+ All Categories
Home > Documents > Juslin_1997b

Juslin_1997b

Date post: 03-Jun-2018
Category:
Upload: rnuevo2
View: 216 times
Download: 0 times
Share this document with a friend

of 37

Transcript
  • 8/12/2019 Juslin_1997b

    1/37

    Emotional Communication in Music Performance: A Functionalist Perspective and Some DataAuthor(s): Patrik N. JuslinSource: Music Perception: An Interdisciplinary Journal, Vol. 14, No. 4 (Summer, 1997), pp.383-418Published by: University of California PressStable URL: http://www.jstor.org/stable/40285731.

    Accessed: 03/04/2014 06:30

    Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at.http://www.jstor.org/page/info/about/policies/terms.jsp

    .JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

    content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

    of scholarship. For more information about JSTOR, please contact [email protected].

    .

    University of California Pressis collaborating with JSTOR to digitize, preserve and extend access toMusic

    Perception: An Interdisciplinary Journal.

    http://www.jstor.org

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/action/showPublisher?publisherCode=ucalhttp://www.jstor.org/stable/40285731?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/40285731?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=ucal
  • 8/12/2019 Juslin_1997b

    2/37

    Music Perception 1997 bythe regents of theSummer 1997, Vol. 14, No. 4, 383-418 university of California

    EmotionalCommunicationn MusicPerformance:A FunctionalistPerspectiveand SomeDataPATRIK N. JUSLINUppsala University, weden

    The firstpartof thispaperpresentsa systematicapplicationof a func-tionalistperspectiveo the studyof emotionalcommunicationn musicperformance.This involvesthe integrationof ideas and conceptsfrompsychological esearchon emotion andnonverbal ommunicationwithBrunswik's1956)probabilisticunctionalism nd a modifiedversionofhis lensmodel.It is argued hat thisapproachmayprovide henecessarytheoretical oundationby generatingusefulquestions,hypotheses,andwaysof evaluating ata fromperformancenalysesandlistening xperi-ments.The secondpartreportsan experimentaltudyin whichprofes-sionalguitarplayerswere instructed o playa shortpieceof musicso asto communicateour basicemotions to listeners.The resultingperfor-manceswereanalyzed egardingarious ues,suchastempo,sound evel,and articulation. t was found that (a) the expressive ntentionsof theperformersffectedallof themeasured uesin theperformances,b)thecues had merelya probabilistic elationto the performers'ntentions,and(c) the cues wereintercorrelated.he performanceswere also vali-datedin a listeningexperimentwhich showed that listenerswere suc-cessfulin decodingthe intendedemotionalexpression,and that therewere no differencesn decodingaccuracybetweenmusically rainedanduntrainedisteners.

    is probably he mostwidelypracticedandappreciated f all artforms.Onepossibleexplanation orthismaybe thedeepemotionaleffectmusichas on people. Music both expressesand induces emotionsand is thus often seen as a powerfulmeans of emotional communication(Dowling&CHarwood, 1986). What is the natureof this communicativeprocess?Unfortunately, ontemporarymusicpsychologyhas little to con-tributeon this matter;especiallywhenit comesto emotionalcommunica-tion in musicperformance.Researchon music performancehas a long history;for a review,seeGabrielssoninpress).This shouldnot besurprisingnview of the fact thatAddress orrespondenceo PatrikN. Juslin,Department f Psychology,UppsalaUniver-sity,Box 1225, S -751 42 Uppsala,Sweden, e-mail:[email protected])

    383

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    3/37

    384 PatrikN. Juslinour experiencesof musicmainlyderivefromlistening o performances fmusic. It follows that the characteristics f musicalperformanceare offundamentalmportanceo thepsychologyof music.Theperformance f apieceof musicrepresentsheovert,physicalmanifestation f thecomposer'sand/or the performer'sntentions.At the sametime, the listener's xperi-enceof the musicdependsverymuchonthespecificproperties f the soundsproducedby the performer Gabrielsson,1988).These factsnotwithstanding,tudiesof communicationn musicperfor-mancearerare.Furthermore,hoseinvestigationshathave studiedperfor-manceandperception ogetherhavemainlybeen concernedwiththecom-municationof dynamics Nakamura,1987), structuralnterpretatione.g.,Clarke,1989; Sloboda,1983; 1985), or level ofgeneral xpression Kendall& Carterette,1990). On the otherhand,virtuallyno studies haveexam-inedhowperformersommunicate motionalcontents o listeners.Behrensand Green(1993), Ohgushiand Hattori (1996), and Senjuand Ohgushi(1987) studied whetherperformers ould communicate pecificemotionsto listeners.However,they did not analyzethe performanceso see whatmeans the performersusedin orderto generate he emotionalexpression.Researchpertinent o this problem s currentlybeingconductedat theDepartment f Psychology,UppsalaUniversity.Whatis investigated,morespecifically, s whether a performercan communicate conic representa-tions of emotions to listeners,as well as what meanstheperformeruses toaccomplish this task. These studies follow a procedure proposed byGabrielsson1985)involving flexibleand terativenterplay etweenanaly-sis andsynthesisof performances.The startingpoint is an analysisof realperformances.The results from this analysis generatehypotheses,whichthen aretestedbymeansof systematicallyaried,synthesized erformancesand listeningexperiments.The focus is thus on the psychophysical ela-tions betweenthe performer's xpressive ntention,objectivepropertiesofthe performance,and the listener's ubjectiveexperience(Gabrielsson&Juslin,1996).

    One of the problemswith studies of communication n musicperfor-manceis that there are few theoriesfor the interpretation f performancedatainrelation o musicalexperience.1Onepromisingapproachmaybe toadopta functionalistperspective Juslin,1995). The firstpartof thispaperis thusconcernedwith a systematicapplicationof a functionalistperspec-tive to the studyof emotionalcommunicationn musicperformance. irst,I describea numberof ideas and conceptsthat are essentialto an under-standingof the functionalistperspectiveon emotionalexpression.ThenIconsidersome of the implicationsof takinga functionalistperspectiveon1. However, heoriesby SusanneLanger 1957) and ManfredClynes(1977)arebrieflydiscussedn Gabrielsson1995).

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    4/37

    EmotionalCommunicationn MusicPerformance 385music performance. Finally,a researchstrategy is outlined that includes (a)an interplaybetween analysis and synthesis of performances, (b) Brunswik's(1956) metatheory,and (c) a modified version of his well-known lens model.The second part of the paper reports an experimental study that investi-gated five questions generated by the functionalist approach.

    A Functionalist PerspectiveFUNCTIONALISM N RESEARCHON EMOTION AND NONVERBAL

    COMMUNICATIONThe past two decades have witnessed a major resurgence of interest intwo related areas: the study of nonverbal communication and the study ofemotion. Several authors have recognized that emotional communicationprobably serves as the foundation of the social order in both animals andhumans (e.g., Buck, 1984; Plutchik, 1994; Scherer,1982). One approachthat has turned out to be heuristic in this context is functionalism (Fridja,1986).First, it may be useful to delineate what is meant by functionalism in thispaper.Human beings are products of biological evolution. Therefore, vari-ous aspects of human functioning should be considered in terms of theiradaptive significance or survival value.2 Functionalism assumes that bothpsychological and biological phenomenon make sense and do not repre-sent mere contingencies. However, this does not imply that absolutely ev-ery aspect of human behavior is or has been adaptive. Moreover, the func-tionalist perspectiveis not a geneticdeterminist position; it rather stressesthe interaction of genetic and environmental forces in the evolution of hu-man attributes (Ruse, 1978).Several authors have stressed the functional significance of emotions,and especially their importance in social interaction:

    I believe heprimaryunctionof emotion s to mobilise heorganism odealquicklywith important nterpersonalncounters,preparedo doso in part,at least,bywhattypesof activityhave beenadaptive n thepast... (Ekman,1992a, p. 171)Most proponents of this approach have postulated the existence of asmall set of basic emotions21rom which all other emotional states can be2. SeeBuss(1995) for a discussionon evolutionarypsychologyas a paradigm or psy-chological cience.3. Basicemotionsare sometimesalso referredo as primary,discrete,or fundamentalemotions.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    5/37

    386 PatrikN. Juslinderived.Basicemotionsareusuallyviewedashavingevolved hrough heiradaptivevaluein dealingwith fundamentalifeissueswithoutresorting otime-consuming rocessing.They typicallyaffectbehavior, hysiology,andsubjectivetates ncomplexways(Ekman,1992a;Johnson-Laird&C atley,1992).Somegeneralcriteria or distinguishingbasicemotionsfrom second-ary or complex emotions have beensuggested.Themostimportantofthese arethefollowing:Basicemotions(a)have distinct unctions hatcon-tributeto individualsurvival,(b) are found in all cultures,(c) areexperi-enced as unique feelingstates, (d) appear early in the course of humandevelopment, e)areassociatedwith distinctautonomiepatternsof physi-ologicalchanges,(f)can be inferredn otherprimates,and(g)have distinctemotional xpressionsEkman, 992a;1992b;Fischer, haver,&cCarnohan,1990; Izard,1977; 1993; Oatley,1992; Plutchik,1994).Although he notion of basicemotionshasbeen the matterof somecon-troversy(cf. Ortony& Turner,1990), a considerablebody of theoryandevidencesupportsthe case for a conceptof basic emotions as definedbythecriteria ustmentioned Ekman,1992b;Oatley,1992; Plutchik,1994).There s also reasonable onsensusamongemotionresearchersoncerningat least four basicemotions;anger,sadness,happiness,and fear(Kemper,1987).

