Licensing long-distance wh-in-situ in Malayalam

46 Pages·2017·0.33 MB·English

by  Athulya Aravind

Licensing long-distance wh-in-situ in Malayalam
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation Aravind, Athulya. “Licensing Long-Distance Wh-in-Situ in
Malayalam.” Natural Language & Linguistic Theory, vol. 36, no.
1, Feb. 2018, pp. 1–43.
As Published http://dx.doi.org/10.1007/s11049-017-9371-2
Publisher Springer Netherlands
Version Author's final manuscript
Accessed Thu Nov 15 11:44:15 EST 2018
Citable Link http://hdl.handle.net/1721.1/113876
Terms of Use Article is made available in accordance with the publisher's policy
and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
Detailed Terms
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 1/45
NonamemanuscriptNo.
(willbeinsertedbytheeditor)
Licensing long-distance wh-in-situ in Malayalam
AthulyaAravind
Received:30May2016/Accepted:20March2017
Abstract Itisgenerallythoughtthatwh-in-situ,likeovertmovement,ispotentially
unbounded.Atthesametime,certainlanguageshavebeenarguedtodisallowlong-
distance wh-in-situ. This paper argues that even in languages that show apparent
clause-boundednesseffects,wh-in-situ,likewh-movement,caninprinciplecrossan
arbitrarynumberofclauses.Failuretolicenseawh-phraseacrossaclauseboundary,
when it occurs, can be shown to result from the interaction between wh-agreement
andindependentoperationsaffectingembeddedclauses.Evidencewillbedrawnpri-
marilyfromMalayalam(Dravidian),whichhasbeenarguedtodisallowlong-distance
wh-in-situwithfiniteembeddedclauses.Iwillshowthattherelevantfactorforwh-
licensingisnotfiniteness,butA¯-movementofembeddedclauses,anoperationthatis
commonwithfiniteCPs.ThecoreoftheproblemliesinthefactthatinterrogativeC
isageneralized[A¯]-probethatcaninteractwithanumberoffeaturallymorespecific
goals,includingthe[A¯]-featuresontheheadofthemovingclause.Itwillbeshown
that this approach can account for a number of facts about Malayalam wh-question
formation,includingselectivetransparencyofcertainfiniteclausesforlong-distance
wh-licensing.
Keywords wh-in-situ·Malayalam·Clause-boundedness·Long-distancequestion-
formation·Relativizedprobes
1 Introduction
Languages can be broadly categorized into two types with respect to wh-question
formation. There are those that overtly move one or all wh-phrases to the front of
the clause and those that leave them in-situ. In principle, both wh-strategies can be
unbounded. Consider the long-distance wh-questions in English (1a) and Japanese
(1b). Though the wh-expressions differ in surface position, the interpretation is the
same—thewh-phraseoriginatingintheembeddedclausetakesscopeovertheentire
sentence.
DepartmentofLinguisticsandPhilosophy,MIT,Cambridge,MA,USA
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 2/45
2 AthulyaAravind
(1) a. WhatdidJohnsay[Marybought ]?
b. Hideki-ga [Kyoko-ga nani-o kat-ta to] it-tano
Hideki-NOM[Kyoko-NOMwhat-ACCbuy-PAST]say-PASTQ
‘WhatdidHidekisayKyokobought?’
Thebasicgeneralizationisthatthistypeofparametricvariationwithrespecttowh-
movement does not create deep differences in the kind of questions speakers of the
differentlanguagescanform.Islandenvironmentsaside,bothmovementandin-situ
strategiesareinprincipleunboundedandallowspeakerstoasklong-distanceques-
tionsofarbitrarylength.
