Cross-Cultural Music Phrase Processing:

Transcription

1 Human Bain Mapping 29: (2008) Coss-Cultual Music Phase Pocessing: An fmri Study Yun Nan, 1,2 Thomas R. Knösche, 1,3 * Stefan Zysset, 1 and Angela D. Fiedeici 1 1 Max Planck Institute fo Human Cognitive and Bain Sciences, Leipzig, Gemany 2 State Key Laboatoy of Cognitive Neuoscience and Leaning, Beijing Nomal Univesity, Beijing, China 3 Key Laboatoy of Mental Health, Institute of Psychology, Chinese Academy of Sciences, Beijing, China Abstact: The cuent study used functional magnetic esonance imaging (fmri) to investigate the neual basis of musical phase bounday pocessing duing the peception of music fom native and nonnative cultues. Geman musicians pefomed a cultual categoization task while listening to phased Westen (native) and Chinese (non-native) musical excepts as well as modified vesions of these, whee the impession of phasing has been educed by emoving the phase bounday making pause (hencefoth called unphased ). Bilateal planum tempoale was found to be associated with an inceased difficulty of identifying phase boundaies in unphased Westen melodies. A netwok involving fontal and paietal egions showed inceased activation fo the phased condition with the obital pat of left infeio fontal gyus pesumably eflecting woking memoy aspects of the tempoal integation between phases, and the middle fontal gyus and intapaietal sulcus pobably eflecting attention pocesses. Aeas moe active in the cultually familia, native (Westen) condition included, in addition to the left planum tempoale and ight vento-medial pefontal cotex, mainly the bilateal moto egions. These latte esults ae intepeted in light of sensoimoto integation. Regions with inceased signal fo the unfamilia, non-native music style (Chinese) included a ight latealized netwok of angula gyus and the middle fontal gyus, possibly eflecting highe demands on attention systems, and the ight posteio insula suggesting highe loads on basic auditoy pocessing. Hum Bain Mapp 29: , VC 2007 Wiley-Liss, Inc. Key wods: music peception; phase stuctue; coss-cultual compaison; fmri INTRODUCTION Auditoy steams, be they speech o music, ae stuctued into phases, theeby facilitating thei peception and futhe pocessing by the human bain. Fo the peception Contact gant sponso: Deutsche Foschungsgemeinschaft; Contact gant numbe: KN 588/1-1 *Coespondence to: Thomas R. Knösche, Max-Planck-Institute fo Human Cognitive and Bain Sciences, Stephanstasse 1a, Leipzig, Gemany. knoesche@cbs.mpg.de Received fo publication 13 May 2006; Revision 9 Febuay 2007; Accepted 18 Febuay 2007 DOI: /hbm Published online 11 May 2007 in Wiley InteScience (www. intescience.wiley.com). of phases, the coect detection of phase boundaies is necessay. Such boundaies ae indicated by specific cues. In the domain of music, these cues include small pauses, lengthening of the phase final note, and the implicit tonal function of the phase limiting tone [see e.g., Riemann, 1900]. While the most elevant of these cues is clealy the pause [Fankland and Cohen, 2004; Neuhaus et al., 2006; Riemann, 1900], the impotance of othe cues has been evidenced by seveal behavioal studies [see e.g., Boltz, 1993; Clake, 1985; Jusczyk and Kumhansl, 1993]. Sensitivity to phasing and phase boundaies can aleady be obseved in young infants, in the domains of both speech [Hish- Pasek et al., 1987] and music [Jusczyk and Kumhansl, 1993; Kumhansl and Jusczyk, 1990]. Recently, EEG and MEG coelates fo the peception of phase boundaies in music have been epoted [Knösche VC 2007 Wiley-Liss, Inc.

2 Coss-Cultual Music Phase Pocessing et al., 2005; Neuhaus et al., 2006]. In EEG, we identified a cento-paietal positive wave appoximately 500 ms afte the onset of the fist tone following a phase bounday. It beas some esemblance to the Closue Positive Shift (CPS), which indexes the pocessing of intonational phase boundaies in speech [Pannekamp et al., 2005; Steinhaue et al., 1999]. It is theefoe called music CPS and has been shown to be sensitive to a numbe of phase bounday cues: pause length, length of the last tone peceding the pause, and hamonic function of this last tone [Neuhaus et al., 2006]. The same study demonstated a elationship between the music CPS and the individual s fomal musical taining [Neuhaus et al., 2006]. Bain Netwoks Undelying Music Pocessing In contast to the behavioal and neuophysiological evidence mentioned above, thee is vey little knowledge on the neual netwoks undelying the pocessing of musical phasing. Othe aspects of music pocessing have eceived consideably moe attention egading thei neuonal substate by means of fmri, MEG souce localization, and lesion studies. In geneal, melody peception is epoted to ecuit the supeio tempoal [Keutz et al., 2003; Moison et al., 2003] and middle tempoal gyi [Platel et al., 2003] as well as the fontal lobe [Benal et al., 2004; Giffiths et al., 1998; Giffiths, 2003]. In addition, paietal, vento-lateal pefontal, and pemoto aeas [Halpen and Zatoe, 1999; Langheim et al., 2002; Zatoe et al., 1996] have also been evidenced to take a ole in the pocess. The neual coelates fo hamony pocessing seem to compise a netwok including anteio cingulate cotex (ACC), tempoal poles, paietal and occipital lobes, and the medial suface of the ceebellum bilateally [Moison et al., 2003; Satoh et al., 2001; Schmithost, 2005]. The pocessing of musical syntax pimaily involves left BA 44 (Boca s aea) and its ight hemisphee homologue, as shown by MEG [Maess et al., 2001] and fmri [Koelsch et al., 2002]. fmri eseach investigating tempoal coheence eveals the impotance of BA 47 fo both linguistic [Dapetto and Bookheime, 1999; Ni et al., 2000] and musical pocessing [Levitin and Menon, 2003, 2005]. Anothe much discussed aspect concening the bain substates of music pocessing is hemispheical asymmety. Despite the suggestions fom some bain-lesion and neuoimaging studies that music might engage the ight hemisphee moe than the left [Benal et al., 2004; Halpen et al., 2004; Maess et al., 2001; Samson and Zatoe, 1994; Tamo and Bhaucha, 1991; Zatoe et al., 1994], thee ae othe studies that have shown a bilateally distibuted netwok fo music pocessing [Keutz et al., 2003; Meiste et al., 2004; Platel et al., 2003; Tillmann et al., 2003]. To ou