On-Phases

1SCIMS“Phases”,NevenWengerFinalsession–26/09/08PhasesEarlyminimalism,rangingfromChomsky(1989)toChomsky’s(1995)MinimalistProgram(MP),incorporatedaweaklyderivationalapproach.Thecomputationalsys-tem(narrowsyntax,CHL1)manipulatesaselectionoflexicalitems(LI)bymeansofastep-by-stepapplicationoftheoperationsMergeandMove,untilSpell-Outoccurs.Then,PFandLFarecreated,thetwolevelsofrepresentationinterfacingwiththesyn-tax-externalmodulesA-PandC-I,respectively.Chomsky’s(2000)MinimalistInquiries(MI)soughttoreducederivationalcomplexitybychoppingthelexicalarray(LA)upintosub-arrays2,eachfeedingCHLtoderiveaparticularphase–aderivationalcycle.Phasesarewell-definedchunksofaderivation–vP,CP,andDP(moreonthereasonsforthischoicebelow)–,eachofwhich,uponcompletion,istransferredtotheinterfaces,andthusdoesnolongerbothersthecomputationwithitsweight.ThisentailsatheoryofMultipleSpell-Out(cf.Uriagereka1999).Thesyntacticobjectsthatqualifyforphasesaretransitivev*Ps(star’sfortransitive),whichcontainanAGENToranEXPERIENCERasanexternalargument(totheexclusionofpassiveandunaccusativevP),andCPs,whicharespecifiedforTenseandForce(Adger’sClauseType)3.Crucially,defectiveTPsandVPsarenophases.Thereasoningbehindthisassumptionisforoneconceptual(phasesrepresentnatural,propositionalob-jects,saturatedexpressions;v*Psrepresentevents,CPsfullpropositions),butalsogrammatical(pseudo-cleftingisnotpossiblewithnon-finiteraising/ECM-complements,whichareTP).CPsarecompleteclausalcomplexesandv*Psarecompletethematiccomplexes.Aratherniceconceptualargumentforphasesconcernstheuninterpretabilityoffeatures:IfSpell-Outhastoremoveun-interpretablefeaturestoavoidacrashattheinterfaces,itmustknowwhichfeaturesareinterpretable,and1HL=HumanLanguage.2TheLAreplacedthenotionofnumeration(NUM)fromminimalistframeworksprevioustoMI(e.g.MP).Technicallyspeaking,aLAisaNUMifitcontainsmorethanoneoccurrenceofoneandthesamelexicalitem(LI),inwhichcasethisitem’sindexislargerthan‘1’(cf.DbP:11).3DPsareconsideredphases,too(Chomsky2005:17f.).WhetherPPsarephasesornorremainstobeinvestigated.LEXICONMergeMoveSPELL-OUTPFLFFig.1:TheY-ModelLEXICONMergeMoveSPELL-OUTPFLFMergeMoveSPELL-OUTPFLFPHASE2PHASE1Fig.2:MultipleSpell-OutModelPHASE1PHASE22whichareuninterpretable.However,narrowsyntaxissupposedtobeblindasabat,andthus,wouldneedtobeabletolookahead(upthetree,toLFandPF)inordertodeterminetheinterpretabilityofafeature.Atransferofderivationalchunkstotheinterfacesremediesthisissue,searchspacebeingreducedtoalocaldomain(aphase).EmpiricalsupportforphasescomesfromtheEPP-featureonT:howdoesCHLdecidebetweenattractingasubjectDPandmerginganexpletivethere?GiventheeconomicpreferenceofMergeoverMove(MoveismorecostlythanMergesinceMoveCopy+Merge),aninsertionofthereshouldbeexpectedineveryinstance,andraisingshouldneverbepossible.Phasescircumventthisissuesincelexicalsub-arrayscanbedefinedforeverycycle.Togetanideaofthetechnicalsideofthisargument,firstconsiderthefollowingtwoexamples,whichillus-trateMerge-over-Move.Theyshareoneandthesamenumeration,butonederivationyieldstheungram-matical(2).(1)Num={there1,T2,seem1,to1,be1,someone1,here1}(2)a.Thereseemstobesomeonehere.b.*Thereseemssomeonetobehere.Let’stakeacloserlookatthestepsofthederivationof(2).A.[TPT[EPP]tobesomeonehere]–MergeTB.[TPthereT[EPP]tobesomeonehere]–Mergethere&checkEPPD.[TPT[EPP][VPseems[TPthereT[EPP]tobesomeonehere]]]–MergeTE.[TPthereT[EPP][VPseems[TPthereT[EPP]tobesomeonehere]]]–Movethere&checkEPPNowconsiderthederivationoftheungrammatical(2)(1),basedonthesamenumeration,takinganotheroptionatstepB.A.[TPT[EPP]tobesomeonehere]–MergeTB.[TPsomeoneT[EPP]tobesomeonehere]–Movesomeone&checkEPPD.[TPT[EPP][VPseems[TPsomeoneT[EPP]tobesomeonehere]]]–MergeTE.[TPthereT[EPP][VPseems[TPsomeoneT[EPP]tobesomeonehere]]]–Mergethere&checkEPPThederivationalstepBof(2)violatesMerge-over-Move,movingsomeoneinsteadofmergingtheexpletivethereavailableintheNum,whichiswhythederivationproducesanill-formedsentence.Definingdifferentsub-arraysfor(2),providedthephasehoodofvPandCP,cancapturethisissuederivationally:(3)a.{{C,T}3{seem,there,T,to}2{be,someone,here}1}.Thereseemstobesomeonehere.–(2)3b.{{C,there,T}3{seem,T,to}2{be,someone,here}1}.*Thereseemssomeonetobehere.–(2)c.{{C,T}3{seem,T,to}2{be,someone,here}1}.Someoneseemstobehere.–w/oexpletiveSofar,sogood.However,intheexamplesabove,appealtoMerge-over-Movewassufficient,withouttheneedtoinvokephases?Butwhatofexampleslikethefollowing(fromBoeckx&Grohmann2005?(4)Therewasarumourthatamanwasintheroom.Here,givenonenumerationforthederivationtogetherwiththeconditionofMerge-over-Move,theembed-dedclause…thatamanwasintheroomcouldneverbederived,sincetheavailabilityoftherewouldal-wayspre-emptraising.(5)LA={C2,there,T2,be2,a2,rumour,that,man,in,the,room}Here,phasalsub-arraysactuallydothetrick!(6)LA={{C,there,T}4{be,a,rumour}3{C,that,T}2{be,a,man,in,the,room}1}Considerthederivation.A.[TPT[EPP]beamanintheroom]–MergeTB.[TPamanwas[EPP]beamanintheroom]–Moveaman&checkEPPC.[CPthatamanwas[EPP]beamanintheroom]–MergeC/thatD.[TPT[EPP][bearumourthatamanwas[EPP]beamanintheroom]]–MergeTE.[CP[TPthere