ContextandCoherence
TheLogicandGrammarofProminence
UNASTOJNIĆ
GreatClarendonStreet,Oxford,OX26DP, UnitedKingdom
OxfordUniversityPressisadepartmentoftheUniversityofOxford. ItfurtherstheUniversity’sobjectiveofexcellenceinresearch,scholarship, andeducationbypublishingworldwide.Oxfordisaregisteredtrademarkof OxfordUniversityPressintheUKandincertainothercountries
©UnaStojnić2021
Themoralrightsoftheauthorhavebeenasserted
FirstEditionpublishedin2021
Impression:1
Allrightsreserved.Nopartofthispublicationmaybereproduced,storedin aretrievalsystem,ortransmitted,inanyformorbyanymeans,withoutthe priorpermissioninwritingofOxfordUniversityPress,orasexpresslypermitted bylaw,bylicenceorundertermsagreedwiththeappropriatereprographics rightsorganization.Enquiriesconcerningreproductionoutsidethescopeofthe aboveshouldbesenttotheRightsDepartment,OxfordUniversityPress,atthe addressabove
Youmustnotcirculatethisworkinanyotherform andyoumustimposethissameconditiononanyacquirer
PublishedintheUnitedStatesofAmericabyOxfordUniversityPress 198MadisonAvenue,NewYork,NY10016,UnitedStatesofAmerica
BritishLibraryCataloguinginPublicationData Dataavailable
LibraryofCongressControlNumber:2020945722
ISBN978–0–19–886546–9
DOI:10.1093/oso/9780198865469.001.0001
PrintedandboundintheUKby TJBooksLimited
LinkstothirdpartywebsitesareprovidedbyOxfordingoodfaithand forinformationonly.Oxforddisclaimsanyresponsibilityforthematerials containedinanythirdpartywebsitereferencedinthiswork.
AppendixA.FormalDefinitionsfortheAttention-Coherence ApproachtoPronounResolution
Preface
Naturallanguagesarevastlycontext-sensitive:whatappearstobeoneandthe samestringofwordscanconveyquitedifferentmeaningsondifferentoccasions ofuse.Yet,weareabletounderstandeachothereffortlessly,inrealtime.What exactlydeterminesthemeaningonanoccasionofuse,andhowarewesogood atrecoveringit?Thisbookoffersaperhapssurprisingresponse:contraryto thedominanttradition,whichmaintainsthatthemeaningofcontext-sensitive languageisatleastpartiallydeterminedbynon-linguisticfeaturesofutterance situation,Iarguethatmeaningisdeterminedentirelybygrammar—byrules oflanguagethathavelargelybeenmissed,andtheeffectsofwhichhavebeen mistakenforextra-linguisticeffectsofanutterancesituationonmeaning.The interpretationiseffortlessbecauseitisunderwrittenbyourknowledgeofthese linguisticrules.
WhilethelinguisticnotionofcontextIdevelopshedsnewlightonthe natureoflinguisticcontent,andtheinteractionbetweencontentandcontext, itstheoreticalinterestextendsbeyondthescopeoftheorizingaboutlanguage. Contexthasbecomeastandardtoolinaphilosopher’stoolkit.Ithasbeen appealedtoinnumerousdebatesinvirtuallyeverysub-fieldofphilosophy:in epistemology,ethics,valuetheory,metaphysics,metaethics,andlogic,among others.Butthenotionofcontextthatthesedebatesassumeistypicallythe traditional,non-linguisticone.Iftheargumentsofthisbookareontheright track,however,thispresuppositionshouldbeabandonedandtheargumentsreexaminedinlightofthelinguisticconstraintsoncontext-sensitivityresolution.It ismyhopethatthisbookwillprovidenewtoolsforconstrainingandevaluating thesephilosophicaldebates,andforre-thinkingtheargumentsinthenewlight.
ThisbookhasgrownoutofaprojectthatIhavebeenworkingonfora numberofyears.Themainideasoriginatedintheworkthatledtomydoctoral thesis(Stojnić,2016).Theprojecthasdevelopedovertheyears,leadingtoa seriesofpublicationswhichhavedealtwithspecificapplicationsofthegeneral theoreticalframeworkIwasdeveloping.Thisbookisanattempttopresent anddefendthisgeneraltheoreticalframeworkindetail,andtoarticulateits keytheoreticalunderpinningsandconsequences.Thebookhasalsobenefited fromthefeedbackthisearlierworkhasgenerated.Itpresentsanattemptto incorporateandaddressmuchofthisfeedbackIhavereceivedovertheyears.
Somepartsofthebookinclude,andbuildon,someofthispreviouswork, althoughineachcase,theyhavebeenmodifiedandextendedtoincorporate
thefeedbackandaddresscriticismsthattheearlierworkhasprovoked.Parts ofPartI,specificallyChapters2–5,includepartsoftheworkIhavedone incollaborationwithMatthewStoneandErnieLepore,andthathasbeen publishedasStojnić,Stone,andLepore(2017)in LinguisticsandPhilosophy.I owetremendousgratitudetomyco-authors.Ialsothankthemforallowingme toincorporatesomeofourearlierworkinthepresentmonograph.Anyfaults introducedinextending,andbuildingon,thisworkare,ofcourse,entirelymy own.PartsofPartIIarebasedonStojnić(2019)whichwaspublishedin Noûs, andPartIIIbuildsonStojnić(2017b)whichwaspublishedin Philosophyand PhenomenologicalResearch.PartsofPartIVoverlapwithStojnić(2018)inthe OxfordUniversityPressvolume Beyond SemanticsandPragmatics,editedby GerhardPreyer,butmostofitisnovel.Thetwoappendices,AandBcontain theformalsystemsdevelopedinStojnić,Stone,andLepore(2017)andStojnić (2017b),slightlyupdatedtoaccommodatesomenovelexamples.
