The authors have declared that no competing interests exist.
Conceived and designed the experiments: CH JK IK RLG. Performed the experiments: CH. Analyzed the data: CH IK RRE RLG. Contributed reagents/materials/analysis tools: JK. Wrote the paper: CH IK KBJ TJK RLG.
Music has pain-relieving effects, but its mechanisms remain unclear. We sought to verify previously studied analgesic components and further elucidate the underpinnings of music analgesia. Using a well-characterized conditioning-enhanced placebo model, we examined whether boosting expectations would enhance or interfere with analgesia from strongly preferred music. A two-session experiment was performed with 48 healthy, pain experiment-naïve participants. In a first cohort, 36 were randomized into 3 treatment groups, including music enhanced with positive expectancy, non-musical sound enhanced with positive expectancy, and no expectancy enhancement. A separate replication cohort of 12 participants received only expectancy-enhanced music following the main experiment to verify the results of expectancy-manipulation on music. Primary outcome measures included the change in subjective pain ratings to calibrated experimental noxious heat stimuli, as well as changes in treatment expectations. Without conditioning, expectations were strongly in favor of music compared to non-musical sound. While measured expectations were enhanced by conditioning, this failed to affect either music or sound analgesia significantly. Strongly preferred music on its own was as pain relieving as conditioning-enhanced strongly preferred music, and more analgesic than enhanced sound. Our results demonstrate the pain-relieving power of personal music even over enhanced expectations.
Clinicaltrials.gov
Music is a treatment that can affect pain perception through a complex set of past and present cues. Juslin and Vastfjall (2008), Salimpoor et al. (2011), and Koelsch (2009), among others, posited many ways through which music may activate regional brain networks mediating reward and anxiolytic effects that also overlap with regions involved in analgesia
No less complex than music are placebo responses, whose definition has evolved to emphasize psychosocial and contextual cues that generate objectively measurable psychobiological effects
To investigate whether the analgesic effect of an emotionally pleasurable stimulus, such as strongly preferred music, could be modified by the well studied conditioning paradigm used within placebo analgesia, we designed an experiment that juxtaposed music with conditioned-expectancy enhancement in two opposing ways - one in which conditioning aimed to enhance expectations of relief for music, and the second in which music would be put up against conditioned expectancy enhancement of a non-musical control sound. We used a well-validated conditioning procedure
All study procedures were carried out with Institutional Review Board approval from MIT COUHES (protocol #1206005109) and MGH (protocol #2012P000969), approved on June 27, 2012 and June 31, 2012 respectively. Data was collected at the Martinos Imaging Center in Charlestown, MA. All participants were taken through written informed consent prior to initiating any study procedures. The protocol for this trial and supporting CONSORT checklist are available as
Participants met the following criteria:
A two-session experiment was performed with 48 healthy, pain experiment-naïve participants. In a first cohort, 36 were randomized (via assignment to outcomes from a random number generator) into 3 treatment groups, including music conditioning, non-musical sound conditioning, and no-conditioning. The primary author completed the random sequence generation, enrollment, and participant assignments. Primary outcome measures included the change in subjective pain ratings to calibrated experimental noxious heat stimuli, as well as changes in treatment expectations on an Expectancy of Relief Scale (ERS). The ERS is a 0–10 scale (with 0 indicating “does not work at all” and 10 indicating “complete pain relief”) used to measure the expectation of treatment pain relief
The two behavioral sessions were separated by a minimum of 2 days and a maximum of 10 days. Prior to coming to the experimental site, all participants were asked to acquire/choose a set of personal music that satisfied certain criteria for the study (see next section). During both sessions, participants received sets of calibrated noxious thermal stimuli using a Thermal Sensory Analyzer (TSA-II) or the Pathway CHEPS model (Contact Heat-Evoked Potential Stimulator) with a 3 cm×3 cm probe (Medoc Advanced Medical Systems, Rimat Yishai, Israel) running proprietary computerized visual analog scale software (COVAS). Each stimulus was 12 seconds long (2.5 second ramps with 7 second plateaus), and was delivered in blocks of 5 with a jittered average of 30 seconds between stimuli. After each stimulus, participants used 0–100 visual analog scales to rate the pain intensity and unpleasantness, with anchors of 0 = no pain and 100 = worst imaginable pain.
In the
Participants who were eligible to continue first received identical baseline stimuli (calibrated individually during Session 1) on three spots of their right arms, in silence (
All subjects first receive identical baseline stimuli on three spots of their right arm. Conditioning groups were then given a verbal suggestion specific to their conditioned audio target, while the no-conditioning group was given a neutral suggestion. All groups then received four blocks of two stimulus intensities, with levels determined during Session 1 (target levels = 20–40 out of 100 and 55–70 out of 100 on the pain intensity VAS). Conditioning groups received lower heat levels when the conditioned audio stimulus was on, while the no-conditioning group received stimuli with silence. All groups were then tested with three audio conditions: music, sound, and silence, in randomized order across participants. Within each set of trials, lead-in arrows represent 1 minute of audio or silence that preceded the start of pain stimuli. In all groups ERS was assessed right after verbal suggestion, then following conditioning/variable pain, and after testing (red arrows). We added one more baseline ERS assessment within the music conditioning replication group at the end of session 1 (orange arrow).
