University Press Scholarship Online Oxford Scholarship ...

Post on 22-Oct-2021

14 views 0 download

Transcript of University Press Scholarship Online Oxford Scholarship ...

Rhythm

Page 1 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

UniversityPressScholarshipOnline

OxfordScholarshipOnline

Music,Language,andtheBrainAniruddhD.Patel

Printpublicationdate:2007PrintISBN-13:9780195123753PublishedtoOxfordScholarshipOnline:March2012DOI:10.1093/acprof:oso/9780195123753.001.0001

Rhythm

AniruddhD.Patel

DOI:10.1093/acprof:oso/9780195123753.003.0003

AbstractandKeywords

Althoughtherehavealreadybeenseveralstudiesthatinvestigatedhowrhythmisintegratedwiththedomainsofbothmusicandspeech,therehaverarelybeencomparativestudiesregardingmusicalrhythmandlinguisticrhythm.Thelackofsuchcomparativeresearchmaybeattributedtohowresearchersofonedomainbarelyhavetimetolookintothecomplexitiesoftheothersinceeachpossessesitsownlargeshareofcomplexities,ratherthanlackofinterest.Assuchresearchmayenableustogainawiderviewonhowrhythmplaysacentralpartinhumancognition,oneofthegoalsofthischapterinvolvesprovidingbothempiricalandconceptualtoolsforresearcherstoexaminethelinksbetweenmusicalandlinguisticrhythm.Thischapterclarifiesthedefinitionof“rhythm”,introducesthenotionofrhythminspeech,andprovidesanaccountregardingcross-domaincomparisonsofrhythm.

Rhythm

Page 2 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Keywords:rhythm,music,speech,musicalrhythm,linguisticrhythm,comparativeresearch

Chapter3Rhythm3.1Introduction963.2RhythminMusic97

3.2.1TheBeat:AStableMentalPeriodicity993.2.2Meter:MultiplePeriodicities1033.2.3Grouping:ThePerceptualSegmentationofEvents1063.2.4DurationalPatterninginMusic112DurationCategoriesinMusic112ExpressiveTiminginMusic1143.2.5ThePsychologicalDimensionsofMusicalRhythm117

3.3RhythminSpeech1183.3.1RhythmicTypology119PeriodicityandTypology119PhonologyandTypology122DurationandTypology126PerceptionandTypology1353.3.2PrinciplesGoverningtheRhythmicShapeofWordsandUtterances138DifferencesBetweenLinguisticandMusicalMetricalGrids141QuestioningthePrincipleofRhythmicAlternationinSpeech1413.3.3ThePerceptionofSpeechRhythm143ThePerceptionofIsochronyinSpeech143TheRoleofRhythmicPredictabilityinSpeechPerception145TheRoleofRhythminSegmentingConnectedSpeech147TheRoleofRhythminthePerceptionofNonnativeAccents1483.3.4FinalCommentsonSpeechRhythm:MovingBeyondIsochrony149

3.4Interlude:RhythminPoetryandSong1543.4.1RhythminPoetry1543.4.2RhythminSong156

3.5NonperiodicAspectsofRhythmasaKeyLink1593.5.1RelationsBetweenMusicalStructureandLinguisticRhythm1593.5.2RelationsBetweenNonlinguisticRhythmPerceptionandSpeechRhythm1683.5.3NeuralRelationshipsBetweenRhythminSpeechandMusic173

3.6Conclusion176

Appendixes

Rhythm

Page 3 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

A.1ThenPVIEquation177A.2MusicalnPVIValuesofDifferentNations178

(p.96) 3.1IntroductionThecomparativestudyofspokenandmusicalrhythmissurprisinglyunderdeveloped.Althoughhundredsofstudieshaveexploredrhythmwithineachdomain,empiricalcomparisonsoflinguisticandmusicalrhythmarerare.Thisdoesnotreflectalackofinterest,becauseresearchershavelongnotedconnectionsbetweentheoriesofrhythminthetwodomains(e.g.,Selkirk,1984;Handel,1989).Thepaucityofcomparativeresearchprobablyreflectsthefactthatspecialistsinonedomainseldomhavethetimetodelveintotheintricaciesoftheother.Thisisregrettable,becausecross-domainworkcanprovideabroaderperspectiveonrhythminhumancognition.Onegoalofthischapteristoequipresearcherswithconceptualandempiricaltoolstoexploretheborderlandbetweenlinguisticandmusicalrhythm.Asweshallsee,thisisafertileareafornewdiscoveries.

Beforeembarking,itisworthaddressingtwooverarchingissues.Thefirstisthedefinitionofrhythm.Theterm“rhythm”occursinmanycontextsbesidesspeechandmusic,suchascircadianrhythms,oscillationsinthebrain,andtherhythmiccallsofcertainanimals.Inmostofthesecontexts,“rhythm”denotesperiodicity,inotherwords,apatternrepeatingregularlyintime.Althoughperiodicityisanimportantaspectofrhythm,itiscrucialtodistinguishbetweenthetwoconcepts.Thecruxofthematterissimplythis:Althoughallperiodicpatternsarerhythmic,notallrhythmicpatternsareperiodic.Thatis,periodicityisbutonetypeofrhythmicorganization.Thispointisespeciallyimportantforunderstandingspeechrhythm,whichhashadalong(andasweshallsee,largelyunfruitful)associationwiththenotionofperiodicity.Thusanydefinitionofrhythmshouldleaveopentheissueofperiodicity.Unfortunately,thereisnouniversallyaccepteddefinitionofrhythm.ThusIwilldefinerhythmasthesystematicpatterningofsoundintermsoftiming,accent,andgrouping.Bothspeechandmusicarecharacterizedbysystematictemporal,accentual,andphrasalpatterning.Howdothesepatternscompare?Whatistheirrelationshipinthemind?

(p.97) Thesecondissueistheverynotionofrhythminspeech,whichmaybeunfamiliartosomereaders.Onewaytoinformallyintroducethisconceptistoconsidertheprocessoflearningaforeignlanguage.Speakingalanguagewithnativefluencyrequiresmorethanmasteringitsphonemes,vocabulary,andgrammar.Onemustalsomasterthepatternsoftimingandaccentuationthatcharacterizetheflowofsyllablesinsentences.Thatis,eachlanguagehasarhythmthatispartofitssonicstructure,andanimplicitknowledgeofthisrhythmispartofaspeaker’scompetenceintheirlanguage.Afailuretoacquirenativerhythmisanimportantfactorincreatingaforeignaccentinspeech(Taylor,1981;Faber,1986;Chela-Flores,1994).

Thefollowingtwosections(3.2and3.3)giveoverviewsofrhythminmusicandspeech,respectively,focusingonissuespertinenttocross-domaincomparisons.(Suchcomparisonsaremadewithineachsectionwhereappropriate.)Theseoverviewsmotivateaparticularwayoflookingatrhythmicrelationsbetweenspeechandmusic.Thisnew

Rhythm

Page 4 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

perspectiveisintroducedinthefinalsectionofthechapter,togetherwithempiricalevidencespanningacoustic,perceptual,andneuralstudies.

3.2RhythminMusicThefollowingdiscussionofrhythminmusicfocusesonmusicthathasaregularlytimedbeat,aperceptuallyisochronouspulsetowhichonecansynchronizewithperiodicmovementssuchastapsorfootfalls.Furthermore,thefocusisonmusicoftheWesternEuropeantradition,inwhichbeatsareorganizedinhierarchiesofbeatstrength,withalternationbetweenstrongerandweakerbeats.Thisformofrhythmicorganizationhasbeenthemostwidelystudiedfromatheoreticalandempiricalstandpoint,andisalsothetypeofrhythmmostoftencomparedwithspeech,eitherimplicitlyorexplicitly(Pike,1945;Liberman,1975;Selkirk,1984).

Itisimportanttorealize,however,thatthisisjustonewayinwhichhumansorganizemusicalrhythm.ItwouldbeconvenientiftherhythmicstructureofWesternmusicindicatedgeneralprinciplesofrhythmicpatterning.Realityismorecomplex,however,andonlyacomparisonofdifferentculturaltraditionscanhelpsiftwhatisuniversalfromwhatisparticular.Toillustratethispoint,onecannotemusicaltraditionsinwhichrhythmisorganizedinratherdifferentwaysthaninmostWesternEuropeanmusic.

OnesuchtraditioninvolvestheCh’in,asevenstringfretlesszitherthathasbeenplayedinChinaforover2,000years(vanGulik,1940).Themusicalnotationforthisinstrumentcontainsnotimemarkingsforindividualnotes,indicatingonlythestringandtypeofgestureusedtoproducethenote(thoughsometimesphraseboundariesaremarked).Theresultingmusichasnosenseofabeat.Instead,ithasaflowingqualityinwhichthetimingofnotesemergesfromthegesturaldynamicsofthehandsratherthanfromanexplicitlyregulated(p.98) temporalscheme.TheCh’inisjustoneofmanyexamplesofunpulsedmusicfromaroundtheglobe,allofwhichshowthatthemindiscapableoforganizingtemporalpatternswithoutreferencetoabeat.

AnothertraditionwhoserhythmsarequitedifferentfromWesternEuropeanmusicisBalkanfolkmusicfromEasternEurope(Singer,1974;London,1995).Thismusichassalientbeats,butthebeatsarenotspacedatregulartemporalintervals.Instead,intervalsbetweenbeatsareeitherlongorshort,withthelongintervalbeing3/2thelengthoftheshorterone.Rhythmiccyclesarebuiltfromrepeatingpatternsoflongandshortintervals,suchasS-S-S-L,S-S-L-S-S(notethatthelongelementisnotconstrainedtooccurattheendofthecycle).Onemightthinkthatsuchanasymmetricstructurewouldmakethemusicdifficulttofolloworsynchronizewith.Infact,listenerswhogrewupwiththismusicareadeptatfollowingthesecomplexmeters(Hannon&Trehub,2005),andmuchofthismusicisactuallydancemusic,inwhichfootfallsaresynchronizedtotheasymmetricbeats.

AsafinalexampleofarhythmictraditionwithadifferentorientationfromWesternEuropeanmusic,GhaniandrumminginWestAfricashowsanumberofinterestingfeatures.First,thebasicrhythmicreferenceisarepeating,non-isochronoustimepatternplayedonasetofhandbells(Locke1982;Pantaleoni,1985).Membersofadrum

Rhythm

Page 5 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

ensemblekeeptheirrhythmicorientationbyhearingtheirpartsinrelationtothebell,ratherthanbyfocusingonanisochronousbeat.Furthermore,thefirstbeatofarhythmiccycleisnotheardasa“downbeat,”inotherwords,aspeciallystrongbeat(asinWesternmusic);ifanything,themostsalientbeatcomesattheendofthecycle(Temperley,2000).Finally,thismusicemphasizesdiversityintermsofthewayitcanbeheard.Asdifferentdrumsenter,eachwithitsowncharacteristicrepeatingtemporalpattern,apolyrhythmictextureiscreatedthatprovidesarichsourceofalternativeperceptualpossibilitiesdependingontherhythmiclayersandrelationshipsonechoosestoattendto(Locke1982;Pressing,2002).ThisisquitedifferentfromtherhythmicframeworkofmostWesternEuropeanmusic,inwhichtheemphasisisonrelativelysimpleandperceptuallyconsensualrhythmicstructures.OnepossiblereasonforthisdifferenceisthatWesternmusichasmajorpreoccupationsinothermusicaldimensions(suchasharmony),andarelativelysimplerhythmicframeworkfacilitatescomplexexplorationsintheseotherareas.AnotherreasonmaybethattempoinWesternEuropeanmusicisoftenflexible,withsalientdecelerationsandaccelerationsofthebeatusedforexpressivepurposes.Afairlysimplebeatstructuremayhelpalistenerstayorientedinthefaceofthesetemporalfluctuations(cf.Temperley,2004).

ThusitwouldbeanerrortoassumethattherhythmicstructureofWesternEuropeanmusicreflectsbasicconstraintsonhowthemindstructuresrhythmicpatternsintermsofproductionorperception.Aswitheverymusicaltradition,therhythmicpatternsofWesternEuropeanmusicreflectthehistoricalandmusicalconcernsofagivenculture.Ontheotherhand,acomparativeperspective(p.99) revealsthatcertainaspectsofrhythminWesternEuropeanmusic(suchasaregularbeatandgroupingofeventsintophrases)arealsofoundinnumerousothercultures,whichsuggeststhattheseaspectsreflectwidespreadcognitiveproclivitiesofthehumanmind.

ThediscussionbelowreliesattimesononeparticularmelodytoillustratevariousaspectsofrhythmicstructureinWesternmusic.Thisisthemelodyofachildren’ssong,indexedasmelodyK0016inadatabaseofBohemianfolkmelodies(Schaffrath,1995;Selfridge-Feld,1995).Figure3.1showsthemelodyinWesternmusicnotationandin“pianoroll”notationwitheachtone’spitchplottedasafunctionoftime(themelodycanbeheardinSoundExample3.1).

ThemelodywaschosenbecauseitishistoricallyrecentandfollowsfamiliarWesternconventions,yetisunlikelytobefamiliartomostreadersandisthusfreeofspecificmemoryassociations.Italsoillustratesbasicaspectsofrhythminasimpleform.Beyondthis,thereisnothingspecialaboutthismelody,andanynumberofothermelodieswouldhaveservedthesamepurpose.

3.2.1TheBeat:AStableMentalPeriodicity

Thephenomenonofamusicalbeatseemssimplebecauseitissofamiliar.Almosteveryonehastappedordancedalongtomusicwithabeat.Aregularbeatiswidespreadinmusicalcultures,anditisworthconsideringwhythismightbeso.Oneobviousfunctionofabeatistocoordinatesynchronizedmovement,suchasdance.(The

Rhythm

Page 6 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

relationshipbetweendanceandmusiciswidespreadinhumansocieties;indeed,someculturesdonotevenhaveseparatetermsformusicanddance.)Asecondobviousfunctionofabeatistoprovideacommontemporalreferenceforensembleperformance.Indeed,across-culturalperspectiverevealsthatensemblemusicwithoutaperiodictemporalframeworkisarareexception.Perlman(1997)pointstoonesuchexceptioninJavanesemusicknownaspathetan,notingthat“Exceptforcertainisolatedphrases,pathetanhasnounifyingmetricframework.…Rhythmicunisonisnotdesired,andthemusiciansneednotmatchtheirattackswiththeprecisionmadepossiblebyadefinitemeter”(p.105).AdetaileddiscussionofpathetanbyBrinner(1995:245–267)suggeststhatitisanexceptionthatprovestherule:withoutametricframe,playerssubstitutecloseattentiontoaleadmelodicinstrument(typicallyarebaborbowedlute)inordertocoordinateandorienttheirperformance.Thuswhenperiodicityinensemblemusiciswithdrawn,itsfunctionalroleisfilledinotherways.

Figure3.1 Asimplemelody(K0016)in(A)musicnotationand(B)pianorollformat.In(B),they-axisshowsthesemitonedistanceofeachpitchfromC4(261.63Hz).

(p.100) Fromalistener’sperspective,perceptionofabeatisoftenlinkedtomovementintheformofsynchronizationtothebeat.Formanypeople,thissynchronizationisanaturalpartofmusicalexperiencerequiringnospecialeffort.Itmaycomeasasurprise,then,thathumansaretheonlyspeciestospontaneouslysynchronizetothebeatofmusic.Althoughsynchronyisknownfromotherpartsoftheanimalkingdom,suchasthechorusingoffrogsorthesynchronizedcallsofinsects(Gerhardt&Huber2002,Ch.8;Strogatz,2003),humansynchronizationwithabeatissingularinanumberofrespects(seeChapter7,section7.5.3,forfurtherdiscussionofthispoint).Ofcourse,beatperceptiondoesnotautomaticallycausemovement(onecanalwayssitstill),butthehumanuniquenessofbeatsynchronizationsuggeststhatbeatperceptionmeritspsychologicalinvestigation.Researchinmusiccognitionhasrevealedseveralinterestingfactsaboutbeatperception.

First,thereisapreferredtemporangeforbeatperception.Peoplehavedifficultyfollowingabeatthatisfasterthanevery200msandslowerthanevery1.2seconds.Withinthisrange,thereisapreferenceforbeatsthatoccurroughlyevery500–700ms(Parncutt,1994;vanNoorden&Moelants,1999).Itisinterestingtonotethatthisisthesamerangeinwhichpeoplearethemostaccurateatmakingdurationjudgments,inother

Rhythm

Page 7 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

words,theyneitheroverestimatenorunderestimatethedurationoftemporalintervals(Eisler,1976;cf.Fraisse,1982).Furthermore,thisistherangeinwhichlistenersarethemostaccurateinjudgingslightdifferencesintempo(Drake&Botte,1993).Itisalsointerestingtonotethatinlanguageswithstressedandunstressedsyllables,theaveragedurationbetweenstressedsyllableshasbeenreportedtobeclosetoorwithinthisrange(Dauer,1983;Lea,1974,describedinLehiste,1977).

Second,althoughpeopleusuallygravitatetowardoneparticularbeattempo,theycantapatothertempithataresimpledivisorsormultiplesoftheirpreferredtappingrate(e.g.,atdoubleorhalftheirpreferredrate;Drake,Jones,&Baruch,2000).Forexample,considerSoundExample3.2,whichpresentsK0016along(p.101) withtwodifferentindicationsofthebeat.Bothareperfectlypossible,anditislikelythatmostpeoplecouldeasilytapateitherleveldependingonwhethertheyfocusonlowerorhigherlevelaspectsofrhythmicstructure.Drake,Jones,andBaruch(2000)haveshownthatpeoplevaryintheleveltheysynchronizewithinmusic,andthattheirpreferredlevelcorrelateswiththeirspontaneoustappingrate.Furthermore,althoughindividualsnaturallygravitatetooneparticularlevel,theycanmovetohigherorlowerlevelsiftheywish(e.g.,bydoublingorhalvingtheirtappingrate)andstillfeelsynchronizedwiththemusic.Thuswhenspeakingof“thebeat”ofapiece,itisimportanttokeepinmindthatwhatalistenerselectsasthebeatisjustonelevel(theirtactus)inahierarchyofbeats.

Third,beatperceptionisrobusttomoderatetempofluctuations.Inmanyformsofmusic,theoveralltimingofeventsslowsdownorspeedsupwithinphrasesorpassagesaspartofexpressiveperformance(Palmer,1997).Peoplearestillabletoperceiveabeatinsuchmusic(Large&Palmer,2002)andsynchronizetoit(Drake,Penel,&Bigand,2000),indicatingthatbeatperceptionisbasedonflexibletimekeepingmechanisms.

Fourth,thereisculturalvariabilityinbeatperception.DrakeandBenElHeni(2003)studiedhowFrenchversusTunisianlistenerstappedtothebeatofFrenchversusTunisianmusic.TheFrenchtappedataslowerratetoFrenchmusicthantoTunisianmusic,whereastheTunisiansshowedtheoppositepattern.DrakeandBenHeniarguethatthisreflectsthefactthatlistenerscanextractlarger-scalestructuralpropertiesinmusicwithwhichtheyarefamiliar.Thesefindingsindicatethatbeatperceptionisnotsimplyapassiveresponseoftheauditorysystemtophysicalperiodicityinsound:Italsoinvolvesculturalinfluencesthatmayrelatetoknowledgeofmusicalstructure(e.g.,sensitivitytohownotesaregroupedintomotives;cf.Toiviainen&Eerola,2003).

Fifth,andofsubstantialinterestfromacognitivesciencestandpoint,aperceivedbeatcantolerateagooddealofcounterevidenceintheformofaccentedeventsatnonbeatlocationsandabsentorweakeventsatbeatlocations,inotherwords,syncopation(Snyder&Krumhansl,2001).Forexample,considerSoundExamples3.3and3.4,twocomplextemporalpatternsstudiedbyPatel,Iversen,etal.(2005)withregardtobeatperceptionandsynchronization.Thepatternsbeginwithanisochronoussequenceof9tonesthatservestoindicatethebeat,whichhasaperiodis800ms.Afterthis“inductionsequence,”thepatternschangeintoamorecomplexrhythmbutwiththesamebeatperiod.Participantswereaskedtosynchronizetheirtapstotheisochronoustonesand

Rhythm

Page 8 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

thencontinuetappingatthesametempoduringthecomplexsequence.Theirsuccessatthistaskwastakenasameasureofhowwelltheywereabletoextractabeatfromthesesequences.Inthe“stronglymetrical”(SM)sequences(SoundExample3.3),atoneoccurredateverybeatposition.Inthe“weaklymetrical”(WM)sequences,however,about1/3ofthebeatpositionsweresilent(SoundExample3.4).(NB:TheSMandWMsequenceshadexactlythesame(p.102) setofinteronsetintervals,justarrangeddifferentlyintime;cf.Povel&Essens,1985.)ThussuccessfulbeatperceptionandsynchronizationinWMsequencesrequiredfrequenttapsatpointswithnosound.

AllparticipantswereabletosynchronizewiththebeatoftheSMsequence:Theirtapswereverycloseintimetotheidealizedbeatlocations.(Infact,tapstypicallyprecededthebeatbyasmallamount,afindingtypicalofbeatsynchronizationstudies,indicatingthatbeatperceptionisanticipatoryratherthanreactive.)OfgreaterinterestwasperformanceontheWMsequences.AlthoughsynchronizationwasnotasaccurateaswiththeSMsequencesasmeasuredbytappingvariability,mostparticipants(eventhemusicallyuntrainedones)wereabletotaptothebeatofthesesequences,thoughfromaphysicalstandpointtherewaslittleperiodicityatthebeatperiod.Thatis,mostpeopletappedtothesilentbeatsasiftheywerephysicallythere,illustratingthatbeatperceptioncantolerateagooddealofcounterevidence.

Theabovefactsindicatethatbeatperceptionisacomplexphenomenonthatlikelyhassophisticatedcognitiveandneuralunderpinnings.Specifically,itinvolvesamentalmodeloftimeinwhichperiodictemporalexpectanciesplayakeyrole(Jones,1976).Thismaybeonereasonwhyitisuniquetohumans.

Beatperceptionisanactiveareaofresearchinmusiccognition,inwhichtherehaslongbeenaninterestinthecueslistenersusetoextractabeat.TemperleyandBartlette(2002)listsixfactorsthatmostresearchersagreeareimportantinbeatfinding(i.e.,ininferringthebeatfromapieceofmusic).Thesecanbeexpressedaspreferences:

1.Forbeatstocoincidewithnoteonsets2.Forbeatstocoincidewithlongernotes3.Forregularityofbeats4.Forbeatstoalignwiththebeginningofmusicalphrases5.Forbeatstoalignwithpointsofharmonicchange6.Forbeatstoalignwiththeonsetsofrepeatingmelodicpatterns

Becausebeatperceptionisfundamentaltomusicandisamenabletoempiricalstudy,ithasattractedcomputational,behavioral,andneuralapproaches(e.g.,Desain,1992;Desain&Honing,1999;Toddetal.,1999;Large,2000;Toiviainen&Snyder,2003;Hannonetal.,2004;Snyder&Large,2005;Zantoetal.,2006)andhasthepotentialtomatureintoasophisticatedbranchofmusiccognitioninwhichdifferentmodelscompetetoexplainacommonsetofbehavioralandneuraldata.Itsstudyisalsoattractivebecauseittouchesonlargerissuesincognitiveneuroscience.Forexample,synchronizationtoabeatprovidesanopportunitytostudyhowdifferentbrainsystemsarecoordinatedinperceptionandbehavior(inthiscase,theauditoryandmotorsystems).Abetter

Rhythm

Page 9 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

understandingofthemechanismsinvolvedinbeatperceptionandsynchronizationcouldhaveapplicationsforphysicaltherapy,inwhichsynchronizationwithabeatisbeingusedtohelppatientswithneuromotordisorders(p.103) (suchasParkinson’sdisease)toinitiateandcoordinatemovement(Thautetal.,1999;cf.Sacks,1984,2007).

3.2.2Meter:MultiplePeriodicities

InWesternEuropeanmusic,beatsarenotallcreatedequal.Instead,somebeatsarestrongerthanothers,andthisservestocreateahigherlevelofperiodicityintermsofthegroupingand/oraccentuationofbeats.Forexample,thebeatsofawaltzaregroupedinthrees,withanaccentonthefirstbeatofeachgroup,whereasinamarchbeatsaregroupedintotwosorfours,withprimaryaccentonthefirstbeat(inafour-beatmarch,thereissecondaryaccentonthethirdbeat).

WaltzesandmarchesarebuttwotypesofmeterinabroaddiversityofmetersusedinWesternEuropeanmusic,buttheyservetoillustratesomegeneralfeaturesofmeterinthistradition.First,themetersofWesternmusicaredominatedbyorganizationintermsofmultiplesoftwoandthreeintermsofhowmanybeatsconstituteabasicunit(themeasure),andhowmanysubdivisionsofeachbeatthereare.Forexample,awaltzhasthreebeatspermeasure,eachofwhichcanbesubdividedintotwoshorterbeats,whereasamarchhastwo(orfour)beatspermeasure,eachofwhichcanalsobesubdividedintotwobeats.Manyotherpossibilitiesexist,forexampletwobeatspermeasure,eachofwhichissubdividedintothreebeats.1Thekeypointisthatmetertypicallyhasatleastonelevelofsubdivisionbelowthebeat(London,2002,2004:34),inadditiontoperiodicityabovethebeatcreatedbythetemporalpatterningofstrongbeats.Onewaytorepresentthisisviaametricalgrid,whichindicateslayersofperiodicityusingrowsofisochronousdots.Oneoftheserowsrepresentsthetactus,withtherowabovethisshowingtheperiodicpatternofaccentuationabovethetactus.Otherrowsaboveorbelowthetactusshowotherpsychologicallyaccessiblelevelsofperiodicity(Figure3.2showsametricalgridforK0016).

Thusoneshouldbeabletotaptoanyoftheselevelsandstillfeelsynchronizedwiththemusic.(Theuseofdotsinmetricalgridsindicatesthatmeterconcernstheperceptualorganizationofpointsintime,whichinphysicaltermswouldcorrespondtotheperceptualattacksoftones;Lerdahl&Jackendoff,1983.)Ingridnotation,therelativestrengthofeachbeatisindicatedbythenumberofdotsaboveit,inotherwords,thenumberoflayersofperiodicityitparticipatesin.Dotsatthehighestandlowestlevelmustfallwithinthe“temporalenvelope”formeter:Periodicitiesfasterthan200msandslowerthan∼4–6sareunlikelytobespontaneouslyperceivedaspartofametricframework.(Notethattheupperendofthisenvelopeissubstantiallylongerthanthe∼1.2secondlimitforfollowingabeatmentionedinsection3.2.1.Thatshorterlimitreferstobeat-to-beatintervals,whereas4–6sreferstothehighestmetricallevels,andislikelytoberelatedtooursenseofthepsychologicalpresent(cf.London,2002,2004:30).2

(p.104)

Rhythm

Page 10 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Figure3.2 MetricalstructureofK0016.Atypicaltactusisshownbythemetricallevellabeled1x.Phraseboundariesareindicatedbelowthepianorollnotation(p1=phrase1,etc.).

Beforemovingon,therelationshipbetweenaccentandmetershouldbediscussed.Thisisanimportantrelationship,becausestrongbeatsareperceptuallyaccentedpointsinthemusic.Thiskindofaccentdoesnotalwaysrelyon(p.105) physicalcuessuchasintensityorduration(note,forexample,thatalltonesinK0016areofequalintensity),andemergesfromthedetectionofperiodicityatmultipletime-scales.Thereareofcoursemanyphysical(or“phenomenal”)accentsinmusicduetoavarietyoffactors,includingduration,intensity,andchangesinmelodiccontour.Therearealso“structuralaccents”duetosalientstructuralpointsinthemusic,forexample,asuddenharmonicshiftorthestartofamusicalphrase(Lerdahl&Jackendoff,1983).Theinterplayofdifferentaccenttypesisoneofthesourcesofcomplexityinmusic(Jones,1993),particularlytheinterplayofmetricalaccentswithoff-beatphenomenalorstructuralaccents.Forexample,syncopationinmusicillustratesthesuccessfuluseofphenomenalaccents“againstthegrain”oftheprevailingmeter.Thisraisesakeypointaboutthemusicalmetricalgrid,namelythatitisamentalpatternofmultipleperiodicitiesinthemindofalistener,andnotsimplyamapoftheaccentualstructureofasequence.Thispointwillbecomerelevantinthediscussionofmetricalgridsinlanguage.

