Music listening and stress recovery in healthy individuals: A systematic review with meta-analysis of experimental studies. Krisna Adiasto et al. PLoS One, June 17, 2022. https://doi.org/10.1371/journal.pone.0270031
Abstract: Effective stress recovery is crucial to prevent the long-term consequences of stress exposure. Studies have suggested that listening to music may be beneficial for stress reduction. Thus, music listening stands to be a promising method to promote effective recovery from exposure to daily stressors. Despite this, empirical support for this opinion has been largely equivocal. As such, to clarify the current literature, we conducted a systematic review with meta-analysis of randomized, controlled experimental studies investigating the effects of music listening on stress recovery in healthy individuals. In fourteen experimental studies, participants (N = 706) were first exposed to an acute laboratory stressor, following which they were either exposed to music or a control condition. A random-effects meta-regression with robust variance estimation demonstrated a non-significant cumulative effect of music listening on stress recovery g = 0.15, 95% CI [-0.21, 0.52], t(13) = 0.92, p = 0.374. In healthy individuals, the effects of music listening on stress recovery seemed to vary depending on musical genre, who selects the music, musical tempo, and type of stress recovery outcome. However, considering the significant heterogeneity between the modest number of included studies, no definite conclusions may currently be drawn about the effects of music listening on the short-term stress recovery process of healthy individuals. Suggestions for future research are discussed.
Discussion
Music listening has the potential to fulfill the promise of effective stress recovery in healthy individuals. However, cumulative evidence from 17 experimental studies suggests that support for the beneficial effect of music listening on stress recovery is currently lacking: for healthy individuals, the effect of music listening on stress recovery may be equivalent to that of other auditory stimuli, or even merely sitting in silence. Furthermore, the effect of music listening on stress recovery is heterogeneous, and moderator analyses suggest the effect may differ in magnitude according to musical genre, whether music is self-selected, musical tempo, and type of stress recovery outcome. Despite this, the limited number of available studies makes it difficult to draw further conclusions from these analyses.
Overall effects of music listening on stress recovery
The results of our review contrast those of previous meta-analyses, which underscore the relevance of music-based interventions for stress-reduction [10, 11]. While previous reviews suggest that music-based interventions may be moderately beneficial for stress-related outcomes, particularly in medical and therapeutic settings, our results suggest that the magnitude of this effect outside of these settings, particularly for healthy individuals under acute, experimentally induced stress, may be more modest. We presume that one of the principal reasons for this difference was our decision to exclude studies conducted in medical and therapeutic settings. In previous reviews, randomized controlled trials of the effects of music-based interventions within medical and therapeutic settings constituted a large portion of included studies: 67 of 79 (85%) studies in de Witte et al. [10], and 15 of 22 (68%) studies in Pelletier [11], making it more likely that overall effect sizes were derived from studies conducted within these settings. Tentatively, the effects of music listening may be more prominent for the stress recovery of individuals in medical or therapeutic contexts, compared to that of individuals under acute stress in an experimental context. Whereas the time course of stress responses and stress recovery in experimental settings can be considered relatively brief [24, 26, 40, 83], the time course of stress responses and stress recovery within medical and therapeutic settings may be significantly more protracted [12, 13]. Thus, within medical and therapeutic settings, music may be exerting its influence on neuroendocrine, physiological, and psychological processes that have been subjected to longer periods of strain [27, 99].
Furthermore, the difference in overall estimated effect sizes may be attributed to differences in the breadth of music activities encompassed by our review and that of de Witte et al. [10]: whereas we included studies in which participants merely listened to music following a stressor, de Witte et al. [10] also included music therapy, along with other unspecified music activities. We speculate that the effect of music on stress recovery may differ depending on whether music is merely listened to, performed, or used within a music therapy setting. However, studies comparing the stress recovery effects of these various music activities are rare [15, 58]. Thus, future investigations into the differential effect of these music activities may therefore provide a more comprehensive picture of the effects of music on stress recovery.
Potential moderating effects
Our review highlights the considerable methodological variety between studies investigating the effects of music listening on stress recovery. This is particularly concerning given the modest number of experimental studies on music listening and stress recovery in current literature. Although we investigated the impact of these methodological differences through moderator analyses, many of the estimated effects at each level of each moderator were either non-significant or originated from single studies. Taken together, meaningful interpretations for these moderating effects are difficult to make. Therefore, for each significant moderator, we instead provide several recommendations for future research, which we believe may help delineate the effects of these potential moderators.