Ihavehadtheopportunitytopresentthematerialincludedinthisbook— sometimesinitsveryearlyform—atvariousconferences,workshops,and venuesoveranumberofyears,includingthephilosophydepartmentsatThe AustralianNationalUniversity,UniversityofBarcelona,UniversityofBelgrade, BostonUniversity,ColumbiaUniversity,CUNY,DartmouthCollege,MIT, NYU,UMassAmherst,UCSD,OxfordUniversity,RutgersUniversity,UniversityofPittsburgh,UniversityofToronto,thelinguisticsdepartmentsatNYU, UniversityofMaryland,andUMassAmherst,theILLCattheUniversityof Amsterdam,theInstituteforLanguage,CognitionandComputationatthe UniversityofEdinburgh,theUniversityofConnecticutLogicGroup,the6th CSLIWorkshoponLogic,Rationality,andIntelligentInteractionatStanford University,PhLiP2016,CUNYCognitiveScienceSpeakerSeries,and2015 PhilosophyofLanguageandLinguisticsConferenceatIUCinDubrovnik.I havetaughtpartsofthebookinagraduateseminarinphilosophyoflanguage atColumbiaUniversityintheFallof2017andtheSpringof2019,agraduate seminaroncontext-sensitivityatPrincetonUniversityinthefallof2018,ina seriesoffourguestlecturesat2017LSASummerInstitute,andat7thNASSLLI. Iamthankfultotheaudiencesandparticipantsattheseeventsfortheirfeedback. Ihavealsobenefitedfromthehelpfulcommentsonearlierwrittenversions ofpartsofthismanuscript,sometimesintheirveryearlyform,fromDaniel Altshuler,JohanvanBenthem,ElisabethCamp,SimonCharlow,CianDorr, AndyEgan,ThonyGillies,MichaelGlanzberg,GilHarman,JohnHawthorne, RobertMay,StephenNeale,CoryNichols,JeffKing,AlexLascarides,Ernie Lepore,JimPryor,StephenSchiffer,MatthewStone,GalaStojnić,andtwo anonymousreadersfortheOUPwhosecommentsandsuggestionssignificantly improvedthefinalversionofthismanuscript.Ilikewiseowegratitudetoa numberofcolleaguesforextensivediscussionsofthematerialincludedin
themanuscript.Iwouldparticularlyliketothank:JoshArmstrong,Nicholas Asher,ChrisBarker,BobBeddor,DavidBeaver,MariaBittner,JustinBledin, PaulBoghossian,DavidBraun,FabrizioCariani,NateCharlow,SamCarter, JonathanCohen,SamCumming,KostaDošen,JaniceDowell,PetervanElswyk, BrandenFitelson,MelissaFusco,GraemeForbes,KathrinGlüer-Pagin,Jeremy Goodman,SimonGoldstein,GabeGreenberg,AlanHájek,DanHarris,Ben Holguín,DanHoek,LucaIncurvati,AndyKehler,AngelikaKratzer,Magdalena Kaufmann,StephenKaufmann,KarenLewis,BarryLoewer,BrianMcLaughlin, EliotMichaelson,MattMoss,SarahMurray,EthanNowak,BarbaraPartee, PaulPietroski,PeterPagin,LewisPowell,BrianRabern,HannahRohde,Jessica Rett,DanielRothschild,CraigeRoberts,DavidRosenthal,PaoloSantorio,Alex Sandgren,JonathanSchaffer,PhilippeSchlenker,TedSider,AdamSennet, BrettSherman,MandySimons,GiuseppeSpolaore,WillStarr,KristenSyrett, ZoltánSzabó,RichThomason,FrankVeltman,BrianWeatherson,Timothy Williamson,MarkWilson,andSethYalcin.
AnetaStojnićofferedinvaluableassistanceinindexingthebook.Iamalso thankfultotwofelines,BettyandYoca,forinspiringsomeoftheexamples.
IamthankfultotheeditorfortheOUP,PeterMomtchiloff,forhissupport, patience,andguidance.
Finally,Ioweimmensegratitudetomyfamily:myparents,JasnaandDanko Stojnić,andmysistersAnetaandGalaStojnić.Withouttheirunconditionallove andsupport,andtheirtirelessencouragement,Iwouldhaveneverbeenableto completethismanuscript.
UnaStojnić, DepartmentofPhilosophy, PrincetonUniversity, May2020.
Introduction
WhatIcommunicatewith‘Givemethat,’pointingatabook,differsfromwhat Icommunicatebytheverysamestringofwordswhenpointingatacup.Your actionsandmyexpectationslikewisedifferinthesetwocases.Atthesametime, thereferentofanyparticularuseof‘that’istypicallyunambiguousandrecovered effortlessly.Whatdeterminesthereferentofexpressionslike‘that’onanoccasion ofuseandwhichresourcespermitustorecoveritsoeasily?
Everyoneagreesthatwhatwecancommunicateistosomedegreeconstrained bygrammar.However,thereceivedview,onethatmightseem primafacie obviouslycorrect,isthatgrammarunderdeterminescontent:mosttheoristsbelieve thattheroleofgrammarislimited,sincetheinterpretationlargelyrelieson generalreasoningaboutthespecificcommunicativesituationsandintentionsof thespeakersinthosesituations.Whatwesaywhenwespeakdependsnotsolely onthewordsandexpressionsweuse,butalsoonthecircumstancesinwhich thesewordsandexpressionsareused,andtheintentionsbehindsuchuses.For instance,whileitisclearthatwhatanutteranceof‘Givemethat’meansispartly determinedbygrammar—bythemeaningofwords‘Give,’‘me,’‘that,’andthe waytheyareputtogetherinaccordancewiththerulesofEnglish—itsmeaning alsoseemstodependontheaccompanyingpointinggesturetowardthedesired object.Indeed,itiscommonlyassumedthatseeminglynon-linguisticfeaturesof utterances,suchasdemonstrativegestures,playacrucialroleindeterminingthe meaningofmostcontext-sensitiveexpressions.Toseewhat’satstake,consider thedifferencebetweenanutteranceof(1)and(2):
(1) Iamhappy.
(2) Sheishappy.
Both(1)and(2)expressdifferentcontentsondifferentoccasionsofuse:ifIwere toutter(1),Iwouldhaveutteredsomethingdifferentthanwhatyouwouldhave, hadyouutteredit,andsimilarly,ifIweretoutter(2)pointingatMary,Iwould havesaidsomethingdifferentthanifIdidsopointingatSue.Andwhile‘that,’ ‘she,’and‘I’allhaveaspecificlinguisticmeaning(their character,toborrowthe terminologyfromKaplan(1989a,b)),alexicallyencodedrulethatspecifiesa referentgivenacontext,‘she’and‘that’requiresomethingbeyondtheirlinguistic
meaninginacontextinordertopickoutaparticularreferent,inawaythat‘I’ doesn’t.Thatis,themeaningof‘I’fullydeterminesitsreferentasafunctionof context:itdoesn’tmatterwhomthespeakerintendstoreferto,orwhomshe ispointingat,orwhoissalientintheutterancesituation,herutteranceof‘I’ alwaysdenotesher.NomatterhowmuchIintendtouse‘I’topickoutsomeone else,Icannotdoso.Bycontrast,thelinguisticmeaningof‘she,’or‘that,’does notbyitselfcompletelydeterminethereferentgivenacontext. Primafacie at least,itdoesnotfullyconstrainwhomthespeakercanrefertobyusing‘she’(or ‘that’),evengivenaspecificsituation.Afterall,shecanchoosetopointatany givenfemale(orthing)inthecontextualsituation,orexploitsomeotherclueto makeherintentiontorefertoaspecificindividualmanifest.Andthisaffectsthe strategiesthattheaudienceexploitsininterpretation:whileallittakestofigure who‘I’picksoutistoseewho’sspeaking,evenifyouknowthatthelinguistic meaningof‘she’requiresthat,roughly,thereferentbeasalientfemale,youstill needtoreasonaboutthecontextualcuesandspeaker’sintentionstodetermine whichreferentisselectedonagivenoccasionofuse.1
Theexpressionsthatsharethisfeaturewith‘I’areknownas pureindexicals, andthosethatbehavelike‘that’and‘she’areknownas truedemonstratives.2 Pureindexicalsarethoughttobefew(perhapsonly‘I’).Mostcontextsensitiveexpressionsareunderstoodtobeliketruedemonstrativesinrequiring someextra-linguistic,contextualsupplementation.Ifthisiscorrect,thenthe interpretationof(most)context-sensitiveexpressionsisunderdeterminedby grammar;onlyoncetheirlinguisticmeaningissupplementedbycontextual,
1 Isn’tthisdistinctionsuperficial?Isn’t,afterall,thefactthatsomeoneisspeakinganon-linguistic featureofthecontext?Andcouldn’tonethinkofthemeaningof‘she’asautomaticallydelivering theprominentfemaleindividual,inthewaythat‘I’automaticallydeliverswhoeverisspeaking?This superficialsimilarityismisleading.Thepointisthatwhileitissufficientforthespeakertoutter‘I’for ittosuccessfullypickoutthespeaker,itisn’tsufficienttomerelyutter‘she’tosuccessfullydetermine aspecificreferent;somefurtherworkisneededtodeterminesomefemaleasthemostprominent, intherelevantsense.Andthisiswhy,thestandardaccountmaintains,onehastoreasonfroma complexsetofdiverseepistemiccuestowardidentifyingwhothespeakerhadinmindinuttering ‘she,’butnotinuttering‘I.’Andwhileitistruethatitisanon-linguisticfactthatIamspeakingwhen Iutter‘I,’thisdoesn’tmeanthatnon-linguisticfeaturesofutterancesituationfixthemeaningof‘I’ onanoccasionofuse:aswith‘I,’onehastolookintotheworldtodeterminewhoMaryiswhenthe speakerutters‘Maryissmart’;thatMaryiswhosheisisanon-linguisticfact,butnotthefactthat determinesthat‘Mary’referstoMary.
