Predictable and unpredictable sources of variable verb and adverb placement in Scandinavian PDF
Preview Predictable and unpredictable sources of variable verb and adverb placement in Scandinavian
Available online at www.sciencedirect.com Lingua120(2010)1062–1088 www.elsevier.com/locate/lingua Predictable and unpredictable sources of variable verb and adverb placement in Scandinavian Ho¨skuldur Thra´insson* DepartmentofIcelandic,UniversityofIceland,IS-101Reykjav´ık,Iceland Received15October2007;receivedinrevisedform22May2008;accepted7April2009 Availableonline18June2009 Abstract RecentresearchhasshownthatthereismorevariationintheplacementofverbsandadverbsinScandinavianthanpreviously thought. This paper argues that the theory of verb movement originally proposed by Bobaljik and Thra´insson [Bobaljik, J.D., Thra´insson,H.,1998.Twoheadsaren’talwaysbetterthanone.Syntax1,37–71]doesinfactpredictmuchoftheobservedvariation, despite recent claims to the contrary. According to this theory, this variation is intimately related to morphological differences betweenthelanguages(ordialects):Languageswithclearlyseparableagreementandtensemorphology(Icelandic,forinstance) haveseparateagreementandtenseprojectionsandthismakesverbmovement(oftheV-to-Itype)obligatorybecauseofthenature ofcheckingoperations.Thisdoesnotmean,however,thatV-to-Icanonlyoccurinlanguageswithrichverbalmorphology.The theoryalsopredictsthatacertainkindofadverbialmodificationcantriggerV-to-I,butthishastypicallybeenoverlookedinthe literature: If adverbs are specifiers of separate functional projections, then V-to-I is necessary, whereas it is not if adverbs are adjuncts.ThismeansthatadverbialadjunctionmustbetheruleinScandinavianandadverbsinseparatefunctionalprojectionsan exception(optionally availablefor somesentenceadverbsbutnot allinRegionalNorthernNorwegian, forinstance). #2009PublishedbyElsevier B.V. Keywords: Verbmovement;Functionalstructure;Checking;Adverbialmodification;Scandinavian 1. Introduction The main purpose of this paper is to demonstrate that a formal theory of verb movement which makes fairly conservativeassumptions about clausal architecture does in fact predict much of the recently observed variation in verbplacementinScandinavian,whereaslessrestrictiveapproachestypicallyfailtomakesuchpredictions.1Inthe * Tel.:+3548628694. E-mailaddress:[email protected]. 1Classicverbmovementaccountsobviouslydependontheassumptionthatheadmovementisapartofthetheoreticalarsenalavailable.Chomsky hasargued(2001)thatheadmovementisnotcompatiblewithcertainrecentassumptionsaboutthenatureofsyntacticderivationandsubsequently Mahajan (2003), Mu¨ller (2004), Koopman and Szabolcsi (2000) and others haveshown that it is possible to derivecertain ‘‘verb movement structures"byassumingsomesortofremnantmovementinstead(seealsoBentzenetal.,2007a,b).Althoughremnantmovementaccountstendto become quite complex and their predictions less than obvious at times, I do not want to maintain that they should be ruled out in general. Nevertheless,thepresentpapercanbeseenasademonstrationofwhatafairlystraightforwardheadmovementaccountcanaccomplish.Ifhead movementisruledoutontheoreticalgrounds,thenonewillsimplyhavetohopethatthepositiveresultsofthepresentaccountcanalsobeobtained byaccountsthatruleoutheadmovement.Ifnot,thentheotherobviousalternativeisthatatheoryrulingoutheadmovementisonthewrongtrack. 