This is a repository copy of Temporal dynamics of musical emotions examined through intersubject synchrony of brain activity..

Transcription

1 This is a eositoy coy of Temoal dynamics of musical emotions examined though intesubject synchony of bain activity.. White Rose Reseach Online URL fo this ae: htt://eints.whiteose.ac.uk/92892/ Vesion: Acceted Vesion Aticle: Tost, W., Fühholz, S., Cochane, T. et al. (2 moe authos) (2015) Temoal dynamics of musical emotions examined though intesubject synchony of bain activity. Social Cognitive and Affective Neuoscience, 10 (12) ISSN htts://doi.og/ /scan/nsv060 Reuse Unless indicated othewise, fulltext items ae otected by coyight with all ights eseved. The coyight excetion in section 29 of the Coyight, Designs and Patents Act 1988 allows the making of a single coy solely fo the uose of non-commecial eseach o ivate study within the limits of fai dealing. The ublishe o othe ights-holde may allow futhe eoduction and e-use of this vesion - efe to the White Rose Reseach Online ecod fo this item. Whee ecods identify the ublishe as the coyight holde, uses can veify any secific tems of use on the ublishe s website. Takedown If you conside content in White Rose Reseach Online to be in beach of UK law, lease notify us by ing eints@whiteose.ac.uk including the URL of the ecod and the eason fo the withdawal equest. eints@whiteose.ac.uk htts://eints.whiteose.ac.uk/

2 TEMPORAL DYNAMICS OF MUSICAL EMOTIONS EXAMINED THROUGH INTER-SUBJECT SYNCHRONY OF BRAIN ACTIVITY Wiebke Tost* 1,2, Sascha Fühholz 1,2, Tom Cochane 3, Yann Cojan 1,4, Patik Vuilleumie 1,4 1) Swiss Cente of Affective Sciences, Univesity of Geneva, Switzeland 2) Neuoscience of emotions and affective dynamics laboatoy, Univesity of Geneva 3) Deatment of Philosohy, Univesity of Sheffield, U.K. 4) Laboatoy fo Behavioal Neuology and Imaging of Cognition, Univesity of Geneva *Coesonding autho: Wiebke J. Tost Swiss Cente of Affective Sciences, Biotech Camus, Univesity of Geneva 9, chemin des Mines, 1202 Geneva, Switzeland Mail: Abstact: To study emotional eactions to music, it is imotant to conside the temoal dynamics of both affective esonses and undelying bain activity. Hee we investigated emotions induced by music using functional magnetic esonance imaging (fmri) with a data-diven aoach based on inte-subject coelations (ISC). This method allowed us to identify moments in the music that oduced simila bain activity (i.e. synchony) among listenes unde elatively natual listening conditions. Continuous atings of subjective leasantness and aousal elicited by the music wee also obtained fo the music outside of the scanne. Ou esults eveal synchonous activations in left amygdala, left insula, and ight caudate nucleus that wee associated with highe aousal, wheeas ositive valence atings coelated with deceases in amygdala and caudate activity. Additional analyses showed that synchonous amygdala esonses wee diven by enegy-elated featues in the music such as ms and dissonance, while synchony in insula was additionally sensitive to acoustic event density. Inte-subject synchony also occued in the left nucleus accumbens, a egion citically imlicated in ewad ocessing. Ou study demonstates the feasibility and usefulness of an aoach based on ISC to exloe the temoal dynamics of music ecetion and emotion in natualistic conditions. Keywods: Inte-subject coelation, musical emotion, musical featues, amygdala, fmri

3 INTRODUCTION The owe of music to aouse stong emotions in listenes is ooted in its essentially temoal natue. This quality enables music to catue the nuance and flux of emotional exeience in a way that listenes can emathically tack fom moment to moment (Cochane, 2010). It is theefoe cucial to conside the temoal dynamics of music ocessing in ode to bette undestand the geneation of emotional eactions and uncove the esonsible bain mechanisms. Howeve, music eesents a vey comlex acoustic stimulus, whose imact on the listene is challenging to chaacteize. Although it has been obseved that musical woks ae caable of stiking a common chod evoking consistent emotions acoss diffeent individuals (Pankse and Benatzky, 2002, Guhn et al., 2007), it emains difficult to edict which kind of musical event o which acoustic featues will tigge a secific emotion. In the cuent study, we tackle this oblem by obing the undelying neual activities that listenes shae while listening to musical woks. Secifically, we investigated the elationshi between the two majo oeties of music, namely, its emotional and temoal dimensions. Moe geneally, valuable insight into music ecetion can be gained by bette undestanding this elationshi and inointing thei substates in the human bain. Seveal neuoimaging studies have investigated emotions induced by music (Blood and Zatoe, 2001, Menon and Levitin, 2005, Salimoo et al., 2011, Tost et al., 2012, Koelsch, 2014). Howeve, these studies tyically used block designs aveaging bain esonses to diffeent kinds of musical excets ove a eiod of seveal seconds o minutes, and wee thus geneally based on stong a ioi assumtions about the emotional imact and homogeneity of a whole musical excet. Hence, even though this eseach has yielded many ecious insights (Koelsch, 2014), it did not take into account the temoal dynamics of how musical emotions may develo in time. Yet, in thei vey essence, emotions ae defined as tansient changes in mental and hysiological states (Schee, 2004). Theefoe ou study eesents a significant advance on these ealie studies by focusing on tansient fluctuations in bain activity evoked by musical stimuli in a consistent manne acoss diffeent individuals. To this aim, we alied a novel data-diven methodological aoach based on inte-subject coelations (ISC). The method of inte-subject coelations to investigate bain activity with fmri has fist been intoduced by Hasson and colleagues (Hasson et al., 2004) to study visual ecetion duing movie viewing unde natualistic conditions. By calculating aiwise coelations of the fmri signal of diffeent aticiants and then aveaging the esults at the gou level, this technique emits the identification of bain aeas that show a simila time-couse of activation acoss subjects, without the need to define events a ioi. Futhe, by using evese-coelation as a mean to tace back in the oiginal stimulus (e.g. movie) when high ISC occued fo secific bain egions (coesonding to moments of inceased, Hasson and colleagues (2004) demonstated that scenes with close-us on faces wee esonsible fo inceased ISC in the fusifom cotex, a visual aea citically imlicated in face ecognition (Kanwishe et al., 1997). Convesely, viewing hand movements accounted fo highe ISC in somatosensoy and moto cotical aeas. Seveal othe studies have also used intesubject coelation analysis as a data-diven aoach to study the attentional and emotional effects of films

4 (Fuman et al., 2007, Hasson et al., 2008a, 2008b, 2010; Jaaskelainen et al., 2008), as well as seech ocessing (Wilson et al., 2008) and moe ecently music ecetion (Allui et al., 2011, Abams et al., 2013). But these studies used this methodology mostly to detemine common fluctuations in bain activity at diffeent fequency anges o to identify lage-scale netwoks ecuited in these task conditions in geneal. Fo examle, Abams and colleagues (2013) comaed natual symhonic music with scambled vesions of the same iece in ode to identify bain egions that would secifically synchonize thei activity duing natual music. These authos found that auditoy cotices, thalamus, moto aeas, as well as fontal and aietal egions synchonize moe acoss aticiants when listening to natual music than to scambled vesions. Howeve, beyond synchonization acoss entie musical ieces, ISC can be alied and be aticulaly infomative ove smalle time scales in ode to identify secific moments at which the sensoy inuts will yield high inte-subject ageement in neual activity. Thus, as successfully used by Hasson et al. (2004), this aoach combined with evese coelation ovides a vey oweful tool to unveil unique events in music ieces that geneate simila bain esonses and consistent emotional eactions in the listenes. In the cuent study, we emloyed a novel ISC aoach to identify moments in time duing which simila changes in bain activity aose while eole listened to the same iece of music, without any a ioi definition of these moments fom music content itself. As music can elicit oweful and consistent emotional eactions (Schee, 2004, Juslin, 2013) we edicted that emotional exeiences shaed by diffeent eole at the same moment of the music should coesond to the ecuitment of simila bain ocesses, and thus lead to enhanced synchony of activation attens between listenes. Futhemoe, if ISC occus at aticula moments in the music, we also edicted that it may eflect the occuence of secific featues in the music (e.g. in tems of acoustic content o subjective affect) and tested which featues dove ISC in diffeent bain egions. This aoach diffes fom the standad GLM aoach based on a model of the BOLD esonse to edefined stimuli in the sense that hee no e-classified stimuli o conditions ae chosen a ioi to detemine the coesonding bain esonses. Instead, moments of coheent bain activity acoss aticiants wee fist detemined and subsequently used to analyze whethe (and which) asects of emotional feelings o acoustical inuts tiggeed these synchonized bain esonses. In addition, unlike a taditional GLM analysis based on the time-couse of acoustic featues o emotion atings, ou measue of ISC may otentially detect musical events eliciting aticula bain esonses even when they ae not stictly simultaneous with an incease in secific music featues o show non-linea elationshis with these featues (e.g. some ISC and coesonding emotional event may aise only afte eaching a cetain level). Pevious neuoimaging studies have aleady detemined bain egions involved in the ocessing of vaious acoustic featues fo seech, osody, and music (Platel et al., 1997, Zatoe et al., 2002, Schime and Kotz, 2006, Allui et al., 2011, Fuhholz et al., 2012). A few of these studies used coelation aoaches that diectly egess bain activity fom each aticiant with aticula chaacteistics of the stimulus (e.g. loudness, o dissonance) ove its whole esentation eiod (Chain et al., 2010, Allui et al., 2011). This wok has established that elementay acoustic featues ae ocessed in imay auditoy aeas, then integated into a coheent ecet

