Does Good Music Sell Itself? The Effect of Prestige on Music Appreciation The Differences between Two Types of Music Listeners

Transcription

1 FACULTY OF ECONOMICS AND BUSINESS Does Good Music Sell Itself? The Effect of Prestige on Music Appreciation The Differences between Two Types of Music Listeners Pieter Schrevens s Thesis submitted to obtain the degree of MASTER OF BUSINESS ECONOMICS Major Marketing Promoter: Prof. Dr. Siegfried Dewitte Assistant: Jan Verpooten Academic year

2 FACULTY OF ECONOMICS AND BUSINESS Does Good Music Sell Itself? The Effect of Prestige on Music Appreciation The Differences between Two Types of Music Listeners Both content and context play an important role in our music appreciation; yet context can sometimes overshadow content, as research has shown that people tend to like songs more just because they are popular. But does music appreciation change when songs are linked to a prestigious music magazine? And if so, are certain people more susceptible for this prestige bias than others? This paper tries to answer these questions by examining two different types of music listeners: savant/enthusiastic and casual/indifferent. In an online survey, 110 participants were asked to rate ten unfamiliar songs with varying amounts of loudness. Loudness was used to measure the effect of content, as it is known to negatively affect perceived quality. After that, a questionnaire determined the listener type for each participant. Results show that casual music listeners gave significantly higher ratings when told that the songs were from a prestigious playlist, while this did not influence enthusiastic music listeners. Although the amount of compression negatively affected music appreciation on an overall level, it did not moderate music appreciation for the two listener types, which means that content is not overridden by context. This paper only scratches the surface of the different possibilities for which prestige can influence our music appreciation; yet it provides a useful insight by distinguishing between two types of listeners. Pieter Schrevens s Thesis submitted to obtain the degree of MASTER OF BUSINESS ECONOMICS Major Marketing Promoter: Prof. Dr. Siegfried Dewitte Assistant: Jan Verpooten Academic year

3 Acknowledgements First of all, I would like to express my sincere gratitude to my promoter Prof. Dr. Siegfried Dewitte and his assistant Jan Verpooten for their meticulous proofreading, guidance and for providing me with useful feedback throughout the entire writing process. Secondly, I would like to thank Toon Stegen and the band Bibles for letting me use one of their songs in my experiment, and also Annie Boyle and Daniel Kellagher for proofreading my final draft. Last but not the least, I would like to thank my parents, my sisters, my grandparents and my friends for supporting me during the writing of this master s thesis. I

4 Table of Contents Acknowledgements... I General introduction Literature review A good song sells itself, or does it? The importance of social influence An evolutionary point of view Content bias Context bias Different types of music listeners The diffusion of innovation Savant vs. indifferent music listeners Combining Roger s theory with dual-inheritance theory Hypothesis Present Research Compression and perceived quality Prestige cue Stimuli and manipulations Methods and results Participants Procedures and measures Results Music expertise Fragment score Main effects and moderations Effects on individual songs Discussion The main findings Limitations and future research Considerations for compression Considerations for prestige Conclusion Appendices List of figures List of tables Sources II

5 General introduction Marketers often say that a good product sells itself, but whether this applies to music is debatable. First of all, there is no definition of a good song. What one person considers to be a great song may be complete rubbish for another person, since everyone has their own musical taste. Secondly, whether a song becomes popular or not is usually very hard to predict, even by the most seasoned music experts. Additionally, whether we like a song or not does not only depend on how good it is but also on how popular (or exclusive) it is and who listens to it. In fact, it is even shown that people tend to like songs more just because they are popular. Cultural evolution theorists call this phenomenon frequency-based bias, referring to the tendency of humans to adapt their behavior to what is most common. Similarly, the tendency to imitate the behavior of a model is called model-based bias, and happens when someone listens to music because other people, who they consider a model, do so too. A special case of model-based bias occurs when the model is particularly prestigious or successful and is called prestige bias. Previous research on how context affects music appreciation has mainly focused on frequency-based bias by studying the effect of popularity on a song s success. However, very little is known about the effects of other context factors like prestige. Does music appreciation for instance change when songs are linked to a prestigious music magazine? And if so, are certain people more susceptible for this prestige bias than others? The aim of this essay is to explore these questions by examining the relationship between prestige and music appreciation for different types of music listeners. Since it is assumed that people who know little about music or so-called casual or indifferent music listeners are more affected by prestige than people with great knowledge or so-called music savants or enthusiasts, who are more confident about their own judgments. This last type, in contrast, is assumed to appreciate music based on the actual content or intrinsic quality of a song, which is referred to as content bias. The key research questions of this paper will thus be: (1) Do casual/indifferent music listeners appreciate music prestige-biased? (2) Do music savants/enthusiasts appreciate music content-biased? This paper is divided into four chapters: the first chapter will give an in-depth review of all the relevant literature that will draw the background against which the study will take place. The second chapter will explain how prestige bias and content bias are measured and will motivate the two specific measures used in this study. The third chapter will give an overview of the methods and results of the experiment, including the experimental setup and procedures, along with an extensive statistical analysis of the results. The fourth and final chapter will discuss the experiment s findings, both predicted and unpredicted, followed by the limitations of the study and corresponding solutions for future research. 1

