Jef Vlegels, Universiteit Antwerpen
Music classification, genres, and taste patterns: a ground-up network analysis on the clustering of artist preferences
In the early 1990s, Peterson (1992; 1996; 1992) introduced the concept of the cultural omnivore, when he discovered a shift in American music taste from a highbrow-lowbrow distinction to a contrast between high-status omnivores–who combine elite and popular culture–and lower-status univores (Bourdieu, 1984; Peterson, 1992). Since then, the cultural omnivore concept has been widely debated in cultural sociology. During the last two decades, researchers around the world have shown the prevalence of cultural omnivorousness in a variety of social settings (e.g. Peterson, 2005; Peterson & Kern, 1996; Stichele & Laermans, 2006; van Eijck, 2001; van Eijck & Lievens, 2008). Some researchers have focused on ‘volume’ measurements, based on a score or a scale that counts the number of genres a person likes in order to quantify the ‘voraciousness’ of a cultural consumer (e.g. Bryson, 1996). Others have used a ‘compositional’ approach, which concentrates on specific combinations of music genres (e.g. van Eijck & Lievens, 2008). Although there is substantial variety in the operationalization of the cultural omnivore in quantitative research, the use of genres as a starting point seems to be beyond discussion. Researchers use the liking of genres, often combined with a dimension-reduction technique such as latent class analysis or factor analysis, to construct cultural omnivorousness measurements in terms of volume or composition of taste. Nevertheless, the use of broad music genre preferences was questioned by Bourdieu (1984), who insisted that music cannot simply be categorized into cultural genres, as there will be differences in the specific types of musical works that are part of these genres. Recent quantitative and qualitative work on boundary drawing around music genres has confirmed this proposition. Genre boundaries are ‘fuzzy’ and the hidden dimensions in genre categories are often overlooked by researchers (e.g. Beer, 2013; Beer & Taylor, 2013; Savage, 2006; Savage & Gayo, 2011; Sonnett, 2016; van Venrooij, 2009). In this presentation, we will show why research on the cultural omnivore based on preferences for a predefined list of music genres might start off on the wrong foot. Inspired by Lamont’s call to study classification systems ‘from the ground up’ (Lamont, 2010, p. 132), we offer an alternative strategy to measure music taste and taste patterns by using an open question about artist preferences. We build on existing knowledge in social network analysis to construct a two-mode network of people and music artists–conforming to the duality between people and cultural items– and we use state-of-the-art network clustering techniques (infinite relational models) to identify clusters of people who have similar relationships to the same set of artists (Breiger, 1974; Dimaggio, 1987). By using this bottom-up approach, taste patterns emerge from the data, an ad hoc list of genres becomes redundant, and we obtain a more detailed measurement of music taste patterns. In general, this analysis also shows how new advances in social network analysis methodology can be relevant for the study of symbolic boundaries in almost every aspect of society.
