The Hypoalgesic Effects of Pleasant Music Are Not Fully Explained by Distraction

ABSTRACT

Background

One explanation for the hypoalgesic effects of pleasant music is that these stimuli impact pain by eliciting positive emotion. However, it is not clear how much of this effect is due to attentional bias, as pleasant music may more effectively capture attention than neutral or aversive control conditions. As pain reduction by distraction and emotion represent distinct mechanistic pathways, differentiating the role of attention from that of emotion and mood is necessary to understand how pleasant music impacts pain.

Methods

If effects are due to attention, performing a challenging cognitive task during music listening should attenuate analgesia through competition for limited cognitive resources. We investigated this possible sub‐additive interaction using a within‐subject factorial design. Pain‐free, healthy participants received thermal stimulations as they completed easy and difficult tasks while listening to music, scrambled music, or silence. We then assessed an interaction effect between the effects of task and music condition to test whether the auditory manipulation's effect on pain varied with task difficulty.

Results

The type of auditory stimulation and type of task had independent effects on pain intensity and unpleasantness, where music reduced pain in comparison to scrambled music and the 2‐back reduced pain compared to an easier task. Tests for a possible interaction between the two factors did not support the hypothesis that music and cognitive tasks both reduce pain by demanding attention.

Conclusions

While listening to music involves attentional processes that facilitate distraction, our results suggest that the emotional impacts of music can operate independently of these processes.

Significance Statement

What makes music hypoalgesic, is it simply the distraction it provides? This study is the first to disentangle the role of attention from other mechanisms, finding that music can reduce pain independently of where attention is directed. These results have direct implications for music therapy, supporting the centrality of other mechanisms such as emotion in therapeutic frameworks and highlight the importance of examining both attentional and emotional pathways to pain relief in future mechanistic research.

1. Introduction

Existing reviews show promising effects of pleasant, preferred music on pain outcomes (Bradt et al. 2024; Garza‐Villarreal et al. 2017; Lee 2016). As musically induced hypoalgesia (MIH) may operate through numerous pathways (Arnold et al. 2024), improving musical interventions requires a better understanding of underlying mechanisms. Two common perspectives are that pleasant music reduces pain by distraction and positive emotion, but whether these effects are independent remains unclear.

When pleasant music reduces pain compared to silence, one parsimonious explanation is that any auditory input can draw attention away from pain. By limiting propagation of nociceptive signalling into awareness, music's hypoalgesic impacts may parallel those of demanding cognitive tasks (Legrain et al. 2009; Vogel et al. 2022). However, positive emotions induced by pleasant music appear to make a distinct, additive contribution to MIH beyond distraction alone, perhaps via neurophysiological processes associated with affective regulation (Bernatzky et al. 2011):Pleasant music outperforms other distracting auditory stimuli such as unpleasant music (Valevicius et al. 2023), pain reductions persist after the auditory input has ended (Pando‐Naude et al. 2019; Roy et al. 2008, 2012) and greater pleasure ratings predict greater pain reduction.

Yet, pleasant music may simply be a more effective distractor, where associated emotional impacts are incidental to pain relief. Existing studies often do not measure which musical stimulus participants paid more attention to, and it is not possible to infer which would be more attention‐grabbing based on stimulus properties alone: attentional capture is not only covert, but depends on the participant's emotional state and motivational goals (Lang 1995; Pool et al. 2016). To test the additive contribution of emotion, one approach is to manipulate attention during music listening. Since challenging cognitive tasks reduce pain via distraction, pairing tasks with music allows us to test if MIH persists when attention is drawn away from pleasant music. While prior work has compared the relative hypoalgesic efficacy of music and tasks (Garza‐Villarreal et al. 2012; Mitchell et al. 2006; Ruscheweyh et al. 2011; Silvestrini et al. 2011), few have manipulated attention during music listening, with mixed results (Arıcan and Soyman 2025) and none have varied both emotion and attention simultaneously.

The present study examined whether music's pleasantness reduces pain via the same distractive processes as cognitive tasks by adapting a paradigm previously applied to placebo hypoalgesia (Buhle et al. 2012). Using a 3 × 2 design, we first sought to replicate hypoalgesic effects of pleasant music relative to scrambled music and silence (Valevicius et al. 2023) and of higher task difficulty on pain (Vogel et al. 2022). Assuming the 2‐back task recruited most attentional resources, we then tested whether MIA depended on attentional demand or an alternative mechanism, such as positive emotion. As attentional capacity is limited (Norman and Bobrow 1975) if effects of music and tasks operate solely via attention, they should interact, producing a sub‐additive effect on pain ratings (Sternberg 1969) (Figure 1A). Alternatively, a lack of interaction would suggest that other mechanisms such as emotion contribute an additive effect on pain (Figure 1B).

