Using appropriate experimental designs for intended hypotheses in ‘song’ playbacks, with examples for testing effects of song repertoire sizes

Abstract

In playback experiments, a mismatch between the experimental design and the intended hypothesis often prevents the intended hypothesis from actually being tested. A review of the literature reveals, for example, that, despite numerous studies, available data do not provide a strong test of the hypothesis that larger song repertoires are especially stimulating to songbird listeners. Single exemplars are often used to represent all possible repertoires of a given size, i.e. the treatment is unreplicated. Repeated response measures to one repertoire are often then used in statistical tests as if those measures were the mean responses to different repertoires of a given size. Because studies usually do not control for variation present in successive renditions of a single song type, they test for the overall effects of ‘song variation’, not specifically for the effects of song type repertoires. Additional concerns include controls for observer-expectancy bias and re-use of either individuals or stimuli to generate inflated sample sizes. Several improved experimental designs are proposed and an increased awareness of these critical issues in experimental research in all subfields of animal behaviour is encouraged.