On Popular Music and Media: Analyzing Changes in Compositional Practices and Music Listening Choice Behavior Using Attention Economy Principles

Home , 2005 in Swiss music, 2017 in American music, Ace of Base, Afrojack, Alesso, All 4 Love

On popular music and media: Analyzing changes in compositional practices and music listening choice behavior using attention economy principles

Dissertation

Presented in Partial Fulﬁllment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Hubert Léveillé Gauvin, M.A., B.Mus.

Graduate Program in Music

The Ohio State University

2018

Dissertation Committee:

David Huron, Advisor Eugenia Costa-Giomi Robert Bond c Copyright by

Hubert Léveillé Gauvin

2018 Abstract

This dissertation investigates popular music compositional practices and music listening choice behavior in studies that attempt to uncover whether technological changes have had an impact on songwriting techniques, and whether music listening choice behavior can be predicted by these compositional devices. The underlying theory unifying this work is the theory of attention economy: how attention can be analyzed using supply and demand principles to explain and predict information content and human behavior.

Two corpus studies were conducted to evaluate whether compositional practices in popular music have changed in the last decades in a way that is consistent with the theory of attention economy. The results suggest that, as predicted, compositional practices have changed in a way that favors attention-grabbing behavior. The strongest change observed was the near elimination of instrumental introductions between 1986 and 2015.

Building on these results, two behavioral experiments were conducted to test whether these changes are eﬀective ways to grab listeners’ attention. In the ﬁrst experiment, participants were asked to listen to randomly sampled songs in an undivided attention setting, while in the second experiment, an independent group of participants was asked to listen to a subset of the same songs in a divided attention setting. The results suggest that music listening choice behavior (i.e. how and when a

ii listener decides to skip when listening to music) is mediated by the listening context, and that compositional devices, along with musical preferences and familiarity with a song, can predict choice behavior in listeners.

iii This dissertation is dedicated to my loving and supportive wife, Vanessa.

iv Acknowledgments

I would like to thank numerous parties for the help and support with this research project. In no particular order:

David Huron, advisor on this project, for constantly giving me generous feedback and comments, and for his invaluable help with the funding process. Eugenia Costa-

Giomi, David Clampitt, Johanna Devaney, and Anna Gawboy, for making the OSU

School of Music an exciting teaching and research institution. Robert Bond, for his support and help on this project. My oﬃce mates and friends, Erin Allen, Claire

Arthur, Andrew Brinkman, SongHui Chon, Nat Condit-Schultz, Dana DeVlieger,

Niels Ch. Hansen, Kirsten Nisula, David Orvek, Lissa Reed, Lindsey Reymore, An- drea Schiavio, Nicholas Shea, and Caitlyn Trevor. Special thanks to Lindsay War- renburg for her constant help and support. Thanks to CSML alumni Joe Plazak and

Paul von Hippel. Thanks to Misti Crane for writing a really awesome press release about my research–what an exciting couple of months that was! I would also like to thank the many members of the OSU community that oﬀered useful ideas and suggestions.

v This research project has been made possible in part by the ﬁnancial support of the Fonds de recherche du Québec - Société et culture.

Finally, special thanks to my wife Vanessa for the continuous encouragement and support. I could not have done it without you.

vi Vita

2012 ...... B.Mus. Music Theory, McGill University 2015 ...... M.A.MusicTheory, McGill University 2013-2015 ...... Graduate Teaching Associate, McGill University 2015-present ...... Graduate Teaching Associate, The Ohio State University

Publications

Research Publications

Duinker, B. & Léveillé Gauvin, H. (2017). Changing content in ﬂagship music theory journals, 1979–2014. Music Theory Online, 23 (4).

Devaney, J., & Léveillé Gauvin, H. (2017). Encoding music performance data in Humdrum and MEI. International Journal on Digital Libraries, Published Online First October 23, 2017.

Léveillé Gauvin, H. (2017). Drawing listener attention in popular music: Testing ﬁve musical features arising from the theory of attention economy. Musicae Scientiae, Published Online First March 1, 2017.

Devaney, J., & Léveillé Gauvin, H. (2016). Representing and linking music performance data with score information. In B. Fields & K. Page (Eds.), Proceedings of the 3rd International Workshop on Digital Libraries for Musicology (pp. 1-8). New York: ACM ICPS.

vii Léveillé Gauvin, H., Huron, D., & Shanahan, D. (2016). On the role of semitone intervals in melodic organization: Yearning vs. baby steps. In Proceedings of the 14th International Conference on Music Perception and Cognition (pp. 727-731). San Francisco, CA.

Léveillé Gauvin, H. (2016). Changing use of seventh chords: A replication of Mauch et al. (2015). Empirical Musicology Review, 11 (1): 103-107.

Léveillé Gauvin, H. (2015). “The times they were a-changin’:” A database-driven approach to the evolution of musical syntax in popular music from the 1960s. Empirical Musicology Review, 10 (3): 215-238.

Fields of Study

Major Field: Music

viii Table of Contents

Page

Abstract ...... ii

Dedication ...... iv

Acknowledgments ...... v

Vita...... vii

List of Tables ...... xii

List of Figures ...... xiii

1. Introduction ...... 1

2. Capturing and Sustaining Attention in the Attention Economy Era . . . 4

2.1 On what basis is the music entertaining or engaging? ...... 8 2.1.1 Arousal ...... 8 2.1.2 Predictability ...... 10 2.2 How is the music structured and what role(s) does this structure serve? 11 2.2.1 AABA and Verse-Chorus Forms ...... 11 2.2.2 Instrumental Introductions ...... 12 2.3 How does the music achieve memorability? ...... 15 2.3.1 Repetition & Memory ...... 15 2.3.2 Earworms ...... 17 2.4 How does the use of language contribute to the music’s poetic or emotive appeal? ...... 19 2.4.1 Self-Focus, Narcissism & Anti-Social Behavior ...... 19 2.4.2 Seriousness & Meaningfulness ...... 20 2.4.3 Sexual Content ...... 21

ix 2.5 To whom is this music aimed? ...... 21 2.5.1 Style & Genre ...... 21 2.5.2 Personality Traits ...... 24 2.5.3 Tasks & Activities ...... 25 2.6 How does the music establish its credibility or authority? ...... 26 2.6.1 Guest Appearances ...... 26 2.6.2 Musical Quotation & Sampling ...... 27 2.7 Concluding Remarks ...... 27

3. Testing Five Musical Features Arising from the Theory of Attention Economy 31

3.1 Introduction ...... 31 3.2 Study 1: Practices between 1986 and 2015 ...... 36 3.2.1 Hypotheses ...... 36 3.2.2 Sample ...... 39 3.2.3 Method ...... 39 3.2.4 Results ...... 40 3.2.5 Discussion ...... 42 3.3 Study 2: Within-artist comparisons of success ...... 44 3.3.1 Hypotheses ...... 47 3.3.2 Sample ...... 48 3.3.3 Results ...... 49 3.3.4 Discussion ...... 51 3.4 General Discussion ...... 51

4. Testing the Inﬂuence of Attention Economy Principles, Musical Prefer- ences, and Familiarity on Music Listening Choice Behavior ...... 54

4.1 Introduction ...... 54 4.2 Hypotheses ...... 58 4.3 Experiment 1: Undivided Attention Music Listening Setting .... 59 4.3.1 Method ...... 60 4.3.1.1 Participants ...... 60 4.3.1.2 Material ...... 60 4.3.2 Procedure ...... 61 4.3.3 Results ...... 62 4.3.4 Discussion ...... 64 4.4 Experiment 2: Divided Attention Music Listening Setting ..... 68 4.4.1 Method ...... 69 4.4.1.1 Participants ...... 69 4.4.1.2 Material ...... 69 4.4.1.3 Procedure ...... 70

x 4.4.2 Results ...... 71 4.4.3 Discussion ...... 72 4.5 General Discussion ...... 75 4.6 Conclusion ...... 78

5. General Summary ...... 80

5.1 Recapitulation ...... 80 5.2 Implications for the Music Industry ...... 82 5.3FutureResearch...... 86 5.4 Conclusion ...... 87

Appendices 88

A. List of the 303 songs analyzed in the ﬁrst study (Section 3.2) of Chapter 3 88

B. List of the 120 songs analyzed in the second study (Section 3.3) of Chapter 3 97

C. Visual representation of the participants’ skipping times in both the undivided and divided attention experimental setting (Chapter 4) superimposed onto each song’s formal diagram ...... 103

Bibliography ...... 129

xi List of Tables

Table Page

2.1 Summary of the literature discussed in this chapter...... 28

3.1 Comparison between ﬁrst-person singular pronoun usage in the present study and in DeWall et al. (2011)...... 45

3.2 Pearson correlation matrix for ﬁve variables...... 46

3.3 Paired t-test results comparing most popular and less popular songs according to ﬁve variables...... 50

4.1 Summary of multiple regression for undivided attention...... 63

4.2 Summary of participants’ responses in the undivided attention condition to the prompt “Please brieﬂy explain why you skipped the previous song you heard.” ...... 67

4.3 Summary of multiple regression for divided attention...... 72

xii List of Figures

Figure Page

2.1 The attention economy model predicts that, as the disparity between available human attention and information content increases, so does the value of attention. This ﬁgure illustrates the attention economy model. The x-axis represents time. The y-axis represents the sum of the information content of all attention-grabbing products (e.g. web pages, images, music, ads) available to a single individual. The dashed- line represents the limited amount of attention a single person has. The scarcity-based increasing value of attention is represented by the darkening of the shaded area: as the gap between information content and human intention increases, so does the value of human attention. 6

2.2 “Blue Moon.” Music by Richard Rogers. Lyrics by Lorenz Hart. Pub- lished in 1935. An example of a song from the ‘Great American Song- book’ written following an AABA structure...... 13

2.3 Proportion of AABA and Verse-Chorus song forms between 1958 and 1971. The graph shows a decline in popularity of the AABA form in the 1960s in favor of the verse-chorus form (where verse and chorus exhibit diﬀerent harmonic structures). This shift seems to occur between 1964– 65, with a drastic and quasi-steady decline in frequency of the AABA form over the second half of the decade, going from being present in roughly 40% of the songs in 1964 to being almost nonexistent four years later in 1967. Figure adapted from Léveillé Gauvin (2015, Figure 3.4). 14

xiii 2.4 Proposed cognitive model of Music Listening Choice Behavior (MLCB) featuring a three-stage screening process. In the first, familiarity- related stage (∼ 0–1 sec), listeners may recognize a specific musical work and decide whether the given work is of interest. If the answer is yes, the song passes the familiarity-related screening process and the listening continues, starting the style-related screening process; if the answer no, the song fails to pass the familiarity screening process and the skip button is pressed. In the second, style-related stage (∼ 0–5 sec), listeners quickly recognize the style/genre and assess their interest in listening to music of that style/genre. They then assess whether they want to listen to a song belonging to that category. If the answer is yes, the song passes the style-related screening process and the listening continues, starting the third, song-related screening process (∼ 5–20 sec); if the answer no, the song fails to pass the style-related screening process and the skip button is pressed. Finally, in the third, song-related stage, listeners assess whether they enjoy they like the musical work, based on their current mood and context. In this stage, characteristics that are specific to this song (such as the melody, the harmony, the lyrics) might come into consideration. Although visually illustrated as three independent stages, the screening processes presented in the MLCB model are likely to temporally overlap with one another...... 23

3.1 Relationship between the year a song appeared on the Billboard Year- End Hot 100 chart and a) the number of words in its title (r = -.27, p < .001), b) its main tempo, in BPM (r =.18,p = .002), c) the elapsed time before the voice enters, in seconds (r = -.36, p < .001), and d) the elapsed time before the title of the song is mentioned, in seconds (r = -.18, p < .001). For Figure 1a, dots of varying sizes have been used to illustrate overlapping data points, as indicated by the graph’s legend. These tests remain signiﬁcant after a Bonferroni adjusted alpha level of .01 per test (.05/5). Not shown is the lack of correlation between the year a song was popular and its self-focus score (r = .01, p = .897). The results are consistent with the attention economy principles. . . . 41

xiv 4.1 A visual summary of the multiple regression for undivided attention, F (9, 565) = 7.13, p < .001, R2adj = .09. The black dots represent the beta values, the horizontal lines represent +1 SE and -1 SE.Ifthe standard error bars cross vertical line at x=0, it means that the beta value is not significantly different from zero. Three predictors are statistically significant in predicting listening times: the number of words in the title, the amount of time before the voice enters, and whether or not a person is familiar with the song. For every additional word in a song title, listening times shortened by 5.88 seconds, on average (SE = 3.92). For each additional second of instrumental introduction, listening times increased, on average, by 1.06 seconds (SE = 0.28). Songs familiar to participants were listened for an extra 21.27 seconds, on average (SE = 7.01)...... 65

4.2 Multiple regression for divided attention, F (9, 506) = 12.46, p < .001, R2adj = .17. Four predictors are statistically signiﬁcant in predicting listening times: the amount of time before the voice enters, two out of four musical preferences dimensions (‘Upbeat & Conventional’ and ‘Energetic and Rhythmic’) and whether or not a person was familiar with the song. For each additional second of instrumental introduction, listening times increased, on average, by 1.42 seconds (SE = 0.34). The inﬂuence of musical preferences on listening times in the divided attention setting were considerable. For every point increase in the ‘Energetic & Rhythmic’ music preferences dimension (i.e. Electronic, Hip-hop/Rap, Soul), listening times increased by 15.61 seconds, on average (SE = 3.92). Similarly, for every point increase in the ‘Upbeat & Conventional’ music dimension (i.e. Country, Pop, Religious, and Sound Tracks), listening times increased by 12.10 seconds, on average (SE = 5.54). Familiarity remained the most important predictor, with songs familiar to participants being listened to for an extra 21.90 seconds, on average (SE = 8.54)...... 74

xv Chapter 1: Introduction

This dissertation examines how technological changes in the last decades have inﬂuenced the way we consume music, not only granting immediate access to a much larger collection of songs than ever before, but also allowing us to instantly skip songs.

This new reality can be partially explained by the theory of attention economy which posits that attention is the currency of the information age, since it is both scarce and valuable. The goal of this dissertation is to examine whether popular music compositional practices have changed in the last decades in a way that is consistent with attention economy principles, and whether music listening choice behavior can be predicted by these principles.

Chapter 2 examines the advertisement function of popular music from an attention economy perspective. It argues that this new economic reality has amplified the advertising nature of popular music in recent years by encouraging music makers to modify their compositional practices in a way that favors attention grabbing. We begin by defining what the attention economy theory is, and we highlight why it is relevant to the analysis of popular music. Then, using the analytic framework proposed by Huron (1989), a survey of the relevant literature on the topic is offered, with emphasis on the repercussion of the attention economy theory on popular music.

A summary of this survey is presented in Table 2.1.

1 Chapter 3 examines whether popular music compositional practices have changed between 1986 and 2015 in a way that is consistent with attention economy principles.

The results of two corpus studies are reported. In the ﬁrst study, 303 U.S. top-ten singles from 1986 to 2015 were analyzed according to ﬁve parameters: number of words in title, main tempo, time before the voice enters, time before the title is mentioned, and self-focus in lyrical content. The results revealed that popular music has been changing in a way that favors attention grabbing, consistent with attention economy principles. In the second study, 60 popular songs from 2015 were paired with 60 less popular songs from the same artists. The same parameters were evaluated. The data were not consistent with any of the hypotheses regarding the relationship between attention economy principles within a comparison of popular and less popular music.

Chapter 4 examines music listening behavior in relation to the attention economy principles identiﬁed in the preceding chapter. Speciﬁcally, the goal is to test whether attention economy principles (i.e. the number of words in title, the main tempo, the time before the voice enters, and the time before the title is heard), musical preferences, and familiarity can predict how long a listener will listen to a song before skipping to the next one. The results of two behavioral experiments are reported.

In the ﬁrst experiment, participants were asked to listen to randomly sampled songs, with no distraction task. In the second experiment, an independent group of participants was asked to listen to a subset of the same songs while participating in a distraction task. The results from these experiments suggest that music listening choice behavior is moderated by the listening context and can be predicted, at least partially, by the participant’s familiarity with the song, their musical preferences, and

2 some compositional techniques such as the use of an instrumental introduction and the number of words in the title of a song.

Finally, Chapter 5 reiterates the attention economy model discussed in this dissertation, summarizes the ﬁndings of the studies presented, and discusses possible avenues for future research. In addition, three appendices are included: Appendix A contains a list of the 303 songs analyzed in the ﬁrst study (Section 3.2) of Chapter

3, Appendix B contains a list of the 120 songs analyzed in the second study (Section

3.3) of Chapter 3, and Appendix C oﬀers a visual representation of the participants’ skipping times as discussed in Chapter 4.

3 Chapter 2: Capturing and Sustaining Attention in the Attention Economy Era

We live in a world where information is abundant, but attention is scarce. New information is constantly being generated in the form of books, web pages, movies, songs, etc. But as Nobel laureate Herbert Simon put it, “a wealth of information creates a poverty of attention and a need to allocate that attention eﬃciently among the overabundance of information sources that might consume it” (1971, p. 40–41).

The scarcity of attention in an information-rich world has yielded the concept of

‘economy of attention’ (e.g. Thorngate, 1990; Goldhaber, 1997; Davenport & Beck,

2001), an economic theory suggesting that human attention, being both scarce and valuable, can be thought of as a currency requiring its own supply and demand model.

Indeed, even the language we use to talk about attention echoes this perspective. As

Kahneman points out, “[t]he often-used phrase ‘pay attention’ is apt: you dispose of a limited budget of attention that you can allocate to activities, and if you try to go beyond your budget, you will fail” (2011, p. 23).

The attention economy theory is based on a commodiﬁcation of the human capacity for attention. It relies on the idea that information is plentiful and attention is the scarce resource. In this ecosystem, the main attention metric is ‘time’ (Thorn- gate, 1990). That is, when we refer to attention as a currency, it must be understood

4 as a time-based attention currency. Regarding attention as a time-based currency

is not simply a theoretical construct. Media companies are already experimenting

with time-based sales with advertisers. For example, publishers such as The Finan-

cial Times, The Wall Street Journal and The Economist are oﬀering advertisers the option to buy ads on a cost-per-hour (CPH) basis (Merritt, 2017).

Attention also has a scarcity-based value (as opposed to a cost-determined value), meaning that its value is driven by demand. Because information generation is constantly increasing, so is the value of attention. This theoretical model is illustrated in Figure 2.1.

Historically, the main product populating the attention marketplace has been advertising. Modern advertising began in the 17th century, and for three centuries, advertising was closely associated with print media (e.g. newspapers, periodicals, posters). By the 20th century, however, the attention marketplace had expanded to include a variety of new venues, such as radio shows and television programs.

As consumers, we tend to think of radio stations as a means for delivering music entertainment, but from a commercial perspective, “commercial radio stations exist to broadcast adverts, not music” (North & Hargreaves, 2011, p. 914). The same is true for television channels. In 2004, Patrick Le Lay, then chairman and CEO of France’s TF1 Group—the holding company of France’s most popular television channel—created a national controversy when he candidly stated that TF1’s main job was to help Coca-Cola sell its products by selling available human brain time to the multinational corporation (Lay, 2004, p. 92). While Le Lay’s comments might have been culturally insensitive, his observation nevertheless perfectly describes the attention economy framework in which we live.

5 Information

Increasing value of attention

Human Attention

Time (in years)

Figure 2.1. The attention economy model predicts that, as the disparity between available human attention and information content increases, so does the value of attention. This ﬁgure illustrates the attention economy model. The x-axis represents time. The y-axis represents the sum of the information content of all attention- grabbing products (e.g. web pages, images, music, ads) available to a single individual. The dashed-line represents the limited amount of attention a single person has. The scarcity-based increasing value of attention is represented by the darkening of the shaded area: as the gap between information content and human intention increases, so does the value of human attention.

6 Perhaps one of the most important manifestations of the attention economy model is the striking growth in popularity of online streaming platforms. Music streaming services such as Spotify and Apple Music make little sense for artists when thinking in terms of traditional currencies, with royalties from online streaming being significantly smaller than from music sales (CDs or digital). However, when one considers attention as a currency, online streaming becomes a significant tool to reap listeners’ attention, and thus offers new possibilities.

Since every information product aspires to engage attention in some way or another, it is useful to think of information products in terms of advertising. Huron

(1989) proposed that a musical work acts as a kind of advertisement for itself. We might also consider a song as an advertisement promoting an artist’s personal brand.

Of course, this phenomenon was not caused by online streaming. When a song is played on broadcast radio, it similarly acts as an advertisement for the artist. The principal diﬀerence between the two media is the extraordinary competitiveness that characterizes online streaming. With millions of songs available instantaneously for free, technological changes in recent decades have made the war for attention more intense than ever.

Huron (1989) identiﬁed six basic ways in which music can enhance advertising:

1) Entertainment, 2) Structure/Continuity, 3) Memorability, 4) Lyrical Language, 5)

Targeting, and 6) Authority Establishment. These six categories can be summarized in six seminal questions:

1) On what basis is the music entertaining or engaging?

2) How is the music structured and what role(s) does this structure serve?

3) How does the music achieve memorability?

7 4) How does the use of language contribute to its poetic or emotive appeal?

5) To whom is this music aimed?

6) How does the music establish its credibility or authority? (Huron, 1989,

p. 572).

Using these six questions as an analytic framework, this chapter examines the advertisement function of popular music from an attention economy perspective.

Specifically, it will be argued that this new economic reality has amplified the self- advertising nature of popular music in recent years by encouraging popular music makers to adapt their compositional practices in a way that favors attention grabbing. For each question raised by Huron, a survey of the relevant literature on the topic will be offered, with emphasis on the attention economy theory.

2.1 On what basis is the music entertaining or engaging?

2.1.1 Arousal

Despite its ubiquitous use in everyday discourse, entertainment is a poorly-deﬁned

concept. In this chapter, we follow in Huron’s footsteps by deﬁning entertainment as

any experience that is engaging.

Studies have shown that arousal can help our brains identify information as being

important through subjective experience and neural activation of sympathetic nerve

ﬁbers that liberate epinephrine or norepinephrine (Storbeck & Clore, 2008). Futher-

more, studies have shown that arousal can facilitate memory by enhancing attention

during the encoding process (Christianson & Loftus, 1991) and that increased arousal

supports slower forgetting (Sharot & Phelps, 2004).

8 Exposure to audio stimuli with faster tempi has been linked with increases in listeners’ levels of arousal (e.g. Husain et al., 2002; Kellaris & Kent, 1993). All other things being equal, songs with faster tempi should be more memorable. As such, it might be useful to think of tempo as a compositional device that can be used to increase listener’s arousal, and thus increase the memorability of a song. Considering the premise of this chapter that the increasing generation of information content has created a much more competitive environment for music makers, we should expect songs with faster tempi to be more arousing, and thus more memorable. Since familiar songs tend to be preferred (Ward et al., 2014) and since familiarity with a song yields longer listening times (see Chapter 4 of this document), an increase in the average tempo of popular songs in the last decades would be consistent with the attention economy theory. Empirical research on the topic shows mixed results, with a general decrease in average tempo from the 1960s to 1990s, followed by a general increase in tempo between the 1990s and the 2000s (Schellenberg & von Scheve, 2012;

Léveillé Gauvin, 2017).