FIGURE 1.

Diagram illustrating hypothesised patterns of interaction between musical valence and cognitive distraction on pain. Cognitive distraction was manipulated via computerised tasks at two difficulty levels (low‐demand and high‐demand). Musical valence was manipulated by comparing pleasant music with scrambled versions of the same songs. Scrambling removes structural elements that contribute to musical pleasure like tempo, rhythm and phrasing (see methods), producing a more neutral acoustic condition that did not reduce pain compared to music or silence in previous studies. (A) Sub‐additive interaction. If pleasant music reduces pain solely by recruiting limited attentional resources, we should see hypoalgesic effects of pleasant music (vs. controls) during a low‐demand task where attentional resources are available, but not during a high‐demand task where there is greater competition for attention. In this case, we would expect a ceiling effect where music would not reduce pain beyond the distractive effects of the harder task. (B) Additive effects. If pleasant music reduces pain through a mechanism separate from attentional modulation (e.g., by inducing positive emotional responses) combining music with cognitive demand should produce additive hypoalgesic effects. We should observe effects of pleasant music during both low‐ and high‐demand tasks, with music providing additional pain relief above and beyond the effects of distraction by the task.

2. Methods

2.1. Participants

Seventy one healthy participants (47 women, mean age = 21.067, SD = 1.818) were recruited from poster advertisements and the McGill University Department of Psychology participant pool, receiving either monetary compensation or course credits for their time. Data collection took place during the spring of 2019. Three participants withdrew participation during the experiment, eight participants were excluded after rating the majority of trials as nonpainful and one participant was excluded due to an error in the computerised task. Analyses were performed on a sample of 59 participants. Informed written consent was obtained from participants prior to beginning the experiment, and all procedures were approved by the Research Ethics Board Office of McGill University. Criteria for exclusion from the studies included a history or current diagnosis of neurological or psychiatric disorder, diagnosis of chronic pain syndrome or neuropathy, history of alcohol or substance abuse and regular (> 2 weekly) use of pharmacological analgesics. Participants abstained from consuming alcoholic beverages and psychoactive or analgesic substances in the 12 h prior to the session.

2.2. Questionnaires

As the efficacy of distracting tasks may be less effective when pain is experienced as highly salient (Verhoeven et al. 2012), we obtained measures of constructs linked to exaggerated negative cognitive and emotional responses to pain in order to control for their effects in the main task. Participants completed the Pain Catastrophizing Scale (Sullivan et al. 1995), as well as the 40‐item State and Trait Anxiety Inventory (STAI) (Spielberger et al. 1983) developed to diagnose anxiety disorders. The STAI‐S captures situational anxiety and the STAI‐T assesses general, dispositional anxiety. Fear of pain was measured with the 30‐item FPQ‐III that asks participants to rate their fearfulness of specific painful situations such as having a tooth pulled (McNeil and Rainwater 1998).

2.3. Auditory Stimuli

All auditory stimuli were delivered via Cyber Acoustics ACM‐90 over‐ear headphones. Participants chose between seven instrumental musical pieces provided by MUSIC CARE composed for use in music therapy and supported by several clinical studies for management of acute pain, chronic pain, anxiety and depressive symptoms (e.g., Guétin et al. 2005, 2012, 2016). Participants chose between Jamaicare, Légende Celtique, Musique de Film, Nuit Cubaine, Reggae Calédonien, Sega Mizik Kèr and Cotton Blues. Each track is 30 min in total and is structured in three distinct phases that form a ‘U’ shape of intended changes in physiological arousal, where intensity and arousal of the music gradually decrease to induce relaxation before increasing again to restore alertness. The first 15 min of the participants selected piece was used in the experiment, corresponding to the beginning of the relaxation stage. The volume was set to a comfortable level at the beginning of the experiment.

2.4. Scrambled Music Controls

This manipulation was chosen based on existing work showing that scrambled music did not reduce pain compared to either favourite music or silence, and was rated as neutral to slightly unpleasant (Valevicius et al. 2023). Scrambling creates an acoustic control condition that disrupts the structure of the music while leaving spectral characteristics intact. For each one of the seven musical pieces, a corresponding ‘scrambled’ version was created by segmenting the musical pieces into segments of 0.7 s, which were reordered in random local permutations using a MATLAB script (Ellis 2010) (Figure 2A). To re‐order the segments, a sliding window is passed over the course of the stimulus and segments are rearranged locally within that sliding window.

FIGURE 2.

Experimental protocol. Illustration of experimental procedure. (A) Music choice paradigm. Participants first chose between 7 songs designed for music therapy, which were then scrambled. Participants then completed three testing blocks. (B) Diagram of a single block. In each 15 min block, participants listened to three minutes of music, scrambled music or silence before completing 20 trials of alternating easy and challenging cognitive tasks. After each block, (C) Diagram of a single trial. Each task was accompanied by a painful stimulation that was then rated for intensity and unpleasantness.

