CanfocusaccentingbeeliminatedinfavorofdeaccentingGivenconstituents?Man - PDF document

CanfocusaccentingbeeliminatedinfavorofdeaccentingGivenconstituents?ManfredKrifkaHumboldt-UniversitätBerlinZentrumfürAllgemeineSprachwissenschaftBerlin1AimsofthispaperWhataretherulesthatgovernthedistributionofsentenceaccent?Amongthevariousfactorsthathavebeendiscussed,suchaswordaccent,rhythmrulesandtheformationofprosodicphrases,onehasarousedparticularinterestsamonglinguists:focus.TheideathatfocusisexpressedbysentenceaccentcanbetracedbacktoPaul(1880),whoarguedthatwhathecalledthepsychologicalpredicatereceivesthemainaccent.Uptotoday,ithasbeenassumedbymostifnotallresearchersthatputtinganexpressioninfocusmeansputtinganaccentonit(perhapsinadditiontodoingotherthings,suchasmovingittoaparticularpositioninthesentence).Butthereisalsoageneralawarenessthatthefocusaccentrulescanbedistortedbyotherrulesthatrequiregivenexpressionsnottocarryaccents.Deaccentingrules,whichhavebeenintroducedbyLadd(1980,1996),gureintheworkofmanyresearchers,suchasSelkirk(1984,1995),Gussenhoven(1983,1992),Uhmann(1991),Féry(1993)andJacobs(1991).TheydosomostprominentlyintheinuentialtheoryofSchwarzschild(1999),whoevenproclaims:Byestablishinggivennessasthemainstayofourtheory,webreakrankswiththosewhoassumethatfocusprovokesinterpretation.Onequestionthathasnotbeeninvestigatedyet,tothebestofmyknowledge,iswhetherwecaneliminaterulesofaccentingfocusedexpressionscompletelyinfavorofrulesofdeaccentinggivenexpressions.Inmyownresearch,Ifoundthattheaccentsensitivityoftheinterpretationofsentenceswithadverbialquantiers,whichisexplainedbyfocusrulesinRooth(1985,1995)andKrifka(1995),tobebetterhandledbydeaccentingrules,cf.Krifka(2001).Sothereisatemptationtogetridoffocusaccentrulesaltogether,andjustworkwithrulesthatdeaccentgivenexpressions.Thenalresultofthispaperisthatweshouldresistthistemptation,eventhoughworkingwithdeaccentingrulesaloneleadsusfartherthanwemaythink.IwillstartwithremindingreadersofthetheoryofSchwarzschild(1999),whocomesclosesttoatheorythatworkswithdeaccentinggivenconstituents,butwhodoesnotquitegoasfarassuggestingatheorybasedsolelyondeaccenting.Iwillshowthatbyeliminatingthenotionoffocus,examplesofthekindthathetreatsareactuallyexplainedinawaythatismorecongenialtohisenterprise.Attheend,though,anumberofproblemsforSchwarzschild(1999)andtheradicaltheoryofGivennesswillappearthat,tomymind,suggestthatwebetterworkwithbothfocusandgivennessasbasicnotionsthatexerttheirinuenceontheprosodyofsentences.Henceoneresultofthispaperisthatthegreatmajorityofresearchersisindeedontherighttrack.Iamawarethatthisarmationofthereceivedviewisperhapsnotearth-shattering,butIcanpromisethattheexcursioninthelandwithoutfocusisworthwhileandthattherewillbenewreasonstobelievein.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated?107 thereceivedview.Also,patientreaderswillndananswertothelong-standingquestionwhyfocusprojectsfromtheargument,andnotfromthehead.2ThetheoryofSchwarzschild(1999)WhileSchwarzschild(1999)makesuseoffocusandgivennessasbasicnotions,hehasonlyonerulegoverningaccentuation:Focusisexpressedbyaccent.Thisishowthetheoryworks.First,thewell-knownfocusprojectionrulesofSelkirk(1984,1995)areassumedac-cordingtowhichfocusisexpressedbyafeatureFthatislicensedbyaccentandprojected:(1)a.F-Assignment:AnaccentedwordisF-marked.b.F-Projection:i.F-markingoftheheadofaphraselicensesF-markingofthephraseii.F-markingoftheinternalargumentoftheheadlicensesF-markingofthehead.Inthefollowingexample,F-markingonBillislicensedbyaccent,whichinturnlicensesF-markingontheheadpraised,whichnallylicensesF-markingonpraisedBill.(2)A:WhatdidMarydo?B:[She[praisedFBíllF]F]Whydoesfocusprojectiontakethedetourviathehead,thatis,whydoesn'ttheargumentprojectthefocusdirectly?Thisistoaccountforcaseslike(3),wherethedirectargumenthimisgiven,andtheabsenceofF-markingappearstoindicatethat.