0024-3841/$–seefrontmatter#2009PublishedbyElsevierB.V. doi:10.1016/j.lingua.2009.04.003 H.Thra´insson/Lingua120(2010)1062–1088 1063 process I shall show that the orders Vfin-Adv (finite verb preceding a sentence adverb) and Adv-Vfin can arise in different ways in Scandinavian and they will have different properties depending on the derivation and resulting structure.Inparticular,theorderAdv-Vfindoesnotnecessarilymeanthatthefiniteverbhasremainedinsitu,asoften erroneously assumed in the discussion of such orders in Icelandic, for instance. The structure of this paper is as follows. After a brief review of the basic facts I shall review the theory of verb movementproposedbyBobaljikandThra´insson(1998,henceforthB&T)anddrawattentiontocertainpropertiesofthis theorythathavetypicallygoneunnoticed.Iwillthenshowthatthesepropertiesdoinfactpredictmoretypesofvariation inverb placementthan usually assumed. I shall discusstowhat extent thesetypescanbe distinguished, both by the linguist analyzing a particular language and the child acquiring it. I shall then argue that certain basic properties of variableverbplacementinScandinavianareinfactpredictedbyB&T’stheory.ButIshallalsomaintainthatcertain aspects of the observed variation are not predicted under any present theory, in particular apparent cross-linguistic differencesinadverbialmodificationandadverbplacement,andhenceIshallarguethatmoreresearchisneededonthis topic. Tosetthestageforthe followingdiscussion, themain predictionsandclaimsoftheanalysisadvocatedhereare listed in (1): (1) a. Verb movement of the V-to-I type is obligatory in languages2 that have a split agreement and tense morphology.ThisfollowsfromB&T’stheoryofverbmovement.InsuchlanguagesV-to-Iwillapplyacrossall sentenceadverbs,i.e.medialadverbslike‘never’,‘always’,‘not’,etc.HencetheorderVfin-Advwillbethe defaultorderinallembeddedclausesinsuchlanguages,regardlessofthetypeofsentenceadverbinvolved.3 b. ApparentexceptionstoV-to-Iinlanguageswithrichverbalmorphologyoftheappropriatekindarejustthat, apparent.Theappearanceistheresultofexceptionaladverbplacementabovethefiniteverb.Itfollowsfrom thisthattheAdv-Vfinorderwillnotbethedefaultorderinlanguagesofthiskindbutonlyacceptableunder special circumstances, depending on the type of adverb and the type of embedded clause involved. c. V-to-Iisnotonlytriggeredbyrichverbalmorphology(moreprecisely,bya‘‘splitIP’’underB&T’stheory). Itcanalsobetriggeredbyaspecialtypeofadverbialmodificationwhichmakesverbmovementnecessary inlanguageswith‘‘poorverbalmorphology’’.Althoughthishastypicallybeenoverlooked,italsofollows B&T’s theory of verb movement. d. Finally, the Vfin-Advorder can also result from aV-to-C-type movementinembedded clauses thathave main-clause properties(e.g. bridgeverbcomplements, certain typesofadverbialclauses,etc.), asis well knownandextensivelyreviewedbyVikner(1995a)andRohrbacher(1999),forinstance.Sincemain-clause propertiesarenottypicalpropertiesofembeddedclauses,itfollowsthataVfin-AdvorderresultingfromV- to-Cwillbeexceptionalinembeddedclausesinalllanguages(paceBentzenetal.,2007a,b,forinstance). Sincesomeofthepredictionsandclaimsin(1)mayseemunexpectedorcontroversial,thereadershouldnowbe very excited to read the following discussion and argumentation. 2. The basic facts and some background assumptions 2.1. The standard analysis of verb placement variation in Scandinavian Examples like the following have standardly been taken to show that there is V-to-I movement in Icelandic: 2 Notethatthewordlanguagecan,ofcourse,mean‘dialect’inthiscontext—andelsewhereinthispaper. 3 Wordorderdifferencesbetweenmainclausesandembeddedclauseswillbediscussedbelow,aswillthespecialpropertiesofcertainnon-finite clauses.Iamlimitingthediscussiontofiniteembeddedclausesforthemomentinordertosimplifythepresentation. 