Presuppositional Analyses of Definiteness

Transcription

1 InstituteofCognitiveScience Albrechtstrasse Osnabrück,Germany osnabrueck.de PresuppositionalAnalysesofDefiniteness MasterThesis by MariaCieschinger osnabrueck.de Sep.11 th,2008 Supervisors:Prof.Dr.PeterBosch,Dr.CorneliaEndriss

2 TABLEOFCONTENTS 1 INTRODUCTION 4 2 PRESUPPOSITIONALANALYSESOFDEFINITENESS TOWARDSAPRESUPPOSITIONALANALYSISOFTHEDEFINITEARTICLE Russell(1905) PresuppositionalAnalyses TheContextDependenceofDefiniteDescriptions STALNAKER Stalnaker(1974) Stalnaker(2002) Summary COMMONGROUND:HOWCANITBEESTABLISHED? InductionSchemaforMutualKnowledge: Clark&Marshall(1981) VarietiesofMutualKnowledge ComparisonbetweenStalnaker(2002)and Clark&Marshall(1981) Summary UNIQUENESS:SEMANTICVS.INFORMATIONALUNIQUENESS SemanticUniqueness:Kadmon(1990) InformationalUniqueness:Roberts(2003) Semanticvs.InformationalUniqueness SUMMARY PROBLEMATICCASES Bosch(2007) Poesio's'WeakDefinites'(1994) Carlsonetal.'s'WeakDefinites'(2006) Preposition ArticleContractionsinGerman:Cieschinger(2006) CONCLUSION 58 REFERENCES 59 2

3 ABSTRACT Inthispaper,weareconcernedwiththemeaningofthedefinitearticle.Itwillbe arguedthatapresuppositionalanalysis,suchastheoneproposedbystalnaker (1974,2002),iscapableofaccountingforalargenumberofusesofdefinitedescriptions.InordertomakesomeofStalnaker'sideasmoreexplicit,wewillalso draw on insights from Clark& Marshall(1981) and Roberts(2003), which are regarded as elaborations on Stalnaker's original ideas, and not as competitors. Thetheorypresentedinthispaperanalysesdefinitedescriptionsasimplyinga uniquenesspresuppositiononpartofthespeaker,andtheuniquenessinquestionissatisfiedifadiscoursereferentcanbeidentifieduniquelyinthecommon ground, i.e., we assume a notion of informational, not semantic, uniqueness. In this sense, the notion of common ground is discourse based, not based on a model of the world. But there are certain uses of definites that seem to evade suchananalysis:asbosch's(2007)examplesseemtoindicate,ananalysislike the one proposed in this paper fails to account for certain cases of associative anaphora,andtheobservationsprovidedbypoesio(1994),carlsonetal.(2006), and Cieschinger (2006) suggest that not all felicitous uses of definite descriptionsrequiretheirreferentstobeuniqueinthecommonground.whatthislatter set of examples seems to suggest is that the definite article is ambiguous betweena'unique'readingalongthelinesofstalnaker'soriginalideasandan'arbitrary' reading which does not require the uniqueness of a referent in the commonground. 3

4 1 INTRODUCTION The meaning of the definite article, and definite reference more generally, is a classictopicinlinguisticsandinthephilosophyoflanguage.oneoftheusesof the definite article is exemplified in this passage from Terry Pratchett's novel MovingPictures(Pratchett2002:187,emphasesadded): "Cut me own Throat Dibbler, or C.M.O.T. as he liked to be called, sat up in bed and staredatthedarkness[...]hefumbledhurriedlybesidehisbedforthematches,man agedtolightthecandle,andeventuallylocatedapen." Why is the definite article used with candle but not with pen? Intuitively, the situation described seems to contain only one candle that is standing by Dibbler'sbed,whereastheremaybemorethanonepenaround.Thedefinitearticle thewouldthen,incontrasttotheindefinitearticlea,implysomekindofuniquenessofaparticularobjectinagivensituation. Inthispaper,Iwanttogiveanoverviewoverexistinganalysesofthemeaning of the definite article and over classes of definite descriptions that have been broughttoattentionintherecentliteratureandthatseemtoevadeastandard analysis.thetheoriesiwillpresentallfollowtheintuitionthatdefinitereference is related to uniqueness of some kind. More specifically, the use of definite descriptionsisanalysedasaspeaker'spresupposingtheuniqueexistenceofareferentinthecommonground.thisgeneralideagoesbacktostalnaker(1974),but wewillalsobeconcernedwithvariationsonthistheme,includingamorerecent proposal of Stalnaker's (2002), Clark & Marshall's (1981) specifications of the central notion of common ground, and Roberts' (2003) account of definite descriptions. The central points of an analysis such as Stalnaker's can be summarizedasfollows: (i) Whenusingadefinitedescription,speakerspresuppose,ratherthanas sert,certainthings. (ii) Whatispresupposedistheuniquenessofadiscoursereferentthatthe definitedescriptionisusedtoreferto. (iii) Thisdiscoursereferentisuniqueinthecommongroundofthepartici pantsofaconversation,notuniqueintheworldorinamodel,andthe commongroundis,roughly,thesetofpropositionsthataresharedbya speakerandhisaudienceandthatarerecognizedtobesoshared. A large number of uses of definite descriptions can be accounted for in such a framework. But, as we will see, there are also cases which cannot easily be explained. Section2introducesStalnaker likeanalysesofdefinitedescriptions,andaims atgettingaclearunderstandingoftheimportantnotionsofcommongroundand uniqueness.wewillbegin,insection2.1,bysketchingarussellianapproachto definitenessandshowthatapresuppositionalanalysis,togetherwithsomekind of domain restriction, is preferable. Section 2.2 will be concerned with Stalnaker's proposals(1974and 2002),introducingthe notion ofcommonground. InSection2.3,wewilltakeacloserlookatthisnotionbyconsideringdifferent 4

5 ways,aspresentedbyclark&marshall(1981),inwhichthecommongroundcan bethoughtofasbeingcommonamongspeechactparticipants.wewillthenproceedtoexaminingthenotionofuniquenessmorecloselyinsection2.4,contrastingtheviewsofroberts(2003)andkadmon(1990):kadmonproposesavariant ofrussellianuniqueness,i.e.,theuniqueexistenceofanobjectintheworldorina modelispresupposed,whereasroberts,moreinkeepingwithstalnaker'soriginalideas,arguesthattheuniquenessofdiscoursereferentsispresupposed.uses ofdefinitedescriptionsthatdonotseemtofittheunifiedaccountpresentedin Section2willthenbediscussedinSection3.Thisdiscussionstartswithexamples provided by Bosch (2007) in Section 3.1, showing that uniqueness in the commongroundaloneisnotenoughtoexplaincertaincasesofassociativeanaphora.thesubsequentsectionswillbeconcernedwithdifferentsetsofexamples thatdonotseemtorequiretheuniqueidentifiabilityofareferentatall:poesio's (1994)'weakdefinites'(Section3.2),Carlsonetal.'s(2006)'weakdefinites'(Section3.3),andpreposition articlecontractionsingerman(cieschinger2006,section3.4).section4concludesthispaper. 5