However,ithasbeenobservedthatinanumberoflanguages,wh-in-situappears
to be subject to locality restrictions that are unique to this question-formation strat-
egy, an observation which, if true, raises doubts about the generalization above. In
languageslikeHindi-Urdu(Mahajan1990;Dayal1996),Bangla(Bayer1997;Simp-
sonandBhattacharya2003)andIraqiArabic(Wahba1991;Ouhalla1996;Simpson
2000), in-situ wh-expressions inside finite complements seem restricted to clause-
bound scope. On the basis of the apparent strict locality of wh-in-situ in these lan-
guages, a number of researchers have suggested that languages can parametrically
vary with respect to the locality conditions on wh-licensing (Ouhalla 1996; Naka-
mura1998;Simpson2000).
Simpson (2000), for example, argues that the domain for wh-feature checking
mayvaryacrosslanguages,proposingthefollowingthree-wayclassification(2).
(2) Wh-licensingdomains(Simpson2000:109,ex.(58))
a. TypeA:Romanian,Bulgarian Spec,CP
b. TypeB:IraqiArabic,Hindi-UrduImmediatetensedomainofthe+QC0
c. TypeC:English Thesentence
OfparticularrelevanceforourpurposesisthedistinctionbetweenTypeBandType
Clanguages.OnSimpson’sproposal,TypeBlanguageslikeIraqiArabicandHindi-
Urdu disallow long-distance wh-in-situ because a wh-phrase that remains inside a
finiteembeddedclausecannotbeaccessedforwh-feature-checking,sincetheinter-
rogative C cannot see past its immediately local tense domain. Long-distance wh-
movement in these languages, then, is in-part motivated by the need to get the em-
beddedwh-phraseintoanaccessiblewh-licensingdomain.
A similar proposal is put forth in Ouhalla (1996), who attributes the locus of
parametricvariationtopropertiesofthewh-phrasesthemselves.Morespecifically,he
argues that wh-phrases in languages like Iraqi Arabic and Hindi-Urdu are anaphors
thatcannotbeboundacrossafiniteclause.
I will refer to those languages claimed to have clause-boundedness restrictions
onwh-in-situ(whetheritberestrictionsonwh-licensingdomainsorbindingrequire-
mentsonthewh-phraseitself)asrestricted-scopelanguages,andlanguageslikeEn-
glish,Japanese,MandarinandKoreanasunrestricted-scopelanguages.Thepropos-
alsoutlinedabove,thoughpresentingstraightforwardexplanationsforwhywh-in-situ
may be more constrained in restricted scope languages than wh-movement, presup-
posethatfinitenessistherelevantfactorintheseasymmetries.Oneofthemainclaims
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 3/45
Wh-in-situinMalayalam 3
inthispaperisthatfinitenessisnotdirectlycorrelatedwithclause-boundednessef-
fects in restricted scope languages. I will argue instead that wh-in-situ is uniformly
non-clause-bound,andthatapparentclause-boundednesseffectsareepiphenomenal.
IdrawprimarilyonevidencefromMalayalam(Dravidian),alanguagethathasbeen
argued in the past to have only clause-bound wh-in-situ (Hany Babu 1997; Sriku-
mar1992;Jayaseelan2003,2004;Srikumar2007;Madhavan2013).Iwillshowthat
therestrictedscopeofin-situwh-phrasesisnotduetoageneralclause-boundedness
constraint,butisinsteadtheresultofaperniciousinteractionbetweenwh-Agreement
andA¯-operationsaffectingembeddedclauses.Specifically,licensingofawh-phrase
isdisruptedbyA¯-movementoftheclausethatcontainsit,anoperationthatishighly
favoredandsometimesobligatorywithfiniteCPs.IdevelopaMinimality-basedanal-
ysisthatexplainsthisinteraction.Licensingin-situwh-phrasesrequiresestablishing
anAgreerelationshipbetweentheinterrogativeCandthewh-phrase1(Simpson2000;
Watanabe2006).Thefeaturesontheheadofthemovingclause,however,aresuffi-
cientlysimilarto[wh]-featurestointerveneforAgreebetweenthehigherCandthe
embeddedwh,aninterventionthathascascadingeffectsonthederivationasawhole.
Crucially, when thisillicit configuration can be avoided, long-distancewh-in-situ is
possible in Malayalam, demonstrating that potential differences in the scope-taking
abilitiesofembeddedwh-phrasesincertainenvironmentsneednotreflectaradically
differentsyntaxofwh-in-situinthislanguage.