6 1 Literature review Music is an art form that has been around for ages, and is practically innate to our existence. It is the soundtrack of our lives, as late American television host Dick Clark famously said, probably because it has the ability to tap into our emotions (Juslin & Sloboda, 2001). It follows logically that musical taste is primarily determined by our personality and to a lesser extent by the mood we re in (Vuoskoski & Eerola, 2011). In the past, a great number of studies have been conducted uncovering all sorts of relationships between certain personality traits and what type of music we prefer (e.g., Dollinger, 1993; Rentfrow & Gosling, 2003). For instance, energetic and extraverted people are shown to prefer happy, upbeat and conventional music, whereas imaginative and intellectually curious people prefer artistic and intricate music (North & Hargreaves, 1999; Rentfrow & Gosling, 2003). 1.1 A good song sells itself, or does it? Nevertheless, since music makes up such a large part of human culture, it would be shortsighted to consider musical taste only on an individual level, and not to include the social aspect. After all, people often define and convey their identity with music, and music plays an important role in interpersonal perception when, for instance, meeting strangers (North & Hargreaves, 1999; Rentfrow & Gosling, 2006). When it comes to popular music, one could assume that whether a certain song becomes successful or not, is determined by the sum of independent individual decisions based on personal musical tastes. In other words, one could assume that a song s success would be a direct translation of its intrinsic qualities. According to an economic behavioral model of Rosen (1981) this deterministic mapping between quality and success exists, yet it is bad predictor for a song s success due to its convex relationship (i.e., small differences in quality become magnified in larger differences in success). More so, Adler (1985) argued that no such relationship exists, and that a song s success relies on sheer luck. Consider for instance a group of a twenty people, who all go to a local music shop in search of a new album to broaden their musical horizons. There are ten new albums on display, that are all perceived to have had been made by equally talented artists. Let s assume that through luck one album gets bought more frequently than the other nine. Afterwards, the group meets and discusses their freshly purchased albums, but since one album has been bought more often than the others, it gets discussed the most, making it the most popular album at the meeting. As it is easier to find a discussing partner when more people are familiar with the same album, upon the next meeting, more people will also have bought the album, again increasing its popularity, which, if repeated a number of times, results in a snowball effect. This example is based on Adler s (1985) economic behavioral model that explains why only a limited number of artists/songs can be successful at one given time, independent from talent or quality, also called the superstar effect. As demonstrated by the example, an artist s success is the result of a small initial advantage that is based on luck (later on we will see that there are other things involved as well), which will snowball into popularity because people prefer to talk about the same artists/songs. This model makes use of Stigler and Becker s (1977) assumption that in a music market every consumer builds 2

7 consumption capital. Whether it is through the reading of a magazine, or through discussions with friends, consumers will always aim to increase their knowledge (i.e., consumption capital), because this results in greater enjoyment of each following encounter with the art and the artist (Stigler & Becker, 1977). Consequently, consumers will feel the need to listen to the same popular music others do, regardless of talent and quality The importance of social influence Elaborating on this model, Salganik, Dodds and Watts (2006) put the theory to the test by setting up an online experimental study. The goal of their research was to test the (un)predictability of a song s success in an artificial music market. The participants were shown previously unknown songs that they had to rate and that could be downloaded if requested. In the first condition, participants were tested independently from one another, whereas in the second condition, which was tested over different independent subgroups, participants were informed of previous participants choices (i.e., the number of downloads performed). Not only did the unpredictability of a song s success increase, when comparing the two conditions, but also the inequality, which confirms the previously mentioned superstar effect, as the best songs rarely did poorly and the worst rarely did well, but any other result was possible (Salganik et al, 2006, p. 1). In a follow-up study, Salganik and Watts (2008) went even further by giving the participants false information on previous participants downloads, namely by completely inverting the perceived success of a song (i.e., the most successful songs became the most unsuccessful songs). The results suggested a self-fulfilling prophecy as the (falsely) perceived popularity became real over time; yet the market is not entirely self-fulfilling as the very best songs regained popularity in the long run. 1.2 An evolutionary point of view In a divergent research approach, music (or art) is studied from an evolutionary point of view. In this approach, people s art preferences are explained through the use of dualinheritance theory (a.k.a. gene-culture coevolution), a theory that suggests that human behavior and evolution are not solely the result of genetics, but also of cultural transmission of information (Laland & Brown, 2002). The two major rules inherent to this theory are content bias and context bias Content bias Content biases are the direct result of one s genetic preferences or culturally acquired mental representations, and can thus be called direct biases (Boyd & Richerson, 1985). Consider for instance the previous example of twenty people buying an album. Suppose that they do a blind test by listening to different excerpts from all ten albums, so that they are not being influenced by any external factors (e.g., title, album cover, artist s name, etc.); this way they have to rely solely on the actual content of an album. After having listened to all excerpts, each individual chooses the album that they thought was the most pleasing aesthetically, based on each individual s unique personal taste. This example is a simple demonstration of content bias, which refers to the fact that humans will prefer 3

8 the content they implicitly like best, based on the interaction of its characteristics with our human evolved psychology (Boyd & Richerson, 1985) Context bias Context biases, in contrast, result from the information embodied in both who holds ideas and how common the ideas are, without considering the actual content. These two major categories are called model-based biases (i.e., who is modeling the behavior), and frequency-dependent biases (i.e., commonality or rarity of a behavior). In the previous example, there were no external factors at play when buying the album. However, this is obviously not the case in real life, where people will usually assess both content and context aspects of the available albums. Content aspects referring to the actual aesthetic qualities (e.g., tempo, mode, timbre), and context cues referring to the kind of people buying the album and how popular (or exclusive) it is. However, as previously mentioned, people do not always assess both aspects when buying an album. In fact, in some cases context bias can outweigh content bias; for example, when someone likes a song merely because it is popular (Henrich & Boyd, 2002). In a particular case of context bias, when the model displaying the behavior enjoys great social influence, this is called prestige bias. Though, it should not be mistaken with dominance (i.e., being influential because of authority), since prestige is usually acquired through someone s past achievements, and is freely conferred instead of enforced (Henrich & Gil-White, 2001). 1.3 Different types of music listeners The diffusion of innovation What is lacking in this approach is the fact that the adoption of a new behavior can differ from one learner to another. In what can now be considered a classic work, Rogers (1962) explains how an innovation spreads through a specific population or social system. Although his work is conceived from a sociological approach, it has since been applied in many different domains, but most commonly in marketing (e.g., launching a new product). Adaption to a new idea, behavior or product does not happen simultaneously in different people; rather it s a process whereby some people are more apt to adopt than others. Therefore, Rogers (1962) distinguishes between five adopter categories with different speeds of adoption: innovators, early adopters, early majority, late majority, and laggards. Especially the second category (i.e., early adopters) plays an important role in the diffusion of an innovation as they have the greatest degree of opinion leadership of all the adopters. The early adopter is respected by his or her peers, and is the embodiment of successful, discrete use of new ideas. They know that to continue to earn this esteem, he or she must make judicious innovation-decisions. As a role model or a trendsetter, they can catalyze the diffusion of an innovation, since possible adopters (especially the early majority) look to them for advice and information about the innovation. This is parallel to the aforementioned prestige bias, where the model that is being copied (or the early adopter) enjoys great social influence because of prestige. In the previous example, early adopters would be the people buying an album just after it has been 4

