Multiple subregions in superior temporal cortex are differentially sensitive to vocal expressions: A quantitative meta-analysis

(1)

Article

Reference

Multiple subregions in superior temporal cortex are differentially sensitive to vocal expressions: A quantitative meta-analysis

FRÜHHOLZ, Sascha, GRANDJEAN, Didier Maurice

Abstract

Vocal expressions of emotions consistently activate regions in the superior temporal cortex (STC), including regions in the primary and secondary auditory cortex (AC). Studies usually report broadly extended functional activations in response to vocal expressions, with considerable variation in peak locations across several auditory subregions. This might suggest different and distributed functional roles across these subregions instead of a uniform role for the decoding of vocal emotions. We reviewed recent studies and conducted an activation likelihood estimation meta-analysis summarizing recent fMRI and PET studies dealing with the processing of vocal expressions in the STC and AC. We included two stimulus-specific factors (paraverbal/nonverbal expression, stimulus valence) and one task-specific factor (attentional focus) in the analysis. These factors considerably influenced whether functional activity was located in the AC or STC (influence of valence and attentional focus), the laterality of activations (influence of paraverbal/nonverbal expressions), and the anterior-posterior location of STC activity (influence of [...]

FRÜHHOLZ, Sascha, GRANDJEAN, Didier Maurice. Multiple subregions in superior temporal cortex are differentially sensitive to vocal expressions: A quantitative meta-analysis.

Neuroscience and Biobehavioral Reviews , 2013, vol. 37, no. 1, p. 24-35

DOI : 10.1016/j.neubiorev.2012.11.002 PMID : 23153796

Available at:

http://archive-ouverte.unige.ch/unige:84144

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

ContentslistsavailableatSciVerseScienceDirect

Neuroscience and Biobehavioral Reviews

j o u r n al hom ep age :w w w . e l s e v i e r . c o m / l o c a t e / n e u b i o r e v

Review

Multiple subregions in superior temporal cortex are differentially sensitive to vocal expressions: A quantitative meta-analysis

Sascha Frühholz

^a,b,∗

, Didier Grandjean

^a,b

aNeuroscienceofEmotionandAffectiveDynamics(NEAD)Laboratory,DepartmentofPsychology,UniversityofGeneva,40bdduPontd’Arve,CH-1205Geneva,Switzerland

bSwissCenterforAffectiveSciences,UniversityofGeneva,Geneva,Switzerland

a r t i c l e i n f o

Articlehistory:

Received6August2012

Receivedinrevisedform8October2012 Accepted4November2012

Keywords:

Vocalexpressions Voice

Prosody

Paraverbal/nonverbalexpressions Valence

Attentionalfocus

a b s t r a c t

Vocalexpressionsofemotionsconsistentlyactivateregionsinthesuperiortemporalcortex(STC),includ- ingregionsintheprimaryandsecondaryauditorycortex(AC).Studiesusuallyreportbroadlyextended functionalactivationsinresponsetovocalexpressions,withconsiderablevariationinpeaklocations acrossseveralauditorysubregions.Thismightsuggestdifferentanddistributedfunctionalrolesacross thesesubregionsinsteadofauniformroleforthedecodingofvocalemotions.Wereviewedrecentstudies andconductedanactivationlikelihoodestimationmeta-analysissummarizingrecentfMRIandPETstud- iesdealingwiththeprocessingofvocalexpressionsintheSTCandAC.Weincludedtwostimulus-specific factors(paraverbal/nonverbalexpression,stimulusvalence)andonetask-specificfactor(attentional focus)intheanalysis.Thesefactorsconsiderablyinfluencedwhetherfunctionalactivitywaslocatedin theACorSTC(influenceofvalenceandattentionalfocus),thelateralityofactivations(influenceofpar- averbal/nonverbalexpressions),andtheanterior-posteriorlocationofSTCactivity(influenceofvalence).

Thesedatasuggestdistributedfunctionalrolesandadifferentiatednetworkofauditorysubregionsin responsetovocalexpressions.

Contents

1. Introduction... 25

2. Materialsandmethods. . ... 26

2.1. Literaturesearchanddatacollection... 26

2.2. Dataset... 26

2.3. Dataanalysis... 26

3. Results... 28

3.1. Paraverbalandnonverbalemotionalvocalexpressions... 29

3.2. Positiveandnegativevocalexpressions... 29

3.3. Explicitandimplicitprocessingofemotionalvocalexpressions... 29

4. Discussion... 31

4.1. Themediumofemotionalvocalexpressionsinﬂuencesthelateralityofactivity... 32

4.2. Valenceinﬂuenceslower-levelandhigher-levelauditorycorticalactivity... 32

4.3. Thearousallevelofvocalexpressions... 33

4.4. Attentionalfocielicitdifferentactivationsinlow-levelauditoryregions... 33

5. Conclusions... 34

Acknowledgments ... 34

References... 34

∗Correspondingauthorat:NeuroscienceofEmotionandAffectiveDynamics(NEAD)Laboratory,DepartmentofPsychology,UniversityofGeneva,40bdduPontd’Arve, CH-1205Geneve,Switzerland.Tel.:+41223799455;fax:+41223799219.

E-mailaddresses:fruehholz@gmail.com,sascha.fruehholz@unige.ch(S.Frühholz).

http://dx.doi.org/10.1016/j.neubiorev.2012.11.002

(3)

1. Introduction

Emotionalstatescanconsiderablyinﬂuencethetoneofavoice bymodifyingthedifferentcomponentsofthevocaltractduring vocalexpressions,fromwhichindividualsinferemotionalstates inotherpersons(BanseandScherer,1996).Therefore,emotional featuresofvocalintonationsareimportantforsocialinteractions.