Another way to increase arousal is through increased loudness. Serrà and colleagues (2012) used an auditory model in order to estimate the perceived loudness from audio signals for more than 450,000 distinct popular recordings from 1955 to

2010. The results showed growing levels of loudness over time (the so-called ‘loudness war’), a phenomenon that is also consistent with the attention economy theory.

9 2.1.2 Predictability

Meyer (1956) argued that the secret to composing engaging music was to strike a balance between predictability and surprise. Similarly, Berlyne (1970, 1971), inspired by the Yerkes-Dodson law (1908), proposed that an inverted U-shaped relationship exists between the complexity of a stimuli and how pleasurable it is perceived. Huron

(2013) suggests the eﬀect arises from two well-known psychological phenomena: ‘processing ﬂuency,’ where easily processed stimuli are experienced as positive, and ‘habituation,’ where redundant stimuli lead to reduced responsiveness and subsequent boredom. Hence, we might imagine an inverted U-shaped curve where the upper part represents a ‘sweet spot’ between predictability and surprise.

North and Hargreaves (1995) tested this conjecture by asking 75 participants to rate excerpts of popular music for liking, subjective complexity, and familiarity.

As predicted, they observed an inverted-U relationship between liking and subjective complexity, suggesting that listeners prefer songs that balance predictability and surprise. A similar study was conducted by Miles et al. (2017), this time using information content instead of participants’ subjective perception of complexity. The authors analyzed the harmonic content of over 500 songs popular between 1958 and

1991 in order to determine whether the popularity of a song is correlated with the predictability of the chord progression it uses. Their results suggest that the most popular songs tend to feature more surprising harmonic events. Moreover, the most popular songs also tend to feature unpredictable verses followed by predictable choruses. Though not decisive, the combined results of North and Hargreaves (1995) and Miles et al. (2017), along with others (e.g. Vitz, 1966; Heyduk, 1975; Tan et al.,

10 2006), suggest that, as Meyer posited, listeners might prefer music that strikes a balance between predictability and surprise. However, converging evidence from further research is needed before making stronger claims.

2.2 How is the music structured and what role(s) does this structure serve?

2.2.1 AABA and Verse-Chorus Forms

Two main formal structures have dominated popular music throughout the 20th century: the AABA form and the verse-chorus form. The AABA form is closely associated with the ﬁrst half of the century and was a ubiquitous vehicle for Tin-Pan Alley,

Broadway, and Hollywood songwriters. Famous examples of AABA songs include

“Blue Moon” (as shown in Figure 2.3), “Heart and Soul,” and “The Christmas Song.”

Von Happen and Frei-Hauenschild have referred to the AABA form as “the pop song form par excellence” during this period (2015, p. 14). The A-section verses often, though not always, end with a refrain, a recurring line of text—often the title of the piece—usually supported by a cadential progression. To this day, the AABA form is still widely associated with the so-called ‘Great American Songbook.’ Of interest is the abrupt decline in popularity of the AABA form in the early 1960s. This shift from AABA from to verse-chorus form is illustrated in Figure 2.3 using data from

Léveillé Gauvin (2015). Famous examples of verse-chorus songs include “Blue Moon”

(as shown in Figure 2.3), “Heart and Soul,” and “The Christmas Song.”What seems clear from this ﬁgure is that the AABA form did not simply decline; it got replaced.

Von Happen and Frei-Hauenschild (2015) suggest that the verse-chorus form evolved

11 directly from the AABA form, with the embedded refrain becoming the chorus, a fully independent section. The obvious advantage of this shift from refrain to chorus is drawing greater attention to the chorus of the song, which typically includes the so-called ‘hook’ of the song.

Although the verse-chorus form aﬀords opportunities to emphasize the hook of a song, one must ask: What happened in the 60s to motivate this paradigmatic shift?

The attention economy theory oﬀers a possible explanation. In 1949, RCA Victor introduced the vinyl 7-inch 45-rpm record. Unlike shellac 78-rpm discs, the 7-inch single was light, portable, and durable. Moreover, this new technology was more aﬀordable, making it a better format for millions of post-war baby-boom teenagers.

In eﬀect, the 7-inch single increased accessibility to a larger collection of music, increasing competition, and consequently encouraging artists to make songs that are more attention-grabbing.

2.2.2 Instrumental Introductions

Between the mid-1980s and the mid-2010s, another important structural change that happened is the near disappearance of the instrumental introduction. Léveillé Gauvin

(2017) reported that instrumental introductions dramatically shortened between 1986 and 2015, from an average of 23 seconds to just ﬁve seconds. Considering that vocal music is more attention-grabbing than instrumental music (e.g. Radano, 1989; Allan,

2006), this change is consistent with the attention economy theory. Furthermore, results from the same study suggest that the title hook appears earlier than in previous decades, which is also consistent with the attention economy theory.

12 E¨ C‹ F‹ B¨7 E¨ C‹ F‹ B¨7 b A &b bC œw™ œ œ œ œ œ œ œ œw œ œ œ œ œ œ œ œ Blue Moon you saw me stand-ing a- lone With- out a dream in my heart,

5 E¨ C‹ F‹ B¨11 E¨ A¨ E¨ B¨7 b &b b w œ œ œ œ œ œ œ œw ˙ Œ œ With- out aloveofmyown Blue

9 E¨ C‹ F‹ B¨7 E¨ C‹ F‹ B¨7 bb w œ œ œ œ œ œ œ œ œ œ™ ˙ œ œ A & b J œ œ œ œ œ œ Moon you knew just what I was there for You heard me say-ing a pray'r

13 E¨ C‹ F‹ B¨11 E¨ A¨ E¨ bb j j j & b œ œ™ ˙ œ œ œ œ œ œ œ œ œ œ™ ˙ ˙ ‰ œ œ œ for Some- one I really could care for And then there

17 F‹7 B¨7 E¨ F‹7 B¨7 E¨ b j B &b b œ œ œ œ œ œ œ œ™ œ œ œ œ œ œ œ œ œ œ ˙ œœ œ œ sud-den-ly appeared before me The only one my arms will ev-er hold, I heard some

21 A¨‹7 D¨7 G¨ B¨ F7 F‹7 B¨7 bbb j ˙™ œ & œ œ œ œ œ œ bœ œ™ œ œ œ œ nœ™ œ œ™ œ œ œ bod-y whis- per,"Please a- dore me," And when Ilooked, the moon had turn to gold! Blue

25 E¨ C‹ F‹ B¨7 E¨ C‹ F‹ B¨7 b œ A &b b w œ œ œ œ œ œ œw œ œ œ œ œ œ œ œ Moon now I'm no long-er a- lone With- out a dream in my hear,

29 E¨ C‹ F‹ B¨11 B¨ A¨ E¨ bb & b w œ œ œ œ œ œ œ œw ˙ Ó With- out a love of my own.

Figure 2.2. “Blue Moon.” Music by Richard Rogers. Lyrics by Lorenz Hart. Pub- lished in 1935. An example of a song from the ‘Great American Songbook’ written following an AABA structure.

13 60% AABA

50% Verse-Chorus

40%

30%

20% 14

10% Proportion of songs by formal structures

0% 1958 1959 1961 1965 1967 1968 1969 1971 1960 1963 1964 1962 1966 1970

Figure 2.3. Proportion of AABA and Verse-Chorus song forms between 1958 and 1971. The graph shows a decline in popularity of the AABA form in the 1960s in favor of the verse-chorus form (where verse and chorus exhibit diﬀerent harmonic structures). This shift seems to occur between 1964–65, with a drastic and quasi-steady decline in frequency of the AABA form over the second half of the decade, going from being present in roughly 40% of the songs in 1964 to being almost nonexistent four years later in 1967. Figure adapted from Léveillé Gauvin (2015, Figure 3.4). 2.3 How does the music achieve memorability?

2.3.1 Repetition & Memory

Arguably, the most memorable part of a pop song is the hook. The term hook is informally used to describe a musical passage which lingers in memory. From a research perspective, the term is problematic since a same passage may or may not be memorable depending on the listener (Traut, 2005). The term is most often used to refer to the portion of a song whose lyrics relay the title of the song (Davis, 1985).

Therefore one might operationalize hook as passage associated with title. In AABA songs, this passage is often located in the A section. For example, in “Blue Moon”

(see Figure 2.3), the beginning of each A section could be operationalized as the hook of the song. Since the rise in popularity of the verse-chorus form in the 1960s, the hook has often been associated with the chorus of a song. Corey et al. (2017) asked participants to rate their continuous enjoyment of eight popular songs using sliders.

The results suggest that enjoyment depends on the formal section, with choruses being rated as more enjoyable than other sections. It is not clear, however, of the greater enjoyment is caused by chorus repetition or if the correlation is merely coincidental.

Chorus repetition also seems to play a role in the popularity of a song. Nunes et al.

(2015) showed that greater repetition of choruses within a song increases both the likelihood of a song reaching the number-one position in the charts and the speed at which it reaches this position.

In discussing the role of climaxes in popular music, Osborn (2013) suggests that climactic sections such as choruses often feature, among other things, repeated lyrical and melodic hooks, modulations, simple harmonic progressions and a ‘laid-back’

15 rhythmic section. Osborn’s model echoes Temperley’s “loose verse/tight chorus”

(LVTC) form (2007), in which choruses are more tightly-knit than verses in terms of rhythm, texture, and tonality. Biamonte suggests that, in the LVTC model, “the chorus functions both as a contrast to the verse and a resolution of it, and is often repeated multiple times at the end of the song to create a de facto coda” (2014, par. 7.7).

Research shows that the interaction between rhythm and pitch aﬀects memorability such that melodies are easier to parse and remember when tones occur on metrically accented moments (Jones et al., 1982; Dowling et al., 1987; Dowling, 1990). In addition to being repeated within a song, the tight interaction between rhythm and tonality in choruses might be expected to contribute to their memorability. Although systematic longitudinal research has yet to investigate this, a tightening of choruses in the last decades would be consistent with the attention economy theory.

Another important aspect of popular music that uses repetition as a driving force is lyrical content. Nunes et al. (2015) analyzed lyrical repetition in hit songs from

1958 to 2012 and observed that songs using greater lyrical repetition were more likely to reach the number one position in the charts, reached this position faster, and were more likely to debut in the top-40 charts. One caveat, however, is that the positive eﬀect of a chorus repetition on a song (see above) is negated if the degree of lyrical repetition is too high (Nunes et al., 2015). This eﬀect is consistent with the existence of a ‘sweet spot’ concept balancing predictability and surprise discussed in section

2.1.2.

The work of Nunes et al. (2015) raises the question of whether pop lyrics are getting more repetitive. Morris (2017) conducted a corpus-study analysis of 15,000 lyrics from songs that charted on the Billboard’s Year End Hot 100 charts between 1958

16 and 2017. The data showed an increasing use of lyrical repetition over time. Further- more, there appears to be a clear distinction between songs reaching the top-10 of the charts and other songs, with top-10 songs exhibiting more lyrical repetition. Once again, these changes in writing practices are consistent with the attention economy model.

2.3.2 Earworms

Music can be tenacious in human memory, perhaps more than any other type of stimuli. From an advertising perspective, this cognitive stickiness potential represents an important tool for music makers. In popular discourse, this phenomenon is commonly discussed under the rubric of ‘earworms,’ where a person involuntary imagines a musical sound or passage in the absence of any actual sound. A large survey conducted in

Finland (Liikkanen, 2008) suggests that 90% of the population experience earworms on a weekly basis, and 33% on a daily basis. Other research suggests that people who consider music to be important in their life tend to experience longer and more tenacious auditory imagery (Levitin, 2006; Beaman & Williams, 2010).

In one of the ﬁrst empirical works aiming to identify musical properties associated with earworms, Kellaris (2001, reviewed in Kellaris, 2008) suggested that features such as simplicity and repetitiveness might contribute to the cognitive stickiness of a musical passage. Beaman and Williams (2010) have speculated that simple and repetitive tunes are easier to overlearn, and thus more likely to become earworms.

Similarly, Williamson and Müllensiefen (2012) compared 29 songs frequently identiﬁed

17 as earworms to 29 control songs and found that songs that are easier to sing (specifically, songs containing longer notes and smaller pitch intervals) are more likely to be identiﬁed as earworms. Building on this work, Jakubowski and colleagues (2017) analyzed a set of 100 songs commonly identiﬁed as earworms and compared them with a control sample of 100 songs with matching popularity and style. Their results suggest that songs featuring common melodic contours (i.e. arch-shaped phrases, see

Huron, 1996) are more likely to be identified as earworms. Moreover, songs with unexpected leaps and repeated notes are also more likely to create involuntary musical imagery. They also observed a difference in tempo between the two groups, with songs commonly identified as earworms featuring, on average, faster tempi than the songs in the control group. More generally, song choruses are more likely to be the part of a song involved in musical imagery (Bailes, 2007; Beaman & Williams, 2010).

If changes in music delivery over recent decades favor music that most conforms with attention economy principles, then we would predict that more recently created music might have a greater propensity to induce earworms in listeners. This conjecture, however, is confounded by the greater repetition or greater exposure that appears to characterize modern listening habits. Sacks (2007) hypothesized that the omnipresence of music in modern lives, both in public spaces (e.g. background music at restaurants) and in private spaces (e.g. portable music players), has increased the incidence of earworms. Similarly, Margulis notes how recent technologies aﬀord “a degree and pervasiveness of repetition that was previously unheard” (2014, p. 77).

We know that earworms are almost always existing melodies that are familiar to people (Beaman & Williams, 2010), that songs heard aloud are more likely candidates

18 (Hyman et al., 2013; Liikkanen, 2012), and that recent exposure seems to trigger auditory imagery (Bailes, 2015; Floridou & Müllensiefen, 2015; Jakubowski et al., 2015;

Williamson et al., 2012). However, no longitudinal study has yet been conducted to track the incidence of earworms over long periods of time.

2.4 How does the use of language contribute to the music’s poetic or emotive appeal?

2.4.1 Self-Focus, Narcissism & Anti-Social Behavior

Recent research in sociology and psychology suggests that individualistic personality traits in U.S. culture have been increasing in recent decades (Roberts & Helson, 1997;

Twenge & Campbell, 2001, 2008; Twenge & Foster, 2010; Twenge et al., 2008). For instance, self-congruity theory (Sirgy, 1986; Johar & Sirgy, 1991, for summary, see

Klipfel et al., 2014) suggests that individuals are driven by self-consistency motives when purchasing goods, favoring products that reﬂect the way they perceive themselves (Cisek et al., 2014). From a music perspective, this suggests that songs whose lyrics reﬂect the way listeners perceive themselves are likely to be preferred. Research also shows that people scoring high in private self-consciousness scales (i.e. scales measuring one’s tendency for introspection) tend to recognize self-relevant words more quickly (Eichstaedt & Silvia, 2003). Again, this suggest that the use of self-focused lyrics in popular music could potentially be used as an attention-grabbing compositional device.

DeWall et al. (2011) analyzed lyrics of popular songs from 1980–2007. Their

ﬁndings suggest an increasing use of self-focused words (e.g. I, me, mine) words in

19 the studied period, consistent with the self-congruity theory. However, a replication study by Léveillé Gauvin (2017) investigating popular songs from 1986–2015 failed to replicate these ﬁndings. DeWall et al. (2011) also observed that words in lyrics related to anger and antisocial behavior have also increased over time, suggesting increases in narcissism and social rejection (Bushman & Baumeister, 1998; Twenge

& Campbell, 2003; DeWall et al., 2009).

2.4.2 Seriousness & Meaningfulness

The environmental security hypothesis (ESH, for a review see Nelson et al., 2007) suggests that what we ﬁnd most desirable depends on social and economic conditions, such that meaningful themes, traditionally associated with safety and security, are preferred during uncertain times, whereas fun and carefree themes are preferred during stable times. Pettijohn and Sacco (2009a, b) analyzed lyrics from songs popular between 1955 and 2003 to see if social and economic conditions are reﬂected in lyrical content in a way that is consistent with the ESH theory. Their results suggest that, when social and economic times are more threatening or uncertain, songs tend to be longer in duration, have more words per sentence and have less carefree lyrics

(i.e. more future references and more coverage of social processes). In addition, songs with more comforting and romantic lyrics also tend to be more popular during threatening times. Similar results consistent with the ESH theory have been observed in songs reaching the top-5 positions of the U.K. weekly sales charts between 1960 and

2011 (North et al., 2017). These results replicate similar trends observed in popular television shows (McIntosh et al., 2000).

20 2.4.3 Sexual Content

Another important theme often explored in popular music lyrics is sexual desire and lust. The axiom that ‘sex sells’ is supported by numerous studies. In general, there is agreement that sexual content draws consumers’ automatic attention (e.g. Aaker &

Stayman, 1990; Brown & Stayman, 1992; Lang et al., 2003; Schroeder & McDonagh,

2006; Wolin, 2003), although this eﬀect is moderated by culture (Sawang, 2010).

Considering that using sexual content in advertisements can increase the advertising effectiveness, and considering further the main thesis of this chapter that popular music has a self-advertisement function, we might predict an increase in sexual content in popular lyrics over time. Madanikia and Bartholomew (2014) investigated love and lust content in lyrics for top-40 songs from Billboard’s Year End Hot 100 singles spanning 1971 to 2011. While the first three decades (1970s through 1990s) featured most commonly lyrics about love without sexual references (and to a lesser extent, lyrics combining love and sex), the authors observed a shift from 2001 onward, with songs alluding to sex in the absence of love becoming more common. These findings parallel an increase in sexual imagery in recent decades in other media forms, such as TV programming and films (Farrar et al., 2003).

2.5 To whom is this music aimed?

2.5.1 Style & Genre

In marketing, a target market is a group of people toward whom marketing eﬀorts and products are aimed. When viewed from an advertising perspective, musical styles and genres can act as targeting tools. Schellenberg et al. (1999) found that participants

21 exposed to 100 ms and 200 ms music excerpts were able to identify popular songs by name and Gjerdingen and Perrott (2008) found that participants could identify speciﬁc styles listening to 250 ms excerpts. Building upon these studies, Plazak and

Huron (2011) proposed a chronology of the musical knowledge acquired by participants upon hearing excerpts ranging from 50ms to 3000ms. The stimuli were collected from genre-speciﬁc radio stations available via XM-Sirius Radio. The results suggest that it takes less than 3000 ms for listeners to identify the instrumentation, genre, mood, density, pleasantness, geographic origin, and tempo. What all three studies indicate is that style or genre is highly salient and that broader musical categories based on instrumentation, geographic origin, and mood (e.g. ‘Piano Music,’ ‘Coun- try,’ ‘Portuguese Music,’ ‘Indie Instrumentals,’ ‘Sunday Morning,’ ‘Party Hits’) are very quickly identiﬁed—between 2.5 and 3 seconds. Since genre and style are the most salient demographic markers, these musical features can be used as targeting tools and as such, might be expected to be indicated soon after the music begins.

Lamere (2014) analyzed music listening habits on Spotify for billions of plays and observed that one out of five song was skipped within the first five seconds and one out of three within the first 20 seconds. Lamere’s observations, in conjunction with the studies mentioned above, suggest that listeners may be using a screening approach when listening to music.

A cognitive model of Music Listening Choice Behavior (MLCB) featuring a three- stage screening process is proposed in Figure 2.4. In the ﬁrst, familiarity-related stage (0–1 sec), listeners may recognize a speciﬁc musical work and decide whether the given work is of interest. If the answer is yes, the song passes the familiarity- related screening process and the listening continues, starting the style/genre-related

22 Do I recognize Yes Do I want to Yes Yes Play Does this song meet this song? listen to this song? my expectations?

Do I want to listen No Yes to a song belonging to this style or genre?

No No No

Skip Skip

Familiarity-Related Screening Process Style/Genre-Related Screening Process Song-Related Screening Process 23

0 sec 1 sec 5 sec 20 sec

Figure 2.4. Proposed cognitive model of Music Listening Choice Behavior (MLCB) featuring a three-stage screening process. In the first, familiarity-related stage (∼ 0–1 sec), listeners may recognize a specific musical work and decide whether the given work is of interest. If the answer is yes, the song passes the familiarity-related screening process and the listening continues, starting the style-related screening process; if the answer no, the song fails to pass the familiarity screening process and the skip button is pressed. In the second, style-related stage (∼ 0–5 sec), listeners quickly recognize the style/genre and assess their interest in listening to music of that style/genre. They then assess whether they want to listen to a song belonging to that category. If the answer is yes, the song passes the style-related screening process and the listening continues, starting the third, song-related screening process (∼ 5–20 sec); if the answer no, the song fails to pass the style-related screening process and the skip button is pressed. Finally, in the third, song-related stage, listeners assess whether they enjoy they like the musical work, based on their current mood and context. In this stage, characteristics that are specific to this song (such as the melody, the harmony, the lyrics) might come into consideration. Although visually illustrated as three independent stages, the screening processes presented in the MLCB model are likely to temporally overlap with one another. screening process; if the answer no, the song fails to pass the familiarity screening process and the skip button is pressed. In the second, style/genre-related stage (0–5 sec), listeners quickly recognize the style/genre and assess their interest in listening to music of that style/genre. For example, a listener might acquire the following knowledge after a 5 second excerpt: banjo, violin, female voice, twang, C&W, and Texas.

Based on that knowledge, they might form a mental category, such as ‘American

Country Music.’ They then assess whether they want to listen to a song belonging to that category. If the answer is yes, the song passes the style/genre-related screening process and the listening continues, starting the third, song-related screening process

(5–20 sec); if the answer no, the song fails to pass the style/genre-related screening process and the skip button is pressed. Finally, in the third, song-related stage, listeners assess whether they enjoy they like the musical work, based on their current mood and context. In this stage, characteristics that are speciﬁc to this song (such as the melody, the harmony, the lyrics) might come into consideration.

Since there is a high probability of a song being skipped within the ﬁrst ﬁve seconds, artists wishing to clearly target their audience might be expected to begin their song with material that is representative of the overall song. As discussed earlier, the disappearance of the instrumental introduction in favor of beginning with the chorus Léveillé Gauvin (2017) is consistent with this conjecture.

2.5.2 Personality Traits

Targeting can also be done in terms of music preferences and personality traits.

Groundbreaking work by Rentfrow and Gosling (2003) and subsequent analyses by

24 Rentfrow et al. (2011) suggest that musical preferences can be analyzed using a ﬁve-

dimension model: Mellow (e.g. pop, soft rock, and soul/R&B); Unpretentious (e.g. country and singer-songwriter); Sophisticated (e.g. classical, jazz, and world); Intense

(e.g. heavy metal, punk, and rock); and Contemporary (e.g. rap and electronica music). Music listeners also tend to prefer songs that correspond to their emotional state (Rentfrow & Gosling, 2003). For example, fans of heavy metal, rock, and other arousing music tend to have higher resting arousal, tend to be more sensation seeking, and are more likely to have antisocial personality traits (Litle & Zuckerman, 1986;

McNamara & Ballard, 1999). Similarly, when listening to heavy metal, fans of this genre show increasing arousal levels beyond those of country music fans (Gowensmith

& Bloom, 1997). Other musical preferences tend to be associated with personality dimensions. For example, individuals scoring high on extraversion and psychoticism tend to prefer rap and dance music (McCown et al., 1997), while those associated with openness to experience and empathy are more likely to enjoy sad music (Vuoskoski et al., 2012).