    What, then,is the relationshipbetweenemotionandits expression?Al-thoughmost researchers cknowledge his link as beingan intrinsicone,the exact natureof therelationships seldomfurtherexplained.However,anumberofresearchers aveproposed hat basicemotionsare mplementedin the form of innateemotionprograms hat functionto initiateandorga-nizeemotionalreactionsand emotionalexpressions Ekman,1984;Izard,1977; Tomkins,1962).Clynes(1977), followingTomkins(1962), haspostulated he existenceof biologicallyprogrammedpatiotemporal atternsor the nonverbal om-municationof specificemotions.According o his theory,emotion anditsexpressionforma singleexistentialentity.Each basic emotion has a spe-cificexpressive ormregardless f theoutputmodality;a gesture,a tone ofvoice,a dancestep,or a musicalphrase.Furthermore,heproductionandrecognitionof expressive ormsare seen as biologicallycoordinatedby in-herentdataprocessingprograms.This idea, that is, that both expressionand recognitionhave a common biologicalbasis, is echoed by otherre-searchers. t makessense,becausethe overtexpressionof emotion wouldbe of littleevolutionaryvalueif othermembersof the groupfailed to de-code andrespondappropriatelyo thedisplay(Buck,1984;Darwin, 1872;Dimberg,1988). The implication s that a thoroughunderstanding f thecommunicative rocessrequireshatexpressionandrecognitionaspectsbestudied n a combined ashion.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    6/37

    EmotionalCommunicationn MusicPerformance 387According o Mayr(1974), communicativebehavior s the categoryofbehavior hat is most influencedby closedgenetic programs.One reason

    maybe that the earlymother-infantnteractiondependsalmostsolelyonnonverbal ommunication.Atthis criticalstage,when thenewbornorgan-ism is most vulnerableo dangers, here is simplyno time to learn a com-municative ode;the communication atternsmustwork the firsttimetheyareused(Plutchik,1994). However,eventhough manyemotionalexpres-sionsareunlearned, ncetheyhaveoccurred, hey maybevoluntarilyusedto communicatenformation rom one individual o another(Buck,1984;Darwin,1872). Schneider,Hastorf,and Ellsworth 1979) note that infer-encesbasedon nonverbal ues areprimarilynferencesaboutrelationshipsandfeelings,and thus areamongthe most importantnferenceswe make(p. 142).Nonverbalcommunicationof emotionimplies(a)a sociallysharedsig-nalsystem, hatis, a code, (b)an encoderwho makessomethingpublicviathat code, and (c) a decoderwho respondssystematically o that code(Wiener,Devoe,Rubinow,& Geller,1972). Thisprocess s amenable o adescriptionn termsof Shannon ndWeaver's1949)communicationmodel.In this model,specialattention s paidto the qualityof reproduction s amessage n theform of informations passedfromone element o another.This is limitedbytwo factors:channelcapacityandnoise. Channelcapac-ity refers o how muchinformation an be transmitted ver a givenchan-nel in a giventime. Whenever he capacityof a channelis less than therichnessof variation romthe sourcefrom which it acceptsmessages, hechannel s overloaded. nthiscase,therewillalwaysbe someuncertaintyor noise concerningwhat the messagewas. However, he uncertaintyofthe signalmay be decreasedby the use of redundancyn the coding pro-cess.A widespreadassumptionexiststhat nonverbal ignalsareparticularlysuitedfor the communicationof emotiveinformation.As Scherer 1982)has pointed out, nonverbal signals are coded continuously andprobabilistically.4ifferentdegreesof strengthof an underlying tate canthusbe communicatedhrough ontinuously hanging ignals,whileproba-bilisticcodingensuresa high degreeof negotiabilityof the meaningcon-veyed.Musiccan obviouslybe regardedas a form of nonverbalcommunica-tion.Perhaps,hen,we cangainmuchfromviewingmusicperformancenthiswiderperspective,hat is, as being partof nonverbalcommunicationingeneral.Whataretheimplicationsof a functionalistperspective n emo-tionalcommunicationn musicperformance?

    4. Probabilistic odingrefers o the fact that the code involvesseveralpartiallyredun-dantbutnot perfectly eliable andhenceprobabilistic)ues.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    7/37

    388 Patrik N. JuslinA FUNCTIONALISTPERSPECTIVEON MUSICPERFORMANCE:SOME

    IMPLICATIONSCrucial o the functionalistperspectives theconsideration f what rela-tionshipthe phenomenonunder nvestigationhas to the ecologyin whichtheorganismhasevolvedandto whichit hasadapted.Does thefunctional-ist viewseemfar-fetchedwhendealingwithmusicperformance?Whatmustbe emphasizeds that the functionalistperspective,as formulatedhere,isnotconcernedwithexplaining hepossible unctionality f music.5nstead,it concerns hoseaspectsof nonverbal ommunication f emotionthat canbeconceivedof as beingrather early nthephylogeneticdevelopment fhumanfunctioning. t is likelythatourspeechless orefathershad to make

    do formillenniawith the same nonverbal ignal systemusedbymostsub-humanspecies (Scherer,1982). Furthermore,nonverbalcommunicationgestures,whosefunctionality romptedheir ncorporationnthegenepoola verylong timeago, may retainsome of that functionality oday simplybecause he ecologyof close social encountershas not changedverymuch(hman& Dimberg,1984). Itmaythus be hypothesizedhatperformers,at least to some extent,can capitalizeon a generalnonverbalcode whentheycommunicate conicrepresentationsf basicemotions o listeners.Thefunctionalistperspective n musicperformance asa numberof importantimplications:(a) Encodingand decodingshould be studied in a combined fashion.Performance nd perceptionarereciprocally elatedprocesses:To under-stand performancedata from studies of emotionalexpression,we mustconsiderthemin relationto the performer'sxpressive ntentionand thelistener's xperience Gabrielsson& Juslin,1996).(b) Basicemotionsshould be easier to communicate han other emo-tions. The primaryquestionis whetherperformers re able to communi-cate iconicrepresentations f emotions to listenersat all. In approachingthisquestion, t maybewiseto base theselectionof contents o be commu-nicated on a theoryof emotion.6Studiesshouldbe guidedby explicitas-sumptionsregardingwo fundamentalssues: 1)What can a performer,na fairlyreliableway,communicate o hisor herlisteners?2)What s it thatmakesthiscommunicative rocesspossible?Functionalistheoriesof emo-tion giveus reasonto expectthat certainemotions areeasier o communi-cate than others.Withoutexcludingcultural nfluences,we may hypoth-

    5. However,eeDowlingandHarwood 1986,pp.235-238), Hodges 1989),andScherer(1991) forspeculations n theevolutionary riginof music.6. Note thatSenjuandOhgushi 1987) foundvery imitedcorrespondenceetween heperformer'sxpressivententionand theimpression f thelisteners,when theviolinistSenjutriedto communicate uchabstract ontentsas deep,beautiful,andsophisticatedo listen-ers.Greater orrespondence asfound na subsequent tudyemployingourbasicemotionlabels, .e., anger,happiness, orrow,andfear(Ohgushi&cHattori, 1996).

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    8/37

    EmotionalCommunicationn MusicPerformance 389esize that basic emotions- because of their phylogenetic history and intrin-sic relationship to expression- should be easier to communicate than otheremotions. Performance studies should accordingly begin by studying theexpression of such proposed basic emotions as happiness, sadness, anger,and fear, followed by a gradual inclusion of the expressions of more com-plex emotions.(c) There should be similarities of expressions across different modali-ties. For instance, there is already some evidence of similarities betweenemotional expressions in music performance and in the nonverbal aspectsof speech(Baroni& Finarelli, 1994; Scherer&cOshinsky, 1977). This shouldperhaps not be surprising, because it is a plausible speculation that musicand speech share common building blocks in vocal expression of emotion(cf. Scherer, 1991; Sundberg, 1982). Anyway, it seems likely that evidencefrom comparative studies of music, speech, and body language may castsome light on the communicative process.(d) The code should involve several probabilistic and partially redun-dant cues. It is known from nonverbal communication research that non-verbal signals are coded probabilistically to a large extent. Several studiesof vocal expression, in particular,indicate that researchers must consider awhole range of probabilistic cues (van Bezooijen, 1984; Frick, 1985). Weshould thus expect to find that emotional communication in music perfor-mance also involves a number of probabilistic and partially redundantcuesin the performance.(e)Decoding accuracy should be independent of musical training. This isbecause, according to the functionalist view, emotional communication inmusic performance argelyuses the same spatiotemporalcode as other formsof nonverbal communication. We may therefore hypothesize that musi-cally trainedlisteners and untrained listenersdo not differsignificantly withrespect to decoding accuracy for expressions of basic emotions in musicperformance. However, this does not imply that every performershould beable to communicate emotions to listeners, because he or she may not havethe requiredtechnical skill for expressing emotions in musical performance.The fact that it is probably easier to recognize an authentic emotional ex-pression in a performance than to produce it (Clynes, 1977) implies that itis preferable to use professional performers in studies of emotional com-munication.(f) There should be gender differences in decoding accuracy. A host ofstudies in nonverbal communication research have found that females aresuperiorto males at identifying emotion from nonverbal cues of face, body,and voice (cf. Eisenberg& Lennon, 1980; Hall, 1978). It should obviouslybe determined whether this relationship holds for the decoding of emo-tional expression in music performance. This question is also of theoreticalinterest because similar patterns of gender differences across different mo-

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    9/37

    390 PatrikN. Juslindalitiesof nonverbalcommunicationmaybe taken as support or the ideaof a commonunderlyingmechanism, hat is, emotionprograms.

