Anearlierversionofthisstudywaspresentedatthe143rdmeetingoftheAcousticalSocietyofAmericaPittsburghPAJune2002Electronicmailtbentnorthwesternedu1600JAcoustSocAm1143September2003000149662 ID: 298592
Download Pdf The PPT/PDF document "Theinterlanguagespeechintelligibilityben..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Theinterlanguagespeechintelligibilitybene®tTessaBentandAnnR.BradlowDepartmentofLinguistics,NorthwesternUniversity,2016SheridanRoad,Evanston,Illinois60208Received20December2002;revised6June2003;accepted16June2003Thisstudyinvestigatedhownativelanguagebackgroundin¯uencestheintelligibilityofspeechbynon-nativetalkersfornon-nativelistenersfromeitherthesameoradifferentnativelanguagebackgroundasthetalker.NativetalkersofChinese(2),Korean(2),andEnglish( Anearlierversionofthisstudywaspresentedatthe143rdmeetingoftheAcousticalSocietyofAmerica,Pittsburgh,PA,June2002.Electronicmail:t-bent@northwestern.edu1600J.Acoust.Soc.Am.114(3),September20030001-4966/2003/114(3)/1600/11/$19.002003AcousticalSocietyofAmerica withanotherspeaker''p.4.Similarly,Wein-states,``Whentheotherinterlocutorisalsobilingual,therequirementsofintelligibility...aredrasticallyreduced.Undersuchcircumstances,thereishardlyanylimittointerference;formscanbetransferredfreelyfromonelanguagetotheotherandoftenusedinun-adaptedshape''p.81.SmithandRa®qzadsomeexperimentalevidenceinsupportofthisclaiminastudyinwhichforeign-accentedtalkersofEnglishfrommul-tiplenativelanguagebackgroundsandonenativeEnglishtalkerrecordedEnglishpassagesoftheirownchoosing.Then,listenersfromvariouslanguagebackgrounds,includ-ingnativelanguagesthatmatchedandmismatchedthenativelanguagesofthetalkers,completedaclozeproceduretestinwhichthelisteners®lledinwordsthathadbeenremovedfromtranscriptsoftherecordedpassages.Resultsshowedthatthenon-nativelistenersfoundthenativetalkerequallyorlessintelligiblethanthenon-nativetalkers.Nevertheless,atalker±listenermatchwithrespecttonativelanguagedidnotnecessarilyproducethehighestintelligibilityscores.Un-fortunately,thepassagesinthisstudydifferedsigni®cantlyindif®cultyandastrongpositivecorrelationbetweenpassagedif®cultyandintelligibilitywasreported.Thiscorrelationsuggeststhattheobservedintelligibilitydifferencesmayhavebeenduetothedif®cultyofthepassageratherthantothelanguagebackgroundsofthetalkersandlisteners.Morerecently,vanWijngaardenandvanWijn-etal.providedstrongevidenceinsupportoftheclaimthatundercertainconditionsnon-nativelisteners®ndsentencesproducedbynon-nativetalkersatleastasin-telligibleassentencesproducedbynativetalkers.Speci®-cally,vanWijngaardenfoundthatfornativelisteners,allfourofthenon-nativetalkersincludedinthestudywerelessintelligiblethaneachofthefournativetalkersincludedinthestudy.However,thetwohigherintelligibilitynon-nativetalkersbasedontheirrankingforthenativelistenersweremoreintelligiblefornon-nativelistenersthaneachofthefournativetalkers.vanWijngaardenetal.suredintelligibilityofDutch,English,andGermansentencesproducedbynativeandnon-nativetalkersofthesethreelan-guagesfortrilinguallistenersnativelanguageDutch,sec-ondlanguageEnglish,thirdlanguageGerman).Thelis-tenerswereallmorepro®cientinEnglishthaninGermanasevidencedbythefactthattheylearnedEnglishearlier,usedEnglishmorefrequently,andgavehigherself-reportedscoresonoverallpro®ciencyinEnglishthaninGerman.WhenlisteningtoEnglishtheirhigherpro®ciencyforeignthelistenersfoundthenativeEnglishtalkersmoreintelligiblethanthenon-nativetalkers.However,whenlis-teningtoGermantheirlowerpro®ciencyforeignlanguagetheyfoundthenon-nativetalkersmoreintelligiblethanthenativetalkers.The®ndingsofthesestudiesdemonstrateanintelligibilityadvantagebetweennon-nativetalker±listenerpairswhenthelistenerisatarelativelyearlystageoftargetlanguageacquisitionandthetalkerisatarelativelyadvancedstageoftargetlanguageacquisition.Similarly,Imaietal.foundanadvantageofsharedlanguagebackgroundforwordrecognition.Theyinvestigatedtheabilityofnativeandnon-nativelistenerswitha®rstlanguageofSpanishtorecognizeEnglishwordsspokenbyonenativetalkerandonenon-nativetalkeralsowithSpanishasa®rstlanguage.Na-tivelistenersperformedbetterthanthenon-nativelistenersi.e.,recognizedmorewordsforthewordsproducedbythenativetalker,whereasthenon-nativelistenersperformedbet-terthanthenativelistenersforwordsproducedbythenon-nativetalker.Animportant®ndingfromthisstudywasthatthenon-nativelistenersoutperformedthenativelistenersonlyforthewordsproducedbythenon-nativetalkerthatcamefromdenselexicalneighborhoodsi.e.,wordsthathavemanysimilarsounding``neighbors''withwhichtheycaneasilybeconfused.This®ndingisconsistentwiththe®nd-ingofBradlowandPisonithatwordrecognitionbynon-nativelistenersisworsefor``hard''wordsfrequencywordsinhigh-densityneighborhoodsthanfor``easy''wordshigh-frequencywordsinsparseneighbor-.Sincelexicalneighborhoodcharacteristicsarede-®nedintermsofsegmentssharedacrosswords,these®nd-ingssuggestaconnectionbetweenlower-levelprocessingatthesegmentallevelandhigher-levelprocessingatthewordOtherworkonnon-nativespeechperceptionhasshownthatthesuccessofacquiringthesecondlanguagesystemdependsonmanyfactors,includingbutnotlimitedtoofacquisition,durationofexposure,amountofcontinued®rstlanguageusage,andquantityandqualityofsecondlan-guageinpute.g.,Flege,2002.Althoughstudieshaveshownthatadultsarecapableoflearningnovelphoneticcontrastsetal.,1991;Livelyetal.,1993,1994;Bradlowetal.,1997,theimpactofthenativelanguagehasbeenshowntoin¯uencespeechperceptionevenforveryearlybilinguals.Forexample,Sebastian-GallesandSoto-Faracoshowedthatbilingualswholearnedtheirsecondlan-guageasearlyas4yearsofagewerelessef®cientatpro-cessingphonemiccontrastson-linethannativelisteners.Theperceptionandproductionofnon-nativecontrastsinlatebi-lingualsandnaivelistenersisevenmorestronglyin¯uencedbythe®rstlanguage.However,animportantgeneralizationtoemergefromstudiesofnon-nativephonemeperceptionisthatnotallnon-nativecontrastsareequallydif®cultforsec-ondlanguagelearners.Therelativedif®cultywithwhichnon-nativelistenersperceivenon-nativesoundsisrelatedtotherelationshipbetweenthestatusofsoundsintheoverallsystemsofphonemiccontrastsinthe®rstandsecondlan-e.g.,Flegeetal.,1999;Bestetal.,2001.Thesedif®cultiesinphonemeperceptionhavebeenshowntoaffectlexicalrepresentationforearlybilingualsetal.andtoimpairwordrecognitionforearlyandlatebi-etal.,2000.Experimentswithnon-nativelistenersandnativetalkershaveshownthatnon-nativelis-tenerscanperformwithnative-likepro®ciencyinthequiet,butevenearlybilingualsareimpairedinnoisecomparedtonativelistenerse.g.,Mayoetal.,1997;McAllister,1990Thecauseforthisdeclineisnotfullyunderstood,butacon-tributingfactormaybethefactthatnon-nativelistenersarelessabletotakeadvantageofcontextualcuesforwordiden-etal.,1997,andsowheninthepresenceofnoisetheyhavelesssignal-independentinformationtorelyonasanaidtosentencerecognition.J.