Transcription of VOICING IN ENGLISH: A CATALOG OF ACOUSTIC FEATURES ...
1 " VOICING " ,d,glandIp,t,klarenowusuallyreferredtoas voicedandvoicelessstops,respectively,alt houghitisrecognizedthatmembershipinthese setsisnotentirelydeterminedbywhether,acc ordingtocommonlyaccepteddefinitions,agiv enphoneticelementisvoicedorvoiceless; ,thenitwillbeassignedtotheIb,d,glset,and ifitisavoicelessstop,itmay,butneednotbe, assignedtoIp,t,k/.Acontextinwhichthestop membersofthetwophonologicalsetsmaybedist inguishedsimplyonthebasisofvoicing(asnar rowlydefinedwithrespecttostopconsonants) isbetweenvowels, ,however, , ,thatis, , ,thoughitisofcoursetruethatdetailsofthea erodynamic, *LanguageandSpeech,(1986), [HASKINSLABORATORIES:StatusReportonSpeec hResearchSR-86/87(1986)J45 Lisker.]
2 " VOICING " , ( ,theInternationalPhoneticAssociation) "voiced"iftheclosureismarkedbylaryngealb uzz,anditis"voiceless" ,andthatauditoryjudgmentandacousticrecor dmaynotalwaysagree, *[bJ-[pJGiventhespellingconventionsandth edefinitionofstopvoicingtowhichlinguists appeargenerallytosubscribe,aphoneticunit representedas[bJisavoicedstop,while[ ,forexample,Dutch,Italian, [bJand[pJ,thatis, ,amongthemEnglish,therearephonologicalca tegories,alsorepresentedasIblandIpl,whos erelationtothephoneticcategories[bJand[ "voiced"setarenotinvariablyvoiced,thatis ,Iblmaybeinitially[pJ(thoughmoreofteniti sspelled"phonetically[bJ,withnoclearindi cationofwhetherthepreferenceforthelatter Ospellingisdictatedbyphoneticorphonologi calconsiderations),andprepausallyaswelli ts"voicelessnessismarked,"thatis,readily detectedbyear(Trager&Smith,1951),itfollo wsthatthesearchfortheacousticpropertiesc ueingtheIb/- [ voicedJdifference, "problem"ofEnglishstopvoicingresideslarg elyinthefactthattheObservedvariabilityof theIb/- ~[bJ-[pJAcontextinwhichthecontrastbetwee nstopmembersofEnglishIblandIplseemsmostn earlytobeoneinvolving[bJvs.]]]]]]]]]]]]] ]
3 [pJismediallyinwordsbeforeanunstressedsy llable, ,then, ,forexample,rabidandrapiddifferinstopvoi cingalone,thentheacousticpropertiesaffec tingtheiridentificationby46 Lisker:" VOICING " ,ofthetwootherfeaturesthathavetraditiona llyfiguredinaccountsoftheEnglishstops,[ aspirated]and[ fortis].thereisgeneralagreementthatthefi rstoftheseplaysnosignificantroleindiffer entiatingrabidandrapid,atleastinAmerican ifnotinstandardsouthernBritishEngli~rons tein,1960;Jones,1956;Trager&Smith,1951). Asforthesecond,asidefromitscontroversial natureasaphoneticfeatureonaparwiththeoth ers(Lisker,1963), ,forTragerandSmith(1951)the[p]ofrapidisf ortis,whileHeffner(1950)followsJespersen indescribingtheAmericanpronunciationof/ ,ifthedurationaldifferencesinclosureandp reclosureintervalsbetweenrabidandrapidar econstruedasevidenceofa[ fortis]distinction,thenitmustbegrantedth atnotalltheacousticcuestothelexicaldisti nctioncanbe,strictlyspeaking,cuesto[ voiced].]
4 Despitethesestrictures, ,althoughinmedialpositionthephoneticdiff erencebetweenEnglish/b/and/p/maywellbesm allerthanelsewhere,thenumberofreadilyiso latedacousticpatternpropertieswhosevaria tionmightbeexpectedtoaffecttheidentifica tionofastimulusasrabidorrapidislarger.(I tfarexceedsthesixlistedinKlatt,1975,forw ord-initialbututterance-medialintervocal icposition,andisinfactmore,bytwo,thanthe fourteenlistedbyEdwards,1981,forthesamep osition,wherethephoneticbasisforthe"voic ingdistinction"ispossiblymaximal.)Howeve r, ,sinceutterance-initialstopscannotbecued bypropertiesoftheintervalprecedingclosur e(exceptforthepre-speechsilence), ,orcanplausiblybesupposedtocuetheidentif icationofaformasrabidorrapid,sevenaretob efoundinthesignalprecedingthemedialclosu re,threeareclosureproperties, :Closure1)durationofclosure2)durationofg lottalsignal3)intensityofglottalsignalPr e-closure4)durationofvowel5)durationoffi rst-formant(F1)transition6)F1offsetfrequ ency7)F1transitionoffsettime(Le.