5 in seconday auditoy egions, and finally categoized as a function of cognitive and affective meaning in highelevel cotical aeas such as the infeio fontal gyus and obitofontal cotex, lus cotical and subcotical moto stuctues (Schime and Kotz, 2006, Koelsch, 2011). Meanwhile, fom sychological studies, it is also known that cetain acoustic featues of music ae linked with aticula emotional esonses (Gabielson and Juslin, 2003). Fo examle, fast temo and high volume both induce highe subjective aousal (Schubet, 2004, Etzel et al., 2006, Gomez and Danuse, 2007), wheeas dissonance is closely elated to eceived (un)leasantness (Blood et al., 1999, Koelsch et al., 2006). Dissonance has also been associated with evocations of excitement, fea, and ange, as well as with sadness (Kumhansl, 1997). Timbe shaness coelate with ecetion of ange and timbe bightness with hainess (Juslin and Laukka, 2000), wheeas a flat sectum with many hamonics has been elated to disgust and fea (Schee and Oshinsky, 1977). Finally, hythmic stability is linked with elaxation, eacefulness, o hainess (Hevne, 1936). Hee, by detemining acoustic featues and emotional esonses concomitant to inceased synchony of bain activity between listenes, we wee able to gain bette knowledge about both the auditoy chaacteistic and neual stuctues undelying music-induced emotions. Although the exact music dimensions esonsible fo diffeentiated affective exeiences emain ooly known (Juslin, 2013, Koelsch, 2013), the involvement of distibuted netwoks of sensoy, moto, and limbic egions showing distinctive attens of activation accoding to tye of emotion elicited by the music have been obseved (e.g. Tost et al., 2012, Koelsch and Skouas, 2013). The vental stiatum, but also ACC, insula, and olandic aietal oeculum ae geneally ecuited by leasant emotions induced by music such as joy o wonde (Blood and Zatoe, 2001, Koelsch et al., 2006, Mitteschiffthale et al., 2007, Tost et al., 2012). On the othe hand, negative musical emotions such as sadness, fea o tension, have been associated with the amygdala, hiocamus, and vental ACC (Koelsch et al., 2006, Geen et al., 2008). Amygdala activity also coelates negatively with the intensity of eceived chills (Blood and Zatoe, 2001). While these studies eesent significant ogess on the neual bases of musical emotions, they mostly focused on emotional exeiences eoted fo the whole duation of a music iece, athe than moe tansient emotions evoked at ecise time oints duing listening. In contast ou study allowed us to investigate emotions induced by music at diffeent moments in a iece using an ecological design duing elatively natual listening conditions. Paticiants head a full movement fom a wok in the classical eetoie (9-14 minutes) while undegoing fmri. Using a data-diven aoach based on ISC, we fist identified those moments in the music whee common bain esonses wee evoked acoss diffeent listenes. These eaks in ISC wee then analyzed in tems of the musical featues esent at the same moment and the coesonding emotional atings made by listenes using a continuous dynamic judgment ocedue. We edicted that distinct bain egions, including not only sensoy cotices, but also limbic stuctues and basal ganglia, would diffeentially be ecuited by diffeent emotion dimensions and esond to diffeent music featues. Note that ou method aimed at identifying common and synchonized bain esonses acoss all listenes, but not moe idiosyncatic eactions of individual aticiants that did not occu in othe aticiants at the same moment. Imotantly, howeve, this novel aoach shaes the

6 vitue of avoiding a ioi assumtions about secific dimensions of the stimuli, as used by othe studies (Hasson et al., 2004, Hasson et al., 2010) to investigate vision. Thus, by extacting meaningful infomation fom ongoing bain activity in natualistic conditions, measues of ISC ovide a oweful data-diven methodology that may offe valuable infomation about music ocessing and music-induced emotions. METHOD Subjects Seventeen healthy aticiants (mean age: 25.4 yeas SD , 9 females) with no histoy of neuological o sychiatic disease took at in the fmri exeiment. Anothe fouteen aticiants (mean age: 28.7 yeas SD , 8 females) took at in the behavioal exeiment. All aticiants eoted nomal heaing, a liking fo classical music, and had a minimum of 5 yeas of actical musical taining. Paticiants in the fmri exeiment wee all ight-handed. All subjects wee aid fo thei aticiation and gave infomed consent in accod with the egulation of the local ethics committee. The study was aoved by the local ethics committee. Stimulus mateial Musical stimuli consisted of thee comlete movements fom the classical music eetoie (see sulementay table S1). The entie movements wee esented to study music listening unde ecologically valid conditions. To avoid the otential influence of lyics, only instumental ieces wee chosen. The duation of each iece was between 9 and 14 minutes. All ieces wee in the s, which allowed segmenting the music in a well-defined manne. The numbe of vaiations vaied between the ieces (Mendelssohn: 18 segments, Pokofiev: 8 segments, Schubet: 14 segments). Futhemoe the ieces diffeed a geat deal in the style of comosition, the comose, and instumentation. These diffeences wee chosen in ode to avoid a biased measue focusing only on one aticula musical style and theeby allowing moe eliable infeences to be dawn about geneal emotional qualities shaed by these woks Exeimental design of fmri exeiment Fo the fmri exeiment, aticiants wee instucted and familiaized with the task io to scanning. They had to listen attentively to the music thoughout the esentation of each iece, while keeing thei eyes oen and diected towad the middle of a dimly-lit sceen. Duing this time, they wee asked to let thei emotions natually flow with the music, while constaining ovet bodily movements. At the end of evey musical stimulus, aticiants answeed a shot questionnaie asking fo thei global subjective evaluation of the eviously head music. A seies of 6 questions wee esented on the sceen, one afte the othe, immediately afte the end of the music excet. Paticiants had to ate (1) how much attention they aid to the iece, (2) how stongly they eacted emotionally, and (3) how much the envionment influenced them thoughout the esentation of the eviously esented iece. In addition they also had to ate (4) how much they liked the iece, and (5) how familia they wee with it.

7 Finally, fo each iece, the aticiants had to evaluate thei subjective emotional exeience along the nine GEMS dimensions of the Geneva Emotion Music Scale (Zentne et al., 2008), including wonde, joy, owe, tendeness, eacefulness, nostalgia, tension, sadness, and tanscendence. All evaluations I I had to be made on a hoizontal 7-oint scale anging fom 0 (not at all) to 6 (absolutely). Duing the esentation of the questionnaies, image acquisition by the scanne stoed and thee was no time limit fo answeing. Answeing all questions took less than 2 minutes on aveage. Duing music listening, the sceen was gey and luminosity inside the scanne oom was ket low and constant. Visual instuctions wee esented on a sceen back-ojected on a headcoil-mounted mio. Musical stimuli wee esented binaually with a high-quality MRI-comatible headhone system (CONFON HP-SC 01 and DAP-cente mkii, MR confon GmbH, Gemany). Paticiants used a MRI comatible esonse button box (HH-1 4- CR, Cuent Designs Inc., USA) to give thei answes. The fmri exeiment consisted of 3 scanning uns and ended with the acquisition of the stuctual MRI scan. The esentation ode of the 3 musical ieces was andomized between aticiants. Pio to scanning, the loudness of the music was adjusted fo each aticiant individually to ensue that both the loudest and softest moments of the musical woks could be comfotably head. FMRI data acquisition and eocessing MRI images wee acquied using a 3T whole body MRI scanne (Tio TIM, Siemens, Gemany) togethe with a 12 channel head coil. A high-esolution T1-weighted stuctual image (1 x 1 x 1 mm 3 ) was obtained using a magnetization-eaed aid acquisition gadient echo sequence (eetition time [TR] = 1.9 s, echo time [TE] = 2.27 ms, time to invesion [TI] = 900 ms). Functional images wee obtained using the following aametes: 35 slices, slice thickness 3.2 mm, slice ga 20%, TR = 2 s, TE = 30 ms, field of view = 205 x 205 mm 2, 64 x 64 matix, fli angle: 80. FMRI data was analyzed using Statistical Paametic Maing (SPM8; Wellcome Tust Cente fo Imaging, London, UK; htt:// Data ocessing included ealignment, unwaing, slice timing, nomalization to the Monteal Neuological Institute sace using an EPI temlate (esamling voxel size: 2 x 2 x 2 mm), and satial smoothing (5 mm full-width at half-maximum Gaussian Filte). FMRI data analysis Inte-subject coelations (ISCs) in the time-couse of the eocessed data wee calculated using the inte-subject coelations toolbox (isc-toolbox, elease 1.1) develoed by (Kaui et al., 2010, Kaui et al., 2014). Each iece was fist analyzed seaately ove its entie time-couse to obtain global ISC mas of bain aeas geneally activated duing music listening. Dynamic ISCs wee also calculated using a sliding time-windows with a bin size of ten scans and a ste size of one scan. Significance thesholds wee detemined by non-aametic emutation tests fo the global mas and fo the time-window analysis fo each iece individually. The dynamic time-window based ISCs wee used to identify eak moments of high inte-subject ageement in fmri activity ove the whole bain. These eaks wee defined such that at least 10 voxels o 10 ecent of the voxels in a secific egion of inteest (ROI)