    INTERPLAYBETWEENANALYSISAND SYNTHESISOF MUSICALPERFORMANCESDeterminingwhether it is possibleto communicaterepresentations fspecificemotions throughmusicalperformancerequiresa definitionofcommunication.According o informationheory,behaviors communica-tive to the extent that it reducesuncertaintyn the behaviorof another.However,this simpledefinition fails to distinguishbetweeninstances nwhich information s involuntarilyand unconsciouslyconveyedand in-stances nwhichinformation s intentionally onveyed.Accordingo Grice

    (1986), communicationcan only be said to have been successful f thesender'scommunicative ntention has becomeknown to both the senderand the receiver.Giventhis definition,how are we to discover he specificcode that is usedin thecommunicativeprocess?The firstimportantstepis to identifythe expressivecuesthat makeupthecommunicative ode. To determinewhichcuesare usedbyperformersandlisteners, t is necessaryo use a flexible nterplaybetweenanalysisandsynthesisof performancesGabrielsson, 985). Thereasons or this shouldperhapsbe clarified.It has been shown in a host of studies that musicperformancenvolvescertain deviations rom the nominalvalues of themusicalscore(Gabrielsson,n press).It is generallybelieved hat someofthesevariationsare usedin order o conveythe intendedmusicalstructureto the listener(Clarke, 1988). However,it has also been suggestedthatsome of the variationsmaycontribute o the emotionalexpressionof theperformanceGabrielsson,985; Seashore, 938/1967;Shaffer, 992; 1995).Theproblem s to find out which of the variations n theperformance reimportant ortheexpression.As it turnsout,performers remostlyunableto informus on this matter.First,the researcherhas to ascertain hat the found variationsaresys-tematic as opposed to random variations, or noise (cf. Bengtsson,Gabrielsson,& Thorsn, 1969; Sloboda, 1994). Thereliabilityof theper-formanceshouldthereforebe testedby askingtheperformero repeat hesameinterpretation f the pieceseveral imesandthencheckingwhich ofthe variations end to remain he samein successiveperformances.Second,the variationsshould be partof the performer's xpressive n-tention,which can be testedby manipulatinghis or her intention.Thosevariations hat remain he sameregardless f theprescribed motionalex-pressioncan presumablynot be partof the performer'sxpressive nten-tion. Nor can theybe usedby the listener n orderto decode differentex-pressions.Lastbut not least,the researcher as to make sure that the found varia-tions areused by the listener n his or her judgmentconcerning he emo-

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    10/37

    EmotionalCommunicationn MusicPerformance 391tionalexpression.Thismustbe investigatedby meansof listeningexperi-ments hatusesynthesizedndsystematicallyaried oundsequencesJuslin,1996a; 1996b; Scherer& Oshinsky,1977) or digitallyrecordedperfor-mancesthat aremanipulatedwith respectto variousphysicalparameters(Repp,1995).Provided hat we manageto identifythe expressivecues, how can weconceptualize ndquantify he process n a way that relatesall aspectsofthe communicative hain?Is there a metatheoreticalramework hat canguidethe selection of problemsto study,as well as providesuitable de-scriptionof variablesmanipulated nd relationsobserved?

    A METATHEORETICALRAMEWORK:BRUNSWIK'SPROBABILISTICFUNCTIONALISM

    Egon Brunswik 1955, 1956) pioneereda functionalisticapproachtoperception.His theoreticalorientation s a formof functionalismbecauseits main focus is on the adaptiveinterrelationof the organismwith theenvironmentCooksey,1996; Hammond,1966).Brunswik mphasized he inherentlyuncertainnatureof the perceiver'sworld.Perceptions seenas an indirectprocessthat involves the achieve-mentof a distalvariablebymeansof a set of proximalcues.However, hevariouscues arisingfrom the environmentare not fully dependable; heyhavemerelya probabilistic elation o the distalvariables.Because uesareinterchangeablendcommonlyhave reliabilities f lessthan1.0, theenvironment an be describedas being vicariouslymediated.Theresponse o vicariousmediation s vicarious unctioningon thepartoftheperceiver.norderto survive, he perceivermust act much like an in-tuitivestatistician. He or she must weigh and combinecues, and shiftfrom ones that are not availableto othersthat are available.Perceptionmaythusbe describedas uncertainty-geared;t aims for smallnessof er-ror at theexpenseof thehighestfrequencyof precision Brunswik,1956,p. 146).According o Brunswik,psychologicalanalysisshould concentrateonstudying he degree o whichtheorganismattainsrelevantdistalvariablesandthe strategies hatthe organismusesto accomplish his task. That is,research houldfocuswherebehavior ocuses,Brunswik eferred o this asthe principleof behavior-researchsomorphy.As pointedout by Brehmer(1988), this importantprinciple mpliesthatif we are to understandwhatsubjectsdo, we haveto know what theirgoals and intentionsare.Animportant eatureof Brunswik'sheory s the lens model of behavior(Brunswik, 956). The modeldefines he unit forpsychologicalanalysisasa systemwith two subsystems the environmentand the organism thathave a common interface: he proximalcues. This model was meant toillustratehow perceptionnvolvesa kind of focusing, n that the scattered

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    11/37

    392 Patrik N. Juslinand mutually interchangeablecues arisingfrom the environment must some-how be recombined in order to attain the hidden regularitiesof a lawfulworld (Figure 1). Brunswik believed that this model could guide the quan-titative assessment of particularperceptual achievements and thus provideresearcherswith an useful analytical tool.Because cues have a probabilistic relation to the distal object intended, itfollows that the degree of the relationship can be described in quantitativemeasures. Thus, the correlation between a distal variable and a proximalcue describes the potential usefulness the ecological validity of that par-ticular cue in the environment. The next step is to establish the functionalvalidity of the cue; that is, the extent to which the cue is actually used bythe perceiver.Note that Brunswik's ideas and Shannon and Weaver's (1949) commu-nication model are somewhat similar: Vicarious functioning of cues maybe viewed as a special case of receiving or sending messages through redun-dant channels, thus reducing the uncertainty.Brunswikfocused on the uncertainnatureof the perceiver'sworld. How-ever, uncertainty is also a prominent feature of social systems (Brehmer,1976). Furthermore,nonverbal signals appearto be coded probabilistically.Scherer (1982) has thus proposed a modified lens model as suitable fornonverbalcommunication research.Similarly,Juslin (1995) has shown thata modified lens model is applicable to music performance. It can serve tohighlight that music performance involves a communicative chain consist-ing of (a) the performer'sexpressive intention, (b) the musical performance,and (c) the listener'sexperience of the performance.

    Fig. 1. Brunswik'slens model for behavior.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    12/37

    Emotional Communication in Music Performance 393A MODIFIEDLENS MODEL FOR EMOTIONALCOMMUNICATIONIN MUSIC

    PERFORMANCEBrunswik'sramework eems suitable ordescribing motionalcommu-nicationin musicperformanceor severalreasons.First,the performer'sexpressive ntentionhas a marked effect on several differentcues in theperformance. econd, t seems that successfulcommunicationof a certainemotioncan be achievedin somewhat differentways dependingon theexpressivemeansavailableon particularnstruments r on thecharacteris-ticencoding trategies f individual erformersGabrielsson&c uslin,1996).Third, listening experimentswith synthesizedand systematicallyvariedperformancesndicate hat listenersuse a wholerangeof cuesin theiremo-tion judgmentsbut that none of these cueshavemore than a probabilisticrelationto the listeners' udgmentsof the emotionalexpression7 Juslin,1996a; 1996b; Scherer& Oshinsky,1977). It may thus be hypothesizedthat the communication ystemadheres o the principleof smallnessoferrorat the expenseof thehighestprecision. Thecueredundancy houldmakethe communication ystempowerful.Thatis, the systemshould notbe sensitiveto the noise causedby (a) individualdifferencesamong per-formers,or (b) structuralconstraintsof specificmelodies. On the otherhand,the limited nformation apacityof the systemshould make it diffi-

    cult to communicate ubtle nuances n a reliableway.In the precedingdiscussionaboutanalysisandsynthesis, t was impliedthattheexpressiveues namusicalperformancexistintwo distinct enses;as potentially usable information and as actually used information.Brunswikianonceptsmaybe suitable orreferringo these two aspects.Inourseriesof studies,the expressive ntentionof the performers manipu-lated. Objectivemeasuresof variouscues in the performance,as well assubjectivemeasuresof the listener'sudgmentare then obtained(Figure2).Inthiscontext,theecologicalvalidity(r )refers o therelationshipbetweenthe performer'sxpressive ntentionand an objectivelymeasurable ue intheperformance.neffect,it is an indexof the potentialusefulnessof thatparticular ue.Incontrast, hefunctionalvalidity(rf)refers o the relation-shipbetweena particular ue in the performance nd the judgmentmadeby the listener; hat is, the extent to which the cue is actuallyused by thelistener n his or herjudgment. norderto establish he functionalvalidityof an expressivecue, it is necessary o conductlisteningexperimentswithsystematicallyariedperformances.inally,unctional chievement(rj refersto the relationshipbetweenthe expressive ntentionof the performerand7. Independent evidence that listeners use a number of probabilistic cues in emotionattribution from auditory stimuli comes from studies of emotional expression in the non-verbal aspects of speech (van Bezooijen, 1984; Frick, 1985).

