Discovering Individual Music-Listening Profiles: Relationship between Music-Induced Emotions, Music-Listening Behaviour, Musical Reward Experiences and Individual Differences

Discovering Individual Music-Listening Profiles: Relationship between Music-Induced Emotions, Music-Listening Behaviour, Musical Reward Experiences and Individual

Differences

Eline L. Bekkers University of Toronto

Abstract

Emotional experiences with music are highly variable. Therefore, it is important to study the complex nature of these experiences, by looking into the relationship between several aspects of music-induced emotions, musical pleasure experiences, music-listening behaviour and individual differences. Sixteen participants listened to 30 short Western classical/opera musical excerpts while continuously indicating their emotional experiences on a valance-arousal grid. From this, average frequencies and intensities of music-induced emotions within each of the four valance-arousal quadrants were calculated. Additionally, participants‟ musical reward experiences were measured using the Barcelona Musical Reward Questionnaire (BMRQ). Lastly, six aspects of music-listening behaviour (e.g., frequency of music listening and the importance of music in people‟s lives) were measured in an online questionnaire, as well as the individual factors of gender, level of musical training and music preferences. A Partial Least Squares analysis including these factors revealed one significant latent variable. The pattern shows that

participants who tended to rate our music as Pleasant-Relaxing use music to regulate their mood, listen to music frequently, and have a preference for country, R&B/soul and top 40 music. These findings uniquely show a specific way in which music-induced emotions can be rewarding and how this relates to music-listening behaviour and individual differences. It is a good start in the quest to characterise individual profiles of emotional musical experiences. We hope more such profiles can be discovered, leading to more in-depth knowledge on why human populations across the world participate in musical activities.

Introduction

The development and persistence of musical behaviour across all cultures is a fascinating phenomenon, the cause of which has not yet been discovered. Steven Pinker (1997) states that music has no benefit for survival. He calls music „auditory cheesecake‟, a means to “tickle the sensitive spots … of our mental faculties” (p. 534), such as language and auditory scene analysis. On the other hand, music/musicality is often seen as an evolutionary adaptation, as it may have been beneficial for social bonding and group cohesion (Kirschner & Tomasello, 2010; Trehub, Becker & Morley, 2015).

One thing that both parties agree on is that cross-culturally, individuals tend to engage in musical activities because they enjoy it. A study on people‟s concept of pleasure showed that music was rated on the eighth place, just after „sun‟ and „sex‟ when participants were rating how pleasurable they found different experiences (Dubé & Le Bel, 2003).

Several factors may contribute to these so-called musical reward experiences, such as music‟s capacity to bring individuals together in groups (e.g., Cross, 2001; Freeman, 1998) or its capacity to stimulate sensory-motor brain networks (Zatorre, Cheen & Penhune, 2007). However, the most extensively researched factor is music‟s strong impact on human emotions. Music-Induced Emotions are Rewarding

While discussing music and emotions, it is important to distinguish perceived and felt emotion: “perceived emotion … refers to the perception of an intended or expressed emotional character, whereas felt emotions reflect the introspective perception of psychophysiological changes, which are often associated with emotional self-regulation” (Kreutz, Ott, Techmann, Osawa & Vaitl, 2007, p. 3). As such, the emotion that people perceive while listening to a piece of music does not necessarily agree with the emotion that people feel. For example, some individuals might experience positive emotions such as nostalgia, peacefulness and wonder while listening to music that expresses sadness (Vuoskoski, Thompson, McIlwain & Eerola, 2012).

Although some researchers have criticised the notion that music is able to induce emotions in the listener, evidence from physiological and neuroimaging studies clearly support this idea. Psychophysiological studies have shown how music expressing different emotions elicits different patterns of cardiac, temperature and skin conductance measures (Krumhansl, 1997). Neuroimaging studies have demonstrated how musical pieces with varying degrees of pleasantness are correlated with changes in regional cerebral blood flow (Blood, Zatorre, Bermudez & Evans, 1999). Moreover, intensely pleasurable experiences with music engage neural circuitry important for reward and motivation, such as the ventral striatum and

dorsomedial midbrain (Blood & Zatorre, 2001), and these experiences coincide with dopamine release in the mesolimbic reward system (Salimpoor, Benovoy, Larcher, Dagher, & Zatorre, 2011). For a more complete overview of evidence and measurements of musically induced emotions, see Juslin & Västfjäll (2008), p. 562-563.

Several studies have investigated the nature, prevalence and intensity of these music-induced emotions, such as Juslin and Laukka‟s survey study (2004) as well as studies using Experience Sampling Method (ESM; Juslin, Liljeström, Västfjäll, Barradas & Silva, 2008). In these studies, participants reported that they experience emotions about 55-64% of the time while listening to music. Music appears to be able to most easily induce basic emotions, such as happiness and sadness, although it is not uncommon to experience more complex emotions such as nostalgia, longing and frustration (Juslin & Laukka, 2004). These studies also indicate that positive emotions are most commonly experienced while listening to music, as participants named „happy‟, „relaxed‟, „calm‟, „moved‟ and „nostalgic‟ as the top five most common emotions in the Justlin and Laukka study. In the ESM study by Juslin et al. (2008), the emotional categories „calm-contentment‟ and „happiness-elation‟ were most commonly experienced. This is in concurrence with another self-report study, in which negative emotions, such as guilt, shame, disgust and embarrassment were identified as „nonmusical emotions‟ (Zetner, Grandjean & Scherer, 2008). Although people in the Western world tend to experience more positive than negative emotions in non-musical settings (Diener & Diener, 1996), it is interesting to note that feelings of joy, elation, nostalgia, peacefulness and amazement are significantly more commonly experienced during musical episodes compared to non-musical ones (Juslin & Laukka, 2004; Zetner et al., 2008). Individual emotional responses to music can be very powerful, with people reporting intense, euphoric responses accompanied by physical experiences such as shivers, tears, lump in the throat, goose bumps and increased heart rate (Gabrielsson & Wik, 2003; Sloboda, 1991).

There is some evidence that suggests how one‟s emotional experiences with music can actually drive their music-listening behaviour. In the study by Juslin and Laukka (2004) 64% of their sample (N = 141) stated they listened to music daily (versus Once a day 18%; A couple of times a week 16%; Once a week 1%; A couple of times a month 1%; A couple of times a year 1%). More interestingly, people reported they used music to “vent their emotions” (Never 3%, Seldom 33%, Often 61%, Always 3%) and have music “evoke strong, emotional memories” (Never 2%, Seldom 37%, Often 59%, Always 2%) . For the basic question “why do you listen to music?” the analysis of participants‟ free responses (inter-coder agreement, k = 0.77) showed the largest category was “to express, release, and influence emotions” (47%). Furthermore,

factor-analyses on reasons-for-music-listening items in two studies with adolescent participants (ages 13 to 16 years old) revealed three factors (North, Hargreaves & O'Neill, 2000; Tarrant, North & Hargreaves, 2000). In both studies, these factors could be interpreted as „fulfilling emotional needs‟, „creating external expression‟ and „fulfilling social needs‟, with the emotional needs factor explaining 28.0% and 16.7% of the variance in both studies respectively. Moreover, in a follow-up study on people‟s concept of pleasure (Dubé & Le Bel, 2003), 59% of the participants assigned music to the category of „emotional pleasure‟, versus 15% for „social pleasure‟, 11.5% for „intellectual pleasure‟, and less than 10% for both „general pleasure‟ and „physical pleasure‟.

