This is a sel -a chi ed e sion o an o iginal a icle. This e sion
may di e om he o iginal in pagina ion and ypog aphic de ails.
Au ho (s):
Ti le:
Yea :
Ve sion:
Copy igh :
Righ s:
Righ s u l:
Please ci e he o iginal e sion:
CC BY-NC-ND 4.0
h ps://c ea i ecommons.o g/licenses/by-nc-nd/4.0/
The ch onnec ome o musical bea
© 2019 Else ie Inc.
Published e sion
Toi iainen, Pe i; Bu una , Iballa; B a ico, El i a; Vuus , Pe e ; Allu i, Vinoo
Toi iainen, P., Bu una , I., B a ico, E., Vuus , P., & Allu i, V. (2020). The ch onnec ome o
musical bea . Neu oimage, 216, A icle 116191.
h ps://doi.o g/10.1016/j.neu oimage.2019.116191
2020
The ch onnec ome o musical bea
Pe i Toi iainen
a
,
*
, Iballa Bu una
a
, El i a B a ico
b
, Pe e Vuus
b
, Vinoo Allu i
c
a
Finnish Cen e o In e disciplina y Music Resea ch, Depa men o Music, A and Cul u e S udies, Uni e si y o Jy €
askyl€
a, Finland
b
Cen e o Music in he B ain (MIB), Depa men o Clinical Medicine, Aa hus Uni e si y, The Royal Academy o Music Aa hus/Aalbo g (RAMA), Aa hus, Denma k
c
Cogni i e Science Lab, Kohli Cen e o In elligen Sys ems, In e na ional Ins i u e o In o ma ion Technology, Hyde abad, India
ARTICLE INFO
Keywo ds:
Music
Bea
MRI
Na u alis ic imaging
Music in o ma ion e ie al
Dynamic connec i i y
ABSTRACT
Keeping ime is undamen al o ou e e yday exis ence. Va ious isoch onous ac i i ies, such as locomo ion,
equi e us o use in e nal imekeeping. This phenomenon comes in o play also in o he human pu sui s such as
dance and music. When lis ening o music, we spon aneously pe cei e and p edic i s bea . The p ocess o bea
pe cep ion comp ises bo h bea in e ence and bea main enance, hei ela i e impo ance depending on he
salience o bea in he music. To s udy unc ional connec i i y associa ed wi h hese p ocesses in a na u alis ic
si ua ion, we used unc ional magne ic esonance imaging o measu e b ain esponses o pa icipan s while hey
we e lis ening o a piece o music con aining s ong con as s in bea salience. Subsequen ly, we u ilized dynamic
g aph analysis and psychophysiological in e ac ions (PPI) analysis in connec ion wi h compu a ional modelling o
bea salience o in es iga e how unc ional connec i i y mani es s hese p ocesses. As he main e ec , co ela ion
analyses be ween he ob ained dynamic g aph measu es and he bea salience measu e e ealed inc eased cen-
ali y in audi o y-mo o co ices, ce ebellum, and ex as ia e isual a eas du ing low bea salience, whe eas
egions o he de aul mode- and cen al execu i e ne wo ks displayed high cen ali y du ing high bea salience.
PPI analyses e ealed pa ial dissocia ion o unc ional ne wo ks belonging o his pa hway indica ing comple-
men a y neu al mechanisms c ucial in bea in e ence and main enance, p ocesses pi o al o ex ac ing and
p edic ing empo al egula i ies in ou en i onmen .
1. In oduc ion
Humans ha e a p opensi y o engage in isoch onous ac ions, which
may be oo ed in e olu iona y adap a ion (Me ke e al., 2009). In
addi ion o p oducing sound pa e ns ha a e pe iodic o e ime, we a e
able o ex ac empo al egula i ies om he audi o y en i onmen , be i
language o music (Ko z e al., 2009;G ube e al., 2013). We a e hus
capable o ex ac ing a egula pulse om he audi o y en i onmen ,
which allows us o p edic u u e hy hmical e en s o which we can hen
synch onize ou mo emen s (D ake e al., 2000;Honing, 2012;Maes
e al., 2014).
When lis ening o music, we a e spon aneously d i en o find a bea ,
which makes bea pe cep ion undamen al o ou expe ience o music
and he e o e key o explaining musical beha io (La ge and Snyde ,
2009;Vuus e al., 2018). To da e ou ema kable and possibly uniquely
human abili y o e o lessly pe cei e he bea in music keeps challenging
cogni i e scien is s, who pu sue o unde s and i s unde lying b ain p o-
cesses (La ge and Palme , 2002). A deepe unde s anding he eo would
be key o explaining a ange o musical beha io s, such as why mo ing
o music is pleasu able (La ge and Palme , 2002;La ge and Snyde , 2009;
Za o e and Salimpoo , 2013), o why lis ening o hy hms has a a o -
able e ec on pa ien s wi h neu odegene a i e gai diso de s (e.g., Pa -
kinson’s disease; Nombela e al., 2013;Ashoo i e al., 2015;Thau e al.,
2018), s oke pa ien s (B ad e al., 2010), and au is ic indi iduals (Wan
e al., 2011). Such he apeu ic po en ial is he eason why he neu o-
science o hy hm has de eloped in o a specificfield o esea ch among
clinicians and schola s om many disciplines (An a i e al., 2002;
F ançois e al., 2015).
Music wi h egula pe iodic pulse exis s in e e y cul u e, a o ding
coo dina ion be ween pe o me s and o en a he same ime inducing a
synch onized mo o esponse om lis ene s (Ne l, 2000). An ob ious
indica o o pulse sensi i i y is ou ins inc i e, almos subconscious, skill
o mo o synch oniza ion o music which mani es s a a young age
(T ehub, 2001;T aino , 2005). Undeniably, ou p ocli i y o hy hmic
mo emen is e idenced in in ancy and o a g ea e ex en o music and
o he me ically egula sounds han o speech (Zen ne and Ee ola,
2010). Such a p edisposi ion o mo e wi h music is specula ed o be
oo ed in he e olu ion o en ainmen and social beha io , as
* Co esponding au ho .
E-mail add ess: pe i. oi iainen@jyu.fi(P. Toi iainen).
