XeroxPARCProceedingsoftheLFG01ConferenceUniversityofHongKong,HongKongMiriamButtandTracyHollowayKing(Editors)
2001
CSLIPublications
http://csli-publications.stanford.edu/
Abstract
Weprovideanewdefinitionofthelinearprominenceconstraintsbetweenpronounsandoperators(wh-wordsandquantifiers)whichcorrectlyrulesoutexamplesthatviolateweakcrossover.Previousanalysesofweakcrossoverreliedonthepresenceofatraceintheextractionsiteofawh-question;incontrast,ouranalysisenablesatracelessaccountofexamplespreviouslycitedinsupportoftraces.Sincenootherincontrovertibleevidencefortraceshasbeenputforward,ouraccountallowsare-turntothetracelessandthereforemoreconstrainedaccountoflong-distancebind-inginLFGproposedbyKaplanandZaenen(1989).1
Long-distancedependenciesinLFG
Traceswereintroducedintransformationalgrammar(Chomsky,1973)inordertogiveaphrasalaccountforawiderangeoflong-distancedependencyphenomena:along-distancedependencybetweenadisplacedelementinawh-questionorrel-ativeclausewasassumedtoarisebymovementofthedisplacedelement,whichleavesbehinda‘trace’initsoriginalposition.TheoriginalLFGtreatmentoflongdistancedependencies(KaplanandBresnan,1982)wasbasedonanadaptationofthetransformational/phrase-structurescheme.Therelationbetweenthedisplacedwh-phraseanditswithin-clausefunctionwasdefinedintermsofarelationbetweenthewh-wordandanemptyc-structureconstituent,atrace,withintheclause.Inexample(1),therelationbetweentheNPnodedominatingwhoandtheNPnodedominatingitstrace,representedby,ensuresthatthewh-phrasewhoisboththeFOCUSandtheOBJofthesentence:
(1)
NPCP
CPRED‘greetSUBJ,OBJ’WhoCIPFOCUSPRED‘who’didNPISUBJPRED‘Bill’BillVVPNPOBJtgreetSubsequently,however,KaplanandZaenen(1989)proposedthatconstraintsonlongdistancedependenciesarebeststatedinfunctionalandnotphrasalterms.Assuch,functionaluncertaintyoffersamoreaccurateanddirectcharacterizationof
thelongdistancedependency.Aruleliketheoneinexample(2)establishestworolesfortheNPdaughterofCP:itistheFOCUS,anditplaysagrammaticalroledefinedbythefunctionaluncertaintypathCOMPOBJ:
(2)CP
NP(FOCUS)=(COMP*OBJ)=
C=
Inexample(3),thepathconsistingsimplyofOBJischosen,andtheFOCUSf-structurealsoappearsastheOBJofgreet:(3)WhodidBillgreet?
PRED‘greetSUBJ,OBJ’
FOCUS
PRED‘who’
SUBJ
PRED‘Bill’
OBJ
Withthisaccount,longdistancedependenciesnolongerprovideargumentsthattrace-likedevicesarenecessaryinthetheoryofgrammar.Andinfact,atracelesstheoryispreferredfortworeasons:itismoreadequatedescriptively,anditofferstheoreticaladvantages.Weoutlinethesereasonsbelowbeforediscussingexamplesinvolvingweakcrossoverandhowatracelesstheorycanaccountforthem.
Atracelesstheoryisdescriptivelymoreadequate.Thatis,itcandescribedatathatthetracefultheorycannot,asarguedindetailbyKaplanandZaenen(1989).KaplanandZaenen(1989)citeevidencefromwanderingadverbsinIcelandic,cross-categorialdependenciesinEnglish,andcross-conjunctrelativizationconstraintsinJapanesethatcanbeaccountedfornaturallyinatraceless,butnotatraceful,anal-ysis.Morerecently,Sag(1998)hasprovidedadditionalevidenceagainsttracesinhisanalysisoftheconjunctconstraintandoffloatingquantifiersandadverbs.