Aprioritizeddecodingoftheseemotionalfeaturesbythehuman brainshouldallowfastandadaptivebehaviorinaccordancewith thesocialcontext.Regionsinthebilateralsuperiortemporalcortex (STC)ofthehumanbraincomprisingthesuperiortemporalgyrus (STG)andthemiddletemporalgyrus(MTG),aswellasthesupe- riortemporalsulcus(STS),aregenerallysensitivetohumanvoices compared withotherenvironmentalsounds(Belinet al.,2000;

Schirmeretal.,2012).Furthermore,subregionswithinthisvoice- sensitivecortexarealsomoreresponsivetotheemotionalthan totheneutraltoneofavoice(i.e.Bachetal.,2008;Ethoferetal., 2009b;2012;Grandjeanetal.,2005;Sanderetal.,2005).These emotion-sensitiveareasaretypicallylocatedinthemid-posterior STC.STCactivityinresponsetovocalexpressionsisaconsistent ﬁndingacrossmanystudies(Bachetal.,2008;Frühholzetal.,2012;

Grandjeanetal.,2005;Sanderetal.,2005),leadingtotheproposal ofageneral“emotionalvoicearea”(EVA)locatedintheSTC(Ethofer etal.,2012).TheEVAmighthaveageneralfunctionalroleforthe decodingofemotionalcuesfromvocalexpressions.

Thought tobe locatedin the posterior STC (pSTC), the EVA notonlycomprisesareasinthehigher-levelauditorycortex(AC), but also areas in the primary and secondary AC, especially in therighthemisphere(Ethoferetal.,2012).Severalstudieshave demonstratedthisbroadextensionofemotion-sensitivecortical regions(Beaucousinetal.,2007;Leitmanetal.,2010b;Sanderetal., 2005;Wildgruberetal.,2004),whichcoversregionsacrossdiffer- entlevelsofauditoryprocessing.Thisﬁndingmightnotsuggest auniqueand generalrole oftheemotion-sensitiveSTCand AC, butratheradistributionoffunctionalrolesacrossthesedifferent areas.Lower-levelauditorycorticesmightdecodeacousticprop- ertiesofvocalexpression,whereashigher-levelauditoryregions mightsubserveasensory-integrativerole,suchasbuildingupper- ceptsofvocalexpressionbasedonpre-processedauditoryfeatures and representingan acousticobject (Schirmer and Kotz,2006).

Thisassumptionofafunctionaldistributionisalsoindicatedacross studiesthatshowaconsiderablevariationofpeakactivationsin responsetovocalemotionalexpressions.Thesestudiesreportpeak activationsforemotionalcomparedwithneutralvocalexpressions thatrangefromthebilateralpSTC(Beaucousinetal.,2007;Ethofer et al., 2006a, 2009a;Mitchell et al., 2003; Sanderet al., 2005;

Wildgruberetal.,2005)andbilateralmid-STC(mSTC;Bachetal., 2008;Ethoferetal.,2006b,2007,2009a;Grandjeanetal.,2005;

Leitmanetal.,2010b;Sanderetal.,2005;Wiethoffetal.,2008)to therightanteriorSTG(aSTC;Bachetal.,2008).Studiesalsoreport peakactivationsintheprimaryandsecondaryACinresponseto vocalexpressions(Fecteauetal.,2007;Leitmanetal.,2010b;Meyer etal.,2005;Phillipsetal.,1998;Sanderetal.,2005;Warrenetal., 2006a;Wildgruberetal.,2005).

Together, thebroad extension of emotion-sensitive areas in theSTCandACandthevariationinpeakactivitymightsuggest different and distributed functional roles rather than a unique functionalroleforalloftheseregions.Specifically,theactivityof thesedifferentregionsandtheirfunctionalrolesmightdependon severalstimulus-as wellastask-specificmodalities, which fig- ureasbottom-upandtop-downfactors,respectively,duringthe processingofvocalexpression.Withrespecttotop-downfactors, onefrequentlyinvestigatedfactorduringtheprocessingofvocal expressionsisattentionalfocus,sometimesreferredtoasthelevel ofprocessingofvocalexpressions (Bach etal.,2008;Grandjean etal.,2005;Sanderetal.,2005).Theemotionalcueofavoicecan

bepresentedinsidethefocusofattention(explicitprocessingof emotion)when participantsareaskedtoexplicitlydiscriminate, classify,orevaluatetheemotional toneof avoice(Alba-Ferrara etal.,2011;Beaucousinetal.,2007;Ethoferetal.,2006a,2009a;

Frühholzetal.,2012;Grandjeanetal.,2005;Johnstoneetal.,2006;

Kotzetal.,2003;Leitmanetal.,2010b;Mitchell,2006;Sanderetal., 2005;Szameitatetal.,2010;Wildgruberetal.,2005;Wittfothetal., 2010).Thisemotionalcuecanalsobepresentedoutsidethefocus ofattention(implicitprocessingofemotion)whenparticipantsare, forexample,askedtodiscriminatethegenderofonevoicefrom another(Alba-Ferraraetal.,2011;Bachetal.,2008;Ethoferetal., 2012;Fecteauetal.,2007;Frühholzetal.,2012;Morrisetal.,1999;

Mothes-Laschetal.,2011)ortomakealinguisticdecisionabout anonemotionallinguisticfeature(Buchananetal.,2000;Ethofer et al., 2006a;Mitchell et al., 2003), or when emotional voices are presented outside thespatialfocus of attention(Grandjean etal.,2005;Sanderetal.,2005).Forthelatter,Grandjeanandcol- leagues(2005),forexample,haveshownthataregionintheright pSTCissensitivetovocalexpressionsindependentofattentional focus,whereas anadjacentregionin therightmSTC isrespon- sivetovocalexpressiononlywhentheemotionalcueisinsidethe focusofattention.Werecentlyreplicated thisﬁndingbyshow- ingthattherightmSTCwasactiveonlywhentheemotionalcue wasinsidethefocusofattention,whereasregionsintherightpSTC weregenerally active duringtheprocessing of vocalemotional expressions(Frühholzetal.,2012).Althoughthesestudiesshow adependencyofSTCactivityinresponsetovocalexpressions,the dependencyofprimaryorsecondaryACactivityisnotyetknown orhasnotbeeninvestigatedsystematically.Thereare,however, reasonstobelievethatevenlow-levelACactivitymightdependon attentionalfocustowardorawayfromvocalexpressions.Studies usuallyreportprimaryandsecondaryACactivityspeciﬁcallyduring implicitattentiontowardvocalexpressions(Fecteauetal.,2007;

Phillipsetal.,1998),butsometimesalsoduringexplicitattention (Leitmanetal.,2010b).