Theinfluenceofmusicalmeteronbehavior,perception,andbrainsignalshasbeendemonstratedinanumberofways.Sloboda(1983)hadpianistsperformthesamesequenceofnotessettodifferenttimesignatures(inmusic,thetimesignatureindicatesthegroupingandaccentuationpatternofbeats,i.e.,themeter).Thedurationalpatterningoftheperformancesdifferedsubstantiallydependingonthemeter,andinmanycasesagivenpianistdidnotevenrealizetheywereplayingthesamenotesequenceintwodifferentmeters.Ademonstrationofmeter’seffectonsynchronizationcomesfromPatel,Iversen,etal.(2005),whoshowedthattappingtoametricalpatterndiffersfromtappingtoasimplemetronomeatthesamebeatperiod.Specifically,tapstothefirstbeatofeachmetriccycle(i.e.,the“downbeats”inthestronglymetricalsequencesofSoundExample

Rhythm

Page 11 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

3.3)wereclosertothephysicalbeatthantapsonotherbeats.Importantly,thesedownbeats(whichoccurredeveryfourbeats)wereidenticaltoothertonesintermsintensityandduration,sothattheinfluenceofdownbeatsontappingwasnotduetoanyphysicalaccentbuttotheirroleincreatingafour-beatperiodicstructureinthemindsoflisteners.

Intermsofmeter’sinfluenceonperception,PalmerandKrumhansl(1990)hadparticipantslistentoasequenceofisochronoustonesandimaginethateacheventformedthefirstbeatofgroupsoftwo,three,fourorsixbeats.Afterafewrepetitions,aprobetonewassoundedandparticipantshadtoindicatehowwellitfitwiththeimaginedmeter.Theratingsreflectedahierarchyofbeatstrength(cf.Jongsmaetal.,2004).Turningtoneuralstudies,Iversen,Repp,andPatel(2009)hadmusicallytrainedparticipantslistentoametricallyambiguousrepeatingtwo-notepatternandmentallyimposeadownbeatinaparticularplace.Specifically,inhalfofthesequencestheyimaginedthatthefirsttonewasthedownbeat,andintheotherhalftheyimaginedthatthesecondtonewasthedownbeat.Participantswereinstructednottomoveortoengageinmotorimagery.Measurementofbrainsignalsfromauditoryregionsusing(p.106)magnetoencephalography(MEG)revealedthatwhenanotewasinterpretedasthedownbeat,itevokedanincreasedamountofneuralactivityinaparticularfrequencyband(beta,20–30Hz)comparedtowhenitwasnotadownbeat(eventhoughthetoneswerephysicallyidenticalinthetwoconditions).3Acontrolexperimentshowedthatthepatternofincreasedactivitycloselyresembledthepatternobservedwhenthenoteinquestionwasinfactphysicallyaccented(Figure3.3).Theseresultssuggestthattheperceptionofmeterinvolvestheactiveshapingofincomingsignalsbyamentalperiodictemporal-accentualscheme.

3.2.3Grouping:ThePerceptualSegmentationofEvents

Groupingreferstotheperceptionofboundaries,withelementsbetweenboundariesclusteringtogethertoformatemporalunit.ThiscanbeillustratedwithK0016.Inlisteningtothismelody,thereisaclearsensethatitisdividedintophrases,schematicallymarkedinFigure3.4.

Theperceptualboundariesofthefirsttwophrasesaremarkedbysilences(musicalrests).Ofgreaterinterestaretheboundariesattheendofthethirdandthefourthphrases,whicharenotmarkedbyanyphysicaldiscontinuityinthetonesequence,butareneverthelesssalientperceptualbreakpoints.

AsemphasizedbyLerdahlandJackendoff(1983),groupingisdistinctfrommeter,andtheinteractionofthesetworhythmicdimensionsplaysanimportantroleinshapingtherhythmicfeelofmusic.Forexample,anacrusis,orupbeat,isarhythmicallysalientphenomenoninvolvingaslightmisalignmentbetweengroupingandmeter,inotherwords,aphrasestartingonaweakbeat(suchasphrase2ofK0016).

Psychologicalevidenceforperceptualgroupinginmusiccomesfromanumberofsources.Memoryexperimentsshowthatifalistenerisaskedtoindicatewhetherabrief

Rhythm

Page 12 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

tonesequencewasembeddedinapreviouslyheardlongertonesequence,performanceisbetterwhentheexcerptendsatagroupboundaryintheoriginalsequencethanwhenitstraddlesagroupboundary(Dowling,1973;Peretz,1989).Thissuggeststhatgroupinginfluencesthementalchunkingofsoundsinmemory.Furtherevidenceforgroupingcomesfromstudiesthatshowhowgroupingwarpstheperceptionoftime.Forexample,clicksplacednearphraseboundariesinmusicalsequencesperceptuallymigratetothoseboundariesandareheardascoincidingwiththem(Sloboda&Gregory,1980;Stoffer,1985).MoreevidenceforperceptualwarpingbasedongroupingcomesfromastudybyRepp(1992a),inwhichparticipantsrepeatedlylistenedtoacomputer-generatedisochronousversionoftheopeningofaBeethovenminuet.Thetaskwastodetectalengtheningin1of47possiblepositionsinthemusic.Detectionperformancewasparticularlypooratphraseboundaries,probablyreflectinganexpectationforlengtheningatthesepoints(Reppfurthershowedthatthesewerethepointsatwhichhumanperformerstypicallysloweddowntomarkthephrasestructure).Finally,inagatingstudyofrecognitionforfamiliarmelodiesinwhichsuccessivelylongerfragmentsoftuneswerehearduntiltheywerecorrectlyidentified,Schulkindetal.(2003)foundthatidentificationperformancewashighestatphraseboundaries.4Thusthereisabundantevidencethatgroupingplaysaroleinmusicalperception.

(p.107)

Rhythm

Page 13 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Figure3.3 (A)Repeatingtwo-noterhythmicpattern,inwhichthelistenerimaginesthedownbeatoneitherthefirsttone(left)orsecondtone(right).(B)Evokedneuralresponses(measuredoverauditorybrainregions)tothetwo-tonepatternsubjectivelyinterpretedintwodifferentways,inotherwords,withthedownbeatontone1versustone2.(Theonsettimesoftones1and2areindicatedbythin,vertical,graylinesat0and0.2s).Thesolidanddashedblacklinesshowacross-subjectmeansforthetwoimaginedbeatconditions(solid=beatimaginedontone1,dashed=beatimaginedontone2).Dataarefromthebetafrequencyrange(20–30Hz).Thedifferenceisshownbythedottedline,withshadingindicating1standarderror.(C)Evokedneuralresponsesinthebetafrequencyrangetoatwo-tonepatternphysicallyaccentedintwodifferentways,withtheaccentontone1(solidline)versustone2(dashedline).

(p.108)

Figure3.4 K0016segmentedintomelodicphrases(p1=phrase1,etc.).

Rhythm

Page 14 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Whatcuesdolistenersuseininferringgroupingstructureinmusic?ReturningagaintoK0016,theendsofphrases3and4aremarkedbylocaldurationallengtheningandloweringofpitch.Itisnotablethatthesecueshavebeenfoundtobeimportantintheprosodicmarkingofclauseendingsinspeech(Cooper&Sorensen,1977).Eveninfantsshowsensitivitytotheseboundarycuesinbothspeechandmusic(Hirsh-Paseketal.,1987;Krumhansl&Jusczyk,1990;Jusczyk&Krumhansl,1993).Forexample,infantsprefertolistentomusicalsequencesinwhichpausesareinsertedafterlongerandlowersoundsratherthanatotherlocations,presumablybecauseintheformercasethepausescoincidewithperceptualboundaries.Ofcourse,thereismuchmoretogroupingthanjustthesetwocues.Deliège(1987)foundthatsalientchangesinintensity,duration,pitch,andtimbrecanallplayaroleindemarcatingtheedgesofgroups.Anotherfactorthatislikelytobeimportantismotivicrepetition,forexample,arepeatingpatternof(p.109)thesameoveralldurationandinternaldurationalpatterning.Whenthesedifferentcuesareinconflict,peoplecandisagreeaboutwheretheyheargroupingboundaries(Peretz,1989).Theinteractionofdifferentfactorsingroupingperceptionisatopicthatdrawscontinuinginterest,becausedataontheperceivedsegmentationofpiecesisrelativelyeasytocollect(Clarke&Krumhansl,1990;Deliègeetal.,1996;Frankland&Cohen,2004;Schaeferetal.,2004).

Groupingplaysaprominentroleinmoderncognitivetheoriesofmusic,inwhichitisconceivedofashierarchical,withlowerlevelgroupsnestedwithinhigherlevelones.Forexample,atheoreticalanalysisofgroupinginK0016wouldaddlayersaboveandbelowthephraselayer.Belowthephraselayer,eachphrasewouldbeparsedintosmallergroups(motives);abovethephraselayer,phraseswouldbelinkedintohigherlevelstructures.Forexample,onemightunitePhrases1and2intoagroup,followedbyagroupconsistingofPhrases3and4,andafinalgroupcoincidentwithPhrase5.OneofthemostdevelopedtheoreticaltreatmentsofhierarchicalgroupinginmusicisthatofLerdahlandJackendoff(1983),whoproposecertainbasicconstraintsongroupingstructuresuchastheconstraintthatapiecemustbefullyparsedintogroupsateachhierarchicallevel,andthatboundariesathigherlevelsmustcoincidewiththoseatlowerlevels.EvidenceformultiplelayersofgroupingstructureinmusiccomesfromresearchbyTodd(1985),whoshowedthattheamountoflengtheningatagivenphraseboundaryinmusicispredictedbythepositionofthatboundaryinahierarchicalphrasestructureofapiece.

Thehierarchicalviewofgroupingstructureinmusicshowsstrongparallelstotheoriesofprosodicstructureinmodernlinguistictheory,notablytheconceptofthe“prosodichierarchy”(Selkirk,1981,Nespor&Vogel,1983).Theprosodichierarchyreferstotheorganizationofsonicgroupingsatmultiplelevelsinspeech,rangingfromthesyllableuptotheutterance.Akeyconceptualpointmadebyallsuchtheoriesisthatthesegroupingsarenotsimplereflectionsofsyntacticorganization.Totakeawell-knownexample,considerthedifferencebetweenthesyntacticbracketingofasentencein3.1aversusitsprosodicphrasalbracketing3.1b(Chomsky&Halle,1968):

(3.1a)Thisis[thecat[thatcaught[therat[thatstole[thecheese]]]]]

Rhythm

Page 15 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(3.1b)[Thisisthecat][thatcaughttherat][thatstolethecheese]

Prosodicgroupingreflectsaseparatephonologicalleveloforganizationthatisnotdirectlydeterminedbysyntacticstructure.Instead,otherlinguisticfactorsplayanimportantrole,suchasthesemanticrelationsbetweenwordsandthedesiretoplacefocusoncertainelements(Marcus&Hindle,1990;Ferreira,1991).Furthermore,therearethoughttobepurelyrhythmicfactorssuchasatendencytoavoidgroupsthatareveryshortorverylong,andatendencytobalancethelengthsofgroups(Gee&Grosjean,1983;ZellnerKeller,2002).Theprosodicgroupingstructureofasentenceisbynomeansset(p.110) instone:Therearedifferencesamongindividualsintermsofhowtheygroupthewordsofthesamesentence,andthegroupingstructureofasentencecanvarywithspeechrate(Fougeron&Jun,1998).Nevertheless,groupingisnottotallyidiosyncratic,andpsycholinguistshavemadegoodprogressinpredictingwherespeakersplaceprosodicboundariesinasentencebasedonsyntacticanalysesofsentences(Watson&Gibson,2004).

AlthoughExample3.1aboveonlyshowsonelevelofprosodicphrasing,moderntheoriesoftheprosodichierarchypositmultiplelevelsnestedinsideoneanother.Theoriesvaryinthenumberoflevelstheypropose(Shattuck-Hufnagel&Turk,1996),5soforillustrativepurposesonlyonesuchtheoryisdiscussedhere.Hayes(1989)positsafive-levelhierarchycomprisedofwords,cliticgroups,phonologicalphrases,intonationalphrases,andutterances.Figure3.5showsaprosodichierarchyforasentenceaccordingtothistheory,withthesyntacticstructurealsoshownforcomparison.(Notethatacliticgroupcombinesalexicalwordthathasastressedsyllablewithanadjacentfunctionword—anunstressedsyllable—intoasingleprosodicunit.SeeHayes,1989,fordefnitionsofotherunits.)

Oneformofevidenceofferedfortheexistenceofagivenlevelintheprosodichierarchyisasystematicvariationintherealizationofaphonemicsegmentthatdependsonprosodicstructureatthatlevel.Forexample,Hayes(1989)discusses/v/deletioninEnglishspeechasanexampleofarulethatoperateswithinthecliticgroup.Thusitisacceptabletodeletethe/v/inAmericanEnglishwhensaying,“Willyou[saveme]aseat?”because“saveme”isacliticgroup.(Thatis,ifyoulistencarefullytoanAmericanEnglishspeakersaythisphraserapidly,“save”isoftenacousticallyrealizedas“say,”thoughitisintendedas—andheardas—“save”.)Incontrast,the/v/isnotdeletedwhensaying“[save][mom]”because[save]and[mom]aretwoseparatecliticgroups.Otherevidencethathasbeenadducedforprosodicconstituentsincludespreferencesforinterruptionpointsbetween,ratherthanwithin,constituents(Pilon,1981),andspeededwordspottingattheboundariesofconstituents(Kim,2003).

Evidencethattheprosodichierarchyhasmultiplelevelscomesfromphoneticmodificationsofspeechthatvaryinaparametricfashionwiththeheightoftheprosodicboundaryatthephoneme’slocation(thiscorrespondstothenumberofcoincidentprosodicboundariesatthatpoint,ashigherlevelboundariesarealwayscoincidentwithlowerlevelones).Forexample,ChoandKeating(2001)showedthatinKorean,thevoice-onsettimeofstopconsonantsislargerathigherprosodicboundaries,andDilley,

Rhythm

Page 16 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Shattuck-Hufnagel,andOstendorf(1996)haveshownthattheamountofglottalizationofword-onsetvowelsisgreaterathigherlevelboundaries.

(p.111)

Figure3.5 (A)Syntacticand(B)prosodichierarchyforasentenceofEnglish.Abbreviationsfor(A):S=sentence,PP=prepositionalphrase,NP=nounphrase,VP=verbphrase,Det=Determiner,A=adjective,N=Noun,V=Verb.Abbreviationsfor(B):U=utterance,I=Intonationphrase,P=phonologicalphrase,C=cliticgroup,W=word.AdaptedfromHayes,1989.

Anotherphenomenonthatsupportsthenotionofhierarchicalgroupinginspeechisvariationinperceivedjuncturebetweenwords(Jun,2003).Inconnectedspeech,wordsareacousticallyruntogetherandthesilentintervalsthatdooccur(e.g.,duetostopconsonants)arenotnecessarilylocatedatwordboundaries(cf.Chapter2,section2.3.3,subsection“ABriefIntroductiontotheSpectrogram”).Nevertheless,wordsareperceivedasseparatedfromoneanother.Unlikewithwrittenlanguage,however,theperceiveddegreeofspacingisnotidenticalbetweeneachpairofwords:Rather,somewordboundariesseemstrongerthanothers.Forexample,thesentencein3.1cbelowcontainsjuncturemarkingsfromasystemdevisedbyPriceetal.(1991).Inthissystem,aresearcherlistensrepeatedlytoagivensentenceandplacesnumeralsfrom0to6betweeneachpairofwordstoindicatethedegreeofperceivedseparationbetweenthem.A“breakindex”of0indicatestheweakestperceivedjuncture,inotherwords,betweenthewordsofacliticgroup.Attheoppositeextreme,abreakindexof6indicatestheendofasentence.

(3.1c)Only1one4remembered3the0lady1in1red6.

Wightmanetal.(1992)studiedtherelationshipbetweenthesebreakindicesandspeechdurationpatternsinalargespeechcorpus,andfoundacorrelationbetweenperceivedboundarystrengthandamountoflengtheningofthesyllableprecedingtheboundary(cf.Gussenhoven&Rietveld,1992).6Thisfindingis(p.112) strikinglyreminiscentoftheresearchonmusicbyTodd(1985)describedabove.Anotherparalleltomusicisthatdurationallengtheninginteractswithpitchandamplitudecuesindeterminingtheperceivedstrengthofprosodicboundaries(Streeter,1978;dePijper&Sanderman,1994).

Inconclusion,groupingisafundamentalrhythmicphenomenonthatappliestoboth

Rhythm

Page 17 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

musicalandlinguisticsequences.Inbothdomains,themindparsescomplexacousticpatternsintomultiplelevelsofphrasalstructure,andmusicandlanguageshareanumberofacousticcuesformarkingphraseboundaries.Thesesimilaritiespointtosharedcognitiveprocessforgroupingacrossthetwodomains,andindicatethatgroupingmayproveafruitfulareaforcomparativeresearch.Asdiscussedinsection3.5,empiricalworkisprovingtheseintuitionscorrect.

3.2.4DurationalPatterninginMusic

Uptothispoint,thediscussionofmusicalrhythmhasbeenconcernedwithpointsandedgesintime:beatsandgroupingboundaries.Adifferentsetofissuesinrhythmresearchconcernshowtimegetsfilled,inotherwords,thedurationalpatterningofevents.

DurationCategoriesinMusicInmusic,thedurationalpatterningofeventsistypicallymeasuredbythetimeintervalsbetweeneventonsetswithinaparticulareventstream:Thisdefinesasequenceofinteronsetintervals(IOIs).Forexample,thesequenceofIOIsbetweenthetonesofamelodydefinesthedurationalpatterningofthatmelody.Typically,durationstendtobeclusteredaroundcertainvaluesreflectingtheorganizationoftimeinmusicintodiscretecategories.Fraisse(1982)pointedoutthattwocategoriesthatfigureprominentlyinWesternmusicalsequencesareshorttimesof200–300msandlongtimesof450–900ms(cf.Ross,1989).Hearguedthatthesetwodurationcategorieswerenotonlyquantitativelydifferentbutalsodifferentintermsoftheirperceptualproperties:Longintervalsareperceivedasindividualunitswithdistinctdurations,whereasshortintervalsareperceivedcollectivelyintermsoftheirgroupingpatternsratherthanintermsofindividualdurations.

Thereisempiricalevidencethatdurationsinmusicalrhythmsareperceivedintermsofcategories(Clarke,1987;Schulze,1989).Forexample,Clarke(1987)hadmusicstudentsperformacategoricalperceptionexperimentonrhythm.Participantsheardshortsequencesoftonesinwhichthelasttwotoneshadaratiothatvariedbetween1:1and1:2.Listenershadtoidentifythefinaldurationratioasoneortheotherofthese,andalsohadtocompleteataskthatrequiredthemtodiscriminatebetweendifferentratios.Theresultsshowedasteeptransitionintheidentificationfunction,andincreaseddiscriminationwhenstimuliwereneartheboundaryversuswithinagivenregion.(Clarkealso(p.113) foundthatthelocationoftheboundarydependedonthemetricalcontextinwhichthesequencewasperceived,providinganotherexampleoftheinfluenceofmeteronperception;cf.section3.2.2.)

Inspeech,thedurationofbasiclinguisticelements(suchasphonemesandsyllables)isinfluencedbyanumberoffactors.Forexample,therearearticulatoryconstraintsonhowfastdifferentsoundscanbeproduced,whichcreatesdifferentminimumdurationsfordifferentsounds(Klatt,1979).Therearealsosystematicphonologicalfactorsthatmakesomesoundslongerthanothers.Forexample,inEnglish,thesamevoweltendstobelongerifitoccursbeforeafinalstopconsonantthatisvoicedratherthanunvoiced

Rhythm

Page 18 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(e.g.,the/i/in“bead”vs.“beet”),andthisdifferenceinfluencestheperceptionofthefinalstopasvoicedorvoiceless(Klatt,1976).Asimplephonologicalfactorthatinfluencessyllabledurationisthenumberofphonemesinthesyllable:Syllableswithmorephonemestendtobelongerthanthosewithfewerphonemes(e.g.,“splash”vs.“sash”;Williams&Hiller,1994).Atopthesesourcesofvariationareothersourcesincludingvariationsinspeakingstyle(casualvs.clear),andvariationsinspeechraterelatedtodiscoursefactors,suchasspeedingupneartheendofasentenceto“holdthefloor”inaconversation(Schegloff,1982;Smiljanic&Bradlow,2005).Givenallthesefactors,itisnotsurprisingthatthedurationsofspeechelementsdonottendtoclusterarounddiscretevalues.Instead,measurementsofsyllableorphonemedurationtypicallyrevealacontinuousdistributionwithonemainpeak.Forexample,Figure3.6ashowsahistogramofsyllabledurationsforasampleofspokenEnglish.

Figure3.6a HistogramofsyllabledurationsinacorpusofspontaneousspeechinAmericanEnglish.Dataarefromapproximately16,000syllables.Meansyllableduration=191ms,sd=125ms.Syllableswithduration〉750msarenotshown(〈1%oftotal).Histogrambinsize=10ms.AnalysisbasedondatafromGreenberg,1996.

(p.114) Havingsaidthis,itisimportanttonotethatdurationalcategoriesdooccurinsomelanguages.Forexample,therearelanguageswithphonemiclengthcontrastsinwhichthesamewordcanmeanentirelydifferentthingswhenashortversuslongversionofthesamevowelorconsonantisused.Insomelanguages,suchasEstonian,therecanevenbethree-waylengthcontrasts.Forexample,“sata”canmeanthreeentirelydifferentthings(“hundred,”“send,”and“get”)dependingonthelengthofthefirst/a/.Itwouldbeinterestingtostudylengthcontrastsinagivenvowelphonemeandexaminetheamountoftemporalvariabilitywithineachdurationcategoryinconnectedspeech.Thiscouldbecomparedtotemporalvariabilityofagivendurationcategoryinmusic,toseewhethertheperceptualsystemhasasimilartoleranceforwithin-categoryvariabilityinthetwodomains.7

Rhythm

Page 19 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

ExpressiveTiminginMusicIftheperceptualsystemcaredonlyaboutmusicaldurationsasasequenceofdiscretecategories,thencomputerrenditionsofmusicalpiecesbasedonexactrenderingsofmusicnotationwouldbeperfectlyacceptabletolisteners.Althoughsuchmechanicalperformancesdooccurinsomesettings(e.g.,rhythmtracksinsomemodernpopularmusic),inothercontexts,suchastheclassicalpianorepertoire,suchperformancesarerejectedasunmusical.Notsurprisinglythen,physicalmeasurementsofhumanperformancesrevealconsiderabledeviationsfromnotateddurations.Forexample,Figure3.6bshowsahistogramofIOIs,allofwhichrepresentrealizationsofnoteswiththesamenotatedduration(aneighthnoteorquaver)fromafamouspianist’srenditionofSchumann’sTräumerei(Repp,1992b).8Hadthepiecebeenperformedbyamachine,alloftheseIOIswouldbeasinglevalue.Instead,considerablevariationisseen.Thecrucialfactaboutthisvariationisthatitisnot“noise”:Itlargelyrepresentsstructuredvariationrelatedtotheperformer’sinterpretationofthepiece(Palmer,1997;Ashley,2002).Forexample,Repp(1992b)studiedseveralfamouspianists’renderingsofTräumereiandfoundthatallshowedslowingoftempoatstructuralboundaries,withtheamountofslowingproportionaltotheimportanceoftheboundary(cf.Todd,1985).Atafinertimescale,Reppfoundthatwithinindividualmelodicphrasestherewasatendencytoaccelerateatthebeginningandslowneartheend,withthepatternofIOIsfollowingasmoothparabolicfunction.Reppspeculatedthatthispatternmayreflectprinciplesofhumanlocomotion,inotherwords,amusicalallusiontophysicalmovement(cf.Kronman&Sundberg,1987).

(p.115)

Figure3.6b HistogramofdurationsofeighthnotesfromaperformanceofSchumann’sTräumereibyClaudioArrau.Thelargevaluesintherighttailofthehistogramareduetophrase-finalritards.Dataarefromapproximately170eighthnotes.Meannoteduration=652ms,sd=227ms.Noteswithduration〉1,600msarenotshown(〈1%oftotal).Histogrambinsize=50ms.

Rhythm

Page 20 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

TheaboveparagraphfocusesontheroleofIOIsinexpressivetiming.IOIsarethebasisof“expressivetimingprofiles,”timeseriesthatshowtheactualpatternofeventtimingversustheidealizedpatternbasedonnotatedduration.Althoughstudiesoftheseprofileshavedominatedresearchonexpressivetiming,itisimportantnottooverlookanotheraspectofexpressivetiming,namely,articulation.AlthoughIOIreferstothetimeintervalbetweentheonsetsofsuccessivetones,articulationreferstothetimebetweentheoffsetofonetoneandtheonsetofthenext.Ifthereislittletimebetweentheseevents(orifthetonesoverlapsothattheoffsetofthepriortoneoccursaftertheonsetofthefollowingtone,whichispossibleinpianomusic),thisisconsidered“legato”articulation.Inthistypeofarticulation,onetoneisheardasflowingsmoothlyintothenext.Incontrast,staccatoarticulationinvolvesasalientgapbetweenoffsetandonset,givingthetonesarhythmicallypunctuatedfeel.InadditiontoIOIandarticulationpatterns,anotherimportantcuetomusicalexpressionisthepatterningoftoneintensity.

Duetothefactthattiming,articulationandintensityinmusiccanbemeasuredwithgreatprecisionusingmoderntechnology(e.g.,usingpianoswithdigitalinterfaces,suchastheYamahaDisklavier),expressionhasbeenafruitfulareaofresearchinstudiesofmusicproduction.Therehasalsobeensomeresearch(p.116) onexpressivefeaturesinperception.Forexample,listenerscanreliablyidentifyperformancesofthesamemusicasexpressive,deadpan(mechanical),orexaggerated(Kendall&Carterette,1990),andcanidentifytheperformer’sintendedemotiononthebasisofexpressivefeatures(Gabrielsson&Juslin,1996).

Palmer(1996)hasshownthatmusicallytrainedlistenerscanidentifyaperformer’sintendedmetricalandphrasestructureonthebasisofexpressivecues.OnecleverdemonstrationoftheperceptualimportanceofexpressivetimingwasprovidedbyClarke(1993),whousednaturallyperformedshortmelodies.Foreachmelody,Clarkeextracteditsexpressivetimingprofile,manipulatedit,andthenreimposeditonamechanicalperformanceofthemelody,thuscreatingaFrankensteinianmelodywithstructureandexpressionmismatched.Forexample,inoneconditiontheoriginalnote-by-noteexpressivetimingprofilewasshiftedseveralnotestotheright.Musiciansjudgedtheoriginalsversusthemismatchedmelodiesintermsofthequalityofperformance,andfavoredtheoriginals.Thuslistenersaresensitivetothewayexpressivetimingalignswiththestructureofmusicalpassages.

Expressivetiminginmusichasaninterestingrelationshiptoprosodicstructureinspeech.Justasamusicalpassageplayedbydifferentperformerswillhavedifferentexpressivetimingpatterns,thesamesentencespokenbydifferentspeakerswillhaveadifferenttemporalpatterningofsyllablesandphonemes.Inthepast,researchershavesuggestedthattheseindividualisticaspectsofperformanceare“normalizedaway”inmemoryformusicalandspokensequences,arguingthattheabstractmemoryrepresentationfavorsalessdetailed,morecategoricalstructure(Largeetal.,1995;Pisoni,1997).Morerecentresearch,however,suggeststhatlistenersretainsometemporalinformationinmemoryforspeechandmusic(Bradlowetal.,1999;Palmeretal.,2001).Forexample,Palmeretal.(2001)familiarizedlistenerswithparticularperformancesofshortmelodicsequences,

Rhythm

Page 21 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

andthenlatertestedtheabilitytorecognizetheseperformancesagainstotherperformancesofthesamesequences.Thedifferentperformancesweregeneratedbyapianistwhoproducedthesameshortmelodicsequencesaspartoflongermelodiesthatdifferedintheirmetricalstructure(3/4vs.4/4time).Asaresultofthedifferingmetricalstructure,thesamemelodicsequencewasproducedwithdifferentpatternsofarticulationandintensity.Foreachsuchmelodicsequence,bothmusiciansandnonmusicianswereabletorecognizetheoriginalversiontheyhadheardwhenpresentedwithitversusanotherversion.Furthermore,even10-month-oldinfantsdiscriminatedbetweenfamiliarandunfamiliarperformances,orientinglongertowardtheformer.Palmeretal.relatethesefindingstoresearchinspeechperceptionshowingthatlistenersretainstimulus-specificacousticpropertiesofwordsalongwithabstractlinguisticproperties(Luce&Lyons,1998).

