Functional diversity in human song

Abstract

Functional diversity-i.e., the diversity of morphophysiological characteristics of species in a biological community-revolutionized ecology in recent decades, shifting the focus of the field from species to ecosystems. While its ecological applications are known, its adaptability to other disciplines, specifically music, is explored here. We retrieved fourteen characteristics of 12,944 songs by the top 100 artists of the 2010s decade on four streaming platforms. Then, we calculated the three main components of functional diversity-richness, evenness, and divergence-to each artist using probabilistic hypervolumes. Furthermore, we investigated to what extent functional diversity and the traits of an artist, its albums and songs has an effect on their popularity across streaming platforms such as Spotify. High functional richness, where an artist's songs differ greatly sonically, correlated with increased listens of up to 244,300,000. This would lead to estimated profit earnings exceeding $1,000,000 per richness gain. Danceable, highly-energetic, melodic, pop, and, notably, melancholic songs, albums, and artists are more listened to than their counterparts in streaming services. We captured how patterns in human song might reflects the social state of human societies in recent years and demonstrate the potential of applying functional diversity concepts and tools across scientific and economic domains, extending its relevance beyond ecology. By demonstrating applications of state-of-the-art functional diversity metrics using music as a case study, we intent to communicate the often-complex concepts of functional diversity using the familiar realm of music, which is an intrinsic trait of human cultures across the globe.

Fig 1. Functional space of top and bottom artists.

Two-dimension functional space of the characteristics of the songs from the top and bottom five artists in terms of functional richness (panels in the left side), evenness (panels in the center), and divergence (panels in the right). The squares in the graph represents the centroid distribution of the traits of artists. Arrows in the center of the graph indicate which is the most defining characteristic of songs in that direction (i.e., only for traits in which loadings were higher than |0.3|). Variable loadings are provided in S5 Table in S1 File.

Fig 2

Associations between functional (a) richness, (b) evenness (c), and divergence of artists. Points are colored according to the year when artists launched their career. Second Y axis represents total profit considering that 1 stream on Spotify leads to $0.00437 of profit [40]. Total number of streams and profit are represented in a logarithm scale for better visualization of data distribution. Lines represent the direction of significant associations between the two variables and the shaded area represent its 95% confidence interval.

Fig 3. Associations between popularity and traits.

Standardized effect size of the association between the selected traits and the number of times that (a) artists, (b) their albums, and (c) songs were played in Spotify. Whiskers represent the 95% confidence interval of the mean estimate, which is represented by the points. Color of whiskers represents the size of the effect in the negative direction (blue whiskers) and the positive direction (red whiskers). Traits that do not have a whisker in a plot were not selected in the linear model after stepwise regression (selected models in S8 and S9 Tables in S1 File).