2.5. Procedure

The study took place over two sessions. After obtaining informed consent, the first session consisted of an explanation of the procedures followed by calibration of the thermal stimulations and task difficulty for use in the main task. In the second session, participants selected their music, performed the task and completed questionnaires. First, participants indicated their music selection after listening to 30‐s samples of each MUSIC CARE track. The main task consisted of three 15 min testing blocks (music, scrambled music, silence) presented in a within‐subjects design and counterbalanced between participants. MUSIC CARE's compositions, created for the purposes of music therapy, follow a U‐shaped sequence, which has three phases: induction (stimulating rhythm), relaxation (slow rhythm) and awakening rising rhythm; (Guétin and Giniès 2004) The musical excerpts in our study followed an L‐sequence with two phases (induction and relaxation).

Each of the three blocks consisted of 3 min of passive, uninterrupted listening before participants began 20 trials of alternating cognitive tasks (10 trials of 2‐back and 10 trials of left–right) presented in a pseudorandom order. Each individual trial lasted 20s, and was paired with a thermal stimulation for a total of 20 stimulations (Figure 2B). After each trial, the auditory stimulus would pause while participants rated pain perception using computerised visual analogue scales (VAS).

2.6. Thermal Stimuli

All thermal stimuli were administered using a 3 × 3 cm contact thermode (TSA II Neurosensory analyser, Medoc Ltd. Advanced Medical System, Israel). During the main task, calibrated thermal stimulations alternated between four areas on the inner surface of the left forearm. Each stimulus lasted 13 s in total, with a 2.5 s rise, an 8 s plateau and a 2.5 s fall from the baseline temperature of 32°C. Participants were instructed to stare at a fixation cross on the computer screen while receiving the stimuli and were free to move their arm away from the hot surface if it exceeded their pain tolerance.

2.7. Sensory Calibration

To control for interindividual differences in subjective pain experience, participants underwent a sensory calibration prior to the main task where a series of 28 stimulations were delivered to one of four possible areas on the left forearm. Seven temperatures ranging from 40°C to 49°C were administered, and participants were assigned to one of four testing orders where combinations of stimulation area and temperature were varied. Ratings of pain or warmth for each stimulation were fitted with a generalised linear regression model to obtain a corresponding temperature for moderate pain (60/100 on a VAS scale of pain) based on the pain tolerance of each participant for use in the main task (mean temperature = 48°C, SD = 0.83). Six participants calibrated at a threshold that exceeded 49°C, and a maximal temperature of 49°C was applied during the main task to minimise the risk of skin damage. Participants were not informed that only one temperature level would be used.

2.8. Pain Measurement

Throughout the sensory calibration and the main task, participants rated the sensation associated with each thermal stimulation on horizontal visual analogue scales (VAS) presented on a computer screen. First, participants made a categorical judgement on whether the stimulus was painful or not (right mouse button for painful, left mouse button for not painful). If the stimulus was judged as painful, participants rated pain as two separate components of the pain experience: pain intensity (reflecting the sensory component of pain) and pain unpleasantness (reflecting the affective component). This model of pain permits nuanced theoretical understanding of the underlying mechanisms of music‐induced hypoanalgesia and has been validated over several experiments (Rainville et al. 1999; Roy et al. 2012). Pain intensity was rated as 0 = ‘not intense at all’ to 100 = ‘extremely intense’ and pain unpleasantness as 0 = ‘not unpleasant at all’ to 100 = ‘extremely unpleasant.’ If the stimulus was judged as not painful, participants rated its warmth on a sliding VAS (0 = ‘no warmth at all’ to 100 = ‘very hot, without pain’). Warmth ratings were then merged with pain intensity ratings to create a 0–200 scale used to model overall pain intensity. In the exploratory analysis to investigate the impact of rating type on pain perception, the 0–100 raw intensity rating for warm and painful trials was used.