Anotherlineofresearchrelatingtiminginmusictospeechprosodyconcerns“tempopersistence.”Jungersetal.(2002)hadpianistsalternatebetweenlisteningtoshortmelodiesandsight-readingdifferentshortmelodies.The(p.117) participantsweretoldtoattendtoboththeheardandperformedmelodiesforalatermemorytest.Infact,therealquestionofinterestwastherelationshipbetweenthetempooftheheardandperformedmelodies.Theheardmelodiesoccurredinblocksofslowandfasttempi,andJungersetal.foundthatthetempoofperformedmelodieswasinfluencedbythetempoofheardmelodies:Thepianistsplayedmoreslowlyafterslowmelodiesandfasterafterfastmelodies.Asimilarexperimentusingspokensentencesratherthanmelodiesshowedasimilartempopersistenceeffectinspeech.Thesefindingsarereminiscentofresearchon“accommodation”insociolinguistics,whichhasshownthatwhenpeopleofdifferentsocialbackgroundsmeet,theirspeechbecomesmorealike(cf.Gilesetal.,1991).

Inaninterestingfollow-upstudy,DallaBellaetal.(2003)studiedtempopersistenceacrossmodalities.Listeners(bothmusiciansandnonmusicians)alternatedbetweenhearingmelodiesandreadingsentencesaloud.Themusiciansshowedatempopersistenceeffect:Theyspokefasterafterhearingfastermelodies.However,thenonmusiciansshowednosucheffect.Furthermore,whenthemusiciansdidthereverseexperiment(inwhichtheyalternatedbetweenhearingsentencesandsight-readingmelodies),therewasnoevidenceoftempopersistence.DallaBellaetal.suggestthatmusiciansmaybebetterthannonmusiciansatbeatextractioninmusic,andthatthismaydrivetheeffectoftempopersistenceseenintheirfirststudy.Followingthislogic,Iwouldsuggestthatthelackofaneffectintheirsecondstudyindicatesthatspeechperceptiondoesnotinvolveextractionofabeat.

3.2.5ThePsychologicalDimensionsofMusicalRhythm

Theinteractionsofbeat,meter,accent,grouping,andexpressivetimingmakemusicalrhythmapsychologicallyrichphenomenon(andthisisjustwithintheconfinesofWesternEuropeanmusic!).SomeideaofthisrichnessissuggestedbytheworkofGabrielsson,whohasconductedstudiesinwhichavarietyofrhythmsarecomparedandclassifiedbylistenersusingsimilarityjudgmentsandadjectiveratings(reviewedinGabrielsson,1993).Statisticaltechniquessuchasmultidimensionalscalingandfactoranalysisareused

Rhythm

Page 22 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

touncovertheperceptualdimensionsinvolvedintheexperienceofmusicalrhythms.Thisresearchhasrevealedanastonishinglylargenumberofdimensions(15),whichgroupbroadlyintothoseconcernedwithstructure(e.g.,meter,simplicityvs.complexity),motion(e.g.,swinging,graceful),andemotion(e.g.,solemnityvs.playfulness).Althoughmuchofthecognitivescienceofmusicalrhythmfocusesonstructuralissues,itisimportanttokeepthelinkstomotionandemotioninmind,fortheseconnectionsarepartofwhatdistinguishesmusicalrhythmfromspeechrhythm,apointtowhichIwillreturnattheendofthediscussionofrhythminspeech.

(p.118) 3.3RhythminSpeechAlthoughthestudyofrhythminpoetryhasalonghistory,datingbacktoancientGreekandIndiantexts,thestudyofrhythminordinarylanguageisarelativelyrecentendeavorinlinguistics.Researchershavetakenatleastthreeapproachestothistopic.Thefirstapproachistypological,andseekstounderstandtherhythmicsimilaritiesanddifferencesamonghumanlanguages.Thedrivingforcebehindthisworkhasbeenideathatlinguisticrhythmsfallintodistinctcategories.Forexample,inonewidespreadtypologicalscheme(discussedinthenextsection),English,Arabic,andThaiareallmembersofasinglerhythmicclass(“stress-timedlanguages”),whereasFrench,Hindi,andYorubaaremembersofadifferentclass(“syllable-timedlanguages”).Asisevidentfromthisexample,membershipinarhythmicclassisnotdeterminedbythehistoricalrelationshipoflanguages;rhythmcanunitelanguagesthatareotherwisequitedistantbothhistoricallyandgeographically.

Thesecondapproachtospeechrhythmistheoretical,andseekstouncovertheprinciplesthatgoverntherhythmicshapeofwordsandutterancesinagivenlanguageorlanguages.Thisresearch,whichincludesanareacalled“metricalphonology,”seekstobringthestudyofthelinguisticrhythminlinewiththerestofmodernlinguisticsbyusingformalizedrulesandrepresentationstoderivetheobservedrhythmicpatterningofutterances.

Thethirdapproachisperceptual,andexaminestherolethatrhythmplaysintheperceptionofordinaryspeech.Oneprominentlineofresearchinthisareaconcernstheperceptualsegmentationofwordsfromconnectedspeech.Another,smallerlineofresearchexaminestheeffectsofrhythmicpredictabilityinspeechperception.

Thegoalofthispartofthechapteristointroduceeachoftheseareasandmakecomparisonstomusicalrhythmwhenappropriate.Beforecommencing,itisworthintroducingaconceptthatoccursineachsection:thenotionofprominenceinspeech.Inmanylanguages,itisnormaltoproducethesyllablesofanutterancewithdifferingdegreesofprominence.Thisistrueevenwhenasentenceissaidwithnospecialemphasisonanyparticularword.Forexample,whenspeakingthefollowingsentence,notehowthesyllablesmarkedbyanxaremoreprominentthantheirneighbors:

Rhythm

Page 23 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Themostimportantphysicalcorrelatesofprominenceareduration,pitchmovement,vowelquality,andloudness.9Prominenceinspeechraisesmany(p.119) interestingquestions.Howmanydifferentdegreesofprominencecanlistenersreliablydistinguish(Shattuck-Hufnagel&Turk,1996)?Dolanguagesdifferintheextenttowhichtheyrelyonparticularacousticcuestoprominenceinproductionandperception(Berinstein,1979;Lehiste&Fox,1992)?Empiricaldataontheseissuesisstillrelativelysparse,andwewillnotdelveintothemhere.Instead,mostsectionsbelowtreatprominenceasabinaryquantityreferredtoas“stress,”followingthetraditionofmuchworkonspeechrhythm.Anexceptionoccursinsection3.3.2,wheredegreesofprominencearediscussedinthecontextofmodernlinguistictheoriesofspeechrhythm.

Beforeembarkingonthefollowingsections,awordshouldbesaidabouttheconceptofstressinlinguistics.Stressisrecognizedasoneaspectofwordprosodyinhumanlanguages;toneandlexicalpitchaccentaretwootheraspects.Justasnotalllanguageshavelexicaltone(cf.Chapter2foradiscussionoftonelanguages)orlexicalpitchaccent,10notalllanguageshavelexicalstress,inotherwords,asystematicmarkingofcertainsyllableswithinawordasmoreprominentthanothers.Importantly,thesethreeaspectsofwordprosodyarenotmutuallyexclusive.Forexample,therearetonelanguageswithstress(e.g.,Mandarin)andwithoutit(e.g.,Cantonese),andpitchaccentlanguageswithorwithoutstress(e.g.,SwedishandJapanese,respectively;Jun,2005).Thusinthediscussionbelow,itshouldbekeptinmindthatstressisawidespreadbutnotuniversalfeatureofhumanlanguage.

3.3.1RhythmicTypology

Fourapproachestorhythmictypologyaredescribedbelow.Behindalloftheseapproachesisacommondesiretounderstandtherelationshipsoftheworld’slinguisticrhythms.

PeriodicityandTypologyThemostinfluentialtypologyoflanguagerhythmtodateisbasedonthenotionofperiodicityinspeech.ThistypologyhasitsrootsintheworkofKennethPike(1945),whoproposedatheoryofspeechrhythmbasedonadichotomybetweenlanguagesintermsofsyllableandstresspatterns.Hedubbedcertainlanguages(suchasSpanish)“syllable-timed,”basedontheideathatsyllables(p.120) markoffroughlyequaltemporalintervals.Thesestoodincontrastto“stress-timed”languagessuchasEnglish,whichwerecharacterizedbyroughlyequaltemporalintervalsbetweenstresses.Toillustratestress-timedrhythm,Pikeinvitedthereaderto“noticethemoreorlessequallapsesoftimebetweenthestressesinthesentence”:

Pikethenaskedthereadertocomparethetimingofstressesintheabovesentencewiththefollowingone,andnoticethesimilarity“despitethedifferentnumberofsyllables”(p.34):

Rhythm

Page 24 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Pikearguedthatinstress-timedlanguagestheintervalsbetweenstressedsyllables(referredtoas“feet”)wereapproximatelyequaldespitechangingnumbersofsyllablesperfoot.Toachieveevenlytimedfeet,speakerswouldstretchorcompresssyllablestofitintothetypicalfootduration.Pikebelievedthatlearningtherhythmofalanguagewasessentialtocorrectpronunciation.Henoted,forexample,thatSpanishspeakerslearningEnglish“mustabandontheirsharp-cutsyllable-by-syllablepronunciationandjamtogether—orlengthenwherenecessary—Englishvowelsandconsonantssoastoobtainrhythmunitsofthestress-timingtype”(p.35).

Abercrombie(1967:34–36,96–98)wentfurtherthanPikeandproposedaphysiologicalbasisforstressversussyllabletiming.Thisboldstepwasbasedonaspecifichypothesisforhowsyllablesareproduced.Abercrombiebelievedthateachsyllablewasassociatedwithacontractionofmusclesassociatedwithexhalation(theintercostalmusclesoftheribcage),andthatsomecontractionswereespeciallystrong:Theselattercontractionsproducedstressedsyllables.Hereferredtothesetwotypesofcontractionsas“chestpulses”and“stresspulses”(thusonlysomechestpulseswerestresspulses;cf.Stetson1951).Abercrombieproposedthatinanygivenlanguage,oneortheotherkindofpulseoccurredrhythmically.Hethenequatedrhythmwithperiodicity:“Rhythm,inspeechasinotherhumanactivities,arisesoutoftheperiodicrecurrenceofsomesortofmovement…”(p.96).Furthermore,heclaimedthat“asfarasisknown,everylanguageintheworldisspokenwithonekindofrhythmorwiththeother”(p.97),namingEnglish,Russian,andArabicasexamplesofstress-timedlanguages,andFrench,Telugu,andYorubaasexamplesofsyllable-timedlanguages.JustasPikehaddone,henotedthatalanguagecouldnotbebothstress-timedandsyllable-timed.Becausetherearevariablenumbersofsyllablesbetweenstresses,equalizingthedurationofinterstressintervalsmeantthat“therateofsyllablesuccessionhastobecontinuallyadjusted,inordertofitvaryingnumbersofsyllablesintothesametimeinterval.”

(p.121) ItishardtooverestimatetheimpactofPikeandAbercrombieonthestudyofrhythminspeech.Theterms“stress-timed”and“syllable-timed”havebecomepartofthestandardvocabularyoflinguistics.Athirdcategory,“moratiming,”isalsoinstandarduse,andisusedtodescribetherhythmofJapanesespeech.Themoraisaunitthatissmallerthanthesyllable,usuallyconsistingofaconsonantandvowel,butsometimescontainingonlyasingleconsonantorvowel.Ladefoged(1975:224)statedthat“eachmoratakesaboutthesamelengthoftimetosay,”thusarguingfortheroughisochronyofmorae.11SincethepublicationofAbercrombie’sbook,manylanguageshavebeenclassifiedintooneofthesetwocategories(Dauer,1983;Grabe&Low,2002),andmanyresearchstudieshaveexaminedtheissueofisochronyinspeech.Inthissense,thestressversussyllable-timedtheoryofspeechrhythmhasbeenveryfruitful.Itprovidedaclear,empiricallytestablehypothesistogetherwithaphysiologicaljustification.

Inanothersense,however,thetheoryhasbeenanutterfailure.Empirical

Rhythm

Page 25 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

measurementsofspeechhavefailedtoprovideanysupportfortheisochronyofsyllablesorstresses(seereferencesinBertinetto,1989).12Totakejustafewexamplesfromthemanypapersthathavetestedtheisochronyhypothesis,Dauer(1983)showedthatEnglishstressfeetgrowindurationwithincreasingnumberofsyllables,ratherthanmaintainingtheevendurationnecessaryforisochrony(cf.Levelt,1989:393).Roach(1982)comparedEnglish,Russian,andArabictoFrench,Telugu,andYorubaanddemonstratedthattheformerstress-timedlanguagescouldnotbediscriminatedfromthelattersyllable-timedonesonthebasisofthetimingofinterstressintervals.Finally,Beckman(1982)andHoequist(1983)showedthatmoraearenotofequaldurationinJapanese.

Giventhatthenotionofperiodicityinordinaryspeechwasempiricallyfalsifiedover20yearsago,whydothelabelsofstress-timing,syllable-timing,andmoratimingpersist?Onereasonmaybethatitmatchessubjectiveintuitionsaboutrhythm.Forexample,Abercrombiehimself(1967:171)notedthattheideaofisochronousstressinEnglishdatesbacktothe18thcentury.AnotherreasonissuggestedbyBeckman(1992),whoarguesthatthistripartiteschemepersistsbecauseitcorrectlygroupstogetherlanguagesthatareperceivedasrhythmicallysimilar,evenifthephysicalbasisforthisgroupingisnotclearlyunderstood(andisnotisochronyofanykind).

(p.122) Thekeypointofthecurrentsection,then,isthatperiodicity,whichplayssuchanimportantroleinmuchmusicalrhythm,isnotpartoftherhythmofordinaryspeech.Thenextsectionexploresadifferentapproachtospeechrhythm,onethatsetsasidenotionsofisochrony.

PhonologyandTypologyThefactthatspeechisnotisochronousshouldnotleadustodiscardtheideaofspeechrhythm.Thatis,researchcanmoveforwardifonethinksofrhythmassystematictiming,accentuation,andgroupingpatternsinalanguagethatmayhavenothingtodowithisochrony.Oneproductiveapproachinthisframeworkisthephonologicalapproachtorhythmictypology.Thefundamentalideaofthisapproachisthattherhythmofalanguageistheproductofitslinguisticstructure,notanorganizationalprinciplesuchasstressorsyllableisochrony(Dauer,1983;cf.Dasher&Bolinger,1982).Inthisview,languagesarerhythmicallydifferentbecausetheydifferinphonologicalpropertiesthatinfluencehowtheyareorganizedaspatternsintime.OneclearexpositionofthisideaisthatofDauer(1983,1987),whopositedseveralfactorsthatinfluencespeechrhythm.

Thefirstfactoristhediversityofsyllablestructuresinalanguage.13Languagesvarysubstantiallyintheirinventoryofsyllabletypes.Forexample,Englishhassyllablesrangingfromasinglephoneme(e.g.,theword“a”)uptosevenphonemes(asin“strengths”),andallowsuptothreeconsonantsinonsetandcoda.Insharpcontrast,languagessuchasJapanese(andmanyPolynesianlanguages)allowfewsyllabletypesandaredominatedbysimpleCVsyllables.RomancelanguagessuchasSpanishandFrenchhavemoresyllabletypesthanJapaneseorHawaiianbutavoidthecomplexsyllablesfoundinlanguagessuchasEnglishandDutch,andinfactshowactiveprocessesthatbreakuporpreventthecreationofsyllableswithmanysegments(Dauer,1987).

Rhythm

Page 26 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Thediversityofsyllablesavailabletoalanguageinfluencesthediversityofsyllabletypesinspokensentences.Forexample,Dauer(1983)foundthatinasampleofcolloquialFrench,overhalfthesyllabletokenshadasimpleCVstructure,whereasinasimilarEnglishsample,CVsyllablesaccountedforonlyaboutone-thirdofthesyllabletokens.Thesedifferencesarerelevanttorhythmbecausesyllabledurationiscorrelatedwiththenumberofphonemespersyllable,suggestingthatsentencesofEnglishshouldhavemorevariablesyllabledurations(onaverage)thanFrenchsentences.

(p.123) Thesecondfactoraffectingspeechrhythmisvowelreduction.Insomelanguages,suchasEnglish,unstressedsyllablesoftenhavevowelsthatareacousticallycentralizedandshortinduration(linguistscommonlyrefertothissoundas“schwa,”aneutralvowelsoundinglike“uh”).Incontrast,inotherlanguages(suchasSpanish)thevowelsofunstressedsyllablesarerarelyifeverreduced,contributingtoalessvariablepatternofvoweldurationbetweenstressedandunstressedsyllables.

ThethirdrhythmicfactorproposedbyDaueristheinfluenceofstressonvowelduration.Insomelanguages,stresshasastrongeffectonthedurationofavowelinasyllable.Forexample,onerecentmeasurementofspokenEnglishfindsthatvowelsinstressedsyllablesareabout60%longerthanthesamevowelsinunstressedsyllables(Greenberg,2006).Incontrast,studiesofSpanishsuggestthatstressdoesnotconditionvoweldurationtothesamedegree(Delattre,1966).

Dauersuggestedthatlanguagestraditionallyclassifiedasstress-timedversussyllable-timeddifferintheabovephonologicalfeatures,withstress-timedlanguagesusingabroaderrangeofsyllabletypes,havingasystemofreducedvowels,andexhibitingastronginfluenceofstressonvowelduration.Thisnicelyillustratestheperspectiveofspeechrhythmasaproductofphonology,ratherthanacausalprinciple(e.g.,involvingperiodicity).14

Dauer’sproposalleadstotestablepredictions.Specifically,thethreefactorssheoutlines(diversityinsyllablestructure,vowelreduction,andtheinfluenceofstressonvowelduration)shouldallcontributetoagreaterdegreeofdurationalvariabilityamongthesyllablesofstress-timedversussyllable-timedutterances.Surprisingly,thereislittlepublisheddataondurationalvariabilityofsyllablesinsentencesofstressversussyllable-timedlanguages.Onereasonforthismaybethatthedemarcationofsyllableboundariesinspeechisnotalwaysstraightforward.Althoughpeoplegenerallyagreeonhowmanysyllablesawordorutterancehas,therecanbedisagreementaboutwheretheboundariesbetweensyllablesare,evenamonglinguists.Forexampledoesthefirst“l”intheword“syllable”belongtotheendofthefirstsyllableortothebeginningofthesecondsyllable,orisit“ambisyllabic,”belongingtobothsyllables?Althoughitistruethatsyllablemeasurementsaresubjecttodecisionsthatmayvaryfromoneresearchertothenext,thisshouldnotimpedeempiricalresearch:Itsimplymeansthatmeasurementsshouldbeaccompaniedbyanindicationofwhereeachsyllableboundarywasplaced.Ireturntothispointbelow.

Rhythm

Page 27 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(p.124) Beforeturningtoanotherphonologicalapproachtospeechrhythm,itisworthnotingthatthephonologicalpropertieslistedbyDauerdonotalwaysco-occur.ThusDauerarguedagainsttheideaofdiscreterhythmicclassesandforthenotionofarhythmiccontinuum.Insupportofthisidea,Nespor(1990)hasnotedthatPolishhascomplexsyllablestructurebutnovowelreduction(atnormalspeechrates),andCatalanhassimplesyllablestructurebutdoeshavevowelreduction.Thusthereiscurrentlyadebateinthefieldofspeechrhythmastowhetherlanguagesreallydofallintodiscreterhythmclassesorwhetherthereisacontinuumbasedonthepatternofco-occurrenceofrhythmicallyrelevantphonologicalfactors(cf.Arvaniti,1994;Grabe&Low,2002).Onlyfurtherresearchcanresolvethisissue,particularlyperceptualresearch(asdiscussedbelowinsection3.3.1,subsection“PerceptionandTypology”).

Inowturnbrieflytoadifferentphonologicaltheoryofspeechrhythm,proposedbyDwightBolinger(1981).AlthoughBolingerfocusedonEnglish,hisideasarequiterelevanttotypologicalissues.ThefoundationofBolinger’stheoryisthenotionthattherearetwodistinctsetsofvowelsinEnglish:fullvowelsandreducedvowels.By“reduced”vowelsBolingerdoesnotsimplymeanvowelsinunstressedsyllablesthatareshortandacousticallycentralized(i.e.,aphoneticdefinition).HearguesforaphonologicalclassofreducedvowelsinEnglish,whichbehavedifferentlyfromothervowels.Bolingerplacesthreevowelsinthisclass,an“ih”-likevowel,and“uh”-likevowel,anda“oh”-likevowel(moresimilarto“uh”thantothefullvowel“o”).Phoneticallyallofthesevowelsoccurinthecentralregionofvowelspace,neartheschwavowel/ə/ofEnglish(seeFigure2.19:Bolinger’s“ih”and“oh”vowelarenotshowninthatfigure,buttheformerwouldoccurjusttotheleftandupfrom/ə/,andthelatterwouldoccurjusttotherightandupfrom/ə/).Bolinger(1981:3–9)presentsargumentstosupportthenotionthatthesevowelsareaphonologicallydistinctsubclass,inotherwords,thattheybehaveincertainwaysthatfullvowelsdonot.Spacelimitationspreventadetaileddiscussionofthesearguments.HereIwillfocusontwoclaimsBolingermakesaboutfullandreducedvowelsthatarerelevantforspeechrhythm.

First,heclaimsthatsyllablescontainingfullandreducedvowelstendtoalternateinEnglishsentences.Second,heclaimsthatthereisa“lengtheningrule”suchthat“whenalongsyllableisfollowedbyashortone,theshortoneborrowstimefromitandmakesitrelativelyshort”(p.18).(Bya“long”syllable,hemeansasyllablewithafullvowel,andbya“short”syllable,hemeansasyllablewithareducedvowel;thereisnoclaimforaparticulardurationratiobetweenthetwotypesofsyllables.)Toillustratethisrule,Bolingeroffersthefollowingexample(notethatthefirstsentenceisfromanadforaspecialtypeofsoap):

(p.125)

Rhythm

Page 28 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Intheexampleabove,IhaveindicatedtheshortenedL’softhesecondsentencebyL-(afterFaber,1986).ThepointofthisexampleisthateachL-ofsentence3.6isshorterthantheL’sofsentence3.5,andthisoccurs(accordingtoBolinger)becauseeachS“borrowstime”fromtheprecedingL.Notethatsentence3.6hasstrictalternationbetweenLandS.Thisisaspecialcase:Bolingermakesnoclaimsforstrictalternation,onlyaclaimforatendency(thussequencessuchasLLSSSLSLL…areperfectlypossible).IsuspectBolingerchosethesentencesin3.5and3.6asexamplesbecausehefeltthateach(L-S)pairinsentence3.6isnotterriblydifferentindurationfromeachLinsentence3.5:ThisissuggestedbyhisgraphicalplacementoftheL’sinthetwosentencesaboveoneanother,inhisoriginaltext.However,thedurationalequivalenceofLand(L-S)isnotpartofBolinger’sclaim.Thisisanimportantpoint.Bolinger’stheorymayberelevanttothesubjectiveimpressionofiochrony(becauseoftheroughalternationofLandSandthelengtheningrule),butithasnoisochronyprinciple.

Faber(1986)arguesthatBolinger’stheoryissuperiortostress-timingtheorywhenitcomestoteachingtherhythmofEnglishtoforeignstudents(cf.Chela-Flores,1994).HealsopointsoutthatBolinger’stheorycanbeusedtoexplaincharacteristictimingpatternsthatstress-timingtheorycannotaccountfor,suchaswhy“cart”isshorterin:

thanin:

Orwhy“man”isshorterin:

thanin:

Bolinger’stheoryofspeechrhythmisdistinctfromthetheoryoutlinedbyDauerinthatitdealsnotjustwiththevariabilitysyllabledurationbutwiththepatterningofduration.Specifically,BolingerarguesthatthecharacteristicrhythmofEnglishisduetotheroughalternationofsyllableswithfullandreducedvowels,andtothewayfullvowelschangedurationwheninterveningreduced(p.126) syllablesareadded.Thisisalreadyenoughtosuggestabasisfortypologicaldistinctionsbetweenlanguages.Forexample,onemighttesttheideathatstress-timedlanguageshavemorecontrastinadjacentvoweldurationsthandosyllabletimedlanguages,andthatstress-timedlanguageshavelengtheningrules

Rhythm

Page 29 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

ofthetypesuggestedbyBolingerforEnglish(Bolingerhimselfdoesnotsuggesttheseideas,buttheyareanobviouscorollaryofhiswork).IfBolingerhadstoppedhere,hewouldalreadyhavemadeavaluablecontributiontospeechrhythmresearch.Bolinger’stheoryhasonefurthercomponent,however,thatrepresentsafundamentaldivergencefromthetheoryoutlinedbyDauer.

OnceagainfocusingonEnglish,Bolingersuggestedthatjustastherearetwokindsofvowels(fullandreduced),therearealsotwokindsofrhythm.Thefirstistherhythmicpatterningalreadydescribed,inotherwords,theroughalternationoflongandshortsyllablesandthelengtheningrule.Abovethislevel,however,isasecondlevelofrhythmicpatterningconcernedwithtemporalrelationsbetweenaccentscuedbypitch.Notethatthisideaentailsthenotionthatsyllabicrhythmisfundamentallyaboutdurationanddoesnotrelyonpitchasacue.Inotherwords,“thereisabasicleveloftemporalpatterningthatisindependentoftonalpatterning”(Bolinger1981:24,citingBruceBruce,1981).Bolingerarguesthatthistemporalpatterningwouldbeobservableeveninspeechspokenonamonotone.Speechisnotspokenonamonotone,however,andBolingerarguesthatsyllablesaccentedbypitchformasecondlevelofrhythmicpatterninginwhichthefundamentalruleisatendencytoseparatepitchaccentssothattheydonotoccurtoocloselytogetherintime.Themechanismforavoiding“accentclash”istomoveadjacentaccentsawayfromeachother,aphenomenonsometimescalled“stress-shift”inEnglish.(Oneoft-citedexampleofstressshiftiswhen“thirtéen”becomes“thírteenmén”;Liberman&Prince,1977.)Theterm“stress-shift”issomewhatunfortunate,becausethereisevidencethatwhatisshiftingispitchaccent,notsyllabledurationoramplitude(Shattuck-Hufnageletal.,1994).

Theideathatspeechrhythminvolvestemporalpatterningattwodistinctlinguisticlevelsmeritsfarmoreempiricalresearchthanithasgarneredtodate.Iwillreturntothisideainsection3.3.4.

DurationandTypologyUntilveryrecently,themeasurementofdurationhashadalargelynegativeroleinthestudyofspeechrhythm,namelyinfalsifyingclaimsfortheperiodicityofstressesorsyllables.Theinsightsofthephonologicalapproach,however,havecreatedanewpositiverolefordurationalmeasurements.Akeyfeatureofthisworkhasbeentheabandonmentofanysearchforisochrony,andafocusondurationalcorrelatesofphonologicalphenomenainvolvedinspeechrhythm.Ramusandcolleagues(1999),inspiredbytheinsightsofDauer,examinedthedurationalpatteringofvowelsandconsonantsinspeech,basedonideasabouthowsyllablestructureshouldinfluencethispatterning.Forexample,languages(p.127) thatuseagreatervarietyofsyllabletypes(i.e.,stress-timedlanguages)arelikelytohaverelativelylesstimedevotedtovowelsinsentencesthanlanguagesdominatedbysimplesyllables,duetothefrequentconsonantclustersintheformerlanguages.Bysimilarreasoning,thedurationalvariabilityofconsonantalintervalsinsentences(definedassequencesofconsonantsbetweenvowels,irrespectiveofsyllableorwordboundaries)shouldbegreaterforlanguageswithmorediversesyllablestructures.Thislatterpointisschematicallyillustratedin3.11,inwhich

Rhythm

Page 30 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

boundariesbetweensyllablesaremarkedwithadotandconsonantalintervalsareunderlined:

Notehowthegreaterdiversityofsyllabletypesin3.11aleadstogreatervariationinthenumberofconsonantsbetweenvowels(likelytotranslateintogreaterdurationalvariabilityofconsonantalintervals)aswellasalowervoweltoconsonantratio(likelytotranslateintoalowerfractionofutterancedurationspentonvowels).

Theseideaswereborneoutbyempiricalmeasurements.Figure3.7(fromRamusetal.,1999)showsagraphwithpercentofdurationoccupiedbyvowels(%V)versusconsonantalintervalvariability(ΔC)withinsentencesineightlanguages.(Thedataforeachlanguagecamefrom20sentencesreadbyfourspeakers,i.e.,fivesentencesperspeaker.)

Figure3.7 Percentageofsentencedurationoccupiedbyvowelsversusthestandarddeviationofconsonantalintervalswithinsentencesfor8languages.(CA=Catalan,DU=Dutch,EN=English,FR=French,IT=Italian,JA=Japanese,PO=Polish,SP=Spanish.)Errorbarsshow+/-1standarderror.FromRamus,Nespor,&Mehler,1999.