6 2 PRESUPPOSITIONALANALYSESOFDEFINITENESS Definitedescriptionsareoftenanalysedasimplyingtheuniqueexistenceofareferentinthecommongroundoftheinterlocutors.Inthissection,wewillfollow thisintuitionandfocusonstalnaker's(1974,2002)accountofdefinitedescriptions:hesuggeststhatthespeakerpresupposestheuniquenessofareferentin thecommongroundwhenutteringadefinitedescription.asshownbyclark& Marshall(1981),thecommongroundcanbeinferredfromvariouskindsofevidence, including community membership and physical, linguistic, and indirect copresence of the speaker, his audience and some object. Furthermore, we assumethattheuniquenessinvolvedintheinterpretationandutteranceofdefinite descriptionsisinformationaluniquenessinthesenseofroberts(2003). This section is structured as follows: We begin, in Section 2.1, by outlining Russell's(1905)andHeim&Kratzer's(1997)approachtodefinitedescriptions, andconcludethatapresuppositionalanalysisofdefinitesisdesirable,together withsomekindofrestrictionofthedomainofevaluation.insection2.2,theproposalputforwardbystalnakerwillbepresented:wewillconsiderbothanearly versionofthisproposal(1974,section2.2.1)andamorerecentone(2002,section2.2.2),andwilltrytoshowthatthemorerecentproposalhasconsiderable advantagesovertheearlierversioninsection2.2.3.togetabetterunderstanding of the important notion of common ground, we will present Clark & Marshall's (1981) proposal, focussing on different ways in which the common groundcanbeestablished.insection2.4,wewillturntothekindofuniqueness that is presupposed by a speaker when uttering a definite description: We will contrastthenotionofsemanticuniquenessasproposedbykadmon(1990,section 2.4.1) with Roberts' (2003) notion of informational uniqueness (Section 2.4.2),andconcludethatitisthelatterthatseemstocorrespondtoStalnaker's originalideas.abriefsummarywillbegiveninsection2.5beforeweproceed,in Section3,tovariousexamplesofdefinitedescriptionsthatappeartoevadethe analysispresentedinthissection. 6

7 2.1 TOWARDSAPRESUPPOSITIONALANALYSISOFTHEDEFINITEARTICLE Asafirststeptowardsananalysisofthedefinitearticle,considerthefollowing passagefromdouglasadam'shitchhiker'sguidetothegalaxy: (ZaphodBeeblebrox,thetwo headedandveryfamouspresidentofthegalaxy,enters thehitchhiker sguide'sheadquarters.hetellsthereceptionisthisname,butthere ceptionistdoesnotbelievehimuntilzaphodtellshimtolookathistwoheads.) Receptionist:YouareZaphodBeeblebrox?[...]TheZaphodBeeblebrox? Zaphod:No,justaZaphodBeeblebrox.Didn'tyouhearIcomeinsix packs?[...]now, wheredoifindzarniwoop? (FittheSeventh,broadcastonBBCRadio4on24Dec1978) ThereceptionistusesthedefinitearticletheandinthegivencontextitisabsolutelyclearthatthereisoneandonlyoneZaphodBeeblebrox,whereasZaphod's use of the indefinite article a seems to suggest that there is more than one Zaphod Beeblebrox. In the following, we will not be concerned with proper namesthatareprecededbythedefinitearticle,buttheexamplenonethelessillustrates an important intuition about the meaning of the definite article: It seemsthatthedefinitearticlethe,incontrasttotheindefinitearticlea,implies somekindofuniquenessofaparticularobjectinagivensituation. Thisintuitionis,ofcourse,notnew:SinceRussell'sOnDenoting(1905),ithas been widely discussed in the linguistic and philosophic literature. In Section 2.1.1, we will present Russell's own proposal and some of the problems that arise within this framework. We will then proceed to sketching a presuppositionalanalysisofthedefinitearticleinsection2.1.2.inthefinalsection,wewill brieflyconsideropenproblemsoftheapproachsketchedin Russell(1905) According to Russell(1905/2001), a sentence containing a definite description oftheformthex(isf)assertsboththeexistenceandtheuniquenessoftheobject describedbyx,andpredicatessomethingaboutthisobject,viz.thatitisf. 1 So, for a sentence like(1), we would get a logical representation along the lines of (2): (1) ThepresentQueenofEnglandisbald. (2) x(qoe(x) ( y(qoe(y) x=y) Bald(x)) The formula in(2) can be paraphrased as'there is a queen of England and not morethanone,andsheisbald'.thesentencein(1)wouldthenbetrueifthereis auniquequeenofenglandandifshebald,andfalseotherwise.inotherwords,if 1Notethat,accordingtoRussell,itissentences,phrases,orlinguisticexpressionsthatdenotean object(cf.,e.g.,1905:212,214). 7

8 oneoftheconjunctsofthecomplexexistentialin(2)werefalse,thewholeexistential would be false, and hence also the sentence (1) corresponding to this existential.itisintuitivelyclearthatasentenceshouldbefalseifthelastofthe threeconjunctsinourrepresentation,thepredicationofacertainpropertyofan object x, is false, but, as Strawson (1950/2001) pointed out in his response to Russell'sproposal,theintuitionsaboutthetruthorfalsityofasentenceareless clearincaseswhereoneorbothofthefirsttwoconjunctsarefalse(cf.section 2.1.2),i.e.,ifthereisnoobjectthatfitsthedescriptivecontentofthedefinitedescriptionorifthereismorethanoneobjectthatfitsthedescription.Considerthe sentencesin(3)and(4): (3) ThepresentKingofFranceisbald. (4) ThesonofPrinceCharlesis11feettall. Francedoesnothaveacurrentking,sothereisnouniqueobjectthatthedefinite description the present King of France in (3) could apply to. Similarly in (4), PrinceCharleshastwosons,HarryandWilliam,sothesonofPrinceCharlesdoes notdenoteauniqueobjecteither.accordingtoarusselliananalysis,thesentenceswouldsimplybefalsesincetheassertionthatthereisoneandonlyoneobject fitting the descriptive content would be false (1905:214). If this analysis werecorrect,then,assumingatwo valuedlogic,thenegationofthesentencesin (3)and(4)shouldbetrue(cf.(5)and(6),respectively). (5) ThepresentKingofFranceisnotbald. (6) ThesonofPrinceCharlesisnot11feettall. But surely these two sentences would not be judged true either. Russell's way out of this dilemma is to remind us of the underlying logical structure.(7) and (8)areroughrepresentationsofthesentencesin(5)and(6),respectively. (7) a. x(kof(x) ( y(kof(y) x=y) Bald(x)) b. x(kof(x) ( y(kof(y) x=y) Bald(x)) (8) a. x(sopc(x) ( y(sopc(y) x=y) Tall(x)) b. x(sopc(x) ( y(sopc(y) x=y) Tall(x)) NotethatRussellassumesthattherearetwodifferentwaysinwhichtheabove sentencescanbetransformedintologicalformulae(1905:218):thescopeofthe negationcancontainasinglepredicateonly(asin(7a)and(8a)),oritcantake scopeoverthewholeexistential(asin(7b)and(8b)).thea versionwouldthen stillcomeoutfalse,accordingtorussell,becausetheassertionsofexistenceand ofuniquenessarefalse,whereastheb versionswouldbetrue.are translationof (7b) into natural language, for example, would be something like 'it is not the casethatthereisauniquepresentkingoffranceandthatheisbald',and,intuitivelythissentenceistrue. 2 Wesaidabovethattheintuitionsaboutthetruthorfalsityofsentencesthat containdefinitedescriptionsthatdonotdenoteauniqueobjectarenotasclear 2Russellclaimsthat"'thekingofFranceisnotbald'isfalseiftheoccurrenceof'theKingof France'isprimary,andtrueifitissecondary"(1905:218).Wewill,however,notbeconcerned withthenotionsofprimaryandsecondaryoccurrenceinthefollowing. 8