knowledge, thee is no evidence suggesting diffeent bain loci fo pocessing music of cultually familia and unfamilia styles, as suggested by the only published fmri study on coss-cultual music peception [Moison et al., 2003]. Objectives and Hypotheses in This Study The goal of the pesent study is to specify: (1) the neual substates of musical phase bounday pocessing, (2) the possible diffeences in neual netwoks undelying the pocessing of cultually familia and unfamilia music in geneal, and (3) the inteaction between 1 and 2, asking the questions of whethe and how the pocessing of musical phase boundaies might be influenced by the cultual familiaity of the music. Bain netwoks of musical phasing Although thee is, to ou knowledge, only vey little published bain imaging wok concened with the neual substates undelying the pocessing of phases in music o in speech posody, ERP and MEG studies [Knösche et al., 2005; Steinhaue et al., 1999] suggest that thee should be consideable activation of neuonal assemblies in esponse to phase boundaies. Moeove, data fom studies on elated topics might allow the establishment of some hypotheses on the bain netwoks involved in the pocessing of phase boundaies. These ae studies that investigate the extaction of featues fom the acoustic input and those investigating integative pocessing of sequential events. A fist possible candidate aea fo detecting phase bounday cues and ecognizing the pesence of a phase bounday is the planum tempoale (PT). The PT contains the auditoy association cotex, which is known to play a ole in highe ode pocesses of speech [Jäncke et al., 2002a] as well as pocesses that ae common to speech and nonspeech input [Binde et al., 1996, 2000]. A ecent study on the peception of nomal speech, degaded speech accompanied with nomal intonation, and flattened speech lacking the intonational contou [Meye et al., 2004] led to the conclusion that the PT might suppot the integation of apidly and slowly changing infomation. Accoding to this view, the involvement of the PT in the pocessing of phase boundaies in music as well as in speech would be vey likely, since the integation of local phase bounday makes with global infomation, like tonality and pitch contou, must be pefomed. Activation of the PT has also been obseved in a numbe of studies on music peception associated with the ecognition of musical featues, like timbe, contou, key, inteval, and hythm [e.g. Liégeois- Chauvel et al., 1998; Menon et al., 2002; Ohnishi et al., 2001]. It could, theefoe, be pedicted that PT exhibits inceased activity duing the pocessing of phase boundaies, suppoting the integation of slow (global) infomation, which pedicts the phase bounday and apid (local) infomation, which maks it. If, howeve, some of the phase bounday makes ae emoved (e.g., the pause; see Methods section), thee would be a conflict between both types of infomation and the identification of the phase boundaies would equie moe effot. This could then be eflected by a futhe incease in PT activity. Indeed, 313

3 Nan et al. inceased activity of the ight PT associated with flattened speech (without posody) as compaed to nomal speech has been epoted ecently [Fiedeici and Alte, 2004; Meye et al., 2004]. The pocessing of integational sequencing (timing and odinality of events) involves stuctues of the so-called moto cicuit, that is, medial and lateal pemoto cotex, fontal opeculum, stiatum, and the intapaietal sulcus [Schubotz et al., 2000]. Within this netwok, the opecula pat of the infeio fontal gyus (IFG) has not only been associated with the pocessing of sequential ules in language syntax [Fiedeici, 2002], posody [Meye et al., 2004], and music syntax [Maess et al., 2001], but also with the peception of visually pesented motion hythms [Schubotz and von Camon, 2001] and imagey of motion [Binkofski et al., 2000]. The obital pat of IFG has been linked to the pocessing of tempoal stuctue in music [Levitin and Menon, 2003; Vuust et al., 2006]. Moe geneally, the IFG seems to be involved in vaious pocesses, equiing the integation of infomation ove time. Hence, it appeas that inceased activity in vaious subegions of IFG as well as othe constituents of the moto cicuit has been obseved, when the stuctue of sequences had to be pocessed. Since phase bounday makes add to the amount of stuctue in a musical sequence, it would be easonable to assume that the peception of phased melodies causes inceased activity in these egions. In summay, we hypothesize that thee will be inceased activation of planum tempoale, when the phase bounday making pause has been emoved. Moeove, inceased activity in constituents of the moto cicuit and, in paticula, of IFG is likely to be elated to phase bounday pocessing. Cultual dependence of bain activation due to music pocessing Thee has aleady been one attempt to isolate diffeences between pocessing native and non-native music [Moison et al., 2003], which yielded a null esult. Thee ae two majo methodological concens with this study, which we stived to avoid in the cuent wok. Fist, it has been acknowledged that thee is geat vaiability in individuals bain esponses to music fo people with diffeent musicality levels [Aydin et al., 2005; Koelsch et al., 2005; Moison et al., 2003; Patel et al., 1998; Satoh et al., 2001, 2003; Seung et al., 2005; Shahin et al., 2003], with o without the pedisposition of absolute pitch [Banea et al., 1994; Bemudez and Zatoe, 2005; Cumme et al., 1994; Itoh et al., 2005; Ohnishi et al., 2001; Peetz and Zatoe, 2005; Schlaug et al., 1995], and of diffeent gende [Johnson et al., 1996; Koelsch et al., 2003a,b; Lee et al., 2003]. In ode to specify the bain loci fo cetain functions by the fmri methodology, a elatively stict selection of subjects is equied in ode to educe intesubject vaiance. In the cuent study, only faily young Geman female musicians without absolute pitch wee ecuited as paticipants. Secondly, elatively long musical excepts pose challenges to the esticted measuing time of an fmri study. Vey small tial numbes (e.g. 3 tials as in the study of Moison et al. [2003]) might pevent the detection of effects [Huettel and McCathy, 2001]. By using as many tials as possible (20 tials fo each condition), the cuent study attempts to obtain a BOLD signal