ebook img

THE COORDINATE STRUCTURE CONSTRAINT AS A DISCOURSE-ORIENTED PRINCIPLE ... PDF

34 Pages·2015·0.35 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview THE COORDINATE STRUCTURE CONSTRAINT AS A DISCOURSE-ORIENTED PRINCIPLE ...

THECOORDINATESTRUCTURECONSTRAINTASADISCOURSE-ORIENTED PRINCIPLE:FURTHEREVIDENCEFROMJAPANESEANDKOREAN YusukeKubota JungmeeLee UniversityofTsukubaand SeoulNationalUniversityof TheOhioStateUniversity ScienceandTechnology coordinate structure constraint We reexamine the status of the (CSC; Ross 1967) by drawingonevidencefromJapaneseandKorean.ContrarytothestandardviewthattheCSCisa syntacticconstraint,theempiricalpatternsfromthetwolanguagesshowthatitshouldinsteadbe viewedasapragmaticprinciple.Weproposeapragmaticanalysisbybuildingonandextendinga previousproposalbyKehler(2002).ExaminingtheJapaneseandKoreandataturnsouttobevital inthecomparisonofthesyntacticandpragmaticapproaches,sincethesyntacticdifferencesbe- tweentherelevantconstructionsinthetwolanguagesandtheircounterpartsinEnglishcrucially distinguishthepredictionsofthetwoapproaches.* Keywords:coordinatestructureconstraint,Japanese,Korean,coordination,subordination,island constraints,discourserelation Introduction coordinatestruc- 1. .SinceRoss1967,thestandardviewaboutthe tureconstraint (CSC)hasbeenthatitisasyntacticconstraint.Detailsvary,butall major syntactic theories (both derivational and nonderivational) have adopted some variantofRoss’s(1967)originalformulationin1,1togetherwithits‘across-the-board’ (ATB)exception—whichsaysthat1canbeviolatedifextractionoccursfromallofthe conjuncts—toexplaincontrastsbetweensentenceslikethosein2.2 (1) In a coordinate structure, no conjunct may be moved nor may any element containedinaconjunctbemovedoutofthatconjunct.(Ross1967:89) (2) a. #Thisisthemagazinethat[Johnbought__]and[Maryboughtthebook]. b. #Thisisthemagazinethat[Johnbought__]and[Marydidn’tbuy__]. However,exceptionstotheCSCwerealreadynotedbyRosshimself:ifsomesemantic relation other than ‘pure logical conjunction’ obtains between the two conjoined clauses, extraction from a single conjunct is allowed, as exemplified by the following well-knownexamples. WewouldliketothankBobLevineforhiscontinuedencouragementandnumerousinputs(ofvarious sorts)onthisworkthroughoutitslonggestation.Earlierversionsofthisarticlewerepresentedatthe15thIn- ternationalConferenceonHead-DrivenPhraseStructureGrammar(Keihanna,Japan,2008)andthe83rd annualmeetingoftheLinguisticSocietyofAmerica(SanFrancisco,2009);wewouldliketothankthepar- * ticipantsoftheseconferencesforfeedback.Thanksarealsoduetothemembersofthesyntax-semanticsdis- cussiongroup‘Synners’attheOSULinguisticsDepartment.Wewouldalsoliketothankfouranonymous Languagerefereesfortheircomments,whichhelpedimproveboththecontentandpresentationofthearticle greatly.Lastbutnotleast,wewouldliketothankLanguageeditorGregCarlsonforhisverythoughtfuledi- torialguidanceandvarioususefulremarksonthiswork.ThefirstauthorwassupportedbytheJapanSociety for the Promotion of Science (JSPS; PD 2010–2013 and Postdoctoral Fellowship for Research Abroad 2013–2014)whenheworkedontherevisionsofthisarticle,andwouldliketothanktheJSPSforitsfinancial support. 1Inthisarticle,wedealonlywiththeso-called‘elementconstraint’oftheCSC(i.e.whatthesecondpart from of1says,whichprohibitsextractionofelements conjuncts).TheotherpartoftheCSC,the‘conjunct of constraint’(i.e.thefirstpartof1,prohibitingextraction conjunctsthemselves),isadifferentissue.Yatabe (2003)discussessomescramblingdatainJapanesethatsuggestthatJapanesenominalcoordinationdoesnot obeytheconjunctconstraint. 2Throughoutthearticle,wemark‘CSCviolation’exampleswith#insteadof*inconformitywithour claimthattheirunacceptabilityisduetopragmaticinfelicity.?isusedformildlydegradedbutbasicallyac- ceptableexamples.Theunderscore(__)formissingmaterialispurelyforexpositorypurposeswithoutany theoreticalimplications. 642 PrintedwiththepermissionofYusukeKubota&JungmeeLee.