Clitics and ClauseStructure

Cleo CondoravdiXEROX PARC andStanfordUniversityPaul KiparskyStanfordUniversity

In lateMedievalGreekandmanymoderndialects,pronominalclitics aresyntac-tically adjoinedto an IP projection. In anothersetof dialectstheyhavebecomesyntacticallyadjoinedto averbalhead.In themostinnovatingdialects(which in-cludeStandardGreek)theyareagreementaffixes.ExtendingtheFontana/Halpernclitic typology, we proposea trajectoryof lexicalizationfrom Xmax clitics via X0

clitics to lexical affixes.Theevolutionof clitic placementalsorevealstheriseofa compositefunctionalprojectionΣP.

Keywords: clitics, functionalprojections,prosodicinversion,topic, fo-cus,typology, Greekdialects,medievalGreek,historicalsyntax,syntacticchange,Pontic,Cappadocian.

1 Introduction

1.1 Syntacticchangein Greek

Researchon syntaxhasdemonstratedtheutility of a comparative-historicalapproachwhich usesco-variationandco-changeto diagnosestructuralrelationships.Suchanapproachis particularlysuitedto a languagesuchas Greek,which during its overthreemillennia of recordedhistory haschangedradically in its syntax. Moreover,Greekshowsconsiderablediversityof dialects.While nearlyall aredescendedfromtheKoinestandardof theHellenisticperiod,theyneverthelessdivergesyntacticallyinsignificantways.

Thisstudyaddressesthedialectalvariationin phrasestructureandwordorder,es-pecially asit relatesto the verbalfunctionalprojections,andattemptsa preliminaryreconstructionof its historicalorigins.1 As is well known,cliticization is oneof thebestdiagnosticsof clausestructurein languageswith relatively freeword order,andmuchof ourargumentationwill bebasedonthepositioningof verbalargumentclitics.We showthat,with respectto thepropertiesof clitics themselves,modernGreekdi-alectsfall into threemajorsyntactictypes.Ouranalysislocatesthedifferencebetweenthemin thestatusof theclitics as,respectively, Xmax, X0, andlexical. We arguethatthesestages,in thatorder,reflectashift from syntacticto lexical statusin two steps.2

Clitic placementconvergeswith othersyntacticevidencein identifying distinctsyntacticprojectionsfor TenseandMood,namelyTNSP andΣP. Thesesyntacticpro-jections,we suggest,werenot presentin the stageof Greekwhich obeyedWacker-nagel’s Law, but hademerged,in that order,at leastby the medievalperiod. ThusGreekhasundergoneadevelopmentsimilar to thatwhichhasbeenproposedfor otherbranchesof Indo-European(Kiparsky1995,1997,Deo2001),wherebytherealizationof functionalcategoriesshiftedfrom verbalinflectionsto syntax.

Thefactthattheevolutionof Greeksyntax,in theserespectsatleast,broadlyparal-lels thatof otherIndo-Europeanlanguages,is of someinterestin thecontextof recentcontroversieson the actuationof syntacticchange. On oneview, syntacticchangeproceedsin smallbut discretesteps;this view is oftenassociatedwith theassumptionthatchangehasastructurally/functionallymotivateddirectionality. An opposingview(Lightfoot 1999)advocates“catastrophic”reanalysiswith no intrinsic directionalityasthebasicmechanismof change.Ourfindingstendto supporttheformerview. Thelexicalization/grammaticalizationof clitics and the renewalof morphologicalcate-goriesin thesyntaxthroughtheriseof newsyntacticfunctionalprojectionsareclassicinstancesof unidirectional(or at leasttypically unidirectional)changewhich, astheGreekdataconfirm,proceedsin smallbutdiscreteincrements.3

1.2 Theclitic typology

In modernGreek,verbalargumentclitics arealwaysadjacentto a finite verb,but insomedialectsthey alwaysfollow or alwaysprecedeit, and in othersthey precedeor follow it dependingon whatothermaterialis presenton the left peripheryof theclause.We arguethatclitics in modernGreekdialectsareof threedistincttypes:4

TypeA (Easterntype): Xmax clitics, syntacticallyadjoinedto a maximal projection.They appear,in invariantform, both in preverbalandpostverbalpositionde-pendingonthesyntacticcontext;underourproposedanalysis,all areenclitic.

The clitics of the following dialectsareof the Xmax type: inland Asia Minor(Cappadocia,Bithynia), the Cyclades,someDodekaneseislands(Karpathos,Kos,Astipalaia),two localitiesonLesbos(Ajassos,Plomari),theTauro-Roumeicdialectsof Ukraine(Marioupoli/Azov). LateMedievalGreekis of this typeaswell.

TypeB (Pontic/Kozanitype): X0 clitics, syntacticallyadjoinedto a lexicalhead.

This type of clitic occursin two forms. X0 encliticsare found in the Ponticdialects,originally spokenon the Black Seacoastboth in Turkey (until 1922,with a small populationof Greek-speakingMoslemsremainingaroundOf inTurkey)andin Russia.X0 proclitics arefoundin Kozaniin GreekMacedonia.

TypeC (Westerntype): lexical clitics, affixedto words.

Theclitics of standardGreekareprefixeswhichattachto prosodicwordsin thelexicon. So arethoseof mostmoderndialectsof mainlandGreeceandof theWesternislands,aswell asthedialectsof Italy.

In general,all clitics in anygivendialectbehavethesameway, regardlessof gender,person,andnumber. All clitics thatareargumentsof a finite verbareconsistentlyoftypeA, typeB, or typeC.5 Thereforewecanalsospeakof typeA, typeB, andtypeCdialects.6

Halpern& Fontana1994proposea two-waydistinctionbetweenXmax andX0 cli-tics. Xmax clitics aremaximalprojectionswhich adjointo a phrasalprojectionanddonot requirea prosodichostof a particularsyntacticcategory. X0 clitics, by contrast,requirea hostof a particularsyntacticcategory. We takethis to be themaincharac-teristicof X0 clitics. Halpern& Fontana,moreover,claim thatX0 clitics arein effectinflectionalaffixes. We will arguethat their categoryof X0 clitics conflatestwo dis-tinct typesof clitics, thosethataresyntacticallyadjoinedto a lexical head(for whichwereservethetermX0 clitics) andthosethatcombinewith ahostword in thelexicon(whichwe call affixal clitics).7

1.3 Thephrasestructure

Our analysisof clitic positioningin typeA dialectsis basedon certainassumptionsabouttheirphrasestructure.In this sectionwebriefly motivatethoseassumptions.

Theclausestructureof typeA dialectsis strikingly similar in certainrespectstothatof standardModernGreek(SMG).Specifically, theysharethefollowing proper-tieswith SMG:

[1] a. theyallow verb-initialclauses;

b. theyhavethesamedistributionof negationandmoodparticles;

c. a single focusedXP or a single emphaticnegativeelementcan appearpreverballywithin theIP;8

d. they allow for multiple topics, which trigger clitic doubling if they arenonsubjectarguments,asin SMG;9

e. a preverbalfocusedXP or emphaticnegativeis alwaysto theright of anypreverbaltopics;

f. noargumentor adjunctXP canintervenebetweenapreverbalfocusedXPor emphaticnegativeandtheverb.

Thediscussionof thedistributionof clitics in section2 illustratesall theseproperties.

[2] illustratesTopic — Focus— Verborder(properties(c) and(e)) for Cappado-cian,a typeA dialect,andshowstheparallelismwith standardModernGreek.Capi-talizationmarkstheemphaticnegativeelement,andclitics areunderlined.

[2] a. Tthe

or owork

mmine

sto

KAN INAnoone

dennot

doit

xerenısko.entrust

‘I entrustmy work to noone.’ (Axos,Cappadocia;M & K 182)

b. Tithe

�ouljawork

mumine

seto

KANENANnoone

�ennot

tinit

embistevome.entrust

‘I entrustmy work to noone.’ (SMG)

We assume,uncontroversially, thatargumentsoriginatewithin theVP, andthatfi-niteverbsin GreekmovefromV to theheadof TNSP.FollowingLaka1990andPinon1993,we assumethat thehighestinflectionalprojectionis ΣP, a compositeof NegP,MoodP, andFocusP. It is headedby negation(mi, den,mina), if present,andby modalparticles(na,�a, as). FocusedXPsor emphaticnegativescanmoveto its specifierpo-sition. ModernGreekhasno V-to-C movement(Drachman& Klidi 1992),hencenowordorderasymmetrybetweenmainclausesandsubordinateclauses.Topicalizationis adjunctionto thehighestIP projection(our ΣP) andto CP (Philippaki-Warburton1985,Tsimpli 1995amongothers).

For all Greekdialectswith Xmax clitics we posit the phrasestructurein [3] (foreconomyof space,adjunctionof topicsto ΣP andto CPis notshown).

