Implicit Arguments - University of Massachusetts Amherst

27 Pages·2004·0.11 MB·English

Implicit Arguments
RajeshBhatt,RoumyanaPancheva
[e-mail],[e-mail]
1 Overview
Referencetonon-overtargumentshasbeenmadeinthedescriptionofawide
rangeofsyntacticphenomena. Someofthem (PRO,pro,A/A-traces)arerel-
0
atively well-understood and there exists a certain consensus regarding their
analysis. There is another class of non-overt arguments, often referred to as
implicitarguments,forwhichnosuchconsensusprevails. Implicitarguments
do not seem to form a unified class. To appreciate this let us examine some
caseswhichhavebeenarguedtoinvolveimplicitarguments.
(1) Implicitagentsofpassives(vs.middlesandunnacusatives)
a. Thisshipwassunk[PROtocollecttheinsurance].(Passive)
b. #Thisshipsank[PROtocollecttheinsurance].(Unaccusative)
c. *Thisshipsinkseasily[PROtocollecttheinsurance].(Middle)
(2) Benefactiveargumentsofadjectives(fromRoeper(1987))
a. Itisnecessary/*inevitable[PROtogo].
b. Itiswise/*probable[PROtogo].
(3) Thebeareroftheobligationofadeonticmodal
a. Thebooks canbe sold[without PRO readingthem]. (fromChom-
sky(1982)viaWilliams(1985))
b. *Thebooksmighthavebeensold[withoutPROreadingthem].(from
Kratzer(1991))
(4) Implicitagentsofnouns
a. thenegotiations[PROtoachieveapeacefulsettlement]
b. thetakingofdrugs[PROtobecomehappy]
c. theplayingofthegame[PROtoproveapoint]
(5) Implicitagentsofagentivesuffixes(e.g.-able)
Goodsareexportable[PROtoimprovetheeconomy].
(6) Nullobjects(cf.Rizzi(1986))
a. Questoconduce(la gente) alla seguente conclusione
this leads thepeopleto-thefollowingconclusion
‘Thisleads(people)tothefollowingconclusion.’
1
b. Questoconduce(la gente) a [PROconcluderequantosegue].
this leads thepeopleto conclude what follows
‘Thisleadspeopletoconcludewhatfollows.’
Theabovelistincludestheimplicitagentofapassive( 2),thebenefactivear-
gumentofanadjective( 5),thebeareroftheobligationxargumentofadeontic
modal( 6),theimplicitaxgentofanoun( 3),theimplicitagentassociatedwith
agentivexsuffixeslike-able,andnullobjecxts( 4). 1 Whatunifiesthisclass? Itis
feltthatalloftheseexamplesinvolveamissinxgnominalelement.Theevidence
forthismissing nominal elementcomesfromthefactthat(1-6)allinvolvean
infinitival with a PRO subject. Something, it is argued, must be controlling
thesePROs. ThereisnoNPargumentintherelevantstructuresthatcouldbe
doingso. Theelementheldresponsibleforcontrolistheimplicitargument.
Inprinciple,nullsubjects(PRO,pro)couldhavebeencalledimplicitargu-
ments, given that they are non-overt and indisputably arguments. Further-
more there have been analyses in the literature according to which PRO/pro
are not syntactically expressed (for PRO see Partee and Bach (1980), Chier-
chia(1984),KleinandSag(1985)a.o,andforproseeAlexiadouandAnagnos-
topoulou (1998)a.o.). However, byconvention, PRO/pro arenot groupedto-
getherwiththecasesofimplicitargumentsin(1-6). Thisiswhyinthissurvey
wedonotdiscussPRO/pro.
From its inception, the literature on implicit arguments has defined them
assyntacticallyactiveelementsthatneverthelessdonotoccupyasyntactically
projectedposition. Considerforexample,thefollowingdefinitionsforimplicit
argumentsthathavebeenproposedintheliterature.