9 astheyshouldbeunderarusselliananalysis. 3 Itseemsthat,whenpressedforan answertothequestion'isasentencelike(3)trueorfalse?',manywouldsaythat thesentenceisneithertruenorfalse.thisclearlyconflictswithrussell'saccount andinthenextsectionwewillhavealittlemoretosayaboutthisintuition PresuppositionalAnalyses InresponsetoRussell(1905),Strawson(1950/2001)arguesthat,firstofall,itis not sentences that denote or refer to something, but rather that utterances of sentences are used to denote or refer to something, and that, secondly, utterancesofsentenceslike(3) (6)areregardedasbeingneithertruenorfalse.AccordingtoStrawson,"'[mentioning]',or'referring',isnotsomethinganexpressiondoes;itissomethingthatsomeonecanuseanexpressiontodo"(1950:232). Heemphasizesthatoneandthesamesentenceorexpressioncanbeutteredat different occasions and that the truth or falsity of such an utterance can be assessed,butnotthegeneraltruthorfalsityofthesentenceorexpressionitself. 4 To illustrate this, consider someone uttering(3), repeated from above, in 1789 duringthereignoflouisxvi. (3) ThepresentkingofFranceisbald. There existed a unique king of France in 1789, so the definite description the present king of France could then be used to refer to something, viz. Louis XVI, whereassuchanutterancecouldnot(properly)beusedtorefertosomethingin, say,1905.strawsonfurtherarguesthatutterancesofsentencescontainingdefinite descriptions that cannot be used to refer to a unique individual, as in (3) when uttered in 1905, are not false or true as Russell's account would predict: Whenaskedwhetheryouwouldsaythatthisutteranceistrueorfalse,Strawson thinksthat"youwouldbeinclined,withsomehesitation,tosaythatyoudidnot doeither;thatthequestionofwhether[this]statementwastrueorfalsesimply did not arise, because there was no such person as the king of France" (1950:234).Giventhisintuition,itseemsthattheuniqueexistenceofthereferent of a definite description is not asserted, but that it is something like a preconditionforthetruthorfalsityofanutterance.strawsonhimselfdoesnotuse the term presupposition, but it has been used in later literature to refer to this phenomenon:theuniqueexistenceofthereferentisthusnotasserted,butpresupposed.forasentencecontainingadefinitedescriptionlikethexthiswould mean that this sentence would lack a truth value unless the presupposition P 'thereisoneandonlyonex'istrue. ThisisalsothesolutionthatHeim&Kratzer(1997)seemtoadopt.Theyproposetheanalysisin(9)forthedefinitearticle(1997:75): 3Note,however,thatRussellwasnot,infact,concernedwithprovidinganaccurateanalysisof definitedescriptionsastheyareusedinnaturallanguage(cf.alsorussell(1957/1999).nonetheless,hisideashavebeen,andstillare,veryinfluentialinthelinguisticandphilosophicliterature. 4Exceptmaybeforanalyticorgenericstatements,butStrawsonisnotconcernedwiththese. 9

10 (9) [[the]]=λf:f D<e,t>andthereisexactlyonexsuchthatf(x)=1. theuniqueysuchthatf(y)=1. Here,thedefinitearticlethedenotesapartialfunctionthatyieldsareferentonly in those cases where there is a unique object that satisfies the description f(x). The predictions of such an approach seem to correspond well to speakers' judgementsofsentenceslike(3) (6).Itshouldbenoted,however,thatthisanalysis considers expressions or sentences as denoting something (cf., e.g., 1997:74), which is contrary to what Strawson and others(cf. Sections 2.2, 2.3, and2.4)haveclaimed TheContextDependenceofDefiniteDescriptions Presuppositional analyses like Strawson's (1950) or Heim & Kratzer's (1997) seemtoaccountwellforspeakers'intuitionsabouttheappropriateuseofdefinite descriptions. It is obvious, however, that a notion like Heim& Kratzer's is way too broad. A sentence like (10) might well be uttered felicitously, even thoughthereissurelymorethanonetableintheworld. (10) WeboughtthetableatIKEA. Whetherornotanutteranceofthedefinitedescriptionthetableisregardedas felicitous usually depends on the context of the utterance. If, for example, Felix andannahavebeentalkingaboutaparticulartableforsometime,oriftheyare inaroomwithasingletableinit,anna'sutteranceof(10)willbeappropriate. Consider, on the other hand, the following situation: Anna approaches a completestrangerinthestreet,felix,andutters(10).inthiscontext,felixwillmost probablyconsideranna'sutteranceasinappropriate.whatappearstobeneeded in order to account for the variations of acceptability in different contexts is somekindofdomainrestriction.thismeansthatweneedtorestrictthedomain inwhichagivendefinitedescriptionistobeevaluated:doesitcontainonlythe objectsinaparticularroom,onlytheobjectsthatweretalkedaboutearlier,only objectsinosnabrück,ineurope,orinfranceinthe18thcentury? Onewaytodealwiththisistoclaimthatadefinitedescriptionlikethetableis notusedtorefertosomeuniquetableintheworld,butratherthatitis,oneach occasion of its utterance, used to refer to some unique table that is present on this occasion. Heim& Kratzer's modification of the representation for the definitearticleseemstosuggestthiskindofreasoning(1997:81): (11) [[the]]=λf:f D<e,t>andthereisexactlyonex Csuchthatf(x)=1. theuniquey Csuchthatf(y)=1,whereCisacontextuallysalient subsetofd. Thedefinitearticleherestilldenotesapartialfunction,butwearenotrequiring thereferentofthedescriptionf(x)tobeuniqueinthewholeworld(d<e,t>),but onlytobeuniqueinsomesubsetofdthatissomehowprominentinthecurrent context.asimilarideaisexpressedinlewis(1979:348,emphasisadded):"'the F'denotesxifandonlyifxisthemostsalientFinthedomainofdiscourse,ac 10

