Abstract
The relationship between music and emotions has often been investigated. It has been shown that there are differences between the emotions aroused in listeners by music and the emotional content or quality of music, as perceived by listeners, but the effect on listeners of ascribing emotions to the composer of the music has not yet been studied. Since emotional responses to music are modulated by personal attitudes to managing emotion, the relationships among emotions aroused in listeners, perceived emotional quality of the music, and emotions ascribed to the composer should vary depending on the extent to which the listener has mastered emotion-regulation skills. We therefore investigated potential differences between these three different types of emotion attribution. We also investigated the effect of emotional dysregulation on emotional responses to music by comparing the emotion attributions of adolescents with borderline personality disorder (BPD) traits with those of a healthy control group. Participants listened to 10 excerpts from pieces of music selected for their different emotional qualities and rated them for felt emotion (myself), perceived emotional quality (song), and the emotion ascribed to the composer (composer). Participants’ skin conductance level was measured to assess their physiological arousal while they were listening to each excerpt. In comparison to controls, participants with BPD traits demonstrated significantly lower levels of arousal. There were no differences between the emotion attributions of the two groups but, overall, participants gave higher ratings for perceived emotional quality and higher ratings for excerpts representing happiness and anger. For excerpts representing fear, higher ratings were given for felt emotion and perceived emotional quality than the emotion ascribed to the composer. These findings contribute to the understanding of emotional responses to music in adolescence and reveal patterns of responses to music representing emotions that are not affected by emotional dysregulation.
The complex relationship between music and human emotions has often been investigated (e.g., Eerola & Vuoskoski, 2012; Juslin & Sloboda, 2010) and it is still a flourishing research field (e.g., De Benedetto et al., 2024; Liu, 2024; Park & Kim, 2024; Piccardi et al., 2024; Trost et al., 2024; Valdés-Alemán et al., 2024; Wang et al., 2024). Numerous studies have shown that listeners appraise music based on its capacity to elicit emotions and that one of the most common purposes of music engagement—whether through passive (i.e., listening) or active involvement (i.e., playing)—is to regulate the emotions of individuals who are seeking comfort and relief from stress (Juslin & Västfjäll, 2008; Sloboda et al., 2001).
Even so, the concept of music-related emotions remains controversial and the mechanisms by which music triggers emotional responses are not well understood. Research on music-evoked responses suggests that listening to music can alter the state of the large-scale neural networks associated with emotional processes, particularly the amygdala, the insula, the cingulate cortex, and the ventral striatum (Habibi & Damasio, 2014). Four parameters of music, in particular, have been found to modulate brain activity and cardiovascular dynamics: tempo, consonance, timbre, and loudness (Schaefer, 2017).
Juslin (2013) developed a multi-level framework to explain the intricacies of the roles played by a variety of psychological mechanisms underlying emotional responses to music, including visual imagery, memory, expectation, and aesthetic judgment. This has enabled a shift away from the notion that emotional responses to music are located primarily at the neural level toward a more comprehensive explanation suggesting that emotional responses to music depend on a constellation of factors including the music, the listener, and the context in which the music is heard.
Juslin and Laukka (2003) had previously shown that emotions are conveyed by music in the same way that emotions are conveyed by nonverbal elements of speech. Similarly, people may recognize the emotions expressed in performances of music using their everyday emotional intelligence (Resnicow et al., 2004). This suggestion is supported by evidence that children and adolescents with autism spectrum disorders (ASDs), who show weaknesses in emotional intelligence, find it more difficult than their typically developing peers to discern differences between pieces of music with differing emotional qualities (Bhatara et al., 2010). Also, music has been shown to affect the regions of the brain that are directly involved in emotion processing (Moore, 2013).
To understand the complex relationship between music and emotion, it is necessary to distinguish between the listener’s emotional response to the music and the emotional content or quality of that music as perceived by the listener. It has been shown in several studies that listeners can recognize and identify the emotional quality of a piece of music and—if it has been selected for the purposes of an experiment to represent a particular emotion, such as happiness or sadness—they can identify it correctly (Gabrielsson, 2001). If the pitch, tempo, mode, and/or especially timbre of the music is modified, listeners adjust their responses to identify a different emotion (Hailstone et al., 2009). Men and women, professional musicians and non-musicians alike, have been shown to recognize and identify the same emotional quality in a piece of music (Robazza et al., 1994), but these abilities are impaired in people who generally find it difficult to identify and express emotions, such as patients with ASD and/or alexithymia (Pedregal & Heaton, 2021).