(7) Implicitargumentsarenotthemysteriousshadowypresencestheyare
sometimesmadeouttobe.Theyarereallynothingmorethantheargu-
mentslotsintheargumentstructure,...... A‘weak’ -criterionisallthat
(cid:18)
isneededtogiveimplicitarguments,sincethesearenothingmorethan
unlinkedargumentroles.(pg.314fromWilliams(1985))
(8) Animplicitargumentisaconceptualargumentthatisneitherexpressed
syntacticallynorboundtoanargumentthatisexpressedsyntactically.
(pg.409fromJackendoff(1987))
However,notallresearchershavefollowedthisunderstandingofimplicit ar-
guments. Somehaveproposedthatwhattheycallimplicitargumentsarereal-
izedaspro(cf.Epstein(1984),Rizzi(1986),Borer(1998))andPRO(cf.Roeper(1987)).
1Larson(1988)usestheterm‘implicitargument’moregenerallytocoverinadditiontothecases
discussedinthemaintext,optionalbutnon-iterablephrasessuchasphrasesofsource,path,goal
andphrasesofinstrumentality.
i. a. Johnran( fromthehouse)( tothestore)( alongtheriver).
Source Goal Path
b. Johncutthesalami( withaknife).
Instrument
2
Butevenifitisassumedthatallofthenon-overtargumentsin(1-6)aresyntac-
ticallyrealizedsyntacticallyaspro/PRO,giventhat(1-6)arenotthecanonical
environments for pro/PRO, it is justified to include them in a discussion of
implicitarguments. Bygeneralassumptionthen,thetermimplicitarguments
isreservedforthosecovertelementsaboutwhosesyntacticrepresentationwe
stillhavedoubts.
In(1-6),wehavealistofenvironmentswhichhavebeenclaimedtoinvolve
implicit arguments. That these different environments have been argued to
involveimplicitargumentsshouldnot,byitself,betakentoindicatethatthey
shouldreceiveauniformanalysis.Ithasbeenquestionedwhethersomeofthe
membersofthelistin(1-6)aretrulyimplicit. Forexample,Bakeretal.(1989)
argue the passive suffix -ed is the agent of the passive i.e. it is not implicit.
Even if an argument is implicit, the question of how it is to be represented
stays open. The implicit argument may be representedat a level of thematic
structure, or as anull PRO/pro subject or object. For example, Roeper (1987)
arguesthatthesubjectsofnounsarerealizedbyPRO.SimilarlyEpstein(1984),
Rizzi (1986), and Borer (1998)argue that the implicit argument of evaluative
predicates,nullobjectsinItalian,andimplicitexternalargumentofthepassive,
respectively, are realized by pro. In what follows, the cases in (1-6) and the
analysesproposedforthemwillbediscussedonebyone. Throughoutwewill
focusonthediagnosticsthathavebeenproposedfordetectingthepresenceof
animplicitargument.Wewillexaminewhatexactlythesediagnosticstest.
2 Implicit Arguments in Passives and Middles
2.1 Implicit Argumentsin Passives
Aclassiccasewhereanimplicitargumenthasbeenarguedtobepresentisthat
of an ‘agentless’ passive. Passives have been contrasted with unaccusatives,
whichincontrasttopassiveshavebeenarguedtonotinvolveanimplicitagent.
(9) (fromRoeper(1987))
a. Theshipwassunk.
b. Theshipsank.
Intuitively,itseemsclearthatthepassivein(9a)hasimpliedagencyaspartof
itsmeaningwhiletheunaccusativein(9b)doesnot. Incasetheshipspranga
leakonitsown andsank,(9b)wouldbetruebutnot(9a). For(9a)tobetrue,
theremusthavebeensomeonewhowasresponsibleforsinkingtheship.