Theremainderofthispaperisorganizedasfollows.Ibeginbypresentingagen-
eralpictureofwh-questionformationinMalayalam,includingthefactthatincertain
multi-clausalconstructions,long-distancewh-in-situisdisallowed.Iwillthenshow
that the ungrammaticality of long-distance wh-questions reflects a general inability
ofwh-phrasestotakescopeoutofembeddedclauses—finiteornon-finite—thatun-
dergo clausal fronting, an A¯-operation common in the language. I then present the
coreofmyanalysis.IarguethatinterrogativeCinMalayalamisageneralizedprobe
that can interact with any number of featurally more specific goals, including those
on the head of the fronting clause. As a result, in configurations where the clause
dominating the wh-phrase bears [A¯]-features relevant for fronting, the interrogative
C erroneously makes contact with it. Next, I turn to cleft questions, which provide
furtherevidencethatrestrictedscopebehaviorofwh-in-situistheresultoffatalin-
teractionsamongA¯-operations.Incleftconfigurations,unlikesimpleclauses,in-situ
wh-phrases can take scope out of finite clauses, a pattern I will argue is due to the
factthattheillicitinteractionconfigurationiscircumventedinclefts.Iconcludewith
aadiscussionofpossibleextensionsofthepresentaccounttootherrestrictedscope
languages,aswellasunrestrictedscopelanguageslikeJapanese.
1Thesyntaxofin-situwh-expressionsismuch-debated.Atleastthreecampsofanalysiscanbefoundin
theliterature:(i)wh-phrasescovertlymovetoC(e.g.Huang1982),(ii)someQ-relatedoperator,notthe
wh-elementitself,undergoesmovement(Hagstrom1998;Cable2010)and(iii)thereisnomovement
atall(Baker1970;Reinhart1998).Aswillbeshowninsection2,acovert-movementanalysisisnot
supportedbytheMalayalamdata.Forpresentpurposes,IwillassumethatMalayalamwh-phrasesremain
intheirbaseposition.Thesyntacticlinkbetweenthewh-expressionanditsscopepositionistakentobe
establishedviaAgree.
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 4/45
4 AthulyaAravind
2 In-situwh-expressionsinMalayalam
ThemainquestionformationstrategyinMalayalaminvolveswh-in-situ.Insimplex
sentences,wh-phrasescanbeleftintheirbasepositionandyieldaquestioninterpre-
tation.Weseethisin(3).
(3) a. EethukuTTiee pustakamvaayi-chu?
which childthis book read-PERF
‘Whichboyreadthisbook?’
b. Ramaneethupustakamvaayi-chu?
Ramanwhich book read-PERF
‘WhichbookdidRamanread?’
In-situ wh-phrases in Malayalam do not appear to be island-sensitive, as illus-
trated by the fact that they are acceptable inside relative clauses (4a) and adjuncts
(4b),bothislandsforovertmovement:
(4) a. Raman[ aaruthann-a] pustakamvaayichu?
RelC
Raman[ who gave-REL]book read?
‘WhichxissuchthatxgaveRamanabookandRamanreadthatbook?’
b. NeeRaman-e [ enthuparanja-ppol]shakaarichu?
TempC
you Raman-ACC[ what said-while] scolded?
‘WhatissuchthatyouscoldedRamanwhenhesaidit?’
Thissectionwillpresentafullerpictureofwh-questionformationinMalayalam
as necessary background. I will start out by providing evidence showing that wh-
phrases in Malayalam remain in their base position and do not covertly move to C.
I will then show that despite not involving movement, in-situ wh-phrases appear to
showpuzzlinglocalityrestrictionsthatappearuniquetothisstrategy.