9 released, so they can be among the first to hear it (i.e., frequency-dependent bias of exclusivity). They feel a thrill (and status) to being the first to explore, adopt and adapt. To spread their opinions, they will write about new album releases on blogs (e.g., Pitchfork) or magazines (e.g., Rolling Stone), which in turn have great influence on potential future buyers Savant vs. indifferent music listeners In the light of music customization services, a 2003 study by the East Midland Allied Press (EMAP) suggested that there are four categories of music listeners: Savants (9%) for whom everything in life seems to be tied up with music. Enthusiasts (16%) music is a key part of life but is balanced by other interests. Casuals (26%) music plays a welcome role, but other things are far more important. Indifferents (49%) would not lose much sleep if music ceased to exist. Savants are, in Stigler and Becker s (1977) assumption that in a music market every consumer builds consumption capital, people who have a large consumption capital and invest a lot of time expanding this capital by searching for new music (e.g., reading magazines/blogs, creating playlists, etc.). They feel the need to keep track of everything that s published so they do not miss out on anything. In Rogers (1962) theoretical framework, they would be considered innovators/early adopters as they are among the very first to explore and adopt new music. Enthusiasts have a smaller consumption capital, and will invest less time searching for new music in comparison to savants. Though, they can still be considered part of the early adopters. Casuals, on the other hand, have an even smaller consumption capital and do not invest any time actively searching for new music themselves, but will instead rely on music listening services (e.g., itunes, Spotify, etc.) that provide personalized playlists and the ability to shuffle new music. In Rogers (1962) theory, they would be considered early majority, since they will only adopt a new song after seeing it used successfully by either innovators (or savants) or early adopters (or enthusiasts). Lastly, indifferents hardly have any consumption capital when it comes to music, and therefore do not benefit from searching for new music. They may only experience music on the radio or television, and can thus be considered late majority/laggards, being the last ones to adopt a new song when it has already been around for a substantial amount of time Combining Roger s theory with dual-inheritance theory In conjunction with the previously discussed dual-inheritance theory, Rogers theory of innovation and these four corresponding categories provide an interesting paradigm to test if and when prestige bias occurs when adopting new music. Seeing that prestige bias is predominantly employed by inexperienced individuals, as it is assumed to be an effective strategy to efficiently acquire better-than-average information in a domain (Henrich & Gil-White, 2001), we can claim that casual and indifferent music listeners rely more on context when adopting new music than music savants and enthusiasts. Music savants and enthusiasts, on the other hand, being part of the early adopters, have a great opinion leadership resulting in a content-biased approach when adopting new music. 5

10 1.4 Hypothesis The first aim of this paper is to verify whether casual/indifferent music listeners employ prestige bias when evaluating new music. This means that content biases may be overridden, as it is not affected by human evolved psychology. The second aim is to verify whether music savants/enthusiasts, in contrast, do not override their content biases by employing prestige bias, and thus their music preferences closely match genetically evolved preferences. In short, the key research questions are as follows: (1) Do casual/indifferent music listeners appreciate music prestige-biased? (2) Do music savants/enthusiasts appreciate music content-biased? 6

11 2 Present Research Measuring a content bias for something as subjective as music is quite challenging considering that individual music preferences are so specific to one s personality. Therefore, it is required to look for a more objective way to measure the intrinsic quality of music. Instead of controlling the actual musical content (e.g., tempo, mode, timbre, etc.), one could also manipulate the volume (i.e., loudness) of a song, since this has been proven to affect the perceived quality (Croghart, Arehart & Kates, 2012). 2.1 Compression and perceived quality Ever since the 1940s, when jukeboxes were first introduced to the general public, record engineers have been trying to make their record stand out by making it sound loud in relation to others. The message here being the louder, the better. Yet it was not until the early 90s with the introduction of a new digital format (i.e., Compact Disks) that this trend would result in the infamous loudness war. This somewhat dramatic sounding term is used to describe the arms race that has been taking place over the last 25 years of steadily increasing the loudness of commercial music. But since digital music has a maximum available peak level that cannot be exceeded (i.e., 0 dbfs), mastering engineers have to resort to dynamic-range compression (DRC) to make their song louder. In short, DRC reduces the difference between the quietest and loudest parts of an audio signal (i.e. dynamic range) by reducing loud sounds above a certain threshold while not affecting quiet sounds. There is, however, a major downside to the use of DRC especially when pushed to the extreme, as it has been demonstrated that heavy degrees of DRC harmfully affect clarity and sound quality of music, and makes everything sound cluttered up and less exciting (Katz, 2007). This is the reason why in the last decade many members of the music industry including consumers have become increasingly opposed to this loudness race. Nevertheless, it is innate to humans to prefer a louder signal when comparing two songs ceteris paribus (i.e., when the same song is used, and no DRC is applied that can degrade the audio quality) (Maempel & Gawlick, 2009). This is, in part, caused on a neurological level by a tiny primitive organ in the inner ear called the sacculus, which reacts to low frequencies over 90dB between 100hz to 300hz and has neural connections to those parts of the brain that are responsive to all forms of pleasure (Todd & Cody, 2000; Todd, 2001). Furthermore, as we increase the volume of a song, we become functionally deaf to everything but the music, which intensifies the listening experience (Blesser, 2008). Yet, this relies on the assumption that all other factors remain equal, and that only the listening volume is changed. When taking into account the effects on quality caused by DRC, it appears that an increase in loudness (and a subsequent loss of dynamic range) would initially elevate perceived audio quality up until the point that the loss of dynamic range (and quality) starts to outweigh the benefits of increased loudness, so that the relationship between quality and loudness is expressed as an inverted-u function (Croghart, Arehart & Kates, 2012). 7

