Generalizing case-based analyses in the study of global music consumption

Generalizing case-based analyses in the study of global music consumption

Matthew Woolhouse, McMaster University: woolhouse@mcmaster.ca

James Renwick, University of Toronto: james-497@hotmail.com

Peer-reviewed by: Ian Milligan, University of Waterloo; Ichiro Fujinaga, McGill University.


Abstract / Résumé

Using large-scale analysis of a music-download database provided by MixRadio, a leading online music-service provider formerly under Nokia ownership, this paper investigates the following in relation to global-music consumption: (1) differences in the downloading trajectories of various musical genres; (2) the extent to which downloading trajectories are invariant with respect to countries and/or genres; and (3) possible links between the downloading behaviour of various genre-defined user subgroups and pre-existing music-personality studies. Substantial differences were observed between download trajectories pertaining to pop, rap, rock and metal—metal, in particular, was seen to exaggerate features of rock's trajectory. Of the genres studied, metal was found to have the only invariant trajectory, seemingly impervious to the local conditions of the country in which it was downloaded. Similarly, musical styles within Brazil were found to be downloaded in more or less the same way, in contrast to the UK where download trajectories varied with musical genre. Lastly, the analysis demonstrated a statistical link between pre-existing music-personality research and patterns of downloading within the metadata, suggesting that different downloading behaviours are due, in part, to differences in personality.

Au moyen d’une analyse à grande échelle d’une base de données de téléchargement de musique fournie par MixRadio, un fournisseur principal de service de musique en ligne qui appartenait auparavant à Nokia, cet article examine ce qui suit en ce qui concerne la consommation musicale : (1) les différences de trajectoires de téléchargement en ce qui concerne les divers genres de musique; (2) la mesure dans laquelle les trajectoires de téléchargement sont invariantes selon les pays et/ou les genres; et (3) des liens possibles entre le comportement de téléchargement de divers sous-groupes d’utilisateurs dont le genre est défini et les études de personnalité musicale actuelles. Des différences importantes ont été observées entre les trajectoires de téléchargement en ce qui concerne les genres pop, rap, rock et métal — métal, en particulier, a été perçu comme une exagération des caractéristiques de la trajectoire du rock. De tous les genres étudiés, on a constaté que seul le métal avait une trajectoire invariante, apparemment insensible aux conditions locales des pays où il a été téléchargé. De la même façon, on a constaté que les styles musicaux téléchargés au Brésil l’étaient plus ou moins de la même façon, contrairement au Royaume-Uni, où les trajectoires de téléchargement variaient selon le genre musical. Enfin, l’analyse a démontré un lien statistique entre les recherches de personnalité musicale actuelles et les modèles de téléchargement au sein des métadonnées, ce qui suggère que différents comportements de téléchargement sont dus, en partie, aux différences de personnalité.

Keywords / Mots Clés

Music, download, genre, personality, databases, rock, pop, metal, rap



Introduction

Music-downloading and live-stream services such as MixRadio, Last.fm and Spotify are providing researchers with unprecedented access to global music-listening data. Part of the research appeal of these rapidly expanding databases undoubtedly lies in the collaborative filtering these online music-service providers employ. Whether to navigate, find resources or engage in fortuitous browsing, mood and genre labels together with play and download information are a rich resource through which discoveries are being made concerning the influence on global music consumption of extra-musical, sociocultural factors, including human personality.

In this article, worldwide music-download patterns are explored through "case multiplication"—a simple, iterative method whereby multiple tracks are sampled, representing many of millions of downloads. Our aim is to produce a representative sample for various musical genres and countries, thereby reducing the chance of false-positive/negative results due to sampling error. For each large-scale study reported in this paper, we first briefly recapitulate the findings of various relevant case studies, previously reported in Woolhouse, Renwick and Tidhar (2014), outlining the motivations underpinning our searches, queries and questions. Second, we investigate differences in download trajectories of various genres—our approach is to (1) identify the subgroups of users associated with each genre, (2) examine their downloading behaviour in terms of genre exclusivity, and (3) link this to pre-existing music-preference and personality studies. This method aimed to demonstrate how personality can, albeit relatively weakly, influence downloading behaviour, and in so doing enrich our understanding of what motivates people from across the world to listen to the music they do.

This paper significantly expands upon a series of music-consumption case studies, presented in an earlier volume of this journal (Woolhouse, Renwick and Tidhar 2014). The case studies in question reported a number of intriguing findings showing significant differences in (1) download trajectories of tracks from different genres, and (2) patterns of track downloads in different countries. Woolhouse, Renwick and Tidhar (2014) also investigated how distinct events, categorized as either "musical" or "extra-musical" depending on whether they related to an artist's creative output, impacted music downloading. Here, we go well beyond the scope of our previous research and uncover significant music-consumption differences, not only for subgroups of users with dissimilar musical tastes but also downloading patterns between entire countries. Two notable examples, which we discuss in detail, are heavy-metal music and Brazil, both of which are found to be remarkably impervious to the usual "boom and bust" consumption trajectories of most musical genres in most countries.

At the heart of our study is the following question: do people with different music-genre preferences exhibit different behaviours with respect to downloading, and if so, why? Although answering the first part of this question was relatively straightforward—the issue was addressed by querying and recording the download patterns of genre-defined user subgroups—why different musical-genre preferences might be linked to different behaviours is messy and complex, involving, as it does, human perception, culture and aspects of personality. The issues of musical genre, and links between music and personality are now briefly discussed.

