 1 Hz and 100 Hz and phase matched at 10 Hz. The authors calculated total respiratory resistance R and elastance E using multiple linear regression, and ensemble-average power spectral density using the FFT with a rectangular time window and padding to 2048 points. Power spectra were compensated for non-unity transducer and anti-alias filter responses up to 60 Hz. Satisfactory results were obtained from more than eight breaths in 18 infants. It is concluded that the bandwidth of the flow waveform increases with decreasing compliance and mechanical time constant.