Forseveralreasons,atracelesstheoryisalsotheoreticallypreferred.Thatis,evenifthetracelessandtracefultheoriesaccountedforexactlythesamedata,thetracelesstheoryispreferable.First,thetracelesstheoryislessredundantinthatthereisonlyasinglewayofdealingwithlongdistancedependencies;itisagen-eralscientificprinciplethatlessredundanttheoriesarepreferredtomoreredundantones.Second,itismorerestrictiveinthatithasfewerpowerfulmechanismsandismoretractablemathematically.Third,itavoidsspuriouslyambiguousanalysesforfillerswithout‘canonical’phrase-structurepositions.Forexample,inextrac-tionofobliquesandadverbsthereismorethanonepossibleextractionsite,evenin
arelativelyfixedwordorderlanguagelikeEnglish,asshownin(4).(4)a.Obliqueextr.:AboutwhatdidJohntalk
toMary
?
b.Adverbextr.:HowoftendidJohntalk
toMaryaboutapples?
Finally,withthetracelesstheory,thelistener’sperceptualproblemissimpler:thelistenermustguessonlythefunctionofthedisplacedelement,notbothafunctionandaposition.Thus,therearebothempiricalandtheoreticalreasonstopreferthetracelessaccount.2
Weakcrossover:Acounterexample?
Despitetheappealofatracelessanalysisoflongdistancedependencies,thereisevidencethatseemstosuggestthattracescannotbeeliminated:weakcrossover(Postal,1971;Wasow,1979)seemstoindicatethatthelinearpositionencodedbyatracedoesaffectgrammaticality.Thesentencein(5)exemplifiesaweakcrossoverviolation:
(5)*Whodidhismothergreet?
(cannotmean:Whosemothergreetedhim?)
Thename‘crossover’comesfromthetransformationalanalysisofwh-questionfor-mation:inacrossoverviolation,thewh-phrase‘crossesover’acoreferentialpro-nounwhenitismovedtothebeginningofasentence.Inlateranalyses,theviolationhasbeenreformulatedintermsofthepositionofthetrace:acrossoverviolationensueswhen,asin(5),acoreferentialpronounprecedesthetrace.Weconcentratehereonweakcrossoverviolations,thoseinwhichthepronounprecedesbutdoesnotc-commandthetrace.
Crossovereffectsarefoundinoperatorbinding,inparticular,bindingbyWH-operatorsandquantifiers(Reinhart,1983):(6)a.*Whodidhismothergreet?
b.*Hismothergreetedeveryone.
Ourmainfocushereisonthequestionofthenecessityoftraces,andourdiscussionwillcenteraroundoperatorbindinginwh-questions.Wemakeonlysomebriefre-marksbelowaboutquantifierbindinginGerman.
SomerecentLFGaccountsofoperatorbindingandweakcrossoverproposeareturntoatheoryoflongdistancedependencywithtraces(Bresnan,1994,2001;
Berman,2000);thesetheoriesadopttheviewthattherelationbetweenthepronounandatraceofthedisplacedwh-phraseistheimportantoneforcharacterizingweakcrossoverviolations.Thesetheoriesproposearepresentationliketheoneinex-ample(7)foraquestionlikeWhodidhismothergreet?,andruleoutcoreferencebetweenthepronounhisandtheoperatorwhobyreferencetotherelationbetweenthepositionofthepronounandthepositionofthetraceofwho:(7)
NPCP
CPRED‘greetSUBJ,OBJ’WhoCIPFOCUSPRED‘who’didDetNPNISUBJSPECPRED‘pro’VPVNPPRED‘mother’hismotherOBJgreettInthefollowing,weprovideareformulationoftheconstraintsontherelationbe-tweenapronounandanoperatorthatallowanaccountoftheseexampleswithouttraces.3
Prominence
Bresnan(1994,1995,2001)proposesthatoperatorbindingrelationsareconstrainedby(atleast)twodimensionsofprominence:syntacticprominencedependsonthefunctionalrankoftheoperatorandthepronounitbinds,andlinearprominencede-pendsonthelinearorderbetweenanoperatorandapronounthatitbinds.WefirstdiscussBresnan’sdefinitionsandthenprovidearevisedsetofdefinitionsofthesetwoprominencedimensions.3.1
Syntacticandlinearprominence
Bresnan(2001)definesoperatorbindingrequirementsinthefollowingway:
(8)SyntacticProminence(Aunitcontaining)thepronounmaynotbehigher
than(aunitcontaining)theoperatoronthegrammaticalfunctionhier-archy:
SUBJOBJCOMP...