©2015. Thecoordinatestructureconstraintasadiscourse-orientedprinciple 643 (3) Here’sthewhiskeywhich[Iwenttothestore]and[bought__]. (Ross1967) (4) That’s the stuff that the guys in the Caucasus [drink __ ] and [live to be a hundred]. (Schmerling1972) Such examples have typically been dismissed as ‘apparent’exceptions involving a special, ‘asymmetrical’use of and that is exempt from the CSC. In fact, the very first suchattemptwasmadebyRoss(1967)himself,accordingtowhomtheunderlyingsen- tencefortherelativeclausein3isderivedfromtheparaphraseIwenttothestoretobuy thewhiskey,andthusdoesnotinvolvetruecoordination.Thisaccountwasthoroughly criticizedbySchmerling(1972),whonotedthatthetwosentencesarenotsynonymous. Despitethefactthatlaterauthors,likeRoss,havefailedtoprovideanyclearevidence thatthisso-calledasymmetricalcoordinationwassyntacticallydifferentfrom‘canoni- cal’coordination (as noted by Kehler (2002), unlike ordinary subordinate clauses, the secondconjunctcanneverprecedethefirstconjunctinthistypeofcoordination),one seesvariousincarnationsofthisideathroughouttheliterature(see,forexample,Steed- man2012forthemostrecentproposalalongtheselines). Amongvariousislandconstraints,thenatureoftheCSCisespeciallyimportantthe- oretically.Notefirstofallthat(despiteblatantexceptionssuchas3and4)theCSChas oftenbeentakentobeoneofthemorerobustcasesofislandconstraints(forotheris- landconstraintssuchastheadjunctconstraintandthecomplexNPconstraint,various sorts of amelioration effects have been discussed in the literature; cf. Kluender 1992, 1998,Hofmeister&Sag2010).Second,theCSChassometimesbeenusedtocrucially distinguishbetweenpredictionsofdifferenttheoreticalapproaches;forexample,oneof generalizedphrasestructuregrammar theclaimedbigsuccessesof (asopposed toderivationaltheories,inwhichthepreciseformulationoftheCSChasalwaysbeena hugeproblem;Sag2000)wasthatit‘predicted’theCSCanditsATBexceptionviathe slash-feature treatment of extraction (a similar claim survives to date in a related the- ory—cf.Steedman&Baldridge2011).Butifthephenomenonisnotsyntactictobegin with,suchanargumentsimplylosesitsforce. Infact,apragmaticaccountoftheCSCpatterns3hasbeenproposedbyauthorssuch as Lakoff (1986), Deane (1991), and Kehler (2002), based on data like 3 and 4, and buildingonrelatedworkbySchmerling(1972,1975)andLevinandPrince(1986)(the gapping latterisastudyon ,butitdiscussespropertiesof‘asymmetrical’coordination relevant for the CSC). Details vary, but the gist of their proposals is essentially the same:thekeydifferencebetweenexampleslike2andthoselike3and4isnotsyntac- tic,butispurelypragmatic.Theformerexpresssituationsoreventsthatareparallelto eachother,whereasinthelatter,thetwoconjunctsdonothavesuchparallelstatuses.In 3,theconjoinedclausesstandinakindof‘stage-setting’relationthatoneserveswith respecttotheother,andin4,thereisacausalrelationbetweenthetwoevents.Extrac- tion is associated with the pragmatic effect of treating the extracted element as being prominent (or designated).Then, extraction from a single conjunct leads to unaccept- ability in examples like 2 since it destroys the parallel between the two conjuncts. By contrast,in3and4,sincenosuchparallelrelationobtainsbetweenthetwoconjunctsto beginwith,non-ATBextractiondoesnotinvokeanyanomaly. Itisworthnotingherethat,thoughmuchless oftendiscussed,thesamepatternex- tendstootherconjunctionwordssuchasorandbut(BobLevine,p.c.). 3Inwhatfollows,weusetheterms‘CSCpatterns’and‘CSCeffects’torefertothedescriptivegeneral- syntactic izationofthesortexemplifiedby2and‘CSC’torefertoa constraintthatismeanttocapturethis generalization. 644 LANGUAGE,VOLUME91,NUMBER3(2015) (5) a. #WhichcitywillRobingotoSeattlethisweekorvisit__nextweek? b. #WhodidJohnvotefor__butMaryvotedforObama? (6) a. [He] regards the limitless abundance of language as its most important property,onethatanytheoryoflanguagemustaccountfor__orbedis- carded. (Campbell1982:183) b. WhatdidJohngotothestorebutforgettobuy__? Hereagain,aparallelrelationblocksnon-ATBextraction,butnonparallelrelationsdo not. These examples are especially troublesome for the type of accounts, alluded to above,thatattempttoexplainawaytheanomalyoftheclassicalRoss/Schmerlingdata in3and4ascasesofasymmetricalcoordination,attributingtheexceptionalCSCvio- lation to some special property of nonlogical and. For example, Sag and colleagues (1985) have a section (§4.1) devoted to asymmetrical coordination, in which the