(p.128) Whatisinterestingaboutthisgraphisthatlanguagestraditionallyclassifiedasstress-timed(EnglishandDutch)havelow%VandhighΔCvalues,andoccupyadifferentregionofthegraphthanlanguagestraditionallyclassifiedassyllabletimed(French,Italian,andSpanish).Furthermore,Japanese,whichlinguistsplaceinadifferentrhythmiccategory(moratimed)isisolatedfromtheotherlanguages.(ThelocationofPolishandCatalaninthisgraphisdiscussedinthenextsection,onperception.)Thisdemonstratedanempiricalcorrelateoftraditionallinguisticrhythmicclasses,andhasinspiredotherresearcherstoexaminemorelanguagesinthisframework.OneinterestingstudyisthatofFrotaandVigário(2001),whoexaminedtherhythmofBrazilianPortugueseversusEuropeanPortuguese(henceforthBPandEP).Linguistshadoftenclaimedthatthesetwo

Rhythm

Page 31 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

varietieswererhythmicallydifferent,withEPbeingstress-timed,andBPbeingsyllable-timedorhavingmixedrhythmiccharacteristics.ThismakesPortugueseafascinatingtopicforspeechrhythmresearch,becauseonecanstudysentenceswithexactlythesamewordsbutspokenwithdifferentrhythms.(BritishEnglishandSingaporeEnglishprovideanothersuchopportunity,becausetheformerisstress-timedandthelatterhasbeendescribedassyllable-timed;seeLowetal.,2000.)FrotaandVigáriocomparedsentencesspokenbyEuropeanandBrazilianspeakersofPortuguese,andfoundthatEPhadsignificantlyahigherΔCandlower%VthanBP,aspredictedbyRamusetal.’sfindings.15

OneimportantquestionaboutthislineofresearchconcernstheperceptualrelevanceofΔCand%V.Ramusetal.focusedonthesemeasuresbecauseoftheirinterestintheroleofrhythmininfantspeechperception.Thereisevidencethatnewbornsandyounginfantscandiscriminatelanguagesthatbelongtodifferentrhythmicclasses(Mehleretal.,1988,1996;seealsothenextsection).Mehlerandcolleagues(1996)havearguedthatthisabilityhelpsbootstraplanguageacquisition:Onceagivenrhythmicclassisdetected,class-specificacquisitionmechanismscanbetriggeredthatdirectattentiontotheunitsthatarerelevantforsegmentingwordsfromconnectedspeech(e.g.,stressesinthecaseofEnglish,syllablesinthecaseofFrench,asdiscussedinsection3.3.3,subsection“TheRoleofRhythminSegmentingConnectedSpeech”).Forthistheorytowork,infantsmusthavesomebasisfordiscrimi(p.129) natingrhythmicclass.ThusRamusetal.(1999)soughtanacousticcorrelateofrhythmicclassthatwouldrequireminimalknowledgeaboutlinguisticunits.ΔCand%Varetwosuchparameters,becauseoneonlyneedassumethattheinfantcandistinguishbetweenvowelsandconsonants(seeGalvesetal.,2002,foranacousticcorrelateofΔCthatdoesnotevenrequiresegmentationintovowelsandconsonants).

Onemayask,however,ifΔCand%Varedirectlyrelevanttotheperceptionofspeechrhythm,oriftheyaresimplycorrelatedwithanotherfeaturethatismorerelevanttorhythmperception.Thatis,onecouldarguethatthesemeasuresareglobalstatisticsreflectingvariabilityinsyllablestructure,andarenotthemselvesthebasisofrhythmperceptioninspeech(cf.Barryetal.,2003).Amorepromisingcandidateforperceptualrelevancemaybevariabilityinsyllableduration,whichislikelytobecorrelatedwithvariabilityinsyllablestructureandwithvowelreduction.Becausethesyllableiswidelyregardedasafundamentalunitinspeechrhythm,andbecausebothadultsandinfantsaresensitivetosyllablepatterning(e.g.,vanOoyenetal.,1997),itwouldbeworthexaminingthecorpusofsentencesusedbyRamusetal.forsyllabledurationvariabilitytoseeifthisparameterdifferentiatestraditionalrhythmicclasses.ThiswouldalsobeastraightforwardtestofDauer’sideas,asthephonologicalfactorssheoutlinesimplythatsyllabledurationvariabilityshouldbehigherinsentencesofstress-timedthanofsyllable-timedlanguages.

Surprisingly,therehasbeenlittleempiricalworkcomparingsentence-levelvariabilityinsyllabledurationamongdifferentlanguages.Asnotedintheprevioussection,thismayreflectthedifficultiesofassigningsyllableboundariesinconnectedspeech.Fromapurelypracticalstandpoint,itiseasiertodefinephonemeboundaries,usingcriteriaagreed

Rhythm

Page 32 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

uponbymostphoneticians(e.g.,Peterson&Lehiste,1960).However,thisshouldnotstopresearchintosyllabicdurationpatterns,becausethesepatternsarelikelytobeperceptuallyrelevant.Toillustrateboththefeasibilityandthechallengesofasyllable-basedapproach,examples3.11canddbelowshowasentenceofEnglishandFrenchsegmentedatsyllableboundaries(thesegmentationsweredonebymyselfandFranckRamus,respectively).Periodsindicatesyllableboundariesthatwefeltwereclear,whereassquarebracketsindicatephonemesthatseemedambiguousintermsoftheirsyllabicaffiliation.Inthelattercase,onemustdecidewheretoplacethesyllableboundary.Forexample,ifthephonemesoundsambisyllabicthentheboundarycanbeplacedinthemiddleofthephoneme,orifitsoundslikeithasbeenresyllabifiedwiththefollowingvowel,theboundarycanbeplacedbeforethephoneme.

(3.11c)The.last.con.cert.gi.ven.at.the.o[p]era.was.a.tre.men.dous.suc.cess

(3.11d)Il.fau.dra.beau.coup.plus.d’ar.gent.pour.me.ne[r]à.bien.ce.pro.jet

Itislikelythatdifferentresearcherswillvaryinhowtheymakethesejudgmentcalls.Nevertheless,thisisnotaninsurmountableproblemforrhythmresearch.Infact,ifdifferentresearchersdefinesyllableboundariesinslightlydifferentwaysbutneverthelessconvergeontherhythmicdifferencestheyfindbetweenlanguages,thisisstrongevidencethattheobserveddifferencesarerobust.16Figure3.8aand3.8bshowmymarkingsofsyllableboundariesinthewaveformandspectrogramsofthesetwosentences(thesentencescanbeheardinSoundExamples3.5aandb;notethatinsentence3.5a,“opera”ispronounced“opra”).

(p.130)

Figure3.8a AsentenceofBritishEnglishsegmentedintosyllables.(Notethat“opera”ispronounced“opra”bythisspeaker.)

Forthesesentences,thevariabilityofsyllabledurationsasmeasuredbythecoefficientofvariation(thestandarddeviationdividedbythemean)is.53fortheEnglishsentence

Rhythm

Page 33 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

and.42fortheFrenchsentence.MakingsimilarmeasurementsonalltheEnglishandFrenchsentencesintheRamusdatabaseyieldsthedatainFigure3.8c.Ascanbeseen,onaverageEnglishsentenceshavemorevariablesyllabledurationsthandoFrenchsentences(thedifferenceisstatisticallysignificant,p〈0.01,Mann-WhitneyUtest).Itwouldbeinterestingtohavesimilarvariabilitymeasurementsfornumerouslanguagesthathavebeenclassifiedasstress-versussyllable-timed:Wouldthesemeasurementsdividethelanguagesintotheirtraditionalrhythmicclasses?(SeeWagner&Dellwo,2004,forapromisingstart.)

(p.131)

Figure3.8b AsentenceofFrenchsegmentedintosyllables.

TurningnowtotheideasofDwightBolinger,recallBolinger’sclaimthatsyllablescontainingfullandreducedvowelstendtoalternateinEnglish.Thisleadstoanempiricalprediction,namelythatthedurationalcontrastbetweenadjacentvoweldurationsinEnglishsentencesshouldbegreaterthaninlanguagesofadifferentrhythmicclass,suchasFrenchorSpanish.Infact,thereisresearchsupportingthisprediction,thoughitwasnotinspiredbyBolinger’sworkbutbyaninterestintherolethatvowelreductionplaysintherhythmofstress-versussyllable-timedlanguages.Low,Grabe,andNolan(2000)setouttoexploretheideathatvowelreductioncontributestotheimpressionofstress-timingviaitsimpactonvoweldurationvariabilityinsentences.Theytestedthisideabyexaminingvoweldurationpatterninginastress-timedversusasyllable-timedvarietyofEnglish(Britishvs.SingaporeEnglish).Crucially,theydevelopedanindexofvariabilitythatwassensitivetothepatterningofduration.Their“normalizedpairwisevariabilityindex”(nPVI)measuresthedegreeofcontrastbetweensuccessivedurationsinanutterance.AnintuitionforthenPVIcanbegainedbyexaminingFigure3.9,whichschematicallydepictstwosequencesofeventsofvaryingduration(thelengthofeachbarcorrespondstothedurationoftheevent).

(p.132)

Rhythm

Page 34 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Figure3.8c Thecoefficientofvariation(CV)ofsyllabledurationin20Englishand20Frenchsentences.Errorbarsshow+/-1standarderror.

InsequenceA,neighboringevents(e.g.,events1and2,2and3)tendtohavealargecontrastinduration,andhencethesequencewouldhavealargenPVI.NowconsidersequenceB,whichhasthesamesetofdurationsassequenceA,arrangedinadifferenttemporalorder.Nowneighboringeventstendtohavelowcontrastinduration,givingthesequencealownPVIvalue.Hencethetwosequenceshaveasharpdifferenceindurationalcontrastiveness,eventhoughtheyhaveexactlythesameoverallamountofdurationalvariability,forexample,asmeasuredbythestandarddeviationofdurations.(Seethischapter’sappendix1forthenPVIequation.)

Figure3.9 Schematicofsequencesofeventswithvaryingduration,toillustratethenPVI(longerbars=longerdurations).Seetextfordetails.

(p.133) BecausethenPVIisfundamentallyameasureofcontrast,theuseoftheterm“variability”initsnameissomewhatunfortunate,asvariabilityandcontrastarenotnecessarilycorrelated,asshowninFigure3.9.Infact,itisquitepossibletohavetwosequencesAandBinwhichthevariabilityofdurationsinAisgreaterthanB,butthenPVIofdurationsisgreaterinBthanofA(anexampleisgiveninsection3.5.1).Thusabettertermforthismeasuremighthavebeenthe“normalizedpairwisecontrastivenessindex.”

Rhythm

Page 35 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

IhavedelvedintothedetailsofthenPVIbecauseithasprovenquitefruitfulinthestudyofspeechrhythmandinthecomparativestudyoflinguisticandmusicalrhythm(discussedin3.5.1).GrabeandLow(2002)haveusedthenPVItoexaminethepatterningofvoweldurationsinsentencesofanumberoflanguages,andhaveshownthatseverallanguagestraditionallyclassifiedasstress-timed(suchasGerman,Dutch,BritishEnglish,andThai)havealargervocalicnPVIthananumberofotherlanguagestraditionallyclassifiedassyllabletimed(suchasFrench,Italian,andSpanish).ThissupportsBolinger’sideathatdurationalalternationofvowelsisimportanttostress-timedrhythm.17Inspiredbythiswork,Ramus(2002b)measuredthevowelnPVIforalleightlanguagesinhisdatabaseandfoundtheresultsshowninFigure3.10.

Figure3.10plotsthenPVIforvocalicintervalsagainsttherPVIforintervocalicintervals(i.e.,consonantalintervals).(TherPVI,or“rawpairwisevariabilityindex,”iscomputedinthesamewayasthenPVIbutwithoutthenormalizationterminthedenominator;cf.thischapter’sappendix1.GrabeandLow[2002]arguethatnormalizationisnotdesirableforconsonantalintervalsbecauseitwouldnormalizeforcross-languagedifferencesinsyllablestructure.)FocusingonthenPVIdimension,thestress-timedlanguages(EnglishandDutch)areseparatedfromthesyllable-timedlanguages(Spanish,Italian,andFrench),whichprovidesadditionalsupportforBolinger’sideas.18Furthermore,Polishisnowfarfromthestress-timedlanguages,whichisinterestingbecausethereisperceptualevidencethatPolishisrhythmicallydifferentfromtheselanguagesduetoitslackofvowelreduction(seesection3.3.1,subsection“PerceptionandTypology”).JapaneseissimilartoFrenchintermsofnPVI,however,suggestingthatnPVIaloneisnotenoughtosortlanguagesintotraditionalrhythmicclasses.AddingaseconddimensionofrPVIforconsonantalintervals,however,doessegregateoutJapanese,whichhasverylowdurationalcontrastbetweensuccessiveconsonantalintervals.Thissuggeststhatatleasttwophoneticdimensionsmaybeneededtocapturedifferencesbetweenrhythmicclasses(seealsoRamusetal.,1999).

(p.134)

Rhythm

Page 36 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Figure3.10 VocalicnPVIversusConsonantal(intervocalic)rPVIforsentencesineightlanguages.(CA=Catalan,DU=Dutch,EN=English,FR=French,IT=Italian,JA=Japanese,PO=Polish,SP=Spanish.)Errorbarsshow+/-1standarderror.FromRamus,2002a.

OneinterestinglinguisticapplicationofthenPVIhasbeentotheontogenyofspeechrhythm.IthasbeenclaimedthattherhythmofEnglish-speakingchildrenissyllable-timedincontrasttothestress-timedrhythmofadultspeech(Allen&Hawkins,1978).Grabeetal.(1999)conductedannPVIstudythat(p.135) supportedthisclaim.TheymeasuredthenPVIofvowelsinthespeechofEnglishversusFrenchspeaking4-year-oldsandtheirmothers.TheyfoundthatEnglishchildrenhadsignificantlylowernPVIvaluesthantheirmothers,whereasFrenchchildrenresembledtheirmothersinhavingalownPVIs.Thatis,bothEnglishandFrenchchildrenspokewithasyllable-timedrhythm(thoughthenPVIoftheEnglishchildrenwasalreadylargerthanthatoftheirFrenchcounterparts).ItwouldbeinterestingtotracknPVIasafunctionofageinEnglishandFrenchchildren,tostudythedevelopmentaltimecourseofspeechrhythminthetwolanguages.

AllnPVIstudiestodatehavefocusedonasinglerhythmiclayerinlanguage:thetemporalpatterningofvowelsorconsonants.InthespiritofBolinger’sideathatrhythmmayinvolvemultiplelevelsoftemporalorganization,itwouldbeworthusingthenPVItoexploretherelationshipofdurationalpatternsatvariousrhythmicallyrelevantlevelsinspeech(cf.Asu&Nolan,2006).Forexample,withinEnglishsentences,onecouldcomputethenPVIofinterstressintervals(ISIs)relativetothenPVIofsyllabledurationsbymeasuringbothofthesequantitiesineachsentenceandthentakingtheratiooftheformertothelatter.ItmaybethatthesubjectiveimpressionofisochronyinEnglisharisesinpartfromalowerdurationalcontrastbetweenISIsthanbetweensyllables,whichwouldmakethisratiosignificantlylessthan1.Ireturntothisideainsection3.3.4.

Thissectionhasreviewedafewdifferentacousticcorrelatesofspeechrhythm.Duetothesuccessofthiswork,itseemscertainthatmoresuchcorrelateswillbeproposedandexploredinthefuture(e.g.,Gut,2005).Ultimately,theusefulnessofsuchmeasureswilldependonwhethertheygrouptogetherlanguagesthatareperceivedasrhythmicallysimilaranddividelanguagesperceivedasrhythmicallydifferent.Perceptualstudiesare

Rhythm

Page 37 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

thusfundamentaltoresearchonrhythmictypology,anditistosuchstudiesthatweturnnext.

PerceptionandTypologyAlltypologicaltheoriesoflanguagerhythmareultimatelyrootedinperception.Inthepast,linguistshavedefinedrhythmcategories(suchasstressvs.syllabletiming)basedontheirauditoryimpressionsoflanguages,andthenresearchershavesoughttoidentifyphonologicalandacousticcorrelatesoftheseclasses.Therecentsuccessinfindingdurationalcorrelatesoftraditionalrhythmclassesisatestamenttotheintuitionoflinguistsintheirauralrhythmicjudgments.However,itisalsoapparentthattheoldcategorizationsystemhasitsshortcomings.Forexample,somelanguagesstraddledifferentcategories(e.g.,PolishandCatalan,seeabove),andmanylanguagesdonotfitneatlyintoanyoftheexistingcategories(Grabe&Low,2002).Thustheoldsystemiscrackingattheseams,andanewscienceofrhythmclassificationiscalledfor.Suchasciencemusthaveasitsfoundationabodyofperceptualdatathatprovidesa(p.136)measureoftherhythmicsimilaritiesanddifferencesbetweenlanguages.Thesedatawillallowresearcherstoconstructaperceptualmapoflanguagerhythmsanddeterminetowhatextenttherhythmsofhumanlanguagesfallintodistinctclusters(vs.formingacontinuum).Itwillalsohelpsuggestnewavenuesforempiricalresearchintotheacousticfoundationsofspeechrhythm.

Fortunately,perceptualworkontherhythmicdifferencesbetweenlanguageshasalreadybegun.AninnovativestudybyRamusandMehler(1999)devisedamethodforstudyingtheperceptionofspeechrhythmpositedontheideathatifalistenercantelltwolanguagesapartwhentheonlycuesarerhythmic,thenthelanguagesbelongtodistinctrhythmicclasses.Speechresynthesistechniqueswereusedtoselectivelyremovevariousphoneticdifferencesbetweenlanguagesandfocusattentiononrhythm.SoundExamples3.6and3.7illustrateRamusandMehler’stechniqueonasentenceofEnglishandJapanese.Eachsentenceispresentedinfourversions,whichconverttheoriginalsentencetoanincreasinglyabstracttemporalpatternofvowelsandconsonants.Inthefirsttransformation,eachphonemeisreplacedbyaparticularmemberofitsclass:allfricativesreplacedby/s/,vowelsby/a/,liquids(l&r)by/l/,plosivesby/t/,nasalsby/n/,andglidesby/ai/(aconditiontheycalled“saltanaj,”pronounced“sal-tan-ai”).Theoriginalintonationofeachsentenceispreserved.Inthesecondtransformation,allconsonantsarereplacedby/s/andallvowelsby/a/(aconditiontheycall“sasasa”).Inthefinaltransformation,thevoicepitchisflattenedtoamonotone,leavingthetemporalpatternofvowelsandconsonantsastheonlydifferencebetweenthelanguages(aconditiontheauthorsrefertoas“flatsasasa”).RamusandMehlerfoundthatFrenchadultscoulddiscriminatebetweenEnglishandJapaneseinallthreeconditions,supportingthehypothesisthattherhythmsofEnglishandJapaneseareindeedperceptuallydistinct.

Focusingontheflatsasasatransformation,Ramusetal.(2003)alsotestedFrenchadults’abilitytodiscriminatetherhythmsofEnglish,Polish,Spanish,andCatalan.TheresultsindicatedthatPolishcouldbediscriminatedfromtheotherlanguages,whereasCatalancouldnotbediscriminatedfromSpanish,thoughitwasdistinctfromEnglishandPolish.

Rhythm

Page 38 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(Recallthatonphonologicalgrounds,PolishandCatalanseemedintermediatebetweenstress-timedandsyllable-timedlanguages;cf.section3.3.1,subsection“PhonologyandTypology.”)TheseperceptualdatasuggestthatPolishdoesbelonginaseparaterhythmiccategorythanEnglish,whereasCatalanbelongsinthesamecategoryasSpanish.Thisfindinghasimplicationsformapsoftheacousticcorrelatesofspeechrhythm,suchasFigure3.7.Inthatmap,Polishclusteredwithstress-timedlanguages,indicatingthatadifferentacousticdimensionisneededtoseparateperceivedrhythmicclasses.Indeed,Ramusetal.(1999,2003)havenotedthatPolishcanbeseparatedoutfromallotherlanguagesintheiroriginalstudyonadimensionthatmeasuresthevariabilityofvoweldurationinasentence,ΔV,becausePolishhasaverylowvoweldurationvariabilitycomparedtoallotherlanguagesin(p.137) theirsample.19Thusperceptualworkonspeechhasalreadysuggestedthatifonewishestopreservethenotionofrhythmclassesinlanguage,atleastfourclassesareneeded:stress-timed,syllable-timed,mora-timed(representedbyJapanese),andoneotheryet-to-be-namedcategoryrepresentedbyPolish.20

Anotherlineofperceptualresearchconcernedwithrhythmictypologyhasfocusedonnewbornsandinfants.Thischoiceofsubjectsmayseemsurprising,butthesestudiesaremotivatedbytheideathatveryyounghumansaresensitivetospeechrhythmanduseittoguidelearningoffine-grainedsoundpatternsoflanguage(Mehleretal.,1996).Nazzietal.(1998)studiednewbornrhythmperceptionusinglow-passfilteredspeech.Thisremovesmostofthephoneticinformationbutpreservessyllable,stress,andpitchpatterns.TheyshowedthatFrenchnewbornsareabletodiscriminateEnglishfromJapanese,butnotEnglishfromDutch,suggestingthatthelatteraremembersofthesamerhythmicclass.TheyalsoshowedthatthenewbornscoulddiscriminateEnglishandDutchfromSpanishandItalian,butnotEnglishandSpanishfromDutchandItalian,suggestingthattheformerpairingsmoreaccuratelycaptureperceptualrhythmicclasses(cf.Nazzietal.,2000,forconvergingfindingswith5-month-oldinfants).Thesefindingssupporttheauthors’hypothesisthatbabiescandiscriminatelanguagesonlywhentheybelongtodifferentrhythmicclasses,anotionthattheydubthe“rhythmhypothesis”forlanguageacquisition.Ifthisistrue,thentheearsofinfantsmaybeparticularlyimportantinstrumentsinmappinghumanspeechrhythmsinfutureresearch.21

ThestudiesofRamus,Nazzi,andcolleaguesraiseanumberofpointsforfutureresearchontheperceptionofspeechrhythm.First,itisimportanttodesignstimuliandtasksthatfocusattentiononthoseaspectsofspeechrhythmthatplayaroleinnormalspeechperception.Forexample,adangeroftheflatsasasaconditionitthatwhenalanguagewithahighlyvariablesyllablestructure(suchasEnglish)iscomparedtoalanguagedominatedbysimplesyllables(suchasJapanese),asalientperceptualdifferencebetweentheresultingflat(p.138) sasasastimuliisthemorefrequentoccurrenceoflong-duration/s/soundsintheformerstimulus(whichresultfromtransformingconsonantclustersintosingle,long/s/sounds).Thusdiscriminationcouldsimplybebasedonlisteningforfrequentlong/s/soundsratherthanonattendingtotemporalstructure.

Rhythm

Page 39 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Second,researchontheperceptualtaxonomyoflanguagerhythmsshouldnotonlybebasedondiscriminationtasks,butshouldalsoincorporatesimilarityjudgments.Musicalstudiesofrhythmicsimilarityprovideagoodmodelforsuchresearch(Gabrielsson,1973,1993).Inthisresearch,rhythmsarepresentedinapairwisefashionandlistenersratetheirperceivedsimilarityusinganumericalscale.Theresultingratingsarestudiedusingmultidimensionalscalingtouncoverperceptualdimensionsusedbylistenersinclassifyingrhythms.Thisparadigmcouldeasilybeadaptedtostudyspeechrhythm,usinglow-passfilteredspeechwithminimalpitchvariationasstimuli.Suchstudiesshouldbesensitivetotheideathattheimportantperceptualdimensionsforrhythmmayberelational,forexample,ahighcontrastivenessbetweensuccessivesyllabledurationswhilesimultaneouslyhavingalowerdurationalcontrastivenessbetweeninterstressintervals(cf.theendofsection3.3.1,subsection“DurationandTypology”).

Finally,afundamentalissueforallfuturestudiesofrhythmictypologyistheextenttowhichperceivedrhythmicsimilaritiesanddifferencesbetweenlanguagesdependonthenativelanguageofthelistener.Thetheoryofstress,syllable,andmoratimingwasproposedbynativeEnglishspeakers,anditisanopenquestionwhetherspeakersofotherlanguagesperceiverhythmiccuesinthesamewaythatEnglishspeakersdo.Forexample,ithasrecentlybeendemonstratedthatFrenchlistenershavesomedifficultydistinguishingnonsensewordsthatdifferonlyinthelocationofstress,whereasSpanishlistenershavenosuchdifficulty(Dupouxetal.,2001).ThislikelyreflectsthefactthatSpanishhascontrastivestress:Twowordscanhavethesamephonemesbutadifferentstresspattern,andthiscanchangethemeaningofthewordentirely(e.g.,sábanavs.sabána,whichmean“sheet”and“savannah”respectively;cf.Soto-Faracoetal.,2001).Frenchdoesnothavethisproperty,andDupouxetal.suggestthatthisdifferenceisresponsibleforthe“stressdeafness”theyfoundintheirFrenchlisteners.Resultssuchasthisraiseafundamentalquestion:Isthereasinglemapofperceivedrhythmicsimilaritiesanddifferencesamonglanguages,ordoesthegeographyofthemapdifferaccordingtothenativelanguageofthelistener?Onlyempiricalworkcanresolvethisissue,butitseemsarealpossibilitythatnativelanguageinfluencestheperceptionofrhythmicrelationsbetweenlanguages.

3.3.2PrinciplesGoverningtheRhythmicShapeofWordsandUtterances

Forthoseinterestedincomparingrhythminlanguageandmusic,itisimportanttobefamiliarwithabranchoftheoreticallinguisticsknownas“metrical(p.139) phonology.”Metricalphonologydealswithspeechrhythm,butitdoessoinamannerquitedifferentfromtheapproachesdescribedsofar.Firstandforemost,rhythmicprominenceistreatedashierarchical.Thatis,prominenceisincrementallyassignedateachleveloftheprosodichierarchyaccordingtosystematicprinciples.Forexample,inagiventheoryitmaybethecasethatallsyllablesbeginwithabasicamountofprominence,thenthelexicallystressedsyllableofeachword(orcliticgroup)isassignedanadditionaldegreeofprominence,thenaphrase-levelprominenceisaddedtoaparticularwordofaphrase(e.g.,the“nuclearstressrule”inEnglish),andsoon.Inthisview,prominenceisnotsimplyabinaryphoneticfeaturecalled“stress”thatsyllableseitherhaveordonot.Rather,prominenceisanacousticprojectionofthehierarchicalprosodicstructureofan

Rhythm

Page 40 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

utterance,andassuch,hasseveraldegreesthatservetoindicateasyllable’spositioninanutterance’srhythmichierarchy(Halle&Vergnaud,1987;Halle&Idsardi,1996;Shattuck-Hufnagel&Turk,1996).22

OneoftheclearestexpositionsofmetricalphonologyisinSelkirk’s1984book,PhonologyandSyntax:TheRelationBetweenSoundandStructure.Onegoalofthisbookistoshowhowonecangofromastringofwordstoarepresentationofthesyllabicprominencepatternofthespokenutteranceinarule-governedfashion.Therelativeprominenceofsyllablesisrepresentedusinga“metricalgrid”thattreatseachsyllableasapointinabstracttime(Figure3.11),meaningthatprominencepatternsareconsideredwithoutregardtotheirexacttiming.

Twoaspectsofthelinguisticmetricalgrid,introducedbyLiberman(1975),embody“theclaimthattherhythmicorganizationofspeechisquiteanalogoustothatofmusic”(Selkirk,1984:9).First,asnotedabove,prominenceistreatedhierarchically,analogouslytohierarchicaltheoriesofmusicalmeter(Cooper&Meyer,1960;Lerdahl&Jackendoff,1983).Abovethebasiclevelofthesyllableareseveralotherlevels.Thesecondlevelmarksstressedsyllables,andisthelevelofthebasic“beat,”inanalogytothetactusinmusic(Selkirk,1984:10,40).Thethirdlevelmarkstheprimarylexicalstressofeachword,andthefourthlevelmarksthemainaccentofeachphrase.This“text-to-grid”assignmentofbeatsprovidestheinputonwhichrhythmicprinciplesoperate.Theseprinciples,whichrepresentthesecondlinktomusicalmeter,amounttoatendencytoalternatebetweenstrongerandweakerelementsateachlevelofthehierarchy.Theprinciplesareenforcedbyrulesthatcanadd,delete,andmovebeatstomakethepatternateachlevelmorecongruouswithanalternatingpattern.Forexample,aruleof“beataddition”mightaddabeatatthesecondleveltoavoidalongseriesofunstressedsyllables.Atthethirdlevel,aruleof“beatmovement”mightshifttheprimarystress/accentofwordtoavoidtheadjacencyofprimarylexicalstress/accent(aswhen“thirtéen”becomes“thírteenmén”;LibermanandPrince,1977,Shattuck-Hufnageletal.,1994;Grabe&Warren,1995).Theidealgoalisthatstrongbeatsatanygivenlevelareseparatedbynomorethantwoweakbeatsatthatlevel(the“principleofrhythmicalternation”).Thusmetricalphonologyderivestheprominencepatternofasentenceusingideasdirectlyinspiredbytheoriesofmusicalmeter.