BesidesthistaskdependencyofSTCactivityinresponsetovocal expressionsasatop-downfactor,STCresponsestovocalexpres- sionsmightalsodependonstimulus-speciﬁcfeaturesasbottom-up factors.WerecentlydemonstratedadependencyofSTCactivityon acousticfeaturessuchasthepitch andintensityofa voicedur- ingvocalemotionalexpressions(Frühholzetal.,2012).Wefound lesssensitivitytoacousticfeaturesinmSTCcomparedwithpSTC.

Besides basic acoustic features, two more important stimulus- speciﬁcfeaturesof emotionalvoicesmightbepresent.Theﬁrst featureconcerns theemotional dimensionofvocalexpressions, representedbythefactorsvalenceandarousal,whichdetermine thepleasantnessandtheexcitementofperceivedvocalemotions, respectively.Theeffectsofarousalhaveseldombeeninvestigated, but the few existing resultspoint to a bilateralSTC sensitivity tothelevelofarousalin vocalemotions(Warrenetal.,2006b;

Wiethoffetal.,2008).Withrespecttovalence,vocalexpression canbeneutralorofpositive(i.e.happy)ornegativevalence(i.e.

anger, fear,disgust).Like theeffects ofarousal,thequestionof howvocalexpressionsofdifferentvalenceelicitactivityintheSTC hasbeenrarelyinvestigated.Ethoferetal.(2009b)reportdiffer- entvoxelpatternsinthebilateralvoice-sensitivecortexintheSTC, includinghigher-levelauditoryregionsintheSTGandlower-level auditoryregionsinthetransversetemporalgyrusinresponseto differentvocalexpressions,withoutprovidingaclearpictureabout thespatialdistributionofactivationmapsacrossdifferentvocal expressions.OtherstudiesreportbilateralSTCactivityelicitedby bothpositiveandnegativevocalexpressions(Fecteauetal.,2007;

Frühholzetal.,2012;Grandjeanetal.,2005;Johnstoneetal.,2006;

Kotzetal.,2003;Leitmanetal.,2010b;Meyeretal.,2005;Mothes- Laschetal.,2011;Phillipsetal.,1998;Sanderetal.,2005;Warren etal.,2006a),butonlyintheprimaryandsecondaryACinresponse

(4)

toanegativeexpression(Fecteauetal.,2007;Kotzetal.,2003;

Leitmanet al.,2010b; Phillipset al.,1998;Sanderetal., 2005).

However,positiveexpressionscanalsosometimeselicitprimary andsecondaryACactivity,althoughsurprisinglyonlyforpositive nonverbalexpressionsratherthanforvocalexpressionssuperim- posedinspeechutterances (Meyeretal.,2005;Szameitatetal., 2010).

Thelastfindingis relatedtothesecondimportantstimulus- specificfactor,themediumofvocalexpressions,whichcoversboth nonverbalexpressionsand emotionalintonations superimposed onspeechand speech-likeverbalutterances.For thelatter,the emotionaltoneofvoicecanmodulateverbalspeechutterances, includingbothneutralandemotionalsemantics(Beaucousinetal., 2007;Ethoferetal.,2006a;Leitmanetal.,2010b),butitcanalso modulatevocalutterancesbasedonpseudolanguageaswellason interjectionsthatsoundlikespeechbutaredevoidofanyseman- ticmeaning(Bachetal.,2008;Dietrichetal.,2008;Frühholzetal., 2012;Grandjeanetal.,2005;Sanderetal.,2005).Thesespeech- based intonations are usually referred to as emotional speech prosodyandgenerallyrepresentaparaverbalemotionalcue.We therefore subsume these cases under the category of paraverbal expressions. Unlike these speech or speech-like paraverbal expressionofemotions,vocalexpressioncanalsobethenonver- balexpressionofemotionalstates(Dietrichetal.,2007;Fecteau etal.,2007;Meyeretal.,2005;Morrisetal.,1999;Phillipsetal., 1998;Szameitatetal.,2010;Warrenetal.,2006a).Thisissueof themediumasrepresentedbythedistinctionbetweenparaver- balandnonverbalexpressionsseemsimportantintermsoflower- andhigher-levelactivityintheAC.Thetypeofmediumcaninflu- encetheacousticrichnessanddiversityofacousticandemotional cuesandcanchangethewayweextracttheemotionalvaluefrom thesedifferenttypesofvocalemotional expressions.Nonverbal expressionsseemtobedecodedaccordingtoalimitedrangeof acousticfeatures,suchasthepitchorthefundamentalfrequency (f0),thespectralcenterofgravity,andthenumberofamplitude onsets(Sauteretal.,2010).Paraverbalexpressionsareclassified anddiscriminatedaccordingtoseveralfeaturesofthef0,theinten- sity,andthespectralandtemporalproperties(BanseandScherer, 1996;JuslinandLaukka,2003;Leitmanetal.,2010a;Pateletal., 2011).Thediversityofacousticfeaturesfromwhichtheemotional toneofaparaverbalexpressionisdecodedanddiscriminatedmight involvestrongerprimaryandsecondaryACactivity.Anotherfac- torconcerning paraverbal and nonverbal expressions might be thelateralityofSTCactivity,withtheassumptionthatparaver- balexpressionsmightmorestronglyactivatethelefthemisphere, whilenonverbalmediumsmightmorestronglyactivatetheright hemisphere.Theonlyrecentstudyprovidesmixedandinconsistent evidenceforthisassumption,specificallyinrelationtoSTCactivity (Kotzetal.,2003).

