On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial:...
Transcript of On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial:...
![Page 1: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/1.jpg)
1
OnBeingTrivial:Grammarvs.Logic
GennaroChierchiaHarvardUniversity
January2019;RevisedSeptember2019ForthcominginGil Sagi & Jack Woods (eds.), The Semantic Conception of Logic: Essays on Consequence, Invariance, and Meaning. Cambridge, Britain: Cambridge University Press
Abstract.Thereisincreasingconsensusontheideathatcertainsentencesperceivedas“ungrammatical”owetheirstatusnottobeingsyntacticallyill-formed,buttotheirbeingL(ogically)-determinateandhenceinformationallytrivial.Clearly,however,noteveryL-determinatesentenceisperceivedasungrammatical,whichraisesthequestionofwhetherthereisaprincipledwayofsiftingamongtheL-determinatesentencesthosethatgiverisetoungrammaticalityfromthosethatdonot.Severalinterestingattemptshavebeenmadeinthisconnection(Gajewski,DelPinal),which,however,wearguefallshortofthetask.Weproposeamodificationandgeneralizationofsuchproposalsbasedonthenotionof‘modulation’ofwhataretermed‘thereferentialpoints’ofsentences(i.e.theirnonlogicalvocabularyandtheirvariables).Thisapproachhasfarreachingconsequencesforourunderstandingofthedividebetweenlogicalandnonlogicalvocabularyandfortheverynotionofsemanticcompetence.
![Page 2: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/2.jpg)
2
OnBeingTrivial:Grammarvs.Logic
1.Introduction:whatisitforasentencetobe‘trivial’?
Thedifferencebetweencontent(or‘openclass’,e.g.cat,table,run,..)wordsandfunction(or‘closedclass’,e.g.if,every,only,…)wordsisfundamentalintheinquiryintonaturallanguage.Thisdistinctionhasahighdegreeofcorrelationwiththevexedcontrastbetweenlogicalvs.nonlogicalwords.Thelatterisinturncentralinfiguringoutwhatlogicis.Myangleonthefunction/contentdistinctioniscenteredontheissueofwhataspectsoflanguageareuniversal,partofabiologicallydeterminedlanguagefaculty,andcanthereforebethoughtofascomponentsoftheininitialstateofthelanguagelearner(akaasUniversalGrammar–UG)vs.whichaspectsofitaretheresultofsociallearningandhave,therefore,apredominantlyhistoricaldimension.Addressingtheuniversalvs.languageparticulardistinctionnecessarilyinvolvesidentifyingfamiliesofwordsandstructuresthatmaybegoodcandidatesatbeingcomponentsoftheinitialstate;andlogicalwordsliketheBooleanfunctionsorthequantifiers,togetherwithotherlesscanonicaloneslike,say,evenoronly,areuncontroversiallycandidatesatbeingpartofUG.Theimportanceofthefunction/contentdichotomymanifestsitselfintherecurrentandarguablysuccessfulattemptsatexplainingaspectsofgrammarintermsoflogic.Theclaimemergingmoreandmoreforcefullyinthisconnectionisthatmanycasesoflinguisticdeviance/ungrammaticalityowetheirstatusnottotheviolationofsomesyntacticwellformednessconditionbuttothefacttherelevantstructuresarelogicallydetermined(i.e.logicallytrueorlogicallyfalse),andhenceinsomesense‘trivial’.Sinceclassicaltautologiesorcontradictions(oftheformIsJohnsmart?Well,heisandheisn’t)arelogicallydeterminedbutarenotperceivedasungrammatical,oneimmediatelyfacestheissueofhowtoteaseapart(possibly,asamatterofprinciple)trivialitiesrootedingrammarandperceivedasungrammatical(whichIwillcallG-trivialities)fromclassicaltautologiesandcontradictions(L-determinedsentences).Inwhatfollows,Iwillgooveranexample,rathercompellinginmyview,ofaclassofphenomenabestexplainedintermsofG-triviality.IwillthenpresentanddiscussthewayofconceptualizingthespecialstatusofG-trivialitiesvis-à-visstandardtautologiesputforthbyJ.Gajewski(2002).IwillpointoutaproblemwithGajewski’sapproachandsketchasolutionthatembodiesasomewhatdifferentviewoftherelationshipbetweengrammarandlogic. Asiswellknown,NegativePolarityItems(NPIs)likeeveroranyarerestrictedintheirdistributionto(roughly)‘DownwardEntailing’(DE)contexts,i.e.contextsthatlicense‘subsetinferences’suchasthosein(1a)asopposedto(1b):(1) a.Ididn’teatpizzaàIdidn’teatpizzawithanchovies Subsetinference(DE) b.IatepizzawithanchoviesàIatepizza Supersetinference(UE)ArepresentativesetofcontrastsinvolvingNPIsisgivenin(2)vs.(3):
![Page 3: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/3.jpg)
3
(2) a.Thereisn’tanypizzaleftb. Idoubtthatthereisanypizzaleft
c.Ifthereisanypizzaleft,wewon’tgohungry(3) a.*Thereisanypizzaleft b.*Ibelievethatthereisanypizzaleft. c.*Ifyouarehungry,thereisanypizzaleftOneandthesamestringofwords,namelythereisanypizzaleftisdeviantinanonDEenvironment(e.g.underbelieveorintheconsequentofaconditional)andbecomesperfectinaDEone(e.g.embeddedundernegation,orunderdoubtorintheantecedentofaconditional).Noticethatdroppingtheitemanyinthesentencesin(3)rendersthemgrammatical,whichconfirmsthattheoccurrenceofanyistheculpritfortheirdegradedstatusin(3).Thedevianceofthesentencesin(3)isquitesevere,comparabletoanagreementmismatchorabasicwordorderviolationoftheformboythewalkedin.Inspiteofthis,athesisthathasconsistentlygainedcredibilityisthatthereisnothingwrongwiththesyntaxofthesentencesin(3).Theproblemiswhollysemantic:Thesesentencesareunrescuablytrivial.Letmefleshoutthisprimafacieimplausibleclaim,startingwithaseeminglyunrelatedexampleoflinguisticdeviance. Considerthedialoguein(4):(4)SpeakerA:Howmanyofthe20papersyouhavetogradedoyouthinkyouwill
havegradedin2hours?SpeakerB:a.Possibly,eventen b.Possiblynotevenone c.*Possibly,evenone
Youwillagreethatwhile(4a)and(4b)arenaturalanswerstoSpeakerA’squestion,(4c)isdistinctlydeviant.Why?Useofevengenerallyconveysthatthepropositionevenappliesto(i.e.theprejacent)isregardedastheleastlikelyamongsomerelevantsetofalternatives:(5) a.IunderstoodevenChomsky’spaper b.Alternativesunderconsideration:GivensomesetAofcontextuallysalient
individuals,ALT={Iunderstooda’spaper:aÎA} c.PresuppositionofSentence(a):
UnderstandingChomsky’spaper<LIKELYUnderstandinga’spaper(foranyaÎA)Thewayinwhichthealternativesaretypicallyindividuatedisdeterminedbythecontext(e.g.,bysomequestionunderdiscussion),whichinturnisoftencodedintothefocalstructureofsentences.Inthecontextofthequestionin(4),theanswerin(4a)willevokeascaleoftheform:
![Page 4: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/4.jpg)
4
(6) <(intwohours)Iwillhavegraded10ormorepapers,
Iwillhavegraded9ormorepapers, …, Iwillhavegraded1ormorepapers>
Thearrowin(6)indicatestheentailmentpatternholdingamongthealternatives,whereifXisstrongerthanY,XistrueinfewerworldsandhencenecessarilylesslikelythanY(oratmostaslikelyasY).So,sentence(4a)statesthatIwillgradeprobably10assignmentsandthisistheleastlikelyandbestscenariooptionamongthealternativesthatstandachanceatbeingtrue,andthisyieldsafelicitousresponse.Considernext(4b).Thepresenceofnegationin(4b)reversesthescalein(6):(7)<(twohoursfromnow)Iwon’thavegradeda(single)paper,
Iwon’thavegraded2ormorepapers,…,
Iwon’thavegraded10ormorepapers>Inthisreversedscale,gradingnopaperbecomesthestrongest,andhenceleastlikelymemberoftherelevantalternativeset,whichagainmakesuseofevenappropriate.Atthispoint,thereasonwhy(4c)soundsweirdbecomesapparent.Withrespecttothescalein(6),naturallyassociatedwithpositivesentences,thesentenceIgradedonepaperistheweakestmemberandhenceitcannotbetheleastlikely.Claimingthecontraryresultsinacontradiction(forapropositioncannotbelesslikelythanitsentailments).So,itlookslikebehindtheimmediacyofourreactionto(4c)thereisarapidcomputationthatleadstothefollowingconclusion:sentence(4c)triggersacontradictory(i.e.trivial)presupposition.ThisseeminglyobscurecornerofthegrammarofevenconstitutesanillustrationofwhatImeanbyG-triviality.ThisexampleofG-trivialityisdirectlyrelevanttoNPIs.Supposethatanymeans
somethinglikeeven+one,andassociateswithascaleanalogousto(6).Thiswouldimmediatelyexplainwhythesentencesin(2)arefineandwhy.Morespecifically,replacinganyin(2)withsomethinglikeeven+one(single)isgrammaticalandyieldsvirtualsynonymsoftheoriginalsentences:(8) a.Thereisn’tevenone(singlepieceof)pizzaleft
b.Idoubtthereisevenone(singlepieceof)pizzaleftc.Ifthereisevenone(singlepieceof)pizzaleft,wewon’tgohungry.