12 2.2 Prestige cue To measure prestige bias, additional information has to be given on the context of the song that acts as a proxy of prestigious individuals. A prestigious online music blog could be used for this purpose, since they have great opinion leadership as they can be considered early adopters by analogy with Rogers (1962) theory of innovation. In parallel with Adler s (1985) previously mentioned theory that a song s disproportionate success to other songs is the result of a snowball effect triggered by a small initial advantage, Dewan and Ramprasad (2009) researched the existence of a feedback loop between blog buzz and album sales. Since more sales implies that a larger percentage of the population is aware of the music, it results in more buzz, and inversely, more blog buzz means that more people are aware of the music, which in turn increases its demand. Their findings confirmed this theory as blog buzz was proven to positively affect album sales, especially when it comes to independently released music. This is because the amount of independently released albums per year is much higher than the amount of albums released by major labels, rendering it impossible for an individual to keep track of all new releases. This advocates the need for a filter, in the form of frequency-dependent biases (i.e., how popular is the song) and model-based biases (i.e., who likes the song). 2.3 Stimuli and manipulations A total of ten stimuli were used, consisting of six songs randomly selected from the Best New Tracks -list of Pitchfork s 1 website, which were all released within two months prior to the study, three unknown non-published demo recordings and one song made by a relatively unknown Belgian band called Bibles (Appendix 1). This made sure that familiarity effects would be minimized, since the songs were either recently released or had not been released yet. From each song, a fifteen second fragment was selected (usually a part of the chorus either preceded or followed by an instrumental or the verse) in order to paint a clear picture of the song without having to listen to it in its entirety. This was partly motivated by the aim of keeping the study short and interesting, but also to conceal the fact that the demo songs were not real. Furthermore, to find out whether loudness influenced the perceived quality of a song, all songs were equalized in terms of loudness (= the uncompressed condition) and were then increased in loudness by adding DRC (= the compressed condition). First, a professional loudness metering plugin by software manufacturer Waves was used in Logic Pro X, a Digital Audio Workstation by Apple, to measure the long-term loudness (i.e., average loudness) of every individual song. The measured long-term loudness 1 Pitchfork has been around since the early days of the Internet, and has proven to be an influential factor in making and breaking artists by what is called the Pitchfork Effect. This term has been coined by Dave Itzkoff (2006), a reporter of the New York Times, to describe the impact, both constructive and destructive, Pitchfork can have with their album reviews. While it started out small as an online fanzine written by the magazine s founder Ryan Schneider, they ve grown out to be the most successful independent-focused music publication online (Itzkoff, 2006). 8

13 varied between -14 and -10 LUFS 2 with an average of -12 LUFS, which is fairly compressed but normal to today s standards. Next, the loudness was made equal for all songs to -12 LUFS either by making the quieter fragments louder with a level maximizing plugin called L3 Ultramaximizer by software manufacturer Waves or by making the louder fragments quieter with a simple gain adjustment. These fragments were used in the uncompressed condition. For the compressed condition, however, an additional step involved the use of aforementioned level maximizing plugin with a threshold of -7.5 db to increase the long-term loudness to -6 LUFS, which is considered very loud and overly compressed, but is often the case on commercial radio. To hide the fact that prestige bias was being assessed, the design of this research was constructed in such a way that the participants would be randomly added to either the neutral or the pitchfork condition with an even presentation of both conditions, and thus, testing prestige between subjects. Those in the neutral condition were merely told in the introduction screen that they were going to have to rate some songs, while those in the prestige condition were told in the introduction screen that the songs they have to rate were selected from the Best New Tracks -list of Pitchfork s website. In the supposition that not all participants would know Pitchfork, the introduction gave some background information (i.e., that it is the most popular independent-focused music publication online and has a lot of influence). In contrast, content bias was tested within subjects by randomly displaying either the uncompressed or the compressed version of a music fragment, with the order of fragments being randomized as well. However, the first fragment all participants heard was predetermined and was uncompressed. This gave the participants the opportunity to adjust the level for comfortable listening, and made sure that the following compressed, thus louder, fragments would have the predicted impact on perceived quality. 2 Loudness units relative to Full Scale (or LUFS) is a term introduced in 2014 by the European Broadcasting Union to serve as an international broadcast standards in order to address the problem of loudness inconsistencies between different broadcast channels (EBU, 2014). It is an absolute measure of loudness, meaning that is has an absolute reference, and one unit of LUFS is equal to one db. 9

14 3 Methods and results 3.1 Participants A total of one hundred and twenty-nine people participated in this study. A number of ways were used to recruit participants for this study, primarily through the use of social media such as Facebook and twitter, but also by hanging up posters and handing out flyers in record stores (Bilbo and Fnac), a concert hall (Het Depot) and a music store (Leo Caerts), and by promoting it on Poppunt. Nineteen participants, who did not finish the study, were excluded from the analysis. The resulting 110 participants, of which 73 were male and 37 were female, ranged in age from 15 to 62 (M=29.42, SD=10.55). 3.2 Procedures and measures To verify the hypotheses, an online experimental study was set up in Qualtrics that consisted of four blocks in a fixed order: an introduction, the fragments, a music questionnaire and some questions about demographics. After opening the survey, participants were first presented to an introduction screen stating that the study is a music preference study that will ask for feedback on ten songs and was going to take about three to five minutes. Next, one half of the participants, who were randomly assigned to the Pitchfork condition, were given additional information on an extra screen stating that the songs were all selected from Pitchfork s Best New Tracks -list of website, including a short description of Pitchfork, whereas the other half, who were randomly assigned to the neutral condition, were not given any extra information (Appendix 2). Figure 1: The interface for participants in the prestige condition After the introduction screens, the first fragment appeared in the form of a small media player. In the Pitchfork condition, there was a small logo of Pitchfork displayed next to the media player, while in the neutral condition, there was no such logo displayed (Figure 1). Below each fragment, it was stated I appreciate this song followed by a seven-point 10

15 Likert scale ranging from not at all (=1) to very much (=7). Only after the participant had completed all statements could he or she move on to the next song, and so on. The order of the songs was randomized, and whether the song was compressed or not was randomized too, resulting in a practically unique survey for each participant. Subsequently, all participants were presented to a short questionnaire, consisting of one set of 6 statements such as music plays a key role in my social life that were rated on a seven-point Likert scale, one set of 6 questions asking about the frequency of certain music related activities such as how often do you attend live musical performances?, which were also rated on a seven-point Likert scale, and an additional question asking about the usual method of learning about new music. These questions were carefully selected to test the participant s subjective involvement with music. Finally, the participants were asked to give additional information on their gender and age, after which the survey was finished (Appendix 3). 3.3 Results Music expertise In order to obtain a scale that represents one involvement or expertise in music, the internal consistency of the questionnaire was tested through the use of Cronbach s Alpha. When including all 18 questions of the questionnaire, Cronbach s Alpha was 0.839, which denotes a good internal consistency and allows for summing up the individual scales. However, to further improve internal consistency, six questions were discarded, as statistical reliability analysis had shown that Cronbach s Alpha would improve when removed (Cronbach s Alpha = 0.875). The scales of the remaining 12 questions were totaled up, resulting in a continuous scale ranging from 18 to 63 (M=46.03, SD=9.08, Median=47) that represents the music expertise score. 3 Nevertheless, since the goal of the questionnaire was to divide the participants into two groups, namely the savant/enthusiast vs. the casual/indifferent music listeners, a continuous variable will not suffice. A median split could be executed to divide the continuous scale into two groups (a group with a score lower than 47 and a group with a score higher or equal to 47), but this method 4 should be avoided, as it would assume that a value just above the median is considered the same as values way at the end. Therefore, it is desirable to find a categorical variable that has actual relevance for this study. When analyzing the results, it became apparent that the question I actively seek out new music myself was checked by half of the participants, namely 55, and since this is one of the main characteristics of a savant/enthusiastic music listener, this served as a 3 When music expertise score was used as a continuous independent covariate in a Generalized Linear model, no significant effects were found. However, when the score was interpreted as an ordinal variable, we found a significant effect of prestige (F(1,57)=19.01, P <.01), of compression (F(1,57)=6.02, P =.017). 4 Statistical analysis using a dummy variable based on a median split also resulted in significant results for prestige (F(1,106)=8.60, P <.01) and compression (F(1,106)=4.41, P =.038). 11