2.9. Cognitive Tasks and Calibration

Task presentation was controlled using E‐Prime software (E‐Prime 3.0, Psychology Software Tools Inc., 2017). The 2‐back working memory task was implemented as the high cognitive load or ‘difficult’ task. During the 2‐back, a series of letters (C, F, J, N, Q, S, V, X) is presented in sequence. The presentation of the letters consisted of a fixation cross (250 ms), followed by a letter (500 ms), followed by a blank interval of variable duration derived from the participants' individual task calibration. For each letter, participants indicate whether or not the current letter is the same as the letter displayed two letters back (See Figure 2A). Participants were asked to press the left mouse button to answer ‘same’ and the right mouse button for ‘different’ before the next fixation cross appeared. Each trial lasted 20 s and presented between 6 and 26 letters depending on the presentation speed. To ensure similar task difficulty across participants, the blank inter‐stimulus interval (ISI) was adjusted based on a calibration session. Participants first completed three practice rounds with feedback, followed by 18 calibration trials in which the blank interval was fine‐tuned using a staircase procedure. Starting from an initial duration of 2583 ms, the blank interval was increased or decreased to achieve a performance accuracy (A) between 0.75 and 0.85. The average calibrated ISI used in the experiment was 670 ms. As the 2‐back requires continuous updating and maintenance of working memory content, it is not possible to perform the task accurately while allocating attention intermittently. However, because task accuracy indexes relative difficulty rather than the proportion of total attentional capacity consumed, an accuracy measure alone cannot confirm that the 2‐back task occupied all resources relevant to pain modulation.

The left–right task was used as a low cognitive load control. In the left–right task, a series of left or right arrows is sequentially presented and participants indicate whether the arrow points to the left or to the right by pressing the corresponding mouse button. Each trial lasted 20 s and presented between 6 and 26 arrows. To standardise the length of each cognitive task trial, the timing of the presentation of the arrows in the left–right task was set to correspond to the timing of the presentation of the letters in the 2‐back task.

2.10. Data Analysis

2.10.1. Cognitive Task Performance

Performance in the 2‐back task and the left–right task was computed as the nonparametric sensitivity measure ‘A’, allowing hit and false alarm rates to be at or near 0 or 1 (Zhang and Mueller 2005). A score of 1.00 corresponds to perfect performance, 0.5 corresponds to chance performance and a score of 0 indicates that all responses were incorrect.

2.10.2. Analysis of Variance

Data from 8 participants were excluded after rating the majority of trials as nonpainful. All analyses on the remaining sample of 59 participants were performed using R (R Core Team 2025). Pain perception was analysed using linear mixed‐effects models estimated with the lmerTest package (Kuznetsova et al. 2017). To control for pain sensitisation, where pain ratings linearly increase in response to repeated stimulations, trial number was added as a covariate. To control for pain habituation, where initial stimulations in an experiment are typically rated as much more painful before plateauing at a lower value, the logarithm of trial number was added as a covariate. The model additionally included arm stimulation site and individual difference measures (pain catastrophising, trait anxiety, fear of pain) as covariates to control for their potential influence on pain ratings. Participant number was included as a random intercept to account for within‐subject dependence across repeated trials. All numeric variables were standardised prior to model estimation to provide beta estimates expressed in units of standard deviation.

First, two mixed effects models were estimated to predict pain intensity and unpleasantness, estimating the fixed effects of task type (Control, 2‐back), acoustic condition (Music, Scrambled, Silence) and their interaction. These models were then inputted into omnibus tests of fixed effects using Type III analysis of variance (ANOVA) with Satterthwaite‐adjusted degrees of freedom as implemented in the car package (Fox and Weisberg 2019). Type III tests assess each fixed effect after accounting for all other terms in the model, including interactions, and are invariant to the order of predictors. Marginal and conditional R ² of each mixed effects model were computed as a measure of fit using the performance package (Lüdecke et al. 2021) and interpreted in line with recommendations made by Cohen (2013) (< 0.02—Very weak, 0.02–0.13—Weak, 0.13–0.26—Moderate ≥ 0.26—Large). The pain intensity model ICC was 0.46, indicating a large degree of clustering. The model conditional R ² was 0.03, and the marginal R ² was 0.47%. For pain unpleasantness, the ICC of the linear mixed effects model was 55%, again indicating a large degree of clustering. The model conditional R ² was 0.04, and the marginal R ² was 0.57.

Task performance was assessed using the same procedure. A linear mixed effects model was estimated to predict the sensitivity metric A from music condition, trial type and their interaction. The ICC was 0.15, indicating a moderate degree of clustering. Fixed effects explained 20% of the variance (marginal R ² = 0.20) and the model explained 32% of variance when including random effects (conditional R ² = 0.32). This model was then inputted into a Type III repeated‐measures ANOVA to examine omnibus effects. Effect sizes of main effects and interactions were computed as partial eta squared using the effectsize package and interpreted using the same guidelines.

To further probe a possible null interaction effect between the impacts of task difficulty and musical valence on pain, we conducted a Bayesian model comparison of two nested models: the same lmer mixed model described above that included an interaction effect, and a null model with no interaction. Bayes factors (BF₀₁ for null/full) in support of the null were estimated using the Bayesian Information Criterion approximation for frequentist models implemented in the bayestestR package (Makowski et al. 2019; Wagenmakers 2007). These factors quantify the evidence for one model over another, with values > 1 indicating support for the null model with no interaction and values < 1 supporting the null model. Interpretation of estimates was guided by Jeffreys 1961 guidelines as implemented by the effectsize package (Jeffreys 1961).