1064 H.Thra´insson/Lingua120(2010)1062–1088 In(2a)thenon-finitemainverbsy´nt‘shown’followsthesententialnegationekki‘not’andisthusarguablyinsitu insidetheVP.4In(2b)themainverbisfiniteandprecedesthenegationandbecausetheembeddedclauseisofthekind thatisknownnottohavemain-clauselikeproperties,i.e.arelativeclause,ithasstandardlybeenassumedthatthis movementistotheI-position(ortosomeI-position,ifoneassumesasplitIPasoriginallysuggestedbyPollock,1989) ratherthantotheC-position(asitcouldbeinthecaseofembeddedclausesknowntohavemain-clauselikeproperties, such as bridge verb complements). Conversely, facts of the following sort have been taken to show that V-to-I movementdoesnotoccurinDanish,andcomparablefactsfromstandardNorwegianandSwedishhavebeenpresented (forextensivereviewssee,e.g.Vikner,1995a,131ff.;Vikner,1995b;HolmbergandPlatzack,1995;Thra´insson,2007, 27ff., 58ff. and references cited there): SothestandardbeliefhasbeenthatIcelandichasV-to-ImovementwhereasMainlandScandinavian(MSc)does not.This,ofcourse,raisesthequestionwhatcouldbethereasonforthisvariationorwhatitcouldberelatedto.Thatis obviouslyapartofamoregeneralproblem:Towhatextentcanlanguagesvaryandwhydothey?Beforewelookat variable verb placement in more detail, it is useful to review some basic ideas about variation. 2.2. Sources of variation, cues and triggers It is probably fair to say that much recent work on syntactic variation does not consider in any detail the basic questionofwhatcouldbethesourceoftheobservedcross-linguisticorcross-dialectaldifferences.Ifwebelievethata majorsourceofvariationisthedifferent‘‘choices’’thatchildrenmakeduringtheacquisitionperiod,thenonemust thinkseriouslyaboutthequestionwhatcouldtriggerthesedifferentchoices.Someyearsago,Chomskyproposedthe following answer to this question (1995, 169)—and I shall refer to this proposal as The Trigger Condition: (4) Variation must be determined by what is ‘‘visible’’ to the child acquiring language, that is by the PLD [Primary Linguistic Data]. WhatmayvaryisthenstandardlytakentobedeterminedbyUG.Onequestion,forinstance,iswhetherortowhat extentthemake-upoffunctionalcategoriescanvaryfromonelanguageordialecttoanother.Whilesomelinguists believethatthefunctionalskeletonofsentencesmustbeidenticalinalllanguagesanddialects,withvariationlimited to the so-called ‘‘strength’’ of individual projections or heads, a somewhat different way of thinking about this is outlined in The Limited Diversity Hypothesis proposed by Thra´insson (1996, 257): (5) a. UG defines a set of possible functional projections. b. The languages oftheworld‘‘select’’elements fromthis set, justlike theyselectfrom the set ofpossible phonological features. c. Hence it is not the case that all functional projections occur in all languages. Thisobviouslyraisesthequestionhowchildrenacquiringaparticularlanguageknowwhichfunctionalprojections occurinthelanguagetheyareacquiring.AccordingtoThra´insson(1996,261),childrenfollowTheRealMinimalist Principle when trying to figure out which functional projections play a role in their language: (6) Only assume the functional projections that you have evidence for. IfoneassumestheTriggerConditiondescribedin(4),thismeans,then,thatvariationsinthefunctionalmake-upof dialects and languages would be limited by what can be triggered during acquisition by visible differences. 4HereandelsewhereinthispaperIwilljustusethetraditionallabelVPratherthan,say,vP,asnothinginmyargumentationwilldependonthe detailedstructureoftheVP-domain. H.Thra´insson/Lingua120(2010)1062–1088 1065 Thisemphasisontheimportanceofvisibleevidenceduringtheacquisitionperiodcanbefoundinanumberofother works. It can, for instance, be related to Lightfoot’s concept of cues, described, e.g. as follows by him (2006, 78): (7) Acueisapieceofstructure,anelementofI-language,whichisderivedfromtheinput,butitisnotasentence [...]