LinearProminenceThepronounmustnotf-precedetheoperator.Bresnanfurtherproposesthatlanguagescanvaryinwhichoftheseconstraintsap-ply:somelanguagesimposebothkindsofprominenceconstraints,somerequireonlyoneortheother,somerequireadisjunctionofthetwo.
WebelievethatBresnan’sbasicinsightiscorrect:prominencerequirementsbe-tweenanoperatorandapronounaremultifaceted,anddifferentlanguagescanim-posedifferentkindsofprominencerequirementsbetweenanoperatorandthepro-nounitbinds.Itturnsout,however,thatasimpleredefinitionofthelinearpromi-nenceconditionmakesmanyofthesamepredictionsasthedefinitionsabove,in-cludingforthecrucialcasesofweakcrossover,withoutassumingtraces.3.2
Rethinkinglinearprominence
Theintuitionbehindourreformulationisthatlinearprecedencerequirementsbe-tweenanoperatorandapronounaredeterminedbytheovertmaterialwhichindi-catesthesyntacticroleofthedisplacedphrase.ThisintuitionfollowstheproposalofSag(1998),thoughitdiffersindetail.
Atheorywithtracesaccountsfortheunacceptabilityofanexamplelike*WhodidSuetalkabouthismotherto?byreferencetotherelationbetweenthepronounhisandthetraceintheobjectpositionofthepronounto,asin(9a).Incontrast,ourtheory(likeSag’s)referstotherelationbetweenthepositionofthepronounandthepositionofthestrandedprepositionto.Thesameresultsobtain,butwithoutpositingatrace,asin(9b).
(9)a.Traces:*WhodidSuetalkabouthismotherto?
b.Notraces:*WhodidSuetalkabouthismother[to]?Wefirstdefinethenotionofcoargumenttoencompasstheargumentsaswellastheadjunctsofasinglepredicate:(10)Coarguments:
theargumentsandadjunctsofasinglepredicate.
Wealsoassumethedefinitionoff-precedencein(11),thoughourproposedredefini-tionoflinearprecedencedoesnotdependonadoptingthisparticularformulationoff-precedenceasopposedtootherdefinitionsoff-precedencethathavebeenadoptedintheliterature(e.g.Bresnan,1995).(11)F-precedence:
f-precedesifandonlyifallc-structurenodescorrespondingtocedeallnodescorrespondingto.(e.g.Kaplan,1989)
pre-
Wenowpresentourrevisedprominencerequirements:
(12)LetCoargOpandCoargProbecoargumentf-structuressuchthatCoargOp
containsOandCoargProcontainsP.Then:
SyntacticProminenceAnoperatorOismoreprominentthanapronounP
ifandonlyifCoargOpisatleastashighasCoargProonthefunctionalhierarchy.LinearProminenceAnoperatorOismoreprominentthanapronounPif
andonlyifCoargOpf-precedesP.Themostimportantdifferencebetweenourreviseddefinitionsandthosein(8)isthatlinearprominencedependsonthef-precedencepropertiesofCoargOp,anf-structurecontainingtheoperator,notontheoperator.4
Evaluatingtheproposal
Inthissection,weshowhowourreviseddefinitionscanaccountforweakcrossoverphenomenainEnglish,German,andMalayalam,withoutpositingtraces.4.1
English
FollowingBresnan(1995),weassumethatinEnglish,theoperatormustoutrankanypronounitbindsinbothLinearProminenceandSyntacticProminence.Withthis,wecorrectlypredicttheunacceptabilityofexample(13):
(13)*Whodidhismothergreet?
CP
NPCPRED‘greetSUBJ,OBJ’
*WhoCIPOBJCoargOp:PRED‘who’didDetNPNISUBJCoargPro:SPECPRED‘pro’VPVPRED‘mother’hismotherFOCUSgreetToapplyourdefinitionsofprominence,wemustfirstlocatethef-structurecoargu-mentsthatcontaintheoperatorandthepronoun.Weobservethefollowing:
SUBJandOBJofgreetarecoarguments
pronounishis
SUBJofgreetcontainsthepronoun,soitisCoargPro
OBJofgreetcontainstheoperator,soitisCoargOp
Onthisbasis,wecheckwhetherbothLinearProminenceandSyntacticProminencearesatisfied.In(13),theOBJwhof-precedesthepronoun,sinceeveryc-structurenodecorrespondingtotheOBJf-structureprecedesthepronoun;thus,theLinearProminencerequirementissatisfied.However,theSyntacticProminencerequire-mentisviolated,sinceCoargPro(SUBJ)outranksCoargOp(OBJ)onthefunctionalhierarchy.Thisaccountfortheunacceptabilityoftheexample.