2.5.3 Tasks & Activities

Finally, targeting can be achieved by considering when the music will be listened

to. The Yerkes-Dodson law (1908) states that the optimum arousal when performing

some task depends on task complexity, where the optimum arousal for complex tasks

is lower than for simple tasks. In terms of popular music, the Yerkes-Dodson law

suggests that highly arousing music (e.g. fast-tempo music in a major mode, with

easy listening harmony, see Juslin, 2001) should be preferred for simple activities such

25 as exercising, cleaning, or dancing, whereas low arousing music should be preferred for complex activities, like reading or studying. Online streaming platforms rely on the

Yerkes-Dodson law when designing activity-based playlists by selecting low-arousal music for task-oriented playlists like ‘Reading’ and ‘Studying’ and high-arousal music for playlists like ‘Working Out’ and ‘Running.’

2.6 How does the music establish its credibility or authority?

2.6.1 Guest Appearances

Celebrity endorsement is another common marketing strategy. In a meta-analysis of the literature on the topic, Amos et al. (2008) identified that trustworthiness and ex- pertise play a significant role in celebrity-endorsement effectiveness. Expert endorse- ments also solidify the link between memory and preference, and improve product recall by modifying the encoding process (Klucharev et al., 2008). We observe an analogous phenomenon in popular music, where artists often team up with one another in the hope of generating excitement from the public (Edwards, 2007). If guest appearances in popular music (sometimes referred to as ‘featuring credits’) function as a way to draw listener attention, then the attention economy theory proposed in this chapter would predict an increasing use of this practice in the last decades. In- deed, a study by Tang (2015) analyzing Billboard’s Year End Hot 100 charts between

1990 and 2014 is consistent with this interpretation.

26 2.6.2 Musical Quotation & Sampling

Another way to establish authority is through musical quotation. Musical quotations can take the form of a topical thematic reference (e.g. patriotic, nationalistic, religious) or of a musical borrowing. Topical thematic references are used to create a speciﬁc atmosphere and do not rely on the listener’s familiarity with the sources.

Musical borrowings, however, rely on the listener’s familiarity with the cited material.

As Keppler suggests, “[t]he very existence of such hints [i.e. musical borrowings] is

. . . that the composer is oﬀering to play a game with anyone who knows the rules”

(1956, p. 478). These types of quotations are often intellectually appealing to the listeners who can aptly identify them, as they suggest a shared cultural experience between the listener and the composer. A specific type of musical borrowing common in popular music is sampling. Generally, sampling can be defined as “the use of a rec- ognizable musical figure from another recording” (Boone, 2013, par. 3.1). Sampling is especially important in hip-hop music, where shared background knowledge is often necessary for a proper semantic interpretation, thus acting as a social bonding agent within a community (Schloss, 2014).

2.7 Concluding Remarks

In this chapter, it has been argued that the attention economy has exerted a profound impact on popular music in recent decades. Speciﬁcally, it was proposed that technological changes have dramatically increased the volume of music available, effectively forcing music makers to modify their compositional practices in a way that

27 favors attention grabbing. In eﬀect, songs can be thought of as advertisements pro-

moting an artist’s personal brand. A survey of the empirical literature on popular

music suggests that these changes can be proﬁtably understood from the perspective

of attention economy theory. A summary of this survey is presented in Table 2.1.

Technological changes in the last decades have yielded a dramatic increase in the

volume of information products being created and distributed. Music has been es-

pecially aﬀected by these technological changes, with music listening media changing

every decade or so. The attention economy theory oﬀers a compelling perspective

from which to reinterpret much of the existing empirical literature on popular music.

However, as with any post hoc analysis of the literature, caution is warranted. More

studies employing a priori attention-economy hypotheses are needed to better un-

derstand how technology aﬀects compositional practices. This cautionary note, the

attention economy theory oﬀers an exciting new lens through which popular music

compositional practices may be analyzed.

Table 2.1

Summary of the literature discussed in this chapter. Category Summary of Findings Source(s) Arousal General decrease in tempo between Schellenberg & von 1965–1995, followed by increase in Scheve (2012) tempo between 1995–2009. General increase in tempo between Léveillé Gauvin 1986–2015. (2017) General increase in loudness between Serrà et al. (2012) 1956–2010. Predictability Listeners prefer popular songs that North & Hargreaves balance predictability and surprise. (1995)

28 The most popular songs tend to fea- Miles et al. (2017) ture more surprising harmonic events and unpredictable verses followed by predictable choruses. Form AABA was the most popular formal von Appen & structure between 1920–1949, but was Frei-Hauenschild replaced by verse-chorus form in the (2015); 1960s. Léveillé Gauvin (2015) Reduced duration of instrumental in- Léveillé Gauvin troductions between 1986–2015. (2017) Repetition & Choruses tend to be more enjoyable Corey et al. (2017) Memory than other sections and therefore contribute to the popularity of a song. Melodies are easier to parse and re- Jones et al. (1982); member when tones occur on metri- Dowling et al. cally accented moments. (1987); Dowling (1990) Greater lyrical repetition increases Nunes et al. (2015) the likelihood of a song becoming a number-one hit, the speed at which a song reaches this position in the charts, and the likelihood of debuting in the top-40 charts. Lyrical repetition increases between Morris (2017) 1958–2017. Simplicity and repetitiveness may con- Kellaris (2001, tribute to the memorability of a pas- 2008); Beaman & sage. Williams (2010) Earworms are more likely to fea- Williamson & ture longer notes, smaller pitch inter- Müllensiefen (2012); vals, common melodic contours, un- Jakubowski et al. expected leaps, repeated notes, and (2017) faster tempi. Choruses are more likely to be the part Bailes (2007); of a song involved in auditory imagery. Beaman & Williams (2010) Self-Focus, Increase in frequency of anger- and DeWall et al. (2011) Narcissism & antisocial-related lyrics between 1980– Anti-Social 2007. Behavior

29 Seriousness & During threatening social and eco- Pettijohn & Sacco Meaningful- nomic times, songs tend to be longer, (2009a,b) ness have more words per sentence, and have less carefree and more comforting and romantic lyrics. Sexual Lyrics concerning love without sexual Madanikia & Content references were most common between Bartholomew (2014) 1971–2001. Lyrics concerning sex in the absence of love were most common between 2001–2011. Style & Genre Listeners can recognize speciﬁc genres, Gjerdingen & styles, instrumentation, tempo, and Perrott (2008); identify popular songs by name within Schellenberg et al. 3 seconds. (1999); Plazak & Huron (2011) Personality Listeners prefer songs that concur with Rentfrow & Gosling Traits their personality traits and emotional (2003); Rentfrow states. et al. (2011); Litle & Zuckerman (1986); McNamara & Ballard (1999); McCown et al. (1997); Vuoskoski et al. (2012) Guest Increase in frequency of guest musician Tang (2015) Appearances appearances between 1990–2014.

30 Chapter 3: Testing Five Musical Features Arising from the Theory of Attention Economy

3.1 Introduction

Technological changes in the last 30 years have drastically inﬂuenced the way we enjoy recorded music.1 Portable music devices grew in popularity–from the portable cassette player in the 1980s and the portable CD player in the 1990s to the mp3 player in the 2000s–allowing the music lover for an easy access to a large collection of music.

Furthermore, with the advent in the last 10 years of Internet-based platforms such as

YouTube, Apple Music, and Spotify, gone are the days where one needed to own a specific vinyl record, CD, or even mp3 file to enjoy a specific song. Record collections, once the only way to decide what music you wanted to hear, are now obsolete as we can enjoy almost any commercial recording legally and for free in only a few seconds.

This ever-increasing availability of new music—Spotify’s catalogue includes over 30

1This chapter was ﬁrst published in its full form as:

Léveillé Gauvin, H. (2017). Drawing listener attention in popular music: Testing ﬁve musical features arising from the theory of attention economy. Musicae Scientiae, Prepublished March 1, 2017.

31 million songs—has had a profound impact on the music industry, with music sales in the U.S. decreasing steadily in the last years (Ingham, 2015).

This radical paradigmatic change is not limited to the way we consume music. In an inﬂuential letter published in Wired Magazine in 1997, Goldhaber anticipated the impact of ‘cyberpsace’ on economic practices:

We’ve turned a corner toward an economy where an increasing number of workers are no longer involved directly in the production, transporta- tion, and distribution of material goods, but instead earn their living managing or dealing with information in some form. Most call this an “information economy.” Yet, ours is not truly an information economy. By definition, economics is the study of how a society uses its scarce resources. And information is not scarce—especially on the Net, where it is not only abundant, but overflowing. We are drowning in information, yet constantly increasing our generation of it. So a key question arises: Is there something else that flows through cyberspace, something that is scarce and desirable? There is. No one would put anything on the In- ternet without the hope of obtaining some. It’s called attention. And the economy of attention—not information—is the natural economy of cyberspace. (Goldhaber, 1997, n.p.)

Goldhaber was not the ﬁrst one to come up with this idea. While the concept of attention economy dates back to the early 1970s (Simon, 1971), the idea that attention is worth money has always been the premise behind modern advertising.

Perhaps the best example is the newspaper industry, which for two centuries has been lowering the cost of their product and diversifying their revenues through publicity.

But Goldhaber aptly predicted that “[a]s the Net becomes an increasingly strong presence in the overall economy, the ﬂow of attention will not only anticipate the ﬂow of money, but eventually replace it altogether” (Goldhaber, 1997, n.p.).

The model proposed by online streaming platforms makes little sense in an economic system based on traditional currencies (e.g. USD, EUR, JPY), with artist’s

32 revenues estimated at less than $0.001 USD per listening on a service like Spotify

(Dredge, 2015). Indeed, several artists have voiced concerns over the low royalties linked with such services (e.g. Swift, 2014). And yet, the fact that almost all of the artists do oﬀer their music through some online streaming platform conﬁrms

Goldhaber’s thoughts: in an economic system treating attention as the preeminent currency, online streaming services oﬀer tremendous possibilities.

While the World Wide Web has become ubiquitous since the new millennium, it would be misleading to attribute this paradigmatic shift to the Internet alone.

Free newspapers such as Metro have proven that revenues from advertising alone can be suﬃcient to make a business model proﬁtable. Perhaps more relevant to our discussion is the tremendous success of MTV in the 1980s. In many ways, the early

MTV model—a television channel broadcasting music 24/7—was based on the idea that “the origin and purpose of music videos is promotional; they are themselves advertisements” (Huron, 1989, p. 570, emphasis in original).

While the meaning of the term may seem intuitive, it might be appropriate to formally deﬁne what we mean by attention. In their book on attention economy,

Davenport and Beck deﬁne attention as follows: “Attention is focused mental engagement on a particular item of information” (Davenport & Beck, 2001, p. 20). Key to this deﬁnition is the notion of “focused mental engagement,” which distinguishes attention from mere awareness. The transition from awareness to attention occurs when “information reaches a threshold of meaning in our brains and spurs the potential for action” (Davenport & Beck, 2001, p. 22). These actions may be unconscious

(e.g. orienting response), or conscious (e.g. buying a speciﬁc product after viewing an advertisement).

33 Formally, we may operationalize attention economy principles as principles that favor focused mental engagement in order to elicit an action that is beneﬁcial for the information generator. Because attention is scarce, it can be convenient to think of attention economy principles as attention-grabbing principles.

A supply and demand economic model predicts that if the demand exceeds supply, the value of the supply increases. In the attention economy model, attention acts as a currency. As such, if the demand for attention is greater than the attention supply

(i.e., more things than one can attend to), then the scarcity of attention increases in value (see Figure 2.1 of Chapter 2). Furthermore, since an increase in demand indicates high economic value, a supply and demand model predicts an increase of supply (Ciampaglia et al., 2015). In other words, since attention is both valuable and scarce, the number of products trying to grab someone’s attention increases, making attention scarcer, and thus even more valuable. This phenomenon is especially relevant in the context of digital content such as web-pages or online streaming.

Research investigating Internet browsing habits showed that “users adopt a ‘screen- and- glean’ browsing behavior where they vet the page prior to more detailed exam- ination” (Liu et al., 2010, p. 386). Moreover, the same study indicated that there appears to be a critical window where the probability of abandoning the page is high, but once this window has passed, the abandonment rate diminishes. The authors describe this window in terms of “screening process:” if a page survives the screening process, the browsing time is high; if it doesn’t, the browsing time is low. These results were consistent with previous studies that showed that web-users visit many pages within seconds (Cockburn & McKenzie, 2001), and that more than half of the visits are shorter than 10 seconds (Weinreich et al., 2008)

34 A similar process can be observed with music listening habits. Lamere (2014) parsed the Spotify data to investigate how millions of music consumers use the skip button. Twenty-one percent of the billions of plays analyzed were skipped in the ﬁrst

5 seconds. That number reaches 34% after the ﬁrst 20 seconds. Moreover, only 51% of the songs are listened to in their entirety. As Lamere points out, “most of the song skips happen within the ﬁrst 20 seconds or so of the song, and after that there’s a relatively small but steady skipping rate” (2014, n.p.). There appears to be a similar critical time period akin to the screening window for web browsing described earlier, although for music, the data suggest that it might be best to understand music listening behavior as a multi-window process, as discussed in Section 2.5.1 of

Chapter 2.

In the present research, several characteristics of popular songs have been investigated using a database-driven approach. Two related studies are presented that aim to examine compositional practices. The ﬁrst study analyzes some 300 of the most popular songs between 1986 and 2015 to investigate whether compositional and production practices in popular music have changed in recent decades in accordance with attention economics. The results will show that most of the parameters evaluated support the idea that compositional practices have changed in the last 30 years in a way that is consistent with the proposed theory. The second study investigates whether the most popular songs of a given artist exhibit attention economics principles at a higher level than less popular songs from the same artist. A total of 120 songs from 2015 were analyzed, representing 60 diﬀerent artists. To anticipate our

35 results, the data were not consistent with any of the hypotheses regarding the relationship between attention economy principles within a comparison of popular and less popular music.

3.2 Study 1: Practices between 1986 and 2015

3.2.1 Hypotheses

An essential aspect of conducting research is to aptly operationalize our proposed conjecture. Five musical parameters were identiﬁed as proxies to evaluate whether or not a song conforms to attention economy principles: number of words in the title, main tempo, time before the voice enters, time before the title is mentioned, and self-focus in lyrical content.

(1) Number of words in title: Several popular music commentators have

recently written about the tendency for popular singles to have short titles.

Billboard Magazine coined the term “One-Word Wonders”—a direct allusion to

the popular expression “One-Hit Wonder,” which refers to an artist with a single

top-40 hit (Trust, 2014). Kopf (2016) aﬃrmed that the average number of words

per song title in the Billboard Hot 100 charts has systematically decreased in

the last several decades. Kopf also suggested that songs in Billboard’s Top 20

charts are more likely to have one-word titles than songs that charted between

the 21st and 100th position. Research has shown that memory span is inversely

related to word length (Baddeley et al., 1975). Consequently, shorter titles

should be more memorable, consistent with the attention economy theory. We

36 thus hypothesize that the average number of words in song titles will decrease over the studied time period.

(2) Main tempo: Exposure to audio stimuli with faster tempi has been linked with increases in listeners’ levels of arousal (e.g Husain et al., 2002; Kellaris

& Kent, 1993). Furthermore, studies have shown that arousal can enhance memory by modifying the attention during the encoding process (Christianson

& Loftus, 1991), and that increased arousal supports slower forgetting (Sharot

& Phelps, 2004). It is thus reasonable to imagine that composers and producers might use tempo to increase the listener’s arousal, and thus increase attention and memorability. Hence, we hypothesize that the average tempo will increase over the studied time period.

(3) Time before the voice enters: The human voice has long been thought to draw attention. When designing functional music for the workspace, Muzak famously omitted the vocal parts and lyrics from their arrangements of popular songs as they believed that those components would draw attention to themselves (Radano, 1989). Moreover, Allan (2006) examined the eﬀect of popular music in advertising and showed that ads featuring vocal music were more ef- fective than ads featuring instrumental music on both attention and memory.

Consequently, since vocal music can help grab listeners’ attention, we hypothesize that the average time elapsed before the voice enters will decrease over the studied time period.

(4) Time before the title is mentioned: A particularly important aspect of popular music is the so-called “hook.” As aptly pointed out by Traut (2005),

37 the very concept of hook—while ubiquitous in popular music discourses—is highly problematic for scholars, since what “hooks” one listener might not “hook” another listener. Delson and Hurst (1980) define the hook as the recurring part, sometimes the title of a piece—also known as a “title hook” (Traut, 2005). While other types of hook exist (see Burns, 1987), the title of a piece is a useful way to operationalize—at least partially—the subjective concept of hook. Hence, for this study, we will operationalize the hook as the title hook, and measure the time before the title of a song is mentioned as an indicator of the first occurrence of the hook. Kaneshiro and Baker (2016) showed that Shazam users more commonly use the app during the first occurrence of the chorus, suggesting that this part caught their attention sufficiently for them to seek more information about the music. Since the title hook is usually located within the chorus or at the end of the preceding section (Davis, 1985), it can act as an attention- grabbing device. We thus hypothesize that the average time elapsed from the beginning of the track to the first appearance of the song title will decrease over the studied time period.

(5) Self-focused lyrical content: DeWall et al. (2011) conducted linguistic analyses of the most popular songs from 1980 to 2007. Their ﬁndings suggest that lyrical content of popular music changed over the studied time period in a way that mirrors psychological changes in U.S. society. Among other things, the authors observed an increase of self-focused and negatively valenced words, paired with a decrease of positively valenced words. Considering that several studies indicate that individualistic traits have increased in the U.S. population in recent decades (e.g. Roberts & Helson, 1997; Twenge & Campbell, 2001,

38 2008; Twenge & Foster, 2010; Twenge et al., 2008), and that people scoring

high in private self-consciousness scale tend to recognize self-relevant words

more quickly (Eichstaedt & Silvia, 2003), it is reasonable to anticipate that

songs featuring self-focused lyrical content will draw self-focused attention from

the listeners. As such, we hypothesize that lyrics will become more self-focused

over the studied time period.

3.2.2 Sample

A sample covering 30 years of music from 1986 to 2015 was assembled using Billboard

Magazine’s Year-End Hot 100 charts. For each selected year, the top-10 singles were sampled. Three of those singles were released as “double A-side,” creating a total sample of 303 entries. A list of the 303 songs included in this sample is available in

Appendix A.

3.2.3 Method

The number of words in the title was tabulated using the titles provided by the Bill- board charts. If part of the title was enclosed by parentheses, this part was not tabulated (e.g. “Stronger (What Doesn’t Kill You)” was reduced to “Stronger”). Two songs had titles based on acronyms (e.g. “E.T.”), and as such were encoded as missing data. Main tempo was calculated using a “tap along” method, which has been used previously in other musicological studies (e.g. Cook, 1995). One song was encoded as missing data since it oscillated between two diﬀerent tempi. Timestamps were manually calculated using spectrographic representation of the soundwaves generated by

39 Sonic Visualiser (Cannam et al., 2010). Five songs did not feature the title in the lyrics, and as such were encoded as missing data. Lyrical content analyses were performed using the LIWC2015 software, a popular program that tabulates statistics in a body of text based on curated dictionaries (Pennebaker et al., 2015). The same software was used by DeWall et al. (2011) to conduct linguistic analyses of popular songs. Following their methodology, self-focus was evaluated using the first person singular pronoun category in LIWC2015. This category consists of a curated dictionary containing 24 words (e.g. I, me, mine), with a corrected alpha score of .81, using the Spearman-Brown prediction formula (Pennebaker et al., 2015). Lyrics were collected from the web and manually curated to match the audio files. LIWC2015 was used to calculate the presence of first-person pronouns. Specifically, the self- focus score of a text represents the ratio of words in the text also present in the

ﬁrst person singular pronoun dictionary. For sure, the analytical values computed by

LIWC2015—just like those computed by virtually all text analysis programs using a similar methodology—are limited. For example, Mehl et al. (2012) have argued that psychological correlates of words are dependent upon the context in which they are used. Nevertheless, automated text analysis oﬀers a practical and objective way to operationalize complex constructs. One song was sung in a language other than

English, and as such was encoded as missing data.

3.2.4 Results

The results are summarized in Figure 3.1. A signiﬁcant correlation was observed for four of the ﬁve studied parameters discussed above: number of words in the title (r

40 Figure 3.1. Relationship between the year a song appeared on the Billboard Year- End Hot 100 chart and a) the number of words in its title (r = -.27, p < .001), b) its main tempo, in BPM (r = .18, p = .002), c) the elapsed time before the voice enters, in seconds (r = -.36, p < .001), and d) the elapsed time before the title of the song is mentioned, in seconds (r = -.18, p < .001). For Figure 1a, dots of varying sizes have been used to illustrate overlapping data points, as indicated by the graph’s legend. These tests remain signiﬁcant after a Bonferroni adjusted alpha level of .01 per test (.05/5). Not shown is the lack of correlation between the year a song was popular and its self-focus score (r = .01, p = .897). The results are consistent with the attention economy principles.

41 = -.27, p < .001), main tempo (r = .18, p = .002), time before the voice enters (r =

-.36, p < .001), and time before the title is mentioned (r = -.18, p < .001). These

tests remain signiﬁcant after a Bonferroni adjusted alpha level of .01 per test (.05/5).

However, no signiﬁcant correlation was observed between the year a song was popular

and its self-focus score (r = .01, p = .897). In general, the results are consistent with the view that popular music compositional practices have evolved in a way that favors attention grabbing.

3.2.5 Discussion

Technological changes in recent decades have had an important impact on almost

every aspect of our lives, including the way we enjoy music. However, the results

presented above suggest that this paradigmatic change is not limited to the way we

consume popular music, but has also aﬀected the way musicians create music. A

post-hoc linear regression analysis was conducted to evaluate the overall ﬁt of the

model. A signiﬁcant regression model was found, F (5, 288) = 16.15, p < .001, with

an adjusted R2 of .21. The eﬀect size is surprising, given the number of factors we

might expect to inﬂuence the popularity of a song (e.g. the quality of the song, the

instrumentation, the popularity of the artist, the image and media presence). As a

matter of fact, many songs in the present corpus were successful without conforming

to the attentional principles described earlier. For example, Gotye’s 2012 number-one

hit “Somebody That I Used to Know” features a title almost three times longer than

the other top-10 songs from that year, a 20-second instrumental introduction (roughly

four times longer than the average other hits from 2012), and the listener must wait

42 two whole minutes before hearing the singer mention the title of the song (the average

for the other 2012 hit songs is 37 seconds). Nevertheless, while not exceptional, these

examples are rare.

While four of the ﬁve variables behave in the predicted way, the strengths of

relationship were modest, with observed r values ranging between ±.18 and ±.36.

As visually illustrated in Figure 3.1, this suggests a lot of variability in the data. A limitation to the present corpus is that it provides no indication regarding musical style. Considering that the Billboard charts are diverse in terms of musical style, and that the time period under study was punctuated by periods of rapid musical changes (Mauch et al., 2015), it is possible that most or all of the variability observed is attributable to just one or two substyles, with other styles not being affected by attention economy principles. Nevertheless, even without taking musical style into consideration, we still observe a low but significant relationship for four of the five studied parameters (number of words in the title, main tempo, time before the voice enters, and time before the title is mentioned).

No correlation was found between self-focused lyrics and the year a song charted.