8 would suvive the calculated statistical theshold fo the time-window ISC. Subsequently, the magnitude of fmri activity duing the coesonding time-window was extacted fom these voxels, aveaged ove all aticiants, and submitted to futhe coelation analyses with the emotional atings and with the acoustic musical featues identified fo the coesonding time-window (second-ode coelation). Given a window size of 10 scans, 10 time oints wee thus taken into account fo the coelation. Fo these coelation analyses, a delay of 4 seconds was assumed in ode to take into account the delay of the BOLD signal. Fo each ROI, the obtained coelation coefficients wee Fishes-z-tansfomed and tested fo significance using one-samle t-tests. The anatomical definition of ROIs was based on the HavadOxfod atlas (htt:// We selected 29 egions including auditoy, moto, and efontal aeas, as well as subcotical egions, accoding to ou main hyotheses about music ecetion and emotional ocessing (see sulementay table S4 and S5 fo the comlete list of ROIs). Fo all ROIs, the left and ight side wee analyzed seaately. ROIs wee ket fo futhe analyses only when thee wee moe than 5 time-windows that suvived the statistical theshold acoss the thee ieces. Musical featue analysis The thee music ieces wee analyzed in tems of acoustic featues using the MIR toolbox (Latillot and Toiviainen, 2007) imlemented in Matlab. We selected diffeent acoustic featues descibing seveal timbal and hythmic chaacteistics of music (see sulementay table S2). These featues included two long-tem chaacteistics (ulse claity and event density) that catue stuctual and context-deendent asects of hythmic vaiations (Latillot et al., 2008), as well as six shot-tem chaacteistics that concen loudness, timbe and fequency content (ms, dissonance, zeo-cossing, sectal entoy, sectal centoid, bightness). The loudness of the music is detemined by the oot mean squae (ms) of the signal. Zeo-cossing and entoy ae measues indicating the noisiness of the acoustic signal, wheeas centoid descibes the cental tendency of the sectum of the signal. Bightness eesents the amount of high-fequency enegy contained in the signal. Fo long-tem featues, the music was analyzed in time-windows of 3 seconds that wee shifted in 1 second stes. Shot-tem featues wee measued in successive windows of 50 ms with shifts of 25 ms. These aametes wee chosen in accodance with evious ublications using a simila methodology fo musical featue analysis (Allui et al., 2011). In addition, the absolute temo of ieces was extacted using built-in functions of the softwae Melodyne (vesion 3.2, Celemony Softwae, GmbH, München, Gemany, efined to catue beat-by-beat temo fluctuations. All musical featues wee consequently down-samled to 0.5 Hz. Given that some featues wee coelated with each othe, a facto analysis was conducted in ode to comae the diffeent tyes of featues and identify ossible highe-ode factos coesonding to coheent categoies of inte-elated featues. Exeimental design of behavioal exeiment

9 The behavioal study was conducted in a simila manne as the fmri exeiment, using identical musical stimuli as fo the fmri exeiment. The main diffeence was that instead of assive listening, the aticiants had now to efom an active continuous evaluation task. The task consisted of dynamic judgments of subjectively felt emotions in esonse to the music ove time. Paticiants wee instucted to use an in-lab ogammed inteface (Adobe Flash CS3 Pofessional, ActionScit 3 in Flash, htt:// which allowed them to indicate the level of thei affective exeience by moving the comute mouse vetically along an intensity axis (see sulementay figue 2). The evaluation was dislayed continuously on the comute sceen. The intensity esonses wee ecoded at a fequency of 4 Hz. The exeiment was conducted in two diffeent vesions. One gou of aticiants (n=14) was asked to evaluate thei subjective - -excite. A second gou of aticiants (n=17) had to evaluate thei affective state along the valence dimension, going fom - - T ose aticiants who aticiated in the fmri exeiment wee econtacted and asked to efom the valence ating task 3 months afte scanning, wheeas a new gou of diffeent aticiants was invited to efom the aousal ating task. Both gous of aticiants wee instucted to listen attentively to the music at the same time as they intosectively monitoed and judged the felt level of valence o aousal evoked in them by the music. It was exlicitly ointed out that they should not evaluate what they thought that the comosito o musicians wanted to exess. In addition, as in the fmri exeiment, each musical iece was immediately followed by a shot questionnaie of the same 6 questions concening thei global feelings as exeienced oveall duing the whole duation of the iece. Paticiants in the behavioal exeiment wee comfotably installed in font of the comute and the musical stimuli wee esented via high-quality headhones (Sennheise, HD280 o). Ratings wee made by moving (with the ight hand) the mouse u o down elative to a neutal baseline esented on the sceen. Note that obtaining valence and aousal atings fom diffeent gous allowed us to take into account individual vaiability in the subjective leasantness of the diffeent music ieces, while minimizing habituation in aousal measues due to eeated exosue to the same stimuli. Pesenting the same ieces thee times fo diffeent atings to the same gou of aticiants would be likely to bias the data by changing the emotional effects oduced by the music when it is head seveal times. The entie duation of the exeiment took aound 45 min. Thee was a taining tial fist, to adjust the volume and get familiaized with the task and the inteface. Paticiants could eeat the taining seveal times if necessay. We used the tems 'enegy' and 'aletness' as definition of the aousal dimension in ode to be cleae fo naive aticiants and ensue that they undestood well the natue of this dimension (see sulementay figue 2). Behavioal data analysis To analyze the ating data of this exeiment, the continuous time-couses of valence and aousal values wee z- tansfomed and aveaged in each of the two gous. To assess the vaiability of evaluations between aticiants,

10 we calculated inte-subject coelations (ISC) between all ais of subjects and aveaged e subject the coelation coefficients with all othe aticiants, afte tansfoming the coefficients using the Fishes z-tansfomation. Data fom the shot questionnaies at the end of each block wee used to comae whethe the emotional ecetion of the musical ieces was simila between gous in the behavioal and fmri exeiments, and whethe the emotional exeience of the fmri gou would be comaable afte thee months. We used a mixed ANOVA with context (inside/outside of scanne) as categoical vaiable and deending on the analysis iece o GEMS as single within facto. These statistical analyses wee done in ode to test if all the gous of the aticiants did the exeiment in simila conditions and had simila emotional eactions to the music. RESULTS Behavioal esults The global behavioal atings given at the end of each music iece wee comaed between the diffeent gous of aticiants and the diffeent exeimental contexts (inside o outside the scanne). Paticiants in both context conditions eoted a simila amount of emotional eactions to the esentation of the thee ieces and eoted simila familiaity with these ieces (table 1a, sulementay figue 1). Howeve, aticiants in the scanne eoted to be moe distacted by the envionment and to be slightly less attentive to the music than the aticiants outside of the scanne, as could be exected due to the MRI envionment and noise. Howeve, inside the scanne, aticiants evaluated all ieces as moe leasant than the aticiants outside. This shows that desite the distaction due to the scanne, aticiants wee able to aeciate the music. Thee wee no diffeences between the ieces concening the attentional level and envionmental distaction. Howeve, the amount of the emotional eactions, as well as atings of leasantness and familiaity diffeed between the ieces. The Schubet iece was evaluated as moe leasant, moe familia and elicited moe emotional eactions than eithe of the othe ieces (table 1a, sulementay figue 1). None of these atings showed any significant inteaction between the context condition and the musical ieces (table 1a). Likewise, emotion atings along the 9 categoies of the GEMS (Zentne et al., 2008) did not vay between the diffeent context conditions fo the ieces of Pokofiev and Schubet, but thee was an inteaction of context by categoy fo the iece of Mendelssohn eflecting the fact that aticiants in the fmri context evaluated this iece as owe elative to othe contexts (table 1b, sulementay figue 1). Because this iece ends with a oweful finale, this might have been oduced by a ecency effect that was attenuated in behavioal exeiment conditions whee aticiants had aleady ovided atings on othe dimensions fo the duation of the wok. The continuous judgments wee analyzed fo consistency acoss aticiants fo each music iece (table 2, figue 1). Fo aousal, aticiants evaluated the ieces vey similaly (aveage Conbach alha = 0.967). Fo valence, thee was a geate vaiability between aticiants (aveage Conbach alha = 0.600), in keeing with lage individual diffeences in hedonic aaisals (Blood and Zatoe, 2001, Salimoo et al., 2009). Nevetheless,

11 the high eliability of continuous judgments fo aousal and global atings above was sufficient to allow a diect comaison of behavioal esults obtained outside the scanne with the neuoimaging data obtained in the fmri exeiment gou. Moeove, as shown in figue 1, continuous atings fo aousal and valence did not follow a simila time couse and thus aeaed not to shae a lage amount of vaiance. Theefoe, these evaluations wee not othogonalized fo futhe analyses. Musical featues The facto analysis on all nine acoustic featues extacted with the MIR toolbox (see Sulementay Table S2) showed that two factos ae sufficient to descibe 79.2% of the vaiance among these musical featues. The facto loadings ointed to thee distinct families of featues that can be goued togethe along these two main dimensions. One subgou contains moe timbe elated featues (zeo-cossing, centoid, bightness, entoy), wheeas anothe gou subsumes event elated o enegy dominated featues (ms, event density, dissonance and temo). Pulse claity seems to constitute a unique featue distinct fom both subgous above. Adding the emotional atings into a simila analysis showed a vitually identical factoial segegation, with two simila dimensions sufficient to exlain 72.2% of the vaiance. This second analysis futhe indicated that aousal atings goued togethe with the event/enegy-elated musical featues, wheeas valence aeaed to constitute a seaate categoy close to ulse claity (see figue 2). Additional multile egession analyses with the musical featues and the emotional dimensions showed that aousal and valence atings could be edicted by the 9 musical featues (see sulementay table S3). Fo aousal, all featues excet centoid significantly contibuted to the ediction, wheeas only the thee featues ms, ulse claity and entoy showed a significant elation to valence. FMRI esults Global ISCs ove the entie time-couse of music ieces showed that activity in auditoy egions (HG, STG, PPt, PPo), as well as distibuted netwoks in seveal cotical and subcotical stuctues, coelated significantly acoss aticiants. Figue 3 shows the thee global ISC mas calculated fo each musical iece at a FDR coected significance theshold. As can be seen, ISC mas wee highly simila fo all 3 ieces. Besides auditoy egions, eaks of synchony acoss the listenes aose bilateally in occiital visual aeas, osteio aietal cotex, emoto aeas, ceebellum, medial efontal and obitofontal cotex, insula, caudate and vental stiatum. Coelations in the amygdala wee only found fo the ieces by Mendelssohn and Schubet. These esults demonstate that BOLD activity in these bain egions was diven in a individuals when listening to the same music ieces. Fo the dynamic ISC analysis with moving time-windows, a FDR coected significance theshold fo coelation values fom each window was defined fo each iece seaately. As non-aametic esamling methods wee used, the significance theshold diffes between ieces (at =0.05, ISC values () with FDR coection