. A sentence expresses a cue if the cue is unambiguously required for the analysis of the sentence. AsLightfootpointsout,theconceptoftriggersisasimilarone,albeitnotalwaysdefinedinaformalfashion,and herefersinthisconnectiontoJanetFodor’sconceptoftriggersaspiecesofstructureor‘‘treelets’’,whereatreeletis‘‘a small piece of tree structure...that is made available by UG and is adopted into a learner’s grammar if it proves essential for parsing input sentences’’ (Fodor, 1998, 6). Wewillhavereasontocomebacktosomeoftheseideasaboutthenatureandpossibleextentofvariationinthe followingsections,concentratingonvariationintherelativeplacementofverbsandadverbs.Butitshouldbepointed out that the approach advocated here is very different from various popular ‘‘radical universalist’’ approaches to syntactic structure which assume that all languages and dialects share the same basic clausal skeleton, e.g. the so- called cartography approach initiated by Rizzi (1997, see also Rizzi, 2004, for instance).5 3. Relating verb placement variation to variation in functional structure 3.1. The relevance of ‘‘rich verbal morphology’’ Giventhisbackground,wecannowaskwhatitcouldmeanto‘‘haveevidencefor’’aparticularfunctionalprojection (cf.(6)above).Theanswertothisquestionwillobviouslydependonone’sconceptoffunctionalprojectionsandtheir nature,inparticulartheirrelationshiptotheinterfacesbetweensyntaxandothercomponentsofthegrammar.Butifone takesmorphologyseriouslyandifonetakestheproposedlabelsoffunctionalprojectionslikeAgreementPhrase(AgrP) andTensePhrase(TP)seriouslytoo,thenonemightthinkthatthereissomerelationshipbetweentheseprojectionsand thestructureofverbalmorphology.ThisisthecentralideabehindThra´insson’sSplitIPParameter(SIP,cf.Thra´insson, 1996,262)6: (8) Languages that have a positive value for the SIP have AgrSP and TP as separate functional projections. Languages with a negative value of SIP are characterized by an unsplit (pre-Pollockian) IP. Crucial evidence triggering the positive value of this parameter was supposed to be ‘‘independent tense and agreement morphology’’ (Thra´insson, 1996, 269). No such trigger is available to the child acquiring standard Mainland Scandinavian, e.g. Danish, whereas it should be quite obvious in the morphology of regular verbs in Icelandic. This is illustrated below: As shown here, Icelandic has a regular past tense marker (a dental obstruent, here represented by /ð/, which alternateswith/d/and/t/asapasttensemarkerinIcelandic)easilyseparablefromtheagreementmarkers,whichvary fromonepersonandnumbertoanother.InDanishthereisjustoneformforthepresenttenseandanotherforthepast 5 Asiswellknown,thecartographyapproachassumesaverycloserelationshipbetweensyntacticandsemanticstructure.Thusitassumes,for instance, that categories like topic, focus, force, etc. project separate hierarchically ordered phrases in the ‘‘CP domain" and that SpecTopP, SpecFocPandSpecForceParethendesignatedpositionsforelementshavingthesemantic/discourserolesoftopic,focusandforce,respectively.The presentaccountismoreinlinewithapproachesthatassumethe‘‘autonomyofsyntax".Thisshouldbecomeclearbelow. 6 A similar idea about the relationship between morphology and functional structure had been proposed independently by Johnson in an unpublishedpaper1990. 1066 H.Thra´insson/Lingua120(2010)1062–1088 tense. This, then, was taken as an argument for the claim that children acquiring Danish would develop a grammar containingafusedIPofthesortillustratedin(10a)(unlesstheywereexposedtoothertypesofvisibletriggersforasplit IP)whereaschildrenacquiringIcelandicwouldhaveevidenceforthemorecomplexfunctionalstructureof(10b)7: One obvious difference between these structures is the fact that the Split IP structure of Icelandic makes two potentialsubjectpositionsavailableinthe‘‘IP-domain’’(i.e.SpecAgrPandSpecTP)whilethereisonlyoneinMSc (namelySpecIP).AspointedoutbyThra´insson(1996,273,passim;seealsoB&T,1998,52ff.),thiswouldprovidean explanationfortheclaimmadebyJonasandBobaljik(1993)andBobaljikandJonas(1996)thatMScdoesnotseemto haveSpecTP‘‘available’’asasubjectpositionwhereasIcelandicdoes.Wewillreturntothisissueinsection4.2below. WhiletheSplitIPparameterwasnotoriginallysuggestedasawayofaccountingforvariationinverbplacement, B&T(1998)showedthatgivenfairlystandardassumptionsaboutfeaturechecking,theSIPdidinfactmakeinteresting and testable predictions about verbplacement variation and these predictions were borne out by the observed MSc facts. The assumptions are as follows (B&T, 1998, 39): (11) a. The features of a projection are those of its head. b. Movement occurs solely for the purposes of feature checking. c. Featuresarecheckedinallandonlylocalrelationstoahead(viz.,head-specifier,head-complement, head-head(adjoinedheads)). Giventheseassumptions,B&TpredicttheobserveddifferencesbetweenDanishandIcelandic.First,letusconsider the reasons for the lack of V-to-I in standard MSc, as illustrated by an example from Danish above: 7Theideaisthatthisparameterisset‘‘onceandforall"foreachlanguageanddoesnot,say,dependonthemorphologicalmake-upofaparticular verb in a given sentence. Thus clauses containing irregular verbs will have split IP although irregular verbs themselves do not provide any morphologicalcuethatcouldhelpthechildsettheparameter.Similarly,nothingshouldruleouttheexistenceofasplitIPinanon-finiteclause, althoughitisnotclearthatallnon-finiteclausescontainafullfunctionalstructure.Weshallreturntothisissueinsection5.1. H.Thra´insson/Lingua120(2010)1062–1088 1067 (cid:2) VP contains all the inflectional features of læste ‘read’ (cf. (11a)). (cid:2) VP is in the local domain of I (it is the complement of I). (cid:2) Hence it is not necessary to move læste out of the VP to check the verbal features against I. (cid:2) Hence there is no V-to-I in Danish (cf. (3) above, possibly for reasons of economy). Note,however,thatitiscrucialforthisaccountofthelackofV-to-IinMScthatsentenceadverbslikethenegationare adjoinedtoVPandarenotdominatedbya(functional)projectiondifferentfromVP.OtherwisetheverbinsidetheVP wouldnotbeinthelocaldomainofI,sinceaprojectiondifferentfromVPwouldbeinterveningbetweenIandVP,and hencetheverbwouldhavetomoveoutoftheVPtocheckitsfeaturesagainstI.Thispropertyoftheanalysiswillfigure prominently in the discussion below, but it has gone relatively unnoticed in recent discussion of verb placement in Scandinavian. In Icelandic, on the other hand, the split IP forces the verb to move out of the VP as illustrated in (13): (cid:2) VP contains all the inflectional features of læsi ‘read’. (cid:2) VP is not in the local domain of Agr (it is not the complement of Agr, TP is). (cid:2) Hence læsi has to move out of the VP to make feature checking against Agr possible. (cid:2) Thus we have V-to-I (more specifically, V-to-T) in Icelandic in all types of embedded clauses (cf. (2) above).8 ItisimportanttonoteinthisconnectionthattherelationshipoftheB&TaccountofV-to-Itotheso-calledRich MorphologyHypothesis(RMH)isratherindirect.TheB&Ttheorydoesnot,asmostvariantsoftheRMH,relyonthe (typicallycircular)notionof‘‘strongfeatures’’noronsomeadhocwayofdefiningthenotionof‘‘richmorphology’’in termsofthenumberofmorphologicaldistinctionsintheverbalparadigm.ThemorphologicalaspectofB&T’stheory simplysaysthatiftheverbalmorphologyhas‘‘split’’tenseandagreementmorphemes,thenthatconstitutesevidence forsplittenseandagreementprojectionsinthefunctionalstructure,i.e.triggeringevidence,oracue,fortheSIP.The splitIPinturnforcestheverbtomoveoutoftheVPifithastocheckfeaturesagainstAgrSbecauseotherwisetheverb willnotbeinthelocaldomainofAgrS,whichisnecessaryforfeaturecheckingtotakeplace.Thusitisthenumberand natureoffunctionalprojectionsthatforcestheverbmovement,notfeaturestrengthsomehowrelatedtomorphological richness. The clear separation of tense and agreement morphemes is the only relevant notion of morphological 8 NotethatthistheoryimpliesthattheverbneednotmoveallthewaytoAgrtocheckAgr-features.Asweshallseeinsection4.5,thismakes interestingpredictionsaboutpossiblevariationinIcelandicwordorder(seealsoB&T,1998,62ff.;Thra´insson,2003,181ff.)