(p.140)

Figure3.11 AmetricalgridforasentenceofEnglish.FromSelkirk,1984.

Thenotionthatspeechhasmultiplerhythmicallyrelevantlevelsisaninterestingabstractsimilaritybetweenrhythminlanguageandmusic,becauseafundamentalpropertyof

Rhythm

Page 41 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

musicalmeteristheexistenceofperceptuallysalienttemporalpatteringonmultipletimescales(cf.section3.2.2).Furthermore,justasmusicalmeterinvolvesatleastonepsychologicallyaccessiblerhythmiclevelbelowthebeatandoneortwolevelsaboveit,metricalphonologyproposesrhythmiclevelsbelowandabovethe“beat”ofstressedsyllables.Thatis,boththeoriesconcernthepatterningoftimeintervalsatseveraltimescales.

Althoughthepicturepaintedbymetricalphonologyisanelegantone,itshouldbenotedthatitsclaimsarebynomeansuniversallyacceptedbyspeechscientists(Cooper&Eady,1986),andthatthepatternsofprominenceitproposesaretypicallyconstructedfromtheintuitionsoflinguistsratherthanfromacousticandperceptualdatacollectedinlaboratorysettings.However,thereishopethatthefieldcanbeputonanempiricalfooting.Forexample,thereisphoneticevidenceforfourdegreesofprominenceinspeech(atleastinstress-timedlanguages),correspondingtoreducedvowels,fullvowels,stressedsyllables,andaccentedsyllables(Terken&Hermes,2000).Forthecurrentpurposes,metricalphonologyisinterestingbecauseitdrawsattentiontoanumberofissuesinwhichcomparisonsoflinguisticandmusicalrhythmareinstructive.Oneoftheseissues(multiplelayeringinrhythmicstructure)leadstoideasforempiricalcomparativestudiesofrhythminspeechandmusic,andisdiscussedfurtherinsection3.3.4.Twootherissuesarediscussedbelow.

(p.141) DifferencesbetweenLinguisticandMusicalMetricalGridsAlthoughthehierarchiespositedbymetricalphonologywereinspiredbyWesternmusic,someveryimportantdifferencesbetweenthemetersofmusicandlanguagearereadilyapparent.Mostnotably,temporalperiodicityinmusicalmeterismuchstricterthananythingfoundinspeech,andthisdifferencehasdramaticcognitiveconsequences.Theregularperiodicitiesofmusicallowmetertoserveasamentalframeworkforsoundperception,suchthataneventcanbeperceivedasmetricallyprominentevenifisphysicallyquiteweak,asinsyncopatedrhythms.Bycomparison,theprominencesoflanguagearenotregularenoughtoallowforanythingasabstractassyncopation.Asaresult,linguisticmetricalgridsarenotabstractperiodicmentalpatterns(likemusicalmetricalgrids)butaresimplymapsofheardprominences,fulloftemporalirregularities.Forexample,Dauer(1983)reportsthattheaverageinterstressintervalinspeechwasaround450ms,withastandarddeviationofapproximately150ms.Dividingthestandarddeviationbythemeanyieldsacoefficientofvariationofabout33%.Thisvariabilityismarkedlydifferentfrommusic,inwhichperceivedbeatsoccurinamuchmoreevenlyspacedfashion.Forexample,whentappingtomusic,adultsshowacoefficientofvariationofabout5%.Thus“metricalgrids”inlanguageshouldperhapsbecalled“prominencegrids,”toavoidtheimplicationofanabstractmentalperiodicity.

QuestioningthePrincipleofRhythmicAlternationinSpeechSettingasidequestionsoftemporalperiodicity,onecanaskifspeechandmusicshareamoreabstractsimilarityintermsofatendencytoarrangeprominencesinalternatingpatternsofstrongandweakelements.Ifso,thismightsuggestabasiccognitiverelationshipbetweenrhythminlanguageandmusic.Evidenceinfavorofaprincipleof

Rhythm

Page 42 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

alternationinlanguagecomesfromstudiesshowingthatEnglishspeakersadjustprominencepatternstomakethemmoreregular.Forexample,KellyandBock(1988)hadspeakerspronouncenonsensewordsembeddedinsentences,suchas:

(3.12a)Thefullteplezdecreased.

(3.12b)Throwtheteplezbadly.

Thefocusofinterestwaswhetherspeakersstressedthefirstorsecondsyllableofthenonsenseword.Overall,speakerstendedtostressthefirstsyllable,inaccordancewithageneraltrendinEnglishfordisyllabicnounstohaveinitialstress.However,thistendencywassignificantlyweakerwhenthenonsensewordwasprecededbyastressedsyllable(asinsentence3.12a),asifspeakerswantedtospreadouttheoccurrenceofstresses.Furtherevidenceforalternation(p.142) ofstresspatternsinspeechproductioncomesfromCutler(1980),whoexaminedsentencesinwhichspeakersinadvertentlyomittedasyllable,suchas:

versustheintendedsentence:

Muchmoreoftenthanchance,theerrorsservedtoshortenalongrunofunstressedsyllables,thustendingtopromotethealternationofstressedandunstressedsyllables.

Althoughthesefindingsseemtosupportapositiveprincipleofalternation,itisalsopossiblethattheyreflecttheactionofanegativeprinciplethatseekstobreakupclustersofprominentsyllablesorclustersofnonprominentsyllables(i.e.,“stressclashes”and“stresslapses,”Nespor&Vogel,1989).SomesupportforthelatterviewcomesfromtheobservationthattheregularizingtendenciesreportedbyKellyandBock(1988)andCutler(1980)areactuallyratherweak.Intheformerstudy,subjectsplacedinitialstressonthetargetnonsensewordinthemajorityofcases,whetherornottheimmediatelyprecedingsyllablewasstressed.Thepresenceofapriorstressedsyllablesimplyloweredtheproportionofinitialstressfrom80%to70%,suggestingonlyamildtendencytomaintainanalternatingstresspatterninspeech.Similarly,Cutler’sstudyisbasedoncollectingrelativelyraresyllableomissionerrors,meaningthatspeakersusuallymanagequitewellwithirregularprominencepatterns.

Thusatthecurrenttimeitisimpossibletoruleoutthehypothesisthatthetendencytoalternatestrongerandweakersyllablesinspeechistheresultofnonrhythmicforcesthatseektokeepprominencesatacomfortabledistancefromeachother.Infact,researchonGreeksuggeststhatthealternationofprominencemaynotevenbeauniversalpatternforhumanlanguages,becauseGreektolerateslongsequencesofunstressedsyllables

Rhythm

Page 43 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(Arvaniti,1994).Thusitmaybethattheonlyuniversalprincipleregardingprominencepatternsinlanguageisthatprominencesthataretooclosetogetheraresubjecttolinguisticmechanismsforclashavoidance(seeArvaniti,1994,forevidencefromGreek;andNespor,1990,forreferencestostudiesofclashavoidancemechanismsinnumerouslanguages).Thereasonsuchmechanismsexistmayultimatelyberootedinthemechanicsofarticulation.Stressedsyllablestendtobemadewithlargerjawmovementsthanunstressedsyllables(deJong,1995),anditmaybebiomechanicallyadvantageoustoavoidcrowdinglargerjawmovementstogetherwhenspeakingatthefastratestypicalofnormalconversation.

(p.143) 3.3.3ThePerceptionofSpeechRhythmTheroleofrhythminspeechperceptionhasbeenthefocusofatleastfourdifferentlinesofresearch.Twoofthesehaveobviousconceptualconnectionstomusicalrhythm:thestudyofperceivedisochronyinspeechandtheinvestigationoftheroleofrhythmicpredictabilityinperception.Thethirdlinehasinvestigatedtheroleofspeechrhythmintheperceptualsegmentationofwordsfromconnectedspeech.Althoughnotobviousatfirst,thisresearchisinfactquitepertinenttocomparativeperceptualstudiesofrhythminlanguageandmusic.Thefinal(andmostrecent)lineofworkconcernstherolethatrhythmplaysinperceptionofnonnativeaccents.Althoughnoconceptuallinkhasyetbeenmadebetweenthisworkandmusicresearch,itisbrieflydescribedbecauseitisapromisingnewareaforempiricalworkonspeechrhythm.

ThePerceptionofIsochronyinSpeechAsnotedinsection3.3.1(subsection“PeriodicityandTypology”),theideathatlinguisticrhythminvolvesregulartemporalintervals(e.g.,betweenstressesorsyllables)hasreceivednoempiricalsupportfrommeasurementsofspeech.However,allsuchmeasurementshavebeenbasedondatafromspeechproduction,inotherwords,onwaveformsorspectrogramsofspokenutterances.Inaninfluentialpaper,Lehiste(1977)madetheinterestingsuggestionthatperiodicitymaybestrongerinperceptionthaninproduction.Thatis,theearmayignoreorcompensateforsurfaceirregularitiesinjudgingperiodicityinspeech.Shebasedthisideaonempiricalworkinwhichsheexaminedtheabilityoflistenerstoidentifythelongestorshortestinterstressinterval(ISI)inshortsentencesoffourISIs,andtodothesametaskonnonspeechanalogsofthesentencesinwhichthestresseswerereplacedbyclicksandthespeechbynoise.Shefoundthatlistenersperformedbetterinthenonlinguisticcondition,andsuggestedthatiflistenershaddifficultyjudgingISIdurationdifferencesinspeech,thiswouldleadtoasensethatISIsweresimilarinduration,inotherwords,animpressionofisochrony.(Ofcourse,itmaynotbespeechpersethatmakessmalldurationdifferencesdifficulttodetect;itcouldbethatthepresenceofsemanticmeaninginlanguagepreoccupiesthelistener’sattentionsothatfinedurationjudgmentsaredifficult.Thusanimportantcontrolconditionforfuturestudiesofthissortistousealanguagewithwhichthelistenerisunfamiliar.)

Lehistewentontostudythejustnoticeabledifference(JND)indurationforsequencesoffournoise-filledintervals,reasoningthatthiswouldestablishaconservativeestimateof

Rhythm

Page 44 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

theJNDsforISIsinspeech.Sheusedthreebasicreferencedurationsinhernoisesequences(300,400,and500ms).Ineachsequence,threeofthefourintervalshadthesameduration,andthefourthwasincreasedordecreasedinnine10-mssteps.Shefoundthatreliablejudgmentsidentifyingoneintervalaslongerorshorterthantheothersrequiredchangesofbetween(p.144) 30and100ms.ShearguedthatJNDsforISIsinspeecharenobetterthanthisandarelikelytobeworse,andthusthatphysicalmeasurementsofisochronyneedtotakethis“perceptualtolerance”intoaccount(cf.Kristofferson,1980).

Lehiste’sworkisinterestingbecauseitraisesthepossibilitythatlistenershearmoreisochronythanisreallythereinspeech.SomeevidenceofferedinfavorofthisargumentcomesfromDonovanandDarwin(1979),whohadindividualslistentoEnglishsentencesandthenimitatethetimingofeachsentence’sstresspatternbytapping.Thesubjectsalsoperformedthistaskwithsequencesofnoiseswhosetimingmimickedthestresspatternofsentences.Thecriticalfindingwasthatwhenimitatingspeech,subjectstappedwithlesstemporalvariabilitythantheactualtimingofstressedsyllables,whereaswhenimitatingnoisetheydidnotshowthispattern.

Althoughthesefindingsareintriguing,furtherworkhassuggestedthatthisparadigmmaybeflawed.Scottetal.(1985)replicatedthefindingsofDonovanandDarwinforEnglish,butalsofoundthatsubjectsshowedregularizationoftappingtoFrench(whichisnotconsideredtohaveperiodicstress),aswellastogarbledspeech.Thissuggeststhattheobservedregularizationmaybeasideconsequenceofthegreaterdifficultyofrememberingacousticallycomplexstimuliversusnoisepatterns.

ThusLehiste’sideasmeritfurtherinvestigation,becausetheyraisetheimportantissueofhowperceivedtimingpatternsrelatetothephysicalintervalsmeasuredinspeech.Nevertheless,thereisnothinginLehiste’sworkthatsupportstheideathatspeechisperceivedasisochronousunderordinarycircumstances.Asnotedinsection3.3.2(subsection“DifferencesBetweenLinguisticandMusicalMetricalGrids”),thevariabilityofISIdurationsinspeechisontheorderof33%.Fora500-msaverageISI,thisis150ms,whichisabovethethresholdforsubjectiveisochronysuggestedbyLehiste.

Lehiste’sideasdopointinoneinterestingdirectionintermsofcomparativestudiesofspeechandmusic,namelyadirectcomparisonofthethresholdfordetectingtemporalirregularitiesinperceptuallyisochronoussequences(e.g.,arepeatingsyllable“tatatata”)versusarepeatingmusicalsoundofequivalentacousticcomplexity.23Ifspeechcantoleratemoredurationalvariabilitythanmusicandstillsoundisochronous,thisraisesinterestingquestionsaboutdifferentmechanismsfortimeperceptioninthetwodomains.

Onecouldalsoexaminethethresholdfordetectingtempochangeinisochronoussequencesinspeechandmusic.Currentdatasuggestthatfornonmusicianlisteners,thethresholdfortempochangedetectioninsequencesofmusicalsoundsis5%–8%(Drake&Botte,1993;cf.Rivenezetal.,2003).Wouldthethresholdbehigherifspeechsoundswereused?

Rhythm

Page 45 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(p.145) TheRoleofRhythmicPredictabilityinSpeechPerceptionAnumberofresearchershavearguedthattheabilitytopredictthelocationofstressedsyllablesinEnglishisperceptuallybeneficial(e.g.,Martin,1972;Shieldsetal.,1974;Cutler&Foss,1977).Thereasoningbehindthisideaisbasedoncertainassumptions:Stressedsyllablescarryimportantsemanticinformation,andalistener’sattentionislimited,sothatitisusefultoexpendattentionalresourcesonthosepointsintimewherestressesoccur.Thustheabilitytoanticipatestresslocationcanhelpguideattentioninanefficientmanner.Thisideasuggestsapointofcontactbetweenrhythmperceptioninspeechandmusic,becausetherearetheorieslinkingrhythmandattentioninmusicpsychology(Jones,1976;Large&Jones,1999;Barnes&Jones,2000).Inordertodetermineifthisisreallyameaningfulparallel,however,twoquestionsmustbeanswered.First,isthereevidencethatrhythmicpredictabilityplaysanimportantroleinspeechperception?Second,arethemechanismsforrhythmicpredictionsimilarinspeechandmusic?

Thebestevidenceforaroleofrhythmicpredictabilityinspeechperceptioncomesfromstudiesusingphoneme-monitoringtasks.Intheseexperiments,listenersaretoldtolistentoonesentenceatatimeandpressabuttonwhentheyhearatargetphoneme(suchas/d/).CutlerandDarwin(1981)conductedastudyinwhichsentenceswererecordedthathadhigh,low,orneutralemphasisonagiventargetword.Forexample,sentences3.14aandbbelowwereusedtorecordhighversuslowemphasisontheword“dirt”(inthesentencesbelow,thewordbearingthemainemphasisofthesentenceisitalicized):

(3.14a)Shemanagedtoremovethedirtfromtherug,butnotthegrassstains.

(3.14b)Shemanagedtoremovethedirtfromtherug,butnotfromtheirclothes.

CutlerandDarwinthensplicedtheneutralversionofthetargetwordintohighandlowemphasissentences,sothatthetargetphoneme/d/(andtherestofthewordthatbeganwiththisphoneme)wereacousticallyidenticalinthetwocases.Afasterreactiontimetothetargetphonemeinhigh-emphasissentenceswouldindicatethatthepredictionofstresswasinfluencingspeechprocessing.Thisispreciselywhatwasfound:Listenerswerereliablyfasterindetectingthetargetphonemeinhigh-stresssentences.Ofparticularinterestisthatthisdifferencepersistedevenwhenfundamentalfrequencyvariationwasremovedfromthetwosentencetypes,suggestingthatpatternsofdurationandamplitudeweresufficienttopredicttheupcomingstress.

CutlerandDarwin’sstudyfocusedontargetwordsthateitherdidordidnotbearthemaincontrastivestressoftheentiresentence.Thatis,theywerenotstudyingtheperceptionofjustanystressedsyllable,butofaparticularlysalientstressedsyllableinasentence.PittandSamuel(1990)conductedastudyinwhichthecontextmanipulationwasnotsoextreme:Theyusedsentencesthat(p.146) predictedstressornonstressatatargetpointduetorhythmicandsyntacticfactors,forexample,thefirstsyllableoftheword“permit”in:

(3.15a)Theguardaskedthevisitorifshehadapermittoenterthebuilding.

Rhythm

Page 46 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(3.15b)Thewaiterdecidedhecouldnotpermitanyoneelseintherestaurant.

Insentence3.15a,thecontextleadsonetoexpectstressedsyllableatthetargetlocation,bothbecausethesyntaxofthesentencepredictsanoun(awordcategorythattendstostartwithastressedsyllableinEnglish)andfortherhythmicreasonthatthepriorstressinthesentenceisquitefaraway(thefirstsyllableof“visitor”).Insentence3.15b,thecontextleadsonenottopredictastressedsyllable,bothbecausethesyntaxpredictsaverb(awordcategorythattendstostartwithaweaksyllableinEnglish),andbecausethepriorstressisquitenearby(on“could”or“not”).

LikeCutlerandDarwin,PittandSamuelusedasplicingtechniquetoensurethatthephysicaltargetwordwasthesameinthetwocontexts,andaskedlistenerstorespondwhentheyheardatargetphoneme(e.g.,/p/intheaboveexample).UnlikeCutlerandDarwin,however,theyfoundnosignificantdifferenceinreactiontimetothetargetphonemeasfunctionoftheprecedingcontext.Thusitappearsthatalthoughrhythmmayhelplistenerspredictsentencelevelemphasis,itdoesnotplayastrongroleinpredictinglexicalstress,evenwhenreinforcedbysyntax.Thiscastssomedoubtontheideathatrhythmplaysanimportantroleinguidingattentiontothemajorityofstressedsyllablesinspokensentences.Clearly,moreworkisneededtodeterminetowhatextentstressispredictableundernormalcircumstances.

Evenifasignificantroleforrhythmicpredictioninspeechisdemonstrated,however,itisquitepossiblethatthemechanismsthatunderlierhythmicpredictioninspeechandmusicarequitedifferent.Inmusic,rhythmicpredictabilityreflectstheperiodicstructureoftemporalintervals.Inspeech,thebasisforrhythmicpredictability(e.g.,predictingwhenastresswilloccur)isunlikelytoinvolveperiodictimeintervals,becausethereisnoevidencethatsuchintervalsexistinnormalspeech.AfirststepinstudyingthebasisofrhythmicpredictioninspeechwouldbetostudythestimuliusedbyCutlerandDarwin(1981),especiallythosestimuliinwhichfundamentalfrequencyvariationwasremoved.Whattemporaland/oramplitudepatternshelpedguidelisteners’expectationsinthatstudy?

Thusatthecurrenttime,thehypothesisthatrhythmicpredictabilityinspeechconfersanadvantagebyguidingattentiontosemanticallyimportantpartsofutterancesisnotwellsupportedbyempiricalevidence.Inmusic,itisclearthatrhythmicpredictabilityhasanadaptivevalue:itallowstheformationofatemporalexpectancyschemethatplaysanimportantroleinmusicalperception(e.g.,beatperception),anditguidesthecoordinationofensembleperformanceandthesynchronizationofmovementsindance.Becausespeechdoesnothavearegularbeat,whatfunctionalrolewouldrhythmicpredictabilityplay?OneideasuggestedbyLehiste(1977)isthatitplaysaroleinsignalingphrase(p.147) boundariesinspeech.Specifically,shesuggestedthatonemethodspeakershavefordisambiguatingsyntacticallyambiguoussentencesisbysignalingastructuralboundaryvialengtheningofaninterstressinterval(ISI).Forexample,shestudiedspeakers’productionsofsentencessuchas“Theoldmenandwomenstayedathome,”whichissyntacticallyambiguous(eitherjustthemenwereoldorboththemenandwomenwereold).Shefoundthatwhenspeakerssaidthissentenceinsuchawaytomake

Rhythm

Page 47 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

oneortheotherinterpretationclear,thesequence“menandwomen”wasverydifferentinduration,beingsubstantiallylongerwhenasyntacticboundarywasintendedbetween“men”and“women.”Furthermore,sheconductedafollow-upstudyinwhichthesamesentenceswereresynthesizedusingamonotone,andthedurationofthecriticalISIwasmanipulatedbyuniformlyexpandingthedurationofthephonemeswithinit,sothattherelativedurationsofsegmentsremainedthesame.ShefoundthatlistenerswereabletoperceivetheintendedmeaningsolelyonthebasisofthelengthofthecriticalISI,suggestingthatISIdurationcansignalaphraseboundary.

AsubsequentstudybyScott(1982)setouttotestLehiste’shypothesisagainstthemoreconventionalnotionthatphraseboundariesaresignaledbyphrase-finallengthening.ShefoundevidenceforaweakversionofLehiste’shypothesis,inthatthereappearedtobesomecasesinwhichlistenersusedISIpatterns,butothersinwhichtheyreliedontraditionalphrase-finallengthening.Nevertheless,theevidencewassuggestiveenoughforthislineofresearchtomeritfurtherstudy.Akeyconceptualpoint,however,isthatevidencethatISIdurationplaysaroleincreatingperceivedboundariesinspeechisnotequivalenttoevidenceforisochrony.TheexpectationforhowlonganISIshouldbeneednotbebasedonanexpectationforisochrony,butcouldbebasedonexpectationsforhowlongagivenISIshouldbegiventhenumber(andtype)ofsyllableswithinitandthecurrentspeechrate(cf.Campbell,1993).Accordingtothisview,aprosodicbreakismorelikelytobeheardwhenanISIissignificantlylongerthanexpected,andrhythmicpredictabilityissimplyimplicitknowledgeofthestatisticalrelationbetweenISIdurationandthenumberandtypeofsyllablesinanISI.Thiswouldallowafunctionalroleforrhythmicpredictabilitywithoutanyrecoursetonotionsofisochrony.

TheRoleofRhythminSegmentingConnectedSpeechToanativelistener,spokensentencesconsistofasuccessionofdiscretewords,yetthisperceptionisanillusion.AspointedoutinChapter2,wordboundariesinlanguagedonotmapinanysimplewayontoacousticbreaksinthespeechsignal,andasanyonewhohaslistenedtosentencesinaforeignlanguagecanattest,itisfarfromobviouswherethewordboundariesinconnectedspeechare.Thisproblemisparticularlyrelevantforinfants,whoareconstantlyfacedwithmultiwordutterances(vandeWeijer,1999)andwhodonothavethe(p.148) benefitofanexistingvocabularytohelpthemidentifywhereonewordendsandthenextbegins.

Asubstantialbodyofresearchinpsycholinguisticsindicatesthattherhythmicpropertiesofalanguageassistalistenerinsegmentingspeech.WorkonEnglish,forexample,haspointedtoasegmentationstrategybasedonstress:Listenersexpectstrongsyllablestobeword-initial.ThislikelyreflectsthepredominanceofwordswithinitialstressintheEnglishlexicon(Cutler&Carter,1987),andmanifestsitselfinanumberofdifferentwaysinperception.Forexample,CutlerandButterfield(1992)showedthatwhenlistenersmissegmentspeech,theytendtoplacewordboundariesbeforestressedsyllables,aswhen“bylooseanalogy”ismisheardas“byLuceandAllergy.”Furthermore,whenEnglishspeakersareaskedtospotrealmonosyllabicwordsembeddedinlargerpolysyllabicnonsensewords,theyfinditeasierwhentherealworddoesnotstraddle

Rhythm

Page 48 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

twostressedsyllables.“Mint,”forexample,iseasiertospotin“mintef”thanin“mintayf,”presumablybecauseinthelatterwordthestrongsecondsyllable“tayf”triggerssegmentation,thussplitting“mint”intotwoparts(Cutler&Norris,1988).Cutler(1990)hasdubbedthestrategyofpositingawordonsetateachstrongsyllablethe“metricalsegmentationstrategy.”

Researchonsegmentationinotherlanguageshasrevealedthatstress-basedsegmentationisbynomeansuniversal.FrenchandSpanishspeakers,forexample,favorsyllabicallybasedsegmentation(Mehleretal.,1981;Pallieretal.,1993),whereasJapanesespeakersfavormoraicsegmentation(Otakeetal.,1993).Thussegmentationreliesonunitsthatarephonologicallyimportantinthenativelanguage.Onestrikingfindingofthiscross-linguisticresearchisthatthenativelanguage’ssegmentationstrategiesareappliedevenwhenlisteningtoaforeignlanguage,showingthatsegmentationtendenciesarenotsimplyareactiontoaparticularspeechrhythm,butaperceptualhabitofalistener(seeCutler,2000,forareview).OnepossibilitysuggestedbyCutleristhatthishabitisaresidueofearlylanguagelearning,whenrhythmicsegmentationplayedanimportantroleinbootstrappinglexicalacquisition.

Therelevanceofthisresearchtocomparativestudiesoflanguageandmusicisthatitshowsthatexperiencewithalanguage’srhythmleavesapermanentinfluenceonalistenerintermsofsegmentingspeechpatterns,whetherornotthesepatternscomefromthenativelanguage.Fromthisobservationitisbutonesteptoaskifexperiencewiththenativelanguageinfluenceshowonesegmentsnon-linguisticrhythmicpatterns.Thisquestionistakenupinsection3.5.2below.

TheRoleofRhythminthePerceptionofNonnativeAccentsWhenapersonlistenstotheirnativelanguage,s/heusuallyhasakeensenseofwhetherornotitisbeingspokenwithanativeaccent.Recentresearchonspeechrhythmhastakenadvantageofthisfactbyhavinglistenersjudgethedegreeofperceived“foreignaccentedness”inutterancesspokenbynonnativespeakers.(p.149) Empiricalrhythmicmeasurementsarethentakenofspeechofthedifferentnon-nativespeakers.Byexaminingthecorrelationbetweenperceiveddegreeofforeignaccentandthequantitativerhythmicmeasures,researchershopetoidentifytheperceptualcueslistenersuseingaugingspeechrhythmpatterns.

Usingthisapproach,WhiteandMattys(2007)examinedSpanishspeakersofEnglishandfoundthatthegreatertheirvoweldurationvariabilitywithinsentences,themorenative-soundingtheywereratedbyEnglishspeakers.Thisprobablyreflectsvowelreduction:Spanishspeakerswholearntoreducevowelsinunstressedsyllables(acharacteristicofEnglish,butnotofSpanish;cf.section3.3.1,subsection“PhonologyandTypology”)arelikelytosoundmorelikenativespeakers.Aconsequenceofvowelreductionwithinsentencesisthatvoweldurationvariabilityincreases,becausesomevowelsbecomeveryshort.

Asnotedinsection3.3.1(subsection“DurationandTypology”),anotherempiricalmeasureofrhythminfluencedbyvowelreductionisthenPVI,whichmeasuresthe

Rhythm

Page 49 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

degreeofdurationalcontrastbetweenadjacentvowelsinasentenceratherthanoveralldurationalvariability.WhiteandMattysfoundthatthevowelnPVIofSpanishspeakersofEnglishwaspositivelycorrelatedwithhownativetheysounded.Crucially,however,voweldurationvariabilitywasabetterpredictorofaccentjudgmentthanwasnPVI.Thissuggeststhatvoweldurationvariabilitymaybemoreperceptuallyrelevantforspeechrhythmthandurationalcontrastiveness.

Thisisaverypromisingapproach,becausedifferentrhythmicmeasurescanbepittedagainsteachothertoseewhichbestpredictsperceptualdata.However,thefindingstodatemustbeconsideredtentativebecauseofanuncontrolledvariable.Thisisvariabilityinthedegreetowhichnonnativespeakersaccuratelyproducethephonemesofthesecondlanguage(i.e.,theindividualvowelsandconsonants).Whenjudgingaspeaker’sdegreeofforeignaccent,listenersalmostcertainlybasetheirjudgmentsonsomecombinationofsegmentalandsuprasegmentalcues.Thisisaproblembecausesomenonnativespeakersmayproducenative-soundingprosodybutnonnativesoundingsegmentalmaterial,orviceversa.Tocompoundtheproblem,differentlistenersmayvaryintheextenttowhichtheyweightsegmentalversussuprasegmentalcuesinjudginghow“foreign”agivennonnativespeakersounds.Thustotrulyfocuslisteners’attentiononrhythm,segmentalcuesmustbemadeuniform.Resynthesistechniques,suchasthoseusedbyRamusandcolleagues(cf.section3.3.1,subsection“PerceptionandTypology”)mightprovideonewaytomakesentencesspokenbydifferentnonnativespeakersuniformintermsofphonemicmaterialwhilepreservingprosodicdifferences.