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t Twoimportantissuesregardingspeechintelligibilitybe-tweennativeandnon-nativetalkersandlistenersremainun-resolvedincontrolledlaboratorystudies.First,allofthestudieswhereboththetalkerandlistenerwerenon-nativehaveinvolvedasharednativelanguagebetweentalkerandlistener.TheonlyexceptionisSmithandRa®qzadwheretheuncontrolledvariableofpassagedif®cultycon-foundedtheresults.Therefore,westilldonotknowhowtheintelligibilityofnon-nativetalkersfornon-nativelistenersvariesdependingonwhetherthereisamatchormismatchinnativelanguagebackgroundbetweentalkerandlistener.Dataaddressingthisissuewillallowustoassesstheextenttowhichthepossiblenon-nativetalker±listenerintelligibilitybene®tisduetoaparticularsharedinterlanguageortocer-tainmoregeneralfeaturesofnon-nativelanguageproductionandperceptionthatareindependentoftheparticularnativeandtargetlanguagesinquestion.Second,the®ndingsofvanWijngaardenandvanWijngaardenetal.gestedthattalkerandlistenerpro®ciencyinthetargetlan-guageisimportantfordeterminingtherelativeintelligibilityofnativeandnon-nativetalkers,butthisfactordeservesmoreattention.Byincludingnon-nativetalkerswhoareknowntovaryintargetlanguageproductionpro®ciency,thepresentstudyaimedtogainfurtherinsightintothefactorsthatcontributetoaninterlanguagespeechintelligibilityben-Usingadatabaseofsentencerecordingsbynon-nativetalkersfromvariousnativelanguagebackgrounds,thepresentstudyinvestigatedhowthetalker±listenermatchormismatchwithrespecttolanguagebackgroundandvariationinsecondlanguagepro®ciencyaffectednon-nativetalkerin-telligibility.Wemadetwospeci®cpredictions.First,wepre-dicted,basedonthe®ndingsofvanWijngaardenvanWijngaardenetal.,thatanon-nativetalkerwithrelativelyhighpro®ciencyinEnglishspeechproductionwouldbeatleastasintelligibleasanativetalkerfornon-nativelistenersfromthesamenativelanguagebackground.Thispredictionsupposesthatnon-nativelanguagelearnersfromagivennativelanguagehaveabroadbaseofsharedlinguisticandphoneticknowledgethatfacilitatesspeechcommunicationinthenon-nativelanguage.Weexpectedthisinterlanguageintelligibilitybene®ttobeattenuatedinthecaseoflesspro®cientnon-nativetalkersduetothefactthattheirspeechproductionmaystraysofarfromthetargetlan-guagenormthatimportantlexicalcontrastsmaynotbead-equatelyconveyed,resultinginwordandsentenceintelligi-bilitythatislowenoughtoproduceverylowoverallintelligibilityforbothnativeandnon-nativelisteners.Sec-ond,wepredictedthatarelativelyhighpro®ciencynon-nativetalkerwillalsobeatleastasintelligibleasanativetalkerfornon-nativelistenersfromadifferentnativelan-guagebackground.Thispredictionsupposesthat,regardlessofnativelanguagebackground,certainfeaturesofnon-nativespeechwillmakenon-nativetalkersmoreintelligibletoallnon-nativelisteners.Forexample,non-nativetalkersmaybelesslikelythannativetalkerstoapplycertainreduc-tionphenomenathatcharacterizenativeaccentedrunningspeechinEnglishsuchasalveolar¯appingandfailuretorelease®nalstops.Whilethesefeaturesofnon-nativeEn-glishspeechmaycontributetotheoverallimpressionofaforeignaccentfornativelisteners,theymayinfactbeben-e®cialforspeechperceptionbyallnon-nativelistenersofEnglishregardlessofnativelanguagebackground.II.THENORTHWESTERNUNIVERSITYFOREIGNACCENTEDENGLISHSPEECHDATABASETheNorthwesternUniversityForeignAccentedEnglishSpeechDatabaseNUFAESDcontainsrecordingsof64sen-tencesproducedby32non-nativetalkersforatotalof2048recordedsentences.Alongwiththeserecordings,thedata-baseincludesdemographicinformationabouteachtalkerandanoverallintelligibilityscoreforeachtalkerasmeasuredbyaperceptiontestwithnativeEnglishlisteners.Thesentencesinthisdatabaseweretakenfromthesen-tencelistsincludedintherevisedBamford±Kowal±BenchStandardSentenceTestdevelopedbytheCochlearCorporationforusewithAmericanchildren.TheoriginalBamford±Kowal±BenchStandardSentenceTestwasdevel-opedforusewithBritishchildrenBamfordandWilson,1979;BenchandBamford,1979.Thesesentenceswerecho-senforthisdatabasebecausetheyincludewordsthatarehighlyfamiliartonon-nativesandaresyntacticallysimple.Eachlistconsistsof16simple,declarativesentenceswith3or4keywordsforatotalof50keywordsperlist.Fromthe21listsincludedintheBKB-Rtest,fourlistslists7,8,9,and10wereselectedfortheNUFAESD.Theselistswereselectedbasedontheirequivalentintelligibilityscoresfornormal-hearingchildrenasreportedinBamfordandWilson.Ratingsofageofacquisition,writtenandverbalfre-quency,imagability,concreteness,andfamiliarityforthe155keywordsinthesefourlistsweregatheredfromtheMRCPsycholinguisticDatabase:MachineUsableDictionaryVer-sion2.00andareshowninTableI.Becausesomeofthekeywordsarerepeatedwithinandacrosslists,therearenot200uniquekeywords.ThetalkersrecordedforthisdatabasewererecruitedfromtheNorthwesternUniversityInternationalSummerIn-stituteandEnglishasaSecondLanguageoverthecourseofthesummerof2000andthe2000±2001academicyear.TheInternationalSummerInstituteprovidesincominginternationalgraduatestudentsfromacrosstheuni-versitywithintensiveEnglishlanguagetrainingaswellasageneralintroductiontoacademiclifeintheUSAduringthemonthbeforetheybegintheirgraduatestudiesatNorthwest-ernUniversity.Alloftheparticipantsinthisprogramhadalreadybeenadmittedtoadoctoralprogramandhadthere-foredemonstratedahighlevelofpro®ciencywithwrittenEnglishcommunication.However,thesestudentsallhadlim-itedexperiencewithspokenEnglishcommunication.ThesubjectsrecruitedfromtheESLprogramallcametotheprogramduetotheirownortheirdepartment'srecognitionoftheirneedtoimprovetheiroralandauralEnglishskills.Thetalkerscamefromavarietyoflanguagebackgrounds:20),Korean(5),Bengali(1),Hindi1),Japanese(1),Romanian(1),Slovakian(1),Spanish(1),andThai(1).Thetalkersrangedinagefrom22to32yearswithanaverageageof25.5years.Theyhadspentonaverage2.7monthsintheUnitedStates1602J.