Content uploaded by Jason Corey
Author content
All content in this area was uploaded by Jason Corey on Mar 12, 2018
Content may be subject to copyright.
https://doi.org/10.1177/0305735617751050
Psychology of Music
1 –15
© The Author(s) 2018
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0305735617751050
journals.sagepub.com/home/pom
Extreme re-listening: Songs
people love . . . and continue
to love
Frederick Conrad, Jason Corey, Samantha
Goldstein, Joseph Ostrow and Michael Sadowsky
Abstract
Despite the lack of surprise each time they listen to their favorite song, people re-listen to these songs
many times. We explored “extreme re-listening” by conducting a survey about the song to which
participants were “listening most often these days.” We questioned participants about their listening
experience, e.g., the deepness of their connection to the song, which aspects of the song draw them back,
how much of the song they were able to hear in their heads, and how (in their own words) the song made
them feel, which we classified as “happy,” “calm,” and “bittersweet.” More participants whose favorite
song made them feel happy reported being drawn back because of its beat/rhythm. Participants whose
favorite song made them feel bittersweet reported having a deeper connection to the song than those
whose favorite song evoked other feelings. These patterns held irrespective of musical training. Finally,
we found that the more times they listened to their favorite song, the more of the song listeners could
hear internally. People’s affection for songs to which they voluntarily listen at high rates appears not to
wane as it does for songs to which their exposure is ambient as is the case with the hit parade.
Keywords
favorite songs, listening experience, musical affect, pop songs, re-listening
On my computer, the play count for the song “Randy Described Eternity” is 406. But I’ve also
listened to it in my car, on the subway and on YouTube. The song is from the 1997 Built to Spill
album Perfect From Now On, which turns 15 this year. And apart from playing a few other Built to
Spill records for variety…, I haven’t voluntarily listened to anything … [else] since May 2011.
– Arnold-Ratliff (2012)
Recorded music is ubiquitous and makes it easy to repeatedly listen to exactly the same song.
Never has it been so easy to intentionally listen to a favorite song over and over, such as through
streaming radio and music services. People may listen to favorite songs very frequently, even
University of Michigan, Ann Arbor, MI, USA
Corresponding author:
Frederick Conrad, University of Michigan, 426 Thompson Street, Ann Arbor, MI 48109, USA.
Email: fconrad@umich.edu
751050POM0010.1177/0305735617751050Psychology of MusicConrad et al.
research-article2017
Article
2 Psychology of Music 00(0)
though listening to familiar songs yields few surprises – and surprise has been identified as one
reason people find songs enjoyable (e.g., Huron, 2006). Yet, we assume listeners love the songs
they listen to most often; otherwise it seems unlikely they would continue to listen to them.
Re-listening and musical enjoyment
Margulis (2014) has identified re-listening to songs as a special case of experiencing repetition
in music and suggests that repeated exposure to a pleasurable stimulus produces a pattern of
enjoyment that follows the Wundt curve (e.g., Anand & Holbrook, 1986; Berlyne, 1971;
Sluckin, Colman, & Hargreaves, 1980) in which pleasure (hedonic value) increases with famili-
arity (exposure) until reaching a ceiling after which hedonic value diminishes (see Figure 1).
For example, Jakobovits (1966) demonstrated that songs played frequently on the radio and on
jukeboxes in the 1960s (the “Hit Parade”) grew in popularity until a critical point when this
growth reversed despite continued exposure. He demonstrated that this reversal was more
rapid the more times a song was played during its rise in popularity. He suggests that such
“exposure saturation” produces semantic satiation – an inhibition in neural response due to
rapid, repeated firing of neurons in response to repeated presentation of a stimulus.
So how is it that someone like Katie Arnold-Ratliff (2012) can listen to a song more than
400 times and still not get bored, as she describes in the epigraph? Jakobovits (1966) predicted
just the opposite – because more frequent exposure hastens the decline in popularity in general,
it will presumably reduce listening frequency for someone like Arnold-Ratliff in particular.
What might account for the difference in her experience from that of listeners contributing to
the Hit Parade in the 1960s? Is Arnold-Ratliff’s experience an extreme case or an example of a
different phenomenon than Jakobovits examined? One possibility is that, unlike most radio lis-
teners in the 1960s, most listeners today, including Arnold-Ratliffe, have access to their favorite
songs at all times and can listen to them on demand, affording them the time to become deeply
acquainted with and connected to such songs. As listeners replay a song, what might have
seemed simple initially turns out to be more complex as the listener discovers new layers of the
song after listening many times. Margulis (2014) proposes that the more complex the piece of
music, the longer the time required to traverse the Wundt curve, presumably because it takes
longer to get bored (Szpunar, Schellenberg & Pliner, 2004). This kind of intentional and active
repeated listening may produce a great deal of pleasure for prolonged periods despite increased
Figure 1. Wundt curve.
Conrad et al. 3
exposure, as listeners luxuriate in their favorite parts of the song. This should lead to a relation-
ship between hedonic value and familiarity that is shaped differently than the classic Wundt
curve, specifically one in which there is a plateau at peak enjoyment rather than a single inflec-
tion point, despite continued listening. It is this phenomenon – high levels of enjoyment over
extended periods of intentional exposure to the same recording – that we explore in this article.
We refer to the phenomenon as extreme re-listening.
The study
The main question we focus on is how listeners can intentionally listen to the same song over
and over, without losing interest. To this end, we examine the behaviors and emotional states
that accompany extreme re-listening, primarily through self-reports in response to questions in
a survey.