Theany=even+onehypothesisalsoexplainswhythesentencesin(3)areungrammatical:Theyareallcontradictory,justlike(4c)is.Forexample,thelogicalformof(3a),repeatedhereas(9a),wouldbesomethinglike(9b):(9) a.Thereisanypizzaleft
![Page 5: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/5.jpg)
5
b.evenALT[$x[one(x)Ùpizza(x)Ùleft(x)]]»thereisevenONEpieceofpizzaleft
c.ALT={[$x[n(x)Ùpizza(x)Ùleft(x)]:nÎN}wheretheentailmentin(c)goesfromnton-m,foranynandm.
It’sasif(9a)wasinterpretedasthereisevenONE(singlepieceof)pizzaleftinreplytoahowmany-question:(9a)triggersacontradictorypresupposition.ThisaccountextendstoallDEenvironmentsandprovidesanarguablyelegantandsimpleexplanationforthedistributionofNPIs,thatdescendsdirectlyfromtheir(hypothesized)semantics. Thereisonestrikingfactthatseemstosupportthishypothesis.InmanylanguagesNPIsareexplicitlyformedbycomposingfocussensitiveadditiveparticlesthatmeanroughlyevenwithsomeitemthatexpressesalowquantitylikethefirstnumeral‘one’.Hereisarepresentativesample:(10) Hindi:ekbhii oneeven/also Tagalog:anu-ma-ng wh-even-CASE Italian:neancheuno negativeagr+alsoone(11)AHindiexample(fromLahiri1998): i.*ekbhiiaadmiiaayaa oneevenmancame ‘anymancame’ ii.ekbhiiaadmiinahiiNaayaa oneevenmannotcame ‘nomancame’Historicalchangeprovidesfurtherevidenceinfavorofthisview.AnycomesfromOldEnglishænig(lit."one-y,"“one-like”),whichinturnisderivedfromProtoGermanic*ainagas.TheProtoIndoEuropeansourceforthisclassofwordsis*oinos,thewordforone.Now,any’sGermancousineinigremainedaplainvanillaindefinitewithoutanegativepolarityuse.TheItaliancounterpartofany/einig,namelyalcuno/alcuni,hasasplitbehavior:thepluralisaregularindefinite,whilethesingularisanNPI.ItisnotimplausibletoconjecturethatwhenexpressionsofminimalamountstartbeingusedasNPIs,itisthroughtheassociationwithanadverbialparticlelikeeven,associationwhichsometimesgoesunexpressed. Allinall,itisclearthatthecombinationeven+oneisawidespreadsourceforNPIbehavior,possiblyeveninEnglish.Whichdirectlyleadustothefollowing(admittedlyrhetorical)question:ifthereasonforwhyeven+oneisrestrictedtoDEenvironmentsisnotanalogoustothereasonwhy(4c)isdeviant,thenwhatisit?1 Andherecomestheissueofinteresttoourpresentconcerns.OuraccountforNPI-violationsreliesonthefactthattheyturnouttobe,underaplausiblesemantics,contradictions/trivialities.Whichimmediatelyraisesthequestionofwhy
1TheLandscapeofPolaritySensitiveItemsismuchricherthantheonesketchedinthetext.ButIthinkthatdealingwithitinmoredetailstillrequiresdealingindepthwiththeproblemofhowlogicalityaffectsgrammar.Cf.,e.g.,Chierchia(2013)andreferencesthereinforamorethoroughinvestigationoftherelevantissues.
![Page 6: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/6.jpg)
6
aren’talltrivialitiesungrammatical.Whyisitthecasethat(12a.ii)isoddinreplyto(12a.i),while(12b.ii)isaperfectlynaturalanswerto(12b.i)?(12) a.i.Howmanypaperscanyougradeby5? ii.*Oh,evenone b.i.IsJohnsmart? ii.Well,heisandheisn’tWhymoreoverthecontradictionin(12a)requiresanalysistounveilitscontradictorynature,whiletheonein(12b)isreadilyaccessibletointrospection?Asitturnsout,thesequestionshaveprincipledanswers,aswewillsee.2.‘Modulated’LogicalForms:arestrictivecontextualism.
WhatwehaveseensofaristhattherearereasonstobelievethatNPIsinnonDEcontextsaretoberuledoutnotbecauseofsomesyntacticviolation,butbecausetheyarecontradictions(andhenceuselessincommunicating).However,thereareplentyofcontradictionsthatsoundfineandareinfactusedinconcretecommunicativesituations.So,howcantheideathatNPIsinnonDEcontextsarejustcontradictionsberight?Gajewski(2002)arguesthatthesolutionisrootedinthedistinctionbetweenfunctionandcontentwordsandwearenowgoingtoreviewhisproposal.Letusconsiderthekeysentencesin(13a)vs.(13b)andtheirrespectivesyntacticstructures.(13) a.[IsJohnsmart?]Heissmartandheisn’tsmart b. TP TP and TP hei T’ hei T’ is SC isn’t SC [tiAP] [tiAP]
Key:T=Tense;T’,TP=TensePhrases;SC=SmallClause;crossedoutconstituentsgounpronouncedbutareusedforinterpretivepurposes.