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t witharangefrom0.25monthsto24monthsatthetimeofrecording.TheaverageageatwhichtheparticipantsbegantostudyEnglishinschoolwas12.0yearsandtheaveragelengthofEnglishlearningwas9.8years.Alltalkershadnoknownspeechorhearingimpairments.Alltalkerswerepaidfortheirparticipation.Thetalkerswererecordedinasound-attenuatedbooth.TherecordingweremadeonanArielProportA/DsoundboardwithaShureSM81microphone.Allsubjectsreadthefoursetsofsentencesinthesameorder.Aftertherecording,thesound®leswereconvertedtotheWAVformatwitha16-kHzsamplingrateandtransferredtoaPC-basedcom-puter.Thedigitalspeech®leswerethensegmentedintosen-tencelength®lesandtheroot-mean-squareamplitudewasequatedacrossall®les.Forassessingtheintelligibilityofthetalkers,thedigi-tizedrecordingswereembeddedinwhitenoise,yieldingaspeech-plus-noise®lewitha5-dBsignal-to-noiseratio.Eachofthestimulus®lesconsistedofa400-millisecondsilentleader,followedby500msofnoise,followedbythespeech-plus-noise®le,andendingwith500msofnoiseonly.InordertokeepthenumberofnativeEnglishlistenersmanageablewhilestillevaluatingtheintelligibilityofallthenon-nativetalkers,eighttestconditionswith®velistenerseachwereconstructed.EachnativeEnglishlistenerevalu-ated64sentencesfromthefourBKB-Rlistsandall32non-nativetalkersdescribedabove.Eachofthe32non-nativetalkerssuppliedtwodifferentsentencerecordingstoeachofeighttestconditionsforatotalof64sentencespercondi-.Therefore,eachtalker'sintelligibilitywasevaluatedonthebasisof16sentences(8conditions2sen-tencesperconditionfromonefullBKB-Rlistinamultipletalkerpresentationformati.e.,ineachcondition,listenersevaluatedsentencesbyall32talkers.ThenativeEnglishlistenerswerepresentedwitheachsentenceonceandtheirtaskwastotranscribethesentenceinstandardEnglishor-thography.Intelligibilityscoresforeachtalkerwerebasedonthenumberofkeywordscorrectlytranscribedacrossallofthe16sentencessubmittedtoperceptualevaluationbythenativeEnglishlisteners.Therawpercent-correctscoreswereconvertedtorationalizedarcsinetransformunits.Thistransformationplacesthescoresonalinearandadditivescale,thusfacilitatingmeaningfulstatisticalcomparisonsacrosstheentirerangeofthescaleStudebaker,1985.Therauscalerangesfrom23to123.Themiddleofthescalecorrespondscloselytopercentcorrect,butattheedgesofthescale,below15%andabove85%,therauscaledivergesfrompercentcorrect.Forexample,80%correspondsto79.9raubut95%correspondsto102rau.Thenon-nativesubjects'sentenceproductionscoresi.e.,theaveragepercentageofkeywordscorrectlytranscribedbythenativeEnglishlisten-rangedfrom43to93rau.III.TESTOFFOREIGN-ACCENTEDENGLISHUsingmaterialsfromtheNUFAESD,wedesignedanexperimenttoinvestigatetheperceptionofforeign-accentedEnglishbylistenersfromvariouslanguagebackgrounds.TheoveralldesigninvolvedpresentingEnglishsentencespro-ducedby®vetalkerstofourlistenergroups.ThetalkerswerenativetalkersofChinese(2),Korean(2),andEn-glish(1).ThelistenerswerenativetalkersofChinese21),Korean(10),English(21),andamixedgroupofnon-nativetalkersofEnglishfromvariousnativelanguagebackgrounds(12).Noneofthenon-nativelis-tenershadservedastalkersinthecollectionoftheNU-FAESDmaterialsandnoneofthenativeEnglishlistenershadevaluatedthesentencesintheNUFAESD.Ofparticularinterestforthisstudywastheeffectofalistener±talkermatchormismatchinnativelanguagebackground,andtheinteractionofthislistener±talkermatchormismatchwithtalkerpro®ciencyinthetargetlanguagei.e.,EnglishA.Method1.TalkersThetalkersselectedforthisexperimentwerefromthreelanguagebackgrounds:monolingualEnglish(1),non-talkersofEnglishwitha®rstlanguageofChi-nese(2),andNNtalkersofEnglishwitha®rstlanguageofKorean(2).Thefournon-nativetalkers'productionsweretakenfromtheNUFAESD.Thesetalkerswereselectedbasedontheir®rstlanguageeitherKoreanorChinese,gen-,andproductionintelligibilityscoresobtainedfromtheperceptualevaluationtestconductedatthetimeofthedatabasecompilation.FortheChineseandKoreantalk-ers,onefromeachlanguagebackgroundwasofhigherpro-®ciencyandoneoflowerpro®ciencyasde®nedbytheirproductionintelligibilityscores.TheChinesehigh-andlow-pro®ciencytalkershadintelligibilityscoresaveragepercentofkeywordscorrectlyidenti®edconvertedtotherauscaleof80and43rau,respectively.TheKoreanhigh-andlow- TABLEI.Frequencyandpsychologicalratingsforkeywords.VerbalAgeofacquisitionWrittenAverage12023158858752450339558226635104122Numberof1343713410610497Verbalfrequency:frequencyofoccurrenceinverballanguagefromBrownAgeofacquisition:agemultipliedby100toproducearangefrom100to700.Writtenfrequency:thenormsofKucera±FrancisConcreteness,familiarity,andimageabilityvaluesrange100to700.J.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t pro®ciencytalkershadscoresof90and55rau,respectively.Aneffortwasmadetomatchtheintelligibilityscoresforthehigh-pro®ciencyChineseandKoreanandthelow-pro®ciencyChineseandKorean.However,notalkersinoursamplewiththeselecteddemographicvariableshadidenticalproductionintelligibilityscores.Thetalkerswerenotmatchedforanyotherfeaturesoftheirspeechandwedidnotcontrolforspeakingrateineithertherecordingofthetalkersortheselectionofthetalkers.ThesameprocedureandequipmentwasusedtorecordthemonolingualEnglishtalkeraswasusedinthecompilationoftheNUFAESD.Noanalo-gousintelligibilityscoreisavailableforthenativeEnglishtalkersinceshewasnotincludedinanypreviousintelligibil-itytesting.2.ListenersAtotalof65adultswithnormalspeechandhearingparticipatedintheexperiment.Thelistenerscamefromfourdifferentlanguagebackgrounds:monolingualEnglish(21),non-nativespeakersofEnglishwitha®rstlanguageofChinese21),non-nativespeakersofEn-glishwitha®rstlanguageofKorean10),andnon-nativespeakersofEnglishwithnativelanguagebackgroundsotherthanChineseorKorean12).TheNN-mixedgroupincludedindividualsfromthefollowingnativelanguagebackgrounds:Bulgarian(Dutch(1),French/Douala(1),German(Greek(2),Hindi(1),Japanese(2),Serbian(1),Spanish(1),andTamil(1).Thenon-nativelistenerswererecruitedfromtheNorthwesternUniversityInternationalSummerInstituteandESLprogramoverthecourseofthesummerof2001andthe2001±2002academicyear.Alllistenerswerepaidfortheirparticipationinthestudy.Additionaldataforthethreegroupsofnon-nativelis-tenersareshownaboveinTableII.ThemonolingualEnglishlistenerswereallundergradu-atesatNorthwesternUniversityandrangedinagefrom17to22yearswithanaverageageof19.1years.ThenativeEn-glishlistenersweresigni®cantlyyoungerthanthenon-native(59).TheywererecruitedfromtheLinguisticsDepartmentsubjectpoolandreceivedcoursecreditfortheirparticipationinthestudy.Noneofthelistenersreportedanyspeechorhearingproblems.3.StimuliandprocedureSentencesfromthefourBKB-Rlistsatotalof60sen-weredividedinto®velistsof12sentencescovering37keywordseach11sentenceswith3keywordsand1sen-tencewith4keywords.SeetheAppendixforthesentencelistswiththekeywordsunderlined.Subjectswereseatedinfrontofacomputermonitorinasound-attenuatedbooth.Stimuluspresentationwascon-trolledbyspecial-purposeexperimentrunningsoftwareSUPERLABPRO2.01.Theaudio®leswereplayedoutthroughthecomputersoundcardSoundBlasterLiveoverhead-SennheiserHD580.Thesubject'staskwastolistentothesentencestimulusandtowritedownwhateversheorheheardonspeciallypreparedanswersheets.Aftereachtrial,thesubjectpressedabuttononaresponseboxaspartoftheSUPERLABPRO2.01toelicitthenexttrial.Eachtrialwaspresentedonlyonce,butsubjectscouldtakeaslongastheyneededtorecordtheirresponses.Eachsubjectheardall®vetalkersandall®vesentencelistsinablockedformat.ThemonolingualEnglishtalkerwasalwaysinthethirdposition.Thehigh-pro®ciencyChi-neseandKoreantalkerswereeither®rstorsecondandthelow-pro®ciencyChineseandKoreantalkerswereeitherfourthor®fth.TheNN-ChineseandmonolingualEnglishlistenersalwaysheardthetalkersinthefollowingorder:Chi-nesehighpro®ciency,Koreanhighpro®ciency,nativeEn-glish,Chineselowpro®ciency,Koreanlowpro®ciency.FortheNN-KoreanlistenerstheorderoftheChineseandKoreantalkerswasreversed.TherationalebehindthisorderingofthetalkerswastoensureconsistencyacrosstheNN-ChineseandNN-Koreangroupswithrespecttotheorderingofthelistener±talkernativelanguagematchandmismatch.Fur-thermore,thehigh-pro®ciencynon-nativetalkerswereor-deredbeforethenativetalkersothatsuperiorperformance TABLEII.Generalinformationaboutnon-nativelistenergroups.Meanandstandarddeviationsinparenthe-areshown.TheNorthwesternUniversityGraduateSchoolminimumforTOEFLis560or220differencesTOEFLn/a(TOEFLÐTimeinUS10.11TOEFLTestofEnglishasaForeignLanguage.ageofacquisition.lengthofacquisitioni.e.,Englishstudy1604J.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t onthenon-nativetalkerscouldnotbeattributedtoapracticeeffect.HalfoftheNN-mixedlistenergroupheardthelistsintheorderfortheNN-ChinesegroupwhiletheotherhalfheardthelistsintheorderfortheNN-Koreangroup.Theparticularsentencelistreadbyeachtalkerwascounterbal-ancedacrosslisteners.Inaseparatesessionaftertheperceptiontest,awordfamiliaritytestwasadministered.Forthistest,eachofthe144uniquekeywordsinthecompletesetofsentencesusedinthisstudywaspresentedtothesubjectsforafamiliarityratingonascaleof1to7where1``Idon'tknowthisword,''4``IrecognizethisasanEnglishwordbutIdon'tknowitsmeaning,''and7``Iknowthisword.''Asetof75®lleritemswasalsopresentedaspartofthistest.These®lleritemswereselectedfromlistsofwordsthatweregivenlow,medium,andhighfamiliarityratingsbynativelistenersinetal.andthatwereusedinprevioustestswithbothnativeandnon-nativelistenersBradlowandPisoni,1999.Theinclusionofthesewordsensuredthatthefullrangeofthefamiliarityscalewouldberepresentedbytheitemsinthistest.Anadditional128wordswerealsoincludedinthistestforthepurposeofaseparatestudynotreportedhere.Oneachtrial,thetargetwordwaspresentedinstandardAmericanEnglishorthographyonthecomputerSUPERLABPRO2.01,andthesubjectenteredhisorherfamiliarityratingbypressingtheappropri-atebuttononthekeyboard.Theitemremainedonthescreenuntilaresponsewasrecorded,whichthentriggeredthestartofthenexttrial.Theorderofpresentationoftheitemswas4.DataanalysisSentence-in-noiseperceptionscoresweredeterminedbyastrictkeyword-correctcount.Foreachsetofsentencesheardbyeachlistener,thetalkercouldreceiveascorefrom0to37keywordscorrect.Thisscorewasobtainedbycount-ingthenumberofkeywordstranscribedperfectly.Wordswithaddedordeletedmorphemeswerecountedasincorrect.However,obviousspellingerrorswerenotcountedasincor-rect.Rawintelligibilityscoreswereconvertedtopercentcor-rectandthentorationalizedarcsineunitsB.Results1.WordfamiliarityDatafromtheword-familiarityratingtaskshowedthatthevastmajorityofwordswashighlyfamiliartothevastmajorityofthenon-nativelisteners.Familiaritydatafromtwoofthe44listenersweremissing:onedidnotreturnfortheseconddatacollectionsessionwhentheword-familiaritytestwasadministeredandtheother'sdatahadtobedis-cardedduetoacomputererror.Oftheremaining42listen-ers,allgavehighratings5orgreatertoatleast94%ofthewords.Thirty-twolistenersgavescoresoflessthan5tonomorethantwowords;sixlistenersgavescoresoflessthan5tothreeto®vewords;fourlistenersgavescoresoflessthan5tosixtoninewords,andnolistenersgavescoresoflessthan5tomorethanninewords.Furthermore,therewasnocorrelationbetweenthelisteners'averagefamiliarityratingscoreandtheiraveragescoreonthesentence-in-noiseper-ceptiontest(rho0.47).Avastmajorityofthewords,79%,wasgivenaratingof7byallsubjects.Addi-tionally,only®vewordshadaveragescoreslowerthan6:,buckets,janitor,jug,andsauce-.Ofthe144targetitems,only19wordsweregivenanyscoreslowerthan5.Ninewordsweregivenascoreunder5byonelistener;threewordsweregivenscoresunder5bytwolisteners;andsevenwordsweregivenscoreslowerthan5bymorethantwolistenersfrom5to18listenersLast,thenon-natives'scoresonthelow-,mid-,andhigh-familiarity®lleritemsweresimilartoscoresgivenbynativelistenersinBradlowandPisoni.Thelow-,mid-,andhigh-familiarity®lleritemsweregivenscoresof1.83,3.88,and6.93,respectively,bythenativelistenersinBradlowandPisoni'sstudyandweregivenscoresof2.85,3.79,and6.53,respectively,bythenon-nativelistenersinthecurrentstudy.Therefore,weperformedallanalysesofthesentence-in-noiseperceptiontestwiththeassumptionthatthenon-nativelistenerswereallsuf®cientlyfamiliarwiththekeywordstoensurethatthistestprovidedavalidmeasureoftheirabilitytoperceivesentencesinnoiseindependentlyofwordfamil-iarity.2.Foreign-accentedEnglishperceptionTableIIIsummarizesthe®vetalkers'intelligibilityscoresforeachofthefourlistenergroups.ArepeatedmeasuresANOVAwithlistenernativeEn-glish,NN-Chinese,NN-Korean,NN-mixedasthebetween-subjectsfactorandtalkerhigh-pro®ciencyChinese,low-pro®ciencyChinese,high-pro®ciencyKorean,low-pro®ciencyKorean,nativeEnglishasthewithin-subjectsfactorshowedhighlysigni®cantmaineffectsoflistener(3,240)andtalker(4,240).Theinteractionoftalkerandlistenerwashighlysigni®cant(12,240).Foreachlistenergroupthelow-pro®ciencyChineseandKoreantalkerswerelessintelligiblethanthenativetalkerandthehigh-pro®ciencyChineseandKoreantalkers.However,therankingsforthetwohigh-pro®ciencynon-nativesandthenativetalkerdependedonthelanguagebackgroundofthelistenergroup.PosthocpairwisecomparisonsBonferroni/Dunntestsoftalkerintelligibilitywithineachlistenergroupwerecon-ducted.Duetothelargenumberofpairedcomparisonsforeachlistenergroup,thevaluemustbelessthan0.005tobesigni®cant.Pairwisecomparisonswithinatalkeracrosslistenergroupswerenotconductedasthosecomparisonswerenotofprimaryinterestforthisstudy.FortheNN-Chineselisteners,thehigh-pro®ciencyChinesetalkerwasnotsigni®cantlydifferentfromthehigh-pro®ciencyKoreantalkerorthenativeEnglishtalker.However,thehigh-pro®ciencyKoreantalkerwassigni®cantlymoreintelligiblethanthenativeEnglishtalker(0.001).Allthreeofthesetalkersweresigni®cantlymoreintelligiblefortheselistenersthanthelow-pro®ciencyChineseandKoreantalkers(0.001).FortheNN-Koreanlisteners,therewerenosig-ni®cantdifferencesbetweenthehigh-pro®ciencyChinese,high-pro®ciencyKorean,andnativeEnglishtalkers.Addi-J.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t tionally,thehigh-pro®ciencyChineseandnativeEnglishtalkerswerenotsigni®cantlydifferentfromthelow-pro®ciencyKoreantalker.However,thehigh-pro®ciencyKoreantalkerwassigni®cantlymoreintelligiblethanthelow-pro®ciencyKoreantalker(0.001).FortheNN-mixedlistenergroup,thehigh-pro®ciencyChinese,high-pro®ciencyKorean,andthenativeEnglishtalkerswerenotsigni®cantlydifferentfromoneanother,andwereallsigni®-cantlymoreintelligiblethanthelow-pro®ciencyChineseandlow-pro®ciencyKoreantalkers(0.0001).ForthenativeEnglishlisteners,thenativeEnglishtalkerwassigni®cantlymoreintelligiblethanalltheothertalkers(0.001).Insummary,thenativelistenersfoundthenativetalkermoreintelligiblethananyoftheothertalkers,andalllisten-ersgenerallyfoundthelower-pro®ciencynon-nativetalkerslessintelligiblethaneithertheirhigh-pro®ciencycounter-partsorthenativetalker.Ofparticularinterestforthisstudywasthe®ndingthatthenon-nativelistenersfoundthehigh-pro®ciencynon-nativetalkerswithwhomtheysharednativelanguageasintelligibleasthenativeEnglishtalker.This®ndingalsoextended,inonecase,toalow-pro®ciencynon-nativetalkerinthatthelow-pro®ciencyKoreantalkerwasasintelligibleasthenativeEnglishtalkerfortheNN-Koreanlistenergroup.This®ndingdemonstratesamatchedinterlan-guagespeechintelligibilitybene®t,suchthatanativelan-guagematchbetweenanon-nativetalkerandanon-nativelistenerfacilitatesspeechintelligibility.Furthermore,whenthenon-nativelistenersandhigh-pro®ciencynon-nativetalk-ersdidnotshareanativelanguage,thenon-nativelistenersfoundthenon-nativetalkersequallyasormoreintelligiblethanthenativetalker.Thisgeneral®ndingsuggeststhattheinterlanguagebene®tcanextendtothesituationofatalker±listenernativelanguagemismatch,demonstratingamatchedinterlanguagespeechintelligibilitybene®tToinvestigatethepossiblecontributionofspeakingratetotheintelligibilityresults,wemeasuredsentencedurationforeachofthe®vespeakers.Allpairwisecomparisonsamongstthe®vetalkersweresigni®cant(0.005)exceptforthehigh-pro®ciencyKoreanandthelow-pro®ciencyChi-nesetalker,whoseaveragesentencedurationsdidnotdiffersigni®cantly.Theaveragesentencedurationsforthe®vetalkersareshowninTableIV.Foreachofthenon-nativelistenergroups,theaveragesentencedurationsforthe®vetalkersdidnotsigni®cantlycorrelatewiththeirintelligibilityscoresSpearmanrankcor-relation,NN-Chineselisteners:rho0.55;NN-Koreanlisteners:rho0.84;NN-Mixedlisteners:0.84;NativeEnglishlisteners:rho0.55).Forexample,thelow-pro®ciencyKo-reantalkerhadlongersentencedurationsthanthehigh-pro®ciencyKorean,yetalmostallthenon-nativelistenersexceptforthreeNN-Koreanlistenersfoundthehigh-pro®ciencyKoreanmoreintelligiblethanthelow-pro®ciencyKorean.Thisanalysissuggeststhattheobservedinterlanguageintelligibilitybene®tisnotsimplyduetovari-abilityinspeakingrateacrossthetalkers.IV.SUMMARYANDDISCUSSIONFourgroupsoflistenersmonolingualEnglish,NN-Chinese,NN-Korean,andNN-mixedtranscribedsentencesproducedbyanativeEnglishtalker,twoChinesenon-nativetalkersofEnglish,andtwoKoreannon-nativetalkersofEn-glish.Themajor®ndingofthisstudywasthattherelativeintelligibilityofeachtalkerdependedonthelanguageback-groundofthelistenersuchthat TABLEIII.Percentkeywordscorrectlytranscribedfortalkersandlistenergroupsinrau.Standarddeviationsareshowninparentheses.Scoresinboldaresigni®cantlyhigherthantheotherscoresintherowandarenotsigni®cantlydifferentfromoneanotherexceptfortheunderlinedscores.Speci®cally,fortheNN-Chineselisteners,thehigh-pro®ciencyKoreantalkerissigni®cantlymoreintelligiblethanthenativeEnglish,andfortheNN-Koreanlistenersthehigh-pro®ciencyKoreantalkerissigni®cantlymoreintelligiblethanthelowpro®-ciencyKoreantalker.Talkerpro®ciencyAlltalkers 11.7 …56 „10.4 …30~12.1!41~9.8!51~17.4!NN-Koreann5„15.5…74„15.8 11.711.4 11.0NativeEnglishAlllisteners67 TABLEIV.Averagesentencedurationsforthe®vetalkersinorderofin-creasingduration.Meansinmillisecondsandstandarddeviationsinparen-areshown.AveragesentenceNativeEnglish1223msLow-pro®ciencyChinese1512msHigh-pro®ciencyKorean1587msLow-pro®ciencyKorean1680msHigh-pro®ciencyChinese1717ms1606J.