wookiee statements, semanticism, and reasonable assertionâ€
TRANSCRIPT
1
WookieeStatements,Semanticism,andReasonableAssertion†
EduardoGarcíaRamírezInstitutodeInvestigacionesFilosóficas
UniversidadNacionalAutónomadeMéxico
Abstract
It is assumed that the content of an assertion is determinedeitherbythesemanticallydefinedcontentorbytheinteractionofthelatterwiththecontext.HereIpresentacounterexamplebymeansoftheWookieeproblem.AfterconsideringseveraloptionsIofferwhatappearstobeitsmostsatisfactorysolution.Thisrequiresthatwegiveuptheassumptionin favor of a view according to which it may be that semanticinformationdoesnot at all determine the content of an assertion, notevenpartlyso.
Keywords:assertion,semanticcontent,context,presupposition.
1Assertion
Allconversations takeplace inacontextandconveysomethingbyutteringsomeor
other sentence.On the receivedview (see Stalnaker1978and1987) the goal of an
assertion is to communicatewhat is said. The context and content of an assertion
interactinordertodeterminewhatissaid.
In order to properly describe this interaction Stalnaker 1978 proposes a two‐
dimensional model. One dimension is determined by the contextually relevant
information:e.g.,whoisthespeaker,whicharethesalientobjects,andwhatdoesthe
world look like. The other dimension is determined by semantic information: the
meaningofthewordsusedbytheparticipants.
†TheauthorwouldliketothankAndyEgan,Shen‐YiLiao,andJoshDeverforveryhelpfulcomments.
2
Thismodelhasbeenusedandinterpretedindifferentways.Stalnaker2004claims
heintendedtousehismodelwitha“metasemantic”interpretationaccordingtowhich
the model offers a mere description of the context‐content interaction within
conversations.Others(seeChalmersandJackson2001andChalmers2002)intendto
usethemodelwitha“semantic”interpretacionaccordingtowhichthemodeloffersa
descriptionofthesemanticandconceptualknowledgethatcompetentspeakershave.
Both interpretations,however, agreeonwhat I call “the semanticist assumption”or
SAforshort:
SemanticistAssumption(SA):thecontentofanassertionisdeterminedeitherby
the semantically defined content of the expression used or by the
interactionbetweenthelatterandthecontext.
It is easy to see howSAworkswithin two‐dimensional semantics. According to
Chalmers2002,forexample,everylinguistictermhastwodifferentkindsofmeaning
or content. On the one hand, A‐intensions track epistemic dependence. Epistemic
dependencetellsushowthesentence’struth‐valuedependsonthespeaker’sapriori
(i.e.,semantic)knowledge.Forexample,forallIknowaprioriapurelyindexicaluse1
of theEnglishsentence ‘Iamhere’ is true ifandonly if thespeaker is locatedat the
place of the utterance. On the other hand, C‐intensions track factual dependence.
Factualdependencetellsushowthesentence'struth‐valuedependsonthefacts.For
example,ifMuses`Iamhere'indexicallywhilebeinginCambridge,whatMsayswill
betrueifandonlyifMisinCambridge.
1Imean“purelyindexical”inKaplan’swayinordertoexcludedemonstrativeusesof‘here’wherethespeakerispointingataparticularlocationdistinctfromhers;e.g.,whensomeoneistryingtogiveherlocationbypointingatamapandsaying‘Ilivehere’.
3
Two‐dimensionalsemantics(seeStalnaker1978,Chalmers2002,2004,and2006)
canbeunderstoodasclaimingthatforeverydeclarativeassertionorthoughtthereis
atwo‐dimensionalintensionthatmapspairsofworlds(oneconsideredasactualand
theotherascounterfactual)intotruth‐values.Theresultisatwo‐dimensionalmatrix
where the horizontal lines describe propositions that are determined by the
interactionbetweencontentandthecontext.Stalnaker1978famouslyarguesthatthe
corresponding two‐dimensional matrix includes a second distinguishable kind of
proposition:thatwhichisdescribedbythediagonalofthematrix.2
Thecentralclaimoftwo‐dimensionalsemanticsisthatthecontentofanassertion
willbeeitherahorizontalpropositionor thediagonalone.There isnothirdoption.
Now, given that both, horizontals and diagonal, are determined either by the
semanticallydefinedcontentorby the interactionof the latterwith thecontext, the
two‐dimensionalviewiscommittedtoSA.
SemanticistAssumption(SA):thecontentofanassertionisdeterminedeitherby
the semantically defined content of the expression used or by the
interactionbetweenthelatterandthecontext.
Whether two‐dimensionalist or not, almost everyone seems to agree with
Stalnaker’s 1978 claim that the content of an assertion is determined in the way
describedbySA.Somesuchassumptionseemstounderlie, forexample, thecurrent
debateconcerningrelativism(seeEgan2005andCappelenandHawthorne2009).