11 cordingtosomecontextuallydeterminedsalienceranking". 5 Itisclearthatthe context, or utterance occasion, is crucial for the interpretation of definite descriptions, whether or not we accept a Lewis like salience theory. But what kind(s) of information does the context provide and how could it be represented? According to von Heusinger (2000:2), the context of an utterance provides threekindsofinformation 6 :Linguisticinformation(lexicalcontent),information that was given linguistically in the preceding discourse (e.g., objects or events that were talked about earlier), and extra linguistic information (background andsituationalknowledgeofthespeakerandhisaudience).thelexicalcontent ofsingleexpressionsismostprobablystoredinsomekindofmentallexicon,and thelexicalcontentofphrasesorsentencesiscomputedfromthecontentofsingle expressions with the help of general interpretation rules. Previously mentionedinformationcanbecapturednicelyinframeworkslikethoseprovidedby Discourse Representation Theory (DRT) or File Change Semantics (FCS). And, e.g., Stalnaker's (1974, 2002) notion of common ground or Clark & Marshall's (1981) mutual knowledge can be regarded as conceptualizations of extralinguistic information that is shared by discourse participants (but both Stalnaker'sandClark&Marshall'sproposalscanalsoaccountforinformationgiven explicitly in the preceding discourse). We will take a closer look at Stalnaker's notions of common ground and of presuppositions in Section 2.2, and will discussclark&marshall'sproposalinsection2.3. 5See,e.g.,vonHeusinger(1997,2000)foramoredetailedsaliencetheory. 6WeareadoptingvonHeusinger'sterminologyhere. 11

12 2.2 STALNAKER Attheendoftheprevioussection,wesawthatweneedsomewayorotherof representingthecontextofalinguisticutteranceinordertorestrictthedomain ofevaluationforadefinitedescription.morespecifically,wewouldwanttoexplainhowinformationthatwasgivenintheprecedingdiscourseandhowextralinguisticinformation,i.e.,backgroundandsituationalknowledge,influencesthe interpretation of a particular expression, or makes it even possible in the first place.onesuchaccountisprovidedbyrobertstalnaker.heintroducedthenotion of'common ground' in the early 1970s, a notion that is still prominent in currentresearchondefiniteness. WewillbeginthissectionbyoutlininganearlyversionofStalnaker'sproposal (Stalnaker 1974/1999), focussing on his characterizations of common ground andpresuppositions.insection2.2.2,wewillthendiscussamorerecentversion (Stalnaker2002),andhighlightsomeofthewaysinwhichthisversionismore fine grained and detailed than the 1974 proposal. In particular, the notion of common ground is more formally based on the notion of common belief in the laterversion,andstalnakeralsogivesanaccountofaccommodationthatcovers amuchbroaderrangeoflinguisticphenomenathantheearlyversion Stalnaker(1974) In Pragmatic Presupposition (1974), Stalnaker gives, as the title suggests, a pragmatic account of the phenomenon of presupposition. Roughly, presuppositions,instalnaker'ssense,arebeliefsorassumptionsofthespeakerthatarepart ofthecommonground.butletusbeginbylookingmorecloselyatthenotionof commonground. What Stalnaker calls the common ground, then, can be summarized as "the background beliefs or assumptions which are shared by the speaker and his audience,anwhicharerecognizedbythemtobesoshared"(1974:48).theintuition behind this idea is that, in general, communication always takes place againstacertainbackgroundofbeliefsorfactsthatthecommunicatorstakefor granted.consider,forexample,aconversationbetweentwostudentsaboutthe progresstheyare(orarenot)makingontheirtheses.usually,bothstudentswill have certain beliefs about their surroundings,about the general structures and proceduresoftheiruniversity,abouttherequirementsinformandcontentona thesis,abouttheirindividualabilitiesandworkingpreferences,abouthowmuch timetheyhaveleft,etc.,andmanyofthosebeliefsmaycoincide.iftheydo,i.e.,if both students believe or assume that they share certain beliefs or assumptions andiftheybothrecognizethatthesebeliefsorassumptionsaresharedbythem, thenthesebeliefsorassumptionsarepartofthecommonground.thecommon ground thus contains all those propositions that a speaker can reasonably take forgrantedwhenmakinganutterance.asageneralrule,informationthatisalreadypartofthecommongroundneednotbeassertedexplicitly,therebymak 12

13 ing conversation the more efficient the larger the common ground is. Those propositionsthatareassertedinthecourseofaconversationarethenaddedto the common ground and can henceforth be taken for granted. The addition of propositions to the common ground is only possible, however, if they are accepted(i.e.,iftheirtruthisnotchallenged)bytheparticipants(1974:49). 7 The common ground does then not only contain extra linguistic information, like backgroundandsituationalknowledge,butalsoallofthepropositionsthatwere asserted earlier, i.e., information that was given linguistically in the preceding discourse.buthowdoesallthisrelatetothepresuppositionproblemalludedto above(cf.section2.1)? "ApropositionPisapragmaticpresuppositionofaspeakerinagivencontextjustin casethespeakerassumesorbelievesthatp,assumesorbelievesthathisaddressee assumes or believes that P, and assumes or believes that his addressee recognizes thatheismakingtheseassumptions,orhasthesebeliefs." According to Stalnaker, the relation between the common ground and presuppositionscanbecharacterizedasfollows(1974:49): Ifagivenpropositionispartofthecommonground,itcanfunctionasapresupposition.So,ifaspeakerpresupposessomething,hetakesitforgranted,heassumesthatitisalreadypartofthecommonground.Ifitisnot,then,usually,the utterance will be regarded as being inappropriate. Stalnaker's account of presuppositionscanbeappliedtoawiderangeoflinguisticexpressions(e.g.thedifference in presuppositional behaviour of factive and semi factive verbs like regret,forget,knowandrealize,discover,believerespectively(cf.1974:56ff),orthe behaviouroftheadverbtoo(2002:719f.)).butsinceweareconcernedwithdefinitedescriptionsinthispaper,letustakeacloserlookathowthisaccountcan beappliedtoexpressionslikethetablein(10),repeatedherefromabove: (10) WeboughtthetableatIKEA. Aspeakerwillutterasentencelike(10)onlyifthepropositionP'inthecurrent situation, there is a unique table' or something similar is already part of the common ground(cf. 1974:52, Footnote 2). So, if a speaker can reasonably take suchapropositionforgranted,hecanusethedefinitedescriptionthetableeven though both he and his audience presumably know that there are many more tablesintheworldthanjust,say,thesingletableinthespeaker'skitchenorthe particulartablethatthespeakerandhisaudiencehavebeentalkingaboutfora while. Consider, on the other hand, a situation in which someone, Felix, utters (10) to a complete stranger in the street, Anna. Anna would most probably regardfelix'utteranceasinappropriate,andrightlyso.eventhoughthepropositionpispartoffelix'assumptionsetorbeliefsystem,hecannotreasonablyexpectanna'smentalstatetobesimilarenoughtotakepforgranted,i.e.,hecannot reasonablyexpectptobepartofthecommonground. 7Stalnakerdoesnotelaborateonwhatexactlyitmeansforapropositiontobeacceptedbythe participantsinstalnaker(1974).allhesaysisthat,asaspeaker,"iwillnotassertthingsincompatiblewiththecommonbackground,sincethatwouldbeself defeating"(1974:49).instalnaker (2002),however,amoredetailedcharacterizationof'acceptance'isgiven.Wewillreturntothis issuebelow. 13