If byimplicit argumentwe mean anon-overt element that is nevertheless
partoftheinterpretationofapredicate,thenthefactsabouttheinterpretation
of (9a)areenough to show that passives involve an implicit agent and unac-
cusativesdonot. However,theliteratureonimplicitargumentsaimstogobe-
yondjustshowingthataparticularnon-overtargumentispartofthesemantic
representation.Itattemptstoshowthatimplicitargumentstakepartinsyntac-
ticprocessesandthatthereforeimplicitargumentsaresyntacticallyreal. Once
3
itisgrantedthatimplicitargumentsaresyntacticallyreal,additionalquestions
arisethatpertaintotherepresentationofthesearguments.
The primarytests used to demonstratethe syntactic realityofthe implicit
agentofapassivearelicensing of by-phrases,the abilitytocontrol, andcom-
patibilitywithadverbslikedeliberately.Theargumentfromby-phraselicensing
goesasfollows.Passivesallowforanovertby-phrasewhileunaccusativeverbs
donot(cf.10). Theimplicitagentinthepassiveistakentoberesponsiblefor
thislicensing.
(10) (fromRoeper(1987))
a. TheshipwassunkbyBill.
b. *TheshipsankbyBill.
Nextweturntotheargumentfromcontrol. Theimplicitagentofapassive
can control the PRO subject of a rationale clause. Unaccusativesdo not have
animplicitagentargumentandsocontrolisnotpossible.
(11) (fromManzini(1983))
a. *Theboatsanktocollecttheinsurance.
b. Theboatwassunktocollecttheinsurance.
If we take it as given that only syntactically active elements can control PRO
thenitfollowsthattheimplicitagentofthepassivemustbesyntacticallyreal.
A related point is made by the fact that adverbs like deliberately can ap-
pearinapassivesentenceandbeassociatednotwiththesyntacticsubjectbut
withtheimplicitagent. Incontrast,in(12b)thereisnoimplicitargumentand
deliberatelycanonlybeassociatedwiththeloneargumentofsink,theboat.Con-
sequently(12b)isperceivedaspragmaticallydeviant.
(12) a. Theboatwassunkdeliberately.
b. #Theboatsankdeliberately.
Ifweassumethatadverbslikedeliberatelycanonlybeassociatedwithsyntac-
ticallyrealelements,itfollowsthattheimplicitagentofthepassiveissyntac-
ticallyreal.
Theargumentforthesyntactic realityofimplicit argumentthereforerests
uponouracceptanceoftheassumptionsbehindtheargumentfromby-phrase
licensing,theabilitytocontrol,andcompatibilitywithadverbslikedeliberately.
Wealreadyknowthatpassiveshaveimplicitagentsaspartoftheirsemantics.
It is not clear that postulating that this implicit agent is representedsyntacti-
cally is necessary for explaining the possibility of by-phrases. It may be the
casethatsyntacticallyunexpressedagentivityisallthatisneededforlicensing
aby-phrase.
Nextweturntothe argumentfromtheacceptabilityof deliberately. Again
itisnotclearthatanythingmorethanthepresenceofanimplicitagentinthe
semanticrepresentationisneeded.Weareleftwiththeargumentfromcontrol.
TheimplicitargumentisabletocontrolthePROsubjectofarationaleclause.
4
So the question is whether the ability to control requires the controller to be
syntacticallyrealized.
Williams (1985) provides several arguments in support of the claim that
controllers need not be syntactically realized. The first argument is basedon
theobservationthattheimplicitagentofanounlikeattemptcontrolsthePRO
subjectofthecomplementclauseinboth(13a)and(13b).
(13) a. Attempts[PROtoleave]
b. Yesterday’sattempts[PROtoleave]
Ifitisassumedthattheimplicitagentofattemptin(13a)isaPROthatoccupies
thepositionoccupiedbyyesterdayin(13b),weexpectthePROtobeabsentin
(13b). However, the implicit agent of attempt controls the PRO subject of the
infinitivalclauseinboth(13a)and(13b). Williams(1985)takesthistosuggest
that the implicit agent in (13b) is able to control the subject of the infinitival
clausewithoutbeingsyntacticallyrealized.Williams’argumentdependsupon
theassumption thatthePROthatrealizestheimplicit agentwouldappearin
thepositionoccupiedbyyesterday(presumably[Spec,DP])andnot[Spec,NP].