Despite musicologist Richard Middleton's assertion that "[n]eat divisions between 'folk' and 'popular,' and 'popular' and 'art,' are impossible to find... arbitrary criteria [which are used] to define the complement of 'popular" (Middleton 1990, 4), there is a long and strong tradition of genre classification within music, dating at least to Claudio Monteverdi's style distinctions prima and seconda pratica of the early sixteenth century (Palisca 2014). Used taxonomically, genre can be an effective genealogical tool, tracing the development of new styles from existing genres, or showing how multiple genres have coalesced to form something new and distinguishable (Pachet and Cazaly 2000). Nor are these distinctions merely theoretical or academic; perceptual studies show that humans are highly sensitive to style differences within music, even when presented with musical fragments as short as 300 milliseconds in length (Bigand et al. 2011; Krumhansl 2010).

Musicologically defined, genres are consensus-derived style categories that enable pieces of music to be grouped according to a shared set of musical features, including instrumentation, texture, melodic and chordal patterning, lyric content, gender and playing style of performers (Samson 2014). Quantifiable audio features contributing to genre have allowed researchers to develop machine-learning tools designed to estimate musical genre automatically and objectively (McKay and Fujinaga 2004; Tzanetakis and Cook 2002). However, despite the relative success of this approach, given that genre depends to some extent upon the social context in which the music is heard (i.e., cultural reception is key; Rentfrow and Gosling 2007), human judgments and computationally derived genre classifications are unlikely ever to align precisely (Lippens, Martens and Mulder 2004). That is to say, although subject to variance, human genre classifications that take into account the complexity of human culture provide an important ground truth with respect to the categorisation of music. Genre information in the database used in this study originates from the various record labels holding rights to particular artists and songs, and is therefore derived entirely from human judgements. And while some genres are admittedly very broad, subsuming multiple sub-genres, they are at least applied by individuals with extensive experience and style-knowledge within the music industry. In total, 64 genres exist within the database, ranging from ubiquitous pop, rock, rap and dance to less-well represented genres such as flamenco, khaliji (a style of music characteristic of Arab states of the Persian Gulf) and ambient/new age music.

Music's power to communicate emotions (Eerola and Vuoskoski 2013; Madsen 1997) and carry extra-musical associations (DeNora 1986) may partly explain why people believe that it indicates more about someone than, for example, preferences for TV shows, movies, books, clothes or food (Rentfrow and Gosling 2003). Similarly, the development of personal identities, particularly among adolescents, including the formation of social and political attitudes, are moulded and reinforced through association with particular musical genres (Bennett 2000; Doak 2003; North, Hargreaves and O'Neill 2000; for an overview see MacDonald, Hargreaves and Miell (2009)). While the process whereby individuals choose to align themselves with certain types of music is undoubtedly complex, possibly unfathomably so, over the past two decades studies have established consistent links between musical preference and various attributes of personality (for overviews see North and Hargreaves 2008; Rentfrow and McDonald 2009). Although not all studies fully agree on specific music-personality mappings, sufficiently robust trends have emerged to suggest that it is possible to deduce something about aspects of personality based on preference for particular genres. For example, McNamara and Ballard (1999) found that people who exhibit sensation-seeking behaviour prefer music that is highly arousing, possibly due to needing greater sensory intensity (see also Litle and Zuckerman (1989)). While many genres can be arousing, it is undoubtedly the case that some are more so than others—metal, for example. Extroverts and those with higher levels of psychoticism tend to like musical styles with exaggerated bass, such as rap, dance and hard rock (McCown et al. 1997; Rawlings et al. 1995). And North (2010) correlated a wide variety of musical styles with a range of personality dimensions, including self-esteem, creativity, attitude to work, outgoingness and gentleness. In his study, people with high self-esteem positively correlated with jazz, opera and rock, amongst other genres, whereas those who negatively correlated included fans of heavy metal and indie music.

Our study, using pre-existing research into musical preference and personality, described in detail below, sought to establish whether a key feature of the Big-Five Inventory of Personality (John, Naumann and Soto 2008)—openness—correlated with levels of genre exclusivity for individuals within the database. Previous research has indeed found that people with high openness scores tend to like more musical genres than those with lower scores (Rawlings and Ciancarelli 1997). Our working hypothesis, therefore, was that users within the database with preferences for musical genres associated with openness should also have higher degrees of heterogeneity with respect to all musical genres. Or, alternatively put, the dominant genres of people with relatively homogeneous music collections should be those associated with lower openness within music and personality studies.

Before continuing, however, an important caveat should be mentioned. Despite the success of the studies referred to above highlighting personality, other factors can be at least as important in defining people's musical tastes. For example, although North (2010) successfully correlated various genres with a range of personality dimensions, liking for musical styles and participants' reasons for listening to particular pieces of music were more strongly predicted by age and gender. Therefore, although personality is a factor in musical preference, it is but one amongst potentially many components contributing to people's listening habits. Consequently, within our study we only expected to find evidence of broad trends consistent with our hypothesis, rather than unequivocal, definitive support. That is to say, due to the potential influence of multiple factors underpinning people's downloading behaviours, the relationship of genre to personality measures was likely to be relatively weak, if it were to exist at all. Before addressing this issue, we report our large-scale studies designed to expand upon previous case-based queries investigating the download trajectories of tracks from different genres and patterns of downloads in different countries.