These results suggest that music‟s ability to induce emotions in the listener is a contributing factor for why people listen to music, and might therefore be a key factor in the development and persistence of musical activities in human populations.

How are Music-Induced Emotions Rewarding?

Several studies have investigated how music can induce an emotional response in the listener and how this leads to the experience of pleasure during music listening. These pleasurable emotional experiences with music can arise through several mechanisms. Firstly, certain acoustic characteristics of a musical piece, such as volume and tempo, may regulate arousal through activation of the brainstem. This brain structure is involved in the control and mediation of the cardiovascular and respiratory systems (Paxinos & Mai, 2004) and receives input from limbic structures involved in emotional arousal (e.g., the amygdala; Silvestri & Kapp, 1998) and structures involved in early auditory processing (e.g., the thalamus; Koelsch & Siebel, 2005). This makes it possible for us to quickly and automatically respond to sounds that might imply danger or threat, such as sudden, loud and dissonant noises (Ploog 1992). Such sounds will increase arousal and induce feelings of unpleasantness (Burt, Bartolome, Burdette & Comstock, 1995; Halpern, Blake & Hillenbrand, 1986). However, up- or down-regulation of arousal in response to sounds may also induce positive, rewarding feelings such as vitality and peacefulness respectively (Zald & Zatorre, 2011). For instance, one might use music to increase energy levels during exercise. On the other hand, music can be used to soothe, as when singing a lullaby.

Additionally, music listening can be rewarding through direct induction of positive emotions. This can happen as a result of emotional conditioning, where a certain musical piece has repeatedly co-occurred with an emotional event (e.g., Blair & Shimp, 1992; Razran, 1954), thus triggering a certain emotion upon hearing the piece again even in the absence of the initial event it was paired with. For instance, the film score of an emotional movie scene may induce

emotions even when heard outside the context of the movie. Such associations are formed automatically and unconsciously (Fulcher & Hammerl, 2005).

Direct induction of emotions can also occur through emotional contagion, meaning that we „catch‟ the emotion that the music is expressing because we internally mimic the perceived emotion (e.g., Koelsch et al., 2002; Lundqvist, Carlsson, Hilmersson & Juslin, 2008). This may lead to direct activation of the relevant emotional representation in the brain (Juslin & Västfjäll, 2008). On the other hand, emotional contagion may arise through the mirror neuron system, which has been thoroughly studied in the field of emotion perception and social cognition (de Gelder, 2006). It is hypothesised that certain characteristics of the music may show similarities to the expression of a certain emotion (Kivy, 1981). For example, the expression of happiness and excitement in music often happens through loud, fast sounds with abrupt changes, which are arguably similar to the physical expression of happiness or excitement (fast, high-energy movements). Therefore it is possible that emotionally expressive music activates mirror neurons in the motor system, thereby activating motor patterns associated with the emotion, which then causes the internal experience of that emotion through peripheral feedback from the muscles.

A final mechanism of direct induction of emotion through music-listening deals with the problem of sad-sounding, yet rewarding music. For a long time, researchers have been struggling with why people voluntarily choose to listen to music which may cause negative affect. Scherer (2004) argues that the negative affect caused by sad music should be seen as an aesthetic emotion, which is detached from concerns about our goals, bodily needs or social values, and is more reactive in nature than other „utilitarian‟ emotions. These types of emotions allow us to engage with an art form on a deeper level, which can make it very rewarding despite the negative affect we might experience.

Music can also induce emotions indirectly. Listeners may conjure up visual images while listening, which can function as internal triggers for emotional responses (Quittner & Glueckauf, 1983; Plutchik, 1984). Additionally, music can trigger a specific memory from a person‟s life, which is accompanied by the emotion that is associated with it (e.g., Baumgartner, 1992). Lastly, listeners may experience emotions from music listening through the confirmation, violation or delay of musical expectancies (Huron, 2006; Meyer, 2008), which are formed based on the syntactic structure of the music (Patel, 2003), and are highly dependent on previous experiences with music (Sloboda, 1989). The dopamine-striatal reward system is often mentioned as the source of these expectancy-dependent affective responses to music, as it is responsible for the release of dopamine in response to repetition (facilitating expectancy confirmation; Wilson,

1979), unexpected events (leading to violation of expectancies; Schultz, 1998) and delay of resolution (creating stronger „yearning‟; Huron, 2006).

Musical Experiences are Highly Variable

Self-report studies on music-listening demonstrate how the high prevalence of musical behaviour across populations may, to a large extent, be caused by music‟s effectiveness in inducing emotions in the listener, which can be experienced as pleasurable through several different mechanisms.

However, perhaps the most interesting fact about musical experiences is the wide variation that exists between individuals. The self-report measure by Juslin and Laukka (2004) demonstrates that the proportion of time people experience strong emotions while listening to music ranged from 5% to 100%. Moreover, the study shows that there is wide variation in the type of music that induces emotions in the listener, with participants naming „popular music‟ (47%), „classical music‟ (29%) and „calm music‟ (22%), as the three types most likely to evoke emotions. On the other hand, a single piece of music can have a range of effects on different people. Some individuals may experience positive emotions while listening to sad-sounding music (i.e. music which expresses sadness), while others may not. This effect has been widely studied, as it shows that the induction of emotions during music listening is not as clear-cut as one might expect. Garrido and Schubert (2011) found that half of their sample (N = 59) experienced pleasure to some degree while listening to sad-sounding music, while the other half did not. Additionally, there are a number of different ways in which music-listening can be rewarding. A confirmatory factor analysis of a questionnaire about musical reward experiences has identified five reliable scales of musical reward (Mas-Herrero, Marco-Pallares, Lonrenzo-Seva, Zatorre & Rodriguez-Fornells, 2013). According to the study, music can be rewarding because of its emotional impact, its ability to regulate our mood, its stimulation of our sensory-motor system, its effectiveness for bonding individuals into groups and the increase of musical knowledge in the listener. Lastly, there are differences in the way people listen to music, such as how often they listen to music, the settings in which they like to do so and the amount of money they spend on music.

The variation in emotional responses to music has been related to several individual differences, such as personality (Kallinen & Ravaja, 2004), musical training (Kawakami, Furukawa, Katahira, Kamiyama & Okanoya, 2013b), music preferences (Kreutz et al., 2007) and gender (Nater, Abbruzzese, Krebs & Ehlert, 2006). Research on these factors‟ relationship with music-induced emotion is quite scarce and the results are far from conclusive. Studies on gender differences have mostly focussed on physiological measures. For instance, women reported

„chills‟ in response to music more frequently than men (Panksepp, 1995). Moreover, sex differences in psychophysiological measures in response to music have been found, showing that women tend to be hypersensitive to aversive musical stimuli (Nater et al., 2006). The results on the influence of musical experience are not univocal. Where some studies found that participants with higher levels of musical experience were more likely to report feeling pleasant emotions in response to „sad‟, minor-key music (Kawakami et al., 2013b), others found no differences between people with high and low levels of musical experience at all (Bigand, Vieillard, Madurell, Marozeau & Dacquet, 2005; Kawakami et al., 2013a). Lastly, musical preferences seem to have an influence on the type of emotions people experience and the intensity of these emotions. In a study using only classical music excerpts, participants‟ ratings of both the intensity and specificity of positive emotions (i.e. happiness and peacefulness) were positively correlated with their preference for classical music (Kreutz et al., 2007).