(14) a.Thereisanypizzaleft b. TP
There T’ is SC DP VP
AnyNP
![Page 7: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/7.jpg)
7
Thetreesin(13b)and(14b)constitutetheroughsyntacticanalysisofthetwokeyexamplesentences,alongrelativelyuncontroversiallines,easytotranslateintootherpopularapproachestotheseconstructions.Theymaybeviewedasthefunctionalskeletaofsentences(13a)and(14a)respectively.I.e.theyconstitutethekindofstructurethatgrammarwouldgenerateoutofthefunctionalelementsalone.Thepronounheiistreatedasavariable,whosevalueisassumedtobecontextuallyset,andvariablesqualifyaslogicalelements.WithintheframeworkofDistributedMorphology(see,e.g.,HalleandMarantz1993sentencesarecomposedbyassemblingtheirfunctionalstructurefirst;contentwordsareinsertedatalaterstagesoastotakeintoaccountthecontributionoffunctionalstructure.Thisdesignismeanttomakesenseofthefactthatthefinalshapeofcontentwordsissensitivetothefunctionalstructureinwhichtheyareinserted:thinkforexampleofthecommonplaceobservationthatPAST+VsequencespellsoutinEnglishasV-edifthechosenverbisregularlikelovedorwalked,butthesamegrammaticalinformationspellsoutdifferentlyiftheverbisirregular,likewentorhit.WithinDistributedMorphology,whichadoptsthis‘latelexicalinsertion’strategy,structuresroughlylike(13c)-(14c)correspondtoanactualphaseofthederivationoftherelevantsentences. Withthenotionoffunctionalskeletoninplace,Gajewski’sapproachtoG-trivialityisbasedonthefollowingcentralgeneralization.Asentencelike(14a)turnsouttobecontradictoryregardlessofthechoiceofcontentwordsoneinsertsinthecorrespondingstructure(14c),i.e.regardlessofwhichNandwhichVoneeventuallyselectsfromthelexicon.Asentencelike(13a),ontheotherhand,comesoutascontradictoryonlyifonechoosesthesameadjectivefromthelexiconforthetwoinstantiationsofthecategoryAP.Thisisthesourceofthedistinctionbetweenungrammaticalvs.grammaticalcontradictions:G-trivialitiesarecontradictoryforanychoiceofcontentwords. Letusputthisinsightinslightlydifferentterms.Considertheinterpretationsof(13a)/(14b)‘minustheircontentwords’,i.e.replacingthelatterwithvariablesoftheappropriatetype:(15) a.P(xi)Ù¬P’(xi) b.evenALT($x[one(x)ÙP(x)ÙP’(x)]) whereALT={$x[n(x)ÙP(x)ÙP’(x)]:nÎN}Formula(15b)iscontradictorynomatterhowthevariablesPandP’areinterpreted,while(15a)iscontradictoryonlyifPandP’aremappedontothesameproperty.Gajewski’soriginalalgorithmisa‘syntacticreplacement’versionofthisverysameidea:heproposesreplacingintherelevantLogicalForms(i.e.thecompletesyntactictreesfor(13a)and(14a))alloccurrencesofcontentwordswithdistinctvariablesofthesametype.Iftheresultiscontradictory,thesentenceisdeemedasungrammatical. Thisaccount,besidesbeingalgorithmic,hasanarguablynaturalfunctionalbasisinthepragmaticsofcommunication.Whatsuggestsitselfisthatthereasonwhysentenceslike(13a)areperceivedasgrammaticalisbecauseitispossible,and
![Page 8: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/8.jpg)
8
indeednatural,toreinterpretthetwooccurrencesofthe(unpronounced)adjectivesmartinslightlydifferentways(e.g.as“Johniscleverathisjob,butheisnotsavvyinthewayhemanagespeople,”orthelike).Butnosuchstrategycanbeofanyhelpwith(14a):nomatterhowwere-interprettheNortheV,contradictorinesspersists. Gajewski’sproposalprovidesuswithaprincipledwayofdistinguishingbetweengrammaticalandungrammaticaltrivialities.AppealtothisdistinctionisbynomeanslimitedtothecaseofNPIs.OtherphenomenathathavebeenarguedtorequireanaccountintermsofG-trivialityincludethedistributionoffor-vsin-Xtimeadverbials(Dowty1979),thedefinitenesseffectinthere-sentences(BarwiseandCooper1981),exceptiveconstructions(vonFintel1993),thedistributionofquantifiersincomparativeconstructions(Gajewski2008),weakislandviolations(Abrusan2014),andmore.Thisisjustapointertosomeoftherelevantliteratureandsoundslikealaundrylist.Theimportantfactisthatresortingtoungrammaticalcontradictionstoexplainpropertiesofgrammarisawidespreadpracticethatisprovingtobemoreandmorefruitfulinfiguringouthownaturallanguageworls. Thedistinctionbetweenfunction-vs.content-wordsissomewhatvagueandobviouslyinneedoffurtherclarification,butitplaysacrucialrolebothintraditionalaswellasincuttingedgelinguistictheories.Wewillcomebacktothefunction/contentdistinctioninlaterpartsofthispaper.Fornow,whatisimportantisthatweneedtoreconsiderthewidespreadstancethatsyntaxdetermineswell-formednessandsemanticsdetermineshowwell-formedsentencesareinterpreted.OntheviewIseemyselfforcedtoadopthere,grammaticalsentencesareconstitutedbywell-formedstructuresthatarenonG-trivial,anddeterminingthesetofG-trivialsentencesinvolvesresortingtoanempiricallydeterminedcastof(interpreted)function(/logical?)words,asperGajewski’salgorithm. DelPinal(2017)arguesforaninterestingmodificationofGajewski’sproposal.Hisproposedmodificationismeanttodirectlyreflectthefunctionalistic/pragmaticinterpretationofGajewski’sproposal.DelPinal’sideacanbeillustratedbywayofexample,representingtheinterpretationsofourtwokeyillustrativesentences(13)and(14)asfollows:(16) a.g(smart)(xi)Ù¬g’(smart)(xi) b.evenALT($x[one(x)Ùg(pizza)(x)Ùg’(left)(x)]) whereALT={$x[n(x)Ùg(pizza)(x)Ùg’(left)(x)]:nÎN}Theassumptionhereisthattheinterpretationofcontentwordscanbemodulatedthroughthe(optional)insertionoffunctionsg,g’,…whicharecontextuallydeterminedandmapanysemanticobjectintosomethingofthesamelogicaltype.Thedefaultinterpretationofthesemodulatingfunctions(wheneverpresent)issimplytheidentitymap.However,whenthedefaultinterpretationleadstoacontradiction,as,say,withsentenceslikeJohnisandisn’tsmart,theoffendingitem,iinthiscaseaproperty,typicallygetsmodulatedonthebasisoftheintentions,communicativegoals,etc.oftheillocutionaryagents,forexampleasin(17):(17) Johnisg(smart)andisn’tg’(smart)
![Page 9: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/9.jpg)
9