Next,weexaminethegrammaticalsentenceWhodidSuetalktoabouthismother?:
(14)WhodidSuetalktoabouthismother?
PRED‘talkSUBJ,OBLto,OBLabout’
FOCUSPRED‘who’CPSUBJPRED‘Sue’NPCOBLtoCoargOp:PRED‘toOBJ’WhoCIPOBJPRED‘aboutOBJ’
didNPIOBLaboutCoargPro:OBJSPECPRED‘pro’
SueVPPPVPPPPPRED‘mother’talkNPDetNtoabouthis
Inthisexample:
mother
OBLtoandOBLaboutoftalkarecoarguments
pronounishis
OBLaboutoftalkcontainsthepronoun,soitisCoargPro
OBLtooftalkcontainstheoperator,soitisCoargOp
Again,wechecktomakesurethatbothLinearProminenceandSyntacticPromi-nencearesatisfied.Thesyntacticprominencerequirementissatisfied:weassumethatallobliquesoccupythesamerankonthefunctionalhierarchy,sothatCoargOp(OBLto)isatleastashighasCoargPro(OBLabout)onthefunctionalhierarchy.TheLinearProminencerequirementisalsomet.TheCoargOpf-structurecorre-spondstothec-structurenodesPPandPdominatingto.CoargOpf-precedesthepronounf-structure,andtheLinearPrecedenceconditionismet.Sincebothre-quirementsaresatisfied,theexampleiscorrectlypredictedtobegrammatical.
Finally,weexaminetheungrammaticalexample*WhodidSuetalkabouthismotherto?:
(15)*WhodidSuetalkabouthismotherto?
PRED‘talkSUBJ,OBLto,OBLabout’
FOCUSPRED‘who’CPSUBJPRED‘Sue’NPCOBLtoCoargOp:PRED‘toOBJ’WhoCIPOBJPRED‘aboutOBJ’
didNPIOBLaboutCoargPro:OBJSPECPRED‘pro’
SueVVPPPPPNPDetNPPRED‘mother’talkPabouttohismother
(15)hasthesamecoargumentsas(14).Again,theSyntacticProminencerequire-mentismet,sincethetwoobliquesoccupythesamerankonthefunctionalhierar-chy.Thec-structurenodescorrespondingtoCoargOparethePPandPdominatingto;thesenodesdonotf-precedethepronoun,theLinearPrecedencerequirementisnotmet,andtheexampleiscorrectlyclassifiedasungrammatical.
Thus,wehaveseenthattracesarenotnecessarytoaccountfortheEnglishweakcrossoverdataoncecoargumentsaretakenintoaccount.4.2
German
FollowingBerman(2000)andBresnan(2001),weassumethatoperatorbindinginGermanrequiresthateithertheLinearProminenceortheSyntacticProminencerequirementmustbemet;unlikeEnglish,meetingasinglerequirementsufficesforgrammaticality.
First,weconsidersomepatternsofquantifierbinding.Bothoftheexamplesin(16)aregrammatical:
(16)...dass[jeder][seineMutter]mag
thateveryone-NOMhismotherlikes
(CoargOp,SUBJ)(CoargPro,OBJ)...dass[seineMutter][jeder]magthathismothereveryone-NOMlikes
(CoargPro,OBJ)(CoargOp,SUBJ)‘...thateveryonelikeshismother’
Inexample(17),thequantifierjeder‘everyone’istheCoargOpSUBJ,andCoargProistheOBJ:
SUBJandOBJofmag‘like’arecoarguments
pronounisseine
OBJofmag‘like’isCoargPro
SUBJofmag‘like’isCoargOp
TheSyntacticProminencerequirementismetinbothexamples,sinceSUBJout-ranksOBJ.Sinceonlyasinglerequirementmustbesatisfied,theLinearPromi-nencerequirementneednothold,andtheSUBJandOBJmayappearineitheror-der.