Current Study

The large variation in musical experiences calls for a more detailed investigation of the link between behaviour, emotions and pleasure experiences. By looking at the way specific aspects of these phenomena influence each other, we will be able to better characterise the individual profiles of musical experiences. Firstly, it remains unclear which aspects of music-induced emotions, such as the type of emotion and the frequency and intensity of these emotional experiences are related to the musical pleasure experience. Secondly, although evidence from self-report measures suggests that people engage in musical behaviour to fulfil emotional needs, it remains unclear which aspects of musical behaviour are reinforced by emotion induction (e.g., the frequency with which people listen to music, the amount of money they spend on music or how often they seek out new music). Furthermore, it is important to investigate how emotional experiences with music are related to different aspects of musical pleasure experiences. How, if at all, do these distinct factors of musical pleasure experiences  in particular the emotional impact and mood regulation factors  relate to music-induced emotions and music-listening behaviour? Lastly, as suggested by evidence mentioned above, it is important to look into the influence of individual differences, such as gender, music preferences and level of musical training.

The current study uniquely investigates the relationship between these more specific aspects of an individual‟s experiences with music. Based on the highly variable nature of these experiences, we expected to identify several ways in which music-induced emotions are related to pleasure experiences and music-listening behaviour, thus discovering unique profiles of musical behaviour which are possibly mediated by individual differences.

Method Participants

Sixteen participants between 18 and 35 years of age participated in this study (8 males, 8 females; mean age = 26.6). They had no known hearing problems or neurological problems. People with varying levels of musical training participated, ranging from no musical training at all to professional musicians. Participants were recruited through the Baycrest Health Sciences participant database and contacted via phone or email.

Apparatus and Materials

The test session took place in a sound isolation room. Participants sat in an arm chair in front of a 19 inch computer monitor (1280x1024 pixels; 60Hz), in reach of a computer mouse that they used to provide the emotional ratings. Two loudspeakers were situated on either side of the monitor and were used to play the musical pieces.

Participants rated their emotions continuously, as the music played, by moving a mouse cursor in a two-dimensional grid, with the emotional valence on the x-axis (Unpleasant and Pleasant) and the emotional arousal on the y-axis (Relaxing and Stimulating). The use of such a two-dimensional valence-arousal model for measuring emotional states is very common in the affective sciences and is well represented in research on music-induced emotions (Scherer, 2004). This model is often found more desirable than discrete emotional models (i.e., using labels such as „happiness‟, „sadness‟ and „anger‟), because it can capture a very wide range of emotions, while still being very intuitive for participants, without the need for discrete, predefined labels. Furthermore, it does not require participants to verbalise their emotional states, which would interfere with the listening process. Because this method it is very intuitive for participants, it yields highly reliable responses. Secondly, it allows participants to continuously rate their feelings, providing researchers with a very detailed account of the moment-to-moment emotional responses to each musical excerpt. In terms of the analysis, it nicely shows similarities between emotions in terms of neighbourhood in space (Feldman, Barrett & Russell, 1999). Lastly, it is very useful for quantifying the strength of the experienced emotions by looking into the position of the ratings within the grid‟s quadrants.

The valance-arousal grid and the musical excerpts were presented on the computer monitor and via the loudspeakers through Presentation software (version 16.3, 2004). The same software was used to record the position of the mouse cursor (X and Y coordinates) every 33.33ms as participants moved it to any place in the grid, corresponding to how the music made them feel.

Participants‟ musical-listening behaviour, level of music training and music preferences were measured in an online questionnaire. They answered six questions about their music-listening behaviour on a 4- to 6-point scale. Additionally, they reported their level of musical training on a 5-point scale and rated on another 5-point scale how often they listen to 21 different musical genres. An overview of the questions is shown in Table 1.

Table 1.

Questions on Music-Listening Behaviour, Musical Training and Musical Preferences in the Online Questionnaire.

Category Questions

Music-listening behaviour How important is music in your life? How often do you listen to music? How large is your music collection?

How much money do you spend on music each month? How important is discovering new music to you? Have you ever experienced “shivers-down-the-spine”, “goosebumps” or “chills” while listening to music? Musical training How much musical training have you had in the past? Music preferences What type of music do you listen to? (Rated for the

following music genres) Popular (top 40) 80‟s Retro Rock Classic Rock Hard Rock Alternative/Indie R&B/Soul HipHop/Rap Blues Christian/Gospel Classical Country Folk Jazz Reggae/Ska Metal Vocal/Easy-Listening World Latin Dance/Electronic Oldies (50‟s/60‟s)

Additionally, participants filled out the Barcelona Musical Reward Questionnaire (BMRQ; Mas-Herrero et al., 2013). This questionnaire consists of 20 statements measuring people‟s reward experiences from music listening on five scales: Emotion Evocation, Mood Regulation, Sensory-Motor reward, Social Reward and Music Seeking. Each scale is measured by averaging the 5-point Likert scale ratings of four statements („completely disagree‟  „completely agree‟), which are presented in a mixed order. An overview of each of the scales with one exemplary statement is shown in Table 2. The authors tested the validity of these factors using confirmatory factor-analysis and concluded there was an acceptable fit (CFI = .99, GFI = .99, RMSEA = .07 with 90% confidence interval .06; .09). Filling out the online questionnaire and BMRQ took approximately fifteen minutes. The full BMRQ can be found in the Appendix.

Table 2.

Overview of BMRQ Scales with an Example Statement taken from the Questionnaire

BMRQ scale Example statement

Emotional Evocation: „I get emotional listening to certain pieces of music.‟ to induce emotions

Mood Regulation: „Music calms and relaxes me‟

to improve mood/to relax

Sensory-Motor: „When I hear a tune I like a lot I can‟t help to move body to the beat tapping or moving to its beat.‟

Social Reward: „When I share music with someone I feel

for group bonding a special connection with that person.‟ Music Seeking: „I inform myself about music I like.‟ to discover or know more about music

Stimuli

The valance-arousal grid was presented against a black background at the centre of the screen at the beginning of each trial. It consisted of a light-grey rectangle of 800x700 pixels with an x- and y-axis dividing the grid in four equal quadrants. The ends of the axes were labelled with white letters just outside the grid, at the left, right, top and bottom. The labels were „unpleasant‟, „pleasant‟, „stimulating‟ and „relaxing‟. The mouse cursor was shown as a red

square (20x20 pixels), and started in the middle of the grid (point 0, 0) for each trial. An example of the grid at the beginning of a trial is shown in Figure 1.

Figure 1. Valance-arousal grid as shown at the beginning of a trial.