Expectancy was assessed with the ERS at time points including post-verbal suggestion, post-conditioning/variable pain, and post-testing (
In the music conditioning replication cohort that was enrolled after completion of the initial study, we added an explicit reminder that they would be receiving stimuli identical to that from baseline during the conditioning block to better ensure high expectations of music analgesia. The purpose of this group was to verify the effects of expectancy-enhancement on music analgesia in the first music conditioning group. The same ERS measures were taken as in the original cohort, and in addition we obtained one further ERS score at the end of Session 1 (orange arrow in
To determine individual song choices, we used a list of criteria designed to specify a particular controlled set of participants’ personal music choices for analgesia. Prior to coming to the first study visit, all participants were asked to provide a set of songs (each at least 4 minutes long) that they love, had been familiar with for at least a few years, could listen to repeatedly, varied across a range from very relaxing to very energizing, did not evoke specific memories or chills, and that they had not seen the music video for. We excluded songs associated with specific visual memories including music videos in order to concentrate our study more on strong emotions associated with the music itself
The non-musical sound samples consisted of alternating, frequency-filtered noise clips that were chosen based on studies where participants perceived filtered noise to be less distressing than white noise
The primary outcomes were pre-post changes in pain intensity and unpleasantness. For the first cohort these data were analyzed using a 3×3 (conditioning by audio type) mixed model analysis of variance (ANOVA) with conditioning (music, sound, or none) as the between-subjects factor, and audio modality (music, sound or silence) as the within-subjects factor. We then compared pain intensity and unpleasantness in the replication cohort with the original music group using a 2×3 (group by audio) mixed model ANOVA. Bonferroni corrections were used for post hoc comparisons following significant main effects.
We examined expectancy in two ways. Participants in the no-conditioning group, where music and sound were presented equally, rated expected relief for both music and sound. These data were analyzed using a 3×2 (time by audio type) repeated measures ANOVA. Participants in the music and sound conditioning groups rated expected relief only for the audio type that was the focus of the verbal suggestion and conditioning. These data were analyzed using a 3×2 (time by conditioning) mixed model ANOVA. Statistical analyses were done using IBM SPSS Statistics version 20. Final sample sizes for sufficient power to detect outcomes were determined during an interim analyses conducted during data collection, based on previous work completed with the current study’s methods by the laboratory. Finally, to analyze the semi-free form qualitative interview and survey data, we performed thematic analyses following the procedure and recommendations of Braun and Clark
We enrolled 58 participants, from which 10 were dropped due to the following factors: the requirement of stable and reliable responses to pain stimuli necessary to perform the quantitative experiments, not wishing to continue, or not following study procedure guidelines. We studied 12 participants in each of four groups for a total of 48 participants (32 females), with average age 27±7. 36 of these comprised the initial cohort for which the main experiment was conducted. 12 participants served as a replication sample for the music conditioning group. All participants completing study procedures were included in the final analyses within originally assigned groups.
Examination of the distribution of changes in pain intensity and unpleasantness ratings revealed extreme positive kurtosis for intensity change scores for music (3.99). A box plot revealed one extreme outlier in the music conditioning group. With data from this participant removed, skewness and kurtosis for these six variables were all between +/−1. Hence, subsequent analyses were conducted with data where this one outlier was removed.
N = 35. Shown here are pre (baseline) minus post (testing) pain difference scores (mean ± SEM); positive values indicate analgesia. Panel A displays Pain Intensity, panel B displays Pain Unpleasantness.
Conditioning | Post Audio | Pain Intensity | Pain Unpleasantness | ||
None | Music | 55(13) | 47(15) | 46(20) | 37(20) |
Sound | 56(16) | 56(15) | 46(22) | 47(20) | |
Silence | 58(15) | 61(14) | 48(23) | 54(23) | |
Music | Music | 59(19) | 59(16) | 50(19) | 44(16) |
Sound | 59(22) | 63(16) | 48(19) | 51(15) | |
Silence | 57(16) | 66(15) | 49(16) | 54(15) | |
Sound | Music | 50(17) | 47(17) | 45(23) | 42(24) |
Sound | 49(14) | 49(19) | 44(21) | 43(23) | |
Silence | 51(19) | 55(20) | 44(24) | 51(25) | |
Music replication | Music | 55(18) | 48(15) | 43(24) | 35(19) |
Sound | 55(15) | 55(14) | 43(21) | 44(22) | |
Silence | 57(18) | 61(14) | 44(26) | 49(23) |
Mean (SD) ratings of pain intensity and unpleasantness for each audio type before and after the conditioning/variable pain phase for both cohorts. Note that all pre- pain stimuli occurred in silence (i.e. pre- ratings for music and sound are given for the corresponding skin spot of stimulation). One outlier subject has been removed from the original music conditioning group (as noted in text).