Con en s lis s a ailable a ScienceDi ec
Neu oImage
jou nal homepage: www.else ie .com/loca e/neu oimage
h ps://doi.o g/10.1016/j.neu oimage.2019.116191
Recei ed 20 June 2019; Recei ed in e ised o m 10 Sep embe 2019; Accep ed 11 Sep embe 2019
A ailable online 13 Sep embe 2019
1053-8119/©2019 Else ie Inc. This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/).
Neu oImage 216 (2020) 116191
synch oniza ion wi h a bea is he simples o m o en ainmen (Jana a
e al., 2012;Madison, 2006;Madison e al., 2011). I is he e o e no
su p ising ha music and dance ha e in some cul u es e ol ed oge he
(C oss, 2008;Wallin e al., 2000).
Bea pe cep ion comp ises wo di e en ypes o subp ocesses
(Came on and G ahn, 2016). Fi s , bea induc ion (bea finding) can be
defined as gene a ion o pu a i e bea s so as o en ain o synch onize o
he isoch onous de ia ions caused due o exp essi e iming in music o ,
in o he wo ds, ac i e in e ence o pe cei ed bea . Second, bea main-
enance (bea con inua ion) can be desc ibed as mechanical, au onomous
ime keeping, which in ol es con inual, sus ained measu emen o p e-
dic able in e als wi hou significan need o a en ion.
The salience o musical bea a ies be ween and wi hin musical
pieces, depending on he deg ee o which he music con ains epe i i e
hy hmic pa e ns wi h a sui able a e. High bea salience is e oked by
musical e en s ha occu a egula in e als a a sui able empo and
ha e sha p a acks. The op imal empo o bea salience has been ound
o lie in he icini y o 100–120 e en s pe minu e, co esponding o a
bea pe iod o 500–600 ms (Pa ncu , 1994;Toi iainen and Snyde ,
2003; an Noo den and Moelan s, 1999), which closely co esponds o
he spon aneous empo o walking (McDougall and Moo e, 2005) and
apping (F aisse, 1982). Bea salience has been ound o be associa ed
wi h he u ge o mo e o music (Madison e al., 2011). Compu a ional
models o bea salience p oposed so a can be di ided in o wo ca e-
go ies based on he kind o music ep esen a ion hey use: symbolic
models (e.g., Pa ncu , 1994;Toi iainen and Snyde , 2003), which
ope a e on no a ion-like music ep esen a ions, such as MIDI, and audio
models (e.g., La illo e al., 2008;Madison e al., 2011), which ake
music eco dings as inpu .
I can be assumed ha , du ing con inuous lis ening o music, bea
in e ence and bea main enance exis concu en ly, wi h bea in e ence
being mo e p ominen du ing momen s o low bea salience, and bea
main enance du ing momen s o high bea salience. Namely, wi hin he
amewo k o ac i e pe cep ion, bea in e ence and main enance can be
ega ded as p edic i e coding (Koelsch e al., 2019) wi h low and high
ce ain y, espec i ely. Wi hin his amewo k, he wo p ocesses could
be conside ed o exis on a con inuum ins ead o being dicho omous. This
wa an s he use o a con inuous a iable ha measu es he deg ee o
bea salience h oughou he musical pieces a each ime poin . The
model by La illo e al. (2008), applied in he p esen s udy, allows he
ex ac ion dynamic a ia ion o bea salience. The s eng h o his model
is ha i has been pe cep ually alida ed and he e o e a o ds in e -
p e abili y o he ound neu al co ela es.
Came on and G ahn (2016) p o ide a comp ehensi e e iew high-
ligh ing he oles o co ical, subco ical, and ce ebella egions in bea
pe cep ion, including in e ence and main enance. They highligh he
need o dynamically in es iga ing he ime cou se o hy hm and bea
pe cep ion. While neu al co ela es o musical bea ha e been in es i-
ga ed using na u alis ic s imula ion on bo h oxel-le el (Allu i e al.,
2012,2013;Toi iainen e al., 2014) and ne wo k-le el (Bu una e al.,
2017), e ec s o musical bea salience on he empo al dynamics o
o e all unc ional connec i i y in he b ain ha e no been s udied so a .
1.1. Bea and neu al ac i a ion
When s udying he b ain p ocesses which sus ain iming, se e al
mo o - ela ed b ain egions ha e been ound o be in ol ed (G ahn,
2009). This has been conside ed a g oundb eaking finding in music
neu oscience, as me ely lis ening o music in he absence o any o e
mo emen engages mo o esou ces o he b ain. These include p emo o
co ex (PMC), supplemen a y mo o a ea (SMA), and basal ganglia
(G ahn and B e , 2007;Chen e al., 2008;Beng sson e al., 2009;G ahn
and Rowe, 2009;Geise e al., 2012;Teki e al., 2012;Kung e al., 2013).
Rhy hm hus ac i a es he neu al ci cui s in ol ed in mo o p ocessing,
which p o ides a possible accoun o he in ima e connec ion be ween
music and mo emen (Za o e and Salimpoo , 2013).
Bo h he basal ganglia (especially he pu amen) and he SMA seem o
be pa icula ly c ucial o bea pe cep ion (Came on and G ahn, 2016;
G ahn and B e , 2007). Mo o a eas, pa icula ly he SMA, pe o m ime
p edic ion ia op-down connec ions o he audi o y co ex (Cadena--
Valencia e al., 2018;Me chan and Honing, 2014;Me chan e al.,
2015a,b;Me chan and Ya ow, 2016). The basal ganglia ha e been
implica ed in bo h bea finding (G ahn and Rowe, 2009) and bea
con inua ion (Kung e al., 2013). Mo e specifically, inc eased ac i a ion
in he basal ganglia du ing hy hm pe cep ion has been epo ed o
eflec he in e nal gene a ion o upcoming bea s once he empo al
s uc u e o he music has been es ablished and i emains egula . In line
wi h his, ac i a ion in he basal ganglia dec eases in esponse o i eg-
ula hy hms, as hese lead o modifica ions in bea p edic ion (G ahn
and Rowe, 2013), while ce ebella ac i i y inc eases, which is consis en
wi h a dis inc ole o he ce ebellum in absolu e iming (G ube e al.,
2010;Teki e al., 2012;Came on and G ahn, 2016). In line wi h his,
Spence and I y (2013) e iew li e a u e on ce ebellum and iming and
conclude ha i s ole is key in “p o iding a p ecise, me ical ep esen-
a ion o defined empo al in e als” a he han gi e ise o abs ac
empo al ep esen a ions o unde lying hy hms.
