Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
1
HowtoNeutralizeaFiniteClauseBoundary:
PhaseTheoryandtheGrammarofBoundPronouns
ThomasGrano
HowardLasnik
Abstract:ThispaperisconcernedwithafactaboutEnglishsyntaxthathasbeen
brieflyobservedinahandfulofplacesintheliteraturebutneversystematically
investigated:aboundpronouninthesubjectpositionofafiniteembeddedclause
renderstheclauseboundaryrelativelytransparenttoprocessesorrelations
ordinarilyconfinedtomonoclausal,control,andraisingconfigurations.Forexample,
too/enoughmovementstructuresinvolvingafiniteclauseboundaryaredegradedin
sentenceslike*Thisbookistoolong[forJohntoclaim[thatBillread__inaday]]but
improvedwhenthefiniteclausehasaboundpronominalsubjectasin?Thisbookis
toolong[forJohn1toclaim[thathe1read__inaday]].Ontheempiricalside,we
showthatthisboundpronouneffectholdsacrossawiderangeofphenomena
includingbutnotlimitedtotoo/enoughmovement,toughmovement,gapping,
comparativedeletion,antecedent-containeddeletion,quantifierscopeinteraction,
multiplequestions,pseudogapping,reciprocalbinding,andmultiplesluicing;and
weconfirmtheeffectviaasentenceacceptabilityexperimenttargetingsomeof
thesephenomena.Onthetheoreticalside,weproposeanaccountofthebound
pronouneffectthathastwocrucialingredients:(1)boundpronounsoptionally
To appear in Linguistic Inquiry
2
enterthederivationwithunvaluedphi-featuresand(2)phasesaredefinedinpart
byconvergence,sothatundercertainconditions,unvaluedfeaturesvoidthephasal
statusofCPandtherebyextendthelocalitydomainforsyntacticoperations.
Keywords:boundpronouns,phasetheory,clause-mateconditions
1Introduction
Englishexhibitsanumberofwellstudiedsyntacticphenomenaallinvolvingsome
kindofoperationorrelationthatcanbecharacterizedtoaveryroughfirst
approximationasclause-bound,i.e.,unabletospanaclauseboundary.These
phenomenaincludebutarenotlimitedtotoo/enoughmovement,gapping,
comparativedeletion,antecedent-containeddeletion,quantifierscopeinteraction,
andmultiplequestions.Thus,thesentencesin(1)areallperfectlyacceptable,
whereastheminimalvariantsin(2)arealldegradedinvirtueoftheclause
boundaryfoundinthebracketedportionofeachsentence.1
(1) a.Thismagazineistoolow-brow[forJohntoread__]. too/enoughMVMT
b.Johnlikesapplesand[Bill<likes>oranges]. GAPPING
c.Morepeoplelikeapplesthan[<like>oranges]. CMPTV.DELETION
d.Johnreadseverything[Billdoes<read>]. ACD
e.[Atleastoneprofessorreadseveryjournal].(∀>∃) QSCOPE
f.Tellme[whoreadswhichjournal]. MULTIPLEQUEST.
3
(2) a.*Thismagazineistoolow-brow[forJohntoclaimthatBillreads__].
b.*JohnclaimsthatMarklikesapplesand[Bill<claimsthatMarklikes>
oranges].
c.*MorepeopleclaimthatBilllikesapples[than<claimthatBilllikes>
oranges].
d.*JohnclaimsthatMarkreadseverything[Billdoes<claimsthatMark
reads>].
e.*[AtleastoneprofessorclaimsthatBillreadseveryjournal].(∀>∃)
f.*Tellme[whoclaimsthatJohnreadswhichjournal].
Onewellknownexceptiontotheclause-boundednesswitnessedin(1)-(2)is
thatnonfinitecontrolandraisingclauseboundariesdonothavethesame
deleteriouseffectasdofiniteclauseboundaries.Thus,ifwecomparetheexamples
in(2)tominimalvariantsinwhichthefiniteclauseboundaryisreplacedbya
nonfiniteclauseintroducedbycontrolverbclaimorraisingverbtend,wefindthat
theexamplesbecomeacceptableonceagain.Thisisshownin(3).2
(3) a.Thismagazineistoolow-brow[forJohntoclaim/tendtoread__].
b.Johnclaims/tendstolikeapplesand[Bill<claims/tendstolike>oranges].
c.Morepeopleclaim/tendtolikeapples[than<claim/tendtolike>oranges].
d.Johnclaims/tendstoreadeverything[Billdoes<claim/tendtoread>].
e.[Atleastoneprofessorclaims/tendstoreadeveryjournal].(∀>∃)
f.Tellme[whoclaims/tendstoreadwhichjournal].
4
Alesswidelyrecognizedexceptiontoclause-boundedness—andthefocus
ofthispaper—istheobservationthatevenfiniteclauseboundariescanbe
renderedrelativelyinnocuousinthephenomenainquestion,providedthatthe
subjectoftheembeddedfiniteclauseisaboundpronoun.WecallthistheBOUND
PRONOUNEFFECT.Itisillustratedin(4).3
(4) a.?Thismagazineistoolow-brow[forJohn1toclaimthathe1reads__].
b.?John1claimsthathe1likesapplesand[Bill2<claimsthathe2likes>
oranges].
c.?Morepeople1claimthatthey1likeapples[than<claimthatthey1like>
oranges].
d.?John1claimsthathe1readseverything[Bill2does<claimsthathe2
reads>].
e.?[Atleastoneprofessor1claimsthathe1readseveryjournal].(∀>∃)
f.?Tellme[who1claimsthathe1readswhichjournal].
Inthispaper,wepresentwhatistoourknowledgethefirstexperimental
documentationoftheboundpronouneffect—intheformofasentence
acceptabilityexperimentdesignedusingthetoolsdescribedinErlewine&Kotek
(2016)andconductedviaAmazon’sMechanicalTurk—aswellasthefirstattempt
atanaccountofit.Theaccounthastwocrucialingredients:first,boundpronouns
optionallyenterthederivationwithunvaluedphi-features(cf.Kratzer1998a,2009;
5
Rullmann2004;Heim2008;Landau2016).Second,phasesaredefinedinpartby
convergence,sothatundercertainconditions,anunvaluedfeaturevoidsthe
otherwisephasalstatusofCP(cf.Chomsky2000:107andFelser2004)andthereby
extendsthelocalitydomainforsyntacticrelations.
Theorganizationoftherestofthepaperisasfollows.Insection2,we
providesomebackground,situatingtheboundpronouneffectwithrespectto
previousliteratureandtothebroaderlandscapeofclause-boundednessandrelated
phenomena.Insection3,welayoutthecoredatathatmotivateourtheoretical
proposalsanddescribeourexperimentalfindings.Insection4,wepresentour
phase-theoreticaccountoftheboundpronouneffectandshowhowitcanbe
embeddedintoexistingformulationsofthePhaseImpenetrabilityConditionsoasto
derivethecrucialfacts.Insection5,weoffersomepreliminaryremarksonthe
boundpronouneffectasitrelatestoislandphenomena.Finally,weconcludein
section6.Anappendixlaysoutingreaterdetailthestatisticalanalysisthatwe
performedonourexperimentalresults.
2Somehistoricalandempiricalcontext
2.1Clause-boundednessandthefinite/nonfinitedistinction
Theideathatsomesyntacticprocessesandrelationscannotcrossaclauseboundary
hasplayedaroleingenerativetheorizingsincethe1950s.(SeeLasnik2002foran
overview.)Ithasalsolongbeenobservedthatnotallclauseboundariesarecreated
6
equal:Chomsky’s(1973)TensedSentenceConditionandSpecifiedSubject
Conditionbothacknowledgetherelativeweaknessofnonfiniteclauseboundaries.
Inarelatedvein,Postal(1974)usestheterm“quasi-clause”(acoinageheattributes
toDavidPerlmutter)forraisingandcontrolcomplements,andhesuggeststhat
quasi-clauseboundariesare“notasstrongabarriertoatleastsomesyntactic
phenomenaasfullclauseboundaries”(p.232).Postalinvokesquasi-clausesin
discussingarangeofprocessesandrelationsincludingcomplexNPshift
(extraposition),comparativedeletion,toughmovement,multiplequestions(cf.also
Kuno&Robinson1972),anddoublenegation(thelatterobeyingan“anti-clause-
mate”condition).
Otherphenomenaforwhichclause-boundednessandthefinite/nonfinite
distinctionhavesincebeenfoundtoberelevantincludereciprocalbinding
(Higginbotham1981),slangNPIlicensing(Lasnik2002),multiplesluicing
(Nishigauchi1998;Merchant2001;Barrie2008;Lasnik2014),“familyofquestions”
readingsofinterrogativesentencesthatcontainauniversalquantifier(May1977,
1985;Williams1986;Sloan&Uriagereka1988;Sloan1991;Lasnik2006;Agüero-
Bautista2007),andACDandquantifierscopeinteraction(May1985;Larson&May
1990;Hornstein1994;Farkas&Giannakidou1996;Kennedy1997;Wilder1997;
Kayne1998;Fox2000;Cecchetto2004;Moulton2008;Hackl,Koster-Hale&
Varvoutis2012;Wurmbrand2013,2015a).
7
2.2ComplicationOne:Restructuring
Forsomephenomena,notallauthorsclaimthattherelevantdistinctionissimply
betweenfiniteclausesandnonfiniteclauses.Onetrendintheliteraturebuildson
Rizzi’s(1978)seminalworkonItalian,wherecliticclimbingandrelatedphenomena
areshowntobeordinarilyclause-boundexceptinsomebutnotallsentences
involvingnonfinitecomplementation.Crucially,Rizzishowedthattheavailabilityof
cliticclimbingacrossanonfiniteclauseboundaryisconditionedbythechoiceofthe
embeddingverb,generalizingthatonlymodal,motion,andaspectualverbsextend
locality.4Thereisnowasizeableliteratureonrestructuringthatcorroborates
versionsofthisclaimforanalogouseffectsinSpanish(Aissen&Perlmutter1983),
German(Wurmbrand2001andLee-Schoenfeld2007),andpotentiallyamuchwider
rangeoflanguagesaswell(Cinque2004;Grano2015;Wurmbrand2015b).
DrawingonthistraditionandbuildingalsoonJohnson(1996),Lechner
(2001)claimsthatgappingandcomparativedeletioninEnglishonlyapplyacross
nonfiniteclauseboundariesiftheembeddingverbisarestructuringverb.Hornstein
(1994)makesthesameclaimforquantifierscopeinteractionandACD,althoughthis
viewisquestionedbyKennedy(1997)andWurmbrand(2013);seealsoMoulton
2008forrelevantexperimentalworkoninversescope.Anotherphenomenonfor
whichrestructuringhasbeeninvokedinEnglishisinfinitivaltocontractionin
locutionslikewanna(fromwantto)orhafta(fromhaveto):seeGoodall1991.5
8
2.3ComplicationTwo:TheBoundPronounEffect
Theboundpronouneffectinvestigatedinthispaperconstitutesyetanother
challengetothecharacterizationoflocalitydomainsintermsofasimple
finite/nonfinitecontrast:justastherestructuringliteraturehasshownthatnotall
nonfiniteclausesarecreatedequal,theboundpronouneffectsshowsthatnotall
finiteclausesarecreatedequal.Asmentionedinnote3above,variousscholarshave
observedtheboundpronouneffectasitpertainstoparticularphenomena(Sloan
1991;Nishigauchi1998;Merchant2001;Lasnik2006,2014;Syrett&Lidz2011),
butthefullrangeofdatabearingontheeffecthasyettobesystematically
documentedandaccountedfor.Wenowturnourattentiontothis.
3Thedata
Inthissection,welayoutthedatathatwillinformouranalysisinsection4below.
Webeginbydocumentingtherangeofphenomenathatexhibittheboundpronoun
effect(section3.1).Wethenshowthattheeffectissubject-oriented(onlybound
pronounsthatareinsubjectpositioninducetheeffect)(section3.2)andthatnon-
boundpronounsdonotinducetheeffect(section3.3).Wethenpresentour
experimentalfindingsthatconfirmtheseobservations(section3.4)andmake
explicitourhypothesisabouthowtheobservedcontrastsinacceptabilitymaponto
thegrammaticalitycutsmadebytheaccountwepursue(section3.5).
9
3.1Therangeofphenomenathatexhibittheboundpronouneffect
In(1)-(4)above,weillustratedtheboundpronouneffectforsixphenomena:
too/enoughmovement,gapping,comparativedeletion,ACD,inversescope,and
multiplequestions.Hereweillustratetheeffectforsixadditionalphenomena,
namely,pseudogapping,reciprocalbinding,multiplesluicing,“familyofquestions”
readings,extraposition,andtoughmovement.Baselinemonoclausalexamplesare
illustratedinthebracketedportionsof(5a-f)respectively.Aswiththeothersix
phenomena,weseedegradedacceptabilitywhenthebracketedportionofthe
sentenceincludesafiniteclauseboundary(6),substantialimprovementwhenthe
embeddedclauseisanonfiniteclauseintroducedbyacontrolorraisingverb(7),
andmoderateimprovementwhentheembeddedfiniteclausehasabound
pronominalsubject(8).
(5) a.Johnlikesapplesand[Billdoes<like>oranges].
b.[JohnandBilllikeeachother.]
c.SomeoneisworriedaboutsomethingbutIdon’tknow[who<isworried>
aboutwhat].
d.[Whichjournaldoeseveryoneread__]?
Anticipatedanswertype:JohnreadsLI,BillreadsNLLT,etc.
e.[Johnreads__verycarefully]—allthemajorlinguisticsjournals.
f.Thisbookiseasy[forJohntoread___].
10
(6) a.*JohnclaimsthatMarklikesapplesand[Billdoes<claimthatMarklikes>
oranges].
b.*[JohnandBillclaimthatMarklikeseachother].
Intendedreading:JohnclaimsthatMarklikesBillandBillclaimsthat
MarklikesJohn.
c.*SomeoneclaimsthatJohnisworriedaboutsomethingbutIdon’tknow
[who<claimsthatJohnisworried>aboutwhat].
d.*[WhichjournaldoeseveryoneclaimthatJohnreads__]?
Anticipatedanswertype:BillclaimsthatJohnreadsLI,Timclaimsthat
JohnreadsNLLT,etc.
e.*[JohnclaimsthatBillreads__everytimeIaskaboutit]—allthe
majorlinguisticsjournals.
f.*Thisbookiseasy[forJohntoclaimthatBillread___].
(7) a.Johnclaims/tendstolikeapplesand[Billdoes<claim/tendtolike>
oranges].
b.[JohnandBillclaim/tendtolikeeachother].
Intendedreading:Johnclaims/tendstolikeBillandBillclaims/tendsto
likeJohn.
c.Someoneclaims/tendstobeworriedaboutsomethingbutIdon’tknow
[who<claims/tendstobeworried>aboutwhat].
d.[Whichjournaldoeseveryoneclaim/tendtoread__]?
Anticipatedanswertype:Johnclaims/tendstoreadLI,Billclaims/tendsto
11
readNLLT,etc.
e.[Johnclaims/tendstoread__everytimeIaskaboutit]—allthemajor
linguisticsjournals.
f.Thisbookiseasy[forJohnto{claimtohaveread___/tendtoread___}].
(8) a.?John1claimsthathe1likesapplesand[Bill2does<claimthathe2likes>
oranges].
b.?[[JohnandBill]1claimthatthey1likeeachother].
Intendedreading:John1claimsthathe1likesBillandBill2claimsthathe2
likesJohn.
c.?Someone1claimsthatthey1areworriedaboutsomethingbutIdon’t
know[who2<claimsthatthey2areworried>aboutwhat].
d.?[Whichjournaldoeseveryone1claimthatthey1read__]?
e.?[John1claimsthathe1reads__everytimeIaskaboutit]—allthe
majorlinguisticsjournals.
f.?Thisbookiseasy[forJohn1toclaimthathe1read].
Thetwelvephenomenaexemplifiedin(1)-(4)oftheintroductionand(5)-(8)here
constitutewhatweconsiderthecorecasesoftheboundpronouneffectforthe
purposeofthispaper,butseealsosection5belowforapreliminarydiscussionof
theboundpronouneffectasitpertainstoislandphenomena.6
12
3.2Subjectorientation
Inthisandthenextsubsectionwediscussinturntwodimensionsofthebound
pronouneffectthatwetaketobecrucialindevelopingouraccountofit:thesubject
orientationoftheeffectandthefactthatthereisnocomparable“non-bound
pronouneffect”.
Bysubjectorientation,wemeanthatonlyboundpronounsthatareinsubject
positiongiverisetotheboundpronouneffect.Thisisillustratedin(9)-(13)for
too/enoughmovement,multiplequestions,andcomparativedeletion,respectively.
(Althoughwebelievesubjectorientationholdsforalltherelevantphenomena,we
restrictourattentionherejusttothesethreecasesforreasonsofspace.)(9)shows
thebaselinemonoclausalexamples,(10)showstheminimalvariantswithfinite
clauseboundariesandnoboundpronoun,and(11)showstheboundpronouneffect
forsubject-positionboundpronouns.Crucially,whatweseein(12)isthatabound
pronouninapositionlowerintheclausethansubjectpositiondoesnothavethe
sameacceptability-boostingeffectasdoesaboundpronouninsubjectposition.Ina
similarvein,(13)showsthattheboundpronouneffectisalsonotoperativefor
subject-internalboundpossessors:fromthisweconcludethattheboundpronoun
hastobetheentiresubjectoftherelevantclauseinorderfortheeffecttobe
operative.Seesection3.4belowforexperimentaldocumentationoftheclineof
acceptabilityimpliedbythejudgmentmarksweusein(9)-(13):sentenceslike(9)
aremoreacceptablethansentenceslike(11)whichareinturnmoreacceptable
thansentenceslike(10),(12),and(13).7
13
(9) a.Thisbookistoovaluable[forJamestolend__toBill].
b.Sandywondered[whichmanboughtGeorgewhichshirt].
c.Moreteachersgavethestudentspencilsthan[gavethestudentspens].