Experiencing emotional responses is an essential component of listening to music. In one study, a high proportion of participants reported experiencing emotional responses at least half of the time they listen to music (Juslin & Laukka, 2004). Instrumental music has been shown to induce emotions in healthy adults (Kreutz et al., 2008). It is crucial, however, to consider the potential influence of individual differences, particularly on the intensity of the emotions elicited by music (Vuoskoski & Eerola, 2011). Even people who find it difficult to recognize and identify the emotional quality of a piece of music, such as patients with ASD and/or alexithymia, experience emotional responses to music and may in fact rely on music to trigger such responses (Lyvers et al., 2020).
Listeners are active in constructing their own emotional experiences when they listen to music, whether they are feeling emotions themselves, or if they are perceiving the emotional quality of the music. They do so by focusing on the expressive elements of music and the sensory stimuli it provides and using empathy, resulting in a state often referred to as absorption (Levitin & Tirovolas, 2009; Sandstrom & Russo, 2013). The importance of engaging with music on multiple levels is supported by the fact that effective interventions for improving emotion recognition in adolescents with poor emotion-recognition skills require participants to interact with music at cognitive, personal, and emotional levels (Pedregal & Heaton, 2021).
Listening to music may evoke emotions that do not align with the listener’s current emotional state, underlining the difference between perception and feeling (Kallinen & Ravaja, 2006). Listeners may embrace or ignore the difference between what they are perceiving and what they are feeling, or they may resolve the difference by ascribing the emotions they perceive in the music to its composer (Peters, 2015).
To the best of our knowledge, no research has been conducted in a single study of music-related emotions including those that are felt, perceived, and ascribed to the composer by the listener. We aimed to clarify the relationship between music and emotion by exploring relationships between these types of music-related emotion attribution. We also aimed to compare the music-related emotion attributions of a clinical sample and a control group since the beneficial effects of music on psychological well-being have been studied extensively and these effects are grounded in the health-related autonomic, neuronal, and endocrinological outcomes of music-evoked emotions (Koelsch & Stegemann, 2012). A clinical population for whom music therapy has been considered a standard of care includes patients with borderline personality disorder (BPD), primarily due to the symptoms of impaired emotion regulation associated with this diagnosis (Strehlow & Lindner, 2016). One predictor of the success of music-therapy interventions with BPD patients is their overall musical competence, including the perception and recognition of emotions in music. As patients’ competencies improve, they become better at regulating their emotions and develop a healthier self-concept (Kenner et al., 2020). Given the high prevalence of young people meeting the diagnostic criteria for BPD (Courtney-Seidler et al., 2013) and the significant role that music plays in promoting emotional well-being, social interactions, and identity formation during adolescence (McFerran, 2012; McFerran et al., 2015), adolescents with BPD traits are an appropriate experimental group for studying emotional responses to music. Also, a more detailed understanding of their complex emotional responses to different types of music could contribute to the design of more effective music-therapy interventions targeting this clinical population.
Accordingly, the aims of the present study were (1) to investigate differences between music-related emotion attributions representing (a) the emotional responses felt by adolescents listening to music (myself), (b) their perceptions of the emotional quality of the music (song), and (c) the emotions they ascribed to the composer of the music (composer) and (2) to explore the effects of emotional dysregulation on music-related emotion attributions by comparing those of healthy adolescents and adolescents with BPD traits.
Method
Design
The study was based on a 3 × 2 × 5 factorial design including a within-subject independent variable (emotion attribution: myself vs. song vs. composer) and two between-subject independent variables: group (healthy controls vs. clinical sample, namely, participants with BPD traits) and type of music (depending on the emotion it represents: anger vs. fear vs. sadness vs. happiness vs. tenderness).
Dependent variables included a set of scores representing emotion attribution and a self-report measure of arousal and skin conductance levels (SCLs).