1.2. Bea and unc ional connec i i y
Keeping ime is likely o be an eme gen ne wo k p ope y, and hence
me ely localizing i o pa icula egions p o ides an incomple e pic u e
o i s neu al unde pinnings. Consequen ly, unc ional connec i i y
s udies in hy hm pe cep ion ha e p o ided new insigh s by iden i ying
associa ed audi o y- and mo o - ela ed ne wo ks, which may acili a e
he empo al p edic ions in ol ed in hy hm pe cep ion (Za o e e al.,
2007;Pa el and I e sen, 2014). Howe e , unc ional connec i i y s udies
a e a e and ha e no su ficien ly in es iga ed his phenomenon beyond
he Gene al Linea Model (GLM) app oach. In pa icula , a unc ional
ne wo k o co ical and subco ical s uc u es, consis ing o audi o y
co ex, PMC, pu amen, and SMA, seems o play a ole du ing bea
pe cep ion, especially when in e nal gene a ion o he bea is equi ed
(G ahn and Rowe, 2009). This co obo a es he hypo hesis ha he PMC
and do sal audi o y egions mainly e ol ed o sus ain he mechanisms o
empo al sequencing (Lea e e al., 2009;Rauschecke and Sco , 2009).
Howe e , as indica ed in Me chan e al. (2015), while one s udy e-
po ed inc eased audi o y-p emo o coupling as a unc ion o bea
salience in isoch onous sequences (Chen e al., 2006; see also Assaneo
and Poeppel, 2018) ano he s udy ound his coupling o dec ease as
me ic complexi y inc eased (Kung e al., 2013). Because eal music is a
me ical s imulus, which comp ises di e en ially accen ed g oupings and
subdi isions o bea and hence di e en ially salien bea s, we should
expec a pa e n o connec i i y in ela ion o bea salience ha is
di e en om wha p e iously has been obse ed in s udies using
isoch onous and equally-salien bea s. On a mo e gene al no e, he lack o
consensus does no cu en ly ega d which b ain a eas a e in ol ed in he
s udy o iming and hy hm, bu mo e specifically, wha connec ions exis
amongs he ele an a eas ha gi e ise o pa icula iming beha io s
(G ahn, 2009). Mo eo e , since bea pe cep ion un olds o e ime when
one lis ens o music (Came on and G ahn, 2016), i is c ucial o s udy
how b ain ne wo ks dynamically o ganize hemsel es in con inuous
na u alis ic music lis ening si ua ions.
Recen esea ch on bea p ocessing beyond he GLM app oach
in es iga ed bea saliency using blind sou ce sepa a ion echniques by
means o Independen Componen Analysis (ICA) in combina ion wi h
Music In o ma ion Re ie al (MIR) me hods (Bu una e al., 2017).
Findings e ealed a obus ac ion-pe cep ion unc ional ne wo k o
audi o y and mo o egions associa ed wi h a ia ions in bea salience,
which was concealed om a compa a i e GLM analysis. In pa icula , he
mo o a eas o he ne wo k we e ound o co ela e nega i ely wi h bea
salience. Fu he elucida ion o empo al dynamics o he b ain mecha-
nisms pe aining o bea p ocessing will be necessa y o ully cha ac e ize
i s na u e.
P. Toi iainen e al. Neu oImage 216 (2020) 116191
2
1.3. Bea and ch onnec ome
G aph-based me hods p o ide a use ul ool o spa ially cha ac e -
izing ne wo k s uc u es (Bullmo e and Spo ns, 2009), by ep esen ing
egions o oxels in he b ain as nodes in a g aph. In he domain o music,
g aph heo y has been used o in es iga e unc ional ne wo ks du ing
music lis ening and hei dependence on musical p e e ence (Wilkins
e al., 2014) and aining (Allu i e al., 2017). A g aph- heo e ical mea-
su e commonly used in hese s udies is node cen ali y, which indica es
he deg ee o which a egion comp ises an impo an hub in he b ain’s
ne wo k. Va ious me hods ha e been p oposed o quan i y node cen-
ali y. Closeness and be weenness cen ali y a e o en used in analyses
conside ing a mos a ew hund ed selec ed egions in he b ain (Loh-
mann e al., 2010), bu due o hei compu a ional complexi y hey a e
no easible o oxel-based whole-b ain analysis, an app oach necessa y
o disen angling unc ional ne wo k s uc u es wi h su ficien spa ial
accu acy and o a oid dependence on a pa icula choice o b ain pa -
cella ion. Fo his pu pose, Eigen ec o Cen ali y (EC; Bonacich, 2007)
is mo e app op ia e. To es ima e EC, a non-nega i e unc ional
oxel-by- oxel connec i i y ma ix is fi s calcula ed. A common choice
o he connec i i y measu e is Pea son’s co ela ion added by one.
Subsequen ly, Singula Value Decomposi ion is applied o he ob ained
connec i i y ma ix o ob ain he fi s eigen ec o . This ec o con ains
he oxelwise cen ali y alues. EC a ibu es high cen ali y alues o
oxels ha a e s ongly connec ed o a la ge numbe o oxels ha
hemsel es a e cen al, he eby p o iding a global indi ec measu e o
unc ional connec i i y. I he e o e indica es o a gi en condi ion o
ask he impo an compu a ional hubs in he b ain’s unc ional ne wo k.
An appealing ea u e o EC is ha , unlike closeness and be weenness
cen ali y, i does no ely on h esholding and hus is a pa ame e - ee
me hod.
In p e ious s udies, EC has been ound o change as a unc ion o
s imulus (Koelsch and Skou as, 2014), ask (Ga cía-Ga cía e al., 2015),
and condi ion (Binnewijzend e al., 2013;Lou e al., 2015). While
unc ional connec i i y s udies adi ionally ha e assumed connec i i y
pa e ns ha a e s a ic o e ime, dynamic unc ional connec i i y s udies
ha e become inc easingly common ( o a e iew, see P e i e al., 2017).