2 Thisterminology,andthecorrespondingdistinctionbetweenthesetwoclassesofcontextsensitiveexpressionsisduetoKaplan(1989b).Thereisasignificantcontroversyoverwhetherany expressionsaregenuinepureindexicals.Eventhepureindexicalstatusof‘I’hasbeencontested (cf.,e.g.,Nunberg,1993).However,evenif‘I’canallowforcertainkindsofnon-rigiduses,such usesareconstrainedinwaysunlikeanythingwewitnesswithdemonstratives.Specifically,evenin suchcases,theinterpretationofthefirst-personindexicalissignificantlyconstrainedbyitslinguistic meaning,notrequiring,orevenallowing,theopen-endedabductiveinterpretiveprocessinvolvedin recoveringtheintendedreferentof‘she’;speakerintentionscannotdetermineanarbitraryreferent asthereferentof‘I’byexploitingtheflexibilityoftheopen-endedsetofcontextualcuesthatwould revealsuchanintention.
extra-linguisticmechanisms—forexample,speakerintentions,orother contextualcues—dotheyhaveaspecificinterpretation.3Sothequestion becomes:whatmechanismsplaytheroleofsupplementationthatfixesthe meaningoftruedemonstratives?
Themostcommonansweristhat,ultimately,itisthespeaker’sreferential orcommunicativeintentionsthatplaythisrole.Thethoughtisthatwhatan utteranceof(2)meansdependsonwhomthespeakerintendedtotalkabout, whomsheintendedtorefertoorhadinmind.Onthisunderstanding,various epistemiccues—forinstance,pointinggestures,orsalienceofanindividual intheutterancesituation—serveasguidestowardidentifyingthismeaningdeterminingintention,butitistheintentionthatdeterminesthemeaning.⁴The ideais,roughly,thatifthespeakerisintendingtorefertoMary,sheshould utter(2)toconveythatMaryishappy,justincaseinthiscontext,thereare enoughresourcesforheraudiencetorecognizeherintention,thatis,justincase herintentionismadeappropriatelymanifest.Correspondingly,whentasked withinterpretation,heraudiencecandrawonanyepistemiccuetheutterance situationmakesavailableinthecourseofidentifyingthespeaker’sintention.But itisbecausethespeakerintendedtotalkaboutMarythat‘she’picksoutMary.
Theideathatinterpretationreliesonmind-reading—onfiguringoutspeakers’ communicativeintentions—goesback,atleast,toLocke(1689).Locke,however, thoughtofmind-readingasmediatedsolelybyexpressionsuttered,maintaining thatcommunicationisentirelyamatterofencodinganddecodingcontent throughlinguisticforms.Thisiscompatiblewiththemeaningfullybeingdeterminedbythegrammar.Thecurrenttheorizing,bycontrast,followsinthefootstepsofGrice(1975)andKaplan(1989a),inunderstandingthemind-readingto
3 Whichexpressionsarecontext-sensitiveisamatterofcontroversy;somebelieveonlypure indexicals(e.g.,‘I’)anddemonstrativepronouns(e.g.,‘she,’‘that’)arecontext-sensitive(Cappelen andLepore,2005),othersbelievethatvirtuallyallexpressionsarecontext-sensitive(Travis,1989). Mostwouldagree,however,thatmorethanjustafew,butfewerthan(almost)allexpressions arecontext-sensitive.Yet,thereisstillgreatdisagreementoverwhichparticularexpressionsare (andinwhatway).(See,forinstance,thedebatesoverwhethermodals(Egan,Hawthorne,and Weatherson2005;Yalcin2007;vonFintelandGillies2008;KolodnyandMacFarlane2010;Dowell, 2011),orpredicatesofpersonaltaste(CappelenandHawthorne2009;Egan2010;MacFarlane 2014),orknowledgeascriptions(DeRose,1995,2009;Cohen,1998;Hawthorne,2004;Stanley,2005; SchafferandSzabó,2013),arecontext-sensitive.)Howeverthesedebatesaresettled,mostauthors maintainthatinsofarastherearesomecontext-sensitiveexpressions,mostofthosebehaveliketrue demonstrativesintherelevantsense—thatis,theyrequireextra-linguisticsupplementation.
⁴ Ofcourse,whilemostauthorswouldmaintainthatintendingtorefertoanobject o isnecessary forfixing o asthereferentofanutteranceofademonstrative,fewwouldthinkthat,unlessqualified insomeway,thiswouldalsobesufficient.FewwouldbitethebulletandsaythatifIpointtothe statueinfrontofmesaying‘That’smyfavoriteobject,’itissufficientthatIintendedtheMoontobe thereferentof‘that,’formyutterancetomeanthattheMoonismyfavoriteobject.Myintentionhas tobeinprinciplerecognizable.Similarly,somewouldarguethelinguisticmeaningofanexpression constrainswhatonecanreasonablyintendwithitsutterance.Forinstance,Icannotreasonably intendtopickoutBillwith‘she,’ifheisaman.Forfurtherdiscussiononhowtoqualifytherequisite intentions,see,forinstance,Schiffer(1972,1981,2005);Neale(2004);King(2014b).
involvenotmeredecoding,butreasoningonthebasisofnon-linguisticfeatures ofutterancesituation,becauseeithersuchfeatures,ortheintentionstheyreveal, partlydeterminethemeaningofexpressions.OnaLockeanunderstanding,the speakerwhointendstoconveyamessageshouldchoosethelinguisticformthat expressesthatmessage(inagivencontext),andtheaudience,inperceivingthata particularformwasuttered,istorecoverthemessagebyconsultingthelinguistic rulesthatassignthemessageasthemeaningoftheformuttered.Inother words,theyinferthatthespeakerintendedthemessagebecausetheyuttered theformthatexpressesit.OntheGricean/Kaplaneanpicture,instead,the audiencehastodeterminethespeakerintentionsfirstinordertoevendetermine whatthemessageexpressedbytheformutteredinthecontextis.Itisthis Gricean/Kaplaneanmodelthathasdominatedthediscussionsaboutcontextsensitivityresolution,anditisthismodelthatIwillchallengeinthisbook.