aveaging output of good quality. Although, as indicated above, thee seems to be no published wok epoting bain aeas associated with the cultual familiaity of music, studies on the ecognition of familiaity in geneal could help to geneate hypotheses, if familiaity effects ae domain geneal. Shah et al. [2001] identified the etosplenial cotex as being associated with the pocessing of both familia voices and faces in contast to unfamilia ones. The authos intepet thei findings such that this pat of the limbic system is involved in the ecognition of familiaity. Howeve, it is not clea whethe the activation of the etosplenial cotex elates to the detection of familiaity in geneal, o to the ecognition of a paticula voice/face, that is, to the identification of the paticula peson, o to emotional esponses to the peson pesonally known to the peceive. Wheele and Buckne [2004] as well as Iidaka et al. [2006] studied cotical netwoks associated with consciously emembeing wods pesented about one day befoe. Iidaka et al. [2006] identified a bilateal netwok involving the middle fontal gyus and infeio paietal lobule. The activity of the ight paietal egion in esponse to a epeated item was modulated by the epetition lag, leading the authos to the conclusion that this aea would be citical fo familiaity-based judgment [see also Yonelinas et al., 2005]. Fom these studies, we might hypothesize that ight paietal cotex nea the infeio paietal lobule and possibly the left paietal lobe and etosplenial cotex might be also associated with familiaity mediated by cultual musical style. An expeiment tageting the familiaity of paticula pieces of music showed activations in the left obital infeio fontal gyus and the supeio tempoal gyus [Platel et al., 1997]. In a late study of the same goup [Platel et al., 2003], semantic memoy of melodies was associated with activation in medial and obital fontal lobe, left anteio middle tempoal gyus (extending into the obital pat of the infeio fontal gyus) as well as the left angula gyus. It theefoe seems that semantic memoy of paticula music pieces is associated with a mainly left hemispheic fonto-tempoo-paietal netwok. This cognitive function, howeve, may diffe fom the moe geneal familiaity based on a cetain cultual style. Since in musicians, familia music is likely to be epesented not only in the sensoy, but also in the moto domain [Banget and Altenmülle, 2003; Haueisen and Knösche, 2001], listening to cultually familia (though not individually known) melodies might activate both moto and auditoy bain aeas. This might hold in paticula, if the music belongs to the subjects own pefomance aea (see, e.g., the compaison between pianists and chous singes listening to piano music, studied by Haueisen and 314

4 Coss-Cultual Music Phase Pocessing Knösche [2001]). It would, theefoe, be easonable to expect geate activation in these aeas fo music pieces that ae cultually familia to the listene, all the moe if the listenes ae musicians and the cultually familia music belongs to thei own aea of pefomance. Apat fom the cultual familiaity issue as such, we wanted to exploe whethe thee ae diffeences in the bain netwoks undelying phase bounday peception as a function of cultual familiaity. Fom a pevious ERP study [Nan et al., 2006], it seems that the pocesses of listening to cultually familia and unfamilia music ae quite simila, with some modulation at ealy latencies, but not of the CPS. Hence, one goal of this study was to find out whethe and to what extent this elative cultual univesality also holds fo the undelying neual netwoks. The above fomulated questions wee addessed by pesenting melodies of two diffeent cultual styles (Chinese and Westen), each in two vesions (phased/unphased; see Methods section fo details), to a goup of Geman musicians. METHODS Paticipants Twenty Geman female musicians with nomal heaing abilities wee paid fo paticipation in this study. They did not have symptoms of any neuological, psychiatic, o intenal disease, no had they taken any medication fo at least 3 days pio to the examinations. Because of thei poo behavioal pefomance (hit ate lowe than 80% fo Westen music), two subjects wee excluded fom the final analysis (as suggested by the epeated Gubb s outlie test at the level of P < 1%). The emaining 18 paticipants wee ight-handed (aveage handedness: ) accoding to the Edinbugh Handedness Inventoy [Oldfield, 1971]. None of them was an absolute pitch possesso accoding to the Absolute Pitch (AP) test (see below). The mean age was yeas (age ange 19 28). The aveage stating age fo instument playing was yeas. Fouteen out of the eighteen paticipants epoted that they wee able to play at least two instuments. The pesent main instuments wee the flute (7), piano (7), violin (3), and keyboad (1). Paticipants epoted that they execised h pe day. All of the musicians had the oppotunity to pefom in public (on aveage, 7 times/ yea). They mainly played music fom Baoque, Classical, and Romantic eas. This was the case fo both taining and spae-time playing. All musicians epoted that they pedominantly listened to classical music ( h/week). Infomed witten consent was obtained fom paticipants pio to the investigation. Absolute Pitch Test The ability to identify the pitch of a given sound without any efeence (absolute pitch) was assessed fo all of the musicians with a mini-ap test intoduced by Zatoe [2003]. Each test sound was a compute-geneated sine wave tone of 1 s duation with 50-ms attack and decay ates, and it was pesented binaually. All musicians got a taining session consisting of tones until they wee familia with the sound. The main AP test contained 10 tones, which wee diffeent fom the taining tones. They wee pesented twice in a andomized ode with a 5-s intestimulus inteval, within which the musicians wee asked to wite down the coesponding tone name. No efeence tone was available to the musicians duing the test. The esponse was egaded as coect if it was within a 1/2 tone distance of the answe tone [Keenan et al., 2001]. Consistent with the pio-epoted non-ap abilities, none of the musicians was able to pefom the task. Afte the fist few notes, 17 of them gave up the test; the othe 3 musicians completed the task with low pefomance (coect esponse: 0, 0, 2 out of 10 testing