Takentogether,theseﬁndingssuggestthat,insteadofaunique emotion-sensitiveareaintheSTCandACwithageneralroledur- ingtheprocessingofvocalexpressions,severalsubregionsmight bepresentwithdifferentfunctionalrolesduringthedecodingof vocalexpressions.Thelocationoffunctionalactivities,aswellas thefunctionalroleofthesesubregions,mightdependontop-down andbottom-upfactorsduringtheprocessingofvocalexpressions.

Althoughrecentstudiesonvocalexpressionsprovideevidencefor afunctionaldistribution,thisissuehasnotyetbeeninvestigated systematically.Thecurrentmeta-analysisthereforeaimedatdeter- miningthedependencyanddistributionoffunctionalactivityinthe STCandACinresponsetovocalexpressionsbasedontheattentional focusasatop-downfactor,aswellasonstimulusvalence(negative, positive)andstimulusmedium(paraverbalexpressions,nonverbal expressions)asbottom-upfactors,duringtheprocessingofvocal emotions.Wedidnotincludestimulusarousalasanotherfactor inthismeta-analysisbecauseofaverysmallamountofactivation

focireportedinpreviousexperiments(seeSection2.3).Wecol- lectedpeakactivations across29functionalmagneticresonance imaging(fMRI)andpositronemissiontomography(PET)studies, including128peakactivationsthatdealtwiththeprocessingof vocalemotionalexpressionsandtheirinﬂuenceonbilateralSTC activity.

2. Materialsandmethods

2.1. Literaturesearchanddatacollection

ToﬁndbothfMRIandPETstudiesdealingwithvocalexpressions ofemotions,wethoroughlysearchedtheBrainMapdatabaseusing Sleuth(http://brainmap.org/sleuth/;seeLairdetal.,2005),aswell asPubMed(www.pubmed.org),usingthefollowingsearchstrings:

“vocalAND(emotionsORexpressions)”,“(emotionalORaffective ORemotion)AND(prosodyORprosodicORspeechORvoicesOR intonationORvocalization)”,“speechAND (melodyORprosody ORintonation)”,“(angerORfearORdisgustORsadORhappyOR arousal) AND prosody”,“auditory emotions.”We includedonly thosestudiesthat werepublishedinEnglishbeforeApril 2012, includedonlyhealthyyoungandmiddle-agedhumanadults,and usedauditorystimulationandascanningspacecoveringatmini- mumthewholebilateralSTCandACinafullposterior-to-anterior direction.Studieswereexcludedthatwerebasedsolelyonpre- deﬁnedregionsofinterest.Peaklocationshad tobereportedin standardstereotacticcoordinatespace(TalairachorMontrealNeu- rologicalInstitute[MNI]space).Wewerespeciﬁcallyinterestedin activationfociintheSTC,includingthelateralSTGandtheSTSand thesuperiorportionoftheMTG,aswellasintheentireplaneof theSTChiddeninthelateralsulcus.Thelatterregionsincludedthe transversetemporalgyrus,aswellastheplanumtemporale(PTe;

mostlycoveredbyBrodmannarea[BA]42)andtheplanumpolare (PPo;mostlycoveredbyBA52).Thelatterregionsrepresentpri- maryandsecondaryAC(seeFig.1foradeﬁnitionofallofthese regions).

2.2. Dataset

Thismeta-analysisdealtwiththeprocessingofvocalexpression inthehumanSTCandAC.Wethereforeexcludedallstudiesthat reportedtheprocessingofvocalexpressionsinprimateandnonpri- mateanimals,aswellasstudiesthatusedanimalvocalizationsin humanparticipants.Wealsoexcludedstudiesorpeakactivations instudiesthatcomparedvocalexpressionsonlyagainstbaseline activityinsteadofcomparingthemagainstotheremotionalorneu- tralvocal expressions. Thiswasdone toensurethat functional activitywasspeciﬁcfortheemotionaldimensioninvocalexpres- sion.Theﬁnal datasetincluded27 independentstudieswitha totalof489participantsreporting128activationfoci(seeTable1).

PleasenotethatTable1reports29studies;however,intwocases, wecombinedtheactivationfocioftwostudiesintoasinglestudy becausetheactivationfociwereacquiredwiththesamesampleof participants.AllcoordinatesweretakeninortoMNIspace.Peak activationsreportedinTalairachspaceweretransformedtoMNI spaceusingalineartransformation(Lancasteretal.,2007).

2.3. Dataanalysis

First,weenteredall128activationfociderivedfromallstudies includedinthismeta-analysisinageneralanalysistoﬁndactiva- tionlikelihoodestimation(ALE)mapsthatprovidedanestimation ofthespatialextentofemotion-sensitiveregionsintheSTCand AC,independentofstimulus-andtask-speciﬁcfactors.Inasecond approach,peakactivationswerecategorizedaccordingtothethree mainfactorsofinterestinthismeta-analysis(Table2).First,allpeak

(5)

Fig.1.(A)TheSTCandtheACasthemainregionsofinterest.TheACisdistinguishedintoprimaryAC(red)andsecondaryAC(green).TheprimaryACisrepresentedby thesubregionsTe1.0,Te1.1,andTe1.2,whicharepredominantlylocatedonthetransversetemporalgyrus.TheseACsubregionsweredeﬁnedaccordingtoMorosanetal.

(2001),asimplementedintheAnatomytoolbox(Eickhoffetal.,2005)forSPM8.ThesecondaryACislocatedanteriorandposteriortotheprimaryACandisrepresented bytheplanumpolare(PPo)andtheplanumtemporale(PTe;BA42),respectively.BA42wasdefinedaccordingtoitslocationrelativetotheprimaryauditorycortexand higher-levelauditoryregionsinBA22(seeHackett,2011).TheSTCismainlyrepresentedbyBA21and22,whichweredefinedaccordingtotheCARETatlas(VanEssenetal., 2001).ThewhitedottedlinesdenoteBAareas.Alongtheanterior-to-posteriordirection,wedefinedthesubregionsofaSTC(y>10),mSTC(10<y<−15),andpSTC(y<−15).

