Large-scale online iterated singing experiments reveal oral transmission effects in the evolution of human song

Abstract

**Author(s):** Anglada-Tort, M; Harrison, P M C; Jacoby, N Introduction: In most human cultures, music and speech are sustained through oral transmission. How do cognitive and motor constraints in this process shape the evolution of music? Music corpus studies have been particularly useful to characterize common and diverse features observed in human songs across cultures (Mehr et al., 2019; Savage, 2015), yet these studies fail to quantify how and why these features emerge. Alternatively, iterated learning experiments provide a powerful quantitative tool for studying how transmission biases shape the evolution of language and music. However, such experiments have traditionally been conducted in nonautomated laboratory settings, making data collection costly while limiting the ability to explore the full space of evolutionary possibilities. Methods: Here we present the first large-scale cultural evolution study on human song, quantifying the evolution of thousands of melodies transmitted orally across participants. Our method is applicable to individuals from any cultural background and works efficiently on computers available to most participants online. We perform many variations of this paradigm, including varying the total number of notes per melody (3-12 notes), adding memory load and motor constraints in the singing task, or naturalistic singing. Results: We show that cognitive and motor constraints in oral transmission play a profound role in the emergence of melodic structure. Specifically, initially random tones develop into more structured systems that increasingly reuse and combine fewer elements, making melodies easier to learn and transmit over time. Importantly, melodic features that emerged artificially from our experiments are largely consistent with melodic universals observed cross-culturally (e.g., small intervals, discrete scales, and tonal structures). Parallel to this, we find that evolving melodies tend to align with melodic pleasantness (estimated by a separate rating experiment with Western listeners) and musical exposure (estimated by a large corpus of popular Western melodies). Discussion: Taken together, our results provide the first quantitative characterization of the rich collection of biases that oral transmission imposes on music evolution, giving us a new understanding of the emergence of musical universals and the resulting vast structural diversity of human song.