(10) a.*Thisbookistoovaluable[forJamestoclaimthatMarklent__toBill].
b.*Sandywondered[whichmanclaimedthatJamesboughtGeorgewhich
shirt].
c.*Moreteachersclaimedthattheprincipalgavethestudentspencilsthan
[claimedthattheprincipalgavethestudentspens].
(11) a.?Thisbookistoovaluable[forJames1toclaimthathe1lent__toBill].
b.?Sandywondered[whichman1claimedthathe1boughtGeorgewhich
shirt].
c.?Moreteachers1claimedthatthey1gavethestudentspencilsthan[claimed
thatthey1gavethestudentspens].
(12) a.*Thisbookistoovaluable[forJames1toclaimthatBilllent__tohim1].
b.*Sandywondered[whichman1claimedthatBillboughthim1which
shirt].
c.*Moreteachers1claimedthattheprincipalgavethem1pencilsthan
[claimedthattheprincipalgavethem1pens].
14
(13) a.*Thisbookistoovaluable[forJames1toclaimthathis1fatherlent__to
Maria].
b.*Sandywondered[whichman1claimedthathis1fatherboughtGeorge
whichshirt].
c.*Moreteachers1claimedthattheir1assistantsgavethestudentspencils
than[claimedthattheir1assistantsgavethestudentspens].
3.3Non-boundpronouns
In(14)-(16),weseethattheboundpronouneffectgoesawayiftherelevant
pronounisnotbound,thatis,ifthepronounisafreethird-personpronoun(14),a
first-personpronoun(15),orasecond-personpronoun(16).Althoughour
experimentalinvestigationdiscussedinsection3.4belowdoesnotinclude
sentenceswithfreethird-personpronounslike(14),itdoesincludesentenceswith
first-andsecond-personpronounslike(15)-(16),andtheresultsareconsistent
withtheviewthatsentenceslike(15)-(16)areindeednomoreacceptablethanthe
correspondingvariantsin(10)abovethathavefulllexicalsubjects.
(14) a.*Thisbookistoovaluable[forJames1toclaimthatshe2lent__toBill].
b.*Sandywondered[whichman1claimedthatshe2boughtGeorgewhich
shirt].
c.*Moreteachers1claimedthatshe2gavethestudentspencilsthan[claimed
thatshe2gavethestudentspens].
15
(15) a.*Thisbookistoovaluable[forJames1toclaimthatIlent__toBill].
b.*Sandywondered[whichman1claimedthatIboughtGeorgewhich
shirt].
c.*Moreteachers1claimedthatIgavethestudentspencilsthan[claimedthat
Igavethestudentspens].
(16) a.*Thisbookistoovaluable[forJames1toclaimthatyoulent__toBill].
b.*Sandywondered[whichman1claimedthatyouboughtGeorgewhich
shirt].
c.*Moreteachers1claimedthatyougavethestudentspencilsthan[claimed
thatyougavethestudentspens].
Theabsenceoftheeffectforfirst-andsecond-personpronounsasseenin
(15)-(16)isparticularlyimportantbecauseitspeaksagainstanalternative
characterizationoftheboundpronouneffectthatwouldlenditselftoanaccount
substantiallydifferentfromwhatweproposebelow.Inparticular,Syrett&Lidz
(2011),notingthedegradedstatusofACDacrossafiniteclauseboundaryandthe
amelioratingeffectofaboundpronoun,speculatethatthefactscouldbedueto“the
extraprocessingloadintroducedbytheinterpretationofTenseandanewdiscourse
referentinthesubjectoftheembeddedclause”(p.330).Inotherwords,onthis
view,whatiscrucialaboutboundpronounsisthattheydonotintroduceanew
discoursereferentandhenceleadtoeasierprocessingofdependenciesthatspan
16
them.Similarly,JasonMerchant(pers.comm.),reportingondiscussionwithIvan
Sag,suggeststhepossibilitythattheboundpronouneffectisnotreallyaboutthe
boundstatusofthepronounpersebutratheraboutaccessibilityinthesenseof
Ariel(1988):roughly,theeasieritistoresolvetheunderstoodreferentofthe
pronoun,themoreinnocuoustheclauseboundaryistothephenomenainquestion.
Consistentwiththissuggestion,MerchantreportsthatSag’slinguisticintuitionwas
thatfirst-andsecond-personpronounswerejustaseffectiveasboundpronounsin
amelioratingfiniteclauseboundariesinphenomenasuchasgapping.Inyetanother
relatedvein,ananonymousreviewerclaimsthatspeakersfindgappingacrossa
finiteclauseboundarytobejustasacceptablewithafreepronounasitiswitha
boundpronouninexampleslike(17).Cruciallyin(17)thereisadiscourse-salient
antecedentfortherelevantpronounandhence(17)isconsistentwiththeSyrett-
Lidz-Sag-Merchantlineofreasoning.
(17)a.WhatdidtheysayaboutBill’spreferences?
b.?Joeclaimsthathelikesapplesbetter,andTim<claimsthathelikes>oranges.
Insofarasdiscourseparticipantsarealwayssalientoraccessibleinthe
relevantsense,wetakeourexperimentalfindingsconcerningfirst-andsecond-
personpronominalsubjectsasevidenceagainsttheviewthattheboundpronoun
effectissubsumedunderamoregeneralaccessibilityordiscourse-salienceeffect.
Thatbeingsaid,wedonotmeantodenythepossibilitythataccessibilityor
somethinglikeitcouldexistasanindependentfactorthataffectsacceptability
17
ratingsforthekindsofsentencesinquestion.Weleaveafullinvestigationofthis
issuetofuturework.
3.4Anexperimentalinvestigationoftheboundpronouneffect
Here,wereportontwosentenceacceptabilityexperimentsdesignedusingthetools
describedinErlewine&Kotek(2016)andconductedviaAmazon’sMechanical
Turk.TheprotocolfortheexperimentsdescribedherewasgrantedExemptstatus
bythe[REDACTED]HumanSubjectsOfficeonJune9,2016(Protocol#1605885354,
‘Anexperimentalinvestigationoftheboundpronouneffect’,[REDACTED],PI).In
whatfollows,wediscussinturntheconstructionofstimuli(section3.4.1),
recruitmentofparticipantsandexperimentalapparatus(section3.4.2),results
(section3.4.3),andsummaryofthestatisticalanalysis(section3.4.4).Thecomplete
detailsassociatedwiththestatisticalanalysisaresuppliedintheappendix.
3.4.1Constructionofstimuli
Wetargetedthreephenomenaforinvestigation:multiplequestions,too/enough
movement,andcomparativedeletion.Oneofthereasonsforchoosingthesethree
particularphenomenaisthat,unlikewhathappenswithsomeoftheother
phenomena,thereisnoneedtocontrolfortheavailabilityofuncontroversially
grammaticalbutirrelevantparsesofsomeofthetargetsentences.Forexample,the
degradedinstancesofinversescope,ACD,andgappingin(18)aresurface-string
18
identicaltotherespectivegrammaticalparsesin(19).Bycontrast,thestrings
associatedwiththedegradedinstancesofmultiplequestions,too/enough
movement,andcomparativedeletionin(20)havenoalternativeuncontroversially
grammaticalparse.
(18) a.*[AtleastoneprofessorclaimsthatBillreadseveryjournal].(∀>∃)
b.*JohnclaimsthatMarkreadseverything[Billdoes<claimsthatMark
reads>].
c.*JohnclaimsthatMarklikesapplesand[Bill<claimsthatMarklikes>
oranges].
(19) a.[AtleastoneprofessorclaimsthatBillreadseveryjournal].(∃>∀)
b.JohnclaimsthatMarkreadseverything[Billdoesread>].
c.JohnclaimsthatMarklikesapplesand[Bill<likes>oranges].
(20) a.*Sandywondered[whichmanclaimedthatJamesboughtGeorgewhich
shirt].
b.*Thisbookistoovaluable[forJamestoclaimthatMarklent__toBill].
c.*Moreteachersclaimedthattheprincipalgavethestudentspencilsthan
[claimedthattheprincipalgavethestudentspens].
19
Foreachofthe3targetedphenomena,weconstructed5baselinesentences
inwhichtherelevantdependencyoccursinamonoclausalframe,yieldingatotalof
15sentences,shownin(21)-(23).
(21)BaseLine(Multiplequestions)
a.SandywonderedwhichmanboughtGeorgewhichshirt.
b.Kimdoesn’trememberwhichmanlentJillwhichmagazine.
c.AbbyfoundoutwhichmantoldBarrywhichjoke.
d.MaryaskedwhichmansentFredwhichletter.
e.MarkdiscoveredwhichmanthrewBobbywhichball.
(22)BaseLine(too/enoughmovement)
a.ThisbookistoovaluableforJamestolendtoMaria.
b.ThisballistooheavyforLindatothrowatBrian.
c.ThisjokeistooinappropriateforPaultotelltoSteve.
d.ThisshirtistooexpensiveforBarbaratobuyforMike.
e.ThisletteristoooutrageousforJennifertosendtoSusie.
(23)BaseLine(Comparativedeletion)
a.Moreteachersgavethestudentspencilsthanpens.
b.Morepoliticianssentthevoterspostcardsthanletters.
c.Moreemployeestoldthebossstoriesthanjokes.
d.Moreauthorssentthepublishernovelsthanshortstories.
20
e.MorepeopleboughtJakeshoesthansocks.
Foreachthese15baselinesentences,weconstructed10variantsthatinvolvea
biclausalconfigurationatthecrucialsite.5ofthe10variantsuseclaimasthe
embeddingverbandtheother5usepromise.Eachgroupof5representsthe5
crucialconditionstestedinExperiment1:anonfiniteclauseboundary,afinite
clauseboundarywithaboundsubject,afiniteclauseboundarywithabound
(prepositional)object,afiniteclauseboundarywithaboundsubject-internal
possessor,andafiniteclauseboundarywithnoboundpronoun.Forexample,the10
variantsconstructedaroundthebaselinesentencein(22a)areasgivenin(24)-
(25).8
(24) Embeddingverb=claim
a.NonFinite
ThisbookistoovaluableforJamestoclaimtohavelenttoMaria.
b.BoundSubj
ThisbookistoovaluableforJamestoclaimthathelenttoMaria.
c.BoundObj
ThisbookistoovaluableforJamestoclaimthatMarialenttohim.
d.BoundPoss
ThisbookistoovaluableforJamestoclaimthathisfatherlenttoMaria.
e.NoBinding
ThisbookistoovaluableforJamestoclaimthatKarenlenttoMaria.
21
(25) Embeddingverb=promise
a.NonFinite
ThisbookistoovaluableforJamestopromisetolendtoMaria.
b.BoundSubj
ThisbookistoovaluableforJamestopromisethathewilllendtoMaria.
c.BoundObj
ThisbookistoovaluableforJamestopromisethatMariawilllendtohim.
d.BoundPoss
ThismovieistoovaluableforJamestopromisethathisfatherwilllendto
Maria.
e.NoBinding
ThisbookistoovaluableforJamestopromisethatBillwilllendtoMaria.
3phenomenaX5sentenceframesX(1baselinecondition+[5non-baseline
conditionsX2embeddingverbs])yieldsatotalof165testsentences.Weused
these165sentencestocreate75listsinsuchawaythateachlistcontains33test
sentences,eachtestsentenceappearson15lists,andnolistcontains2non-baseline
testsentencesthatinstantiatethesamephenomenonandthatvaryalongonlyone
factor(therelevantfactorsbeingsentenceframe,condition,andembeddingverb).
UsingErlewine&Kotek’s(2016)Turktoolssoftware(availableat
http://turktools.net),eachlistwasseparatelyrandomizedandinterspersedwith33
22
fillersentencesofvaryingdegreesofacceptabilityandofcomplexitysimilartothat
ofthetestsentences.
Forthesecondexperiment,theconstructionofstimuliandlistswasidentical
tothatforthefirstexperiment,exceptthattheBoundObjandBoundPossconditions
werereplacedbyconditionsinwhichthefiniteembeddedclausecontainedno
boundpronounbutinsteadcontainedafirst-personsingularpronominalsubject
andasecond-personpronominalsubject,respectively.Forexample,(24c-d)/(25c-
d)fromabovewerereplacedwith(26a-b)-(27a-b),respectively.
(26) a.1pSubj
ThisbookistoovaluableforJamestoclaimthatIlenttoMaria.
b.2pSubj
ThisbookistoovaluableforJamestoclaimthatyoulenttoMaria.
(27) a.1pSubj
ThisbookistoovaluableforJamestopromisethatIwilllendtoMaria.
b.2pSubj
ThisbookistoovaluableforJamestopromisethatyouwilllendtoMaria.
3.4.2Recruitmentofparticipantsandexperimentalapparatus
75experimentalparticipantswererecruitedviaAmazon’sMechanicalTurk,sothat
eachofthe75listswasseenbyoneparticipant.Participantswererequiredtobe
23
nativespeakersofAmericanEnglishatleast18yearsofageandresidinginthe
UnitedStates.Participantswereaskedtorateeachsentenceonascaleof1(least
acceptable)to7(mostacceptable),wherean“acceptable”sentencewasdefinedin
theinstructionsas“somethingthatanativespeakerofEnglishwouldsay,evenifthe
situationthesentencedescribessoundsimplausible”.9
75participantswhohadnotparticipatedinExperiment1wererecruitedto
participateinExperiment2.Recruitmentandinstructionswereotherwiseidentical
tothoseforExperiment1.
3.4.3Results
TheresultsofExperiment1areshowninTable1,whichindicatestheraw
distributionofeachratingforeachofthecrucialconditionsasinstantiatedbyeach
ofthethreephenomenatested.
Table1:Experiment1Results
Key:BL=BaseLine;NF=NonFinite;BS=BoundSubj;
BO=BoundObj;BP=BoundPoss;NB=NoBinding
EXPERIMENT1RESULTSComparativeDeletion MultipleQuestions ƚŽŽͬĞŶŽƵŐŚMovement
RATING BL NF BS BO BP NB BL NF BS BO BP NB BL NF BS BO BP NB1 1 7 6 12 12 10 3 8 21 24 25 22 1 5 8 18 27 202 2 10 12 23 19 14 3 14 32 30 30 32 0 12 17 30 17 303 3 12 21 16 26 34 7 14 33 31 35 30 3 15 22 32 40 354 7 15 14 22 25 21 4 28 19 25 20 19 1 19 21 19 25 155 10 24 28 27 17 34 18 39 20 23 22 23 7 33 32 28 24 276 20 35 42 25 37 25 29 36 19 15 13 21 16 42 33 15 11 187 32 47 27 25 14 12 11 11 6 2 5 3 47 24 17 8 6 5
TOTAL 75 150 150 150 150 150 75 150 150 150 150 150 75 150 150 150 150 150
EXPERIMENT2RESULTSComparative Deletion MultipleQuestions Tough Movement
RATING BL NF BS 1P 2P NB BL NF BS 1P 2P NB BL NF BS 1P 2P NB1 3 5 5 15 17 10 2 6 18 29 30 24 0 7 9 20 23 222 4 6 6 26 26 24 7 9 23 32 38 30 2 11 25 34 28 283 5 16 21 29 22 29 8 30 38 27 32 30 2 18 23 26 25 334 7 25 28 20 34 22 12 29 29 28 22 25 2 17 23 25 23 235 11 23 26 22 16 38 15 33 28 23 21 25 11 32 30 20 32 276 14 39 44 26 24 18 17 30 10 9 5 14 8 35 27 17 15 137 31 36 20 11 9 9 14 13 4 2 2 2 50 30 13 8 4 4
TOTAL 75 150 150 149 148 150 75 150 150 150 150 150 75 150 150 150 150 150
24
TheresultsofExperiment2areshowninTable2.10
Table2:Experiment2Results
Key:BL=BaseLine;NF=NonFinite;BS=BoundSubj;
1P=1pSubj;2P=2pSubj;NB=NoBinding
3.4.4Summaryofstatisticalanalysis
StatisticalanalysiswasperformedinIBMSPSSStatisticsVersion24.Asdescribedin
greaterdetailintheAppendix,anIndependent-SamplesKruskal-WallisTest,when
appliedtotheresultsofExperiment1,indicatesthatthedistributionofratingsis
notthesameacrossthedifferentconditions(𝑋!(5)=325.701,p<0.01).More
specifically,pairwisecomparisonsshowthateachconditiongivesrisetoarating
profilethatissignificantlydifferentfromeachothercondition(p<0.01),exceptfor
theBoundPoss,BoundObj,andNoBindingconditions,whicharenotsignificantly
differentfromoneanother(p=1).Asschematizedin(28),BaseLinesentenceswere
ratedasmostacceptable(MeanRank=1841.69),followedbysentenceswitha
EXPERIMENT1RESULTSComparative Deletion MultipleQuestions Tough Movement
RATING BL NF BS BO BP NB BL NF BS BO BP NB BL NF BS BO BP NB1 1 7 6 12 12 10 3 8 21 24 25 22 1 5 8 18 27 202 2 10 12 23 19 14 3 14 32 30 30 32 0 12 17 30 17 303 3 12 21 16 26 34 7 14 33 31 35 30 3 15 22 32 40 354 7 15 14 22 25 21 4 28 19 25 20 19 1 19 21 19 25 155 10 24 28 27 17 34 18 39 20 23 22 23 7 33 32 28 24 276 20 35 42 25 37 25 29 36 19 15 13 21 16 42 33 15 11 187 32 47 27 25 14 12 11 11 6 2 5 3 47 24 17 8 6 5
TOTAL 75 150 150 150 150 150 75 150 150 150 150 150 75 150 150 150 150 150
EXPERIMENT2RESULTSComparativeDeletion MultipleQuestions ƚŽŽͬĞŶŽƵŐŚMovement
RATING BL NF BS 1P 2P NB BL NF BS 1P 2P NB BL NF BS 1P 2P NB1 3 5 5 15 17 10 2 6 18 29 30 24 0 7 9 20 23 222 4 6 6 26 26 24 7 9 23 32 38 30 2 11 25 34 28 283 5 16 21 29 22 29 8 30 38 27 32 30 2 18 23 26 25 334 7 25 28 20 34 22 12 29 29 28 22 25 2 17 23 25 23 235 11 23 26 22 16 38 15 33 28 23 21 25 11 32 30 20 32 276 14 39 44 26 24 18 17 30 10 9 5 14 8 35 27 17 15 137 31 36 20 11 9 9 14 13 4 2 2 2 50 30 13 8 4 4
TOTAL 75 150 150 149 148 150 75 150 150 150 150 150 75 150 150 150 150 150
25
nonfiniteembeddedclause(MeanRank=1494.58),followedbysentenceswitha
finiteembeddedclausecontainingaboundpronominalsubject(MeanRank=
1258.58).Atthelowendoftheacceptabilityscalearesentenceswithanembedded
finiteclausecontainingaboundpronominalobject(MeanRank=1064.88),abound
subject-internalpossessor(MeanRank=1024.02),ornoboundpronounatall
(MeanRank=1046.09).Theselatterthreeconditionsgiverisetoratingsthatare
notsignificantlydifferentfromoneanother.