For consistency, the data used in this study are from the same database as Woolhouse, Renwick and Tidhar (2014). Briefly, the database comprises over 180 million downloads, each containing metadata relating to a musical track downloaded by a mobile-phone user accessing Nokia's online music stores around the world. User attributes include date and (local) time of download, (anonymized) user ID, start date, download count, and country; track attributes include artist, genre, track name, duration, album, label, rights holder, and so on. As previously, these data are arranged into a relational database management system and queried using the open-source MySQL implementation of SQL (Weinberg, Groff and Oppel 2010). In addition, a Python Database API (Lemburg 2008) was used to run automated iterative queries, extracting and comparing information from multiple tracks and countries within the database. The Python Database API in particular allowed us to execute far more extensive analyses than had previously been undertaken; that is, to take our research from the particular to the general, as indicated in the title of this paper.

Given that downloads were made specifically onto Nokia mobile phones, the users in our study constitute a self-selected group. Despite this, Nokia has historically made a range of phone models to appeal to different market sectors. In 2013 Nokia released its Asha 501 series, a low-end phone that subsequently sold in the tens of millions. In fact, in a pattern replicated in numerous emerging economies, India's population predominantly accesses the Internet via smart phones, rather than desktop computers (Fontevecchia 2014). Therefore, although the self-selected users in the database may not be fully representative, it is assumed that they are relatively widely distributed throughout the workforce of the countries within our study (certainly more so than, for example, downloads onto iPhones, which in developing countries are viewed predominantly as top-end luxury products; Heneghan 2014).

Questions and results

Do different genres have different download trajectories?

Our previous study asked whether three popular songs typical of pop, rap and rock genres gave rise to different download trajectories (Woolhouse, Renwick and Tidhar 2014). The aim was to investigate whether differences in musical categorization based upon "style" translated into differences in music-consumption behaviour. To recapitulate our earlier results very briefly, significantly different download patterns were found between the tracks: the pop and rap tracks peaked quickly, within four to five months after release; the rock track's popularity grew relatively gradually, peaking 15 to 18 months after release.

Although based on hundreds of thousands of downloads, the genre study of Woolhouse, Renwick and Tidhar (2014) was case-based in that it investigated individual tracks by individual artists—the extent to which the analysis fully represented genre download trajectories was therefore not entirely clear. In order to address this sampling issue, here we are not concerned simply with individual tracks (representative of certain genres), but rather tracks en masse belonging to certain genres. To execute the analysis, the top 100 newly released tracks within each specified genre were extracted from the database; previously released or re-released tracks were excluded as prior public knowledge of these tracks is likely to cause modifications to their download trajectories. Second, a 12-month timeline for each extracted track consisting of monthly download counts was constructed. Third, a normalization procedure divided the monthly download count for each track by the track's total downloads. Finally, a mean download trajectory for each genre was calculated by averaging the monthly downloads of the 100 tracks within each genre. Results were graphed and error bars calculated at 95% confidence intervals.

Four genres were studied: pop, rap and rock, as per Woolhouse, Renwick and Tidhar (2014), and a fourth, new genre—metal. The rational for extending the genre list to include metal was as follows: given that metal can be viewed as an extreme form of rock (Straw 1984; Weinstein 2000), we hypothesized that the conservative features of the rock track's trajectory from our previous study might be exaggerated in tracks belonging to the metal genre. That is, that the download patterns associated with metal music might be "extreme" versions of rock's download patterns, reflecting the view that metal is an extreme form of rock. Figure 1 shows the averaged download trajectories from time of release for the top 100 track within each genre—pop, rap, rock and metal—covering a 12-month time period.

Figure 1: Average download trajectories for the top 100 tracks within genres pop, rap, rock and metal. The data cover a 12-month period from time of track release.

Average download trajectories for the top 100 tracks within genres pop, rap, rock and metal. The data cover a 12-month period from time of track release.

The different distributions evident in Figure 1 are reflected in the standard deviation (SD) of each genre's download trajectory, which in descending order are: rap, 5.11; pop, 3.04; metal, 2.93; rock, 1.72. This broadly replicates the standard deviation of each track's download trajectory found in our previous study (rap, 5.04; pop, 4.45; rock, 1.82), and, thus, arguably verifies our earlier case-based findings. Prima facie, metal's higher SD with respect to rock (2.93 versus 1.72) appears to refute our current hypothesis that the more conservative trajectory of the rock track reported in Woolhouse, Renwick and Tidhar (2014) might be exaggerated in the metal genre. Upon closer inspection, however, despite the two genres' SDs, metal's trajectory is, in many respects, indeed an exaggeration of rock. In comparison to rap and pop, which fall away quickly after steep initial ascents, rock remains largely flat, whereas metal actually "increases" throughout the twelve-month period. The exaggerated nature of metal's download trajectory in respect of other genres can be seen in Table 1, which contains coefficients from Pearson product-moment correlation analyses examining pairwise relationships between the genres' download trajectories. Coefficients in respect of pop's trajectory—the most ubiquitous genre in the database—are particularly revealing. Unsurprisingly, given their similar trajectories, the coefficient between rap and pop is high, 0.93; rock's coefficient is lower, although still positive, 0.47, while metal's is negative, -0.15. And in this respect, there is a neat linear trend moving from similarity (rap), via an intermediate step (rock), to dissimilarly (metal).