To conduct post hoc comparisons, estimated marginal means were retrieved from the fitted linear mixed‐effects models using the emmeans package (Lenth and Piaskowski 2025), and pairwise contrasts using the residual (within‐subject) variance from the mixed‐effects models were used to calculate differences in predicted mean pain ratings between every pair of music conditions and between trial types, holding all other model factors constant. To estimate each set of contrasts, means for each factor were averaged across the levels of the other factor. Estimates were corrected using the False Discovery Rate (FDR) correction for multiple comparisons. Effect size for the changes in EMMs was then computed as Cohen's d using the emmeans package, where standardised mean differences were obtained by dividing the estimated marginal mean differences by the residual standard deviation from the mixed‐effects model.

Finally, we performed an exploratory analysis to further investigate whether observed impacts of music and tasks in our study differentially impacted pain intensity and unpleasantness, in order to make comparisons with previous research. In some cases, manipulation of attention with tasks has primarily impacted pain intensity only, with nonsignificant effects on pain unpleasantness (Villemure et al. 2003; Villemure and Bushnell 2009), whereas pain unpleasantness was independently modulated by emotion and did not alter pain intensity. While these two dimensions of pain are driven by activity in dissociable structures (Rainville et al. 1997), other studies have shown a different relationship where behavioural ratings are not doubly dissociated. For example, using hypnotic suggestion, altering intensity has a corresponding downstream impact on pain unpleasantness (Hofbauer et al. 2001; Rainville et al. 1999). Attentional manipulations that strongly impact intensity should therefore also reduce unpleasantness, but this has not been clearly demonstrated when altering pain intensity with cognitive tasks. Furthermore, while previous investigations primarily examined effects on intensity and unpleasantness in separate models, comparing results across models is not a formal statistical test of how effects differed between rating types. We aimed to extend these findings by modelling rating type as an interaction term with the impacts of music and tasks to investigate whether each manipulation would impact one dimension and not the other, or whether a manipulation that altered pain intensity would also reduce pain unpleasantness. A significant interaction would indicate that a manipulation's effect on pain depended on which dimension was being rated.

A hierarchical mixed effects model assessed differences in visual analogue scale (0–100) ratings of both warm and painful trials predicted by the dimension of pain that was assessed. The model then included an interaction term between the type of rating and each of the two experimental factors. The ICC of the model suggested a large degree of clustering (ICC = 0.48); the model conditional R ² was 0.48, and the marginal R ² was 0.02. We then created a nested model with a possible three‐way interaction between the factors of music, task type and pain rating and performed a likelihood ratio test to assess whether including the three‐way interaction improved model fit. Two exploratory Type III analyses of variance (ANOVA) were performed on these models using the same procedure outlined above, and the simple effects of each manipulation were estimated using emmeans to quantify posthoc effects of music and pain within each rating type.

3. Results

3.1. Task Performance

Participants showed far greater accuracy in the easier, left–right task (A = 0.98) compared to the more challenging N‐back task (A = 0.87), F(1, 3470.9) = 1049.23, p < 0.001, ${\eta }_p^2$ = 0.23, 95% CI [0.21, 1.00]. Conversely, the type of auditory condition that participants heard during the cognitive tasks did not impact performance (F(2, 3470.9) = 1.14, p = 0.32, ${\eta }_p^2$ = 0.0007, 95% CI [0.00, 1.00]). Furthermore, the effects of auditory condition and tasks did not depend on each other, showing no interaction effect. F(2, 3470.9) = 0.80, p = 0.45. Raw data are displayed in Figure 3A.

FIGURE 3.

Raw means of pain ratings per condition, interaction result, and accuracy between conditions. (A) Raw means for each combination of music condition and task type demonstrating the lack of interaction between the effects of music and tasks. Post hoc tests confirmed a significant hypoanalgesic effect of music that was not dependent on task type, where music significantly reduced pain intensity compared to scrambled music, and pain unpleasantness compared to scrambled music and silence. Error bars indicate 95% confidence intervals. Colours and markers differentiate task type (purple circles = Control; blue squares = 2‐back). (B) Performance was impacted by task type but not music condition, indicating that the auditory condition did not impact the attention participants used to focus on task performance. Bar plots display average accuracies in each combination of task and music condition, showing highest accuracy during the left–right (LR) task.