5 ,"F1cutback,"or,moreprecisely,"F1cutforw ard")47 Lisker:" VOICING "inEnglish8)timingofvoice offset9)fundamentalfrequency(Fe)contour1 0) )releaseburstintensity12)timingofvoiceon set(vaT)13)onsetofF1transition("F1"cutba ck")14)F1onsetfrequency15)F1transitiondu ration16)FecontourThislistdoesnotfullyex hausttheinventoryofpropertiesthatpossibl yaffectlisteners'labelingbehavior,forwem ightimaginethatfactorscontributingtothe" prominence"ofthesecondsyllablerelativeto thefirst( ,thestresscontourattributedtotheform) ,asaresultofacousticalterationseffecting astressshift,beperceivedtoincludeIblrath erthanIpl,sinceanaturaltokenofthederivat iveofthefirstword,rapidity,callsforthevo icelessaspirate[ph],whereasa[p] ,forexample,higherformants,maymakecontri butionstolexicalidentity,eventhoughsuche ffectsmightnotbereadilyexplained(Lisker, 1975)
6 Intheinventoryjustlistedsixteenacousticp ropertieswereenumerated,andseveralmoresu ggested,buttheprecisenumbercannotbetaken veryseriously, ,atleastforpurposesofexperimentation,the ymaynotbeacousticallydistinct, ,forexample,property#12mightbeanalyzedas twoproperties,voice-onsettimeandaspirati on(followingKlatt,1975),sinceadelayinvoi ceonsetcanbeaccompaniedbyasilentinterval (perejectivearticulation) ,items#2and#8arecountedastworatherthanon e,notonanacousticbasis,butonlybecauseofa priorsegmentationofthespeechpatternswher ebytheteststimuliwerepartitionedintopre- closure,closureandpost-closureintervals. (Asimilarsegmentationunderliesthecommond istinctiondrawnbetweenthephoneticfeature sofstopvoicingandvoicelessaspirationinEn glishandsomeotherlanguages,andtheirsubse quenttreatmentasindependentpropertiesofs topconsonants.)
7 AcousticPropertiesasContext-variableLexi calCuesOftheabove-listedacousticproperti esthatmightaffecttheidentificationofasig nalasrabidorrapid,itisprobablytruethatno neisindispensible, ,areportedrabidneednotmeanthatthemedialc losurewasvoiced(Lisker,1957),whilealongc losuredurationdoesnotinvariablyelicitara pidlabelingresponse(Lisker,1981).Atprese ntwemayonlysaythatsomeofthepropertiesdem onstrablyaffectwordperceptionundercertai nconditions,and48 Lisker:" VOICING "inEnglishthattherestofth emare"candidatecues," ,itcannotinprinciplebeproventhatanyconce ivableacousticpropertyofaspeechorspeechl ikesignalisincapableofaffectingthepercep tionofanacousticsignalasaparticularlingu isticmessage;ontheotherhand,wehavenorigh ttoassumeaprincipleof"onceacue,alwaysacu e.
8 "Thus,forexample,thelinguisticirrelevanc eofthe[ voiced]differenceinthecaseofinitialIbldo esnotmeanthattheidentificationofastopasI bliseverywhereunaffectedbywhetheritsclos ureisvoicedorvoiceless(althoughitisjustt hisnonsequiturthatunderliestheassertionb yJakobsonandHalle,1956,thatthe"distincti vefeature"distinguishingthecategorysetsI b,d,glandIp,t,klisoneofarticulatoryforce andnotvoicing).Theaimofmostresearchintot heprocessesofspeechperceptionhasbeentoun coveralltheacousticpropertiesthatcansome whereserveascues,andnotsomuchtospecifyth econditionsunderwhichanyoneofthemdoesand doesnotservethatfunction, "cue"donotinvolveademonstrableconformity withnature--itisenoughthatpatternsbedevi sedsothatmanipulatingthepropertyeffectsa significantshiftinlisteners'wordidentifi cation,forexample, ,itwouldseem,thateitherthe~ #1listedabove,thedurationoftheformantles sintervalcorrespondingtooralclosure,serv esasacuetotherabid-rapidcontrastonlyinth eabsenceofglottalsignalovermostofthatint erval.
9 Varyingtheclosuredurationwillhavenoeffec tonlisteners'word-labelingbehavior;atmos tsometokensofthereportedrabidmaystriketh elistenersashavinganabnormallylongIb/. ,apatternsynthesizedwithverylowvaluesofF 1offsetandonsetfrequencies(properties#6a nd#14)islikelytobereportedasrabidnomatte rhowlongtheclosure(cLRepp,1978).Undersom econditions,then,closuredurationoperates asacuetostop" VOICING ,"thatis,lexicaliden tity; [ Voiced]CuesTheacousticpropertiesthatcans erveascuestolistenersindecidingwhetheran auditorystimulusisaninstanceofrabidorrap idaresaidtobecuestotheIb/-/plcontrastinm edialpositionbecausetheacceptedphonologi calrepresentationsoftheseforms,I'r~bIdan dI'r~pId/, ,then,thattheyarecuestothevoiced-voicele ssdistinctionaspreciselydefined?
10 Areasonableanswerwouldbethat,ifthelexica ldistinctionisequivalentperceptuallytoaI b/-/pldifferenceandthatinturnisamatterof closurevoicing,thentheacousticcuesarecue stothe[ voiced] ,itisbynomeansgenerally49 Lisker:" VOICING "inEnglishagreedthatphono logicalrepresentationsdointhemselvesamou nttoclaimsabouttheperceptualnatureofapho neticdistinction,anditcanbearguedthatthe phoneticspellings[ :a:!: bofd]and['.A:a:!pofd ]moredirectIyreflectlinguists'judgmentsa boutitsperceptualbasis,thatis, ,evenifwechoosetoviewtherabid-rapiddisti nctionasequivalenttoadifferenceintheirst opconsonants,itcanbearguedthatinordertob ecountedasacuetothevoicingstatusofthesto p,itisnotenoughthatagivenacousticpropert yshouldsignificantlydeterminealistener's lexicaldecision.
