Listeners show remarkable abilities to distinguish music and speech when asked but the essence of such categories is arguable. Here, using recordings of dùndún drumming (a West-African drum also used as a speech surrogate), we first replicate standard speech-music categorization results (N=108, sample size based on a prior study), then depart from the typical experimental procedure by asking participants (N=180) to freely categorize and label these recordings. Hierarchical clustering of participants’ stimulus groupings shows that the speech/music distinction emerges, but is not primary. Analysis of participants' labels in the free-response task converges with acoustic predictors of the categories, supporting the effect of priming in music/speech discrimination, and thereby providing a new perspective on the categorisation of such common auditory signals.