2For several reasons, Chalmers 2004 and Stalnaker (see Block and Stalnaker, 1999, and Stalnaker,2004) distinguish this from Chalmers’ A‐proposition. The distinction, however, is irrelevant for thepurposeofthisdiscussion.
4
Some reject the idea that all truth‐conditionally relevant information is
semanticallyencodedinfavorofwhatisknownastruth‐conditionalpragmatics(see
Chomsky1996,Travis1997,Bezuidenhout2002andAsherandLascarides2003).On
this view, there is truth‐conditionally relevant contextual information that is not
semanticallyencoded.Still,itisassumedthecontentofanassertionisatleastpartly
determinedbythesemanticallyencodedinformation.SomethinglikeSAistakentobe
trueevenwithintruth‐conditionalpragmatics.
The case is clearer for Neo‐Rusellians like Soames 2007 who reject the two‐
dimensional approach just described. On this view, the content of an assertion is
always the semantically defined proposition. This suggests an even stronger
endorsementofSA.
Soitseemsfairtosayphilosophersoflanguagehaveadoptedasemanticistattitude
by assuming that something like SA is true. As I will show, this assumption is
mistaken.Thereare informativeassertions the contentofwhich isnot semantically
determinedatall,notevenpartlyso.
2Theoreticconstraints
Speakers seem to utter sentence tokens, more often than not, for communicative
purposes. It is plausible to think that utterances of declarative sentences count as
assertions. And it is plausible to think that what the assertion expresses for the
speakerisidentifiedwithwhatthespeakerwantstocommunicatewithherassertion.
Ifthisisaccepted,thentheremustbeawayinwhichatheorythatassumesSAhelps
usunderstandthisphenomenon.
5
Atfirstglanceitseemslikeitcanreadilydoso.Allitneedstodoisfindoutwhich
content, either the horizontal or the diagonal, gets to play the role of the asserted
content.But,whenitcomestocommunication,therearesomeconstraints.Itakeitfor
granted that any account—for sure, any two‐dimensional account— of
communicationmustobservethefollowingprinciplesowedtoStalnaker(1978).
Informative:Apropositionasserted(orthought)isalwaystrueinsomebutnotallof
thepossibleworldsinthecontext.
Nogaps:Anyassertiveutterance(orthought)shouldexpressaproposition,relative
toeachworld in thecontext,and thatpropositionshouldhavea truth‐
valueineachworldinthecontext.
No ambiguities: The same proposition is expressed (or thought) relative to each
worldinthecontext.
These principles work under the assumption that the goal of an assertion is to
reducethesetofpossibilitiesinthecontextset.Thiscannotbeachievediftherelevant
propositioniseithernecessarilytrueornecessarilyfalse,that’swhatweneedthefirst
principlefor.Thesamehappenswiththesecondprinciple.Ifthereisaworldwhere
theproposition in questionhasno truth‐value thenwewill not knowwhether that
possibility shouldbe carvedoutof the spaceofpossibilities; that’swhat the second
principleisfor.Similarconsiderationsapplytothethirdprinciple.Propositionscanbe
understoodasrecipesforlogicalcarving.Ifthematrixoffersmanyofthemwewillnot
knowwhichrecipetofollow.3
3Thethirdprinciplecomeswithawrinkle.Ontheepistemicinterpretation,theassertion(orassertive
thought)willcorrespondto twopropositionsatonce. Ifpresent,anambiguitywillberepresented in
6
It shouldbeclear thatany theory thatacceptsSAmustembrace theseprinciples.
Theydeterminewhatagivenspeakerexpressesbymeansofherassertiveutterance.
So here comes the challenge, can a theory that accepts SA account for the way in
whichspeakersuttersentencestocommunicateinformationwhilebeingconstrained
bythethreeprinciplesabove?Theanswer,Ibelieve,isnegative.
3TheWookieeproblem
Therearesuccessful(i.e., informative)assertionswhereneitherahorizontalnorthe
diagonalproposition–i.e., neither theC‐intensionnor theA‐intension, onChalmers’
view–canbesaidtodeterminethecontentoftheassertionin lightoftheprinciples
above.
Considerthefollowingsituation.IamlookingforTom.Weagreedtomeetatthe
departmentatnoon, either in the commons roomor inhisoffice. It’snoonandas I
walkthroughthehallwayIcalloutp:“Tom,whereareyou?”Tomrepliesbyuttering
q:“Iamhere”fromwithinhisoffice.Tom’sspeechactobservestheprinciples.Itlets
meknowwhereheis(i.e.,notambiguous),itiscontingent(i.e.,informative),anddoes
notseemtobeincomplete(i.e.,hasnotruth‐valueormeaninggaps).Thequestionis,
then,howshouldwerepresentthissituation?
thehorizontallines,whichcorrespondtotheexternalcontent.Followersoftheviewcouldsimplyreply
that there is no ambiguity since, in those cases, there is still a single proposition, the diagonal or A
proposition, represented by the matrix. If this is so, then so be it. The point is irrelevant for the
purposesofthepresentdiscussion.
