Correspondence: D. E. Kroodsma, Department of Biology, University of Massachusetts, Amherst, MA 01003, U.S.A. (e-mail: kroodsma@bio.umass.edu). About 10 years ago, several papers in Animal Behaviour addressed the quality of experimental designs in ‘playback’ experiments (Kroodsma 1989a, b, 1990a, 1992; Searcy 1989; Weary & Mountjoy 1992), and this debate culminated in a consensus report by McGregor et al. (1992). The key issue was ‘pseudoreplication’, defined by Hurlbert (1984, page 187) as ‘the use of inferential statistics to test for treatment effects with data from experiments where either treatments are not replicated (though samples may be) or replicates are not statistically independent’. McGregor et al. (1992, page 2) offered their own simplified definition, ‘the use of an n (sample size) in a statistical test that is not appropriate to the hypothesis being tested’. McGregor et al. agreed that pseudoreplication was a serious issue, and that designing and implementing good experimental designs was a worthy and attainable goal. What effect did the debate and subsequent consensus report have on the quality of experimental designs used in animal behaviour? To answer that question, we surveyed the designs used in 50 papers published during the last several years. The papers were chosen by searching electronic databases for 25 papers that cited a key paper on pseudoreplication and for 25 other ‘playback’ papers that did not explicitly cite or address pseudoreplication issues. We reasoned that these two samples of papers should provide an index to the experimental designs and logic currently being used by investigators in animal behaviour. (Note: we chose not to review playback experiments that used synthetic stimuli. Although use of synthetic stimuli may solve some problems (e.g. see McGregor et al. 1992), we encountered a number of papers in which we felt that interpretations based on large sets of synthesized playback variants exceeded the permitted inferential space.)