3.2. Main Effects

ANOVA analyses showed that ratings of pain intensity varied significantly across auditory conditions, F(2, 3407) = 20.14, p < 0.001, but the effect size was very weak (${\eta }_p^2$ = 0.01, 95% CI [0.01, 1.00]). Pain intensity also significantly differed depending on the types of task participants concurrently performed, F(1, 3407) = 50.46, p < 0.001, ${\eta }_p^2$ = 0.01, 95% CI [0.01, 1.00]. We saw the same pattern of results for pain unpleasantness, finding that both the auditory condition F(2, 3407) = 39.65, p < 0.001, ${\eta }_p^2$ = 0.02 [0.01, 1.00] and the difficulty of the task F(1, 3407) = 54.13, p < 0.001, ${\eta }_p^2$ = 0.02 95% CI [0.01, 1.00] were associated with small, significant differences in pain ratings.

3.3. Post Hoc Comparisons

To investigate which conditions reduced pain intensity, we performed contrasts between estimated marginal means. As auditory condition and task type were manipulated simultaneously, effects within each factor were estimated while controlling for the other. A first set of analyses examined the effects of auditory condition on while controlling for the type of task, and found that listening to intact music reduced pain intensity compared to scrambled music (ΔEMM = −0.19, SE = 0.03, 95% CI [−0.26, −0.11], t(3407) = −6.13, p ≤ 0.01, d = −0.25, 95% CI [−0.33, −0.17]) and listening to scrambled music increased pain intensity relative to silence (ΔEMM = 0.13, SE = 0.03, 95% CI [0.06, 0.211], t(3407) = 4.49, p > 0.001, d = 0.19, 95% CI [0.10, 0.27]). While pain intensity ratings trended lower during music compared to silence, the difference between the conditions was not significant (ΔEMM = −0.05, SE = 0.03, 95% CI [−0.12, 0.02], t(3407) = −1.64, p = 0.101, d = −0.07, 95% CI [−0.15, 0.02]). See Table S1 for a full table of results. A second set of analyses found that when controlling for the auditory condition, the more challenging N‐back task reduced pain intensity compared to the easier, left–right task (ΔEMM = −0.19, SE = 0.02, 95% CI [−0.13, −0.24], t(3703) = −7.10, p < 0.001, d = −0.25, 95% CI [0.18, 0.33]). A full table of results may be found in Table S2.

We then assessed the post hoc effects of auditory condition and tasks on pain unpleasantness. Controlling for task type, music again reduced pain unpleasantness compared to scrambled music (ΔEMM = −0.24, SE = 0.03, 95% CI [−0.30, −0.17], t(3407) = −8.74, p < 0.001, d = −0.36, 95% CI [−0.45, −0.28]) and there was a significant difference between music and silence (ΔEMM = −0.08, SE = 0.03, 95% CI [−0.15, −0.01], t(3407) = −2.92, p < 0.01, d = −0.12, 95% CI [−0.20, −0.04]). Pain unpleasantness increased during scrambled music compared to silence (ΔEMM = 0.16, SE = 0.02, 95% CI [0.09, 0.22], t(3407) = 5.81, p < 0.001, d = 0.24, 95% CI [0.16, 0.33]). A full table of results may be found in Table S3. Similarly to effects on pain intensity, N‐back trials reduced pain unpleasantness compared to left–right trials (ΔEMM = −0.18, SE = 0.01, 95% CI [−0.12, −0.06], t(3407) = −6.54, p < 0.001, d = −0.23, 95% CI [0.16, 0.31]). A full table of results may be found in Table S4. Figure 3B depicts raw (unadjusted) group means. These raw means differ from model‐adjusted estimates of marginal means because the models account for the effects of several covariates and the underlying clustered structure of the data. For transparency we opted to depict the data as observed, independently of modelling choices.

3.4. Interaction Between Conditions

Given that both pleasant music and cognitive tasks reduced pain, we then examined their interaction to test whether these effects were independent. The ANOVA tests showed no significant interaction between auditory condition and task for either pain intensity (F(2, 3407) = 1.72, p = 0.17, ${\eta }_p^2$ = 0.001) or pain unpleasantness (F(2, 3407) = 2.14, p = 0.09, ${\eta }_p^2$ = 0.001). Bayesian model comparisons further evaluated the observed lack of interaction. For both pain intensity (BF₀₁ = 617.69) and pain unpleasantness (BF₀₁ = 308.79), there was strong evidence favouring models without the interaction terms, indicating that adding the interaction did not improve model fit.