7
Thereare twoworlds in the setofpossibilities:worldw1 inwhichTom is in the
commons room and world w2 in which Tom is in his office. The purpose of the
conversation is to let me knowwhich of these worlds is the actual one. Matrix A1
representsthetwo‐dimensionalpropositionofTom’sutteranceq.
A1 w1 w2
w1 T F
w2 F T
So,whichpropositiondidTomjustassert?Therearetwocandidates:thediagonal
andahorizontal.Butnoneof themobservetheprinciples.Thediagonalproposition
fails tomeet Informative because the sentenceTomuttered isbound tobe true in
everypossibleworldinwhichheuttersit.Inshort,ithasanecessarydiagonal,which
istrivial.Thehorizontalpropositions,determinedbytheC‐intensions,goagainstNo
ambiguities because Tom’s utterance expresses different propositions relative to
differentworlds.Sotherearedifferenthorizontalpropositions.Hence,neitherofthe
twocandidatesexplainshowTom’sutteranceobservestheprinciples.
Matrix‐levelupdate?
Ifyouareatwo‐dimensionalistyoumightbethinking:“butwhyshouldwebeforced
to pick one, either a horizontal or a diagonal proposition. Why not update on the
wholematrix?”
Theonlykindof“matrix‐level”updatethatIknowofisStalnaker’sdiagonalization
bymeansofthedaggeroperator.Stalnaker1978introducesthedaggerasapragmatic
8
tool, but there is no problem ifwewant to interpret it semantically (see Chalmers
2002).Thedagger†worksinaverystraightforwardfashion: ittakesamatrixasan
input and gives a matrix as its output by taking the diagonal of the former and
projectingitonthehorizontallinesofthelatter.Toillustrate,theresultofapplying†
toTom’soriginalproblematicMatrixA1isgivenbyMatrix†A1.
†A1 w1 w2
w1 T T
w2 T T
Asyoumay see, the result isnogood. Idonot seem tohave learnedanything in
virtue of the semantics of Tom’s utterance. Matrix †A1 does not offer any update
recipeforTom.Ofcourse,theremightbeothermoreinteresting,perhapsevenmore
semantic,kindsof“matrix‐level”update.Sofar,thereseemstobenoalternativeone.
Furthermore, it’sunlikely thatanynon‐arbitrarymatrix‐level strategywill solve the
problem.
It goes without saying that the pragmatic use of † is in no better standing (see
Stalnaker1978).Sofar,Tom’sfelicitousassertionisproblematicforbothsortsoftwo‐
dimensionalism.
Amorenaturalresponse
It is commonly agreed (see Stalnaker 1978 and 2002) that there are at least two
processes by means of which one can be informative. Accommodation, on the one
hand,takesplacepriortotheacceptanceofthecontentandaftertheutterancetakes
place. It simply adds on one more presupposition to the context, namely, that the
9
speechacttookplace.Reduction,ontheotherhand,takesplaceaftertheacceptance
of the asserted content by eliminating incompatible worlds. This gives the two‐
dimensionalista response:weare in troublebecausewe leftoutan importantstep:
wehavenotyetaccommodatedthespeechactintothecontext.
The speech act is an observable fact. The fact that we communicate verbally
presupposesthatwebothareabletohearthesoundsproducedbyoneanother.More
specifically, Tom presupposes that I presuppose that he presupposes that I can
identify the sound of his speech act and where it comes from. Tom is correct. I
presupposethis.Furthermore,fromtheobservablefactthatTomhasutteredasound
Icaninferthattheutterancecomesfromhisoffice.
Thisgivesusaderivedsetofpossibilities: the resultof accommodation from the
observedfactthattheutterancetookplace.ThisisthecontextinwhichIamsupposed
tointerpretTom’sutteranceanddetermineitscontent.Itincludesthepresupposition
(Φ),fromwhichwecanderive(Υ).
(Φ)thatTomisspeakingfromwithinhisoffice.
(Υ)thatTomisinhisoffice.
But(Υ)isincompatiblewithw1,theworldinwhichTomisinthecommonsroom.SoI
must get rid of it. The accommodated context includes onlyw2, theworld inwhich
Tomis inhisoffice.This,apparently, solvestheproblem.Theaccommodatedmatrix
representsasinglecontingentproposition:i.e.,(Υ).Suchpropositioncanexplainhow
Tom’sspeechactobservestheprinciplesofcommunication.
