The derivation of highest subject questions and the nature of the EPP

This squib argues that matrix subject wh-phrases reside in SpecCP and moreover during the course of the derivation, never move to SpecTP. This observation raises an immediate problem for languages with a strong EPP requirement on T, such as English. I argue that Chomsky (2013)’s approach to EPP effects (and other similar approaches) predict the observed pattern, while other prominent theories of the EPP fail to account for it.

Linguists have learned a great deal about (internal) language by looking at complex constructionsinanumberofdifferent(external)languages,butsometimesconstructions that seem simple on the surface can also shed light on the workings of the grammar and fundamental theoretical issues. Take for example the English highest subject question in (1).
2 It should be noted that the disagreement about the placement of the wh-phrase in highest subject questions is especially acute in English because the lack of do-supportandalsothelackofotherobviouseffects onwordorder.Inotherlanguages,suchasMainlandScandinavianlanguages,thereismuchmorerobust evidence that favors a representation that places the wh-phraseinthespecifierofCP.Takeforexample theSwedishcasesin(i).In(ia),theorderofthefiniteverbhar and negation inte indicate that the verb has raisedtotheV2positioninC.Hence,asthewh-phrase vemprecedestheverb,itindicatesthatvem occupies SpecCP.Likewiseinembeddedclauses,vemprecedesthecomplementizersom,againindicatingthatvem residesinSpecCP(seeHolmberg1986andmuchsubsequentwork).
(i) a. Vem har inte läst boken? who has not read book-the 'Who has not read the book?' b. Hon undrar vem som inte har läst boken. she wonders who that not has read book-the 'She asks/wonders who has not read the book.' WhilethisisnotasobviousfromthewordorderinEnglish,theevidencedocumentedinthenextsection shows that English is like Swedish in that highest subject wh-phrasesalsoresideinSpecCP.

Arguments for direct movement to CP
In this section, I will present arguments that the correct derivation for highest subject questions involves direct movement from the wh-phrase'sbasepositiontothespecifier ofCP.Iwillfirstpresentthreeargumentsagainstthewh-staying-in-SpecTPapproachand thentwoargumentsagainsttheintermediatemovementtothespecifierofTPapproach. 3
(5) a. What the hell did you buy? b. *Who bought what the hell?
(6) Who the hell bought a car?
(9) A: Someone bought a car. B: Who was it?
Second,onemaywonderwhetherthecorrectgeneralizationisthatanymovedwh-phrase can license the deletion of its sister. This would allow for the example in (8) to not involve ellipsisofTPasisstandardlyassumed,butratherT'. Stjepanović(1999),however,provides evidence against such an analysis. Stjepanović shows that wh-phrases in Serbo-Croatian may front below CP, but in cases where sluicing applies, the wh-phrase must resideinSpecCP. ThefinalargumentIwilldiscussinvolvesapparentcoordinatestructureconstraint(CSC) violationssuchas(12)discussedinBošković(2018).
(12) *Who left and John went to the store?
If subject wh-phrases were able to stay in the specifier of TP, as schematized in (13a) All these data suggest that subject wh-phrases do not stay in the specifier of TP and instead,likeallotherwh-phrases,movetothespecifierofCP.Thenextsectionwillfurther provide evidence that not only do subject wh-phrasesnotstayinthespecifierofTP,they nevermovetothatpositioninthefirstplace.

Arguments against intermediate movement to TP
I will present two arguments that movement of highest subject questions does not stop offinthespecifierofTP.ThefirstargumentIwillpresentcomesfromthedistribution ofquantifierfloatinWestUlsterEnglish,asdescribedinMcCloskey (2000)andfurther elaboratedinFitzpatrick (2006) These data suggest that the proper treatment of highest subject sentences is one in which the wh-phrasemovesdirectlyfromitsbasepositiontothespecifierofCPwithoutinter-mediatemovementtothespecifierofTP. 5 The second argument comes from extraction out of British do ellipsis as discussed in denDikken&Griffiths(2018).IthasbeennotedthatA-movementisallowedoutofsuch ellipsis while wh-extractionisnotallowed.Thisisshownin(17).In(17a),wehavean exampleofNP-raisingoutoftheelidedVPandtheexampleisgrammaticalevenwithdo suggestingthatA-movementispossibleoutofsuchconstructions.In(17b),ontheother hand,wehavewh-movement of an object out of the ellipsis site and the result is ungrammatical with do,suggestingthatthistypeofextractionisdisallowed.

An analysis of the Lack of EPP
Whenthesubject,aDP,mergeswiththevP,wehavethecaseofmergewithtwophrases.
There is then no head that can project the label so the label needs to be provided by either feature sharing or movement. Since the vPandDPdonotsharefeatureshere,theonly possibility is movement. The derivation continues with the T being merged into the derivation;asahead,Twillprojectasthelabeloftheobjectinquestion.TheDPthenmoves tothespecifierofTP,allowingforthev to project a label in the lower part of the structure. SincethetheDPandTshareφ-features,thenewobjectislabeledwiththosefeatures.
Returning to highest subject questions, again when the subject is first merged, the resulting structure cannot be labeled. Let us now investigate how the standard approach would deal with such data using Chomsky(2015)asamodelthoughthisholdsofanyvariantofthestandardapproach. ChomskyproposesthatTis"tooweak"toprovidealabelonitsown,whichmeansthat theresultofmergein(23a)cannotbelabeled.Trequiresmergerwithanotherelement thatitagreeswithinitsspecifierinorderforlabelingin(23b)tooccur.
Nowinhighestsubjectquestions,ifthesubjectwh-phrase were to move directly to the specifierofCP,aproblemarisesunderthisapproach.BecauseTistooweaktolabelonits own,theoutputofitsmergewithvPinthishypotheticalrepresentationisneverlabeled and the derivation does not converge as a result.
Hence,theonlyconvergentrepresentationunderthisapproachwouldhavethesubject wh-phrasemovetothespecifierofTPbeforemovingtothespecifierofCP,butthedata from allfloatanddo-ellipsis extraction presented in the previous section demonstrate that such a movement step does not in fact take place. Inthissection,IhaveshownthatthelackofmovementtothespecifierofTPinhighest subject questions can be naturally captured by approaches where movement to the speci-fierofTPisnottiedtoadeficiencyofT.Approachesthattiethemovementinquestionto suchadeficiencyfaceaproblemwiththederivationofsubjectquestions.
A number of authors (Richards & Biberauer 2005;Deal 2009; Alexiadou & Schaefer 2011; Wu 2018), however, provide evidence that there must first be merged at the edge of vP phase. 8 If this is truly the case, then the local instability of the labeling algorithm of v and there would again cause theretomovetothespecifierofTP.Tand there would agree in φ-features and label the resulting structure (there shares φ-features withtheassociateDPviaanindependentagreementrelationseeDeal2009).Itagain appears that the local instability approach can account for the data without the need ofanEPP. 9