WhentheLinearProminencerequirementismet,theSyntacticProminencere-quirementneednotbemet.Inexample(17),CoargOpistheOBJjeden‘everyone’,andCoargProistheSUBJseineMutter‘hismother’.ThoughtheSyntacticPromi-nencerequirementisnotmet,theCoargOpf-precedesthepronoun,andtheexampleisgrammatical:
(17)...dass[jeden]seineMuttermag
thateveryone-ACChismotherlikes
(CoargOp)
‘...thathismotherlikeseveryone’
SUBJandOBJofmag‘like’arecoarguments
pronounisseine
SUBJofmag‘like’isCoargPro
OBJofmag‘like’isCoargOp
Inexample(18),whichhasthesamecoargumentstructureof(17),neitherrequire-mentismet,andthesentenceisungrammatical:(18)*...dass[seineMutter][jeden]mag
thathismothereveryone-ACClikes
(CoargPro,SUBJ)(CoargOp,OBJ)
‘...thathismotherlikeseveryone’
Wenowturntoexamplesinvolvinglongdistancedependencies.Inexample(19),CoargProistheSUBJof‘say’,seineMutter‘hismother’.CoargOpistheCOMPof‘say’.SinceSUBJoutranksCOMP,CoargOpisnotmoresyntacticallyprominentthanCoargPro.AndsinceCoargOpdoesnotf-precedethepronoun,thelinearprominencerequirementdoesnotholdeither.Sinceneitherrequirementholds,thesentenceisclassifiedasungrammatical.Notethatthereasonfortheungrammat-icalityofexample(19)isnotthatthereisatraceinthesubordinateclause;instead,thesubordinateclauseitselfistreatedastheimportantconstituentindeterminingprecedencerequirementsbetweentheoperatorandthepronoun.(19)*jeden/wenmeinte[seineMutter],[habesiegetr¨ostet]
everyone/whosaidhismotherhassheconsoled
(CoargPro,SUBJ)(CoargOp,COMP)
‘Everyone,hismothersaidthatsheconsoled./Whodidhismothersaythatsheconsoled?’
SUBJandCOMPofmeinte‘say’arecoarguments
pronounisseine
SUBJofmeinte‘say’isCoargPro
COMPofmeinte‘say’isCoargOp
Inexample(20),weevaluatetheprominenceconditionswithrespecttothecoar-gumentSUBJandOBJofgetr¨ostet‘consoled’:theSUBJseineMutter‘hismother’isCoargPro,andtheOBJisCoargOp.TheSyntacticProminenceconditiondoesnothold,sincetheCoargProSUBJoutrankstheCoargOpOBJ.ButtheLinearPromi-nenceconditionholds,sincetheCoargOpf-precedesthepronoun.Thus,theop-eratorbindingconditionsforGermanaremet,andthesentenceispredictedtobegrammatical:
(20)[jeden/wen]sagtesie,habeseineMuttergetr¨ostet
everyone/whosaidshehashismotherconsoled(CoargOp)
‘Everyone,shesaidthathismotherconsoled./Whodidshesaythathismotherconsoled?’
SUBJandOBJofgetr¨ostet‘consoled’arecoarguments
pronounisseine
SUBJofgetr¨ostet‘consoled’isCoargPro
OBJofgetr¨ostet‘consoled’isCoargOp
Thus,weseethatinGermansimplexandcomplexclauses,thereisnoneedtoposittracestoaccountfortheweakcrossoverdata.4.3
Malayalam
Finally,webrieflyexaminethebehaviorof‘nullpronouns’,phonologicallyunre-alizedpronominalelements.AsshownbyMohanan(1982)andBresnan(2001),onlytheLinearProminenceconditionisrelevantinMalayalam:theCoargOpmustf-precedethepronoun,butnosyntacticprominenceconditionisrelevant.Example(21)isungrammaticalwithanovertpronoun:
¯rookut(21)*innaleawa¯resakaa¯riccastriikal.einn[oo¯.t.iyum]sahaayiccu¯¯¯¯¯
helpedyesterdaytheyscoldedwomantodayeachchild
(CoargOp)
‘Todayeachchildhelpedthewomanwhoscoldedthemyesterday.’