10
Anothertroublingfact
Unfortunately, the problem is still standing. For now that we have updated by
accommodatingthefactoftheutterancethesemanticallydeterminedcontentofTom’s
utterance doesn’t have any communicativework to do. Any utterance—indeed any
productionofnoises—wouldhavethesamecommunicativeeffect.Sonowweareleft
wondering why would Tom—and for that matter, anyone—would ever make that
utterancetoconveythatinformation.
What we learn from accommodation, by definition, has nothing to do with the
contentoftheutteredsentence.Thus,itisnotsomethingthatcanbeexplainedbyany
semantically determined proposition. However, we get too much from
accommodating these presuppositions. So much that prior to the acceptance of
content I already know that the actualworld isw2. This preempts the semantically
determinedcontentfrombeinginformative.
Itfollowsthat,accordingtothetheory,thesemanticallydeterminedcontentofany
appropriateuse4of theEnglishsentence ‘Iamhere’ isuninformative.Giventhat the
assertion is informative, it followsthatthecontentoftheassertionisnot inanyway
semanticallydetermined.
4Itwouldbeterriblyinappropriate(uncooperative)ofTomtosay,“Iamhere”,whenIaskhim,“Where
areyou?”bythephone.Amongotherreasons,becausesomecentralpresuppositionsarenotsatisfied:
e.g., thepresupposition that I be able to infer the locationof theutterance from the fact that it took
place.
11
TheWookieeproblem
Let memake this more dramatic. Suppose that Tom replies, instead, by uttering a
sentence‐token from a different language, say, r: “Ich bin hier”. Is this a successful
communicative act? Tompresupposes that I presuppose that he presupposes that I
can identify the soundshe isutteringand,hence, that I can infer the locationofhis
utterance from the observable fact that it took place. This presupposition is easily
satisfied.
Thistimethingslookabitworse.Notonlyisthecontentinformativelyirrelevant.It
isnotevenrelevantwhethertheparticipantsarecompetentintheuseofthelanguage
ofwhichasentence‐tokenhasbeenuttered.Howisitthenthatspeakersalwaystryto
conveythesemanticallydeterminedcontent?
Why not twist this a little bitmore? Tom is a StarWars fan. In particular, he is
pretty knowledgeable about Chewbacca and the Wookiee species. He knows, for
example,thattheyspeakShyriiwook.Likemanyotherfans,heknowsprettywellhow
to utter a sentence‐token of Shyriiwook. You only have to utter two vocal sounds
withoutusingyourvocalchords.Iwalkthroughthehallwayandask,p:“Tom,where
areyou?”CouldTomnotreplywithhisfavoriteShyriiwooksentence?IfhedoesIcan
dothesamesortofaccommodationdescribedaboveandfindoutwherehe is.Heis
not even speaking a human language (Yes, Tom is amazing). Yet, his utterance is
informativeandsuccessfulinmeetingthegoaloftheconversation.
WhatistheWookieproblemabout?Theproblemisnotthatthetheoryisunableto
explainthosecaseswhereTomdecidestospeaklikeChewbacca.Theproblem,rather,
isthatthemeaningof‘Iamhere’isnotplayinganyinterestingroleinthestoryabout
12
howTomusesittocommunicate.AssumingthetruthofSA,onewouldhavethought
that ‘I amhere’ is, in virtueof its semantics, uniquelywell suited for conveying the
sortofinformationthatTomistryingtoconvey.Butitseemsthatitisnot.
That is the Wookiee Problem. Thus far, the theory cannot tell us howWookiee
utterances–e.g,Tom’sq:“Iamhere”–managetobeinformativeinvirtueofasserting
anysortofsemanticallydeterminedcontent.
4Areasonablesolution
Speakers usually observe certain principles when communicating. It is good
communicativepractice, forexample,toberelevant,clear,notambiguous,brief,and
orderly. So, when I ask p: “Tom, where are you?” Tom should not reply with “No
worries!”unlesshewantstochangethetopic;orwith“Ichbinhier”,unlesshewants
tobeobscure.Similarly,Tomshouldnotreply inShyriiwookbecausethatwouldbe
obscure,unclear,andirrelevant.
Thisgivesusasimplesolution.Whyisq,anutteranceoftheEnglishsentence‘Iam
here’,uniquelywellsuitedforTomtoconveythesortofinformationthathewantsto
communicate? The answer is: because q is uniquely well‐suited, in virtue of its
semantics,forTomtosaysomethingrelevant,clear,brief,etc.
The example we are concerned with involves a context in which by simply
accommodating the fact of the utterance we manage to update all the relevant
information.Yet,itwillbebest,forcommunicativeconsiderations,ifthespeakerpicks
an utterance the semantically determined content of which coincides with the
informationwe have contextually inferred. That way wewill avoid confusions and
13
misunderstandings.Forexample,anutteranceofq semanticallyexpresses thesame
propositionthatgetscommunicated,therebyallowingTomtoberelevant.