Theviewwhichassociatesspeakerintentionswithameaningdetermining roleis,indeed,byfarthemostdominantone.⁵Butitisimportanttonote thateventheviewsthatdissentfromit,stillinsistonthecrucialroleof extra-linguisticfactorsintheresolutionofcontext-sensitivity.Forinstance,the mostnotablealternativemaintainsthatthenon-linguisticfeaturesofcontext— objective,speaker-independent,aspectsofthereal-worldsituationinwhichan utterancetakesplace—arewhatultimatelyfixesthemeaningofcontext-sensitive expressions(Wettstein,1984;Glanzberg,2007).Onthiskindofviewthefacts aboutthereal-worldsituation—whichgesturesaccompanytheutterance,who orwhatissalient,orrelevant,orotherkindsofreal-worldknowledgetheagents candrawon—don’tmerelyserveasepistemiccuesthathelpidentifyreferential intentions,theythemselvesfixthemeaningofcontext-sensitiveexpressions. So,thepresenceofareferentialintentionisnotnecessaryfordetermining reference:ifMaryisobjectivelymostsalient(wheresalienceisspelledout independentlyofspeaker’smentalstates),thentheutteranceof‘she’refersto her,regardlessofwhomthespeakerhadinmind.⁶Glossingoverthenuances differentiatingvariousversionsofthedominantapproach,Ishallcallthetypeof viewthatmaintainsthatthespeaker’sreferentialorcommunicativeintentions arenecessaryfordeterminingthemeaningofacontext-sensitiveexpression, the Intentionalist view.Ishallcalltheviewthatmaintainsthattheobjective non-linguisticfeaturesofcontextcanfixthemeaningofacontext-sensitive expression,the Objectivist account.Asbothtypesofaccountemphasizethe
⁵ See,forexample,Schiffer(1972,1981,2005);Kaplan(1989a);Neale(1990,2004);King (2014a,b); interalia
⁶ See,forexample,Wettstein(1984),Reimer(1992),Glanzberg(2007),Lewis(2019)fordiscussionofthistypeofview.Onecanalsoholdahybridview:sometimesthespeaker’smentalstates, andsometimestheobjectivefeaturesofthereal-worldsituationfixthesemanticcontent,orperhaps, differentmechanismsarerelevantfordifferenttypesofcontext-sensitiveexpressions.
underdeterminationofmeaningofacontext-sensitiveexpressionbygrammar, andtheneedforextra-linguisticsupplementation,Ishalldubsuchaccounts Extra-linguistic.
Thisbookarguesforanovelalternativetotheextra-linguisticapproach.I shallarguethatcontext-sensitivityresolutionisamatteroflinguistic,ratherthan extra-linguistic,mechanisms;itisgovernedbylinguisticrules,whichdetermine particularvaluesofcontextualparametersonwhichcontext-sensitiveitems dependfortheirmeaningatanygivenpointindiscourse,independentlyof speakerintentions, and non-linguisticfeaturesofutterancesituation.Against thetradition,Ishallarguethatthereasonwecaninterpretcontext-sensitivity soeffortlesslyisbecauselanguage—grammar—itselfisfarmoreexpressiveand pervasivethanhasbeenassumed;theresolutionofcontext-sensitivityisentirely amatteroflinguisticconvention.Itisthelinguisticrulesthatdeterminewhat ‘that’or‘she’picksoutonanoccasionofuse.And,moreover,inrecoveringthis contentitisthisnarrowsetoflinguisticcuesthatweexploit.
TheviewIshalldevelopanddefendpositslinguisticconventionsthathave goneunnoticed,becausetheirprincipaldomainsareentirediscourses,notjust theirconstituentwordsandsentences.Whileitisnotcontroversialthatthe waysentencesareconstructeddependsonconventionsofsyntaxandsemantics whichspecifytherulesbywhichindividualexpressionscombine,Isimilarly argueforrules—discourseconventions—thatspecifyhowindividualsentences combinetoformadiscourse.Theseconventionsgovernhowspeakersorganize utterancesintolargerunitsthataddressparticulartopicsandanswerquestions aboutthem;Iarguetheyalsodeterminetheresolutionofcontext-sensitivity. TheimportantsetofmechanismsIwilldrawonarethesocalledmechanismsof discoursecoherence,whichprovideanimplicitorganizationtothediscoursethat signalshowindividualutterancesareorganizedintoacoherentwhole(Hobbs, 1979;Kehler,2002;AsherandLascarides,2003).Toillustratewhatkindsof mechanismsIhaveinmind,consider(3):
(3) PhiltickledStanley.Lizpokedhim.(Smyth,1994)
Outoftheblue,speakerstendtounderstandanutteranceof(3)inoneof twoways:eitherLiz’sactionisaresultofPhil’s(perhapsconveyingthatshe’s actingindisapproval),orheractionisdescribedassimilartoPhil’s.Inone case,thediscourseisorganizedaroundacause-effect(i.e.,event-result)relation, describingoneeventualityasresultingfromtheother,andintheotheraround aresemblanceone,comparingtwoeventualities.This,crucially,hasaneffect ontheinterpretationofthedemonstrativepronoun‘him’:intheformercase ‘him’isunderstoodtorefertoPhil,andinthelattertoStanley.Thiskind ofcorrelationbetweenestablishingcoherenceindiscourseandtheresolution
demonstrativepronounshasbeenwellconfirmed.⁷Howevermostextanttheoriestreatitasapragmaticdefault.Standardcoherencetheoreticaccounts interpretthiscorrelationasevidenceofaninferentialrelationshipbetweena speaker’sintentioninorganizingthediscourseandherreferentialintentions.I arguethisisamistake:thereisatighterconnectionbetweendiscoursecoherence andpronounresolution,oneunderscoredbylinguisticconvention.Moreover,I willarguethatthistightconnectionextendstoothertypesofcontext-sensitive expressions,beyondtheresolutionofdemonstrativepronouns.Inshort,Iargue thatdiscourserelationsthatconnectandorganizeutterancesareapartofthe grammarofalanguage,andthattheygoverntheresolutionofcontext-sensitivity asamatterofgrammar,too.
Whileatfirstglanceitmayseemradical,Ishallarguemyaccountisboth empiricallysuperiortotheextra-linguisticalternativesandhaswelcometheoreticalconsequences.Conversely,whiletheextra-linguisticapproach prima facie appearsobviouslyontherighttrack,andthoughitsendorsementisalmost universal,Ishallarguethatitisbothempiricallyinadequateandleadstoradical andtroublingconsequencesfortheunderstandingofthenatureofsemantic content,itsinterplaywithcontext,anditsrelationtologic.
Inthisregard,Iwillexaminetherecentandgrowingtrendintheliteraturewhichadvocatesabandoningthestandardtruth-conditional,propositionalaccountsofmeaning.⁸Thetraditionalparadigm,accordingtowhichthe meaningofasentenceisidentifiedwithapropositionthatis,ordetermines, atruth-conditionwhichrepresentstheworldasbeingacertainwayhasbeen hugelyinfluential.Theidentificationofmeaningwithrepresentational,truthconditionalcontentgoesbacktoatleastFrege(1892,1918),andisreflectedin someofthemostdominantresearchprogramsinphilosophicalandlinguistic theorizingaboutmeaningtothisday.Themotivationforthedeparturefromthe truth-conditionalparadigmstemsfromthepuzzlingbehaviorofcertaincontextsensitiveexpressions,whichappeartoevadethestandard,extra-linguistic, contextualtreatment:context,asstandardlyunderstood,failstodetermine
⁷ Consult,forexample,Kehler(2002);Wolf,Gibson,andDesmet(2004);Kehleretal.(2008); Kaiser(2009);andreferencestherein.