If the implicit agent can be generated in [Spec,NP], both (13a) and (13b) can
havePROimplicitagentsin[Spec,NP]andtheparallelismbetween(13a)and
(13b) with respect to control follows. In other words, the argument against
realizingtheimplicitagentasaPROsubjectvanishes.
Williams(1985)arguesthatcontrolofrationaleclausesdoesnotrequirethe
controller to be syntactically realized. In fact, the examples that he provides
demonstratethateventhepresenceofanimplicitargumentisnotnecessary.
(14) Grass is green[to promote photosynthesis]. (fromWilliams (1974)via
Williams(1985))
In (14), grass is not a sensible controller. Green does not have any implicit ar-
gument either. Instead what promotes photosynthesis is the circumstance of
grassbeinggreenandthiscircumstanceisunderthecontrolofsomepurpose-
fulagentsuchasevolutionorGod. Williams(1985)alsopointsoutthatinthe
rightcontexttheungrammaticaltheboatsankinordertoimpressthekingbecomes
acceptable.Considerasituationwhereaplaywrightisrationalizingthedesign
ofhisplayandutters:
(15) The boat sank in orderto impress the queen and move her to murder
herhusbandbytheendofactiii. (fromWilliams(1985))
SinceWilliams’ examplesshow thatthePROsubjects ofrationaleclausescan
becontrolledeveninthe absenceofanyimplicit argument, theproposalthat
allcontrollersmustbesyntacticallyrealizedisweakened.
2.2 Implicit Agents in Middles
Thecontrastbetweenpassivesandergativescarriesovertomiddles. Middles
areincompatiblewithovertby-phrases.
5
(16) (ex.75,pg.406fromKeyserandRoeper(1984))
a. Bureaucratswerebribedbymanagers.
b. *Bureaucratsbribeeasilybymanagers.
Theyeitherlackanimplicitagentargument,ortheimplicitargumentofamid-
dleisunabletocontroltheunderstoodsubjectofarationaleclauseorasubject-
orientedadverb.
(17) (fromBakeretal.(1989))
a. Thisbureaucratwasbribed[PROtoavoidthedraft].
b. *Thisbureaucratbribeseasily[PROtoavoidthedraft].
(18) (fromBakeretal.(1989))
a. Thisbureaucratwasbribeddeliberately.
b. *Thisbureaucratbribesdeliberately.
Despite these tests, the English middle construction has been assumed to in-
volve an implicit agent. For example, Fiengo (1980), Condoravdi (1989), Fa-
gan (1992), Zribi-Hertz (1993), and Ackema and Schoorlemmer (1995) argue
thatthattheimplicitagentofthemiddleispresentata(lexical)-semanticlevel,
whileStroik(1992)andHoekstraandRoberts(1993)proposethattheagentof
themiddleissyntacticallyrealized. Ofcourseanytheorywhichproposesthat
anagentispresentinmiddlesneedstoprovideanaccountofwhymiddlesfail
thetestsforagentivityshownin(16-18),alloftheimplicitagentofthepassive
passes.
Thereareseveralreasonsthathavebeenadvancedforassumingthatmid-
dlesinvolveanimplicitagent. Thefirstisthecontrastinmeaningthatisper-
ceivedbetween(19a,b).
(19) a. Theclotheshangeasily.(fromKeyserandRoeper(1984))
b. Theclothesarehangingontheline. (fromMarantz(1981))
Keyser and Roeper (1984)observe that (19a)implies that it is easy for some-
onetohangclothes,whereasthereisabsolutelynoimpliedagentintheerga-
tive/unaccusative (19b). This is why the addition of easily to (19b) is infelic-
itous. The second reason for postulating an implicit agent in the middle is
thateventhoughmiddlesdonotlicenseby-phrases,theycancontainafor-PP
whoseargumentseemstobeidenticaltotheagentofthemiddleverb.