This is surprising, considering that DeWall et al. (2011) reported positive association using a very similar sample (the 10 most popular U.S. songs between 1980 and 2007, according to the Billboard Hot 100 Year-End chart) and using a very similar methodology (a more recent version of the LIWC software was used in the present study).

Table 3.1 compares the ﬁrst-person singular pronoun usage data from the current study with the data reported in DeWall et al. (2011, p. 3, Table 1). The overlapping time period (i.e. 1986–2007) is highly correlated (r = .86), which suggests that our data are similarly reliable. Based on the data reported in DeWall et al., it appears

43 that the moderately low score for a single, high-leverage year (1980) is responsible for the significant results. This is not to say their results are invalid. Without investigating lyrics from the 1970s, it is difficult to assess whether the year 1980 is an outlier or if it is representative of a larger trend. Nevertheless, it seems surprising that a sample ranging from 1986 to 2015 would produce such different results than a sample ranging from 1980 to 2007.

A closer look at the results suggests other trends in compositional practices. Table

3.2 presents a correlation matrix for the ﬁve studied variables. Of particular musical interest is the moderate negative correlation between the tempo of a song and the self-focus quality of its lyrics. This suggests that slower songs such as ballads would tend to be more self-focused or personal, while up-tempo songs such as dance music would be less self-focused. This makes sense from a musical perspective, as we tend to view ballads as more introspective than up-tempo dance music.

3.3 Study 2: Within-artist comparisons of success

The results of the ﬁrst study suggest that popular music changed in the last 30 years in a way that is consistent with attention economy principles. However, it is unknown whether this phenomenon is limited to the most popular songs (e.g. top-10 songs), or can be generalized to popular music in general. Thus, the aim of our second study is to investigate whether more popular songs exhibit a greater aﬃnity for attention economy principles than less popular songs.

44 Table 3.1

Comparison between ﬁrst-person singular pronoun usage in the present study and in DeWall et al. (2011). Year Study #1 (mean) DeWall et al. (2011) (mean) 1980 6.88 1981 9.39 1982 8.62 1983 5.92 1984 7.51 1985 9.70 1986 8.48 8.46 1987 5.72 6.40 1988 9.01 8.92 1989 11.64 11.04 1990 6.35 6.59 1991 10.64 10.28 1992 10.06 9.63 1993 10.48 10.78 1994 10.73 12.83 1995 9.01 8.69 1996 11.68 10.23 1997 11.89 12.49 1998 11.30 9.61 1999 11.10 8.68 2000 10.99 8.88 2001 12.33 12.50 2002 10.44 10.38 2003 12.80 13.03 2004 8.86 8.72 2005 10.00 9.50 2006 7.51 7.35 2007 9.02 9.18 2008 9.20 2009 7.45 2010 9.70 2011 8.54 2012 9.59 2013 8.31 2014 10.90 2015 10.90

45 Table 3.2

Pearson correlation matrix for ﬁve variables. Variable name Words in title Main tempo Time before voice enters Time before title is mentioned Self-focus in lyrics 46 Words in title 1.00 Main tempo -0.09 1.00 Time before voice enters 0.10 <0.01 1.00 Time before title is mentioned 0.19 -0.10 0.28 1.00 Self-focus in lyrics 0.03 -0.25 -0.04 0.08 1.00 3.3.1 Hypotheses

If we postulate that compositional practices in popular music are inﬂuenced by the attention economy theory, it is reasonable to assume that the most popular songs take better advantage of these principles than the less popular songs. Using the variables identiﬁed in Study 1 and validated through a post-hoc linear regression model, we hypothesize the following:

(1) Number of words in title: We hypothesize that the most popular songs

will tend to have, on average, shorter titles than the less popular songs.

(2) Main tempo: We hypothesize that the most popular songs will tend to

have, on average, faster tempi than the less popular songs.

(3) Time before the voice enters: We hypothesize that the average time

elapsed before the voice enters will be shorter for the most popular songs than

for the less popular songs.

(4) Time before the title is mentioned: We hypothesize that the average

time elapsed before the title is heard will be shorter for the most popular songs

than for the less popular songs.

(5) Self-focused lyrical content: We hypothesize that the most popular

songs will tend to feature, on average, more self-focused lyrics than the less

popular songs.

47 3.3.2 Sample

The most-streamed songs on Spotify for 2015 were elected as representative candidates of the most popular songs. We shall refer to this ﬁrst group as “most popular songs.”

In considering what might be the less popular songs, there are several confounds that arise. One might choose songs that are very rarely played on Spotify. However, these songs might be rarely played for numerous reasons. For example, the music might be out of tune, or badly recorded, or badly arranged, or sung in an unpopular style. Ideally, our comparison group should be similar in all musical respects, with the exception of those factors related to attentional economy that are the subject of this study. One way of reducing these unwanted confounds is to select recordings by the same artist that are simply much less popular. Consequently, the most popular and less popular works are more likely to be matched for singer, tuning, recording quality, quality of arrangements, etc. As such, for every “most popular song,” we used the least streamed song by that same artist from the same album as a control, based on

Spotify’s streaming statistics. We will refer to this group as the “less popular songs.”

This sampling method allows us not only to control for genre, but also to control for the impact of an artist’s fame over the popularity of a song.

A potential caveat regarding Spotify’s streaming statistics needs to be acknowl- edged. As the data from Lamere (2014) shows, Spotify users have a high skipping rate. Jonze (2014) states that tracks need to be listened to for 30 seconds to count towards the total number of plays. Considering that one of our hypotheses relates to lyrics, one might wonder how much analytical weight should be given to the text if

48 listeners are likely to only partially hear it. This is a valid concern. However, consid-

ering we have no reason to believe that some songs are systematically more skipped

than others, this caveat should not aﬀect our data in any systematic way. In other

words, while Spotify’s play count methodology does introduce noise in our data, it

likely does so in a randomized way.

Some albums exist in both standard and “deluxe” or “extended” version. When

this is the case, the “deluxe” version was used to elect the “less popular” song. Some

popular songs are not part of an album, but are rather released as singles. Since

these songs can’t be paired with a “less popular” song using the same methodology,

they have been omitted from the sample. Similarly, songs only released as part of a

compilation of songs from diﬀerent artists (e.g. a movie soundtrack) have also been

omitted. Finally, each artist is only represented once in each group. Subsequent songs

by the same artist were omitted.

Using this methodology, 120 songs were included in our sample. The average

number of plays for the “most popular” group is 350,761,781 and 6,313,174 for the

“less popular” group.2 A list of the 120 songs included in this sample is available in

Appendix B.

3.3.3 Results

Five paired t-tests were conducted to evaluate whether the “most popular” and the

“less popular” songs from each album diﬀered signiﬁcantly from each other. The

results are presented in Table 3.3.

2Data collected on June 2–8, 2016.

49 Table 3.3

Paired t-test results comparing most popular and less popular songs according to ﬁve variables. Variable Name NM SD 99%CI tdfpCohen’s d Number of words in title Most popular 58 2.47 1.30 [-0.48, 0.91] 0.83 55 .413 0.11 Less popular 58 2.69 1.43 Main tempo (in BPM) Most popular 60 115.43 22.03 [-25.17, 0.59] -2.54 57 .014 0.33

50 Less popular 58 102.47 25.63 Time before the voice enters (in sec) Most popular 60 8.61 7.64 [-1.72, 9.42] 1.84 58 .071 0.24 Less popular 59 12.46 15.27 Time before the title is mentioned (in sec) Most popular 58 38.34 21.26 [-7.03, 17.39] 1.14 47 .260 0.16 Less popular 49 44.53 23.75 Self-focus in lyrical content Most popular 60 11.24 5.83 [-5.28, 0.45] -2.25 58 .029 .029 Less popular 59 8.91 5.43 Note. 99% CIs reflect the Bonferroni-adjusted alpha level of .01 per test (.05/5). Using a 95% confidence level, two variables were identified as significantly different between the two groups: the main tempo and the self-focus score of the lyrical content.

However, none of the variables remain statistically signiﬁcant with a Bonferroni- adjusted alpha level of .01 per test (.05/5).

3.3.4 Discussion

Recall that our hypotheses predicted that the most popular songs take better advantage of attention economy principles than the less popular songs. However, the data were not consistent with any of the hypotheses regarding the relationship between these principles and popularity.

3.4 General Discussion

The results of the ﬁrst study provide evidence suggesting a change in compositional practice in popular music over the last 30 years. This apparent change is consistent with attention economy principles. The results of the second study, however, failed to support any of the hypotheses regarding the relationship between attention economy principles within a comparison of popular and less popular music.

This capacity for popular music creators to adapt themselves—consciously or not—to a paradigm shift should not come as a surprise. Another much-discussed example is the so-called “loudness war,” a tendency for popular music engineers to increase the perceived loudness of a recording through dynamic compression. This tendency to increase Root-Mean-Square (RMS) levels since 1980, paired with a steady

51 decrease in average dynamic range (Vickers, 2010) is consistent with the attention economy theory described above.

Several other examples in recent history illustrate ways in which music makers— either artists or producers—had to adapt themselves to consumer practices. Perhaps the most obvious example is the so-called “3-minutes rule,” which stated that, in order to be commercially successful, a single had to have a maximum duration of 3 minutes. The 3-minutes rule originates from the “Drake Clock,” a structured time usage formula elaborated by Bill Drake and Ron Jacobs at KHJ where “[d]eejays, songs, even commercials, and news were designed to ﬁt the mold; if they did not, they were not aired” (Denisoﬀ, 1989, p. 84). This system eventually governed the entire radio programming in America (Osborne, 2012).

The rise in popularity of karaoke in Japan gives another example of how cultural practices can inﬂuence the music industry. Karaoke emerged in the 1970s and became widely popular during the late 1980s and 1990s, especially through small private rooms called “karaoke boxes.” During the 1990s, the number of multi-million seller singles in Japan increased signiﬁcantly compared to the preceding decade. Ogawa

(1998) suggests that the growing popularity of karaoke boxes is directly responsible for these multi-million sellers:

The songs often sung in karaoke boxes occupy the top 40. Now those who make the top 40 are not just passive listeners but the users of karaoke. This means that one of the important elements for hit songs changed from “good to listen to” to “good to sing.” (Ogawa, 1998, p. 49)

Ogawa gives the example of Tetsuya Komuro, a singer-songwriter and producer who composes songs speciﬁcally with karaoke singers in mind. Similarly, Cantopop composer Angel Lam uses “narrower melodic ranges, fewer diﬃcult leaps, and shorter

52 phrases in order to accommodate the amateur singers’ typically modest abilities”

(Katz, 2011, p. 468).

All the above examples—the loudness war, the “Drake Clock,” the karaoke boxes, and the attention economy principles in contemporary popular music—highlight the musicians’ and music producers’ impressive capacity to adapt themselves to their rapidly-changing environment. Today’s online streaming platforms create a highly saturated ecosystem that encourages a high level of competition for the listener’s attention. This chapter highlighted four main parameters that have changed in the last three decades in a way that is consistent with the proposed theory of attention economy: the number of words in song titles has decreased, the average tempo has increased, the time elapsed before the initial entry of the voice has shortened, and similarly the time before the title of a song is heard has also shortened. However, the data were not consistent with any of the hypotheses regarding the relationship between attention economy principles within a comparison of popular and less popular music. What remains to be demonstrated is whether these compositional changes are an eﬀective way to grab listeners’ attention.

53 Chapter 4: Testing the Inﬂuence of Attention Economy Principles, Musical Preferences, and Familiarity on Music Listening Choice Behavior

4.1 Introduction

Both scarce and valuable, attention is a currency constantly growing in importance.

In the introduction to his book on this topic, Tim Wu argues that “if the attention merchants were once primitive, one-man operations, the game of harvesting human attention and reselling it to advertisers has become a major part of our economy...Beginning with radio, each new medium would attain its commercial viability through the resale of what attention it could capture in exchange for ‘free’ content”

(2016, p. 6). In the music world, this attention economy model (e.g. Goldhaber, 1997;

Davenport & Beck, 2001; Léveillé Gauvin, 2017) is especially relevant with the emer- gence of free online streaming platforms such as Spotify, Pandora, and Apple Music.

To paraphrase Wu, a streaming platform like Spotify harvests listeners’ attention and redistributes it to advertisers and artists.

The rise in popularity of digital content in the last few decades has created a much more competitive environment for music makers. Spotify’s catalogue contains

54 roughly 30 million songs, each available for free with the simple click of a button.

Moreover, millions of these songs have never been played (Planas Rego, 2013). From an economic perspective, this indicates that the supply of songs is disproportionately big compared to the demand, making listener’s attention scarcer and thus even more valuable. The abundance of music available for online streaming has thus decreased the monetary value of music itself, encouraging producers to use popular songs as promotional vehicles for artists rather than only as cultural products. As such, in an economic model using human attention as its main currency, it may be useful to consider songs as self-advocating advertisements for artists (Huron, 1989).

Research by Datta, Knox, and Bronnenberg (2017) shows how consumers’ adoption of online streaming affects their general listening behavior. Their results suggest that, by providing music free of charge, online streaming services like Spotify affect both the quantity and the diversity of listeners’ music consumption. In other words, since users of streaming services are not financially penalized for experimenting with new music, they are willing to take risks that they wouldn’t take if they had to pay for every song they listen to. An important corollary to this new music distribution model is that, since there is no financial cost for listening to a song, there can be no financial cost for skipping a song. Consequently, listeners have little incentive to listen to songs they do not like. Indeed, roughly one-third of the songs on Spotify are skipped after the first 20 seconds (Lamere, 2014).

Sustaining listeners’ attention and postponing the moment they skip to the next song is especially important for an online streaming platform like Spotify, where a song has to be listened to for 30 seconds or more to count as one play and collect royalties

55 (Jonze, 2014). In such a competitive environment, an attention economy theory predicts that artists will adapt themselves by making compositional decisions that make their songs more attention grabbing. Building on this conjecture, Léveillé Gauvin

(2017) investigated top-10 hits from 1986 to 2015 and observed that popular music has been changing in a way that is consistent with attention economy principles.

Specifically, he noted that song titles have become shorter, tempi have increased, instrumental introductions have become shorter, and the time before the title of a song is heard has also become shorter. Although this study suggests that popular music has changed in the last 30 years in a way that favors attention grabbing, it is unclear whether these attention-grabbing principles have an actual impact on music listening choice behavior. For example, it is possible that song producers and artists are engaged in an attention-war with one another, and that music listeners are simply witnessing this constant increase in attention-grabbing compositional practices with no real preferences one way or another. To this point, several pop songs have had considerable success in the last few years despite having long introductions or slower than usual tempi, suggesting that listeners might not weigh attention economy principles as heavily when they choose which song they want to listen to as music makers do when they make compositional decisions. For instance, it could be that artists put a lot of effort on making sure the hook of a song appears early on, but that listeners are more interested in timbre and instrumentation than when they hear the hook. Other elements are likely to be factored in when determining what prevents a listener from skipping to the next song. For example, music listeners tend to prefer certain genres and styles over others, meaning that musical preferences are likely to influence skipping behavior by listeners. Similarly, although people tend to say they

56 prefer novelty, research by Ward et al. (2014) suggest that song familiarity is as good, if not a better predictor of choice, as liking.

Another important factor to consider when analyzing music listening choice behavior is the listening setting. From a music perspective, divided attention to background music has been shown to inﬂuence choice behavior in a shopping context (for summary on this topic, see Renko & Gregur, 2017). It is unclear, however, how divided attention to music aﬀects music listening choice behavior itself.

In the present research, music listening choice behavior is examined in relation to attention economy principles, musical preferences, and familiarity. Speciﬁcally, the goal is to test whether attention economy principles (i.e. the number of words in title, the main tempo, the time before the voice enters, and the time before the title is heard), musical preferences, and familiarity can predict how long a listener will listen to a song before skipping to the next one.

Two experiments were conducted. In the first experiment, participants were asked to listen to randomly sampled songs, with no distraction task. This group is referred to as the undivided attention group. To anticipate our results, a significant multiple regression model accounting for 8.8% of the skipping behavior was found. In the second experiment, an independent group of participants was asked to listen to a subset of the same songs while participating in a distraction task. This group is referred to as the divided attention group. Again, a significant multiple regression model was found, explaining 16.7% of the skipping behavior.

57 4.2 Hypotheses

Formally, we may state our main hypothesis as follows:

H1 We hypothesize that songs conforming to attention economy principles (i.e.

short titles, fast tempi, short time before the voice enters, and short time before

the title is mentioned) will be listened to longer than songs that do not conform

to these principles.

According to the attention economy theory proposed in Léveillé Gauvin (2017), each of these parameters is expected to behave in a speciﬁc way. Speciﬁcally, we expect the following:

H1a We expect a negative relationship between the number of words in the

title of a song and the time elapsed before a song is skipped, such that songs

featuring shorter titles will be listened to longer, on average, than other songs.

H1b We expect a positive relationship between the main tempo of a song and

the time elapsed before a song is skipped, such that songs featuring faster tempi

will be listened to longer, on average, than other songs.

H1c We expect a negative relationship between the time elapsed before the voice

enters and the time elapsed before a song is skipped, such that songs featuring

shorter instrumental introductions will be listened to longer, on average, than

other songs.

H1d We expect a negative relationship between the time elapsed before the

title of a song is mentioned and the time elapsed before a song is skipped, such

58 that songs mentioning the title of a song earlier will be listened to longer, on

average, than other songs.

In addition to attention economy principles, we predict that musical preferences might influence skipping behavior, such that songs performed in a genre corresponding to a listener’s preference will be listened to for a longer time period than songs in other genres. Since the musical preferences of our participants are diverse and since the songs in our music sample cover different genres, we do not make any predictions on the way musical preferences might influence skipping behavior, but rather simply predict that they will affect listening times. Formally:

H2 We hypothesize that listeners’ musical preferences will predict the time

elapsed before a participant skips to the following song in queue.

Finally, we also expect familiarity with a song to inﬂuence skipping behavior.

Considering the power of familiarity on music choices (Ward et al., 2014), we expect familiar songs to be preferred to unfamiliar songs. Speciﬁcally:

H3 We hypothesize that familiar songs will be listened to for a longer time than

unfamiliar songs.

4.3 Experiment 1: Undivided Attention Music Listening Set- ting

This ﬁrst experiment aims to investigate the relationship between attention economy

principles, musical preferences, and familiarity on music listening choice behavior

when participants are paying undivided attention to music. Following the approach

59 used by Parkin et al. (1990) and others, we broadly make a distinction between

divided and undivided attention by deﬁning undivided attention as accomplishing a

task without conducting any concurrent activities.

4.3.1 Method

4.3.1.1 Participants

After receiving IRB approval, a total of 23 participants were recruited for this

study, all second-year undergraduate music majors at the Ohio State University who

participated for course credits. The age of the participants ranged from 19 to 24

years (M = 20.13, SD = 1.10). The participants’ musical training ranged from 2 to

15 years of regular instrumental or vocal practice (M = 6.32, SD = 3.29).

4.3.1.2 Material

A randomly sampled subset of 25 songs was assembled from the sample used in

Léveillé Gauvin (2017). The original sample covered top-10 singles for 30 years of music from 1986 to 2015, using Billboard Magazine’s Year-End Hot 100 charts.

In addition to tracking participants’ music listening behavior (i.e. how long it took them to skip to the next song), music preferences were assessed with the Short Test of Music Preferences (STOMP) (Rentfrow & Gosling, 2003). This test uses a 7-point

Likert-type scale to assess the extent to which participants like each of the following

14 diﬀerent musical genres: alternative, blues, classical, country, electronic, folk, heavy metal, jazz, pop, rap, religious, rock, soul and soundtracks. These 14 genres load onto four dimensions: Reﬂective & Complex, Intense & Rebellious, Upbeat &

60 Conventional, and Energetic & Rhythmic. Each dimension is represented by a score ranging between 1 and 7. Familiarity was coded categorically as ’YES’ or ’NO.’

4.3.2 Procedure

After providing informed consent, participants sat at a computer terminal in a soundproofed booth and listened to stimuli through headphones.

Participants received the following instructions:

The purpose of this study is to examine your musical preferences while listening to music. At the end of the experiment, you will be debriefed about our speciﬁc goals, but for now it is best that you understand that we are simply interested in the way listeners use the ’skip’ button when listening to a new playlist. We have created a playlist of several popular songs from the last four decades. You will hear the songs in a randomized order. The title and artist of the song currently playing will be displayed on the screen. At any time during the song, you can use the spacebar as a ’skip’ button to move to the next song in the playlist. Try to use the skip button in the same way you would if you were listening to a new playlist on a music streaming platform (e.g. Spotify). After each sound example, you will be asked to specify whether you were familiar with the song you just heard and to brieﬂy explain what made you skip to the next song in the playlist. Feel free to talk about anything at all, but try to keep your answer brief. After listening to all examples, you will be asked to answer short demographic questions about your age, music training, and musical preferences.

After a practice trial, each participant heard a total of 25 songs in a randomized order. The artist and title of each song was printed on the screen as the song was playing, but the upcoming songs were unknown to the participant. Participants were invited to use the spacebar to skip to the following song on the playlist. After using the spacebar to skip to the next song in queue, participants were asked to indicate

61 whether they were familiar with the song they heard, and to brieﬂy explain how (or

if) they used the skip button. The time between the beginning of the audio ﬁle and

the moment a participant pressed the spacebar was calculated and was considered to

be the variable of interest to the study.

After completing these tasks, participants were asked to complete the Short Test

of Music Preferences (Rentfrow & Gosling, 2003). Once the entire experiment was

completed, participants were thanked for participating in the study and were de-

briefed.

4.3.3 Results

The average listening time in the undivided setting was 71.81 seconds (M = 71.81,

SD = 79.27). The data were analyzed using a multiple regression model, as shown in Table 4.1. The model explains 8.8% of the skipping behavior, F (9, 565) = 7.13, p

< .001, R2adj = .088. Speciﬁcally, three variables played a signiﬁcant role: number

of words in title, time before the voice enters, and familiarity with the song. The

time elapsed before the voice enters, however, contributed to the model in opposite

direction of our predictions, such that songs with longer introductions tended to be

listened to for a longer period.

A visual representation of the participants’ skipping times in the undivided atten-

tion setting superimposed onto each song’s formal diagram is available in Appendix

62 Table 4.1

Summary of multiple regression for undivided attention. Variable Name BSEBt p F dfp R2 R2adj 7.14 9, 565 > .001 .102 .088 Number of words in title -5.88 2.46 -2.39 .017

63 Main tempo (in BPM) 0.19 0.16 1.19 .235 Time before the voice enters (in sec) 1.06 0.28 3.73 >.001 Time before the title is mentioned (in sec) 0.11 0.14 0.78 .435 Reﬂective & Complex 5.25 3.92 1.34 .182 Intense & Rebellious 1.81 2.74 0.66 .510 Upbeat & Conventional 3.13 3.00 1.04 .298 Energetic & Rhythmic 4.77 3.12 1.53 .127 Familiarity 21.27 7.00 3.04 .003 4.3.4 Discussion

Figure 4.1 uses the approach proposed by Kastellec and Leoni (2007) to visually illustrate the information presented in Table 4.1. Predictors are displayed on the y- axis, while skipping times (in seconds) are displayed on the x-axis. Each dot represents the beta coeﬃcient associated with a predictor, so that the vertical placement of the dots indicates the magnitude of the eﬀect size. Dots on the left side of the dashed line indicate that a predictor is negatively related to the time before skipping to the next song. Conversely, dots on the right side of the dashed line indicate that a predictor is positively related to listening times. Horizontal lines indicate

95% conﬁdence intervals. A line overlapping with the dashed line indicates that the predictor is not statistically signiﬁcant.