au- thorsdiscussexampleslike3and4andproposetotreatthemviaspecialtypesofphrase structurerules(distinctfromtheirmoregeneralcoordinationrule)specificallyapplica- ble to the conjunction word and. (See also Goldsmith 1985 for a similar idea but one castintermsofsyntacticreanalysis.)Similarly,Steedman(2012:95)continueswiththis lineofanalysis,suggestingtotreatsuchexamples‘byassigningadditionalindependent categoriestoand,supportingextractionfromleftandrightconjunctsrespectively’.But onallsuchaccounts,itremainscompletelymysteriouswhythesamesensitivitytodis- courserelationsextendstootherconjunctionmarkers. WeargueinsteadforapurelypragmaticaccountoftheCSCpatterns,building(pri- marily)onKehler’s(2002)work(indoingso,weintroducesomerefinementsnecessary forextendinghisapproachtotheJapaneseandKoreandatadiscussedbelow).Although thepreviousproposalsbyLakoff(1986),Deane(1991),andKehler(2002)(towhichthe morerecentChaves2007,2012canbeadded)offeranintriguingalternativetothecur- rentlydominantsyntactictreatmentoftheCSC,wethinkthattheystilldonotshowthe superiority of the former over the latter convincingly enough.The data considered by these authors are limited (mostly) to English, and this limitation in empirical domain standsinthewayofevaluatingthetruestrengthofthispragmaticalternative.4Infact,we believethatthisisatleastoneofthereasonsthattheroutinelyinvokedrecoursetothe lexicalidiosyncrasyofasymmetricalandnotedaboveisstillprevalentintheliterature. Notealsothatconductingextensivecrosslinguisticworkisespeciallyimportantforthe discussion of the CSC, since one of the central motivations for syntactic island con- straintsispreciselytoexplaintherecurrentpatternsof(un)grammaticalityacrosstypo- logicallydiverselanguages.Suchafact,theargumentgoes,canbeexplainedonlyifitis amanifestationoftheuniversalgrammar.Thus,anadvocateofapragmaticalternativeis responsibleforshowingthatitscrosslinguisticapplicabilityatleastmatchesthatofthe morepopularsyntacticview. Butmerelybroadeningtherangeoflanguagesconsidereddoesnotnecessarilylead to theoretically relevant findings. In order to evaluate the relative advantages of the pragmaticapproachvis-à-visthesyntacticapproach,itiscrucialtoexaminedatafrom 4NaandHuck’s(1992)workonEnglishandKoreanCSCdataisanotableexception,andisanimportant precursorforthepresentwork.AspointedoutbyKehler(2002),however,theirproposedconditionisappli- cabletoasymmetricalcoordinationonlyandneedstobesupplementedwithaseparateprinciplefordealing withsymmetricalcoordination(infact,itseemsthatNaandHucktaketheCSCasasyntacticconstrainttobe responsibleforit).Moreover,theydonottakeintoaccountthedifferencebetweenEnglishextractioncon- structionsandextraction-likeconstructionsinKoreanandsimplytreatthelatterasacaseofextraction.Thus, eventhoughthesetofdatathatNaandHuck(1992)discussoverlapswithourstoacertainextent,theulti- mateconclusionswedrawarequitedifferentfromtheirs. Thecoordinatestructureconstraintasadiscourse-orientedprinciple 645 languages that are typologically different from English, in particular, languages in whichthekindsofmeaningthataretypicallyexpressedbycoordinationinEnglishare expressedbyconstructionsthatdonotexhibitthecharacteristicpropertiesofcoordina- tion and in which the kinds of semantic/pragmatic functions that English extraction servesarebornebyconstructionsthatdonotdisplaythecanonicalpropertiesofextrac- tion.JapaneseandKorean,thetwolanguagesthatwetakeupinthisarticle,turnoutto be ideal candidates in this respect, since they satisfy both of these criteria. Coordina- tion-likeconstructionsinJapaneseandKorean(i.e.onesthatexpressthesamekindsof semantic relations as English coordination) are morphosyntactically subordination ratherthancoordination(importantly,weexcludetheCSCpatternsfromthesetofdi- agnostics for coordination, since its inclusion would beg the question). Displacement constructionsintheselanguages(ofwhichweconsiderboththe‘overtmovement’type andthe‘covertmovement’type)alsohavedifferentsyntacticpropertiesfromEnglish extraction constructions, suggesting that (at least on the null hypothesis) they should notbetreatedintermsofsyntacticmovement. ThesedifferencesbetweenJapaneseandKoreanontheonehandandEnglishonthe othermakeitpossibletocomparethesyntacticandpragmaticapproacheswithrespectto