(20) a. Frenchbooksreadeasilyforeducatedpeople.
b. LatintextsdonottranslateeasilyforBill.
Thelicensingofagentivefor-PP’shasbeenusedtoargueforthesyntacticpres-
ence of an agent in the middle. For Stroik (1992), the for-phrase is an overt
realizationoftheagentargumentrealizedasPROthatforhimisalwayssyn-
tacticallypresentinthemiddle.
Thethirdargumentforagentivity comesfromtheincompatibility ofmid-
dleswithaphraselikeallbyitself.
6
(21) a. *Thiskindofbreadcutseasilyallbyitself.
b. *Thiswoodcarveseasilyallbyitself.
c. *Thisicecrusheseasilyallbyitself.
The ungrammaticality of (21a-c)has been used by Keyser and Roeper (1984)
andFagan(1992)asanargumentinfavorofthemiddle’sagentivity.
However, Rapoport (1999)points out that thereis a serious problem with
theabovetestsforagentivity:theydonotholdforallmiddles.AsAckemaand
Schoorlemmer(1995)noted,notallmiddlesallowforfor-phrases.
(22) a. Thesebooksdon’tsell(*fortheaverageshopkeeper).
b. (Onshoechest:)Stowsonfloororshelf(*fortidypeople).
Rapoport (1999)further points out that many middles are in fact compatible
withallbyitself.
(23) a. Thiskindofglassbreakseasilyallbyitself.
b. Milkchocolatemeltssmoothlyallbyitself.
c. Theseheavywindowsopeneasilyallbythemselves.
d. Thesecomicbookssell(easily)allbythemselves.
Rapoport(1999)thereforeconcludesthatmiddlesdonothaveanimplicitagent.
Thefactthattheimplicitagentofthemiddleissyntacticallyinactive(cf.16-18)
receivesa very natural explanation under Rapoport (1999)’sproposal – there
simplyisn’tanimplicitagentintheEnglishmiddle. TheEnglishmiddleisnot
inherentlyagentive.
As for the licensing of for-phrases in (20)and the unacceptability of all by
itself with certain middles in (21) is related by Rapoport (1999)to the Instru-
mental/Manner(I/M)component in the meanings of certainverbs. Inherent
inthemeaningofcut,carve,andcrush,theverbsin(21),isthemeansormanner
involvedin theactiondescribedbytheverb(theI/Mcomponent). Rapoport
arguesthattheI/Mcomponentbringsalongwithitanimplicationofaproto-
agent and this implication is responsible for the agentivity effects discussed
above. Not all verbs have an I/M component as part of their meaning and
withsuchverbsthereisnoagentivityeffect(cf.22,23). Tosumup,wedonot
needtopostulateanimplicitagentinmiddlesacrosstheboard.Somemiddles
don’t haveany agentivityeffectsandthe agentivityeffectsin theone that do
canbeaccountedforwithoutpostulatinganimplicitargument.2
2Thereisa richliterature onmiddleandpassiveconstructions involving siintheRomance
languages.ThediscussioninCinque(1988)seemstosuggestthatsiconstructionsinItalian,which
canbebothmiddlesandpassives,allowcontrolwhenthesiconstructioncanbepassivebutnot
whenitcanonlybeamiddle.SeeCinque(1988),Vinet(1988)andDobrovie-Sorin(1994)fordetails.
7
3 Implicit Arguments of Nouns
Oneargumentforthesyntacticvisibilityofimplicitargumentsofnounscomes
fromthefactthattheyparticipateinbindingtheory.