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    13/37

    394 PatrikN. Juslin

    Fig.2. Modified ensmodel for theconceptualizationf musicperformance.

    theemotionjudgmentmadebythe listener,hatis, the extentto whichthecommunicational ct is successfulor not.The various cues in the performancemay be rankedfor theirrelativeimportancenthecommunicative rocess.Theserankings,n termsof eco-logicaland functionalvalidities,probablydifferdependingon the emotioncommunicated.An importantquestionis whetherthe expressivecues areintercorrelated.f they are, it means that two performers an be equallysuccessful n communicatinga particularemotion, despitedifferencesnhow they use the expressivecues. Similarly, ue intercorrelations nablelisteners o have a highlevel of decodingaccuracydespite ndividualdiffer-encesin cue utilization.Thelens modelallows us to exploresome other mportantaspectsof thecommunicative rocess.Forinstance,Brunswik ssumed hatgoodadjust-ment wouldrequire hatthe functionalvalidityof the cue for theorganismbe approximatelyhe sameas the ecologicalvalidity. notherwords, theorganism houldutilize hecue to whateverextent t is ecologicallyvalid oruseful (Hammond,1966, p. 33). This matching ssumptionmightjustas well be investigatedwithina musicalcontext.Byuse of multipleregres-sionmethods, t is possible o discoverwhethereachexpressive ue is givenanappropriate ubjectiveweight bythe listener.Listenerswithhighdecod-ing accuracy houldobviouslyhave bettermatchedweightsthanlistenerswith low decodingaccuracy.Furthermore,he encodinganddecodingas-pects may be mathematically elatedusingthe lens model equation pre-sentedin an influentialpaper by Hursch,Hammond,andHursch(1964).

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    14/37

    EmotionalCommunicationn MusicPerformance 395They pointedout that without a knowledgeof the limitationsplacedonfunctionalachievementby the statisticalcharacteristics f the (encodinganddecoding)systems, t is impossible o (a)evaluatea subject'sachieve-mentwithinthat system,(b)comparea subject'sachievementacross eco-logicalsituations hat have different tatisticalcharacteristics, nd (c) un-derstandwhy the subject'sachievementwas as high or low as it was. Inaddition,the lens model equation permitsthe researcher o describethedecodingprocess n terms of an organizingprinciple, hat is, whethertheexpressive uesareintegratedn an additiveor a configuraiway (Brehmer,1976; Cooksey,1996).Theapplicationof the lensmodelequationto musicperformanceaisessomeproblems, or example,with regard o the fact that someof the ex-pressivecues are not constant but typicallyvary continuouslyover time.However, uslin(1996b)was able to providesomepreliminaryvidenceonthe functionalvalidities of some expressivecues:tempo, sound level, ar-ticulation,attack,and timbre.This was doneina listeningexperimentwithsynthesized ndsystematically ariedperformances f a shortmelody,us-ingmultipleregressiono capturevariousaspectsof thelistener'sudgmentpolicy.

    RESEARCHQUESTIONSAND FURTHERCOMMENTSAccording o the functionalistperspective,performersmay communi-cate iconicrepresentationsf specificemotions to listenersby means of anumberof probabilisticand partially nterchangeableues in the perfor-mance.Thesecuesare used in accordancewith a generalnonverbalcodeimplementedn biologicalprograms orrespondingo the basic emotions.Cueredundancyn the codingmakesthe system ess sensitive o the noisecausedbyindividualdifferences mongperformers.On the otherhand,thelimitedinformationcapacityof the systemlimits successfulcommunica-tion to the basic emotioncategories.The functionalistperspective eneratesa numberof researchquestions:Can a performer ommunicate epresentations f specificemotionsto hisorher isteners?Are basicemotionseasier o communicatehan otheremo-tions?Do musically rained istenershavea higherdecodingaccuracy hanuntrained isteners?Are there genderdifferences n decodingaccuracy?Whicharetheexpressive uesinvolved n thecommunicationprocess?Arethesecues ntercorrelated?Whichcueshavethehighestecologicaland func-tionalvalidities?Arelisteners'uses of cues matchedto the ecologicalva-liditiesof the cues?Whichorganizingprinciple s used by listeners?Howconsistentare listeners n theirjudgments?Are theresimilaritiesof emo-tionalexpressionsacrossdifferentmodalities?What is therelationshipbe-

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    15/37

    396 PatrikN. Juslintweenthe encodinganddecodingaspectsof theprocess?Can the commu-nicativeprocesssomehowbe improved hroughthe use of the lens modelparadigm?As pointedout byJessorandJessor(1977), the presentationof a theo-reticalposition engagesthe possibilityof a numberof misunderstandings,for instance, (a) that it is the only usefulway of construing he eventsatissue, (b)that it represents priorclaim to understandingather hanoneto bedemonstrated, r (c)that itspurview s relatively osmologicalratherthan limited. None of these implications s intendedhere. The proposedtheoretical ramework hould be seen as complementaryo other modelsof music performance e.g., Clarke, 1988; Clynes, 1987; Friberg,1995;Kendall& Carterette,1990; Sundberg,1988; Todd, 1985; 1989), whichmainlyhave dealt withvariousstructural spectsof performance.Thisis inaccordancewith Clarke's 1996) view that (a)expressivevariationsn mu-sicalperformance recharacterized y a multiplicityof functions,and (b)no singletheorymaybe sufficient o account orallaspectsof thiscomplexphenomenon.This is onlya preliminary ttempt o outline thepossibilitiesnherentnthe functionalistapproach,and we must also consider heprofound nflu-ence of culturalfactors.Nevertheless, t should be evidentthat the func-tionalistperspectivecan make a powerfulheuristic n the studyof emo-tionalcommunication n musicperformance;t generatesusefulquestions,hypotheses,and ways of evaluatingdata fromperformance nalysesandlisteningexperiments.Emotional Communication Through Electric Guitar Performance

    INTRODUCTIONInthis section,somefindingsfrom a performance nalysisand a listen-

    ing experimentarereported.Themainaim of the studywas to investigate(a)whetherguitarplayersare ableto communicate epresentationsf ba-sic emotions to listeners,(b) how the performer'sxpressive ntentionaf-fects expressive cues in the performance, (c) whether the cues areintercorrelated,d)whether here aregenderdifferencesn decodingaccu-racy,and (c)whethermusically rained istenershavehigherdecodingac-curacy hanuntrainedisteners.Threeprofessionalguitarplayerswereinstructed o playa melodyso asto communicateanger,happiness, adness,or fear to listeners. naddition,theywere instructed o try to play the melodywithoutexpression.Theirperformanceswerefirstrecordedandstored ncomputermemoryand thenvalidated na listeningexperimenthatcompared hedecodingaccuracy flisteners with regardto genderand musical training.The effect of the

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    16/37

    Emotional Communication in Music Performance 397performer'sxpressivententionon the resultingperformancewas investi-gated by means of an analysis focusingon three acousticalparameters:soundlevel,tempo,and articulation.Inthe firstpartof the study, heexpressivententionof the performersthe independent ariable,andthe resultingperformance as measured ntheanalysis is thedependentvariable. nthe secondpartof the study, heexpressiveperformances the independentvariable,and the listeners'rat-ingsof the sameperformances the dependentvariable.Theencodinganddecodingaspectsof the expressionare thus studied n a combined ashion.Theparticularnstrumentused,the electricguitar,waschosen because tprovides heperformerwith a numberof expressivemeans;he or shemayvarythetempo,articulation, oudness, imbre, iming, ntonation,vibrato,or attack.Also, the presentauthor is especiallyfamiliarwith this instru-ment.However, t shouldbe noted that our earlierstudies have includedseveralother nstrumentsnd differentmelodies.The results romthisstudyshouldthusbe considered n relationto our earlier indings(Gabrielsson,1994;1995;Gabrielsson&Juslin,1996;Gabrielsson, uslincLindstrm,1994; Gabrielsson&cLindstrm,1995;Juslin,1993).

    THE ENCODING EXPERIMENTMethodsThe PerformersThree professional guitar players (males), 25-45 years old, participated in the study.They were mainly performing rock, blues, jazz, and folk music, and their experience ofplaying the guitar ranged from 10 to 30 years. They were paid for their anonymous contri-bution and used their own guitars for the recording to ensure that they were familiar withthe instrument.Apparatus and Recording ConditionsThe recording took place in a laboratory room (about 6x3 m), partly furnished as aliving room with tables, sofas, and a large carpet on the floor. The performances wererecorded by means of a microphone put close to the speaker of a guitar amplifier, PeveyClassic 50. The amplifier was placed approximately 2-1/2 m from the performer,who wasable to monitor the sound directly from the amplifier.The guitar sound was clean, that is, devoid of any external effect devices. However,because electric guitarists rarely play with a completely dry sound, it was decided to usea small amount of reverberationin order to give a natural feeling of ambience to the sound.Monitor level, reverberationlevel, and recording level were the same for all the performers.The recordingswere made on a Revox B88 tape recorder at 19 cm/s and on a Yamaha MT120 four-channel cassette recorder that was used as a back-up.ProcedureIt was decided that the performersshould have a chance to prepare themselves for theirtask a few days before the actual recording procedure. About 5 days before the recording,the players received written instructions and scores for the five melodies they were to per-form.The melodies were Greensleeves, Nobody Knows, Let ItBe, Whenthe Saints,

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    17/37

    398 PatrikN. Juslinand What Shall we Do with the Drunken Sailor. These were chosen so as to representsomewhat varying emotional characterwith regardto the melodic structure.However, onlythe results for When the Saints will be treated in this paper.The notation of this melodyis shown in Figure 3.When the performerarrived,he first had a chance to warm up on his instrument andto familiarize himself with the sound of the amplifier.Meanwhile, the recording levels wereadjusted. Then, the performer received the instructions once more. They were given asfollows: You are now going to play each of the five melodies so that they sound as follows:happy, ad,angry,earful,andwithoutexpression.Youarenot allowed o change hepitchesof the melody, or the guitar sound, other than the obvious timbre differences due to differ-ent styles of playing. You are, however, otherwise free to vary all other aspects of the perfor-mance as you like: tempo, loudness, rhythm, articulation, attack, vibrato, phrasing, and soon. Theimportant hing s thatyouachieve heprescribed motionalexpression. magine,for example, that you are going to demonstrate for some listeners how to play in order tocommunicate the emotions mentioned above. You play each version two times, as similarlyas possible, with a few seconds break in-between. You are allowed to repeat the recordinguntil you are satisfied with each version. After the recording of each expressive versionwas completed for the first melody, the same procedure was repeated for each of the otherfour melodies. The recording session took approximately 2 hr.