The battery of instruments included a test to assess the participants’ competence in Theory of Mind (ToM), along with a self-report questionnaire to evaluate their emotion-regulation skills. Additionally, sociodemographic data and information regarding the participants’ level of musical expertise were collected.
Participants
A total of 32 adolescents (20 f, 9 m, 3 non-binary) aged 13–17 years (M = 15.2, SD = 1.3) took part in the study. They were divided into 2 groups with 16 participants each: the experimental group with BPD traits (BPD group) and the healthy control group (HC group). The two groups were matched for age (t = 1.52; p = .14) and gender (χ2 = 4; p = .13). All participants were native Italian speakers.
The BPD group was recruited in psychiatric rehabilitation centers in northern Italy via referrals from mental health professionals at the centers who screened the clinical profiles of each potential participant to verify that they met the inclusion and exclusion criteria. The inclusion criteria for the experimental group were established in collaboration with the clinical staff at the psychiatric centers involved in the study, deriving from the International Statistical Classification of Diseases and Related Health Problems–Tenth Edition (ICD-10; World Health Organization, 1992) and the Diagnostic and Statistical Manual of Mental Disorders (5th ed., text rev.; DSM-5-TR; American Psychiatric Association, 2022) criteria for BPD, which were then adapted for the adolescent population. To select participants for the experimental group, we used a polythetic approach consistent with the DSM-5-TR criteria. To meet eligibility criteria, individuals had to have exhibited at least three characteristic symptoms during the past 12 months: (1) emotional and behavioral dysregulation (e.g., impulsivity, self-harm, suicide attempts); (2) repetitive and resistant patterns of antisocial, aggressive, or defiant behavior; (3) persistent aggressive, antisocial, or defiant behavior in the presence of depressive symptoms, anxiety, or other emotional disorders; (4) depressive symptoms, anxiety, or other emotional disorders; (5) changes in social functioning with onset in childhood and adolescence; and (6) identity disturbance involving a persistent instability in self-image or sense of self. Participants with a diagnosis of ASD, intellectual disability (TIQ < 70), and/or psychotic symptoms were excluded. No international consensus has yet been reached as to how to diagnose BPD in adolescents. Our rationale for using and adapting the criteria for BPD in adults was based on evidence of large overlaps between BPD symptomatology in adults and adolescents, including emotional dysregulation (Sharp & Fonagy, 2015). The HC group (n = 16) was recruited through advertisements shared in several secondary schools and recreational centers for young people in northern Italy. Potential participants were screened for the following exclusion criteria: (1) psychiatric disorders; (2) ASD; and (3) intellectual disability (TIQ < 70).
Ethical approval was sought and obtained from the Ethics Committee of the Catholic University of the Sacred Heart in Milan, Italy (Approval code: 66-22, December 15, 2022) and the study was conducted according to the standards of the Helsinki Declaration. The parents of all participants gave their signed informed consent prior to enrollment in the study and the participants themselves gave oral consent before data collection began.
Measures
Sociodemographic information and music expertise
Participants’ age, gender, level of music expertise, and music preferences were assessed using an ad hoc self-report questionnaire. Specifically, music expertise was measured via a multiple-choice question on the type of music courses the participant had attended (none, music school, private lessons, conservatory, other) and two Likert-type scales measuring levels of music theory and knowledge of music notation, from 1 (none) to 4 (advanced). Finally, participants were asked to report how long they listened to music every day and list their preferred music genre(s).
Emotion dysregulation
The Italian adaptation of the Difficulties in Emotion Regulation Scale (DERS) (Sighinolfi et al., 2010) was used to measure clinically relevant difficulties in emotion regulation. The self-report scale includes 36 items measuring 6 facets of emotion regulation: non-acceptance (i.e., refusal to accept emotional responses); goals (i.e., difficulty in adopting goal-oriented behaviors); impulse (i.e., difficulty in controlling impulses); awareness (i.e., lack of emotional awareness); strategies (i.e., access to emotional regulation strategies); and clarity (i.e., lack of emotional clarity).
Physiological arousal
Measuring electrodermal activity or SCL is the most commonly used method of assessing physiological arousal (Kreibig, 2010). We recorded participants’ SCL using the Biofeedback 2000 X-pert Schuhfried data acquisition system, having placed a multi-channel sensor on the index finger of their non-dominant hand.