SUBJandOBJofsahaayiccu‘helped’arecoarguments
pronounisawa¯re
OBJofsahaayiccu‘helped’isCoargPro
SUBJofsahaayiccu‘helped’isCoargOp
Inthisexample,CoargOpisthequantifierphraseoo¯rookuttiyum.CoargOpdoes..notf-precedetheovertpronounawa¯re,andsotheLinearProminencerequirementisnotmet.
However,withanullpronoun,thesentenceisgrammatical:
e
¯(22)innalerookutsakaa¯riccastriikal.einn[oo¯.t.iyum]sahaayiccu¯¯¯¯¯
yesterday(they)scoldedwomantodayeachchildhelped
(CoargOp)
‘Todayeachchildhelpedthewomanwhoscoldedthemyesterday.’
Thesamecoargumentinformationholdsfor(22)asfor(21).However,thenullpro-noundoesnotcorrespondtoanyc-structurenodes,andsoaccordingtothedefi-nitionoff-precedencein(11),itvacuouslyf-precedesandisf-precededbyeveryotherf-structure.Inparticular,itvacuouslysatisfiestheLinearProminencecondi-tion.Thisexplainsthedifferencebetweentheovertandnullpronounsinexamples(21)and(22).5
Empiricaldifferencesbetweentraceandtracelessaccounts
ThoughthedatafromEnglish,German,andMalayalamdiscussedaboveareclassi-fiedcorrectlybyourtheoryofprominenceaswellasbytheoriesthatassumetraces,thereareempiricallytestabledifferencesbetweenthetwoapproaches;inparticu-lar,theavailabilityofcertainsetsofdatawouldallowustodistinguishbetweenourtracelessaccountandtheaccountsofBresnan(1994),Bresnan(2001),andBerman(2000),whichassumetraces.Thoughwehavenotyetbeenabletofindthecrucialdatathatwouldallowustodecidebetweenthetwotheories,weprovidecharacter-izationsofsomeofthesedifferencesinthefollowing,inhopesthatsomeofthesecrucialdatacanbefoundandexamined.5.1
Language1
WefirstexaminewhatwewillcallLanguage1,alanguageinwhichtheSyntacticProminenceconditiondoesnotapply.Morespecifically,Language1hasthefol-lowingcharacteristics:
Fixedwordorder
Wh-phrasesaredisplacedtothebeginningofthesentence
OnlyLinearProminenceisrelevant
InLanguage1,ourtheorypredictsthatanexamplelike(23)wouldbegrammatical,sinceCoargOpprecedesthepronoun,andbyhypothesistheSyntacticProminenceconditiondoesnotapply:
(23)[who]did[hismother]see()
(CoargOp,OBJ)(CoargPro,SUBJ)
e
Incontrast,thetracefultheorypredictsungrammaticality,sincethepronounout-ranksthe(traceofthe)operatorinLinearProminence.5.2
Language2
WenextexamineLanguage2,alanguageinwhichtheobjectprecedesthesubject.Language2hasthefollowingcharacteristics:
Fixedwordorder;Objectprecedessubject
Wh-phrasesaredisplacedtothebeginningofthesentence
BothLinearProminenceandSyntacticProminencemusthold
InLanguage2,ourtheorypredictsgrammaticalityforexample(24):sawhisbook()(24)[who]
(CoargOp,SUBJ)(CoargPro,OBJ)
HeretheSyntacticProminencerequirementholds,sinceCoargOpisSUBJandCoargProisOBJ.TheLinearProminencerequirementisalsomet,sinceCoargOpprecedesthepronoun.Incontrast,thetracefultheorypredictsungrammaticality,sincethepronounprecedesthetrace.5.3
Language3
Finally,weexamineLanguage3,alanguageinwhicheitherprominencerequire-mentmusthold.Language3hasthefollowingcharacteristics:
Fixedwordorder;Subjectprecedesobject
Wh-phrasesaredisplacedtothebeginningofthesentence
EitherLinearProminenceorSyntacticProminencemusthold
Ourtheorypredictsgrammaticalityforexample(25):sawitsowner()(25)[what]
(CoargOp,OBJ)(CoargPro,SUBJ)
TheSyntacticProminencerequirementisnotmetinthisexample,sinceCoargOpisOBJandCoargProisSUBJ.However,theLinearProminencerequirementismet,sinceCoargOpprecedesthepronoun.Thetracefultheorypredictsthatthisexampleisungrammatical,however,sinceonceagainthepronounprecedesthetrace.