3.3.4FinalCommentsonSpeechRhythm:MovingbeyondIsochrony

Althoughthehistoryofspeechrhythmresearchistightlyboundupwithnotionsofperiodicity(e.g.,theisochronyofstressesorsyllables),theevidence(p.150) reviewedabovesuggeststhatthecaseforperiodicityinspeechisextremelyweak.Thusprogressinthestudyofspeechrhythmrequiresconceptuallydecoupling“rhythm”and“periodicity,”apointmadeintheintroductionofthischapter.Itisquiteclearthatspeechhasrhythminthesenseofsystematictemporal,accentual,andgroupingpatternsofsound,andlanguagescanbesimilarordifferentintermsofthesepatterns.However,therhythmsoflanguagearenotbasedontheperiodicoccurrenceofanylinguisticunit.Instead,thepatterningislargelytheby-productofphonologicalphenomena,suchasthestructureofsyllables,vowelreduction,thelocationoflexicalprominence,stressclashavoidance,andtheprosodicphrasingofsentences.Thesephenomenaleadtodifferencesinthewayutterancesareorganizedintime.

Thenotionthatrhythminlanguageisprimarilyconsequenceratherthanconstructstandsinsharpcontrasttorhythminmusic,inwhichpatternsoftimingandaccentareafocusofconsciousdesign.Anothersalientdifferencebetweenrhythminspeechandmusic,relatedtothelackofaperiodicframeworkforspeechrhythm,isthefactthatspeechrhythmconveysnosenseofmotiontoalistener(cf.section3.2.5).Dothesedifferencesmeanthatrhythminlanguageandmusiccannotbemeaningfullycompared?Absolutelynot.Asdemonstratedinsection3.5below,empiricalcomparisonsarenotonlypossible,theycanbequitefruitful.Theyhavehadnothingtodowithperiodicity,however.

Rhythm

Page 50 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Forthoseinterestedincross-domainstudiesofrhythm,itishearteningtonotethatthereisrenewedinterestinempiricalstudiesofrhythminspeechproductionandspeechperception(Ramusetal.,1999;Ramus&Mehler,1999;Lowetal.,2000;Grabe&Low,2002;Lee&Todd,2004,White&Mattys,2007),andthatthereismuchroomforfurtherwork.Forexample,thereisaneedformoreempiricaldataonlisteners’judgmentsofhownative-soundingaforeignspeakers’utterancesare,fromthestandpointofrhythm.Suchstudieswillneedtoemploycreativewaysofisolatingtherhythmofspeechfromotherphoneticdimensionsoflanguage,forexample,usingresynthesizedspeechinwhichphoneticcontentandpitchcontourscanbecompletelycontrolled(cf.Ramus&Mehler,1999).Thereisalsoaneedforstudiesthatmeasuretemporalpatterningatmultiplelinguisticlevelsandthatquantifyrelationsbetweenlevels.Itmaybethatimportantperceptualdimensionsofspeechrhythmarerelational,suchashavingahighdegreeofcontrastbetweenadjacentsyllabledurationswhilesimultaneouslyhavingalowdegreeofcontrastbetweenthedurationofinterstressintervals(somedatapertinenttothisideaaregivenattheendofthissection).Thisisanareainwhichcollaborationsbetweenlinguistsandmusicresearcherswouldbeespeciallyuseful.

Intheremainderofthissection,Iwouldliketoconsiderwhyperiodicityhasbeen(andcontinuestobe)suchanenduringconceptinspeechrhythmresearch.BelowIofferseveralreasonsforthishistoricalphenomenon.

Thesimplestreason,ofcourse,isthemistakennotionthatrhythmisperiodicity,orthatrhythmisaregularalternationbetweenstrongandweakbeats,(p.151) ratherthanthebroadernotionofrhythmassystematictemporal,accentual,andphrasalpatterningofsound,whetherornotthispatterningisperiodic.Indeed,oneneednotlookbeyondmusictoseethatadefinitionofrhythmasperiodicityorasstrong-weakbeatalternationisoverlysimplistic:Manywidespreadmusicalformslackoneand/ortheotherofthesefeaturesyetarerhythmicallyorganized(cf.section3.2).

Thesecondreasonthatthenotionofperiodicityhasenduredmaybetheideathatithasausefulfunctioninspeechperception,suchasmakingsalientinformationpredictableintime.Therearepsychologicaltheoriesofauditoryperceptionthatproposethatattentioncanbeallocatedmoreefficientlywheneventsaretemporallypredictable,basedontheideathatauditoryattentionemploysinternaloscillatoryprocessesthatsynchronizewithexternalrhythmicpatterns(e.g.,Jones,1976;Large&Jones,1999).Suchtheoriesprovidearationaleforthoseinterestedintheideathatperiodicityinspeechisperceptuallyadaptive.Alternatively,thoseinterestedinperiodicitymightclaimthatitisusefulbecauseitcreatesaframeworkwithinwhichdeviationsaremeaningful.ThisisthebasisofLehiste’sideathatlengtheningofinterstressintervalsinEnglishcanbeusedtomarkphraseboundaries(cf.section3.3.3,subsection“TheRoleofRhythmicPredictabilityinSpeechPerception”).Theprincipaldrawbackoftheseperception-basedargumentsforperiodicityisthattheevidenceforthemisveryweak.Althoughfurtherresearchisneeded,thecurrentevidencesuggeststhatperiodicitydoesnothaveanimportantroletoplayinnormalspeechperception.Thisshouldnotbesurprising:The

Rhythm

Page 51 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

comprehensionofspeechshouldberobusttovariationinthetimingofsalientevents,becausesuchvariationscanoccurforanumberofreasons.Forexample,aspeakermaysuddenlyspeeduporslowdownforrhetoricalreasonswithinaconversation.Undersuchconditions,relyingonperiodicityforcomprehensionseemsamaladaptivestrategy.

Thethirdreasonforperiodicity’salluremaybethebeliefthatbecausevarioustemporalpatternsinhumanphysiology(e.g.,heartbeat,walking,chewing)exhibitperiodicstructure,speechisalsolikelytobeperiodic,perhapsevengovernedbyrhythmicpatterngenerators.However,theuseofrhythmicneuralcircuitsforspeechisnotparticularlyplausible.Theconstantuseofnovelutterancesinlanguagemeansthatarticulatorsmustbecoordinatedindifferentwayseachtimeanewsentenceisproduced.Furthermore,themaneuversthatproduceparticularspeechsoundsdependonthelocalcontextinwhichtheyoccur.Thusthemotorpatternsofspeechcannotbepredictedinadvancewithahighdegreeofprecision.Withoutstereotypedmovementpatterns,evolutionhasnogroundsforplacingthecontrolofspeechinarhythmicneuralcircuit.Ananalogyhereistomultifingeredtouch-typing,abehaviorinvolvingthesequencingofoverlappingmovementsofmultiplearticulators(thefingers).Althoughtouch-typingishighlytemporallyorganized,theresultingsequencesarenotbasedonperiodicmovements.

(p.152) SofarIhavefocusedonnegativereasonsforthepersistenceoftheconceptofperiodicityinspeech.Iwillnowbrieflyspeculateonthepositivereasonsforthepersistenceofthisconcept,inotherwords,whyperiodicityinspeechhasbeensuchanintuitivelyappealingnotiontospeechresearchers,particularlythosewhosenativelanguageisEnglish.(ItisnotablethattheideaofperiodicityinspeechwaspromulgatedbylinguistswhowereEnglishspeakers,andthatargumentsforstressisochronyinEnglishhavebeenpresentsinceatleastthe18thcentury;cf.Abercrombie,1967:171;Kassler,2005).First,itseemsthatEnglishspeakersfindtheinterstressinterval(ISI)tobeasalienttemporalunitinspeech.Forexample,inapreliminarystudy,Cummins(2002)askedEnglishlistenerstorepeatnonsensephrasessuchas“manningthemiddle”intimewithanexternalpacingcue,sothatthetwostressedsyllableswereperceptuallyalignedwithaperiodicallyrepeatingtwo-tonepattern(forexample,“man”wouldalignwithahightone,and“mid”withalowtone;cf.Cummins&Port,1998).ThisisequivalenttoaligningthestartandendoftheISIwithtwotones.CumminsalsotestedspeakersofItalianandSpanishbecausetheselanguageshavelexicalstress,permittingphrasestobeconstructedinamanneranalogoustotheEnglishphrases(suchas“BUScaalMOto”inSpanish,stressindicatedbycapitalization).CumminsobservedthatalthoughEnglishspeakerslearnedthetaskquicklyandperformedaccurately,SpanishandItalianspeakerstookmuchlonger,wereuncomfortablewiththetask,andproducedagreatdealofvariabilityintheirresults.CumminssuggeststhatthisdifferenceisduetothefactthattheISIisnotasalientperceptualunitforspeakersofItalianandSpanish,despitethefactthatthereislexicalstressintheselanguages.

ThisintriguingfindingraisesthequestionofwhyISIissalienttoEnglishlisteners.Doesitplaysomefunctionallinguisticrole?Asdiscussedinsection3.3.3(subsection“TheRoleof

Rhythm

Page 52 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

RhythmicPredictabilityinSpeechPerception”),ISIdurationmayplayaroleinsignalinglinguisticboundariestoEnglishlisteners,evenifISIsarenotisochronous.CurrentlythereisnotenoughevidencetosayconfidentlywhatroletheISIplays,butletusassumeforamomentthatEnglishspeakersandlistenersaresensitivetoitasanentity.GiventheempiricalobservationsaboutthelargevariabilityinISIdurationinEnglish(e.g.,coefficientsofvariationaround33%;Dauer,1983),whywouldlistenerseverfeelthatISIswereisochronous?OneanswermayconcerntherelativedegreeofvariabilityinISIdurationscomparedtosyllabledurations.Assuggestedabove,theimpressionofisochronymaybedueinparttoalowerdegreeofcontrastbetweensuccessiveISIdurationsrelativetosuccessivesyllabledurations.Forexample,considerFigure3.12,whichshowsthesamesentenceasFigure3.8aandSoundExample3.5a(“thelastconcertgivenattheoperawasatremendoussuccess”).

Syllableboundariesaremarkedwithverticallines(asinFigure3.8),butnowstressedsyllables(indicatedinbyboldfaceabove)havebeenmarkedwithanasterisk.Theasteriskwasplacedatthevowelonsetofthestressedsyllable,inotherwords,nearitsperceptualattack(its“P-center”;Mortonetal.,1976;Pateletal.,1999).Inthissentence,thenPVIofsyllabledurationsis59.3,andthenPVIofISIsis28.4,makingtherationPVIISI/nPVIsyllequalto0.48.Thusinthisparticularcase,theamountofdurationalcontrastbetweenadjacentISIsisonlyabouthalfofthatbetweenadjacentsyllables.

(p.153)

Figure3.12 TheEnglishsentenceofFigure3.8a,withstressesmarkedbyasterisks(*).

Ihavecomputedsimilarratiosforeachofthe20sentencesofBritishEnglishfromRamus’sdatabase.24For15outof20sentences,thisratiowaslessthan1.Theoverallmeanratioacrossthe20sentenceswas.83(std=.45),andwassignificantlylessthan1(p〈.0001byaone-tailedt-test).AnevenstrongereffectwasobservedwhenonecomputestheratioofISIdurationvariabilitytosyllabledurationvariability(usingthecoefficientofvariation,i.e.,CVISI/CVsyll)Here17outof20sentenceshadavaluelessthan1,andthemeanwas.69(std=.25),againsignificantlylessthan1.Thusiftheearissensitiveto

Rhythm

Page 53 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

temporalpatterningatthelevelsofbothsyllablesandstresses,thelowdurationalvariabilityofISIsrelativetothevariabilityofsyllabledurationsmightcontributetoasensethatstressesaretemporallyregular.Ofcourse,forthisexplanationtohaveanymerititmustbeshownthattheratioofISItosyllablevariabilitydifferentiatesstress-timedfromsyllable-timed(p.154) languages.LanguagessuchasItalianandSpanishwouldbegoodcandidatesfortestingthishypothesis,becausetheyaresyllable-timedlanguagesinwhichstresscanbereliablymarked.

Althoughthenotionofisochronyinspeechcontinuestobeguileresearchers,Isuspectthatitwillhavelittleornoroleinthemostfruitfulresearchonspeechrhythminthecomingyears.Isochronywasimportanttothebirthofspeechrhythmstudies,butitisaconceptwhoseusefulnessisexhausted.Itistimetomoveontoaricherviewofspeechrhythm.

3.4Interlude:RhythminPoetryandSongAsintherestofthisbook,thefocusofthischapterisoncomparingordinaryspeechtoinstrumentalmusic.However,nocomparisonofrhythminlanguageandmusiciscompletewithoutadiscussionofpoetryandsong.Intheseartforms,wordsarecarefullychosenandconsciouslypatternedforrhythmiceffect.Ofcourse,poetryandsongarebuttwoofnumerousvocalgenreswithorganizedrhythms.IntheUnitedStatescertainstylesofpreachinginAfricanAmericanchurchesarenotablefortheirrhythmicpatterning,atasteofwhichcanbeheardinMartinLutherKingJr.’sfamous“IHaveaDream”speech.Inothercultures,itispossibletoidentifynumerousgenresofspeechinwhichrhythmicdesignplaysarole(seeAgawu,1995,forafascinatingcasestudyoftherangeofrhythmicallyregulatedformsofspeechinanAfricansociety).Thefocushereisonpoetryandsong,however,becausethesehavereceivedthegreatestamountofempiricalresearchintermsofrhythm.

3.4.1RhythminPoetry

Thestudyofpoeticrhythmhasbeenthefocusofagooddealofresearchbyliteraryscholars(forintroductions,seeGross,1979;Fussell,1979;Hollander,2001).Inthistradition,poetic“meter”referstotheabstractpatterningschemewhichgovernsthetemporalstructureofapoem,whereas“rhythm”referstotheactualpatterningofdurationsandaccents.Forexample,agreatdealofEnglishverseiswritteniniambicpentameter,averseformconsistingof5iambicfeet,inwhichaniambisa(weak+strong)syllablepattern.Naturallytherearemanyexceptionstothispatternwithiniambicpentameterpoetry:Aparticularlycommononeisthesubstitutionofatrochaicfoot,or(strong+weak)syllablepatternattheonsetofaline.Thustherhythmofaparticularlinemayviolatetheoverallmeterofthepoem.

Literaryprosodistsarguethatlistenersinternalizetheregularitiesofmeterandperceivedeparturesfromthisschemeasvariationfromastablebackground(Richards,1979:69;Adams,1997:12).Thatis,meterisseenashaving(p.155) anintimaterelationshipwithexpectancy.Thisideaisrelatedtothenotionofmusicalmeterasanabstractmentalscheme,butdiffersfrommusicalmeterinanimportantway.Musicalmeterrefersto

Rhythm

Page 54 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

temporalperiodicity,whereaspoeticmeterinvolvesconfigurationalperiodicity,inotherwords,thefocusisontherepetitionofsomebasicprosodicunitratherthanontemporalperiodicityperse.Forexample,iniambicpentameteritistheweak+strongconfigurationoftheiambicfootthatisthedesignfocus,nottheisochronyofstressedsyllables.InvariousformsofFrenchandChineseverse,thenumberofsyllablesperlineisstrictlyregulated,butthereisnofocusonmakingsyllablesperiodic(equalinduration).

Itisinterestingtonotethatdifferentlanguagestendtofavordifferentkindsofpoeticmeters.Forexample,Englishversehasoftentendedtowardpurelystress-basedformsinwhichregulationofthenumberofstressesperlineisthefocus,independentofthenumberofsyllables(e.g.,themeterofBeowulf,withfourstressesperline).Incontrast,Englishversebasedonregulatingthenumberofsyllablesperlinewithoutregardtostressisrare(Fussell,1979:62–75).ThislikelyreflectsthepowerfulroleofstressinordinaryEnglishspeechrhythm.Indeed,Fussellhasarguedthat“ametercustomaryinagivenlanguageiscustomaryjustbecauseit‘measures’themostcharacteristicqualityofthelanguage”(Fussell,1974:498).ThusstressplaysadominantroleinEnglishpoetry,butlittleroleinFrench,inwhichthenumberofsyllablesperlineisamorecommonconcern.Japanese,inturn,oftenregulatesthenumberofmoraeperline,asinthe5–7–5morastructureofthehaiku.

LerdahlandHalle(1991)andLerdahl(2003)havesoughttounifythetheoreticaltreatmentofrhythminpoetryandmusic,usingsharedconceptssuchashierarchicalgroupingstructureandmetricalgrids.Ourfocushere,however,isonempiricalresearch.Overthepastfewdecades,poeticrhythmhasattractedtheinterestofanumberofspeechscientists,whohavemadequantitativemeasurementsofthetemporalpatternsofpoetry.Forexample,thephoneticianGunnarFantandcolleagueshavestudiedtheacousticsofiambicversustrochaiclinesofverseinSwedish(1991b).Theresearchersfoundthatiniambicfeet,theweaksyllableisabout50%aslongasthefollowingstrongsyllable,whereasintrochaicfeet,theweaksyllableisabout80%ofthedurationoftheprecedingstrongsyllable(cf.Nordetal.,1990).Thisdifferenceislikelyduetopreboundarylengthening,whichactstoincreasethedurationofthefinalsyllableineachfoot(i.e.,thestrongsyllableinaniambandtheweaksyllableinatrochee).Thusiambicandtrochaicfeetarenotsimplymirrorimagesofeachotherintermsoftheirtemporalprofiles:Iambicfeetaremuchmoretemporallyasymmetric.

Theseobservationsmayberelevanttothestudyoftheaestheticeffectofthetwokindsoffeetinpoeticlines.Forexample,Adams(1997:55–57)notesthattrochaicmetersareoftenassociatedwithaweandthesuspensionofreality,as(p.156) inBlake’spoem,“TheTyger,”inwhichtrochaicpatternsdominatethefirstthreelinesofthefirststanza:

(3.16)

Tyger!Tyger!burningbright

Intheforestsofthenight

Rhythm

Page 55 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Whatimmortalhandoreye

Couldframethyfearfulsymmetry?

Thisaestheticpropertyoftrochaicmetermaybepartlyduetoitsmoreuniformprofileofsyllabicdurations,whichgoesagainstthegrainofnormalEnglishspeechrhythmandthusgivestheresultingspeechanincantatoryfeel.

AnotherprominentphoneticianwhohaslongconductedresearchonpoeticrhythmisIlseLehiste(1991).Inonesetofstudies,Lehisteexaminedtherelationshipbetweenthetimingoffeetandofthelinesinwhichfeetareembedded.Shefoundthatthetemporalvariabilityoflinesislowerthanonewouldpredictbasedonthevariabilityoffeetduration,suggestingthatspeakersmaketemporalcompensationsbetweenfeetinordertokeeplineswithinacertainduration.Thatis,linesactasaunitoftemporalprogrammingintherecitationofpoetry(Lehiste,1990).RossandLehiste(1998,2001)havealsoexaminedtheinterplayoflinguisticandpoeticrhythminframingthetemporalpatternsofEstonianverseandfolksongs.

3.4.2RhythminSong

Forlanguageswithclearlydefinedstress,suchasEnglish,eachphraseorsentencecomeswithadistinctpatternofstrongerandweakersyllables.Whenwordsintheselanguagesaresettometricalmusic,arelationshipisestablishedbetweenthesyllabicaccentpatternsandmusicalmetricalaccentpatterns.Sensitivitytotheserelationshipsispartoftheskillofwritingmusicwithwords,andempiricalresearchsuggeststhatcomposersexploitthisrelationshipforartisticends.

PalmerandKelly(1992)studiedvocallinesinthemesfromGilbertandSullivan’s14operettas,focusingoncompoundnouns(likethesingleword“blackbird”)andadjective-nounpairs(likethetwowordphrase“blackbird”).InEnglish,compoundnounsreceivestressonthefirstsyllable,whereasadjective-nounpairsreceivestressonthesecondsyllable.Theystudiedhowsuchwordswerealignedwiththemetricalstructureofthemusic,andfoundthatthestressedsyllabletendedtoalignwithametricallystrongbeatinthemusic.GiventhecomplextextsofGilbertandSullivan’ssongs,thisstrategyofalignmentmaycontributeasenseofprecisionandbalancetothelyricsoftheseoperettas.

Temperley(1999)lookedatadifferentgenreofvocalmusic,namelyrocksongs.IncontrasttoPalmerandKelly,hefoundthatverbalstressfrequentlyanticipatedmetricalaccentinrocksongsbyafractionofabeat,asinthe(p.157) Beatles’“HereComestheSun”(Figure3.13).Thissystematicanticipationcontributesasenseofsyncopationandrhythmicenergytothesong,andprovidesanexampleofhowthesystematicmisalignmentofverbalandmusicalstressaddsdynamicenergytomusic.

Therelationshipbetweenrhythminspeechandsongisafertileareathatmeritsmuchmoreempiricalinvestigationthanithasreceivedtodate.Intheremainderofthissection,Ioutlinethreedirectionsthatthisresearchcouldtake.Thefirstpertainstocultural

Rhythm

Page 56 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

differencesintheprevalenceofcertaintypesofmusicalrhythms.Forexample,Yamomoto(1996)notesthatchildren’ssongsbasedontriplerhythms(e.g.,6/8timesignature)arerareinJapanbutcommoninBritain,andsuggeststhatthismightbeduetodifferencesinEnglishversusJapanesespeechrhythm.IfYamomotoiscorrect,onewouldpredictthatJapanese-versusEnglish-speakingchildrenwoulddifferinhoweasilytheycanlearnsongsinthesemeters.AnotherlanguagewithwhichonecouldtestasimilarideaisGreek.Recallfromsection3.3.2(subsection“QuestioningthePrincipleofRhythmicAlternationinSpeech”)thatArvaniti(1994)showedthatGreek,alanguageoftheBalkanregion,toleratesamoreirregularalternationbetweenstressedandunstressedsyllablesthandoesEnglish.Alsorecallfromsection3.2thattheBalkanregionfeaturesmusicwithirregularlyspacedbeats.WouldGreek-speakingchildrenfinditeasiertolearntheirregularmetersofBalkansongsthanEnglish-speakingchildren(cf.Hannon&Trehub,2005)?Inthestudiesoutlinedabove,itwouldofcoursebeessentialthatthetwogroupsofchildrenbematchedforpriormusicalexposuretodifferentmusicalmeters.ItmaythusbebesttoworkwithimmigrantswhospeakthenativelanguageathomebutwhosechildrenareexposedtoWesternmusic.Insuchacase,iflearningexperimentsrevealthepredictedculturaldifferences,thiswouldsupporttheinterestinghypothesisthataculture’sspeechrhythmpredisposesittowardorawayfromcertainmusicalrhythms.

Aseconddirectionforresearchinthisareaistoexamineverballyimprovisedmusicthatisaccompaniedbyarhythmicmusicalcontext,suchascontemporaryrapmusic.Ifthevocalandmusicallinescanberecordedondifferentaudiotracks,andpointsofverbalandmusicalstresscanbeindependentlyidentified,thenonecouldstudytemporalrelationsbetweenverbalandmusicalaccentpointsasapieceunfoldsintime.Itwouldbeparticularlyinterestingtostudytheserelationsinnoviceversusexpertrapmusicians,toseeifpartofbeinganexpertinthisgenreisgreaterflexibilityand/orprecisioninthemannerinwhichthealignmentofthetwotypesofaccentsishandled.Inastudysuchasthis,identifyingprecisepointsintimeforverbalandmusicalaccentwillbeessential,andtheissueoftheperceptualattacktimeofsyllablesandofmusicaltones(“P-centers”)comestothefore(Mortonetal.,1976;Gordon,1987;Pateletal.,1999).

Figure3.13 AmusicalmetricalgridforaportionoftheBeatles’“HereComestheSun.”Thelyricsarealignedbelowthegrid,andlinguisticallystressedsyllablesareindicatedinboldface:Notehowmostsuchsyllablesslightlyprecedestrongmetricalpositionsinthemusic.FromTemperley,1999.

(p.158) AfinalpossiblelineofresearchissuggestedbyacorrespondencebetweenRichardStraussandRomainRollandin1905aboutmusicaltextsetting(Myers,1968).25Straussemphasizestheclear-cutrelationshipbetweensyllabicaccentinspeechandmetricalaccentinmusic:“InGerman‘she’onthestrongbeatofabarisabsolutely

Rhythm

Page 57 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

impossible.Forexample,inabarof4/4,thefirstandthirdbeatalwayshaveanecessarystresswhichcanonlybemadeontheradical[stressed]syllableofeachword.”HealsoexpresseshisfrustrationovervariabilityinthealignmentofwordstressandmusicalstressinFrenchopera:“Yesterday,IagainreadsomeofDebussy’sPélleasetMélisande,andIamoncemoreveryuncertainabouttheprincipleofthedeclamationofFrenchwhensung.Thusonpage113,Ifound:‘Cheveúx,chéveux,décheveux.’Forheaven’ssake,Iaskyou,ofthesethreewaystherecanallthesameonlybeonewhichisright.”

RollandrepliesbyemphasizingthemutabilityandsubtletyofFrenchwordaccent:

Thenaturalvalueof“cheveux”ischevéux.Butamaninlovewill,whensayingthisword,putquiteaspecialstressonit:“teschéveux.”…Yousee,thegreatdifficultywithourlanguageisthatforaverylargenumberofwords,accentuationisvariable,—neverarbitrary,butinaccordancewithlogicalorpsychologicalreasons.Whenyousaytome:…”Ofthese3(cheveux)onlyonecanberight,whatyousayisdoubtlesstrueofGerman,butnotforFrench.

WhatthiscorrespondencesuggestsisthatGermantextsettingismorerigidinitsalignmentofverbalandmusicalaccent,whereasFrenchismorepermissiveintermsofaccentalignmentbetweentextandmusic.Indeed,DellandHalle(inpress;cf.Dell,1989)reportthatFrenchtext-settingisquitetolerantofmismatchesbetweenverbalandmusicalaccent,exceptattheendsoflines,wherealignmenttendstobeenforced.Theycontrastthishighdegreeof“mismatchtolerance”inFrenchsongstoamuchlowerdegreefoundinEnglishsongs.ThecorrespondencebetweenStraussandRolland,andtheworkofDellandHalle,suggestthatlanguageshavesalientdifferencesinthewaytheyalignmusicandtextintermsofrhythmicproperties,thoughquantitativeworkis(p.159)neededtoconfirmthis.Thesewritingsalsoleadtoideasforcross-culturalperceptualstudiestestingsensitivitytoaccentmismatchesinsongs.Specifically,listenerscouldbepresentedwithdifferentversionsofasongthathavetextandtunealignedindifferentways,withoneversionhavingmanymoreaccentmismatches.Listenerscouldthenbeaskedtojudgeinwhichversionthewordsandmusicgobesttogether.OnemightpredictthatGermanlistenersjudgingGermansongs(orEnglishlistenersjudgingEnglishsongs)wouldbemoreselectiveintermsofthepairingsthatsoundacceptablethanFrenchlistenersjudgingFrenchsongs.

3.5NonperiodicAspectsofRhythmasaKeyLinkOnemajorthemeofthischapteristhatlanguageshaverhythm(systematictemporal,accentual,andgroupingpatterns),butthatthisrhythmdoesnotinvolvetheperiodicrecurrenceofstresses,syllables,oranyotherlinguisticunit.Initiallyitmayseemthat“givinguponperiodicityinspeech”wouldmeanthatthereislittlebasisforcomparingrhythminmusicandlanguage.Infact,theoppositeistrue.Byabandoningafixationonperiodicityoneisfreedtothinkmorebroadlyaboutspeechrhythmanditsrelationshiptomusicalrhythm.Asweshallseebelow,afocusonnonperiodicaspectsoflinguisticrhythmisprovingfruitfulintermsofcomparinglanguageandmusicatstructuralandneurallevels.

Rhythm

Page 58 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

3.5.1RelationsbetweenMusicalStructureandLinguisticRhythm

Thenotionthatanation’sinstrumentalmusicreflectstheprosodyofitslanguagehaslongintriguedmusicscholars,especiallythoseinterestedin“nationalcharacter”inmusic.GeraldAbrahamexploredthisideaatlength(1974,Ch.4),notingasoneexampleanobservationofRalphKirkpatrickonFrenchkeyboard-music:“BothCouperinandRameau,likeFauréandDebussy,arethoroughlyconditionedbythenuancesandinflectionsofspokenFrench.OnnoWesternmusichastheinfluenceoflanguagebeenstronger”(p.83).Inamoresuccinctexpressionofasimilarsentiment,Glinka(inTheaterArts,June1958)wrote:“Anationcreatesmusic,thecomposeronlyarrangesit”(citedinGiddings,1984:91).