Table 1: Pearson product-moment correlation coefficients between pairs of genre-download trajectories.

Pop Rap Rock Metal
Pop 1 0.93 0.47 -0.15
Rap 0.93 1 0.26 -0.41
Rock 0.47 0.26 1 0.63
Metal -0.15 -0.41 0.63 1

As previously conjectured, marked differences between genre download trajectories may be due to a number of factors, including commercial advertising activities of the music industry (Tschmuck 2012), the extent to which local and national radio stations promote tracks belonging to particular genres (Negus 1993), and the role played by social media in spreading information about new releases. For research investigating the relationship between social media and music popularity, see Hauger and Schedl (2012), Koenigstein and Shavitt (2009), Schedl (2011) and Schedl et al. (2010). The speculative approach we take in this paper is to assume that some of the variance reported in relation to pop, rap, rock and metal may be due to personality differences between genre-defined user subgroups, which, in turn, influence the manner and extent to which new music is discovered, shared and ultimately downloaded. However, before reporting music-personality aspects of our research, we further explore the issue of genre by investigating differences between countries in respect of musical style.

Do genres have different download trajectories in different countries?

Expanding upon and generalizing the question in Woolhouse, Renwick and Tidhar (2014)—which investigated whether a popular song, released in multiple countries at the same time, produced different downloading patterns—here we ask whether songs from certain genres are downloaded differently en masse in different countries. To recapitulate briefly, in our previous study, European countries' download trajectories resembled those of pop and rap shown in Figure 1 above: following and initial steep ascent, downloading falls away rapidly. In contrast, however, Brazil's trajectory took eight months to peak, containing a secondary peak a year after release. That is, in comparison to the European countries in the study, the popularity of the song in Brazil emerged far more gradually, over an extended time period.

Our present intention was to examine this unusual aspect of Brazil's consumption of pop music, and to investigate whether this might only be a feature of specific genres or music consumption in Brazil in general, irrespective of musical style. Of particular interest was metal, shown above to be downloaded in a somewhat contrary manner to pop and rap. We wished to ascertain whether metal was "country-invariant" (downloaded the same everywhere) as opposed to the pop genre for example, which given our previous findings in respect of Lady Gaga we suspected might differ markedly based upon country.

Results of three genres downloaded in two nations are reported: pop, rap, and metal in Brazil, and the United Kingdom of Great Britain and Northern Ireland (UK). Brazil's 200-million-strong population is ethnically diverse. Consisting of 48% white (of European and Levantine descent), 44% pardo (multiracial Brazilians and assimilated, westernized Amerindians), and 7% black of African ancestry (Central Intelligence Agency, 2014), Brazil's varied cultural roots have given rise to a myriad of distinctive and attractive musical styles. Often subsumed under the supra-genre of Latin music, Samba, Bossa Nova, Setanejo, Pagoda, Música Popular Brasileira, Tropicalia and Choro (to mention only a fraction; McGowan and Pessanha 1998) are as significant culturally in Brazil as mainstream genres more familiar to western listeners such as pop, rap and rock (Stroud 2008). As a result, Brazil is musically eclectic, reflecting the multiple traditions that forged the modern state in the post-colonial aftermath of Portuguese rule (Perrone and Dunn 2002).

Some 9000 thousand kilometers northeast across the Atlantic, the UK is also host to a variety of musical styles, but for very different historical reasons to those of Brazil. Despite being 87% white, of native British, Irish and European origin (Central Intelligence Agency 2014), recent immigration from Asia and previously held territories in the Caribbean has greatly added to the cultural (and culinary) life of many of its inhabitants (Cook 2001), including musically. An example is reggae and its precursor ska, originating in Jamaica, which can be found in 60's Beatles songs such as "I call your name" and "Ob-La-Di, Ob-La-Da" (O'Brien Chang and Chen 1998, 44). Later British bands fully appropriating reggae's distinctive offbeat accents included UB40 (Hebdige 2003, 94), Paul McCartney's Wings and Police (Alleyne 2000). Bhangra, a musical style associated with Punjabi culture, is increasingly popular in Britain and is now regularly found in the works of non-South Asian musicians (Dixon 2003). And latterly, Bollywood, a filmic musical style used extensively within the Hindi-language cinema industry, has gained mainstream acceptance not only within the UK, but broadly throughout the west (Thussu 2008, 97). The UK, then, in addition to mainstays of rock and pop, has since the 50's assimilated a number of foreign musical idioms, many emanating from countries within the Commonwealth of Nations.

The analytical procedure was similar to the genre study shown in Figure 1. First, the top 100 newly released tracks per genre within each country were extracted from the database. Second, a 12-month timeline for each extracted track was constructed consisting of monthly download counts (normalized by dividing the monthly count for each track by the track's total downloads). Third, a mean download trajectory was calculated by averaging the monthly downloads of the 100 tracks per genre within each country. Figure 2 shows the results of the analysis; error bars are at 95% confidence intervals.

Figure 2: Average download trajectories for pop, rap and metal within Brazil and UK. The data cover a 12-month period from time of track release.

Average download trajectories for pop, rap and metal within Brazil and UK. The data cover a 12-month period from time of track release.