[3] CP

C′

ΣP

Σ′

TNSP

TNSP

TNS′

Spec C Spec Σ0 Cl Spec TNS0 VP{Wh-RelWh-Qu

} {FocXPEmphNeg

} {NegModPrt

}Vj . . . tj . . .

We assumethatclitics adjoin to TNSP, andthat [SPEC,TNSP] remainsempty.10 Weare agnosticas to whetherthereare functionalprojectionsother than the oneswehaveindicatedin [3].11 We leavefor furtherinvestigationthequestionwhethernon-focusedpreverbalsubjectsappearin [SPEC,ΣP]. Preliminaryindicationssuggestdi-alectalvariationonthispoint,which(in typeA dialectsat least)shouldcorrelatewithclitic positioning,accordingto ouranalysis.

2 TypeA Dialects: Xmax Clitics

2.1 Clitics in A dialects:thedescriptivegeneralizations

In typeA dialects,clitics appearimmediatelybeforeor immediatelyafterafinite verb.Thebasicdistributionof clitics wasworkedout by Dawkinsin his remarkable1916studyof thenowextinctCappadociandialects.AdditionalobservationsweremadebyJanse1998,basedonDawkins’ texts.Their rulescanbesummarizedasfollows:

[4] a. Main rule: Clitics directly follow V.

b. Specialrule: Clitics directlyprecedeV in thefollowing cases:

1. aftera negation,2. in subjunctiveor futuretenseclauses,3. afterinterrogativewh-phrases,4. afterrelativepronouns(Janse1998),5. aftersubordinatingcomplementizers(Janse1998),6. afterpreverbalphrasesin focus(Janse1998).

Thesegeneralizationshold not only for Cappadocian,but for our type A dialectsingeneral.We illustratethemherewith datafrom theTauro-Roumeicdialectsof Mari-upoli/Azov(South-EasternUkraine),citedfrom Pappou-Zouravliova1999.

[5] a. Minot

zmonatforget-2pl

namıNA me

ankevit.remember-2pl

‘Do not forgetto rememberme.’

b. Anwith

duthe

tiwhat

naNA

tuit

fsa nu?kill-1sg

Zurl enwith effort

efsaksan-du.killed-3pl it

‘What amI goingto kill it with? With difficulty theymanagedto kill it.’

Significantly, all thesedialectsconformto thegeneralizationsin [1], which diagnosethepresenceof at leastonefunctionalIP projectionaboveTNSP,accordingto usΣP.

Assumingthephrasestructurein [3], thedistributionof postverbalclitics in dialectA is characterizedby thefollowing descriptivegeneralization:

[6] Clitics arepostverbalif andonly if thereis no non-adjoinedconstituentwithinthesameCPto theleft of theclitic.

We showbelowthatthesyntacticassumptionsin section1.3accountfor thedescrip-tive generalization[6]. To do that, we demonstratethat, undertheseassumptions,clitics arepostverbalexactlywhentheycannotbepreverbalbecausethereis no hostfor themin thatposition.

2.2 Explainingthedistribution

Onthesurface,it appearsthatin TypeA dialectstheclitic or theverbappearin at leasttwo differentsyntacticpositions.We arguethattheclitic andtheverbalwaysappearin thepositionindicatedin [3]. Clitics areadjoinedto a functionalprojectionwhoseheadthe verb movesto, namelyTNSP. The distributionof clitics is a consequenceof their syntacticand prosodicproperties. Clitics prosodicallysubcategorizefor aprosodicword on their left within thesameCP. Adjoinedconstituentsarenot visiblefor cliticization.12 If thereis no availableprosodichostto their left, theyencliticizeontotheadjacentwordon their right by PROSODICINVERSION (Halpern1995).13

We view prosodicinversionasanoptimizationstrategywhich ensuresbestsatis-faction of the cliticization requirementplus the twin syntacticconstraintsthat inputorderof clitics mustbepreservedandthatclitics remainwithin thesameCP. Accord-ing to ourproposal,then,postverbalpositioningof clitics is thespecialcase(contrasttheformulationin [4], which in effecttreatsit asthedefault).14

An interestingalternativeis theideaof Boskovic2001thatclitics areplacedby thesyntaxin morethanoneposition,andthatphonologicalconstraintsfunctionasfilterswhich selectamongthosepositions.Boskovic showsthat it worksstraightforwardlyfor Serbo-Croatian.For Bulgarianclitics, which aresimilar to the clitics of type AGreekdialects,Boskovic suggeststhatclitics form a movementchainwhoseheadispronounced,exceptif thiswould leadto aphonologicalviolation,in whichcasesomelower copyof theclitic in thechainis pronounced.This accounthastheconceptualvirtue of locatingall movementin thesyntax,andmakessomeinterestingempiricalpredictions.It maywork for theGreekXmax dialects.We do notadoptit herebecausewe do not seehow to extendit to the dialectstreatedbelow, including crucially theKozanidialect(section5.4).

It would be desirableto havea theoryof clitic placementin which all systemswould find their place. From the Optimality-Theoreticperspective,the differencebetweenthe Serbo-Croatiansystemandthe Greek/Bulgariansystemcanbe seenasa matterof constraintdomination;in Serbo-Croatian,syntacticconstraintsdominatethe phonologicalconstraints;in type A Greek,phonologicalconstraintsconverselydominatethesyntacticconstraints,forcingprosodicinversion.

A lesssatisfactoryalternativeis thatthepostverbalpositioningof clitics is duetosyntacticfactorsalone,suchasverbmovementovertheclitic. For instance,onecouldassumethattheverbmovesto Σ0 if theΣ andC projectionsaredevoidof anylexicalmaterial.Whatwouldbethesyntacticmotivationof suchamovement?Terzi1999,inananalysisof thepositioningof Cypriot clitics, which appearsto belike thatof typeA dialects,arguesthat theclitics needa syntacticlicenserandin theabsenceof anyotherlicensertheverbmovesto the highestprojectionwithin the IP, in heranalysistheMoodP, in orderto licensetheclitics. Onereasonwe do not adoptthis proposalis that themotivationfor syntacticlicensingseemsratherweak. Thesetof licensersincludesboth functionalheads,like negation,modalparticles,andcomplementizers,

aswell asheadsof non-functionalprojections,suchastheheadof a preverbalfocusphraseor of a wh-phrase.It would bea strangesyntacticlicensingrequirementthatcouldbesatisfiedby sodisparateasetof licensers.15

2.3 PreverbalClitics

Clitics arepreverbalif andonly if thereis somenon-adjoinedconstituentwithin thesameCP to the left of the clitic. This may be a complementizer(in C0), a Wh-element(in [Spec,CP]),a negativeor modalparticle(in Σ0), or a focusedconstituent(in [Spec,ΣP]). We takeup eachof thesecasesin turn. (As before,the clitics areunderlinedin ourexamples.)

2.3.1 Complementizers

Thefirst predictionis thata lexical (overt)complementizerin C0 hostsa clitic whichis syntacticallyto its immediateright. Theexamplesin [7] demonstratethispreverbalpositioningof theclitic afteravarietyof subordinatingconjunctions.Notethecontrastin clitic orderingbetweenthetwo clausesin [7c,g].

[7] a. Opwhile

tohim

paisge,take-PastImp-3sg

ırtecame-3sg

enaa

binarspring

kundanear

‘As hewastakinghim, hecameneara spring.’ (Ulaghatsh,Cappadocia;D 366)

b. tonwhen

doit

emaxenlearned-3sg

‘whenhelearnedit’ (Axos,Cappadocia;M & K 216)

c. mongewhen

tonhim

ı�esaw-3sg

etsout

e�oksendonsent-3sgawayhim

‘Whenhesawhim, hethrewhim out.’ (Demirdesi;Dang174)

d. ospuuntil

tupanhim said-3pl

ithe

�askalıteachers

tuhis

‘until his teacherstold him.’ (Astypalaia,Dodekanese;DD 56)

e. �arobelieve-1sg

posthat

tafaethemate-3sg

tathe

pit akjalittle pies

‘I believethatheatethepies.’ (Pyli, Kos,Dodekanese;DD 230)

f. tothe

pult ınbird

tuthe-Gen

�ndzamantsudiamond-Gen

puthat

muferesmebrought-2sg

‘thebird with thediamondthatyoubroughtme’ (Astypalaia,Dodekanese;DD 59)

g. Iferabrought-1sg

to ,him

jat ıbecause

tonhim

ı�ela.wanted-1sg

‘I broughthim becauseI wantedhim.’ (Karpathos,Dodekanese;Minas1970)

2.3.2 Wh-phrases

In relativeclausesandin matrixor embeddedwh-interrogatives,thespecifierof CPisoccupiedby a relativepronounor by an interrogativewh-phrase.Therefore,a cliticwill alwaysappearpreverballyin relativeclauses,asin [8a], or in wh-questions,asin[8b,c]. As in StandardModernGreek,the(CP-adjoined)topic in [8c] is to theleft ofthewh-element(in [Spec,CP]).