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t Fornativelisteners,intelligibilityofthenativetalkerwasgreaterthantheintelligibilityofanyofthenon-nativetalkers.Fornon-nativelisteners,intelligibilityofahigh-pro®ciencynon-nativetalkerandinonecasealow-pro®ciencytalkerfromthesamenativelanguageback-groundwasequaltotheintelligibilityofthenativetalker.Thisisthe``matchedinterlanguagespeechintel-ligibilitybene®t.''Fornon-nativelisteners,intelligibilityofahigh-pro®ciencynon-nativetalkerfromadifferentnativelan-guagebackgroundwasgreaterthanorequaltotheintel-ligibilityofthenativetalker.Thisisthe``mismatchedinterlanguagespeechintelligibilitybene®t.''Thematchedinterlanguagespeechintelligibilitybene®tcanbeexplainedbythefactthatnon-nativespeechproduc-tionandperceptionarebothsystematicallylinkedtonativelanguagesoundstructureforawealthofsupportingresearchandtheoryseeStrange,1995;Best,1994,1995;Flege,1992,1995;KuhlandIverson,1995andmanyothers.Thus,thespeechofanon-nativetalkerismoreintelligibletonon-nativelistenerswithwhomtheyshareanativelanguagethanfornativelistenersduetothefactthattheoverallsharedphoneticandphonologicalknowledgebetweenthenon-nativetalkerandnon-nativelistenersfromthesamelan-guagebackgroundislikelytobemoreextensivethananative/non-nativepair.Forthenon-nativeswhoshareana-tivelanguage,theirlinguisticknowledgecoversaspectsofboththenativeandtargetlanguages,whereasforthenon-native/nativepairthesharedknowledgebaseincludesonlytheirknowledgeofthetargetlanguageinsofarasitisdevel-opedinthenon-nativetalker.Thissharedknowledgebaseincludesthesystemofconsonantandvowelcategories,pho-notactics,stresspatterns,andintonationaswellasotherfea-turesofthesoundsystem.Thus,anon-nativelisteneriswellequippedtointerpretcertainacoustic±phoneticfeaturesofthespeechofanative-language-matchednon-nativetalkerasthetalkerintendedthemtobeinterpreted,eventhoughtheymaydeviatemarkedlyfromthetargetlanguagenorm,whereasnativelistenersarebetterequippedtointerpretthespeechofanativetalker.Forexample,evenforeign-accentedtalkerswhohavegainedcontroloverproducingthefullin-ventoryofvowelcontrastsofthetargetlanguagemaypro-ducethevowelsofthetargetlanguageintheregionofthevowelspacei.e.,withabaseofarticulationthatistypicalofthenativelanguageratherthanofthetargetlanguage.Whilethismayresultinmisinterpretationofaparticularvowelfornativelistenersornon-nativelistenersfromadif-ferentnativelanguagebackground,non-nativelistenersfromthesamenativelanguagebackgroundasthetalkerwillbemorelikelytoaccessthecorrectvowelcategory,therebycontributingtothematchedinterlanguagespeechintelligibil-itybene®tthatweobservedinthisstudy.Apossibleexplanationforthemismatchedinterlanguagespeechintelligibilitybene®tisthatitresultsfromthetalker'sandlistener'ssharedknowledgeofthestructureofthetargetlanguageinconjunctionwiththein¯uenceofgeneralstrate-giesthatlistenersandtalkersapplywhenlearningtoproduceandperceiveaforeignlanguage.Forexample,innative-accentedAmericanEnglishword-®nalstopconsonantsarefrequentlyunreleased.Nativelistenersknowtolistenforcuestothepresenceandidentityofword-®nalconsonantsinotherpartsofthesignalandtointerpretthelackofasalientword-®nalstopreleaseasre¯ectingstructuraland/orcontex-tualin¯uencesatthephraseanddiscourselevels,butnon-nativelistenersmaymissthesecues,therebycompromisingtheiroverallcomprehensionofnative-accentedEnglish.However,intheirownEnglishspeech,non-nativetalkerswhohavenotyetmasteredallthedetailsofAmericanEn-glishallophonymayproduceparticularlysalientword-®nalstopconsonantreleases,therebyfacilitatingtheintelligibilityoftheirspeechforothernon-nativelistenersfromawiderangeofnativelanguagebackgroundsthoughnotnecessar-ilyfornativelisteners.Forexample,Smith,Bradlow,anddemonstratedthatnon-nativelistenersfromvariousnativelanguagebackgroundsarebetteratidentifyingwordsinminimalpairsthatdifferonlyinthevoicingofthe®nalstope.g.,capvscab,pickvspigproducedbynon-nativetalkersthanproducedbynativetalkers.Inthiscase,eventhoughthenativelistenersgenerallyperformedbetterthanthenon-nativelisteners,theperformanceofthenon-nativelistenerssurpassedthatofthenativelistenersforwordsproducedbyonenon-nativetalker,indicatingthatthenon-nativelistenersmustbelisteningforcertaincuesthatnativelistenersarenotattendingto.Alternatively,themismatchedinterlanguagespeechin-telligibilitybene®tthatweobservedinthisstudymaystemfromsimilaritiesinthesoundstructureofthetwolanguagesthatweinvestigated,inwhichcaseitisinfactjustanothermanifestationofthematchedinterlanguagebene®tratherthanaseparatephenomenon.Forexample,bothChineseandKoreanhaveamuchmoreconstrainedsyllablestructurethanEnglish,includingaconstraintagainst®nalconsonantclus-ters.Thetransferfromalanguagewhichdoesnotallow®nalconsonantclusterssuchasChineseorKoreantoalanguagethatdoesallowcodaclusterssuchasEnglishmayresultinsimilarfeaturesofChinese-accentedandKorean-accentedEnglishwhichservetofacilitateperceptionofEnglishspeechbetweennativeChineseandnativeKoreanlisteners.However,evidenceagainstthisaccountbasedonstructuralsimilaritiesbetweenChineseandKoreancomesfromthere-sultsofthetestwiththeNN-mixedlistenergroupwhichincludedverysmallnumberoflistenersonly1or2eachofawiderangeoflanguagebackgrounds.Sevenofthese12listenersfoundthehigh-pro®ciencyChinesetalkerequallyormoreintelligiblethantheEnglishtalker,and®veofthe12listenersfromthishighlyheterogeneouslistenergroupfoundthehigh-pro®ciencyKoreantalkermoreintelli-giblethanthenativeEnglishtalker.TheselistenersthosefromtheNN-mixedgroupwhoshowedamismatchedinterlanguagespeechintelligibilitybene®tcamefromadi-versegroupofnativelanguagebackgroundsincludingBul-garian,Dutch,French/Douala,Greek,Hindi,Japanese,andSpanish.This®ndingsuggeststhattheob-servedmismatchedinterlanguagespeechintelligibilityben-e®tisunlikelytobetheresultofstructuralsimilaritiesbe-tweenthenativelanguagesofthetalkersandlisteners,butJ.