3.5. Exploratory Interaction With Rating Type

Results of a Type III ANOVA with Satterthwaite‐adjusted degrees of freedom showed no significant main effect of rating type F(1, 6878) = 0.06, p = 0.80, ${\eta }_p^2$ = 0.000003 [0.00, 1.00]. However, there were significant interaction effects between the type of rating and the effects of music type (F(2, 6878) = 9.79, p ≤ 0.001, ${\eta }_p^2$ = 0.003 [0.01, 1.00]) and task type (F(1, 6878) = 5.11, p = 0.024, ${\eta }_p^2$ = 0.0007 [0.01, 1.00]). Means of pain ratings between conditions for each dimension of pain are plotted in Figure 4. For the second model estimating a three way interaction between rating type, tasks and auditory condition, this interaction was not significant in this model (F(2, 6876) = 0.61, p = 0.54, ${\eta }_p^2$ = 0.0001 [0.01, 1.00]) and including the interaction did not significantly improve fit, χ ²(2) = 1.23, p = 0.54, indicating no evidence for a three‐way interaction.

FIGURE 4.

Interactions with rating type. Mean ratings of pain intensity and unpleasantness across the effects of music type and task type. ANOVA analyses showed differential impacts of music and tasks on each dimension, where music conditions only significantly impacted pain unpleasantness, and tasks impacted both pain intensity and unpleasantness. Error bars represent 95% confidence intervals.

Post hoc analyses of simple effects revealed that auditory conditions and tasks differentially impacted pain unpleasantness and intensity. As in previous models, music reduced pain unpleasantness compared to scrambled music (ΔEMM = −0.24, SE = 0.03, t(6878) = −7.95, p < 0.0001, d = −0.33, 95% CI [−0.41, −0.25]) and silence (ΔEMM = −0.08, SE = 0.03, t(6878) = −2.62, p = 0.024, d = −0.11, 95% CI [−0.19, −0.03]) and scrambled music increased pain relative to silence(ΔEMM = 0.16, SE = 0.03, t(6878) = 5.32, p < 0.0001, d = 0.22, 95% CI [0.14, 0.30]). However, auditory conditions did not have any significant effects on pain intensity (all comparisons p > 0.09). In contrast, increasing task difficulty had a significant hypoalgesic effect on both intensity (ΔEMM = 0.09, SE = 0.03, t(6878) = 3.71, p = 0.0002, d = −0.13, 95% CI [0.06, 0.20]) and unpleasantness (ΔEMM = 0.17, SE = 0.03, t(6878) = 6.82, p < 0.0001, d = −0.24, 95% CI [0.17, 0.31]).

4. Discussion

To disentangle attentional effects of music from other possible mechanisms, we implemented a paradigm where music and tasks would theoretically compete with each other for attention. We first replicated independent hypoalgesic effects of each manipulation, comparing music, scrambled music and silence across two levels of task difficulty. The more difficult task reduced pain relative to the easy task, and pleasant music significantly reduced pain unpleasantness relative to both silence and scrambled music and pain intensity compared to scrambled music. Task accuracy did not differ significantly by auditory condition at either difficulty level, indicating that differences in pain ratings across conditions are unlikely to be explained by differences in task performance. We then found that pleasant music and task difficulty had additive hypoalgesic effects on pain unpleasantness and pain intensity with no interaction effects in either frequentist and bayesian analyses. This would suggest that while music can provide some distraction through cognitive processes, it may also provide additional hypoalgesia via a more automatic, emotional route.

We also aimed to gain better insight into the stage of pain processing impacted by each manipulation by comparing our pattern of results to previous findings. Across modalities, positive emotions elicited by music, odours and images have been shown to predominantly alter the secondary, affective‐motivational component of pain thought to drive ‘unpleasantness’ that is distinct from sensory‐discriminative qualities such as location and intensity driven by primary nociception (Loggia et al. 2008; Rainville et al. 1999; Valevicius et al. 2023; Villemure and Bushnell 2009). Notably, while music's effects were more pronounced for pain unpleasantness than for pain intensity, performing a hard cognitive task reduced both measures. This conflicts with some comparisons of emotional and cognitive manipulations where tasks and emotions each modulated only one dimension of pain (Villemure et al. 2003), which may have been impacted by a lack of statistical power. These results align instead with serial models of pain processing where intensity has downstream impacts on unpleasantness, but pain unpleasantness, as a ‘later’ part of the pathway, can be modulated independently. While music likely reduces pain through emotional and attentional mechanisms, emotional impacts represent an independent contribution to pain relief by reducing pain unpleasantness that is not redundant with that of attention.

Surprisingly, we also observed a surprising but consistent effect of scrambled music increasing both pain intensity and unpleasantness. This condition was intended to be a neutral control, sharing pleasant music's spectral content while being devoid of musicality (i.e., rhythm, melody) as an alternative to the heavily dissonant music employed by previous studies (Roy et al. 2008, 2012). These impacts on pain may have been due to the hyperalgesic effects of negative mood (Meagher et al. 2001; Villemure and Bushnell 2002; Wiech and Tracey 2009), as removing these qualities in the relaxing MusicCare pieces may have induced negative emotionality. Additionally, as the success of a pain management strategy relies to some extent on the subject's self‐efficacy expectations (Perlini and Viita 1996), both scrambled music and music could have been perceived as irritating if trying to complete a concurrent demanding task. These findings ultimately lend further support for the effects of emotion on pain.