Method
One approach to studying extreme re-listening would be to present listeners with songs known to
have certain properties and measure their listening experience. Under these circumstances, the
participants would be asked to report experiences in their own words or choose from descriptions
provided by the researchers (as in Song, Dixon, Pearce, & Halpern, 2016). This approach has the
advantage of controlling the stimulus and promoting comparability across participants. However,
it is unlikely the experimental stimulus to which they are exposed will happen to be their favorite
song. A more ecologically valid approach is to measure a participant’s experience with the song
they report listening to most often at the time of the study. The drawback to this approach is that
it becomes impractical to measure listeners’ experience in real time because this would require
presenting a very large number of songs as stimuli, perhaps as many songs as there are partici-
pants. Nonetheless, we adopt the second approach and ask participants for summary or generic
reports about their experience while listening to these particular songs.
More specifically, our approach was to ask listeners questions about the one recorded song
they were “listening to most often these days.” While we did not ask listeners how much they
like the song they reported, we presume that they like it more than any other song at that time,
i.e., that hedonic value is high despite repeated listening. If listeners do eventually get bored
with the song, they are not bored with it at the time of the study. We will refer to these songs as
their “favorite” songs. Although it is possible that in absolute terms they have not listened to
this song very often, just more than others, the play counts reported below are large and give us
confidence that we are in fact sampling instances of extreme re-listening. Moreover, while it is
possible these songs were not actually their favorites at the time of the study, the data we report
later suggest that listeners afford these songs a preferred status.
We administered an online questionnaire1 from September 30 to October 9, 2013 to 204
“workers” from Amazon Mechanical Turk (https://www.mturk.com/mturk/welcome) in
which we asked 27 questions about that one song concerning objective behaviors (e.g., listen-
ing frequency) and subjective judgments (e.g., how the song made the listener feel). In addition,
we asked seven questions about background information (e.g., the amount of formal musical
training the participant had received) and standard demographics. The full questionnaire
appears in Appendix A in Supplemental Materials online. The participants were U.S. residents,
about evenly split between males (53%) and females (47%), and generally young with 51%
below 29 years of age, 26% between 30 and 39 years old, 14% between 40 and 49 years, and
9% above 50 years old. Fourteen percent had a high school education or less, 39% had some
4 Psychology of Music 00(0)
post-secondary education, 37% had earned a bachelor’s degree, and 10% had a graduate or
professional degree.
The questionnaire was programmed so that the song title provided by each listener was
inserted into his or her respective questions. We recommended to the participants (listeners)
that they identify the song with the highest play count in their music library as a way to select
the song about which they answered the questions but they were free to use other criteria.
We assumed that different listeners have different emotions in response to their favorite
songs. For example, if listeners generally feel happy when they repeatedly play a song they
might attend to different features of the song (e.g., the beat or the melody) than when they feel
calm. A critical step in the study was to see what emotions listeners experienced when listening
to the song by asking them “How does the song make you feel?” and then coding their open
responses into a small number of high-level affect categories. Because we did not know what
affect categories would emerge, the initial part of the study was exploratory. Having completed
the coding, we were able to develop hypotheses about how other aspects of the listening experi-
ence (based on answers to different questionnaire items) might vary depending on which affect
category was associated with a particular song.
Results
We first provide descriptive data about the listeners’ favorite songs as well as how frequently
they listened to the songs and engaged in certain behaviors while listening. We then present
results of the affect coding which enabled us to generate several hypotheses. Finally, we report
the outcomes of analyses conducted to test the hypotheses.
Descriptive data
About the songs. A complete list of listeners’ favorite songs can be found in Appendix B in Sup-
plemental Materials online. The 204 participants reported listening most often to 179 different
songs. Two songs were reported by four listeners each, one song by three listeners, and eight
songs by two listeners. In other words, only 11 songs were reported by more than one listener.
Some of the listeners’ favorite songs were in the charts at the time of the study, i.e., in the
9/21/13 Billboard Hot 100, but most of the songs were not. Of the 11 songs (out of the 179
unique songs) that were the favorite of more than one listener, 8 were in the Billboard Hot 100
at the time. However, of the remaining 168 songs only 11 were in the Hot 100 at that time. So,
while media exposure at the time of the study may have contributed to a small number of the
songs being listened to most often, it seems that in general listeners actively chose to listen to
most of the songs rather than being passively exposed to them.2
The songs fell into 10 genre categories according to AllMusic.com (see Figure 2), the major-
ity being pop/rock music. Note that none of the listeners’ favorite songs were Classical. Because
we asked about “songs,” some listeners might have excluded Classical music from considera-
tion. According to the participants, 94.6% (193 out of 204) of the songs had lyrics. The mean
duration of the songs was 258 s (i.e., 4 m 18 s, SD =114.8 s).
Listeners reported listening quite often – over 86% listened daily or weekly – to their favorite
songs. Table 1 presents the frequency of listening, i.e., percentages of listeners across five levels
of listening frequency. Most of the listeners listened to their favorite song quite frequently: for
the 43% of participants who reported listening at least once a day (n = 87), the median number
of listens per day was 3.00, that is, these listeners voluntarily played the exact same song three
times a day, every day. Clearly, these listeners were very engaged with these songs.
Conrad et al. 5
Sixty percent (n = 120) of all listeners listened to the song again, at least once, immediately
after listening once and about half of these participants (n = 57) listened more than twice in a
row. When asked “About how many times in total would you say you have listened to [the song]?”
respondents’ mean answer was 303.7 (SD = 754.0,3 Mdn = 100). Taken together these data sug-
gest there is an urgency for at least some listeners to hear their favorite song. In fact, when asked
“How urgently do you want to hear the song immediately before you play it?” the mean rating
(where 1 was not at all urgently and 5 was extremely urgently) was 3.53 (SD = 0.895), and 56.7%
of listeners rated their urgency as either 4 or 5. It is clear that relatively high levels of urgency to
hear one’s current favorite song often accompany the extreme re-listening we observed here.