g(smart)=cleverathisjob g’(smart)=savvyinmanagingpeopleThestrikethroughindicatesthatthefirstinstanceoftheAPsmartiselidedunderidentitywiththefirst,butpresentforinterpretivepurposes.Onemighttrytoexplaintheacceptabilityoftheprimafaciecontradictorysentencesofthissortindifferentmanners,e.g.throughtheresettingofsomecontextualparameterimplicitinadjectiveslikesmart,thatrequirea‘comparisonclass’(smartwithrespecttowhat?).However,thisalternativeaccountdoesn’textendinanyobviouswaytoexampleslikeHowistheweather?Well,itrainsanditdoesn’train(»itrainsonandoff).Moreover,resettinggrammaticallydeterminedvariables(suchascomparisonclasses)isgenerallybannedinVP-ellipsisenvironmentssuchasthosein(17).Thepresentproposaldoesn’tsufferfromthesedrawbacksandisthussuperiortoalternativesrelyingsolelyongrammaticallydeterminedparameters. Wecalllogicalforms/interpretationssuchasthosein(16)‘modulatedlogicalforms’.2WemayregardDelPinal’sproposalasvariantofGajewski’sthatembedsthelatterwithinacontextualiststanceaccordingtowhichthestandardinterpretationofcontentwordsiscontextdependent.OnDelPinal’smodification,G-trivialsentencesarethosethataretrue/falseforanyvalueofthemodulatingfunctions,whileclassicaltautologies/contradictionsarethosethataretrue/falsewhenallmodulatingfunctionsareinterpretedasidentitymaps.Obviously,G-trivialsentencesareapropersubsetoftheclassicaltautologies/contradictions.Thisapproachrequiresconstraintsonmodulation.MappingsomecontentwordWintosomethingg(W)ofthesametypeisnotenough.Themappingsappealedtomustbeinsomesense‘natural’,addressthecommunicativeintentionsoftheillocutionaryagentsinpragmaticallysensibleways,e.g.byresolvingappropriatelythequestionsunderdiscussion(howsmartisJohn?).Nobodyhasafullyworkedouttheoryofwhatmakesamodulation‘natural’,beyondappealingtocontext,questionunderdiscussion,andthelike.3Still,withtheselimitsacknowledged,DelPinal’sproposalprovidesausefulcharacterizationofG-trivialityandismoregeneralthanGajewsky’s,sinceresettingofbasicwordmeaningshappensextensively.4Herearetwocases.Onecanbeillustratedbyafamousexample,duetoG.Nunberg:2DelPinalusestheterms‘rescaling’and‘rescaledlogicalforms’inthisconnection.Iprefertheterm‘modulation’.ThisterminologicalchoiceforeshadowsageneralizationofDelPinal’sapproachintwoways,whichwillbedevelopedinSection4below.FirstDelPinallimitsrescalingtopredicates(ortypesthat‘endin’thetype<e,t>),whileIgeneralizeitalsotoindividuals.Second,Ithinkthatthepresentproposalfitswithandaccommodatesalsoatreatmentofderebelief.3DelPinalsuggeststhatmodulationmaybesubsective,i.e.mapapropertyintosomesubproperty.Ibelievethisconstrainttobetoorestrictiveinlightofexampleslike(18)andothersconsideredbelow.4NoticethatDelPinal’sproposaldoesnotmakecontradictionsinexpressibleinEnglish,aworryexpressedbyananonymousreferee.Italldependsontheintendedinterpretationofremodulation.Inthefollowingquote,fromAristotle’sMetaphysics,forexample,acontradictionisclearlyintendedandcommunicativelyeffective:
![Page 10: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/10.jpg)
10
(18) a.[Awaitertoafellowwaiter:]Thehamsandwichwantshisbill b.wants(hisbill)(ix[g(hamsandwich)(x)]) whereg(hamsandwich)=personthatorderedthehamsandwichThelogicalformin(18b)illustrateshowNunberg’sexamplemightbehandledonanapproachbasedonmodulation.Asecondclassofcasesisexemplifiedbysentenceslike(19a),duetoB.Partee:(19) a.Tommybelievesthatcloudsarealive b."w[BELTOMMY,w0(w)®"x[g(cloud)(w)(x)®alive(w)(x)]] Whereforanyindividualuandworldw,BELu,wisthe(characteristic
functionofthe)setofworldscompatiblewithu’sbeliefsinw.Asentencelike(19a)expressesa(typicallydere)beliefaboutinstancesofthecloud-kind,namelythattheyhavelife.Givenhowlifeisunderstoodinourlinguisticcommunity,thebeliefattributedtoTommyconstitutesametaphysicalimpossibilitythatfailsineverypossibleworld(muchlikeHesperuscannotbedifferentfromPhosphorus,givenhownamesworketc.).Hencethesemanticsfor(19a),say(19b),wouldcondemnTommy’sbeliefstatetoincoherence(underthedefaultinterpretationofgasidentity).Modulationoffersawayout.Theg-functionin(19b)maymapcloudsintosomecloud-likelivingcreature,forexample. DelPinal’sproposal,besidesprovidingaconceptualembeddingofGajewski’sapproachwithinanindependentlyplausibleformofcontextualism,alsohas,perhaps,afurthertechnicaladvantage.ConsideryetagainourtoyNPIviolation,repeatedhere.(20) a.*Thereisanypizzaleft
b.evenALT($x[one(x)Ùg(pizza)(x)Ùg’(left)(x)]) c.ALT={$x[n(x)Ùg(pizza)(x)Ùg’(left)(x)]:nÎN}Itiscrucialthattheinterpretationofthenonlogicalwords(pizza,left)bekeptconstantintheassertion(19b)andacrossallofthealternativesin(19c).OnasubstitutionalapproachlikeGajewski’s,wherelogicalskeletaareobtainedbyreplacingeachoccurrenceofthenonlogicalwordswithdistinctvariables,someworkisrequiredtoensurethatthereplacementisuniformacrossthealternatives.
(a) Itisimpossiblethatthesamethingcanatthesametimebothbelongand
notbelongtothesameobjectandinthesamerespect,andallother specificationsthatmightbemade,letthembeaddedtomeetlocalobjections (Aristotle,Metaphysics,1005b19-23)In(a),thewordbelonghastoberemodulatedviatheidentitymapandhencethesentenceinboldfaceexpressesagenuinelogicalcontradiction,asthatisthechoicethatmakespragmaticsense.
![Page 11: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/11.jpg)
11
ItisnothardtoimagineadefinitionoflogicalskeletathatwouldNOThavesuchaproperty,therebyyieldingwrongpredictions.OnDelPinal’sapproachthisissuedoesn’tarise. Onthebasisoftheseconsiderations,IconcludethatDelPinal’sproposalconstitutesafriendlyandusefulamendmenttoGajewski’soriginalapproach.ButbeforeattemptingsomegeneralreflectionsonwhatthistakeonG-trivialitytellsabouttherelationbetweenlogicandgrammar,weneedtoaddressaproblemthatbothGajewski’sandDelPinal’sapproachleaveopen.3.Theproblemofboundvariables. Considersentencesofthefollowingform:(21) a.Johnisneverhimself b.Yesterday,JohnmanagedtobemoreeloquentthanhimselfThefirstrelevantobservationisthatthesesentencesareperfectlygrammaticalandcommunicativelyuseful.Thesecondnoticeablepointisthat,takenliterally,theyarecontradictory.Third,theseexamplesarebeyondrepaironthemodulationapproachadoptedhere,forthefollowingreasons.Reflexivepronouns,comparativemorphemes(more),andnegation(never)areprototypicalfunctionalitems.Thusthemodulatedstructureof(21a,b)isgoingtoberoughlyasfollows:(22) a.g(John)isneverhimself b.g(John)g(managed)tobemoreg(eloquent)thanhimselfThe(pseudo)formulaein(22)arecontradictoryforanychoiceofg.Hence,sentences(21a,b)shouldbeungrammaticalaccordingtothecharacterizationofG-trivialityweareadopting.Buttheyarenot;theyclearlydonothavethesamestatusasthereareanycookiesleft.5Ourproposal,asitstands,seemsthereforetoruleouttoomuch. Thesourceoftheproblemseemstolieinthefactthatreflexivesareinterpretedasvariablesboundtosomesuitableantecedentintheirlocalsyntacticenvironment;andboundvariablesarefunctional/logicalitems,ifanythingis.Hencetheyshouldnotbetargetedbyreplacementormodulation.ThisiswhytheGajewski/DelPinalapproachappearstofailinitsjobofsiftingungrammaticalvs.grammaticaltrivialitiesinthecasesathand.Howcanwemodifysuchanapproach,soastoretainitsmainmeritsanditsprincipledcharacter?InthepresentsectionIaddressthisproblem. Thesyntaxandsemanticsofreflexivesandcomparativesisacomplexmatter.Inwhatfollows,Iwillbasemyproposalonreflexives,sketchingasmuchoftheir
5ThisversionoftheboundvariablesproblemwaspointedouttomebyRichardLarson,atatalkIgaveatSUNYStonybrookin2015.Gajewski(2002,Section4.2)isclearlyawareofit.AlsoDelPinal(2017)discussesitexplicitly(cf.fn.9below).