2.1 Malayalamwh-phrasesasgenuinelyin-situ
FocusInterventioneffectscanestablishthatin-situwh-phrasesinMalayalamremain
in their base position at the time of interpretation. Focus intervention is the name
giventothephenomenonwherein-situwh-phrasesareforbiddenfromappearingin
thescopeofcertainscopalorfocus-relatedelements,likenegationoronly.InBeck
(2006), focus intervention is argued to be at the source of the ungrammaticality of
the Korean sentence in (5), where the wh-phrase nuku-lul ‘who-ACC’ occurs in the
scope of the focus sensitive operator man ‘only’. Observe that scrambling the wh-
phrasepasttheoffendingelementtoahigherposition,asin(6),seemstorepairthe
otherwiseillicitstructure.
(5) *Minsu-mannuku-lul po-ss-ni?
Minsu-only who-ACCsee-PAST-Q
‘WhodidonlyMinsusee?’
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 5/45
Wh-in-situinMalayalam 5
(6) Nuku-lul Minsu-man po-ss-ni?
who-ACCMinsu-onlysee-PAST-Q
‘WhodidonlyMinsusee?’ (BeckandKim2006)
InBeck’s(2006)analysisofsucheffects,whichbuildsonearlierworkbyRooth
(1985) and Hamblin (1973), both wh-phrases and focus phrases introduce focus-
alternatives into the computation. Intervention occurs when the alternatives intro-
duced by the wh-phrase are “picked up” by another operator before they can reach
theappropriateinterrogativeoperator.Itisstraightforwardwhymovement,overtor
covert,should repairthe illicitconfiguration—thealternatives generatedby thewh-
phrasearenolongerinthescopeoftheoffendingoperator,andthuscannotbeacci-
dentallyaccessedbyit.
FocusInterventioneffectshavebeenarguedtodistinguishbetweencovertmove-
ment and in-situ composition of wh-phrases, sometimes within the very same lan-
guage(e.g.Pesetsky2000;Cable2010;Kotek2014).Forinstance,observingthatin-
situwh-phrasesinEnglishSuperiority-violatingquestions(7a),butnotSuperiority-
obeying(7b)questions,weresubjecttointerventioneffects,Pesetsky(2000)argued
that the two types of questions differed in where the in-situ wh-phrases were inter-
preted.
(7) a. *WhichprofessordidonlyFredintroducewhichstudentto ?
b. WhichstudentdidonlyFredintroduceto whichprofessor?
Though in both cases, the second wh-phrase is pronounced in its base position, in
caseslike(7b),itwasarguedtoundergocovertmovement,suchthatatthepointof
interpretation,itisnolongerinthescopeofthepotentialintervener.
Iftheseargumentsareontherighttrack,focusinterventioneffectscanserveasa
usefultoolfordeterminingthepositionofawh-phraseatthetimeofinterpretation,
even when this is not transparent from the actual pronunciation site. When we ap-
plythediagnostictoMalayalam,wefindthatthelanguagepatternswithKoreanand
Superiority-violatingEnglishquestionsindisplayingtheinterventioneffect.When-
ever a focus-related element occurs to the left of an in-situ wh-phrase, the resulting
structureisungrammatical,asillustratedbythecontrastbetween(8a)and(8b).Asin
Korean,scramblingthewh-phrasetoahigherpositionrepairstheviolation(8c).
(8) a. Rajanpustakangalaar-kke koduthu?
Rajanbooks-ACC who-DATgave
‘WhoalldidRajangivebooksto?’
b. *Rajanpustakangalmaatramaar-kke koduthu?
Rajan books only who-DATgave
‘WhoalldidRajangiveonlybooksto?’
c. Rajanaar-kke pustakangalmaatramkoduthu?
Rajanwho-DATbooks only gave
‘WhoalldidRajangiveonlybooksto?’
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 6/45
6 AthulyaAravind
IftheresearchtraditionofPesetsky(2000).Beck(2006)andtheirsuccessorsiscor-
rect,thesusceptibilityofMalayalamwh-phrasestointerventioneffectssuggeststhat
theydonotraisetoC,overtlyorcovertly.