(28)BaseLine>NonFinite>BoundSubj>{BoundObj=BoundPoss=NoBinding}
AppliedtotheresultsofExperiment2,theIndependent-SamplesKruskal-
WallisTestsimilarlyshowsthatthedistributionofratingsisnotthesameacrossthe
differentconditions(𝑋!(5)=349.406,p<0.01).Pairwisecomparisonsshowthat
eachconditiongivesrisetoaratingprofilethatissignificantlydifferentfromeach
othercondition(p<0.01),exceptforthe1pSubj,2pSubj,andNoBindingconditions,
whicharenotsignificantlydifferentfromoneanother(p=1).Asschematizedin
(29),BaseLinesentenceswereratedasmostacceptable(MeanRank=1802.87),
followedbysentenceswithanonfiniteembeddedclause(MeanRank=1534.98),
followedbysentenceswithafiniteembeddedclausecontainingabound
pronominalsubject(MeanRank=1285.35).Atthelowendoftheacceptabilityscale
aresentenceswithanembeddedfiniteclausecontainingafirst-personpronominal
subject(MeanRank=1030.64),asecond-personpronominalsubject(MeanRank=
26
992.41),ornopronounatall(MeanRank=1054.39).Theselatterthreeconditions
giverisetoratingsthatarenotsignificantlydifferentfromoneanother.
(29)BaseLine>Nonfinite>BoundSubj>{1pSubj=2pSubj=NoBinding}
Takentogether,thetwoexperimentssupporttheconclusionthatthebound
pronouneffectisreal(Experiments1and2),thatitissubject-oriented(Experiment
1),andthatnocomparableeffectholdsforfirst-orsecond-personpronominal
subjects(Experiment2).FormoredetailsconcerningtheresultsoftheKruskal-
WallisTest,aswellasindependentconfirmationofthecoreresultsusingamore
powerfulstatisticaltechnique(inparticular,aGeneralizedEstimatingEquations
analysis),seetheAppendix.
3.5Therelationshipbetweenthedataweobserveandthetheorywepursue
ThegoalfortherestofthispaperistodevelopatheorythattreatstheBaseLine,
NonFinite,andBoundSubjsentencesasgrammaticaltotheexclusionofthe
BoundObj,BoundPoss,1pSubj,2pSubj,andNoBindingsentences,assummarizedin
(30).
(30)Summaryofthegrammaticalitycutsmadebythetheorywepursue
a.Grammatical:BaseLine,NonFinite,BoundSubj
b.Ungrammatical:BoundObj,BoundPoss,1pSubj,2pSubj,NoBinding
27
Wepursueatheorythatmakestheseparticularcutsingrammaticalitynotbecause
wethinkthatthesecutscanbereaddirectlyoffthedata(obviously,datacannottell
uswhatisgrammatical,onlywhatismoreorlessacceptable)butratherbecause
thesecutsareconsistentbothwiththeobservedclineofacceptabilityandwithan
independentlyplausibletheoryofphasesandofphi-featurevaluationonbound
pronouns.
Tobesure,(30)isonlyoneofmanyconceivablewaysofconstructinga
theorythatisconsistentwiththeobserveddata,andweleaveittofutureworkto
pursueotherpossibilities.11Onematterinparticularthatourtheorywillnothave
anythingtosayaboutiswhy,amongthosesentencetypesthatourtheorytreatsas
grammatical,thereisaclineofacceptabilitythatcanbecharacterizedas:BaseLine>
NonFinite>BoundSubj.Butwewouldliketosayafewwordsaboutthiscline,
beforeconfiningourselvesintheremainderofthepapertobuildingatheorythat
derives(30).Inparticular,wethinkitisplausiblethattheBaseLine>NonFinite>
BoundSubjclinereflectsdifferencesinprocessingcost.Thereissomeprecedentfor
theideathatacceptabilityjudgmentsareaffectedbyfactorsconcerningthematerial
inbetweentwoelementsinvolvedinadependency,evenwhenthehypothesized
grammaticalityofthedependencyisheldconstant.Forexample,Pickering&Barry
(1991)arguethatthedistancebetweenagapanditsfilleraffectsacceptability.Ina
similarvein,Kluender&Kutas(1993)arguethatwh-movementacrossaclause
boundaryincursaprocessingcost,andSprouse,Wagers&Phillips(2012)showthat
wh-dependenciesthatcrossafiniteclauseboundaryareindeedsomewhatless
28
acceptablethanonesthatdonot.Finally,McElree,Foraker&Dyer(2003)showthat
accuracyinrejectingungrammaticaldependenciesdecreaseswiththenumberof
interveningclauses.Therearestillmanyopenquestionshere:itisnotclearwhether
theeffectsdocumentedintheseworksreflectprocessingcostsassociatedwith
clauseboundariesinparticularorwithmoregeneralfactorssuchaslengthortime.
Butinanyevent,wetaketheseworkstoreinforcetheplausibilityofviewingthe
BaseLine>NonFinite>BoundSubjclineasreflectingdifferingprocessingcosts
amonggrammaticalsentencesratherthanreflectingcutsingrammaticality.
4Aphase-theoreticaccountoftheboundpronouneffect
4.1Theaccountinanutshell
Anattractivefirstapproximationofanaccountoftheboundpronouneffectwould
takethebasicshapein(31).Ifthelocalitydomainforthephenomenaofinterestis
thephase,andboundpronounsenterthederivationinawaythatvoidsthe
otherwisephasalstatusofthecomplementclause,thentheboundpronouneffect
fallsout.
(31) Accountoftheboundpronouneffect(version1of3)
a.Unvaluedphi-featuresvoidphasehood.
b.Thelocalitydomainforthephenomenathatgiverisetothebound
pronouneffectisthephase.
29
c.Boundpronounsenterthederivationwithunvaluedphi-features.12
Butasitstands,(31)overgenerates.Asschematizedinformallyin(32),althoughit
accuratelypredictsthecutbetweenfiniteclauseswithnoboundpronoun(32a)and
finiteclauseswithaboundpronominalsubject(32b),italsoincorrectlyrulesin
caseswheretherelevantfiniteclausehasaboundpronounsomewhereotherthan
subjectposition(32c-d).
(32)Predictionsoftheaccountin(31)
a.Thebookistoodear[forJimtoclaim[PHASEthatMarklent__toBill]].
àcorrectlyruledout
b.Thebookistoodear[forJim1toclaim[NON-PHASEthathe1lent__toBill].
àcorrectlyruledin
c.Thebookistoodear[forJim1toclaim[NON-PHASEthatBilllent__tohim1]].
àincorrectlyruledin
d.Thebookistoodear[forJim1toclaim[NON-PHASEthathis1dadlent__toBill]].
àincorrectlyruledin
Toremedythis,weproposeamodificationto(31)(andwearegratefulforHisa
Kitaharaforsuggestingthisapproachtous):
(33) Accountoftheboundpronouneffect(version2of3)
a.Unvaluedfeaturesontheheadofthecomplementtothephaseheadkeep
30
thephaseopen.
b.Thelocalitydomainforthephenomenathatgiverisetothebound
pronouneffectisthephase.
c.Boundpronounsenterthederivationwithunvaluedphi-features.
By(33),itwillnotdotohaveanunvaluedfeaturejustanywhereinthecandidate
phase;rather,theunvaluedfeaturemustbesufficientlyclosetotheedgeofthe
phase,morespecifically,ontheheadofthecomplementtothephasehead.The
crucialpropertyofthesubjectposition,onthisview,isthatthephi-featuresofthe
subjectvaluethephi-featuresonT(subject-verbagreement).Soifthesubject’sphi-
featuresareunvalued,thentheagreeingphi-featuresonTarenecessarilyalso
unvalued.AndTistheheadofthecomplementtothephaseheadC.Asschematized
in(34),thisrevisedaccountmakesalltherightcuts.
(34)Predictionsoftheaccountin(33)
a.Thebookistoodear[forJimtoclaim[PHASEthatMarklent__toBill]].
àcorrectlyruledout
b.Thebookistoodear[forJim1toclaim[NON-PHASEthathe1lent__toBill].
àcorrectlyruledin
c.Thebookistoodear[forJim1toclaim[PHASEthatBilllent__tohim1]].
àcorrectlyruledout
d.Thebookistoodear[forJim1toclaim[PHASEthathis1dadlent__toBill]].
àcorrectlyruledout
31
Butwearestillnotquitedoneyet,becausealthough(33)makestheright
predictionsforthesentencesofinterest,italsoseverelyundergenerateswhenit
comestomorebasicsentences.Inparticular,wewanttomakesurethatinprinciple
(whennophase-bounddependenciesareatstake),aboundpronouncanbe
arbitrarilydistantfromitsantecedent,asin(35).Butifboundpronounsobligatorily
enterthederivationwithunvaluedphi-featuresandifthesephi-featuresare
eventuallyvaluedwhentheantecedentismergedin,thesimplestviewwouldbe
thatthisvaluationprocedureisitselfphase-bound.Andthenweincorrectlyruleout
(35).
(35)John1said[PHASEthatMarythought[PHASEthatKimsawhim1]].
Soweproposeonefinalrevisiontoouraccount:
(36) Accountoftheboundpronouneffect(version3of3)
a.Unvaluedfeaturesontheheadofthecomplementtothephaseheadkeep
thephaseopen.
b.Thelocalitydomainforthephenomenathatgiverisetothebound
pronouneffectisthephase.
c.Boundpronounsoptionallyenterthederivationwithunvaluedphi-
features.
32
Ifboundpronounshavetheoptionofenteringthederivationwithvaluedphi-
features,thensentenceslike(35)havealicitderivation.
Havingsketchedthebasicgistofouraccount,wenowdiscusseachofthe
threeingredientsin(36)ingreaterformaldetail.Inparticular,wewanttorelate
themtopreviousideasintheliterature,andwewanttoshowthattheycanbe
successfullycombinedbothwithexistingformulationsofthePhaseImpenetrability
Conditionandwithmoreconcreteassumptionsaboutthesyntaxoftherelevant
phenomena.
4.2Phasehoodandfeaturevaluation
Chomsky(2000:107)entertainstwopotentialwaysofdefiningphases:
(37) a.Phasesarepropositional.
b.Phasesareconvergent.
ForChomsky,propositionalobjectsaredefineddisjunctivelyaseither“averb
phraseinwhichallθ-rolesareassigned”(p.106)(i.e.,avP)or“afullclause
includingtenseandforce”(p.106)(i.e.,aCP).13Bycontrast,aconvergentobjectis
onethatitislegibleatallinterfaces(p.95).So,onewayinwhichanobjectcouldfail
tobeconvergentwouldbetocontainunvaluedphi-features.Thepresenceof
unvaluedphi-featureswouldmaketheobjectillegibletothePFinterface,since
33
thesephi-featuresareneededtodeterminethemorphologicalshapeoftheoutput
form.
Theoreticalparsimonyfavorstheviewthatifeitheroftheseapproachesto
phasehoodiscorrect,thenitiscompleteinitself;i.e.,aphraseXPisaphaseifand
onlyifitispropositional(on37a)orifandonlyifitisconvergent(on37b).And
indeed,Chomskyultimatelyarguesinfavorof(37a)andagainst(37b).Butthe
boundpronouneffectanditssubjectorientationleadsustoentertaintheviewthat
propositionalityandconvergencebothneedtoplayaroleindefiningwhataphase
is.14Inparticular,wehypothesizethatthelocalitydomainfordependenciessuchas
too/enoughmovementin(38)isviolatedinvirtueoftheCPstatusofthebracketed
constituenteventhoughthatCPcontainsunvaluedfeatures(ontheboundpronoun
him),whereasthelocalitydomainfortoo/enoughmovementissatisfiedin(39)
becauseeventhoughthebracketedconstituentisaCP,thatCPcontainsunvalued
features(ontheboundpronounhe)thatareinasufficientlylocalconfigurationwith
theedgeofthephasesoastokeepthephaseopen.
(38)*ThisbookistoovaluableforJames1toclaim[thatBilllent__tohim1].
(39)?ThisbookistoovaluableforJames1toclaim[thathe1lent__toBill].
Whatcountsas“sufficientlylocal”?Supposethatacandidatephasedoesnotcount
asaphaseiftheheadofthecomplementtotheheadofthecandidatephasecontains
unvaluedfeatures.Then,thecontrastbetween(38)and(39)followsinvirtueofthe
34
factthatTismergedintothederivationwithunvaluedphi-featuresthatarevalued
inagreementwiththesubject.In(38),therelevantTisvaluedbyBill,sotheCP
countsasaphase.In(39)bycontrast,adoptingthehypothesisthatheentersthe
derivationwithunvaluedphi-features,therelevantTremainsunvalued,sotheCP
doesnotcountasaphase.
Inthefollowingsubsection,weembedthisideaintomoreconcrete
assumptionsabouthowandwhyphasesmatterforthephenomenainquestion.
Beforedoingthis,though,wewouldliketobrieflyconsiderChomsky’s(2000)
empiricalargumentagainstdefiningphasesintermsofconvergenceandexplain
whyitdoesnotunderminethehybridapproachtophasehoodthatwepropose.
Chomsky’sargumentagainsttheconvergenceapproachgoesasfollows.First,
Chomskyconsidersthecontrastinacceptabilitybetween(40a)and(40b)and
appealstoa“MergeoverMove”principletoexplainit:atthepointinthederivation
whenembedded[Spec,TP]isbuilt,thenaïveexpectationwouldbethatthegrammar
canfilliteitherbyraisingmanylinguistsorbymerginginexpletivetherefromthe
lexicalarray.Butapparently,onlythelatteroptionyieldsagrammaticalderivation,
whichfollowsifMergeoverMoveexistsasageneralprinciple.
(40) a.Therearelikely[TPtheretobemanylinguistsatthisconference].
b.*Therearelikely[TPmanylinguiststobemanylinguistsatthisconference].
Butnowconsider(41).Here,atthepointinthederivationwhenembedded
[Spec,TP]isbuilt,raisingmanylinguistsinsteadofmerginginthereyieldsa
35
grammaticalderivation,againsttheexpectationsofMergeoverMove.Thisisnota
problem,however,ifwefurtherassumethatMergeoverMoveappliesonlyover
lexicalsubarrays,andlexicalsubarraysareorganizedintophases.Ifwefurther
supposethattheembeddedCPin(41)isaphase,thenthereisinaseparatelexical
subarrayandisnoteligibletobemergedinattherelevantpointinthederivation.
Consequently,Movedoesnotcompetewithitandthederivationgoesthrough.
(41) Thereissomelikelihood[CPthat[TPmanylinguistswillbemanylinguistsat
thisconference]].
Butifthisisright,then(42)isunderivableonaconvergence-basedapproachto
phasehood,providedthatwh-phrasesenterthederivationwithuninterpretable
features:theuninterpretablefeatureonwhichshouldvoidthephasalstatusofthe
embeddedCP.Consequently,givenMergeoverMove,thereshouldbeforcedto
mergeinat[Spec,TP]ofwill.Butevidentlyitisnotsoforcedbecausethisdoesnot
happenin(42)andyet(42)isgrammatical.Ontheviewthatphasesaredefinedas
vPandCP,ontheotherhand,thereisnoproblemderiving(42),sincetheembedded
CPisphasaldespiteitsuninterpretablefeature,therebypreventingasituation
whereinMergeoverMoveappliesandforcesprematuremergingofthere.
(42) Whichconferenceistheresomelikelihood[CPthatmanylinguistswillbeat
whichconference]?