    Data TreatmentThe recorded material was stored on tapes. The audio recordings were then transferredto a computer memory, using a sampling frequency of 22 kHz, and analyzed by means ofthe Soundswell program (Ternstrm, 1992). This permits visual displays of each perfor-mance in as much detail as needed to identify onsets and offsets, as well as the amplitudeenvelope for each tone. The onset was set where the waveform of a new tone appeared and

    the offset at the point where this regularwaveform disappeared;this is relatively easy to dowith signals from the electric guitar.Two durations were measured for each tone; the dura-tion from the onset of a tone until the onset of the next tone (d.J,and the duration from theonset of a tone until its offset (dio)(cf. Bengtsson & Gabrielsson, 1980). These durationswere used to calculate the dio:d atio, that is, the articulation of each tone. If the ratio is1.00 or almost so, this reflects egato articulation. Ratios less than 1.00 reflect variousdegrees of staccato or portato articulation. The sound level was measured in terms of thedB/Leq value (loudness equivalent level) of the Soundswell program, which is computedas the standard deviation but is expressed in decibels relative to a selected reference level (0dB). The mean tempo for each performance was obtained by dividing the total duration ofthe tune, until its final note, by the number of beats, and then calculating the number ofbeats (i.e., quarternotes) per minute (bpm). The last note was omitted from the calculationsof mean tempo because its d..value cannot be determined.

    Fig. 3. The notation of the melody When the Saints.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    18/37

    EmotionalCommunicationn MusicPerformance 399Table 1Sound Level (dB Leq) for Each Emotional Expression and Performer

    Performer A PerformerB PerformerC M {SD)ExpressionAngry 0 (reference) -I -2 -1 (1.0)Sad -10 -7 -10 -9(1.7)Happy -3 -4 -10 -6(3.8)Fearful -14 -18 -20 -17(3.1)No expression -11 -8 -9 -9(1.5)

    Note. The values ndicate he sound evelof eachperformancen decibelsrelative o thereferenceevel(0) set for the loudestof theperformancesperformerA, angryversion).Results and DiscussionTheexpressivententionof theperformerthe ndependent ariable)hada markedeffecton allof the measured uesin theperformancethedepen-dentvariables).Table1 presents he relativesoundlevel (dBLeq)of the performancesfor each emotionalexpressionand performer.The angryversionswereclearlythe loudestof the performances,ollowed by the happyversions.The sad andwithoutexpressionversionswere softer,whereasthe fearfulversionswere the softestof all. A one-wayanalysisof variance ANOVA,repeatedmeasurements) evealeda statistically significanteffect of theperformer'sxpressive ntention on the relativesoundlevel of the perfor-mances[F(4,8) =9.49;p < .01]. Furthermore, lannedpairwisecompari-sonsof themeanvalues ndicated hat theangry/fearfulontrastwas statis-tically significant F(l, 2) =184.69;p < .01], as was the angry/sad ontrast[F(l, 2) =48.00;p < .05], th bbbbehappy/fearfulontrast F(l, 2) = 94.23;p < .05), and the angry/without xpression[F(l, 2) = 39.06; p < .05] con-trast.However, he remaining ontrastswerenot statistically ignificant.8Table2 showsthe meantempoof the performancesor each emotional

    expressionand performer.As can be seen, the happy(M = 303 beatsperminute)andthe angry(M = 294 bpm)versionsweregenerallyplayedthefastest,followedby the withoutexpressionversions(M = 264 bpm).Thefearfulversionswere slower (M = 186 bpm),and slowest of all were thesadversions M= 98 bpm).Note further hateventhoughtherewerelargeindividualdifferences mongtheperformers,hegeneral rendwithin eachperformerwas prettymuchthe same.A one-wayANOVA(repeatedmeasurements)was computed,but theeffectof the performer'sxpressive ntentionon the meantempofailed toreachstatistical ignificance F(4,8) = 2.13; p < .16]. Thiswas due to the8. However,note thatthe powerwas low in these(andin the following)specificcom-parisonsdueto the few degreesof freedom.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    19/37

    400 PatrikN. JuslinTable 2Mean Tempo for Each Emotional Expression and Performer

    Performer A PerformerB PerformerC M (SD)ExpressionAngry 455 226 201 294(139)Sad 109 73 113 98(22)Happy 260 273 375 303 (63)Fearful 231 137 191 186(47)No expression 135 225 431 264(152)M /performer 238 187 262

    Note. The values indicate he meantempo expressed n beats (i.e., quarternotes)perminute.

    large variabilityamongthe performers ndthe smallnumberof observa-tions.9It is, however,readily apparent romTable2 that the performer'sintentionhad a decisiverolein the choiceof tempo.Plannedpairwisecom-parisonsrevealed hat the happy/sadcontrastwas statistically ignificant[F(l, 2) =40.48; p < .05], as was also the sad/fearfulcontrast[F(l, 2) =25.83; p < .05]. The remaining ontrastswere not significant.Table3 shows the mean and standarddeviationof the articulation thed.Jd..ratio) of the first 16 notes (the first half of the melody)for eachemotionalexpression ndperformer. hemeanarticulation crossperform-ers was mainlylegato for the sad (M = 88%) and angry(M = 80%) ver-sions,whereas t was slightlymore staccatofor thehappy(M= 70%)and,in particular,he fearful(M = 50%) versions.Furthermore,he withoutexpressionversionsweremostlyplayedwithlegatoarticulationM= 81%).Notably, in contrast to findingsobtainedwith other instruments,guitarplayersseem to preferto play angryversionswith legato instead of stac-cato articulation Gabrielsson& Juslin,1996;Juslin,1993). That the dif-ferencesbetweenthe expressionswere not greater s explainedby the factthat the articulationvariationsbroughtaboutby the performerswereap-pliedonlyto certainnotes of themelody,whereasother notes wereplayedmore or less legatoregardless f the expression.However, herewere alsolarge ndividualdifferences mongtheperformers.The articulationmeansof performersA and C were in agreementwith the trend of the mean val-ues, whereasperformerB deviated rom this patternby playingthe angry9. Actually, ffectsizemaybe a moreuseful ndex(Kirk,1996),becauset tellshowlargean effectis, disregardingignificanceevel andsamplesize.Cohen's 1988) d is definedasthedifferencebetween wo meansdividedby theircommonstandarddeviation.Thus,de-spite arge ndividual ifferences mong heperformers,herewerevery argeeffectsizesfor

    themeansof thehappy/sad 1.71 SD), fearful/sad1.51 SD), happy/fearful1.45 SD), an-gry/sad 1.40 SD), andwithoutexpression/sad1.25 SD) versions.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    20/37

    Emotional Communication in Music Performance 401Table 3Mean and StandardDeviation (%) of the Articulation of the First Sixteen

    Notes of the Melody When the Saints for Each PerformerandEmotional ExpressionM{SD)

    PerformerA PerformerB PerformerC ExpressionAngry 70 (24) 91 (8) 80 (25 ) 80 (22)Sad 91(11) 83(14) 90(15) 88(15)Happy 67(16) 67(28) 75(27) 70(24)Fearful 43 (19) 69 (27) 39 (32) 50 (29)No expression 66(13) 86(13) 90 (6) 81(15)

    versionmorelegatothanthe sad version,as well as by playingthe happyversionmorestaccatothan the fearfulversion.A one-wayANOVA (repeatedmeasurements)ndicated a statisticallysignificant veralleffectof theperformer'sxpressivententiononthe meanarticulation f theperformancesF(4,8) = 6.39;p < .05]. Further, lannedpairwise omparisonsndicated hatthehappy/sad ontrastwassignificant[F(l, 2) = 41.3; p < .05], as was also the angry/fearfulontrast[JF(1,2)27.84;p < .05]. None of theremainingpairwisecontrastsweresignificant,although he sad/fearful ontrastwas fairlyclose [P(l, 2) = 10.08;p < .08].Figure4 presentsrepresentative xamplesof amplitudeenvelopesforthe firsteighttonesof fiveperformanceswith differentemotionalexpres-sion (thenotationis shownin Figure3). Theamplitude arbitrary oltageunits) is displayedalong the verticalaxis, and time (seconds)along thehorizontalaxis. The different ime spans should be taken into consider-ationwhencomparing he performances.Thesefiguressimultaneouslyl-lustratethe differences n dynamics,articulation,onsets, and timingde-pendingon the emotionalexpression ntendedbythe performer.Consider,forinstance, heveryabruptonsets of theangryversion; harpand noisyattackshave been shown to affect the frequencyspectrumof the tones(Gabrielsson& Juslin,1996). In contrast,the tones of the fearfulversionare so weak thatthey barelymake a visible mpacton the amplitudeenve-lope.Moreover,hetimingseems farfrom the nominalvaluesof thenota-tion,andthe articulations extremely taccato.Thefiguresalso show thatthe sadversionwas playedrelativelymorelegatothan the happyversion.Finally,note the tidy appearanceof the without expressionversion.Ithasthemost uniform ound evelof allversions,and thetimingseems o becloseto the nominalvaluesof the notation.A more detaileddiscussionof

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    21/37

    402 Patrik N. Juslin

    Fig.4. Representativeexamples of amplitudeenvelopes for the firsteight notes of the melodyWhen the Saints in performances with different emotional expression.

    variousaspectsof timing nexpressive lectricguitarperformancewaspro-videdby GabrielssonandJuslin(1996).SomePreliminarieson theEcologicalValiditiesof theExpressiveCuesAlthoughthe numberof performers,melodies,and cues investigatednthisstudy s limited, t maystillbeof some interest o examinetherelation-