2.2 Wh-asymmetries
Thewh-in-situstrategy,whichwesawtobeunproblematicinmono-clausalconstruc-
tions,neverthelessfailsincertainbi-clausalconstructions.Inparticular,wh-phrases
inside finite clausal complements appear to be highly restricted.2 In (9), where the
wh-phraseisembeddedinthecomplementofknow,whichcantakebothinterrogative
anddeclarativecomplements,onlytheembeddedquestioninterpretationisavailable.
(9) [Sitaeethupustakamvaayikk-umennu]Raman-u ariyaam
[Sitawhichbook read-FUT that] Raman-DATknow
EmbeddedQ:‘RamanknowswhichbookSitawillread.’
(cid:55)MatrixQ:‘ForwhichbookxdoesRamanknowthatSitawillreadx?’
A verb like vicaarikk- ‘think’, on the other hand, cannot take question comple-
ments. When an embedded wh-phrase occurs inside its complement, the resulting
structureisungrammatical,asneitheranarrowscopenorwidescopereadingseems
tobeavailable.
(10) *[Sitaeethupustakamvaayikk-umennu]Ramanvicaarichu?
[Sita whichbook read-FUT that] Ramanthought
(cid:55)EmbeddedQ:‘RamanthoughtwhichbookSitawillread.’
(cid:55)MatrixQ:‘WhichbookdidRamanthinkSitawillread?’
Thisrestrictionispuzzlinginlightofthefactthatthesameclauseistransparentfor
overt extraction, at least in the case of bridge verbs like say or think. Example (11)
showsthatlong-distancerelativizationfromfinitecomplementsispossible.In(12),
weseethatclefting3fromafiniteembeddedclauseispossible.
(11) [[Sita vaayikk-umennu]Ramanparanj-a ]pustakamivideunde
[[Sita read-FUT that ]Ramansaid-REL]book here COP
‘ThebookthatRamansaidSitawillreadishere.’
(12) WarandPeaceaane[Sita vaayikk-umennu]Ramanvicaaarich-athe
WarandPeaceCOP [Sita read-FUT that] Ramanthought-NOMNL
‘It’sWarandPeacethatRamanthoughtSitawillread.’
In addition, as we see below in (34), embedded wh-phrases seem to be able to
takematrixscopeoutofnon-finiteclauses.
2ThereisdebateastowhatcountsasafiniteclauseinMalayalam.Forinstance,AmritavalliandJayaseelan
(2005)arguesthatonlyclausesthatcanhostcertainmodals,moodmorphologyand“high”negationcan
beconsideredfinite.Thoughthisdebateisnotcrucialtotheissuesinthispaper—itwillbeshownshortly
thatfinitenessisnotarelevantfactorforwh-scope—therelevantfiniteclausesintheexamplesusedin
thissectionsatisfytheaforementionedcriteriaoffiniteness.
3Seesection5forevidencethatMalayalamcleftsinvolveovertmovement.
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 7/45
Wh-in-situinMalayalam 7
(13) Raman[eethupustakamvaayikk-aan]shramichu?
Raman[whichbook read-INF] tried
‘WhichbookdidRamantrytoread?’
OnemightjumptotheconclusionatthispointthatitisacorefactaboutMalay-
alam that finite clauses are scope islands for wh-expressions. One possibility, pur-
sued by some previous analyses of the phenomena (e.g. Madhavan 2013), resorts
to the Phase Impenetrability Condition (PIC) (Chomsky 2001) as an explanation.
Perhaps the matrix C cannot Agree with an embedded wh-in-situ because it is too
low within another phase to be accessible. However, there are cross-linguistic and
Malayalam-specific reasons to believe that languages have mechanisms for accom-
plishing long-distance Agree without violating the PIC. For instance, as mentioned
earlier,languageslikeJapaneseandKoreandoallowlong-distancewh-in-situacross
a phase boundary. More crucially for present purposes, even in Malayalam, long-
distance wh-in-situ across a finite-clause boundary is sometimes possible. The in-
tendedquestionin(10)canbeformedbycleftingtheentireembeddedclause.Aswe
see in (14), a wh-phrase inside a finite embedded clause can receive wide scope in
thisconfiguration.
