Abstract
Crossley (1988, Anim. Behav., 36, 1098–1109) and Ewing (1988, Anim. Behav., 36, 1091–1097), have criticized the use of non-linear regression in testing for song cycles in Drosophila courtship songs. In the present study a large body of data was reanalysed with appropriate time series analyses. The results confirm the original conclusions based on regression. Evidence suggests that the apparent discrepancies between our results may arise from: (1) the size of mating cell used by Ewing (1988) to record songs; (2) the way the song traces are interpreted; (3) the way mean inter-pulse intervals are assessed, especially when there are missing data points; and (4) the method used to assign statistical significance.