The fact that many listeners expressed urgency to hear the songs and likely chose to listen to
them rather than being exposed through commercial channels, suggests that people have a
relatively deep connection to these songs. In order to examine this, we asked listeners “How
deep is your connection to [the song]?” and asked them to respond on a 5-point scale from not
at all deep to extremely deep. The mean rating was 3.73 (SD = 0.93) and 65.6% of the partici-
pants rated their connection as a 4 or 5.
While different listeners probably define “deep connection” in different ways, one behavior
that is associated with the depth of listeners’ connection to their favorite song is closing their
eyes while listening. Of those who had a deep connection with the song (rating of 4 or 5),
20.3% (24/118) reported closing their eyes while listening but only 8.1% (5/62) of those
whose connection was not deep reported closing their eyes, χ2(1) = 4.53, p < .05. This seems
likely to reflect listeners’ attempt to focus their attention on the music by reducing visual dis-
traction (see Koreimann, Gula, & Vitouch, 2014).
What brought listeners back to songs? There can be many reasons why people listen to the same
song repeatedly. One possibility is that listeners are particularly attracted to certain attributes
Figure 2. Percentage of listeners’ songs that fell into each of 10 musical genres.
Table 1. Frequency of self-reported listening.
% Listeners n
At least once a day 42.6 87
A few times a week 44.1 90
Once a week 5.9 12
A few times a month 3.9 8
Once a month or less 1.5 3
6 Psychology of Music 00(0)
of a song. To determine which attributes were most important for repeated listening, we pre-
sented listeners with a list of nine song attributes and asked them “Thinking about the song,
what is most important in repeatedly bringing you back to it?” As is evident in Figure 3, the
three attributes selected most often were “melody,” “beat or rhythm,” and “lyrics,” each of
which was selected by more than 50% of the participants. Consistent with the idea that the
melody was particularly important in attracting listeners back to the song, 71.1% reported that
they sing or hum along while listening to the song.
Summary of descriptive results. Our participants listened to their favorite (mostly pop/rock) songs
very frequently; many listened daily and many of those listened multiple times per day, includ-
ing multiple times in sequence. More than half reported an urgent desire to hear the song. And
more than 65% indicated their relationship to the song was “deep.” More of these listeners
closed their eyes while listening than their counterparts whose connection to the song was not
deep. The listeners re-listened mostly because of the lyrics, melody, and beat or rhythm.
Affective response and listening experience
Affect coding. By almost all accounts, listening to a song creates some emotional response in the
listener. For example, Krumhansl (1997) demonstrated that, based on physiological measures,
people can experience emotions when listening to music (the “emotivist” view) and do not sim-
ply recognize that a song expresses a particular emotion such as “sad,” “fear,” “happy” (the
“cognitivist” view). Listeners report intense emotional reactions in what Gabrielsson calls their
“strongest experience related to music” (e.g., Gabrielsson & Wik, 2003). Although the song one
is listening to most often these days does not necessarily involve listeners’ strongest musical
experience, one’s emotional response to a song seems likely to be magnified the more one lis-
tens to it, and participants in the current study listened many times (mean total listens was
303.7).
Based on listeners’ answers to the open question “Please tell us briefly how [the song] makes
you feel” two coders (who are authors) assigned each song to one or more of 34 affect catego-
ries (kappa = 0.81; 23 of the 204 cases were uncodable) identified by reading through the open
responses. We then identified three high-level affect categories that subsumed the 34 more
Figure 3. Percentage of listeners indicating song attributes most important in their decision to listen
repeatedly. Listeners could choose up to three attributes.
Conrad et al. 7
detailed categories: Happy/Energetic, Calm/Relaxed, and Bittersweet/Melancholy/Nostalgic. In
a second pass, the coders worked together to assign each listener’s open response to one of these
three categories.4 The majority (69%) of songs were assigned to the Happy/Energetic category
with the remainder evenly split between the other two categories. Table 2 shows examples of
how the song made listeners feel.
A song was assigned to the Happy/Energetic category (abbreviated as “Happy”) if the lis-
tener used words or phrases about positive affect such as “motivated,” “mood booster,” “fun,”
“happy,” “high energy,” “party,” and “upbeat.” A song was coded as making the listener feel
Calm/Relaxed (abbreviated as “Calm”) if the listener used terms like “calm,” “relaxed,” “sooth-
ing,” “peaceful,” and “at ease.” Finally, if listeners described their experience as a mix of emo-
tions such as feeling happy and sad at the same time, using words such as “bittersweet,”
“melancholy,” “nostalgic,” “wistful,” “haunting,” or “reminisce,” the song was assigned to the
Bittersweet/Melancholy/Nostalgic category (abbreviated “Bittersweet”).
Relationship between affect and listening experience. Having established the major categories into
which the emotional impact of the songs can be classified, we were able to develop hypotheses
about how listeners engage with and experience the songs that create these emotions. The lis-
tener affect emerging from open descriptions may be related to other aspects of the listening
experience. For example, it has been suggested that a fast tempo and regular rhythm are related
to feeling happy when listening to music (Gabrielsson, 2016; Gabrielsson & Juslin, 2003).
Table 2. Example responses to “How does [the song] make you feel?”, assigned to the three high-level
affect categories.
Happy/Energetic
(69%, n = 108)
Calm/Relaxed
(15%, n = 23)
Bittersweet/Melancholy/Nostalgic
(16%, n = 26)
“It makes me feel very upbeat
and energetic.”