Calhoun e al. (2014) ha e coined he e m “ch onnec ome” o e e o
ime- a ying whole-b ain connec i i y pa e ns. Dynamic unc ional
connec i i y s udies a e mos commonly based on sliding window ap-
p oaches (P e i e al., 2017) o phase synch oniza ion es ima es (e.g.
Gle ean e al., 2012).
The p esen s udy in es iga es he e ec o bea salience on he dy-
namics o whole-b ain unc ional ne wo ks du ing con inuous lis ening
o music o in es iga e p ocesses unde lying bea in e ence (p e alen
du ing low bea salience) and bea main enance (p e alen du ing high
bea salience). To his end, i employs dynamic g aph analysis (Eigen-
ec o Cen ali y) in andem wi h compu a ional es ima ion o bea
salience om he musical s imulus, o ob ain an o e all iew o how
unc ional ne wo k cen ali y changes dynamically as a unc ion o
musical bea salience. Subsequen ly, a ge ed unc ional connec i i y
analyses (Psychophysical In e ac ion) a e ca ied ou o in es iga e how
unc ional connec i i y a ies and is modula ed by bea salience wi hin
and be ween b ain egions ha ha e p e iously been ound o be
in ol ed in bea pe cep ion.
Due o sca ci y and lack o consensus in p e ious esea ch, specific
hypo heses abou he mani es a ions o he p ocesses o bea in e ence
and main enance in whole-b ain unc ional ne wo k connec i i y a e
di ficul o d aw. Howe e , we assume ha ne wo k cen ali y in a eas
p e iously ound o be ac i a ed by bea pe cep ion, such as co ical,
ce ebella , and s ia al mo o a eas, would be modula ed by he p ocess
o in e ing a salien bea . Fo specific hypo heses ega ding he depen-
dence o unc ional connec i i y on bea salience, we assume, based on
esul s ob ained by Kung e al. (2013), ha because music is a me ical
s imulus, audi o y-mo o connec i i y inc eases wi h low bea salience.
Doya (2000) posi s ha he ce ebello- halamo-s ia o-co ical mo o
pa hway can be di ided in o smalle ne wo ks ha play complemen a y
oles in lea ning and mo o con ol. The e iew pape highligh s he ole
o he ce ebellum as one ha pe o ms supe ised lea ning whe ein an
in e nal ep esen a ion o he da a is al eady a ailable. The co ex’s ole
is desc ibed as one ha pe o ms unsupe ised lea ning, which can be
defined as da a-d i en app oach o lea ning/and esponding o he ex-
is ence o lack o pa e ns in i /guided by on he con ex / he s a is ical
p ope ies o he s imulus (Bos an e al. 2013). In his app oach, he
pa e n o s uc u e has o be lea n based on he a ailable da a and
con ex .
In his amewo k, one could su mise ha he ce ebello- halamo-
co ical mo o pa hway in conjunc ion wi h he audi o y co ex is
esponsible o de eloping an in e nal ep esen a ion o bea . Howe e ,
once his empla e is o med, he ce ebellum con inues o p ocess he
bea and ma ch he esul wi h he o med empla e. One could hen
hypo hesize ha dec eases in bea salience engende de ia ions om he
empla e, which would hen ec ui highe -o de co ical mo o pa h-
ways o ake o e and lea n how bes o fill in he gaps based on he
audi o y in o ma ion and sho - e m empo al con ex a ailable. In o he
wo ds, in he case o low bea salience, highe le el co ical ne wo ks
would be ec ui ed o upda e he p edic ion minimizing p edic ion e o
ed o wa d om lo e le el a eas. Once he bea is in e ed, he main-
enance may no ely on o wa d p edic ion e o om he lowe le els.
2. Me hods
2.1. MRI da a acquisi ion
Pa icipan s: Se en y-se en heal hy pa icipan s (age 29.6 8.9, 43
emales, 71 igh -handed, 1 ambidex ous) wi h no his o y o neu olog-
ical o psychological diso de s pa icipa ed in he MRI expe imen . The
pa icipan s we e sc eened o inclusion c i e ia be o e admission o he
expe imen (no e omagne ic ma e ial in hei body; no a oo o ecen
pe manen colo ing; no p egnancy o b eas eeding; no ch onic pha -
macological medica ion; no claus ophobia). The pa icipan s had a i-
able le els o music educa ion, wi h a median o 3.5 yea s and an
in e qua ile ange o 11.25 yea s o o mal music educa ion.
S imulus & ask: Pa icipan s’b ain esponses we e acqui ed wi h
MRI while hey lis ened o an 8-min long piece o A gen inean ango,
Adi
os Nonino by As o Piazzolla. This s imulus was chosen because i
con ains wide a ia ions in he salience o musical bea . Pa icipan s’
only ask was o a en i ely lis en o he music deli e ed ia high-quali y
MR-compa ible inse ea phones (A o ec, S ua , FL, USA) while keeping
hei eyes open. Foam was used o a enua e he g adien noise. The
sound le el o he s imuli was indi idually adjus ed so ha hey we e
audible abo e he scanne noise bu he olume s ayed wi hin sa e y
limi s (below 80 dB). The s udy p o ocol p oceeded on accep ance by he
e hics commi ee o he Coo dina ing Boa d o he Helsinki and Uusimaa
Hospi al Dis ic . The da a collec ion was pa o a b oade p ojec
(Tun ee ) in ol ing addi ional es s and neu oimaging and neu ophysi-
ological measu es (Allu i e al., 2015;Boge e al., 2016;Bu una e al.,
2014,2015;Ca lson e al., 2015;Haumann e al., 2016;Kliuchko e al.,
2015,2016).