(3)A:WhatdidJohn'smotherdo?B:[She[práisedFhim]F]However,as(4)shows,givennessiscompatiblewithF-marking:(4)A:WhatdidJohn'smotherdo?B:[She[praisedhímF]]Schwarzschildproposesthat(1)issupplementedbythefollowingtworules:(5)a.Givenness:IfaconstituentisnotF-marked,itmustbegiven.b.AvoidF:F-markaslittleaspossible,withoutviolatingGivenness.NoticethatGivennessallowsforgivenconstituentstobeF-marked,asin(4).GivennessalsoforcesF-marking,asitallowsfornon-F-markingonlyincaseconstituentsaregiven.Itiscrucialthatgivenisdenedforreferentialaswellasforpredicationalorpropositionalconstituents:(6)a.AnutteranceUisGivenifithasasalientantecedentAsuchthati.IfUreferstoanentity,thenUandAcorefer;ii.otherwise,AentailstheexistentialF-closureofU.Case(6a-i),isevident.Case(6a-ii)presupposesthatUhasameaningwhichisbasedonthetypeoftruthvalues.ItasksustoformtheexistentialclosureoverallopenargumentsofthemeaningofU,andtoreplacethefocusexpressionsbyvariablesandexistentiallyclosethemaswell.Thiscouldbedoneinthestructuredmeaningaccountoffocus(cf.vonStechow1991)orintheaccountofAlternativesemantics(cf.Rooth1985,1992);wedisregardthesedistinctionshere.Forexample,consider:108.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated? (7)ExistentialF-closureof[praisedJóhnF]:9y9x[praised(y)(x)]Themeaningof[praisedJóhnF]isx[praised(john)(x)].Existentialclosureoveritsfreeargumentpositiongivesus9x[praised(john)(x)].Replacingtheiteminfocusbyafreevariableleadsto9x[praised(y)(x)];existentialclosureoverthisvariableresultsin9y9x[praised(y)(x)].ThefollowingexamplesillustratehowSchwarzschild'sruleswork,startingwithnar-rowfocus.(8)a.A:WhodidMarypraise?b.B:[Mary[praisedJóhnF]](8a)introducestheexistencepresupposition9x[praised(x)(mary)](or,apointtowhichwewillreturninsection4,9x[praised(x)(mary)^person(x)]).Givennessisappliedonallsyntacticnodes.AsthewholeanswerisnotF-marked,itmustbeGiven,whichisthecaseastheexistentialclosureof[Mary[praisedJóhnF]]]is9x[praised(x)(mary)],whichfollowsfromtheexistencepresuppositionofthequestion.Astheconstituent[praisedJóhnF]isnotfocusedmarked,itmustbegivenaswell,whichisthecaseasitsexistentialclosure9x9y[praise(x)(y)]followsfromtheexistencepresuppositionofthequestion.TheconstituentJóhnisF-marked,henceGivennessisnotapplicable,butwehavetocheckwhether,byAvoidF,ithastobefocus-marked.Thisisindeedthecase,fortheanswer[Mary[praisedJohn]],withoutanyfocusmarking,wouldhavetosatisfyGivenness,butitdoesn't:Itsexistentialclosurepraised(john)(mary)doesnotfollowfromthepresuppositionofthequestion.CouldweplaceFsomewhereelse,onMaryoronpraised?No.Forexample,[MáryF[praisedJohn]]wouldhaveasitsexistentialclosure9y[praised(john)(y)],whichdoesnotfollowfromthepresuppositionofthequestion.WecouldtryoutwidefocusontheVP,whichisgeneratedas[Mary[praisedFJóhnF]F]accordingtoSelkirk'srules.Theexistentialclosureforthissentenceis9P[P(mary)],i.e.,Maryhassome(contextuallyrestricted)property,andthiscertainlyfollowsfromtheexistentialpresupposition.However,AvoidFltersoutVPfocus,asitinvolvesmorefocithantheoriginalanswerin(8).ThefollowingexampleillustratesthatagivenexpressioncanbeF-marked.Theargumentisexactlyparalleltotheonegivenin(8).NoticethatnothingpreventsF-markingonagivenconstituent.(9)A:WhodidJohn'smotherpraise?B:[She[praisedhimF]]ConsidernowanexampleinvolvingVPfocus:(10)A:WhatdidMarydo?B:[She[praisedFJóhnF]F](10)presupposesthatMarydidsomething:9P[P(mary)^activity(P)].ThesentenceisnotF-marked,whichisneasitisgiven:It'sexistentialF-closureis9P[P(mary)],orperhapsmorespecically9P[P(mary)^activity(P)],andthisclearlyisentailedbythepresuppositionofthequestion.TheVPisF-marked,whichisrequiredbyGivenness.Toseethis,rstconsiderthecasewithoutanyF-marking,[She[praisedJóhnF]].TheexistentialF-closureisthepropositionpraised(john)(mary),whichdoesnotfollowfromthepresuppositionof(9).Second,considerthecasethatthereisF-markingonlyonJohn,asin[She[praisedJóhnF]].ExistentialF-closureleadsto9x[praised(x)(mary)],.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated?109 whichagaindoesnotfollowfromthepresupposition.NowconsiderF-markingonlyonpraised,asin[She[práisedFJohn].ExistentialF-closureleadsto9R[R(john)(mary)],thatis,MaryandJohnstandinsomerelationtoeachother,whichagaindoesnotfollowfromthepresuppositionofthequestion.NextconsiderF-markingonpraised,projectedtotheVP:[She[práisedFJohn]F].NowJohnisnotF-marked,henceitmustbegiven,butinfactitisn't,asitisnotmentionedinthepreviousdiscourse.Letusnallytry[Mary[práisedFJóhnF]].ThisleadstoanexistentialF-closure9x9R[R(x)(mary)],thatMarystandsinsomerelationtosomeindividualx,whichagaindoesnotfollowfromthepresuppositionofthequestion.Henceweareforcedtoassumethethreefociintheanswerof(10),asonlythenallconditionscanbesatised.Whatwillhappenincasetheargumentisgiven,asinthefollowingexample:(11)A:WhatdidJohn'smotherdo?B:[She[práisedFhim]F]Theargumentationisthesameasinthepreviousexample,withtheonlydierencethathimisgiven.NowwecandropF-markingonhim,followingAvoidF.IncontrasttoSelkirk'sproposal,wedonotneedaseparatetreatmentofF-markedconstituentswithinotherF-markedconstituents.3AGivennesstheoryofaccentuationSchwarzschild'sexplanationofsentenceaccentisremarkablebecauseoftheimportantroleitassignstoGivenness,whereasprevioustheoriesmostlyconsideredFocusasthedecisivefactor.ForSchwarzschild,GivennessisnearlycomplementarytoFocus:WhatisnotF-markedmustbeGiven,accordingtorule(5a).Furthermore,hestatesthatthemarkingofFocusshouldbeavoided(5b),whichcanbetakenassayingthatGivennessshouldbeexpressed.ThissuggestsareformulationofSchwarzschild'srulesthatessentiallybuildonGivennessinsteadofFocustodescribethedistributionofaccents.Insteadof(5a,5b),wecantryoutthefollowingrules,whichhavetheadvantageofbeingformulatedaspositivestatements:(12)a.GivennessD:IfaconstituentisD-marked,thenitisGivenD.b.Deaccent!:D-markasmuchaspossible.Somethinglike(12a)hasbeententativelysuggestedbyBüring(2006),inacontraposedform:IfaconstituentisnotGIVEN,itmustbeprominent(i.e.,notD-marked).Theformulationheregoesonestepfarther,asitworkswithnon-prominence,orD-marking,asabasicnotion.F-markingdoesn'tplayaroleintherules(12),anditdoesnotgureintherevisednotionofgivennesseither,whichissimplerthantheonein(6):(13)a.AnutteranceUisGivenDifithasasalientantecedentAsuchthati.IfUreferstoanentity,thenUandAcorefer;ii.otherwise,AentailstheexistentialclosureofU.D-markingresultsindeaccentuation.Theserulesassumeanewviewofhowaccentisdetermined.Asmentioned,mostresearchers,includingSelkirk(1984,1995),Gussenhoven(1983)andJacobs(1991)consideraccentaresultoffocusmarking:Constituentsthatarefocusedareaccented.Inthecurrentrulesystem,thepositionofaccentsfollowsfromrulesthatidentifyconstituentsthataretobedeaccented;constituentsthatcannotbedeaccentedemergeastheonesthatareaccented.110.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated? JustaswithF-marking,D-markingprojects.IncontrasttotherulesofSelkirkin(1),therulesgoverningtheprojectionofgivennessissimpleandintuitive:(14)IfallthesubconstituentsofacomplexconstituentareD-marked,thenthiscon-stituentisD-marked.Thatis,[DD])[DD]D.Butwewillneedoneadditionalruletodealwiththeargument/headasymmetry.Beforeweintroducethat,letusdiscusshowtheGivennesstheoryofaccentuationworks.Webeginwithexample(8),repeatedhereunderitsnewanalysis.(15)A:WhodidMarypraise?B:[SheD[praisedDJóhn]]MaryandpraisedcarryadeaccentuationfeatureDthatleadstoasuppressionofaccent;Johndoesnotcarrythisfeature,andthereforeitwillcarryaccent.TheD-markingin(15)istheonlyonethatiscompatiblewiththerules(12):First,noticethattheD-markingisindeedcompatiblewiththerules.SheandpraisedareD-marked,andbothconstituentsareGivenD.Inthecaseofshe,anexpressionoftypee,thereisasalientexpressionthatreferstothesameentity,namelyMary.Inthecaseofpraised,theexistentialclosure,9x9y[praised(x)(y)],followsfromthepresuppositionofthequestion.Second,noticethatD-markingofJohnisnotpossible;thiswouldviolateGivennessD,asJohnisnotgiven.Third,whilelackofD-markingonpraisedorMarywouldbecompatiblewithGivennessD,itwouldviolateDeaccent!,asdeaccentingwouldnotbemaximized.Ofcourse,D-markingontheVPoronthewholesentenceisnotpossibleeither,astheconstituents[praisedDJóhn]and[sheD[praisedDJóhn]]arenotGivenD.Nextconsideranexamplewithanaccentedconstituentthatisgiven:(16)A:WhodidJohn'smotherpraise?B:[SheD[praisedDhím]F]LackofD-markingonhimiscompatiblewithGivennessD,asthisonlystatessomethingaboutD-markedconstituents.IslackofD-markingalsorequired?Yesindeed:AssumethathimwereD-marked,asin[praisedDhimD];thenthewholeVPwouldbeD-marked,following(14),resultinginastructure[praisedDhimD]D.FollowingGivennessD,thisVPmustbeGivenD,butitsexistentialclosure9x[praised(john)(x)]doesnotfollowfromthepresuppositionofthequestion.ThenecessitytoD-marksheandpraisedfollowsfromthesamereasonsastheonesdiscussedfor(15).Nowletusreconsiderexample(11)thatwehavedescribedasonewithbroadfocus:(17)A:WhatdidMarydo?B:[SheDpraisedJóhn]]ItisevidentthatthisistheonlyD-markingcompatiblewiththerules.ShecanbeD-marked,followingGivennessD,andithastobeD-marked,followingDeaccent!.Furthermore,otherconstituentscouldnotbedeaccented.Forexample,[praisedDJóhn]isnotapossibleD-marking,asitviolatesGivennessD.ItwouldrequirepraisedtobeGivenD,thatis,9x9y[praised(x)(y)]tobeinferrable,fromthecontext,whichisnotthecase.However,itnowbecomesevidentthattherulesystemproposedsofarisincomplete,asitdoesnotpredictthatpraisedisrealizedasifitweredeaccented,withJohnbearingthemainaccent.Weobviouslyhavetoappealtosomeequivalenttofocusprojectionruleshere,likethefollowing:.