The download trajectories in Figure 2 fall into two distinct groups: UK pop and rap, and the remaining trajectories, UK metal and all genres pertaining to Brazil. The distinction between the groups can be illustrated by considering the SD of each trajectory, which for UK pop and rap (first group) are 8.57 and 9.27 respectively, and UK metal and Brazil pop, rap and metal (second group) 1.09, 3.09, 2.84 and 1.03 respectively. The difference in the groups' SDs reflects the fact that UK pop and rap peak and fall away relatively rapidly, whereas UK metal and all of Brazil's genre trajectories are relatively flat. The contrast in this large-scale study between the consumption in Brazil and the UK of pop is largely consistent with the finding of our previous case-base analysis of Lady Gaga's "Bad romance," which produced a similarly flat profile in Brazil, whilst spiking quickly in Europe (in Germany and Italy). In this analysis we also sought to discover whether the flat nature of Brazil's pop-music consumption might also occur in other genres within the country, and indeed from Figure 2 this indeed appears to be the case. The SDs of all of Brazil's genre trajectories are relatively low, ranging from 1.03 (metal) to 3.09 (pop), indicating that Brazilian musical culture constraints how music is consumed irrespective of genre. This constraint, however, is not found within the UK, in which there are marked differences between the genres. Most noticeable is the fact that metal, in contrast to pop and rap, is flat (matching its trajectory in Brazil). And, therefore, there is evidence to suggest that metal downloading is country-invariant, i.e., that metal exerts constraints on how it is consumed irrespective of the country in which it is downloaded.

In addition to generalizing and expanding the case-based studies of Woolhouse, Renwick and Tidhar (2014) through the use of large, representative samples of musical genres within each country, our aim was also to explore how music downloading might be linked to aspects of personality. Consequently, the next section presents an analysis in which a key aspect of downloading behaviour—genre exclusivity within people's music collections—is compared with data from pre-existing research exploring music-genre preferences and personality.

Is there a link between music downloading and personality?

As outlined in the Introduction to this paper, our analysis investigated whether a key feature of the Big-Five Inventory of personality—openness—correlated with levels of genre exclusivity for individuals within the database. In brief, the Big-Five Inventory consists of openness, conscientiousness, extroversion, agreeableness, and neuroticism. Openness is associated with attitudes to new experiences, intelligence, creativity and independence, whereas extroversion, for example, measures sociability, which includes features such as the desire to interact with others, talkativeness, and assertiveness (for overview, see John, Naumann and Soto 2008). Our analysis sought to discover whether users within the database with preferences for musical genres associated with openness had greater heterogeneity within their music collections in respect of genre. That is to say, we hypothesized that there is a positive relationship between openness and a willingness to encounter different musical genres.

Our analytical method was briefly as follows. First, each user within the database was classified as an "x-head," where x was the predominant genre within their download collection. For example, a user with 60% rock and 40% classical would be classified as a "rock-head"; 70% jazz and 30% folk a "jazz-head," and so on. Second, subgroups of users were created based on x, i.e., the most popular genre within each user's download collection. Third, a degree of genre exclusivity was ascertained for each user by calculating the SD of the genre proportions within their collection. To use the somewhat simplified examples above, the SD of a user with 60% rock and 40% classical is 14, whereas the SD of a user with 70% jazz and 30% folk is 28. The higher the SD, the more exclusive that user is with respect to genre. And thus, in this example, the jazz-head is more exclusive than the rock-head, 28 to 14. Fourth, a median SD per x-head subgroup per country was calculated. Fifth, these values were correlated with levels of openness from a pre-existing studying associating this aspect of personality with particular genres.

The music-personality data used in our study was from an earlier study by Dollinger (1993), whose population demographics and genres aligned well with those in our database. In brief, Dollinger predicted how likely certain personality types are to prefer specific musical genres by correlating results from personality (the NEO-PI; Costa and McCrae 1992) and music-preference surveys completed by the same individuals. In his study, openness was found to have a positive relationship with preference for a variety of non-mainstream genres such as classical and jazz. In addition, Dollinger also found that extraversion positively correlated with a preference for highly arousing music such as upbeat jazz; people who scored well in "excitement seeking" tended to prefer metal and hard rock. Given our music-exclusivity-based hypothesis, our focus was on musical genres associated with openness. Figure 3 is a scatterplot showing median SDs of x-heads' music collections per country (i.e., genre exclusivity for each genre within each country) against openness levels associated with specific genres from Dollinger (1993). [1]

Figure 3: Scatterplot of openness against genre exclusivity. X-axis shows median SD per x-head subgroup per country; y-axis shows openness levels per genre from Dollinger (1993).

Scatterplot of openness against genre exclusivity.
                                        X-axis shows median SD per x-head subgroup per country;
                                        y-axis shows openness levels per genre from Dollinger (1993).

The scatterplot and linear trend line in Figure 3 shows a negative correlation between the genre-openness associations reported in Dollinger (1993) and genre exclusivity within the database, calculated as the median SD per x-head subgroup per country: r(109) = -0.4; p < 0.001. Recall that higher SD values, caused by a greater skew in favour of one genre over another, represents a tendency for exclusivity within someone's download collection. The association therefore of x-head subgroups with higher SD values, such as pop, rock and country, with lower openness scores in Dollinger (1993) is consistent with our hypothesis that there is a relationship between personality and downloading behaviour. Specifically, that users within the database with preferences for musical genres associated with lower openness have a tendency for greater homogeneity within their downloads, and vice versa.