Musical genre.
Although comparisons between musical genres seem relatively straightforward, investigating the differential effects of musical genres may be particularly challenging: the conceptualization of musical genres, along with the songs they encompass, tends to be somewhat arbitrary [69, 75, 100, 101]. Indeed, studies display considerable variation in musical stimuli, even within the same genre (Table 4). A notable example of this is the study by Sandstrom and Russo [53], which utilized four ‘classical’ songs, each at different extremes of valence and arousal. It should also be considered that new music is continuously being released which may not completely fit with the definition of any existing genre [9].
As such, an alternative approach to the investigation of musical genre involves describing these genres according to their musical features, such as tempo, timbre, and loudness, and subsequently investigating the effects of these individual musical features on stress recovery [9, 101]. For example, classical music may be described as rhythmically complex, with mellow timbre and fluctuating loudness. Comparatively, though equally rhythmically complex, heavy metal possesses sharper timbre and more pronounced loudness. Investigating the differential effects of these musical features on stress recovery may provide relevant insight into the differential effects of listening to various musical genres on stress recovery.
Self- versus experimenter selection.
In investigating the effects of self- versus experimenter selected music on stress recovery in healthy individuals, studies typically request participants to select music they consider ‘relaxing’ prior to an experiment [3, 17, 18]. Although this approach is viable, it precludes the potential role of perceived control in the relationship between music listening and stress recovery, since allowing participants to self-select their own music may already be helpful for stress recovery due to a restoration of perceived control [15]. Our results were not able to provide a significant contribution to this discussion, as hardly any experimental studies in our review have attempted to account for the potential effects of perceived control. As such, when contrasting the effects of self- and experimenter selected music on stress recovery, future studies may benefit from the inclusion of perceived control as an additional variable in their theoretical models.
It should also be noted that allowing participants to self-select their own music will result in a considerable variety of musical stimuli. Given that each of these musical stimuli may possess a different combination of musical features, the use of self-selected music may generate confounding effects that should preferably be accounted for. Arguably, self-selected music may produce consistent effects on stress-recovery regardless of underlying musical features, given that individuals tend to select music in service of personal self-regulatory goals [64, 75, 76]. However, given that variations in specific musical features, such as tempo, pitch, and loudness have been related to various physiological (e.g., heart rate) [73] and psychological stress recovery outcomes (e.g., positive and negative affect) [100–102], future studies may benefit from ensuring that musical features are consistent between self- and experimenter selected musical stimuli. This may be done, for instance, by comparing expert ratings of musical features [18]. Alternatively, there may be value in allowing participants to self-select music from a list provided by experimenters [21], as this would allow experimenters to standardize musical features a-priori, which may further help disentangle the effects of music listening from that of perceived control.
The comparison of musical features between self-selected and experimenter selected music may also offer a more nuanced perspective on the role of preference and familiarity. Specifically, preferences and familiarity towards certain songs could be described in terms of specific (combinations of) musical features. For example, an individual may prefer music with slow tempo, mellow timbre, and moderate loudness. This approach is often leveraged by music recommender systems, such as those implemented by music streaming platforms (e.g., Spotify, Deezer, Apple Music, etc.), with the goal of recommending songs that listeners are likely to engage with. Future studies could investigate the extent to which preference and familiarity might differ between self-selected and experimenter selected music with similar combinations of musical features, to further clarify the role of selection in the relationship between music listening and stress recovery.
Musical tempo.
The systematic review portion of our results demonstrates that no studies have directly compared the effect of different musical tempi on stress recovery in healthy individuals. As such, the most straightforward approach to delineate the effects of musical tempo on stress recovery would be to adopt procedures in which participants listen to the same musical stimulus post-stressor, which is then varied in tempo across experimental conditions. Furthermore, even when the goal of a particular study on music listening and stress recovery is not to clarify the effects of musical tempo, we suggest that tempo values for each musical stimulus should be noted down and reported, as this would facilitate the comparison of the differential effects of musical tempo on stress recovery in future meta-synthesis of the literature.