![Page 12: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/12.jpg)
12
grammarasneeded,inanasuncontroversialmanneraspossible.Iwillthenindicatehowmyproposalextendstocomparatives. LetussaythatreflexivesaregovernedbyPrincipleAofChomsky’s(1981)bindingtheoryaccordingtowhichtheymustbeboundtoanantecedentintheirlocalsyntacticenvironment,whichforourpurposescansimplybethesmallestsentencecontainingthereflexives.PrincipleAisasyntacticaxiomwithsemanticconsequences.Bindingisachieved,letusassumefortheargument’ssake,byassigningscopetotheantecedent(viaQuantifierRaising–QR–ortheequivalent),whichcreatesanabstractthatbindsthereflexivesasillustratedinwhatfollows.(23) a.Johnis(not)himselfi b.Johni[tiis(not)himselfi]èJohnlxi[xiis(not)himselfi] wheretiisthetraceleftbehindby(stringvacuous)raisingofthesubject.Followingawidespreadpractice,weanalyzetheindexonJohnasanabstractorthatcreatesthederived(reflexive)predicatein(23b).6PrincipleAensuresthattheindexonthesubjectin(23b)bethesameastheanaphoricindexonthereflexive.Weassumethatthecopulawindsupbeinginterpretedhereasidentity. Wemustbrieflyconsiderafurtheroptionatthisjuncture.Onewayofaddressingourproblemandmaking(23a)nonG-trivialmightconsistoftreatingthecopulaasacontentitemandmodulateitbymappingidentityintosomeotherrelationthatdoesn’tyieldacontradiction.Ithinkthismoveisimplausible.First,thecopulais,syntacticallyspeaking,aprototypicalfunctionalitem.Inmanylanguages,copularsentenceslike(23a)donotexploitanyovertitemliketheverbtobebutareassembledbymereconcatenation.ThisistrueevenofsomeEnglishpredicative‘smallclause’constructionssuchas:(24) a.Iconsider[SCJohnagoodplayer/hisownworstenemy/finallyhimselfagain] b.Iregard[SCJohnasmybestfriend/hisoldselfagain]Sentenceslike(24)seemtoyieldmanifestationsofthesameproblemas(23),inwaysthatdoesnotrelyonanovertcopularverb.Moreover,injustaboutanylanguagetheitemusedincopularsentences,whenattested,typicallydoublesup,justasinEnglish,asamereexpressionoftenseandaspect(asinJohnwasinthebathtub/agoodfriend).Thisbehaviorissymptomaticoffunctionalelements.Infact,themostdetailedattemptatanalyzingthesemanticsideofcopularconstructionnamelyPartee(1986),analyzescopularconstructionsasinvolvinga(restricted)setof‘logical’type-shiftingdevices.Second,semanticcriteriasuchasidentityunderdomainpermutationsputidentityamongthelogicalconstants.7Third,treatingidentityasacontentitemwouldnothelpwiththecaseofcomparatives,wheretheidentityrelationassuchisnotinvolved.Wemightaswelllookforasolutionthat
6Seee.g.HeimandKratzer(1998)7SeeMcFarlane(2017)andreferencestherein–esp.McGee(1996)andSher(2003).Cf.alsothediscussionSection4below.
![Page 13: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/13.jpg)
13
coversalsoreflexivesincomparatives,asthediagnosisofthesourceoftheproblem(namely,thepresenceofboundvariables)seemstobethesame. Iftinkeringwiththecopulaisofnohelp,theonlyotherwaytogoistomodulatetheboundvariableitself,e.g.asfollows:(25) a.Johnis(not)himselfi b.Johnlxi[xiis(not)g(himselfi)] c.Johnisnotthepersonheusuallyis/thewayheusuallyis d.lxi¬[xi=ix[xbehaves(inw)mostsimilarlytohowxiusuallybehaves]]In(25c)Iexemplifytypicalwaysofunderstandingsentenceslike(25a).Themodulationofvariableshastohaveanintensionalcharacter,whichspelledoutinafullfledgedcompositionalsystemwouldyield,forexample,somethinglike(25d).8Noticethattheoutcomeofmodulationofthereflexivein(24d)isacontingentproperty,whichinturnensuresthatsentence(25a)won’tcomeoutasG-trivial.9 Whilethismodificationperhapsyieldsanempiricallyadequatesolutiontotheproblemofboundvariables,itseemstogiveuponthefunctional/logicalvs.contentdistinctionthatwehavebeenrelyingonsofar,andappearstobelessprincipledthantheoriginal.Notonlycontent/nonlogicalitemsneedtobemodulated.Variablesneedtobemodulatedaswell:(26) NewdefinitionofG-triviality. a.Modulation:optionallyinsertamodulationfunctiongonanycontentword orvariable(»boundpronounortrace). b.AsentenceisG-trivialiffiscomesoutastrue/falseonanymodulation. c.AsentenceisL-trivialiffitcomesoutastrue/falseforthedefault
8ThesemanticmetalanguageIhaveinmindisGallin’s(1975)TY2withovertworldvariables.Iamassumingthatpredicatescarryaworldvariables(e.g.thatisred=red(that)(w)abbreviatedasredw(that));modulationofindividualvariablesmapsindividuals(oftypee)intoindividualconceptsoftype<s,e>,e.g.inthecaseathand,itpicks“theindividualthatbehavesmostsimilarlytohowJohnusuallybehaves”.Suchconceptwindsupbeingappliedintheendtotheactualworld.Inotherwords,thepropositionassociatedwith(24)issomethinglike:
(i) lw¬[j=ix[xbehaves(inw)mostsimilarlytohowjusuallybehaves]Whilemakingthisfullyexplicitmayrequiresomework,Itrustthatthechiefideaisclearenoughforourpresentpurposes.9DelPinal(2017)proposestoaddresstheproblemwithreflexivesbymodulatingthepropertyobtainedbyabstractingoverthereflexivepronounroughlyalongthefollowinglines:
(i) Johng(lx[x=x])Thistantamountstotreatingidentityasacontentitem,whichIhavearguedagainst.,compoundedwiththecomplicationthatthepropertyin(i)isasingletonproperty,thatineveryworldmapseachindividualintothepropertyofbeingidenticaltoitself.Itisunclearhowmodulationofsuchpropertycanyieldtherightresults.
![Page 14: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/14.jpg)
14
valueofmodulations(asidentity).Underthisnewdefinition,sentenceslikethosein(21)comeout,correctly,asnonG-trivial.Ithinkthat,inspiteofappearances,thisproposalinfactretainsoriginalinspirationandprincipledcharacterofGajewski’sandDelPinal’s.Variablesconstitutestand-insforcontentexpressions.Ifyouthinkinmodeltheoretictermsandconsideracanonicalintensionalmodel<U,W,F>,withUasetofindividuals,WasetofworldsandFaninterpretationfunction,theitemsthatcanbemodulatedarethevaluesofFandoftheassignmentstovariables.Logicalconstants,ontheotherhand,havevaluesthatremainconstantacrossmodelsandcannotbemodulated.Obviously,Iamnotgivinghereacharacterizationofthelogical/nonlogicaldivide.Iamsimplyadoptingastandardsemanticpracticeandpointingouthowmydefinitionofmodulatedlogicalformsfallswithrespecttoit.Hereistheguidingprincipleofourproposal:(27) Thereferentialpointsofalogicalform(/LFtree),namelythenonlogical
constantsandvariables,maybemodulated.Thesolutionjustsketcheddoesextendtocomparativeconstructions.Theroughlogicalformofasentencelike(21b),repeatedinsimplifiedformin(28a),isasin(28b):(28) a.Johnwasmoreeloquentthanhimself b.Johnilti[tiwasMORE(eloquent)thanhimselfi] =lxi[MORE(eloquent)(xi)(xi)](j) whereforanyu,MORE(eloquent)(u)isthepropertyofbeingmoreeloquent
thanudefinedasfollows:u’hasthepropertyofbeingmoreeloquentthanuiffthereissomedegreedsuchthatu’isatleastd-eloquentanduisnot.
Theanalysissketchedin(28)reliesonadegreesemanticsforcomparatives,suchtheoneexploredinKennedy(2007)andmuchrelatedwork.Accordingtoit,adjectivescorrespondtorelationsbetweenindividualsanddegrees:Johnisd-talliffJohn’sheightisatleastd.Thecomparativemorpheme,thus,sayssomethingabouttherespectivemaximaldegreestowhichtwoindividualshaveacertaingradableproperty.Thelogicalformin(28b)canbemodulatedjustlikeothersentencesinvolvingreflexives:(29) a.Johni[tiwasMORE(eloquent)thang(himselfi)] =lxi[MORE(eloquent)(g(xi))(xi)](j) b.Johnis(today)moreeloquentthanthedegreetowhichheusuallyis c.JohnismoreeloquentthantheindividualwhoseeloquenceismostsimilartoJohn’susualone.