Stimuli
We selected 10 excerpts from the Film Music Stimulus Set (FMSS) (Eerola & Vuoskoski, 2011), each one lasting ∽15 s. Each of five discrete emotions (anger, fear, sadness, happiness, and tenderness) was represented by two excerpts, chosen because they were highly representative of that emotion and had the highest values for emotion intensity reported by Eerola and Vuoskoski.
Music-related emotion-attribution task
After hearing each excerpt, participants were asked to rate the intensity of the emotion (1) “that you felt,” (2) “expressed by the song,” and (3) “of the composer while they were creating the song.” They did this five times, once for each of the five discrete emotions (anger, fear, sadness, happiness, and tenderness), using Likert-type rating scales from 1 (not at all) to 7 (extremely intense).
Self-perceived arousal
After hearing each excerpt, participants used the single-item picture-based Self-Assessment Manikin (SAM) (Bradley & Lang, 1994) to report their own perceived arousal, on a scale from low to high.
ToM
Given the reported excess of ToM (hypermentalization) in adolescents with BPD traits (Sharp et al., 2013), we aimed to exclude a potential bias in the music-related emotion-attribution task, such that we asked participants to rate the intensity of the emotion of the composer while creating the song, by asking participants to complete the children’s version of the Reading the Mind in the Eyes Test (RME-T; Baron-Cohen et al., 2001). In this task, participants were shown 28 photographs of children’s eye regions and asked to match each photograph with the most appropriate of 4 descriptors of the child’s mental state.
Procedure
Participants undertook one 45-min testing session under the supervision of a trained researcher with expertise in adolescent psychiatry and music therapy. Testing sessions were conducted individually in a quiet room. First, participants were asked to complete the sociodemographic and music-expertise questionnaires. Next, the SCL sensor was placed on the index finger of the participant’s non-dominant hand and they were asked to wear over-ear noise-canceling headphones. SC was recorded for a 3-min resting period (baseline) during which participants were asked to relax while observing a fixation cross on a blank computer screen. Thereafter, participants were presented with the 10 excerpts selected from the FMSS. Trial order was randomized. SCL was recorded as participants were listening to each excerpt. They carried out the music-related emotion-attribution task and reported self-perceived arousal after hearing each excerpt. When they had finished, the headphones and SCL sensor were removed and participants completed the DERS and children’s version of the RME-T. Finally, participants were debriefed and thanked for their participation.
Analyses
We used independent t-tests to compare the scores of the BPD and HC groups on the DERS and children’s version of the RME-T and checked for differences between their music expertise, music theory and music notation knowledge, daily music-listening time, and preferred music genres using independent t-tests for continuous variables and chi-squared tests for categorical variables.
We compared baseline SCL of the two groups using independent t-tests and measured physiological arousal in response to each excerpt by calculating difference scores between music-induced SCL and baseline SCL. We computed measures of physiological (SCL) and self-perceived arousal (SAM) by averaging the scores for the two excerpts representing the same discrete emotion (type of music), resulting in five SCL scores and five SAM scores. We conducted 2 × 5 mixed analyses of variance (ANOVAs) to test the effect of group on SAM for type of music. Homogeneity of variance could not be assumed for all variables tested, so we conducted multiple non-parametric pairwise comparisons using the Man–Whitney U-test for differences between the SCL of each group in response to type of music.
Finally, we conducted a 2 × 3 mixed analysis of covariance (ANCOVA) to test the effect of group (BPD vs. HC) and music-related emotion attribution (myself, song, composer), and interactions between them, on participants’ ratings of each type of music. Since comparisons of differences between the SCL of each group revealed significantly lower physiological arousal in the BPD group, SCL scores were included as a covariate to control for the effect of physiological arousal on ratings. We excluded multicollinearity by calculating bivariate correlations between SCL scores and participants’ ratings (all r < .20). Finally, we applied Tukey’s correction for multiple comparisons.