Untilveryrecently,evidenceforthisideahasbeenlargelyanecdotal.Forexample,Garfias(1987)hasnotedthatinHungarianeachwordstartswithastressedsyllable,andthatHungarianmusicalmelodiestypicallystartonstrongbeats(i.e.,anacrusis,orupbeat,israre).Althoughthisisaninterestingobservation,itispossiblethatthisisduetothefactthatmanysuchmelodiescomefromfolksongs.Inthiscase,thelinguisticinfluenceonmusicalrhythmwouldbemediated(p.160) bytext.Amoreinterestingissue,impliedbyKirkpatrick,iswhetherlinguisticrhythminfluencestherhythmofinstrumentalmusic,inotherwords,musicthatisnotvocallyconceived.

OneapproachtothisquestionwassuggestedbyWenk(1987).Heproposedthatcultureswithrhythmicallydistinctlanguagesshouldbeexaminedtoseeifdifferencesinmusicalrhythmreflectdifferencesinspeechrhythm.WenkfocusedonEnglishandFrench,prototypicalexamplesofastress-timedversusasyllable-timedlanguage.WenkandWioland(1982)hadpreviouslyarguedthatasalientrhythmicdifferencebetweenthetwolanguageswasthatEnglishgroupedsyllablesintounitsbeginningwithastressedsyllable,whereasFrenchgroupedsyllablesintounitsendingwithastressedsyllable,asin:

WenkandWiolandfurtherarguedthatthestressattheendsofrhythmicgroupsinFrenchwasmarkedprimarilybydurationallengthening.Basedonthisidea,Wenk(1987)predictedthatphrase-finallengtheningwouldbemorecommoninFrenchversusEnglishinstrumentalmusic.HetestedthisideabyhavingaprofessionalmusicianmarkphraseboundariesinEnglishversusFrenchclassicalmusic.Thenumberofphrasesinwhichthefinalnotewasthelongestnoteinthephrasewasthentalliedforbothcultures.WenkfoundthatitwasindeedthecasethatmoresuchphrasesoccurredinFrenchthaninEnglishmusic.

Wenk’sstudywaspioneeringinitsempiricalorientation,butitalsohadlimitationsthatmakeitdifficulttoacceptthesefindingsasafirmanswertothequestionofinterest.Only

Rhythm

Page 59 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

onecomposerfromeachculturewasexamined(FrancisPoulencandBenjaminBritten),andfromtheoeuvreofeachcomposer,onlyonemovementfromonepiecewasselected.Furthermore,nocomparableempiricaldataforlanguagerhythmwerecollected(e.g.,thedegreeofphrase-finallengtheninginEnglishvs.Frenchspeech).

Despiteitslimitations,Wenk’sstudyoutlinedausefulapproach,namelytoidentifyempiricalrhythmicdifferencesbetweentwolanguagesandthendetermineifthesedifferencesarereflectedinthemusicofthetwocultures.Pursuingthisideainarigorousfashionentailsthreerequirements.First,anempiricalmeasureofspeechrhythmisneededtoquantifyrhythmicdifferencesbetweenlanguages.Second,thissamemeasureshouldbeapplicabletomusicsothatlanguageandmusiccouldbecomparedinacommonframework.Third,boththelinguisticandmusicalsamplesneededtobebroadenoughtoinsurethatthefindingsarenotidiosyncratictoafewspeakersorcomposers.

(p.161) JosephDanieleandIconductedastudythatsetouttomeetthesecriteria(Patel&Daniele,2003a).LikeWenk,wefocusedonBritishEnglishandFrenchduetotheirdistinctspeechrhythmsandbecausetheyhavebeenthelocusofstrongintuitionsaboutlinksbetweenprosodyandinstrumentalmusic(e.g.,Hall,1953;Abraham,1974;Wenk1987).Ourworkwasinspiredbyrecentphoneticresearchonempiricalcorrelatesofstress-timedversussyllable-timedspeechrhythm(cf.section3.3.1,subsection“DurationandTypology”).Inparticular,theworkofLow,Grabe,andNolan(2000)attractedourattentionbecauseitfocusedonsomethingthatcouldbemeasuredinbothspeechandmusic,namelythedurationalcontrastbetweensuccessiveelementsinasequence.Theirmeasure,calledthenormalizedpairwisevariabilityindex,ornPVI,hadbeenappliedtovowelsinsentencesfromstress-timedandsyllable-timedlanguages,andhadbeenshowntobehigherinstress-timedlanguages,likelyduetothegreaterdegreeofvowelreductionintheselanguages(Grabe&Low,1990;Ramus,2002a;Lee&Todd,2004;seetheabove-mentionedsubsectionof3.3.1forbackgroundonthenPVI).

Twoaspectsofthismeasuremadeitappealingforusewithmusic.First,thenPVIaispurelyrelativemeasureofcontrast.Thatis,thedurationaldifferencebetweeneachpairofintervalsismeasuredrelativetotheaveragedurationofthepair.Thisnormalization,whichwasoriginallyintroducedtocontrolforfluctuationsinspeechrate,makesthenPVIadimensionlessquantitythatcanbeappliedtobothlanguageandmusic.(Forexample,nPVIcanbecomputedfromspeechdurationsmeasuredinsecondsandfrommusicaldurationsmeasuredinfractionsofabeat.)Second,thenPVIhasbeenappliedtovowels.Vowelsformthecoreofsyllables,whichcaninturnbecomparedtomusicaltones(i.e.,insettingwordstomusicitisquitecommonforeachnotetobeassignedtoonesyllable).26Ourstrategy,then,wastoapplythenPVItotonesequencesfromBritishandFrenchinstrumentalmusic,todetermineifdifferencesemergedthatreflectedtherhythmicdifferencesbetweenBritishEnglishandFrenchspeech.

Figure3.14showsthenPVItoBritishEnglishversuscontinentalFrenchspeech,basedonmeasurementsofvoweldurationsinsentencesutteredbynativespeakersofeachlanguage.(Thesentencesareshort,news-likeutterancesfromthecorpusofNazzietal.,1998.)27

Rhythm

Page 60 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(p.162)

Figure3.14 ThenPVIofBritishEnglishandFrenchsentences.Errorbarsshow+/-1standarderror.DatafromPatel,Iversen,&Rosenberg,2006.

ThenPVIissignificantlyhigherforEnglishthanforFrenchspeech.Figure3.15givesanintuitionforwhythisisthecasebyillustratingthepatternofvoweldurationforoneEnglishandFrenchsentenceinthiscorpus(cf.SoundExamples3.8a,b).

Forexample,inthetoppanel,thefirsttwovalues(about120msand40ms)arethedurationsofthefirsttwovowelsinthesentence(i.e.,thevowelsin“Finding”),andsoon.NotehowsuccessivevowelstendtodiffermoreindurationfortheEnglishsentencethanfortheFrenchsentence.IntheEnglishsentence,somevowelsareveryshort(oftenduetovowelreduction),whereasothervowelsarequitelong(oftenduetostress).Thisleadstoagreatertendencyfordurationalcontrastbetweenneighboringvowels,whichisreflectedinthenPVIscore.

Asmentionedabove,anappealingaspectofthenPVIisthatitcanbeappliedtomusicinordertomeasurethedurationalcontrastbetweensuccessivenotes.Westernmusicnotationindicatestherelativedurationofnotesinanunambiguousfashion,asshowninFigure3.16.

Inthefigure,thefirstnoteofeachthemeisarbitrarilyassignedadurationof1,andthedurationsoftheremainingnotesareexpressedasamultipleorfractionofthisvalue.(AnynumericalcodingschemethatpreservesrelativedurationofnoteswouldyieldthesamenPVI,becauseitisanormalizedmeasure.)Inthisexample,thenPVIoftheDebussythemeislowerthanthatoftheElgartheme,eventhoughtherawvariabilityofnotedurationintheDebussythemeisgreaterthanthatintheElgartheme(asmeasuredbythecoefficientofvariation,inotherwords,thestandarddeviationdividedbythemean).ThisemphasizesthefactthatthenPVIindexesthedegreeofcontrastbetweensuccessiveelementsinasequence,nottheoverallvariabilityofthoseelements.

(p.163)

Rhythm

Page 61 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Figure3.15 VoweldurationsinanEnglishandaFrenchsentence.Notethegreaterdegreeofshort-longcontrastintheEnglishsentencebetweenadjacentvoweldurations.ThenPVIfortheEnglishsentenceis54.9,andfortheFrenchsentenceis30.0.

Oursourceofmusicalmaterialwasastandardreferenceworkinmusicology,ADictionaryofMusicalThemes,SecondEdition(Barlow&Morgen-stern,1983),whichfocusesontheinstrumentalmusicofWesternEuropeancomposers.Inchoosingcomposerstoincludeinourstudywewereguidedbytwofactors.First,thecomposershadtobefromarelativelyrecentmusicalerabecausemeasurementsofspeechprosodyarebasedoncontemporaryspeech,andlanguagesareknowntochangeovertimeintermsofsoundstructure.Second,thecomposershadtobenativespeakersofBritishEnglishorFrenchwholivedandworkedinEnglandorFrance.Usingtheseguidelines,weexaminedallEnglishandFrenchcomposersfromBarlowandMorgensternwhowereborninthe1800sanddiedinthe1900s,andwhohadatleastfivemusicalthemesinthedictionarythatwereeligibleforinclusioninthestudy(seePatel&Daniele,2003a,forinclusioncriteria,andfortherationaleofusingmusicnotationratherthanrecordedmusicfornPVIanalysis).Wechosecomposerswhospannedtheturnofthecenturybecausethiseraisnotedbymusicologistsasatimeof“musicalnationalism”inEurope.

(p.164)

Rhythm

Page 62 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Figure3.16 Twomusicalthemeswiththerelativedurationsofeachnotemarked.nPVIofD122=42.2,ofE72=57.1.ThemesarefromBarlow&Morgenstern,1983.FromPatel&Daniele,2003a.

Basedonourcriteria,16composerswereincludedinthestudy,includingEnglishcomposerssuchasElgar,Delius,andVaughanWilliams,andFrenchcomposerssuchasDebussy,Poulenc,andSaint-Saëns.About300musicalthemeswererepresented,andonemusicalnPVIvaluewascomputedforeachtheme.TheresultsofouranalysisofmusicalnPVIareshowninFigure3.17,alongwiththespeechnPVIvalues.Remarkably,thetwocultureshavesignificantlydifferentmusicalnPVIvalues,withthedifferencebeinginthesamedirectionasthelinguisticnPVIdifference(seePatel&Daniele,2003a,andPateletal.,2006,forfurtherdetails).

Thusthereisempiricalevidencethatspeechrhythmisreflectedinmusicalrhythm,atleastinturn-of-thecenturyclassicalmusicfromEnglandandFrance.Howisthisconnectionbetweenlanguageandmusicmediated?Somemusicologistshaveproposedthatnationalcharacterarisesfromcomposersadaptingfolkmelodiesintotheircompositions.Becausesuchmelodiesaretypicallyfromsongs,itmaybethattherhythmofwordsinfluencestherhythmofthesemelodies,thusgivingthemelodiesalanguage-likerhythmicpattern.However,webelievethatthismaynotbethebestexplanationforourfinding,becauseourstudyincludednumerouscomposerswhoarenotthoughttobestronglyinfluencedbyfolkmusic,suchasElgarandDebussy(Grout&Palisca,2000).Instead,wefeeltheremaybeamoredirectroutefromlanguagetomusic.Itisknownfromstudiesoflanguageacquisitionthattheperceptualsystemissensitivetotherhythmicpatternsoflanguagefromaveryearlyage(Nazzietal.,1998;Ramus,2002b).Composers,likeothermembersoftheirculture,internalizethesepatternsaspartoflearningtospeaktheirnativelanguage.(Onemechanismforthisinternalizationisaprocesscalledstatisticallearning,whichisdiscussedinmoredetailinthenextchapter.)Wesuggestthatwhencomposerswritemusic,linguisticrhythmsare“intheirears,”andtheycanconsciouslyorunconsciouslydrawonthesepatternsinweavingthesonicfabricoftheirmusic.Thisdoesnotimplythattheconnectionbetweenlinguisticandmusicalrhythmisobligatory.Rather,thislinkislikelytobegreaterinhistoricalepochswherecomposersseekanationalcharacterfortheirmusic.

(p.165)

Rhythm

Page 63 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Figure3.17 ThenPVIofBritishEnglishandFrenchmusicalthemes.Errorbarsshow+/-1standarderror.DatafromPatel&Daniele,2003a,andPatel,Iversen,&Rosenberg,2006.

OurfindingsforEnglishandFrenchspeechandmusicimmediatelyraisedtwoquestions.WouldthemusicalnPVIdifferencebeobservedifabroadersampleofEnglishandFrenchthemesandcomposerswerestudied?Perhapsmoreimportantly,wouldourresultgeneralizetootherculturesinwhichstress-versussyllable-timedlanguagesarespoken?Fortunately,HuronandOllen(2003)providedanswerstothesequestions.UsinganelectronicversionofADictionaryofMusicalThemescreatedbyHuron,theycomputedthe(p.166) nPVIofamuchlargersampleofEnglishandFrenchmusicalthemes(about2000themes,composedbetweenthemid-1500sandmid-1900s).TheyconfirmedthatthenPVIofEnglishmusicwassignificantlyhigherthanthatofFrenchmusic,thoughthedifferencewassmallerthanthatfoundbyPatelandDaniele(likelyduetolessstringentsamplingcriteria).TheyalsocomputedthemusicalnPVIforarangeofothernations,analyzingalmost8,000themesfrom12nationalitiesovermorethan3centuries.Ofthenationalitiestheyexamined,fivecanbeassignedtostress-timedlanguagesandthreetosyllabletimedlanguages(Fantetal.,1991a;Grabe&Low,2002;Ramus,2002b).ThesearelistedinTable3.1alongwiththeirmusicalnPVIvalues.(ThedatainTable3.1representcorrectedvaluesoftheoriginaltableinHuron&Ollen,2003,kindlyprovidedbyDavidHuron.Seethischapter’sappendix2fordatafrommorecultures.)

Fouroutofthefivenationswithstress-timedlanguages(American,Austrian,English,andSwedish)doindeedhavehighermusicalnPVIvaluesthanthethreenationswithsyllable-timedlanguages,providingsupportfortheideathatstress-timedandsyllable-timedlanguagesareassociatedwithdistinctivemusicalrhythms.However,Germanmusicisanotableexception:IthasalowmusicalnPVIvaluedespitethefactthatGermanisastress-timedlanguagewithahighnPVIvalueforspeech(Grabe&Low,2002;Dellwo,2004).

However,theremaybeahistoricalreasonwhyGermanmusichasalownPVI,namelythewell-knowninfluenceofItalianmusiconGermanmusic(Kmetzetal.,2001).Because

Rhythm

Page 64 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

ItalianmusichasalownPVI,stylisticimitationofthismusicmightoutweighanylinguisticinfluenceoftheGermanlanguageonthenPVIofGermanmusic.OnewaytotestthisideaistoexaminethenPVIinhistoricalperspective,forexample,asafunctionofeachcomposer’sbirthyear.Whenthemesfrom14Germancomposerswereexaminedinthisfashionastrikingtrendemerged,asshowninFigure3.18(Patel&Daniele,2003b;Daniele&Patel,2004).

Table3.1MusicalnPviValuesforEightDifferentNationalities MusicalnPVI

Mean S.E.NationalitiesWithStress-TimedLanguages American 46.7 1.0Austrian 45.1 0.6English 45.6 0.9German 43.2 0.6Swedish 50.0 2.4NationalitiesWithSyllable-TimedLanguages French 43.4 0.7Italian 41.4 1.0Spanish 42.5 1.9

(p.167) Overthecourseof250years,nPVIalmostdoubled,atrendthatishighlystatisticallysignificant.(Interestingly,thistrendisalsoevidentforthesixAustriancomposersweincludedinourstudy.)GivenwhatisknownaboutthehistoryoftheGermanlanguage,thisisunlikelytoreflectachangeintherhythmofGermanfromsyllable-timedtostress-timedduringthisperiod(C.Heeschen,personalcommunication).Instead,itmostlikelyreflectshistoricalchangesinmusicalstyle,perhapsincludingawaninginfluenceofItalianmusiconGermanmusicoverthisperiod.Infact,thefindingwouldbeconsistentwiththeideathatItalianmusichadastronginfluenceonGermanmusicduringtheBaroqueera(1600–1750),lessinfluenceduringtheClassicalera(1750–1825),andtheleastinfluenceduringtheRomanticera(1825–1900).Moregenerally,itsuggeststhatinstudyinglinguisticinfluencesonspeechrhythm,itisimportanttokeepinmindhistoricalinfluencesthatcanruncountertolinguisticinfluences.28

Rhythm

Page 65 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Figure3.18 nPVIasafunctionofcomposerbirthyearfor20composers.(Soliddots=Germancomposers;opendots=Austriancomposers.)Thebest-fittinglinearregressionlineisshown.FromPatel&Daniele,2003b.

(p.168) Takingastepback,musicalnPVIresearchdemonstratesthattherhythmicstructureofspeechandmusiccanbefruitfullycomparedwithoutanyresorttonotionsofperiodicity.Itisworthnotingthattheresearchdonesofarhardlyexhaustswhatcanbedoneusingthismeasure.Forexample,onecouldapplythenPVItorecordingsofperformedmusicratherthantomusicnotation.OnecouldalsoexaminethenPVIofperformancesofthesamepieceofinstrumentalmusicbymusicianswhospeakstress-timedversussyllable-timedlanguages,toseeifthenativelanguageinfluencestemporalpatternsinmusicperformance(cf.Ohgushi,2002).Finally,onecouldstudyimprovisedmusic,forexample,bystudyingjazzmusicianswhospeakdifferentdialectswithdifferentrhythmicqualities(e.g.,intheUnitedStates,perhapsanortheastdialectversusasoutherndialect).Inthiscase,thenPVIcouldbeusedtoinvestigatewhetherthetemporalpatternofspeechisreflectedintherhythmofimprovisedmusic.

3.5.2RelationsbetweenNonlinguisticRhythmPerceptionandSpeechRhythm

Theideathatnonlinguisticrhythmperceptioncanbeinfluencedbyone’snativelanguagehasbeenarticulatedbybothlinguistsandmusicresearchers.Over50yearsago,Jakobson,Fant,andHalle(1952:10–11)madethefollowingclaim:

Interferencebythelanguagepatternaffectsevenourresponsestononspeechsounds.Knocksproducedatevenintervals,witheverythirdlouder,areperceivedasgroupsofthreeseparatedbyapause.ThepauseisusuallyclaimedbyaCzechtofallbeforethelouderknock,byaFrenchmantofallafterthelouder;whileaPolehearsthepauseoneknockafterthelouder.Thedifferentperceptionscorrespondexactlytothepositionofwordstressinthelanguagesinvolved:inCzechthestressisontheinitialsyllable,inFrench,onthefinalandinPolish,onthepenult.

ThegroupingssuggestedbyJakobsonetal.canbeschematicallyrepresentedasfollows,

Rhythm

Page 66 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

inwhicheachxrepresentaknockandtheuppercaseX’sarelouder:

(p.169) TheclaimofJakobsonetal.iscertainlyprovocative,buttherehasbeennoempiricalevidencetosupportit.Nevertheless,theideaofalinkbetweennativelanguageandnonlinguisticrhythmperceptionpersists.Forexample,StobartandCross(2000)havedocumentedaformofmusicfromtheViachapeopleoftheBolivianhighlandsinwhichthelocalmannerofmarkingthebeatisdifferentfromwhatmostEnglish-speakinglistenersperceive.SoundExample3.9illustratesthismusicwithanEastersongplayedonasmallguitar(charango).ThepositionatwhichtheViachaclaportaptheirfoottothebeatcanbeheardattheendoftheexcerpt.ThistendencytomarktheshortereventineachgroupoftwonotesasthebeatiscontrarytothetendencyofEnglishspeakerstohearthepatterniambically,thatis,withthebeatonthesecondeventofeachpair.StobartandCrossspeculatethatthetendencytomarkthebeattrochaically(onthefirstmemberofeachgroup)isrelatedtostresspatternsinwordsofthelocallanguage,Quechua.

Thetwoexamplesabovedifferinthattheformerconcernssegmentation(rhythmicgrouping),whereasthelatterconcernsbeatperception.Note,however,thatneitherreferstonotionsofperiodicityinspeech.Instead,theybothrefertopatternsoflexicalstressandhowthisinfluencesnonlinguisticauditoryperception.Thusonceagainweseethatinterestingclaimsaboutrhythmicrelationsbetweenmusicandlanguagecanbemadewithoutanyreferencetoperiodicityinspeech.

Howcanoneassesswhetherthenativelanguageinfluencestheperceptionofnonlinguisticrhythm?Asafirststep,itisnecessarytodemonstratethatthereareculturaldifferencesinnonlinguisticrhythmperception.Rhythmicsegmentationorgroupingisofparticularinterestinthisregard,asintimatedbyJakobsonetal.(1952).Thisisbecausepsycholinguisticresearchindicatesthatthattherhythmofone’snativelanguageleadstosegmentationstrategiesthatareappliedevenwhenlisteningtoaforeignlanguage.Theideathatthenativelanguagecaninfluencenonlinguisticrhythmicsegmentationisthusjustonestepawayfromthisidea(cf.section3.3.3,subsection“TheRoleofRhythminSegmentingConnectedSpeech”).

Yetatthecurrenttime,itiswidelybelievedthatelementarygroupingoperationsreflectgeneralauditorybiasesnotinfluencedbyculture.Thisbeliefstemsfromacentury-oldlineofresearchinwhichresearchershaveinvestigatedrhythmicgroupingusingsimpletonesequences(Bolton,1894;Wood-row,1909).Forexample,listenersarepresentedwithtonesthatalternateinloudness(…loud-soft-loud-soft…)orduration(…long-short-long-short…)andareaskedtoindicatetheirperceivedgrouping.Twoprinciplesestablishedacenturyago,andconfirmedinnumerousstudiessince,arewidelyaccepted:

1.Aloudersoundtendstomarkthebeginningofagroup.

Rhythm

Page 67 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

2.Alengthenedsoundtendstomarktheendofagroup.

(p.170) Theseprincipleshavecometobeviewedasuniversallawsofperception,underlyingtherhythmsofbothspeechandmusic(Hayes,1995b;Hay&Diehl,2007).However,thecross-culturaldatahavecomefromalimitedrangeofcultures(American,Dutch,andFrench).Aretheprinciplestrulyuniversal?AstudybyKusumotoandMoreton(1997)suggestedotherwise,findingthatAmericanversusJapaneselistenersdifferedwithregardtoPrinciple2above.ThisstudymotivatedareplicationandextensionofthisworkbyIversen,Patel,andOhgushi(2008),describedbelow.

Iversenetal.hadnativespeakersofJapaneseandnativespeakersofAmericanEnglishlistentosequencesoftones.Thetonesalternatedinloudness(“amplitude”sequences,SoundExample3.10a)orinduration(“duration”sequences,SoundExample3.10b),asshownschematicallyinFigure3.19.

Listenerstoldtheexperimentershowtheyperceivedthegrouping.TheresultsrevealedthatJapaneseandEnglishspeakersagreedwithprinciple1):bothreportedthattheyheardrepeatingloud-softgroups.However,thelistenersshowedasharpdifferencewhenitcametoprinciple2.)AlthoughEnglishspeakersperceivedthe“universal”short-longgrouping,manyJapaneselistenersstronglyperceivedtheoppositepattern,inotherwords,repeatinglong-shortgroups.(cf.Figure3.19).Becausethisfindingwassurprisingandcontradicteda“law”ofperception,Iversenetal.replicateditwithlistenersfromdifferentpartsofJapan.Thefindingisrobustandcallsforanexplanation.WhywouldnativeEnglishandJapanesespeakersdifferinthisway?

Figure3.19 Leftside:Schematicofsoundsequencesusedintheperceptionexperiment.Thesesequencesconsistoftonesalternatinginloudness(“amplitudesequence,”top),orduration(“durationsequence,”bottom).Intheamplitudesequence,thinbarscorrespondtosoftersoundsandthickbarscorrespondtoloudersounds.Inthedurationsequence,shortbarscorrespondtobriefersoundsandlongbarscorrespondtolongersounds.Thedotsbeforeandafterthesequencesindicatethatonlyanexcerptofalongersequenceofalternatingtonesisshown.Rightside:PerceivedrhythmicgroupingbyAmericanandJapaneselisteners,indicatedbyovals.Solidblackovalsindicatepreferencesthatfollow“universal”principlesofperception,whilethedashedblackovalindicatesapreferencethatviolatesthepurporteduniversals.

(p.171) Assumingthatthatthesedifferentperceptualbiasesarenotinnate,theykeyquestioniswhataspectofauditoryexperiencemightberesponsibleforthisdifference.

Rhythm

Page 68 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Twoobviouscandidatesaremusicandspeech,becausethesesoundpatternssurroundhumansthroughouttheirlife.Bothpatternspresenttheearwithsequencesofsoundthatmustbebrokenintosmallercoherentchunks,suchasphrasesinmusic,orphrasesandwordsinspeech.Mightthetemporalrhythmofthesechunksdifferformusicorspeechinthetwocultures?Thatis,mightshort-longpatternsbemorecommoninAmericanmusicorspeech,andlong-shortbemorecommoninJapanesemusicorspeech?Ifso,thenlearningthesepatternsmightinfluenceauditorysegmentationgenerally,andexplainthedifferencesweobserve.

Focusingfirstonmusic,onerelevantissueconcernstherhythmofhowmusicalphrasesbegininthetwocultures.Forexample,ifmostphrasesinAmericanmusicstartwithashort-longpattern(e.g.,a“pick-upnote”),andmostphrasesinJapanesemusicstartwithalong-shortpattern,thenlistenersmightlearntousethesepatternsassegmentationcues.Totestthisidea,weexaminedphrasesinAmericanandJapanesechildren’ssongs(becausewebelievetheseperceptualbiasesareprobablylaiddownearlyinlife).Weexamined50songsperculture,andforeachphrasewecomputedthedurationratioofthefirsttothesecondnoteandthencountedhowoftenphrasesstartedwithashort-longpatternversusotherpossiblepatterns(e.g.,long-short,orequalduration).WefoundthatAmericansongsshownobiastostartphraseswithashort-longpattern.Interestingly,Japanesesongsshowabiastostartphraseswithalong-shortpattern,consistentwithourperceptualfindings.However,themusicaldataalonecannotexplaintheculturaldifferencesweobserve,becausethisdatacannotexplaintheshort-longgroupingbiasofAmericanlisteners.

Turningtolanguage,onebasicdifferencebetweenEnglishandJapaneseconcernswordorder(Baker,2001).Forexample,inEnglish,shortgrammatical(or“function”)wordssuchas“the,”“a,”“to,”andsoforth,comeatthebeginningofphrasesandcombinewithlongermeaningful(or“content”)words(suchasanounorverb).Functionwordsaretypically“reduced,”havingshortdurationandlowstress.Thiscreatesfrequentlinguisticchunksthatstartwithashortelementandendwithalongone,suchas“thedog,”“toeat,”“abigdesk,”andsoforth.ThisfactaboutEnglishhaslongbeenexploitedbypoetsincreatingtheEnglishlanguage’smostcommonverseform,iambicpentameter.

Japanese,incontrast,placesfunctionwordsattheendsofphrases.CommonfunctionwordsinJapaneseinclude“casemarkers,”shortsoundsthatcanindicatewhetheranounisasubject,directobject,indirectobject,andsoforth.Forexample,inthesentence“John-san-gaMari-san-nihon-woagemashita,”(“JohngaveabooktoMari”)thesuffixes“ga,”“ni,”and“wo”arecasemarkersindicatingthatJohnisthesubject,Mariistheindirectobjectand“hon”(book)isthedirectobject.Placingfunctionwordsattheendsofphrasescreatesfrequentchunksthatstartwithalongelementandendwithashortone,(p.172) whichisjusttheoppositeoftherhythmofshortphrasesinEnglish(cf.Morganetal.,1987).

Apartfromshortphrases,theothershortmeaningfulchunksinlanguagearewords.Becauseourperceptionexperimentfocusedontwo-elementgroups,weexaminedthetemporalshapeofcommondisyllabicwordsinEnglishandJapanese.Englishdisyllabic

Rhythm

Page 69 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

wordstendtobestressedonthefirstsyllable(e.g.,MO-ney,MAY-be;Cutler&Carter,1987),whichmightleadonetothinkthattheywouldhavealong-shortrhythmicpatternofsyllableduration.Totestthis,weexaminedsyllabledurationpatternsforthe50mostcommondisyllabicwordsinthelanguage(fromacorpusofspontaneousspeech),andmeasuredtherelativedurationofthetwosyllables.Surprisingly,commonwordswithstressonthefirstsyllabledidnothaveastrongbiastowardalong-shortdurationpattern.Incontrast,commonwordswithstressonthesecondsyllable,suchas“a-BOUT,”“be-CAUSE,”and“be-FORE,”hadaverystrongshort-longdurationpattern.Thustheaveragedurationpatternforcommontwo-syllablewordsinEnglishwasshort-long(Figure3.20).