36
WehavetwothingstosayinresponsetoChomsky’sargument.First,aspointedout
byFelser(2004),theargumentreliesontheviewthatmovementcompeteswith
mergingofthere.But,Felserpointsout,itcouldbethatthereisnotatrueexpletive
andhencethatitsmergesiteisconstrainedthematicallybythechoiceofthe
predicate.Inthiscase,itdoesnotcompetewithmovement,andsotheargument
doesnotgothrough.Second,evenifweassumeforthesakeofargumentthatthere
isatrueexpletive,ourhybridapproachtophasesisimmunetoChomsky’s
argument,becausein(42),theuninterpretablefeatureonwhichhasnobearingon
thestatusofT.TheTheadoftheembeddedCPisfullyvalued,therebyensuringthe
phasalstatusofCP.Consequently,thereisnotintherelevantlexicalsub-array,so
MergeoverMovedoesnotpreventraisingofmanylinguists,andthederivationis
successful.
4.3Phasesaslocalitydomains
In(43),welistallofthephenomenathatwereshowninsections1and2togiverise
totheboundpronouneffect.
(43) a.too/enoughmovement
b.gapping
c.comparativedeletion
d.ACD
e.quantifierscopeinteraction
37
f.multiplequestions
g.pseudogapping
h.reciprocalbinding
i.multiplesluicing
j.familyofquestions
k.extraposition
l.toughmovement
Eachofthesephenomenahasinspiredaliteraturemuchtoovasttodojusticeto
here.Butoneconsistentthemeemerges:allofthesephenomenahavebeenargued
toinvolvesomekindofmovementdependency.See,amongmany,manyothers:
Chomsky1977ontoo/enoughmovementandtoughmovement;Pesetsky1982on
gapping;Kennedy2002oncomparativedeletion;May1985,Hornstein1994and
Kennedy1997onACDandquantifierscopeinteraction;Saito1994andKotek2014
onmultiplequestions;Lasnik2002andGengel2013onpseudogapping;Heim,
Lasnik&May1991onreciprocalbinding;Lasnik2014onmultiplesluicingand
extraposition,andLasnik&Saito1992onfamilyofquestions.
Wetakeitthatthemovementdependenciesinvolvedinthephenomenain
(43)aresubjecttophase-theoreticlocalityconstraints;morespecifically,for
concreteness,weassumethatsomeversionofChomsky’s(2000,2001)Phase
ImpenetrabilityCondition(PIC)asdepictedin(44)holds.Chomsky(2001)
entertainstwovariantsofthePIC,onewhereinthecomplementtoaphaseheadH
becomesinaccessibletosubsequentsyntacticoperationsassoonasHPisbuilt(the
38
so-calledstrongPIC)andonewhereinthecomplementtoHbecomesinaccessible
oncethephraseheadedbythenexthighestphaseheadisbuilt(theso-calledweak
PIC).Foranoverviewofmanyoftheissuesatstakeindecidingbetweenthetwo
variants,seeCitko(2014);wewillultimatelyconcludethatthestrongPICmakesfor
thebestoverallfitwithourtheoreticalaims.
(44)PhaseImpenetrabilityCondition
ThecomplementtoaphaseheadHisnotaccessibletooperations…
a. outsideHP.(strongPIC)
b.atZP(whereZPisheadedbythenexthighestphasehead)(weakPIC)15
(adaptedfromChomsky2001:13-14)
Whatcountsasaphasehead?Inlinewiththeproposalfromtheprevious
subsection,wedefinephaseheadasin(45).
(45)AheadXisaphaseheadiff:
a.Xisacandidatephasehead,and
b.TheheadofthecomplementtoXhasnounvaluedfeatures.
Whatcountsasacandidatephasehead?Wecontrasttwopossibilities,statedin
(46).Ononeview,candidatephaseheadsincludeatleastCandv,whereasonthe
otherview,candidatephaseheadsincludeatleastCbutnotv.SeeLegate(2003)for
39
argumentsinfavorofthephasalstatusofvanddenDikken(2006)forareply.The
viewthatvisnota(candidate)phaseheadisadmittedlynonstandard,butasitturns
out,wewillseebelowthatthisviewfitsthebestwiththerestofourassumptionsto
derivetheboundpronouneffect.Itwillbebeyondthescopeofthispaperto
reconcilethisconclusionwiththeargumentsinthepreviousliteratureinfavorofv’s
phasalstatus,butseenote14belowforasuggestion.Italsobearsnotingthatsince
ourconcernhereiswithclausalsyntax,wetakenostanceonwhatotherkindsof
categories,suchasPorD,mightalsocountascandidatephaseheads;seeCitko
(2014)fordiscussion.
(46)Candidatephaseheadsinclude:
a.Hypothesisa:atleastCandv
b.Hypothesisb:atleastCbutnotv
CrossingthetwovariantsofthePICin(44)withthetwohypothesesabout
theinventoryofcandidatephaseheadsin(46)yieldsthefourtheoretical
possibilitieslistedin(47).
(47)a.StrongPIC;CandidatephaseheadsincludeCandv
b.StrongPIC;CandidatephaseheadsincludeCbutnotv
c.WeakPIC;CandidatephaseheadsincludeCandv
d.WeakPIC;CandidatephaseheadsincludeCbutnotv
40
Inwhatfollows,weassesshowsuccessfuleachofthepossibilitiesin(47)is
ininteractingwithourotherassumptionstoderivetheboundpronouneffect.For
concreteness,wefocusontoo/enoughmovement,andconsidertwocrucialsyntactic
configurations:oneinwhichmovementproceedsfromafiniteclausethatdoesnot
containaboundpronominalsubject(whichwewantourtheorytoruleout)andone
inwhichmovementproceedsfromafiniteclausethatcontainsaboundpronominal
subject(whichwewantourtheorytorulein).Bywayofpreview,wewillconclude
thatfortoo/enoughmovement,(47b)and(47d)succeedbut(47a)and(47c)donot.
Wethenconsiderhowsuccessful(47b)and(47d)areinscalinguptotheother
phenomenathatexemplifytheboundpronouneffect,ultimatelyconcludingthat
(47b)yieldsthebestfitwithourdataandtheory.
Beforeproceeding,twonotesareinorder.First,weintendthisexercisenot
asaforcefulargumentinfavorofaparticularformulationofthePICbutratherasa
“proofofconcept”thatourphase-theoreticaccountoftheboundpronouneffectand
ourviewthatphasesaredefinedinpartbyconvergencearecompatiblewithsome
existingvariantsofthePIC.Weareopentothepossibilitythattherecouldbeother
variantsofthePIC(orevenotherapproachestophasesthatdonotrelyonanything
likethePIC)thatwouldalsointeractwithourotherassumptionsinawaythat
wouldderivethecrucialfacts.Second,weassumeinwhatfollowsthatthe
movementoperationsinquestiondonothavetheoptionofproceedingina
successivecyclicfashion:theelementundergoingmovementmustimmediately
targetitsfinallandingsite.Wewillreturntoandelaborateonthisassumptionafter
wewalkthroughthederivations.
41
Considerfirstthetoo/enoughmovementstructurein(48)thatwewantour
theorytoruleout.Weassumethattoo/enoughstructuresinstantiateA-bar
movementofanoperatorto[Spec,CP]ofthecomplementtotheembedding
predicate(inthiscase,theembeddingpredicateisvaluable).Priortothis
movement,thebracketedportionof(48)hasthestructureindicatedin(49).
(48)*Thisbookistoovaluable[forJohntopromisethatBillwillbuy__].
(49)[CP1C1for[TPJohnTto[vP1v1[VPpromise[CP2C2that[TPBillTwill[vP2v2[VPbuyOp]]]]]]]]
Thequestionthatwenowwanttoaskis:whichofthefourtheoreticalpossibilities
listedin(47)accuratelyrulesout(48)byensuringthatmovementofOpto
[Spec,CP1]in(49)inducesaPICviolation?Inwhatfollows,weconsidereach
possibilityinturn.
a.StrongPIC;CandidatephaseheadsincludeCandv:Onthisview,v2isa
phaseheadanditscomplementbecomesinaccessibleassoonasvP2iscomplete.
Therefore,Opcannottarget[Spec,CP1]andthederivationiscorrectlyruledout.We
schematizethisin(50),where|-----|indicatestheportionofthestructurethat
becomesinaccessibleatthesiteindicatedbytheasterisk(*).
(50)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPBillTwill[vP2v2[VPbuyOp]]]]]]]]
*|----------|
42
b.StrongPIC;CandidatephaseheadsincludeCbutnotv:Onthisview,
thefirstphaseheadencounteredisC2,thecomplementofwhichbecomes
inaccessibleassoonasCP2iscomplete.Asinthepreviousscenario,Opcannottarget
[Spec,CP1]andthederivationiscorrectlyruledout.Thisisschematizedin(51).
(51)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPBillTwill[vP2v2[VPbuyOp]]]]]]]]
* |-------------------------------------|
c.WeakPIC;CandidatephaseheadsincludeCandv:Onthisview,v2isa
phaseheadanditscomplementbecomesinaccessiblewhenthenexthighestphase
head,namelyC2,ismergedin.Asschematizedin(52),Opcannottarget[Spec,CP1]
andthederivationiscorrectlyruledout.
(52)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPBillTwill[vP2v2[VPbuyOp]]]]]]]]
* |------------|
d.WeakPIC;CandidatephaseheadsincludeCbutnotv:Finally,onthe
mostpermissiveview,thefirstphaseheadencounteredisC2,anditscomplement
becomesinaccessibleoncethenexthighestphasehead,namelyC1,ismergedin.As
intheotherthreescenarios,Opcannottarget[Spec,CP1]andthederivationis
correctlyruledout.Thisisschematizedin(53).
(53)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPBillTwill[vP2v2[VPbuyOp]]]]]]]]
* |-------------------------------------|
43
Theinterimconclusion,then,isthatregardlessofwhetherweadoptthe
strongortheweakversionofthePICandregardlessofwhetherwecountCandvor
justvascandidatephaseheads,(48)iscorrectlyruledout.
Considernow,bycontrast,thetoo/enoughmovementsentencein(54),
whichinstantiatestheboundpronouneffectandwhichwewantourtheorytotreat
asgrammatical;i.e.,wewantourtheorytobeabletoderiveitwithoutincurringa
PICviolation.Thebracketedportionof(54)isasindicatedin(55).Therelevant
parseisoneinwhichtheboundpronounentersthederivationwithunvaluedphi-
features,forwhichweusethenotation“pro[ɸ:__]”in(55).
(54)?Thisbookistoovaluable[forJohntopromisethathe1willbuy__].
(55)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPpro[ɸ:__]Twill[vP2v2[VPbuyOp]]]]]]]]
Similarlytowhatwedidforthepreviousstructure,whatwenowwanttodo
isassesswhichofthetheoreticaloptionsin(47)correctlypredictthatOpcantarget
[Spec,CP1]in(55)withoutincurringaPICviolation.
a.StrongPIC;CandidatephaseheadsincludeCandv:Onthisview,v2isa
phasehead,anditscomplementbecomesinaccessibleassoonasvP2iscomplete.
ThereforeOpcannottarget[Spec,CP1]andthestructureisincorrectlyruledout.
Thisisschematizedin(56).
44
(56)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPpro[ɸ:__]Twill[vP2v2[VPbuyOp]]]]]]]]
*|-----------|
b.StrongPIC;CandidatephaseheadsincludeCbutnotv:AlthoughC2isa
candidatephasehead,itsstatusasanactualphaseheadisvoidheresincethe
elementsittinginthelower[Spec,TP]position,andbyextensionTitself,has
unvaluedfeatures.Consequently,thefirstphaseheadencounteredisC1.Its
complementbecomesinaccessibleonceCP1isbuilt,whichiscruciallylateenoughin
thederivationforOptotarget[Spec,CP1].Asschematizedin(57),thestructureis
correctlyruledin.
(57)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPpro[ɸ:__]Twill[vP2v2[VPbuyOp]]]]]]]]
*|---------------------------------------------------------------------------------------------------|
c.WeakPIC;CandidatephaseheadsincludeCandv:Onthisview,the
firstphaseencounteredisv2.Itscomplementbecomesinaccessibleassoonasv1is
mergedin.Therefore,Opcannottarget[Spec,CP1],andthestructureisincorrectly
ruledout.Thisisshownin(58).
(58)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPpro[ɸ:__]Twill[vP2v2[VPbuyOp]]]]]]]]
* |------------|
45
d.WeakPIC;CandidatephaseheadsincludeCbutnotv:Onthisview,the
firstactualphaseheadencounteredisC1,anditscomplementisspelledoutassoon
asthenexthighestphasehead(callitC0)ismergedin.Consequently,Opcantarget
[Spec,CP1]andthestructureiscorrectlyruledin,asschematizedin(59).
(59)C0…[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2that[TPpro[ɸ:__]Twill[vP2v2[VPbuyOp]]]]]]]]
* |-----------------------------------------------------------------------------------------|
Insummary,inorderforthetoo/enoughsentencesthatinstantiatethebound
pronouneffecttosatisfythePICandbecorrectlyruledin,boththestrongandthe
weakvariantsofthePICareviable,butitmustbethecasethatvisnotacandidate
phasehead.16
Itisalsoimportanttoverifythatouranalysiscorrectlypredictsthatcontrol
andraisingclausesextendlocality.Forcontrolclauses,twoanalyticaloptionsare
available.Thefirstistotakethepositionthatcontrolledsubjectsareinstantiatedby
PROandPROisanunvaluedpronoun,àlaKratzer2009;Landau2015,asin(60).
Then,evenifcontrolclauseshaveallthesamephasalpropertiesasfiniteclauses,
localityisextendedinvirtueofthesameproposalsthatsupportedouranalysisof
theboundpronouneffect.
(60)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise[CP2C2[TPPRO[[ɸ:__]Tto[vP2v2[VPbuyOp]]]]]]]]
46
TheotheroptionforcontrolclausesistoadopttheMovementTheoryofControl
(seee.g.Hornstein1999)wherebycontrolledsubjectsdonotharborunvalued
featuresbutratheraretheresidueofmovement.Inthatsituation,itwouldhaveto
bethecasethatcontrolclauseseitherlackCorhaveanon-phasalCinorderto
ensurethatlocalityisextended.Thisissketchedin(61).Onthisview,thereisno
candidatephaseheadbetweenOpandC1,soweaccuratelypredictthattherelevant
movementdependencycanbeestablished.
(61)[CP1C1for[TPJohn[Tto[vP1v1[VPpromise([C2)[TPJohnTto[vP2v2[VPbuyOp]]]]]]]]
Asforraisingclauses,weadoptthestandardviewthattheyprojectonlyTP,
asin(62).Aswiththesecondofthetwoapproachestocontrolclauses,thismeans
thatthereisnocandidatephaseheadbetweenOpandC1,soweagainaccurately
predictthattherelevantmovementdependencycanbeestablished.
(62)[CP1C1for[TPJohn[Tto[vP1v1[VPtend[TPJohnTto[vP2v2[VPbuyOp]]]]]]]
Thenextquestiontoaddressis:willtheanalysissketchedabovefor
too/enoughmovementextendstraightforwardlytotheotherelevenphenomena
thatwehaveidentifiedasgivingrisetotheboundpronouneffect?Sincetheseother
phenomenaseemtopatterninexactlythesamewaywithrespecttothebound
pronouneffect,itisverytemptingtotrytoaccountfortheminthesameway.At
leastsomeofthephenomenafallinlinestraightforwardlywithtoo/enough
47
movementsincetheyalsoinvolveamovementoperationthattargets[Spec,CP];
theseincludetoughmovementandpossiblyalsocomparativedeletion.Asan
anonymousreviewerpointsouttous,however,someoftheotherphenomenahave
beenanalyzedasinvolvingamovementoperationthattargetssomepositionbelow
CP.Gengel(2013),forexample,analyzespseudogappingasmovementofthe
remnanttoa[Spec,FocP]positionbetweenTPandvP,sothatthebracketedportion
of(63)hasastructurelike(64).
(63)Johnlikesapplesand[Billdoes<like>oranges].
(64)[TPBillTdoes[FocPoranges1[vPv[VPliket1]]]]
Ifthisisright,thenthesuccessoftheanalysiswehavejustsketchedinaccurately
predictingthat(65)cannotbegeneratedbut(66)candependsonensuringthatin
(67),orangescannottarget[Spec,FocP]whereasin(68)itcan.Continuingtoassume
thatvisnota(candidate)phasehead,theweakPICenablesorangestotarget
[Spec,FocP]in(67);orangesisembeddedinaCPthatwillnotbecomeinaccessible
untilthenextphaseheadismergedin.ThestrongPIC,ontheotherhand,accurately
rulesoutthemovementin(67),sincetheCPthatembedsorangeswillbecome
inaccessibleassoonasCPisbuilt,beforeFocPentersthederivation.Meanwhile,in
(68),theboundpronounvoidsthephasalstatusofCsothattherearenophase
headsbetweenorangesandFocP,enablingmovement.
48
(65)*JohnclaimsthatMarylikesapplesand[Billdoes<claimthatMarylikes>
oranges].
(66)?John1claimsthathe1likesapplesand[Bill2does<claimthathe2likes>
oranges].
(67)[TPBillTdoes[FocP[vPv[VPclaim[CPthat[TPMary[vP[VPlikesoranges]]]]]]]]
(68)[TPBillTdoes[FocP[vPv[VPclaim[CPthat[TPpro[ɸ:__][vP[VPlikesoranges]]]]]]]]
Wethereforeconcludethatthephase-theoreticaccountsketchedabovefor
too/enoughmovementstructurescanbesuccessfullyextendedtothosephenomena
thatinvolvemovementtoapositionlowerthan[Spec,CP],providedweadoptthe
strongratherthantheweakvariantofthePIC.