“The song makes me feel like
I want to party.”
“It just makes me happy, it
puts me in a good mood.”
“Pumped up! Excited! Ready
to dance, sing, and love!”
“Just gets me pumped up, it’s
a really upbeat song and it
gets me going.”
“It makes me feel happy. Feel
upbeat and good about life
and myself.”
“It makes me feel calm and
at ease.”
“The song just calms me.
The combination of the
melody and Sting’s voice are
soothing.”
“Just helps relieve stress.”
“It makes me feel relaxed,
makes me feel young, makes
me miss my husband when
he’s not home.”
“It makes me feel at ease,
calm, and helps me to put
things into perspective.”
“It makes me feel more
relaxed than I was before.”
“It makes me feel sad. But not the bad
kind of sad, and I like singing with it.”
“It fills me with bittersweet feelings,
the sadness of losing those times, but
finding the joy in the memories.”
“It makes me feel like I’m understood,
but it also makes me think about a sad
part of my past.”
“It makes me feel nostalgic and
wistful. It makes me think about
people and places I used to know.”
“It brings me back to the first time I
heard it. A simpler time when I was
young.”
“It makes me feel emotional, with
mixed feelings, beauty and bittersweet
at the same time. I think that’s what
makes it haunting for me.”
8 Psychology of Music 00(0)
Although we did not ask listeners about tempo and regularity of rhythm, we did ask them if the
feature of the song that most led to repeated listening was “the beat or rhythm” (among other
options) so our first hypothesis was:
Hypothesis 1: For songs that make listeners feel happy, listeners are more likely to focus on
beat or rhythm than for songs that produce other feelings.
The Bittersweet descriptions seemed to arise from more moving and personally meaningful
experiences than Happy descriptions. For example (see Table 2), the listener who explained his/
her Bittersweet experience as “It makes me feel like I’m understood, but it also makes me think
about a sad part of my past” seems to be more introspective and internally oriented than is the
listener who described his Happy experience as “Just gets me pumped up, it’s a really upbeat
song and it gets me going.”
The idea that listeners can experience mixed emotions has been observed elsewhere
(e.g., Hunter, Schellenberg, & Schimmack, 2008; Larsen & Stastny, 2011) and suggests a
more complex and nuanced experience than what listeners’ experience when a song
makes them happy. We did not ask specifically about complexity or nuance of the experi-
ence but we did ask about the depth of the listener’s connection to his or her favorite song
which has some of this character: a complex emotional experience is more likely to be
significant and memorable5 than one which is superficial. Therefore, our second
hypothesis was:
Hypothesis 2: When songs make listeners feel bittersweet they are likely to have a “deeper
connection” to the song than are listeners whose favorite songs make them happy.
Because the songs we asked listeners about were the songs they were listening to most often at
the time of the study, it seemed possible that listeners had developed relatively detailed internal
representations of these songs. Such musical imagery has been described as “hearing music
apart from the presence of external sound” (Cohen, 2015). If so, listening to and imagining the
song could be relatively isomorphic. Bailes (2015) captured the richness and completeness of
musical imagery in an experience sampling study of the general public, eliciting reports such as
“I was singing it in my head from beginning to end” (p. 61). One question this raises is whether
musical imagery becomes more veridical as listeners are increasingly exposed to the song.6 It
has been demonstrated that memory for other material is improved with more exposure, e.g.,
memory for individual words (von Hippel & Hawkins, 1994) or for advertisements (Goldstein,
McAfee, & Suri, 2012). It seems likely this is also the case for music, i.e., the more a listener lis-
tens to her favorite song the more of the song she will be able to “hear” in her head. This led to
our third hypothesis:
Hypothesis 3a: The more times people listen to the song, the more of the song they can hear
in their heads.
The number of times people listen to a song might be affected by the depth of their relationship
with the song. If they feel a deep connection they are likely to listen more times which, if
Hypothesis 2 is supported, leads to the following:
Hypothesis 3b: When a song makes listeners feel bittersweet they can hear more of the song
in their heads than when the song creates other feelings.
Conrad et al. 9
In the next section we test these hypotheses.
Song attributes and listeners’ affect (Hypothesis 1). Consistent with Hypothesis 1, more listeners
identified the beat or rhythm as the key reason they listened to the song repeatedly if the song
made them feel happy than if the song made them feel bittersweet (59% for happy vs. 19% for
bittersweet), t(45) = 4.35, p < .001. Marginally fewer listeners who indicated that their
favorite song made them feel bittersweet reported listening because of the beat or rhythm
than if the songs made them feel calm (19% for bittersweet vs. 43% for calm), t(41.8) = 1.84,
p = .07. Consistent with the finding that listeners who return to songs that made them feel
happy are attracted by the beat or rhythm, 44.4% of listeners reported that they tap their
feet, clap their hands, or drum on furniture when the song made them feel happy versus
17.4% when the song made them feel calm and 23.1% when the song made them feel bit-
tersweet, χ2(2) = 8.53, p = .014. This demonstrates that the song attributes that listeners
indicated were important in repeatedly bringing them back to the song differed based on how
the song made them feel.