Thirty-six music excerpts of 40 to 60 seconds were selected by a musician, so that they would likely evoke emotions corresponding to each of the four quadrants in the valance-arousal grid. All of the excerpts came from Western classical/opera music, since this type of music is very effective for evoking a wide range of emotions and is often designed to communicate and induce emotions in the listener (Kreutz et al., 2007). The music selection consisted of a mix of quite well-known pieces as well as more obscure ones, ranging from solo instrumental music to full orchestral, and from baroque to post-modern eras. The music excerpts were cut from the full musical piece using GarageBand software (Apple Inc., 2009) and chosen so that they would have the most emotional impact within the 40 to 60 seconds. Thirty excerpts were used for the experimental trials. Additionally, six pieces were used in practice trials. An overview of the musical pieces featured in this study is presented in Table 3. The pieces used in the experimental phase were presented in three different orders, counterbalanced across participants. The order of the practice trials was fixed.

Table 3.

List of Musical Pieces and Composers used in the Current Study

Task phase Musical piece Composer

Practice Recomposed Vivaldi – Spring Max Richter

Tissues Philip Glass

Chants d'Auvergne – Bailero Joseph Canteloube

Elan Linda Bouchard

Il Barbiere di Siviglia – Sestetto Gioachino Rossini

Pierrot Lunaire Arnold Schoenberg

Experiment Les Boréades  Act 4: Entrée de Polymie Jean Philippe Rameau Armide  Air sicilien Christoph Willibald Gluck Liebeslied – Widmung arranged by Franz Liszt

Zefiro Torna Claudio Monteverdi

Johannes-Passion – Es ist vollbracht J.S. Bach Cosi fan tutte  Soave sia il vento W.A. Mozart La Boheme  O soave fanciulla Giacomo Pucini

Adagio for Strings Samuel Barber

When David Heard Eric Whitacre

Eternal Light – Belief Howard Goodall

Rite of Spring Igor Stravinsky

Totentanz Franz Liszt

Requiem – Dies Irae Giuseppe Verdi Threnody for the Victims of Hiroshima Krzysztof Penderecki Tristan und Isolde – Act 3 Prelude Richard Wagner

Spiegel im Spiegel Arvo Pärt

Recuerdos de la Alhambra Francisco Tàrrega

Field of Stars Oliver Schroer

Opus 118 Intermezzo Johannes Brahms

Passion of Angels Marjan Mozetich

Nimrod Edward Elgar

Lakmé  Dôme épais, le jasmin Lèo Delibes

Nixon in China – Act 1 John Adams

Requiem – Tuba Mirum Giuseppe Verdi

Die Walküre Richard Wagner

The Housatonic at Stockbridge Charles Ives Recomposed Vivaldi – Summer 2 Max Richter

Disappointment Lake Charles Ives

Solstice Songs  No. 2 Interlude Andrew Staniland Il Barbiere di Siviglia – Largo al Factotum Gioachino Rossini

Procedure

Participants were asked to fill out the online Music-Listening Habits Questionnaire at home before coming in for the test session. At the session, they first read and signed the consent form and filled out the BMRQ. Then they took place in the sound proof room in which the music-listening task took place. Firstly, a short pure-tone audiogram was collected for both ears to make sure participants‟ hearing capabilities were sufficient for the music-listening task. Hearing capabilities were tested for frequencies of 250, 500, 1000, 2000, 4000 and 8000 Hz, starting out at an easily detectable dB level (30dB) and lowering until the just audible dB level was found1. The tones were presented through stereo in-ear earplugs one ear at a time, starting with the right. Participants were instructed to press a button whenever they thought they heard a tone.

For the music-listening task, the participants were instructed to rate their emotional response to the music. They were told that we were after their personal experiences with the music, meaning that there were no right or wrong answers. We asked them not to compare the pieces to each other and to take as long as they needed in between trials. They were instructed to try to rate how the music actually made them feel, instead of how they thought the music should make them feel. This was done to emphasize the difference between felt and perceived emotions. The task started with six practice trials, followed by the test trials. Each trial started with the appearance of the grid on the screen, with a randomly set delay between the grid onset and music onset of 4 to 6 seconds. The music clips had a fade-out of about 4 seconds. In between trials a small fixation cross was presented in the centre of the screen. Participants could move to the next trial by pressing the left mouse button.

Data Analysis

The 30Hz mouse position data was processed using RStudio software (RStudio team, 2013). For each trial, all of the x- and y-coordinate samples were categorized into one of four quadrants, ignoring the first couple of (0, 0) coordinates, as it took participants some time to evaluate their assessment of the music and start responding. The number of samples sorted into each quadrant was then divided by the total number of samples, in order to get the proportion of time spent in each quadrant. Additionally, the average x- and y-values in each of the quadrants were calculated. These calculated proportions and x- and y-values were then averaged over trials, to get the mean proportion of time spent in each quadrant and the mean location of the mouse

1

cursor in each quadrant for each participant. The averaged x- and y-values were combined into one value by calculating the Euclidian distance from point (0, 0). The mean proportion of time spent in each quadrant indicates the frequency with which participants experiences a certain emotion and the mean location of the mouse cursor (distance from (0, 0)) indicates the average intensity of that emotion.

For the BMRQ questionnaire, an average score on each of the five scales was calculated by averaging the scores on the items belonging to those scales. The music-listening behaviour items were treated individually, as were the scores on musical training and preferences questions.

Results Music-Listening Behaviour

Most of our participants stated music is quite important to them, with the largest amount of participants (n = 8) stating that music is one of the top 5 five most important things in their lives. Although there were two participants who seldom engage themselves with music, no-one in our sample stated that music is not important to them at all. The largest number of participants (n = 7) reported that they listen to music between 1 and 4 hours a day, while other answers ranged from just occasional radio listening to more than 4 hours of music listening a day. None of our participants stated they never listen to music. The ratings of the size of participants‟ music collections spanned all answer categories, from having no music collection at all to having over 5,000 tracks of music. Similarly, the amount of money participants spend on music every month varies greatly, although it is slightly skewed towards lower amounts, with five participants saying they do not spend any money on music. Our sample predominantly reported being open to hearing new music but said they do not go out of their way to expose themselves to it (n = 10). Lastly, there was a lot of variation in the experience of physical reactions while music listening, such as shivers, chills or goose bumps, with some participants never having such experiences while others reported having them very commonly and at specific moments in a song. A detailed overview of participants‟ answers to the music-listening questions can be seen in Figure 2.

Figure 2. Histograms of participants‟ music-listening habits in the current study, as measured in an online questionnaire.

Musical Pleasure Experiences

Participants scored above average (above three on a 5-point Likert-scale) on each of the five BMRQ scales. The highest score was for the Mood Regulation scale (M = 4.20). The lowest scores were for the Music Seeking scale (M = 3.44). An overview of the mean scores, ranges and standard deviations of the BMRQ scores is shown in Table 4. A paired-samples t-test showed that the high score on the Mood Regulation scale did not significantly exceed the scores on the Emotion Evocation, Sensory-Motor and Social Reward scales. It only differed significantly from the Music Seeking score mentioned above, t (15) = 4.279, p = .001, Cohen‟s d = 2.21. The complete t-test table can be found in the Table 5.

Table 4.