—itshouldbenoted here,however,thatB&TassumedthataninterveningAgrO-projectionplayedaroleinverbmovementinIcelandic(see,e.g.B&T,1998,63),butas showninthepresentarticle,thisisnotanecessaryassumption. 1068 H.Thra´insson/Lingua120(2010)1062–1088 ‘‘richness’’forB&T’stheoryoffunctional projectionsandthe reason isthat itis thisseparationthatfunctionsasa trigger for a positive setting of the SIP.9 3.2. Other structural sources of verb movement out of the VP Asshownabove,themostimportantpropertyofB&T’stheoryisitsclaimthatverbmovementismovementofthe verbintothelocaldomainoftheheadthatitneedstocheckfeaturesagainst.ThustheverbneedstomoveoutoftheVP in a language with a split IP if it needs to check features against Agr. This is the source of ‘‘V-to-I’’ in Icelandic, according to B&T. But contrary to common assumptions, B&T do not maintain that ‘‘V-to-I’’ can only occur in languageswithasplitIP.Theyjustsaythatitwillnotoccurunlessafunctionalheadintervenesbetweentherelevant I-headandtheV.InlanguageswithasplitIPthis‘‘intervening’’functionalheadwillbetheT-headoftheTPandhence theVwillhavetomoveoutoftheVPifitneedstocheckfeaturesagainstAgr(seetheillustrationin(13)above).Butin a language with an unsplit (or fused) IP the V would have to move out of the VP if some other functional head intervenedandtheVhadtocheckfeaturesagainsttheI.Thisisillustratedin(14),whereFPisthe(fornowarbitrary) functional projection headed by the functional head F (see also B&T, 1998, 42): Inastructurelikethis,theVcouldnotcheckfeaturesagainstIwithoutmovingoutoftheVPsinceotherwiseit wouldnotbeinthelocaldomainofI.Asweshallseebelow,thereissomeevidencethatthiskindofV-movementis found in Scandinavian dialects that would be expected to have an unsplit (or fused) IP. Similarly,B&T’stheoryalsopredictsthatiftheVhasfeaturestocheckagainstC,itwillhavetomoveoutoftheVP in languages with an unsplit IP because otherwise it would not be in the local domain of C. This is shown in (15): Whilethismayseemtrivialatthemoment,weshallseebelowthatsomeveryrobustandinterestingevidencefor this prediction can be found in the Scandinavian languages. 9Thisobviouslyraisesthequestionofwhetherornototherkindsof‘‘separated’’inflectionalverbalmorphologymightconstituteevidenceforasplit IPandhence‘‘force’’V-movementunderB&T’stheory.Thepredictionwouldbethatifmorphologicalcategoriesofthiskindaremirroredbyfunctional projectionsabovetheVPinsyntacticstructure,thenthatshouldtriggerV-movement.AspointedoutbyAlexiadouandFanselow(2002,230),Greek tenseandaspectmorphologymightbesuchacase:Greekhasseparatetenseandaspectmorphemesanditarguablyhasverbmovementoftherelevant kindtoo. H.Thra´insson/Lingua120(2010)1062–1088 1069 4. Some relatively well known facts supporting B&T’s theory 4.1. Historical evidence for the relevance of split tense and agreement morphology The diachronic development of the Scandinavian languages is consistent with the claim that split tense and agreementmorphologyisanimportanttriggerforapositivesettingoftheSIP:OldNorseandtheoldScandinavian languagesallhadsimilartenseandagreementmorphologyasModernIcelandicandtheyallseemtohavehadV-to-Iof