amuchbroaderrangeofdatathanhasbeendoneinthepreviousliterature.And,moreim- portantly,theyenableustocomparethepredictionsofthetwoapproachesmoresharply. On the syntactic approach (unless with ad hoc extensions), one would expect that the extraction coordinate CSC effects are found only in from structures, whereas the pragmaticapproachpredictsthatthesamepatternswillbefoundinconstructionsthat sharesimilarsemanticandpragmaticproperties.Topreviewtheconclusion,wewillsee thatbasicallythesamepatternsarefoundinallofthethreelanguages,despitethesyn- tacticdissimilaritiesintherelevantconstructionsinvolved.Moreover,theJapaneseand Koreandataweconsiderinthisarticledisplayadditionalcomplexities(duetotheexis- tenceofthetensedvs.tenselessvariantsofthecoordination-likeconstructioninKorean andtheso-called‘gapless’variantsofdisplacementconstructionsinthetwolanguages) ofakindnotfoundintheEnglishdatadiscussedinthepreviousliterature,whichulti- matelyprovidefurthersupportforthepragmaticapproach.Theseconsiderationsleadus totheconclusionthatthepragmaticapproachtotheCSCeffectsnotonlyisviable,butis alsomoreexplanatorythanthemorefamiliarsyntacticapproach. Before moving on, we would like to briefly comment on two broader issues. First, datalike3and4thatcastdoubtonthesyntacticnatureoftheCSChavebeentakenby someauthors—mostnotablyLakoff(1986)—tounderminetheautonomyofsyntaxand toargueforan‘integrated’modelofgrammarinwhichsyntactic,semantic,andprag- maticinformationisrepresentedsimultaneously(cf.Chaves2007,2012forarecentex- plicitproposalalongtheselines).Wethinkthatthisistoostrongaconclusiontodraw (seeKehler2002:§5.6foradiscussiononarelatedpoint).TheCSCpatterns(including theJapaneseandKoreanfactsdiscussedbelow)canbeaccountedforadequatelyinany theoryasapragmaticeffectaslongassomereasonablyarticulatesyntax-pragmaticsin- terfaceisprovided.Second,anoteisinorderinrelationtorecentprocessing-oriented accounts of other island phenomena (e.g. Kluender 1992, 1998, Hofmeister & Sag 2010).Atagenerallevel,ourfindingsarecompatiblewiththeseaccounts,andwebe- lievethatthediscourserelationsthatplaycrucialrolesinouraccountshouldultimately begivensomepsycholinguisticbasisinasufficientlyelaboratetheoryofdiscoursepro- cessing(alongthelines,forexample,thattheunacceptabilityofCSCviolationexam- ples like 2a stems from the fact that the impossibility of establishing an appropriate discourserelationincursanextraprocessingcosttoanalreadycostlyprocessofparsing 646 LANGUAGE,VOLUME91,NUMBER3(2015) structuresinvolvingdisplacement).However,giventhelackofsuchamodelatpresent, wehavechosentocouchouranalysisinthetermsofmoretraditionallinguisticseman- ticsandpragmatics. Thearticleisstructuredasfollows.First,therelevantCSCpatternsinJapaneseand Koreanarepresented(§2),andthentheprevioussyntacticapproachestotheCSCinthe twolanguagesarereviewed(§3).OurpragmaticanalysisoftheCSCdataispresented in§4,followedbyaconclusion(§5). The CSC patterns in japanese and korean 2. . In this section, we examine the CSC patterns in Japanese and Korean. However, since both the extraction-like con- structions and the coordination-like constructions in the two languages have syntactic properties that are distinct from those of the corresponding English constructions, we reviewthepropertiesoftheseconstructionsbeforeexaminingtheCSCpatterns. Displacement and coordination-like constructions in japanese and 2.1. korean. Displacementconstructions .ForthecounterpartofEnglishextraction,wecon- sider three overt ‘displacement’constructions in Japanese and Korean (relativization, wh topicalization, and cleft) and one covert ‘displacement’construction ( -questions). Theterm‘displacement(construction)’isusedherejustasatheoreticallyneutralterm forgroupingtogethertheseconstructionsbasedonacertainsyntacticsimilarity(while avoidingthetheoreticallyloadedterms‘extraction’,‘movement’,or‘filler-gapdepen- dencies’):inalloftheovertdisplacementconstructions,somematerialinthesentence appears in a ‘displaced’position rather than in its canonical position within the sen- wh wh tence.And we call the -question constructions in these languages (both being - in-situ)‘covertdisplacement’,justforterminologicalconsistency. WhilethethreeovertdisplacementconstructionsinJapaneseandKoreanareappar- entlysimilartoextractionconstructionsinEnglish,theirsyntacticpropertiesareactu- allyquitedifferentfromtypicalextractionconstructions,asdiscussedbyKuno(1973), Yoon(1993),andMatsumoto(1997),amongothers.First,unlikeextraction,thesedis- placement constructions do not obey syntactic island constraints (except possibly for theCSC).Second,theseconstructionshavetheso-calledgaplessvariants,thatis,sen- tencesthatdonotcontainasyntacticgapcorrespondingtothe‘displaced’element.The wh covert movement of -questions is also different from extraction in that it does not obeyislandconstraints(seeAppendixAfortherelevantdata;gaplessexamplescanbe wh constructedonlyforovertdisplacementandarethusnotprovidedfor -questions). Asforislandsensitivity,ithasoftenbeenpointedoutthatrelativeclausesandtopi- calizationinJapaneseandKoreandonotobeysyntacticislandconstraints(Kuno1973, Yoon1993,Matsumoto1997).Examples7and8showthatrelativizationoutofcom- plexNPs(theinternalbracketedpartin7isarelativeclausethatmodifiesthefollowing noun)andoutofadjunctsispossibleinbothJapanese((a)-examples)andKorean((b)- examples).AnalogousexampleswithtopicalizationandcleftaregiveninAppendixA.5 (7) a. [[__ki-te i-ru] yoohuku-ga kitanai] sinsi teprog npst nom npst [[__wear- - clothes- dirty. gentleman ‘thegentlemansuchthattheclothesthatheiswearingaredirty’(=‘the gentlemanwhoseclothesaredirty’) 5Thefollowingglossesareusedinthisarticle:acc:accusative,cop:copula,dat:dative,decl:declara- i gen hon ko neg nmlz nom npst tive, :-i, :genitive, :honorific, :-ko, :negation, :nominalizer, :nominative, : pass pst pl prog q rel te top nonpast, :passive, :past, :plural, :progressive, :interrogative, :relativizer, :-te, : topic. Thecoordinatestructureconstraintasadiscourse-orientedprinciple 647 b. [[__ip-koiss-nun] yangpok-itelep-un] sinsa prog rel nom rel [[__wear- - suit- be.dirty- gentleman ‘thegentlemansuchthatthesuitthatheiswearingisdirty’ (8) a. [[__sin-da ato] mina-gakanasin-da]zyosei pst nom pst [[__die- afterall- miss- woman ‘thewomanthatallmissedaftershedied’ b. [[__cwuk-unhwu-ey]motwu-kakuliuweha-n]yeca rel nom rel [[__die- after all- miss- woman ‘thewomanthatallmissedaftershedied’ An even more radical difference from typical extraction constructions is that these displacement constructions have so-called gapless variants where there is no missing position in the sentence corresponding to the syntactically ‘displaced’element, as ex- emplified by the following relative clauses ((a)-examples are Japanese, (b)-examples Korean;again,forexampleswithotherconstructions,seeAppendixA). (9) a. atama-ga yoku-naru hon nom npst head- good-become. book ‘abookthatonebecomessmartbyreading’ b. meli-ka cohaci-nun chayk nom rel head- become.good- book ‘abookthatonebecomessmartbyreading’ (10) a. watasi-de-saeinemuri-deki-nakat-taeiga cop neg pst I- -even doze.off-can- - movie ‘the movie that (was too exciting that) even I could not doze off (by watching)’ b. na-to col swu-ka ep-ess-ten yenghwa nom pst-rel I-evendoze.offpossibility- not.exist- movie ‘the movie that (was too exciting that) even I could not doze off (by watching)’ (11) a. yakedosi-ta kizu pst get.burned- scar ‘aburnscar’(ascarcausedbygettingburned) b. John-i pwul-eytey-n sangche nom rel John- fire-at get.burned- scar ‘John’sburnscar’ Matsumoto(1997)studiesthepropertiesofthesegaplessrelativesindetailandshows that they involve a link between the event denoted by the relative clause and another eventinvolvingthereferentoftheheadnoun.Thelattereventisnotovertlyexpressed bythesentencebutisinvokedbythelexicalcontentoftheheadnounorthemainpred- icateoftherelativeclause.Thetwoeventsarerelatedviaadiscourserelationinstanti- ating the cause-effect category (broadly construed and including the nonprototypical ‘violationofexpectation’typerelation).Forexample,in9,theheadnounhon‘book’in- vokes the event of reading a book via its lexical information (in some theories of the generative lexicon lexicon such as Pustejovsky’s (1995) (GL), such information (‘telicrole’intheGLterminology)ispartofthemeaningoftheword).Fromthegen- eralworldknowledgethatreadingabookisanintellectualactivity,acause-effectrela- tion