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated?111 (18)Ifinaconstituent[]withaheadandaninternalargumentneithernorareD-marked,thenD-markthehead!ThereisanobviousproblemwiththisrulewhenwecombineitwithGivennessD,asitisthenrequiredthattheheadisGivenD.Wehavetwooptionshere.Eitherwecanunderstand(18)asarulethatisoperativeonlyafterGivennessDhasbeenchecked.Or,ifwedon'tlikeextrinsicruleorderings,wecanintroducetheconceptofd-markingandsaythatif[]isnotD-marked,thentheheadisd-marked,whered-markingisinterpretedprosodicallyjustasD-marking,withoutbeinglinkedtogivenness.Asthislatterwayisnotationallyclearer,Iwillmakeuseofithere,andreplace(18)bythefollowingrule:(19)Ifinaconstituent[]withaheadandaninternalargumentneithernorareD-marked,thend-markthehead!Example(17)thenhastobeanalyzedasfollows:(20)A:WhatdidMarydo?B:[SheD[praiseddJóhn]]Rule(19)canbereformulatedsothatitcangoverntheaccentdistributionwithinacomplexconstituentwhosesubconstituentsareD-marked,cf.(21),whichleadstoanalysesasin(22).(21)Ifinaconstituent[]withaheadandaninternalargumentbothandhavethesamestatusastoD-marking,thend-markthehead!(22)A:WhatdidMarydoaftershepraisedJohn?B:After[sheD[praisedD;dJóhnD]]D,shegavehimakiss.Theadditionald-markingofpraisedleadstoarelativeaccentuationofJohn.Arecursivedenitionofd-markingappearspossiblethatmimickstherecursivedenitionoffocusmarkinginJacobs(1991),butIwillnotattempttoimplementthishere.Letusnowconsiderexample(11),undertheGivennessanalysis.(23)A:WhatdidJohn'smotherdo?B:[SheD[práisedhimD]]ItisevidentthatD-markingonsheandhimispossible,followingGivennessD,asbothexpressionsaregiven.ItisalsoevidentthatD-markingonpraisedisnotpossible,asitisnotGivenD;thepresuppositionofthequestiondoesnotentail9x9y[praised(x)(y)].TheGivennesstheoryofaccentuationyieldstherightresultincasesthathavebeenanalyzedasdoublefocus(whichSchwarzschild'stheorycandealwithaswell):(24)FirstJohncalledBillaRepublican,andthen[Bíll[insultedDJóhn]]D-markingoninsultedisjustiediftherstsentenceisaccommodatedinsuchawaythat,ifxcallsyaRepublican,thenxinsultsy.Theexistentialclosureofinsultedis9x9y[insult(x)(y)],andthisfollowsfromtherstclause.CouldwealsoD-markBill,orJohn,orboth,astheyaregiventoo?No:TheVP[insultedDJohnD]isnotacceptable,asitsexistentialclosure,9x[insult(john)(x)],isnotentailedbythecontext.Neitheristheexistentialclosureof[BillD[insultedDJóhn]],northeoneof[BillD[insultedDJohnD]],ascanbeeasilychecked.112.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated? 4AcomparisonoftheoriesWhichofthetwotheories,Schwarzschild(1999)ortheGivennesstheory,istobepre-ferred?Theymakerathersimilarempiricalpredictionsinmanycases.Also,itappearsthattheinterfacerulesthatleadtospell-outofF-markedorD-markedconstituentsinprosodyareofsimilarcomplexity.ThepossibilityforapositiveformulationofGivennessDby[D-marked()!GivenD()],insteadofthenegativeformulationinSchwarzschild'soriginaldenitionofGivennessby[:F-marked()!Given()],shouldbeaconsideredaconceptualadvantageoftheGivennesstheory.AmoresubstantialadvantageisthefactthattheGivennesstheoryreliesonasimplertheoryofGiveness,namelyGivenD,whichdoesnotrefertotheF-featureatall(infact,thereisnoF-featureinthistheory).GivenDcanbeexpressedbysimpleexistentialclosure,whereasGIVENinvolvesbothreplacingthefocusexpressionbyavariableandexistentialclosure.TheGivennesstheorymightbeconsideredmorecomplexbecauseithastoresorttoeitherrule-orderingorasecondfeaturedtodealwithfocusprojectioncasessuchas(20).Butthenthisrelativelysimplerulereplacesthetworulesoffocusprojection(1b).Furthermore,oncloserinspection,Schwarzschild'stheoryneedsanadditionruleaswell.InSelkirk'sruleprojection,nothingpreventstheaccentstructureofthefollowinganswer:(25)A:WhatdidMarydo?B:[She[práisedFJóhnF]F]Here,praisedisaccented,whichmotivatestheFfeaturethatitwouldhavegottenanywaybythefactthattheargumentJohnisinfocus.AccentonpraisedisnotexcludedbyAvoidF,asthisconstrainsonlytheassignmentofF-features,notaccentuation.Hencewewouldneedanadditionalconstraint,avoidaccent.ThisconstraintcouldnotsimplyreplaceAvoidF,asweneedthat,forexample,toblocktheVPaccentincasesofdoublefocus,suchas[She[[práisedFJóhnF]and[condémnedFBíllF]]].ItshouldberemarkedthatSchwarzschildhimself,attheendofhispaper,feelscompelledtosketchanalternativeversionofhistheoryinwhichF-markingisfreebutcheckedbyaviolableconstraintsayingthatheadsarelessprominentthanarguments(his(59));thisisexactlywhat(19)wantstoinforce,andSchwarzschild'sconstraintactuallycouldreplace(19).Büring(2006)hasanequivalentruleofhorizontalfocusprojection.IfwecomparethetwopreferencerulesAvoidFandDeaccent!,itisdiculttoarguethatoneisintrinsicallybetterthantheother.AvoidFcanbeseenasarulethatavoidsthelinguisticcomplexitythatisaresultofF-marking,whichresultsinF-markingonlywhennecessary.Deaccent!canbeseenasarulethatprefersthegreaterlinguisticsimplicitythatcomeswithdeaccentuation,whichisblockedonlyincasetheaddresseeistobedirectedtoinformationthatisnotyetderivablefromthecontext.Iconsiderthesetwoviewsequallyplausible.Whiletheargumentsdiscussedsofarleadtotheconclusionthatboththeorieshavesimilarcomplexity,therearesomethatshowthatGivennesstheorymayactuallybesimpler.First,considernon-accentableexpressionssuchassomeone.(26)A:WhatdidMarydo?B:[She[práisedFsomeoneF]F]Thedirectargumentsomeonecannotbeaccented,andhenceitisnotF-marked.FollowingGivenness,itshouldbegiven.ApplyingSchwarzschild'sdenitionofGivenrequiresustoformtheexistentialclosureoverthemeaningofsomeone,whichis9P[person\P=;],.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated?113 whichistrueithereisatleastoneperson.Thiscertainlyfollowsfromthepresuppositionofthequestionin(26),asMaryisaperson.However,itdoesfollowfromthequestionofthefollowingcases:(27)A:Whatdidthedogdo?B:[It[bítFsomeoneF]F](28)A:Whathappened?B:[PéterF[cameinF]]Fvs.[Someone[cameínF]]FPerhapsthepresuppositionthatsomeoneexistsisafairlyinnocentonethatcanalwaysbeassumed.NoticethatindeniteNPsbasedongeneralnouns,suchasaperson,maybehaveinasimilarway:(29)A:Whatdidthedogdo?B:[It[bítF[aperson]]F]/[It[bitF[apérson]F]F]Butthiscannotbethewholestory.Thewordspersonandhumanbeingpresumablyareextensionallyequivalentforourpurposes,butcertainlyhumanbeingcanneverbedeaccentedinItbitahumanbeing,asananswerto(29).Webetterassumethatitisagrammaticalpropertyofexpressionslikesomeone,andofcertainusesofaperson,thattheycannotbeaccented.WecanexpressthisinSchwarzschild'stheorybystipulatingthattheynevercanbefocused,whichpreventsthemfrombeingaccented,asfocusonawordhastobelicensedbyaccent.ButthentheydonotsatisfytheGivennessconditionascribedtonon-focusedexpressions.Inthecurrenttheorywecanstipulatethatsomeonehasthefeaturedaspartofitslexicalspecication,whichisirrelevantforGivennessD.Wegetthefollowinganalysis:(30)A:WhatdidMarydo?B:[SheD[práisedsomeoned]]NoticethatpraisedisnotD-marked,andtheVP[práisedsomeoned]isnotD-markedeither.Weknowthatpraisedand[práisedsomeoned]couldnotbeD-marked,astheseconstituentsarenotGivenD.ItappearsthattheGivennesstheorycandealwithnon-accentableexpressionslikesomeonebetter,bytheassignmentofad-featurewhichisneededforindependentreasons.AnothertypeofinstancethatfavorstheGivennesstheoryarecasesinwhichfocusappearstoprojectfromanembeddedconstituent.Schwarzschild(1999)hasdiscussedsuchcases,andBüring(2006)hasaddedmorethatallshowthat,whilefocusonacomplexexpressionmustbegroundedinafocused(andaccented)word,theselectionofthiswordismainlydeterminedbygivenness.Schwarzschild'sexamplereceivesthefollowinganalysisintheGivennesstheory:(31)A:JohndroveMary'sredconvertible.Whatdidhedrivebeforethat?B:[He[drove[herblúeFconvertible]]]Schwarzschildwouldhavefocusonblue,andhearguesthatallotherconstituentsareGiven.However,itisapurecoincidencethatthechosencontextentailstheexistentialF-closureofthesentence,thatJohndroveaconvertibleofMary.InacontextlikeMarydroveherredconvertible.WhatdidJohndrive?,theF-closureoftheanswerisnotgiven,asitdoesnotentailthatJohndroveaconvertible.Yetthesamefocusstructureisrequiredinthiscontext.HereiswhatGivennesstheorysays:114.