12 wee consideed as significant when =0.18 fo Mendelssohn; =0.16 fo Pokofiev; =0.16 fo Schubet (Pajula et al., 2012). Fo time windows in which the ISC suvived this theshold, the aveaged BOLD signal time-couse was extacted fom each ROI and was then coelated with the time-couse of emotion atings, as well as with acoustic indices comuted fo each music iece (second-ode coelation). The etansfomed coelation coefficients and the coesonding -values fom one-samle t-tests ae eoted fo the emotion atings (aousal and valence) in sulementay table S4, and fo the musical featues in sulementay table 5. Fo this analysis, an abitay theshold of =0.001 was adoted that allowed etaining the most eliable coelations between BOLD signal fom synchonized bain egions and eithe acoustic o emotional dimensions of music, while minimizing false ositives due to multile testing. These analyses evealed significant second-ode coelations fo the two emotional dimensions that wee evaluated continuously (aousal and valence), as well as fo all 9 musical featues that wee detemined by MIR analysis. These esults ae descibed in the next sections and summaized gahically in figues 4-8. Emotional exeience We fist tested fo second-ode coelations of bain activity (in synchonized bain egions) with the degee of emotional aousal and valence duing each music iece, as obtained fom the continuous ating ocedue. Fo aousal, significant ositive coelations wee found in seveal bilateal auditoy aeas along STG, including lanum temoale and olae. But also moe citically, seveal subcotical egions coelated ositively with aousal, notably the left insula, left amygdala, left thalamus, and ight caudate nucleus, as well as the ight fontal oecula egions (figue 4). Fo valence, mainly negative coelations wee found. These involved only a few seconday auditoy egions on the left side, togethe with negative coelations in bilateal fontal oeculum, left ACC, and left ostcental gyus. Moeove, negative coelations wee also found fo subcotical egions including the left amygdala and ight caudate. The only ositive coelation with valence was found in the ight SMA (see figue 4 and sulementay table S5). Thus, moe leasant moments in the music wee associated with tansient eductions in key aeas of the limbic system at both cotical (ACC) and subcotical (amygdala and caudate) levels; wheeas convesely subjective inceases in aousal wee associated with inceases in atly ovelaing limbic bain aeas. Acoustic featues As shown by the facto analysis above, the 9 diffeent musical featues wee goued into a few distinct clustes of co-occuence when the entie time-couse of the ieces was taken into account. The coelation analysis of bain activity (in synchonized aeas) fo each of these acoustic featues confimed a simila gouing at the anatomical level but also evealed slight diffeences. Bain aeas esonding to these acoustic featues duing moments of inte-subject synchony ae descibed below.

13 Fo the featues centoid, bightness, and zeo-cossing, which descibe the noisiness and homogeneity of the sound, significant ositive coelations duing ISC eaks wee found in shaed bain egions including bilateal auditoy aeas, bilateal infeio fontal gyus, as well as sueio and middle left fontal gyus, lus left anteio cingulate cotex (see figue 5 and sulementay table S5). Only zeo-cossing also coelated with the left sueio fontal sulcus. Additional ositive coelations wee found fo these featues in subcotical egions including thalamus and bilateal utamen, while the left nucleus accumbens showed a negative coelation with the same featues (see figue 4 and sulementay table S4). Thus, activity in these egions tended to incease as a function of highe noisiness in music, excet fo the nucleus accumbens that showed a eduction instead. A second gou of featues showing a common atten of coelation duing inte-subject synchony was obseved fo entoy and event density. Besides auditoy egions, these featues coelated ositively with activity in left e-cental gyus, ight IFG as oeculais, left insula, thalamus (event density only ight) and utamen (event density only left). In addition, fo entoy, significant coelations wee also found in left ACC and ight ecuneus (see figue 6 and sulementay table S5), indicating that these egions wee secifically sensitive to sound comlexity. No negative coelation was obseved. Patly simila to the evious gou, ms and dissonance aeaed howeve to constitute a diffeent subset of featues, which we descibe as enegy elated. Besides eaks in auditoy and oecula fontal egions, we found ositive coelations in bilateal thalamus, ight ecuneus, and left insula. Futhemoe, ms also showed selective ositive coelations in the ight utamen and ight IFG as oeculais, wheeas dissonance was uniquely elated to the ight SMA. Convesely, negative coelations with these featues wee found in the left amygdala and left nucleus accumbens (see figue 7 and sulementay table S5). Thus, the loude and the moe dissonant the music, the moe activity in these limbic subcotical stuctues deceased when they synchonized between aticiants, but the moe the activity inceased in auditoy aeas, emoto cicuits, and insula. Finally, the emaining two featues, temo and ulse claity, showed slightly diffeent coelation attens as comaed with othe featues (see figue 8 and sulementay table S5). Notably, less coelation was obseved in imay auditoy aeas. Instead, temo significantly coelated with activity in bilateal IFG and ight OFC. In subcotical egions, significant coelations occued in left thalamus (fo temo only) and in utamen (fo ulse claity only left side), but none elated to the limbic system. DISCUSSION In the esent study, we investigated music ecetion and in aticula musical emotion ocessing using intesubject coelations (ISC) of bain activity, and then elated the latte to the time-couse of diffeent musical featues. We obseved significant synchony acoss listenes in distibuted bain netwoks including not only auditoy cotices (as exected given exosue to simila auditoy inuts) but also, moe emakably, seveal aeas imlicated in visual, moto, attention, o affective ocesses (see figue 3). In aticula, inte-subject synchony occued in key egions associated with ewad and subjective feeling states such as the vental stiatum and insula.

14 Synchony in the amygdala was only found fo the ieces of Mendelssohn and Schubet, which wee the ieces which wee moe liked and induced stonge emotional eactions (sulementay figue 1). In addition, we found that bain activation duing synchony eiods coelated with the esence of secific acoustic featues in the music, with diffeent featues being associated with inceases o deceases in diffeent egions. Again, these effects involved widesead aeas beyond auditoy cotex, including cotical and sub-cotical limbic stuctues such as the insula and amygdala, the basal ganglia (caudate and utamen), efontal aeas, as well as moto and emoto cotices. Moeove, these diffeent musical featues and the coesonding bain esonses aeaed to constitute diffeent clustes, based on thei time-couse (acoss the diffeent music ieces) as well as accoding to thei elation with subjective emotional judgments of aousal and leasantness. The thee main gous of musical featues wee timbe-elated (centoid, bightness, zeo-cossing), enegy- o event- elated (entoy, event density, ms, dissonance), and hythm-elated attibutes (temo, ulse claity). Aousal effects showed activation attens esembling mostly the enegy- and hythm-elated featues, wheeas leasantness shaed activations with both aousal and timbe featues. These esults accod with the fact that emotional eactions to music ae esumably not elicited by single acoustical chaacteistics, but athe by a combination of diffeent musical featues. Effects of emotional valence and aousal duing ISC The main goal of ou study was to investigate whethe inte-subject synchony in bain activity elicited by music would involve egions associated with emotions. In addition, we sought to detemine whethe synchony in secific bain egions would eflect aticula acoustic featues associated with musical emotions. Although seveal limbic egions and the basal ganglia wee found to coelate with the subjective continuous atings of aousal and valence, we did not obseve a significant elation between activity of the nucleus accumbens (a egion citically imlicated in ewad and leasue; (Beidge and Robinson, 1998, Salimoo et al., 2013) and subjective eots of leasantness. This contasts with evious studies eoting activation of nucleus accumbens duing eaks of leasuable music (Blood and Zatoe, 2001, Menon and Levitin, 2005), but othe studies on musical leasantness have also failed to find consistent activation of the vental stiatum (Koelsch et al., 2006). That we did not find ositive coelations between activity of the nucleus accumbens and valence might be due to ou exeimental design, which imaily tested fo coheent bain activity between aticiants ove time. Howeve, musical leasantness eesents a vey subjective exeience and it is ossible that ou aticiants did not always exeience ositive emotions at the exact same time but showed moe idiosyncatic esonses at vaious times, which would educe the owe of ISC measues. Nevetheless, as futhe discussed below, we found that nucleus accumbens coelated negatively with timbe- and enegy-elated featues that eflect the acoustic fequency distibution, noisiness, and dissonance of music (see figue 5 and 7). In addition, the ight caudate, anothe key comonent of the basal ganglia, showed a selective ositive coelation with affective judgments of aousal but negative coelation with valence. Remakably, unlike the