is established between this implicit event and the event of one’s becoming smart (i.e.thedenotationoftherelativeclause).In10,therelationbetweentherelativeclause eventandtheimpliciteventofwatchingthemovie(again,invokedviathetelicroleof 648 LANGUAGE,VOLUME91,NUMBER3(2015) eiga‘movie’)isnotthatofcause-effectbut,rather,violationofexpectedconsequence. Thisexamplereliesontheknowledge(sharedbytheinterlocutors)thatthespeakeris notveryinterestedinmoviesingeneralandoftenfallsasleepbywatchingthem.With thiscontextualsupport,thecontentoftherelativeclausebecomesrelevantasadescrip- tionofthemoviesinceitcanbeinterpretedasindirectlyendorsingitsunexpectedhigh quality.Finally,in11,theimpliciteventinvolvingtheheadnounisinvoked(primarily) bythemainpredicateoftherelativeclauseratherthantheheadnoun.Yakedosuru‘get burned’is a creation verb and entails a result state of a burn scar being created. Note alsoherethatthecause-effectrelationbetweentherelativeclauseeventandtheimplicit eventistheoppositefrom9.Here,therelativeclauseeventisthecauseandtheimplicit eventistheresult. The establishment of a discourse relation between the relative clause event and the implicit event contributed by the head noun depends heavily on shared knowledge among interlocutors. The following example from Matsumoto 1997 illustrates this pointnicelyforJapanese. (12) #atama-ga yoku-naru kuruma nom #head- good-becomecar intended:‘acarthatonebecomessmartbydriving’ Example 12 is minimally different from the grammatical 9 in its head noun (kuruma ‘car’),butisunacceptablebecause,givenordinaryworldknowledge,itisdifficulttoes- tablish a causal relation between driving a car and becoming smart.We come back to thispointofpragmaticsensitivitywhenwediscusstheCSCeffectsinexamplesinvolv- inggaplessrelativesattheendof§2.2.Notsurprisingly,wewillseetherethatthefelic- ityof‘CSCviolation’examplesinvolvinggaplessrelativesisalsocruciallyaffectedby thissensitivitytopragmaticinformationandcommonknowledgeestablishedamongin- terlocutorsthatcharacterizesthisconstruction. The island insensitivity and the existence of gapless variants suggest that Japanese and Korean displacement constructions do not involve syntactic filler-gap dependen- cies. Given this evidence, the null hypothesis is that the relationship between the dis- placed element and the rest of the sentence is instead established pragmatically, as argued by Matsumoto (1997) andYoon (1993).6Thus, followingYoon (1993), we as- sumeastructureroughlyalongthelinesofthatinFigure1forrelativeclausesinJapa- head-drivenphrasestructuregrammar neseandKorean.Weadopt (HPSG)since itenablesustostatetherelevantassumptionsaboutthesyntax-semanticsinterfaceex- plicitly and perspicuously. However, it should be noted that the pragmatic account of theCSCpatternsweformulatebelowisentirelytheory-neutralanddoesnotdependon anyspecificaspectsofthesyntaxweassume(also,hereandinwhatfollows,wewrite lambdatermsforsemantictranslationsforbetterreadability,insteadofadoptingsitua- minimal recursion semantics tion semantics or —the two more commonly adopted semantic frameworks in HPSG—but nothing hinges on this choice; (≡ λPλQλx. P(x)∧Q(x))standsforParteeandRooth’s(1983)generalizedconjunction). subcat Here, crucially, the relative clause is a fully saturated sentence ( 〈〉) with an slash empty feature(thus,nofiller-gaplinkageisinvolved).Aunaryruleprojectsthis (cid:4) sentencetoaprenominalmodifier(forKorean,inwhichthereisanovertrelativizer,we 6Thisdoesnotexcludeapossibilityofanalyzingtheseconstructionsintermsofafiller-gapdependency mechanisminthesyntax.Buttheburdenofproofisonthosewhowouldmaintainthislatterassumptionto explainthedifferenceamongtherelevantconstructionsacrosslanguagesastotheirisland(in)sensitivity. Thecoordinatestructureconstraintasadiscourse-orientedprinciple 649 N(cid:2)HEAD 5(cid:3) CONT 4 SUBCAT (cid:5)(cid:6)  N(cid:2)HEAD 5 (cid:3) SLASH {} 3 CONT 2 λx.book(x) S  MOD 3    CONT 4(cid:7) 2 (cid:2)λx.propertyC(x,1)(cid:8) ‘movie’ SUBCAT (cid:5)(cid:6)  S SLASH {} CONT 1¬♦doze(I) ‘Icouldn’tdozeoff’ Figure 1.StructureofrelativeclausesinJapaneseandKorean. could dispense with this unary projection by encoding the relevant information in the