Deepness of connection and affect category of song (Hypothesis 2). While listeners were relatively
unlikely to focus on the beat or rhythm of songs that made them feel bittersweet, substantially
more of these listeners reported having a deep connection (ratings of 4 or 5) to these songs
(88%) than did listeners whose favorite song made them feel either happy (57%), t(53.07) =
4.4, p < .001, or calm (65%), t(45.24) = 2.23, p = .03. The fact that more listeners reported a
deep connection to songs that made them feel bittersweet than to songs that created other emo-
tions is consistent with the idea that they felt moved when listening to these songs: a moving
experience is inherently more profound emotionally (much like the experience of ter Bogt,
Mulder, Raaijmakers, & Nic Gabhainn’s [2011] “high involved” listeners) than one which gets
the listener “pumped up.” Recall that listeners are more likely to close their eyes when listening
to a song with which they have a deep connection so there seems to be something personal and
transformative about the experience of songs that create bittersweet feelings.
Predicting affect category. The fact that more listeners have a deep connection to songs that make
them feel bittersweet, i.e., whose songs’ beat or rhythm is not responsible for re-listening, sug-
gests that there is an inverse relationship between deepness of connection and beat or rhythm.
To test this idea we modeled the relationship between beat/rhythm and deepness of connection
on the one hand and the likelihood that a song creates feelings from a particular category on
the other hand. More specifically, in a logistic regression model, the relationship between feel-
ing bittersweet and re-listening because of the beat/rhythm is negative, β = -1.664, p = 0.01,
but the relationship between feeling bittersweet and deepness of connection is positive, β =
0.940, p = 0.003.
It is possible that these relationships are affected by listeners’ awareness of the songs’ attrib-
utes, which might in turn be affected by listeners’ musical background. For example, experi-
enced musicians may well be able to detect – and name – structural details in songs that
non-musicians cannot distinguish (see Sloboda, Davidson, & Howe, 1994). In fact, we found no
interaction between musical background (measured on a 4-point scale borrowed from Honing
& Ladinig, 2009) and beat/rhythm or deepness of connection suggesting that these relation-
ships are not restricted to musicians.
Internal listening and amount of exposure. (Hypothesis 3a). Because of the large number of times the
participants listened to these songs it is likely their memory for the songs is quite good. The
10 Psychology of Music 00(0)
question is, how good? Listeners with the best memory for the song should be able to “hear”
large amounts of the song in their heads, potentially including all the instrumental and vocal
sounds. In fact, 98.5% of listeners said they could hear the song when not listening to it. We
then asked these participants how much of the song they could hear in their heads (“a few sec-
onds,” “10–20 seconds,” “up to a minute,” “more than a minute but not the whole song/piece,”
“the whole song/piece”) and modeled their answers on the basis of the total reported listens.
Consistent with Hypothesis 3a, as the number of times they listened to the song increased, so
did the amount of the song they could hear in their heads. Figure 4 plots the amount of the
song that is internally audible against total reported listens, log transformed because the range
was so large. The relationship in an ordinal logistic regression is highly significant, t(162) =
4.728, p < .0001. The listeners who reported the highest number of listens could generally
hear more than a minute and in many cases all of the song in their heads, although, because
the data are log-transformed, this suggests that disproportionately more exposure is required to
increase one’s memory for the song the more times one listens. Similar non-linear relationships
between exposure and recall have been reported in domains as diverse as advertising (Anand &
Holbrook, 1986) and playing chess (Gobet & Simon, 2000).
Presumably hearing the song in one’s head requires some auditory representation of the
song, and presumably the more one can hear in one’s head, the richer the listener’s auditory
memory for the song. A possible reflection of a listener’s auditory representation for a song is
the ability to sing or hum the song while listening to it. A large percent of listeners (72.2%)
reported singing or humming along to the song; of these, a larger percent (82.4%) reported
being able to hear more than a minute of the song internally than those (66.7%) who reported
being able to hear a minute or less in their heads, χ2(1) = 5.41, p = .02.
Internal listening and affect category. (Hypothesis 3b). While the songs the participants identified
were those they were listening to most often at the time of the study, the number of times they
ever listened to the song could well have been related to how the song made listeners’ feel.
This was in fact the case. Listeners whose favorite songs made them feel bittersweet reported
listening to the song 790.2 times on average, whereas when the song made them feel calm lis-
teners reported listening 515.0 times on average, and when the song made them feel happy the
mean reported number of listens was 174.9, F(2, 127) = 5.98, p = .004.7 Clearly, the number
Figure 4. Amount listeners could hear in their heads as a function of the (log) number of reported
listens.
Conrad et al. 11
of times these listeners listened to their favorite song was related to how it made them feel. And
it seems that “bittersweet” was a feeling they chose to reproduce substantially more often (and
an astonishingly large number of times) than “calm,” which, in turn, was a feeling they chose
to re-experience more than songs that made them happy. Thus it seems likely that people want
to hear songs that make them feel bittersweet more frequently than songs that make them feel
happy.8 Perhaps because the number of listens was highest for songs that made listeners feel
bittersweet, more listeners from this group (30.8%) were able to hear the entire song in their
heads than in the other groups (12% and 13% for happy and calm, respectively), χ2(2) = 5.78,
p = .06, consistent with Hypothesis 3b.
Summary of affective response and listening experience. Listening to their favorite songs made par-
ticipants feel primarily happy, calm or bittersweet. When songs made listeners feel happy they
were most likely to focus on the beat or rhythm. Listeners had a deeper connection to songs that
made them feel bittersweet than to those that made them feel happy or calm. The more times
participants listened to their favorite song, the more of the song they could hear in their heads
although the effect of repeated listens on the amount that was internally audible weakens as
the number of listens increases. Listeners reported listening most to songs that make them feel
bittersweet (i.e., to which they had the deepest connection), followed by songs that make them
feel calm, listening least to songs that make them happy. Finally, more listeners were able to
hear the entire song in their heads if the song made them feel bittersweet than if it made them
feel happy or calm.