16 dummy variable for the two groups. An independent-sample t test showed that the previously calculated music expertise score was significantly higher among the group who actively seeks out new music (M=51.47, SD=6.22), than those who do not (M=40.48, SD=8.17 and F(1,107)=7.91, p <.01), which further backed up the legitimacy of this dummy Fragment score In order to test the within-subjects factor of compression, the scores of the different songs have to be grouped by compression for each participant. The first fragment was excluded because it served as a test for participants to adjust the volume for comfortable listening, leaving nine fragments to consider. A Cronbach s Alpha of suggests an acceptable internal consistency, allowing the summation of the individual scores for each condition (uncompressed and compressed). Since the compression condition was randomized for each individual fragment and the number of fragments was uneven, some participants were more often exposed to compressed songs than uncompressed songs, and viceversa, which resulted in the use of means to calculate the individual score for the two conditions Main effects and moderations Generalized Linear Mixed Models (GLMMs) with compression (uncompressed vs. compressed) as a within-subjects factor and prestige (neutral vs. Pitchfork) and music expertise (as a dummy variable reflecting enthusiasts vs. casual music listeners) as between-subjects factors indicated an overall effect of prestige (F(1,106) = 9.58, p <.01), expertise (F(1,106) = 9.82, p <.01), and compression (F(1,106) = 5.82, p <.05) (Table 1). Table 1: The output from the Generalized Linear Mixed Model SOURCE F DF1 DF2 SIG. Corrected Model 4, ,000 Prestige 9, ,003 Expertise 9, ,002 Compression 5, ,018 Prestige * Listener Type 5, ,019 Prestige * Compression 0, ,870 Expertise * Compression 0, ,574 Prestige * Listener Type * Compression 0, ,889 Regarding the overall prestige effect, participants appreciated the songs more in the Pitchfork condition (M=3.78, SD=0.12) compared to the neutral condition (M=3.27, SD=0.10). The overall effect of expertise is due to the fact that savants/enthusiasts (M=3.3, SD=0.11) appreciated the songs less than casual/indifferents (M=3.75, SD=0.11). Lastly, the overall effect of compression is caused by the fact that the participants preferred the uncompressed version (M=3.69, SD=0.11) to the compressed version (M=3.45, SD=0.11) (see figure 1). The GLMMs further revealed that music 12

17 expertise moderated prestige (F(1,210) = 4.10, p =.04), however, it did not moderate compression (F(1,210) = 0.01, p =.94), contrary to the predictions (Figure 2). 3,8 3,6 3,4 3,2 3 Normal Compressed Casual Enthusiast Neutral Pitchfork Compression Music listener type Prestige Figure 2: Significant main effects on music appreciation Hypothesis 1 predicted that music appreciation would be moderated by prestige for casual and indifferent music listeners and not for music savants and enthusiasts. Simple contrast tests confirmed this hypothesis, by revealing that the moderating effect of music expertise on prestige was caused by the fact that music casuals appreciated the songs in the prestige condition (M=4.23, SD =0.16) more than those in the neutral condition (M =3.39, SD = 0.15; F(1,106) = 14.93, p <.01, whereas music enthusiasts appreciation was not influenced by prestige (M=3.39, SD=0.17 vs. M=3.28, SD=0.14; F(1,106) =0.26, p=.613). The findings are displayed in Figure ,5 3 Non-specified Pitchfork 2,5 2 Casuals/Indifferents Savants/Enthusiasts Figure 3: The effect of prestige on music appreciation for the different listener types 5 Adding age and gender into the model did not alter the results, nor did it have any significant effect or moderation of interest to our purpose (F(1,104)=0.05, p=.819; F(1,104)=0.06, p=.801) 13

18 3.3.4 Effects on individual songs To further explore the two hypotheses, General Linear Models (GLMs) were employed to analyze the songs independently, with compression (uncompressed vs. compressed), prestige (neutral vs. Pitchfork) and music expertise (enthusiastic vs. casual music listeners) as between-subjects factors. When Cronbach s Alpha is lower than 0.7, internal consistency is questionable, and, however it is not to the case here (Cronbach s Alpha is 0.703), it is recommended to test the songs on an individual level as well. Since each individual song could have been in either a compressed or an uncompressed state for each participant, this allows for a different approach than the previous model, where compression was a within-subjects factor. Table 2 shows all the significant main effects and moderations (P <.05) for each individual song, except for three songs that had no significant effects. Four of the songs appreciation scores showed a positive relationship with prestige, of which two had an additional negative relationship with expertise. Moderating effects were found for two songs, both related to compression. One song, Damn That Valley, showed a significant moderating effect of music expertise on compression caused by the fact that casuals appreciated the uncompressed state more than music enthusiasts (M=3.58, SD=0.28 vs. M=2.68, SD=0.31; F(1,106)=4.55, P <.05), which is the opposite of what hypothesis 2 predicted. Another song, The Lost Drum Beat, showed a highly significant moderating effect of compression on prestige caused by the fact that the compressed version was appreciated more in the prestige condition than in the neutral condition (M=4.27, SD=0.42 vs. M=1.97, SD=0.30; F(1,106)=20.48, P <.001). This effect, however, was not predicted by either one of the hypotheses. Table 2: The main effects and moderations on individual songs. SORT ANNIE PRESTIGE (P) (+) COMPRES- SION (C) LISTENER TYPE (LT) (P) * (C) (P) * (LT) (LT) * (C) THREE SIDES OF AUDREY THE LOST DRUM BEAT DAMN THAT VALLEY SHOW YOU CASIO VOICES WHERE DID YOU GO? GROW WINGDANCE (+) (+) (+) (-) (+) (-) (+) (-) (-) (+) 14