(B)EmotionalvocalexpressionsgenerallyelicitedextendedactivitiesintheSTCindependentofstimulusmodality,task,andemotionalvalence.Activityisrenderedonthe left(L)andright(R)corticalﬂatsurfaceofthehumanColinatlasimplementedinCARETsoftware(VanEssenetal.,2001).Abbreviations:insinsula,mtgmiddletemporal gyrus,stgsuperiortemporalgyrus,stssuperiortemporalsulcus.(Forinterpretationofthereferencestocolourinthisﬁgurelegend,thereaderisreferredtothewebversion ofthearticle.)

activations weredividedaccordingtotheiractivityin response to“paraverbal”or“nonverbal”vocalexpressions(factorstimulus medium).Paraverbalexpressionscomprisedvocalexpressionthat usedspeechbasedonpseudolanguage,aswellassemanticneu- traloremotionalspeech.Allothertypesofvocalexpressionswere classiﬁedasnonverbalvocalexpressions.Second,peakactivations wereclassiﬁedasbeingelicitedby“positive”or“negative”expressions(factorvalence).Positiveexpressionscomprisedexpressions ofhappinessorlaughter.Negativeexpressionscomprisedexpres- sionsofanger,fear,ordisgust.Onlythosepeakactivationswere includedthatcouldbeuniquelycategorizedinoneofthesecate- gories.Wedidnotincludepeakactivationsthatwerereportedto beacombinedresponsetobothpositiveandnegativeexpressions.

Unlikethestudiesthatreportedfociclassiﬁcationaccordingto valenceasoneemotionalfeatureofvocalexpressions,onlytwo studies(Warrenetal.,2006b;Wiethoffetal.,2008)reportedthree differentactivationfociforarousalastheotheremotionalfeature.

OnlyoneleftandonerightSTCfocuswerelocatedwithinaproper brainmask.ThesetwofocididnotallowavalidALEmapestimation ofthearousaleffectsofvocalemotions,andthusthefactorarousal wasnotincludedinthismeta-analysis.

Finally,weseparatedpeakactivationsaccordingtotheatten- tional focus (factor attention). Peak activations in the category

“explicitattention” weretaken fromstudiesthat askedpartici- pantstodirectly focus ontheemotional cue of a voiceand to explicitlycategorizevocalexpressionaccordingtotheirvalence.

Peakactivationsduring“implicitattention”weretakenfromstud- iesinwhichparticipantswereaskedtofocusonanonemotional stimulus feature, suchas thegender of a voice or a nonemo- tional linguistic feature of spoken speech, or when emotional expressionwaspresentedoutside thespatialfocusof attention.

Studiesthatreportedpeakactivationduringpassivelisteningor inwhich activations werereportedtobe combinedfor explicit

and implicit attentiontasks were not taken into account here.

ALEmapswerecalculated foreachofthesesixcategories (paraverbal/nonverbal, negative/positive,explicit/implicit attention).

ALE maps were also calculated for all three subcategories of verbalexpression(pseudo-language-basedspeech,semanticneutralspeech,semanticemotionalspeech),ofnegativeexpressions (anger,fear,disgust),andoftheimplicitattentioncondition(focus ongender,focusonlinguisticfeature,spatialmanipulation).How- ever, the resulting ALE maps for fear and disgust have to be consideredwithcautionbecauseonlyone study(Phillips etal., 1998)reportedafewpeakactivationsspeciﬁcallyfortheexpres- sionsoffear(n=5)anddisgust(n=10).

ALE mapswereestimated usingthealgorithm implemented in the GingerALE toolbox (http://brainmap.org), allowing a coordinate-basedmeta-analysisofneuroimagingresultsreported asactivationfoci(Eickhoffetal.,2009).Theseactivationfociare modeledasthree-dimensionalGaussiandistributionsrepresented bymodeled activation(MA)mapsofactivationlikelihoods.The number ofsubjectsin eachstudyis takeninto accounttoesti- matethespatialuncertaintyofeachactivationfocus.Subsequently, ALEmapsweregeneratedbycombiningtheMAmapsofallstud- iesintoonecategory.ALEmapsrepresentthelikelihoodofeach voxeltobeactivatedacrossthestudiesenteredintotheALEalgo- rithm.TheseALEmapsarecomparedagainstanulldistributionof therandomspatialassociationoftheMAmapsofeachstudy,as revealedbyapermutationprocedure.TheoriginalALEmapswere thentestedagainsttheALEmapsresultingfromthenulldistribu- tion.Theresultingstatisticalmapswereeachthresholdedatafalse discoveryrate(FDR)correctedp-valueof<0.05andaclustervol- umeof20mm³.Thisrelativelylowclustervolumethresholdwas chosensothatwewouldbeespeciallyabletoidentifysubregions intheSTCandACthatmightbesmallcomparedwiththebroad extentofthegeneralemotion-sensitivecortex.

(6)

Table1

Studiesincludedinthismeta-analysiswiththenumberofsubjects(N),stimulusmaterial,imagingmethodused,numberofactivationfoci,andcontrastsfromwhichfoci werederived.