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t ratherduetocertaintendenciesinforeign-accentedEnglishregardlessofnativelanguagebackground.Nevertheless,datafromstudieswithadditionalcarefullyselectedlanguagesareneededinordertoruleoutthisalternativeaccount.Fromthecurrentstudy,wecannotdetermineatwhichstageofspokenlanguageprocessingtheinterlanguageben-e®tarisessinceperceivingwordsinsentencesrequirespro-cessingonmanydifferentlevelsandourtaskwasanoff-linemeasureofcomprehensionthatrepresentstheaccumulationofprocessingatmultiplelevels.Becausewecontrolledforthelinguisticcontentofthesentences,wecanassumethattheadvantagethatmanyofthenon-nativelistenersreceivedwhenlisteningtonon-nativespeechcamefromthediffer-encesintheacousticsignalandnotfromdifferencesinlexi-calchoicesorsyntacticstructures.Therefore,itisverylikelythatthesourceoftheobservedinterlanguageeffectwasatarelativelyearly,phoneticstageofprocessing.Nevertheless,itispossiblethattheinterlanguagebene®talsooperatesathigherlevelsofsentencecomprehensionwhereotheraspectsoflinguisticstructureandofextralinguisticfactorscomeintoplay.Additionalteststhatspeci®callytapintovariouslevelsofspokenlanguageprocessingwillberequiredtodeterminewhethertheearlyprocessesaremostimportantfortheinter-languagebene®torifthelocalizationofthephenomenonoccurslaterintheprocessingofspokenlanguageaswell.Largeindividualdifferenceswerefoundinthemagni-tudeoftheinterlanguagebene®t.Thedifferenceinintelligi-bilitybetweenthehigh-pro®ciencynon-nativetalkerandthenativetalkerrangedfrom23to52rauforthematchedinterlanguagebene®tandfrom24to39rauforthemis-matchedinterlanguagebene®t.Thebasisoftheselargeindi-vidualdifferencesremainsunknown.Thesubjectsinthisstudywereratherhomogeneousintermsofdemographice.g.,ageofEnglishstudyonset,lengthofresi-dence,etc.andintermsofEnglishpro®ciencye.g.,TOEFLscoresrangedonlyfrom573±677and220±290.Futureresearchwillbeneededtodeterminethesourceofthesedifferences.Forexample,independenttestsoflanguagepro®ciency,particularlywithrespecttotargetlanguagespeechperception,willhelptodeterminehowlis-tenerpro®ciencyinthenon-nativelanguagein¯uencesthepreferencefornon-nativeovernativespeech.The®ndingsfromthepresentstudyareconsistentwiththe®ndingsofvanWijngaardenandvanWijngaardenetal.,whichdemonstratedanon-nativetalkerspeechintelligibilityadvantagefornon-nativelistenersatarela-tivelyearlystageoftargetlanguageacquisition.Twoimpor-tantmethodologicaldifferencesbetweenthepresentstudyandthestudiesbyvanWijngaardenandcolleaguesarethetargetlanguageEnglishvsDutchandtheintelligibilitymeasurementtechnique:vanWijngaardenandcolleaguesmeasuredthesignal-to-noiseratiorequiredfora50%-correctthespeechreceptionthreshold,orSRT,whereasthepresentstudymeasuredintelligibilityintermsofakey-wordrecognitionaccuracyscoreforsentencespresentedata®xedsignal-to-noiseratio.Takentogether,thepresentstudyandthoseofvanWijngaardenandcolleaguesprovidecon-vergingevidenceforaninterlanguagespeechintelligibilitybene®tanddemonstratethatanymeasureofspeechintelli-gibilitymusttakeintoaccountbothtalker-andlistener-relatedfactors.Weconcludebynotinganimplicationofthepresent®ndingsforlanguagevariationandchange.Thedemonstra-tionoftheinterlanguagespeechintelligibilitybene®tsug-gestsamechanismthatmayunderlietheestablishmentofnewpronunciationnormsacrossacommunityofnon-nativetalkers.Duetovarioussocial,political,andhistoricalcir-cumstances,non-nativetalkerswhoshareanativelanguagemaysometimescommunicateinasharedforeignlanguage.Thissituationcanariseinsettingswherethesharedforeignlanguagedominatesinthebroadercontext.Forexample,inmanyuniversityresearchlaboratorieswherethedirectorandexperimentersshareanativelanguage,thetypicallanguageofthelaboratorymaybeEnglishduetothefactthatthebroaderscienti®ccommunityisEnglishdominated.Thissituationcanoccuronanevenlargerscalesuchasinacoun-trylikeIndia,whereEnglishiswidelyspokenasasecondlanguageand,incertainsettings,eventalkersfromthesamenativelanguagebackgroundwillcommunicateinEnglish.Undersuchcircumstances,acharacteristicvarietyofthetar-getlanguagee.g.,IndianEnglishmaydevelopasaresultoftheinterlanguagespeechintelligibilitybene®twhichwillcausecertaininterlanguagefeaturestobecome®rmlyen-trenchedinthespeechofthecommunity.Thisintelligibility-basedfactorwilllikelyoperateinconjunctionwithothersociolinguisticfactorstoreinforcetheestablishmentofanewandlastingvarietyofthetargetlanguage.WearegratefultoMartinArmstrong,LylaMiller,MengtingShieh,andSaundraWrightfordatacollectionas-sistance.ThisresearchwassupportedbyNIH-NIDCDGrantNo.DC03762.APPENDIX:FOREIGN-ACCENTEDENGLISHPERCEPTIONTESTMATERIALSBasedonBKB-RmaterialsBamfordandWilson,1979;BenchandBamford,1979List1 children droppedthe bag.~2!The dog came back.~3!The ¯oor looked clean.~4! She foundher purse.~5!The fruit isonthe ground.~6! Mother gota saucepan.~7! They washedin cold water.~8!The young peopleare dancing.~9!The bus left early.~10!The ballis bouncingvery high.~ Father forgotthe bread.~12!The girlhasa picture List2 boy forgothis book.~2!A friend camefor lunch.~3!The match boxesare 1608J.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t ~4! He climbedhis ladder.~5!The family boughta house.~6!The jug isonthe shelf.~7!The ball brokethe window.~8! Theyare shoppingfor cheese.~9!The pond wateris dirty.~10! They hearda funny noise.~The policeare clearingthe road.~12!The bus stopped List3 book tellsa story.~2!The young boy left home.~3! Theyare climbingthe tree.~4! She stoodnearher window.~5!The tablehas three legs.~6!A letter fellonthe ¯oor.~7!The ®ve menare working.~8!The shoeswere very dirty.~9! They wentona vacation.~10!The baby brokehis cup.~The lady packedher bag.~12!The dinner plateis List4 dish towelisbythe sink.~2! She looked inher mirror.~3!The good boyis helping.~4! They followedthe path.~5!The kitchen clockwas wrong.~6! Someoneis crossingthe road.~7!The mailman broughta letter.~8! Theyare ridingtheir bicycles.~9! He brokehis leg.~10!The milkwasbythe front door.~The shirtsare hanginginthe closet.~12!The chicken laidsome List5 orangewas very sweet.~2! Heis holdinghis nose.~3!The new roadisonthe map.~4! She writestoher brother.~5!The football player losta shoe.~6!The three girlsare listening.~7!The coat isona chair.