14 By requiring the uniqueness presupposition of a definite description to be part of the common ground, we can bring about the desired restriction of the domainofevaluation.thecommongroundisadjustedtotherelevantconversationbytheadditionofpropositionsassertedinthediscourseandofbeliefsand assumptionsaboutthegeneralsituationofutterance.so,if,inagivendiscourse, aparticulartableisprominentorhasbeenmentionedpreviously,thenadefinite descriptionlikethetableisevaluatedagainstthethenrelevantcommonground, notagainstthedomainofallexisting(andpossible)objectsintheworld.heim& Kratzer'sstipulationthatsomethinglikethetabledoesnotdenotesomeunique table in the world, but ratherthat it denotes, on each occasionofits utterance, someuniquetablethatispresentonthisoccasion(cf.section2.2),seemstobe metbystalnaker'sproposal. Whatisstillmissing,however,isanaccountofsalienceeffects.Asanexample forthis,considerthefollowingpassagefromlewis(1979:348,emphasesadded): "ImagineyourselfwithmeasIwritethesewords.Intheroomisacat,Bruce,whohas been making himself very salient by dashing madly about. He is the only cat in the room,orinsight,orinearshot.istarttospeaktoyou: Thecatisinthecarton.Thecatwillnevermeetourothercat,becauseourother catlivesinnewzealand.ournewzealandcatliveswiththecresswells.andthere he'llstay,becausemiriamwouldbesadifthecatwentaway." Tosimplifymatters,letusconcentrateonthedefinitedescriptionthecatonly.It seemsobviousthatthefirsttwooccurrencesofthecatareusedtorefertobruce, thecatthatisinthesameroomasdavidlewis.thethirduseofthecat,however, denotesthecatthatlivesinnewzealand,notbruce.itisnotclearhowthisphenomenoncouldbeexplainedinstalnaker'sframework:twocatshavebeenmentionedinthediscourseprecedingthelastoccurrenceofthecat,sotwopropositionsalongthelinesofp'thereisauniquecatnamedbruce'andp''thereisa uniquecatthatlivesinnewzealandandthatisnotcalledbruce'shouldbepart ofthecommongroundbythen.whyexactlythenewzealandcatispickedoutas the right referent cannot be accounted for by Stalnaker's proposal. In his later paper(2002),stalnakeradmitsthatsomekindofsaliencetheoryisneeded:"a certainwomanmustbesalientfor[the]useof'she'tobeappropriateandsuccessful(where salience is presumably to be explained in terms of common belief)"(2002:711).hearguesthatpointinggesturescanmakeanindividualsalient,butthisissurelynotsufficienttoexplainlewis'example.here,thereisno overt pointing gesture involved and the underlying mechanisms that could explain the use of the definite the cat in this example appear to be more complicatedthanisexplicitlyacknowledgedbystalnaker. 8 Twopointsshouldbenotedatthispoint:First,onStalnaker'saccountitisnot sentences, propositions, or utterances that have or make presuppositions, it is persons.presuppositionsarepartofthecommongroundaspeakershareswith his audience, so they are part of the speaker's background beliefs, his assumptions(1974:48,50).and,secondly,stalnakergivespresuppositionsapragmatic analysis,notasemanticone.astatement'shavingatruthvaluedoesnotdepend onthetruthorfalsityofthepresuppositionsinvolved(as,e.g.,heim&kratzer's 8Cf.,e.g.,vonHeusinger(1997,2000)forasalience basedapproachtodefiniteness. 14

15 analysis in (11) suggests). In contexts in which a given proposition cannot be takenforgranted,isnotpartofthecommonground,anassertionalludingtothat proposition will be regarded as inappropriate, not as lacking a truth value (1974:48, 53, 61). Keeping these considerations in mind, it is clear why Stalnaker, in the passage cited above, speaks of "pragmatic presuppositions of a speaker". BeforeweproceedtoadiscussionofStalnaker(2002),letusbrieflyconsiderin more detail how propositions can be added to the common ground. Stalnaker argues that assertions should be distinguished from propositions. He gives the followingexampleasacaseinpoint:ifaspeakeruttersthesentencethequeen ofenglandisbald,thenhepresupposesthatenglandhasauniquequeen,andhe asserts that she is bald(1974:47). As we saw above, it seems obvious for Stalnaker that assertions can modify the common ground: "assertions [...] always have consequences which are part of the common background" (1974:49). An assertionisusuallymadeinordertoconveysomeinformationthatthespeaker believeshisaddresseenottohaveatthepointofutterance(1974:51).stalnaker claims that "[someone] will not say things that are already taken for granted, sincethatwouldberedundant"(1974:49),sopropositionsthatarealreadypart of the common ground need (or should) not be asserted explicitly. But what aboutpropositionsthatarenotassertedbutonlypresupposedbyaspeaker,and thatarenotalreadypartofthecommonground? It seems that in "normal, straightforward serious conversational contexts" (1974:51),ifanutteranceindicatesthatthespeakerhasapresuppositionthatis notalreadypartofthecommonground,thenthisutterancewillberegardedas beinginappropriate.thisistheviewthatwepresentedabove.thereare,however,situationsinwhichthisgeneralrulecanbeoverridden(1974:51): "[A]speakermayactasifcertainpropositionsarepartofthecommonbackground whenheknowsthattheyarenot.hemaywanttocommunicateaproposition indi rectly, and do this by presupposing it in such a way that the auditor will be able to infer that it is presupposed. In such a case, a speaker tell his auditor something in partbypretendingthathisauditoralreadyknowsit." Theaudience'sabilitytoinferthatsomethingisbeingpresupposeddependson knowledgeabouthowcertainlinguisticexpressionsbehaveinnormalconversation(e.g.thatdefinitedescriptionsusuallyrequireauniquenesspresupposition) andontherecognitionthatthespeakerismakinguseofthese"rulesgoverning normal conversation in order to communicate something which is not exactly said"(1974:52,footnote1).anexamplethatillustratesthispointcanbefound in Stalnaker (2002:710), where a daughter informs her father about her engagementbyuttering(12) 9 : (12) ODad,IforgottotellyouthatmyfiancéandIaremovingtoSeattle. Stalnaker acknowledges that this kind of exploitation of the 'normal' use of an expressioncannotbeaccountedforinhisframeworkandsuggeststhathischaracterization of presuppositions should therefore be regarded as an approxima 9AccordingtoStalnaker,thisexampleoriginallycomesfromKaivonFintel. 15