Authors N Stimulus Method Foci Contrasts

Alba-Ferraraetal.(2011) 19 Neutralsemanticspeech fMRI 4 Emotionalvs.neutralvoices^c;complexemotionsvs.simpleemotions Bachetal.(2008) 16 Pseudo-language-basedspeech fMRI 5 Emotionalvs.neutralvoices^c;implicit(gender^d)vs.explicit(valence^e) Beaucousinetal.(2007) 23 Emotionalsemanticspeech fMRI 3 Explicit(valence^e)vs.implicit(linguistic^f);humanemotionalvoices

vs.syntheticemotionalvoices

Buchananetal.(2000) 10 Neutralsemanticspeech fMRI 1 Implicit(linguistic^f)vs.explicit(valence^e) Dietrichetal.(2007) 16 Nonverbalexpressions fMRI 1 Affectiveburstsvs.vocalgestures Dietrichetal.(2008) 16 Pseudo-language-basedspeech fMRI 2 Emotionalvs.neutralvoices^c

Ethoferetal.(2006a)â 24 Emotionalsemanticspeech fMRI 2 Implicit(linguistic^f)vs.explicit(valenceê);explicit(valenceê)vs.

implicit(linguistic^f)

Ethoferetal.(2006b)^a 24 Emotionalsemanticspeech fMRI 2 Increasingemotionalprosodicintensity Ethoferetal.(2007) 20 Neutralsemanticspeech fMRI 4 Emotionalvs.neutralvoices^c

Ethoferetal.(2009a) 24 Neutralsemanticspeech fMRI 3 Emotionalvs.neutralvoices^c;explicit(valence^e)vs.implicit (linguistic^f)

Ethoferetal.(2012) 22 Pseudo-language-basedspeech fMRI 2 Emotionalvs.neutralvoices^c

Fecteauetal.(2007) 14 Nonverbalexpressions fMRI 7 Emotionalvs.neutralvoices^c;positivevs.neutralvoices;negativevs.

neutralvoices;negativevs.positivevoices

Frühholzetal.(2012) 17 Pseudo-language-basedspeech fMRI 6 Angervs.neutralvoices;explicit(valence^e):angervs.neutral;implicit (gender^d):angervs.neutral

Grandjeanetal.(2005)^b 15 Pseudo-language-basedspeech fMRI 3 Angervs.neutralvoices;explicit(spatial^g)vs.implicit(nospatial) Johnstoneetal.(2006) 40 Neutralsemanticspeech fMRI 3 Happyvs.angryvoices

Kotzetal.(2003) 12 Neutralsemanticspeech fMRI 5 Normalspeech:positivevs.negativevoices;normalspeech:negative vs.neutralvoices;prosodic(delexicalized)speech:positivevs.neutral voices;prosodic(delexicalized)speech:negativevs.neutralvoices Leitmanetal.(2010b) 20 Neutralsemanticspeech fMRI 17 Emotionalvs.neutralvoices^c;emotionalvoicesandacousticcue

interaction;angervoiceandacousticcueinteraction;happy+fear voicesandcueinteraction

Meyeretal.(2005) 12 Nonverbalexpressions fMRI 2 Laughtervs.sound

Mitchelletal.(2003) 13 Emotionalsemanticspeech fMRI 2 Emotionalvs.neutralvoices^c;implicit(linguistic^f)vs.explicit (valence^e)

Mitchell(2006) 28 Emotionalsemanticspeech fMRI 6 Prosodywithcongruent;semanticvs.prosody-only;prosodywith incongruent;semanticvs.prosody-only

Morrisetal.(1999) 6 Nonverbalexpressions PET 1 Emotionalvs.neutralvoices^c Mothes-Laschetal.(2011) 24 Neutralsemanticspeech fMRI 2 Angervs.neutralvoices

Phillipsetal.(1998) 6 Nonverbalexpressions fMRI 15 Fearfulvs.mildhappyvoices;disgustvs.mildhappyvoices;fearfulvs.

disgustvoices;disgustvs.fearfulvoices

Sanderetal.(2005)^b 15 Pseudo-language-basedspeech fMRI 8 Angervs.neutralvoices;anger(spatial^g)vs.neutralvoices;anger(no spatial)vs.neutralvoices;anger(spatial^g)vs.anger(nospatial) Szameitatetal.(2010) 18 Nonverbalexpressions fMRI 6 Ticklevs.joy;ticklevs.taunt;ticklevs.(taunt+joy)

Warrenetal.(2006a,b) 20 Nonverbalexpressions fMRI 4 F-contrastbetweenemotions;correlationwithpositivevalence;

correlationwitharousal

Wiethoffetal.(2008) 24 Neutralsemanticspeech fMRI 9 Emotionalvs.neutralvoices^c;correlationwithintensity,f0,duration, arousal

Wildgruberetal.(2005) 10 Neutralsemanticspeech fMRI 1 Explicit(valence^e)vs.implicit(linguistic^f) Wittfothetal.(2010) 20 Emotionalsemanticspeech fMRI 2 Emotionalvs.neutralvoices^c

aTheactivationfociofthesetwostudieswerecombinedsoastorepresentasinglestudybecausetheresultsreportedinthesetwostudieswereacquiredwiththesame sampleofparticipants.

b Sameasabove.

c Thecontrast“emotionalvs.neutralvoices”includesthecomparisonofemotionalexpressionofdifferentvalencewithneutralexpressions.Studiesdifferintheamount andthekindofemotionalexpression,whichwereenteredinthiscontrast.

d “Gender”indicatesanimplicitgenderdiscriminationtask.

e“Valence”indicatesanexplicitvalencediscriminationtask.

f “Linguistic”indicatesanimplicitlinguisticdiscriminationtask.

g“Spatial”indicatesanexplicitprocessingofemotionalvoicesbydirectingattentiontothespatiallocationofthesevoices(operationalizedasattentionalfocustotheleft orrightear).“Nospatial”meansdirectionofattentionawayfromthespatiallocationofemotionalvoices.

ResultingALEmapswererenderedonacorticalflatsurfaceof thehumanColinatlasimplementedinCARETsoftware(VanEssen etal.,2001)toallowtheexactcorticalrepresentationinrelation totheboundariesof auditory-relatedBA21, 22, 41,and 42, as wellasaccordingtotheprimaryACsubregionsTe1.0,Te1.1,and Te1.2,asdefinedby(EconomoandKoskinas,1925).Thelatterthree regionsofinterestwerecreatedaccordingto(Morosanetal.,2001), asimplementedintheAnatomytoolbox(Eickhoffetal.,2005)for theSPMsoftware(Version8;WelcomeDepartmentofCognitive Neurology,London,UK).BA42wasdefinedaccordingtoitsrela- tivelocationtoprimaryACandBA22(seeHackett,2011).Finally, wesubdividedSTCactivityinananterior-to-posteriordirectionat approximatey-coordinateboundariesintheanteriorSTC(aSTC) (y>0),mSTC (−20>y>0),andpSTC (y>−20)(seealsoFrühholz etal.,2012;Grandjeanetal.,2005).

