We developed a technique that assists in objective identification of pulse and sine components in the Drosophila melanogaster courtship song. The song was recorded digitally and subjected to wavelet analysis, which decomposed the signal into a series of bands of decreasing frequency in which acoustic power appeared as a function of time. This time-frequency analysis expresses characteristic 'fingerprints' of the pulses, which appear in all frequency bands, and characteristic 'fingerprints' of the sine song, which appear only in the band containing its frequency. Means of the interpulse intervals (IPIs) for each second of 10-min recordings of song in which pulse singing occurred constituted an irregularly sampled time series, intractable to common spectral analysis techniques. Therefore, we took the discrete Fourier transform of this series, which retained all the spectral information in the Fourier coefficients, and used the inverse Fourier transform of these coefficients to yield a new series that was regularly spaced, with an estimate of IPI for each second in the interval. We observed an IPI period of 67.9 s in wild type, 31.1 s in per0, 45.9 s in pers, and 72.0 s in perL. Copyright 1998 The Association for the Study of Animal Behaviour.