Participants. A naive random sample 59 participants was included in the data analysis (age range 19 to 35 years, M = 22.15 years, SD = 3.22, 29 females), based on an a-priori sample size estimation using G*Power (version 3.1.9.6)40 which indicated that for the detection of a small to medium effect size (d = 0.4)23 with an α-level of 0.05 and a statistical power of 90% (one-tailed), a sample of n = 59 is required. Exclusion criteria were use of any prescription drugs, psychiatric or neurological disease, hearing impairment, and history of substance dependence (according to self-report). Participants did not consume alcohol nor any medicine for the treatment of pain at least 24 hours prior to the experiment. None of the participants had musical anhedonia according to the Barcelona Music Reward Questionnaire (M = 81.06, SD = 11.37)41. The study was carried out in accordance with the guidelines of the Declaration of Helsinki and approved by the Regional Committee for Medical and Health Research Ethics for Western Norway (Reference Number: 2019/1031). The study was registered as a clinical trial at ClinicalTrials.gov (registration number NCT05267795), and the trial was first posted on 04/03/2022. All participants provided written informed consent before enrollment and received a monetary compensation of 200 NOK following participation in the experiment.
Stimuli. Music stimuli were presented with an average sound pressure of approximately 60 dB SPL over Beyerdynamic DT 770-PRO 250 Ohm headphones. 10 instrumental music experts were selected, each 30 seconds long (song characteristics are provided in Supplementary Table S2). Each excerpt was played twice during the experimental task, once during the active task and once during the passive task. Stimuli were delivered using the Matlab-based toolbox Cogent 2000 (version 1.33).
Pressure Algometry. Pain was applied by the experimenter using a Wagner Force One Pressure Algometer (Wagner Instruments, Greenwich, USA). The equipment head (used to apply the pressure) had a diameter of 12mm and a surface area of 113mm2. Pain stimuli were delivered to index, middle and ring fingers of both hands. For each participant, pain thresholds were determined separately for each of the six fingers used for pain delivery, prior to the actual experiment. The average value of these pain thresholds was then computed, and for the actual experiment, only 50% of the participant’s average pain threshold was delivered as pain stimulus in each of the 40 experimental trials (e.g., if a participant had an average pain threshold of 5 kg, then a pressure stimulus of 2.5 kg was applied in the actual experiment). During the entire experiment, the applied pressure was measured with a 10 Hz sampling rate using MESURgauge Plus by Mark-10 (version 2.0.5). The statistical analysis of these data showed that neither the applied pressure, nor the duration of applied pressure, differed significantly for the within-subjects factors Condition (music, silence; pressure intensity: F(1,58) = 0.33, p = .568; pressure duration: F(1,58) = 1.62, p = .209), nor Task (active, passive; pressure intensity: F(1,58) = 0.33, p = .568; pressure duration: F(1,58) = 0.66, p = .419). Likewise, no significant interaction between Condition and Task was observed for the pressure intensity (F(1,58) = 1.00, p = .321), nor for pressure duration (F(1,58) < 0.01, p > .999). Thus, we can exclude the possibility that the pain ratings of participants were simply an artifact of faulty stimulus delivery (note that the experimenter was also blinded to avoid any systematic differences of pain stimulus delivery between the experimental trial types). For further information see Supplementary Information and Supplementary Table S3.
Procedure. 40 experimental trials were delivered, 10 for each experimental trial type (Music Active, Music Passive, Silence Active, Silence Passive). The experimenter was blinded during the entire experimental procedure. Each trial started with an instruction screen where participants were either instructed to tap their right foot like a metronome (active tapping task) or to relax (passive control task). Then, participants pressed the space bar and either music started to play, or a silent period started, and participants either had to tap or relax (depending on the trial type), while looking at a fixation cross in the middle of the screen. After 20 seconds the experimenter applied pressure with the Algometer to a fingernail for 10 seconds. Then (after the pain application had stopped), participants indicated the pain they felt at the end of the pain application on a 9-point scale (1 = very little, 5 = medium, 9 = very strong), followed by a pleasantness rating (1 = very uncomfortable, 5 = medium, 9 = very comfortable) and an arousal rating (1 = very calm, 5 = medium, 9 = very activated). In music trials, participants also rated to which extent they were familiar with the music (1 = not at all, 5 = partially known, 9 = well known). Participants were aware that the experimenter could not see their ratings (so that they would not feel influenced in any way). Each trial lasted about one minute (thus, the experiment had a duration of approximately 40 minutes). After completing all 40 experimental trials, participants listened to the 10 music excerpts again and indicated their preference for each excerpt on a 9-point scale (1 = strongly disliked, 5 = medium, 9 = strongly liked).
Experimental Design & Data Analysis. Data analysis was computed using R (version 4.0.4) including the R Stats, R Base, and ggplot 2 (version 3.3.3)42 packages. Figures were graphically edited with CorelDraw Graphics Suite version 21.0.0.593 (Corel Corporation, Ottawa, Ontario, Canada). We conducted an LME analysis on single trial level, as this analysis allowed to reduce interindividual variance introduced by participants and music excerpts, and decrease the risk for false positives43. The analysis was applied by the use of the lme4package (version 1.1-26)44. The model included the two categorical fixed-effects factors Condition (music [+0.5] and silence [-0.5]) and Task (active [+0.5] and passive [-0.5]) as well as their interaction as predictors for the perceived pain in each individual experimental trial (rated on a scale ranging from 1 to 9). Furthermore, the model included the Participants and music Excerpts as random-effects factors. The model was estimated by using a restricted maximum likelihood approach45. The α-level for significance was 0.05. We estimated degrees of freedom and p-values with the Satterthwaite approximation implemented in the lmerTest package (version 3.1)46. We used the R package influence.ME (version 0.9-9)47 to test the data for statistical outliers on the participant level. No outliers were detected, and the model was conducted with a sample of 59 participants. Especially in clinical contexts the lack of statistical significance does not necessarily mean that a treatment has no effect48,49. Therefore, and to comprehensively explore the pattern of results, we resolved the interaction between Condition and Task with a simple slope LME analysis implemented in the R package jtool (version 2.1.2)50. In addition, we calculated Cohen's d effect size23 by the use of the R package lsr (version 0.5)51 because this effect size measure is taken as one of the most important indicators of clinical significance49.