items). This was in contast to about 90% coect esponses epoted fo AP musicians in liteatue [Keenan et al., 2001]. Mateials Foty shot music pieces (20 pieces of Westen music, 20 pieces of Chinese music) wee chosen fom pe-existing, although not well-known, melodies fo the pesent study. These melodies (labeled as phased) wee clealy stuctued into two phases (each with 4 bas), with pauses indicating phase boundaies. The lengths of the excepts anged fom 9 to 21 s. In ode to obtain anothe set of matched unphased melodies, the phase bounday was modified by a pofessional musician by filling it with one o moe musically plausible notes. This technique has aleady been used in ealie studies [Knösche et al., 2005; Nan et al., 2006; Neuhaus et al., 2006]. It ensues that the mete emains intact, which would not be the case afte just eliminating the pause. It is clea that the manipulated pieces cannot be consideed completely unphased in the stict sense, since phase bounday makes othe than the pause (e.g., pefinal lengthening, hamonic closue, melodic contou, etc.) ae still in place. Theefoe, the condition might have been called less phased. Howeve, fo the sake of bevity and consistence with ealie publications, we will call the expeimental conditions phased and unphased heeafte. An additional set of melodies was ceated by combining two fagments fom diffeent examples, which, howeve, belonged to the same style (Westen o Chinese). Fagments wee ceated by splitting a musical piece into two at a andomly chosen splitting point. The combined pieces wee always composed of a font and a tail fagment. These stimuli seved as task items and wee excluded fom the final analysis. All of the melodies wee synthesized with piano-like timbe. Note that the change of timbe fom a Chinese instument (such as ehu) to a piano fo Chinese music, though necessay, compomises Chinese music employed in the cuent study as entiely epesentative. As a esult, Chinese music would not be ecognized by its diffeent 315

5 Nan et al. TABLE I. Compaison of phase bounday featues between Chinese and Westen melodies Chinese melodies timbe, but athe by its diffeent tuning and scale system as compaed with Westen music. A bief summay of the acoustic featues of both goups of stimuli is listed in Table I. Examples of stimuli ae shown as scoes in Figue 1. Pocedue Westen melodies Length of the peceding 8,012 (441) 7,649 (349) phase (ms) Pause length (ms) 549 (39) 500 (39) Length of the whole melody (ms) 15,399 (850) 14,422 (640) Values ae aveages, with 95% confidence intevals in paentheses. The expeimental paadigm is sketched in Figue 2. All melodies wee pesented in pseudo-andom ode, in ode to avoid sequence effects. Thee was one expeimental un fo each paticipant. Each un consisted of 40 Westen and 40 Chinese musical pieces (half phased and half unphased) togethe with 8 combined task items. The task was to categoize the ongoing melodies into thee classes: combined task melodies (which could be eithe Chinese o Westen), Chinese melodies, and Westen melodies. Befoe the eal measuement, a detailed instuction and a taining session wee given to assue that the paticipants undestood the task well. The answes wee collected by delayed key-pess (a visual pompt sentence was given afte the pesentation of each melody and also a feedback fo the coectness of the answe was povided immediately). The same esponse patten was assigned to each paticipant: ight index key-pess fo Chinese, ight middle fo Westen, and no key-pess fo the combined melodies. The combined items wee intoduced into the expeiment to make the task moe difficult and elying on the entie mateial and not only the beginnings. Since the combined items could be sewn togethe at any point, sub- Figue 1. Example scoes of stimuli. The last line demonstates how the combined taget items ae ceated. It is composed of a head fagment of w46p and a tail fagment of w60p. The seaming point is indicated by the inteuption in the scoe. 316

6 Coss-Cultual Music Phase Pocessing Figue 2. Sketch of the expeimental paadigm. [Colo figue can be viewed in the online issue, which is available at jects had to attend to the end of a melody befoe making a decision. Subjects wee not infomed that the pupose of the expeiment was the investigation of the peception of phasing in a coss-cultual context. MR Imaging MRI data wee collected by a 3T scanne (Siemens TRIO, Elangen, Gemany). Twenty axial slices (19.2 cm field of view (FOV) with a matix, 4 mm thickness, 1 mm inteslice distance), paallel to the AC-PC plane and coveing the whole bain wee acquied using a single shot gadient ecalled EPI sequence (TR 2000 ms, TE 30 ms, 90 flip angle). One functional un with 982 scans was caied out, with each scan sampling ove the 20 slices. Befoe functional imaging, 20 anatomical T1-weighted MDEFT [Nois, 2000; Ugubil et al., 1993] images (data matix , TR 1.3 s, TE 10 ms) and 20 T1-weighted EPI images with the same spatial oientation as the functional data wee acquied. Additionally, in ode to coegiste functional scans, high-esolution whole-head 3-D MDEFT bain scans (128 sagittal slices, 1.5 mm thickness, FOV cm 3, data matix of voxels) wee acquied in a sepaate session. Paticipants wee comfotably placed in the scanne in supine position. Cushions wee used to educe head motion. The auditoy stimuli wee pesented binaually though specially constucted, MR compatible headphones (Commande XG, Resonance Technology, Nothidge). The combination of soundpoof headphones and special ea plugs attenuated the scanne noise without educing the quality of sound stimulation. The loudness of the stimulation system was adapted so that the paticipants could peceive the stimuli without any poblems (90 db). The visual cues wee displayed with an LCD pojecto on a back-pojection sceen mounted in the boe of the magnet behind the paticipant s head. Paticipants viewed the sceen weaing mio glasses. Data Analysis The softwae package LIPSIA was employed fo MRI data analysis [Lohmann et al., 2001]. The seies of pocesses included pepocessing, egistation, nomalization, and statistical evaluation of the MRI data. Pepocessing was fist stated with an offline motion-coection of the functional data with the Siemens motion coection potocol (Siemens, Elangen, Gemany). Then a cubic-splineintepolation