![Page 15: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/15.jpg)
15
Thesentencesin(29b-c)arepossibleinformalrenderingsoftheeffectofmodulatingthereflexivepronounin(29a).10 ThetreatmentofvariablesIamproposingbearsanonaccidentalconnection,Ithink,totheissueofdere(anddese)belief.Ihavealreadyhintedatthisinconnectionwithexample(19)above,Johnbelievesthatcloudsarealive.Letmeoutlinetheconnectionmorefullyherethroughasimpleexample.Consider:(30) Johnbelievesthathisbrotherisnothisbrother. a.Johnbelievesthathisactualbrotherisinfactanimpostortryingtosteal John’sinheritance. b.Johnisatthedentist.Whilesittingonhisdentist’soperatingchair,hespots amanactingasanaidtothemaindoctor.Heformsthebeliefthatthat personisthenewassistanttohisdentist,withoutrecognizingthatheisin factJohn’sownbrother.Whileknowingthathisbrotherisadentisttoo,John doesn’tthinkthatthepersonassistinghisdentistishisbrother.Onitsnoncontradictoryinterpretation,sentence(30)maybeusedtoreportaderebeliefofJohn’stowardshisactualbrother,compatiblewith(andappropriateto)avarietyofscenarios,suchasforexamplethosein(30a-b).Theissueofderebeliefisofcourseintricate.Oneimportanttradition11addressestheproblembyappealingtoconceptsthroughwhichtherelevantresisaccessedbytheattitudeholder.Abeliefisdereaboutanindividualuwheneverureliablyinducesaconceptaboutuinthebeliefholdera,whichidentifiesuforaina’sbeliefstate.Suchconceptsforexample(30)mightbe,say,themanthatwantstoshareJohn’sinheritanceforcontext(30a)andthemanJohnisseeingfor(30b).CharlowandSharvit(2014)haveproposedanimplementationofthiskindofapproachtodereinwhichlogicalformsforderebeliefsemploy‘conceptgenerators’areinsertedinthesyntacticspotoftheresanddrivepragmaticallythepropositionalcontentofthebelief.Inthecaseof(30a),forexample,wemightgoforalogicalformlike(31a),withtheg-functionspelledoutasin(31b-c):(31) a."w[BELj,w0(w)®¬brotherw(g(ix.brotherw0(j)(x))(w)] wherej=Johnandbrotherw0(j)(x)=xisbrotherofjinw0.10Than-complementsaresometimesclausal.Forexample,(i)isbestanalyzedas(ii)
(i) JohnistallerthanBillis(ii) JohnistallerthanBillistall
Ifallthan-complementswereclausal,modulationofvariableswouldnotbenecessary,foronemightachievetheintendedresultsbymodulatingthetwooccurrencesoftheadjectivetallin(ii).However,itisnotclearwhetherthisisrightforsentencesinvolvingreflexives,forthesourceisungrammatical:
(iii) *JohnistallerthanhimselfistallMoreover,therearelanguagesthathavebeenarguedtolackclausalcomparativessuchasFijan(cf.Pearson2010,andforageneraloverviewofvariationincomparisonconstructions,Beck2011).11SeeinparticularQuine(1956),Kaplan(1968),CresswellandvonStechow(1982).
![Page 16: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/16.jpg)
16
b.LetubeJohn’sbrotherintheactualworld.Then: g(u)=lw.ix[xwantstosharej’sinheritanceinw] c."w[BELJohn,w0(w)®¬brother(ix[xisthepersonwhowantstosharej’s inheritance)(w)] =JohnbelievesofthepersonwhowantstoshareJohn’sinheritance (namelyhisactualbrother)thatheisnothisbrother.Asinthecaseofvariables,thedefinitedescriptionhisbrother(evaluatedintheactualworld)ismodulatedviaaconceptthatmediatesbetweenJohn,theattitudeholder,andthereshisbeliefisabout.TheuseofmodulationforindividualexpressionsproposedherecanthusbeviewedasanextensionofCharlowandSharvit’sproposalforthesemanticsofderebeliefingeneral. Insum,thepresentproposalisthatlogicalforms(whichdrivethecompositionalinterpretationofsentences)canandsometimesmustbemodulatedbytheinsertionof‘replacementfunctions’intheirreferentialpoints.ReferentialpointsofanLFarethecontentwords(‘nonlogical’)andvariables(whosevaluesrangeonthedenotationsofcontentwords).Modulationisnecessaryforavarietyofreasons,mostprominentlytomakesenseofourbelief-statesandtoresolvecontradictionsinacommunicativelyeffectiveway,explainingwhysometimescontradictionscanbeusefulcommunicationtools.Thereare,however,sentencesthatcannotberescuedinthisway.TheirLFsturnouttobecontradictoryforanymodulation.Thesesentencesareuselessandcanberegardedasonparwithsyntacticallyill-formedsentences.TheoutcomeisanarguablygeneralandprincipledproposalinwhichacharacterizationofG-trivialitystemsfromtheindependentneedofre-interpretingcertainsentencesincontext. 4.GrammarvsLogic. Ourapproachtomodulationisrootedinthedistinctionbetweenfunctionand
contentwords,wherefunctionwordssubsumelogicalwords.Inthepresentsection,Igooversomeissuesinthecharacterizationofthisdichotomy. Asiswellknown,functionalitemsdonothaveaclearcut,absolutecharacterization,butthereanumberofsyntacticandsemanticcriteriathatarereliablyrelevanttothedistinctionfunctionalvs.content.Startingatthesyntacticendofthings,hereisa(partial)listoffunctionalcategoriesandmorphemes:(32) TypicalFunctionalcategoriesandsubcategories a.Determiners,Quantifiers,Classifiers,Complementizers,Coordinations, Negation,Comparative/Superlativemarkers,TenseandAspectmarkers (»‘Auxiliaries’),Modals,FocusandTopicmarkers,DiscourseParticles,… b.Gender,Number,Case,(in)definitenessmarkers,Verb/Noun-class markers,pronouns,wh-elements,…Thosein(31a)aretypicalfunctionalcategoriesandsubcategories;thosein(31b)involve‘features’specifictocertainitems(likepronouns)oractiveinagreement
![Page 17: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/17.jpg)
17
patterns.Toseewhattheitemsin(31)haveincommon,considerthemainbasicordersinthelanguagesoftheworld:(33) a.SOV b.SVO c.VSO [whereS=subject,O=Object,V=verb) Typicalfunctionalelementssuchasthosein(32)tendoccurattheedgesofthemainclausalconstituentsin(33),relativetothebasicwordorderalanguagechooses,wheretheymayberealizedasboundmorphemesorasautonomouswords(‘free’morphemes).Forexample,theexpressionofPAST-nessoccursattheperipheryoftheVPandcanberealizedasaboundmorphemelov-ed(through,say,incorporationoftheVintothePASTmorpheme)orasanindependentmorphemeasinhaswalked(withsemanticdifferencesbetweenthetwooptions).Toillustratefurther,coordinators(and,or)connecttypicallyclausesandhencetheytendtooccurattheedgeofclausalstructuresratherthaninthemiddleofthem.Similarly,DiscourseParticles(e.g.,Germandoch,Greekµen,etc.)havetodowithsignalingdiscoursejuncturesrelatedtotopicality,backgrounding,etc.andareoftenplacedatornearmajorconstituentboundaries.12ThewayfunctionalcategoriesareconceptualizedwithincurrentgenerativeapproachesisasaseriesofheadsattheedgeofNPsorVPs,formingtheso-called‘functionalspines’or‘extendedprojections’ofthelatter.Iprovideanexamplein(34):(34) DP D #P the PL NP boysThestructurein(34)representsthedefinitepluralDPtheboys,where#P(‘numberphrase’)isthelayerdrivinginformationaboutnumber.DPand#Pare(partof)thefunctionalspineorextendedprojectionofNP. Functionallayersaredrawnfromahypothesizeduniversalinventoryandaresubjectto(limited)parametricvariations.Variationhastodowithwhichmembersoftheuniversalinventoryareexploitedbyagivenlanguage,mattersofwordorder,andwhetheralanguageadmitsphonologicallyunrealizeditemsofspecificsorts.Thespecialcharacteroffunctionalitemsdeterminesafurtherseriesofassociatedproperties,schematically:(35) a.Frequency:Highestfrequencyinanylanguageisassociatedwith functionality.Themostfrequent50lemmasinEnglishincludenomorethan 3or4contentwords(withsayinthelead,atthe19thposition).13 b.Historicalchange:contentwordschangeconstantlywithoutaffecting12Thisroughcharacterizationneedstotakeintoaccountthefactthatconstituentscanbemovedfromtheirbaseposition.13Cf.www.wordfrequency.info.