Scanning: Scanning was pe o med using a 3T MAGNETOM Sky a
whole-body scanne (Siemens Heal hca e, E langen, Ge many) and a
s anda d 32-channel head-neck coil, a he Ad anced Magne ic Imaging
(AMI) Cen e (Aal o Uni e si y, Espoo, Finland). Using a single-sho
g adien echo plana imaging (EPI) sequence hi y- h ee oblique slices
(field o iew: 192 192 mm; 64 64 ma ix; slice hickness: 4 mm,
in e slice skip: 0 mm; echo ime: 32 ms; flip angle: 75; oxel size:
222mm
3
) we e acqui ed e e y 2 s, p o iding whole-b ain co e age
pe pa icipan . T1-weigh ed s uc u al images (176 slices; field o iew:
256 256 mm; ma ix: 256 256; slice hickness: 1 mm; in e slice skip:
0 mm; pulse sequence: MPRAGE) we e also collec ed o indi idual
co egis a ion.
P ep ocessing: Func ional MRI scans we e p ep ocessed on a Ma lab
P. Toi iainen e al. Neu oImage 216 (2020) 116191
3
pla o m using SPM8 (S a is ical Pa ame ic Mapping), VBM5 o SPM
(Voxel Based Mo phome y; Wellcome Depa men o Imaging Neu o-
science, London, UK), and cus omized sc ip s de eloped by he p esen
au ho s. Fo each pa icipan , low- esolu ion images we e ealigned on
six dimensions using igid body ans o ma ions ( ansla ion and o a ion
co ec ions did no exceed 2 mm and 2, espec i ely), segmen ed in o
g ey ma e , whi e ma e , and ce eb ospinal fluid, and egis e ed o he
co esponding segmen ed high- esolu ion T1-weigh ed s uc u al im-
ages. These we e in u n no malized o he MNI (Mon eal Neu ological
Ins i u e) segmen ed s anda d a p io i issue empla es using a 12-pa am-
e e a fine ans o ma ion. Func ional images we e hen blu ed o bes
accommoda e ana omical and unc ional a ia ions ac oss pa icipan s as
well as o enhance he signal- o-noise by means o spa ial smoo hing
using an 8 mm ull-wid h-a -hal -maximum Gaussian fil e . Mo emen -
ela ed a iance componen s in MRI ime se ies esul ing om esid-
ual mo ion a i ac s, assessed by he six pa ame e s o he igid body
ans o ma ion in he ealignmen s age we e eg essed ou om each
oxel ime se ies. Following his, spline in e pola ion was used o de end
he MRI da a. Nex , empo al fil e ing was pe o med by Gaussian
smoo hing (ke nel wid h: 4 s), as i p o ides a good comp omise be ween
e ficiency and bias (F is on e al., 2000).
2.2. Musical ea u e ex ac ion
Bea salience o he used s imulus was compu a ionally es ima ed
using he model p oposed and pe cep ually alida ed by La illo e al.
(La illo e al., 2008). The model is based on he use o a mo ing analysis
window o ob ain a dynamic es ima e o bea salience. Fo each analysis
window, he signal is fi s passed h ough a gamma one fil e bank, a e
which he onse s a e de ec ed o each equency channel using ull-wa e
ec ifica ion, low-pass fil e ing, ime di e en ia ion and hal -wa e
ec ifica ion. Following his, he au oco ela ion unc ion o he en e-
lope is calcula ed o each equency channel and hese ob ained unc-
ions a e summed ac oss channels o es ima e pe iodici ies p esen in i s.
Subsequen ly, he summa y au oco ela ion unc ion is mul iplied by a
esonance cu e (Toi iainen and Snyde , 2003), wi h he maximum a
he lag o 500 ms, co esponding o he p e e ed empo a he icini y o
120 bpm. Finally, ins an aneous bea salience is es ima ed as he heigh
o he highes peak in he au oco ela ion unc ion. The me hod is
schema ically p esen ed in Fig. 1. Bea salience depends on se e al
musical pa ame e s, including hy hmic complexi y, imb e, and empo.
High bea salience is e oked by music ha con ains ones wi h sha p
a acks occu ing egula ly a a es close o 120 bpm, while music ha
con ains slow a acks o syncopa ion, o is played a low o high empo,
yields a lowe bea salience. The p esen bea salience model has been
alida ed using a a ing s udy wi h 100 musical excep s ep esen ing a
ange o di e en musical s yles (La illo e al., 2008). The pe cep ual
da a showed high in e -subjec consis ency and he model could p edic
hem wi h high accu acy. This sugges s ha pe cei ed bea salience o
he p esen s imulus can be p edic ed by he p esen model.
In he p esen s udy we es ima ed he s imulus’bea salience using a
3-s window and a 1-s hop. These ob ained ime se ies had he mean alue
o 0.25 and s d o 0.11. Subsequen ly, he ob ained ime se ies was
con ol ed wi h canonical double-gamma hemodynamic esponse unc-
ion and downsampled o con o m wi h he sampling a e o he MRI
da a. The o iginal and con ol ed ime se ies a e p esen ed in Fig. 2.
In addi ion o bea salience, 24 o he musical ea u es we e ex ac ed,
ep esen ing a ange o musical dimensions. Fo a ull desc ip ion o hese
a iables, see Supplemen a y ma e ial. None o hese ea u es co ela ed
significan ly wi h bea salience, sugges ing he absence o con ounding
ac o s in o he musical dimensions.
2.3. Cen ali y analysis
Pa cella ion: To educe he compu a ional bu den, he da a o each
pa icipan was spa ially downsampled by a linea ac o o wo, yielding
28,542 cubical oxels wi h 4-mm edges. The connec i i y analysis used
in his s udy is schema ically p esen ed in Fig. 3.
Windowing: Fo dynamic analysis o connec i i y, we employed a
sliding window app oach, which is he mos commonly used app oach o
es ima e dynamic unc ional connec i i y (P e i e al., 2017). To his end,
each pa icipan ’s da a was windowed using a mo ing Gaussian window
wi h a s anda d de ia ion o 10 scan in e als (20 s) and a hop size o one
scan in e al. This window has he e ec i e wid h o app ox. 31 secs
Fig. 1. Bea salience es ima ion me hod used in
he p esen s udy: (a) audio signal; (b) windowed
exce p ; (c) ou pu o gamma one fil e bank; (d)
onse de ec ion cu es pe equency channel; (e)
summa y au oco ela ion unc ion o onse
de ec ion unc ions; ( ) au oco ela ion unc ion
mul iplied by a esonance cu e (Toi iainen and
Snyde , 2003). The la e is displayed as a do ed
line. Ins an aneous bea salience (BS) is es ima ed
as he heigh o he maximum o he au oco e-
la ion unc ion.