Alternatively, the notion that music with slow tempo is more beneficial for stress recovery compared to music with fast tempo is supported by the assumption that physiological parameters will entrain to musical rhythms [63, 68]. As such, a more accurate approach to investigate the effects of musical tempo on stress recovery would be to leverage the dynamic, temporal nature of both music and physiological parameters through use of non-linear analyses of continuous data [52, 103]. For example, cross-recurrence quantification analysis (CRQA) [104, 105] may enable future studies to quantify the magnitude and duration of rhythmic entrainment for each participant. These indexes of magnitude and duration could then be compared between different musical tempi. Studies have utilized CRQA to investigate cardiac entrainment between participants of collective rituals [106] and the entrainment of an audience’s heart rate to a live musical performance [107]. This analytical approach may therefore yield a more nuanced understanding of the effect of musical tempo on the recovery of autonomic parameters.
Stress recovery outcomes.
During short-term stress responses, catecholamine- and cortisol-mediated stress responses follow temporally specific patterns: catecholamines rapidly exert their influence on ANS activity, and these changes tend to normalize within 30–60 minutes [26]. Meanwhile, decreases in cortisol that may be attributed to stress recovery will only become noticeable after recovery-related changes in autonomic activity have begun to occur [24]. As such, to further clarify the effect of music listening on various stress recovery outcomes, we recommend future studies to be more sensitive towards the innate, intricate, and temporally specific changes of each stress recovery outcome.
Furthermore, multiple studies included in our review have opted to analyze continuous data by means of multivariate analyses of variance, after averaging participants’ observed stress recovery outcomes at multiple time points (e.g., pre-stress, post-stress, post-recovery). Although this approach is practical, doing so may over-simplify the complex changes that may occur during the stress response and subsequent stress recovery, such as the temporal dynamics of different physiological responses [52] and emotion regulation strategies [108]. As such, we again suggest future studies to utilize non-linear analyses of data when appropriate, particularly when investigating the effects of music listening on the recovery of autonomic activity post-stressor. The idea of using non-linear analyses, such as time-series analysis, to investigate the stress recovery process is not new [5]. However, few studies on music listening and stress recovery have utilized this analytical approach.
Additional recommendations
Two studies with unreported stress induction procedures were still included in the review [17, 84], as reported means for certain recovery outcomes still suggested an increase from baseline that participants could recover from. For example, with the information reported in Gan et al. [84], assuming a correlation of 0.5 between baseline and post-stressor measures of state anxiety, we estimated that their stress induction procedure elicited a significant increase in state anxiety in their sedative music (t(34) = 5.87, p < .001, mdiff = 8.17, SDdiff = 8.24), stimulative music (t(34) = 8.21, p < .001, mdiff = 12.42, SDdiff = 8.95), and control (t(34) = 13.15, p < .001, mdiff = 15.83, SDdiff = 7.12) conditions. As the overall estimated effect of music listening on the recovery process of healthy individuals following laboratory stressors may be relatively modest, it becomes particularly important to ensure that a sufficient stress response is elicited, to provide a larger window of opportunity in which the effect of music listening may be exerted on participants’ recovery processes. We thus encourage future studies to adopt validated, (variations of) well-known stress tasks, such as the TSST [109], SECPT [110], or CO2 stress task [111], which have been demonstrated to consistently elicit marked physiological and psychological stress-related responses in laboratory settings. Furthermore, we remind future studies to candidly report the results of their stress induction procedures to facilitate subsequent meta-syntheses of the effects of music listening on stress recovery.
As the current review focused on the effects of music listening after a stressor, studies where music was played before or during a stressor were omitted from our analyses. However, several studies suggest that the timing at which music is played (i.e., before, during, or after a stressor) may influence its effects on stress recovery. For example, in Burns et al. [48], participants who listened to classical music while anticipating a stressful task exhibited lower post-music heart rate compared to participants who anticipated the stressor in silence. Similarly, concentrations of salivary cortisol were lower for participants who watched a stressful visual stimulus while listening to music compared to those who watched the same stimulus without music [112]. Together, these findings hint that, when listened during a stressor, music may attenuate cortisol responses [9, 113], thus reducing the subsequent need for recovery. On the other hand, Thoma et al. [9] reported that participants who listened to music prior to a stressor exhibited higher post-stressor cortisol compared to participants who listened to an audio control. Interestingly, despite the stronger stress response, Thoma et al. [9] noted a trend for quicker ANS recovery among participants who listened to music, particularly with regards to salivary alpha-amylase activity. This pattern of findings is consistent with the notion forwarded by Koelsch et al. [61], in that music listening may promote a more adaptive stress response, thus facilitating subsequent stress recovery processes. To date, research on timing differences in the context of music listening and stress recovery is scarce. Thus, future studies could further examine the influence of such timing differences to better understand their role in the relationship between music listening and stress recovery.