and a tempoal highpass filte with a cutoff fequency of 1/280 Hz wee applied fo an fmri slice acquisition sequence and baseline coections. The last step of pepocessing was spatially smoothing with a Gaussian kenel of 5.65 mm full-width half-maximum (FWHM). Afte the pepocessing, a igid linea egistation with six degees of feedom (3 otational, 3 tanslational) was pefomed in ode to align the functional data slices onto a 3D steeotactic coodinate efeence system. The otational and tanslational paametes wee acquied on the basis of the MDEFT [Ugubil et al., 1993] and EPI-T1 slices to achieve an optimum match between these slices and the individual 3D efeence data set. This 3D efeence data set was acquied fo each subject duing a pevious scanning session. The MDEFT volume data set with 160 slices and 1 mm slice thickness was standadized to the Talaiach steeotactic space [Talaiach and Tounoux, 1988], that is, scaled to match the extent of the Talaiach bain and otated and tanslated into the Talaiach coodinate system with the oigin at the anteio commissue (AC) and the anteio-posteio commissual line defining the y-axis. The otational and tanslational paametes wee then used to tansfom the functional slices using tilinea intepolation, so that the esulting functional slices wee aligned with the steeotactic system. Statistical evaluation was based on a least-squaes estimation using the geneal linea model fo seially auto-coelated obsevations [see also Aguie et al., 1997; Fiston et al., 1995; Wosley and Fiston, 1995; Zaahn et al., 1997]. Fo each music style (Westen, Chinese), a box-ca function with the same duation as the music excepts was convolved with a hemodynamic esponse function [HRF, constucted by a gamma density function; Glove, 1999] in ode to geneate the egessos. In ode to model the pocess undelying the phasing, an HRF with a esponse delay of 6 s, shifted by the latency of the phase bounday offset, was used fo the design matix. Additionally, combined melodies wee included in the model as an independent egesso (constucted the same way as fo Westen o Chinese melodies). The model equation, including the obsevation data, the design matix and the eo tem, was convolved with a Gaussian kenel of dispesion of 4 s FWHM to account fo 317

7 Nan et al. the tempoal autocoelation [Wosley and Fiston, 1995]. Fo each paticipant, two contast maps (i.e., estimates of the aw-scoe diffeences of the beta coefficients between specified conditions) wee geneated, which epesented the main effects of: (1) Westen vesus Chinese music and (2) phased vesus unphased melodies. As the individual functional datasets wee all aligned to the same steeotactic efeence space, the single-paticipant contast-images wee then enteed into a second-level andom effects analysis fo each of the contasts. The goup analysis consisted of a one-sample t-test acoss the contast images of all paticipants that indicated whethe obseved diffeences wee significantly distinct fom zeo [Holmes and Fiston, 1998]. Subsequently, t values wee tansfomed to Z scoes. Images wee thesholded at z > 3.09 (P < 0.001, uncoected). Moeove, a egion was consideed significant only if it contained a cluste of 5 o moe continuous voxels [Bave and Bongiolatti, 2002; Foman et al., 1995]. This double theshold coesponds to a 5% multiple compaisons adjusted pobability of falsely identifying one o moe activated voxel clustes on the basis of Monte Calo simulations (Alphasim/AFNI). In addition, a time couse analysis of the fmri signal was conducted. Tial-aveaged time couses (phase bounday offset locked) fo all of the fou conditions (2 music styles 2 phase conditions) wee extacted fom the pepocessed data fo each paticipant at a sampling ate of 1 s. In ode to detect the possible inteaction between music style and phasing conditions, a subsequent ROI analysis was conducted. All significant aeas of both contasts wee employed as ROIs. The 4 s pio to phase bounday onset was consideed as baseline. Pecent signal change was calculated elative to this baseline. Mean activation fom the peak voxel ove a 6 s time window (2 though 8 s afte phase bounday onsets) was computed fo each paticipant. A two-way ANOVA with the factos STYLE (Westen vs. Chinese) and COND (phased vs. unphased) was pefomed fo each ROI. RESULTS Behavioal Results We pefomed a two-way ANOVA with the factos STYLE (Westen vs. Chinese) and COND (phased vs. unphased). All musicians pefomed bette fo the Westen than fo the Chinese melodies, as indicated by a significant main effect of STYLE (F (1,17) ¼ 44.9, P < 0.001). The mean ecognition ate fo Chinese melodies was 72.1% (standad eo: 2.4%), while Westen ones yielded a mean ecognition ate of 90.6% (standad eo: 1.1%). No significant main effect o inteaction involving the phasing condition (facto COND) was found. Figue 3. Diect contast maps of phase vs. unphased melody listening. Red colos indicate stonge activity fo phased, while blue colos indicate stonge activity fo unphased melodies. The abbeviations of the aeas ae the same as in Table II. In Figues 3 and 6, functional intesubject activation (N ¼ 18) is supeimposed onto a nomalized 3D efeence bain. All of the shown aeas have thei maximal Z- values above 3.09 (P < 0.001, uncoected) and exceed a minimal size of 135 mm 3. Phasing effect Bain aeas that exhibited significant diffeences between phased and unphased melodies ae depicted in Figue 3 and listed in Table II. Listening to unphased melodies esulted in inceased activation in the bilateal supa-tempoal plane, appaently compising the anteio planum tempoale (PT) and possibly involving the posteio Heschl s gyus. 1 Time couses show that the effect is pesent in both Westen and Chinese music fo elatively anteio pats of the PT (Fig. 4A, second ow), while somewhat moe posteioly and medially (Fig. 4B, second ow) in the left hemisphee the phasing effect shows up only fo the Westen music. The time couses also suggest an effect of musical style in that Westen music causes highe activity, but this effect only poved significant fo left posteio PT activation (see next subsection). On the othe hand, phased melodies poduced geate levels of activation in Imaging Results The locations of the clustes in the z-maps wee identified in a Talaiach scaled anatomical bain atlas [May et al., 2004]. 1 In the ight hemisphee, the cluste was situated within PT, while in the left hemisphee, it was centeed between Heschl s gyus, PT, and paietal opeculum. 318