P. Toi iainen e al. Neu oImage 216 (2020) 116191
4

Fig. 2. Top: bea salience es ima ed om he s imulus used in he s udy. Bo om: bea salience con ol ed wi h hemodynamic esponse unc ion. Uni s o he y axis a e
a bi a y and hence a e ma ked as low and high.
Fig. 3. Schema ic ep esen a ion o he dynamic
cen ali y analysis: (a) BOLD ime se ies da a o
oxels i ¼1, …,V, T ime poin s, oxel size
444mm
3
; (b) Gaussian window cen e ed a
ime poin
τ
; (c) windowed BOLD ime se ies
da a; (d) ins an aneous non-nega i e unc ional
connec i i y ma ix a ime poin
τ
, ob ained
using Spea man co ela ion
ρ
; (e) fi s eigen ec o
o he connec i i y ma ix, ob ained by sol ing
he eigenequa ion, ep esen ing cen ali y map a
ime poin
τ
; ( ) eigen ec o cen ali y ime se ies
ob ained by ca ying ou s eps b-e o all ime
poin s.
P. Toi iainen e al. Neu oImage 216 (2020) 116191
5
(Leona di &Van De Ville, 2015), which lies wi hin he ange o 30–60 s
ecommended by P e i e al. (2017). A Gaussian window was used
because o i s supe io spec al selec i i y o e a ec angula one (Leo-
na di &Van De Ville, 2015). The lowe end o he ecommended ange
was used, because he ecommenda ion is based on es ing-s a e da a,
and i can be assumed ha connec i i y dynamics wi h he p esen
s imula ion a e as e han wi h es ing s a e. Fo each window loca ion, a
unc ional connec i i y ma ix was es ima ed by calcula ing he
Spea man co ela ion be ween each pai o windowed ime se ies and
adding uni y o each elemen o make he ma ix nonnega i e.
Eigen ec o cen ali y mapping: Fo each o he hus ob ained ma ices,
he fi s eigen ec o was ob ained by powe i e a ion me hod ( on Mises,
1929). Fo each pa icipan , his yielded a se ies o EC maps co e-
sponding o each o he imepoin s in he da a.
Co ela ion wi h eg esso : To assess he ex en o which he dynamic
EC o each oxel was modula ed by he bea salience o he musical
s imulus, he EC ime se ies we e co ela ed wi h he bea salience ime
se ies using Spea man co ela ion. Fo each pa icipan , his yielded a
spa ial map indica ing he s eng h and di ec ion o modula ion a each
oxel. Fo significance es ima ion o he hus ob ained co ela ion maps,
a Mon e Ca lo simula ion was ca ied ou in which he cen ali y and
co ela ion analysis was applied o BOLD ime se ies phase-sc ambled
wi h FFT. This was epea ed 10,000 imes o ob ain an empi ical dis i-
bu ion o co ela ion alues, which was hen used o con e he co e-
la ion o z s a is ics. Subsequen ly, hese maps will be e e ed o as he
Cen ali y Modula ion Maps (CMM).
Second-le el analysis: Fo g oup-le el in e ence, he pa icipan -le el
CMMs we e pooled using S ou e ’s Z-sco e me hod (S ou e , S.,
DeVinney, L. &Suchmen, 1949). This yielded a g oup-le el CMM o z
s a is ics. Subsequen ly, wo ypes o analysis we e pe o med. Fo
ne wo k-le el analysis, he g oup-le el CMM was co ela ed wi h
es ing-s a e ne wo ks. Fo egional analysis, he map was h esholded,
and co ec ion o mul iple compa isons was pe o med wi h clus e -size
h esholding, whe e he espec i e h esholds we e ob ained using a null
dis ibu ion ob ained wi h a pe mu a ion es . Specifically, we pe o med
10.000 i e a ions, in which we co ela ed he g oup-le el CMM wi h a
phase-sc ambled e sion o he bea salience eg esso , pooled and
h esholded he hus ob ained map and kep he maximal clus e size.
2.4. Psychophysiological in e ac ions analysis
Since cen ali y analysis as such does no p o ide any di ec measu e
o in e connec i i y be ween a eas in he b ain and how his migh be
modula ed by bea salience, subsequen Psychophysiological In e ac ions
(PPI) analyses we e ca ied ou . PPI e alua es how connec i i y be ween
wo b ain a eas depends on an ex e nal a iable, ha is, how he p es-
ence o absence o an ex e nal a iable modula es he connec i i y. To
his end, i employs a mul iple eg ession model ha p edic s he ime-
cou se o a a ge egion by he ime-cou se o a seed egion, an
ex e nal a iable, and he p oduc he eo . The be a weigh s o he
p oduc a iable a e assumed o indica e he s eng h o connec i i y
modula ion by he ex e nal a iable ( o mo e de ails, see (F is on e al.,
1997)). As audi o y-mo o connec i i y was o key in e es in he p esen
s udy, we es ic ed his analysis o egions associa ed wi h he p o-
cessing o musical bea belonging o he
ce ebello- halamic-s ia o-co ical mo o pa hway, specifically he mo o
and audi o y co ices, basal ganglia, halamus, and ce ebellum. The seeds
employed in he PPI analyses we e selec ed in a da a-d i en ashion
based on he g oup CMM ob ained om he EC analysis. Subsequen ly,
o each seed eg esso , PPI analysis was pe o med wi h all oxels wi hin
he mo o co ex (MC), audi o y co ex (AC), basal ganglia (BG), hal-
amus (TH), and ce ebellum (CE). Fi e sepa a e PPI analyses we e hus
ca ied ou , in each o which he a ge egion comp ised he union o he
fi e a o emen ioned egions. Significance o pa icipan -le el PPI maps
was es ima ed wi h an empi ical dis ibu ion ob ained om a Mon e
Ca lo simula ion based on FFT phase-sc ambling o he bea salience ime
se ies. Pooling o he maps and clus e -size h esholding we e ca ied ou
ollowing he p ocedu e desc ibed in chap e 2.4.2.