Finally,beforemovingon,animportantquestionwhichwealludedtoabove
andwhichstillneedstobeaddressedis:sincemost,ifnotall,ofthephenomenathat
instantiatetheboundpronouneffectinvolveA-barmovement,whycan’talong-
distancedependencybeestablishedinaccordancewiththePICviasuccessivecyclic
A-barmovementthroughintermediate[Spec,CP]positions,asisthecasefor
ordinarywh-movement?17Infact,thisistheoppositeofwhatFelser(2004:547)
callsthetriggeringproblemforwh-movement:
49
(69)TheTriggeringProblem:
Ontheassumptionthatagreement(andhence,movement)istriggeredby
matchingbutuninterpretablefeaturesoftheprobe,whattriggersmovement
ofawh-expressiontothespecifierofintermediatenon-interrogativeheads?
(Felser2004:547)
Seenfromthisperspective,therealpuzzleisnotwhytoo/enoughmovementandthe
otherphenomenathatinstantiatetheboundpronouneffectdisallowsuccessive
cyclicitybutratherwhyordinarywh-movementasfoundinstructuressuchas
interrogativeandrelativeclausesdoesallowit.Andthisfar-reachingpuzzleiswell
beyondthescopeofthispaper.Buthereisonepotentiallyfruitfulpossibilityto
explore.Supposeintermediatenon-interrogativeCheadshaveanoptionalwh-
featurethatattractswh-elementsto[Spec,CP].ThisissimilartoChomsky’s(2000)
proposalthatphaseheadshaveanoptionalEPPfeature.Butifweinsteadtreatthe
relevantoptionalfeatureasawh-feature,thisprovidesabasisfordistinguishingwh-
movementinthestrictsensefromotherkindsofA-barmovement:successive
cyclicityisavailableforthoseelementsundergoingA-barmovementthat
themselveshavematchingwh-features,butitisnotavailableforthekindsof
operatorsandphrasesthatundergomovementinthephenomenathattriggerthe
boundpronouneffect.Thisstrikesusasaplausibleavenuetopursue,thoughit
remainstobeseenwhetheritisultimatelyworkable.Oneissuethatwouldneedto
beworkedoutishowitisdeterminedwhetheramovedelementhaswh-features.
50
Theexistenceofwh-featuresonthemovedelementinawh-questionis
straightforward,asistheexistenceofsuchfeaturesonrelativeoperators,giventhat
theysometimesinvolveanovertwh-constituent.Buttopicalizationseemstopattern
likewh-questionsandrelativeclauseformationinbeingunbounded,despitethe
apparentabsenceofanyindependentevidenceforwh-featuresonthemoved
elementintopicalization.AnotherrelevantconsiderationhastodowithQR,which
hasbeenarguedtoallowforsuccessivecyclicmovement,albeitonlywheneachstep
inthemovementissemanticallymotivated(Fox2000;Cecchetto2004).Hencewe
leavethisasanareaforfurtherinvestigation.18
4.4Phi-featuresonboundpronouns
Finally,werevisitthelastcrucialpieceofourproposal,namelythatbound
pronounshavetheoptionofenteringthederivationwithunvaluedphi-features.By
wayofbackground,considerasentencelike(70)onitsboundvariable
interpretation.Bywhatprincipleis(61)ruledinbutthegendermismatchedvariant
in(71)ruledout?
(70)Everyman1thinksthathe1isagenius.
(71)*Everyman1thinksthatshe1isagenius.
51
Ingeneraltherearetwokindsofapproachesthatcanbeentertained.Ononeview,
schematizedin(72),aboundpronounentersthederivationwithunvaluedphi-
features(72a),andacquiresthosefeaturesviatransmissionfromitsbinderata
laterstageinthederivation(72b).19Sincethebinderinthiscasehasmasculine
gender,thepronounisultimatelyspelledoutasheratherthanshe(72c).Onthis
view,(71)isruledoutbecauseitcanneverbegeneratedinthefirstplace.For
variousversionsofthisview,seeKratzer1998a,2009;Rullmann2004;Heim
2008.20
(72)a.[Everyman][ɸ:3.sg.masc]thinksthatpro[ɸ:__]isagenius
b.[Everyman][ɸ:3.sg.masc]thinksthatpro[ɸ:3.sg.masc]isagenius
|__________________________↑
c.pro[ɸ:3.sg.masc]→he
Onanotherview,boundpronounsenterthederivationwithfullyspecified
phi-features,sothatastructurelike(71)caninprinciplebebuiltbythesyntax.But
itisdeviantbecauseitinducesapresuppositionfailure.Inparticular,supposethat,
justasisoftenassumedforfreepronouns,phi-featuresonboundpronounsactas
presuppositionalfiltersthatrestricttherangeofvaluesthatthevariabledenotedby
thepronouncantake(seeespeciallyCooper1983andHeim2008).Onthisview,
focusingjustongenderfeaturesandignoringpersonandnumberfeatures,shehas
thedenotationin(73).
52
(73)Foranyassignmentfunctiongandindexn:
[[shen]]gisdefinedonlyifg(n)isfemale.Wheredefined,[[shen]]g=g(n)
(cf.Heim2008:36)
ThispresuppositionprojectsupthroughthestructuresothatthematrixVPendsup
denotingthepartialfunctionin(74)whosedomainisrestrictedtothesetof
females.21
(74)[[1t1thinksthatshe1isagenius]]g=λx:xisfemale.xthinksthatxisagenius
(cf.Heim2008:38)
FollowingHeim(2008:39)inassumingthateverycomeswithitsown
presupposition,namelythatthesetassociatedwithitsNPargumentisasubsetof
thedomainassociatedwithitsVPargument,asin(75),thesentencein(71)endsup
presupposingthatallmenarefemale,asin(76).Thisfaultypresuppositionthen
accountsfortheperceiveddevianceofthesentence.
(75)[[every]]=λPλQ:{x:P(x)=1}⊆dom(Q).{x:P(x)=1}⊆{x:Q(x)=1}
(Heim2008:39)
53
(76)[[Everyman1t1thinksthatshe1isagenius]]gisdefinedonlyifthesetofmen
isasubsetofthesetoffemales.Wheredefined,…
Aswasthecaseforthetwopotentialdefinitionsofphasesentertainedin
section4.2above(phasesaspropositionalobjectsvs.phasesasconvergentobjects),
theoreticalparsimonyfavorstheviewthatifeitheroftheseapproachestobound
pronounsiscorrect,itiscorrectinallcasesandtheotheroneisalwaysincorrect.
Butaswasthecasewithphases,wethinkthattheboundpronouneffectpoints
towardtheviewthatbothoftheseapproachesarecorrect:inprinciple,pronouns
havetheoptionofenteringthederivationeitherwithorwithoutphi-features.22If
theyenterthederivationwithoutphi-features,thentheyhavetobeboundsothat
theirfeaturescanbedetermined(cf.Kratzer’s2009:195FeatureTransmission
underBinding).Butcrucially,bindingisalsoconsistentwithaconfigurationin
whichthepronounentersthederivationwithitsphi-featuresalreadyvaluedand
theappearanceofphi-featureagreementisachievedviatheworkingsof
presuppositionprojection.Weneedtheformeroptionaspartofouraccountofthe
boundpronouneffect,andweneedthelatteroptioninordertoensurethat
structureslike(77)canbebuilt.In(77),theboundpronounisseparatedfromits
antecedentbyatleasttwophaseheads,andsoshouldnotbeaccessibleforfeature
transmission.23Instead,wegetwhatlookslikelong-distancephi-feature
“agreement”asaconsequenceofthefactthatpresuppositionprojectionisnot
subjecttothePIC.
54
(77)Everyman1thinks[CPthatAnnsaid[CPthatMarysawhim1]]
5Islands
Beforeconcluding,weoffersomepreliminaryremarksinthissectiononthe
relevanceoftheboundpronouneffecttoislandphenomena.Inparticular,forat
leastsomeislandtypesincludingadjunctislandsandwh-islands,weseethesame
clineofacceptabilityfamiliarfromthephenomenawefocusedoninthispaper:
extractionoutofanonfiniteclauseisfairlyacceptable(78a/79a),extractionoutofa
finiteclausewithaboundpronominalsubjectissomewhatdegraded(78b/79b),
andextractionoutofafiniteclausewithnoboundpronominalsubjectisthemost
degraded(78c/79c).24
(78) a.What2didJohn1gohome[afterPRO1readingt2]?
b.?What2didJohn1gohome[afterhe1readt2]?
c.*What2didJohngohome[afterMaryreadt2]?
(79) a.What2didJohn1wonder[whetherPRO1toreadt2]?
b.?What2didJohn1wonder[whetherhe1shouldreadt2]?
c.*What2didJohnwonder[whetherBillshouldreadt2]?
Inthisconnection,itisinterestingtonotethatRoss(1967)questioned
Chomsky’s(1964)wh-islandconstraintonthebasisthatitwastoostrong,andthe
55
dataRossofferedtosupportthispositioninvolvedcontrolledinfinitivalembedded
questions(80a-d)aswellassixexamplesofembeddedfinitequestionswithbound
pronominalsubjects(80e-g,81).(AlthoughRossdidnotactuallysaythatthese
pronounswerebound,thisissurelytheintendedinterpretation,sincethereisno
contexttosupportafreereading.)Thedatain(80)-(81)aretakenfromRoss
1967:27withthejudgmentmarksastheyappearintheoriginal.
(80)HetoldmeaboutabookwhichIcan’tfigureout
a.whethertobuyornot.
b.howtoread.
c.wheretoobtain.
d.whattodoabout.
e.whyheread.
f.?whetherIshouldread.
g.??whenIshouldread.
(81)Whichbooksdidhetellyou
a.whyhewantedtoread?
b.?whetherhewantedtoread?
c.??whenhewantedtoread?
Rossnotedthatextractionoutofinfinitivalembeddedquestionsseemedtobemore
acceptablethanextractionoutoffiniteembeddedquestions.Healsonoted
56
regardinghisexamplesofextractionoutofafiniteembeddedquestionthat“there
aremanysentenceswhichdifferinnowaywhichIcandiscernfromthosein[80e-g,
81]butwhichIfindtotallyunacceptable.(Chomsky’sexample,‘*whatdidhe
wonderwhereJohnput?’isagoodcaseinpoint)”(p.27).Theboundpronouneffect
suggestsasolutiontoRoss’spuzzle:(80e-g,81)allcontainaboundpronominal
subject.
Wehypothesizethattheboundpronouneffectasmanifestinislandsis
amenabletothesamekindofphase-theoreticproposalweadvancedforthecore
casesconsideredinthispaper.Inparticular,supposethatwhatmakesaclausean
islandforextractionisthatithassomepropertythatdisablesmovementtoitsedge.
Thisisofcoursetheclassictreatmentofwh-islands,andmayalsoextendtoatleast
someadjunctislandsifweadoptLarson’s(1990)proposalthatsomeadjunct-
introducingprepositionslikebeforeandaftercombinewithCPcomplementswhose
Specpositionisfilledbyanoperator.If[Spec,CP]isalreadyfilled,thenitisnot
availableasanintermediatelandingsite.Consequently,iftheCPisaphase,then
extractionwillincuraPICviolation.Butifaboundpronominalsubjectvoidsthe
phasalstatusofCP,aswehaveproposed,thenextractioncanproceedwithoutthe
needforanintermediatelandingsite.
Thatbeingsaid,asDavidPesetsky(pers.comm.)remindsus,itremainsthe
casethatextractionofadjunctsoutofislandsisrobustlyungrammatical,regardless
ofthestatusoftheembeddedsubject,asillustratedin(82).Consequently,weleave
forfutureworkamorecompleteinvestigationofboundpronounsinislands.25
57
(82) a.*How2didJohn1gohome[afterPRO1solvingtheproblemt2]?
b.*How2didJohn1gohome[afterhe1solvedtheproblemt2]?
c.*How2didJohngohome[afterMarysolvedtheproblemt2]?
6Conclusions
Thispaperbeganwiththeobservationthataboundpronouninthesubjectposition
ofafiniteembeddedclauserenderstheclauseboundaryrelativelytransparentto
syntacticprocessesandrelationsordinarilyconfinedtomonoclausal,control,and
raisingconfigurations.Weshowedthatthiseffectholdsforawiderangeof“quasi-
clause-bound”phenomenaincludingtoo/enoughmovement,gapping,comparative
deletion,ACD,quantifierscopeinteraction,multiplequestions,pseudogapping,
reciprocalbinding,multiplesluicing,familyofquestions,extraposition,andtough
movement.Andwedocumentedtheeffectexperimentallyfortoo/enough
movement,comparativedeletion,andmultiplequestions.
Towardanexplanation,wesuggestedthattherelevantlocalitydomainforall
ofthesephenomenaisthephase,andthatboundpronounshavetheoptionof
enteringthederivationwithunvaluedphi-features,therebyvoidingphasehood.
Thisbasicpictureiscomplicatedbythefactthattheboundpronounmustbein
subjectpositioninordertoextendthelocalitydomain,andinresponsetothiswe
entertainedtheviewthatonlyunvaluedfeaturesthatstandinasufficientlylocal
relationshiptothephasehead(inparticular,ahead-to-headrelationship)void
phasehood.
58
Thisaccounthastwoprimarytheoreticalimplications.Thefirstisthatnotall
boundpronounsarecreatedequal:boundpronounscaneitherenterthederivation
phi-complete,orenterthederivationunvaluedandtherebyinteractwithcore
grammaticalprocesses.ThisconclusionechoesChomsky’s(1955/1975)treatment
ofthird-personpronounsaswellasmorerecentworkonboundpronounsby
Kratzer(2009)(seenote20).Thesecondtheoreticalimplicationisthatnotallfinite
CPsarecreatedequal,specificallywithrespecttotheirphasalstatus.Thebound
pronouneffectoffersnovelevidencefortheviewthatfeaturevaluationhasaroleto
playinphasetheory.
ReferencesAgüero-Bautista, Calixto. 2007. Diagnosing cyclicity in sluicing. Linguistic Inquiry
38:413–443.
Aissen, Judith, and David Perlmutter. 1983. Postscript to republication of “Clause
reduction in Spanish”. In Studies inRelationalGrammar, ed. David Perlmutter,
383–396.Chicago:UniversityofChicagoPress.
Ariel,Mira.1988.Referringandaccessibility.JournalofLinguistics24:65-87.
Barrie,Michael.2008.Controlandwh-infinitivals.InNewhorizonsintheanalysisof
control and raising, ed. William D. Davies and Stanley Dubinsky, 263– 279.
Dordrecht:Springer.
Cable, Seth. 2005. Binding local person pronouns without semantically empty
features.Ms.,MIT.
Cecchetto, Carlo. 2004.Explaining the locality conditionsofQR:Consequences for
thetheoryofphases.NaturalLanguageSemantics12:345-397.
59
Chomsky,Noam.1964.Currentissuesinlinguistictheory.TheHague:Mouton.
Chomsky, Noam. 1973. Conditions on transformations. In A festschrift for Morris
Halle, ed. Stephen Anderson and Paul Kiparsky, 232–286. New York: Holt,
RinehartandWinston.
Chomsky,Noam.1975.Thelogicalstructureoflinguistictheory.NewYork:Plenum.
(Originalworkpublished1955)
Chomsky,Noam.1977.Onwh-movement.InFormalsyntax,ed.PeterCulicover,Tom
Wasow,andAdrianAkmajian,71–132.NewYork:AcademicPress.
Chomsky, Noam. 1981. Lectures on Government and Binding: The Pisa Lectures.
Holland:ForisPublications.
Chomsky,Noam.1986.Barriers.Cambridge:MITPress.
Chomsky, Noam. 2000. Minimalist inquiries. In Step by step: Essays onminimalist
syntax in honor ofHoward Lasnik, ed. Roger Martin, David Michaels, and Juan
Uriagereka,89–155.Cambridge:MITPress.
Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A life in language, ed.
MichaelKenstowicz,1–52.Cambridge:MITPress.
Cinque,Guglielmo.2004.‘Restructuring’andfunctionalstructure.InStructuresand
beyond: The cartography of syntactic structures, volume 3, ed. Adriana Belletti,
132–191.NewYork:OxfordUniversityPress.
Citko, Barbara. 2014. Phase Theory: An Introduction. Cambridge: Cambridge
UniversityPress.
Cooper,Robin.1983.QuantificationandSyntacticTheory.Dordrecht:Reidel.
60
denDikken,Marcel.2006.AreappraisalofvPbeingphasal—AreplytoLegate.ms.,
CUNYGraduateCenter.
Erlewine, Michael and Hadas Kotek. 2016. A streamlined approach to online
linguisticsurveys.NaturalLanguage&LinguisticTheory34:481-495.
Farkas, Donka F., and Anastasia Giannakidou. 1996. How clause-bounded is the
scopeofuniversals? InSALTVI, ed.TeresaGallowayand JustinSpence,35–52.
Ithaca,NY:CornellUniversity.
Felser,Claudia.2004.Wh-copying,phases,andsuccessivecyclicity.Lingua114:543–
574.
Fiengo,Robert,C.-T. JamesHuang,HowardLasnik,andTanyaReinhart.1988.The
syntax of Wh-in-situ. In Proceedings of the Seventh West Coast Conference on
FormalLinguistics,ed.HagitBorer,81-98.
Fox,Danny.2000.Economyandsemanticinterpretation.Cambridge:MITPress.
Frank, Robert and Dennis Storoshenko. 2015. Experiencing scope: Inverted
expectationsofQRinraising.PaperpresentedatLinguisticSocietyofAmerican
AnnualMeeting,Portland,OR,1/8/15.
Gengel,Kirsten.2013.Pseudogappingandellipsis.Oxford:OxfordUniversityPress.