Results
Emotion regulation, ToM, and music expertise
BPD participants scored significantly higher on the DERS subscale (t30 = 2.22, p = .03, η2 = .14), but there were no significant differences between the groups for any other facets of emotion regulation (ps ranging from .09 to .90) nor for ToM (t30 = .30, p = .61). The HC and BPD groups were comparable for music experience (χ24 = 4.95, p = .29), music knowledge (theory: t30 = 0.61, p = .54; notation: t30 = 0; p = 1.0), daily music-listening time (t30 = 1.56, p = .13), and preferred genre (χ25 = 6.60, p = .25) (all descriptive statistics are shown in Table 1).
Descriptive statistics of BPD and HC groups: sociodemographic characteristics, music expertise and preference, emotional regulation, and ToM.
Physiological and self-perceived arousal
While there were no significant differences between the SCL scores of the two groups (t30 = 1.49, p = .15), the scores of the BPD group were significantly lower than those of HC group for music representing fear (U = 69.0; p = .04), sadness (U = 67.0; p = .04), happiness (U = 70.0; p = .049), and tenderness (U = 60.0; p = .02). There were no differences between the SCL scores of the two groups in response to music representing anger (U = 74.0; p = .07).
There were significant main effects of type of music (F4,120 = 27.38, p < .001, η2 = .24) and group (F1,30 = 9.05, p = .005, η2 = .10) on SAM, with medium to large effect sizes, and a significant interaction between type of music and group (F4,120 = 3.61, p = .008, η2 = .03), albeit with a small effect size. Post-hoc pairwise comparisons revealed that the BPD group scored significantly lower in response to music representing happiness (t30 = 2.5, p = .02), anger (t30 = 2.5, p = .02), and fear (t30 = 4.05; p < .001). There were no significant differences between the SAM scores of the two groups in response to music representing tenderness (t30 = 0.24, p = .81) or sadness.
Music-related emotion attributions
Descriptive statistics for music-related emotion attributions are shown in Table 2.
Marginal means and SE of music-related emotion attributions after listening to the corresponding type of music, for each group (BPD, HC) and after controlling for SCL.
There were significant main effects of attributions (myself, song, composer) on ratings of music representing happiness (F2,58 = 12.18, p < .001, η2 = .05) and anger (F2,58 = 8.54, p < .001, η2 = .02), with small effect sizes, and fear, with a large effect size (F2,58 = 98.15, p < .001, η2 = .44; see Figure 1). Post-hoc pairwise comparisons with Tukey correction revealed that the significant main effect of attributions on ratings of music representing happiness was such that lower ratings were given for myself than for song (t29 = 3.78, p = .002) and for myself than for composer (t29 = 4.65, p < .001). Similarly, the significant main effect of attributions on ratings of music representing anger was such that lower ratings were given for myself than for song (t29 = 3.35, p = .006) and for myself than for composer (t29 = 3.87, p = .002). Conversely, the significant main effect of attributions on ratings of music representing fear was such that higher ratings were given for song than for myself (t29 = 3.25, p = .008), for composer than for myself (t29 = 14.96, p < .001), and for myself than for composer (t29 = 10.90, p < .001).
Music-related emotion attributions for music representing happiness, anger, and fear after controlling for SCL.
There was no significant main effect of group (ps ranging from .23 to .72) nor any interaction between group and attributions (ps ranging from .39 to .83). There were no main effects of type of music on attributions for music representing tenderness (F2,58 = 0.17, p = .84) or sadness (F2,58 = 2.87, p = .06).
Discussion
The main goal of the present study was to investigate music-related emotion attributions in a sample of adolescents with BPD traits and a matched control group. Our aims were to (1) investigate differences between music-related emotion attributions representing felt emotional responses, the perceived emotional quality of the music, and the emotions ascribed to the composer and (2) the effects of emotional dysregulation on such attributions. We did this by comparing two groups: a sample of adolescents with BPD traits and HC group, matched for age and gender.
Participants, regardless of whether or not they had BPD traits, correctly identified the emotions represented by the excerpts and also ascribed the same emotions to the composer, particularly for excerpts representing happiness and anger, which are likely to induce high levels of arousal. This finding, which would seem to be unaffected by emotional dysregulation, is consistent with previous findings that people can recognize the emotions intended to be conveyed by music (Hailstone et al., 2009; Robazza et al., 1994) and extends it by showing that they ascribe the same emotions to the composer.