[8] a. opwhoever

toit

dranasees-3sg

‘whoeverseesit’ (Axos,Cappadocia;M & K 57)

b. tseand

rot usenasked-3sg

othe

enasone

tonthe

altonother

ındawhat

tusthem

ıt�elenwanted-3sg

othe

�asiltasking

‘and theywereaskingeachotherwhat theking wantedthemfor’ (Asty-palaia,Dodekanese;DD 57)

c. Etothis

tothe

beırstallion

cıswho

toit

epkenmade-3sg

auca?thus

‘Who madethisstallionlike this?’ (Delmeso,Cappadocia;D 314)

2.3.3 Negationandmodalparticles

Negationandmodalparticles,we assume,areheadsof ΣP. Therefore,whensuchaparticleis present,therightmostlexically filled positionbeforetheclitic is Σ0, whichhoststheclitic; hencenoprosodicinversionis necessary. Observethecontrastin cliticorderbetweenthetwo clausesin [9e].

[9] a. Tucithis

cinthe

goridaughter

zarjanıpresent

tuhis

enekawife

r ennot

cinher

a apisilove-3sg

‘This daughterhispresentwife doesnot love.’ (Silli; D 300)

b. epseyesterday

�enot

seyou

kratsamkept

‘Yesterdaywedidn’t putyouup’ (Demirdesi;Dang176)

c. Enot

siyou

skutonu,kill-1sg

naNA

mime

padreps.marry-2sg

‘I won’t kill yousothatyoufind mea wife.’ (Plomari,Lesvos;K 492)

d. �ennot

tahis,themhave-2sg

�ennot

syou

ehunhave-3pl

‘Let it goandit will let yougo.’ (Asphendiou,Kos,Dodekanese;DD 229)

e. He osGod

esergetsenprotected

dathem

keand

l ıkoswolf

dennot

dathem

ıvren.found

‘God protectedthemandno wolf foundthem.’ (Axos,Cappadocia;M &K 188)

f. asAS

toit

piaso,catch-1sg

asAS

toit

kopso,kill-1sg

keand

kalawell

asAS

toit

fa oeat-1sg

‘Let me catchit, let me kill it, and let me eat it right up.’ (Ulaghatsh,Cappadocia;D 366)

Accordingto ouranalysis,apreverbalclitic encliticizesontotheelementonits leftratherthanprocliticizing onto theverb. Striking evidencefor this comesfrom moodparticleswhich do not constituteprosodicwordson their own,but do togetherwith aclitic (activesubcategorization,Inkelas1989,Halpern1995,Boskovic 2001:160).Forexample,in theCappadociandialectof Ulaghatsh,na is stressedjust in casea cliticfollows it:16

[10] a. Derenow

�a�afather

mmy

naNA

ert,come-3sg

geand

naNA

seyou

rot isask-3sg

. . .

‘Now my fatherwill comeandwill askyou.’ (Ulaghatsh,Cappadocia;D366)

b. naNA

ıtosewere-2sg

koundanear

m,me

tothe

gaidurdonkey

naNA

toit

pj asoumcatch-1pl

ton,were-3sg

keand

naNA

tosell-1pl

pulısumit

ton,were-3sg

naNA

parumget-1pl

liolittle

kirjasmeat

. . .

. . .NaNA

toit

efa ameat-1pl

ton.were-3sg‘Had you beenby me,we would havecaughtthedonkey, we would havesold it to geta little meat. . . .We would haveeatenit.’ (Ulaghatsh,Cap-padocia;D 366)

2.3.4 Focus

Preverbalfocusand emphaticnegativesare housedin [Spec,ΣP]. In [11] and [12]sucha focusedelementconstitutesthe rightmostpre-clitic positionwith lexical ma-terial. The examplesin [12] areanswersto wh-questions,with the focusedphrasecorrespondingto thewh-phraseof thequestion.

[11] a. [FOC Polımuch

] dohim

se�dini sgeloved-3sg

‘Shelovedhim much.’ (Ulaghatsh,Cappadocia;D 366)

b. [FOC Kal awell

] tonhim

elasa.tricked-1sg

‘I trickedhim well.’ (Demirdesi;Dang190)

c. esıyou

tsaozuntil

derenow

ıle essaid-2sg

keand

[FOC e elfobrother

] toit

ehishave-2sg

keand

a apanesloved-2sg

to.it‘Up until nowyouweresayingit (thedeer)wasyourBROTHERandyoulovedit.’ (Axos,Cappadocia;M & K 192)

[12] a. Ehohave-1sg

enaa

kori c,daughter

kand

[FOC ekınoshe

] topkenit said

‘I haveadaughterandSHEsaidit.’ (Delmeso,Cappadocia;D 314)

b. [FOC I oI

] tunhim

anj ksaundressed-1sg

‘ I undressedhim.’ (Plomari,Lesvos;K 493)

2.3.5 Topicversusfocus

A topic aloneneverattractsa clitic to thepreverbalposition.On theotherhand,ele-mentswithin ΣP, suchasfocus,modalparticlesandnegation,which follow preverbaltopics,alwaysattracttheclitic to thepreverbalposition.[13] providesaminimalcon-trast. All threeexampleshavea nominalizedclausalcomplementin topic position,ascanbe seenby the presenceof clitic doubling. [13a] hasan emptyΣP, henceapostverbalclitic, while [13b] hasthespecifierof ΣP lexically filled, and[13c] hastheheadof ΣP lexically filled; in thesetheclitic is preverbal.

[13] a. [TOP tothat

sema eftengot-3sgengaged

] ıpantold-3pl

masus

tait

(TOPIC )

‘That hegotengaged,theytold usaboutit.’(Axos,Cappadocia;M & K 85)

b. [TOP tothat

psofsendied-3sg

tthe

alo ohorse

] [FOC derenow

] tit

akuohear

(TOPIC , FOCUS)

‘That thehorsedied,I only heardit now.’(Axos,Cappadocia;M & K 85)

c. [TOP tothat

naNA

azandesowin-1sg

utsathus

pollamany

] dennot

dit

omzahoped

(TOPIC , NEG)

‘That I wouldwin somany, I didn’t hopefor it.’ (Axos,Cappadocia;M &K 85)

2.4 Postverbalclitics by prosodicinversion

Whenthespecifierandheadpositionsof CPandΣP areempty, minimal compliancewith therequirementthattheclitic mustencliticizeontosomethingforcestheclitic tobeplacedafterthefirst word,which,giventhesyntax,is theverb(prosodicinversion).Thesimplestcaseof postverbalclitics, illustratedin [14] by examplesfrom four typeA dialects,ariseswhentheclitic is syntacticallyCP-initial.

[14] a. Pulsasold-1sg

tathem

tathe

devja.Devs

‘I soldthemto theDevs(spirits).’ (Ulaghatsh,Cappadocia;D 378)

b. Vreıstencalled-3sg

doher

keand

geletzepsan.talked-3pl

‘He calledherandtheytalked.’ (Axos,Cappadocia;M & K 216)

c. �okasegave-3pl

dhim

enaa

ftir owing

‘They gavehim awing’ (Plomari,Lesvos;K 490)

d. Ekame-mas-tomadeusit

enasa

ftoxospoor

erosold man

‘A poorold manmadeit for us.’ (Demirdesi;Dang176)

Forreasonswhichremainto beexplored,cliticizationis blockedacrossanadjunc-tion boundary. A clitic following an adjoinedconstituentundergoesprosodicinver-sionin TypeA dialects.Adjunctionoccursin coordinationandtopicalization.A cliticwhichsyntacticallyfollows aconjunction,suchaske/tse/ce‘and’ andame ‘but’, mustundergoprosodicinversion:

[15] a. Toranow

metanjosa,repent-1sg

keand

ksomoloumeconfess-1sg

suyou

toit

‘Now I repentandI confessit to you.’ (Pyli, Kos,Dodekanese;DD 233)

b. cand

ekanimade-3sg

dunhim

limn j

lake‘andturnedhim into a lake.’ (Ajassos,Lesvos;K 485)

c. Amebut

nzuloftonais jealous

tonhim

tthe

afendikonmaster

tuhis

tseand

�nzoxnisend-3sgaway

to,him

tseand

leitells

tuhim‘But his masteris jealousof him andsendshim awaytelling him’ (Asty-palaia,Dodekanese;DD 56)

A corollaryof thisanalysisis thedifferentialbehaviorof clitics in initial andnon-initial conjuncts.When,for instance,ΣPsareconjoinedwithin aCPwith somelexical

materialin the C projection,a clitic in the first conjunctfinds a host to its left, butclitics in subsequentconjunctsmay be forcedto prosodicallyinvert. Dawkins1950notesjust this effectfor dialectsof theDodekanese,andwe havealsofoundit in hisCappadociantexts.