(14) [Sitaeethupustakamvaayikk-umennu]aaneRamanvicaarich-athe?
[Sitawhichbook read-FUT that] COP Ramanthought-NOMNL
‘WhichbookwasitthatRamanthoughtSitawillread?’
It seems, therefore, that we need another explanation for the ungrammaticality of
questionslike(10).
Observethatexampleslike(10)arespecialinanotherway:inadditiontobeing
finite, the embedded clauses in such examples appear in a preposed position rather
thanthecanonicalobjectposition.Thisraisesanimportantquestion:whichofthetwo
factors—finitenessorfrontedclauseposition—isresponsiblefortheunavailabilityof
wide scope? I will argue that it is fronted clause position and not finiteness that is
directlyrelevantforwh-scope.Wefindthatwhenclausesappearinanimmediately
preverbalobjectposition,wh-phrasesinsidethemcantakematrixscope;whenthey
appearinafrontedposition,thisisnotpossible.Thenextsectionexaminesinmore
detailthecorrelationbetweenclausal-frontingandwh-scope.
3 Clausalfrontingandwh-in-situ
Malayalam is an SOV language, but clausal complements can and sometimes must
appearsentence-initially,resultinginanOSVorder,asalreadyseenin(9)and(10).
Itisoftenthoughtthereisastrictcorrespondencebetweenobligatorinessoffronting
andfiniteness—finiteclausesmustfront,whereasfrontingisoptionalfornon-finite
clauses (e.g. Srikumar 2007; Menon 2011). If this were the case, then it would be
particularly difficult to ascertain which factor is directly responsible for the disrup-
tionofwh-scope.Iwillshow,however,thatfrontingdoesnotdirectlycorrelatewith
finiteness:notallfiniteclausesmustfrontandnotallnon-finiteclauseshavetheop-
tiontoremaininanon-frontedposition.Whatisrelevantforfronting,instead,isthe
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 8/45
8 AthulyaAravind
prosodicheavinessoftheclauseinquestion.Inparticular,Iwillsuggestthatprosodic
constraintsoperativeinthelanguageinteracttoprohibitcertainheavyclausesfrom
appearing in a medial position.4 Later in this section, I will offer a precise charac-
terizationofheavinessrelevanttofronting.Whatiscrucialforusisthatwecannow
construct examples in which finiteness and clausal fronting can be divorced. Once
wedoso,weareabletoestablishthatitisfrontingandnotfinitenessthatisdirectly
responsibleforthewh-scopepatternswefindinthislanguage.
Thissectionisorganizedasfollows.Iwillfirstdemonstratethatfrontingisfeature-
drivensyntacticmovement,specificallyA¯-movement.Next,Idiscussthequestionof
why certain clauses should possess the features relevant for fronting. I will suggest
that the obligatoriness of fronting correlates with prosodic considerations, in par-
ticular, the weight of the embedded subject. Embedded clauses with overt subjects
cannot remain in their base position, for reasons that I will argue relate to prosodic
ill-formedness of such structures. But clauses with unpronounced subjects, finite or
non-finite, can remain in-situ, and in such cases, embedded in-situ wh-expressions
canfreelytakematrixscope.
3.1 Syntacticpropertiesoffrontedclauses
The fronted clause is generally the leftmost element in a given sentence. Previous
analysesoffrontedclausesgenerallytakethemtooccupyaspecificdesignatedleft-
peripheral position (Hany Babu 1997; Srikumar 2007), an analysis I will adopt in
thispaper.Iwillargue,furthermore,thatclausesundergoovertA¯-movementtothis
position. Evidence comes from obligatory reconstruction, syntactic locality effects
andparasiticgaplicensing,whichIdetailbelow.
3.1.1 Frontedclausesobligatorilyreconstructforbinding
Consider(15)below;apronouninsidetheembeddedclausecanbeboundbyaquan-
tifierinthematrix,eventhoughthepronounlinearlyprecedesitsbinder5.Thegram-
maticalityof(15),then,showsusthatfrontedclausescaninprinciplereconstructfor
bindingandthusarguesformovement.