16 tiononly(1974:52,cf.also49).inafootnote,stalnakerbrieflydiscussesanother example that evades his analysis(1974:52, Footnote 2). Suppose Felix is asked bysomeonehejustmetwhetherhewasgoingtolunch,andfelixreplies: (13) No,I'vegottopickupmysister. In using the definite my sister, Felix seems to presuppose that he has a sister, eventhoughhecannotreasonablyexpectapropositionlike'felixhasasister'to bepartofthecommonground.nonetheless,felix'utteranceisappropriate.we willdiscussthisexampleinmoredetailinthenextsection.here,letusjusttakea quicklookatthetwopossiblesolutionsthatstalnakersuggests.first,wecould simply deny that the use of definite descriptions always requires a uniqueness presuppositionofthespeaker,butthisexplanationdoesnotseemtobeonethat Stalnakerwouldwanttosupport.Thesecondoptionwouldbetoexplaintheappropriatenessofsentenceslike(13)intermsofexploitation.Incaseslike(13), however, as opposed to something like (12), it seems that the indirectly communicatedproposition('felixhasasister')isnotreallyinthefocusoffelix'utterance,butratherthatitisjustsomeminorbackgroundinformation.stalnaker suggests, however, that "the explanation of how communication takes place in thisexample[(13)]maybethoughttobesimilarinformtoexplanationsofhow it takes place in the more familiar cases [(12)]: the addressee infers that the speaker accepts that Q from the fact that he says that P because normally one says that P only when it is common background knowledge that Q" (1974:52, Footnote 2). Two things are unclear at this point: First, what does Stalnaker meanby'morefamiliar'?intuitively,itseemsthatcaseslike(13)aremuchmore frequentthancaseslike(12).and,secondly,howexactlywouldthisexplanation fitintostalnaker'sframework?exploitationlikethatin(12)cannotbeaccounted for in Stalnaker's account, so what would we gain from pushing examples like (13)intothesamecategory? Stalnaker(1974)givesapragmaticaccountofpresuppositionsthatseemstobe adequateforthemajorityofusesofdefinitedescriptions.inparticular,thequestionofhowtorestrictthedomainofevaluationforadefinitedescriptioncanbe answeredwiththehelpofhisnotionofcommonground.butwedonotseemto findananswertothequestionofwhether,andunderwhatconditions,propositionsthatarenotassertedbutonlypresupposedcaneverbeaddedtothecommon ground. In cases like (12) and (13), it would be a desirable result if the speaker'spresuppositionswouldendupaspartofthecommonground,andnot justthattheaudiencecaninferthatthespeakerhasthesebackgroundbeliefs(cf. also1974:54):definitedescriptionslikemydaughter'sfiancéorfelix'sister(and pronouns) can be used felicitously after the utterance of (12) or (13), respectively,sotherelevantpropositionsshouldbepartofthecommongroundifwe wanttomaintainthatdefinitedescriptionsrequireauniquenesspresupposition onpartofthespeaker.instalnaker(2002),thenotionofcommongroundisfurtherrefined,andalsotheproblemsaddressedattheendofthissection,i.e.,the addition of (only presupposed) propositions to the common ground, are dealt withinmoredetail. 16

17 2.2.2Stalnaker(2002) Aswesaidearlier,inCommonGround(2002),Stalnakeraimsatdevelopingfurthersomeoftheideasalreadypresentinhis1974account.Morespecifically,the notionofcommongroundisbasedonaformalnotionofcommonbelief,andthe addition of propositions to the common ground is considered in more detail. Stalnakeremphasisesagainthathisaccountisapragmatic,notasemanticone, andthatitisspeakersthathaveormakepresuppositions,notsentences,utterances,orpropositions:"speakerpresuppositionisapropositionalattitudeofthe speaker[...]whatismostdistinctiveaboutthispropositionalattitudeisthatitis asocialorpublicattitude:onepresupposesthatφonlyifonepresupposesthat otherspresupposeitaswell"(2002:701). 10 Thegeneralnotionsofpresuppositionandcommongroundinthe2002paperarealsorathersimilartotheones expressedin1974:"topresupposesomethingistotakeitforgranted,oratleast to act as if one takes it for granted, as background information as common ground among the participants of the conversation."(2002:701). We will begin thissectionbytakingacloserlookattherefinednotionsofcommonground,definedintermsofcommonbelief,andofpresupposition,andwillthenproceedto Stalnaker'sviewofpresuppositionaccommodation,i.e.,oftheadditionofpropositionsthatarenotassertedbutonlypresupposedtothecommonground. As a first step towards a definition of common ground, consider the following passage(2002:704): "Inthesimplepicture,thecommongroundisjustcommonormutualbelief,andwhat a speaker presupposes is what she believes to be common or mutual belief [...] a propositionφiscommonbeliefofagroupofbelieversifandonlyifallinthegroup believethatφ,allbelievethatallbelieveit,allbelievethatallbelievethatallbelieve it,etc." Twoimportantdifferencesbetweenthe2002andthe1974accountarealready apparent here: First, when a speaker presupposes something, he does not believe that the relevant proposition is part of the common ground, but that it is common belief that this proposition holds. So, while in the earlier account, speakershadbeliefsaboutthecommonground,inthemorerecentaccountthey havebeliefsaboutthecommonbelief.and,secondly,theconditionforapropositiontobepartofthecommonbeliefismorecomplexthantheonegivenin1974: "ApropositionPis[apresupposition,i.e.,ispartofthecommonground]justin case the speaker assumes or believes that P, assumes or believes that his addresseeassumesorbelievesthatp,andassumesorbelievesthathisaddressee recognizesthatheismakingtheseassumptions"(1974:49).the'etc.'inthepassagecitedaboveiscrucial,i.e.,whereasintheearlyversiononlythreeconditions are given, the conditions in the later version can potentially continue ad infi 10Actually,whendiscussingexploitationinStalnaker(1974),Stalnakerclaimsthat"[presupposing]is[...]notamentalattitudelikebelieving,butisratheralinguisticdisposition"(1974:52). Thisconflictswithwhathearguesforinthe2002paper,butsincetheremarkjustcitedappears tobesomewhatadhocandisnotexplainedfurther,wehaveignoredthistensionsofarandwill alsoignoreitinthefollowing. 17