Goodall, Grant. 1991. Wanna-contraction as restructuring. In Interdisciplinary
approachestolanguage:EssaysinhonorofS.-Y.Kuroda, ed.CarolGeorgopoulos
andRobertaIshihara,239–254.Dordrecht:Kluwer.
Grano,Thomas.2015.Controlandrestructuring.Oxford:OxfordUniversityPress.
Hackl,Martin,JorieKoster-HaleandJasonVarvoutis.2012.QuantificationandACD:
Evidencefromreal-timesentenceprocessing.JournalofSemantics29:145-206.
61
Heim, Irene. 2008. Features onboundpronouns. InPhi theory:Phi-featuresacross
modulesandinterfaces,ed.DanielHarbour,DavidAdger,andSusanaBéjar,35–
56.Oxford:OxfordUniversityPress.
Heim,IreneandAngelikaKratzer.1998.SemanticsinGenerativeGrammar.Malden,
MA;Blackwell.
Heim, Irene, Howard Lasnik, and Robert May. 1991. Reciprocity and plurality.
LinguisticInquiry22:63-101.
Higginbotham,James.1981.Reciprocalinterpretation.JournalofLinguisticResearch
1:97–117.
Hornstein, Norbert. 1994. An argument for minimalism: The case of antecedent-
containeddeletion.LinguisticInquiry25:455–480.
Hornstein,Norbert.1999.Movementandcontrol.LinguisticInquiry30:69-96.
Huang,C.-T.James.1982a.Movewhinalanguagewithoutwh-movement.The
LinguisticReview1:369-416.
Huang,C.-T.James1982b.LogicalrelationsinChineseandthetheoryofgrammar.
PhDdissertation,MIT.
Jacobson,Pauline.2012.Directcompositionalityand“uninterpretability”:Thecase
of (sometimes) “uninterpretable” features on pronouns. Journal of Semantics
29:305-343.
Johnson, Kyle. 1996. In search of the English middle field. Ms., University of
Massachusetts,Amherst.
Kayne,RichardS.1998.Overtvs.covertmovement.Syntax1:128–191.
62
Kennedy, Christopher. 1997. Antecedent-contained deletion and the syntax of
quantification.LinguisticInquiry28:662–688.
Kennedy,Christopher.2002.Comparativedeletionandoptimalityinsyntax.Natural
Language&LinguisticTheory20:553-621.
Kenward,MichaelG.,EmmanuelLesaffre,andGeertMolenberghs.Anapplicationof
maximum likelihood and generalized estimating equations to the analysis of
ordinaldatafromalongitudinalstudywithcasesmissingatrandom.Biometrics
50:945-953.
Kluender,RobertandMartaKutas.1993.Subjacencyasaprocessingphenomenon.
LanguageandCognitiveProcesses8:573-633.
Kotek,Hadas.2014.Composingquestions.PhDdissertation,MIT.
Kratzer,Angelika.1998a.Morestructuralanalogiesbetweenpronounsandtenses.
InProceedingsofSemanticsandLinguisticTheoryVIII,ed.DevronStrolovitchand
AaronLawson,92–109.CornellUniversity:CLCPublications.
Kratzer,Angelika.1998b.Scopeorpseudoscope?Aretherewide-scopeindefinites?
In Events and Grammar, ed. Susan Rothstein, 163–196. Kluwer Academic
Publishers.
Kratzer, Angelika. 2009. Making a pronoun: Fake indexicals as windows into the
propertiesofpronouns.LinguisticInquiry40:187–237.
Kuno,SusumuandJaneJ.Robinson.1972.MultipleWhQuestions.LinguisticInquiry
3:463-487.
Landau,Idan.2015.Atwo-tieredtheoryofcontrol.Cambridge:MITPress.
63
Landau, Idan. 2016. Agreement at PF: An argument from Partial Control. Syntax
19:79-109.
Larson, Richard K. 1990. Extraction and multiple selection in PP. The Linguistic
Review7:169-182.
Larson, Richard K., and Robert May. 1990. Antecedent containment or vacuous
movement:ReplytoBaltin.LinguisticInquiry21:103–122.
Lasnik,Howard.2002.Clause-mateconditionsrevisited.GlotInternational6:94–96.
Lasnik, Howard. 2002. On repair by ellipsis. In Proceedings of the 2002 LSK
InternationalSummerConference,vol.1,Forumlecturesandpaperpresentations,
23–36.Seoul:I.ThaaehaksaPublishers,KyungHeeUniversity.
Lasnik,Howard.2006.Afamilyofquestions.Handout,USC.
Lasnik,Howard.2014.MultiplesluicinginEnglish?Syntax17:1–20.
Lasnik,HowardandMamoruSaito.1984.Onthenatureofpropergovernment.
LinguisticInquiry15:235–289.
Lasnik,HowardandMamoruSaito.1992.Moveα:ConditionsonItsApplicationand
Output.Cambridge:MITPress.
Lasnik, Howard and Juan Uriagereka. 2005. A course in minimalist syntax:
Foundationsandprospects.Malden,MA:Wiley-Blackwell.
Lechner,Winfried. 2001. Reduced and phrasal comparatives.NaturalLanguage&
LinguisticTheory19:683–735.
Lee-Schoenfeld, Vera. 2007. Beyond coherence: The syntax of opacity in German.
Amsterdam:JohnBenjaminsPublishingCompany.
Lee-Schoenfeld,Vera.2008.Binding,phases,andlocality.Syntax11:281-298.
64
Legate, Julie. 2003. Some interface properties of the phase. Linguistic Inquiry
34:506-515.
May,Robert.1977.Thegrammarofquantification.PhDdissertation,MIT.
May,Robert.1985.Logicalform:Itsstructureandderivation.Cambridge:MITPress.
McElree,Brian,StephaniForaker,andLisbethDyer.2003.Memorystructures that
subservesentencecomprehension.JournalofMemoryandLanguage48:67-91.
Merchant, Jason. 2001. The syntax of silence: Sluicing, islands, and the theory of
ellipsis.Oxford:OxfordUniversityPress.
Moulton,Keir.2008.Clausalcomplementationandthewager-class.InProceedingsof
NELS38,ed.AnisaSchardl,MartinWalkow,andMuhammadAbdurrahman,165–
178.Amherst,MA:GLSAPublications.
Nishigauchi,Taisuke.1998.‘Multiplesluicing’inJapaneseandthefunctionalnature
ofwh-phrases.JournalofEastAsianLinguistics7:121–152.
Partee,BarbaraH.1989.Bindingimplicitvariablesinquantifiedcontexts.InCLS25.
Part one, the general session, ed. Caroline Wiltshire, Randolph Graczyk, and
BradleyMusic,342–365.Chicago:ChicagoLinguisticSociety.
Pesetsky,David.1982.PathsandCategories.PhDdissertation,MIT.
Pickering, M., & Barry, G. (1991). Sentence processing without empty categories.
Languageandcognitiveprocesses6:229-259.
Postal, Paul. 1974. On raising: One rule of English grammar and its theoretical
implications.Cambridge,MA:MITPress.
Reuland, Eric. 2010. Minimal versus not so minimal pronouns: Feature
transmission, featuredeletionandtheroleofeconomyinthe languagesystem.
65
In The Linguistics Enterprise: From knowledge of language to knowledge in
linguistics, ed. Martin B.H. Everaert, Tom Lentz, Hannah de Mulder, Øystein
Nilsen,andArjenZondervan,257–282.Amsterdam:JohnBenjamins.
Rizzi,Luigi.1978.Arestructuringrule inItaliansyntax. InRecenttransformational
studies inEuropean languages, ed. Samuel J. Keyser, 113–158. Cambridge:MIT
Press.
Rizzi, Luigi. 1982. Violations of the wh-island constraint and the subjacency
condition.InIssuesinItalianSyntax,ed.L.Rizzi,49–76.Dordrecht:Foris.
Ross,JohnR.1967.Constraintsonvariablesinsyntax.PhDdissertation,MIT.
Rullmann, Hotze. 2004. First and second person pronouns as bound variables.
LinguisticInquiry35:159–168.
Ruys,E.G.1992.Thescopeofindefinites.PhDdissertation,UtrechtUniversity.
Saito,Mamoru.1994.AdditionalWHeffectsandtheadjunctionsitetheory.Journal
ofEastAsianLinguistics3:195-240.
Sheskin, David. 2003. Handbook of parametric and nonparametric statistical
procedures.BocaRaton,FL:Chapman&Hall/CRC.
Sloan, Kelly. 1991. Quantifier-wh interaction. MIT Working Papers in Linguistics
15:219–237.
Sloan, Kelly, and Juan Uriagereka. 1998. What does ‘everyone’ have scope over?
GLOW.Budapest.
Spathas,Georgios.2010.FocusonAnaphora.Utrecht:LOT.
Sprouse, Jon. 2011. A test of the cognitive assumptions of magnitude estimation:
Commutativitydoesnotholdforacceptabilityjudgments.Language87:274-288.
66
Sprouse, Jon,MattWagers,andColinPhillips.2012.A testof therelationbetween
workingmemorycapacityandsyntacticislandeffects.Language88:82-123.
vonStechow,Arnim.2003.Featuredeletionundersemanticbinding.InProceedings
oftheNorthEastLinguisticSociety(NELS)33,ed.MakotoKadowakiandShigeto
Kawahara,377–403Amherst:UniversityofMassachusetts,GLSA.
Sudo,Yasutada.2012.Onthesemanticsofphifeaturesonpronouns.PhDdissertation,
MIT.
Syrett,KristenandJeffreyLidz.2011.Competence,performance,andthelocalityof
quantifierraising:Evidencefrom4-year-oldchildren.LinguisticInquiry42:305-
337.
White, Aaron Steven, and Thomas Grano. 2014. An experimental investigation of
partial control. InProceedingsof SinnundBedeutung18, eds. Urtzi Etxeberria,
AnamariaFălăuş,AritzIrurtzun,andBryanLeferman,469–486.
Wilder, Chris. 1997. Phrasal movement in LF: de re readings, VP-ellipsis and
binding.InProceedingsoftheNorthEastLinguisticSocietyAnnualMeeting27,ed.
byKiyomiKusumoto,425–439.Amherst:UniversityofMassachusetts,GLSA.
Williams, Edwin. 1986. A reassignment of the function of LF. Linguistic Inquiry
17:265–299.
Wurmbrand, Susi. 2001. Infinitives: Restructuring and clause structure. Berlin:
MoutondeGruyter.
Wurmbrand,Susi.2011.Onagreeandmerge.RevisedcoursenotesfromProblemsin
Syntax(Spring2011),UniversityofConnecticut,accessed5/17/13.
67
Wurmbrand, Susi. 2013. QR and selection: Covert evidence for phasehood. In
ProceedingsoftheNorthEasternLinguisticsSocietyAnnualMeeting42,ed.Stefan
Keine and Shayne Sloggett, 277–290. Amherst: University of Massachusetts,
GLSA.
Wurmbrand, Susi. 2015a. The cost of raising quantifiers. Ms., University of
Connecticut.
Wurmbrand, Susi. 2015b. Restructuring cross-linguistically. In Proceedings of the
North Eastern Linguistic Society Annual Meeting 45, ed. Thuy Bui and Deniz
Özyıldız,227-240.Amherst:UniversityofMassachusetts,GLSA.
(Grano)
DepartmentofLinguistics
IndianaUniversity
BallantineHall844
1020E.KirkwoodAvenue
Bloomington,IN47405
(Lasnik)
DepartmentofLinguistics
UniversityofMaryland
1401MarieMountHall
68
CollegePark,MD20742
Appendix:StatisticalanalysisofexperimentalresultsInthisappendix,wedescribeingreaterdetailthestatisticalanalysisofthe
experimentalinvestigationdescribedinsection3.4above.Inselectingtestsfor
statisticalanalysis,weassumefollowingSprouse(2011)andothersthatsentence
acceptabilityjudgmentsdonotnecessarilyconformtoaratioscale;thatis,we
assumethatparticipantstreatthesevenpointsontheLikertscaleasdefininga
ranking,butwedonotassumethatthedifferencebetweenaratingof2andarating
of3,forexample,isthesameasthedifferencebetweenaratingof3andaratingof
4.Thismeansthattheresultingdatahavetobetreatedasordinaldataratherthan
asratio-scaledata.
TheinputtothestatisticalanalysisforExperiment1is2,475testsentences
ratedonascaleof1to7.Wetreattheratingasthedependentvariable.Eachofthe
2,475sentencesiscodedfortwofactorsthatconstitutetheindependentvariables.
Thephenomenonfactorconsistsofthethreecategorieslistedin(A1a)andthe
conditionfactorconsistsofthesixcategorieslistedin(A1b).
(A1)a.Phenomenon:ComparativeDeletion,MultipleQuestions,too/enough
Movement
b.Condition:BaseLine,NonFinite,BoundSubj,BoundObj,BoundPoss,NoBinding
69
ThefirsttestweemployisanIndependent-SamplesKruskal-WallisTest,a
rank-basednonparametrictestsimilartoaone-wayANOVAbutappropriatefor
ordinal(non-ratio-scale)data(seeSheskin2003).Thistestallowsustodetermine
whetherornotthedistributionofsentenceratingsisthesameacrossthedifferent
categoriesofachosenfactor.Torunthistestandalltheotherstatisticaltests
describedinwhatfollows,weuseIBMSPSSStatisticsVersion24.
Appliedtothephenomenonfactor,theKruskal-WallisTestindicatesthatthe
distributionofratingsisnotthesameacrossthedifferentcategoriesofthe
phenomenonfactor(𝑋!(2)=107.130,p<0.01).Furthermore,pairwisecomparisons
revealthateachphenomenongivesrisetoaratingprofilethatissignificantly
differentfromeachotherphenomenon.Thesepairwisecomparisonsareshownin
TableA1,withsignificancevaluesadjustedbytheBonferronicorrectionformultiple
tests.TakentogetherwiththemeanrankforeachphenomenonindicatedinTable
A2,thisanalysissupportstheconclusionthatthethreephenomenainvestigatedin
theexperimentconformtotheacceptabilityclinein(A2):onthewhole,comparative
deletionsentences(MeanRank=1422.17)wereratedhigher(p<0.01)than
too/enoughmovementsentences(MeanRank=1229.29),whichwereinturnrated
higher(p<0.01)thanmultiplequestions(MeanRank=1062.54.Whileinteresting
andworthyoffurtherstudy,wetakethisresulttobeorthogonaltoourmain
purpose,whichistoestablishhowratingsvaryasafunctionoftheconditionfactor.
Thatbeingsaid,theclinethatemergesheremaybearaninterestingcross-linguistic
connectiontoRizzi’s(1978)claimthatItaliandoesnotallowmultiplequestions.
70
(A2)ComparativeDeletion>too/enoughMovement>MultipleQuestions
Sample1 - Sample2 Test Statistic
Standard Error
Standard Test Statistic Significance
Adjusted Significance
Multiple Questions – too/enough Movement
-166.750 34.776 -4.795 .000 .000***
Multiple Questions - Comparative Deletion
359.627 34.776 10.341 .000 .000***
too/enough Movement - Comparative Deletion
192.877 34.776 5.546 .000 .000***
TableA1:Experiment1pairwisecomparisonsofphenomena
Phenomenon Mean Rank Comparative Deletion 1,422.17 too/enough Movement 1,229.29 Multiple Questions 1,062.54
TableA2:Experiment1meanranksforphenomena
Appliedtotheconditionfactor,theKruskal-WallisTestindicatesthatthe
distributionofratingsisnotthesameacrossthedifferentconditions(𝑋!(5)=
325.701,p<0.01).Pairwisecomparisonsrevealthateachconditiongivesrisetoa
ratingprofilesignificantlydifferentfromeachothercondition(p<0.01),exceptfor
theBoundPoss,BoundObj,andNoBindingconditionswhicharenotsignificantly
differentfromoneanother(p=1).ThesepairwisecomparisonsareshowninTable
A3.Takentogetherwiththemeanranksforeachcondition(TableA4),thisanalysis
supportstheconclusionthatthesixconditionsinvestigatedinExperiment1
conformtotheclineofacceptabilityindicatedin(A3):BaseLinesentenceswere
ratedasmostacceptable(MeanRank=1841.69),followedbysentenceswitha
nonfiniteembeddedclause(MeanRank=1494.58),followedbysentenceswitha
finiteembeddedclausecontainingaboundpronominalsubject(MeanRank=
1258.58).Atthelowendaresentenceswithanembeddedfiniteclausecontaininga
71
boundpronominalobject(MeanRank=1064.88),aboundsubject-internal
possessor(MeanRank=1024.02),ornoboundpronoun(MeanRank=1046.09).
Thesethreegiverisetoratingsnotsignificantlydifferentfromoneanother.