⁸ See,forinstance,Gillies(2004,2010);Swanson(2006);vonFintelandGillies(2007);Yalcin (2007,2015);Willer(2014);Moss(2015);Charlow(2015,2020);Starr(2016); interalia.Mostof theseaccountsdrawcloselyontheinfluentialworkofVeltman(1985,1996),whichoffersaformal framework,aswellasalinguisticargument,forabandoningpropositionalism.(InotethatMoss (2015)issomewhatofanoutlieronthislist,asheraccount,whilenon-propositionalist,maintains thattherelevantbitsofdiscourseexpressinformationalcontent,modeledprobabilistically.Ishall discussthisnuanceinmoredetailbelow.)Theseneo-expressivistshavetheirprecursorsinthework ofearlynon-cognitivists,suchasAyer(1936);Stevenson(1937);Hare(1952),aswellasmorerecently Gibbard(1981),butalsointheworkonconditionalsstemmingfromAdams(1975).Unlikemanyof thesepredecessors,however,theyattempttogivealinguisticargumentagainstthetruth-conditional accounts.
truth-conditionalcontentforsomeutterancesinvolvingsuchexpressions.Thus, thepuzzlingbehaviorisarguedtomotivatenon-propositionalaccountswhich denytheseutterancespropositionalcontent.Instead,theseaccountstendto takethecharacteristicdynamiceffectanutterancehasonthecontexttobethe keyaspectofitsmeaning.Thedynamiceffectoftheproblematicutterances— theircontext-changepotential—isirreducibletopropositionalcontent,andis incompatiblewithsuchutterancesexpressing(non-trivial)propositionalmeaning.⁹Atthesametime,theproblematicexpressions primafacie giveriseto apparentfailuresofcertainclassicallyvalidpatternsofinference,suggesting aneedforadeparturefromclassicallogic.Thisistakenasfurthersupport forthenon-propositionaltreatmentasthevariousimplementationsofnonpropositionalismdepartfromclassicallogicbyinvalidatingatleastsomeofthese patterns.
Ishallarguethatbothofthesetypesofpuzzlingbehaviorthatmotivate non-propositionalist,non-classicalaccountscanbetracedbacktotheunderlyingassumptionoftheextra-linguisticmodelofcontext-sensitivityresolution. Philosophers,operatingunderthetraditionalassumptionsabouttheresolution ofcontext-sensitivity,wereledtoabandonthepropositionalistparadigmandto concludethatthereisadeeptensionbetweenclassicallogicandthesemantics ofanaturallanguagediscourse,oncefacedwithaclassofexpressionsthatdo notfitsuchcontextualisttreatment.Buttheproblemisnot,Ishallargue,in theideathattheseexpressionsexpresstruth-conditionalcontent;theproblem isintheunderlyingassumptionofhowacontextoperatestodeterminethese truth-conditions.Oncewehaveamorenuancedunderstandingofthelinguistic mechanismsdrivingcontext-sensitivityresolution,wewillseethatwecan explainawaythepuzzlingphenomenawhilestillmaintainingthatutterances containingtheproblematicexpressionsexpressnon-trivialrepresentational, truth-conditionalmeaning,and,evenmoresurprisingly,wecandosowithin aframeworkthatpreservesclassicallogic.
However,whileonmyaccountutterancesexpresstruth-conditionalcontent, that’sonlyoneaspectoftheirmeaning.Theaccountshareswithnonpropositionaliststheideathatanimportantaspectofmeaningisthe dynamiceffectanexpressionhasonacontextinwhichitisuttered.Butthe non-propositionalistsaremistakenbothaboutthenatureandtheeffect
⁹ Itisworthbearinginmindthatnon-propositionalistaccountscomeindifferentflavours.Some non-propositionalistswillwanttoargue,forinstance,thattherelevantbitsofdiscourseexpress aspecifictypeofnon-propositionalcontent,ratherthanmerelyinducingaparticulardynamic effectonthecontext(see,e.g.,Schroeder(2013)orMoss(2015,2018)).Ingivingthispreliminary, umbrelladescriptionofthenon-propositionalistapproachtomeaning,Iamthusskirtingaround someimportantdistinctionsbetweendifferentproponentsofnon-propositionalism.Iwilldiscuss thoseinmoredetailinwhatfollows.Thepreliminary,tentativecharacterizationsufficesatthisstage.
ofthisdynamicaspectofmeaning.Thedynamicmeaningencodesthe effectofdiscourseconventions,specifyinghowtheymanipulateandchange parametersofcontext,inawaythatdeterminesthepropositionalcontent expressedbythesubsequentdiscourse.1⁰Thus,farfrombeingincompatible withdynamicmeaning,myaccounthasitthatthedynamicmeaningplaysa crucialroleindeterminingpropositionalcontent.Theaccountthusreconciles thedynamicapproachtomeaning—dynamicsemantics—withthetraditional representational,truth-conditionalapproachtocontent.
Myaccounthasfurtherphilosophicalramifications.Philosophersfrequently identify(purported)context-sensitivityinphilosophicallyinterestingexpressionslike‘know,’or‘good,’andappealtoittoshedlightonproblemsinvolving theconceptstheseexpressionsdenote,oftenleadingtoradicalconclusionsabout theunderlyingphenomena.Buttheyalsoassumethattheseexpressionsget theirvaluesfixedincontextpartlythroughspeakerintentions,ornon-linguistic featuresofutterancecontext.Andoftenthedetailsofsuchargumentsrely ontheseunderlyingassumptionsaboutwhatcontextis,andhowitinteracts withmeaning.Ifmyaccountisontherighttrack,thenahostofdebates overcontextualisttreatmentsofvariousphilosophicallyimportantexpressions restsonaflawedextra-linguistic,flexible,conceptioncontext,andcontextdependenceinphilosophicaltheorizingwillhavetobeinvokedquitedifferently thanhasbecomecustomary.
Ibeginbybrieflyoutliningthekeytheoreticalconceptsandconsiderations, emphasizingthemaindeparturesfromthetraditionalpictureandthemain theoreticalconsequencesoftheviewIshalldevelop.Iwilldelegatedetailed developmentandargumentsformymodeltothelaterchapters.
1.1Context
Traditionally,contexthasbeenunderstoodtocomprisefeaturesofutterance situationuponwhichthecontentofcontext-sensitiveexpressionsdepends, orthosewhichcanserveascuestowardrecognizingspeakerintentionsthat playmeaning-fixingrole.Agentstakethesecues—togetherwiththeirgeneral
1⁰ NotethatwhenIsaythatadiscourseexpressespropositional,truth-conditional,content,Ido notintendtodenythatcertainbitsofdiscourse,forexample,imperativesorinterrogatives,donot simplydenotepropositions.Awiderangeofresearchinlinguisticsandphilosophyoflanguage pointstowaysofintegratingtheideathatsemanticsofmoodmightrequiredistincttypesof content,withtheideathatordinaryassertionsofdeclarativeutterancesexpresspropositionalcontent (irrespectiveofthedisagreementoverwhetheralldeclarativeutterancesexpresspropositional content,whichiswhatnon-propositionalistsdeny).Fordiscussion,seeGinzburg(1994);Portner (2007,2012);CondoravdiandLauer(2012);Kaufmann(2012);Roberts(2012);Charlow(2014); MurrayandStarr(2018);Ciardelli,Groenendijk,andRoelofsen(2019);Starr(2020); interalia
worldknowledge—asinputtoabductivereasoningprocessaimingtorecoverthe overallmostplausibleinterpretation.Someparametersofcontextprovidethe valueofacontext-sensitiveexpressionstraightforwardlyandautomatically—if thespeakersays‘Iamhappy,’thevalueof‘I’willbethespeaker,automatically. Butinmostcases,bothwhichaspectsoftheutterancesituationarerelevant— intentions,salience,variousobjectivefactsabouttheutterancesituation,orsome combinationofthose—andhowtheyshouldbecombinedtogetherandweighed againstoneanother—willrequireabductivereasoning,thegoalofwhichis torecovertheoverallmostplausibleinterpretation.Forinstance,supposeI utter(4):
(4) Sheisfast.