Wookieeutterancesseemtocarrytoomanypresuppositionswiththem;somany
thatthesemanticallydeterminedcontentbecomes irrelevant,butonlypartlyso.The
informativepurposeofthespeechact issatisfiedpriortotheacceptanceofcontent.
But being informative is not the only goal of a conversation. By uttering q Tom
managestobewellbehaved.
Thisisnotthegoodoldtwo‐dimensionalpicture,clearly.However,itmightstillbe
acceptable for two‐dimensionalists. It would be surprising otherwise. They simply
needtoaddsomethinglike:whentheprinciplesofassertionarenotobserved,itmust
bethattheyarebeingexploitedformanners’sake.
Thesolutionis,however,clearlyincompatiblewiththegenerallyacceptedtruthof
SA.
SemanticistAssumption(SA):thecontentofanassertionisdeterminedeitherby
the semantically defined content of the expression used or by the
interactionbetweenthelatterandthecontext.
IfthereasonableaccountoftheWookieeproblemIjustgaveiscorrect,andIseeno
reasontothinkitisnot,thenthecontentoftheassertionisnotdeterminedeitherby
thesemanticallydefinedcontentoftheexpression‘Iamhere’orbytheinteractionof
thelatterwiththecontext.Theonlyrelevantrolethatthesemanticfeaturesof ‘Iam
here’playinthiscaseispurelydiplomatic,toputitsomehow.Semanticinformationis
relevantonlyformanner’ssake,notforcontent’ssake.
14
5Wookieestatementsandstipulations
AppropriateutterancesoftheEnglishsentence‘Iamhere’are,ofcourse,nottheonly
onesthatcanhelpmakethecaseagainstSA.Therearemanyothercases,withinand
across languages,ofWookieestatements.Consider, forexample,appropriateusesof
sentencessuchas‘Iamspeakingnow’, ‘Iamherenow’,and‘Iexist’.5Allappropriate
utterances of these sentences are such that the presuppositions are enough to
determine thecontentof theassertion.Evenbeforedetermining themeaningof the
relevantsentences,assoonasthespeechacttakesplacetheaudienceknowsthatthe
speakerisspeaking,thatthespeakerislocatedattheplaceandtimeoftheutterance,
andthatthespeakerinfactexists.Idubthese“Wookieestatements.”
Wookiee statements have one further feature in common: their associated two‐
dimensionalmatrix ismadeupofdifferentcontingenthorizontalpropositionsanda
necessary diagonal. As such, they are similar to another kind of problematic
statements: stipulations. This is not the place to offer a theory of stipulations. But
reflectingonsomeplausibleconsiderationswillhelpusseethat theyconstituteone
morecaseofWookieestatements.
Awell‐knownexampleisthatofthestandardmeter.Consideracompetentspeaker
making the following stipulation s: “This length is one meter”, while point at a
particularmetalrod.Weknowthat,whatevertheactuallengthoftherelevantmetal
rod is, itwillbeonemeter long.Thus, regardlessof contextualvariations, including
5 Another, yet controversial, example is given by `Honey, I’m home!’. If one considers it to beinappropriatewhenutteredataplacedifferentfromhome,thensuchutterancedoespresentuswiththesameWookieeproblem.Othercontroversialexamplesare‘Iamthinking’and‘Iamthinkingnow’.The latterarecontroversial insofaras it is controversial to claim that it isa commonpresupposition(i.e.,commonground)inordinaryconversationsthatparticipantsarethinking.
15
variationsinlengthoftherelevantmetalrod,thespeakermakingthestipulationswill
assert something true. This gets described in the two‐dimensional matrix as a
necessarydiagonal (seeMatrixB). Suppose the stipulation takesplace in a context
wheretheparticipantsignorewhattheactuallengthofthemetalrodis,andconsider
the following possibilities:worldw1where the length of themetal rod the speaker
referstoisinfactl1,andworldw2whereitisl2,andl1≠l2.
B w1 w2
w1 T F
w2 F T
Not surprisingly, the same Wookiee problem arises. Whoever makes the
stipulation s will be asserting nontrivial information. Which proposition is she
asserting?Itwillnotbethediagonal,foritistrivial;andwecannotpickahorizontal
propositionbecause there isnotenough information, including thesemanticone, to
determinewhetherweare inw1orw2. It isalsonotviable todiagonalize, since the
resultingpropositionwillstillbetrivial.Furthermore,ifwepaycloseattentiontothe
processofaccommodation,bymeansofwhichthefactthatthespeechacthastaken
placeisincludedinthecontext,werealizethatthereisenoughinformationamongthe
presuppositions to preclude the semantic information from being informative and,
thus,fromdeterminingthecontentoftheassertion.