![Page 18: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/18.jpg)
18
languageidentity(whencethecharacterization‘openclass’);changesinthe functionallayersinvolvegrammaticalchange,whichmayaffecttheidentity ofalanguage(whencethelabel‘closedclass’). c.Selectiveimpairments:functionwordsareoftenselectivelyimpairedina varietyoflanguagepathologies,likeagrammatismor‘nonfluent’aphasia. (Seee.g.CaramazzaandHillis(1989),FriedmanandGrodzinsky(1997) amongmanyothers.)Thepropertiesin(35)arefairlyself-explanatory,andperhapsunsurprisinggiventhenatureoffunctionwords.Noticethattheyare‘one-way’generalizations,i.e.conditionals,notbiconditionals:ifxishighfrequencyxhasahighprobabilityofbeingafunctionword.Butthereareofcourserelativelylowfrequencyfunctionwords(e.g.,shall,ought). Thesyntacticcharacterizationoffunctionalitemsjustreviewedreliesprimarilyonthe‘slots’theyoccupywithintheclause.Thismaintraitunavoidablycomeswithaclusterofsemanticproperties.BuildingonvonFintel(1995),Iwillbrieflydiscussherefoursuchproperties,namely:(36) a.Havinghightypes. b.Being‘inferencebased’ c.Beingsubjecttocrosslinguisticallywidespread,sometimesuniversal constraints. d.Beingpermutationinvariant.Startingwith(36a),nounsandverbstypicallyexpressfirstorderpropertiesandrelationsthatsubdividedomainsofdiscourseintoclassesandrelateindividuals,events,etc.tooneanother.Intypetheoreticterms,thisisconceptualizedbypositingabasictypeofindividualse,andrelationsoverindividualsoftype<e,t>,or<e,<e,t>>,etc.Functionalexpressionsfindtheirnaturalconceptualizationathighertypes.Forexample,determinerscanbeviewedasassociatedwithhigherorderrelationsoftype<<e,t>,<<e,t>,t>>betweensetsorclasses.Similarly,onecanthinkofpropositionsascarvinganabstractspaceofpossibilities(say,asetofworlds)intosubregionsandpropositionalconnectivescanberepresentedashigherorderfunctionsonsetsofworlds.Thepropertyin(36b)iseasytograspbuthardtodefine.Thebasicideaisthatwhilethemeaningofeverykindofexpressionisultimatelyrootedinitsentailments,presuppositionsandimplicatures,contentwordsarealsocausallylinkedtofairlytangibleandlocalizedregularitiesinourenvironment(themeaningof“cats”iscausallylinkedtocats,thatof“run”torunningevents,etc.).Incontrastwiththis,themeaningsofevery,or,onlyorevenarewaymoreabstractandonlycharacterizableintermsoftheinferencepatternstheygiveriseto.Suchpatternsaremoreoversubjecttopossiblyuniversalstructuralconstraints.Forexample,onlyandevenarealways‘alternativesensitive’:they
![Page 19: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/19.jpg)
19
requireidentifyingaclassofalternativeswithrespecttowhichtheirprejacentisevaluated;determinersareconservative,14etc. Turningnexttopermutationinvariance,i.e.theideathelogicalwordremainconstantacrossone-onemappingsofthedomainontoitself,15thereislittledoubtthatitisapowerfulcriterionthatidentifiesanaturalsemanticclass.Itemswiththispropertysystematicallyfallwithinthefunctionalsegmentofthelexicon,toanextentthatsimplycan’tbeaccidental.Expressionswithanarguablylogicalmeaningthatbehavelikecontentwordsareexceedinglyfew.Theyincludeverbslikedeny,orexist,nounslikemajority,adjectiveslikemereorformer.Butnotethatthesewordsareallmorphologicallyderived(e.g.,major+ity)andtheytypicallyundergoadriftthatgivesthemsomenonlogicalcontent(e.g.existsisnotjust‘beingthevalueofaboundvariable’,but-driftsintosomethinglike‘havingphysicalexistence”).Sotheclaimthatpermutationinvariantfunctionsareexpressedwithinthefunctionallayerofsyntaxis,Ithink,bornout. Arefunctionwordslimitedtoexpressingpermutationinvariantitems?I’dsayno.Theclearestcaseisperhapsthatofgenderfeatures(andmoregenerallyclassagreementmarkers).16Grammaticalgendersystemscanbequitecomplex;theytypicallycodesomeanthropologicallysalienttraitandextendit,oftenarbitrarily,inordertopartitionorclassifythedomainofindividuals.SinceatleastCooper(1983),17thesemanticsideoffeatureinformationistreatedpresuppositionally:(37)a.i. ||fem||=lxe:female(xe).xe ||male||=lxe:male(xe).xe ii. ||ragazz-a||=lxe:fem(xe).youngadult(xe) iii. ||ragazz-o||=lxe:male(xe).youngadult(xe) b."x[female(x)®¬male(x)]Thefunctionsin(37ai)arerestrictedidentitymaps,definedonlyforfemaleormaleindividuals,respectively;in(37a.ii)youseehowsuchfunctionscanbeusedtorestrictthedenotationofthewordsforgirlvs.boyinalanguagewithgrammaticalgender.Thepredicatein(37a.ii),forexample,isdefinedonlyforfemaleindividuals;wheneverdefined,itistrueofyoungadultsandfalseofnonyoungadults.Useoffeaturesofthissortinducesdisjointnessconstraintssuchas(37b),whichareamongthemostcommonacrosslanguages.Thisseemstorequireanextensionofwhatcountsas‘logical’toconstraintsthatdefine‘subcategories’ofvariouscontentwords.Sagi(2014)providesaminterestinggeneralwayofextendingthenotionoflogicalitybyusingconstraintsofthissorttorelativizepermutationstosubcategoriesofcontentwords.
14AdeterminerDisconservativeiffforanyAandB,D(A)(B)=D(A)(AÇB).SeeBarwiseandCooper(1981).15See,e.g.,McGee(1996),Sher(2003),and,foranoverview,MacFarlane(2017).16Currentterminologyusestheterm‘f-features’forelementsofthissort.17Foramorerecentversionofthepresuppositionaltreatmentofgrammaticalfeatures,cf.,eg.Sudo(2012)andreferencestherein.