Discussion and conclusions
What is it about listeners’ experience when replaying the songs they love that enables them to
listen many times without growing tired of the songs – after any novelty and surprise are sure
to be gone? One possibility is that when listeners feel happy or calm, novelty may not be a prior-
ity. They may use these songs to regulate their mood and may want the same experience each
time, much as they would want mood altering medication to work the same way each time it is
taken. The analogy between music and medication has been drawn before (e.g., Chanda &
Levitin, 2013) and playing music to regulate affect has been well documented (e.g., Baltazar &
Saarikallio, 2016; Saarikallio & Erkkilä, 2007; Skånland, 2013). While this may also be true of
songs that make listeners feel bittersweet, mixed emotions are inherently more complicated
than a single emotion, so there may be other reasons that listeners return to these songs repeat-
edly. One possibility is that because their feelings are complicated, listeners are comforted by
knowing that at least someone else (the artist) has experienced these contradictory feelings.
But how do listeners come to experience contradictory feelings?9 The “tragedy paradox”
(e.g., Sachs, Damasio, & Habibi, 2015) account of mixed emotions in music listening implies
the emotions are experienced distinctly, at least initially: a song creates sad feelings and the lis-
tener likes the song, thus experiencing sadness and pleasure separately but in response to the
same stimulus. Huron (2011) offers an alternative account, which also separates the experi-
ence of sadness and pleasure in response to a song. He proposes that music which makes listen-
ers feel sad triggers a hormonal response in some listeners, which has a psychologically
consoling effect, making the experience both negative and positive. Yet by other accounts (e.g.,
Juslin, 2013) listeners feel conflicting emotions simultaneously and directly from the music.
Consistent with this idea, Kivey (1990) argues that a song can create feelings in listeners that
are simultaneously mournful (negative) and moving (positive). Some listeners in the current
study articulated this kind of experience, e.g, “It makes me feel emotional, with mixed feelings,
beauty and bittersweet at the same time.” And other listeners seem to directly experience
12 Psychology of Music 00(0)
different feelings on different listening occasions: “The song makes me happy sometimes, but
sad other times.” It’s possible that some listeners experience mixed emotions directly from the
musical stimulus while for others the different emotions have distinct origins. Our data – open
responses, provided retrospectively – cannot definitively address this. Follow-up research with
real-time measures may shed light on this issue.
Whether listeners feel one or multiple emotions when listening to a favorite song, do they
experience these feelings the first time they listen, or do their feelings evolve with repeated lis-
tening? Pop songs are generally easy to listen to for the first time, and as Frith (2007) says, pop
music “is designed to appeal to everyone” (p. 169), but we propose that unless they reveal new
features and layers and take on new meaning with repeated listening, listeners will eventually
lose interest consistent with the Hit Parade trajectory observed by Jakobovits (1966). Consider
the experience of this listener in the current study: “The song title is, Royals, by Lorde. As soon
as I heard it, I liked it. At the time the song had a more alternative feel to it. But now, I hear it all
the time on the radio and does not have the appeal it did when I first heard it.” The evolution of
listeners’ emotional reaction to songs they love deserves additional research, likely conducted
longitudinally rather than through the cross-sectional approach used in the current study.
Because we have only asked listeners about their favorite song, our data do not allow us to
compare their listening experience for these songs to that of less beloved songs. We cannot say,
for example, what proportion of all their listening is dedicated to their favorite song and we can-
not say to what degree our findings apply to songs they listen to less often. But to the extent that
one is interested in listeners’ experience when they listen to songs they are fond of and to which
they listen frequently, it makes sense to examine their experience for the songs they are most
fond of and listen to most frequently. The way they represent these songs internally and the role
those representations play while they listen is almost sure to differ from how they represent
songs they listen to less frequently.
The information that listeners represent seems likely to include not only musical attributes
but also associated emotions, thus enabling listeners to re-experience the same emotions each
time they listen. This suggests to us that representations of songs that are listened to extremely
often are likely to be dynamic, triggered each time listeners hear the song. Their associated emo-
tions may be bound to specific aspects of the song such as tempo, pitch, timbre, and song struc-
ture or to the entire song, e.g., “It makes me remember fond[ly] … my years in college” or “It
makes me feel nostalgic and wistful.” The current data cannot distinguish between these pos-
sibilities but it would seem that at least some of the affective experience unfolds during listen-
ing, suggesting that listeners “play” this representation as they listen, allowing them to
anticipate particular feelings as the song progresses. Knowing that a beloved emotional state is
“around the corner” – Huron’s (2006) “sweet anticipation” – may be part of why these emo-
tions are experienced as intensely as they seem to be.
Objective attributes of songs can certainly be related to how they make listeners feel time
after time. In the current data, a song’s salient beat and rhythm seems to contribute to listeners
feeling happy. And in the literature, songs that are associated with sadness have certain attrib-
utes such as lower pitch, slower tempo, minor key tonality, and dark tone qualities (e.g., Sachs
etal., 2015). However, we propose that the objective aspects of the song do not necessarily lead
to the same subjective experience – and internal representation – for all listeners. Instead, it
seems, two listeners could love the same song for very different reasons, e.g., the song’s beat for
one listener and the flood of memories triggered by the lyrics for another. Moreover, personality
may affect re-listening: people suffering from major depressive illness are predisposed to rumi-
nate about their condition which increases the likelihood that they will listen to songs that
prolong their depression (e.g., Garrido, 2009). People’s relationships with their favorite songs
Conrad et al. 13
seem to be individualized and idiosyncratic, in contrast to the mass popularity of hit songs
investigated by Jakobovits (1966). In the current study, it is likely that listeners’ initial exposure
was more serendipitous (recall that only about 6% of the 189 unique songs were in the music
charts at the time of the study) and their continued exposure was more deliberate, i.e., listeners
chose when to expose themselves to the particular song that was their favorite at the time of the
study. This individualized listening experience is an example of what Anderson (2009) calls the
“long tail,” i.e., the transition in the digital era from a small number of hits (determined by
industry “taste makers”) to a vast number of niches (each driven by a small number of indi-
viduals). Niche listening may enable listeners to develop the kind of personally meaningful
relationships with particular songs that allows their affection for those songs to persist across
very large amounts of exposure.