3. Results

Inaﬁrstanalysis,wewereinterestedingeneralactivationinthe STCelicitedbyemotionalvocalexpressionsirrespectiveofstimulus medium,stimulusvalence, andattentionaltaskcondition.Emo- tionalvocalexpressionsgenerallyelicitedextendedactivityinthe bilateraltemporal cortex(Fig.1;Table3).Thisactivityincluded regionsinthebilateralprimaryACinBA41,includingsubregions intheleftTe1.1andTe1.0andintherightTe1.0.Inthelefthemi- sphere,thisprimary ACactivityextended rostrally(BA 52)and caudally(BA42)intothesecondaryAC.Additionalgeneralactiv- ityinresponsetovocalexpressionswasrevealedinthebilateral STC(BAs21,22,38).ThisindicatesawidespreadnetworkofSTC andACregionsthataresensitivetoemotionalvocalexpressions.

However,subregionsinthisnetworkweredifferentiallysensitive

(7)

Table2

Detailsaboutthestudyandpeakvoxelclassiﬁcationofallstudiesincludedintheﬁnalmeta-analysisaccordingto(A)stimulusmedium,(B)stimulusvalence,and(C) attentionalfocus.Thetablereportsnumberofexperiments,numberofparticipants,andnumberofactivationfoci.

Experiments Participants Foci

(A)Stimulusmedium

Nonverbal 6 78 19

Paraverbal Pseudolanguage 5 86 26

Neutralsemanticspeech 10 207 49

Emotionalsemanticspeech 6 132 17

(B)Stimulusvalence

Positive 6 116 11

Negative Anger 4 76 19

Fear 1 6 5

Disgust 1 6 10

(C)Attentionalfocus

Explicit 14 286 50

Implicit Gender 8 117 33

Linguistic 3 47 3

Spatial 1 15 5

tovocalexpressions,dependingon(a)themediumofvocalexpres- sion(paraverbal,nonverbal);(b)thevalenceofvocalexpressions (positive,negative),and(c)theattentionalfocustowardoraway fromvocalexpressions(explicitattention,implicitattention).

3.1. Paraverbalandnonverbalemotionalvocalexpressions

ALEmapsrevealedbilateralactivityinthehigher-levelAC(BA 21,22)forparaverbalexpressionslocatedinthemSTCandpSTC,as wellasfornonverbalvocalexpressionsintherightpSTC(Fig.2A;

Table4AandB).Nonverbalexpressionsadditionallyactivatedthe bilateralprimaryAC(leftTe1.1,rightTe1.0),andparaverbalexpres- sionsadditionallyactivatedtheleftprimary(Te1.0)andsecondary AC(BA42).Sinceparaverbalexpressionsincludedifferenttypesof verbalmediums,wegeneratedALEmapsseparatelyforvocalemo- tionsbasedonpseudolanguage,forneutralsemanticspeech,and foremotionalsemanticspeech(Fig.2B,Table4CandE).Whereas alloftheseparaverbalexpressionsrevealedbilateralactivityinthe STGandtheMTGinthemSTCand pSTC(BA21, 22),only neutral(Te1.0)and pseudo-language-basedspeech(Te1.1)revealed activityintheleftprimaryAC.

3.2. Positiveandnegativevocalexpressions

ALEmapsfornegativevocalexpressionsrevealedbilateralpri- maryACactivity(Te1.0;extendinginthelefthemisphereintoBA 52),secondaryACactivity(BA42),andactivityintherightmSTC andleftpSTC(BA22),whereaspositiveexpressionselicitedactivity onlyinthebilateralleftmSTCandintherightaSTCandpSTC(BA 21,22,38)(Fig.2C;Table5AandB).Negativevocalexpressions in thestudiesincludedin this meta-analysis mainlycomprised

expressionsofanger,disgust,orfear,forwhichwegeneratedALE mapsseparately(Fig.2D;Table5CandE).Angryvoicesrevealed bilateralactivityinthemSTCandpSTC(BA22)andintheleftsec- ondaryAC(BA22/52).Fearfulexpressionsrevealedactivityinthe leftsecondaryAC(BA42)andtherightmSTC(BA22).Expressions ofdisgustrevealedactivityintherightprimary(Te1.0/Te1.1)and secondaryAC(BA42),aswellasintheleftmSTC(BA22).

3.3. Explicitandimplicitprocessingofemotionalvocal expressions

Vocalexpressionspresentedinside(explicitattention)orout- side thefocusof attention(implicit attention)revealed activity mainlyinthebilateralpSTC (BA22),butonlyvocalexpressions presentedoutsidethefocusofattentionrevealedactivityinthe bilateralprimary(bilateralTe1.0,leftTe1.1)andsecondaryAC(BA 42)(Fig.2E;Table6AandB).Duringimplicitattention,participants couldfocustheirattentiononthegenderofavoice,oronalin- guisticfeatureofspokenspeech,oronadifferentspatiallocation (i.e.attentionwasdirectedawayfromvocalexpressions)(Fig.2F;

Table6CandE).Duringimplicitattentiontowardvocalexpressions, mostoftheactivityinthebilateralprimaryAC(bilateralTe1.0,left Te1.1),intherightsecondaryAC(BA42),andintheleftmSTCand rightpSTC(BA22)wasdrivenwhenparticipantsfocusedonthe genderofavoiceinsteadoffocusingonalinguisticfeature(left pSTC,BA22)oradifferentspatiallocation(rightmSTC/pSTC,BA 22).

Allresultsforthemainanalysesonparaverbalandnonverbal, negativeandpositive,andexplicitandimplicitattentiontoward vocalexpressionsaresummarizedinFig.3.