3. Resul s
3.1. EC analysis
The CMM was h esholded a a significance le el o p <.01 ( wo-
ailed, clus e co ec ed a FWE<0.01; clus e size h eshold 100 oxels).
Resul s a e p esen ed in Fig. 4 and Table 1a–b. As can be seen, high bea
salience was associa ed wi h significan ly inc eased cen ali y in se e al
a eas o he DMN, including igh on al a eas, igh angula gy us, and
bila e al p ecuneus. Low bea salience, in u n, was associa ed wi h
significan ly inc eased cen ali y in mo o a eas, including p ima y
mo o co ex, p emo o co ex, and supplemen a y mo o a ea, in igh
audi o y a eas, middle occipi al gy us, as well as in mo o a eas o he
ce ebellum.
To assess he deg ee o which he significan ly co ela ed egions
o e lap wi h es ing-s a e ne wo ks, he b ain was pa cella ed acco ding
o he es ing-s a e ne wo ks epo ed in Damoiseaux e al. (2006) by
assigning each oxel o he ne wo k wi h highes loading and calcula ing,
o posi i ely and nega i ely co ela ing clus e s sepa a ely, he p opo -
ion o hei o e lap wi h each o he ne wo ks. As can be seen in Fig. 5,
highes o e lap o posi i ely co ela ing clus e s was ound wi h DMN
(48%) and igh cen al execu i e ne wo k (33%), and ha o nega i ely
co ela ing clus e s o senso imo o ne wo k (38%) and audi o y
ne wo k (15%).
3.2. Psychophysiological in e ac ions analysis
To in es iga e in de ail he modula o y e ec o bea salience on
unc ional connec i i y be ween and wi hin b ain a eas p e iously ound
o be associa ed wi h musical bea p ocessing, we ca ied ou Psycho-
physiological In e ac ions (PPI) analyses in he mo o co ex (MC),
audi o y co ex (AC), basal ganglia (BG), halamus (TH), and ce ebellum
(CE), wi h he bea salience ime se ies as he psychological eg esso . Fo
seeds we selec ed oxels ha displayed highes inc ease in EC du ing low
bea salience in each o he a o emen ioned a eas. This esul ed in fi e
seed esiding in he le SMA ([-14 -10 64]), igh STG ([58 -14 8]), le
cauda e ([-10 0 12]), halamus ([-10 -4 12]) and igh lobule VI ([34–70
-20]). Fo he pu pose o he PPI analyses, each seed eg esso was ob-
ained as he mean ime se ies o ac oss oxels wi hin a 6-mm adius om
he espec i e seed. Fo each seed eg esso , PPI analysis was pe o med
wi h all oxels wi hin MC, AC, BG, TH, and CE.
The g oup-le el PPI maps o each seed egion we e h esholded a a
significance le el o p <.01 ( wo- ailed, clus e co ec ed a FWE<0.01;
clus e size h eshold 264 mm
3
). To p o ide an o e iew o how bea
salience modula ed o e all unc ional connec i i y wi hin and be ween
he fi e egions o in e es , Fig. 6 displays he s eng h o in e ac ion o
each pai o seed and a ge egion, o bo h nega i e (high connec i i y
du ing low bea salience) and posi i e (high connec i i y du ing high
bea salience) in e ac ion. As can be seen, du ing low bea salience
connec i i y om he AC seed inc eased o a g ea p opo ion o he MC,
and o a somewha lesse deg ee o AC, TH, and CE. Fo high bea
salience, we obse ed a di e en pa e n. Mos no ably, connec i i y
wi hin AC and CE inc eased, and he e was significan connec i i y in-
c ease om he TH seed o MC, AC, and CE, and om he BG seed o AC
and CE.
High bea salience: Clus e s whose unc ional connec i i y wi h he
selec ed seeds mani es ed inc eased unc ional connec i i y du ing high
bea salience a e shown in Table 2a (a he end o he manusc ip ). As can
be seen, mos significan inc ease in unc ional connec i i y was
obse ed wi hin he AC and CE egions (z ¼5.7, z ¼5.0, espec i ely) as
well as om he TH and BG seeds o igh SMA (z ¼4.4, z ¼3.9,
espec i ely). O e all, hese esul s sugges ha unc ional connec i i y
be ween audi o y a eas and senso imo o co ex ends o inc ease when
P. Toi iainen e al. Neu oImage 216 (2020) 116191
6
music has high bea salience.
Low bea salience: Clus e s whose unc ional connec i i y wi h he
selec ed seeds mani es ed inc eased unc ional connec i i y du ing low
bea salience a e shown in Table 2b (a he end o he manusc ip ). As can
be seen, o low bea salience mos significan inc ease in unc ional
connec i i y was ound om he AC seed (R STG) o le PoG (z ¼5.8)
and le STG (z ¼5.7), as well as om he MC seed (L SMA) o le PoG
(z ¼3.5). Mo eo e , g ea e connec i i y was obse ed be ween he
igh AC seed and he le STG (z ¼5.7; n¼138) du ing low bea
salience whe eas high bea salience was associa ed wi hin he igh STG
(z ¼5.7; n¼252) wi h a educed e ec in he le STG (z ¼4.2; n¼71).
4. Discussion
Using dynamic Eigen ec o cen ali y analysis, we he e showed ha
musical bea a ec s he b ain ne wo ks’cen ali y pa e n du ing music
lis ening, wi h high bea salience inc easing cen ali y in he De aul
Mode Ne wo k, and low bea salience inc easing cen ali y in he mo o
and audi o y co ices. Subsequen Psychophysiological In e ac ions An-
alyses wi h seeds selec ed om he Eigen ec o cen ali y analysis esul s
showed ha unc ional connec i i y be ween he mo o and audi o y
co ices ends o inc ease du ing momen s o low bea salience. This is he
fi s s udy o employ dynamic whole-b ain unc ional connec i i y
analysis wi h na u alis ic musical s imula ion. The analysis showed ha
unc ional connec i i y om he basal ganglia and halamus o senso i-
mo o and audi o y co ices as well as ce ebellum ends o inc ease
du ing high bea salience. Finally, he esul s showed inc eased con-
nec i i y wi hin ce ebellum du ing momen s o high bea salience.