The 3×3 ANOVA on changes in pain unpleasantness in the primary cohort revealed a significant main effect for audio type, F (2,64) = 15.43, p<.001, eta2 = .33. Paired contrasts using a Bonferroni adjustment for multiple contrasts revealed that music decreased pain unpleasantness significantly compared to both sound (p = .015) and silence (p<.001), and the difference between sound and silence just reached significance (p = .05). Neither the main effect for conditioning nor the conditioning by audio type interaction reached significance. For this aspect of the pain experience as well, it also appears that the superiority of music over sound, and of sound over silence at a statistical trend level, held regardless of conditioning. The 2×3 ANOVA comparing the original music conditioning group to the replication conditioning group showed a similar main effect for audio type, F (2,42) = 8.44,
The ANOVA in the no-conditioning group revealed a main effect for audio type, F (1,11) = 43.70, p<.001, eta2 = .80). These participants expected much greater pain relief from music (Mean = 5.08; SD = 1.69) than from sound (Mean = 2.89; SD = 1.42), an effect that accounted for 80% of the variance. Neither the main effect for time nor the time by audio type interaction approached significance. The ANOVA in the two conditioning groups revealed a main effect for time, F (2,42) = 6.05, p = .005, eta2 = .22. Polynomial contrast revealed a significant quadratic effect, F (1,21) = 13.31, p = 002, eta2 = .39, in which expected relief increased following conditioning and then decreased after post-conditioning testing. The main effect for conditioning group did not approach significance. The group by time interaction was not significant (p = .096). Expectancy scores are shown in
. Means ± SEM are shown for N = 35. Time 1 occurs after verbal suggestion but prior to the conditioning phase in the conditioning groups, and prior to the matched variable pain phase in the no- conditioning group. Time 2 occurs immediately after the conditioning/variable pain phase, and Time 3 occurs after testing.
Without conditioned-expectancy enhancement, music had significantly higher measured positive expectancy and was most effective at decreasing pain, followed by sound over silence. Expected relief was boosted in the conditioning groups, however typical corresponding increases in analgesia
While some routes to analgesia, such as distraction and positive expectations of pain relief, have been seen to be additive
Additionally, the preferred personal music could have yielded maximal experimentally detectable endogenous analgesia in a ceiling effect. We believe that the strength of the analgesic effects were likely due to the latent conditioned relationships subjects had with their songs, leading to a more complex view of music as an analgesic agent. We suggest it can be viewed in the same framework that placebo responses are increasingly being seen in - that of a rich, nuanced set of contextual personal factors. An individual’s song becomes connected over time to people, places, and circumstances such that there is a robust latent effect from the history and time spent with the music. Indeed, additional analyses of systematically gathered, qualitative participant experiences revealed themes consistent with those seen in the nascent field of neurobehavioral music research
We further propose that enhancing expectations for sound may not have boosted analgesia precisely because the conditioning phase, with lowered pain during sound samples, also included trials with powerfully rewarding music. If conditioning based expectancy modulation works through pathways common to music, then the presence of highly personal music during our attempted generation of increased sound expectancy would certainly interfere with that process. It would be akin to trying to show the glow of a lamp when the sun is shining brilliantly into the room. This cognitive conflict brings to mind the framing of belief and its effects on learning that Corlett et al (2010) put forth: “…sculpted connections…predict subsequent states of the internal and external world and respond adaptively; however, should that next state be surprising, novel or uncertain new learning is required.”
One caveat to mention is the somatotopic specificity that has been shown for placebo analgesia
On the surface it seems that music should be amenable as a modulation target for the same reasons contextual placebo factors are being discussed for maximal clinical effect. Yet here, the combination of music with an expectancy-mediated conditioning procedure showed that music analgesia is not improved by these methods. On its own it performed just as well as it did following conditioning. Thus, despite the likely existence of many converging mechanisms of endogenous analgesia, music and expectancy-based placebo analgesia were not additive or synergistic in this study.
In her review, Tracey
(DOC)
(DOCX)
(DOC)
The authors would like to acknowledge the contributions of Rosa Spaeth, who assisted with data analysis and interpretation, and of Drs. Gregory Fricchione, M. Christian Brown, and Anne Blood, who assisted in the guidance of this work. We also acknowledge Amanda Cook, Alex Cheetham, and Sonya Freeman for their help in facilitating protocol approval and volunteer recruiting.