Summary conclusions

This paper sought to generalize and expand upon some of the issues surrounding the consumption of music, initially presented in Woolhouse, Renwick, and Tidhar (2014) as a series of largely case-based analyses. Here, we began by asking whether people with different music-genre preferences exhibit different behaviours with respect to downloading, and if so, why? As a first step towards analyzing the "why" part of our question, we attempted to link genre exclusivity—a particular type of downloading behaviour—with openness, a dimension of personality associated with attitudes to new experiences, intelligence, creativity and independence. Broadly speaking, the analysis succeeded in demonstrating a statistical link between pre-existing music-personality research and patterns within the metadata of our database, and thus we surmise that why people with different music-genre preferences exhibit different downloading behaviours is due, in part, to differences in personality. However, as discussed earlier, music-personality studies have also highlighted the role of some other factors in guiding people's musical tastes. Therefore, it is likely that future modeling of musical behaviours such as genre exclusivity will require the incorporation of additional independent variables, relating to other personality attributes, or perhaps age and gender.

Prior to exploring music downloading and personality, two related questions were addressed. The first concerned whether different genres had different download trajectories; the second dealt with the issue of genre versus country invariance. Both analyses largely verified earlier findings from Woolhouse, Renwick, and Tidhar (2014). For instance, substantial differences were observed between the 12-month download trajectories pertaining to pop, rap, rock and metal—metal, in particular, was seen to exaggerate features of rock's trajectory, consistent with the view that, stylistically at least, metal is an extreme form of rock (Weinstein 2000). Metal was further explored in the context of Brazil and the UK. Here, of the genres studied, metal was found to have the only invariant trajectory, seemingly impervious to the local conditions of the country in which it was downloaded. Similarly, trajectories associated with Brazil appeared impervious to genre; which is to say, within Brazil styles are downloaded in more or less the same way, in contrast to the UK, as shown in Figure 2.

Why certain genres and countries appear to be invariant with respect to music consumption is not immediately apparent. In the case of metal, Arnett (1996) found empirical support for the notion that metal is a specific subculture, frequented to a significant degree by adolescents who, for a variety of social causes, experience a heightened sense of alienation. If this is true, it is possible that the alienation prevalent within the metal-head community extends to, among other things, the music industry itself, and thus renders it largely immune to top-down commercial activities. As an interesting aside, consistent with Arnett's (1996) theme of alienation, it should be noted that the x-head subgroup with the highest levels of genre exclusivity in Figure 3, along with pop, is metal.

The pluralistic ethnic and cultural landscape pervading Brazilian society, discussed earlier in this paper, may partly explain the slow uptake of newly released tracks—it is difficult for any single genre or artist to become dominant within an intensely heterogeneous society and thus hard to capture a critical mass of media attention which could result in sudden, large-scale popularity spikes seen elsewhere. So, too, does community ownership of local media (usually through license to local civic associations) contribute to creating a widely distributed media system (Boas 2013). This is particularly true of radio, where legal reform beginning in the 90's had, as of June 2003, resulted in the proliferation of over 2,300 community radio stations in Brazil (Boas and Hidalgo 2011), each capable of promoting local artists free from commercial pressures. However, although the factors outlined above may have contributed to the relatively flat downloading trajectories found in Figure 2, further research will have to address other possible causes if the unusual music-consumption patterns found in Brazil are to be thoroughly explicated. Whatever the outcome of that research, the resistance in Brazil to behaving in a herd-like manner, apparent in other countries, is surely a salutary warning to any artist wishing or expecting to take Brazil's rich musical world by storm.

At the opening of this paper, reference was made to the impact on digital scholarship of large music-listening databases. MixRadio's download database, for example, is allowing broad, global trends in music consumption to be studied in a manner that was hitherto virtually impossible. The explanation of trends within global data requires researchers to adopt an interdisciplinary approach, in which connections between consumption patterns may be made with socially embedded statistics and indices, such as those published annually by the United Nations (e.g., the Human Development Index; see Woolhouse and Bansal (2013)). However, before interpretive and critical skills can be applied in full, it behooves those working within digital scholarship to establish bodies of empirically based, verifiable data against which ideas and theories may be reckoned and tested. The research presented in this paper attempts modestly to do both of these things: firstly, to extract concrete longitudinal information relating to consumption patterns of particular musical genres in various countries; and secondly, to interpret these data, with the aim of showing how music maybe seen to relate to socio-cultural factors, including musical preference and personality. The extent to which this sort of approach is ultimately successful will depend upon continued access to high-quality, commercially generated data, and the degree to which those of us working in the field are willing to undertake imaginative, interpretative, and interdisciplinary research.


Acknowledgements / Remerciements

The authors would like to express their thanks to the following people and organizations, who in various ways have contributed to the research presented in this paper: Matthew Harvell Pearson at MixRadio (Microsoft, UK); Nick Rogers and Michael Barone (Research Assistants, Digital Music Lab, McMaster University). In particular, thanks are due to Jotthi Bansal for her significant contribution to the research exploring the relationship between music downloading and personality. This research was supported by the Social Sciences and Humanities Research Council of Canada (SSHRC; Insight Development Grant #430-2012-0835). Some of the research referred to in the paper was first presented at the Annual conference of the Canadian Society for Digital Humanities/Société canadienne des humanités numériques, St Catharines, Ontario, 28-30 May 2014.