[16] emename

anif

meme

baterdeisdip

tr iathree

for as,times

keand

v aliztakeout

meme

sothe

milumill

sothe

sifon,stream

‘if youdip meandtakemeoutthreetimesin themillstream,(Silli, Cappadocia;D 308)

Thereversionto postverbalclitics in non-initial conjunctsis predictedon thepresenttreatment.Thesyntacticstructureis displayedin [17]; theclitic in thesecondconjunctundergoesprosodicinversionwhile theonein thefirst conjunctencliticizesontoan.

[17] [CP [C’ an[ΣP [ΣP [TNSP Cl V . . . ]] ke [ΣP [TNSP Cl V . . . ]]]]]]

Turning to theothercaseof adjunction,namelytopics,a clitic that immediatelyfollows an argumenttopic syntacticallyundergoesprosodicinversionin Type A di-alects.[18a] illustratesthis; comparewith thestandardGreekversionof thesentencein [18b],wherethesameNPis alsoa topic,triggeringclitic doubling.

[18] a. [TOP Tothe

l ıkowolf

] r otsanasked-3pl

dohim

. . .

‘They askedthewolf . . . ’ (Axos,Cappadocia;M & K 182)

b. [TOP Tonthe

likowolf

] tonhim

rotisanasked-3pl

. . . .

‘They askedthewolf . . . ’ (ModernGreek)

Adjuncttopics(suchasadverbialmodifiers)canadjointo ΣP, with prosodicinversionunderexactlythesameconditions:

[19] [TOP sımertoday

tothe

purnomorning

] [TOP pourbefore

naNA

paen,leave-3sg

] ekamendesmade-3sgthem

mwith

enaone

looword

this

enaa

spithouse

‘This morning,beforeheleft, hemadethema housewith oneword.’ (Demird-esi;Dang176)

In orderto justify thisanalysis,it is importantto beableto identifyapreposedcon-stituentasatopic. Topicsservecertaindiscoursefunctions,andnon-subjectargumenttopicstriggerclitic doubling,asin StandardModernGreek.Therefore,aclitic relatedto a topic will appearpostverballyif thereis noappropriatepreverbalmaterialwithin

thesameCPto hostit. Thepredictedcorrelationis documentedfor a rangeof casesin theexamplesbelow. In all of thesecasesModernGreekpermitspreverbaltopicsaswell. [20] illustratesthat subsectionalanaphorsaretopics. A clitic immediatelyfollowing it in thesyntaxis placedaftertheverb.

[20] a. EnasA

patisahosking

ıhehad

tr ıathree

perj a.sons

[TOP Tathe

rj otwo

] dıkisenmarried-3sg

da.them

‘A king hadthreesons. Two of themhe marriedoff.’ (Ghurzono,Cap-padocia;D 340)

b. Enasa

vasiljasking

ihehad

triathree

agorja.sons

[TOP Tathe

diotwo

] tathem

pandrepse.married-3sgoff

‘A king hadthreesons.Two of themhemarriedoff.’ (ModernGreek)

[21] illustratesthatthesameis truefor contrastivetopics:

[21] a. ekınoshe

pırentook

tithe

vasileking

tinthe

gordaughter

keand

[TOP tothe

ambrobridegroom

]

edosandongave-3plhim

tinthe

a�refısister

this

‘He marriedthe king’s daughterandthey married(the would be) bride-groom(of theking’sdaughter)with hissister.’ (Demirdesi;Dang220)

b. Ekinoshe

pandreftikemarried

tinthe

koridaughter

touthe-Gen

vasiljaking-Gen

keand

[TOP tonthe

ambro]bridegroom

tonhim

edosangave-3pl

stinto the

a�elfisister

tuhis

‘He marriedthe king’s daughterandthey married(the would be) bride-groom(of theking’sdaughter)with hissister.’ (ModernGreek)

A topiccanalsointroduceashift in narrativeperspective.If thereis noothermaterialbetweenthetopicandtheclitic, theclitic is againpostverbal.

[22] [TOP Imiswe

] l j p�ıkamefelt sorry

dun,him

pıramitook-1pl

mnj aa

varkaboat

ciand

pı miwent-1pl

ciand

pj asamecaught-1pl

dunhim

‘We felt sorry for him, we took a boatand went and savedhim.’ (Plomari,Lesvos;K 495)

3 TypeC Dialects: clitics asword-level affixes

3.1 Thedistributionof clitics in C dialects

In typeC systems,clitics directly precedethefinite verbwhoseargumentstheyare.Thepropertiesof typeC dialectsarewell knownfrom SMG.Thepatternis illustratedin [23].

[23] a. Tisher-gen

toit-acc

ipa.said-1sg

‘I saidit to her.’

b. Tisher-gen

toit-acc

exohave-1sg

pi.said

‘I’ve saidit to her.’

c. NaNA

sasyou-pl

posay-1sg

tinthe

alithia,truth

othe

jatrosdoctor

mume

egrapsewrote

enaa

polivery

kalogood

siropi.syrup‘To tell you the truth, thedoctorwrotemea prescriptionfor a very goodsyrup.’

d. Moume-Gen

anikseopened

ithe

mitinose

mjaone

xara.joy

‘My noseclearedverywell.’

Thispatternis widespreadin mainlandGreece;theexamplesin [24] and[25] illustratethattheGreekdialectsspokenin Italy alsoconformto it.

[24] Salento(Profili 1999)

a. NaNA

sasyou-pl

posay-1sg

pr obbiotruth

tothe

justo,real

othe

medekommudoctormy

mme

ois

grammenawritten

enaa

scerupposyrup

kathat

eis

ppr obbioreally

kkal o.good

‘To tell you the truth, thedoctorwrotemea prescriptionfor somesyrupwhich is reallygood.’

b. Mume-Gen

svuddhıetedischarge

ethe

mittinose

miaone

bbelletza.beauty

‘My noseis clear,just like that.’

c. Mabut

quaisome

pr amatathings

‘enot

mmasus

eddiaennanepassthrough

mankueven

attifrom the

cciofali.head

But therewerecertainthingsthatwe wouldneverhavethoughtof.’

[25] Calabria(Katsoyannou1999)

a. o�ehere

tinit

ekamandid-3pl

dithe

skolaschool

‘They wentto schoolhere.’

b. E�ohere

ennot

doit

ıserenknew-3sg

ganes.noone

Artenow

toit

mathealearned-3pl

iatıbecause

e assanleft-3pl

osotheso

ceand

otuthus

toit

mathealearned-3pl

‘No-one knew it here. Now they’ve learnedit becausethey’ve left (thevillage)sotheylearnedit thatway.’

c. ıxenhad-3sg

ecınuthose

puthat

selawanted-3pl

naNA

paugo-3pl

mabut

ennot

dusthem

afınnalet-3pl

‘Thereweresomewhowantedto go,but theywouldn’t let them.’

3.2 Derivingthedistributionof clitics in C dialects

We assumethat type C dialectshavethe clausalstructure[3], like type A dialects.Thereis justonedifference:clitics areaffixeswhichsubcategorizefor aphonologicalwordon their right. Therefore,theydonotattachsyntacticallyto TNSP,but lexicallyto theleft of afinite verb.As partof thefinite verb,theymovewith it to TNS0. Impor-tantly, theyareword-levelaffixes(not stem-levelaffixes,like the subjectagreementmorphemesof Greek),thatis to say, theyattachto words,forminglargerwords.

We areadoptingherethefollowing tripartitecategorizationof affixes,empiricallymotivatedand theoreticallyjustified in Kiparsky (to appear):stem-to-stemaffixes,stem-to-wordaffixesandword-to-wordaffixes(lexicalclitics). Thesethreecategoriesareassumedto beuniversal,buttheallocationof morphemesto themis notpredictableandnotall languagesnecessarilyhaveall typesof affixes.

Thatclitics in standardGreekarelexical affixes(albeitnot word-to-wordaffixes)hasbeenarguedfor by Joseph1988on thebasisof phonologicalandmorphologicalevidence.A syntacticargumentis that theydo not combinelexically with non-finiteverbs. It is virtually a definitionalpropertyof agreementmorphemesthat they areaffixedonly to finite verbs. For example,subjectagreementin all Greekdialectsisrestrictedto finite verbs. If objectclitics arelexical agreementmorphemes,we canunderstandwhy theyobeythis restriction;otherwiseit remainsunmotivated.

A secondargumentthatcliticsarelexicalaffixesin typeCdialectsis thatconjoinedfinite verbscannotsharea clitic. If clitics weresyntacticallyadjoinedto a V0 head,thenin principle theyshouldbecapableof beinghostedby a conjoinedV0 head(astheyin factarein thedialectswheretheyareX0 categories,suchasPonticandKozani,seebelow). Sentenceslike [26] are, however,ungrammaticalin C dialects(in theintendedinterpretation).

[26]*toit

eliosemelted

kiand

exaselost

‘Shemeltedit andlost it.’

In contrast,sentenceslike [27] aregrammatical,sincetheauxiliary with theattachedclitic hasscopeoverbothconjuncts.

[27] Ekinothat

tothe

vra�ievening

meme

ixehad

vrisicursed

keand

xtipisihit

‘That eveninghehadcursedmeandhit me.’