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    22/37

    EmotionalCommunicationn MusicPerformance 403

    Fig.4. Continued

    ships amongthe performers'ntentionsand the expressivecues. Thus, toprovidesomepreliminaryvidenceon theecologicalvaliditiesof theexam-inedcues,thecorrelations r)betweenthe intendedexpressionand thecuelevels were computed(Table4). It is seen that none of the cues wereper-fectlyreliable,althoughthe potentialusefulnessof each cue seems to havevariedquite a lot dependingon the intended emotionalexpression.Forinstance, oudnessseems to have been a moreecologicallyvalid cue forthedecodingof angerthantempo,whereasthe oppositeis truefor the decod-ingof happiness.Althoughthesecorrelationsarebasedon limiteddata,it

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    23/37

    404 Parrik N. Juslin

    Fig. 4. Continued

    should be noted that there are many similarities between these ecologicalvalidities and the functional validities (i.e., relative weights in cue utiliza-tion) that were obtained byJuslin (1996a, 1996b). However,a generalanaly-sis of the ecological validities of the cues should include much more data,includingseveralperformers,melodies, cues, and instruments(Juslin, 1997).One of the questions posed by the functionalist approach was whetherthe expressive cues are intercorrelated or not. This seems indeed to be thecase in the present experiment; cue intercorrelations are evident for loud-

    Tabi.i;4The Ecological Validities (r ) of the Expressive Cues in the EncodingExperimentFmotional Expression

    Cue Angry Sad Happy Fearful Overall M rTempo .40 -.66 .44 -.IS I .2 1 ILoudness .66 -.07 .24 -.So1 I .M) IArticulation .29 .56 -.09 -. ~ I .25 I

    Note. The values indicate the product moment correlations between the intended emo-tional expression (dichotomous coding) ,\nd the values of the cues as measured in the result-ing performances. The right-hand column shows the mean coefficient of determination (r1)across emotional expression for each cue (i.e., disregarding the direction of the relation-ship).

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    24/37

    EmotionalCommunicationn MusicPerformance 405ness/tempo(r = .41), loudness/articulationr = .52), and tempo/articula-tion(r=-.25). Intercorrelationsmongcuesshouldbeconsiderednevalu-atingtheirultimateusefulness Cooksey,1996). The exact values of thesecorrelations anobviouslynot begeneralizedbeyondthisexperiment.Theimportantpoint, however,is that there are intercorrelations mong thecues.Presumably,ueredundancy nabled isteners o decode the intendedexpressiondespite arge ndividualdifferences mongthe performers.

    THEDECODING XPERIMENTMethodsParticipantsTwenty-fourtudents,21-52 yearsold (M = 25.4 years),participated n a voluntarybasis.Half of thesubjectswas musically rained, uch as musiciansor musicstudents.Theotherhalfcompriseduntrainedisteners, hat is, peoplewithoutanyclose involvementnmusic.Thesubjectswereevenlydistributedwithrespect o genderand werepaidfor theiranonymousparticipation.ProcedureThesubjectwas instructedo ratethe emotionalexpressionof eachperformance f themelody When he Saints on fiveadjective cales;happy, ad,angry, earful,andexpres-sive.Theregistrationsweremadeby meansof a speciallydesignedcomputerprogram orquantitativeatingsof auditory timuli.The instructionsweregivenas follows:Youwill now listen o severaldifferent ersionsof a well-knownmelody.Yourtaskis to describe heemotionalexpressionof eachversionof themelody,byindicating valueon eachof thefiveadjectivecales hatyoucan seein frontofyou. The more of a particular motion(forinstance happiness ) ou thinkthat a particular ersion s expressing,hehighervalueyou should ndicateonthe corresponding djective cale.Tenrepresentsmaximumof this emotionandzerorepresentsminimum. naddition o the fourwords that refer o spe-cific emotions,thereis also the adjective expressive. This meansthat youshould ndicatehowexpressiveregardlessf theparticularmotions xpressed)youthink thateachversion s on the above-mentionedcale fromzero to ten.Note thatyou do not haveto hurry.Theimportanthingis thatyou aresatis-fiedwithyourratingsof eachversionbeforeyou move on to the next. Whenyouhavecompletedheratingsof a particular ersion,you simplyclick on thenext button;this will start the next version.It may happenthat you find itdifficult o complete heratingsof a particular ersionbefore tsplaybackhascometo an end.Youmaythenrepeat heplayback f thisparticular ersionbyclickingon therepeatbutton.In order oryouto familiarize ourselfwith thetestprocedure, ou will first be givena shortpre-testwith someexamples. nthe actualexperiment, ou will listen to 15 differentversionsof the melody.Theprinciples the samethroughoutheexperiment.Are thereanyquestions?Thesubjectsistened o the stimuli hrougha loudspeakerAudioProA 4-14), andthesoundlevelwas adjusted o that it would roughlycorrespond o the sound level at the

    recordingession.Theorderof the stimuliwas randomizedor eachsubject,andso was theorderof theadjective cales.Anexperimentalession astedapproximately 0 min.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    25/37

    406 Patrik N. JuslinDesign and Data TreatmentA split-plot design was used, with repeated measurements for emotional expression (5

    levels) and performer (3 levels), and between-groups design for the classificatory variablesgender (2 levels) and musical training (2 levels). The data were subjected to analysis ofvariance for each of the five adjective scales featuring four sources: expression, performer,gender,and musical training. The F tests were conducted by using a fixed-effects model.Results and DiscussionThe emotionalexpressionof the performances the independentvari-able)hada considerable ffect on the listeners' atingsof theperformances(thedependentvariable).Figure5 presents he meanratingsof the emotionalexpressionsacrossperformers,eparatelyor eachadjective cale.As can beseen,the listenerswere rather uccessfulndiscriminatinghe intended motionalexpression.Foreach emotionalexpression, he meanratingsareclearlyhigheston theadjective calecorrespondingo the emotionalexpression ntendedby theperformers.This is true for all versionsexcept the ones intendedto bewithoutexpression,which received ow ratingson all adjective cales.Onthewhole, happiness,anger,and sadnessseem to have beenslightlyeasierto communicateo listeners han fear.Obviously, he numberof responseoptionswere limited n this experi-

    ment. On the otherhand,it should be notedthat quantitative atingsareless biased than forced choice judgments n that they do not implythatonly one emotion can be expressedat the time.Thus,it seems fairto saythat the listenersrespondedconsistentlyand categorically o the perfor-mancesandthattheydid so in accordancewith theperformer'sxpressiveintention.The dataweresubjected o a four-wayANOVA(seeData Treatment),anda summaryof the resultsappears n Table5. The main effect of emo-tionalexpressionwashighlysignificant p< .00001) for alladjective cales.In addition,post hoc tests (Fisher'seast significantdifference)revealedthat, for all adjectivescales,the meanratingsof the correct motionalexpressionweresignificantly ifferent romthe meanratingsof all theotherexpressions.The main effect of performerwas also significant,althoughon a lesser evel.Likewise, he interactionsbetweenemotionalexpressionand performerwere significantfor all adjectivescalesexcept the happyscale.However,no significantdifferences nvolved eithermusical rainingor gender. nteractions mongthe fourvariableswerefew,apartfrom theaforementioned erformerby expression nteractions,husmaking he in-terpretation f the maineffectsrather traightforward. heonlyexceptionwas a marginally ignificantmusicaltrainingby emotionalexpression n-teraction[F(4, 80) =2.58;p < .05] on the angerscale.However, his inter-actiondid notreflectanactualdifferencen decodingaccuracybetween hetwo groups.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    26/37

    Fig. 5. Mean ratings of each emotional expression across performers.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    27/37

    408 PatrikN. Juslin

    Fig.5. ContinuedItwas shown earlier hat theperformers sedsomewhatdifferent ncod-ingstrategies.Similarly,he significantperformerbyemotionalexpressioninteractionsndicate hat there were individualdifferencesamongthe per-formers n theirabilityto communicate he various emotions.Theseper-formerbyemotionalexpressionnteractions reshowninFigure6. Clearly,allperformers ad some successat theirtask.However, heversionsplayedby performerA wereslightly ess able to communicateheprescribed mo-

    tions to the listenersthan the versionsby both performersB and C. Thebest versions,in terms of successfulcommunication,are the ones that

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    28/37

    EmotionalCommunicationn MusicPerformance 409Table 5Summary of the Results from the Analysis of Variance for Each

    Adjective ScaleScale and Source F{df) MSEHappinessPerformer 4.13(2,40)* 4.40Emotional xpression 42.32 (4, 80)** 5.36Performer emotion Not significant -SadnessPerformer 4.36 (2, 40)* 4.38Emotional xpression 39.55 (4, 80)*** 6.52Performer emotion 5.75 (8, 160)*** 3.50Anger

    Performer 7.50 (2, 40)** 3.45Emotional xpression 88.56 (4, 80)*** 4.72Performer emotion 2.43 (8, 160)* 3.12FearPerformer 12.30 (2, 40)*** 5.32Emotional xpression 29.98 (4, 80)*** 5.92Performer emotion 5.73 (8, 160)*** 3.33ExpressivenessPerformer 8.14(2,40)** 3.15Emotional xpression 34.31(4,80)*** 4.13Performer emotion 3.44(8,160)** 3.14Note. Significanceevels:> < .05, *> < .01, **>< .00001

    havehighratingson theadjective calecorrespondingo the intendedemo-tionalexpressionand low ratingson all otherscales.Pairwise omparisonsindicated hat themajorityof the differencesbetweenthe performers' er-sions were statisticallysignificantat the p < .01 level. Interestingly,heperformer'successin communicating he emotionsto listenerswas notrelated o how long he had beenplayingthe guitar.The versions ntended o be withoutexpressionwereactuallyratedsig-nificantly ess expressive han the otherversions,despitethe fact that thelistenerswerenot informed hat performancesdevoidof expressionwereincluded n the experiment.This shows that the listenersdid not simplychoose one of the providedadjectivescales.If they found that a certainversiondid not soundverymuchlikeany of the providedemotionlabels,theywere ableto givelow ratingson all of the scales.However, heperfor-manceswithout expressionwere still rated as expressive o some extent.This doesperhapsreflect he melodicstructure f thepiece,or the factthatit may be impossible o generatea performance ompletelydevoid of ex-pression.Anyway, heresultsreplicateKendallandCarterette's1990)find-ing that listenersare ableto discerndifferent evelsof expressive ntent inmusicalperformances.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    29/37