Acknowledgements
We thank Katie Arnold-Ratliff, David Huron, Daniel Levitin, Sile O’Modhrain, Michael Schober and Kaidi
Wu for their advice and helpful suggestions. We thank Carolyn Lau for help with analyses.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this
article.
Ethical approval
Ethical approval for this project was given by University of Michigan Institutional Review Board
HUM00075398.
Notes
1. Via Qualtrics (http://www.qualtrics.com/).
2. It is possible their affection for the songs was waning at the time of the study but it seems more likely
their relationship with the song was ongoing when we questioned them: when asked to describe how
the song made them feel, 202 out of 204 listeners reported their feelings in the present tense. Taken
together, these data suggest to us that the songs about which we questioned listeners were personally
meaningful to them at the time of the study.
3. Mean number of listens and standard deviation were calculated after removing the two observations
that were greater than two standard deviations above the unadjusted mean.
4. The affect category to which a listener’s emotional response is assigned is less about preference for a
musical genre (e.g., Rentfrow, Goldberg, & Levitin, 2011) than about the emotional experience listen-
ing to that one song. In fact most of the songs in the study are from the same broad, Pop/Rock genre,
as shown in Figure 2. Further, the fact that we identified three high-level affect categories into which
all of the listeners’ descriptions could be assigned does not mean that these three categories are the
only categories to which any listener’ experiences can be assigned. For example, Zentner, Grandjean,
and Scherer (2008) present a nine-factor taxonomy (the Geneva Emotional Music Scale) in which
listeners’ emotional responses to music can be classified.
5. Memory is known to be better when the experience is emotional rather than neutral because it
leads to deeper processing of the experience (e.g., Reber, Perrig, Flammer & Walther, 1994; Ritchey,
LaBar & Cabezza, 2011). If so, then processing more complex affect should lead to more memorable
experiences.
6. In this article we are concerned with voluntary as opposed to involuntary musical imagery also
known as “earworms,” “stuck song syndrome,” “brain worms,” and “sticky music” (e.g., Beaman &
Williams, 2010), which also involves internal listening but is generally unintentional and typically
does not involve listeners’ favorite songs.
14 Psychology of Music 00(0)
7. Responses greater than two standard deviations above the mean were removed from the analyses of
total number of listens. There were two such outliers, both involving songs that the listeners reported
made them feel bittersweet.
8. We cannot rule out an explanation of these results that assumes people who like feeling bittersweet
tend – for some reason – to like listening to the same song over and over while those whose favorite
songs make them feel happy may just like more variety or are more open to new experience. What
is indisputable is that the participants in this study listen to songs that make them feel bittersweet
substantially more than songs that create other feelings.
9. The contradictory feelings that these listeners reported may be more than a musical phenome-
non, but one that is experienced across the arts as an aesthetic judgment of the sort discussed by
Juslin, Harmat, and Eerola (2014), among others. For example, very similar phenomena have been
observed among film viewers. Larsen, McGraw, and Cacioppo (2001) reported that viewers experi-
enced mixed emotions more frequently after than before watching the film Life is Beautiful, which
combined humor and tragedy. Hanich, Wagner, Shah, Jacobsen, and Menninghaus (2014) propose
that, at least in the context of sad films, viewers enjoy being moved and the sadness intensifies the
feeling of being moved.
References
Anand, P., & Holbrook, M. B. (1986). Chasing the Wundt curve: An adventure in consumer esthetics. In
Richard J. Lutz (Ed.), Advances in consumer research volume 13 (pp. 655–657). Provo, UT: Association
for Consumer Research.
Anderson, C. (2009). The long tail: Why the future of business is selling less of more. New York, NY: Hyperion.
Arnold-Ratliff, K. (2012, April 24). Songs I can’t let go. Salon. Retrieved from https://www.salon.
com/2012/04/24/songs_i_cant_let_go/
Bailes, F. (2015). Music in mind? An experience sampling study of what and when, towards an under-
standing of why. Psychomusicology, 25(1), 58–68.
Baltazar, M., & Saarikallio, S. (2016). Toward a better understanding and conceptualization of affect self-reg-
ulation through music: A critical, integrative literature review. Psychology of Music, 44(6), 1500–1521.
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. (1971). Aesthetics and psychobiology. New York, NY: Appleton-Century-Crofts.
Chanda, M. L., & Levitin, D. J. (2013). The neurochemistry of music. Trends in Cognitive Sciences, 17(4),
179–193.
Cohen, A. J. (2015). Special issue on musical imagery: About those earworms? Psychomusicology, 25(1),
1–2.
Frith, S. (2007). Taking popular music seriously: Selected essays. New York, NY: Routledge.
Gabrielsson, A. (2016). The relationship between musical structure and perceived expression. In S.
Hallam, I. Cross, & M. Thaut (Eds.), The Oxford handbook of music psychology (pp. 215–232). Oxford,
UK: Oxford University Press.
Gabrielsson, A., & Juslin, P. N. (2003). Emotional expression in music. In R. J. Davidson, K. R. Scherer,
& H. H. Goldsmith (Eds.), Handbook of affective sciences (pp. 503–534). Oxford, UK: Oxford University
Press.