(A3)BaseLine>NonFinite>BoundSubject>{BoundObj=NoBinding=BoundPoss}
Sample1 - Sample2 Test Statistic
Standard Error
Standard Test Statistic Significance
Adjusted Significance
BoundPoss-NoBinding -22.074 47.087 -.469 .639 1BoundPoss-BoundObj 40.861 47.087 .868 .386 1BoundPoss-BoundSubj -234.564 47.087 -4.982 .000 .000***BoundPoss-NonFinite -470.560 47.087 -9.993 .000 .000***BoundPoss-BaseLine 817.672 57.669 14.179 .000 .000***NoBinding-BoundObj 18.787 47.087 .399 .690 1NoBinding-BoundSubj 212.490 47.087 4.513 .000 .000***NoBinding-NonFinite -448.486 47.087 -9.525 .000 .000***NoBinding-BaseLine 795.598 57.669 13.796 .000 .000***BoundObj-BoundSubj -193.703 47.087 -4.114 .000 .000***BoundObj-NonFinite -429.699 47.087 -9.126 .000 .000***BoundObj-BaseLine 776.811 57.669 13.470 .000 .000***BoundSubj-NonFinite -235.996 47.087 -5.012 .000 .000***BoundSubj-BaseLine 583.108 57.669 10.111 .000 .000***NonFinite-BaseLine 347.112 57.669 6.019 .000 .000***
TableA3:Experiment1pairwisecomparisonsofconditions
Condition Mean Rank BaseLine 1,841.69 NonFinite 1,494.58 BoundSubj 1,258.58 BoundObj 1,064.88 NoBinding 1,046.09 BoundPoss 1,024.02
TableA4:Experiment1meanranksforconditions
AlimitationoftheKruskal-WallisTestisthatitonlyallowsustotestone
factoratatime:phenomenonorcondition.Toremedythis,weemployamore
powerfulstatisticaltechnique:aGeneralizedEstimatingEquations(GEE)analysis.
72
GEEisatechniqueappropriateforordinaldatawithmultipleindependent
variables,similartoageneralizedmultiplelinearregressionbutdifferentinthatit
requiresfewerassumptionsaboutthedataanditmodelspopulationaverages
ratherthanyieldingsubject-specificestimates(seee.g.Kenward,Lesaffreand
Molenberghs1994foradiscussionofGEEinthecontextofapsychiatricstudy).
AppliedtothedatainExperiment1,GEEyieldstheresultsindicatedinTableA5.
OfmostrelevancetousaretherowslabeledA-CantherowslabeledD-I
respectively.LookingfirstattherowslabeledA-C,thetoo/enoughMovement
categoryinrowCis(arbitrarily)selectedasabaseline,andtheBcolumnshowsthe
increaseinlogoddsfortheothercategoriesinthisfactor,namelyMultiple
QuestionsandComparativeDeletion,yieldingaratingthatishigherthantherating
foratoo/enoughMovementsentence.TheExp(B)columntranslatesthisfigureinto
anoddsratio:oddsratiosthataregreaterthan1indicateanincreasedlikelihoodof
ahigherratingwhereasratioslessthan1indicateadecreasedlikelihoodofahigher
rating.Hence,weseeconfirmationoftheconclusionfromthepairwisecomparisons
thatratingsfortoo/enoughMovementsentencesaresignificantlyhigherinodds
ratiothanratingsforMultipleQuestionssentences(Exp(B)=0.618,p<0.01)and
significantlylowerinoddsratio(0.62)thanratingsforComparativeDeletion
sentences(Exp(B)=1.570,p<0.01).
TurningtotherowslabeledD-I,theNoBindingconditionis(arbitrarily)
selectedasabaseline,andtheExp(B)columnindicatestheoddsratioforeachofthe
otherconditionsinyieldingaratingthatishigherthanthatforNoBinding.Wesee,
alsoconsistentwiththepairwisecomparisonsshownabove,thattheoddsratiosfor
73
theNoBindingsentencesarenotsignificantlydifferentfromthoseforBoundObj
(Exp(B)=1.016,p=0.917)orBoundPosssentences(Exp(B)=0.959,p=0.786),but
aresignificantlylowerthanthoseforBoundSubj(Exp(B)=1.680,p=0.002),
NonFinite(Exp(B)=3.272,p<0.01),andBaseLine(Exp(B)=9.608,p<0.01)
sentences.
Parameter Estimates
Parameter B Std. Error
95% Wald Confidence Interval Hypothesis Test
Exp(B)
95% Wald Confidence Interval for Exp(B)
Lower Upper Wald Chi-Square df Sig. Lower Upper
Threshold [Choice_DV=1] -2.041 0.1520 -2.339
-1.743
180.362 1 0.000 0.130 0.096 0.175
[Choice_DV=2] -0.939 0.1350 -1.203
-0.674
48.296 1 0.000 0.391 0.300 0.510
[Choice_DV=3] -0.075 0.1382 -0.346
0.196 0.293 1 0.588 0.928 0.708 1.217
[Choice_DV=4] 0.516 0.1363 0.249 0.783 14.317 1 0.000 1.675 1.282 2.188
[Choice_DV=5] 1.375 0.1456 1.090 1.661 89.291 1 0.000 3.957 2.975 5.263
[Choice_DV=6] 2.666 0.1574 2.358 2.975 286.817 1 0.000 14.385 10.566 19.584
A. [Multiple Questions] -0.482 0.1118 -0.701
-0.263
18.553 1 0.000 0.618 0.496 0.769
B. [Comparative Deletion] 0.451 0.1202 0.215 0.686 14.066 1 0.000 1.570 1.240 1.987
C. [too/enough Movement] 0a 1
D. [BaseLine] 2.263 0.2115 1.848 2.677 114.497 1 0.000 9.608 6.348 14.543
E. [NonFinite] 1.185 0.1601 0.872 1.499 54.832 1 0.000 3.272 2.391 4.477
F. [BoundSubj] 0.519 0.1644 0.196 0.841 9.950 1 0.002 1.680 1.217 2.319
G. [BoundObj] 0.016 0.1532 -0.284
0.316 0.011 1 0.917 1.016 0.753 1.372
H. [BoundPoss] -0.042 0.1548 -0.346
0.261 0.074 1 0.786 0.959 0.708 1.299
I. [NoBinding] 0a 1
(Scale) 1
Dependent Variable: Choice_DV Model: (Threshold), Phenomenon, Condition a. Set to zero because this parameter is redundant.
TableA5:Experiment1GeneralizedEstimatingEquationParameterEstimates WenowturnourattentiontotheanalysisofthedatainExperiment2.Since
Experiment2isidenticalinsetuptoExperiment1exceptthatthesentences
instantiatingtheBoundObjandBoundPossconditionsarereplacedbysentences
thatinstantiate1pSubjand2Subjconditions,respectively,weemploythesame
statisticaltests.Asexpected,theKruskal-WallisTestappliedtothephenomenon
74
factorintheExperiment2dataindicatesthatthedistributionofratingsisnotthe
sameacrossthedifferentcategoriesofthephenomenonfactor(𝑋!(2)=86.409,p<
0.01).AsshowninTablesA6andA7,weseethesameclineofacceptability
schematizedin(A2)aswedidfortheExperiment1data.Alsoasexpected,the
Kruskal-WallisTestappliedtotheconditionfactorindicatesthatthedistributionof
ratingsisnotthesameacrossthedifferentconditions(𝑋!(5)=349.406,p<0.01).
ThepairwisecomparisonsandmeanranksareshowninTablesA8-A9.Taken
together,theysupporttheconclusionthatthesentencestestedinExperiment2
conformtotheclineofacceptabilityschematizedin(A4).Ofparticularinterestisthe
observationthatthe1pSubjand2pSubjconditionsgiverisetoratingprofilesthat
arenotsignificantlydifferentfromthatoftheNoBindingcondition.
(A4)BaseLine>NonFinite>BoundSubject>{1pSubj=2pSubj=NoBinding}
Sample1 - Sample2 Test Statistic
Standard Error
Standard Test Statistic Significance
Adjusted Significance
Multiple Questions – too/enough Movement
-197.550 34.744 -5.686 0.000 0.000***
Multiple Questions - Comparative Deletion
320.315 34.775 9.211 0.000 0.000***
too/enough Movement - Comparative Deletion
122.764 34.775 3.530 0.000 0.001***
TableA6:Experiment2pairwisecomparisonsofphenomena
Phenomenon Mean Rank Comparative Deletion 1,384.37 too/enough Movement 1,261.61 Multiple Questions 1,064.06
TableA7:Experiment1meanranksforphenomena
75
Sample1 - Sample2 Test Statistic
Standard Error
Standard Test Statistic Significance
Adjusted Significance
2pSubj-1pSubj 38.236 57.122 .811 .417 12pSubj-NoBinding 61.978 47.096 -1.316 .188 12pSubj-BoundSubj -292.945 47.096 -6.220 .000 .000***2pSubj-NonFinite -542.577 47.096 -11.521 .000 .000***2pSubj-BaseLine -819.462 57.659 -14.056 .000 .000***1pSubj-NoBinding -23.742 47.070 -.504 .614 11pSubj-BoundSubj -254.709 47.070 -5.411 .000 .000***1pSubj-NonFinite -504.341 47.070 -10.715 .000 .000***1pSubj-BaseLine -772.225 57.637 -13.398 .000 .000***NoBinding-BoundSubj 230.967 47.043 4.910 .000 .000***NoBinding-NonFinite -480.599 47.043 -10.216 .000 .000***NoBinding-BaseLine 748.483 57.616 12.991 .000 .000***BoundSubj-NonFinite -249.632 47.043 -5.306 .000 .000***BoundSubj-BaseLine 517.517 57.616 8.982 .000 .000***Nonfinite-BaseLine 267.884 57.616 4.649 .000 .000***
TableA8:Experiment2pairwisecomparisonsofconditions
Condition Mean Rank BaseLine 1,802.87 NonFinite 1,534.98 BoundSubj 1,285.35 NoBinding 1,054.39 1pSubj 1,030.64 2pSubj 992.41
TableA9:Experiment2meanranksforconditions
Finally,theresultsoftheGEEanalysisasappliedtothedatafrom
Experiment2areasindicatedinTableA10.Hereweseeresultsthatareconsistent
withtheconclusionsfromtheKruskal-Wallistest.AsseeninrowsA-C,too/enough
MovementsentencesareratedsignificantlyhigherthanMultipleQuestions(Exp(B)
=0.567,p<0.01)butlowerthanComparativeDeletionsentencesinawaythat
trendstowardsignificance(Exp(B)=1.322,p=0.014).AsseeninrowsD-I,ratings
76
forNoBindingsentencesarenotsignificantlydifferentthanthosefor2pSubj
sentences(Exp(B)=0.848,p=0.302)or1pSubjsentences(Exp(B)=0.919,p=
0.590),butsignificantlylowerthanthoseforBoundSubjsentences(Exp(B)=1.773,
p<0.01),NonFinitesentences(Exp(B)=3.334,p<0.01),andBaseLinesentences
(Exp(B)=8.405,p<0.01).
Parameter Estimates
Parameter B Std. Error
95% Wald Confidence Interval Hypothesis Test
Exp(B)
95% Wald Confidence Interval for Exp(B)
Lower Upper Wald Chi-Square df Sig. Lower Upper
Threshold [Choice_DV=1] -2.076 0.1527 -2.375
-1.777
184.771 1 0.000 0.125 0.093 0.169
[Choice_DV=2] -0.933 0.1357 -1.199
-0.667
47.339 1 0.000 0.393 0.301 0.513
[Choice_DV=3] -0.099 0.1338 -0.361
0.164 0.545 1 0.460 0.906 0.697 1.178
[Choice_DV=4] 0.615 0.1359 0.348 0.881 20.455 1 0.000 1.849 1.417 2.413
[Choice_DV=5] 1.493 0.1418 1.216 1.771 110.921 1 0.000 4.453 3.372 5.879
[Choice_DV=6] 2.610 0.1597 2.297 2.923 267.078 1 0.000 13.602 9.946 18.602
A. [Multiple Questions] -0.568 0.1207 -0.804
-0.331
22.096 1 0.000 0.567 0.447 0.718
B. [Comparative Deletion] 0.279 0.1138 0.056 0.502 6.001 1 0.014 1.322 1.057 1.652
C. [too/enough Movement] 0a 1
D. [BaseLine] 2.129 0.2434 1.652 2.606 76.489 1 0.000 8.405 5.216 13.544
E. [NonFinite] 1.204 0.1409 0.928 1.480 73.029 1 0.000 3.334 2.529 4.394
F. [BoundSubj] 0.572 0.1374 0.303 0.842 17.356 1 0.000 1.773 1.354 2.320
G. [1pSubj] -0.085 0.1572 -0.393
0.223 0.291 1 0.590 0.919 0.675 1.250
H. [2pSubj] -0.164 0.1594 -0.477
0.148 1.064 1 0.302 0.848 0.621 1.160
I. [NoBinding] 0a 1
(Scale) 1
Dependent Variable: Choice_DV Model: (Threshold), Phenomenon, Condition a. Set to zero because this parameter is redundant.
TableA10:Experiment2GeneralizedEstimatingEquationParameterEstimatesAcknowledgments:Forvaluablecommentsonvariousaspectsofthework
presentedhere,wewouldliketothankTomoFujii,NorbertHornstein,NickHuang,
AtakanInce,HisaKitahara,IdanLandau,GesoelMendes,JasonMerchant,Jon
Sprouse,andtwoanonymousLIreviewers.Wearealsogratefultoaudiencesata
numberofvenueswherewehavepresentedversionsoftheworkreportedhere.
77
ThesevenuesincludeHarvard,IndianaUniversity,Krakow(KrakówSyntaxLab
2016,JagiellonianUniversity),RiodeJaneiro(XCongressoInternacionalda
ABRALIN),Princeton(JerseySyntaxCircle:ProspectsfortheTheoryofSyntax),
UniversityofArizona,UCLA,UniversitätLeipzig,UniversityofMaryland,University
ofMassachusettsatAmherst,andUniversitätTübingen(Pronouns@Tübingen2).
Finally,weowespecialthankstoAnnBungerforhelpwithourexperimentaldesign
andanalysis,JeffLidzforcommentsontheexperimentaldesign,andBingyueLiof
theIndianaStatisticalConsultingCenterforindispensableguidanceconcerningthe
statisticalanalysisofourexperimentalresults.Naturally,allremainingerrorsare
ourown.
1Regardingthemultiplequestionin(2f),itbearsnotingthatthereisonestrandin
theliteraturethattakesthepositionthatthewh-elementsinamultiplequestioncan
infactbeseparatedbyafiniteclauseboundaryandthattheinsituwh-elementcan
evenbeembeddedinanisland;see,e.g.,Huang1982a;Lasnik&Saito1984;Fiengo,
Huang,Lasnik&Reinhart1988.Herewedepartfromthisviewandinsteadfollow
Kuno&Robinson(1973)andPostal(1974)intreatingexampleslike(2f)as
ungrammatical.Forrelevantexperimentalfindingsontherelativeacceptabilityof
multiplequestionsthatspanafiniteclauseboundary,seesection3.4andthe
appendix.
2SeePostal1974foranearlyversionofthisobservationinconnectionwithtough
movement,comparativedeletion,andmultiplequestions(onthelattercf.alsoKuno
andRobinson1972).ThegappingfactsarediscussedbyJohnson(1996)and
Lechner(2001).Andthesizeableliteratureonlocalitydomainsforquantifierscope
78
andACDistoovasttodojusticetohere,butincludesMay1977,1985;Larson&May
1990;Hornstein1994;Farkas&Giannakidou1996;Kennedy1997;Wilder1997;
Kayne1998;Fox2000;Cecchetto2004;Moulton2008;Hackl,Koster-Hale&
Varvoutis2012;Wurmbrand2013,2015a.
Incidentally,itisworthaskingwhetherthecrucialdistinctionbetween(2)
and(3)isthe(non-)finitenessoftheembeddedclauseorthenullness/overtnessof
theembeddedsubject.Inprincipleitshouldbepossibletoadjudicatethismatterby
consideringminimalvariantsinwhichtheembeddedclauseisnonfinitebuthasan
overtsubject,asin(i).Inpractice,though,thejudgmentsconcerning(i)arenot
crystal-clearandsowerefrainfromtakingastanceonwhetherthegrammarshould
rulethemout.Thetheoryweendupproposingpredictsthatthesentencesin(i)
shouldbeungrammaticalifthenonfinitecomplementtowantisaphasalcategory,a
questionweleaveopenforfutureresearch.
(i) a.?Thismagazineistoolow-brow[forJohntowantFredtoread__].
b.?JohnwantsFredtolikeapplesand[Bill<wantsFredtolike>oranges].
c.?MorepeoplewantFredtolikeapples[than<wantFredtolike>oranges].
d.?JohnwantsFredtoreadeverything[Billdoes<wantFredtoread>].
e.?[AtleastoneprofessorwantsFredtoreadeveryjournal].(∀>∃)
f.?Tellme[whowantsFredtoreadwhichjournal].
79
3ThisobservationandmanyoftherelevantfactsarelaidoutbyLasnik(2006),who
reportsonmaterialbasedinsubstantialpartonunpublishedjointresearchwith
TomohiroFujiiandNorbertHornstein.Versionsoftheobservationasitrelatesto
particularphenomenaarefoundinscatteredplacesthroughouttheliterature.These
includeSloan1991(onfamilyofquestions),Nishigauchi1998(onmultiple
questions,multiplesluicing,andgapping),Merchant2001:113,note4(ongapping
andmultiplesluicing),Syrett&Lidz2011(onantecedent-containeddeletion),and
Lasnik2014(onmultiplesluicingandextraposition).
4Inthisconnection,itisnoteworthythatthereisnoboundpronouneffectforclitic
climbing;inotherwords,cliticclimbinginlanguageslikeSpanishandItalianis
alwaysbannedacrossafiniteclauseboundary,evenwhenthesubjectofthe
embeddedfiniteclauseisaboundpronoun.Presumablythisisrelatedtothefact
thatunliketheclause-matephenomenaunderinvestigationinthispaper,noteven
allnonfiniteclausessupportcliticclimbing;whateverisresponsibleforthismore
severerestrictionwouldthenalsoaccountfortheabsenceoftheboundpronoun
effect.Wehopetoaddressthisissuefurtherinfuturework.