Tounderstand(4),youwillhavetofigureoutwho‘she’denotes.Evenifyou knowthatthemeaningof‘she’dependsonwhichfemaleIintendedtorefer to,orwhichoneisthemostsalient,youwillstillhavetoengageinreasoning astowhichcontentthisdetermines.Todoso,youwillhavetotakevarious thingsintoaccount.HaveIpointedatsomeone?Issomefemalereferentalready theunique,ormostprominent,topicofdiscussion?Issomefemalereferent (uniquely)salientinoursurrounding?Andsoon.Similarly,youwillhaveto figureoutwhatwasmeantby‘fast.’Arewediscussingasport—swimming,or trackandfield?Ifso,arewetalkingaboutamateursorprofessionalathletes? Maybewearediscussinghorses,orcheetahs?Orarewe,perhaps,engagedin ataskthatwoulddesignatepossessionofsomedegreeofspeedsignificantfor thattask(e.g.,reachingadrowningpersonwithinsomelimitoftime)?Doesthe situationweareinotherwisemakeanyoneoftheseparameterssalient,ordoes itmakeitplausiblethatthespeakerintendedaparticularinterpretation?These factorsarecomplex,open-ended,andneednotallpointinthesamedirection; consequently,reasoningtowardthemostplausibleinterpretationisdefeasible andopen-ended.
Further,featuresofanutterancesituationarenotstatic—aswespeak,differentthingscanbecomesalient,othersrecedeinsalience.Thefollowingexample, fromLewis(1979)illustratesthepoint:
(5) [A]cat,Bruce,[...]hasbeenmakinghimselfverysalientbydashingmadly about.Heistheonlycatintheroom,orinsight,orinearshot.Istartto speaktoyou:Thecatisinthecarton.Thecatwillnevermeetourothercat, becauseourothercatlivesinNewZealand.OurNewZealandcatliveswith theCresswells.Andtherehe’llstay,becauseMiriamwouldbesadifthecat wentaway.
Atthebeginningofthediscourse,Bruceisthemostsalientcat;howeverasthe discourseunfoldsandswitchestothedescriptionoftheNewZealandcat,Bruce recedesinsalience,andtheNewZealandcatbecomesthemostsalientone. Consequently,thedescription‘thecat,’whichdenotesthemostsaliententity thatisacat,nolongerpicksoutBruce,buttheNewZealandcat.11
Onthecommonunderstanding,trackingsuchfeaturesislikewiseamatter ofrational,abductivereasoning:theagentsunderstandthatthevalueofaparticularparameterchanges,bytakingintoaccountallsortsofcues,background knowledge,andconsiderations.(E.g.,itonlymakessensethatthecatthat’smost recentlymentionedandthatwearecurrentlydiscussingisthemostsalientwhen ‘thecat’isinterpreted,justasitmakessensethat—ifwehavenotmentionedany cats—themostsalientone,ortheoneintended,istheonethat’sperceptually presentintheutterancesituation.)Andsuchreasoningisbyitsnatureholistic anddefeasible.
So,onthetraditionalpicture,thecontextcarriesbothlinguisticandnonlinguisticinformationthatonecanequallydrawonininterpretingacontextsensitiveitem,andthatcaninteractincomplexwaysindeterminingthecontent ofacontext-sensitiveitem.Thecontentofanutterance,thus,dependson variableandchangingfeaturesoftheutterancesituation,encodedinsuch information.Ofcourse,thestandinglinguisticmeaningofanexpressionplays aroleinfixingthecontent:that‘she’asamatterofmeaningpicksoutthe third-person,singular,femaleindividualsalientorintended,constrainswhothe expressioncanpickoutinacontext.Butaccordingtothetraditionalpicture thenon-linguisticinformationthecontextcarriesplaysakeyroleinfixingthe meaning,andso—ontheinterpretiveside—theaudiencewillhavetoengagein holistic,abductivereasoningaboutsuchfeaturesinordertorecoverthecontent.
Whilethetraditionalpicturemightbe primafacie natural,itis,Ishallargue, mistaken.Accordingtotheaccountofcontextandcontext-sensitivityIshall develop,thefeaturesofcontextthatfixthemeaningofcontext-sensitiveexpressionsare linguistically determined,throughtheeffectsof discourseconventions, whichtriggersemanticeffectsofthelinguisticitemsanutterance—ormore generallythediscourseinwhichitisembedded—comprises.Thecontext,I shallargue,isarunningrecordof linguistic informationthatiscontributed bydiscourse-internal,linguisticcues;therecord,inturn,fullydeterminesthe interpretationbyfixingthecontentexpressedbythediscourse.12Myconception ofcontextisthus linguistic,ratherthanextra-linguistic.
11 Thecomplexitiesconcerningthesemanticsofdefinites,thoughimportantintheirownright, canbesetasideforthepurposesofthisillustration.
12 Theideaofacontextasaconversationalrecord,orscoreboard,goesbacktoLewis(1979).But whileLewismaintainsthatthecontextisanevolvingrecordofcontextuallyrelevantparameters,he isnotcommittedtotheideathattherecordismaintainedandupdatedentirelythroughlinguistic mechanisms.
Giventhisideaofcontext,acontext-sensitiveitemreceivesitsinterpretation fullyandautomaticallyfromthelinguisticcontext,selectingthevalueofthe relevantcontextualparameterthatthecontextrecordsataparticularpointin discourseatwhichtheitemoccurs.Toillustratethegeneralidea,considerthe followingdiscourse:
(6) Bettycamein,andshesatdown.
Thecontentofthepronoun‘she’in(6),isnotdeterminedindependentlyofwhat camebefore.Thefirstconjunctsetsthestage:itmentionsafemaleantecedent, Betty,thatthepronouncanthenpickupon.Onthetraditionalpicture,we couldexplainthisbysayingthatmentioningsomeonemakesthemsalient,and sothat‘she’naturallypicksoutBetty.13Butsupposethatwhile(6)isuttered,a distinctfemaleindividual,Mary,isthemostsalient:sheisjumpingupanddown, yellingloudly,andclearlymakingherselfthecenterofourattention.Evenso, ‘she’isnaturallyunderstoodasselectingBetty.1⁴Howdoweexplainthis?Ishall argueitisapartofthelinguisticcontributionof(6)thatitforcesthisparticular interpretation:itisapartofthestandingmeaningofapropername,‘Betty,’that itaffectsthecontextsothatitsreferentismadeprominent;anditisbecause(6) exhibitsanarrativestructurethatBettyremainsthuscontextuallyprominentfor theinterpretationofthesubsequentpronoun‘she.’