15 accumbens, this stuctue was not modulated by any of the acoustic featues, suggesting that it does not esond to low-level featues but athe to highe-level affective oeties of the music. Pevious studies eoted that the caudate is moe activated duing hay than neutal music (Mitteschiffthale et al., 2007), o found esonses to both aousal and leasantness (Colibazzi et al., 2010). Howeve, some studies also eoted an involvement of the caudate fo negative emotional exeiences (Caetie et al., 2009), o fo the avoidance of ositive emotions (Hae et al., 2005). Hee, we found the caudate showed oosite coelations fo aousal (ositive) and valence (negative). This finding conveges with a ecent study (Tost et al., 2012) whee we showed that the ight caudate is not only sensitive to highly aousing leasant emotions (e.g. owe) induced by music, elative to low aousing leasantness (e.g. tendeness), but also activates duing highly aousing emotions with a negative valence that ae chaacteized by subjective feelings, that is, an emotion conveying a negative affect state of (Zentne et al., 2008, Tost et al., 2012). Othe studies indicate that the caudate is ecuited duing the anticiation of leasant moments in the music (Salimoo et al., 2011) and mediates the automatic entainment of moto o visuomoto ocesses by musical hythms (Gahn et al., 2008, Tost et al., in ess). Taken togethe, these data suggest that activity in the ight caudate might jointly incease between aticiants in moments associated with the anticiation of secific musical events, such as those associated with a aticula emhasis engaging attention of the listenes and coesonding to highe aousal, but inducing eithe ositive o negative emotions deending on othe concomitant aametes in the musical iece. This would accod with the ole of the ight caudate in encoding exectations o edictions of sensoy o moto events in othe domains (Balleine et al., 2007, Asaad and Eskanda, 2011), as well as its imlication in hythmic ecetion and entainment (Kokal et al., 2011, Tost et al., in ess). We found a simila elation between inte-subject synchony and emotional effects in the left amygdala, with a ositive coelation with aousal but negative coelation with valence judgments. This aousal effect accods with evious findings that the amygdala activates to moments in the music that ae novel, aleting, and associated with stuctual changes (Fische et al., 2003, James et al., 2008, Koelsch et al., 2013), as well as to moe intense stimuli in othe sensoy modalities (Andeson et al., 2003, Winston et al., 2005). The valence effect also conveges with esults showing that amygdala signals coelate negatively with the intensity of music-induced chills (Blood and Zatoe, 2001). Moeove, the amygdala shows geate esonse to unleasant comaed to leasant music (Koelsch et al., 2006) contents (Gosselin et al., 2005, Gosselin et al., 2007). Inteestingly, ou acoustic analysis demonstated that synchonized activity in the left amygdala was negatively elated to the dissonance and ms featues of music, which also coelated negatively with nucleus accumbens (see above), a atten dovetailing with the study of Koelsch and colleagues (2006) who used dissonance to induce unleasantness. Kuma and colleagues (2012) also suggested that the amygdala is diffeentially engaged by acoustical featues associated with emotional unleasantness. It is notewothy that even though aousal is often closely linked with the loudness of the music, ou esults highlight a clea dissociation between these two factos, as we obseved a ositive coelation of amygdala activity with aousal but a negative coelation with ms. This

16 finding indicates that the amygdala does not esond to the highe ms of music in a simila way as to loud events that may be aousing and behavioally salient (Sande et al., 2003). Thus, amygdala eactivity to ms and dissonance seems to eflect the emotional aaisal of the stimulus, athe than meely its acoustic enegy. The only ositive coelation fo valence atings found in ou study concened the ight SMA. As this egion is a key comonent of moto initiation and sequence lanning (Tankus et al., 2009), ou finding might indicate that musically induced leasantness engendes moto action tendencies like the uge to dance o move with the music (Zatoe and Halen, 2005). Moeove, induced mith and smiling have been associated with activity in the SMA (Iwase et al., 2002). Othe studies of music ecetion have obseved a ositive coelation between activation of SMA and the intensity of eceived chills (Blood and Zatoe, 2001). Howeve, we also obseved a selective coelation of ight SMA with dissonance, which is athe associated with unleasantness (Koelsch et al., 2006). It is ossible that moto activation in this case might eflect a covet eaction of avoidance (Sagase et al., 2011), but futhe eseach is needed to claify the significance of SMA inceases duing music ocessing and emotions. Finally, aat fom stong coelations with activity in sensoy auditoy aeas and ight oecula efontal egions, aousal also imlicated the left insula and thalamus. These effects aalleled those of seveal acoustic featues, aticulaly enegy and event density elated featues (i.e., dissonance, ms, o entoy) that also stongly modulate insula activity. As these featues coesond to moe intense auditoy signals, they ae likely to contibute to enhance subjective aousal as eflected by such inceases in sensoy athways and insula. These data accod with studies indicating that the insula lay an imotant ole in ocessing highly aousing and motivating stimuli (Bentson et al., 2011). As ou egession analysis of behavioal and acoustic data showed that the aousal values could be well edicted by the set of musical featues (excet fom centoid), it is not suising to find simila esults also fo the ISC analyses. One egion also showing common coelations with diffeent factos was the ight oecula efontal cotex, which showed synchony with inceasing aousal as well as with all musical featues excet enegy-elated dimensions. This involvement of the fontal oeculum might eflect a disciminative and analytic function of this egion in auditoy ocessing in geneal (Fuhholz and Gandjean, 2013). Inteestingly, a negative coelation in this egion was found fo valence, which convesely might suggest a decease of analytical heaing in moments of highe leasantness. Effects of timbe-elated featues on bain activity and ISC Besides uncoveing inte-subject synchony in bain activity elated to emotional exeiences evoked by music, ou study also identified neual attens associated with the ocessing of secific music featues. These featues modulated seveal egions in auditoy cotices as well as highe-level cotical aeas and subcotical aeas. Fo timbe elated featues, which descibe the sectal quality of auditoy signals, we found diffeential activations in left infeio and sueio fontal gyi, in addition to auditoy cotex and sub-cotical egions. The left IFG B aleady been shown to be involved in timbe moe than loudness ocessing fo simle tones (Reitee et al., 2008). Activity in these lateal efontal aeas might eflect abstact categoization

17 and temoal segmentation ocesses associated with ecetion and aesthetic evaluation of the musical sound quality. Moeove, timbe-elated featues also coelated negatively with the left nucleus accumbens, a key at of the ewad system influenced by doamine and usually imlicated in high leasantness (Menon and Levitin, 2005, Salimoo et al., 2011). As timbe featues measued hee descibe the noisiness and shaness of sounds (see sulementay table S2), thei imact on accumbens activity might eflect a elative unleasantness of the coesonding musical events (i.e. indexed by elevated zeo-cossing, high bightness and enlaged centoid). Effects of event and enegy-elated featues Fo ISC diven by event- and enegy elated musical featues (entoy, event density, ms, and dissonance), aat fom auditoy aeas and thalamus, common inceases wee obseved in the left insula fo all fou featues and in the ight ecuneus fo all featues excet event density. While insula activity seemed elated to aousal (see above), ecuneus activity may coesond to heightened attentional focusing and absotion (Cavanna and Timble, 2006). In suot of the latte inteetation, music studies have eoted that the ecuneus is involved in the selective ocessing of itch (Platel et al., 1997), diecting attention to a secific voice in hamonic music (Satoh et al., 2001), and shifting attention fom one modality to anothe (e.g. fom audition to vision; (Tost et al., in ess). Moeove, in a evious study we found the ecuneus to be selectively esonsive to music inducing feelings of t (Tost et al., 2012), an affective state that also heightens attention. These fou event and enegy-elated musical featues could howeve be futhe distinguished. Entoy and event density coelated ositively with the left ecental gyus, which might encode action tendencies engendeed by these two featues, such as stonge moto tonicity associated with felt tension (Tost et al., 2012), o instead stonge evocation of covet hythmic movements due to inceased event density in the music (Lahav et al., 2007). On the othe hand, ms and dissonance oduced distinctive negative coelations with the left amygdala and left nucleus accumbens, eflecting a diect influence of these two featues on emotion elicitation (Gabielson and Juslin, 2003), as discussed above. Effects of hythm-elated featues Both temo and ulse claity oduced a distinct atten, unlike othe musical featues. Fist, these featues did not affect bain egions whose activity was associated with emotional valence o aousal atings. Second, only a few coelations wee found in auditoy aeas, none of which aose in the imay auditoy cotex (excet on the left side fo ulse claity). This suggests that these featues constitute highe-level oeties, athe than low-level acoustic cues. The lack of effects in the ight imay auditoy cotex might also oint to a lowe temoal esolution elative to the left side (Zatoe and Belin, 2001). In addition, both, temo and ulse claity ae featues that do not change quickly, but athe vay ove lage time scales o stewise. In contast, ou ISC measues likely emhasized moe tansient activations. These factos might also account fo the aucity of coelations with fmri signal changes.

18 Notably, howeve, both temo and ulse claity modulated activity in the utamen, the otion of basal ganglia most closely linked to moto ocesses. Although the utamen was also modulated by othe music featues, it did not coelate with aousal and valence atings. The basal ganglia lay an imotant ole in timing tasks and ecetion of temoal egulaity (Buhusi and Meck, 2005, Geise et al., 2012). Moeove, the utamen is essentially involved in sensoimoto coodination and beat egulaity detection, wheeas the caudate is moe imlicated in moto lanning, anticiation, and cognitive contol (Gahn et al., 2008). In ou study, only caudate activity coelated with emotion exeience (see above), but not utamen. Advantages and limitations of ou method The goal of this study was to identify synchonized bain ocesses between aticiants duing music listening. We then also sought to detemine which emotional dimensions o musical featues could dive this synchonization. Howeve, ou aoach did not aim at, and thus is not caable of, identifying moe globally those bain stuctues that esond to aticula emotional o acoustical featues. Futue studies coelating bain activity with the timecouse of these featues may ovide additional infomation on the latte question (Chain et al., 2010, Allui et al., 2011, Nummenmaa et al., 2012). Hee, in addition to ISC, we used continuous judgments of valence and aousal, which eesent a useful and sensitive tool to study emotional eactions ove time (Schubet, 2010). Continuous judgments have been emloyed in othe neuoimaging studies, allowing eseaches to coelate the entie timecouse of these judgments with bain activity (Chain et al., 2010, Allui et al., 2011, Abams et al., 2013). Howeve, such analyses allow only the identification of bain egions that would tack a secific asect of the music to the same degee fo the entie duation of the stimulus. This method is valid fo constantly changing aametes such as low-level musical featues. Howeve, fo emotional eactions, which instead occu eisodically (but do not fluctuate continuously), these analyses seem less aoiate. Moeove, in standad GLM analyses, a chaacteistic linea esonse of the BOLD signal is assumed to eflect changes in stimulus featues. Howeve, in the case of music, meaningful events ae moe difficult to seaate and may combine non-linealy to oduce ISC. Theefoe ou aoach ovides a useful tool to tack emotional eactions and othe neual ocesses esonding to musical content indeendent of edefined effects on the BOLD signal. Hee, by coelating the fmri signal with eithe emotion atings o musical featues in a esticted time window that coesonded to tansient synchony of bain activity acoss all aticiants, ou measue of ISC should be sensitive to even bief and ae signal changes. On the othe hand, this aoach is less sensitive to slowe vaiations and sustained states. Moeove, the size of the time window used fo such analysis is citical. We chose a window of 10 scans (coesonding to 20 sec in the music), which was a eliable comomise between sufficient statistical owe to efom coelation analyses and adequate musical secificity without adoting a too long time eiod. This comomise was chosen to ensue a sufficient temoal esolution and to allow coelation analyses with a minimum of time oints. Howeve, moe eseach is needed to validate this method moe systematically, also with diffeent window sizes. Futhe methodological wok would also be equied to comae