(15) [avarude kuTTiaaneclass-il onnaamanennu]oroosthree-um vicaarichu
i i
[heri child COP class-infirst that] eachwomani-UMthought
‘Eachwomanthoughtthatherchildisfirstinclass.’
EvidencethatthismovementisoftheA¯-typecomesfromPrincipleCeffects.Itis
well-knownthatreconstructionforPrincipleCisrequiredforA¯-movement(Lebeaux
1988;Chomsky1995).Theungrammaticalityof(16a)illustratesthatfrontedclauses
inMalayalammustreconstruct.
4See Dryer (1991) for a typological survey showing that movement of medial clauses to a peripheral
positioniscommonplaceforSOVlanguages.
5Mohanan(1982)arguesthatwhatmattersforbindinginMalayalamislinearprecedence.Thisdoesnot
seemtobethecaseforthedialectspokenbymyinformants(fromthePathanamthittaregionofKerala,
India).
NALA9371_source [06/07 12:34] SmallExtended, Chicago, NameYear, rh:Option 9/45
Wh-in-situinMalayalam 9
(16) a. *[Raman aaneclass-il onnaamanennu]avan vicaarichu
i i
[Ramani COP class-infirst that] hei thought
‘HethoughtthatRamanisfirstinclass.’
b. ?[avan aaneclass-il onnaamanennu]Raman vicaarichu
i i
[hei COP class-infirst that] Raman thought
The examples in (16) do not contrast perfectly, as the embedded coindexed pro-
nounin(16b)issomewhatdegraded.Thisdegradednessisduetoablockingeffect.
Malayalam has a long-distance reflexive, taan, which takes human antecedents and
is subject-oriented (e.g. Jayaseelan 1998). When the conditions for using taan are
met,astheyareinthesubjectpositionofthecomplementofanattitudeverb,speak-
ers of Malayalam prefer to use the reflexive instead of a bound pronoun. Thus, the
counterpartof(16b)withtaan,asin(17),isjudgedmoreacceptable.
(17) [taan aaneclass-il onnaamanennu]Raman vicaarichu
i i
[SELFi COP class-infirst that] Raman thought
Caseslike(16b)areneverthelessjudgedgrammaticalbyspeakers,soItakeittobe
sufficientevidencethattheembedded clausereconstructsforbinding.Theexample
in (17) gives further support for a reconstruction analysis, since the anaphor can be
boundbyamatrixelement,evenwhenitiscontainedwithinafrontedclause.
3.1.2 Argument2:Clause-frontingisisland-sensitive
IfclausalfrontingisA¯-movement,weexpectittoobeylocalityconstraintsonmove-
ment.Wefindthatthisisindeedthecase.Relativeclausesandtemporaladjunctsare
islandsforA¯-extractioninMalayalam,asillustratedbytheimpossibilityofclefting
outofthemin(18).
(18) a. *SitaaaneRaman[ thann-a] pustakamvaayich-athe
RelC
Sita COP Raman[ gave-REL] book read-NOMNL
Intended:‘It’sSitathatRamanreadabookthatshegavehim.’
b. *SitaaaneRaman[ wann-appol] santhoshich-athe
TempC
Sita COP Raman[ came-when] become.happy-NOMNL
Intended:‘It’sSitathatRamanbecamehappywhenshecame.’
Clausalfrontingissimilarlyrestrictedwhentheclauseisinsideanisland.Asshown
below,frontingoutofrelativeclauses(19a)andtemporaladjuncts(19b)isungram-
matical.
(19) a. *[SitaWarandPeacevaayikk-umennu] njaan[t t paranj-a]
CP i CP
[Sita WarandPeace read-FUT that] I [ said-REL]
aaL-e kandu
i
person-ACCsaw
Intended:‘IsawthepersonwhosaidthatSitawillreadWarandPeace.’