The results of the undivided attention condition were generally consistent with our proposed hypotheses. Speciﬁcally, the following variables contributed signiﬁcantly to the model: number of words in title time before the voice enters, and familiarity.

The model explained 8.8% of the variance in skipping behavior. As shown on Figure

4.1, familiarity with a song had the biggest impact on skipping behavior, with songs familiar to participants being listened to for an extra 21 seconds, on average. Famil- iarity was also the predictor with the largest variability, with a standard error of 7 seconds. Shorter titles also encouraged longer listening times, with each extra word reducing listening times by 6 seconds on average.

While these results were expected, others were surprising. The tempo of a song and the time before the title is mentioned did not statistically inﬂuence music listening choice behavior. Perhaps more surprising, musical preferences did not predict

64 Words in Title

Main Tempo

Time Before Voice

Time Before Title

Reflective & Complex

Intense & Rebellious

Upbeat & Conventional

Energetic & Rhythmic

Familiarity

−12−9−6−30369121518212427303336 Time Before Skip (in sec)

Figure 4.1. A visual summary of the multiple regression for undivided attention, F (9, 565) = 7.13, p < .001, R2adj = .09. The black dots represent the beta values, the horizontal lines represent +1 SE and -1 SE. If the standard error bars cross vertical line at x=0, it means that the beta value is not significantly different from zero. Three predictors are statistically significant in predicting listening times: the number of words in the title, the amount of time before the voice enters, and whether or not a person is familiar with the song. For every additional word in a song title, listening times shortened by 5.88 seconds, on average (SE = 3.92). For each additional second of instrumental introduction, listening times increased, on average, by 1.06 seconds (SE = 0.28). Songs familiar to participants were listened for an extra 21.27 seconds, on average (SE = 7.01).

65 skipping times. The most surprising result arising from Experiment 1, however, is that long instrumental introductions yielded longer listening times. As shown in

Figure 4.1, for each additional second of instrumental introduction, listening times increased, on average, by 1 second. In other words, songs with instrumental introductions did not get penalized by being skipped earlier than other songs. This was especially unexpected considering how counterintuitive these results are compared to general attention economy principles. The shortening of musical introductions was the biggest eﬀect observed by Léveillé Gauvin (2017), averaging 23 seconds in the mid-1980s to 5 seconds by the mid-2010s. This represents a 78% decrease in duration over thirty years.

A possible explanation for this behavior is that listeners might wait to hear the voice of the main singer before deciding whether they want to skip or not. Under this conjecture, the voice of a singer could be a make-or-break characteristic of a song, one that weighs heavily in a listener’s decision. When asked to brieﬂy explain why they skipped the song they heard, 14.9% of the responses mentioned disliking the quality of the vocals. Disliking the voice of the singer was the third most common reason given by participants after disliking the song in general (23.9% of the responses) and losing interest (21.3%). Other responses included liking it but wanting to move on

(10.8%), disliking this speciﬁc style (9.0%), the music sounded dated (5.6%), not being in the mood for this song at the moment (5.0%), disliking the sound of the instruments (3.6%), the music being too slow (2.8%), waiting to hear the chorus

(1.6%), autobiographical reasons (1.6%). These results are summarized in Table 4.2.

Alternatively, another possible explanation for the surprising way instrumental introductions inﬂuenced music listening choice behavior is that these results are due

66 Table 4.2

Summary of participants’ responses in the undivided attention condition to the prompt “Please brieﬂy explain why you skipped the previous song you heard.” Category Response % Valid Records (n = 575; Example of Participant Response Records valid responses = 502; NA Coded to =73) Category

67 Didn’t like it 120 23.9 Hate this song now; way over played. Lost interest 107 21.3 I enjoyed the song but I lost interest after a while. Didn’t like the voice 75 14.9 I liked the intro but not the high-pitched singing. Liked it 54 10.8 I really like this song, I would listen to it again. Don’t like this style 45 9.0 I do not typically enjoy rap/hip-hop. It sounded dated 28 5.6 Sounded like a song on an 80s-movie soundtrack. Not in the mood 25 5.0 I like her voice I just wasn’t in the mindset for this song. Didn’t like how the instruments sounded 18 3.6 I don’t like the keyboard synth sound; why was that ever a thing? Too slow 14 2.8 I don’t like slow songs. Personal reasons 8 1.6 I skipped it because it reminds me of my 6th grade dances and I just really do not want to relive that right now. Skipped after the chorus 8 1.6 Just wanted to hear the chorus. to demand characteristics. While none of the participants mentioned this explicitly in the post-experiment interview, it is possible that the artificial experimental environment does not aptly reflect the typical listening environment of most users of online music streaming platforms, and that, since participants were asked to pay undivided attention to the music, they felt compelled to wait for the singer’s voice before skipping to the next song. Indeed, in everyday life, the vast majority of music listening is done while doing another activity such as driving, working, doing chores, exercising, surfing the web, reading, etc. (Nielsen Company, 2015).

In order to address these potential confounds, a second experiment was conducted with an alternative design in hope of creating a more ecologically-valid environment for participants. Speciﬁcally, in the second experiment, participants were asked to listen to a music playlist while completing a distraction task. This listening context is referred to as divided attention. Moreover, this second experiment was time-limited, in an attempt to standardize the attention budget of each participant.

4.4 Experiment 2: Divided Attention Music Listening Setting

A second experiment paralleling the first experiment was designed to test music listening choice behavior in an ecologically-valid, divided attention context. A distraction task was designed to be completed by the participants while listening to music. The distraction task was meant to be ecologically valid (i.e. something that people might do while listening to music). Since different levels of complexity demand for different optimal levels of arousal (Yerkes & Dodson, 1908), the distraction task should ideally have a constant level of complexity. Considering that the average person spends more

68 than two hours a day on social media and messaging (GlobalWebIndex Company,

2017), asking participants to browse through a social media platform while listening

to music is an ecologically-valid approach. Speciﬁcally, participants were invited to

browse the Instagram social platform, using an account created speciﬁcally for this

experiment. Instagram was chosen over other social media platforms (e.g. Twitter,

Facebook) because its content is mostly visual (i.e. no long texts or articles), which

presumably keeps the complexity level more or less constant. The account’s feed was

populated with images from 150 followed accounts. Considering that the content of

the images could inﬂuence the emotional state of the participants, an eﬀort was made

to populate the Instagram feed with a variety of image types (e.g. food, animals,

traveling, celebrities).

4.4.1 Method

4.4.1.1 Participants

After receiving IRB approval, a total of 26 new participants were recruited for

this study, all second-year undergraduate music majors at the Ohio State University

who participated for course credits. The age of participants ranged from 19 to 25

years (M = 19.73, SD = 1.43). The participants’ musical training ranged from 1 to

15 years of regular instrumental or vocal practice (M = 6.27, SD = 3.74).

4.4.1.2 Material

A subset of 20 songs was randomly sampled from the 25 songs used in the ﬁrst study. The duration of all 20 songs combined was less than 90 minutes, so that even without skipping any songs, each participant would hear every single song. In

69 addition, another subset of 100 unique songs was assembled from the full sample used in Léveillé Gauvin (2017).

Again, participants were asked to complete the Short Test of Music Preferences

(STOMP) (Rentfrow & Gosling, 2003).

4.4.1.3 Procedure

After providing informed consent, participants sat in a soundproofed booth equipped with two computer terminals: one computer was used to listen to stimuli through headphones, and one computer was used to browse Instagram.

Participants received the following instructions:

The purpose of this study is to examine your musical preferences while listening to music. At the end of the experiment, you will be debriefed about our speciﬁc goals, but for now it is best that you understand that we are simply interested in the way listeners use the ‘skip’ button when listening to a new playlist. We have created an Instagram account for this experiment. For the duration of the experiment, we ask that you browse Instagram while listening to music. You are free to use the search bar to look for speciﬁc hashtags, but we ask that you don’t sign in into your own account, nor that you visit your friends’ accounts. While browsing Instagram you will be listening to some music. We have created a playlist of several popular songs from the last four decades. You will hear the songs in a randomized order. The title and artist of the song currently playing will be displayed on the screen. At any time during the song, you can use the spacebar as a ‘skip’ button to move to the next song in the playlist. Try to use the skip button in the same way you would if you were listening to a new playlist on a music streaming platform (e.g. Spotify). After each sound example, you will be asked to specify whether you were familiar with the song you just heard. The playlist will play for 90 minutes. After that time, you will be asked to answer short demographic questions about your age, music training, and musical preferences.

70 While in the ﬁrst experiment participants listened to a ﬁxed number of songs, this

second experiment was designed to last for a ﬁxed time period. As such, the audio

interface consisted of two blocks of songs: Block A, consisting of a ﬁxed set of 20 songs

randomly sampled from the 25 songs used in the ﬁrst study, in a randomized order,

and Block B, consisting of an additional 100 songs, also presented in a randomized

order. Block B was added to make sure that participants did not run out of songs

before the allotted time. Every participant heard Block A before Block B. Only

responses from Block A were analyzed.

Again, the artist and title of each song was printed on the screen as the song

was playing, but the upcoming songs were unknown to the participant. Just like

in the ﬁrst experiment, participants were invited to use the spacebar to skip to the

following song on the playlist. After using the spacebar to skip the next song in

queue, participants were asked to indicate whether they were familiar with the song

they just heard. For time considerations, participants were not asked to explain why

they used the skip button. The time between the beginning of the audio ﬁle and the

moment a participant pressed the spacebar was the variable of interest.

4.4.2 Results

The average listening time in the divided setting was 116.56 seconds (M = 116.56,

SD = 97.60). Just like in the ﬁrst experiment, the data were analyzed using a multiple regression model, as summarized in Table 4.3. The model explains 16.7% of the skipping behavior, F (9, 506) = 12.46, p < .001, R2adj = .167. Speciﬁcally, four

71 Table 4.3

Summary of multiple regression for divided attention. Variable Name BSEBt p F dfp R2 R2adj 12.46 9, 506 > .001 .181 .167 Number of words in title 5.16 3.51 1.469 .142 Main tempo (in BPM) -0.66 0.15 -0.430 .667 Time before the voice enters (in sec) 1.42 0.34 4.164 >.001 Time before the title is mentioned (in sec) 0.03 0.17 0.161 .872 Reﬂective & Complex 11.19 6.53 1.714 .087 Intense & Rebellious -1.47 4.18 -0.351 .726 Upbeat & Conventional 12.11 5.54 2.184 .030 Energetic & Rhythmic 15.61 3.92 3.981 >.001 Familiarity 21.90 8.54 2.566 .011

variables played a signiﬁcant role: time before the voice enters, two out of four musical preferences dimensions (‘Upbeat & Conventional’ and ‘Energetic & Rhythmic’) and familiarity with the song. However, just like in the undivided attention model presented in Experiment 1, the time elapsed before the voice enters contributed in opposite direction to our predictions, such that songs with longer introductions tended to be listened to for a longer period.

Once again, a visual representation of the participants’ skipping times in the divided attention setting superimposed onto each song’s formal diagram is available in Appendix C.

4.4.3 Discussion

Our multiple regression model for music listening choice behavior in a divided attention setting explained 16.7% of the variance in skipping behavior. Generally, the results of this second experiment replicated the results of the ﬁrst study, as illustrated

72 in Figure 4.2. Familiarity remained the most important predictor, with almost ex- actly the same eﬀect size. Similarly, the change in the experimental design did not change the direction of the association between the time before the voice comes in and skipping times, suggesting that this association was not due to demand characteristics.

However, two important elements differ from the first experiment. First, although shorter titles encouraged longer listening times in the undivided attention setting, this effect disappeared when participants were completing a distraction task. While the title of the song was still available to them, it is possible that participants did not consider it as relevant information, or simply failed to pay attention to it. Second, whereas musical preferences did not influence skipping behavior in the divided attention setting, their influence on listening times in the divided attention setting were considerable. Specifically, for every point increase in the “Energetic & Rhythmic” music preferences dimension (i.e. Electronic, Hip-hop/Rap, Soul), listening times increased by 16 seconds, on average. Similarly, for every point increase in the “Up- beat & Conventional” music dimension (i.e. Country, Pop, Religious, and Sound

Tracks), listening times increased by 12 seconds, on average. Scores in the “Reﬂective

& Complex” (i.e. Blues, Classical, Folk, and Jazz) and “Intense & Rebellious” (i.e.

Alternative, Heavy Metal, and Rock) music preferences dimensions did not inﬂuence music listening choice behavior. This is perhaps less surprising considering the nature of our sample, and how unlikely these genres are to be represented on the Billboard charts.

73 Words in Title

Main Tempo

Time Before Voice

Time Before Title

Reflective & Complex

Intense & Rebellious

Upbeat & Conventional

Energetic & Rhythmic

Familiarity

−12−9−6−3036912151821242730333639 Time Before Skip (in sec)

Figure 4.2. Multiple regression for divided attention, F (9, 506) = 12.46, p < .001, R2adj = .17. Four predictors are statistically signiﬁcant in predicting listening times: the amount of time before the voice enters, two out of four musical preferences dimensions (‘Upbeat & Conventional’ and ‘Energetic and Rhythmic’) and whether or not a person was familiar with the song. For each additional second of instrumental introduction, listening times increased, on average, by 1.42 seconds (SE = 0.34). The inﬂuence of musical preferences on listening times in the divided attention setting were considerable. For every point increase in the ‘Energetic & Rhythmic’ music preferences dimension (i.e. Electronic, Hip-hop/Rap, Soul), listening times increased by 15.61 seconds, on average (SE = 3.92). Similarly, for every point increase in the ‘Upbeat & Conventional’ music dimension (i.e. Country, Pop, Religious, and Sound Tracks), listening times increased by 12.10 seconds, on average (SE = 5.54). Fa- miliarity remained the most important predictor, with songs familiar to participants being listened to for an extra 21.90 seconds, on average (SE = 8.54).

74 4.5 General Discussion

Based on the results of the two experiments presented above, we might formulate a

list of principles for popular music makers (e.g. artists, producers, etc.) aiming to

grab and sustain listeners’ attention:

1. Maximize how often your song is played to captive audiences.

In both listening settings tested in this study, familiar songs were listened to longer than unfamiliar songs. Past research by Schellenberg et al. (1999) found that participants are able to identify popular songs by name within a second. Since familiar songs tend to be listened to for a longer time period, increased exposure to captive (or semi-captive) audiences is likely to increase familiarity with listeners. One important group of captive listeners consists of car commuters. Americans spend, on average, 42 hours per year in congested traﬃc (INRIX Company, 2017). This number more than doubles for major cities like San Francisco, New York, and Los Angeles. Considering that 70% of the in-car listening time is devoted to AM/FM radio (Murphy, 2017), artists and producers wishing to grab and sustain listeners’ attention should thus aim to maximize how often their song is heard on broadcasting radio.

2. Favor shorter titles.

There has been a general trend in the last few decades for song titles to become shorter (Kopf, 2016; Léveillé Gauvin, 2017). A possible explanation for this phenomenon is the rising popularity of portable music players with small LCD displays such as the iPod, released in 2001. Since a limited number of characters can be ﬁt on such small displays, shorter titles prevent horizontal text scrolling.

75 In the present study, the number of words in a song title only predicted skipping

times when listeners were paying undivided attention to the music and were looking

directly at the title of the song as they are listening. This suggests that, in situations

where people listen to music while doing something else, having a short song title

might have little impact on listeners’ behavior. However, considering that shortening

song titles is cost-free and that it leads to longer listening times in an undivided

attention listening setting, music makers should favor shorter titles, acknowledging

that the eﬀect size might be modest.

3. Use stylistic markers as targeting tools.

In both experimental settings, the variance explained by attention economy principles and familiarity was of similar size. The biggest diﬀerence we observed came from musical preferences. Musical preferences explained an important part of the variance in skipping behavior, but only in the divided attention context. This suggests that listeners might be able to override their personal musical preferences when asked to pay undivided attention to the music, but that in an everyday setting, styles and genres can act as targeting tools. Past research has shown that participants can identify the style of a song within three seconds (Gjerdingen & Perrott, 2008; Plazak

& Huron, 2011). This suggests that, within the first five seconds or so of a song, listeners are likely to evaluate a song based on their familiarity with the song, and the style or genre to which the song belongs. For example, during the first five seconds of a song, a listener might identify a song as being rap music, and skip to the next song because they generally dislike this type of music. Note how this decision is not based on the characteristics of the song itself, but rather on prior experience with music from this style or genre. Recall that although people think they prefer novelty,

76 they actually do not (Ward et al., 2014). Thus, the beginning of a song should be

composed so that its style is representative of the overall style of the song in general.

Which leads to our next principle:

4. Make use of instrumental introductions.

An important change in compositional practice in the last few decades is the near disappearance of the instrumental introduction Léveillé Gauvin (2017), going from an average of 23 seconds in the mid-1980s to just ﬁve seconds in the mid-

2010s. Contrary to our intuitions, however, postponing the moment the voice enters increased listening times in both the divided and undivided attention settings. The instrumental introduction thus appears to be a conundrum. On the one hand, it lacks the attention-grabbing quality that the voice confers and that appears to be essential in popular music–the last purely instrumental song to reach the number-one position of the Billboard Hot 100 charts was the “Miami Vice Theme” on November

9, 1985. On the other hand, the instrumental introduction contrasts sufficiently with the rest of the song (in terms of instrumentation, but also often in terms of timbre and dynamics) that it can almost be thought of as a different song, a sort of prelude to the main piece. As such, it is possible that, when the voice enters, the contrast between the instrumental material and the new material is such that the listener’s attention span is, if not fully, at least partially restored. The results from both experiments presented in this study are consistent with this theory: every extra second added to an instrumental introduction yielded, on average, an extra second in listening time. Considering the 30-second cutoff to get royalties on an online streaming service like Spotify, if this theory is correct, then a successful strategy could be to have a long instrumental introduction that is interesting enough to keep

77 people engaged for 15-20 seconds, which then means you only need to hold listeners’

attention for 10-15 more seconds to reach the royalty cutoﬀ. Thus, for artists wishing

to increase listening times, eliminating the instrumental introduction might not be the

best strategy. Instead, artists and producers would do better keeping the instrumental

introduction, but making sure that it aptly represents the overall style or genre of the

song or artist.

5. Pay special attention to the vocals.

The voice of a singer could be a make-or-break characteristic for a song, one that weighs heavily in a listener’s decision. Nearly one participant in six in the undivided attention group said they skipped because they did not like the singer’s voice or the way the voice was produced. Thus, popular music makers aiming to grab and sustain listeners’ attention should pay special attention to the vocal track(s) when recording and producing a song. This is especially true for songs featuring longer instrumental introductions. Some listeners might wait to hear the voice of the singer before deciding whether they want to skip or not. If this is the case, making sure that the vocal production matches the listeners’ expectations becomes crucial.

4.6 Conclusion

The present research aimed to test the inﬂuence of attention economy principles, musi-

cal preferences, and familiarity on the skipping behavior of music listeners. Attention

economy principles were deﬁned using the operationalizations used in Léveillé Gau-

vin (2017): shorter titles, faster tempi, shorter time before the entry of the voice,

and shorter times before the title of a song is mentioned. Musical preferences were

78 assessed using the Short Test of Music Preferences (STOMP) (Rentfrow & Gosling,

2003). Familiarity with a song was encoded categorically as ‘YES’ or ‘NO.’ Two different listening contexts were tested independently: undivided and divided attention.

The results from these experiments suggest that music listening choice behavior is moderated by the listening context. The undivided attention model explained 8.8% of the variance in skipping behavior, a medium eﬀect size that “would be perceptible to the naked eye of a reasonably sensitive observer” (Cohen, 1988, p. 80). In comparison, the divided attention model explained 16.7% of the variance in skipping behavior, nearly doubling the eﬀect size.

Based on these results, ﬁve recommendations were made to music makers aiming to grab and sustain listeners’ attention: 1) Maximize how often your song is played to captive audiences, 2) Favor shorter titles, 3) Use stylistic markers as targeting tools,

4) Make use of instrumental introductions, and 5) Pay special attention to the vocals.

Of course, these recommendations are purely based on a mercantile perspective and do not take into consideration other aesthetic decisions that go into creating a song. Nor was it argued that songs following these principles are qualitatively superior than others. Rather, this chapter simply aimed to test the inﬂuence of attention economy principles, musical preferences, and familiarity on music listening choice behavior.

79 Chapter 5: General Summary

5.1 Recapitulation

This dissertation investigated popular music compositional practices and music lis-

tening choice behavior in studies that attempted to uncover whether technological

changes aﬀecting the music industry have had an impact on songwriting techniques,

and whether music listening choice behavior can be predicted by these compositional

devices. The underlying theory unifying this work is the theory of attention economy:

how listener attention can be analyzed using supply and demand principles to explain

and predict both compositional decisions and music listening choice behavior.

Chapter 2 argued that popular songs, in addition to being cultural products, act

as advertisements promoting an artist’s personal brand. A survey of the empirical

literature on popular music was presented, suggesting that technological changes and

changes in compositional practices can beneﬁt from being analyzed from the perspec-

tive of attention economy theory.

The ﬁrst study, presented in Chapter 3, used a corpus of 303 songs popular between

1986 and 2015 to investigate whether compositional practices have changed in a way

that favors attention-grabbing behavior. Five aspects of popular music–subsequently

80 referred to as attention economy principles–were evaluated: number of words in the title, main tempo, time before the voice enters, time before the title is mentioned, and self-focus in lyrical content. The results of this first study generally supported our hypotheses. A regression model explained 21% of the variance in compositional practices, with four of the five parameters (number of words in the title, main tempo, time before the voice enters, and time before the title is mentioned) significantly contributing to predict the year a song appeared on the Billboard charts. The second study presented in Chapter 3 aimed to investigate whether more popular songs exhibit a greater affinity for attention economy principles than less popular songs. One hundred twenty songs were included in this sample: 60 most popular songs and 60 less popular songs. The results were not statistically significant after correcting for multiple testing. However, all the results were in the predicted direction, suggesting that the study might have been statistically underpowered.

The aim of the two experiments presented in Chapter 4 was to behaviorally test whether attention economy principles, musical preferences, and familiarity can predict how long a listener will listen to a song before skipping to the next one. Two music listening settings were tested: undivided attention (i.e. without distraction task) and divided attention (i.e. with distraction task). The results suggest that music listening choice behavior is moderated by the listening context. The undivided attention model explained 9% of the variance in skipping behavior, while the divided attention model explained 17% of the variance in skipping behavior.

81 5.2 Implications for the Music Industry

The results presented in this dissertation–and more generally the attention economy theory–have important implications for the music industry, most notably regarding music recommendation. Recommender systems play a crucial role in the online streaming media industry. Major companies such as Netﬂix and Amazon rely on automated recommendation to suggest content to their clients that is likely to appeal to them. Similarly, online streaming platforms such as Last.fm and Spotify use recommender systems to help listeners navigate seamlessly through their immensely vast catalogue. As Celma aptly puts it, “we are moving towards the Hit vs. Niche paradigm, where there is a large enough availability of choice to satisfy even the most

Progressive-obscure-Spanish-metal fan. The problem, though, is to ﬁlter and present the right artists to the user, according to her musical taste” (2010, p. 87).