Notes

[1] Nine genres in Dollinger's study matched those in our database: classical, country, dance, folk, jazz, metal, pop, rock, and soul/funk; these are listed on the right y-axis in Figure 3. Thirteen countries in our database matched the western-centric demographic of Dollinger's participants: Austria, Finland, France, Germany, Ireland, Italy, Netherlands, Norway, Poland, Portugal Sweden, Switzerland and the UK; the 13 countries per genre are represented by horizontally-positioned markers sharing the same colour. Six outliers were removed from the analysis.


Works Cited / Liste de références

Alleyne, Mike. 2000. "White reggae: Cultural dilution in the record industry." Popular Music and Society 24.1: 15-30.

Arnett, Jeffrey J. 1996. "Metalheads: Heavy metal music and adolescent alienation." Boulder, Colorado: Westview Press.

Bennett, Andrew. 2000. "Popular music and youth culture: Music, identity and place." London: Macmillan Press.

Bigand, Emmanuel, Charles Delbé, Yannick Gérard, and Barbara Tillmann. 2011. "Categorization of extremely brief auditory stimuli: Domain-specific or domain-general processes?"  PloS ONE 6.10: e27024.

Boas, Taylor C. 2013. "Mass media and politics in Latin America." In Constructing democratic governance in latin american, edited by Jorge I. Domínguez and Michael Shifter. 4th ed. 340-364. Baltimore: The John Hopkins University Press.

Boas, Taylor C., and F. Daniel Hidalgo. 2011. "Controlling the airwaves: Incumbency advantage and community radio in Brazil."  American Journal of Political Science 55.4: 869-885.

Central Intelligence Agency. 2014. "Brazil." The World Factbook. Accessed November 1. https://www.cia.gov/library/publications/the-world-factbook/geos/br.html.

Cook, Robin. 2001. "Robin Cook's chicken tikka masala speech."  The Guardian, April 19. Accessed November 1, 2014. http://www.theguardian.com/world/2001/apr/19/race.britishidentity.

Costa, Paul T., and Robert R. McCrae. 1992. "Revised NEO personality inventory (NEO-PI-R) and NEO Five-Factor Inventory (NEO-FFI) professional manual." Odessa, FL: Psychological Assessment Resources.

DeNora, Tia. 1986. "How is extra-musical meaning possible? Music as a place and space for 'work." Sociological Theory 4.1: 84-94.

Dixon, Martha. 2003. "Bhangra fusion gathers support." British Broadcast Corporation News, September 18. Accessed November 1, 2014. http://news.bbc.co.uk/1/hi/uk/3117432.stm.

Doak, Bridget A. 2003. "Relationships between adolescent psychiatric diagnoses, music preferences, and drug preferences."  Music Therapy Perspectives 21.2: 69-76.

Dollinger, Stephen J. 1993. "Research note: Personality and music preference: Extraversion and excitement seeking or openness to experience?" Psychology of Music 21.1: 73-77.

Eerola, Tuomas, and Jonna K. Vuoskoski. 2013. "A review of music and emotion studies: Approaches, emotion models, and stimuli." Music Perception: An Interdisciplinary Journal 30.3: 307-340.

Fontevecchia, Agustino. 2014. "India's 243 million internet users and the mobile e-commerce revolution." Forbes Magazine, July 7. Accessed November 1, 2014. http://www.forbes.com/sites/afontevecchia/2014/07/07/indias-massive-e-commerce-opportunity-and-the-explosion-of-mobile/.

Hauger, David, and Markus Schedl. 2012. "Exploring geospatial music listening patterns in microblog data." Paper published in AMR 2012: 10th international workshop on adaptive multimedia retrieval, Copenhagen, Denmark, October 24-25, 1-15.

Hebdige, Dick. 2003. Cutn'Mix: Culture, identity and caribbean music. New York: Routledge.

Heneghan, Carolyn. 2014. "Apple: Luxury brand in the making?" The Richest, January 26. Accessed November 1. http://www.therichest.com/business/technology/apple-luxury-brand-in-the-making/.

John, Oliver P., Laura P. Naumann, and Christopher J. Soto. 2008. "Paradigm shift to the integrative Big Five trait taxonomy." Handbook of Personality: Theory and Research 3: 114-158.

Koenigstein, Noam, and Yuval Shavitt. 2009. "Song ranking based on piracy in peer-to-peer networks." Paper published in ISMIR: 10th international society for music information retrieval conference, Kobe, Japan, October 26-30, 633-638.

Krumhansl, Carol L. 2010. "Plink: 'Thin slices' of music." Music Perception: An Interdisciplinary Journal 27.5: 337–354.

Lemburg, M. A. 2008. "Python database API specification v2.0."  Python Enhancement Proposal, 249.

Lippens, Stefaan, Jean-Pierre Martens, and Tom De Mulder. 2004. "A comparison of human and automatic musical genre classification." Paper published in Proceedings (ICASSP'04) of IEEE international conference on acoustics, speech, and signal processing, 2004, Montreal, Canada, May 17-21. 4: 233-236.

Litle, Patrick, and Marvin Zuckerman. 1989. "Sensation seeking and music preferences." Personality and Individual Differences 7.4: 575-578.

MacDonald, Raymond, David J. Hargreaves and Dorothy Miell. 2009. "Musical identities." In Oxford handbook of music psychology, edited by Susan Hallam, Ian Cross, and Michael Thaut, 462-470. Oxford: Oxford University Press.

Madsen, Clifford K. 1997. "Emotional response to music." Psychomusicology: A Journal of Research in Music Cognition 16.1-2: 59-67.