8 Coss-Cultual Music Phase Pocessing TABLE II. Phased vs. unphased melodies Location Volume Z-max x y z R IPS L MFG L IFG L PT R PT R, ight; L, left; IPS, intapaietal sulcus; MFG, middle fontal gyus; IFG, infeio fontal gyus, obital pat (BA 47); PT, planum tempoale. the ight intapaietal sulcus, the left middle fontal gyus, and an aea in the obital pat of the left anteio IFG (BA 47, nea the bode to tiangle pat and to the anteio middle fontal gyus). The time couses of the fome two aeas suggest a geneally highe activation and also a geate contibution to the phasing effect fo the Chinese melodies. Howeve, this was not significant. Music style effect The diect contast between Chinese and Westen music listening evealed the activations listed in Table III and depicted in Figue 5. Inceased activation fo Chinese music was found in the ight latealized netwok compising the posteio insula as well as the middle fontal and angula gyi. 2 Pocessing of Westen music was associated with stonge activation in a widely distibuted netwok in both hemisphees. In the left hemisphee, two clustes of inceased activation wee found. One included moto aeas, like the supeio fontal gyus (putative SMA) and the posteio pecental gyus epesenting the putative mouth aea of the pimay moto cotex. The othe cluste was located aound the supeio tempoal gyus and was extended into the PT and Heschl s gyus. In the ight hemisphee, inceased activations wee seen in the posteio bank of pecental gyus in the classical hand aea and the ight homologue of the peviously found mouth aea. Moeove, an inceased activation fo Westen music was obseved in the ight vento-medial pefontal cotex (VMPC). Inteactions between musical style and phasing ROI analysis Among the obseved egions found fo the above two diect contasts, significant inteactions between COND (phased/unphased) and STYLE (Chinese/Westen) wee detected in thee aeas (see Fig. 6). Fo details, see the time 2 Although these aeas exceeded the significance theshold only in the ight hemisphee, loweing the theshold of minimum cluste volume (135 mm 2 ) also evealed activity in the homologous aeas in the left hemisphee (left MFG consisted coveed 27 mm 2, PI and AnG 81 mm 2 at the chosen z theshold of 3.09). couses in Figue 4. An activity cente nea the left PT and posteio Heschl s gyus (see Table III) showed a significant inteaction between music style and phase pocessing (F (1,17) ¼ 4.40, P ¼ 0.05) in that diffeences between the phased and unphased conditions ae pesent only fo the Westen musical style (accoding to post-hoc compaison, P < 0.05). Inteactions wee also obseved along the left and ight pecental gyi/cental sulci in putative mouth and hand aeas, espectively (F (1,17) ¼ 4.45, P ¼ 0.05; F (1,17) ¼ 7.84, P < 0.01). In both aeas, Westen music seems to cause stonge activation when the phase bounday was eliminated, while fo Chinese music, the phased vesions wee associated with stonge activation. Post-hoc compaisons, howeve, suggest that in the left hemisphee the inteaction stems only fom the Westen melodies (P < 0.05 fo Westen, P ¼ 0.35 fo Chinese music), while in the ight hemisphee thee is no diffeence fo Westen melodies (P ¼ 0.53), but a significant diffeence was obseved fo Chinese music (P < 0.01). DISCUSSION The pesent study aimed at (1) identifying the possible bain coelates fo music phase stuctue pocessing, and (2) isolating cultue-dependent diffeences in the pocessing of music in geneal, as well as the pocessing of musical phase boundaies in paticula. With espect to these poposed eseach questions, the data evealed two lines of findings: (1) With espect to music phase stuctue pocessing, bilateal planum tempoale (PT) wee identified as the most pominent bain aeas fo the pocessing of unphased melodies as compaed to phased ones, while a fonto-paietal netwok consisting of left middle/ infeio fontal gyi and ight intapaietal sulcus was obseved in esponse to phased melodies. Inteaction analysis evealed that pat of left PT diffeences oiginate fom the Westen stimuli only. (2) Cultually unfamilia, non-native (Chinese) music elicited moe activation of a ight-dominant netwok compising the angula gyus, posteio insula, and middle fontal gyus, while familia, native music style (Westen music) activated a bain netwok compising left supeio tempoal aeas (PT, HeG, STG), ight VMPC, and bilateal moto egions. An inteaction between the cultual style of the music and phasing conditions was evealed in some egions along the posteio bank of the pecental gyus (significant in left putative mouth and ight putative hand aeas). Phase Stuctue Pocessing The ole of the planum tempoale Thee ae two altenative intepetations of the bilateally inceased PT activation associated with unphased melody pocessing. Fist, compaed to the phased melodies, the unphased melodies contain a few moe notes. Since the PT is known to be associated with basic acoustic 319

9 Nan et al. Figue 4. Time couses fo those aeas with significant effects of the phase bounday (contast phased unphased and inteaction). Gey: Chinese; black: Westen; solid: phased; dotted: unphased. The tigge point (time point zeo) is at the offset of the phase bounday making pause. The backeted numbe tiples denote the Talaiach coodinates of the activity centes. Fo abbeviations, see Tables II and III. A: Aeas with significant contast between phased and unphased items (see Table II). B: Aeas with significant inteaction between music style and phase pocessing fom the ANOVA on the aveage values between 2 and 8 s afte phase bounday offset (see Fig. 6). [Colo figue can be viewed in the online issue, which is available at 320