Itseemsatleastintuitivelycorrecttointerpretthespeakerissuingthestipulation
sasassertingsomethinglike(Β):
(Β):Fromnowon‘onemeter’willbecorreferentialwiththisuse‘thislength’.
16
There are at least two reasons to think that the information has to be so
underdetermined. On the one hand, stipulations are ways of “coining” new terms.
Ludlow2006and2008forcefullyarguesforadynamicviewofthelexiconaccording
to which “discourse participants routinelymint new linguistic items” and also that
what seem to be “common coins […] are typically “thin” – in the sense that their
meanings are underdetermined and fleshed out on a case‐by‐case basis.” [Ludlow,
2008, p. 115] If, as Ludlow suggests, commonly used terms have a very
underdeterminedmeaning, one should expect that new, not previously used, terms
have a radically underdetermined meaning, if any at all, at the moment of their
issuing.Thisclaimisratheruncontroversial:ifwewanttointroducetheterm‘T’into
our linguistic practice we cannot do this by presupposing that ‘T’ meansm, for in
ordertobeunderstoodwewouldhavetopresupposethatouraudiencepresupposes
that ‘T’ means m. That is tantamount to presupposing that ‘T’ has already been
introduced.
Ontheotherhand,andpartly invirtueofthepreviousconsideration,thereisno
way to further determine the information conveyed by s. One might think, for
example,thatwhatthespeakerconveysis,atleast,alittlebitmoredeterminedthan
(B),somethingperhapslike(Δ):
(Δ)Fromnowon‘onemeter’willrefertothislength.
However, the context inwhich s appears offers noway to determinewhat ‘this
length’refersto.All theparticipantscanlearnfromthespeaker’s issuingofs isthat
‘onemeter’referstowhatever it is that the lengthof therelevantmetalrod is.So it
17
seems that the only way of taking (Δ) as appropriately describing the information
conveyedbythespeaker’sissuingofsistotakeittomeansomethinglike(Β).
(Β):Fromnowon‘onemeter’willbecorreferentialwiththisuseof‘thislength’.
Ifso, then itseemsthat thecontentof thespeaker’sstipulations: “This length is
one meter” is in fact determined by sheer accommodation. As it happens with
Wookieestatements,oncetheparticipantsaccommodatethefactthatthespeakerhas
utteredthesentence‘Thislengthisonemeter’,theycaninfer(B)withoutbotheringto
considerthemeaningofthetermsused.Alltheyneedtoknowtodrawthatinference
isthattherelevantspeechactisastipulation,butthatpieceofknowledgeiscertainly
notsemanticbutmerelycontextual.
It seems, then, that stipulations, such as the one described, constitute another
counterexampleagainstthesemanticistassumption.
SemanticistAssumption (SA): the content of an assertion is determined either by
the semantically defined content of the expression used or by the
interactionbetweenthelatterandthecontext.
Foritseemsthattheinformationassertedbythespeaker’sissuingofs:“Thislengthis
onemeter”issomethinglike(B),whichisfullydeterminedbymerelyaccommodating
the fact that thespeechacthas takenplace,prior to theacceptanceofanysemantic
information. Furthermore, it seems that (B) simply does not express any of the
information that the English sentence ‘This length is one meter’ may semantically
encode.
18
6Objectionsandreplies
Could the semanticist philosopher of language not claim that the English sentence
‘Thislengthisonemeter’semanticallyencodestheinformationthat“‘thislength’and
‘one meter’ are correferential”? If so, could she not claim, furthermore, that the
contentofthespeaker’sissuingofsisinfactsemanticallydefined?
Thereisinfactarathersimplemovethatthesemanticistcouldmake:toclaimthat
partof the lexicalmeaningof eachword is theword itself.Basedonevidence from
linguisticpractices,Geurts1998claimsthat“ALLexpressionsfromanylanguageare
equivocal in a way: besides their ordinary meaning(s) they can also be used to
designate themselves.”[Geurts, 1998, p.291] Semanticist philosophers may want to
followGeurtsandclaim,forexample,thateveryreferentialexpressionhasatleasttwo
referents:theobjectitisordinarilyusedtoreferandtheexpressionitself.
Fromthisitwouldfollow,againstwhatIsaidtowardtheendoflastsection,that
the information conveyed by the issuing of the stipulation s: “This length is one
meter”, namely, that ‘this length’ and ‘one meter’ are correferential, can be
semantically determined. In other words, if the semanticist makes the “equivocal”
move described by Geurts 1998, she could claim that the proposition asserted by
issuing s could be either one of the semantically defined contents of the relevant
Englishsentenceortheresulttheinteractionofthelatterwiththecontext.