4.1. Dynamic g aph analysis
As can be obse ed in he esul s, low bea salience in he s imulus
Fig. 4. Regions wi h significan ly inc eased cen ali y du ing (a) high and (b) low bea salience (p <.01, wo- ailed; clus e -size co ec ed a FWE 0.01; clus e -size
h eshold 800 mm
3
).
Table 1a
Regions wi h inc eased cen ali y du ing high bea salience (p <.01, wo- ailed,
FWE<0.01 CS ¼100).
Clus e # L/
R
egion N MNI
(mm)
max(Z) BA addi ional
egions
1 R MFG 211 34,34,44 4.05 9 IFG, SFG
2 R AG 188 34,14,34 4.01 39
3 R IFG 185 6,42,-8 4.14 11 L ACG, R GR,
ACG
4 R P ec 131 2,-54,48 3.93 7 L P ec, MCG
5 R P ec 125 6,-58,24 3.71 23 L P ec, Cun
P. Toi iainen e al. Neu oImage 216 (2020) 116191
7
was associa ed wi h inc eased cen ali y in ce ebellum and co ical
audi o y-mo o egions while he key hubs o he DMN exhibi ed high
cen ali y du ing high bea salience. Specifically, main aining pe cei ed
bea du ing musical momen s cha ac e ized by high bea salience was
ound o mani es as inc eased cen ali y in egions belonging o he DMN
specifically in he bila e al p ecuneus, egions su ounding he igh
angula gy us, MPFC, and bila e al pos e io cingula e gy us. Inc ease in
ac i i y in hese egions has been p e iously implica ed in dec eased
audi o y-cogni i e load (Allu i e al., 2012). Indeed, momen s o high
bea salience can be assumed no o equi e a en i e bea finding and
hence be conside ed as low load condi ions, implying inc eased
mind-wande ing. Ga cía-Ga cía e al. (2015) obse ed inc ease in cen-
ali y o he p ecuneus and angula gy us o he DMN in es condi ions
e sus ask-based condi ions ( isual) which we e associa ed wi h
inc eased cen ali y in isual egions. This esul is in acco dance wi h
he p esen findings, assuming ha bea in e ence is conside ed o be a
ask which equi es a en i e lis ening, and bea main enance, in con as
o bea in e ence, o be a es condi ion.
On he o he hand, inc eased cen ali y du ing low bea salience was
ound in p emo o co ex, SMA, ce ebellum in conjunc ion wi h audi o y
associa ion a eas, sugges ing ha bea in e ence migh be associa ed wi h
inc eased global connec i i y in hese a eas. These esul s a e in line wi h
se e al s udies ha indica e he ole o co ical mo o egions in ac i e
bea finding (Chen e al., 2008;G ahn and Rowe, 2009;Kung e al.,
2013). Specifically, he SMA has been implica ed o be c i ical in in e -
nal/co e mo emen gene a ion, which equi es en aining o a hy hm
(Jenkins e al., 2000;Wu and Halle , 2013). Penhune e al. (1998)
demons a ed he impo ance o he igh STG as a locus in sel -cued
hy hm ep oduc ion om audi o y sequence/sequen ial empo al
memo y (Penhune e al., 1999). Fu he mo e, he pos e io ce ebellum
has been implica ed in iming pe cep ion and adjus men o senso y
e en s ia in e nal p edic ion models (Doya, 2000). Howe e , he ole o
he ce ebellum and basal ganglia in iming pe cep ion has been deba ed
upon.
The e was no e idence o significan co ela ion be ween cen ali y in
he basal ganglia and bea salience, which p obably indica es i s ole bo h
in bea main enance and bea finding. In line wi h his, Kung e al. (2013)
posi ha BG is ac i e in bo h bea finding and bea main enance/ apping
o he bea . Se e al s udies ha e indica ed he BG’s ole in bea p o-
cessing when he bea is highly salien and/o in he “p epa a ion o
p oduc ion o well-lea ned mo o esponses”(Kung e al., 2013). Teki
e al. (2011) demons a e he exis ence o wo dissocia i e ne wo ks
unde lying absolu e and ela i e ime keeping: one wi h he ce ebellum
as a ocal poin /p ecision clock ha media es absolu e iming; he o he a
s ia o- halamo-co ical ne wo k in ol ed in bea -based ime keeping o
Table 1b
Regions wi h inc eased cen ali y du ing low bea salience (p <.01, wo- ailed, FWE<0.01 CS ¼100).
Clus e # L/R egion N MNI (mm) max(Z) BA addi ional egions
1 L PoG 605 18,-34,72 4.61 9 PCL,SPG,P G,SMA,SFG,IPG
2 L PoG 194 50,-18,48 4.86 4 IPG
3 L P G 156 38,-10,60 4.66 6
4 L IFG 141 62,10,12 4.52 6 RO,STG
5 L SOG 118 10,-94,20 3.42 18 Cun, MOG
6 L IFG 113 50,22,4 3.70 45
7 R P G 356 30,.22,60 4.21 6 SMA,PoG,SFG
8 R IOG 306 46,-78,-16 4.95 19 Lob VI, C I-II,FusG
9 R STG 180 58,-14,8 4.05 RO,SMG
Fig. 5. P opo ion o o e lap be ween significan clus e s co ela ing posi i ely
and nega i ely wi h bea salience, and es ing s a e ne wo ks. DMN –de aul
mode ne wo k, AN –audi o y ne wo k, CBLN –ce ebella ne wo k, SN –
salience ne wo k, CEN VR – igh cen al execu i e ne wo k, CEN (L) –le
cen al execu i e ne wo k, SM –senso imo o ne wo k, VIS (med) –medial
isual ne wo k, VIS (la ) –la e al isual ne wo k, VIS (occ) –occipi al isual.
Fig. 6. S eng h o in e ac ion (maximal z s a is ic) o each pai o seed and a ge egion, o bo h nega i e (high connec i i y du ing low bea salience) and posi i e
(high connec i i y du ing high bea salience) in e ac ion. MC –mo o co ex, AC –audi o y co ex, BG –basal ganglia, TH – halamus, CE –ce ebellum.
P. Toi iainen e al. Neu oImage 216 (2020) 116191
8