Given the pervasiveness of stress, Ecological Momentary Assessment (EMA) studies may provide a more intimate outlook on the dynamics of daily music listening behaviour, particularly for the purpose of stress recovery. For example, through an ambulatory assessment study, Linnemann et al. [38] revealed that music produced the most notable reductions in physiological and psychological stress outcomes when it was listened to for the purpose of ‘relaxation’, compared to other reasons such as ‘distraction’, ‘activation’, and ‘reducing boredom’. Indeed, given their high ecological validity, EMA studies may provide further insight into important contextual variables in the relationship between music listening and stress recovery. For example, in an EMA study, listening to music in the presence of others was related to decreased subjective stress, attenuated cortisol secretion, and higher activity of salivary alpha-amylase [55]. Furthermore, physiological responses to music may co-vary between members of a dyad when music is listened to by couples [114]. Thus, given the benefits of EMA studies, we invite future studies to continue exploring the dynamics and contextual factors of music listening behaviour for stress recovery in daily life.
Lastly, we encourage studies to support open science research practices, and to clearly report statistical information that may be relevant for meta-syntheses (e.g., means and standard deviations per time point, per experimental condition, etc.). Additionally, based on our assessment of study quality using the RoB 2, pre-registration of analysis plans can be helpful to ensure that the conducted study is of overall high quality. Next, we encourage studies to note down which specific musical stimuli were used, particularly those self-selected by participants [69, 99], as this enables future exploratory analyses of structural commonalities between different musical stimuli. Musical features from individual songs may be extracted by means of audio information extraction packages, such as MIRtoolbox [115]. Alternatively, individual song titles may be used to query related meta-data from online databases of various music streaming platforms. This meta-data can subsequently be used to obtain additional insight into the effects of music listening on stress recovery.
Limitations of the current review
To our knowledge, our review is the first to comprehensively investigate the effect of music listening on stress recovery within healthy individuals. Given the explicit focus of our review, our meta-analytic data set excluded the more prominent effects of music listening in both medical and therapeutic settings [12, 13], allowing us to obtain results that are tentatively more representative of daily stress recovery processes. Despite this, the present review is not without its limitations:
First, although the specific focus of our review has allowed us to obtain a portrait of the effects of music listening on stress recovery in well-controlled experimental settings, the results of our review may be difficult to generalize to situations in which individuals experience prolonged stress responses. Stress induction procedures in experimental studies are designed to elicit acute stress responses that are meant to subside upon conclusion of an experiment [83]. Although we believe these procedures provide a suitable approximation of typical stressors in daily life, certain stressors in daily life may also persist for a longer time. The manner and magnitude in which music listening influences prolonged stress responses may potentially differ from the way music influences acute, laboratory-induced stress responses [18, 45]. However, studies investigating the effect of music listening on stress recovery in the long-term are particularly rare.
Next, despite our best efforts to obtain relevant meta-analytic information from all studies selected for our review, our meta-analytic data set was ultimately constructed from a subset of fourteen studies. Although the subset allowed us to extract sufficient information to estimate an overall effect of music listening on stress recovery, several estimated effects at moderator level were derived from merely one or two studies (see Table 3). This precluded us from drawing further, meaningful conclusions about the results of our moderator analyses.
Finally, despite our clear focus on the effects of music listening on stress recovery within healthy individuals, there was substantial heterogeneity in our meta-analytic data set that could not be fully explained by the inclusion of moderators. Although the systematic review portion of our results highlighted potential additional sources of between-study heterogeneity, these additional sources could not be evaluated in our meta-analytic data set. We note, for example, that all studies utilized different musical stimuli to investigate the effect of music listening on stress recovery (see Table 4). The differential effects of these musical stimuli were difficult to account for in our meta-analysis, given the limited number of included studies. Overall, the significant heterogeneity in our meta-analytic data set suggests that our moderator analyses should be interpreted with caution.