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated? (32)B:[HeD[droveD[herDblúeconvertibleD]]]NootherconstituentisD-marked.Thatis,thesentenceisnotpredictedtorequireasalientantecedentfromwhichitfollowsthatJohndroveoneofMary'sconvertibles.TheD-markingsin(32)arejustied:Therearesalientantecedentsforheandher;droveisGivenDbecausethecontextentails9x9y[drove(x)(y)];andconvertibleisgivenasthecontextentails9x[convertible(x)].WecannotD-markbluebecauseitdoesnothaveanantecedent.ItalsowouldleadtoD-markingofallconstituents,following(14),whichinturnwouldrequirethatthecontextalreadyentailsthatJohndroveablueconvertible.5WhyweneedFocus,inadditiontoGivennessInthelastsectionIhavetriedtoarguethataSchwarzschild-styletheorycanbemoresuccinctlyexpressedinatheorythatusesD-markinginsteadofF-marking.InthissectionIwouldliketoshowthatweactuallyneedbothrulesofdeaccentuationandrulesofaccentuation,asatheorybasedonlyonGivennessleadstoanumberofproblems.1.TherstproblemisthatwenddeaccentuationintheabsenceofGiven(D)ness,beyondthecaseofsomeonediscussedabove.Letustakeuptheconvertibleexample,inthefollowingform:(33)Asthereweren'tanyredconvertiblesanymore,Johndroveablúeconvertible.Fromtherstclauseitdoesnotfollowthatthereareconvertibles,butbothSchwarzschild'stheoryandtheGivennesstheoryrequirethatthereisone,toaccountforthelackofac-centonconvertible.ThissuggeststhatconvertibleisnotGiven(D)becauseitsexistentialclosurefollowsfromthecontext,butsimplybecausetheconcept`convertible'hasbeenmentionedbefore.Thissuggeststhatweshouldextendthenotionofgivenness.Follow-ingWebber(1978)wecouldassumethatnounsintroducediscoursereferentsforkinds,whichcanbetakenupanaphoricallybyothernouns,wherethelifetimeofsuchdiscoursereferentsisnotrestrictedbythescopeofnegation(cf.theuseofone-anaphora,asinJohndroveablúeone).ButthereareothercaseswhereGiven(D)failstodeterminedeaccentuation.Inthefollowingexample,stoodupisclearlydeaccented,althoughitisnotGiven(D)(thecontextdoesnotentailthatanyonestoodup).(34)Asnoneofherfriendsstoodup,Márystoodup.Wehaveasimilarprobleminothercasesofnon-presupposingconstructions:(35)A:ItispossiblethatMarypraisedsomeone.B:Yes,MarypraisedJóhn.(36)A:IdoubtthatMarypraisedanyone.B:You'rewrong,MarypraisedJóhn.Oncloserinspection,eventheconstituentquestionsthatmotivatedtheaccentuationtheoriesturnouttobeunconvincing,asitisnotclearwhethersuchquestionsgenerallycomewithanexistencepresupposition.Consider(37).Here,praiseisnotGiven(D),andpraisedhimisnotGiven(D),yettheseconstituentsmustbeD-marked(orcannotbeF-marked,inSchwarzschild'stheory).(37)A:Who,ifanybody,praisedJohn?B:Márypraisedhim..LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated?115 PerhapsweshouldextendthedenitionofGiven(D)suchthateverythingthatismen-tionedintheimmediatelyprecedingcontextshouldcountasgiven.Forexample,stoodupisgiveninthesecondclauseof(34)asthisconceptwasmentionedintherstclause.Withthisweareheadingtowardsthenotionofc-construabilityofRochemont(1986).However,thisleadsusintowell-knownproblems,asintheRepublicanexample(24),orincasethealternativesareexplicitlygiven,asin(38):(38)A:WhodidMarypraise,BillorJohn?B:MarypraisedJóhn.Heretheconceptsof`John',of`praisedJohn',andof`MarypraisedJohn'areallc-construable,whichpredictsthatweshouldnotndanyaccent,contrarytofact.Ro-chemont(1986)dealtwithsuchcasesasaspecialtypeoffocus,contrastivefocus,whichfollowsdierentrules.WhileIbelievethatthereiscontrastivefocusasaseparatecase,Ithinkthat(38)isnotaninstanceofthat,aswedonotndtheusualhallmarksofcontrastivefocus,likemorepronouncedaccentorthepossibilityofcleftconstructions.2.TherearesimilarproblemsthathavebeenpointedoutbyFéry&Samek-Lodovici(2006).Forexample,in(39)deaccentingonfarmerandaccentonCanadianisjustiedinSchwarzschild'stheory(aswellasintheGivennesstheory),buttheoptionofaccent-ingAmericanisnot.Similarly,theoptionoffocusingredinin(33)isnotpredicted.Schwarzschildhastoresorttoaspecialcontrastivefocusrelation,justasRochemont.(39)AnAméricanfarmerwastalkingtoaCanádianfarmer.3.Aslightvariationoftheexampleslike(9)thatSchwarzschildhasusedtomotivatefocusrules,andIhaveusedtomotivatedeaccentuationrules,resultsinwrongpredictions:(40)A:WhatdidMarypraise?B:*ShepraisedJóhn.Theanswerdoesnotsatisfythepresuppositionofthequestion,thatMarypraisedathing.Nevertheless,neitherSchwarzschild'stheorynortheGivennesstheorywouldconsiderthisproblematic.Hereiswhy:(40)createsthepresupposition9x[praised(x)(mary)^thing(x)].UnderSchwarzschild'stheory,(40)hasthefocusassignment[She[praisedJóhnF]].AstheVPandthesentencearenotF-marked,theirexistentialF-closure9x9y[praised(x)(y)]and9x[praised(x)(mary)]shouldbeGiven.Andindeedtheyare,astheyarenotrestrictedtopersonsorthings.IntheGivennesstheory,theanswerhasthedeaccentuationpattern[SheD[praisedDJóhn]],whichrequiresthatpraisedisGivenD;thisisindeedthecase,as9x9y[praised(x)(y)]followsfromthepresuppositionofthequestion,asbefore.4.ThemostsevereproblemofaGivennessonlytheoryisthatitleavesthequestion-answerrelationvastlyunderdetermined.While(41a)motivatestheaccentstructureof(41b),whatwehavesaidsofarineithertheorycannotruleoutanswerslike(41c).