![Page 20: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/20.jpg)
20
Sopermutationinvarianceseemstohaveacoreandaperiphery.Inveryroughterms,thecoreisconstitutedbytheitemscharacterizedbysomestrictdefinitionofpermutationinvariance(say,bijectionsamongdomainsofequalcardinality).Theperipherytakesintoaccountmorespecificstructuralconstraintsonnaturalsemanticcategorieslikemodals,tense,massvs.count,etc.allthewaytofairlyidiosyncraticfeature-basedconstraints.Jointly,coreandperipherydeterminewhatmightbeviewedasauniversalnaturallogic,specifictoHomoSapiens. Summing up, the syntax of function words systematically differs from that of content words. These syntactic differences correlate with semantic ones. Permutation invariant items in the strict sense are systematically treated as functional by syntax and are obvious candidates for Universal Grammar membership. The remnant of the functional vocabulary seems to be constituted by a broader class of mostly inference based operations, relations, etc. subject to cross-linguistically stable structural constraints. Ourstartingpointhasbeenthatcertainformsoflinguisticdevianceappeartobebestmadesenseofintermsoflogicaltruthorfalsehoodratherthanintermsofwell-formedness.Assumesomebackgroundlogicalframework,saythetypedlambdacalculus,andenrichittoatheoryNatLogbysomesetofaxioms/structuralconstraintsonmodals,eventstructures,countablevs.uncountableentities,etc.ImaginenextusingNatLogtospecifythesemanticsofanaturallanguage,sayEnglish,intheusualsenseofasystematic,compositionalmappingfromthestructuresconstructedbythegrammarofEnglishintoformulaeorstatementsofNatLog.SomeEnglishsentenceswillbelogicallyfalse/truerelativetotheirLogicalFormsinterpretedinNatLog.AndasubsetoftheNatLog-logicallyfalse/truesentenceswillbeperceivedbyspeakersofEnglishas‘notinthelanguage’,onaparwithsyntacticallydeviantstructures.Wehavecalledsuchsentences‘G(rammatically)-trivial’.Ourproblemwastodetermine:(i)whichsentenceswithintheL(ogically)-determinedonesareG-trivial,and(ii)why.WehaveproposedamodificationofGajewski’sandDelPinal’sproposalthataddressestheseissues. Bethatasitmay,thesearchforthecomponentsofthefunctional/logicallexiconandthewaysinwhichtheymayvary,whilestilldaunting,isdefinitelynolongermerespeculation,butawelldefinedandexcitingresearchprogram,whichdeliversconstantlynewresults.Andthediscoverythatformsof‘ungrammaticality’areinfactduetologicalinference(ratherthantosyntacticill-formedness)isactuallygamechanging,andshowsinverytangibleandfruitfulwayshowinterconnectedgrammarandlogicare.Acknowledgments.Thankstotheparticipantsatthe2018MünichWorkshopTheSemanticConceptionofLogicandtoananonymousrefereeforhelpfulcomments.IamalsogratefultoGuillermoDelPinal,DannyFox,andJonGajewskiforextensivediscussionsofthesemattersovertime.REFERENCESAbrusan,M.(2014)Weakislandsemantics,OxfordUniversityPress,OxfordUK.Barwise,J.andR.Cooper(1981)“Generalizedquantifiersandnatural
Language,”LinguisticsandPhilosophy,4:159–219.
![Page 21: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/21.jpg)
21
Beck,S.(2011)"ComparisonConstructions,"inC.Maienborn,K.vonHeusingerandP.Portner(eds.):Semantics:AnInternationalHandbookofNaturalLanguageMeaning,DeGruyterMouton,Berlin,1341-1389.
Caramazza,A.andA.E.Hillis(1989)”TheDisruptionofSentenceProduction:SomeDissociations,”BrainandLanguage36:625-650
Charlow,S.andY.Sharvit(2014)“Bound‘dere’PronuonsandtheLFofAttitudeReports,”SemanticsandPragmaticshttp://dx.doi.org/10.3765/sp.7.3.
Chierchia,G.(2013)LogicinGrammar,OxfordUniversityPress,Oxford,UK.Chomsky,N.(1981)LecturesonGovernmentandBinding:ThePisaLectures,Foris,
Dordrecht.Cooper,R.1983.QuantificationandSemanticTheory.Dordrecht:Reidel.Cresswell,.andA.vonStechow(1982)“DereBeliefGeneralized,”LinguisticsandPhilosophy,5:503–535.
DelPinal,G.(2017)“TheLogicalityofLanguage:ANewTakeonTriviality,“Ungrammaticality”,andLogicalForm,”Noûs,https://doi.org/10.1111/nous.12235
Dowty,D.(1979)WordMeaninginMontagueGrammar,D.Reidel,Dordrecht.Friedman,N.andY.Grodzinsky(1997)“TenseandAgreementinAgrammatic
Production,”BrainandLanguage,56:397-425.Gallin,D.(1975)IntensionalandHigherOrderModalLogic,NorthHolland,
Amsterdam. Gajewski,J.(2002)L-analyticityandnaturallanguage.Manuscript,Universityof
Connecticut.Gajewski,J.(2008)“Moreonquantifiersincomparativeclauses,”inT.Friedman&S.
Ito(eds.)SemanticsandLinguisticTheoryXVIII,vol.18,340–357.Ithaca,NY.Halle,M.andA.Marantz(1993)"DistributedMorphologyandthePiecesof
Inflection",inK.HaleandJ.Keyser(eds)TheViewfromBuilding20,MITPress,Cambridge,MA.
Heim,I.andA.Kratzer(1998)SemanticsinGenerativeGrammar,Blackwell,Oxford,UK.
Kaplan,D.(1968)“Quantifyingin,”Synthese19:178–214.Kennedy,C.(2007)“VaguenessandGrammar:Thesemanticsofrelativeand
absolutegradableadjective,”LinguisticsandPhilosophy,30:1-45.Kratzer,A.(1981)“TheNotionalCategoryofModality”inH.J.Eikmeyer
andH.Rieser(eds)Words,WorldsandContexts,WalterdeGruyter,Berlin.Lahiri,U.(1998)“FocusandNegativePolarityinHindi”,NaturalLanguage
Semantics,6.1:57-123.McGee,V.(1996)“LogicalOperations,”JournalofPhilosophicalLogic,16:423-443.MacFarlane,J.(2017)"LogicalConstants,"inE.Zalta(ed)TheStanford
EncyclopediaofPhilosophy(Winter2017Edition),<https://plato.stanford.edu/archives/win2017/entries/logical-constants/>.
Partee,B.(1986)“Nounphraseinterpretationandtype-shiftingprinciples,”InJ.Groenendijk,D.deJonghandM.Stokhof(eds.)StudiesinDiscourse
RepresentationTheoryandtheTheoryofGeneralizedQuantifiers,Foris,Dordrecht.
Pearson,H.(2010)“HowtodoComparisoninaLanguagewithoutDegrees:a
![Page 22: On Being Trivial: Grammar vs. Logic - Scholars at Harvard · 2020. 5. 19. · 1 On Being Trivial: Grammar vs. Logic Gennaro Chierchia Harvard University January 2019; Revised September](https://reader033.fdocuments.in/reader033/viewer/2022051914/6005be6c3dc6b77d630bd44d/html5/thumbnails/22.jpg)
22
SemanticsfortheComparativeinFijan”,inSchmitt,V.&Zobel,S.(eds.)ProceedingsofSinnundBedeutung14,356-372.
Quine,W.V(1956)“QuantifiersandPropositionalAttitudes,”TheJournalofPhilosophy53:177–187.
Sagi,G.(2014)“FormalityinLogic:FromLogicalTermstoSemanticConstraints,”LogiqueetAnalyse,227:259-276
Sher,G.(2003)“ACharacterizationofLogicalConstantsIsPossible,”Theoria:RevistadeTeoria,HistoriayFundamendosdelaCiencia,18:189-197.
Sudo,Y.(2012)OntheSemanticsofPhi-FeaturesonPronouns,Ph.D.Dissertation,MIT.
vonFintel,K.(1993)“Exceptiveconstructions,”NaturalLanguageSemantics,1:123–148.
vonFintel,K.(1995)“Theformalsemanticsofgrammaticalization.InProceedingsofNELS25,vol.2,175–189.