theIcelandickind(e.g.movementacrossnegationinallsortsofembeddedclauses,cf.Platzack,1988a,b;Falk,1993; Vikner,1995a;Rohrbacher,1999;Thra´insson,2003;Thra´inssonetal.,2004,etc.).Somerelevantexamplesaregiven below: ThestandardclaimisthatinthedevelopmentoftheScandinavianlanguagessimplificationofverbalmorphology typicallyprecedeslossofV-to-Ibyapproximately200years(seeespeciallythediscussioninPlatzack,1988a,b;Falk, 1993; Vikner, 1995a, 132ff.) whereas Icelandic has kept its split agreement and tense morphology and verb movement. These historical facts constituted an important part of the original arguments linking ‘‘rich verbal morphology’’andV-to-IinScandinavian,probablyfirstinKosmeijer’swork(1986).Althoughthisworktypically operatedwiththenotionof‘‘stronginflection’’relatedsomehowtorichverbalmorphology,thefactsareactually consistentwithB&T’sclaimthatwhatmattersisthepresencevs.absenceofsplittenseandagreementmorphology andnotsomesortofcountingofinflectionaldistinctions.Forreasonsofspace,thereaderisreferredtothesources listed above for relevant paradigms. 4.2. The numberof available subject positions As pointed out in the discussion below (10) above, the split IP structure makes more potential subject positions available than the unsplit one. Interestingly, there is some evidence that Icelandic does in fact have more subject positions available than MSc (see, e.g. Vangsnes, 1995, 1999, 2002a; Bobaljik and Jonas, 1996; Thra´insson, 1996, 2003, 2007; B&T, 1998; Bobaljik, 2002). First, observe the expletive constructions in (17) (based on examples in Vangsnes, 1995, for instance): 1070 H.Thra´insson/Lingua120(2010)1062–1088 Ithasbeenclaimed(see,e.g.BobaljikandJonas,1996;Thra´insson,1996;B&T,1998;Thra´insson,1996,2007,etc.) that facts of this sort show that there are two ‘‘subject positions’’ above the VP in Icelandic, namely SpecAgrP and SpecTP.Thefirstonecanbefilledbytheexpletiveelementþað,theotherbytheassociateoftheexpletive(thelogical subject),cf.(17a).Ifthemainverbofanexpletiveconstructionisintransitive,thentheobjectpositionintheVPisalso availablefortheassociateoftheexpletive,cf.(17b).InalanguagewithanunsplitIPontheotherhand,likeNorwegian, thereisonlyonesubjectpositionavailableabovetheVP,i.e.SpecIP.Hence(17c)isimpossiblebut(17d)isacceptable, wheretheassociateoftheexpletiveispresumablyinsomesortofanobjectposition,whichisavailabletoitsincetheverb isanintransitiveone. Second,thisapproachalso(correctly)predictsthattransitiveexpletivesshouldbepossibleinSplitIP-languageslike IcelandicbutnotinlanguageswithafusedIP,likeMSc,forinstance,sincewithatransitiveverbtheobjectposition willnotbeavailablefortheassociateoftheexpletive.Thispredictionisborneoutbyexamplesliketheonesin(18),for instance (based on Vikner, 1995a, 189): ThusB&T’stheorymakesaninterestingandtestablepredictionabouttherelationshipbetweenverbmovementand availablesubjectpositionsinexpletiveconstructions—andthereissomeevidencethatthispredictioniscorrect(fora relateddiscussionoftheavailabilityofdifferentsubjectpositionsseeAlexiadouandAnagnostopoulou,1998;fora different account of the availability of transitive expletives see, e.g. Richards, 2006). We shall return to this issue below.10 4.3. Acquisition of V-to-C and inflection If main clauses in Germanic V2 languages are characterized by the need of the (finite) verb to check some features against C, then B&T’s theory predicts that the verb should only move out of the VP to check these features if some functional projection intervenes between CP and VP, as explained above. In the adult language there is presumably always such a projection, namely some sort of an IP. Hence B&T do not predict any cross- linguistic variation with respect to this. But if the acquisition of IP is somehow conditioned by, or correlated with, the acquisition