[28] showstherelevantstructuresof [26] and[27].

[28]

V0

V0 ke V0

to+eliose exase

TNSP

VP

V0

me+ixe vrisi ke xtipisi

Sinceweclaimthatclitics areword-levelaffixes,adoptingthedistinctionbetweenstem-levelandword-levelaffixesarguedfor in Kiparsky (to appear),our positionisto be distinguishedfrom that of Joseph.Not only arelexical clitics predictedto beon theoutsideof all inflectionalaffixes,theyare,moreover,predictedto differ frominflectionalaffixesin their phonologicalproperties.The reasonis that stemssatisfythestemphonology, while wordssatisfytheword phonology. We believethatexist-ing phonologicalor morphologicalargumentsagainstthe lexical statusof clitics instandardGreek,suchasthosein Philippaki-Warburton& Spyropoulos(1999),arear-gumentsagainsttheanalysisof clitics asstem-levelaffixes,ratherthanasword-levelaffixes.

4 TypeB Dialects: SyntacticX0 Clitics

4.1 Ponticclitics arealwayspostverbal

In themajority of Ponticdialects,theplacementof clitics is easilystated:clitics arealwayspostverbal(Papadopoulos1955,Oikonomidis1958,Cernyseva1958,Drettas1997),evenin environmentswheretheyarepreverbalin theotherdialects(seesection2.3):17

[29] ti�ennothing

khjnot

le netell-3pl

menme

(NEGATION )

‘They tell menothingatall.’ (Dr 632)

[30] a. asAS

akugnahear

tait

ekfrom

defternsecondtime

(M OOD PARTICLE )

‘Let ushearit a secondtime.’ (Dr 632)

b. prinbefore

apo�andies

pr epmust

naNA

�ijsegive-3Sgyou

vesaættestament

(M OOD PARTICLE )

‘Beforehedies,hemustgiveyouhis testament.’ (Dr 380)

[31] ondaswhen

telionfinish-1sg

atoit

(COMPLEMENTIZER )

‘when I finish it’ (Trapezounda;P 224)

[32] dowhat

lestell-2sg

meme

(WH-INTERROGATIVE )

‘What areyoutelling me?’ (P159)

[33] ekinothat

[FOC e oI

] exerknow

atoit

(FOCUS)

‘Only I knowthat.’ (Trapezounda;P 224)

TheOphiticdialectsspokenby thePonticMoslems(Mackridge1987,1999)representthesamekind of system:

[34] a. Polamuch

nasubj

xarenumesbe-glad

alomıjananother-time

anif

elepumesee

sas.you

‘We shallweverygladif we seeyouagain.’

b. Es’you

apolı umesend-1Pl

sasyou

soto-the

ReppıGod

muas

emanet.pledge

‘We’resendingyouasa pledgeto God.’

4.2 Ponticclitics arenotsuffixesbutX0 enclitics

Clearly, Ponticclitics areenclitic ratherthanproclitic. However,theydiffer from bothtypeA andtypeC clitics in a subtlerway. It would be temptingto seePonticasthemirror imageof SMG,andindeedDrettas1997claimsthattheyareobjectagreementsuffixes(seealsoJanse1998).Wethink thatPonticclitics requireasyntacticanalysis.Our proposalis that theyarephonologicallyenclitic (just asin typeA dialects),but

that they are of categoryX0 rather than of categoryXmax. Consequently, they arehead-adjoinedto V0, ratherthanadjoinedto aphrasalfunctionalprojection,andtheirsyntaxdiffers from thatof TypeA clitics accordingly. TheX0 statusof Ponticcliticsis supportedby thefollowing threearguments.

First, in theperfect,clitics in Ponticareattachedto theinfinitive, not to theauxil-iary:

[35] anif

ihamehad-1pl

ndosnebeaten

se,you

iheshad-2sg

ma�inelearned

tothe

ma�emalesson

syours

‘If we hadbeatenyou, you would havelearnedyour lesson.’ (Trapezounda;P174)

Sincelexical agreementaffixes (morphologicalargumentclitics) go only on finiteverbs(section3.2),thisshowsthatclitics arenotagreementaffixes.

Secondly, conjoinedverbsmayshareasingleclitic, which thenalwaysappearstotheir right.

[36] a. esegenput-3sg

tothe

vutoronbutter

sin

sonthe

furninoven

kand

elisenmelted

kand

exasenlost

a.it

‘She put the butter in the oven and meltedit and lost it.’ (AdissaAr-giroupoleos;P 200)

b. ekombo�enwasduped

kand

eksegkentookout

kand

e�ekenput

atonher

simanear

thim

‘He wasdupedandtook herout andput hernearhim.’ (Trapezounda;P22)

c. Ekınoshe

paPA

eprostaksenordered

nasubj

luzneshampoo

keand

plınwash

atenhim

‘He orderedthemto shampoohimandtowashhim.’ (Imera,Fostiropoulou1938:190)

Thissharplycontrastswith standardGreek,wheretheclitic is obligatorilyrepeatedinsuchcases,aswouldbeexpectedfor anaffix.18 Thebehaviorof clitics in conjunctionthusconfirmsthattheyarelexical in standardGreekandsyntacticin Pontic.

Thethird argumentcomesfrom phonology, whichshowsthatclitics arenotpartofthesamelexicalwordastheirhosts(thoughtheyaresurelypartof thesamepostlexicalword). The argumentis basedon a stresscontrastbetweensimple long wordsandwordswith attachedclitics. In simplelong words,whenthe lexical stressis beforethe antepenultsyllable,rhythmicalternatingstressesareassignedto the word (e.g.,eklapsa,eklapsane,ekimumunestine). However,nosuchadditionalstressesappearinclitic sequences,asexplicitly statedby Papadopoulos1955:32.If clitics werelexicalsuffixes,thisdifferencewouldbeunmotivated.

In supportof his claim that Pontic clitics are affixes, Drettas1997 arguesthattheycombinewith theirhostsin phonologicallyidiosyncraticways.Drettas’principalargumentis thatthethird personobjectformsfaisen,fazætonin [37] cannotbederivedby phonologicalrules.19

[37] ‘you’ (Sg.) ‘him’1Sg. fazo fazosen fazaton2Sg fais —— faisaton3Sg faz faisen fazæton

In orderto assessthis argument,considerfirst theparadigmsof thesimpleverbsfazo‘feed’, pleko‘knit’, siro ‘drag’, andvrexo‘rain’ in Pontic.

[38] /faz-/ /plek-/ /sir-/ /vrex-/1.sg. /-o/ fazo pleko sıro vrexo2.sg. /-is/ fais pleks sırts vreis3.sg. /-i/ faz plek sır vres1.pl. /-omen/ fazomen plekomen sıromen vrexomen2.pl. /-eten/ fazeten pleketen sıreten vreseten3.pl. /-ne/ fazne plekne sırne vrexne

Regularphonologicalprocessesof Ponticaccountfor theseinflectionalpatterns.Thealternationof x ands in vrexo, vrexomen, vrexneversusvresetenis dueto palataliza-tion of x to s beforea front vowel. Thesamepalatalizationprocessalsoaccountsfor3.Sg.vres. Apocopeof final i is a productivephonologicalprocessin Pontic. It ismotivatedby suchcontrastsas/podari/ podar ‘foot’ versus/podari-mu/podarim ‘myfoot’, with retentionof non-final-i in the latter form. Theprocessseemsto beauto-matic, in thatno phonologicalphraseor phonologicalword canendin -i. Thus,weposit thethird personendingas/-i/, which triggerspalatalizationin /vrex-i/→ vresi,andis obligatorilyapocopated.

Thesameregularphonologicalprocessesapplyto clitic combinationsaswell. Thephonologicalderivationsareasfollows.

[39] 1Sg. /faz-o/ → fazosen2Sg. /faz-is/ → fais3Sg. /faz-i/ → faz1Sg.2Sg. /faz-o-sen/ → fazosen1Sg.3Sg. /faz-o-aton/ → fazaton2Sg.3Sg. /faz-is-aton/ → faisaton3Sg.2Sg. /faz-i-sen/ → faisen3Sg.3Sg. /faz-i-aton/ → fazæton

Thesederivationsrequireonly independentlymotivatedgeneralphonologicalpro-cessesof Pontic. The realizationof 2Sg./faz-is/as fais, and3Sg.+2Sg./faz-i-sen/

as faisenis dueto a regularprocessof Pontic,widespreadin otherNorthernGreekdialectsaswell. Apparentlywithout exception,the sequences/-Vsis-/ and/-Vzis-/arerealizedas -Vis- in Pontic. A plausiblederivationis /faz-is/→ fazis→ fazs→fais (Malikouti-DrachmanandDrachman1977,FatimaEloeva,p.c.).20 This process(“anameiosis”,Oikonomides1937,Papadopoulos1955:13,26)appliesevenin under-ived lexical items, suchas the namesAnastasis→ Anastais, kurnazis→ kurnais,Karagiozis→ Karagois, andsimilarly Thanais,Thodois,Kondofois, etc. It is fed bythepreviouslydescribedpalatalizationof /x/ before/i/, e.g.,/vrex-is/vreis(cf. vrexo).