6Summary
Ourtracelesstheoryofweakcrossoveraccountsforthedataonweakcrossoverthatmotivatedpreviousresearcherstoposittracesinlongdistancedependencies.WeaccomplishedthisbyprovidinganewdefinitionofLinearProminencewhichap-pliestof-structureunitsthatcontaintheoperator.Sinceweareawareofnootherevidencesupportingtheexistenceoftracesinlongdistancedependencies,webe-lievethattracesremainunmotivatedinthetheoryofgrammar.Acknowledgments
WearegratefultoYehudaFalkandtheaudienceatLFG2001forhelpfuldiscussionandcomments.References
Berman,Judith.2000.TopicsintheClausalSyntaxofGerman.Ph.D.thesis,Uni-versityofStuttgart,Stuttgart,Germany.Bresnan,Joan.1994.Linearordervs.syntacticrank:evidencefromweakcrossover.InKatieBeals,JeannetteDenton,BobKnippen,LynetteMelnar,HisamiSuzuki,andErikaZeinfeld(editors),PapersfromtheThirtiethRegionalMeetingoftheChicagoLinguisticSociety.UniversityofChicago:ChicagoLinguisticSociety.Bresnan,Joan.1995.Linearorder,syntacticrank,andemptycategories:Onweakcrossover.InMaryDalrymple,RonaldM.Kaplan,JohnT.Maxwell,andAnnieZaenen(editors),FormalIssuesinLexical-FunctionalGrammar,pp.241–274.StanfordUniversity:CSLIPublications.Bresnan,Joan.2001.Lexical-FunctionalSyntax.Oxford:BlackwellPublishers.Chomsky,Noam.1973.Conditionsontransformations.InStephenAndersonandPaulKiparsky(editors),AFestschriftforMorrisHalle.NewYork:Holt,Rein-hart,andWinston.Dalrymple,Mary,RonaldM.Kaplan,JohnT.Maxwell,III,andAnnieZaenen(edi-tors).1995.FormalIssuesinLexical-FunctionalGrammar.StanfordUniversity:CSLIPublications.Kaplan,RonaldM.1989.TheformalarchitectureofLexical-FunctionalGrammar.InChu-RenHuangandKeh-JiannChen(editors),ProceedingsofROCLINGII,pp.3–18.ReprintedinDalrympleetal.(1995,pp.7–27).
Kaplan,RonaldM.andJoanBresnan.1982.Lexical-FunctionalGrammar:Afor-malsystemforgrammaticalrepresentation.InJoanBresnan(editor),TheMentalRepresentationofGrammaticalRelations,pp.173–281.Cambridge,MA:TheMITPress.ReprintedinDalrympleetal.(1995,pp.29–130).Kaplan,RonaldM.andAnnieZaenen.1989.Long-distancedependencies,con-stituentstructure,andfunctionaluncertainty.InMarkBaltinandAnthonyKroch(editors),AlternativeConceptionsofPhraseStructure,pp.17–42.ChicagoUni-versityPress.ReprintedinDalrympleetal.(1995,pp.137–165).Mohanan,K.P.1982.GrammaticalrelationsandclausestructureinMalayalam.InJoanBresnan(editor),TheMentalRepresentationofGrammaticalRelations,pp.504–589.Cambridge,MA:TheMITPress.Postal,PaulM.1971.Cross-OverPhenomena.NewYork:Holt,Rinehart,andWinston.Reinhart,Tanya.1983.AnaphoraandSemanticInterpretation.Chicago:TheUni-versityofChicagoPress.Sag,IvanA.1998.Withoutatrace.
MS,
StanfordUniversity.
Wasow,Thomas.1979.AnaphorainGenerativeGrammar.Ghent:E.Story.Addresses:
dalrymple,kaplan,thking@parc.xerox.comXeroxPARC,PaloAlto,CA94304USA
因篇幅问题不能全部显示，请点此查看更多更全内容