of verbal inflection, then B&T’s theory predicts that there should be a connection between the acquisition of verb movement in main clauses (the so-called V-to-C) and the acquisition of verbal morphology. Althoughtherelevantfactshavebeenknownforalongtime,theyarerarelyifevermentionedinthisconnection. Yet it was pointed out by Platzack manyyears ago (1990, 1992) that there is no evidence for V-to-C in early child Swedishuntilthereisevidenceforinflection.HeinterpretedthisasshowingthatearlychildSwedishhasnoIPandhe gave examples like the following to illustrate this connection11: 10B&T(1998)alsoarguethattheavailabilityofObjectShiftoftheIcelandickindshouldberestrictedtolanguageswithSplitIP,assumingthat ObjectShiftismovementtothespecifierpositionofanobjectagreementprojection(AgrOP)andthattherecannotbeanobjectagreementprojection unlessthereisasubjectagreementprojection.Whilethereisanintriguingcorrelationhere,IhavedecidedtoleaveObjectShiftoutofthediscussion, mainlybecauseitistoocomplexandcontroversial(foranextensiveoverviewoftheissuesseeThra´insson,2001a)—notealsothatitmayverywell bethatsplitIPisanecessarybutnotasufficientconditionfortransitiveexpletives.EvidencecitedbyAlexiadouandFanselow(2002,229–230) suggeststhatthismaybethecase.ThusFrenchandItalian,forinstance,appeartohaverichverbalmorphologyoftherelevantkind,buttheydonot havetransitiveexpletives(norobjectshift)—andthesamemaybetrueofGreek(cf.fn.9above). 11Thenatureofclausestructureinearlychildlanguagewashotlydebatedinthe1990sandthecontributionstoMeisel(ed.1992a,b)giveagood ideaoftheissuesinvolved,especiallytheusefulsummaryinMeisel’sintroductiontothevolume(Meisel,1992a,b).Theideathatearlyclausesin childlanguagehavenofunctionalcategorieswasfirstmadefamousbyRadford(1986)andhearguesforasimilarpositioninhiscontributiontothe Meisel-volume(Radford,1992,seealsoRadford,1990).Othercontributorstothatvolumearguingforatleastsome‘‘functionaldeficiency’’of clausesinearlychildlanguagesincludePlatzack(1992),MeiselandMu¨ller(1992),andClahsenandPenke(1992),whereasHyams(1992),for instance,takesadifferentview.Thereisneitherroomnorreasontogointothiscontroversyhere. H.Thra´insson/Lingua120(2010)1062–1088 1071 Similarly, as Sigrjo´nsdo´ttir has discussed in a number of publications (1999, 2005a, 644, 2005b), so-called root infinitivesoccurinIcelandicchildlanguageandheretoothereisanextremelyclearconnectionbetween(apparent) verb movement and finiteness. Some examples are given in (20) and (21): Asshownhere,thisdataisfromtheperiodwhenthetwogirlsEvaandBirnaareintheprocessofacquiringverbal inflection.Hencetheysometimesusefiniteformsandsometimestheydonot,butthepositioningoftheverbisvirtually always appropriate for the finite verb. This is shown in the overview in (22): It is important to note here that it does not matter for B&T’s theory in this connection whether the children are acquiringasplitIPorafusedone.IneithercasetheypredictthattheverbshouldhavetomoveoutoftheVPifithasto checkfeaturesagainstCbecausetheVPwillnotbeinthelocaldomainofC.UnderB&T’saccount,thismovementhas nothingtodowith‘‘strength’’oftheinflectionalfeaturesnorthestructuralmake-upoftheIPnorwiththenatureofthe sentenceadverbs.AllthatmattersisthefactthattheinterveningIPwillpreventtheVPfrombeinginthelocaldomain ofCandhenceforcetheverbtomoveoutofitifithastocheckfeaturesagainstC.Theso-calledRegionalNorthern Norwegian(ReNN)isaninterestingtestcaseinthisconnection.Inthisdialectthefiniteverbnevermovesacrossthe negationinembeddedclausesalthoughitmaymoveacrossotherkindsofadverbsaswillbediscussedinmoredetail below.Thisisillustratedin(23)(see,e.g.Hro´arsdo´ttiretal.,2006,6passim;Bentzenetal.,2007a;Hrafnbjargarson et al., 2007):
Description:The list of books you might like