Overview of Range, Mean and Standard Deviations of BMRQ Scores for each of the Five Scales BMRQ scale Minimum Maximum M SD

Emotion Evocation 2.25 5.00 3.88 0.88 Sensory-Motor 2.50 5.00 3.92 0.73 Mood Regulation 3.00 5.00 4.20 0.57 Music Seeking 1.50 5.00 3.44 0.85 Social Reward 2.25 4.50 3.84 0.64 Total 2.95 4.65 3.86 0.46 Table 5.

Results of Paired-Samples T-Test of BMRQ Mood Regulation Scores with Each of the Other BMRQ Scales

BMRQ scales Mean Difference SD df t p Mood Regulation - Emotion Evocation 0.33 1.18 15 1.117 .282 Mood Regulation - Sensory-Motor 0.28 0.84 15 1.346 .198 Mood Regulation - Music Seeking 0.77 0.72 15 4.279 .001* Mood Regulation - Social Reward 0.36 0.71 15 2.033 .060 * significant at p > .05.

Emotion Ratings

The continuous emotion ratings provided by participants were analysed to get the frequency with which different types of emotions were experienced and the intensity with which these were experienced. Participants spent the most time in the Pleasant-Stimulating quadrant (mean proportion of time = 0.34), followed by Pleasant-Relaxing and Unpleasant-Stimulating quadrants (mean proportion of time = 0.28 for both quadrants). Participants spent very little time in the Unpleasant-Relaxing quadrant (mean proportion of time = 0.07). The frequencies of Pleasant-Stimulating, Pleasant-Relaxing and Unpleasant-Stimulating ratings did not differ significantly from each other (p = .132 for Unpleasant-Stimulating vs. Pleasant-Stimulating; p = .936 for Unpleasant-Stimulating vs. Pleasant-Relaxing; p = .193 for Pleasant-Stimulating vs. Pleasant-Relaxing). The frequency of Unpleasant-Relaxing ratings did differ significantly from the other three quadrant ratings, with t (15) = 7.520, p < .001, Cohen‟s d = 3.88 for the comparison with Unpleasant-Stimulating ratings, t (15) = 9.712, p < .001, Cohen‟s d = 5.02 for the comparison with Pleasant-Stimulating ratings and t (15) = 7.084, p < .001, Cohen‟s d = 3.66 for the comparison with Pleasant-Relaxing emotions.

The average intensity of emotional ratings in each of the quadrants was highest for the Unpleasant-Stimulating quadrant (mean intensity = 186.73), followed by the Pleasant-Stimulating quadrant (mean intensity = 182.13) and the Pleasant-Relaxing quadrant (mean intensity = 175.37). The mean intensity was the lowest for the Unpleasant-Relaxing quadrant (mean intensity = 97.43). Again, no significant differences were found between the intensities in the Pleasant-Stimulating, Pleasant-Relaxing and Unpleasant-Stimulating quadrants (p = .641 for Unpleasant-Stimulating vs. Pleasant-Stimulating; p = .472 for Unpleasant-Stimulating vs. Pleasant-Relaxing; p = .696 for Pleasant-Stimulating vs. Pleasant-Relaxing). The intensity ratings for the Relaxing quadrant did differ significantly from the Unpleasant-Stimulating quadrant (t (15) = 8.130, p < .001, Cohen‟s d = 4.20), the Pleasant-Unpleasant-Stimulating quadrant (t (15) = 9.005, p < .001, Cohen‟s d = 4.65) and the Pleasant-Relaxing quadrant (t (15) = 5.065, p < .001, Cohen‟s d = 2.62). Table 6 shows an overview of the average frequencies and intensities in each quadrant, along with the value ranges and standard deviations. We observed no significant differences in emotion frequencies and intensities between counterbalance orders (for ANOVA table see Appendix).

Table 6.

Overview of Range, Averages and Standard Deviations of Frequency and Intensity Ratings in Each Quadrant

Emotion rating Minimum Maximum M SD

Frequency U-S* 0.09 0.44 0.28 0.09 Frequency P-S 0.16 0.54 0.34 0.10 Frequency P-R 0.13 0.53 0.28 0.10 Frequency U-R 0.01 0.16 0.07 0.04 Intensity U-S 86.21 305.75 186.73 61.36 Intensity P-S 104.58 290.57 182.13 57.89 Intensity P-R 71.29 348.38 175.37 52.95 Intensity U-R 32.30 198.50 97.43 52.95

* U-S = Unpleasant-Stimulating, P-S = Pleasant-Stimulating, P-R = Pleasant-Relaxing, U-R = Unpleasant-Relaxing

The ranges of frequency and intensity ratings show there was considerable variation in emotional responses between participants. This variation can be seen more clearly in Figure 3 and 4, in which the locations of the mouse cursor throughout a musical piece are plotted for each participant, collapsed over trials. The heat maps show clear differences in the frequencies of certain emotion ratings. For instance, participant 11267 experienced mostly Pleasant-Stimulating emotions, participant 13581 mostly Pleasant-Relaxing emotions and participant 4000 mostly Unpleasant-Stimulating emotions. There is also variation in the intensity the emotions, with some participants keeping the ratings closer to the centre of the grid than others (e.g., participant 12608 vs. 13320).

Figure 3. Heat maps of mouse positions collapsed over trials for participant 1-8. Darker shades of red indicate a higher density of data points. Numbers in the quadrants show the average proportion of time spent in each quadrant. Green squares are average mouse positions in each quadrant.

Figure 4. Heat maps of mouse positions collapsed over trials for participant 9-16. Darker shades of red indicate a higher density of data points. Numbers in the quadrants are the average proportions of time spent in each quadrant. Green squares are average mouse positions in each quadrant.

Pearson correlation showed a strong negative correlation between proportions of time spent in Pleasant-Stimulating and Pleasant-Relaxing quadrant (N = 16, r = -0.65, p = .007). The correlation between Unpleasant-Stimulating and Pleasant-Stimulating proportions neared significance (N = 16, r = -0.48, p = .059). No further significant correlations were found between proportions of time spent in the quadrants. The average intensity measurements for the four quadrants all correlated positively and significantly with each other (N = 16, r = 0.79, p < .001 for Unpleasant-Stimulating and Pleasant-Stimulating intensities, N = 16, r = .70, p = .002 for Unpleasant-Stimulating and Pleasant-Relaxing intensities, N = 16, r = .71, p = .002 for Unpleasant-Stimulating and Unpleasant-Relaxing intensities, N = 16, r = 0.63, p = .009 for Stimulating and Relaxing intensities, N = 16, r = 0.77, p < .001 for Pleasant-Stimulating and Relaxing intensities, and N = 16, r = 0.71, p = .002 for Unpleasant-Relaxing and Pleasant-Unpleasant-Relaxing intensities).he frequency of Pleasant-Stimulating emotions correlated significantly and negatively with Unpleasant-Stimulating (N =16, r = -0.63, p = .010) and Pleasant-Relaxing intensities (N = 16, r = -0.54, p = .032). These results can be found in Figure 5.

Figure 5. Correlations of average frequencies and intensities of emotions in each valance-arousal quadrant. U-S = Unpleasant-Stimulating, P-S = Pleasant-Stimulating, P-R = Pleasant-Relaxing, U-R = Unpleasant-Relaxing.