Song et al. (2012) present a succinct summary of current recommendation systems. Two main models are currently used, often in conjunction with one another: models based on users’ listening behavior, and content-based models. Models using user’s listening behavior typically rely on musical genres and user preferences. The results from Chapter 4 regarding the role musical preferences in predicting music listening choice behavior reinforce this idea. But other personal information can be used to improve recommendations systems. For example, Vigliensoni and Fujinaga

(2016) showed that recommendation accuracy can be improved by 8% by taking into consideration demographics information such as age, country, and gender. Further- more, by taking into consideration the user’s preference regarding ‘mainstreamness’ and ‘exploratoryness,’ the accuracy of the model improves by 12%. However, we can

82 imagine ways to further improve recommendation systems. The results from Chapter

4 suggest that listening settings mediate how different parameters influence music listening choice behavior. For example, musical preferences did not seem to influence music listening choice behavior in the undivided attention setting, but had a significant impact in the divided attention setting. Thus, recommendation systems should, ideally, take into consideration the listening setting of a user before making a recommendation. This can be accomplished in two different (sometimes complimentary) ways. First, if a query is vocally made through the use of an Intelligent Personal As- sistant (IPA) (e.g. Apple’s Siri, Amazon’s Alexa, Google Home), the recommendation system should try to predict the listening setting by evaluating the semantic content of the question asked. Second, considering the importance of establishing a strong user profile, a powerful way to evaluate the current listening setting (and improve music recommendation generally) would be to use web-crawling algorithms to gather personal information on users. For example, social media posts (e.g. Facebook, Twit- ter, Instagram), Internet search history (e.g. Google), buying history (e.g. Amazon, eBay), and biometric data (e.g. Fitbit, Apple Watch) could be combined to create highly-detailed user profiles. Specifically, nearly-continuous updates of these data could help to evaluate a user’s listening setting by making music recommendation matching their current geographical location, activity, emotional mood.

Content-based modelling can also be used to improve recommendation accuracy.

In such models, songs sharing similar acoustic features are considered more likely to be enjoyed by the same listeners than songs with different acoustic features. But content-based modelling could also be used at a song level. Imagine that, instead of sending WAV files to streaming platforms, record labels sent a project file containing

83 a modular version of each song. If this were the case, recommendation systems could recompose songs based on a user profile by reconfiguring the formal organization, shortening or lengthening some sections, changing the instrumentation, etc. This example might seem far-fetched at first, but this sort of targeting has been present in popular music for a long time. Artists regularly re-record songs in a foreign language to please specific markets, and songs are regularly edited for radio broadcasting.

Recommendation systems could similarly adapt the content of a song at a user- level, as opposed to a group-level. Obviously, such an approach raises serious ethical issues: What is the value of privacy? And how much of it are we willing to sacriﬁce to improve automated recommendation systems? These are diﬃcult but important questions that academics, researchers, philosophers, and ultimately legislators will have to address in the near future.

In addition to the implications for music recommendation described above, the present dissertation raises the question of what is the value of music. By moving away from physical entities such as records and brick-and-mortar record stores to a digital production and distribution of music, we effectively entered what Chris Anderson coined the era of “long tailed economics” (2006). Because brick-and-mortar retailers have limited shelf-space, and because shelf space is expensive, only a limited number of unique items can be carried. This means that, for every new release carried in store, an older, less popular item is removed from the inventory. In effect, limited shelf space artificially limited the demand for attention, keeping its value more or less stable. By shifting to a purely digital distribution model, online streaming companies essentially eliminated shelving costs, thus creating an ever increasing catalog of songs and, consequently, drove the demand for attention up.

84 The attention economy theory attributes a scarcity-based value to attention based on the demand. If the demand goes up, so the does the value of attention. And when the value of attention is high, the amount of attention a listener is willing to pay for a song decreases, which in turns decreases the monetary value of the song itself.

We’ve reached a point where the demand for attention is such that the monetary value a listener is willing to pay for a song is $0 (or approaches $0). If the demand for attention keeps rising (and we currently have no reason to believe it won’t), we must ask ourselves how will the online system adapt itself.

Current monetary policies may help us imagine what the ﬁnancial future of online streaming will look like. Traditionally, central banks have used interest rates to stimulate economic growth. In a recent attempt to prevent deﬂation and kick-start economic growth, the European Central Bank, along with central banks in Denmark,

Sweden, Switzerland, and Japan, have introduced negative interest rates, eﬀectively charging fees to banks wishing to sit on their money instead of lending it.

Applying the same logic to online streaming would mean that listeners would eﬀectively be compensated for listening to music online. Although a direct monetary compensation is unlikely–imagine receiving a check at the end of the month from

Spotify!–other types of compensation are more realistic. In 2015, the online streaming giant Spotify announced a partnership with Starbucks that allowed Spotify users to earn Starbucks Rewards stars–Starbucks’ reward program–for subscribing to Spotify

Premium. Based on the theories discussed in this dissertation, we predict that this type of partnership is likely to increase in future years.

85 5.3 Future Research

In addition to the suggestions already made above, there are many other avenues for future research on attention economy and music listening choice behavior. In this dissertation, I favored an empirical distant reading of popular music in the hope of discovering global patterns and trends. However, I believe that some of the principles and ﬁndings discussed could inform close readings of single pieces. In particular, the

Music Listening Choice Behavior (MLCB) model proposed in Section 2.5.1 of Chapter

2 could be applied to both phenomenological and analytical studies of popular music.

Much of the research on popular music, including the present work, is biased towards Anglo-popular music. Future research should look at popular music from different cultures and sung in different languages. The theories proposed in this document should be revised to take into consideration the specificities of other cultural groups. There are roughly 360 million native speakers of English worldwide. By comparison, there are roughly 9 million native speakers of Swedish. This suggests that the number of information products in English is likely to be much greater than those in Swedish. Recall the theoretical model proposed in Chapter 2 assigns a scarcity-based value to attention based on the difference between the supply of attention and the demand for attention. If there are more information products available in English, this should increase the value of attention in this specific market and thus accelerate changes based on the economy of attention. As such, I suspect that isolated cultural groups–either geographically or linguistically–should be more strongly immune to the economic pressure of the attention economy model than the

86 dominant cultural groups.3 This conjecture could be tested by replicating the ﬁrst study of Chapter 3 with several samples representing cultural minorities.

Finally, the attention economy theory discussed in this dissertation should be applied to the studies of other cultural products. While I focused on music, the same phenomenon should be observed in television, movies, and to a lesser extent in books.

More work is needed in that domain.

5.4 Conclusion

Musicologists often use socio-cultural phenomena to analyze and explain changes and trends in compositional practices. However, the important role of technology is often overlooked. With this work, I hope I highlighted how the technology we use to create, distribute, and enjoy music aﬀords certain behavior. To this end, I would like to close by quoting Marshall McLuhan:

In a culture like ours, long accustomed to splitting and dividing all things as a means of control, it is sometimes a bit of a shock to be reminded that, in operational and practical fact, the medium is the message. This is merely to say that the personal and social consequences of any medium– that is, of any extension of ourselves–result from the new scale that is introduced into our aﬀairs by each extension of ourselves, or by any new technology. (McLuhan, 1964/1994, p. 7)

3Obviously, this is an over-simplified example. Not only native speakers generate cultural products in a given language, and information products from isolated cultural groups are likely to also compete with those of the dominant group. Furthermore, there is probably a ceiling effect: I suspect that past a certain threshold, the number of information products available does not significantly affect the attention value. Nevertheless, this is still something to keep in mind when discussing attention economy.

87 Appendix A: List of the 303 songs analyzed in the ﬁrst study (Section 3.2) of Chapter 3

Included in this appendix is a list of the 303 songs analyzed in the ﬁrst study

(Section 3.2) of Chapter 3.

Year Position Title Artist 1986 1 That’s What Friends Are For Dionne Warwick & Friends 1986 2 Say You, Say Me Lionel Richie 1986 3 I Miss You Klymaxx 1986 4 On My Own Patti LaBelle and Michael McDonald 1986 5 Broken Wings Mr. Mister 1986 6 How Will I Know Whitney Houston 1986 7 Party All The Time Eddy Murphy 1986 8 Burning Heart Survivor 1986 9 Kyrie Mr. Mister 1986 10 Addicted to Love Robert Palmer 1987 1 Walk Like An Egyptian The Bangles 1987 2 Alone Heart 1987 3 Shake You Down Gregory Abbott 1987 4 I Wanna Dance With Whitney Houston Somebody (Who Loves Me) 1987 5 Nothing’s Gonna Stop Us Starship Now 1987 6 C’est La Vie Robbie Nevil 1987 7 Here I Go Again Whitesnake 1987 8 The Way It Is Bruce Hornsby and the Range 1987 9 Shakedown Bob Seger 1987 10 Livin’ on a Prayer Bon Jovi

88 1988 1 Faith George Michael 1988 2 Need You Tonight INXS 1988 3 Got My Mind Set On You George Harrison 1988 4 Never Gonna Give You Up Rick Astley 1988 5 Sweet Child O’ Mine Guns n’ Roses 1988 6 So Emotional Whitney Houston 1988 7 Heaven Is a Place on Earth Belinda Carlisle 1988 8 Could’ve Been Tiﬀany 1988 9 Hands to Heaven Breathe 1988 10 Roll With It Steve Winwood 1989 1 Look Away Chicago 1989 2 My Prerogative Bobby Brown 1989 3 Every Rose Has Its Thorn Poison 1989 4 Straight up Paula Abdul 1989 5 Miss You Much Janet Jackson 1989 6 Cold Hearted Paula Abdul 1989 7 Wind Beneath My Wings Bette Midler 1989 8 Girl You Know It’s True Milli Vanilli 1989 9 Baby, I Love Your Will to Power Way:Freebird Medley 1989 10 Giving You The Best That I Anita Baker Got 1990 1 Hold On Wilson Phillips 1990 2 It Must Have Been Love Roxette 1990 3 Nothing Compares 2 U Sinead O’Connor 1990 4 Poison Bell Biv DeVoe 1990 5 Vogue Madonna 1990 6 Vision Of Love Mariah Carey 1990 7 Another Day In Paradise Phil Collins 1990 8 Hold On En Vogue 1990 9 Cradle Of Love Billy Idol 1990 10 Blaze Of Glory Jon Bon Jovi 1991 1 (Everything I Do) I Do It for Bryan Adams You 1991 2 I Wanna Sex You Up Color Me Badd 1991 3 Gonna Make You Sweat C&C Music Factory (Everybody Dance Now) 1991 4 Rush, Rush Paula Abdul 1991 5 One More Try Timmy T 1991 6 Unbelievable EMF 1991 7 More Than Words Extreme

89 1991 8 I Like the Way (The Kissing Hi-Five Game) 1991 9 The First Time Surface 1991 10 Baby, Baby Amy Grant 1992 1 End Of The Road Boys II Men 1992 2 Baby Got Back Sir Mix A Lot 1992 3 Jump Kriss Kross 1992 4 Save The Best For Last Vanessa Williams 1992 5 Baby-Baby-Baby TLC 1992 6 Tears In Heaven Eric Clapton 1992 7 My Lovin’ (You’re Never En Vogue Gonna Get It) 1992 8 Under the Bridge Red Hot Chili Peppers 1992 9 All 4 Love Color Me Badd 1992 10 Just Another Day Jon Secada 1993 1 I Will Always Love You Whitney Houston 1993 2 Whoomp! (There It Is) Tag Team 1993 3 Can’t Help Falling In Love UB40 1993 4 That’s The Way Love Goes Janet Jackson 1993 5 Freak Me Silk 1993 6 Weak SWV 1993 7 If I Ever Fall in Love Shai 1993 8 Dreamlover Mariah Carey 1993 9 Rump Shaker Wreckx N Eﬀect 1993 10 Informer Snow 1994 1 The Sign Ace of Base 1994 2 I Swear All-4-One 1994 3 I’ll Make Love To You Boyz II Men 1994 4 The Power of Love Celine Dion 1994 5 Hero Mariah Carey 1994 6 Stay (I Missed You) Lisa Loeb and Nine Stories 1994 7 Breathe Again Toni Braxton 1994 8 All for Love Bryan Adams, Rod Stewart and Sting 1994 9 All That She Wants Ace of Base 1994 10 Don’t Turn Around Ace of Base 1995 1 Gangsta’s Paradise Coolio ft. L.V. 1995 2 Waterfalls TLC 1995 3 Creep TLC 1995 4 Kiss From A Rose Seal 1995 5 On Bended Knee Boyz II Men 1995 6 Another Night Real McCoy

90 1995 7 Fantasy Mariah Carey 1995 8 Take a Bow Madonna 1995 9 Don’t Take It Personal (Just Monica One of Dem Days) 1995 10 This Is How We Do It Montell Jordan 1996 1 Macarena Los Del Rio 1996 2 One Sweet Day Mariah Carey and Boyz II Men 1996 3 Because You Loved Me Celine Dion 1996 4 Nobody Knows The Tony Rich Project 1996 5 Always Be My Baby Mariah Carey 1996 6 Give Me One Reason Tracy Chapman 1996 7 Tha Crossroads Bone Thugs-N-Crossroads 1996 8 I Love You Always Forever Donna Lewis 1996 9 Let It Flow Toni Braxton 1996 9 You’re Makin’ Me High Toni Braxton 1996 10 Twisted Keith Sweat 1997 1 Candle In the Wind Elton John 1997 1 Something About The Way Elton John You Look Tonight 1997 2 Foolish Games Jewel 1997 2 You Were Meant For Me Jewel 1997 3 I’ll Be Missing You Puﬀ Daddy ft. Faith Evans and 112 1997 4 Un-Break My Heart Toni Braxton 1997 5 Can’t Nobody Hold Me Puﬀ Daddy ft. Mase Down 1997 6 I Believe I Can Fly R. Kelly 1997 7 Don’t Let Go (Love) En Vogue 1997 8 Return of the Mack Mark Morrison 1997 9 How Do I Live LeAnn Rimes 1997 10 Wannabe Spice Girls 1998 1 Too Close Next 1998 2 The Boy Is Mine Brandy & Monica 1998 3 You’re Still The One Shania Twain 1998 4 Truly Madly Deeply Savage Garden 1998 5 How Do I Live LeAnn Rimes 1998 6 Together Again Janet 1998 7 All My Life K-Ci & Jojo 1998 8 Candle In the Wind Elton John 1998 9 Nice & Slow Usher 1998 10 I Don’t Want To Wait Paula Cole

91 1999 1 Believe Cher 1999 2 No Scrubs TLC 1999 3 Angel Of Mine Monica 1999 4 Heartbreak Hotel Whitney Houston ft Faith Evans and Kelly Price 1999 5 Baby One More Time Britney Spear 1999 6 Kiss Me Sixpence None the Richer 1999 7 Genie in a Bottle Christina Aguilera 1999 8 Every Morning Sugar Ray 1999 9 Nobody’s Supposed To Be Deborah Cox Here 1999 10 Livin’ la Vida Loca Ricky Martin 2000 1 Breathe Faith Hill 2000 2 Smooth Santana ft. Rob Thomas 2000 3 Maria Maria Santana ft. The Product G&B 2000 4 I Wanna Know Joe 2000 5 Everything You Want Vertical Horizon 2000 6 Say My Name Destiny’s Child 2000 7 I Knew I Loved You Savage Garden 2000 8 Amazed Lonestar 2000 9 Bent Matchbox Twenty 2000 10 He Wasn’t Man Enough Toni Braxton 2001 1 Hanging By A Moment Lifehouse 2001 2 Fallin’ Alicia Keys 2001 3 All For You (Video Single Janet Mix) 2001 4 Drops Of Jupiter (Tell Me) Train 2001 5 I’m Real Jennifer Lopez ft. Ja Rule 2001 6 If You’re Gone Matchbox Twenty 2001 7 Let Me Blow Ya Mind Eve ft. Gwen Stefani 2001 8 Thank You Dido 2001 9 Again Lenny Kravitz 2001 10 Independent Women Destiny’s Child 2002 1 How You Remind Me NickelBack 2002 2 Foolish Ashanti 2002 3 Hot In Herre Nelly 2002 4 Dilemma Nelly ft. Kelly Rowland 2002 5 Wherever You Will Go The Calling 2002 6 A Thousand Miles Vanessa Carlton 2002 7 In The End Linkin Park 2002 8 What’s Luv? Fat Joe ft. Ashanti

92 2002 9 U Got It Bad Usher 2002 10 Blurry Puddle of Mudd 2003 1 In Da Club 50 Cent 2003 2 Ignition (Remix) R. Kelly 2003 3 Get Busy Sean Paul 2003 4 Crazy In Love Beyonce ft. Jay Z 2003 5 When I’m Gone 3 Doors Down 2003 6 Unwell Matchbox Twenty 2003 7 Right Thurr Chingy 2003 8 Miss You Aaliyah 2003 9 Picture Kid Rock ft. Sheryl Crow 2003 10 Bring Me To Life Evanescence 2004 1 Yeah! Usher Featuring Lil Jon & Ludacris 2004 2 Burn Usher 2004 3 If I Ain’t Got You Alicia Keys 2004 4 This Love Maroon 5 2004 5 The Way You Move Outkast ft. Sleepy Brown 2004 6 The Reason Hoobastank 2004 7 I Don’t Wanna Know Mario Winans ft. Enya and P.Diddy 2004 8 Hey Ya! Outkast 2004 9 Goodies Ciara ft. Petey Pablo 2004 10 Lean Back Terror Squad 2005 1 We Belong Together Mariah Carey 2005 2 Hollaback Girl Gwen Stefani 2005 3 Let Me Love You Mario 2005 4 Since U Been Gone Kelly Clarkson 2005 5 1,2 Step Ciara ft. Missy Elliott 2005 6 Gold Digger Kanye West ft. Jamie Foxx 2005 7 Boulevard Of Broken Dreams Green Day 2005 8 Candy Shop 50 Cent ft. Olivia 2005 9 Don’t Cha Pussycat Dolls ft. Busta Rhymes 2005 10 Behind These Hazel Eyes Kelly Clarkson 2006 1 Bad Day Daniel Powter 2006 2 Temperature Sean Paul 2006 3 Promiscuous Nelly Furtado ft. Timbaland 2006 4 You’re Beautiful James Blunt 2006 5 Hips Don’t Lie Shakira ft. Wyclef Jean 2006 6 Unwritten Natasha Bedingﬁeld 2006 7 Crazy Gnarls Barkley

93 2006 8 Ridin’ Chamillionaire ft. Krayzie Bone 2006 9 Sexyback Justin Timberlake ft. Timbaland 2006 10 Check on It Beyonce ft. Slim Thug 2007 1 Irreplaceable Beyonce 2007 2 Umbrella Rihanna ft.Jay-Z 2007 3 The Sweet Escape Gwen Stefani ft. Akon 2007 4 Big Girls Don’t Cry Fergie 2007 5 Buy U a Drank (Shawty T-Pain ft. Yunc Joc Snappin) 2007 6 Before He Cheats Carrie Underwood 2007 7 Hey There Deliah Plain White T’s 2007 8 I Wanna Love You Akon ft. Snoop Dogg 2007 9 Say It Right Nelly Furtado 2007 10 Glamorous Fergie ft. Ludacris 2008 1 Low Flo Rida Ft. T-Pain 2008 2 Bleeding Love Leona Lewis 2008 3 No One Alicia Keys 2008 4 Lollipop Lil Wayne ft. Static Major 2008 5 Apologize Timbaland ft. OneRepublic 2008 6 No Air Jordin Sparks and Chris Brown 2008 7 Love Song Sara Bareilles 2008 8 Love In This Club Usher ft. Young Jeezy 2008 9 With You Chris Brown 2008 10 Forever Chris Brown 2009 1 Boom Boom Pow Black Eyed Peas 2009 3 Just Dance Lady Gaga 2009 3 Poker Face Lady Gaga ft. Colby O’Donis 2009 4 I Gotta Feeling The Black Eyed Peas 2009 5 Love Story Taylor Swift 2009 6 Right Round Flo Rida 2009 7 I’m Yours Jason Mraz 2009 8 Single Ladies (Put a Ring on Beyonce It) 2009 9 Heartless Kanye West 2009 10 Gives You Hell The All-American Rejects 2010 1 TiK ToK Kesha 2010 2 Need You Now Lady Antebellum 2010 3 Hey, Soul Sister Train 2010 4 California Gurls Katy Perry ft. Snoop Dogg

94 2010 5 OMG Usher ft. will.i.am 2010 6 Airplanes B.o.B ft. Hayley Williams 2010 7 Love the Way You Lie Eminem ft. Rihanna 2010 8 Bad Romance Lady Gaga 2010 9 Dynamite Taio Cruz 2010 10 Break Your Heart Taio Cruz ft. Ludacris 2011 1 Rolling In the Deep Adele 2011 2 Party Rock Anthem LMFAO ft. Lauren Bennett &GoonRock 2011 3 Firework Katy Perry 2011 4 E.T. Katy Perry ft. Kanye West 2011 5 Give Me Everything Pitbull ft. Ne-Yo, Afrojack & Nayer 2011 6 Grenade Bruno Mars 2011 7 Fuck You (Forget You) Cee Lo Green 2011 8 Super Bass Nicki Minaj 2011 9 Moves Like Jagger Maroon 5 ft. Christina Aguilera 2011 10 Just Can’t Get Enough The Black Eyed Peas 2012 1 Somebody That I Used to Gotye ft. Kimbra Know 2012 2 Call Me Maybe Carly Rae Jepsen 2012 3 We Are Young fun ft. Janelle Monae 2012 4 Payphone Maroon 5 ft. Wiz Khalifa 2012 5 Lights Ellie Goulding 2012 6 Glad You Came The Wanted 2012 7 Stronger (What Doesn’t Kill Kelly Clarkson You) 2012 8 We Found Love Rihanna ft. Calvin Harris 2012 9 Starships Nicki Minaj 2012 10 What Makes You Beautiful One Direction 2013 1 Thrift Shop Macklemore & Ryan Lewis ft. Wanz 2013 2 Blurred Lines Robin Thicke ft. T.I. and Pharell 2013 3 Radioactive Imagine Dragons 2013 4 Harlem Shake Baauer 2013 5 Can’t Hold Us Macklemore & Ryan Lewis ft. Ray Dalton 2013 6 Mirrors Justin Timberlake 2013 7 Just Give Me a Reason P!nk ft. Nate Ruess 2013 8 When I Was Your Man Bruno Mars

95 2013 9 Cruise (Remix) Florida Georgia Line ft. Nelly 2013 10 Roar Katy Perry 2014 1 Happy Pharrell Williams 2014 2 Dark Horse Katy Perry 2014 3 All Of Me John Legend 2014 4 Fancy Iggy Azaela ft. Charli XCX 2014 5 Counting Stars OneRepublic 2014 6 Talk Dirty Jason Derulo ft. 2 Chainz 2014 7 Rude MAGIC! 2014 8 All About That Bass Meghan Trainor 2014 9 Problem Ariana Grande ft. Iggy Azaela 2014 10 Stay With Me Sam Smith 2015 1 Uptown Funk Mark Ronson ft. Bruno Mars 2015 2 Thinking Out Loud Ed Sheeran 2015 3 See You Again Wiz Khalifa 2015 4 Trap Queen Fetty Wap 2015 5 Sugar Maroon 5 2015 6 Shut Up And Dance Walk The Moon 2015 7 Blank Space Taylor Swift 2015 8 Watch Me Silento 2015 9 Earned It (Fifty Shades Of The Weeknd Grey) 2015 10 Cheerleader (Felix Jaehn OMI Remix Radio Edit)

96 Appendix B: List of the 120 songs analyzed in the second study (Section 3.3) of Chapter 3

Included in this appendix is a list of the 120 songs analyzed in the second study

(Section 3.3) of Chapter 3.