19 activations identified by ISC and taditional GLM analysis based on the time-couse of elevant featues. It would also be inteesting to exloe ISC and temoal dynamics of moe extended bain netwoks athe than single bain aeas, and elate them to secific emotional esonses. Futue studies might also use ISC and evese coelation techniques to identify moe comlex elements in music stuctues that edict synchonized esonses and emotional exeiences. In ou study, an infomal exloation evealed only atial coesondences between ISC eaks and theme tansitions within each music ieces (see Sul. Figue 3), but moe advanced musicology measues might ovide innovative ways to elate ISC in aticula bain egions o netwoks to well-defined musical oeties. In any case, ou esults aleady demonstate fo the fist time that ISC measues could successfully identify secific attens of bain activity and synchony among individuals, and then link these attens with distinct emotion dimensions and aticula music featues. Conclusion By using a novel aoach based on ISC duing music listening, we wee able to study the neual substates and musical featues undelying tansient musical emotions without any a ioi assumtions about thei temoal onset o duation. Ou esults show that synchony of bain activity among listenes eflects basic dimensions of emotions evoked by music such as aousal and valence, as well as aticula acoustic aametes. ISC concomitant with emotional aousal was accomanied by inceases in auditoy aeas and limbic aeas including the amygdala, caudate, and insula. On the othe hand, ISC associated with highe leasantness judgments was coelated with educed activity in atly simila limbic aeas, including the amygdala, caudate, and anteio cingulate cotex, but with inceased activity in emoto aeas (SMA). Most of these effects wee concomitant with vaiations in enegyelated featues of the music such as ms and dissonance. Futhe, ou esults eveal a dissociation in the left amygdala esonses fo the ocessing of ms and aousal (inducing decease vs inceased activity, esectively), which challenges the notion that sensoy intensity and subjective aousal ae always tightly linked (Andeson et al., 2003). Rathe, amygdala esonses imaily eflect the affective significance of incoming stimuli which may be modulated by thei intensity (Winston et al., 2005). In addition, we found that the nucleus accumbens (a subcotical limbic egion cucially imlicated in ewad and leasue) showed no diect modulation by aousal o valence atings, but a selective negative coelation with both enegy and timbe-elated featues that descibe the noisiness and sectal content of music. This atten of activity in accumbens dovetails well with the fact that euhonic timbe and consonant hamony ae tyically associated with leasuable music and ositive emotions activating vental stiatum and doaminegic athways (Salimoo et al., 2011, Tost et al., 2012). Oveall, these data oint to a majo ole fo seveal subcotical stuctues (in amygdala and basal ganglia) in shaing emotional esonses induced by music. Howeve, seveal cotical aeas wee also involved in ocessing music featues not diectly associated with musical emotions. Besides auditoy aeas ecuited by a ange of featues, the left doso-lateal and infeio efontal cotex was selectively sensitive to sound quality and timbe,

20 wheeas the insula and ecuneus esonded to both enegy and event-elated featues that tend to enhance aletness and attention. Finally, hythmical featues diffeed fom the othes by thei lack of coelation with the ight imay auditoy cotex, but edominant effects on infeio efontal aeas and subcotical moto cicuits (utamen). Taken togethe, these findings highlight that new methodological aoaches using ISC can be usefully alied on vaious musical ieces and styles in ode to bette undestand the ole of secific bain egions in ocessing music infomation and thei elation to secific emotional o cognitive esonses. Acknowledgements: The authos would like to thank Dimiti Van De Ville and Jukka Kaui fo thei suot with the inte-subject coelation analyses and Claie Senelonge fo heling with the behavioal data acquisition. This wok was suoted by a PhD fellowshi awaded to WT by the Lemanic Neuoscience Doctoal School, an awad fom the Geneva Academic Society (Foemane Fund) to PV, and the Swiss National Cente fo Affective Sciences (NCCR SNF No. 51NF ). REFERENCES: Abams DA, Ryali S, Chen T, Chodia P, Khouzam A, Levitin DJ, Menon V Inte-subject synchonization of bain esonses duing natual music listening. Eu J Neuosci 37: Allui V, Toiviainen P, Jaaskelainen IP, Gleean E, Sams M, Battico E Lage-scale bain netwoks emege fom dynamic ocessing of musical timbe, key and hythm. Neuoimage. Andeson AK, Chistoff K, Staen I, Panitz D, Ghahemani DG, Glove G, Gabieli JD, Sobel N Dissociated neual eesentations of intensity and valence in human olfaction. Nat Neuosci 6: Asaad WF, Eskanda EN Encoding of Both Positive and Negative Rewad Pediction Eos by Neuons of the Pimate Lateal Pefontal Cotex and Caudate Nucleus. Jounal of Neuoscience 31: Balleine BW, Delgado MR, Hikosaka O The ole of the dosal stiatum in ewad and decision-making. Jounal of Neuoscience 27: Bentson GG, Noman GJ, Bechaa A, Buss J, Tanel D, Cacioo JT The insula and evaluative ocesses. Psychol Sci 22: Beidge KC, Robinson TE What is the ole of doamine in ewad: hedonic imact, ewad leaning, o incentive salience? Bain Res Bain Res Rev 28: Blood AJ, Zatoe RJ Intensely leasuable esonses to music coelate with activity in bain egions imlicated in ewad and emotion. Poc Natl Acad Sci U S A 98: Blood AJ, Zatoe RJ, Bemudez P, Evans AC Emotional esonses to leasant and unleasant music coelate with activity in aalimbic bain egions. Nat Neuosci 2: Buhusi CV, Meck WH What makes us tick? Functional and neual mechanisms of inteval timing. Natue eviews Neuoscience 6: Caetie L, Rios M, de la Gandaa BS, Taia M, Albet J, Loez-Matin S, Alvaez-Linea J The stiatum beyond ewad: caudate esonds intensely to unleasant ictues. Neuoscience 164: Chain H, Jantzen K, Kelso JA, Steinbeg F, Lage E Dynamic emotional and neual esonses to music deend on efomance exession and listene exeience. PLoS One 5:e Cochane T A simulation theoy of musical exessivity. The Austalasian Jounal of Philosohy 88: Colibazzi T, Posne J, Wang Z, Goman D, Gebe A, Yu S, Zhu H, Kangalu A, Duan Y, Russell JA, Peteson BS Neual systems subseving valence and aousal duing the exeience of induced emotions. Emotion 10: Etzel JA, Johnsen EL, Dickeson J, Tanel D, Adolhs R Cadiovascula and esiatoy esonses duing musical mood induction. Int J Psychohysiol 61:57-69.

21 Fische H, Wight CI, Whalen PJ, McIneney SC, Shin LM, Rauch SL Bain habituation duing eeated exosue to feaful and neutal faces: A functional MRI study. Bain Reseach Bulletin 59: Fuhholz S, Ceavolo L, Gandjean D Secific bain netwoks duing exlicit and imlicit decoding of emotional osody. Ceeb Cotex 22: Fuhholz S, Gandjean D Pocessing of emotional vocalizations in bilateal infeio fontal cotex. Neuosci Biobehav Rev 37: Fuman O, Dofman N, Hasson U, Davachi L, Dudai Y They saw a movie: long-tem memoy fo an extended audiovisual naative. Lean Mem 14: Gabielson A, Juslin PN Emotional exession in music. In: Handbook of Affective Sciences (Davidson, R. J. et al., eds), New Yok: Oxfod Univesity Pess. Geise E, Notte M, Gabieli JD A Coticostiatal Neual System Enhances Auditoy Pecetion though Temoal Context Pocessing. The Jounal of neuoscience : the official jounal of the Society fo Neuoscience 32: Gomez P, Danuse B Relationshis between musical stuctue and sychohysiological measues of emotion. Emotion 7: Gosselin N, Peetz I, Johnsen E, Adolhs R Amygdala damage imais emotion ecognition fom music. Neuosychologia 45: Gosselin N, Peetz I, Noulhiane M, Hasboun D, Beckett C, Baulac M, Samson S Imaied ecognition of scay music following unilateal temoal lobe excision. Bain 128: Gahn JA, Pakinson JA, Owen AM The cognitive functions of the caudate nucleus. Pog Neuobiol 86: Geen AC, Baeentsen KB, Stodkilde-Jogensen H, Wallentin M, Roestoff A, Vuust P Music in mino activates limbic stuctues: a elationshi with dissonance? Neuoeot 19: Guhn M, Hamm A, Zentne M Physiological and musico-acoustic coelates of the chill esonse. Music Pecet 24: Hae TA, Tottenham N, Davidson MC, Glove GH, Casey BJ Contibutions of amygdala and stiatal activity in emotion egulation. Biol Psychiat 57: Hasson U, Fuman O, Clak D, Dudai Y, Davachi L. 2008a. Enhanced intesubject coelations duing movie viewing coelate with successful eisodic encoding. Neuon 57: Hasson U, Malach R, Heege DJ Reliability of cotical activity duing natual stimulation. Tends in cognitive sciences 14: Hasson U, Ni Y, Levy I, Fuhmann G, Malach R Intesubject synchonization of cotical activity duing natual vision. Science 303: Hasson U, Yang E, Vallines I, Heege DJ, Rubin N. 2008b. A hieachy of temoal ecetive windows in human cotex. J Neuosci 28: Hevne K Exeimental studies of the elements of exession in music. Am J Psychol 48: Iwase M, Ouchi Y, Okada H, Yokoyama C, Nobezawa S, Yoshikawa E, Tsukada H, Takeda M, Yamashita K, Yamaguti K, Kuatsune H, Shimizu A, Watanabe Y Neual substates of human facial exession of leasant emotion induced by comic films: a PET Study. NeuoImage 17: Jaaskelainen IP, Koskentalo K, Balk MH, Autti T, Kauamaki J, Pomen C, Sams M Inte-subject synchonization of efontal cotex hemodynamic activity duing natual viewing. Oen Neuoimag J 2: James CE, Bitz J, Vuilleumie P, Hauet CA, Michel CM Ealy neuonal esonses in ight limbic stuctues mediate hamony inconguity ocessing in musical exets. Neuoimage 42: Juslin PN Fom eveyday emotions to aesthetic emotions: towads a unified theoy of musical emotions. Phys Life Rev 10: Juslin PN, Laukka P Imoving emotional communication in music efomance though cognitive feedback. Musicae Scientiae 4: Kanwishe N, McDemott J, Chun MM The fusifom face aea: a module in human extastiate cotex secialized fo face ecetion. The Jounal of neuoscience : the official jounal of the Society fo Neuoscience 17: Kaui JP, Jaaskelainen IP, Sams M, Tohka J Inte-subject coelation of bain hemodynamic esonses duing watching a movie: localization in sace and fequency. Font Neuoinfom 4:5.