Furthermore,thesemanticistcouldmakeasimilarclaimaboutthecasesofwhatI
have called “Wookiee” statements. After all, the information conveyed by Tom’s
assertion of p: “I am here”, namely, that Tom is in his office, coincides with the
propositiondeterminedbytheinteractionofthesemanticallydefinedcontentofthe
19
English sentence ‘I am here’ and the relevant context. So, it seems that even with
Wookiee statements the content of the assertion couldbe semantically determined,
eitherbythesemanticcontentorbytheinteractionofthelatterwiththecontext.
TheseobjectionswillhelpclarifythepointIamtryingtomake.Iamnotclaiming
that the content of Wookiee assertions or statements cannot or could not be
semanticallydeterminedeitherbythesemanticcontentorbyitsinteractionwiththe
context.WhatIamclaimingisthatthecontentofthesepeculiarassertionsisnot,asa
matteroffact,semanticallydeterminedatall,notevenpartlyso.
Similarly,theargumentformycentralthesisdoesnotconsistinshowingthatthe
assertedcontentcannotbysemanticallydetermined.Rather,theargumentconsistsin
showing that the relevant information is contextually conveyed in a purely non‐
semanticmanner.TodothisIhaveshownthatpriortotheacceptanceofthesemantic
contentof thesentenceuttered, i.e.,evenbeforeanyassignmentofsemanticvalues,
theaudiencehasalreadyinferredtherelevantassertedinformation.Thus,evenifthat
sameinformationcanbesemanticallydetermined,thesemanticcontentisredundant
and, hence, useless for truth‐conditional orcontent purposes.As the solution to the
Wookieeproblemshows, theonly role that semantic contentplays in thesecases is
ratherperipheral:ithelpsthespeakerobserveberespectful,clear,etc.
ThatmuchisclearforthecaseofTom’sassertionp,andingeneralforappropriate
assertionsofsentencessuchas‘Iamspeakingnow’,‘Iamherenow’,and‘Iexist’.But
somethingsimilarcanbesaidaboutstipulationssuchastheissuingofs:“Thislength
isonemeter.”Themetalinguisticinformationconveyedbysuchastipulation,namely,
that ‘this length’ is correferential with ‘one meter’, gets conveyed prior to the
20
assignmentofsemanticvaluesto‘Thislengthisonemeter’.Thesemanticinformation
encodedby thisEnglish sentencedoesnotdetermine the contentof the stipulation.
Thatjobhasalreadybeendone.Still,thesentence’ssemanticinformationisusefulfor
manner’s sake: since themetalinguistic information associatedwith itwill coincide
withthealreadydeterminedcontent,itwillhelpthespeakerbeclear,respectful,etc.
7Theinclusiveview
Ihavearguedallalongforamoreinclusiveviewofassertivecontent.Onthisview,the
semantically defined content of the sentence(s) usedmay not at all determine the
content of an assertion, not even partly so. This goes against what I called the
SemanticistAssumption.
SemanticistAssumption (SA): the content of an assertion is determined either by
the semantically defined content of the expression used or by the
interactionbetweenthelatterandthecontext.
OnthismoreinclusiveviewwhatwegetissomethinglikeReasonableAssertion
(RA).
Reasonable Assertion (RA): the content of an assertion may be determined by
differentmeans,onacasebycasebasis,andbywhatappearstobethe
most reasonable manner. It is sometimes determined by the
semantically defined content of the expression used or by the
interaction between the latter and the context, but purely contextual
non‐semanticfeaturesmayalsodetermineit.
ThereareverygoodreasonstoendorseRAinsteadofthetraditionalSA.Firstand
foremost,RAwillallowustohandlewhatwouldotherwisebeproblematicWookiee
21
assertions.Thiswillpotentiallyhave further ramifications.UnlikeSA,RA isopen to
assertionsbeingamoreflexibleactasit,intuitively,appearstobe.Suchaflexibleand
inclusivepositionseems, inprinciple,moreattractive than thestringentsemanticist
one.
Second,RAwillhelpusavoidcertainphilosophicalpuzzles.At leastsinceKripke
1980,somehavethoughtthatcertainspecialassertionsprovidespeakerswithwhat
would otherwise by a rather doubtful kind of knowledge: that of contingent yet a
priori truths. Such knowledge seems philosophically suspect for it purports to be
abouttheworldofourexperienceeventhoughtheaccesstoitisindependentofour
experience.
Kripke1980arguesthatissuingstipulations,suchass:“Thislengthisonemeter”,
providethespeakerwithsuchcontingentaprioriknowledge.Furtherdiscussion(see
Stalnaker1978andChalmers2002)hasportrayedsuchcontingentaprioritruthsas
describing a two‐dimensional matrix with contingent horizontals and a necessary
diagonal.6 If we accept RA and, with it, the reasonable account of the Wookiee
problem (see section 4), we can explainwhy such assertions seem special without
accepting that they provide speakers with anything like contingent a priori
knowledge. These assertions are special because their content is not determined
semantically,notevenpartlyso.Theircontentis,rather,contextuallydetermined.And
itisuncontroversialthatthereisnoaprioriaccesstocontextualinformation:itmust
be acquired through experience. Thus, such assertions may provide speakers with
6Stalnakerappears tono longerendorse thisdiagnosis (seeBlockandStalnaker1999andStalnaker2004).ItisunclearwhetherChalmersstilldoes(seeChalmers2010).