(24) a. ConditionA:
Respectforoneselfisimportant. (fromWilliams(1985))
b. ConditionB:
Admirationofhim(admirer admiree)(fromWilliams(1985))
=
6
c. ConditionC:
TherealizationthatJohnwasunpopular(realizer John)(fromRoss(1969b)
=
viaWilliams(1987)). 6
Anaturalwayto accountforthe syntacticvisibility ofimplicit argumentsin-
volvesprojectingthemsyntacticallyasPROsubjects.Ifwedothat,thebinding
effectsin(24)follow directly. Inaddition toparticipatingin binding, implicit
arguments are also able to control and be controlled/bound themselves (cf.
25a,brespectively).
(25) a. Theattempt[PROtoleave]
(attemptercontrolthePRO)
b. Johnmadeanattempt[PROtoleave].
(Johnistheattempter.)
Williams(1985)andWilliams(1987)notethatthebindingandcontroljudge-
mentsstayunchangedevenifthenouninquestionhasanovertNPinitsspec-
ifierposition.
(26) a. Yesterday’sattempt[PROtoleave]
(attempter=leaver)
b. Yesterday’sdecision[thatJohnwasthebestcandidate]
(decider John)
=
6
Williams assumes that a PRO agent would occupy the position occupied by
yesterday’s. Thus a PRO should be blocked in (26a,b). Since the binding and
controljudgementsstayunchanged,Williamsarguesthatbindingandcontrol
donotrequireasyntacticallyprojectedNP.Heproposesthatnon-syntactically
projected implicit arguments are also visible to binding and control. Non-
syntacticallyprojectedimplicitargumentsaremadevisibletosyntaxbystate-
mentslikethefollowing.
(27) Forattempt,andsimilarnouns,theAgentcontrols(orisassociatedwith)
thesubjectoftheembeddedclause.(ex.14onpg.302ofWilliams(1985))
(28) Animplicitargumentc-commandsXiftheverb(ornoun)ofwhichitis
animplicitargumentc-commandsX.
IfanimplicitargumentiscoindexedwithXandc-commandsX,thenit
bindsX.(ex.17onpg.303ofWilliams(1985))
8
Theresultofthesestatementsisthatthebindingandcontroleffectsdiscussed
abovefollownaturally.
Withhissystem,Williams(1985)isabletoderivethefactthatConditionC
effectssurfacewithrespecttoalltheimplicitargumentsoftriadicpredicates.
(29) a. *ThepromisethatJohnwouldwinwasmadetohimyesterday.
b. *ThepromisethatJohnwouldwinwasmadebyhimyesterday.
SinceWilliams(1985)isarguingagainstaPROsubjectofNP,hetakesthefacts
in(29)tocorrectlyshowthatjustsyntacticallyrepresentingtheAgentisinade-
quate.Thisissobecausein(29a)theConditionCeffectistriggeredbythegoal
implicitargumentandnotbytheagentimplicitargument.ThenWilliamsgoes
ontoconcludethattheimplicitagentshouldnotbesyntacticallyrepresented
atall. Thislattermoveisnotforcedbythefacts. Thefactsareequallycompat-
ible with the syntactic projection of the implicit arguments of promise as null
pronouns. Boththesubjectanddirectobjectofpromisewouldc-commandinto
thecomplementofpromiseandyieldtheobservedConditionCeffects.
Moreover,Williams’conclusionthatimplictargumentsarenotsyntactically
projectedisbasedontheasumptionthatanagentprojectedasPROwouldnec-
essarilybeoccupyingthelocationoccupiedbyyesterday’s,whichispresumably
[Spec,DP].Ifweassume,asisplausible,thattheagentprojectedasPROcould
occupy the [Spec,NP] position (assuming that PRO needs no case), then the
thefactsin(26)donotcomeasasurpriseandWilliams’argumentagainstthe
syntacticprojectionofimplicitargumentsisdefused. Tobesure,thefactsare
compatiblewithWilliams’proposal. Theyarealso,however,compatiblewith
thesyntacticprojectionofimplicitargumentsasnullpronouns.
3.1 Optionality of the Implicit Argumentsof Nouns
The treatment of anaphors within NPs in Chomsky (1986)assumes that NPs
have implicit agents that are projected as subjects which are visible for the
bindingtheory.