Adiscoursecan,ofcourse,beambiguouswithrespecttoitsstructure. Recall(3):
(3) PhiltickledStanley.Lizpokedhim.
Thediscoursein(3)isambiguous:itcouldbeinterpretedsothatLizpokedPhil, orsothatshepokedStanley.1⁵Buttheambiguity,Ishallargue,doesnothavea sourcein(potentiallyopen-endednumberof)possiblecontextualresolutionsof thepronoun;ratherdifferentinterpretationsofthepronounhavetheirsource indifferentcoherencerelationsthatconnectthetwosentencesintoacoherent
13 Admittedly,thisisjustabasic,bare-bonesversionofthetraditionalaccount.Weshallsee, however,thateventhemoresophisticatedwaysofspellingthingsoutwithinthetraditional frameworkmischaracterizethemechanismsofcontext-sensitivityresolution:itislinguistic,rather thanextra-linguisticmechanismsthatplaythemeaning-fixingrole.
1⁴ Iamassumingthatin(6),thepronounisnotstressed,oraccompaniedbyadeicticgesture.One couldgetareadingaccordingtowhich‘she’in(6)selectsMary,ifthepronounwasappropriately stressed.Butprosodicstress(and,Ishallargue,deicticgestures),arepreciselylinguistic,languagespecific,elements,ratherthanextra-linguisticbitsofinformationthatserveasinputstoabduction.
1⁵ Again,Iamassumingnopointinggesturesarepresentin(3).
unit.1⁶IfthediscourseharborsParallel,arelationsignalingcomparisonbetween theeventsdescribedbythetwosentences,thenthepronounisresolvedto Stanley,theobjectoftheprevioussentence,whereasifitisorganizedbyResult relation,signalingthateventdescribedbythesecondsentenceresultedfromthe firstone,thepronounisresolvedtothesubjectoftheprevioussentence,Phil. Theserelations,Ishallargue,areapartofthediscoursein(3)—muchlikethe propername,‘Phil,’orthepasttense,are—and,further,theydictateaparticular resolutionofthepronoun,byaffectingthecontextinsuchawaythatoneorthe othercandidatereferentisdeemedthemostprominentreferentinthelinguistic contextinwhich‘him’isinterpreted.Inthisway,(3)(andsimilarly,(6))lacks dependenceonnon-linguisticcontext:theinterpretationofthepronounsthey featureisfullydeterminedbythepronouns’locallinguisticcontext,whichin turnisdeterminedbythelinguisticitemsinternaltothesediscourses.
Moregenerally,then,whileonthestandardpictureanexpressionhasa standinglinguisticmeaningwhichtogetherwith(potentiallyextra-linguistic) contextualinformationhelpsfixitscontentonanoccasionofuse,onmy accountexpressions,asapartoftheirstandinglinguisticmeaning,createa linguisticcontextthatdeterminesthecontentexpressedbysubsequentcontextsensitiveexpressions.Somewhatmoreprecisely,whileonthestandardpicture expressionsareassignedacharacter—whichinacontextdeliversacontent— onmyaccountexpressionsareinterpretedasinstructionstoupdatethecontext, potentiallychangingthevaluesofcontextualparametersfromtheinputcontext totheoutputone,reflectingthechangesbroughtaboutbythemeaningof eachexpression.Thissetsthevaluesofcontextualparameterswhichhelpfix thesemantic,truth-conditionalcontentofanutterancejustastheywould onthestandardpicture,byconsultingthelinguisticallyassignedcharacterof eachexpression(e.g.,giventhat,peritscharacter,‘she’picksoutthethirdperson,singular,femaleindividual,anutteranceof‘she’willselectthemost prominentelementsatisfyingthischaracterasrecordedinthecurrentstate oftheconversationalrecord).However,sincethecontextualparametersare linguisticallymaintained,thesemanticcontentexpressedislinguisticallydetermined,throughandthrough.Inthisway,anycontextualeffectsarethoseof thislinguisticallymaintainedcontext,tracedbacktotheeffectofsomelinguistic itemwithinthediscourse.
Noticethat,onmyaccount,discoursesarerich:theycomprisenotjust individualsentences(alongwithalltheirovertandcovertelements),butalso discourserelations—coherencerelationsbetweenindividualsentences.While thismightappearstrangelyunparsimonious,Ishallarguethattherearegood
1⁶ SeeHobbs(1979);Kehler(2002);AsherandLascarides(2003); interalia).Muchmorewillbe saidaboutdiscoursecoherenceinsubsequentchapters.
empiricalreasonstoembracethisrichness.1⁷Eachoftheselinguisticelements, inturn,isassociatedwithaninstructiononhowtoupdatetheconversational record,thusgraduallybuildingthelinguisticcontextwhichdeterminesthe interpretation.ItissuchinstructionsthatIcall discourseconventions.Discourse conventionsdeterminethe prominence ofpotentialinterpretationsofcontextsensitiveitems:theyunderscorethegrammarofprominence.Notice,thenotion ofprominenceIargueisrelevantfortheinterpretationofcontext-sensitive expressionsislinguistic;aswehaveseeninourdiscussionof(6),itisdistinctfrom,andcross-cuts,theworldlysalienceemphasizedbythetraditional approach.
1.2Content
Ultimately,wewanttoexplainhowwecommunicatesoeffortlesslywithcontextsensitivelanguage.Wewanttoexplainhowitisthatweareabletoconvey ourthoughtsthroughlanguage,andhowtheinformationconveyedthrough communicationcancometoguideouraction.Forinstance,supposeIwantto conveytoyoumythoughtthatitiscurrentlyraininginNYC.Iutter(7):
(7) Itisraining(inNYC).
Uponhearingmyutterance,youcometounderstandit,andifyoutakemeto bereliableandsincere,youmightcometobelievewhatIsaid,andthismight impactyouraction:youmightdecidetobringanumbrella.Asimilarexchange mighthappenifIutter,forinstance,thefollowing:
(8) Thatisyourumbrella.
whilepointingattheumbrellabythedoor.Howdosuchexchangessucceed?
Asimpleideaisthatmythought,andmyutterance,expressacontent,which representstheworldasbeingacertainway,forexample,assuchthatinit,it israininginNYC(ataparticulartime),orthattheobjectIpointedatisthe addressee’sumbrella.Bothmythoughtandmyutterancearetruejustincase theworldisindeedthisway.Itisbecauseyouunderstandthatmyutterance expressesthiscontent—andunderstandwhatitwouldtakeformyutteranceto betrue—thatyoucanunderstandmyutteranceand,ifyoudeemmereliableand sincere,youcancometobelievewhatIasserted.Thismightfurtherimpactyour
1⁷ Indoingso,Ishallbuildontheliteratureondiscoursecoherence.SeeHobbs(1979,1990); Kehler(2002);AsherandLascarides(2003);Webberetal.(2003); interalia
action(youmightdecidetograbyourumbrella).Propositionsaretakentobe (oratleastdetermine)thistruth-conditional,representationalcontent.