3Sg.+3Sg./faz-i-aton/→ fazætonis derivedby vowel contraction(synalepha)/i,e+a/→ a, /i,e+o/→ o. This is alsoanautomaticpostlexicalphonologicalprocessin Pontic,which appliesacrossword boundariesaswell, asPapadopoulos1955:11makesclear.21 For example:mi a apasaton→ ma apasaton, erxumeontaman→erxumontaman.

Finally, 1Sg.+3Sg./fazo-aton/→ fazatonis a straightforwardcaseof elisionof avowel beforeanothervowel, alsoa processwhich appliesregularlyin Pontic,withinandacrosswords.22

Weconcludethatverb+ clitic combinationsarederivedby phonologicalprocesseswhich applywithin wordsandacrossword boundaries,andwhich areexceptionless,asfar astheevidenceshows.If so,thephonologyof Ponticclitics is consistentwithX0 cliticization,andDrettas’argumentfor theiraffixal statusdoesnotgo through.

5 The diachronic perspective

5.1 Thelexicalizationtrajectory

The generalizationthat syntacticcombinationstendto becomegrammaticalized(orreanalyzed)aslexical,butnotconversely, impliesthatthethreedialectsarehistoricallyrelatedasfollows:

[40] SystemA SystemB SystemCXmax→ X0 X0→ Affix

Accordingly, the systemof the A-type dialectsmustbe the mostarchaicof thethree.Thedialectologicalpictureitself suggeststhisbecauseA-dialectsareperipheralin theGreek-speakingarea.More compellingis the fact thatsomeA-dialectsoccurasenclaveswithin B- andC-dialects,ason Lesbos,presumablyasrelicsof anearlierwider distributionof theA type. But perhapsthemosttelling fact is that thesyntaxof A-dialectsis closestto themedievalGreeksystem,assketchedout in Mackridge(1993).We showthisdirectlybelow, in section5.2.

BasedontheseconsiderationswepositthestartingpointasanA-typedialect,thatis, adialectwith Xmax clitics. TheseXmax clitics developedinto X0 clitics in Pontic,andfurtherbecamelexical clitics in WesternGreek.

However,if this weresimply a directdevelopmentasin [40], we would havenoexplanationfor why PonticdevelopedencliticsandWesternGreekdevelopedprocli-tics. It is morelikely thatthetwo systemssprangfrom earliersystemswith Xmax cliticswhich alreadydifferedsyntactically, in sucha way that“Proto-Pontic”hadpredomi-nantlypostverbalclitics (whichwerelexicalizedasX0 encliticsin modernPontic),and“proto-WesternGreek”hadpredominantlypreverbalclitics (whichdevelopedinto thetypeC proclitics). In thefollowing sectionsweattemptto tracetheserespectivepathsof development.

5.2 MedievalGreek

Accordingto the excellentdescriptionof Mackridge1993,the positionof clitics inlatemedievalGreek(12thto 16thcenturies)is governedasfollows:

[41] a. TheorderV + Cl is moreor lessobligatory:

1. whenthe verbstandsat the beginningof the clauseor immediatelyfollows a coordinatingconjunction.

2. whentheverbcomesimmediatelyafteroneof thefollowing:

thecomplementizeroti

thecausalconjunction�ioti

theconditionalconjunctionei

thenegativeadverbou

3. whentheverbis precededby anobjectwith thesamereferentasthatof theclitic (i.e.,whentheclitic is resumptiveor doubling).

b. TheorderCl + V is moreor lessobligatorywhentheverbcomesimmedi-atelyafteroneof thefollowing:

thefinal conjunctionina

theparticlesna,as,�a

thenegativeadverbsmi, mi�en,ou�en,�en

all interrogativespronounsandadverbs

thecomplementizerpos

all temporalandcomparativeconjunctions

theconditionalconjunctionsan,ean

all relativepronouns

c. The orderCl + V is almostobligatorywhensomesemanticallyempha-sizedwordor phraseprecedestheverb.

d. TheorderCl + V is normalwhenthesubjectprecedestheverb.

e. Thepositionof theclitic beforeor aftertheverbis relativelyfreewhenatemporaladverbprecedestheverb.

f. A coordinatingconjunctionsuchaske removesthe forceof a precedingsubordinatingconjunction,thusrestoringtheorderV + Cl.

FromMackridge’sformulationin [41] it will beseenthatthesystemis basicallyatypeA dialectwith someextrawrinkles.23 We takeup thecasesin turn.

Cases(a1) and(a3)arestraightforward(section2.4). With respectto case(a3),Pappas2001:102makesanadditionalobservation:with respectto atopicalizedobjectwith clitic doubling,“the adjectiveolos[‘all’] behavescontraryto thegeneralpattern[postverbalclitics], sinceonefindspronounsin thepreverbalpositionwhenolosis thedoubledelement.”

[42] testhe

xorescountries

mumy

olo iraall-around

olesall

testhem

afanizidestroy

‘My countriesall around,hedestroysthemall.’ (Pappasp. 79)

Here tesxoresmu olo ira is the topic adjoinedto ΣP, andoles is a focusedfloatedquantifierin [Spec,ΣP]. Thepreverbalpositioningof theclitic, then,is predictedbyouranalysis.

Turningto (b) and(c) in [41], we seethatcomplementizers,wh-elements,nega-tion, modalparticles,andpreposedfocusXPshostclitics, which thenarepreverbal,asin TypeA dialects(section2.3). Whatis surprisingin [41] is thecontrastbetweenthecomplementizersandthenegationin (a2)andthosein (b). It may tentativelybeinterpretedasfollows.

The conjunctionsoti, dioti. . . in [41a2] areold relativepronouns(asHorrocks1997pointsout). It is reasonableto assumethattheyarestill in [Spec,CP]atthispoint,andtriggerV-to-Cmovement.(Pappas2001showsthatin themoreextensivematerialexaminedby him, �ioti hasbecomeregularizedto occurwith preverbalClitics.) Thenegationou alsotriggersV-to-C movement.24 Mackridge1993alsonotesa residualcaseof materialin [Spec,CP]triggeringV-to-Cmovementin earlyMedievalGreek.

[43] tothe

poshow

ikokirevomanagehousehold

moumy

tinthe

apasanentire

ikianhome

‘how I managemy entirehousehold’

Thestructureof [43] is presumably:25

[44] [CP pos[C’ ikokirevo [TNSP Cl . . . ]]]

In the moderntypeA dialects,theseitemshavebeenlargely regularized.However,thedialectof Pharasahaspreservedacognateof thenegationou, in theform jo, whichretainstheold syntacticbehavior(Dawkins1916).As for theconditionalcomplemen-tizerei, Pappas2001notesthatit is almostalwaysfollowedby thecontrastiveparticle�e, whichperhapsintroducedanintonationbreak.

The variation in (d) and (e), confirmedin Pappas’study, can be attributedtostructuralambiguity. Preverbalsubjects(case(d)) fall bothundercase(c) (focusin[Spec,ΣP]), andundercase(a3) (topic adjoinedto ΣP), exceptthat subjectsdo nottriggerclitic doubling,asin all Greekdialects.In texts,theambiguitycansometimesberesolved,however,whenthediscoursestatusof apreverbalsubjectis establishedbycontext.In anycase,preverbalsubjectswould hosta clitic if theyarein thespecifierpositionof a functionalprojection,andnot if theyareadjoined.

Thevariationin case(e) is slightly different. Certainone-wordtemporaladverbsaresimplyX0, andthesecanbewithin theΣ projection.Phrasaladverbs,on theotherhand,would be adjoined. Evidencefor the X0 statusof certaintemporaladverbsinStandardModernGreekis thattheycanintervenebetweenafocusor awh-phraseandtheverb(Alexiadou1994):

[45] a. PjonWho-acc

akomastill

skeftetethinks

othe

Yanis?John-nom

‘Who is Johnstill thinkingabout?’

b. Pjowhich

arthroarticle

idhihas

ehialready

teliosifinished

othe

Yanis?John-nom

‘Which articlehasJohnalreadyfinished?’

[46] a. Tithe

MARIAMary-acc

akomastill

skeftetethink-3sg

oJohn-nom

Yanis.

‘Johnis still thinkingaboutMARY.’

b. Tothe

ARTHROarticle

idhialready

ehihas

teliosifinished

othe

Yanis.John

‘JohnhasalreadyfinishedtheARTICLE.’

Theadverbsidentifiedby Mackridgeasfalling under(e) in latemedievalGreekforma largerclassthanthe X0 temporaladverbialsin SMG; neverthelessit is suggestivethattheyareall one-wordtemporaladverbs,suchaspanta‘always’, tote ‘then’, palin‘again’,ef�is ‘immediately’.