5Asidefromrestructuring,otherpotentialwaysinwhich“nonfiniteness”maybetoo
coarse-grainedanotionincharacterizinglocalitydomainsincludecontrol/raising
asymmetries(thereisagreementthatinversescopeispossibleoutofcontrol
complementsbutdisagreementaboutwhetheritispossibleoutofraising
complements:Wurmbrand2013andFrank&Storoshenko2015)andasymmetries
betweencontrolandraisingcomplementsontheonehandvs.ECMorraising-to-
objectcomplementsontheotherhand.
80
6Forthesakeofcompleteness,wedocumentinthisnoteoneotherpotential
manifestationoftheboundpronouneffect.Kratzer(1998b:5),followingRuys
(1992),observesthatboundpronounsfacilitateintermediatescopereadingsfor
indefinitesinsentenceslike(i)(cf.(ii)forthevariantwithouttheboundpronoun).
Thatis,itiseasierin(i)thanin(ii)tounderstandsomestudentasvaryingfromone
professortothenextbutnotvarying,foreachprofessor,fromoneclasssessionto
thenext.
(i) [Everyprofessor]1gotaheadachewheneversomestudenthe1hatedwas
inclass.
(ii) [Everyprofessor]1gotaheadachewheneversomestudentMaryhated
wasinclass.
Whetherornotthecontrastin(i)/(ii)canbesubsumedunderthesamekindof
phase-theoreticaccountthatweadvanceforthecorecasesoftheboundpronoun
effectisunfortunatelynotsomethingthatwewillbeabletoestablishinthispaper,
butitmaybeaninterestingtopicforfutureinvestigation.
7Anotherpotentialsubjectorientationeffectconcernstheantecedent.Inadditionto
theboundpronounhavingtobeinsubjectpositioninorderfortheboundpronoun
effecttohold,datalike(ia-b)suggestthattheantecedenthastobeinsubject
positionaswell.
81
(i) a. *Joe1 persuaded Bill2 that he2 should read Pride & Prejudice and Tim3
<persuadedBill2thathe2shouldread>Sense&Sensibility.
b.*Joe1promisedBill2 thathe2hadalreadyreadPride&PrejudiceandTim3
<promisedBill2thathe2hadalreadyread>Sense&Sensibility.
Anticipatingourphase-theoreticaccountoftheboundpronouneffectinsection4
below,aninitiallyattractivewayofmakingsenseofthisconstraintonthe
antecedentwouldbetoproposethattransferofacandidatephasehead’s
complementoccursassoonastheboundpronounisvaluedbyitsantecedent,so
thatvaluationoftheboundpronounbysomethinglowerthanthesubjectwouldnot
delaytransferlongenoughtoextendlocality.Unfortunately,however,thisapproach
facesdifficultygiventhatsomeoftheclause-matephenomenathattriggerthe
boundpronouneffect(includingforexampletoughmovement,whichalsoraisesa
numberofotherwellknownproblems)involvedependenciesthatspanacrossthe
valuingantecedentsubject.Sointhesecases,weseethattransfermustcontinueto
bedelayedevenaftertheboundpronounisvalued.Consequently,thisisnot
somethingthatwewillbeabletoaccountforinthispaper,thoughwehopeto
addressitinfuturework.
8Anotablelimitationofthisdesignisthatthesentencesthatinstantiatethevarious
boundpronounconditions(24b-d,25b-d)canalsobereadinsuchawaythatthe
relevantpronounisfreeratherthanbound.Sincewedidnotaskexperimental
participantstoratesentencesrelativetoanyparticularinterpretation,thereisno
guaranteethattheirjudgmentsreflecttheboundreading.Buttheexpectationisthat
82
sinceeachsentenceisjudgedinisolationwithoutacontextthatcouldsupplya
referentforafreepronoun,theonlysalientreadingistheboundreading.
9ThiswordingintheinstructionsisborrowedfromWhite&Grano’s(2014)
experimentalinvestigationofpartialcontrol,whosematerialsareavailableat:
https://github.com/aaronstevenwhite/PartialControlExperiments
10TotalsforComparativeDeletionconditions1Pand2PinTable2areslightlylower
thantheyshouldbe(149and148insteadof150and150,respectively)becauseone
oftheparticipantsinExperiment2neglectedtosupplyaratingforthreeofthe
targetitems.Buttheratingsfortheother30targetitemsthatthisparticipantdid
rateareincludedinthetableandinthestatisticalanalysis.
11Seee.g.Wurmbrand(2015a)onquantifierscopeinteraction.Wurmbrand
proposesthatquantifierraisingisnotclause-boundandthatinstead,quantifier
raisingacrossmultiplefiniteclauseboundariesincursaprocessingcostthat
accountsforitsdegradedacceptability.Onthiskindofapproach,thebound
pronouneffectwouldhavetobeunderstoodassomekindofprocessingfacilitation
ongrammaticalsentencesratherthansomethingthatmakesthedifferencebetween
agrammaticalsentenceandanungrammaticalsentence.Itremainstobeseenhow
suchaprocessingaccountwouldfareincomparisonwiththegrammaticalaccount
weproposebelow.
12Theway(31c)isformulatedpresupposes(possiblyproblematically)thatbound
pronounsaredistinguishedfromfreepronounsinthelexicon(i.e.,itisdetermined
assoonasthepronounismergedintothederivationwhetheritwillbeboundor
not;cf.alsonote20below).Wepresentthingsthiswayforexpositoryconvenience,
83
andsimplywishtonoteherethatthefinalversionofouraccount,statedin(36c)
andfurtherelaboratedinsection4.4,doesnotrequiresuchanassumption.Byway
ofpreview,whatwewillultimatelysayisthatpronouns(irrespectiveofany
free/bounddistinction)optionallyenterthederivationwithunvaluedphi-features.
Ifapronounentersthederivationwithunvaluedphi-featuresandendsupbeing
free,thederivationcrashes,sincephi-featurevaluationpiggybacksonbinding.By
contrast,ifthepronounendsupbeingbound,itsunvaluedfeatureswillbe
determinedbythebinder.
13Itisnotentirelycleartoushowtounderstand“propositional”insuchawaythat
itpicksoutCPandvPasanaturalclasstotheexclusionofothercategoriessuchas
TP.Thisleavesuswitha“listproblem”:thesetofphasalcategorieshastobe
stipulatedratherthanfollowingfromsomethingmoregeneral.Itisinterestingto
notethatversionsofthe“listproblem”arefoundelsewhereinChomsky’swork;for
example,Chomsky’s(1973)TensedSentenceConditionandSpecifiedSubject
ConditionarebothsubsumedunderthenotionofGovernmentinChomsky1981,
butburiedinthedefinitionofGoverningCategoryisthetermofartSUBJECT(all
caps),whichChomskydefineswithalist:finiteAGR(supplantingtheTensed
SentenceCondition)andthesubjectofanonfiniteclause(supplantingtheSpecified
SubjectCondition).Yetanotherexampleofthelistproblemisthedefinitionof
“cyclicnodes”asNPandSinclassicSubjacency,somethingChomsky(1986)
attemptstoremedyinBarriers.Inanycase,thephasalstatusofvPhasnotgone
unquestioned(denDikken2006),andtheanalysiswepursueinthispaperinfact
84
seemstobeabetterfitwiththeviewthatCPsarephases(undersomeconditions)
whereasvPsarenot.Seesection4.3belowforfurtherdiscussion.
14PredecessorsofthisideaincludeFelser2004,whoproposesthat“phasesshould
bestbedefinedintermsofconvergence,withthe‘propositional’categoriesCPand
vPbeingpotentialcandidatesforlocalSpelloutonly”;Wurmbrand2011,who
proposesthat“onlyinterpretationallycompleteunitscanbetransferred…iF:___ina
potentialphaseprojectionpostponestransfer”(where“iF:___”isanunvalued
interpretablefeature)(p.69);andUriagereka(pers.comm.),whosuggests,building
onLasnik&Uriagereka(2005),that“transferissuspendedwhenananaphoric
dependencyisatstake(untiltheantecedententersthepicture)”.
15Ananonymousreviewerpointsoutthat“atZP”in(44b)canbeinterpretedin
morethanoneway:doesinaccessibilityariseassoonasZismergedin,ornotuntil
ZPiscomplete?Inwhatfollows,weadoptCitko’s(2014)interpretation:“Thetwo
definitions[i.e.,strongPICandweakPIC]differwithrespecttowhenthedomainof
thephaseheadHbecomesinaccessible:assoonasHPiscompleteversusatthe
pointthenextphasehead(Z)ismerged”(p.33).Inotherwords,onthestrongPIC,
inaccessibilityariseswhenthephraseassociatedwiththephaseheadiscomplete,
whereasontheweakPIC,inaccessibilityarisesassoonasthenexthighestphase
headismerged.
16Acompromiseisavailable:ourtheoryisconsistentwiththepossibilitythatvisa
candidatephasehead(whichcouldaccountforwhyitseemstopatternlikeaphase
headincertainrespects:seee.g.Legate2003;Lee-Schoenfeld2008;Citko2014).
Butitisneveranactualphasehead,becausetheheadofitscomplement(i.e.,V)
85
alwayshasmorphologicaltenseandagreementfeaturesthatarenotvalueduntil
higherintheclause.
17WealsoassumeintheforegoingthatOpcannottargetintermediate[Spec,vP]
positions.Wemakethissimplifyingassumptionprimarilyforthreereasons.First,if
Opcannottargetintermediate[Spec,CP]positions,thenitseemsreasonableto
hypothesizethatitalsocannottargetintermediate[Spec,vP]positions.Second,if
successivecyclicmovementvia[Spec,vP]dependsonthestatusofvasa(candidate)
phasehead,thenourtentativeconclusionthatvisnota(candidate)phaseheadalso
constitutesareasonnottoconsiderintermediate[Spec,vP]positions.Finally,the
thirdreasonispractical:entertainingintermediate[Spec,vP]landingsiteswould
excessivelymultiplythenumberofanalyticaloptionstobeassessed.Andsinceour
goalhereisa“proofofconcept”ofaphase-theoreticaccountoftheboundpronoun
effect,weneednotconsidereveryconceivablewaythingscouldbe.
18Thereisalsomoretobesaidaboutmultiplequestions.Here,the(covertly)moved
phraseclearlyhaswh-features,reopeningthepuzzleaboutwhyitcannotmovein
successivecyclicfashion.Butregardlessofwhysuccessivecyclicityisblockedin
multiplequestions,thatitisblockedisaconclusionconvergentwithrecentworkon
multiplequestions.Kotek(2014),basedondataverydifferentfromthatwhich
concernsushere,concludesthat“thecovertmovementofthein-situwhin
superiority-obeyingquestionsisnotanunboundedlong-distancemovement,as
oftenassumed,butinsteadashortQR-likemovement,whichisonlyextendedin
extraordinarycases,…”(p.209).Similarly,Saito(1994)proposesthatthelowerwh-
86
elementinamultiplequestionLFadjoinstothehigherwh-elementratherthan
movingtoa[Spec,CP]position.
Inasimilarvein,ananonymousrevieweraskswhysuccessivecyclic
movementfailsforfamilyofquestionsandformultiplesluicingasexemplifiedin(i)
and(ii)respectively.
(i)*[WhichjournaldoeseveryoneclaimthatJohnreads__]?
Anticipatedanswertype:BillclaimsthatJohnreadsLI,Timclaimsthat
JohnreadsNLLT,etc.
(ii)*SomeoneclaimsthatJohnisworriedaboutsomethingbutIdon’tknow
[who<claimsthatJohnisworried>aboutwhat].
Forfamilyofquestions,weassumefollowingSloan(1991)andLasnikandSaito
(1992)thatthecrucialfactoristhestructuralrelationshipbetweenthequantifier
andthetraceofthewh-movement.Consequently,(i)isruledoutbecausealthough
thewh-phrasecanmovesuccessive-cyclically,itstraceisnotinasufficientlylocal
configurationwiththequantifier.Formultiplesluicing,weassumefollowingLasnik
(2014)thatthesecondwh-expressionundergoesrightwardmovement
(extraposition).Plausibly,suchmovementisnotsubjecttosuccessivecyclicityeven
whenthemovedphrasehappenstohavewh-features,thoughweleaveafull
investigationofthisquestiontofutureresearch.SeeLasnik(2014)forsome
speculationaboutwhyextrapositioncannotbesuccessivecyclic.
87
19Weassumehereforconcretenessthatboundpronounsareboundandvaluedby
theirDPantecedents.Butcf.Kratzer(2009)forthealternativeviewthatbound
pronounsareboundbyclause-localverbalfunctionalheadsCandv.Asfaraswecan
tell,thischoicepointisinprincipleorthogonaltotheconcernsofthispaper,
althoughKratzer’sparticularimplementationmaynotbecompatiblewithour
approach,insofarashersystemwouldallowCtoenterthederivationwithvalued
phi-featuresthatwouldimmediatelyvaluea[Spec,TP]pronounandrenderthe
clauseitappearsinphi-completeandhencephasal.
20Oneofthecentralargumentsinvokedinfavorofthiskindofapproachhastodo
withtheobservation,originallyduetoPartee(1989),thatsometimesphi-features
onfirst-andsecond-personboundpronounsappearasthoughtheyareignoredby
thesemantics,suchasinexampleslike(i).Suchfactscanbereadilymadesenseofif
boundpronounslikemyin(i)acquiretheirphi-featuresatastageofthederivation
thatistoolateforthesefeaturestobeinterpretedbythesemantics(cf.alsoLandau
2016fordiscussion).
(i)OnlyIdidmyhomework.
Relevantreading:Iamtheonlyxsuchthatxdidx’shomework.
Thatbeingsaid,theverdictisstilloutonwhetherlatevaluationofphi-featuresis
therightwaytoaccountforsentenceslike(i).Otheranalyticaloptionsthathave
beenentertainedincludephi-featuredeletion(vonStechow2003;Reuland2010)as
88
wellasapproachesinwhichthephi-featuresonmyin(i)areinfactinterpreted
afterall(Cable2005;Spathas2010;Jacobson2012;Sudo2012).So,itmaybe
prematuretotakesentenceslike(i)asstrongevidencethatboundpronounsenter
thederivationwithunvaluedphi-features.
Inasimilarvein,Heim,Lasnik&May(1991)arguethatundersome
conditions,boundpronounsthatrangeoveratomicindividualsarenonetheless
syntacticallypluralduetoasyntacticagreementrequirementwiththeirantecedent.
Takenatfacevalue,thiskindofsituationalsoseemstosupporttheviewthatphi-
featuresonboundpronounsareatleastsometimesvaluedlateandignoredbythe
semantics,thoughotherapproachesareconceivable.
21ForconcretenesswefollowHeim&Kratzer(1998)inassumingthatbindingofa
pronoundependsonQRoftheantecedent,whichtriggersPredicateAbstractionin
thesemantics.Whentheantecedent’smovementindexmatchestheindexonthe
pronoun,bindingresults.
22EssentiallythesameideaisproposedbyKratzer(2009),basedonaverydifferent
setofdata.ThishybridapproachtopronounsisalsoreminiscentofChomsky’s
(1955/1975:519-524)proposalthatthereare“twoelementsheandhe*,withhe*a
propernoun,andheapronounjustlikeI,you”(p.524of1975edition).For
Chomsky,though,thedistinctioncorrelatedwithwhetherthepronounhada
(sentence-local)antecedent,whereasforus,aswellasforKratzer(2009),the
suggestionisthathavinganantecedentisanecessarybutnotasufficientcondition
forhavingenteredthederivationwithunvaluedphi-features.
89
23ItisalsoconceivablethatthereareotherstructuralconstraintsasidefromthePIC
thatlimittheapplicationoffeaturetransmission.Forexample,itcouldbethat
featuretransmissionissubjecttointervention.Considertheminimalpairin(i)-(ii).
Toourear,(ii)soundsratherdegradedincomparisonwith(i),andonepossible
takeonwhyisthattheDPMaryin(ii)intervenesandtherebydisablesfeature
transmissionbetweentheboundpronounanditsantecedent.
(i) ?Thisbookistooexpensive[forJohn1topromisethathe1willbuyOp].
(ii) *Thisbookistooexpensive[forJohn1topromiseMarythathe1willbuy
Op].
Thatbeingsaid,wediscussbelowsomeexamplesfromRoss(1967)thatsuggest
thattheboundpronouneffectmaybeoperativeforsomeislandphenomena,and
someoftherelevantexamples(71e,72a-c)arenotasdegradedaswemighthave
expectedthemtobeiffeaturetransmissionissubjecttointervention.Weleaveto
futureworkamorethoroughinvestigationofthisissue.
24Inarelatedvein,parasiticgapsarewellknowntobebetterinnonfiniteadjuncts
(ia)thaninfiniteadjuncts(ic),anditseemstousthatfiniteadjunctswithbound
pronominalsubjectspatternwithnonfiniteadjunctsinbeingacceptablewitha
parasiticgap(ib).Sothisappearstobeyetanothermanifestationofthebound
pronouneffect.
90
(i)WhichpapersdidJohnreadbefore…
a.…filing?
b.…hefiled?
c.?...Billfiled?
25Classically,itwasproposedthatmovementofargumentsissubjectonlyto
SubjacencywhilemovementofadjunctsissubjectbothtoSubjacencyandtotheECP
(Huang1982b;Lasnik&Saito1984).Whileitisareasonablehypothesisthatwhat
countsasa“barrier”forSubjacencyisthesameasthatfortheECP,itisnota
logicallynecessaryone.Soonewayofinterpretingthefactsin(73)isthatwhatever
principleisresponsibleforECPeffectsisnotsubjecttothekindofphase-theoretic
constraintsthatgiverisetotheboundpronouneffectbutratherobeyssomeother
setofconstraints.