22 Kaui JP, Pajula J, Tohka J A vesatile softwae ackage fo inte-subject coelation based analyses of fmri. Font Neuoinfom 8:2. Koelsch S Towad a neual basis of music ecetion - a eview and udated model. Font Psychol 2:110. Koelsch S Stiking a chod in the bain: neuohysiological coelates of music-evoked ositive emotions. In: The Emotional Powe of Music (Cochane, T. et al., eds), Oxfod: Oxfod Univesity Pess. Koelsch S Bain coelates of music-evoked emotions. Nat Rev Neuosci 15: Koelsch S, Fitz T, V. Camon, Mulle K, Fiedeici AD Investigating emotion with music: an fmri study. Hum Bain Ma 27: Koelsch S, Rohmeie M, Toecuso R, Jentschke S Pocessing of hieachical syntactic stuctue in music. Poc Natl Acad Sci U S A 110: Koelsch S, Skouas S Functional centality of amygdala, stiatum and hyothalamus in a "Small-Wold" netwok undelying joy: an fmri study with music. Hum Bain Ma. Kokal I, Engel A, Kischne S, Keyses C Synchonized dumming enhances activity in the caudate and facilitates osocial commitment--if the hythm comes easily. PLoS One 6:e Kumhansl CL An exloatoy study of musical emotions and sychohysiology. Can J Ex Psychol 51: Kuma S, von Kiegstein K, Fiston K, Giffiths TD Featues vesus feelings: dissociable eesentations of the acoustic featues and valence of avesive sounds. J Neuosci 32: Lahav A, Saltzman E, Schlaug G Action eesentation of sound: audiomoto ecognition netwok while listening to newly acquied actions. J Neuosci 27: Latillot O, Eeola T, Toiviainen P, Fonai J 2008 Multi-featue modeling of ulse claity: Design, calidation, and otimization. In: 9th Intenational Confeence on Music Infomation Retieval (Bello, J. P. et al., eds), Philadelhia, PA. Latillot O, Toiviainen P 2007 A Matlab toolbox fo musical featue extaction fom audio. In: Intenational Confeence on Digital Audio Effects (DAFx-07) Bodeaux, Fance. Menon V, Levitin DJ The ewads of music listening: esonse and hysiological connectivity of the mesolimbic system. Neuoimage 28: Mitteschiffthale MT, Fu CH, Dalton JA, Andew CM, Williams SC A functional MRI study of hay and sad affective states induced by classical music. Hum Bain Ma 28: Nummenmaa L, Gleean E, Viinikainen M, Jaaskelainen IP, Hai R, Sams M Emotions omote social inteaction by synchonizing bain activity acoss individuals. Poc Natl Acad Sci U S A 109: Pajula J, Kaui JP, Tohka J Inte-subject coelation in fmri: method validation against stimulus-model based analysis. PLoS One 7:e Pankse J, Benatzky G Emotional sounds and the bain: the neuo-affective foundations of musical aeciation. Behav Pocess 60: Platel H, Pice C, Baon JC, Wise R, Lambet J, Fackowiak RSJ, Lechevalie B, Eustache F The stuctual comonents of music ecetion - A functional anatomical study. Bain 120: Reitee S, Eb M, Godd W, Wildgube D Ceebal Pocessing of Timbe and Loudness: fmri Evidence fo a Contibution of Boca's Aea to Basic Auditoy Discimination. Bain Imaging Behav 2:1-10. Sagase P, Schwatz S, Vuilleumie P Fea and sto: a ole fo the amygdala in moto inhibition by emotional signals. Neuoimage 55: Salimoo VN, Benovoy M, Lache K, Daghe A, Zatoe RJ Anatomically distinct doamine elease duing anticiation and exeience of eak emotion to music. Nat Neuosci 14: Salimoo VN, Benovoy M, Longo G, Cooestock JR, Zatoe RJ The ewading asects of music listening ae elated to degee of emotional aousal. PLoS One 4:e7487. Salimoo VN, van den Bosch I, Kovacevic N, McIntosh AR, Daghe A, Zatoe RJ Inteactions between the nucleus accumbens and auditoy cotices edict music ewad value. Science 340: Sande D, Gafman J, Zalla T The human amygdala: An evolved system fo elevance detection. Reviews in the Neuosciences 14: Satoh M, Takeda K, Nagata K, Hatazawa J, Kuzuhaa S Activated bain egions in musicians duing an ensemble: a PET study. Bain Res Cogn Bain Res 12: Schee KR Which emotions can be induced by music? What ae the undelying mechanisms? And how can we measue them? Jounal of New Music Reseach 33:

23 Schee KR, Oshinsky JS Cue utilization in emotion attibution fom auditoy stimuli. Motiv Emotion 1: Schime A, Kotz SA Beyond the ight hemishee: bain mechanisms mediating vocal emotional ocessing. Tends in cognitive sciences 10: Schubet E Modeling eceived emotion with continuous musical featues. Music Pecet 21: Schubet E Continuous self-eot methods. In: Handbook of Music and Emotion: Theoy, Reseach, Alications (Juslin, P. N. and Sloboda, J. A., eds) Oxfod (U.K.): Oxfod Univesity Pess. Tankus A, Yeshuun Y, Flash T, Fied I Encoding of seed and diection of movement in the human sulementay moto aea. Jounal of neuosugey 110: Tost W, Ethofe T, Zentne M, Vuilleumie P Maing aesthetic musical emotions in the bain. Ceeb Cotex 22: Tost W, Fuhholz S, Schon D, Labbé C, Pichon S, Gandjean D, Vuilleumie P. in ess. Getting the beat: entainment of bain activity by musical hythm and leasantness. Wilson SM, Molna-Szakacs I, Iacoboni M Beyond sueio temoal cotex: intesubject coelations in naative seech comehension. Ceeb Cotex 18: Winston JS, Gottfied JA, Kilne JM, Dolan RJ Integated neual eesentations of odo intensity and affective valence in human amygdala. J Neuosci 25: Zatoe RJ, Belin P Sectal and temoal ocessing in human auditoy cotex. Ceeb Cotex 11: Zatoe RJ, Belin P, Penhune VB Stuctue and function of auditoy cotex: music and seech. Tends Cogn Sci 6: Zatoe RJ, Halen AR Mental concets: musical imagey and auditoy cotex. Neuon 47:9-12. Zentne M, Gandjean D, Schee KR Emotions evoked by the sound of music: chaacteization, classification, and measuement. Emotion 8: Sulementay figue legends Sulementay figue 2: Schematic view of the inteface used fo the behavioal emotion evaluation fo the emotion dimension valence. Fo U level of enegy induced in yo B feeling of high alet, excitement even o a fenzied state. A low level of enegy on the othe hand, means a feeling of aeasement, being mellow, and sleeiness. Sulementay figue 3: Moments in time with significant ISC eaks fo each of the thee ieces. The black line shows aveage ISC values, The black line deicts the standadized numbe of egions (fom 0 = none to 1 = **) showing a significant ISC eak at each time oint; the blue line eesents the standadized loudness of the music. The vetical bas mak tansitions between diffeent thematic vaiations in the music. As can be seen, ISC eaks did not systematically coesond to tansitions no did they follow volume. Sulementay figue 4: Illustation of ou analysis method using a sliding time window to extact the aveaged BOLD signal in the left amygdala (blue lines) and the aveaged valence atings (ed lines) along continuous time oints duing which the ISC values wee above the significance theshold. The coelation coefficients calculated between the two signals wee then aveaged acoss all significant time windows and esulted in values eoted as eoted in the main manuscit.

