StructurepreservationinintonationInvestigatingtheprosodyofinfant-directedspeechinEnglishandBengali
SameerudDowlaKhan,ReedCollegeMonday25February2013UniversityCollegeLondon
Outline
¡ QuestionsMotivationofstudy
¡ BackgroundIntonationalphonologyofEnglish,Bengali¡ DatacollectionExperiment,recordings,annotation
¡ ResultsIDSvs.non-IDS:similarities,differences¡ DiscussionProposedexplanationsforIDSdifferences
¡ ConclusionsSummary,upcomingextensions
Questions
Infant-directedspeech(IDS)
¡ Infant-directedspeech(IDS)ormothereseischaracterizedasinvolving:l Syntactically:shortersentenceswithsimplerstructure1,focusmovement2
l Lexically:smallervocabulary,paraphrasing3
l Phonetically:expansionofvowelspace4,stopVOTmanipulation5,distinctiveprosody
1O’Grady1997:2502Cristia20113Ferguson1964,O’Grady1997:2504Andruski&Kuhl1996,Burnhametal.2002,Cristia20115Sundberg&Lacerda1999,Sundberg2001
Infant-directedspeech(IDS)
¡ IDSprosodyistraditionallyanalyzedfromanacoustic-phoneticapproach1
l Expansionofpitchrange
l Raisingofpitchmaximuml Exaggerationofcontours
¡ Typicallyseenasanon-phonologicalorevenparalinguistic2phenomenon
1Jacobsonetal.1983,Sternetal.1983,Fernald&Simon1984,Fernaldetal.1989,Fernald&Mazzie1991,Greiser&Kuhl1998,Masataka19992Jacobsonetal.1983,O’Grady1997:250
Infant-directedspeech(IDS)
Infant-directedspeech(IDS)
Basicquestions
¡ It’swell-establishedthatcross-linguistically,IDSprosodyinvolvesawiderpitchrange
¡ Butaretherecategoricalchangesbetweennon-IDSandIDSprosody?
¡ Whatlinguisticinformationmightbeconveyedbythechanges?
¡ Whatcouldbethepressuresmotivatingthesechanges?
Background
Intonationalphonology
¡ Tohelplookbeyondlg.-specificprosodicproperties,Ilookedattwolanguageswithtypologicallydivergentintonation:l MainstreamAmericanEnglish
l BangladeshiStandardBengali1
1ThevarietyillustratedinKhan2010
Intonationalphonology
¡ SharedfeaturesofEnglishandBengali:l Pitchaccents1:tonesmarkingwordsl Intonationphrases(IPs)2:tonally-markedunitsl Boundarytones:tonesmarkingendsofIPs
¡ Language-specificfeatures:l Inventoryofintonationaltonesl Whatisconveyedbyspecifictonesequences
1Thesearepostlexicalpitchaccents,unlikethelexicalpitchaccentsofJapanese,Swedish,etc.,inthattheirpresenceandshapearenotpropertiesoftheword.2These“phrases”areprosodicunits,onlyindirectlyconnectedtosyntacticunits.
English:pitchaccents
¡ Pitchaccentsoccuronthestressedsyllableofprominentwords
l Partialinventory:Default Non-default(rising)H*(high) L+H*(earlyrise)1
L*(low) L*+H(laterise)
l Choiceisrelatedtoattitude2,focusstatus2,andtonalenvironment3
1ManyresearchersconsiderH*andL+H*tobevariantsofthesamepitchaccent.2Beckman&Hirschberg19903Dainora2002,2006
English:intonationphrases
¡ Boundarytonesoccurattheendsofintonationphrases(IPs)l Partialinventory:
¡ L-L%(low)¡ H-H%(high)
l Choiceisrelatedtosentencetype,speakerconfidence,finality1
1Beckman&Hirschberg1990
Bengali:pitchaccents
¡ Pitchaccentsoccurontheinitialsyllableofnearlyallwordsl Partialinventory:
Default Non-default(f-marked)L*(low) fH*(extrahigh) L*+fH(risetoextrahigh)
l Choiceisrelatedtospeakerattitude1,focusstatus/type2,andtonalenvironment
1H*markssarcasmorsurprise,L*otherwise(Khan2008,2013toappear)2Thechoiceoff-markedtonelargelydependsonfocustype(Khan2008,2013toappear)
Bengali:boundarytones
¡ BoundarytonesmarktheendsofIPsl Partialinventory:
L-initial H-initialL%(lowfall) H%(highrise)LH%(lowrise) HL%(highfall) HLH%(highfall-rise)
l Choiceisrelatedtosentencetype,informationstructure,finality1
1Khan2008,2013toappear
Bengali:boundarytones
¡ BoundarytonesmarktheendsofIPsl Partialinventory:
L-initial H-initialL%(lowfall) H%(highrise)LH%(lowrise) HL%(highfall) HLH%(highfall-rise)
l Choiceisrelatedtosentencetype,informationstructure,finality1
Bothcanserveastopicalizers
1Khan2008,2013toappear
Bengali:boundarytones
¡ BoundarytonesmarktheendsofIPsl Partialinventory:
L-initial H-initialL%(lowfall) H%(highrise)LH%(lowrise) HL%(highfall) HLH%(highfall-rise)
l Choiceisrelatedtosentencetype,informationstructure,finality1
Bothcanserveascontinuationrises
1Khan2008,2013toappear
Variablephrasing
¡ Prosodicphrasinginbothlanguagesisaffectedby:l Syntacticstructure,e.g.disambiguation
[old[menandwomen]]<IP>[[oldmen]<IP>andwomen]<IP>
l Informationstructure,e.g.focusIsawyourmotherreadingthemenu.<IP>IsawYOURMOTHER<IP>readingthemenu.<IP>l Otherfactors,e.g.speechrate
Datacollection
Design:subjects
¡ 19subjectsl 9speakersofEnglish(5M,4F)l 10speakersofBengali(5M,5F)
¡ Allwereparentsofyoungchildren¡ AllresideintheLosAngelesarea
Design:materials
¡ Recordedreadingsofthe“NorthWindandSun”fablel SuitableforadultspeechandIDSl Similarsemantics/pragmaticsacrosslanguagesl Consistentsemantics,morphosyntax,segmentalphonologyacrossconditions
Design:conditions
¡ Twoconditionsl Defaultreading(non-IDS):“Readatacomfortablepace.”
l Simulatedinfant-directedreading(IDS):“Readasthoughyouspeakingtoaninfant.”¡ Sametext,illustratedwithchildlikedrawings¡ Stuffedanimalsarrangedaroundspeaker
Experiment:annotation
¡ Englishannotationl Segmental:orthographyl Prosodic:tonelabelsfromMAE_ToBI1
¡ Bengaliannotationl Segmental:phonemicromanizationl Prosodic:tonelabelsfromB-ToBI2
1Beckmanetal.20052Khan2008,2013toappear
Analysis
¡ Acoustic-phoneticmeasurementsl Pitchrange
¡ Categorical/phonologicalmeasurementsl Inventoryoftonesl Numberofpitchaccentsl NumberofIPsl Numberofeachtypeofpitchaccentsl Numberofeachtypeofboundarytones
Results
Previewofresults
¡ What’sthesameacrossconditions:l Foreachlg.,IDSandnon-IDScanbothbeanalyzedusingthesameprosodicmodel
¡ Whatdiffersacrossconditions:l IDShaswiderpitchrange(highermax)l IDShasmoreIPsl SometonesaremorefrequentinIDSl IDShasmoreoverallcontourcomplexity
Structurepreservation
¡ IDSprosodyusesthesametonalinventoryandgrammarasnon-IDSprosodyl Noneedtoproposenew(allo)tonesforIDSl Samerelationsbetweentonesequenceandintonational“meanings”
Structurepreservation
¡ Intonationalstructurepreservationl cf.phonemicstructurepreservation,inwhichaphonologicalprocessresultsinasoundalreadyfoundinthephonemeinventory
l German/d/ à [–voi] / ___]word¡ /t/ alreadyfoundininventory
l Bengali/LH%/ à [HLH%] / [...___...]IDS¡ /HLH%/alreadyfoundininventory
Pitchrange
¡ AllBengalispeakersraisedthef0maxinIDS
¡ f0minnotconsistentlylowered
¡ SamepatternseeninEnglish
¡ Followsfrompreviousstudies
¡ “Authentic”IDS
IDS non-IDS
meanf090%f0range
10%f0range
PreviewofEnglish-specificresults
¡ EnglishIDSinvolves:l Increaseinrisingpitchaccentsl IncreaseinIPs
English:pitchaccents
¡ Onaverage,speakersincreasedthenumberandproportionofrisingpitchaccentsinIDSl 3.9%increaseinL*+Hproportion[p=0.02]l 7.5%increaseinL+H*proportion[p=0.01]
byspkr.