(41)a.A:WhodidMarypraise?b.B:ShepraisedJóhn.c.B:#Sheslépt.116.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated? Thequestionintroducesthepresupposition9x[praised(x)(mary)],fromwhichitdoesnotfollowthatsomeoneslept,hencetheaccentuationin(41c)issatised.CanweuseGivennessasafactortodeterminewhatacongruentanswertoaquestionis?ThebestIcouldcomeupwithisthefollowing,where(ii)isessentiallywhatwehaveassumedsofar.(42)a.Aquestion-answerpairQAiscongruenti:i.QintroducesanexistentialpresuppositionQE,andAentailsQE.ii.theexistential(F-)closureofallD-marked(nonF-marked)constituentsofAfollowfromQE.Forexample,(41b)entailstheexistentialpresuppositionof(41a),thatMarypraisedsomeone,but(41c)doesnot.Thiswayofexplainingquestion/answercongruenceisproblematic,however,inthecaseofnon-presupposingquestions,asin(37).IhavelistedanumberofproblemsthatmakeitveryquestionablethatatheorybasedonGivennessalonewillbesucienttodescribealltheeectsthathavebeenascribedtofocusandfocusaccent.Asitisobviousthatfocusaloneisnotsucienteither,weneedbothconcepts.Howtheyinteractissketchedinthenalsection.6HowFocusaccentingandGivennessdeaccentinginteractLetusassumethenatheorythathasbothF-markingandD-marking,whereF-markingindicatesthepresenceofalternatives,andD-markingindicatesGivenness.(Suchtheorieshavebeenproposedbefore,mostrecentlybyFéry&Samek-Lodovici(2006),whoalsofac-torinconstraintsofphonologicalphraseformation).Asadefaultassumption,F-markingandD-markingaremaximalinthesensethatwheneveralternativestoaconstituentplayaroleininterpretation,thenitisF-marked,andwheneveraconstituentisgiven,itisD-marked.Alternatives,andhenceF-marking,canbeusedforavarietyoffunctions,forex-ampletoexpresscontrast,ortoidentifythedomainofquanticationofonlybywayofalternatives.Inquestion-answersequences,thealternativesintroducedbythequestionmustbeidentiablewiththealternativesoftheanswer.Usingstructuredmeaningsthatrelateabackground,asetofalternatives,andafocustoeachother(cf.Krifka2006),thiscanbeillustratedasfollows:(43)a.A:WhodidMarypraise?hx[praised(x)(mary)];person;ib.B:ShepraisedJóhn.hx[praised(x)(mary)];A;johniThequestionhasanemptyfocus,,asthequestionwordwhojustidentiesthesetofalternatives.TheanswerhasavariableAforthealternativeset,asthisisdeterminedbythecontext.(43a43b)isacoherentquestion-answerpair,asthebackgroundsareidentical,theidenticationofthealternativesetsperson=Aispossible,andjohn2Aholds.AsforGivenness,Isuspectthattherelation`given'basedonentailmentofexistential(F-)closureistoonarrow,asexampleslike(33)and(34)show.Weprobablywillhavetouseawidernotion,suchasRochemont'sc-construability.FocusandGivennessareexpressedinsimplebutcontradictoryways(cf.alsoFéry&Samek-Lodovici2006):(44)a.FOCUSACCENT(toberevised):IfaconstituentisinFocus,itbearsAccent..LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated?117 b.GIVEN-DEACCENT:IfaconstituentisGiven,itisDeaccentend.TheFOCUS-ACCENTruleiscompatiblewithaccentbeingassignedtonon-focusedex-pressions.Wemayreneitinsuchawaythatitsaysthatthatitbearsstrongeraccentthansisterconstituentsthatarenotinfocus,cf.Jacobs(1991),orthatitbearsthestrongestaccentinitsfocusdomainwhichincludesthebackground,cf.Truckenbrodt(1999)andFéry&Samek-Lodovici(2006).TheFOCUS-ACCENTruleoutrankstheGIVEN-DEACCENTrule(cf.Féry&Samek-Lodovici2006).Henceanexpressioninfocusthatisgivenmustbeaccented:(45)A:WhodidMarypraise?B:SheDpraisedD[hímD]FThefocus-accentruleasstatedin(44)isfairlyunspecic,asitdoesnotindicatehowAccentisrealizedincomplexconstructions.Asitiswell-knownsinceGussenhoven(1983)andSelkirk(1984),wehavetodistinguishbetweenhead-argumentconstructions,whichareoftenrealizedbyoneaccent,andotherssuchashead-adjunctconstructionsorcoordinationconstructions,whicharerealizedbymultipleaccent.Ithasbeensuggested(seeGussenhoven1983,1992,Jacobs1991,Truckenbrodt1999andothers)thatthelatterconstructionshavetoformseparatephonologicalphrasesthateachgetanaccent,whereashead-argumentconstructionscanbeintegratedintoonephonologicalphrase.Icannotgointotheintricaciesofphonologicalphraseformation,whichalsodependsonthepresenceandabsenceoffocus.However,itisclearthat(44)shouldbespeciedasfollows:(46)FOCUS-ACCENT:IfaconstituentisinFocus,eachofitsphonologicalphrasesbearsAccent.Finally,weturntothequestionofaccentrealizationinhead-argumentconstructionsthatareintegratedintoasinglephonologicalphrase.Herewehavetostatethewell-knownasymmetry:(46)FOCUS-ACCENT:IfaconstituentisinFocus,eachofitsphonologicalphrasesbearsAccent.