Individual factors

A one-way ANOVA with gender as between-groups factor showed that women reported stronger Pleasant-Stimulating emotions (M = 211.97, SD = 55.81) than men (M = 152.29, SD = 45.05), F (1, 14) = 5.538, p = .034, 2 = 0.01. The intensities of emotions in the other quadrants did not differ significantly between men and women. The same goes for the frequencies with which the emotions were experienced. The full ANOVA table is shown in Table 7.

Table 7.

One-Way ANOVA with Frequencies and Intensities of Emotions in each Quadrant as Dependent Variables and Gender as Between-Subjects Factor

men women Emotion ratings M SD M SD F (1, 14) p Frequency U-S* 0.27 0.11 0.29 0.07 0.189 .670 Frequency P-S 0.34 0.11 0.35 0.11 0.087 .772 Frequency P-R 0.32 0.12 0.25 0.07 1.889 .191 Frequency U-R 0.06 0.05 0.09 0.03 2.321 .150 Intensity U-S 166.70 51.76 206.74 66.87 1.794 .202 Intensity P-S 152.29 45.05 211.97 55.81 5.538 .034** Intensity P-R 166.76 100.31 183.98 76.87 0.149 .706 Intensity U-R 80.43 43.73 114.44 58.61 1.730 .210 *U-S = Unpleasant-Stimulating, P-S = Pleasant-Stimulating, P-R = Pleasant-Relaxing, U-R = Unpleasant-Relaxing. ** significant at p < .05.

Seven participants had no formal training in the past at all and did not play an instrument. Two participants had no formal training but did play an instrument, another two had less than one year of formal training, and another two had between one and five years of formal training. Three participants had more than five years of formal training. We found no effect of the level of musical training on the emotional responses of our participants. Musical training scores did not correlate significantly with either the frequency or the intensity of emotions in any of the quadrants. These results can be found in Table 8.

Table 8.

Correlations of Average Frequencies and Intensities of Emotional Responses in each Arousal-Valance quadrant with the Level of Musical Training

Emotion ratings Correlation coefficient r p

Frequency U-S* 0.18 .512 Frequency P-S 0.11 .700 Frequency P-S -0.35 .182 Frequency U-R -0.07 .788 Intensity U-S -0.01 .980 Intensity P-S 0.06 .833 Intensity P-S -0.09 .737 Intensity U-R -0.26 .334

*U-S = Unpleasant-Stimulating, P-S = Pleasant-Stimulating, P-R = Pleasant-Relaxing, U-R = Unpleasant-Relaxing

Participants on average preferred to listen to some music genres more often than other genres (measured on a 5-point Likert scale). The most commonly listened to genres were top 40, R&B/soul, HipHop/Rap, and Dance/Electronic, with mean scores between 3.13 and 3.56. The least preferred genres were Metal, Latin, Christian/Gospel, World and Jazz, with mean scores ranging from 1.50 to 1.81. There was large variation between participants, though, since the majority of the genres received scores ranging from 1 to 5 (listening to a genre „Never‟ to „Most Often‟). An overview of participants‟ musical preferences can be seen in Figure 6.

Figure 6. Histograms of participants‟ musical preferences in the current study.

Since the musical experts in our study were limited to Western classical/opera music, we expected to find an effect of participants‟ preference for classical music on their emotional responses. However, Pearson correlation did not show any significant relationships between classical music preference scores and the frequency or intensity of emotional responses, as is shown in Table 9.

Table 9.

Correlations of Average Frequencies and Intensities of Emotional Responses in each Arousal-Valance Quadrant with the Preference for Classical Music

Emotion ratings Correlation coefficient r p

Frequency U-S* -0.14 .610 Frequency P-S 0.12 .649 Frequency P-S -0.07 .792 Frequency U-R -0.10 .717 Intensity U-S -0.21 .439 Intensity P-S -0.17 .530 Intensity P-S 0.06 .818 Intensity U-R 0.10 .708

*U-S = Unpleasant-Stimulating, P-S = Pleasant-Stimulating, P-R = Pleasant-Relaxing, U-R = Unpleasant-Relaxing

Partial Least Squares analysis

In order to characterise the relationship between the emotion ratings, music-listening behaviour, musical pleasure experiences and individual factors, we performed a Partial Least Squares analysis. This method can be used to measure distributed patterns, and is particularly suited for data where the number of variables exceeds the number of observations, as is the case in the current study. PLS searches for components (latent vectors; LV) derived from the independent variables, so that these components explain as much of the covariance between the independent and dependent variables as possible (McIntosh & Lobaugh, 2004).

We ran a „behaviour PLS‟ analysis in MATLAB, with the average frequencies and intensities of emotions in the four quadrants as the eight dependent variables and BMRQ scores, music-listening behaviour scores, musical training score and musical preferences scores as 32 independent variables. A permutation test with 500 permutations was included, as well as a bootstrap analysis with 100 samples.

The permutation test revealed one significant LV (p < .001), which reflected a relationship between the frequency of Pleasant-Relaxing emotions, the intensity of emotions in all quadrants, the scores on the Mood Regulation scale of the BMRQ, how often participants listen to music, and preferences for country music, r&b/soul and top 40. The 95% confidence interval calculated by bootstrapping confirmed the reliability of the independent variables Mood Regulation, frequency of music listening, and preference for country, R&B/soul and top 40 are reliable in this pattern. The results are shown in Figure 7.

-1 -0,8 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 0,8 1 C o rr e la ti o n r -4 -3 -2 -1 0 1 2 3 4 5 6 Z-sc o re s

Figure 7. Top: z-scores for the frequency and intensity of emotion ratings for four quadrants (U-S = Stimulating, P-S = Pleasant-Stimulating, P-R = Pleasant-Relaxing, U-R = Unpleasant-Relaxing). Bottom: correlation of BMRQ scale scores, music-listening behaviour scores, level of musical training and music genre preferences with the latent variable, with 95% confidence interval calculated by bootstrapping.

Discussion

In this study, we took a closer look at the relationship between several aspects of music-induced emotions, music-listening behaviour, musical reward experiences and individual factors. We expected our results to provide us with a clearer insight into how music-induced emotions can be rewarding and how this relates to people‟s engagement in musical activities.

Our PLS analysis revealed one significant latent variable, in which the frequency of Pleasant-Relaxing emotions, the overall intensities of all types of emotions, the score on the BMRQ Mood Regulation scale, the frequency of music listening and the preference for country, R&B/soul and top 40 music reliably grouped together (according to bootstrap analysis). This pattern characterises a group of people whose emotional reactions to music are quite strong, who often experience music to be calming, who listen to music quite often, who mostly use music to calm themselves down or to keep themselves company and mostly listen to country, R&B/soul and top 40 music. This pattern uniquely and interestingly ties together aspects of music-induced emotions (frequency of Pleasant-Relaxing emotions) with musical pleasure experiences (Mood Regulation scores), music-listening behaviour (frequency of music listening) and individual factors (music preferences).