10 Coss-Cultual Music Phase Pocessing TABLE III. Westen vs. Chinese music Location Volume Z-max x y z R MFG a R AnG a RPI a L SMA/peSMA L CS oal epesentation 2, R PCGhand epesentation R CS oal epesentation L PT/HeG L PT/STG R VMPC Tables II and III list Talaiach coodinates [Talaiach and Tounoux, 1988], Z-values, and volume (mm 3 ) of the activated egions yielded fom the aveaged contast images based on the individual contast (Figue 5: Chinese vs. Westen; Figue 3: phased vs. unphased). Z-values wee thesholded at Z > 3.09 (P < 0.001, uncoected). Only the activation clustes exceeding a minimal size of 135 mm 3 (5 voxels) wee listed. a These aeas ae also active in the othe hemisphee, but failed to each the theshold fo cluste size (fo left PI and AnG, the blobs consisted of 3 voxels : 81 mm 2, fo left MFG 1 voxel : 27 mm 2 ). Abbeviations: R, ight; L, left; MFG, middle fontal gyus; AnG, angula gyus/infeio paietal lobe; PI, posteio insula; PCG, medial pat of pecental gyus/foot aea; CS, cental sulcus, posteio bank of pecental gyus in mouth and hand egions; PT, planum tempoale; HeG, Heschl s gyus; STG, supeio tempoal gyus; VMPC, vental medial pefontal cotex (supeio ostal gyus). pocessing [e.g. Giffiths and Waen, 2002; Knösche et al., 2003; Lütkenhöne and Steinstäte, 1998], this could have led to an incease in PT activity in ou data. Second, the involvement of bilateal PT in the pocessing of unphased melodies might altenatively be explained by the ecently poposed integation function between apidly and slowly changing infomation [Meye et al., 2004]. Accoding to this view, the obseved PT activity might eflect the identification of phase boundaies tiggeed by local cues (apid infomation) pedicted on the basis of moe global stuctual featues of the music piece (e.g., contou, mete). The absence of a pause makes phase bounday identification moe difficult (but not impossible, because othe cues ae available, like e.g., pefinal lengthening), which esults in stonge PT activity fo the unphased melody vesions. If this intepetation is coect, one would pedict that the obseved effect is mainly o exclusively pesent with cultually familia music. Indeed, at least fo pat of left PT, inteaction analysis shows that inceased activity fo unphased melody vesions was only obseved fo the Westen musical style. Moe geneally, the supeio posteio pat of the bilateal tempoal lobe, including the PT, has been suggested to constitute the pimay substate fo constucting soundbased epesentations of speech [Hickok and Poeppel, 2000, 2004], but also fo nonspeech sound pocessing [e.g. Jäncke et al., 2002b]. Auditoy pocessing is geneally asymmetic between hemisphees depending upon the specific mateial pocessed and the task. It is poposed that the left hemisphee specializes in tempoal analysis and the ight hemisphee specializes in spectal analysis [Zatoe 1997; Zatoe et al. 2002]. Similaly, Poeppel [2001, 2003] suggested the asymmetic sampling in time (AST) model, stating that the left hemisphee has a shote time integation window (20 50 ms) and the ight one has a Figue 5. Diect contast maps of Westen vs. Chinese melody listening. Red colos indicate stonge activity fo Westen, while blue colos indicate stonge activity fo Chinese melodies. The abbeviations of the aeas ae the same as in Table III. 321

11 Nan et al. A fonto-paietal netwok fo phase bounday pocessing Figue 6. Mean pecent signal change and standad eo fo egions with significant inteaction between music style and phase pocessing. CS, cental sulcus; PT, planum tempoale. longe time integation window ( ms). In pinciple, phasing and phase bounday detection ely on both tempoal and spectal cues. Howeve, the distinction of vey fine gained tempoal featues seems to play the most impotant ole [see e.g. Neuhaus et al., 2006]. Fo example, the diffeences in length of the closing tones of the pebounday phase was found to be about ms on aveage, if the same musical pieces wee played by a pianist with and without phase bounday [Knösche et al., 2005; Table I]. Hence, accoding to the Zatoe model, phasing should be pocessed pefeentially in the left hemisphee. Indeed, in ou data, the left PT seems to be consideably moe active (see Table II), which is not only in accodance with the Zatoe model, but also with findings by Ohnishi et al. [2001], demonstating left latealized PT activity in musicians (but not in non-musicians) when listening to music. The pocessing of phased, as contasted to unphased, melodies activated aeas in the left obital IFG (BA47), left middle fontal gyus (MFG), and ight intapaietal sulcus (IPS). The infeio fontal cotex has been shown to be involved in a wide ange of bain pocesses including syntactic, semantic, and phonological aspects of speech pocessing [e.g. Caplan et al., 1999; Fiez et al., 1999; Poldack et al., 1999; Wagne et al., 2001; see also Fiedeici and Alte, 2004; Fiedeici, 2002 fo eview], pocessing of nonlinguistic sequential input [Adams and Janata, 2002; Giffiths et al., 1999; Mülle et al., 2001], music pocessing [Levitin and Menon, 2003, 2005; Maess et al., 2001], and pocessing of timing infomation [Schubotz and von Camon, 2001; Schubotz et al., 2000]. The aea obseved in ou study is well anteio to the bulk of IFG activations epoted in liteatue. Howeve, a simila focus of activation (distance 1.3 cm in Talaiach space) has, e.g., been epoted by Meye et al. [2004] fo the pocessing of degaded speech (all syntactic and semantic contents emoved, while posody is peseved) in contast to nomal and flattened (posody emoved) speech. Vey simila activations have also been epoted by Tillmann et al. [2003] in a musical piming expeiment involving chod sequences (distances 1 2 cm to ou activation). In the light of thei own findings, as well as pevious evidence, Tillmann et al. chaacteize the ole of the infeio fontal cotices as... pocessing and integating infomation ove time and the compaison of olde stoed infomation with newe incoming infomation. Such integation of infomation ove time is also expected to be chaacteistic fo the pocessing of phase boundaies: the musical context established by the fist phase, compising infomation on, fo example, hythm, mete, melodic line, timbe, and tonal key, has to be kept active ove the time of the pause and then integated with the new incoming infomation of the second phase in ode to establish a elation between phases. It is woth pointing out that activity in BA47 has been found to specifically coelate with the pocessing of musical tempoal stuctue [Levitin and Menon, 2003; Vuust et al., 2006], although thee is some vaiability in the exact positions of the epoted activity centes. The question emains, why is the activity in the left IFG stonge fo the phased as compaed to the unphased items. Afte all, one could ague that phasing is moe difficult without the pause and activity should be highe. Howeve, the demand on woking memoy due to the tempoal delay is highe in the phased condition. It could theefoe be that BA 47 eflects the woking memoy aspect of pocessing tempoal stuctue in music. Indeed, it has ecently been shown that the IFG (BA 45) plays a citical ole in syntactic woking memoy duing auditoy language peception [Caplan et al., 1999]. The ight IPS is pat of the poposed dosal fonto-paietal netwok of attention [Cobetta and Shulman, 2002; 322