lexicalentryfortherelativizer,asisdoneinYoon1993),whichcombineswiththehead head-adjunct schema noun via the standard of HPSG. The semantics (specified in cont the feature)ofthisprenominalmodifier(whichisinheritedtothemothernodevia semantic principle the of HPSG; Pollard & Sag 1994) says that the meaning of the < whole relativized nominal is obtained by conjoining the denotation of the head noun with the property λx.property (x, ¬◊doze(I)). Here, property is a context-sensitive C C operatorthatreturnssomeappropriatepropertythatholdsofitsfirstargumentbasedon thepropositiondenotedbyitssecondargument(thiscorrespondstothenotionof‘prag- maticallycoherentproperty’thatwediscussin§4.2below).Forexample,inthecaseof Fig.1,viaworldknowledge,property identifiesitsfirstargumentasanentityxsuch C bywatching thatthespeakercouldnotdozeoff x(wheretherelation‘bywatching’is pragmatically invoked to establish an explicit link between x and the content of the propositionthatisthesecondargumentofproperty ). C Notethat,correspondingtotheabsenceofsyntacticfiller-gaplinkage,thereisnose- manticvariablebindinginvolvedbetweenxandsomeargumentpositionintherelative clause.Thus,therelationisestablishedentirelypragmatically,alongthelinesdiscussed above.Weassume,followingYoon(1993),thatevenincaseswhereitapparentlylooks likethereisvariablebinding,asinthecaseof‘ordinary’relativeclausessuchas13,the linkisestablishedpurelypragmatically. (13) Japanese [John-ga __kat-ta] hon nom pst [John- __buy- book ‘thebookthatJohnbought’ JustliketheabovegaplessrelativeinFig.1,here,too,therelativeclauseisafullysat- uratedsentence(notethatbothJapaneseandKoreanallowfornullpronouns).Compo- sitionally,weobtainatranslationalongthelinesofλx.property (x,buy(z)(j))forthe C relativeclause.Thevariablezistechnicallyafreevariable,buthereitispragmatically identified with x and we obtain an interpretation equivalent to λx.buy(x)(j), since the mostnaturalwayofestablishingacoherentrelationbetweenthedenotationsoftherel- ativeclause(inwhichtheidentityofthedirectobjectargumentisleftunspecified)and 650 LANGUAGE,VOLUME91,NUMBER3(2015) theheadnounistosimplyidentifythismissingargumentintherelativeclausewiththe headnoun. Thisway,eventhoughthereisnofiller-gaplinkage(andthereforenoexplicitseman- ticvariablebinding)betweenanargumentpositionintherelativeclauseandthehead nouninthesyntax(andsemantics)ofrelativeclauses,alinkcanbeestablishedbetween therelativeclauseandtheheadnoun.Importantly,thisanalysisenablesauniformtreat- mentofboth‘ordinary’relativeclausesandthegaplessvariants.Weassumethatasim- ilaranalysisextendstootherdisplacementconstructionsinthetwolanguages. Coordination-likeconstructions .InJapaneseandKorean,whatappeartocor- respond to English verbal (and sentential) coordination are expressed by marking the nonfinalconjunctswiththemorphemes-teor-i(Japanese)and-ko(Korean),asinex- amples14and15. (14) Japanese [John-ga zassi-o kat(*-ta)-te/-i][Mary-ga hon-o kat*(-ta)]. nom acc te i nom acc pst [John- magazine- buy-PST- /- Mary- book- buy- ‘JohnboughtthemagazineandMarydidn’tbuythebook.’ (15)Korean [John-un capci-lul sa(-ss)(*-ta)-ko] [Mary-nunchayk-ul top acc pst ko top acc [John- magazine- buy- -DECL- Mary- book- sa-ss*(-ta)]. pst decl buy- - ‘JohnboughtthemagazineandMaryboughtthebook.’ Onemightbetemptedtoregardtheseconstructionsasinstancesofcoordination,given thesemanticsimilaritytoEnglishcoordination.Butitisimportanttokeepinmindthat syntactic whatisatissueisthe propertiesoftheconstructionsathand,sinceitissuch syntacticpropertiesthatcruciallydeterminewhethertheCSC—asitisformulatedasa syntactic constraint—should be applicable to them. In fact, when we turn to purely (morpho)syntacticproperties,itturnsoutthattheevidenceistothecontrary.7Ascanbe seenin14and15,thefinitenessmarker(namely,thetensemarker-tainJapaneseand themoodmarker-tainKorean)cannotappearinthenonfinalclausesintheseconstruc- tions.Moreover,asshownin16and17,thesenonfinalclauseswithnonfiniteendings cannotstandaloneasfull-fledgedsentences. (16) Japanese *John-ga zassi-o kat-te/-i. nom acc te/-i *John- magazine- buy- intended:‘Johnboughtthemagazine.’ 