18 nitum.wewillnothavemuchtosayaboutthislatterpointinthissection(butcf. Section2.3.1forashortdiscussion),butwillratherfocusontheformer. Stalnakeruses"ahighlyidealizedformalsemanticframework"(2002:706)of possibleworldstodefinethelogicofcommonbelief. 11 Inthisframework,thebeliefs of a discourse participant i are represented by an accessibility relation Ri: WecansaythatibelievesthepropositionPinaworldwonlyifforeveryworldw' thatisaccessiblefromworldw(wriw'),pholdsinw'.furthermore,weassume thateachindividual'sbeliefsaretransparenttotheindividual,i.e.,thatanindividual always has introspective access to his beliefs. There are two kinds of introspection:positiveandnegativeintrospection.positiveintrospectionmeans thatifanindividualbelievesthatp,healsobelievesthathebelievesthatp.negativeintrospection,ontheotherhand,meansthatifanindividualdoesnotbelieve thatp,thenhealsobelievesthathedoesnotbelievethatp.afurtherassumption Stalnaker makes is that individuals have consistent beliefs, i.e., that the set of possible worlds that are accessible from world w is always non empty. To accommodate all these assumptions, the accessibility relation Ri has to be transitive,euclidean,andserial(2002:706): Transitivity: w w' w''(wriw' w'riw'' wriw'') Euclideanity: w w' w''(wriw' wriw'' w'riw'') Seriality: w w'wriw' The transitivity of Ri implies positive introspection, the euclideanity implies negative introspection, and the seriality implies the consistency of an individual'sbeliefs. 12 Stalnakerthendefinesthelogicofcommonbeliefforagroupof individuals with the help of the individuals' beliefs. He claims that the accessibilityrelationforagroupisthetransitiveclosureoftheaccessibilityrelationsfor the individuals. Transitive closure is identified with reachability in Schlenker's representation:"letussaythatw'isreachablefromwjustincasesomepathπ leads from w to w' through a series of steps of the form <wi, wi+1> for which eitherwir1wi+1orwir2wi+1."(2006:4)(here,1and2representdifferentindividuals.)wecanthensaythatapropositionpiscommonbeliefataworldwonlyifp holdsineveryworldthatisreachablefromw.thisisequivalenttothemoreinformal characterization of common belief given above. It should be noted that, even though common belief is defined with the help of individual beliefs, the logicofcommonbeliefisslightlydifferentfromthelogicofindividualbelief.in particular, negative introspection is no longer implied, whereas positive introspectionispreserved.thismeansthat"[if]itiscommonbeliefthatφ,itwillbe commonbeliefthatitiscommonbeliefthatφ,butthefactthatitisnotcommon beliefthatφdoesnotimplythatitiscommonbeliefthatitisnotcommonbelief thatφ"(2002:707).aspeaker'spresupposition,onthispicture,isthenabeliefof thespeakeraboutthecommonbelief.moreprecisely,thesetofpresuppositions of a speaker can be identified with the set of propositions that the speaker be 11Stalnakerhimselfdoesnotusealotoflogicalformulaetoillustratehisclaims,butsinceformulaearesometimesusefultomakethingsmoreexplicit,wewill,inthefollowing,useaformalrepresentationofStalnaker'saccountthatispresentedinSchlenker(2006). 12Forproofsoftheseandthefollowingimplicationsoftheformulae,seeSchlenker(2006:4). 18

19 lieves to be common belief(2002:707). This entails that different speakers can have different presuppositions, i.e., different beliefs about what is common belief,andthateachspeaker'sbeliefsmaydifferfromwhatisactuallycommonbelief in the group. To illustrate this, consider a group whose members are Felix andanna.if,forexample,felixbelievesthatitiscommonbeliefthatenglandhas aqueen,i.e.,thatannaalsobelievesthatenglandhasaqueenandbelievesthat thisiscommonbelief,thenhemaybemistaken.annamightnotbelievethatenglandhasaqueen,orshemaybelievethis,butalsobelievethatfelixdoesnotbelieveit,i.e.,thatitisnotcommonbeliefthatenglandhasaqueen.insuchascenario,theproposition'englandhasaqueen'willnotbecommonbelief,eitherbecausefelixbelievesthispropositiontobecommonbelief,butannadoesnotbelieve that England has a queen in the first place, or because, even though both FelixandAnnamayactuallybelievetheproposition,onlyoneofthem,Felix,believesthatitiscommonbelief. Definitedescriptionsrequiretheuniquenesspresuppositionofthespeakerto bepartofthecommonground(i.e.,commonbeliefonthepresentview).so,sentence(13),repeatedherefromabove,couldbeappropriatelyutteredbyfelixif theproposition'felixhasa(unique)sister' 13 wereinfactcommonbelief. (13) No,I'vegottopickupmysister. ButwhatifitisnotcommonbeliefthatFelixhasasister,asintheexamplewe consideredattheendofsection2.2.1?thesituationwasthefollowing:felixis askedbysomeoneheonlyjustmet,anna,whetherhewasgoingtolunch.herepliesbyuttering(13),and,eventhoughannadidnotbelievethatfelixhadasister,theutteranceisfelicitous.stalnaker(1974)didnotprovideasatisfactoryaccount of these kinds of cases, but, as we will see in the following, this problem seemstobesolvedinstalnaker(2002:709ff). 14 To better understand Stalnaker's proposed solution, we need to make two preparatory assumptions: The first assumption we need to make is that any speakerofalanguageisfamiliarwiththefactthatcertainexpressionsarestandardly used to mean something. Furthermore, once this fact is mutually recognizedbyagroup,itbecomespossibletoexploitthatrecognitiontoconveymore than, or something different from, what is usually conveyed by a given expressionorconstruction(2002:704,cf.alsothediscussionofexploitationintheprevious section, 2.2.1). For our purposes, it is important to note that definite descriptionsarerecognizedbyagrouptorequireauniquenesspresuppositionof thespeaker,i.e.,thattherelevantpropositioniscommonbeliefamongtheparticipantsofadiscourse.secondly,stalnakerassumesthattheoccurrenceofany kind of speech act is always a manifest event, "by which [Stalnaker means] an eventthat,whenitoccurs,ismutuallyrecognizedtohaveoccurred"(2002:708). 13Itseemsthatthedefinitedescriptionmysisterisnotreallythebestchoicetoillustratethegeneralbehaviourofdefinitedescriptions:Asentencelike(13)maybeacceptableandappropriate eveninsituationsinwhichfelixhasmorethanonesister.(roberts(2003:344)alsoseemsto supportthisintuition.)forthesakeofargument,however,wewillignorethistensioninthefollowing. 14InStalnaker(2002),theoriginalexampleisslightlymodified.Insteadofreplying(13)tothe questionwhetherthespeakerwillgotolunch,thespeakersaysican'tcometothemeeting I havetopickupmysisterattheairport(2002:709).wewillusetheoriginalversion. 19