12 Coss-Cultual Music Phase Pocessing Imauoka et al., 2003]. This netwok is esponsible fo topdown signals fo visual attention [Cobetta et al., 2002; Small et al., 2003; Thiel et al., 2004] and auditoy attention [Collette et al., 2005]. In paticula, Thiel et al. [2004] identified a netwok fo eoienting attention in bilateal IPS and MFG, in which the loci of the left MFG and the ight IPS activations ae in close ageement with active aeas found in ou study (distance <1 cm in Talaiach coodinates). MFG activation coelated to speech and tone segmentation pocesses has also been epoted [Buton et al., 2000; LoCasto et al., 2004]. It is theefoe likely that the obseved fonto-paietal netwok (i.e. left MFG and ight IPS) eflects moe geneal cognitive pocesses like selective attention involved in the segmentation of the musical input into phases. Hence, the peception of phased as compaed to unphased melodies in tained musicians is associated with highe activity in the obital pat of the left IFG and lowe activity in bilateal PT. In pinciple, these effects ae pesent fo both cultually familia and unfamilia music, although some activity in the left PT is esticted to the familia music condition. This patten of cultually specific as well as cultually univesal phasing peception as indicated by these obseved bain activations is in line with the findings of Nan et al. [2006] in the sense that music phasing peception as eflected by the ERP also compises cultually geneal (duing the time ange of the CPS) as well as cultual specific subcomponents (at ealie latencies). Aeas Associated With Music Style Pocessing In geneal, it should be noted that thee is a cetain confound between the cultual familiaity of the music and basic acoustic popeties, which ae of couse not (and cannot be) identical between the two musical styles. Theefoe, we have to be cautious when intepeting activations in aeas that ae involved in basic acoustic pocessing, that is, the supeio tempoal lobe (Heschl s gyus, PT and supeio tempoal gyus). Aeas fo familia style Among the aeas found to be moe active fo the familia music style (Westen music), the ole of the ight vental medial pefontal cotex (VMPC) has been suggested to eflect the peceived ease of etieval [Ranganath et al., 2003] o the feeling of knowing [Schnye et al., 2004, 2005]. Consistent with these notions, VMPC showed enhanced activity fo the familia music style (Westen music) as compaed to the unfamilia one (Chinese music) in the cuent study. Because in ou case, the Geman musicians wee not familia with the musical pieces pe se, the etievable infomation would efe to the geneal Westen musical schema athe than explicit knowledge of paticula melodies. Howeve, thee is also evidence that VMPC plays a ole in pleasue and ewad behavios [Blood and Zatoe, 2001]. Hence, the obseved VMPC esponse to Westen music might altenatively be intepeted to indicate that the familia style of music is moe positively evaluated. Some othe constituents of the neual netwok fo familia music pocessing belong to the moto system, namely those in the bilateal pecental gyus (including the SMA). The pocessing of event sequences in geneal (especially complex ones) is known to involve the so-called moto cicuit [e.g. Schubotz et al., 2000; fo a eview see e.g. Janata and Gafton, 2003]. These pocesses ae cetainly influenced by attention and it seems plausible to assume that classical Westen music, being moe appealing to the Geman subjects in compaison to the Chinese examples, attacted moe attention. Hence, the geneal attactiveness of cultually familia music and consequently inceased attention could be esponsible fo the obseved activation inceases in moto aeas. Altenatively, the obseved moto activations cooboate the notion of sensoimoto coupling. Music is not at all a puely peceptual phenomenon, but athe it is closely elated to action, patly diected at poducing the music (singing, instument playing) and patly unelated to poduction (clapping, tapping, swinging, dancing, etc.). Theefoe, one possible explanation fo the obseved moto system activations might involve intenal tapping o swinging, which is not diectly elated to the poduction of the peceived music. On the othe hand, the elation between the peception of a stimulus and the moto behavio necessay to poduce the same stimulus has been attibuted to the mio neuon system [Feai et al., 2003; Rizzolatti et al., 1996]. Activation in the human moto system is not only found duing obsevation of action (classical mio neuons), but also duing listening of othe s actions [audio-visual mio neuons; Aziz-Zadeh et al., 2004], and duing speech listening [echo neuons; Rizzolatti and Caigheo, 2004; see also Fadiga et al., 2002; Sundaa et al., 2001; Watkins and Paus, 2004; Watkins et al., 2003]. Thus, if thee is an echo neuon system in humans, that is, a system that motoically esonates peceived vebal mateial [Rizzolatti and Caigheo, 2004], it seems plausible to assume that othe modes of human communication exhibit a simila phenomenon. Indeed, this has been demonstated fo humans viewing symbolic hand signs [Nakamua et al., 2004] and fo pianists listening to music [Haueisen and Knösche, 2001]. In the pesent study, the Westen musical pieces wee not familia to the listenes as such, but they wee familia in tems of cultual style. Hence, if the echo neuon account fo the activity in the pecental gyus holds, this means that cultual familiaity, that is, the familiaity with cetain abstact ules and conventions, is sufficient to activate the echo neuons. Inteestingly, Koelsch et al. [2006] obseved activity in the Rolandic opeculum nea the putative epesentation of the vocal tact (laynx and phaynx) in non-musicians who wee listening to emotionally pleas- 323