Artist Title Album Popularity Alesso Heroes (we could be) Forever 1 Alesso Immortale Forever 0 Ariana Grande One Last Time My Everything 1 (Deluxe) Ariana Grande Only 1 My Everything 0 (Deluxe) Avicii Waiting For Love Stories 1 Avicii Touch Me Stories 0 Beyoncé 711 Beyonce (Platinum 1 Edition) Beyoncé No Angel Beyonce (Platinum 0 Edition) Big Sean I Don’t Fuck With Dark Sky Paradise 1 You (Deluxe) Big Sean Platinum And Wood Dark Sky Paradise 0 (Deluxe) Calvin Harris Outside Motion 1 Calvin Harris Ecstasy Motion 0 Carly Rae I Really Like You E-mo-tion (Deluxe) 1 Jepsen Carly Rae Black Heart E-mo-tion (Deluxe) 0 Jepsen Charlie Puth Marvin Gaye (feat. Nine Track Mind 1 Meghan Trainor)

97 Charlie Puth As You Are (feat. Shy Nine Track Mind 0 Carter) Chris Brown Ayo Fan Of A Fan The 1 Album (Deluxe Version) Chris Brown It’s Yo Shit Fan Of A Fan The 0 Album (Deluxe Version) Clean Bandit Rather Be (feat. Jess New Eyes 1 Glynne) Clean Bandit Up Again (feat. Rae New Eyes 0 Morris) Coldplay A Sky Full of Stars Ghost Stories 1 Coldplay Oceans Ghost Stories 0 David Guetta Hey Mama (feat. Listen 1 Nicki Minaj & Afro- jack) David Guetta Dangerous (feat. Sam Listen 0 Martin) Demi Lovato Cool for the Summer Conﬁdent (Deluxe 1 Edition) Demi Lovato Cool For The Summer Conﬁdent (Deluxe 0 - Suraci Remix Edition) Drake Hotline Bling Views 1 Drake Summers Over Inter- Views 0 lude Echosmith Cool Kids Talking Dreams 1 (Deluxe Version) Echosmith We’re Not Alone - Talking Dreams 0 Bonus Track (Deluxe Version) Ed Sheeran Thinking Out Loud X (Deluxe Edition) 1 Ed Sheeran Shirtsleeves X (Deluxe Edition) 0 Ellie Goulding “Love Me Like You Do Delirium (Deluxe) 1 - From ’Fifty Shades Of Grey”’ Ellie Goulding Heal Delirium (Deluxe) 0 Fall Out Boy Centuries American Beauty 1 American Psycho Fall Out Boy Favorite Record American Beauty 0 American Psycho Fetty Wap Trap Queen Fetty Wap 1 (Deluxe)

98 Fetty Wap Whateva Fetty Wap 0 (feat.Monty) (Deluxe) Fifth Harmony Worth It Reﬂection 1 Fifth Harmony Body Rock Reﬂection 0 Flo Rida GDFR (feat. Sage My House 1 The Gemini & Lookas) Flo Rida That’s What I Like My House 0 (feat. Fitz) Galantis Runaway (U & I) Pharmacy 1 Galantis Kill ’Em With The Pharmacy 0 Love George Ezra Budapest Wanted On Voyage 1 George Ezra It’s Just My Skin Wanted On Voyage 0 Hozier Take Me To Church Hozier 1 Hozier Foreigner’s God Hozier 0 Imagine Dragons Radioactive Night Visions 1 (Deluxe) Imagine Dragons I Don’t Mind Night Visions 0 (Deluxe) James Bay Hold Back The River Chaos And The 1 Calm (Deluxe Edition) James Bay Need The Sun To Chaos And The 0 Break - The Dark Of Calm (Deluxe The Morning Version Edition) Jason Derulo Want To Want Me Everything Is 4 1 Jason Derulo X2CU Everything Is 4 0 Jess Glynne Hold My Hand I Cry When I 1 Laugh Jess Glynne Saddest Vanilla (feat. I Cry When I 0 Emelis Sandé) Laugh Jessie J Bang Bang Sweet Talker 1 (Deluxe Version) Jessie J Your Loss I’m Found Sweet Talker 0 (Deluxe Version) John Legend All Of Me Love In The Future 1 (Special Edition) John Legend Ordinary People - Love In The Future 0 Live (Special Edition) Justin Bieber What Do You Mean? Purpose (Deluxe) 1 Justin Bieber All In It Purpose (Deluxe) 0

99 Kygo Firestone Cloud Nine 1 Kygo Not Alone Cloud Nine 0 Lilly Wood and Prayer In C - Robin Invincible Friends 1 The Prick Schulz Radio Edit (Edition Robin Schulz Remix) Lilly Wood and Hymn To My Invisible Invincible Friends 0 The Prick Friend (Edition Robin Schulz Remix) MAGIC! Rude Don’t Kill The 1 Magic MAGIC! Little Girl Big World Don’t Kill The 0 Magic Major Lazer Lean On (feat. Mo & Peace Is The 1 DJ Snake) Mission (Extended) Major Lazer Thunder & Lightning Peace Is The 0 (feat. Gent & Jawns) Mission (Extended) Mark Ronson Uptown Funk Uptown Special 1 Mark Ronson "Crack In The Pearl& Uptown Special 0 Pt. II" Maroon 5 Sugar V (Deluxe) 1 Maroon 5 Shoot Love V (Deluxe) 0 Meghan Trainor All About That Bass Title (Deluxe) 1 Meghan Trainor What If I Title (Deluxe) 0 Nick Jonas Jealous Nick Jonas X2 1 Nick Jonas Teacher - Young Nick Jonas X2 0 Bombs Remix Radio Edit Omarion Post To Be (feat. Sex Playlist 1 Chris Brown & Jhene Aiko) Omarion Already Sex Playlist 0 OMI Cheerleader - Felix Me 4 U 1 Jaehn Remix Radio Edit OMI Stir It Me 4 U 0 One Direction Drag Me Down Made In The A.M. 1 (Deluxe Edition) One Direction Hey Angel Made In The A.M. 0 (Deluxe Edition) OneRepublic Counting Stars Native 1

100 OneRepublic Don’t Look Down Native 0 Passenger Let Her Go All The Little 1 Lights (Deluxe Version) Passenger Feather On The Clyde All The Little 0 - Acoustic Lights (Deluxe Version) Pharrell “Happy - From ’Despi- GIRL 1 Williams cable Me 2”’ Pharrell Gush G I R L 0 Williams Pitbull Time Of Our Lives Globalization 1 Pitbull Day Drinking Globalization 0 R. City Locked Away What Dreams Are 1 Made Of R. City Our Story What Dreams Are 0 Made Of Robin Schulz Headlights (feat. Sugar 1 Ilsey) Robin Schulz This Is Your Life Sugar 0 Sam Smith Stay With Me In The Lonely 1 Hour Sam Smith Good Thing In The Lonely 0 Hour Selena Gomez Good For You Revival (Deluxe) 1 Selena Gomez Rise Revival (Deluxe) 0 Shawn Mendes Stitches Handwritten 1 Shawn Mendes I Don’t Even Know Handwritten 0 Your Name Sheppard Geronimo Bombs Away 1 Sheppard Halfway To Hell Bombs Away 0 Sia Elastic Heart 1000 Forms Of 1 Fear (Deluxe Version) Sia Big Girls Cry - 1000 Forms Of 0 Bleachers Remix Fear (Deluxe Version) Skrillex Where Are U Now Skrillex and Diplo 1 present Jack Skrillex Don’t Do Drugs Just Skrillex and Diplo 0 Take Some Jack U present Jack

101 The Weeknd Can’t Feel My Face Beauty Behind 1 The Madness The Weeknd As You Are Beauty Behind 0 The Madness Tove Lo Talking Body Queen Of The 1 Clouds (Blueprint Edition) Tove Lo Crave Queen Of The 0 Clouds (Blueprint Edition) Vance Joy Riptide Dream Your Life 1 Away (Special Edition) Vance Joy Straight Into Your Dream Your Life 0 Arms Away (Special Edition) WALK THE Shut Up and Dance Talking Is Hard 1 MOON WALK THE Come Under The Talking Is Hard 0 MOON Covers Years & Years King Communion 1 (Deluxe) Years & Years 1977 Communion 0 (Deluxe) Zedd I Want You To Know True Colors 1 Zedd Bumble Bee True Colors 0

102 Appendix C: Visual representation of the participants’ skipping times in both the undivided and divided attention experimental setting (Chapter 4) superimposed onto each song’s formal diagram

Included in this appendix is a visual representation of the participants’ skipping times in both the undivided and divided attention experimental setting (Chapter 4) superimposed onto each song’s formal diagram. The x-axis represents listening times, in seconds. Each dot represents a participants, grouped according to their familiarity with the song. The dotted purple line indicates the moment the voice comes in. The dashed yellow line indicates the moment the title of the song is mentioned.

103 104 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided The Way andtheRange Hornsby ItIs−Bruce The Way andtheRange Hornsby ItIs−Bruce 0 0 300 200 300 100 200 0 100 0

intro intro

verse verse

chorus chorus

interlude interlude

verse verse

Time Before Skip(insec) chorus Time Before Skip(insec) chorus

interlude interlude

solo solo

verse verse

chorus chorus

outro outro 105 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Sweet ChildO'Mine−Gunsn'Roses Sweet ChildO'Mine−Gunsn'Roses 0 0 300 200 300 100 200 0 100 0

intro intro

verse verse

chorus chorus

interlude interlude

verse verse Time Before Skip(insec) Time Before Skip(insec) chorus chorus

solo solo

chorus chorus

solo solo

outro outro 106 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Could've Been−Tiffany Could've Been−Tiffany 01010200 150 200 100 150 50 100 0 50 0 intro intro

verse verse

chorus chorus

interlude interlude Time Before Skip(insec) Time Before Skip(insec)

verse verse

chorus chorus

solo solo

chorus chorus

outro outro 107 (b) (a) Familiarity es Ye No iie teto Setting Attention Divided niie teto Setting Attention Undivided Cold Hearted −PaulaCold Hearted Abdul This songwasnotusedinthe dividedattentionsetting. 01010200 150 100 50 0

intro

chorus

verse

chorus

interlude

verse Time Before Skip(insec)

chorus

interlude

bridge 1

solo

bridge 2

outro 108 (b) (a) Familiarity es Ye No iie teto Setting Attention Divided niie teto Setting Attention Undivided Baby, ILove Your Way:Freebird Medley −WilltoPower This songwasnotusedinthe dividedattentionsetting. 0101020250 200 150 100 50 0

intro

verse

pre−chorus

chorus Time Before Skip(insec) verse (Freebird)

verse

pre−chorus

chorus 109 (b) (a) Familiarity es Ye No iie teto Setting Attention Divided niie teto Setting Attention Undivided It MustHave BeenLove −Roxette This songwasnotusedinthe dividedattentionsetting. 0101020250 200 150 100 50 0

intro

verse

pre−chorus

chorus Time Before Skip(insec) verse

pre−chorus

chorus

solo

chorus

outro 110 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Vision OfLove Carey −Mariah Vision OfLove Carey −Mariah 01010200 150 200 100 150 50 100 0 50 0

intro intro

verse verse

chorus chorus

verse verse Time Before Skip(insec) Time Before Skip(insec)

chorus chorus

bridge bridge

verse verse

chorus chorus

outro outro 111 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided The FirstTime−Surface The FirstTime−Surface 0 200 100 200 0 100 0

intro intro

verse verse

pre−chorus pre−chorus

chorus chorus

interlude interlude Time Before Skip(insec) Time Before Skip(insec)

verse verse

pre−chorus pre−chorus

chorus chorus

interlude interlude

solo solo

chorus chorus

outro outro 112 (b) (a) Familiarity es Ye No iie teto Setting Attention Divided niie teto Setting Attention Undivided All ThatSheWants −AceofBase This songwasnotusedinthe dividedattentionsetting. 01010200 150 100 50 0

intro

intro (part 2)

verse

chorus

interlude Time Before Skip(insec)

break

verse

chorus

solo

chorus

outro 113 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Don't Take ItPersonal (Just OneofDemDays) −Monica Don't Take ItPersonal (Just OneofDemDays) −Monica 0 200 100 200 0 100 0

intro intro (based on chorus) (based on chorus)

verse verse

chorus chorus

verse verse Time Before Skip(insec) Time Before Skip(insec)

chorus chorus

interlude interlude

chorus chorus 114 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Because You Loved Me −CelineDion Because You Loved Me −CelineDion 0 200 100 200 0 100 0

verse verse

chorus chorus

interlude interlude

verse verse Time Before Skip(insec) Time Before Skip(insec)

chorus chorus

interlude interlude

bridge bridge

chorus chorus

outro outro 115 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided All MyLife −K−Ci&Jojo All MyLife −K−Ci&Jojo 01010200 150 200 100 150 50 100 0 50 0

intro intro

verse verse

chorus chorus

Time Before Skip(insec) interlude Time Before Skip(insec) interlude

verse verse

chorus chorus

bridge bridge

chorus chorus 116 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Every Morning −SugarRay Morning Every −SugarRay Morning Every 01010200 150 200 100 150 50 100 0 50 0

intro intro

chorus chorus

verse verse

refrain refrain Time Before Skip(insec) Time Before Skip(insec) interlude interlude

chorus chorus

verse verse

refrain refrain

bridge bridge

chorus chorus

outro outro 117 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Let MeBlow Ya Mind−Eve ft.Gwen Stefani Let MeBlow Ya Mind−Eve ft.Gwen Stefani 01010200 150 100 200 50 150 100 0 50 0

intro intro

verse verse

chorus chorus Time Before Skip(insec) Time Before Skip(insec)

verse verse

chorus chorus

verse verse

chorus chorus

outro outro 118 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Again −Lenny Kravitz Again −Lenny Kravitz 01010200 150 100 200 50 150 100 0 50 0

intro intro

verse verse

chorus chorus

verse verse Time Before Skip(insec) Time Before Skip(insec)

chorus chorus

bridge bridge

solo solo

chorus chorus 119 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided In DaClub−50Cent In DaClub−50Cent 01010200 150 100 200 150 50 100 0 50 0

intro intro

chorus chorus

verse verse

Time Before Skip(insec) chorus Time Before Skip(insec) chorus

bridge bridge

verse verse

chorus chorus

outro outro 120 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Irreplaceable −Beyonce Irreplaceable −Beyonce 01010200 150 100 200 50 150 100 0 50 0

intro intro

verse verse

pre−chorus pre−chorus

chorus chorus

verse verse Time Before Skip(insec) Time Before Skip(insec) pre−chorus pre−chorus

chorus chorus

bridge bridge

interlude interlude

chorus chorus 121 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided I Wanna Love You −Akon ft.SnoopDogg I Wanna Love You −Akon ft.SnoopDogg 0101020250 200 150 250 100 200 50 150 0 100 50 0

intro intro

chorus chorus

verse (Snoop Dogg) verse (Snoop Dogg)

chorus chorus

verse (Akon) verse (Akon) Time Before Skip(insec) Time Before Skip(insec)

chorus chorus

verse (Snoop Dogg) verse (Snoop Dogg)

verse (Akon) verse (Akon)

chorus chorus

outro outro 122 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided No Air − Jordin Sparks and Chris Brown andChris No Air−JordinSparks Brown andChris No Air−JordinSparks 010150 100 150 50 100 0 50 0 intro intro

verse verse

pre−chorus pre−chorus

chorus chorus Time Before Skip(insec) Time Before Skip(insec)

verse verse

pre−chorus pre−chorus

chorus chorus

interlude interlude

chorus chorus 123 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Love −Taylor Story Swift Love −Taylor Story Swift 01010200 150 100 200 50 150 0 100 50 0

intro intro

verse verse

pre−chorus pre−chorus

chorus chorus

interlude interlude

Time Before Skip(insec) verse Time Before Skip(insec) verse

pre−chorus pre−chorus

chorus chorus

solo solo

bridge bridge

chorus chorus

outro outro 124 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Dynamite −Taio Cruz Dynamite −Taio Cruz 01010200 150 200 100 150 50 100 0 50 0 intro intro

verse verse

pre−chorus pre−chorus

chorus chorus

verse verse Time Before Skip(insec) Time Before Skip(insec) pre−chorus pre−chorus

chorus chorus

bridge bridge

interlude interlude

chorus chorus 125 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided Starships −NickiMinaj Starships −NickiMinaj 01010200 150 200 100 150 50 100 0 50 0 intro intro

verse verse

pre−chorus pre−chorus

chorus chorus

interlude interlude Time Before Skip(insec) Time Before Skip(insec)

verse verse

pre−chorus pre−chorus

chorus chorus

interlude interlude

chorus chorus

outro outro 126 (b) (a) Familiarity es Ye No iie teto Setting Attention Divided niie teto Setting Attention Undivided Mirrors −Justin Timberlake This songwasnotusedinthe dividedattentionsetting. 0 0 0 0 500 400 300 200 100 0

intro to part 1

verse

pre−chorus

chorus

verse

pre−chorus

chorus Time Before Skip(insec)

bridge

chorus

interlude

outro to part 1

intro to part 2

verse

interlude

verse

outro to part 2 127 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided All AboutThatBass−MeghanTrainor All AboutThatBass−MeghanTrainor 010150 100 150 50 100 0 50 0

chorus chorus

verse verse

pre−chorus pre−chorus Time Before Skip(insec) Time Before Skip(insec)

chorus chorus

verse verse

pre−chorus pre−chorus

chorus chorus

outro outro 128 (b) (a) Familiarity Familiarity es Ye s Ye No No iie teto Setting Attention Divided niie teto Setting Attention Undivided See You Again−WizKhalifa See You Again−WizKhalifa 01010200 150 100 200 50 150 100 0 50 0

intro intro

chorus chorus

verse verse

pre−chorus pre−chorus

chorus chorus Time Before Skip(insec) Time Before Skip(insec)

interlude interlude

verse verse

pre−chorus pre−chorus

bridge bridge

chorus chorus

outro outro Bibliography

Aaker, D. A. & Stayman, D. M. (1990). Measuring audience perceptions of com-

mercials and relating them to ad impact. Journal of Advertising Research, 30 (4),

7–17.

Allan, D. (2006). Eﬀects of popular music in advertising on attention and memory.

Journal of Advertising Research, 46 (4), 434–444.

Amos, C., Holmes, G., & Strutton, D. (2008). Exploring the relationship between

celebrity endorser eﬀects and advertising eﬀectiveness: A quantitative synthesis of

eﬀect size. International Journal of Advertising, 27 (2), 209–234.

Anderson, C. (2006). The long tail: Why the future of business is selling less of more.

Hyperion.

Baddeley, A. D., Thomson, N., & Buchanan, M. (1975). Word length and the struc-

ture of short-term memory. Journal of Verbal Learning and Verbal Behavior, 14 (6),

575–589.

Bailes, F. (2007). The prevalence and nature of imagined music in the everyday lives

of music students. Psychology of Music, 35 (4), 555–570.

129 Bailes, F. (2015). Music in mind? An experience sampling study of what and when,

towards an understanding of why. Psychomusicology: Music, Mind, and Brain,

25 (1), 58–68.

Beaman, C. P. & Williams, T. I. (2010). Earworms (stuck song syndrome): Towards

a natural history of intrusive thoughts. British Journal of Psychology, 101 (4),

637–653.

Berlyne, D. E. (1970). Novelty, complexity, and hedonic value. Attention, Perception,

& Psychophysics, 8 (5), 279–286.

Berlyne, D. E. (1971). Aesthetics and psychobiology. New York: Appleton-Century-

Crofts.

Biamonte, N. (2014). Formal functions of metric dissonance in rock music. Music

Theory Online, 20 (2).

Boone, C. (2013). Mashing: Toward a typology of recycled music. Music Theory

Online, 19 (3).

Brown, S. P. & Stayman, D. M. (1992). Antecedents and consequences of attitude

toward the ad: A meta-analysis. Journal of consumer research, 19 (1), 34–51.

Burns, G. (1987). A typology of ‘hooks’ in popular records. Popular Music, 6 (1),

1–20.

Bushman, B. J. & Baumeister, R. F. (1998). Threatened egotism, narcissism, self-

esteem, and direct and displaced aggression: Does self-love or self-hate lead to

violence? Journal of personality and social psychology, 75 (1), 219–229.

130 Cannam, C., Landone, C., & Sandler, M. (2010). Sonic visualiser: An open source

application for viewing, analysing, and annotating music audio ﬁles. In Proceedings

of the 18th ACM international conference on multimedia (pp. 1467–1468). Firenze,

Italy.

Celma, O. (2010). Music Recommendation and Discovery: The Long Tail, Long Fail,

and Long Play in the Digital Music Space. Springer.

Christianson, S.-Å. & Loftus, E. F. (1991). Remembering emotional events: The fate

of detailed information. Cognition and Emotion, 5 (2), 81–108.

Ciampaglia, G. L., Flammini, A., & Menczer, F. (2015). The production of informa-

tion in the attention economy. Scientiﬁc Reports, 5.

Cisek, S. Z., Sedikides, C., Hart, C. M., Godwin, H. J., Benson, V., & Liversedge, S. P.

(2014). Narcissism and consumer behaviour: A review and preliminary ﬁndings.

Frontiers in psychology, 5 (232).

Cockburn, A. & McKenzie, B. (2001). What do web users do? An empirical analysis

of web use. International Journal of Human-Computer Studies, 54 (6), 903–922.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences. (2nd ed.).

Hilsdale, NJ: Lawrence Earlbaum Associates.

Cook, N. (1995). The conductor and the theorist: Furtwängler, Schenker and the

ﬁrst movement of Beethoven’s Ninth Symphony. In J. Rink (Ed.), The practice of

performance: Studies in musical interpretation (pp. 105–125). Cambridge, England:

Cambridge University Press.

131 Corey, J., Conrad, F., Reichert, H., Goldstein, S., Ostrow, J., & Sadowsky, M. (2017).

Moment to moment listening experience for popular songs. Paper presented at the

annual meeting of the Society for Music Perception and Cognition, San Diego, CA.

Datta, H., Knox, G., & Bronnenberg, B. J. (2017). Changing their tune: How con-

sumers’ adoption of online streaming aﬀects music consumption and discovery.

Marketing Science, Published Online in Articles in Advance 11 September 2017.

https://doi.org/10.1287/mksc.2017.1051.

Davenport, T. & Beck, J. (2001). The attention economy: Understanding the new

economy of business. Cambridge, MA: Harvard Business School Press.

Davis, S. (1985). The craft of lyric writing. Cincinnati, OH: Writer’s Digest Books.

Delson, D. & Hurst, W. E. (1980). Delson’s Dictionary of Radio & Record Industry

Terms. Thousand Oaks, CA: Bradson Press.