19 4 Discussion Previous theoretical and empirical research has suggested that social influence plays an important role when adopting new behavior, along with our own genetically evolved preferences (Henrich & McElreath, 2003). Music is no exception to this rule; since whether we like a song or not is usually the result of both personal taste and social influence like the popularity of a song or the kind of people who like the song (Salganik et al, 2006; Henrich & Boyd, 2002). As for the latter, when the influential source is prestigious (e.g., music magazines, playlists, etc.), we call this prestige bias, which is proven to be an effective strategy for laypeople to efficiently acquire better-than-average information in a domain (Henrich & Gil-White, 2001). Consequently, the purpose of this study was to examine whether casual and indifferent music listeners (i.e., people with a limited music capital) would rely more on prestige when evaluating new songs than music savants and enthusiasts would (i.e., people with a large music capital). Additionally, I also examined that, in doing so, casual and indifferent music listeners would bypass their own genetically evolved personal preferences, as content bias is overridden by prestige bias (Henrich & Boyd, 2002). Music savants and enthusiasts, in contrast, would appreciate music content-biased and not prestige-biased. 4.1 The main findings The analysis suggests three main findings. Firstly, casual and indifferent music listeners give higher ratings when they are told that the songs were selected from the Best New Tracks -list of Pitchfork. In contrast, music savants and enthusiasts are not affected by this information, and give constant ratings in both conditions. This supports the prediction that casual and indifferent music listeners appreciate music prestige-biased, and extends our understanding of how music appreciation is susceptible to prestige. Secondly, songs that have been overly compressed are rated lower than normally compressed songs, which is the expected result of overcompression on perceived quality (Croghart, Arehart & Kates, 2012). However, there is no significant difference for both music listener types, which disproves the prediction that content bias is overridden when prestige bias comes into play. Nevertheless, this finding should not be generalized for the entire population, since there might be other factors at play that have been overlooked (see limitations and further research). Additionally, it is worth considering that the lack of prestige bias for music savants and enthusiasts would automatically suggest that they appreciate music content biased, although there is no direct evidence for this. And thirdly, the overall appreciation differs for the two listener types in that music savants/enthusiasts give lower ratings than casual/indifferent music listeners, on average. This effect was not predicted at all, and provides food for thought, as it seems rather contradictory that people who are presumably more involved with music give new songs significantly lower ratings than people who are less involved with music. It is more logical to assume that their general interest in music would instead positively affect their music appreciation. A first possible explanation for this finding is that music savants and enthusiasts have developed higher standards than casual and indifferent music listeners have, as they have a greater music capital, which could render them more selective and critical when exposed to new music. 15

20 Secondly, how we usually discover new music could also have an impact on the appreciation of unfamiliar music. While a music savant happily spends hours a day scrolling through music blogs in search of new music, the indifferent music listener may not even make the simplest of efforts like turning the radio on. Consequently, casual and indifferent music listeners discover music passively by listening to the radio or shuffling through itunes, whereas music savants and enthusiasts actively seek out new music themselves. This could explain why music savants and enthusiasts appreciate new music less because they are used to assess music, filtering out what s good and what s bad, when casual and indifferent music listeners, on the other hand, are used to lean back and just enjoy what there given, which would render them less critical. The third possible explanation is derived from the two previous explanations. Since I categorized the participant sample by asking if one actively seeks out new music, there could be other hidden personality traits in play that also have an effect on music appreciation. In the domain of music psychology, research has shown that openness to experience positively correlates to the cognitive use and complexity of music (Chamorro- Premuzic & Furnham, 2007; Dunn, de Ruyter & Bouwhuis, 2011). This could be the case for music savants and enthusiasts, as they are more open to discover new music. The lower appreciation could then be explained by their need to be cognitively challenged, making them more critical. Other research has shown that agreeable individuals prefer conventional music and display an intense emotional response to unfamiliar music (Langmeyer, Guglhör-Rudan & Tarnai, 2012; Ladinig & Schellenberg, 2012). This could be the case for casual and indifferent music listeners, as their passive approach for discovering new music makes them more agreeable. The higher appreciation could then be explained by their intense emotional response when hearing new music, making them less critical. 4.2 Limitations and future research Considerations for compression Although the use of compression appears to be a valid way to influence the perceived quality of music, as demonstrated by the results, there are two pitfalls to this approach Controlled environment First, since the experiment did not take place in a controlled environment like a laboratory, it allowed for manually adjustment of the listening volume. This means that when the participant was exposed to a louder song, the positive effect on perceived quality would disappear with the first spontaneous loudness correction (Maempel & Gawlick, 2009). Additionally, again, since there was no controlled environment, the quality of the audio system for each participant was different. Seeing that the positive effect of loudness on participant s appreciation has been linked with low frequencies, the effect would be non-existent when the participant did not have the adequate audio system to play bass (Todd & Cody, 2000). Therefore, it is desirable to control for these flaws in future research by conducting a laboratory-based experiment, so that there are no external factors in play. Alternatively, the obligated use of headphones with the request not to temper with the listening volume could have been a solution, however, this 16