Finally, thenon-parallelismof conjunctsin (f), illustratedin [47], is dueto coor-dinationof internalΣPs,alreadydemonstratedfor typeA dialectsin section[2.4].

[47] ou�enneither

ivre�ikewasfound

kanis,anyone

inato

meme

kataf�asireach

keand

polemisifight

meme

‘neitherhastherebeenanyoneto challengemeandfight me’ (DigenisAkritas,Escorialversion100)

As notedearlier,this sofar unexplainedpeculiarity, recordedin Mackridge1993andin Pappas2001,finds an explanationin our account. The crucial point is that thegeneralizationaboutclitic orderis not to be framedin termsof mainclausesversussubordinateclausesbut in termsof theconcomitantstructures.

5.3 Pontic

In section5.1 we reasonedthat Ponticmustbe descendedfrom a dialect in whichclitics werepredominantlypostverbal.In fact,adialectwith thehypothesized“proto-Pontic” propertiesis alreadyimplicit in our historicalanalysis. It may be identifiedwith a stageof Greekprior to theemergenceof ΣP, whereTNSP wasthehighestIPprojection,andthe finite verb movedto C (recall the residualV-to-C movementofsection5.2). At this stage,clitics wereat theleft edgeof IP, andtheverbcouldbetotheir left. Thisdialectwouldhavehadthephrasestructurein [48].

[48] CP

C′

TNSP

TNSP

TNS′

Spec C Cl TNS0 VP

Vj . . . tj . . .

At the stagebeforeΣP wasintroduced,focusedelementsandnegationwould havebeenfrontedto Spec-CP. Movementof theverb to C in suchcaseswould thenhaveresultedin postverbalpositioningof clitics. Thereis goodevidencethat suchcliticpositioningbeganto arise,asanoptionto inheritedWackernagelcliticization,duringtheclassicalGreekperiod,asillustratedby the following examplesfrom Herodotus(citedfrom Garrett1995).

[49] a. tonthey-GenPl

menFOCUS

dePrt

oudenNEG

prosıetoattracted

minhim

‘Not oneof themattractedhim.’ (Hdt. 2.68.48)

b. ekfrom

toutouthis

deFOCUS

manenaigo-mad

minhim-Acc

nomızousithink-3Pl

SpartietaiSpartans

‘It’ s from this thattheSpartansthink hewentmad.’ (Hdt. 6.84.3)

c. pothenwhere

anPrt

laboimitake-Opt-1Sg

rh emaword

muri amphoronmyriad-measures-holding

hotoiwhich-DatSg

proseipoaddress-Sbj-1Sg

s’you-Acc

‘WherecanI find a zillion-dollar word with which to addressyou?’ (Ar.Pax521-22)

Our hypothesispredictsthat the medievalPontic systemshoulddisplay robustevidenceof thissyntax.Indeed,in deedsto themonasteryof Vazelon(southof Trebi-zond),we find exampleslike thefollowing:

[50] MedievalPontic(OuspenskyandBenechevitch1927)

a. tonthe

dePrt

toponplace

edokamenwe gave

soiyou

eisin

tousthe

eksesfollowing

kaıand

dieneke˜ıseverlasting

xr onousyears

‘we havegivenyouthis landin perpetuity’(deeddated1260)

b. hosonhowmuch

diaphereibelongs

moume-Dat

‘asmuchasis my share’(dated1435)

[50b] is especiallyinterestingbecauseit showsa postverbalclitic in thepresenceof awh-element,differing from thelatemedievalsystemdiscussedin section5.2.AnothercharacteristicPontictrait seenin theseearlytextsis multiplepreposednegation.

[51] tin annobody

tıpotenothing

ounot

xreostoowe-1Sg

‘I don’t oweanyoneanything’(ibid., dated1291)

In a systemsuchas [48] whereV raisesto C in main clauses,the majority ofclitics will endup in postverbalposition. In suchdialects,lexicalizationfrom Xmax toX0 wouldnaturallygiveriseto enclitics,asin Pontic.

After thePonticdialectsdivergedat anearly stageof MedievalGreek,theotherGreekdialectsunderwenta periodof furthercommondevelopment(which includedthe rise of ΣP) beforein turn splitting off into the ancestorsof the Cappadociandi-alects(typeA) andtheWesternGreekdialects(typeC).

[52]

C A B

TheimplicationthatthePontic(typeB) dialectssplit off from therestof Greekquiteearly, andthattypeA andtypeC dialectsunderwentaperiodof commondevelopment,is consistentwith Dawkins’ 1940suggestionthat the Ponticdialectswereseparatedfrom the restof Greekasearly asthe 11thcenturyby the Seljuk conquests,severalcenturiesbeforetherestof Asia Minor’sGreekdialects(includingthoseof Cappado-cia)werecutoff by theOttomanincursion.

5.4 Kozani: themissinglink

If we adopttheview thatsyntacticchangetakesplaceby small stepwiseincrementsratherthanby Lightfootiancatastrophes,we mustassumethatWesternGreekdid notdevelopdirectly from a type A systemwhereclitics areXmax categories,but passedthroughan intermediatePontic-typestageof syntacticX0 cliticization. In fact, weareled to positasthe immediateantecedentof standard/WesternGreeka systemB′,whereclitics alreadyprecedetheverb,butstill retaintheirsyntacticX0 status,like thePonticclitics. Thissystemwouldconstitutethemirror imageof Pontic:

[53] V0

V0 clitic

V

SystemB (Pontic)

V0

clitic V0

V

SystemB′ (inferred)

We shallnowarguethatsystemB′ survivesto thisday.

Kontosopoulos1994:53,101reportsthat clitics areplacedbetweenthe auxiliaryand the participle in two moderndialects: Kozani (Macedonia)andChios (off thecoastof Asia Minor).

[54] a. ıxanhad-3Pl

tsthem

vapspainted

(Kozani) tus ıxan vapsi (Standard Greek)

‘they hadpaintedthem’

b. ıxenhas-3Sg

meme

piasicaught

(Chios) me ıxepiasi (Standard Greek)

‘he hascaughtme’

Thehypothesisthatthesedialectsinstantiateourpredicted“missinglink” betweenbe-tweentypesA andB, that is, X0 procliticsmakesseveralsyntacticandphonologicalpredictions.Kiparsky et al. (2001)wereableto confirm thesepredictionsby inter-viewinga speakerof thedialect.

Kiparskyetal. foundthatin theKozanidialectclitics maybeplacedeitherbeforetheauxiliary, or, asKontosopoulosreports,betweenit andthemainverb.This seemsto be a genuineoption within the dialect itself. The repetitionof the clitic in [55b]showsthattheclitic in [54a]andin [55a]mustattachto V0, ratherthanto theauxiliary.

[55] a. ıxanhad-3Pl

tsthem

vapspainted

kiand

ftiaksfixed

‘they hadpaintedandfixedthem’ (preferred)

b. ıxanhad-3Pl

tsthem

vapspainted

kiand

tsthem

ftiaksfixed

‘they hadpaintedthemandfixed them’

Our first syntacticpredictionis thatconjoinedverbsmayshareanX0 clitic, asinPontic. Specifically, whereasPontic’s sharedencliticsalwaysfollow the verb con-junction(see[36]), in Kozaniwe expectedthatits sharedprocliticswould precedeit.This is whatwe find; see[56]. [56a] (thoughungrammaticalin standardGreek)is infactpreferredover[56c].

[56] a. nher

ıdasaw-1Sg

keand

xeretsagreeted-1Sg

‘I sawherandgreetedher’

b.* ıda‘saw-1Sg

keand

ther

xeretsagreeted-1Sg

‘I sawandgreetedher’

c. nher

ıdasaw-1Sg

keand

ther

xeretsagreeted-1Sg

‘I sawherandgreetedher’

Therespectivestructuresfor PonticandKozaniaregivenin [57].

[57] V0

V0

V0 ke V0 clitic

V V

Pontic

V0

V0

clitic V0 ke V0

V V

Kozani

A secondsyntacticconsequenceis that VP-adverbsmay intervenebetweentheauxiliaryandtheclitic, butnothingmayintervenebetweentheclitic andthefollowingnonfiniteverb.

[58] a. ıxanhad-3Pl

keroalready

tsthem

vapspainted

‘they hadalreadypaintedthem’

b.* ıxanhad-3Pl

tsthem

keroalready

vapspainted

‘they hadalreadypaintedthem’

This follows on theplausibleassumptionthatVP-adverbsareat the left edgeof VP,andthattheVP is thecomplementof theauxiliary.

[59] TNSP

VP

VP

V0

exi kero clitic V

Kiparskyet al. alsogive two phonologicalargumentsthat theclitic is a procliticon theparticipleratherthananencliticon thefinite auxiliary. Onecomesfrom stress.In verbformswhichbearlexicalstressbeforetheantepenult,asecond,equallypromi-nentstressis assignedto thepenult,in orderto avoida stresslapse,asin [60a]. Nosuchstressis assignedin caseslike [60b].