24 Sulementay table S1: Musical stimuli Mendelssohn Pokofiev Schubet Comose Piece Recoding Inteetes Instuments Duation Felix Vaiations Peahia lays Muay Peahia Piano solo 11:32 Mendelssohn- séieuses, o.54, Mendelssohn, Batholdy (on a theme in D Muay Peahia ( ) mino) Label: Sony Classical, 1984 CBS Inc. ASIN: B0014KABVI DDD Segei Pokofiev ( ) Fanz Schubet ( ) Piano Conceto No. 3 in C majo, O. 26, 2 nd movement: Tema con vaiazioni (E mino) Sting quatet No. 14 in D mino T 2 nd movement: Andante con moto (G mino) Pokofiev: Piano Concetos 1-5 Recoded in Studio at Concetgebouw Hall (Amstedam), Label: Chandos 8889 DDD Schubet: The late sting quatets. Emeson Sting Quatet, Studio ecoding 1988, Label: DEUTSCHE GRAMMOPHON ASIN: B0001ZWGI8 DDD Hoacio Gutieez (Piano), Neeme Javi (conducto), Royal Concetgebouw Ochesta Emeson Sting Quatet Ochesta with iano Sting quatet 8:58 14:12

25 Sulementay table S2: Descition of musical featues Featue Function in MIRtoolbox Descition fom the (Latillot and Toiviainen, 2007) MIRtoolbox manual Bightness mibightness Measues the high fequency enegy above 1500 Hz contained in the signal Centoid micentoid Descition of the shae (sectal centoid) of the sectum, gives a measue of cental tendency of the sectum Dissonance mioughness Measues the sensoy dissonance o oughness, ceated by the atio of fequencies contained in a signal Entoy mientoy Measues the Shannon entoy (Shannon, 1948) of the sectum and descibes the flatness of the sectum Event density mieventdensity Indicates the aveage fequency of events, i.e. the numbe of onsets e second Pulse claity miluseclaity Indicates the stength of beats, by detecting beats and calculating thei egulaity Root-mean-squae (RMS) Temo mims Evaluated in Melodyne (Melodyne studio vesion 3.2, Celemony Softwae, GmbH, München, Gemany, Descibes the global enegy of the signal Beats ae detected and counted in a secific time unit. Measued in beats e minute (BPM) Zeo-cossing mizeocoss Counts the numbe of times the signal changes the sign Effect Bightness is an indicato of the shaness of a sound Centoid gives a measue of homogeneity of the sound Indicates the dissonance and the suosed unleasantness of a sound A flat sectum with low entoy eesents a noisie signal Event density indicates how many note onsets haen in one time unit Pulse claity indicates the how clealy the hythm including the mete is detectable Rms is an index of loudness of the sound Temo is an indicato how fast the music is efomed Zeo-cossing is an indicato of noisiness of the sound

26 Sulementay table S3: Multile egession analyses fo the emotional dimensions with the musical featues a) Multile egession fo Aousal: R = 0.86, R^2=0.74, F(9,1031)=324.68, < b* -value Intecet zeocossing dissonance ms entoy bightness ulseclaity temo centoid eventdensity b) Multile egession fo Valence: R = 0.27, R^2=0. 07, F(9,1031)=8.94, < b* -value Intecet zeocossing dissonance ms entoy bightness ulseclaity temo centoid eventdensity Numbes in bold eesent significant values

27 Sulementay table S4: Coelation analysis fo emotion evaluations aousal valence nb of eaks ACC_L ACC_R Amy_L Amy_R Caudate_L Caudate_R FMC_L FMC_R Heschl_L Heschl_R HPC_L HPC_R IFGo_L IFGo_R IFGt_L IFGt_R 0 no eaks Insula_L Insula_R MFG_L MFG_R Nac_L Nac_R OFC_L OFC_R OPc_L OPc_R OPf_L OPf_R OP_L OP_R Pallidum_L no eaks 0 Pallidum_R PaaHPCant_L no eaks 0 PaaHPCant_R no eaks 0 PaaHPCost_L no eaks 0 PaaHPCost_R no eaks 0 PCC_L PCC_R PostCG_L

28 PostCG_R PP_L PP_R PeCG_L PeCG_R Pecuneus_L Pecuneus_R PT_L PT_R Putamen_L Putamen_R SCC_L no eaks 0 SCC_R no eaks 0 SFG_L SFG_R SMA_L SMA_R STGant_L STGant_R STGost_L STGost_R Thalamus_L Thalamus_R Significant coelations above the theshold <0.001 ae maked in bold.

29 Sulementay table S5: Coelation analysis fo musical featues zeocossing cen toid bight ness entoy ms disso nance event density ulse claity temo ACC_L ACC_R Amy_L Amy_R Caudate_L Caudate_R FMC_L FMC_R Heschl_L Heschl_R HPC_L HPC_R

30 zeocossing cen toid bight ness entoy ms disso nance event density ulse claity temo IFGo_L IFGo_R IFGt_L IFGt_R no eaks Insula_L Insula_R MFG_L MFG_R Nac_L Nac_R OFC_L OFC_R

31 zeocossing cen toid bight ness entoy ms disso nance event density ulse claity temo OPf_L OPf_R Pallidum_L no eaks Pallidum_R PaaHPCant_L no eaks PaaHPCant_R no eaks PaaHPCost_L no eaks PaaHPCost_R no eaks PCC_L

32 zeocossing cen toid bight ness entoy ms disso nance event density ulse claity temo PCC_R PostCG_L PostCG_R PP_L PP_R PeCG_L PeCG_R Pecuneus_L Pecuneus_R PT_L PT_R

33 zeocossing cen toid bight ness entoy ms disso nance event density ulse claity temo Putamen_L Putamen_R SCC_L no eaks SCC_R no eaks SFG_L SFG_R SMA_L SMA_R STGant_L STGant_R STGost_L

34 Significant coelations above the theshold <0.001 ae maked in bold. Thalamus_R Thalamus_L STGost_R zeocossing cen toid bight ness entoy ms disso nance event density ulse claity temo

35 aveaged z-scoe of evaluations Mendelssohn time Aousal Valence aveaged z-scoe of evaluations Pokofiev time Aousal Valence aveaged z-scoe of evaluations Schubet time Aousal Valence Figue 1: Aveaged evaluation scoes ove time fo continuous emotional judgments of valence and aousal). The x-axis dislays the time dimension samled at 4 Hz and the y-axis shows the continuous evaluation scoes (between 0 and 300). At the stat osition, the cuso was fixed to 150. These numbes wee not visible to the aticiants. The gay shaded aeas eesent 95% confidence intevals.

36 Figue 2: Factoial analysis fo all musical featues combined with emotion atings.

37 Figue 3: A) Global ISC mas ovelaid on the mean stuctual bain image, combined fo all thee music ieces togethe. Significance theshold: <0.05 (FDR). B) ISC mas fo each music iece seaately. Significance theshold: <0.05 (FDR), coesonding to a minimum theshold of =0.08 and a maximum of =0.36. In all cases, synchony involved not only lage aeas in auditoy cotex, but also visual aeas, sueio aietal aeas, efontal aeas, emoto cotex, ceebellum, and vental stiatum.

38 Figue 4: Schematic coelation ma fo subjective emotional atings, aousal (A) and valence (B). White lettes indicate ositive coelations, black lettes indicate negative coelations. Abbeviations: acc: anteio cingulate, hg: gyus, ins: insula, nac: nucleus accumbens, of: fontal oeculum, og: ostcental gyus, : lanum olae, t: lanum temoale, sma: sulementay moto aea, stg: sueio temoal gyus (anteio and osteio, esectively), thal: thalamus. (C) Aveaged coelation coefficients ae dislayed fo the egions which exceeded the significance theshold of =0.001.

39 Figue 5: (A) Schematic coelation mas fo timbe elated featues (centoid, bightness and zeo cossing). Blue oints indicate coelations common to all thee featues. Light blue oints indicate coelation fo zeo-cossing only, dak blue oints indicate coelations fo all excet fo zeo-cossing. White lettes indicate ositive coelations, black lettes indicate negative coelations. Abbeviations: acc: anteio cingulate, hg: gyus, mfg: middle fontal gyus, nac: nucleus accumbens, oe: infeio fontal gyus, as oeculais, of: fontal oeculum, : lanum olae, t: lanum temoale, ut: utamen, sfg: sueio fontal gyus, stg: sueio temoal gyus (anteio and osteio, esectively), thal: thalamus, ti: infeio fontal gyus, as tiangulais. (B) Aveaged coelation coefficients ae dislayed fo the egions which exceeded the significance theshold of =0.001.

40 Figue 6: (A) Schematic coelation mas fo event elated featues (entoy and event density). Geen oints indicate coelations common to both featues. Light geen oints indicate coelation fo event density only and dak geen oints fo entoy only. Abbeviations: acc: anteio cingulate, hg: gyus, ins: insula, oe: infeio fontal gyus, as oeculais, of: fontal oeculum, c: ecuneus, : lanum olae, g: ecental gyus, t: lanum temoale, ut: utamen, stg: sueio temoal gyus (anteio and osteio, esectively), thal: thalamus. (B) Aveaged coelation coefficients ae dislayed fo the egions which exceeded the significance theshold of =0.001.

41 Figue 7: (A) Schematic coelation ma fo enegy elated featues (ms and dissonance). Geen oints indicate coelations common to both featues. Light geen oints indicate coelation fo dissonance only, dak geen oints fo ms only. White lettes indicate ositive coelations, black lettes indicate negative coelations. Abbeviations: amy: amygdala, hg: gyus, ins: insula, mfg: middle fontal gyus, nac: nucleus accumbens, oe: infeio fontal gyus, as oeculais, of: fontal oeculum, c: ecuneus, : lanum olae, t: lanum temoale, ut: utamen, sfg: sueio fontal gyus, sma: sulementay moto aea, stg: sueio temoal gyus (anteio and osteio, esectively), thal: thalamus, ti: infeio fontal gyus, as tiangulais. (B) Aveaged coelation coefficients ae dislayed fo the egions which exceeded the significance theshold of =0.001.

42 Figue 7: Schematic coelation ma fo hythm elated featues, temo (A) and ulse claity (B). Abbeviations: hg: gyus, oe: infeio fontal gyus, as oeculais, of: fontal oeculum, : lanum olae, t: lanum temoale, ut: utamen, stg: sueio temoal gyus (anteio and osteio, esectively), thal: thalamus, ti: infeio fontal gyus, as tiangulais. (C) Aveaged coelation coefficients ae dislayed fo the egions which exceeded the significance theshold of =0.001.