    410 PatrikN. Juslin

    Fig.6. Theperformer y emotionalexpressionnteractionor eachadjective cale.The most expressive versions, according to the listeners, were the sad (M= 7.0) and angry (At = 6.7) versions, followed by the happy (At= 5.9) andfearful (At= 5.3) versions. The without expression versions received a con-siderably lower rating (At = 3.4). The low ratings to the fearful versionsmay be related to the fact that these versions also were a little more diffi-

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    30/37

    EmotionalCommunicationn MusicPerformance 411

    Fig.6. Continuedcult to decode than the otherversions.Still,considering he currentme-dium,thecommunication f fearworkedsurprisinglywell.Therelationshipbetweendecodingaccuracyandgenderwas not statis-tically significant or any of the five adjectivescales, and no significantinteractions nvolving genderwere found. However,carefulobservation

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    31/37

    412 PatrikN. Juslinrevealed hat the meanratingsby the femalesactuallywereslightlymoreaccuratehan theratingsbythe males.Thiswas the casein 19 (76%)of the25 instances,whereasno differenceswere found in the remaining ondi-tions.Themagnitudeof the differenceswas largest or the happyandsadscales. In a similarstudy,Gabrielssonand Juslin (1996) found that theratingsbythe femalesweremore accurate n 89%of the instances.Again,though, the differencesdid not reach statisticalsignificance,perhapsbe-cause of the largewithin-groupvariance.However, t shouldbenoted thatthe split-plotdesign(whichwas usedin bothstudies) endsto increase hepowerof the repeatedmeasurementsariableat the expenseof the powerof the between-groupsvariable(e.g., Kirk, 1982). The issue of whethertherearegenderdifferencesndecodingaccuracy an thusbe resolvedonlythroughfurtherstudies,preferablyusing point estimatesof the effectsize(Kirk,1996).

    General DiscussionResearchersnterestedn musicperformance ave beenreluctanto turnto thepsychologyof emotionfor theoreticalguidance.One reasonfor thismaybe that most music researchers renot familiarwith the literature n

    emotion. Anotherreasonmaybe thatthepsychologyof emotionfor a longtime has been a ratherconfusingareaof research(e.g., Plutchik,1994).Whatever hereasons or this state of affairs, hepresentpaperconstitutesan attempt o show that the functionalistapproachcan be heuristic n thestudyof emotionalcommunicationn musicperformance.What,then,hasbeen learned rom this study?First, t is evidentthatthe performers' xpressive ntentionshad a deci-sive impacton all of the measuredcues in the performances.The perfor-manceanalysis howed that(a)angerwas associatedwithhighsound evel,fast tempo,and legatoarticulation, b) sadnesswas associatedwith slowtempo,legatoarticulation,and low soundlevel, (c) happinesswas associ-atedwith fast tempo, high soundlevel,and staccatoarticulation,and (d)fear was associatedwith low sound level, staccatoarticulation,and slowtempo.Second,none of the expressivecues hadmore than a probabilistic ela-tion to the expressivententionsof the performers.Although heperform-ers' intentionsand the expressivecues weresystematicallyelated, hepo-tential usefulness of each cue varied a lot dependingon the emotionalexpression ntendedby the performer.This is consistentwith the resultsfrom studiesof vocal expressionof emotion(Sundberg, 982).Third,the expressivecues were intercorrelated.This shows that therewas an elementof redundancyn the communicativeprocess.It mightbe

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    32/37

    EmotionalCommunicationn MusicPerformance 413assumedhat cueredundancy ecreased heuncertaintyntroducedbylargeindividualdifferences mongthe performers.

    Fourth,the listenerswere able to successfullydiscriminate he perfor-mancesexpressinganger,happiness, adness,andfear from eachother,aswell asfromperformancesntended o bewithoutexpression.Thelistenersrespondedconsistentlyand categorically o the expressiveperformances,andtheydid so in accordancewith the performer'sxpressive ntention.Fifth,musically rained istenersdid not show a higher decodingaccu-racy hanuntrained isteners nthedecodingexperiment.Thissupports hehypothesis hat decodingof basicemotionalexpressions n musicperfor-mance s largely ndependent f musical raining.Sixth, hefemaleshadhigherdecodingaccuracyhanthemales,althoughthis differencewas small anddid not reachstatistical ignificance.Slightlylargergenderdifferencesn the same directionwere foundby GabrielssonandJuslin(1996), butthese differencesalso failedto reachstatisticalsig-nificance.This issueawaitsfurtherresearch.Studiesof emotionalcommunicationn musicperformancepresentre-searcherswith a numberof problems.A complicatingactor n thepresentstudyis that the structureof the piecein itself carriesa certainemotionalcharacter,whichpresumablymakes t moreorless difficult o achieveotheremotionalexpressions n the performance.However, t may also be thattheperformancetructure nd themelodystructure perateon partly nde-pendent evels,somewhat iketheverbaland nonverbalaspectsof speech.Because,as Shaffer 1992) notes aboutthe performance f a pieceof mu-sic, aperformer an be faithfulto its structureandat the same time havethe freedom o shape ts moods (p.265). Thetheoretical rameworkpro-posedhere s thuscomplementaryo earliermodels on musicperformancethathave focusedon how performance xpressionreflectsvariousaspectsof themusical tructureClarke,1988;Clynes,1987;Friberg, 995;Kendall&C arterette, 990; Sundberg,1988; Todd, 1985; 1989).Anotherproblemconcernsthe ambiguityof emotion words. It is, forinstance,possiblethat the performersn this studydid not interpret heemotionwordsof the instructionsn exactlythe sameway. Notably,someresearchersavesuggesteda scenarioapproachas awayof coming o termswiththisproblem(e.g.,Baroni& Finarelli,1994; Scherer, 991;Williams& Stevens,1972). Bycreatinga specificcontext, it is hopedthat the per-formersor actorswill interpret he meaningof the emotion words in amoreuniformway.Whether hisactuallyhappensmaybe somewhathardto determine.However,vanBezooijen 1984) found thattheuseof specificcontexts ailedbecause hesuitability f thesecontextsappearedo behighlypersondependent.Furthermore,f a particular cenario s used,does thismeanthat the resultsobtainedcannot be generalizedbeyondthis narrowcontext?If the assumptionsof the functionalistapproacharecorrect,the

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    33/37

    414 PatrikN. Juslinissueon theinterpretation f emotion wordsshould not bethatmuchof aproblem.This is because he involvedcommunicationystemmaynot havethe required requencyof precision n orderto communicate uch subtlenuancesreliablyanyway. tmayverywellbethatthe basicemotioncatego-ries are as exact as this nonverbalcommunication ystemcan get. How-ever,furtherresearch s clearlyneededin orderto definethe limits of theinformation hat can be reliablycommunicated hroughmusicalperfor-mance.It mightreasonablybe assumedthat the ease or difficultywith whichvarious emotions are communicateddepends,at least in part,on the in-strumentused. Each instrumentpresents ts own possibilitiesand limita-tions in termsof the expressivemeans available o the performer. omeoftheexpressivevariations oundheremaybetypicalof theparticularnstru-mentused.Forexample,contrary o the findingsobtainedwith otherin-struments,guitarplayersseem to prefer o playangryversionswith legatoarticulation.Accordingly,t seemsimportant o conductstudiesof emo-tionalexpressionwith a varietyof instruments Behrens&cGreen,1993;Gabrielsson& Juslin,1996). In fact, it seemshighlyunlikely hatwe willfind completelygeneralrules, independentof the melody,performer,n-strument,or listener.An importantquestionwithrespect o the presentstudyis whetherper-formerseverhaveexpressivententionsof the kindusedhere.A reasonableanswer s that musiciansdo not consciously hinkin theseterms,but thattheynonetheless end to use the describedprinciplesmoreor lessimplicitlyin orderto communicateor amplifya particularmood or emotionalex-pression hat seemsappropriateorthemusicalpieceathand.Thus,in anygivenmusicalperformance,we would expectto findthatsomeexpressivevariationsare used to conveythe structuralnterpretation,whereasotherexpressivevariations unctionto providethe performancewith an appro-priateemotionalexpression.Thepresent tudyinvestigatedhow theperformer's xpressivententionaffectstempo,loudness,andarticulationn theperformance.These resultsgenerally eplicate hefindings romour earlier tudies(Gabrielsson, 994;1995;Gabrielsson& Juslin,1996;Gabrielsson& Lindstrm, 995;Juslin,1993). However,systematicvariations n timing,intonation,and timbrehave also beenfoundto be usedbyperformers e.g.,Gabrielsson& Juslin,1996).Wehave hus shownthat a number f expressiveues tempo,soundlevel, timing,timbre, articulation,and intonation arepotentiallyusefulfor the decodingof emotionalexpression n musicperformance.However,althoughperformancenalyses how thatperformers se thesecues to com-municateemotionsto listeners, heycannot determinewhich cues are actu-allyusedby the listener.The next step, according o the functionalistper-spective,s thusto investigatehe listener's se of cues(Juslin, 996a,1996b),

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    34/37

    Emotional Communication in Music Performance 415so that encodingand decoding aspects may be systematically elatedbyusingthe lensmodelequation(Juslin,1997).