McCown, William, Ross Keiser, Shea Mulhearn, and David Williamson. 1997. "The role of personality and gender in preference for exaggerated bass in music." Personality and Individual Differences 23.4: 543-547.

McGowan, Chris, and Ricardo Pessanha. 1998. The Brazilian sound: Samba, Bossa Nova, and the popular music of Brazil. Philadelphia: Temple University Press.

McKay, Cory, and Ichiro Fujinaga. 2004. "Automatic genre classification using large high-level musical feature sets." Paper published in ISMIR: 5th international society for music information retrieval conference, Spain, Barcelona, October 10-15, 525-530.

McNamara, Linda, and Mary E. Ballard. 1999. "Resting arousal, sensation seeking, and music preference." Genetic, Social, and General Psychology Monographs 125.3: 229-251.

Middleton, Richard. 1990. Studying popular music. Berkshire: Open University Press.

Negus, Keith. 1993. "Plugging and programming: Pop radio and record promotion in Britain and the United States." Popular Music 12.1: 57-68.

North, Adrian C. 2010. "Individual Differences in Musical Taste." American Journal of Psychology 123.2: 199-208.

North, Adrian C., and David J. Hargreaves. 2008. The social and applied psychology of music. Oxford: Oxford University Press.

North, Adrian C., David J. Hargreaves, and Susan A. O'Neill. 2000. "The importance of music to adolescents."  British Journal of Educational Psychology 70.2: 255-272.

O'Brien Chang, Kevin, and Wayne Chen. 1998. Reggae routes: The story of Jamaican music. Philadelphia: Temple University Press.

Pachet, François, and Daniel Cazaly. 2000. "A taxonomy of musical genres." Paper published in Content-based multimedia information access conference, Paris, France, March, 1238-1245.

Palisca, Claude V. 2014. "Prima prattica." Grove Music Online, edited by L. Macy. Accessed October 20, 2014. http://www.grovemusic.com/.

Perrone, Charles A., and Christopher Dunn. 2002. "Brazilian popular music and globalization."  Journal of Popular Music Studies 14.2: 163-165.

Rawlings, David, and Vera Ciancarelli. 1997. "Music preference and the Five-Factor Model of the NEO Personality Inventory."  Psychology of Music 25.2: 120-132.

Rawlings, David, Malcolm Hodge, Debbie Sherr, and Anne Dempsey. 1995. "Toughmindedness and preference for musical excerpts, categories and triads." Psychology of Music 23.1: 63-80.

Rentfrow, Peter J., and Samuel D. Gosling. 2003. "The do re mi's of everyday life: The structure and personality correlates of music preferences." Journal of Personality and Social Psychology 84.6: 1236-1256.

Rentfrow, Peter J., and Samuel D. Gosling. 2007. "The content and validity of music-genre stereotypes among college students." Psychology of Music 35.2: 306-326.

Rentfrow, Peter J., and Jennifer A. McDonald. 2009. "Music preferences and personality." In Handbook of music and emotion, edited by P. N. Juslin, and J. Sloboda, 669-695. Oxford: Oxford University Press.

Samson, Jim. "Genre." Grove Music Online, edited by L. Macy. Accessed October 20, 2014. http://www.grovemusic.com/.

Schedl, Markus. 2011. "Analyzing the potential of microblogs for spatio-temporal popularity estimation of music artists." Paper published in IJCAI: International workshop on social web mining, Barcelona, Spain, July 18, 539-553.

Schedl, Markus, Tim Pohle, Noam Koenigstein, and Peter Knees. 2010. "What's hot? Estimating country-specific artist popularity." Paper published in ISMIR: 11th international society for music information retrieval conference, edited by J. Stephen Downie, and Remco C. Veltkamp, Utrecht, Netherlands, August 9-13, 117-122.

Straw, Will. 1984. "Characterizing rock music cultures: The case of heavy metal." Canadian University Music Review 5: 104-122.

Stroud, Sean. 2008. The defense of tradition in Brazilian popular music. Politics, culture and the creation of MPB. London: Ashgate.

Thussu, Daya K. 2008. "The globalization of 'Bollywood' - The hype and hope." In Global Bollywood, edited by A.P. Kavoori, and A. Punathambekar, 97-113. New York: New York University Press.

Tschmuck, Peter. 2012. Creativity and innovation in the music industry. Springer Berlin Heidelberg.

Tzanetakis, George, and Perry Cook. 2002. "Musical genre classification of audio signals."  Transactions on Speech and Audio Processing, Institute of Electrical and Electronics Engineers (IEEE) 10.5: 293-302.

Weinberg, Paul N., James R. Groff, and Andrew J. Oppel. 2010. SQL, The complete reference. New York: McGraw-Hill.

Weinstein, Deena. 2000. Heavy metal: The music and its culture. Boston: Da Capo Press.

Woolhouse, M. H. & Bansal, J. 2013. "Work, rest and (press) play: Music consumption as an indicator of human economic development." Journal of Interdisciplinary Music Studies 7/1&2: 45-71.

Woolhouse, Matthew H., James Renwick and Dan Tidhar. 2014. "Every track you take: Analysing the dynamics of song and genre reception through music downloading." Digital Studies/Le champ numérique 4. http://www.digitalstudies.org/ojs/index.php/digital_studies/article/view/266/321.

Valid XHTML 1.0!



Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 License.