~8!The trainis moving fast.~9!The child dranksome milk.~10!The janitor useda broom.~The groundwas very hard.~12!The buckets hold de®nedaninterlanguageas``aseparatelinguisticsystembasedontheobservableoutputwhichresultsfromalearner'sattemptedproductionofatargetlanguagenorm''p.214AllChinesesubjectswerenativespeakersofMandarinChineseinthesensethatallschoolingincludingUniversityhadbeeninMandarin.However,manyoftheChinesesubjectsspokeanotherdialectofChineseathome.Bamford,J.,andWilson,I..``MethodologicalconsiderationsandpracticalaspectsoftheBKBsentencelists,''inSpeech-hearingTestsandtheSpokenLanguageofHearing-impairedChildren,editedbyJ.BenchandJ.BamfordAcademic,London,pp.148±187.Bench,J.,andBamford,J.,EditorsSpeech-hearingTestsandtheSpokenLanguageofHearing-impairedChildrenAcademic,LondonBest,C.T..``Theemergenceofnative-languagephonologicalin¯u-encesininfants:Aperceptualassimilationmodel,''inTheDevelopmentofSpeechPerception:TheTransitionfromSpeechSoundstoSpokenWordseditedbyJ.GoodmanandH.C.NusbaumMITPress,Cambridge,MApp.167±224.Best,C.T..``Adirectrealistviewofcross-languagespeechpercep-tion,''inSpeechPerceptionandLinguisticExperience:IssuesinCross-languageResearch,editedbyW.StrangeYork,Baltimore,pp.171±204.Best,C.,McRoberts,G.,andGoodell,E..``Discriminationofnon-nativeconsonantcontrastsvaryinginperceptualassimilationtothelisten-er'snativephonologicalsystem,''J.Acoust.Soc.Am.,775±794.Bradlow,A.R.,Pisoni,D.B.,Yamada,R.A.,andTohkura,Y.``TrainingJapaneselistenerstoidentifyEnglish/r/and/l/.IV.Someef-fectsofperceptuallearningonspeechproduction,''J.Acoust.Soc.Am.,2299±2310.Bradlow,A.R.,andPisoni,D.B..``Recognitionofspokenwordsbynativeandnon-nativelisteners:Talker-,listener-,anditem-relatedfac-tors,''J.Acoust.Soc.Am.,2074±2085.Brown,G.D.A..``Afrequencycountof190,000wordsintheLondon-LundcorpusofEnglishconversation,''Behav.Res.MethodsIn-,502±532.Flege,J.E..``Speechlearninginasecondlanguage,''incalDevelopment:Models,Research,andApplications,editedbyC.Fer-guson,L.Menn,andC.Steol-GammonYork,Timonium,MD,pp.565±Flege,J.E..``Secondlanguagespeechlearning:Theory,®ndings,andproblems,''inSpeechPerceptionandLinguisticExperience:IssuesinCross-languageResearch,editedbyW.StrangeYork,Baltimore,pp.233±277.Flege,J.E..``Interactionsbetweenthenativeandsecond-languagephoneticssystems,''inAnIntegratedViewofLanguageDevelopment:PapersinHonorofHenningWode,editedbyP.Burmeister,T.Piske,andA.RohdeTrier,WissenschaftlickerTrierFlege,J.E.,MacKay,I.R.A.,andMeador,D..``NativeItalianspeakers'perceptionandproductionofEnglishvowels.''J.Acoust.Soc.,2973±2987.Imai,S.,Flege,J.E.,andWalley,A..``Spokenwordrecognitionofaccentedandunaccentedspeech:Lexicalfactorsaffectingnativeandnon-nativelisteners,''inProceedingsoftheInternationalCongressonPhoneticScience,Barcelona,Spain.Kucera,H.,andFrancis,W.N.ComputationalAnalysisofPresent-DayAmericanEnglishBrownUniversityPress,ProvidenceKuhl,P.K.,andIverson,P..``Linguisticexperienceandthepercep-tualmagneteffect,''inSpeechPerceptionandLinguisticExperience:Is-suesinCross-languageResearch,editedbyW.StrangeYork,Baltimorepp.121±154.Lewellen,M.J.,Goldinger,S.D.,Pisoni,D.B.,andGreene,B.G.``Lexicalfamiliarityandprocessingef®ciency:Individualdifferencesinnaming,lexicaldecisionandsemanticcategorization.''J.Exp.Psychol.,316±330.Lively,S.E.,Logan,J.S.,andPisoni,D.B..``TrainingJapaneselistenerstoidentifyEnglish/r/and/l/.II.Theroleofphoneticenviron-mentandtalkervariabilityinlearningnewperceptualcategories,''J.Acoust.Soc.Am.,1242±1255.Lively,S.E.,Pisoni,D.B.,Yamada,R.A.,Tohkura,Y.,andYamada,T..``TrainingJapaneselistenerstoidentifyEnglish/r/and/l/.III.Long-termretentionofnewphoneticcategories,''J.Acoust.Soc.Am.2076±2087.Logan,J.S.,Lively,S.E.,andPisoni,D.B..``TrainingJapaneselistenerstoidentifyEnglish/r/and/l/:A®rstreport,''J.Acoust.Soc.Am.,874±886.Mayo,L.H.,Florentine,M.,andBuus,S..``Ageofsecond-languageacquisitionandperceptionofspeechinnoise,''J.SpeechLang.Hear.Res.,686±693.McAllister,R..``Perceptualforeignaccent:L2user'scomprehensionability,''inProceedingsofthe1990AmsterdamSymposiumontheAc-quisitionofSecond-languageSpeech,pp.216±228.Meador,D.,Flege,J.E.,andMacKay,I.R.A..``FactorsaffectingtheJ.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t recognitionofwordsinasecondlanguage,''Bilingualism:Lang.Cogni.55±67.Munro,M.J.,andDerwing,T.M..``Foreignaccent,comprehensi-bility,andintelligibilityinthespeechofsecondlanguagelearners,''Lang.,73±97.Munro,M.J..``Theeffectsofnoiseontheintelligibilityofforeign-accentedspeech,''Stud.SecondLang.Acquisit.,139±153.Nash,R..``IntonationalinterferenceinthespeechofPuertoRicanbilinguals,''J.English,1±42.Pallier,C.,Colome,A.,andSebastian-Galles,N..``Thein¯uenceofnative-languagephonologyonlexicalaccess:Exemplar-basedversusab-stractlexicalentries,''Psychol.Sci.,445±449.Rochet,B..``Perceptionandproductionofsecond-languagespeechsoundsbyadults,''inSpeechPerceptionandLinguisticExperience:IssuesinCross-languageResearch,editedbyW.StrangeYork,Baltimore,pp.379±410.Sebastian-Galles,N.,andSoto-Faraco,S..``Onlineprocessingofnativeandnon-nativephonemiccontrastsinearlybilinguals,''Cognition,111±123.Selinker,L..``Interlanguage.''IRAL,209±231.Smith,B.,Bradlow,A.R.,andBent,T..``Productionandperceptionoftemporalcontrastsinforeign-accentedEnglish,''inProceedingsoftheInternationalCongressonPhoneticScience,Barcelona,Spain.Smith,L.E.,andRa®qzad,K..``Englishforcross-culturalcommu-nication:Thequestionofintelligibility,''TESOLQuarterly,371±380.Strange,W.,EditorSpeechPerceptionandLinguisticExperience:IssuesinCross-languageResearchYork,BaltimoreStudebaker,G.A..``Arationalizedarcsinetransform,''J.SpeechHear.Res.,455±462.vanWijngaarden,S.J..``Intelligibilityofnativeandnon-nativeDutchspeech.''SpeechCommun.,103±113.vanWijngaarden,S.J.,Steeneken,H.J.M.,andHoutgast,T.``Quantifyingtheintelligibilityofspeechinnoisefornon-nativelisten-ers.''J.Acoust.Soc.Am.,1906±1916.Weinreich,U.LanguagesinContact:FindingsandProblemston,TheHague1610J.Acoust.Soc.Am.,Vol.114,No.3,September2003T.BentandA.R.Bradlow:Interlanguageintelligibilitybene®t