    Perhaps he most importantconclusionto be drawnfrom the presentinvestigations that, indeed, it is too simplistic o reduceperformanceothe articulationof structure Clarke,1994, p. 3). It has beenfound thatprofessionalguitarplayers, ar fromjustpresentingwhat is implicit n themusicalscore,are ableto communicatebasicemotionsto listeners hroughtheirspecificperformances f a pieceof music.Thus, the messageof thispapermust be that in orderto trulyunderstandwhy humanbeings per-formmusic n theparticularways theydo, we musttakeinto accounta lotmore than merelythe constraints hat the musical score imposeson theperformance f a given pieceof music.10

    10. I thank Edward Carterette, Roger Kendall, Bruno Repp, and an anonymous re-viewer for useful comments on a preliminaryversion of the manuscript. I am also indebtedto Alf Gabrielsson for his great support and to Anders Winman for valuable help withcomputer programming. Part of this paper was presented at the Conference of DeutscheGesellschaftfrMusikpsychologie and EuropeanSocietyfor the CognitiveSciences of Music,Bremen, Germany, September 12-15, 1995. The research was supported by The Bank ofSweden TercentenaryFoundation through a grant to Alf Gabrielsson.

    ReferencesBaroni, M., & Finarelli,L. (1994). Emotions in spoken language and in vocal music. In: I.Delige (Ed.), Proceedings of the Third International Conference for Music Perceptionand Cognition (pp. 343-345). Lige, Belgium: University of Lige.Behrens, G. A., & Green, S. B. (1993). The ability to identify emotional content of soloimprovisationsperformedvocally and on threedifferent nstruments.Psychology of Music,21, 20-33.Bengtsson, I., & Gabrielsson, A. (1980). Methods for analyzing performances of musicalrhythm. ScandinavianJournal of Psychology, 21, 257-268.Bengtsson, I., Gabrielsson, A., & Thorsn, S. M. (1969). Empirisk rytmforskning (Empiri-cal rhythm research).Swedish Journal of Musicology, 51, 49-118.van Bezooijen, R. A. M. G. (1984). Characteristicsand recognizability of vocal expressionsof emotion. Dordrecht, The Netherlands: ICG Printing.Brehmer,B. (1976). Social judgment theory and the analysis of interpersonal conflict. Psy-chological Bulletin, 83, 985-1003.Brehmer,B. (1988). The development of social judgment theory. In: B. Brehmer,& C. R. B.Joyce (Eds.), Human judgment: The SJTview (pp. 13-40). Amsterdam:North Holland.Brunswik, E. (1955). Representative design and probabilistic theory in a functional psy-chology. Psychological Review, 62, 193-217.Brunswik,E. (1956). Perceptionand the representativedesign of psychological experiments.Berkeley: University of California Press.Buck, R. (1984). The communication of emotion. New York: Guildford Press.Buss, D. M. (1995). Evolutionary psychology: A new paradigm for psychological science.Psychological Inquiry, 1, 1-30.Clarke, E. F. (1988). Generative principles in music performance. In:J. A. Sloboda (Ed.),Generativeprocesses in music. Thepsychology of performance, improvisation, and com-position (pp. 1-26). Oxford: Clarendon Press.Clarke, E. E (1989). The perception of expressive timing in music. Psychological Research,51, 2-9.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    35/37

    416 PatrikN. JuslinClarke,E. E (1994).Perspectivesnexpressionnmusicalperformance.n:I.Delige Ed.),Proceedings f the Third nternationalConferenceor MusicPerception nd Cognition(pp.1-4). Lige,Belgium:University f Lige.Clarke,E. F.(1996).Expressionnperformance:Generativity, erception, nd semiosis. n:J. Rink(Ed.),Thepracticeof performance pp.21-54). Cambridge:CambridgeUniver-sityPress.Clynes,M. (1977).Senties:The touchof emotions.New York:AnchorPress/Doubleday.Clynes,M. (1987). What can a musician earnabout musicperformanceromnewlydis-coveredmicrostructurerinciples PMandPAS)? n: A. GabrielssonEd.),Action andperceptionn rhythmand music(pp.201-233). Stockholm:RoyalSwedishAcademyofMusic,No. 55.Cohen,J. (1988).Statistical oweranalysis orthe behavioral ciences 2nded.).New York:AcademicPress.Cooksey,R. W.(1996)judgmentanalysis.Theory,methods,andapplications.New York:AcademicPress.Darwin,C. (1872). The expressionof the emotions in man and animals. London:JohnMurray.Dimberg,U. (1988).Facialexpressions nd emotionalreactions:A psychobiologicalnaly-sis of humansocial behavior.n:H. L.Wagner Ed.),Socialpsych physiology ndemo-tion:Theoryand clinicalapplication pp. 131-150). New York: ohn Wiley6c Sons.Dowling,W.L, 6c Harwood,D. L. (1986).Musiccognition.New York:AcademicPress.Eisenberg,N., & Lennon,R. (1983). Sex differencesn empathyand relatedcapacities.PsychologicalBulletin,94, 100-131.Ekman,P. (1984). Expressionand the natureof emotion. In: K. R. Scherer,& P. Ekman(Eds.),Approacheso emotion(pp.329-343). Hillsdale,NJ:Erlbaum.Ekman,P.(1992a).Anargumentor basicemotions.CognitionandEmotion,6, 169-200.Ekman,P.(1992b).Are therebasic emotions?PsychologicalReview,99, 550-553.Fischer,K.W, Shaver,P.R., 6c Carnohan,P.(1990).How emotionsdevelopandhow theyorganizedevelopment.CognitionandEmotion,4, 81-127.Friberg,A. (1995).A quantitative ulesystem or musicalperformance.Unpublished oc-toraldissertation,RoyalInstituteof Technology, tockholm.Frick,R. W.(1985).Communicatingmotion:The role of prosodie eatures.PsychologicalBulletin,97, 412-429.Fridja,N. H. (1986). The emotions.Cambridge:CambridgeUniversityPress.Gabrielsson,A. (1985).Interplay etweenanalysisandsynthesisnstudiesof musicperfor-manceandmusicexperience.MusicPerception, , 59-86.Gabrielsson,A. (1988).Timing nmusicperformancenditsrelations o musicexperience.In:J. A. Sloboda(Ed.),Generative rocesses n music. Thepsychologyof performance,improvisation, ndcomposition pp.27-51). Oxford:ClarendonPress.Gabrielsson,A. (1994). Intentionandemotionalexpression n musicperformance.n: A.

    Friberg, . Iwarsson,E.Jansson,6c J. Sundberg Eds.),Proceedings f the StockholmMusicAcousticsConference,uly1993 (pp.108-111). Stockholm:RoyalSwedishAcad-emyof Music.Gabrielsson,A. (1995). Expressivententionandperformance.n: R. SteinbergEd.),Mu-sic and themindmachine.Psych physiologyandpsych pathologyof the senseof music(pp.35-47). Heidelberg: pringerVerlag.Gabrielsson,A. (in press).Music performance.n: D. Deutsch(Ed.),Thepsychologyofmusic(2nded.).New York:AcademicPress.Gabrielsson,A., 6cJuslin,P.N. (1996). Emotionalexpression n musicperformance: e-tween the performer'sntentionandthe listener's xperience.Psychologyof Music,24,68-91.Gabrielsson,A., Juslin,P.N., 6c Lindstrm,E. (1994).Expressivententiongovernsmusicperformance.n:I. Delige(Ed.),Proceedings f the Third nternationalConferenceorMusicPerception nd Cognition pp. 19-20). Lige,Belgium:University f Lige.Gabrielsson, .,6cLindstrm, .(1995).Emotionalxpressionnsynthesizerndsentographperformance. sychomusicology,4, 94-116.

    This content downloaded from 137.222.155.224 on Thu, 3 Apr 2014 06:30:53 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/12/2019 Juslin_1997b

    36/37

    EmotionalCommunicationn MusicPerformance 417Grice,H. P.(1986).Utterer'smeaning, entence-meaning,ndword-meaning. oundationsof Language, , 1-18.Hall,J. A. (1978). Gendereffects n decodingnonverbal ues.PsychologicalBulletin,85,845-857.Hammond,K. R.(Ed.) 1966).Thepsychology f EgonBrunswik.New York:Holt, Rinehart,& Winston.Hodges,D. A. (1989).Whyarewe musical?Speculationsn theevolutionary lausibility fmusicalbehavior.Bulletinof the Council or Researchn MusicEducation,99, 7-22.Hursch,C.J., Hammond,K. R., & Hursch,J. L. (1964). Somemethodological onsider-ations n multiple-cue robability tudies.PsychologicalReview,71, 42-60.Izard,C. E. (1977). The emotions.New York:PlenumPress.Izard,C.E. (1993).Organizationalnd motivational unctionsof discrete motions.In:M.Lewis,& J. M. Haviland(Eds.),Handbookof emotions(pp. 631-643). New York:GuildfordPress.Jessor,R., & Jessor,S. L. (1977).Problembehaviorandpsychosocialdevelopment:A lon-

    gitudinal tudyof youth.New York:AcademicPress.Johnson-Laird, .N., & Oatley,K. (1992). Basicemotions,rationality, nd folk theory.CognitionandEmotion,6, 201-223.Juslin,P. N. (1993). Theinfluenceof expressivententionon electricguitarperformance.UnpublishedC-level hesis,Department f Psychology,UppsalaUniversity, weden.Juslin,P. N. (1995). Emotionalcommunicationn music viewedthrougha Brunswikianlens.In: G. Kleinen Ed.),Proceedings f the Conference f DeutscheGesellschaft rMusikpsychologiendEuropeanSociety ortheCognitiveSciences f Music,September1995 (pp. 21-25). University f Bremen,Germa