20 So, whenever a speaker says something, both he and his audience will believe that he said something, and they both can reasonably assume that the other party believes it as well, i.e., it will become common belief that something has beenuttered.asanextstep,itshouldbenotedthatthefollowinginferenceisa validoneinthelogicofcommonbeliefthatwepresentedabove(2002:709): IfitiscommonbelievethatthespeakerbelievesthatitiscommonbeliefthatP, andtheaudiencebelievesthatp, thenitiscommonbeliefthatp. Wesaidabovethat apropositionpiscommonbeliefiffall participants believe thatp,allbelievethatallbelieveit,etc.so,ifitalreadyiscommonbeliefthat,say, Felix,believesthatPiscommonbelief,thenitisclearthatFelixalsobelievesthat P. If, additionally, the audience also believes that P and believes that Felix believes that P is common belief, then P will actually be common belief, because bothfelixandhisaudiencebelievethatp,bothbelievethattheybothbelieveit, etc.(aformalproofcanbefoundinschlenker(2006:4f.).)nowwearereadyto analyse our example: We said that Felix presupposes that P 'Felix has a sister' whenuttering(13)andthatannadoesnotbelievethatp.so,thepropositionpis notcommonbelief,isnotpartofthecommonground.whenfelixuttersi'vegot topickupmysister,then,firstofall,bothheandannawillbelievethathemade thisutterance.andtheywillalsobothbefamiliarwiththefactthatadefinitedescriptionlikemysisterrequiresauniquenesspresupposition,viz.somethinglike P,andthatthisfactcanbeexploited.So,uponhearingFelix'utterance,Annawill beabletoinferthatfelixispresupposingthathehasasister(andthatheasserts thathehastopickherup).shebelievesthatfelixbelievesthatitiscommonbeliefthatp.thisgivesusthefirstpartoftheaboveinference.butwhataboutthe secondpart?howdoesannacometobelievethatp?stalnakerarguesthat"ina normal situation it will be a prior tacit common belief that [someone] knows whetherornotshehasasister"(2002:710).annawillthuscometobelievethat Felixhasasister,becauseshebelievesthathespeaksappropriatelyanddoesnot trytodeceiveherinuttering(13)(cf.2002:709),whichmeansthatsheis"prepared to add[the proposition that Felix has a sister], without objection, to the contextagainstwhichtheutterance...isevaluated"(2002:710).thisgivesusthe secondpartoftheinferencepatternabove,andthereforeitwillbecommonbelief that P after Felix' utterance. The timing is crucial here (cf. 2002:709ff and Footnote14).Toillustratethisinageneralfashion,wewill,again,makeuseof(a slightlymodifiedversionof)oneofschlenker'srepresentations(2006:3): (i) Attimet,thespeakerbelievesthatthepropositionPistrue,butitisnot commonbeliefthatp.thespeakeralsobelievesthatatt+2itwillbecom monbeliefthatp. (ii) Attimet+1,thespeakerpresupposesthatP(e.g.byusingadefinitedescrip tion),therebymakingitclear(i.e.,makingitcommonbelief)thathebelieves thatp(att+1andafterwards),andthathealsobelievesthatpwillbecom monbeliefatt+2. (iii) Attimet+2,theaudienceupdatestheirbeliefs,takingintoaccountwhat happenedatt+1(thespeaker'sutterance): (a) Becauseofwhathappenedatt+1,itiscommonbeliefatt+2thatthe speakerbelievesatt+2thatitiscommonbeliefatt+2thatpistrue. 20

21 (b) TheaudiencealsobelievesthatP. (c) Giventheinferencepatternabove,itisnowcommonbeliefthatP. (iv) Attimet+3,allparticipantscheckthatPiscommonbelief.Duetot+2thisis indeedthecase,thusjustifyingthespeaker'sbeliefatt. A short remark about the claim that the speaker is justified in believing that P willbecommonbeliefatt+2:whetherornotthisbeliefisjustifieddependson whetherthespeakercanreasonablyexpecthisaudiencetocometobelievethatp uponprocessingthespeaker'sutterance(cf.2002:710). 15 Inourexample,Felix canreasonablyexpectannatobelievethathehasasisterbythistimebecauseof the assumptions that it is (albeit tacit) common belief that someone knows whetherhehas a sisterornot, and that, in general, speakerswill speak appropriately.theimportantobservationfromsentenceslike(13)isthat,sometimes, itisenoughforaspeakertoimplythatheispresupposingsomethingtoensure thathispresuppositionwillendupinthecommonground,willbecommonbelief.instalnaker'sownwords(2002:710): "[In] general, if it is common belief that the addressee can come to know from the manifestutteranceeventboththatthespeakerispresupposingthatφ,andthatφis true, that will suffice to make Φ common belief, and so a presupposition of the ad dresseeaswellasthespeaker." Note,again,thattheantecedentofthisclaimismadetruebytheassumptionthat competent speakers of a language know how to exploit the usual meaning of a certainlinguisticexpressionandareabletorecognizethiskindofexploitationin theutterancesofothers(cf.2002:704). Giventhediscussionsofar,wecanalreadyseethattheproposalpresentedin Stalnaker(2002)issuperiortotheearlierversion:Incertaincases,aspeakercan presuppose something felicitously even if it is not part of the common ground, and this presupposition can be added to the common ground. In Stalnaker (1974), this kind of accommodation of propositions did not seem possible. Accommodation,inthissense,doesnotnecessarilyinvolvetherepairofanalready established common ground; rather, it involves an adjustment of the common ground.adjustmentsofthekindjustpresentedareacommonfeatureofcontext enrichment,andthemechanismsunderlyingpresuppositionaccommodationare not different from those governing 'ordinary' proposition accommodation, i.e., the addition of propositions that are explicitly stated (2002: 711, see also pp ). We already saw in the previous section that information can sometimesbeconveyedindirectlywiththehelpofpresuppositions(theexamplewe gavetherewas(12),odad,iforgottotellyouthatmyfiancéandiaremovingto Seattle (2002:710), uttered in a situation in which the father does not know abouthisdaughter'sengagement),andstalnakercallsthesepresuppositions'informative presuppositions' (2002:710). The indication of a speaker that he is taking an informative presupposition for granted is not always acceptable, though.thedaughterfromourexamplewillmostprobablysucceedinconveying some new information to her father, but the utterance may nonetheless be re 15Unfortunately,whenexactlyitisreasonableforaspeakertoexpecthisaudiencetocometo believehispresuppositionisnotmadeexplicitinstalnaker'saccount,itratherseemsthatstalnakerreliesontheintuitiveforceofthenotionof'reasonability'. 21