Figure3.20 Distributionofsyllabledurationratiosforcommontwo-syllablewordsinspontaneousspeechinAmericanEnglish.Separatehistogramsareshownforinitial-stressversusfinal-stresswordsin(a)and(b),andcombineddataareshownin(c),weightedbywordfrequency.Averagesindicatedbyarrowheads.Theoveralldistributionin(c)hasasignificantshort-longbias(averageratio=1:1.11).

(p.173) Thismeansthatashort-longrhythmpatternisreflectedatboththelevelofsmallphrasesandcommondisyllabicwordsinEnglish.Wealsoexaminedsyllabledurationpatternsinthe50mostcommondisyllabicwordsinJapanese.IncontrasttoEnglish,theaveragedurationpatternforsuchwordswaslong-short.Thusonceagain,linguisticrhythmmirroredtheresultsoftheperceptionexperiment.

Takingastepback,ourresultsshowthattheperceptionofrhythmicgrouping,longthoughttofollowuniversalprinciples,actuallyvariesbyculture.Ourexplanationforthisdifferenceisbasedontherhythmsofspeech.Specifically,wesuspectthatlearningthe

Rhythm

Page 70 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

typicalrhythmicshapeofphrasesandwordsinthenativelanguagehasadeepeffectonrhythmperceptioningeneral.Ifourideaiscorrect,thenrhythmicgroupingpreferencesshouldbepredictablefromthetemporalstructureofsmalllinguisticchunks(phrasesandwords)inalanguage.

Thesefindingshighlighttheneedforcross-culturalworkwhenitcomestotestinggeneralprinciplesofauditoryperception.MuchoftheoriginalworkonrhythmicgroupingoftoneswasdonewithspeakersofWesternEuropeanlanguages(e.g.,English,Dutch,andFrench).Althoughtheselanguagesdoindeedhaveimportantdifferences,theyallfollowthepatternofputtingshortfunctionwordsattheonsetofsmalllinguisticphrases,whichmayaccountforthesimilarityofperceptualgroupinginthesecultures.Amoreglobalperspectiverevealsthatlanguageswithphrase-finalshortfunctionwordsarewidespread,butexistlargelyoutsideofEurope,forexample,inIndiaandEastAsia(Haspelmathetal.,2005).WepredictthatnativespeakersoftheselanguageswillgrouptonesofalternatingdurationlikeJapaneselistenersdo(long-short).

Animportantfuturedirectionforthisworkconcernsthedevelopmentofrhythmicgroupingpreferencesinchildhood.Doinfantshaveaninnatebiasforaparticulargroupingpattern(e.g.,short-long),whichisthenmodifiedbyexperience(cf.Trainor&Adams,2000)?Oraretheyrhythmic“blankslates”?Regardingtheperceptionofrhythmbyadults,ifspeakersofdifferentlanguagesperceivenonlinguisticrhythmdifferently,thiscouldhelpexplainreportsofdifferencesbetweenWesternersandJapaneseintheperformanceofsimplemusicalrhythms(Ohgushi,2002;Sadakataetal.,2004).Thatis,simplerhythmsmaybeperformeddifferentlyindifferentculturesbecausetheyareperceiveddifferentlyduringlearning.Thiswouldindicatethatexperiencewithspeechshapesnonlinguisticrhythmcognitionataverybasiclevel.

3.5.3NeuralRelationshipsbetweenRhythminSpeechandMusic

Inthischapter,Ihaveclaimedthatcertainaspectsofspeechrhythmandmusicalrhythmshowastrikingsimilarity,suchasthegroupingofeventsintophrases,(p.174) whereasotheraspectsarefundamentallydifferent,suchastheroleoftemporalperiodicity.Towhatextentdoneuraldatasupportthisclaim?Isthereevidencethatsomeaspectsofrhythminspeechandmusicarehandledbysimilarbrainsystems,whereasotheraspectsshowlittleneuraloverlap?

Focusingfirstongrouping,thereisevidenceforoverlapinbrainprocessingofphraseboundariesinbothdomains.Thisevidencecomesfromelectricalbrainresponses(event-relatedpotentials,ERPs)innormalindividuals.Steinhaueretal.(1999)demonstratedthattheperceptionofphraseboundariesinlanguageisassociatedwithaparticularERPcomponenttermedthe“closurepositiveshift”(CPS),acentro-parietalpositivityofafewhundredmillisecondsthatstartssoonaftertheendofanintonationalphrase.Furtherstudiesusingfilteredorhummedspeech(toremovelexicalcuesandleaveprosodiccues)showedthattheCPSissensitivetoprosodicratherthansyntacticcuestophraseboundaries(Steinhauer&Friederici,2001;Pannekampetal.,2005).Inspiredbythiswork,Knöscheetal.(2005)examinedtheERPsinmusicianstotheendsofmusical

Rhythm

Page 71 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

phrases,andfoundacomponentsimilartotheCPSreportedbySteinhaueretal.UsingMEG,theyalsoidentifiedbrainareasthatwerelikelytobeinvolvedinthegenerationoftheCPSinmusic.Theseareasincludedtheanteriorandposteriorcingulatecortexandtheposteriorhippocampus.Basedontherolestheseareasplayinattentionandmemory,theresearchersarguethatthemusicalCPSdoesnotreflectthedetectionofaphraseboundaryperse,butmemoryandattentionprocessesassociatedwithshiftingfocusfromonephrasetothenext.

ThestudiesofSteinhaueretal.andKnöscheetal.pointthewaytocomparativeneuralstudiesofgroupinginlanguageandmusic.Thereismuchroomforfurtherwork,however.Forexample,intheKnöscheetal.studythesequenceswithphraseboundarieshaveinternalpauses,whereasthesequenceswithoutphraseboundariesdonot.Itwouldbepreferabletocomparesequenceswithandwithoutphraseboundariesbutwithidenticaltemporalstructure,forexample,usingharmonicstructuretoindicatephrasing(cf.Tanetal.,1981).Thisway,ERPsassociatedwithphraseboundariescannotbeattributedtosimpletemporaldifferencesinthestimuli.Itwouldalsobedesirabletoconductawithin-subjectsstudyofbrainresponsestophrasesinlanguageandmusic.Suchcomparativeworkshouldattendtotheabsolutedurationofmusicalversuslinguisticphrases,astheneuralprocessesinvolvedingroupingmaybeinfluencedbythesizeofthetemporalunitoverwhichinformationisintegrated(Elbertetal.,1991;vonSteinbüchel,1998).

Turningtothequestionofperiodicity,ifspeechrhythmsandperiodicmusicalrhythmsareservedbydifferentneuralmechanisms,thenonewouldpredictneuraldissociationsbetweenlinguisticrhythmicabilityandtheabilitytokeeporfollowabeatinmusic.Theneuropsychologicalliteraturecontainsdescriptionsofindividualswithmusicalrhythmicdisturbanceafterbraindamage,or(p.175) “acquiredarrhythmia”(e.g.,Mavlov,1980;Fries&Swihart,1990;Peretz,1990;Liégeois-Chauveletal.,1998;Schuppertetal.,2000;Wilsonetal.,2002;DiPietroetal.,2003).Twonotablefindingsfromthisliteraturearethatrhythmicabilitiescanbeselectivelydisrupted,leavingpitchprocessingskillsrelativelyintact,andthattherearedissociationsbetweenrhythmictasksrequiringsimplediscriminationoftemporalpatternsandthoserequiringtheevaluationorproductionofperiodicpatterns(e.g.,Peretz,1990).Forexample,Liégeois-Chauveletal.(1998)foundthatpatientswithlesionsintheanteriorpartoftheleftorrightsuperiortemporalgyrusweremuchmoreimpairedonametricaltaskthanonatemporaldiscriminationtask.Themetricaltaskinvolvedidentifyingapassageasawaltzoramarch,whereasthetemporaldiscriminationtaskinvolvedasamedifferentjudgmentonshortmelodicsequencesthatdifferedonlyintermsoftheirdurationpattern.Inthemetricaltask,patientswereencouragedtotapalongwiththeperceivedbeatofthemusictohelpthemintheirdecision.Wilsonetal.(2002)describeacasestudyofamusicianwitharighttemporoparietalstrokewhocoulddiscriminatenonmetricalrhythmsbutwhocouldnotdiscriminatemetricalpatternsorproduceasteadypulse.

Unfortunately,noneofthesestudiesexplicitlysetouttocomparerhythmicabilitiesinspeechandmusic.Thusthefieldiswideopenforcomparativestudiesthatemploy

Rhythm

Page 72 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

quantitativemeasuresofbothspeechandmusicalrhythmafterbraindamage.Itwouldbeparticularlyinterestingtostudyindividualswhowereknowntohavegoodmusicalrhythmicabilitiesandnormalspeechbeforebraindamage,andtoexaminewhetherdisruptionsofspeechrhythmareassociatedwithimpairedtemporalpatterndiscrimination,impairedmetricalabilities,orboth.

Anotherpopulationofindividualswhowouldbeinterestingtostudywithregardtospeechandmusicalrhythmareindividualswith“foreignaccentsyndrome”(Takayamaetal.,1993).Inthisraredisorder,braindamageresultsinchangesinspeechprosodythatgivetheimpressionthatthespeakerhasacquiredaforeignaccent.Itremainstobedeterminedifthisdisorderisassociatedwithsystematicchangesinspeechrhythm,butifso,onecouldexamineifsuchindividualshaveanyabnormalitiesintheirmusicalrhythmicskills.

Ofcourse,adifficultyinstudyingacquiredarrhythmiaandforeignaccentsyndromeisthatsuchcasesarequiterare.Thusitwouldbepreferabletofindlargerpopulationsinwhicheitherspeechrhythmormusicalrhythmicabilitieswereimpaired,inordertoconductcomparativeresearch.Onepopulationthatholdspromiseforcomparativestudiesaretone-deafor“congenitalamusic”individualswhohaveseveredifficultieswithmusicperceptionandproductionwhichcannotbeattributedtohearingloss,lackofexposuretomusic,oranyobviousnonmusicalsocial/cognitiveimpairments(Ayotteetal.,2002).Oneadvantageofworkingwithsuchindividualsisthattheycaneasilybefoundinanylargecommunitythroughaprocessofadvertisingandcareful(p.176) screening(Ayotteetal.,2002;Foxtonetal.,2004).Suchindividualsappeartohaveproblemswithbasicaspectsofpitchprocessing,suchasdiscriminatingsmallpitchchangesordeterminingthedirectionofsmallpitchchanges(i.e.,whetherpitchgoesupordown)(Peretz&Hyde,2003;Foxtonetal.,2004).Interestingly,theydonotseemtobeimpairedindiscriminatingsimpletemporalpatternsandcansynchronizesuccessfullytoasimplemetronome.However,theydohavedifficultysynchronizingtothebeatofmusic(DallaBella&Peretz,2003).Ofcourse,itcouldbethatthedifficultyinsynchronizingwithmusicissimplyduetothedistractioncausedbyastimuluswithpitchvariation,duetodeficitsinpitchprocessing(cf.Foxtonetal.,2006).Thusfuturestudiesofbeatperceptionincongenitalamusiashouldusecomplexrhythmicsequenceswithnopitchvariation,suchasthoseusedinthestudyofPatel,Iversen,etal.(2005)describedinsection3.2.1above(cf.SoundExamples3.3.and3.4).Ifmusicallytone-deafindividualscannotsynchronizetothebeatofsuchsequences,thiswouldsuggestthatthemechanismsinvolvedinkeepingabeatinmusichavenothingtodowithspeechrhythm(becausethespeechofmusicallytone-deafindividualssoundsperfectlynormal).29

Isuspectthatfutureresearchwillreveallittlerelationshipbetweenspeechrhythmabilitiesineitherproductionorperceptionandmusicalrhythmabilitiesinvolvingperiodicity(suchasmetricaldiscriminationorbeatperceptionandsynchronization).Thiswouldsupportthepointthatperiodicitydoesnotplayaroleinspeechrhythm.

3.6ConclusionSpeechandmusicinvolvethesystematictemporal,accentual,andphrasalpatterningof

Rhythm

Page 73 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

sound.Thatis,botharerhythmic,andtheirrhythmsshowbothimportantsimilaritiesanddifferences.Onesimilarityisgroupingstructure:Inbothdomains,elements(suchastonesandwords)aregroupedintohigher(p.177) levelunitssuchasphrases.Akeydifferenceistemporalperiodicity,whichiswidespreadinmusicalrhythmbutlackinginspeechrhythm.Ironically,theideathatspeechhasperiodictemporalstructuredrovemuchoftheearlyresearchonspeechrhythm,andwasthebasisforarhythmictypologyoflanguageswhichpersiststoday(stress-timedvs.syllable-timedlanguages).Itisquiteevident,however,thatthenotionofisochronyinspeechisnotempiricallysupported.Fortunately,muchrecentempiricalresearchonspeechrhythmhasabandonedthenotionofisochrony,andismovingtowardarichernotionofspeechrhythmbasedonhowlanguagesdifferinthetemporalpatterningofvowels,consonants,andsyllables.Akeyideathatmotivatesthisresearchisthatlinguisticrhythmistheproductofavarietyofinteractingphonologicalphenomena,andnotanorganizingprinciple,unlikethecaseofmusic.

Itmayseemthatbreakingthe“periodicitylink”betweenspeechandmusicwoulddiminishthechanceoffindinginterestingrhythmicrelationsbetweenthedomains.Infact,theconverseistrue.Changingthefocusofcomparativeworkfromperiodictononperiodicaspectsofrhythmrevealsnumerousinterestingconnectionsbetweenthedomains,suchasthereflectionofspeechtimingpatternsinmusic,andtheinfluenceofspeechrhythmsonnonlinguisticrhythmicgroupingpreferences.Althoughmanymoreconnectionsawaitexploration,itseemsclearthatsomeofthekeyprocessesthatextractrhythmicstructurefromcomplexacousticsignalsaresharedbymusicandlanguage.

Appendix1:ThenPVIEquation

Thisisanappendixforsection3.3.1,subsection“DurationandTypology.”ThenPVIequationis:

Inthisequation,misthenumberofdurationsinthesequence(e.g.,voweldurationsinasentence)anddkisthedurationofthekthelement.ThenPVIcomputestheabsolutevalueofthedifferencebetweeneachsuccessivepairofdurationsinasequence,normalizedbythemeanofthesetwodurations(thisnormalizationwasoriginallyintroducedtocontrolforfluctuationsinspeechrate).Thisconvertsasequenceofmdurationstoasequenceofm−1contrastivenessscores.Eachofthesescoresrangesbetween0(whenthetwodurationsareidentical)and2(formaximumdurationalcontrast,i.e.,whenoneofthedurationsapproacheszero).Themeanofthesescores,multipliedby100,yieldsthenPVIofthesequence.ThenPVIvalueforasequenceisthusboundedbylowerandupperlimitsof0and200,withhighernumbersindicatingagreaterdegreeofdurationalcontrastbetweenneighboringelements.

(p.178) Appendix2:MusicalnPVIValuesofDifferentNations

Rhythm

Page 74 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

ThisisanappendixforChapter3,section3.5.1,Table3.1.DatakindlyprovidedbyDavidHuron.

Inthetablesbelow,#C=numberofcomposers,sd=standarddeviation.

Nationality Mean #Themes #C sdAmerican 46.7 478 32 22.4Armenian 43.1 33 1 22.2Austrian 45.1 1,636 22 23.7Austro-Hung 45.9 14 1 18.7Belgian 48.7 41 6 19.4Bohemian 44.4 30 1 22.9Brazilian 41.5 5 1 18.3Catalan 48.9 12 1 16.8Cuban 36.0 17 2 17.9Czech 46.9 266 6 24.3Danish 51.0 7 1 24.6English 45.6 741 27 24.4Finnish 44.9 169 2 25.3Flemish 25.2 3 1 3.8French 43.4 1,343 52 25.2German 43.2 2,379 39 26.8Hungarian 45.4 244 8 25.4Irish 44.1 16 3 25.7Italian 41.4 572 46 23.9Mexican 28.4 13 2 19.3Norwegian 45.2 122 2 20.3Polish 50.2 60 8 18.7Romanian 42.4 26 2 21.3Russian 41.3 853 25 23.4Spanish 42.5 108 8 19.3Swedish 50.0 12 3 29.3

(p.179) Datafromtheabovetableregroupedbylanguage:English=American,English,Irish;French=French,Belgian;German=German,Austrian,Austro-Hungarian;Slavic=Russian,Czech,Polish,Bohemian;Spanish=Spanish,Catalan,Cuban,Mexican;Scandinavian=Danish,Norwegian,Swedish(notFinnish)

Rhythm

Page 75 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Language Mean #Themes #C sdEnglish 46.0 1,235 62 23.6French 43.6 1,384 58 25.0German 44.0 4,029 62 25.6Slavic 43.1 1,209 40 23.5Spanish 41.0 150 13 19.4Scandinavian 45.9 141 6 21.3

Notes:

(1)Thiscorrespondstoatimesignatureof6/8,incontrasttoawaltz,whichhasatimesignatureof3/4.Asonecansee,although6/8=3/4inmathematicalterms,theseratiosrefertoratherdifferentformsoforganizationinamusicalcontext.

(2)AnapparentexceptionoccursinlongrhythmiccyclesofIndianclassicalmusic(e.g.,cyclesof16beatsormore),inwhichthestrongaccentonthefirstbeatofeachcyclecanbeseparatedby10secondsormore,yetplaysanimportantperceptualroleinthemusic.However,thismaybeanexceptionthatprovestherule,asexplicitcountingofthebeatsbytheaudienceispartofthelisteningtraditioninthismusic.Thatis,consciouseffortisexpendedinordertokeeptrackofwherethemusicisinitslongmetricalcycle.

(3)Neuralactivityinthebetafrequencybandhasbeenassociatedwiththemotorsystem,raisingthepossibilitythatmeterperceptioninthebraininvolvessomesortofcouplingbetweentheauditoryandmotorsystem,evenintheabsenceofovertmovement.

(4)Schulkindetal.(2003)conductedaninterestinganalysisofmelodicstructurethathelpssuggestwhyphraseboundariesmightbeimportantlandmarksinmelodyrecognition.Theyexaminedthetemporaldistributionofpitchandtemporalaccentsinmelodies,inwhichpitchaccentsweredefinedasnotesthatweremembersofthetonictriadorpointsofcontourchange,andtemporalaccentsweredefinedasrelativelylongnotesandmetricallyaccentednotes.Theyfoundthataccentdensitywashigheratphraseboundariesthanwithinphrases.(Forthoseunfamiliarwiththeconceptofatonictriad,itisexplainedinChapter5).Thustheedgesofphrasesarestructuralslotsthatattractaccentsofvariouskinds.

(5)Inasurveyof21typologicallydifferentlanguages,Jun(2005)foundthatalllanguageshadatleastonegroupinglevelabovetheword,andmosthadtwo.

(6)Morespecifically,thelengtheningwasconfinedtothesyllabicrime(thevowelandfollowingconsonants).Interestingly,thisasymmetricexpansionofthesyllableduetoprosodicboundariesdiffersfromtemporalchangesinasyllableduetostress(Beckmanetal.,1992).

Rhythm

Page 76 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(7)Indoingthisresearch,itwouldbeimportanttobeawareofdurationallengtheninginthevicinityofphraseboundariesinbothspeechandmusic:Iftherearedifferentdegreesofpreboundarylengtheninginthetwodomains,theneventsnearboundariesshouldbeexcludedfromtheanalysisasthiswouldbeconfoundedwithvariabilitymeasures.

(8)IamgratefultoBrunoReppforprovidingmewiththisdata.

(9)Perceivedloudnessincorporatesbothphysicalintensityandthedistributionofenergyacrossdifferentfrequencybands,inotherwords,“spectralbalance.”Thelattermaybeamoresalientandreliablecuethantheformer(Sluijter&vanHeuven,1996,Sluijteretal.,1997).

(10)Inapitch-accentlanguage,awordcanhaveentirelydifferentmeaningdependingonitspitchpattern.Thedifferencebetweenatonelanguageandapitch-accentlanguageisthatintheformerthereisaprescribedpitchforeachsyllable,whereasinpitch-accentlanguagesacertainsyllableofawordmayhavelexicalspecificationforpitch(Jun,2005).

(11)Asanexample,Ladefogedpointsoutthatthewordkakemono(scroll)takesaboutthesameamountoftimetosayas“nippon”(Japan),andattributesthistothefactthatbothwordscontainfourmorae:[kakemono]and[nippon].

(12)Abercrombie’stheoryofsyllablesasrootedinchestpulseshasalsobeenfalsified.ItshouldbenotedthatAbercrombiewasapioneeringscientistwhoestablishedoneofthefirstlaboratoriesdevotedtobasicresearchinphonetics(inEdinburgh).Hisideasaboutspeechrhythmarebutatinysliceofhiswork,andthoughwrong,stimulatedagreatdealofresearch.

(13)Syllablesaregenerallyrecognizedashavingthreestructuralslots:theonsetconsonant(s),thenucleus(usuallyoccupiedbyavowel),andthefollowingconsonants(referredtoasthecoda).AsyllablewithoneconsonantintheonsetandnoneinthecodaisrepresentedbyCV,whereasCCVCmeanstwoconsonantsintheonsetandoneinthecoda,andsoon.

(14)Daueralsosuggestedthatstress-andsyllable-timedlanguageshadadifferentrelationshipbetweenstressandintonation:Intheformer,stressedsyllablesserveasturningpointsintheintonationcontour,whereasinthelatter,intonationandstressaremoreindependent.Asintonationisnotdiscussedinthischapter,thisideawillnotbepursuedhere.

(15)AlthoughRamusetal.(1999)relateddifferencesinΔCand%Vtosyllablestructure,FrotaandVigário(2001)pointoutthatintheBP/EPcase,differencesinthesevariablesaredrivenbyvowelreduction,becausesyllablestructuresaresimilarinthetwovarieties.SeeFrotaandVigário(2001)fordetails.FrotaandVigárioalsoprovideaveryusefuldiscussionoftheΔCparameterandtheneedtonormalizethisvariableforoverallsentenceduration/speechrate(seealsoRamus,2002a).

Rhythm

Page 77 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

(16)Individualresearcherswhoarecomparingsyllabledurationpatternsacrosstwolanguagescanalsohandletheproblemofsyllableboundaryidentificationbymakingallsuchdecisionsinamannerthatisconservativewithregardtothehypothesisathand.Forexample,ifcomparinglanguagesAandBwiththehypothesisthatsyllablesaremorevariableindurationinthesentencesoflanguageA,thenanyjudgmentcallsaboutsyllableboundariesshouldbemadeinsuchawayastoworkagainstthishypothesis.

(17)ThecorrelationbetweennPVIandrhythmclassisnotperfect,however.GrabeandLow(2002)foundahighvowelnPVIvalueforTamil,alanguagewhichhasbeenclassifiedassyllable-timed(cf.Keane,2006).ItshouldbenotedthatGrabeandLow’s(2002)resultsshouldbeconsideredpreliminarybecauseonlyonespeakerperlanguagewasstudied.SubsequentworkhasappliedthenPVItocross-linguisticcorporawithfewerlanguagesbutmorespeakersperlanguage(e.g.,Ramus,2002b;Lee&Todd,2004;Dellwo,2004).

(18)ItshouldbenotedthatbothGrabeandLow(2002)andRamus(2002b)measuredthenPVIofvocalicintervals,definedasvowelsandsequencesofconsecutivevowelsirrespectiveofsyllableandwordboundaries,whereasBolinger’sargumentsarefocusedonindividualvowels.Thisisnotaseriousproblembecausemostvocalicintervalsareindividualvowelsduetothestrongtendencyforvowelstobeseparatedbyconsonantsinspeech.Forexample,inthedatabaseusedbyRamus(eightlanguages,160sentences),thereare2,725vowels,outofwhich2,475(91%)aresingletons,inotherwords,asinglevowelflankedbyaconsonantoneitherside(orifthevowelisthefirstorlastphonemeofthesentence,asinglevowelflankedbyafollowingorprecedingconsonant,respectively).ThusitislikelythatnPVImeasurementsbasedonindividualvowelswouldproducequalitativelysimilarresultsasthosebasedonvocalicintervals(cf.Pateletal.,2006).

(19)PolishalsohasthelowestvocalicnPVIofalllanguagesintheRamusetal.(1999)database(Ramus,2002a;cf.Figure3.10).

(20)Itwouldbedesirableforfutureresearchonrhythmicclassestosuggestnewnamesforrhythmicclasses,asthecurrentnames(stress-timed,syllable-timed,andmora-timed)areimplicitlyboundupwiththe(failed)notionofisochrony.

(21)OnepossibleconfoundintheelegantstudyofNazzietal.(1998)isthepresenceofintonation,whichmayhaveplayedaroleintheinfants’discrimination.Indeed,Ramusetal.(2000)foundthatFrenchnewbornscoulddistinguishDutchfromJapaneseusingresynthesizedsaltanajspeech,butthattheirdiscriminationabilitywasmuchweakerwhentheoriginalF0contoursofthesentenceswerereplacedbythesameartificialcontours(Ramus,2002b).Healsonotesthatintonationcanberemovedentirelyusingflatsasasaresynthesis,butthattheresultingsoundpatternsareproblematicforusewithnewbornsandinfants,whomayfindthemboringordistressing.

(22)AnearlyconceptuallinkbetweenhierarchicaltheoriesoflinguisticrhythmandtheoriesofmusicalstructurewasmadebyJackendoff(1989),whonotedastructural

Rhythm

Page 78 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

equivalencebetweenonetypeofprosodictreestructureusedtodepicthierarchicalprominencerelationsinlanguageandatypeoftreeusedbyLerdahlandJackendoff(1983)toindicatetherelativestructuralimportanceofeventsinaspanofmusicalnotes.Jackendoffspeculatedthatthecoincidenceofthesetwoformalismsmightreflectthefactthatlanguageandmusicusedifferentspecializationsofgeneral-purposementalprinciplesforassigningstructuretotemporalpatterns,inotherwords,principlesthatparsesoundsequencesintorecursivehierarchiesofbinaryoppositionsofstructuralimportance.AsnotedbyJackendoff,however,prosodictreeshavebeenlargelyabandonedintheoriesofspeechrhythm.

(23)Suchastudywillhavetobecarefultotryandmatchacousticpropertiesoftheonsetofthespokenandmusicalsound.Forexample,ifamusicalsoundwithasharpattackisused,suchasapianotone,thenaspeechsoundwithaplosiveonset(suchas/ta/)shouldbeusedratherthanonewithagradualonset(suchas/la/).

(24)IamgratefultoLauraDilleyformarkingstressedsyllablesinthesesentences.

(25)IamgratefultoGraemeBooneforbringingthiscorrespondencetomyattention.

(26)AlthoughthisistrueforEnglishandFrench,itshouldbenotedthatinJapaneseitisthemoraandnotthesyllablethatgetsmappedontoamusicalnote(Hayes,1995a).

(27)ThenPVIvaluesforEnglishandFrenchspeechshowninFigure3.14aretakenfromPateletal.(2006),ratherthanfromPatelandDaniele(2003a).Bothstudiesshowasignificantdifferencebetweenthetwolanguages(EnglishnPVI〉FrenchnPVI),butthe2006studyisbasedonmoreaccuratemeasurements.SeePateletal.(2006)formeasurementdetails,andforalistofallsentencesanalyzed.

(28)UnlikeGermanmusic,EnglishandFrenchmusicdonotshowasignificantincreaseinnPVIovertheequivalenttimeperiod,basedonthemesinBarlowandMorgenstern’sdictionary(Greig,2003).Thisraisesaninterestingmusicologicalpuzzle:WhydoGermanandAustrianmusicshowsuchastronghistoricalchangeinthismeasureofrhythm,whereasEnglishandFrenchmusicdonot?

(29)Ofcourse,beforeanyfirmconclusionscanbedrawn,thespeechrhythmoftone-deafindividualswouldneedtobequantitativelymeasuredtoshowthatitdidnotdifferfromnormalcontrols(forexample,usingthenPVI).Also,itispossiblethattone-deafindividualscannotkeepabeatbecausetheirpitchperceptionproblemhascausedanaversiontomusic,sothattheyhavenothadenoughexposuretomusictolearnhowtokeepabeat.Thusitwouldbepreferabletoworkwithindividualswhohavenormalpitchperceptionandwhoenjoymusic,butwhocannotkeepabeat.Theexistenceofsuch“rhythm-deaf”individualsisintuitivelyplausible,astherearecertainlypeoplewholikemusicbutclaimtohave“twoleftfeet”whenitcomestodancing,and/orwhocannotclapalongwithabeat.Itshouldbepossibletofindapopulationofsuchpeoplethroughaprocessofadvertisingandscreening,akintotheproceduresusedtofindcongenitalamusics.

Rhythm

Page 79 of 79

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.All Rights Reserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of amonograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: UC - SantaCruz; date: 02 May 2015

Accessbroughttoyouby: UC-SantaCruz