The experience of Pleasant-Relaxing emotions in response to our musical excerpts, rated continuously on a two-dimensional Valance-Arousal grid as the music played, was quite frequent. Calculation of the average proportion of time spent in each quadrant of the grid (frequency of emotions) and the average location of the mouse cursor in each quadrant (intensity of emotion) showed that participants most frequently experienced Pleasant-Stimulating emotions, although not significantly more than Pleasant-Relaxing and Unpleasant-Stimulating emotions. Interestingly, Unpleasant-Relaxing emotions were significantly less commonly experienced than each of the other types of emotions. Although we tried to select musical pieces which would induce emotions in each of the four quadrants of the valance-arousal grid, the lack of Unpleasant-Relaxing ratings is not surprising. Scherer (2005) gives a clear and elaborate overview of the specific emotions which fall into the four quadrants of the valance-arousal model (Figure 8). Note that in this grid the pleasant-unpleasant labels are on opposite sides compared to the grid used in the current study. The bottom-right quadrant in the figure represents Unpleasant-Relaxing emotions, which include boredom, depression and sadness as the main emotions. The former emotion was arguably seldom experienced in our study since our musical excerpts were only 60 seconds long, which may not have been long enough to induce boredom in our listeners.

Figure 8. Valance-arousal model of emotions with emotion labels falling into each quadrant, taken from Scherer (2005).

Sadness/depression in response to music listening has been widely studied because of the interesting phenomenon of pleasurable responses to sad music (e.g., Huron, 2011; Vuoskoski et al., 2012). It is probable that a large part of our sample generally experiences pleasure from listening to sad music, therefore not rating the „sad-sounding‟ excerpts that were presented as Unpleasant-Relaxing, but rather as Pleasant (-Relaxing or - Stimulating). In fact, the enjoyment of sad-sounding musical pieces was mentioned by several participants in the debrief interviews we conducted. Moreover, three of the less prevalent emotions in the Unpleasant-Relaxing quadrant (guilt, shame and embarrassment) have previously been identified as „nonmusical‟

emotions, based on participants‟ ratings on the prevalence of emotions that are expressed or induced by music (Zetner et al., 2008). The other three „nonmusical emotions‟ (contempt, jealousy and disgust) fall in the Unpleasant-Stimulating quadrant, suggesting that the prevalent Unpleasant-Stimulating ratings in our study were driven mostly by the more common emotions in this quadrant (e.g., anger, fear and distress). The fact that the prevalence of Unpleasant-Stimulating emotions was equal to that of Pleasant (-Relaxing or - Unpleasant-Stimulating) emotions in the current study seems to contradict the findings on musical experiences in everyday life, which demonstrate that people more commonly experience positive rather than negative emotions while listening to music (Juslin & Laukka, 2004). Additionally, they show that positive emotions are significantly more commonly experienced in musical settings compared to non-musical settings (Juslin et al., 2008; Zetner et al., 2008). However, these contrasting results are easily explained by the fact that participants in our study, contrary to everyday musical experiences, had no control over the music they were exposed to. About this control, Juslin and Laukka (2004) said: “Given this choice, people will tend to prefer to listen to music that they like and that makes them „feel good‟” (p. 231). The high prevalence of positive emotions in their participants‟ experiences is therefore not surprising, nor is the substantial amount of Unpleasant emotional responses in the current study.

The average intensity of emotions was the strongest for the Unpleasant-Stimulating quadrant, though it was not significantly stronger than the intensities in the Pleasant-Stimulating and Pleasant-Relaxing quadrants. The average intensity of Unpleasant-Relaxing emotions was significantly weaker than the intensities in each of the other quadrants, showing that participants experienced these types of emotions not only infrequently but also weakly. The average intensities of emotions in each quadrant correlated positively, suggesting that participants tended to have an overall strong or weak emotional reactivity to the music. It is therefore not surprising that the intensities of emotions in all four quadrants were grouped together in our latent variable pattern.

We found no significant differences in the frequencies of the types of emotions between men and women in our sample. However, women‟s Pleasant-Stimulating emotions were significantly stronger than those of men. Previous research on gender differences in music-induced emotions does not concur with this finding. Nater et al. (2006) found that women had stronger psychophysiological reaction (finger temperature, skin conductance and heart rate) to „aversive‟ musical stimuli (in this case heavy metal music) than men. On the other hand, there are several studies on emotional reaction to music which report no gender differences at all (Lundqvist et al., 2008; Robazza, Macaluso & D'Urso, 1994). A possible explanation for our

findings is that our classical and opera musical pieces were quite calm compared to some other musical genres (e.g., hard rock or metal). Since it is known that women tend to rate calm music more positively than men (Christenson & Peterson, 1988), this might explain the stronger Pleasant-Stimulating emotions experienced by women in our sample. It is interesting, though, that we found a gender-effect only for Stimulating emotions and not for Pleasant-Relaxing ones, meaning that the gender effect is not limited to the valence of the emotional experiences but also influences the arousal experience.

The second reliable factor in our latent variable pattern is the score on the Mood Regulation scale of the BMRQ. The average score of this scale was the highest compared to the other four scales, and was with the average score of 4.20 far above the mean score of 3 (5-point scale). Although previous self-report studies found that “to express, release and influence emotions” (Juslin & Laukka, 2004) and “to fulfil emotional needs” (North et al., 2000; Tarrant et al., 2000) were by far the largest response categories to the question why people listen to music, the average score on the Emotion Evocation BMRQ scale was statistically equal to that of the Mood Regulation, Social Reward and Sensory-Motor reward scales. Moreover, these studies did not identify music‟s mood-regulatory effects as a separate factor that drives music listening. The clear distinction between music‟s effect on emotions and moods emerging from the current data is therefore quite surprising. It may be due to the fact that emotion- and mood-regulation were difficult to distinguish in earlier self-report measures (e.g., Juslin & Laukka, 2004), in which participants‟ answers were less structured compared to the BMRQ scales. In fact, slight rephrasing of the Mood Regulation statements in the BMRQ can transform them into statements that can be easily interpreted as referring to emotion manipulation rather than mood manipulation. For instance, the statement “Music keeps me company when I am alone” can be transformed into „music reduces loneliness‟ and the statement “music calms and relaxes me” into „music reduces feelings of distress”. Note that loneliness and distress are named as emotions in the Unpleasant-Relaxing and Unpleasant-Stimulating quadrants in the valance-arousal model (see Figure 7). Therefore, it is not surprising that previous studies  without the specifically worded statements in the BMRQ  did not distinguish music‟s effect on people‟s moods from its effect on emotions in the answers provided by their participants.

The link between Pleasant-Relaxing emotions and Mood Regulation scores seems quite obvious, since the moods that the BMRQ statements refer to are linked to the emotions in the Pleasant-Relaxing quadrant. For instance, the statements “Music calms and relaxes me” and “Music helps me chill out” are linked to emotions calm, relaxed and at ease, which are some of the main emotions belonging to the Pleasant-Relaxing quadrant, as shown in Figure 7. Therefore,

it is a logical pattern that people who often experience music to be Pleasant-Relaxing are the ones who score high on the Mood Regulation BMRQ scale. It is interesting to see that this pattern is related to the frequency with which people listen to music, as shown by our latent variable results. Although participants in earlier self-report studies mentioned that manipulation of emotions/moods is the main reason they listen to music, our study provides more direct evidence of a relationship between the frequency-aspect of music-listening behaviour and the reward people experience from music listening.