22
contingent truths about the world of their experience, but there is no experience‐
independentaccesstosuchtruths.
I hope this is enough to convince the reader that we need to give up our
semanticist assumption in favor of a more reasonable account of assertion. Those
that are not convinced will at least have to face a substantial challenge: to give a
semanticist‐friendlysolutiontotheWookieeproblem.
23
References
Asher, N., and Lascarides A., 2003: Logics of conversation. Cambridge: Cambridge
UniversityPress.
Bezuidenhout,A.,2002:“Truth‐conditionalpragmatics.”PhilosophicalPerspectives,15,
105‐134.
Block,N.,andStalnaker,R.,1999:“ConceptualAnalysis,Dualism,andtheExplanatory
Gap”inThePhilosophicalReviewVol.108,No.1,1‐46.
Cappelen,H.,andHawthorne, J.,2009:RelativismandMonadicTruth.Oxford:Oxford
UniversityPress.
Chalmers,D.,1996:TheConsciousMind,Oxford:OxfordUniversiityPress.
Chalmers,D.,2002:“TheComponentsofContent(revisedversion)”inChalmers(ed.)
Philosophy of Mind: Classical and Contemporary Readings http://
consc.net/papers/content95.htmlpagenumberstothelatter.
Chalmers,D.,2004: “EpistemicTwo‐dimensional semantics” inPhilosophicalStudies,
118,153‐226.
Chalmers,D.,2006:“Two‐DimensionalSemantics”inE.Lepore,andB.C.Smith,(eds)
2006: Oxford Handbook of the Philosophy of Language. Oxford: Oxford
UniversityPress,[http://consc.net/papers/twodim.pdf]pagenumbersto
thelatter.
Chalmers,D.,2010:“Propositionsandattitudeascriptions:AFregeanaccount.”Nous,
forthcoming.
Chalmers,D., (forthcoming)“TheNatureofEpistemicSpace”unpublished, inhttp://
consc.net/papers/espace.html.
24
Chalmers,D.,andJackson,F.,2001:“ConceptualAnalysisandReductiveExplanation”,
inThePhilosophicalReview,Vol.110,No.3,315‐60.
Chomsky,N.,1996:PowersandProspects.Boston:SouthEndPress.
Egan,A., “Epistemicmodals, relativism,andassertion.”PhilosophicalStudies,133:1‐
22.
Evans,G.,1977“ReferenceandContingency”inTheMonist,62,161‐89.
Frege, G., 1892: “On sense and nominatum.” In A.P. Martinich (ed.) 2002: The
PhilosophyofLanguage,Oxford:OxfordUniversityPress.
Geurts,B.,1998:“Themechanismsofdenial.”Language,74,2,274‐307.
Jackson,F.,1994:“ArmchairMetaphysics” inO’Leary‐Hawthorne, J.,andMichael,M.,
(eds)1994:PhilosophyinMind;theplaceofphilosophyinthestudyofthe
mind.KluwerAcademicPublishing.
Jackson,F.,1998:FromMetaphysicstoEthics.Oxford:OxfordUniversityPress.
Jackson,F.,2004:“WhyweneedA‐intensions.”PhilosophicalStudies,118,257‐77.
Ludlow,P.,2006:“Themythoflanguage.”CroatianJournalofPhilosophy,18,385‐400.
Ludlow,P.,2008:“Cheapcontextualism.”PhilosophicalIssues,18,1,104‐29.
Kripke,S.,1981:NamingandNecessity.Cambridge:HarvardUniversityPress.
Lewis,D.,1986:OnthePluralityofWorlds.Oxford:BasilBlackwell.
Soames,S.,2007:ReferenceandDescription.Princeton:PrincetonUniversityPress.
Stalnaker, R., 1978: “Assertion.” In, R. Stalnaker, 1999:Context andContent. Oxford:
OxfordUniversityPress.
Stalnaker,R.,1987:“Semanticsforbelief.”InR.Stalnaker,1999:ContextandContent.
Oxford:OxfordUniversityPress.
25
Stalnaker,R.,2002:“CommonGround.”LinguisticsandPhilosophy25,701‐21.
Stalnaker, R., 2004: “Assertion revisited: On the interpretation of two‐dimensional
modalsemantics.”PhilosophicalStudies,118,299‐322.
Travis,C.,1997: “Pragmatics.” InB.HaleandC.Wright (eds.)1997:ACompanion to
thePhilosophyofLanguage.Oxford:Blackwell.