English:intonationphrases
¡ Onaverage,Englishspeakersproduced33.3%(=3.44)moreIPsinIDS[p<0.01]
byspkr.
PreviewofBengali-specificresults
¡ BengaliIDSinvolves:l Increaseinf-markedpitchaccentsl IncreaseinIPsl IncreaseinHL%andHLH%boundarytones
Bengali:pitchaccents
¡ f-markedpitchaccentuseishigherinIDSforallbutonespeakerl fH*l L*+fH
byspkr.
fH*,L*+fHinnon-IDS
fH*,L*+fHinIDS
Bengali:intonationphrases
¡ Onaverage,Bengalispeakersproduced49.0%(=8.97)moreIPsinIDS[p<0.01]
byspkr.
Bengali:boundarytones
¡ TheincreaseinIPscanbelargelyattributedtoincreasesinthoseendingin:l HL%(highfalling)l HLH%(highfalling-rising)
byspkr.
HL%,HLH%inIDS
HL%,HLH%innon-IDS
Summaryofresults
¡ TrueforIDSinbothlanguages:l Morenon-defaultpitchaccents
¡ Despitelackofincreaseindefaultpitchaccents
l MoreIPs¡ Certainboundarytonesweremorecommon
¡ So,whydoweseethesemodifications?
Discussion
Why:pitchaccentpatterns
¡ WhydoesIDSinvolveanincreaseinnon-defaultaccents?l EnglishrisingL*+HandL+H*l Bengalif-markedfH*andL*+fH
¡ ThesetonescanmarkfocusedelementsèGreateruseoffocusprosodyinIDS1
1IDScanalsoinvolvegreateruseofsyntacticmovementtoconveyfocus(Cristia2011).
Why:phrasingpatterns
¡ WhydoesIDSinvolvemoreIPs?l IPbreakshelpdisambiguatesyntaxl Boundarytonesconveyinformationstructure
¡ BengaliHL%canmarktopicalization
èMoremarkingofsyntactic/informationstructureinIDS
Why:contourpreference
¡ TherecouldbeotherreasonswhyrisingpitchaccentsandH-initialboundarytonesaremorecommoninIDSl Preferenceformorepitchvariation,to“keepthingsinteresting”fortheinfant1
l PreferencefortonesinvolvingHasinfantspreferhigherpitch2
1Fernald1991,Werker&McLeod19892Kearsley1973,Fernald&Kuhl1981
Why:contourpreference
¡ Butonlytoneswithwell-formedcomplexandhighcounterpartscanundergothissubstitution
¡ IDSbalancesgrabbing/keepinginfant’sattentionwithstructurepreservation
Conclusions
Returningtothequestions
¡ Aretherecategoricalchangesbetweennon-IDSandIDSprosody?l Moresubstitutionofdefaultpitchaccentswithnon-defaultpitchaccents
l HigherlikelihoodtoaddIPbreaksl Moresubstitutionoflower/simplertoneswithhigher/morecomplextones
l But...thebasicprosodicsystemisthesame
Yes(but...)
Returningtothequestions
¡ Whatlinguisticinformationmightbeconveyedbythechanges?l GreateruseofIPbreakscandisambiguatesyntacticstructure
l GreateruseofIPbreaksandnon-defaultpitchaccentscanhighlighttopic,focus
Returningtothequestions
¡ Whatcouldbethepressuresmotivatingthesechanges?l Pressurestowardsgreatercomplexity:desiretoattractinfant’sattention,desiretoclarifycomplexstructures
l Pressureagainstgreatercomplexity:intonationalstructurepreservation
Workinprogress
¡ Moredataontheway!l SecondtranscribercurrentlyannotatingEnglishl SecondtranscriberhasalreadyannotatedBengali,dataiscurrentlybeinganalyzed
Workinprogress
¡ Moreprincipledwayoflookingattheconnectionbetweenphonologicalchangesandsyntax/informationstructurel e.g.isprobabilityofIPbreakhigheratlargesyntacticphrasebreaks?
¡ Agreatareaforcollaborationacrosslinguisticsubfields!
Thankyou!
SpecialthankstomycollaboratorKristineM.Yu,aswellastoJaimePannaRoemer,MeghaSundara,theReedCollegeLinguisticsDepartment,theexperimentalsubjects,andtheaudiencehereatUCL!