Butifmyutteranceof(7)or(8)expressesaproposition,thenthequestionis howisthispropositionconnectedtothesemanticcontentofthesentencein(7), thatis,(8).Thesemantictheory,ofcourse,shouldplaysomeroleindetermining thecontentexpressed.Presumably,thatImanagetoconveymythoughtwith(7), or(8),hassomethingtodowiththesemanticcontentthatthecorrectsemantic theoryforEnglishassignsto(7),thatis,(8)(relativetoacontext).1⁸
Now,thetraditionalpicture,wehaveseen,maintainsthattheproposition expressedbyanutterancesuchas(8)isn’tfullydeterminedbygrammar; theinputofextra-linguisticfeaturesoftheutterancesituationisneededto determine(atleast)what‘that’denotes.1⁹Butevenso,onecouldsubscribeto asimpleanddirectwayofconnectingthecompositionalsemanticcontentwith thecontentexpressedembodiedinwhatonemightcallthe identificationthesis. Theidentificationthesismaintainsthatthecompositionalsemanticcontentof asentence(ofalanguage)inacontextjustisthecontentassertedbyitin thatcontext;thatis,thatthecompositionalsemanticcontentofasentence (ofalanguage,onanoccasionofuse)justisitsso-called assertoriccontent (onthesameoccasionofuse).Indeed,traditionally,identificationhasbeen assumed,whereboththecompositionalsemanticcontent(ofasentence)and theassertoriccontenthavebeenidentifiedwiththenotionof propositional or truth-conditional content.Propositionsarewhatwebelieveandassert;andthey arethecompositionalsemanticcontentsentencesexpressrelativetocontexts.
Assumingthatgrammarunderdeterminestheresolutionofcontext-sensitive itemsandthattheinputoftheextra-linguisticcontextisneededtodetermine thepropositionexpressed,howcanonemaintainidentification?Theideaisthat onecanrelativizeassignmentsofsemanticcontenttoexpressionstocontexts.A semantictheoryforalanguage ℒ assignsacompositionalsemanticcontentto eachexpressionofℒandspecifiestherecursiveruleswherebythesecontentsare compositionallycombinedtoformcontentsoflargerexpressionsof ℒ.Butto handlecontext-sensitivity,onemaintainsthatcompositionalsemanticcontent isassignedonlyrelativetoacontext.Inthisway,theinputofthenon-linguistic contextplaysaroleduringsemanticcomposition,andthuscontext-sensitivity canberesolvedduringthecourseofsemantic-composition,sothatinacontext, asentenceexpressesacompositionallydeterminedproposition.2⁰
1⁸ SeeStojnić(2017a)foramoredetaileddiscussionofthispoint.
1⁹ Indeed,thesamecanbesaidof(7),ifit,too,harborscontext-sensitiveelements:forinstance, ifweadoptacontextualisttreatmentoftense.
2⁰ Noteveryonewouldagreethatcontext-sensitivityshouldbeaccommodatedbyrelativizing semantic contenttocontexts.Someauthorswouldmaintainthatsemanticcontentisratherminimal and(largely)context-insensitive,andthattheapparentcontextual-sensitivityisduetonon-semantic
Identificationwouldthusprovideastraightforwardaccountoftheroleof semanticcontentinthetheoryofcommunication.Unfortunately,ithasbeen forcefullycriticized,aspropositionalcontentturnedouttobeinadequateforthe roleofcompositionalsemanticcontent,duetoitsfailuretoaccountforall theembeddingbehaviorofcompositionalsemanticcontent.21Butevenamong thetheoriststhatrejectidentification,mostwouldstillexpectthatsemantic contentsomehowdeterminestheinformational,propositionalcontent,the contentofassertion.Insteadofidentification,theysubscribeto determination, thethesisthatsemanticcontent(atleastpartially)determinesthecontentof assertionandbelief—thatis,thatthelattercanberecoveredfromtheformer.22 Onethenhastospellouthowthepropositionalcontentisrecovered.Thebroad rangeofresearchoncontextandcontext-sensitivityhasfocusedonspellingthis out.Inspellingitout,contexthasbeenpresumedtoplaythecrucialrole,where effectsofcontext(Bach,1994).Othersarguethatthere’srelativelylittleinterestinpositingthenotion ofsemantic(asopposedtopragmatic)contentaltogether(Recanati,2004).Suchauthors, ipsofacto wouldn’tendorse identification.Fordiscussion,seeKingandStanley(2005).
21 Forcriticismsalongtheselines,seeEvans(1979);Lewis(1980);Kaplan(1989b);Dummett (1993);Stanley(1997a,b);Ninan(2010);Rabern(2012); interalia.SeeKing(2003);Stojnić(2017a) forfurtherdiscussion.Asimplewaytoseethepointisasfollows(cf.Lewis(1980);Kaplan(1989b)). Considerthesentence‘Johnhasfivecoinsinhispocket’.Accordingtothestandardaccount,sucha sentenceistrueinacontextc,justincaseitexpressesatruepropositioninc.Sincepropositions—the kindsofthingswebelieveandassert—aretrue/falserelativetoworlds(theyrepresenttheworldas beingacertainway),thissentenceistruein c,justincasethepropositionitexpressesin c,namely thatJohnhasfivecoinsinhispocket,istrue(ataworldinwhichitisuttered).Now,consider‘Itis possiblethatJohnhasfivecoinsinhispocket’;thetruthofthissentence(inacontext c)dependsnot onwhetherJohnhasfivecoinsinhispocketintheworldofutterance,butonwhetherthereis some (accessible)possibleworldinwhichhehasfivecoinsinhispocket(afterall,thatitispossiblethat hehasfivecoinsinhispocket,doesn’tmeanthathedoes).Onecanaccountforthisbytreating‘It ispossiblethat’asanintensionaloperator,whichtakesthecompositionalsemanticcontentofthe embeddedsentence—thatJohnhasfivecoinsinhispocket—anddeliversapropositionthatistrue inalltheworlds w suchthatinsomeworld w′ accessiblefrom w,Johnhasfivecoinsinhispocket. Inotherwords‘itispossiblethat’shiftstheworldofevaluationforthepropositionexpressedby theembeddedsentence.Butthisonlyworksifthecontentoftheembeddedsentenceissensitiveto worldsofevaluation;otherwise,thecontributionoftheoperatorwouldbevacuous.Now,suppose therewereoperatorsthatshiftnottheworldofevaluation,butsomeotherparameteronwhichthe truthofthecontentoftheembeddedsentencedepends.Forinstance,supposethat‘it’salwaysthe casethat’in‘It’salwaysthecasethatJohnhasfivecoinsinhispocket’isanoperatorthattakes thecontentoftheembeddedsentenceandcheckswhetheritistruerelativeto all times,notjust thetimeofutterance.Forthistowork,thecontentexpressedbytheembeddedsentencehasto betrue/falserelativetotimes,notjustworlds:ifthecontentwereonlysensitivetoworlds,then thecontributionoftheoperator‘always’wouldbevacuous.Butifthisisright,thenthecontentof thesentencecannotjustbeapropositiontraditionallyconstrued,sincepropositionstraditionally construeddonotvaryintruth-valuesrelativetotimes.Idonotmeantoendorsethat‘always’ isatemporaloperator;whetheritisorisn’tisamatterofcontroversy.Buttherecentliterature suggestsEnglishcontainsatleastsomeintensionaloperatorsbeyondtheworld-shiftingones,be itmodals,antecedentsofconditionals,quantifiers,temporaloperators(Yalcin,2007,2015;Ninan, 2010;Rabern,2012).Theseargumentsseemtosuggestthatpropositions—traditionallyconstrued— cannotplaytheroleofcompositionalsemanticcontent.
22 SeeLewis(1980);Kaplan(1989b).