7Thedistinctionbetweencoordinationandsubordinationisahairyissue(seeHaspelmath2007forsome relevantdiscussion),wheresomeauthorstakesemanticpropertiestobe(atleastpartly)relevantforthedis- tinction.Giventhelogicofthesyntacticapproach,however,itshouldbeclearthatwhatoneshouldbelook- ingforhere—soastoavoidcircularityofargument—ispurelysyntacticevidence.Noteonceagaininthis connectionRoss’s(1967)failuretoprovideindependentcriteriaforthenotionofcoordinationinEnglishwith respecttothe‘asymmetrical’coordinationcases,whichhasbeencriticizedinmuchlaterwork(Schmerling 1975,Lakoff1986,Kehler2002). Movabilityandrepeatabilityaresometimesusedascriteriaforcoordination,butwedonotrelyonthese testshere.ThemovabilitytestisinapplicableforJapaneseandKoreansincetheselanguagesareverb-final andthusdonotallowsubordinateclausestoappeartotherightofthemainclause.Asforrepeatability,itis syntactic notclearwhetherthistestreallypertainstothe differencebetweencoordinationandsubordination (notethatcertainsubordinateclausesliketemporaladverbialclausescanberepeatedmultipletimesaslong astheydonotinducesemanticanomaly). Thecoordinatestructureconstraintasadiscourse-orientedprinciple 651 (17) Korean *John-un capci-lul sa(-ss)-ko. top acc pst ko *John- magazine- buy- - intended:‘Johnboughtthemagazine.’ This shows that the -te/-i/-ko-marked clauses are (morphosyntactically) subordinate clauses. Additional(thoughindirect)evidenceforthe(morphosyntactically)subordinatesta- tusoftheseconstructionscomesfromthefactthatverbalformswiththesamemorpho- logicalendingsappearinothersubordinateenvironments,whicharenormallyanalyzed as‘complexpredicate’constructions.Forexample,Japanesehasmorphologicallycom- plex verbs such as V-te miru (‘try V-ing’), V-te morau (benefactive), V-i hazimeru (‘begin to V’), and V-i sugiru (‘over-V’), where the first verb marked in the -te or -i formissyntacticallyandsemanticallyanargumentofthesecondverb.Themorpholog- icalformsofthefirstcomponentofthesecomplexverbsareidenticaltothe-teand-i formsastheyappearinthecoordination-likeconstructionsexemplifiedin14and15.In both constructions, -te undergoes the same voicing assimilation morphophonological process in certain environments, and the -i form is the same conjugation form called Renyookei (lit. ‘predicate-connecting form’, the verbal conjugation used in environ- mentsinwhichapredicatefollows).Koreansimilarlyhascomplexpredicatesinvolving -kosuchasV-kosiph‘wanttoV’,V-koiss‘beintheprocessofV-ing’,andV-komal ‘end up V-ing’. While the existence of morphologically identical forms with subordi- nate functions can provide only an indirect piece of evidence for the subordination view,suchafactistotallyunexpectedonthecoordinationview(theEnglishconjunc- tionwordsandandordonothavesuchhomophonoussubordinationmarkers). Thus, despite the semantic similarity to English coordination, independent mor- phosyntactic evidence suggests that the Japanese and Korean -te/-i/-ko constructions aresubordinationconstructions.Justtobeexplicit,herewespelloutoursyntacticas- sumptions about English coordination and the Japanese and Korean coordination-like constructions,continuingtousetheHPSGnotation.Followingthestandardassumption inHPSG(Pollard&Sag1994,Sagetal.2003,Chaves2012),weassumethatcoordi- nationinEnglishislicensedbyanonheadedschemaoftheforminFigure2. CAT 1  S CONT P∧Q   CONTEXT R(P,Q) S(cid:2)CAT 1(cid:3) Conj S(cid:2)CAT 1(cid:3) CONT P CONT Q Figure 2.CoordinationinEnglish. Thesyntacticcategoriesoftheconjunctsneedtomatch,8andtruth-conditionally,the cont semantics( )issimplytheBooleanconjunctionofthemeaningsoftheconjuncts. context The feature(whichstorespragmaticinformationinHPSG)additionallyspec- ifiesthatthereisacontextuallydeterminedrelationRthatholdsbetweenthemeanings ofthetwoclauses.WeassumethatRisinstantiatedasoneofthediscourserelationsin 8Thisisofcourseasimplification,giventhewell-knownexamplesofunlikecategorycoordination(Saget al.1985,Bayer1996).

Description:
incarnations of this idea throughout the literature (see, for example, Steed- respect to the other, and in 4, there is a causal relation between the two .. seen in 14 and 15, the finiteness marker (namely, the tense marker -ta in like temporal adverbial clauses can be repeated multiple times as
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.