21 is also susceptible for differences in quality, and participants might be more aware that they re tested Other ways to measure content bias A second pitfall to the use of compression is that, although it affects perceived quality, it does this solely on a superficial level; a well-written song doesn t all of the sudden become bad. The actual content (e.g., tempo, mode, lyrics) remains the primary factor when comparing different songs (Maempel & Gawlick, 2009). Future research could tackle this problem by pretesting a set of songs to gain insight in the average appreciating of each individual song. After that, a selection of the worst songs and best songs could be used to test for content bias. A previously mentioned study by Salganik and Watts (2008) used this method to research the effect of self-fulfilling prophecy in an artificial music market. In this study a large number participants were presented with forty-eight songs by unknown bands that they had to rate and could be downloaded when they really liked it. In addition, each song included the download count of previous participants. However, the download count had been falsified by swapping the download count of the most popular songs with the download count of the least popular songs, as had previously been tested in the neutral condition. The results partly suggested a self-fulfilling prophecy, as the falsely perceived popular songs remained popular, but in addition some of the falsely perceived unpopular songs regained popularity. This suggests that not all the participants appreciate music using frequency-based biased (i.e., how popular a song is), and that some participants instead rate the songs content-biased. In conjunction with the research conducted in this paper, where participants were subdivided according to their listening habits, this could provide an interesting basis for future research Considerations for prestige The findings of this research have extended our knowledge of prestige bias by recognizing different effects for different types of listeners. Casual and indifferent music listeners are shown to be more susceptible for prestige bias than enthusiastic and indifferent music listeners within the independent music genre. However, this paper only scratched the surface of the different possibilities in which prestige can influence our music appreciation. A person s susceptibility to prestige bias, for instance, may also stem from different individual factors, what we find to be prestigious can also differ from one person to another, and different genres can result in different effects Are there other types of listeners? Regarding the issue of who is more susceptible for prestige bias, it is possible to hypothesize that prestige bias is more likely to occur for young adolescent people, since they are still developing their musical taste (Levitin, 2011). In parallel with the critical period of language development, it is suggested that fourteen is the magic age for developing our musical taste, in a period that starts at 12 and ends at 16. Coincidentally, social influence plays an important role at this age as musical tastes are influenced and refined by what our peers listen to. Other research suggests that musical taste becomes a badge of identity when we are adolescent, favoring people who like the same music as they do (North & Hargreaves, 1999). In the research conducted for this paper, 17

22 participants were on average 29.4 years old, which means that their musical tastes have already been developed. Future research will have to be conducted to find out whether social influence, prestige bias in particular, is stronger for adolescents than for adults What is prestigious? To account for interpersonal differences in regard to what we find prestigious, further research should be undertaken to investigate if certain prestige cues have a larger impact on music appreciation than others, and how different prestige cues correspond to different types of individuals. Within the limited scope of this research, I obtained a significant prestige effect by using an online music magazine called Pitchfork. Yet in future investigations, it might be possible to use a different prestige cues like other more mainstream music magazines, and compare the relative effects on music appreciation. Or instead of magazines, all sorts of possible prestige cues could be used like an influential model, or a producer, or even radio stations How does prestige affect other genres? By using indie songs, I exploited the fact that prestige influences the success of independently released music more than the success of commercial music. As previously discussed, this is because the amount of independently released music is much higher than commercially released music, which raises the need for a filter when looking for new music (Dewan & Ramprasad, 2009). Subsequently, the question rises whether the findings of this research can be generalized for different types of music. The unpredictability of a song s success seems to be higher for independently released songs than for commercial songs, because of the greater amount of songs released every year. However, mainstream music markets are more susceptible for inequality of a song s success, resulting in a market where only a limited amount so-called superstars have all of the success, whereas other artists struggle to get some attention at all (Adler, 1985; Salganik et al., 2006). Future studies on the current topic for mainstream music, and other genres are therefore recommended. 18

23 Conclusion The purpose of this paper was to determine the effect of prestige on music appreciation for two types of music listeners, namely music savants/enthusiasts and casual/indifferent music listeners. It was predicted that the former appreciates music content biased, meaning that they rely solely on the intrinsic quality, whereas the latter was predicted to appreciate music prestige biased, meaning that they would appreciate music more if there is prestige involved. This also implied that content bias, which was measured by varying the amount of compression applied to the songs within subjects, would be overridden by the occurrence of prestige bias for casual/indifferent music listeners. We examined these predictions by conducting an online experiment with 110 participants. The findings of this experiment generally support the predictions, as it was shown that casual and indifferent music listeners indeed gave higher ratings when the songs were accompanied by a prestige cue (i.e., an online music magazine called Pitchfork), while this did not influence music savants and enthusiasts. The amount of compression had an overall negative impact on music appreciation, as predicted, but it did not moderate music appreciation for the two listener types, in contrast with the predictions. Unexpectedly, the overall ratings given by music savants and enthusiasts were significantly lower than the ratings given by casual/indifferent music listeners, which is presumably the result of higher standards carried by the savants and enthusiasts. This paper only scratches the surface of the different possibilities for which prestige can influence music appreciation. More research should be undertaken to explore whether the findings of this paper are generalizable for other genres, be it more mainstream or niche, or for different prestige cues like mainstream magazines or a music festival, or even multiple prestige cues to compare the different effects on music appreciation. Additionally, to have more control of measuring content bias, laboratory based experiments are desirable as they eliminate unwanted effects that could have occurred by poor execution of the experiment. Future research could also explore other ways to measure content bias by for instance using the worst and best songs of a pretested set of songs. By combining elements from cultural evolution theory with the diffusion of innovation theory, this study provides an interesting framework for future research, as it distinguishes between two distinct listener types that employ different strategies when rating unfamiliar music with an added prestige cue. Casual and indifferent music listeners like music more when there is prestige involved, while music savants and enthusiasts only seem to care about the quality of a music. So to end where we started, whether good music sells itself becomes a matter of what type of music listener listen you are. 19

24 Appendices Appendix 1: Songs used in the survey with their original loudness 6 1. Torres - Cowboy Guilt 2. Neon Indian - Annie 3. Nicolas Jaar - Three Sides of Audrey 4. Mikael Seifu - The Lost Drum Beat 5. U.S. Girls - Damn That Valley 6. ZooGang - Show You 7. Dektro - Casio Voices 8. Demo - Where Did you Go 9. Demo - Grow 10. Bibles - Wingdance Appendix 2: Introduction 7 Thanks for agreeing to participate in this Master Thesis Study. The goal of this study is to give us your feedback of 10 songs, which will be used in a music preference study. The study should take 3-5 minutes. You will be given 10 fragments (15 sec) of songs that you have to listen to, after which you will be asked to rate them. The featured songs are all picked from the "Best New Tracks" list of Pitchfork. Pitchfork is an influential online music magazine that primarily reviews independent music. With an audience of more than 240,000 readers per day, and more than 1.5 million unique visitors per month, it is the most popular independent-focused music publication online. Appendix 3: Questionnaire Q1. Given are 6 statements. Please indicate to what extent they are applicable to you. I'm interested in music Music is an important part of my life Strongly Disagree Disagree Somewhat Disagree Neutral Somewhat Agree Agree Strongly Agree o o o o o o o o o o o o o o 6 The songs written in italics were not selected from the Best New Tracks -list from Pitchfork, of which 7, 8 and 9 are home made demos and 10 is a song by Belgian band Bibles. 7 The text written in italics was only presented to participants of the prestige condition. 20