From a mechanistic perspective, this emotional pathway likely operates through diverse neurophysiological responses elicited by pleasurable music (Lee 2016; Lunde et al. 2019), including arousal (de Witte et al. 2020), endocrine activity (Bernatzky et al. 2011), reward (Zatorre 2015) and reductions in psychiatric symptoms (Garza‐Villarreal et al. 2017; Guétin et al. 2012). For example, soothing singing used by parents to co‐regulate emotions of young infants is proposed to initiate a hormonal cascade involving oxytocin and endogenous opioid release (Trehub and Gudmundsdottir 2015; Ullsten et al. 2018). This suggests that pleasant music may parallel placebo hypoalgesia, which operates via endogenous opioids without requiring executive attention (Buhle et al. 2012; Wager et al. 2007), such that even passively heard background music could reduce pain provided that it elicits a sufficient change in emotional response. As music recruits processes of valuation and reward (Salimpoor et al. 2013), it may also impact the endogenous opioids system through dopaminergic activity (Leknes and Tracey 2008), as interactions between these systems are thought to underlie our ability to prioritise competing goals in spite of concurrent pain (Fields 2006).

4.1. Limitations

First, it is important to note that a lack of interaction does not definitively prove the null, and that our sub‐additive hypothesis assumes that the 2‐back task occupied most or all of the attentional resources relevant to pain modulation. This assumption was supported by relatively lower task accuracy during the 2‐back task, but relative difficulty cannot indicate the proportion of total attentional capacity consumed. It is therefore possible that available cognitive resources were not sufficiently challenged to bring the two manipulations into competition. Second, music was always accompanied by a task, making it challenging to interpret the lack of difference in pain intensity between music and silence. The low‐demand left–right task may have been distracting enough to produce a ceiling effect in the silence condition, though the trend in marginal mean differences was that ratings in the silence condition were higher than that of music. Additionally, with a trial length of 20 s, participants constantly shifted their attention back and forth between the task, the subsequent pain ratings and the pleasant music. These fast alternations may have prevented full immersion in the music and detracted from its attentional or emotional effects. Finally, music used in our study was self‐selected and constrained to seven music therapy tracks following a U‐shaped arousal profile that is initially stimulating before reducing volume, tempo and number of instruments over the course of 10 min before returning to the initial stimulating profile (Guétin et al. 2016). As participants experienced the relaxation phase during the second half of the experiment, the music became increasingly unobtrusive by nature and may not have been salient enough to compete for attentional resources. Self‐selected music from participants' own libraries may produce stronger effects on both emotion and attention by acting as a more attention‐grabbing, salient reward. Future studies could attempt to disentangle possible effects of music as a calming signal of safety, which may operate through endogenous opioids, and music as a highly rewarding stimulus, which may involve more dopaminergic processes (Becker et al. 2013), and be augmented by focused attention (Howlin et al. 2023).

4.2. Conclusion and Future Directions

We observed that the emotional effects of self‐selected pleasant music on pain did not depend on attentional resources recruited by a concurrent demanding task. This supports that positive emotions elicited by music are additive to, not redundant with, the capacity of music to recruit attention away from pain. As music can offer potent emotional impacts that make it a useful tool for studying affective neuroscience (Koelsch 2013), frameworks for using musically induced emotions to reduce pain outlined by music therapists (Hiller 2015) could help us to understand how emotion regulation alleviates suffering. Future studies should further investigate the efficacy and mechanisms of beneficial combinations of attentional and emotional mechanisms that may provide superior hypoalgesia.

Author Contributions

This study was designed by A.D., T.A.V. and M.R. The experiments were performed by A.D. and T.A.V. The data were analysed by É.D., and the results were critically examined by all authors. É.D. had a primary role in preparing the manuscript. All authors have approved the final version of the manuscript and agree to be accountable for all aspects of the work.

Funding

This work was supported by a CGS Doctoral Scholarship from the Natural Sciences and Engineering Research Council of Canada (NSERC) awarded to É.D. who performed analyses and wrote the manuscript, and by an NSERC Discovery Grant (244945; ‘Keeping Pain Out of Mind’) awarded to M.R., which funded data collection and publication of the study. Natural Sciences and Research Engineering Council of Canada (NSERC) Canada Graduate Scholarships 110_2024_2025_Q2_834.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Contrast effects of auditory conditions on pain intensity.

Table S2: Contrast effect of task type on pain intensity.

Table S3: Contrast effect of auditory conditions on pain unpleasantness.

Table S4: Contrast effect of task type on pain unpleasantness.