By contrast, when participants—again, regardless of whether or not they had BPD traits—heard excerpts representing fear, they were more likely to feel fear themselves and perceive it in the music than ascribe it to the composer. It is well documented that music is effective in eliciting fear responses in participants (Siedlecka & Denson, 2019). In neuropsychological research on music-induced emotional responses, fear has been shown to trigger activation in the amygdala (Putkinen et al., 2021), unlike the other emotions represented in the present study. This neural response is assumed to be faster and less mediated than responses involving areas such as the auditory, somatosensory, and motor cortices and explains participants’ tendency to attribute their response to themselves (as a felt emotion) or the immediate stimulus (the emotional quality of the music) rather than an external source (the composer). Participants probably also realized that it would be unlikely for a composer to write a piece of music while experiencing fear, as this would interfere with their ability to compose.
There was no effect of emotional dysregulation on participants’ emotion-attribution skills nor differences between the two groups’ ratings of the emotional quality of excerpts. Deficits in emotion recognition have been documented in BPD patients (e.g., Hanegraaf et al., 2021), but it is worth noting that these deficits appear to be associated with biases in visual attention, primarily affecting the recognition of emotional facial expressions (Kaiser et al., 2019; Kaplan et al., 2020). This justifies the use of music therapy for treating the emotional symptoms of patients with BPD (Strehlow & Lindner, 2016).
There was a difference, however, between the two groups such that adolescents with BPD traits showed less physiological arousal, as measured by SCL, than the HC group, while they were listening to the excerpts. This could mean that their emotional responses were less intense. There is evidence from research on autonomic response patterns for specific affective states that most emotional responses are accompanied by increased SCL (Kreibig, 2010). This may reflect motor preparation, in line with the theory that emotion leads to action tendency (Fredrikson et al., 1998). SCL has been used in the past to demonstrate physiological changes elicited by music listening (e.g., Krumhansl, 1997). While relatively few studies have examined biophysiological activation in patients with BPD (Rosenthal et al., 2008), they tend to exhibit lower skin conductance responses to International Affective Picture System (IAPS) images than healthy controls (Herpertz et al., 2000; Rosenthal et al., 2008), providing preliminary evidence for hypo-arousal in patients with BPD. Our findings are consistent with this evidence.
Limitations and future developments
While this study makes novel contributions to two fields of research—music-related emotion attributions and the emotional responses to music of adolescents with BPD traits—it is important to acknowledge some limitations that need to be addressed in future studies. First, we used music excerpts approximately 15 s long, but it would be worth exploring or controlling for the potential effects of stimuli with varying durations since this has been debated in the literature on music and emotion. In a study using stimuli lasting 9–39 s, MacGregor and Mullensiefen (2019) found that emotional quality was better discriminated in longer excerpts, concluding that shorter excerpts contain fewer emotional cues. This contradicted the findings of earlier studies in which participants had correctly identified intended emotional quality in excerpts as short as 1 s (Bigand, 2005; Goydke et al., 2004). Generally, longer stimuli seem to be preferable when the outcome measure is felt emotion rather than the emotional quality perceived by the listener (Eerola & Vuoskoski, 2012; Ferrer et al., 2013). Second, for a better understanding of the relationships and differences investigated in the present study, a larger sample and a second control group consisting of patients with alexithymia should be used in future research. Third, the potential effects of other variables on emotional responses to music, such as the use of medication and cognitive differences between groups, could be studied.
Notwithstanding these limitations, our study contributes to research on the relationship between music and emotions, with its focus on adolescents with BPD traits. We have shown that a sample of this population can recognize the emotions conveyed by excerpts from film music and also ascribe the same emotions to the composer of the music. This supports existing evidence that auditory stimuli can trigger emotional and cognitive processing and suggests that our findings could serve as a basis for developing innovative music-therapy interventions for adolescents with BPD traits.
Footnotes
Acknowledgements
A special thanks to the healthcare professionals who made the collaboration possible, in particular to Ferruccio Demaestri, music therapist (Paolo VI Center of Casalnoceto), and Francesca Camia, pediatric neuropsychiatrist (“La Finestra sul Porto,” Genoa).
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Data availability statement
The data that support the findings of this study are available on request from the corresponding author, A.C.