25 I'm interested in discovering new, unheard music I actively seek out new music Music plays a key role in my social life Strongly Disagree Disagree Somewhat Disagree Neutral Somewhat Agree Agree Strongly Agree o o o o o o o o o o o o o o o o o o o o o I like making music o o o o o o o Q2. Please select the right answer. Never Rarely Sometimes Often Extremely Often How often do you listen to music o o o o o How often do you purchase music in a digital format (itunes, Napster, etc.) How often do you purchase music in a physical format (CDs, records, cassettes, etc.) How often do you stream music online? (Spotify, Deezer, etc.) How often do you attend live musical performances? o o o o o o o o o o o o o o o o o o o o Q3. How often do you search for new music? Less than 2-3 Times a 2-3 Times a Once a Month Once a Week Daily Once a Month Month Week o o o o o o Q4. How do you usually learn about new music? o Online music streamers (Spotify, Deezer, Google Play, Youtube etc.) o Mainstream media (Morgen, De Standaard, Knack, etc.) o Less mainstream media (Rolling Stone, Pitchfork, Allmusic, etc.) o Word-of mouth or what's new on the radio o Recommendations from friends o I actively seek out new music myself o Other:... Q5. What s your gender? o Male o Female Q6. What year were you born? 21

26 List of figures Figure 1: The interface for participants in the prestige condition Figure 2: Significant main effects on music appreciation Figure 3: The effect of prestige on music appreciation for the different listener types

27 List of tables Table 1: The output from the Generalized Linear Mixed Model Table 2: The main effects and moderations on individual songs

28 Sources Books Boyd, R. and Richerson, P. (1985). Culture and the evolutionary process. Chicago: University of Chicago Press. Juslin, P. and Sloboda, J. (2001). Music and emotion. Oxford: Oxford University Press. Rogers, E. (1962). Diffusion of innovations. Glencoe: Free Press. Articles Adler, M. (1985). Stardom and Talent. American Economic Review, 75(1), pp Blesser, B. (2008). Aural architecture: The missing link. The Journal of the Acoustical Society of America, 124(4), p Chamorro-Premuzic, T. and Furnham, A. (2007). Personality and music: Can traits explain how people use music in everyday life?. British Journal of Psychology, 98(2), pp Croghan, N., Arehart, K. and Kates, J. (2012). Quality and loudness judgments for music subjected to compression limiting. The Journal of the Acoustical Society of America, 132(2), p Dewan, S. and Ramprasad, J. (2009). Chicken And Egg? Interplay Between Music Blog Buzz And Album Sales. PACIS 2009 Proceedings, 87. Dollinger, S. (1993). Research Note: Personality and Music Preference: Extraversion and Excitement Seeking or Openness to Experience?. Psychology of Music, 21(1), pp Dunn, P., de Ruyter, B. and Bouwhuis, D. (2011). Toward a better understanding of the relation between music preference, listening behavior, and personality. Psychology of Music, 40(4), pp Eerola, T. and Vuoskoski, J. (2010). A comparison of the discrete and dimensional models of emotion in music. Psychology of Music, 39(1), pp European Broadcasting Union (2014). R 128: Loudness Normalisation And Permitted Maximum Level Of Audio Signals. Geneva. Henrich, J. and Boyd, R. (2002). On Modeling Cognition and Culture: Why cultural evolution does not require replication of representations. Journal of Cognition and Culture, 2(2), pp Henrich, J. and Gil-White, F. (2001). The evolution of prestige: freely conferred deference as a mechanism for enhancing the benefits of cultural transmission. Evolution and Human Behavior, 22(3), pp Henrich, J. and McElreath, R. (2003). The evolution of cultural evolution. Evol. Anthropol., 12(3), pp

29 Ladinig, O. and Schellenberg, E. (2012). Liking unfamiliar music: Effects of felt emotion and individual differences. Psychology of Aesthetics, Creativity, and the Arts, 6(2), pp Laland, K. and Brown, G. (2002). Sense & Nonsense: Evolutionary Perspectives on Human Behavior. Oxford University Press, p.260. Langmeyer, A., Guglhör-Rudan, A. and Tarnai, C. (2012). What Do Music Preferences Reveal About Personality?. Journal of Individual Differences, 33(2), pp Maempel, H. and Gawlick, F. (2009). The influence of sound processing on listeners program choice in radio broadcasting. 126th Convention of the Audio Engineering Society. Munich, Germany (May 7 9, 2009), Paper #7785. North, A. and Hargreaves, D. (1999). Music and Adolescent Identity. Music Education Research, 1(1), pp Rentfrow, P. and Gosling, S. (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), pp Rentfrow, P. and Gosling, S. (2006). Message in a Ballad: The Role of Music Preferences in Interpersonal Perception. Psychological Science, 17(3), pp Rosen, S. (1981). The Economics of Superstars. The American Economic Review, 71(5), pp Salganik, M. and Watts, D. (2008). Leading the Herd Astray: An Experimental Study of Self-fulfilling Prophecies in an Artificial Cultural Market. Social Psychology Quarterly, 71(4), pp Salganik, M., Dodds, P. and Watts, D. (2006). Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market. Science, 311(5762), pp Stigler, G. and Becker, G. (1979). De gustibus non est disputandum. American Economic Review, 67(3), pp Todd, N. (2001). Evidence for a behavioral significance of saccular acoustic sensitivity in humans. The Journal of the Acoustical Society of America, 110(1), p.380. Todd, N. and Cody, F. (2000). Vestibular responses to loud dance music: A physiological basis of the rock and roll threshold. The Journal of the Acoustical Society of America, 107(1), p.496. Articles in newspapers Levetin, D. (2011) Forever Young? In Some Ways, Yes. The New York Times. [online] Available at: [Accessed 15 Aug. 2015]. Itzkoff, D. (2006). The Pitchfork Effect. Wired. [online] Available at: [Accessed 15 Aug. 2015]. 25

30 FACULTY OF BUSINESS AND ECONOMICS Naamsetraat 69 bus LEUVEN, België tel fax info@econ.kuleuven.be