(30) (fromChomsky(1986),pg.166)
a. They told[ storiesabouteachother ].
i NP i
b. *They told[ mystoriesabouteachother ].
i NP i
c. *They told[ storiesaboutthem ].
i NP i
d. They told[ mystoriesaboutthem ].
i NP i
(31) (fromChomsky(1986),pg.166)
a. They heard[ storiesabouteachother ].
i NP i
b. *They heard[ mystoriesabouteachother ].
i NP i
c. They heard[ storiesaboutthem ].
i NP i
d. They heard[ mystoriesaboutthem ].
i NP i
9
Chomsky(1986)arguesthattheobjectoftellandhearin(30,31a,c)hasanim-
plicit subject on a parallel with the overt subject in (30, 31b, d). Since there
is an overt subject in the object of tell and hear in (30, 31b, d), the object con-
stitutes the binding domain of the anaphor/pronoun. (30, 31b)areruled out
becausetheanaphorsarenotboundintheirbindingdomainand(30,31b)are
acceptablebecausethepronounsarefreeintheirbindingdomain.
(30,31a)bythemselvesarecompatiblewiththerebeinganimplicitsubject
in the story NP bound by the matrix subject and the anaphor bound by this
implicitsubject,orwiththerebeingnoimplicitsubjectandthebindingdomain
of the anaphor being large enough to include the matrix subject, which is its
binder.Therelevantevidenceinfavorofpostulatinganimplicitsubjectcomes
fromthecontrastbetween(30c)and(31c).
Theverbstellandheardifferinhowtheycombine semanticallywith their
objectstory.Considerthecontrastininterpretationbetween(32a)and(32b).
(32) a. Johntold[storiesaboutMary].
b. Johnheard[storiesaboutMary].
In (32a),the subject of tell has to be interpretedas the ‘agent’/narratorof the
story. It is claimed that there is no such requirement in (32b). The process
bywhichtheimplicitargumentsofanounreceivetheirinterpretationwillbe
discussed in the next section ( 3.2). For now, we can representthe difference
betweentellandhearbystipulxating thatthe‘agent’/narratorargumentofthe
complementoftellmustbethesameastheagentoftell. heardoesnotbringin
sucharequirement.Thisisshownin(33).
(33) a. John told[IMP storiesaboutMary].
i i=(cid:3)j
b. John heard[IMP storiesaboutMary].
i i=j
Thecontrastbetween(30c)and(31c),repeatedbelowin(34),nowfollows.
(34) a. They told[IMP storiesaboutthem ].
(IMPiisruledou(cid:3)it=b(cid:3)jyConditionB;IMPiisruledoutbytell.)
i j
b. They heard[IMP storiesaboutthem ].
(IMPiisruledoutjb=(cid:3)yiConditionB;IMP iisallowedbyhear.)
i j
Thepresenceoftheimplicitargumentasthesubjectin(34a,b)makestheobject
NPbecomethebindingdomainforthem. Sincein(34a),theimplicitargument
isobligatorily coreferentwith thematrixsubject, wehaveaviolation ofCon-
dition B in (34a). Since hear in (34b) does not require obligatory coreference,
(34b)has a representationwhere the ‘agent’/narrator of the story is different
fromthesubjectofhear.ThisrepresentationdoesnottriggeraviolationofCon-
ditionB.Condition Bisstillrelevantfor(34b)–itblockstheinterpretationof
(34b),wherethehearersarethenarrators.3
Letusnowre-examinetheanalysisofanaphorsinsidetheNPcomplements
ofanaphorsin(30,31a),repeatedhereas(35a,b).
3Itislikelythatthehearersasnarratorsreadingisruledoutbythepragmaticsoftheverbhear.
ItishardtoconstrueJohnheardastoryaboutMaryasJohnhearingastoryaboutMary,whereJohnis
thenarratorofthestory.Presumablywhatevermakesithardtoconstruethehearerasthenarrator
10