(47)ACCENT-ARGUMENT:Ifanintegratedconstituentconsistingofaheadandanargumentbearsaccent,thenaccentisrealizedontheargument.TheACCENT-ARGUMENTruleisrankedlowerthanFOCUS-ACCENTorGIVEN-DEACCENT.Thispredictsthefollowingdata:(48)A:WhatdidMarydo?B:SheD[praisedJóhn]F(49)A:WhatdidJohn'smotherdo?B:SheD[práisedhimD]Inthelattercase,accenthastoberealized,duetoFOCUS-ACCENT,butitcannotberealizedonhimduetoDEACCENT-GIVEN,andhencehastoberealizedonpraised.Incasebothconstituentsaregiven,thenACCENT-ARGUMENTre-emerges:(50)A:Didyouprepareameal,oroeradrink?B:I[oeredD[adrínk]D]FTheviewthatDEACCENT-GIVENcounteractsACCENT-ARGUMENToersawaytounderstandwhywehavearulelikeACCENT-ARGUMENTtobeginwith,thatis,why118.LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated? accentbydefaultpercolatestotheargument,andnotthehead.IthinkitisaplausibleassumptionthatreferentialexpressionsaremoreoftenGiventhannon-referentialones,likepredicates.IfDeaccentuationsignalsGivenness,thentheusefulnessofthismarkingstrategyismaximizedifreferentialexpressionsareaccentedbydefault.Thisisthecaseifaccentinahead-argumentconstructionisrealizedbydefaultontheargument,whichistypicallyreferentialandoftengiven,andnotonthehead,whichisnon-referentialandmorerarelygiven.Obviously,thedierencesbetweenheadsandargumentswithrespecttogivennessstipulatedherehavetobecheckedinnaturallinguisticcorpora,butitappearsquitelikelythattheindicatedtendencywillindeedhold.referencesBüring,Daniel.2006.Focusprojectionanddefaultprominence.In:ValeriaMolnarandSusanneWinkler(eds.).Thearchitectureoffocus.Berlin:MoutondeGruyter.Féry,Caroline.1993.Germanintonationalpatterns.Tübingen:Niemeyer.Féry,CarolineandVieriSamek-Lodovici.2006.Discussionnotes:Focusprojectionandprosodicpromi-nenceinnestedfoci.Language82:131157.Gussenhoven,Carlos.1983.Focus,mode,andthenucleus.JournalofLinguistics19:377417.Gussenhoven,Carlos.1992.Sentenceaccentsandargumentstructure.In:I.M.Roca(ed.).Thematicstructure.Itsroleingrammar.Foris,Berlin,NewYork.Jacobs,Joachim.1991.Focusambiguities.JournalofSemantics8:136.Krifka,Manfred.1995.Focusandtheinterpretationofgenericsentences.In:GregoryN.CarlsonandFrancisJeryPelletier(eds.).TheGenericBook.ChicagoUniversityPress,Chicago.238264.Krifka,Manfred.2001.Non-novelindenitesinadverbialquantication.In:CleoCondoravdiandGer-ardRenardelderLavalette(eds.).LogicalPerspectivesonLanguageandInformation.CSLIPress,Stanford.140.Krifka,Manfred.2006.Thesemanticsofquestionsandthefocusationofanswers.In:ChungminLee,MatthewGordonandDanielBüring(eds.).Topicandfocus:Across-linguisticperspective.Kluwer,Dordrecht.139151.Ladd,RobertD.1980.Thestructureofintonationalmeaning:EvidencefromEnglish.Bloomington:IndianaUniversityPress.Ladd,RobertD.1996.Intonationalphonology.Cambridge:CambridgeUniversityPress.Paul,Hermann.1880.PrinzipienderSprachgeschichte.Leipzig.Rochemont,MichaelS.1986.Focusingenerativegrammar.Amsterdam/Philadelphia:JohnBenjamins.Rooth,Mats.1985.AssociationWithFocus.Ph.D.thesis.UniversityofMassachusettsatAmherst.Rooth,Mats.1992.Atheoryoffocusinterpretation.NaturalLanguageSemantics1:75116.Rooth,Mats.1995.Indenites,adverbsofquanticationandfocussemantics.In:GregoryN.CarlsonandFrancisJeryPelletier(eds.).TheGenericBook.ChicagoUniversityPress,Chicago.265291.Schwarzschild,Roger.1999.GIVENness,avoidFandotherconstraintsontheplacementofaccent.NaturalLanguageSemantics7:141177.Selkirk,ElisabethO.1984.Phonologyandsyntax:Therelationbetweensoundandstructure.Cambridge,Mass.:MITPress.Selkirk,ElisabethO.1995.Sentenceprosody:Intonation,stress,andphrasing.In:J.A.Goldsmith(ed.).TheHandbookofPhonology.Blackwell,London.550569.vonStechow,Arnimvon.1991.Focusingandbackgroundingoperators.In:WernerAbraham(ed.).Discourseparticles.AmsterdamPhiladelphia:Benjamins.3784.Truckenbrodt,Hubert.1999.Ontherelationbetweensyntacticphrasesandphonologicalphrases.Lin-guisticinquiry30:219255.Uhmann,Susanne.1991.Fokusphonologie.EineAnalysedeutscherIntonationskonturenimRahmendernicht-linearenPhonologie.Tübingen:Niemeyer.Webber,BonnieLynn.1978.AFormalApproachtoDiscourseAnaphora.Ph.D.thesis.HarvardUniver-sity..LoLa9/ManfredKrifka:Canfocusaccentingbeeliminated?119