Lastly, our latent variable pattern included the individual factor of music preference. Our results indicate that people who (frequently) listen to music to regulate their mood preferably listen to country, R&B/soul and top 40 music. A causal relationship between these factors remains speculative from our results. We expected to find an effect of participants‟ preference for classical music, since this type of music was exclusively used in the current study. It seemed logical to assume that people who prefer classical music would have more positive emotional reactions while listening to our musical excerpts compared to people who hardly ever listen to this genre. However, unlike Kreutz et al. (2007), who demonstrated how people with a preference for classical music had more intense emotional responses to their classical music stimuli, we did not find any influence of a preference for classical music in either frequency or intensity of the experienced emotions.

We found no effect of the level of musical training in our analysis. This is not very surprising, given the fact that previous studies on the effects of musical training on emotional experiences with music have been far from conclusive. A common theory is that musically trained people have more experience with dissonant or minor-key music than the average music listener. Therefore, they are said to be more tolerant of dissonant music and might find it more pleasurable to listen to sad-sounding music than untrained people (Webster & Weir, 2005). Given this theory, we expected participants with high levels of musical training to rate our musical excerpts more positively than those with low levels of training. Similar results have been found in a study using minor-key musical stimuli (Kawakami et al., 2013b), in which musically trained participants‟ perceived emotions were significantly more negative than their felt emotions. However, in a second study no such difference between perceived and felt emotions was observed for participants with high levels of musical training (Kawakami at al., 2013a). Moreover, another study hypothesising that individuals with musical expertise would have stronger emotional reactions to music (Kreutz et al., 2007) surprisingly found the opposite effect, where individuals with low levels of musical training reported stronger emotional responses than those with a high level of training. Finally, Bigant et al. (2005) did not find any differences

between musicians and non-musicians and even went on to claim that emotional responses to music “are not subject to strong individual differences” (p. 1130). It should be clear from these findings that the possible effect of musical training on music-induced emotions is still a mystery and calls for more research.

The relaxing, calming effect that music appears to have according to the emerging pattern is most likely caused by the underlying mechanism of arousal-regulation, in which a brainstem reflex is triggered by characteristics of the music (Juslin & Västfjäll, 2008). Down-regulation of arousal is said to be related to feelings of peacefulness, which in turn correlates with positive affect (Zald & Zatorre, 2011). This fits well with the fact that our participants mostly rated music that they found Relaxing as Pleasant rather than Unpleasant. The use of music to soothe may have had a beneficial effect on our survival, mostly in the form of lullabies and child-directed songs, which are very effective for regulating infant arousal (Trehub & Trainor, 1998). Such beneficial effects may have arguably been a relevant factor in the evolution of musical behaviour (Fitch, 2006). In fact, singing lullabies to soothe infants seems to be a universal phenomenon, observed in most cultures (Trehub, 2000). It appears from our findings that music is not only useful in adult-infant interaction, but is also effectively used by adults to soothe themselves.

The highly individual nature of musical experiences suggested that there should exist multiple ways in which music-induced emotions can be related to the experience of pleasure while music listening and music-listening behaviour. However, in the current study only one significant and reliable pattern was discovered. This is possibly due to the relatively small sample size (N = 16), which may have caused a lesser degree of variation in some of our measures. It would be interesting to see whether additional profiles of music-listening could be discovered in a larger sample. Based on the findings that people tend to engage in musical activities to influence or induce emotional reactions, one would expect to find a differential pattern based around the use of music for emotional evocation. Another possible candidate is a more social profile of music listening, as social reward experiences are usually named as a third incentive for musical activity (e.g., North et al., 2000; Tarrant et al., 2000). Moreover, it is possible that such additional patterns will shed some more light on the influence of musical training and music preferences, which was either absent or difficult to explain in the current findings.

Conclusion

Our findings uniquely describe a detailed relationship between aspects of music-listening behaviour, musical pleasure experiences, individual differences and music-induced emotions. Although findings of this nature are complex, they contribute to our understanding of humans‟

highly personal experiences with music, by uncovering the ways in which musical emotions are enjoyable and how this influences our music-related behaviour. This knowledge may bring is closer to discovering why musical behaviour has prevailed over time and is still a universal phenomenon. The highly variable nature of musical experiences suggests that the single pattern emerging from our data is most probably not exclusive. We therefore hope to discover  as more data becomes available  more of these detailed relationships in the future.

Appendix

Barcelona Musical Reward Questionnaire

Each item of this questionnaire is a statement that a person may either agree with or disagree with. For each item, indicate how much you agree or disagree with what the item says. Please respond to all the items; do not leave any blank. Choose only one response to each statement. Please be as accurate and honest as you can be. Respond to each item as if it were the only item. That is, do not worry about being consistent in your responses. Choose from completely disagree (left) to completely agree (right) one of the five options:

[1] – Completely disagree; [2] – disagree; [3] – Neither agree nor disagree; [4] – agree; [5] Completely agree.

1. When I share music with someone I feel a special connection with that person. 2. In my free time I hardly listen to music.

3. I like listening to music that contains emotion. 4. Music keeps me company when I‟m alone. 5. I don‟t like to dance, not even with music I like. 6. Music makes me bond with other people. 7. I inform myself about music I like.

8. I get emotional listening to certain pieces of music. 9. Music calms and relaxes me.

10. Music often makes me dance. 11. I‟m always looking for new music.

12. I can become tearful or cry when I listen to a melody that I like very much. 13. I like to sing or play an instrument with other people.

14. Music helps me chill out.

15. I can‟t help humming or singing along to music that I like. 16. At a concert I feel connected to the performers and the audience. 17. I spend quite a bit of money on music and related items.

18. I sometimes feel chills when I hear a melody that I like. 19. Music comforts me.

Effect of Counterbalance Order

Emotion ratings were collapsed over counterbalance orders, as a one-way ANOVA with emotional frequencies and intensities as dependent variables and counterbalance order as the three-level between-subjects factor did not reveal significant differences for any of the dependent variables. The results are shown in Table 10.

Table 10.

Results of One-Way ANOVA with Counterbalance Order as Between-Subjects Factor and Frequencies and Intensities of Emotions in each Quadrant as Dependent Variables

M SD

Emotion ratings Order1 Order2 Order3 Order1 Order2 Order3 F (2, 13) p Frequency U-S* 0.21 0.31 0.32 0.10 0.08 0.04 3.270 .071 Frequency P-S 0.37 0.34 0.32 0.11 0.15 0.04 0.222 .804 Frequency P-R 0.33 0.25 0.26 0.13 0.10 0.05 0.967 .406 Frequency U-R 0.07 0.08 0.07 0.05 0.04 0.05 0.040 .961 Intensity U-S 169.76 195.77 198.04 53.92 72.45 67.38 0.336 .721 Intensity P-S 168.65 196.16 184.26 49.00 49.67 81.07 0.283 .758 Intensity P-R 195.61 160.41 166.04 93.65 96.85 83.07 0.239 .791 Intensity U-R 87.46 111.10 95.75 53.59 49.45 63.98 0.248 .784 *U-S = Stimulating, P-S = Pleasant-Stimulating, P-R = Pleasant-Relaxing, U-R = Unpleasant-Relaxing.