Gabrielsson, A., & Wik, S. L. (2003). Strong experiences related to music: A descriptive system. Musicae
Scientiae, 7(2), 157–217.
Garrido, S. (2009). Rumination and sad music: A review of the literature and a future direction. In C.
Stevens, E. Schubert, B. Kruithof, K. Buckley, & S. Fazio (Eds.), Proceedings of the 2nd International
Conference on Music Communication Science (pp. 20–23). Sydney, Australia: HCSNet, University of
Western Sydney.
Gobet, F., & Simon, H. A. (2000). Five seconds or sixty? Presentation time in expert memory. Cognitive
Science, 24(4), 651–682.
Goldstein, D. G, McAfee, R. P., & Suri, S. (2012). Improving the effectiveness of time-based display
advertising. Proceedings of the 13th ACM Conference on Electronic Commerce. New York, NY: ACM.
Conrad et al. 15
Hanich, J., Wagner, V., Shah, M., Jacobsen, T., & Menninghaus, W. (2014). Why we like to watch sad
films: The pleasure of being moved in aesthetic experiences. Psychology of Aesthetics, Creativity, and
the Arts, 8(2), 130–143.
Honing, H., & Ladinig, O. (2009). Exposure influences expressive timing judgments in music. Journal of
Experimental Psychology: Human Perception and Performance, 35(1), 281–288.
Hunter, P. G., Schellenberg, E. G., & Schimmack, U. (2008). Mixed affective responses to music with con-
flicting cues. Cognition & Emotion, 22(2), 327–352.
Huron, D. B. (2006). Sweet anticipation: Music and the psychology of expectation. Cambridge, MA: MIT press.
Huron, D. (2011). Why is sad music pleasurable? A possible role for prolactin. Musicae Scientiae, 15(2),
146–158.
Jakobovits, L. A. (1966). Studies of fads: I. the “Hit Parade”. Psychological Reports, 18(2), 443–450.
Juslin, P. N. (2013). From everyday emotions to aesthetic emotions: Towards a unified theory of musical
emotions. Physics of Life Reviews, 10(3), 235–266.
Juslin, P. N., Harmat, L., & Eerola, T. (2014). What makes music emotionally significant? Exploring the
underlying mechanisms. Psychology of Music, 42(4), 599–623.
Kivey, P. (1990). Music alone: Philosophical reflections on the purely musical experience. Ithaca, NY: Cornell
University Press.
Koreimann, S., Gula, B., & Vitouch, O. (2014). Inattentional deafness in music. Psychological Research,
78(3), 304–312.
Krumhansl, C. L. (1997). An exploratory study of musical emotions and psychophysiology. Canadian
Journal of Psychology, 51, 36–52.
Larsen, J. T., McGraw, A. P., & Cacioppo, J. T. (2001). Can people feel happy and sad at the same time?
Journal of Personality and Social Psychology, 81, 684–696.
Larsen, J. T., & Stastny, B. J. (2011). It’s a bittersweet symphony: Simultaneously mixed emotional
responses to music with conflicting cues. Emotion, 11(6), 1469–1473.
Margulis, E. H. (2014). On repeat: How music plays the mind. Oxford, UK: Oxford University Press.
Reber, R., Perrig, W. J., Flammer, A., & Walther, D. (1994). Levels of processing and memory for emo-
tional words. Swiss Journal of Psychology, 53(2), 78–85.
Rentfrow, P. J., Goldberg, L. R., & Levitin, D. J. (2011). The structure of musical preferences: A five-factor
model. Journal of Personality and Social Psychology, 100(6), 1139–1157.
Ritchey, M., LaBar, K. S., & Cabeza, R. (2011). Level of processing modulates the neural correlates of emo-
tional memory formation. Journal of Cognitive Neuroscience, 23(4), 757–771.
Saarikallio, S., & Erkkilä, J. (2007). The role of music in adolescents’ mood regulation. Psychology of Music,
35(1), 88–109.
Sachs, M. E., Damasio, A., & Habibi, A. (2015). The pleasures of sad music: A systematic review. Frontiers
in Human Neuroscience, 9, 404.
Sloboda, J. A., Davidson, J. W., & Howe, M. J. A. (1994). Is everyone musical? The Psychologist, 7(8), 349–354.
Skånland, M. S. (2013). Everyday music listening and affect regulation: The role of MP3 players.
International Journal of Qualitative Studies on Health and Well-Being, 8. doi: 10.3402/qhw.v8i0.20595
Sluckin, W., Colman, A. M., & Hargreaves, D. J. (1980). Liking words as a function of the experienced
frequency of their occurrence. British Journal of Psychology, 71, 163–169.
Song, Y., Dixon, S., Pearce, M. T., & Halpern, A. R. (2016). Perceived and induced emotion responses to
popular music: Categorical and dimensional models. Music Perception, 33(4), 472–492.
Szpunar, K. K., Schellenberg, E. G., & Pliner, P. (2004). Liking and memory for musical stimuli as a func-
tion of exposure. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30(2), 370–381.
ter Bogt, T. F., Mulder, J., Raaijmakers, Q. A., & Nic Gabhainn, S. (2011). Moved by music: A typology of
music listeners. Psychology of Music, 39(2), 147–163.
von Hippel, W., & Hawkins, C. (1994). Stimulus exposure time and perceptual memory. Attention,
Perception, & Psychophysics, 56(5), 525–535.
Zentner, M., Grandjean, D., & Scherer, K. R. (2008). Emotions evoked by the sound of music:
Characterization, classification, and measurement. Emotion, 8(4), 494–521.