Computation of impedances from forced oscillation data during breathing can yield results which reflect not only changes in respiratory mechanics, but artifacts related to the signal analysis. A method has been developed, employing sinusoidal forcing, to determine intra-breath variations of respiratory impedance. The measured pressure and flow waveforms are each the sum of a slowly varying constituent associated with breathing, and a high-frequency oscillatory constituent, whose amplitude and phase vary with time. The signal constituents were separated with a moving-average filter. Characteristic amplitudes and phases of the oscillatory constituents over a short time interval (window) were determined by correlating the constituents with sine waves of the same frequency. Continuous estimates of the time-varying impedance were obtained by moving the window over the data. Two procedures were developed to examine the accuracy of impedances computed with this technique: (1) the analysis of fixed-amplitude sinusoids superimposed on a breathing pattern; and (2) the analysis of a known, time-varying impedance. The effects of forcing frequency, window size, breathing frequency, and the position within the respiratory cycle on the computed impedances were examined. For quiet breathing, the technique can yield impedances which are accurate to within 5% in magnitude, and 5 degrees in phase angle at all instants within the breath. An efficient algorithm was developed for implementing the technique on a computer.