Denisoﬀ, R. S. (1989). The evolution of pop music broadcasting: 1920–1972, (pp.

78–93). American popular music: The age of rock. Bowling Green, OH: Popular

Press.

DeWall, C. N., Jr., R. S. P., Campbell, W. K., & Twenge, J. M. (2011). Tuning in to

psychological change: Linguistic markers of psychological traits and emotions over

time in popular us song lyrics. Psychology of Aesthetics, Creativity, and the Arts,

5 (3), 200–207.

DeWall, C. N., Twenge, J. M., Gitter, S. A., & Baumeister, R. F. (2009). It’s the

thought that counts: The role of hostile cognition in shaping aggressive responses

to social exclusion. Journal of personality and social psychology, 96 (1), 45–59.

132 Dowling, W. J. (1990). Expectancy and attention in melody perception. Psychomu-

sicology: A Journal of Research in Music Cognition, 9 (2), 148–160.

Dowling, W. J., Lung, K. M.-T., & Herrbold, S. (1987). Aiming attention in pitch

and time in the perception of interleaved melodies. Attention, Perception, & Psy-

chophysics, 41 (6), 642–656.

Dredge, S. (2015, April 3). How much do musicians really make from Spo-

tify, iTunes and YouTube? The Guardian. Retrieved from https:

//www.theguardian.com/technology/2015/apr/03/how-much-musicians-

make-spotify-itunes-youtube.

Edwards, M. (2007, November 22). Voices ﬁnd gilt by association. The Aus-

tralian. Retrieved from http://web.archive.org/web/20080202144158/http:

//www.theaustralian.news.com.au/story/0,25197,22799180-16947,00.html.

Eichstaedt, J. & Silvia, P. J. (2003). Noticing the self: Implicit assessment of self-

focused attention using word recognition latencies. Social Cognition, 21 (5), 349–

361.

Farrar, K., Kunkel, D., Biely, E., Eyal, K., Fandrich, R., & Donnerstein, E. (2003).

Sexual messages during prime-time programming. Sexuality & Culture, 7 (3), 7–37.

Floridou, G. A. & Müllensiefen, D. (2015). Environmental and mental conditions

predicting the experience of involuntary musical imagery: An experience sampling

method study. Consciousness and cognition, 33, 472–486.

Gjerdingen, R. O. & Perrott, D. (2008). Scanning the dial: The rapid recognition of

music genres. Journal of New Music Research, 37 (2), 93–100.

133 GlobalWebIndex Company (2017). GWI social summary report Q3 2017. GlobalWe-

bIndex Company. Retrieved from https://insight.globalwebindex.net/hubfs/

GWI%20Social%20Summary%20Q3%202017.pdf?t=1510330069650.

Goldhaber, M. H. (1997, December 1). Attention shoppers! Wired Magazine. Re-

trieved from https://www.wired.com/1997/12/es-attention/.

Gowensmith, W. N. & Bloom, L. J. (1997). The eﬀects of heavy metal music on

arousal and anger. Journal of music therapy, 34 (1), 33–45.

Heyduk, R. G. (1975). Rated preference for musical compositions as it relates to

complexity and exposure frequency. Attention, Perception, & Psychophysics, 17 (1),

84–90.

Huron, D. (1989). Music in advertising: An analytic paradigm. Musical Quarterly,

73 (4), 557–574.

Huron, D. (1996). The melodic arch in Western folksongs. Computing in Musicology,

10, 3–23.

Huron, D. (2013). A psychological approach to musical form: The habituation-ﬂuency

theory of repetition. Current musicology, 96, 7–35.

Husain, G., Thompson, W. F., & Schellenberg, E. G. (2002). Eﬀects of musical

tempo and mode on arousal, mood, and spatial abilities. Music Perception, 20 (2),

151–171.

Hyman, I. E., Burland, N. K., Duskin, H. M., Cook, M. C., Roy, C. M., McGrath,

J. C., & Roundhill, R. F. (2013). Going gaga: Investigating, creating, and manip-

ulating the song stuck in my head. Applied Cognitive Psychology, 27 (2), 204–215.

134 Ingham, T. (2015, January 1). The sad story of the collapse of the al-

bum in America. Music Business Worldwide. Retrieved from http:

//www.musicbusinessworldwide.com/the-sad-story-of-the-collapse-of-

the-album-in-america/.

INRIX Company (2017). INRIX 2016 traﬃc scorecard – U.S. INRIX Company. Re-

trieved from http://inrix.com/resources/inrix-2016-traffic-scorecard-

us/.

Jakubowski, K., Farrugia, N., Halpern, A. R., Sankarpandi, S. K., & Stewart, L.

(2015). The speed of our mental soundtracks: Tracking the tempo of involuntary

musical imagery in everyday life. Memory & cognition, 43 (8), 1229–1242.

Jakubowski, K., Finkel, S., Stewart, L., & Müllensiefen, D. (2017). Dissecting an ear-

worm: Melodic features and song popularity predict involuntary musical imagery.

Psychology of Aesthetics, Creativity, and the Arts, 11 (2), 122–135.

Johar, J. S. & Sirgy, M. J. (1991). Value-expressive versus utilitarian advertising

appeals: When and why to use which appeal. Journal of advertising, 20 (3), 23–33.

Jones, M. R., Boltz, M., & Kidd, G. (1982). Controlled attending as a function

of melodic and temporal context. Attention, Perception, & Psychophysics, 32 (3),

211–218.

Jonze, T. (2014, March 19). How to make money from Spotify by streaming silence.

The Guardian. Retrieved from https://www.theguardian.com/music/musicblog/

2014/mar/19/spotify-streaming-silence-vulpeck-make-money.

135 Juslin, P. N. (2001). Communicating emotion in music performance: a review and

theoretical framework. In P. N. Juslin & J. A. Sloboda (Eds.), Music and Emotion:

Theory and Research (pp. 305–333). New York: Oxford University Press.

Kahneman, D. (2011). Thinking, fast and slow. New York: Farrar, Straus and Giroux.

Kaneshiro, B. & Baker, C. W. (2016). Musical correlates of large-scale song discovery:

A Shazam study. In Proceedings of the 14th International Conference for Music

Perception and Cognition (pp. 658). San Francisco, CA.

Kastellec, J. P. & Leoni, E. L. (2007). Using graphs instead of tables in political

science. Perspectives on politics, 5 (4), 755–771.

Katz, M. (2011). The amateur in the age of mechanical music. In T. Pinch &

K. Bijsterveld (Eds.), The Oxford handbook of sound studies (pp. 459–479). Oxford,

UK: Oxford University Press.

Kellaris, J. J. (2001). Identifying properties of tunes that get stuck in your head:

toward a theory of cognitive itch. In S. E. Heckler & S. Shapiro (Eds.), Proceed-

ings of the Society for Consumer Psychology Winter 2001 Conference (pp. 66–67).

Scottsdale, AZ.

Kellaris, J. J. (2008). Music and consumers. In C. P. Haugtvedt, P. Herr, & F. Kardes

(Eds.), Handbook of consumer psychology (pp. 837–856). New York: Taylor & Fran-

cis.

Kellaris, J. J. & Kent, R. J. (1993). An exploratory investigation of responses elicited

by music varying in tempo, tonality, and texture. Journal of Consumer Psychology,

2 (4), 381–401.

136 Keppler, P. (1956). Some comments on musical quotation. The Musical Quarterly,

42 (4), 473–485.

Klipfel, J. A., Barclay, A. C., & Bockorny, K. M. (2014). Self-congruity: a determinant

of brand personality. Journal of Marketing Development and Competitiveness, 8 (3),

130–143.

Klucharev, V., Smidts, A., & Fernández, G. (2008). Brain mechanisms of persuasion:

how ’expert power’ modulates memory and attitudes. Social cognitive and aﬀective

neuroscience, 3 (4), 353–366.

Kopf, D. (2016, March 9). Say “Hello” to the era of the one-word song title.

Priceonomics. Retrieved from https://priceonomics.com/say-hello-to-the-

era-of-one-word-song-title.

Lamere, P. (2014, May 2). The skip. Music Machinery [blog]. Retrieved from

https://musicmachinery.com/2014/05/02/the-skip/.

Lang, A., Wise, K., Lee, S., & Cai, X. (2003). The eﬀects of sexual appeals on

physiological, cognitive, emotional, and attitudinal responses for product and alcohol

billboard advertising, (pp. 107–132). Lawrence Erlbaum Associates Mahwah, NJ.

Lay, P. L. (2004). Les dirigeants face au changement. Paris: Éditions du 8e jour.

Léveillé Gauvin, H. (2015). Three Empirical Studies on the Evolution of Musical

Syntax in Popular Music from the 1960s (Unpublished master’s thesis). McGill

University, Montreal, Canada.

137 Léveillé Gauvin, H. (2017). Drawing listener attention in popular music: Testing ﬁve

musical features arising from the theory of attention economy. Musicae Scientiae,

Prepublished March 1, 2017. http://dx.doi.org/1029864917698010.

Levitin, D. J. (2006). This is your brain on music: The science of a human obsession.

New York: Dutton/Penguin Books.

Liikkanen, L. A. (2008). Music in everymind: Commonality of involuntary musical

imagery. In Proceedings of the 10th International Conference on Music Perception

and Cognition (pp. 408–412). Sapporo, Japan.

Liikkanen, L. A. (2012). Inducing involuntary musical imagery: An experimental

study. Musicae Scientiae, 16 (2), 217–234.

Litle, P. & Zuckerman, M. (1986). Sensation seeking and music preferences. Person-

ality and individual diﬀerences, 7 (4), 575–578.

Liu, C., White, R. W., & Dumais, S. (2010). Understanding web browsing behaviors

through weibull analysis of dwell time. In 33rd international ACM SIGIR confer-

ence on Research and development in information retrieval (pp. 379–386). Geneva,

Switzerland.

Madanikia, Y. & Bartholomew, K. (2014). Themes of lust and love in popular music

lyrics from 1971 to 2011. SAGE Open, 4 (3), 1–8.

Margulis, E. H. (2014). On repeat: How music plays the mind. Oxford University

Press.

Mauch, M., MacCallum, R. M., Levy, M., & Leroi, A. M. (2015). The evolution of

popular music: USA 1960–2010. Open Science, 2 (5). doi: 10.1098/rsos.150081.

138 McCown, W., Keiser, R., Mulhearn, S., & Williamson, D. (1997). The role of per-

sonality and gender in preference for exaggerated bass in music. Personality and

Individual Diﬀerences, 23 (4), 543–547.

McIntosh, W. D., Schwegler, A. F., & Terry-Murray, R. M. (2000). Threat and

television viewing in the united states, 1960-1990. Media Psychology, 2 (1), 35–46.

McLuhan, M. (1994). Understanding media: The extensions of man. MIT press.

McNamara, L. & Ballard, M. E. (1999). Resting arousal, sensation seeking, and music

preference. Genetic, social, and general psychology monographs, 125 (3), 229–250.

Mehl, M. R., Robbins, M. L., & Holleran, S. E. (2012). How taking a word for

a word can be problematic: Context-dependent linguistic markers of extraversion

and neuroticism. Journal of Methods and Measurement in the Social Sciences, 3 (2),

30–50.

Merritt, B. (2017). The rise of attention metrics: Can a new digi-

tal currency help sustain journalism? [white paper]. Retrieved March 19,

2018, from the George Washington University School of Media and Pub-

lic Aﬀairs: https://smpa.gwu.edu/sites/g/files/zaxdzs2046/f/downloads/

Merritt_whitepaper_for_publication.pdf.

Meyer, L. B. (1956). Emotion and meaning in music. University of Chicago Press:

Chicago.

Miles, S. A., Rosen, D. S., & Grzywacz, N. M. (2017). A statistical analysis of the

relationship between harmonic surprise and preference in popular music. Frontiers

in human neuroscience, 11 (263).

139 Morris, C. (2017, May). Are pop lyrics getting more repetitive? The Pudding [blog].

Retrieved from https://pudding.cool/2017/05/song-repetition/.

Murphy, M. (2017, March 3). Radio: Reaching a captive audience. Katz Media Group.

Retrieved from http://www.katzradiogroup.com/research/Dropoff/Insights/

2017_0303_Radio_Insight_Reaching%20a%20Captive%20Audience.pdf.

Nelson, L. D., Pettijohn, T. F., & Galak, J. (2007). Mate preferences in social cogni-

tive context: When environmental and personal change leads to predictable cross-

cultural variation. In V. Swami & A. Furnham (Eds.), Body beautiful: Evolutionary

and sociocultural perspectives (pp. 185–206). New York: Palgrave.

Nielsen Company (2015, January 7). 2014 Nielsen music U.S. report. The Nielsen

Company. Retrieved from http://www.nielsen.com/us/en/press-room/2015/

2014-nielsen-music-report.html.

North, A. C. & Hargreaves, D. J. (1995). Subjective complexity, familiarity, and

liking for popular music. Psychomusicology, 14 (1-2), 77–93.

North, A. C. & Hargreaves, D. J. (2011). Music and marketing. In P. N. Juslin & J. A.

Sloboda (Eds.), Handbook of Music and Emotion: Theory, Research, Applications

(pp. 909–930). New York: Oxford University Press.

North, A. C., Krause, A. E., Kane, R., & Sheridan, L. (2017). United kingdom “top

5” pop music lyrics. Psychology of Music, Prepublished July 31, 2017. https:

//doi.org/10.1177/0305735617720161.

140 Nunes, J. C., Ordanini, A., & Valsesia, F. (2015). The power of repetition: repetitive

lyrics in a song increase processing ﬂuency and drive market success. Journal of

Consumer Psychology, 25 (2), 187–199.

Ogawa, H. (1998). The eﬀects of karaoke on music in japan. In T. Mitsui &

S. Hosokawa (Eds.), Karaoke around the world: Aăglobal technology, local singing

(pp. 43–52). London, UK: Routledge.

Osborn, B. (2013). Subverting the verse-chorus paradigm: Terminally climactic forms

in recent rock music. Music Theory Spectrum, 35 (1), 23–47.

Osborne, R. (2012). Vinyl: A history of the analogue record. Burlington, VT: Ashgate.

Parkin, A. J., Reid, T. K., & Russo, R. (1990). On the diﬀerential nature of implicit

and explicit memory. Memory & Cognition, 18 (5), 507–514.

Pennebaker, J. W., Boyd, R. L., Jordan, K., & Blackburn, K. (2015).

The development and psychometric properties of LIWC2015. Retrieved

from https://repositories.lib.utexas.edu/bitstream/handle/2152/31333/

LIWC2015_LanguageManual.pdf?sequence=3.

Pettijohn, T. F. & Sacco, D. F. (2009a). The language of lyrics: An analysis of

popular billboard songs across conditions of social and economic threat. Journal

of Language and Social Psychology, 28 (3), 297–311.

Pettijohn, T. F. & Sacco, D. F. (2009b). Tough times, meaningful music, mature

performers: popular billboard songs and performer preferences across social and

economic conditions in the USA. Psychology of Music, 37 (2), 155–179.

141 Planas Rego, P. (2013, October 7). We’ve turned 5–here’s our story so far. Spotify

News [blog]. Retrieved from https://news.spotify.com/us/2013/10/07/the-

spotify-story-so-far/.

Plazak, J. & Huron, D. (2011). The ﬁrst three seconds: Listener knowledge gained

from brief musical excerpts. Musicae Scientiae, 15 (1), 29–44.

Radano, R. M. (1989). Interpreting muzak: Speculations on musical experience in

everyday life. American Music, 7 (4), 448–460.

Renko, S. & Gregur, T. (2017). The aural nature of atmosphere in a retail setting.

In H. R. Kaufmann & M. F. A. K. Panni (Eds.), Socio-economic perspectives on

consumer engagement and buying behavior (pp. 290–311). IGI Global.

Rentfrow, P. J., Goldberg, L. R., & Levitin, D. J. (2011). The structure of musi-

cal preferences: a ﬁve-factor model. Journal of personality and social psychology,

100 (6), 1139–1157.

Rentfrow, P. J. & Gosling, S. D. (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.

Roberts, B. W. & Helson, R. (1997). Changes in culture, changes in personality: The

inﬂuence of individualism in a longitudinal study of women. Journal of Personality

and Social Psychology, 72 (3), 641–651.

Sacks, O. (2007). Musicophilia: Tales of Music and the Brain. New York: Alfred A.

Knopf.

142 Sawang, S. (2010). Sex appeal in advertising: What consumers think. Journal of

Promotion Management, 16 (1-2), 167–187.

Schellenberg, E. G., Iverson, P., & Mckinnon, M. C. (1999). Name that tune: Iden-

tifying popular recordings from brief excerpts. Psychonomic Bulletin & Review,

6 (4), 641–646.

Schellenberg, E. G. & von Scheve, C. (2012). Emotional cues in american popular

music: Five decades of the top 40. Psychology of Aesthetics, Creativity, and the

Arts, 6 (3), 196–203.

Schloss, J. G. (2014). Making beats: The art of sample-based hip-hop. Wesleyan

University Press.

Schroeder, J. E. & McDonagh, P. (2006). The logic of pornography in digital camera

promotion. In T. Reichert & J. Lambiase (Eds.), Sex in Consumer Culture: The

Erotic Content of Media and Marketing (pp. 219–242). Mahwah, NJ: Lawrence

Erlbaum Associates.

Serrà, J., Álvaro Corral, Boguñá, M., Haro, M., & Arcos, J. L. (2012). Measuring

the evolution of contemporary Western popular music. Scientiﬁc reports, 2 (521).

Sharot, T. & Phelps, E. A. (2004). How arousal modulates memory: Disentangling the

eﬀects of attention and retention. Cognitive, Aﬀective, & Behavioral Neuroscience,

4 (3), 294–306.

Simon, H. A. (1971). Designing organizations for an information-rich world. In

M. Greenberger (Ed.), Computers, Communication, and the Public Interest, vol-

ume 37 (pp. 37–72). Baltimore, MD: Johns Hopkins University Press.

143 Sirgy, M. J. (1986). Self-congruity: Toward a theory of personality and cybernetics.

Praeger Publishers/Greenwood Publishing Group.

Song, Y., Dixon, S., & Pearce, M. (2012). A survey of music recommendation sys-

tems and future perspectives. In 9th International Symposium on Computer Music

Modeling and Retrieval (pp. 395–410). London: Queen Mary University of London.

Storbeck, J. & Clore, G. L. (2008). Aﬀective arousal as information: How aﬀec-

tive arousal inﬂuences judgments, learning, and memory. Social and personality

psychology compass, 2 (5), 1824–1843.

Swift, T. (2014, July 7). For Taylor Swift, the future of music is a love story. The

Wall Street Journal. Retrieved from http://www.wsj.com/articles/for-taylor-

swift-the-future-of-music-is-a-love-story-1404763219.

Tan, S.-L., Spackman, M. P., & Peaslee, C. L. (2006). The eﬀects of repeated exposure

on liking and judgments of musical unity of intact and patchwork compositions.

Music Perception: An Interdisciplinary Journal, 23 (5), 407–421.

Tang, S. (2015, September 5). Visualizing artist collaborations in the Billboard

top 10 songs. Medium. Retrieved from https://medium.com/fun-with-data-

and-stats/visualizing-artist-collaborations-in-the-billboard-top-

10-songs-ff6188a0f57b.

Temperley, D. (2007). The melodic-harmonic ’divorce’ in rock. Popular Music, 26 (2),

323–342.

Thorngate, W. (1990). The economy of attention and the development of psychology.

Canadian Psychology/Psychologie canadienne, 31 (3), 262–271.

144 Traut, D. (2005). “Simply Irresistible”: Recurring accent patterns as hooks in main-

stream 1980s music. Popular Music, 24 (1), 57–77.

Trust, G. (2014, March 26). Top 50 single word songs: One-word wonders. Bill-

board. Retrieved from http://www.billboard.com/articles/list/5944890/top-

50-single-word-songs-one-word-wonders.

Twenge, J. M. & Campbell, W. K. (2001). Age and birth cohort diﬀerences in self-

esteem: A cross-temporal meta-analysis. Personality and Social Psychology Review,

5 (4), 321–344.

Twenge, J. M. & Campbell, W. K. (2003). ‘Isn’t it fun to get the respect that we’re

going to deserve?’ Narcissism, social rejection, and aggression. Personality and

Social Psychology Bulletin, 29 (2), 261–272.

Twenge, J. M. & Campbell, W. K. (2008). Increases in positive self-views among high

school students birth-cohort changes in anticipated performance, self-satisfaction,

self-liking, and self-competence. Psychological Science, 19 (11), 1082–1086.

Twenge, J. M. & Foster, J. D. (2010). Birth cohort increases in narcissistic person-

ality traits among american college students, 1982–2009. Social Psychological and

Personality Science, 1 (1), 99–106.

Twenge, J. M., Konrath, S., Foster, J. D., Campbell, W. K., & Bushman, B. J.

(2008). Egos inﬂating over time: A cross-temporal meta-analysis of the narcissistic

personality inventory. Journal of Personality, 76 (4), 875–902.

145 Vickers, E. (2010). The loudness war: Background, speculation, and recommenda-

tions. In Audio Engineering Society 129th Convention (pp. 4–7). San Francisco,

CA.

Vigliensoni, G. & Fujinaga, I. (2016). Automatic music recommendation systems:

Do demographic, proﬁling, and contextual features improve their performance?.

In Proceedings of the 17th International Society for Music Information Retrieval

Conference (pp. 94–100). New York.

Vitz, P. C. (1966). Aﬀect as a function of stimulus variation. Journal of experimental

psychology, 71 (1), 74–79. von Appen, R. & Frei-Hauenschild, M. (2015). AABA, refrain, chorus, bridge, pre-

chorus song forms and their historical development. German Society for Popular

Music Studies e. V., 13, 1–83.

Vuoskoski, J. K., Thompson, W. F., McIlwain, D., & Eerola, T. (2012). Who enjoys

listening to sad music and why? Music Perception: An Interdisciplinary Journal,

29 (3), 311–317.

Ward, M. K., Goodman, J. K., & Irwin, J. R. (2014). The same old song: The power

of familiarity in music choice. Marketing Letters, 25 (1), 1–11.

Weinreich, H., Obendorf, H., Herder, E., & Mayer, M. (2008). Not quite the average:

An empirical study of web use. ACM Transactions on the Web, 2 (1), 1–31.

Williamson, V. J., Jilka, S. R., Fry, J., Finkel, S., Müllensiefen, D., & Stewart,

L. (2012). How do ‘earworms’ start? Classifying the everyday circumstances of

involuntary musical imagery. Psychology of Music, 40 (3), 259–284.

146 Williamson, V. J. & Müllensiefen, D. (2012). Earworms from three angles. In Proceed-

ings of the 12th International Conference on Music Perception and Cognition and

the 8th Triennial Conference of the European Society for the Cognitive Sciences of

Music (pp. 1124–1133). Thessaloniki, Greece.

Wolin, L. D. (2003). Gender issues in advertising—An oversight synthesis of research:

1970–2002. Journal of Advertising Research, 43 (1), 111–129.

Wu, T. (2016). The attention merchants: The epic scramble to get inside our heads.

New York, NY: Alfred A. Knopf.

Yerkes, R. M. & Dodson, J. D. (1908). The relation of strength of stimulus to rapidity

of habit-formation. Journal of comparative neurology, 18 (5), 459–482.

147