Features underlying speech versus music as categories of auditory experience

Classifying whether an auditory signal is music or speech is important for both humans and computer systems and critical for many applications. Though previous literature suggests that music and speech are easily separable categories, our approaches to studying the distinction have arguably biased findings in that direction. Prior studies often use material from different sound sources and provide response-label categories. Here, we use stimulus material from the dùndún drum–a speech surrogate that can signal either speech-related or musical content, rather unfamiliar to Western listeners. We first replicate standard speech-music categorization results (N=108). Then, we depart from the typical experimental procedure by asking new participants (N=180) to freely sort then label the stimulus material. Hierarchical clustering of participants’ stimulus groupings shows that the speech/music distinction emerges but is not primary. By reverse engineering the acoustic features and semantic labels underlying participants’ auditory categories, we provide an empirical argument for moderating the long-standing, but nevertheless disputable, perceptual salience of music and speech categories.