Table3

Peakactivationsinthesuperiortemporalcortexofregions,whichweregenerallysensitivetovocalexpressions.

MNI

Region BA Volume(mm³) x y z

Lsuperiortemporalgyrus 22 9024 −62 −24 0

22 −52 −14 4

41 −52 −32 10

41 −40 −32 8

Lmiddletemporalgyrus 22 24 −56 −44 8

Rsuperiortemporalgyrus 22 7640 64 −32 4

22 62 −12 0

41 62 −22 4

42 46 −32 6

Rtransversetemporalgyrus 41 52 −22 8

Rmiddletemporalgyrus 21 56 −4 −14

Rsuperiortemporalgyrus 38 128 52 8 −16

(8)

Fig.2. Functionalactivationmaps(A)forparaverbalandnonverbalvocalexpressionsand(B)forthedifferenttypesofparaverbalexpression(pseudo-language-basedspeech [“pseudolanguage”],semanticneutralspeech[“neutralsemantic”],andsemanticemotionalspeech[“emotionalsemantic”]).(C)Functionalactivationmapsforpositiveand negativevocalexpressions.(D)Functionalactivationmapsshownseparatelyforthedifferenttypesofnegativevocalexpressions(disgust,fear,anger).(E)Functional activationmapsforvocalexpressionspresentedinside(explicit)oroutsidethefocusofattention(implicit).(F)Duringimplicitattention,mostoftheactivitywasdriven whenparticipantsfocusedonthegenderofavoice,insteadoffocusingonalinguisticfeatureofspeech(linguistic)oradifferentspatiallocation(spatial).Forabbreviations, seeFig.1.

Table4

Peakactivationsfor(A)nonverbaland(B)paraverbalvocalexpressions.Forverbalexpressions,ALEmapswereseparatelycalculatedforstudiesusing(C)pseudospeech,(D) semanticneutralspeech,or(E)semanticemotionalspeech.

MNI

(A)Nonverbal

22 52 −16 8

42 280 64 −28 14

(B)Paraverbal

22 −52 −32 6

22 −52 −12 0

41 −50 −14 6

Rsuperiortemporalgyrus 42/22 56 48 −36 4

22 62 −12 0

Rmiddletemporalgyrus 21 56 −4 −14

(C)Pseudospeech

22 672 −62 −26 0

(D)Neutralsemanticspeech

22 −64 −24 0

41 −50 −14 8

22 80 62 −12 0

(E)Emotionalsemanticspeech

Lsuperiortemporalgyrus 22 80 −54 −18 −8

22 64 −54 −12 0

22 56 −60 0 −4

22 40 −62 −42 16

Lmiddletemporalgyrus 21 32 −48 9 −39

Rmiddletemporalgyrus 21 456 56 −2 −14

22 336 64 −22 2

22 280 66 −42 4

22 32 60 −8 −4

(9)

Table5

Peakactivationsforprocessingvocalexpressionsof(A)positiveand(B)negativeemotionalvalence.Peakactivationsfornegativeexpressionsareseparatelyreportedfor expressionsof(C)anger,(D)fear,or(E)disgust.

MNI

(A)Positiveexpressions

Lmiddletemporalgyrus 21 80 −58 −36 −9

Rmiddletemporalgyrus 22 160 52 −38 −6

21 144 50 −4 −24

(B)Negativeexpressions

22 −50 −12 4

41 296 −52 −30 14

22 −46 −16 −4

Rsuperiortemporalgyrus 22 48 54 −6 −4

42 304 52 −34 12

Rtransversetemporalgyrus 41 160 52 −20 8

(C)Angryexpressions

(D)Fearfulexpressions

Lsuperiortemporalgyrus 41/42 32 −40 −17 −6

Rsuperiortemporalgyrus 22 80 54 −8 −2

(E)Expressionsofdisgust

Rsuperiortemporalgyrus 42 54 −32 14

4. Discussion

Thepresent meta-analyticdata provideevidence,ﬁrst, for a widespreadandextendednetworkofSTCregions,includinglower- levelACregionsandhigher-levelauditoryregionsintheSTG,STS, andMTG,whichareresponsivetovocalexpressionsofemotions.

Theseemotion-sensitiveregions coverareasintheprimaryand secondaryAC,aswellasin higher-levelauditoryregionsinthe STC.Interestingly,this networkofSTCandACregions,whichis

sensitivetobothparaverbalandnonverbalvocalemotions,ismore extendedcomparedwiththesuggestedEVA(Ethoferetal.,2012), whichwassolelydefinedbyparaverballyexpressedvocalemo- tions. Second, and most important,across this widespreadSTC network,wefoundthatdifferentsubregionsseemtobedifferen- tiallysensitivetostimulus-specificbottom-upfeatures(medium, valence)andtask-specifictop-downconditions(attentionalfocus).

These data indicate that different subregions in this emotion- sensitiveSTCnetworkmightrepresentdifferentanddistributed

Table6

Peakactivationsforprocessingvocalexpressionsduring(A)explicitor(B)implicitattentiontotheemotionalcueofavoice.Peakactivationsforimplicitprocessingare separatelyreportedfor(C)genderdecisiontasks,(D)linguistictasks,and(E)tasksinvolvingspatialmanipulations.

MNI

(A)Explicitprocessing

22 264 −52 −32 12

(B)Implicitprocessing

Lsuperiortemporalgyrus 22 824 −52 −10 −2

22 32 −66 −28 4

22 −46 −18 −2

41 880 −40 −32 8

42 −56 −20 4

41 −62 −16 4

22 50 −14 8

(C)Genderdecisiontasks

22 808 −52 −10 −2

41 −56 −20 4

41 −62 −16 4

Ltransversetemporalgyrus 41 −46 −18 −2

41 50 −14 8

(D)Linguisticdecisiontasks

Lmiddletemporalgyrus 22 96 −48 9 −39

(E)Spatialmanipulation