[60] a. efagami‘ate-1Pl’‘we ate’

b. ıxamehad-1Pl

toit

vapspainted

(not * ıxame to vaps)

‘we hadpaintedit’

The reasonis that in [60b], the sequenceixameto is not a word either lexically orpostlexically, accordingto our analysis.Therefore,it cannotbeassignedword stressatanylevelof thephonology.

Thesecondphonologicalargumentcomesfrom a processof voicing assimilationof [s] to [z] before[m] within prosodicwords.26 Thefinal [-s] of averbdoesnotvoicebeforeanobjectclitic beginningwith [m-], whichshowsthatthereis noenclisis.

[61] /exismas�i/ → [exismaz�i] ‘you’ve seenus’

The Kozani dialectalsohasenclitic pronouns,suchaspossessiveclitics. Theseseemto havethestatusof lexical suffixes,just asin standardGreek.

[62] /�ikosmas/→ [�kozmas]‘our own’

The object proclitics of Kozani, then, confirm the hypothesizedB′ system. Itremainsto beseenhowwidespreadit is, andin particularwhethertheChiosdialectissimilar to thatof Kozani.

More importantly, our predictionthat theB′ systemis the immediateantecedentof thestandard/WesternGreekC systemremainsto beverifiedby historicaldatafromearlierstagesof Greek.27

5.5 Summaryof thehistoricaldevelopment

Althoughwewill notattemptto reconstructtheevolutionfrom thetwo medievalsys-temsbackto theHomericlanguagein thispaper,we canoffer somepreliminarysug-gestionsbasedon thereinterpretationof Taylor 1994proposedby Kiparsky1996.InHomericGreek,we supposethat no IP (whetherTNSP or ΣP) is syntacticallypro-jected.Consequently, Xmax clitics at thatstageareadjoinedto CP, wheretheyundergoprosodicinversionif necessaryto satisfytheirenclisisrequirement.This is to saythatHomericclitics aresecondposition(Wackernagel)clitics.

In later classicalGreek,a syntacticIP (specificallya TNSP, we assume)is intro-duced.Clitics (still of theXmax type)adjointo this lowerprojection,while finite verbsmaymoveto C. This is the“proto-Pontic”system,in whichclitics arepredominantlypostverbal.Ponticdevelopsfrom it by thefirst stageof lexicalizationof Xmax clitics,by which theybecameX0 clitics, with encliticstatus.

The dialectsfrom which WesternGreekarosedevelopeda ΣP projection,whilestill at the Xmax stage. This stageis attestedin medievalGreek,andpersistsin themodernTypeA dialects.Fromthisstartingpoint,lexicalizationof Xmax clitics resultedin a TypeB′, suchasstill attestedin Kozani. Thesecondstageof lexicalization,bywhich clitics becameaffixes, then resultedin the Type C systemsof standardandWesternGreek.

This scenariois summarizedin thefollowing syntacticstemmaof Greekdialects.

[63] HomericGreek

riseof TNSP

Classical,“Proto-Pontic”

Pontic

riseof ΣP

Medieval/TypeA Kozanitype SMG,W.Gk.Xmax→ X0 X0→ Affix

Two corollariesof our analysisof clitic positioningare worth stressing. First,joinedwith our assumptionthatsyntacticchangeproceedsin minimal steps,we pre-dict thatsyntacticvariationin clitic positioning,in sofar asit reflectsongoingchange,shouldbebetweensystemswhich areadjacentin theaboveschema.ThevariationinKozanibetweenX0 andaffixal cliticization is anexample.Dialectmixtureor borrow-ing, on theotherhand,couldresultin variationbetweenanysystems.In fact, for oneof thetwo heterogeneousdialectswehavefound,thatof Amisos,whichshowsvaria-tion betweenTypeA andTypeB, thepredicteddialectmixtureis amatterof historicalfact. Spokenin the Pontic(typeB) area,Amisosis knownto havehadan influx ofrefugeesfrom CappadocianCaesarea(whowouldhavespokena typeA dialect)afterthefall of Constantinoplein 1453(Christopoulos1974:179a).

A secondcorollaryis thattheclitics mightbecome(ormightalreadyhavebecome)affixal in somedialectof Pontic.Suchaninnovativedialectof Ponticwould havethefollowing characteristics:

[64] HypotheticalC′ dialect:

a. Clitics arepostverbal(asin Pontic):exi ta ‘he hasthem’

b. Clitics attachto finite verbsonly (unlike Pontic):exi to kani ‘he hasdoneit’

c. Clitics mustberepeatedin eachconjoinedverb(unlike Pontic): *na feroketrog’ a ‘I’ll takeandeatit’

d. Verb+clitic combinationsarestressedlike lexical words(unlike Pontic):* esirenatona‘he threwhim’

e. Verb+cliticcombinationsmayshowlexical idiosyncrasies(unlikePontic).

5.6 Implications

Thedialectevidenceshowsthat thedistinctionbetweenaffixal andX0 clitics is min-imal andirreducible.On theonehand,we foundno intermediatesystemsto support

Janse’s claim (1998)thatthedistinctionbetweenclitics andaffixesis a gradientone.On the other hand,Halpernand Fontana’s two-way classificationof clitics, whichidentifiesX0 clitics with affixes,is not fine-grainedenough,andshouldbe replacedby a ternaryone. StandardGreekclitics arelexical (asJosephproposed),but Ponticclitics aresyntacticX0 (contraDrettas1997).

On thehistoricalside,our findingssuggestthatchangeis neithercatastrophic(asLightfoot claims)nor gradient(aswassuggestedin someearly work on grammati-calization).Rather,changeproceedsin minimaldiscreteincrements. Moreover,it isstriking thatnoneof thechangesthatour theorypositsleadsto abruptdiscontinuitiesin theoutput.Eachstepin thereanalysisor grammaticalizationprocessmodifiesthelanguagein waysthatarenot salientto languagelearners(not to speakof dialectolo-gists).

The recognitionof a three-waydistinctionin clitics, coupledwith our view thatsyntacticchangeproceedsin minimal steps,led usto predicttheexistenceof systemB′, theproclitic counterpartof Pontic.Our discoverythatthepredictedsystemexistsin thedialectof Kozaniconfirmsbothassumptions.

Notes

1Thismaterialwaspresentedatthefirst internationalconferenceonmodernGreekdialectsandlinguistic theoryat theUniversityof Patrasin Oct.2000,atStanfordUni-versityin Nov. 2000,andat theconferenceonhistoricalmorphosyntaxat theUniver-sity of Konstanzin June2001. We aregratefulto theconferenceorganizers,AngelaRalli andAditi Lahiri, andto theaudiencesfor their suggestionsandchallenges.Forvaluablediscussionandcriticismweareindebtedto AndrewGarrett,who,in addition,generouslysharedhis unpublishedhandoutswith us. Commentson previouswrittenversionsby TracyKing, ElaHarrisonWiddows,ananonymousreviewer,andaneditorof this journal,havealsogreatlyimprovedthis paper. PanayiotisPappaskindly sentus a copy of his unpublisheddissertation(2001). It reachedus after our paperwassubstantiallywritten,butwehavebeenableto takesomeof hismostrelevantfindingsinto account;wehopeto continuethedialogueonanotheroccasion.

2Interestingly, a similar trajectory, albeitwith two stages,hasbeenproposedfortheclitics in someRomanceandSlaviclanguages(Fontana1993,Tomic2000).

3Of courseaffixesaresometimes“upgraded”to clitics. In thecaseswe haveex-amined,this happensby analogywith existingclitic patterns.If this is generallytrueof upgrading,wewouldconcludethatit differs from grammaticalization,which,asiswell known,doesnot requireexistingmodels,andmayintroducenewcategoriesandstructures.In general,theunidirectionalityhypothesisshouldbeseenasconceptuallysimilar to thehypothesisthatsoundchangeis exceptionless.Both arefalsifiableonlyin the contextof an entiretheoryof change,not simply by unanalyzedinstancesofchange.

4Thelist doesnotincludedialectsof whichwehavereadnotexts,notablythoseofCrete,Cyprus,andRhodes.Judgingfrom publisheddescriptions,thesedialectsseemto belongin our typeA category.

5An editor of the journal (citing an unpublishedpaperby Mackridgewhich wehavenot seen)remindsus of exceptionalStandardGreekexpressionssuchasgamwse ‘fuck you’, pateic me, patw se ‘pell-mell, a scrum’ (lit. ‘you stepon me,I stepon you’). Thoughidiomatic,with theverbsnot constitutingclausalheads,theseex-pressionsarepresumablystill parsableinto recognizableparts.Theexceptionalwordorderis afixedpropertyof theidiomsthemselves,andnotof anyparticularclitic.

6Dawkins1950 notedthat partsof the Dodekaneseshowa mix of type A andtypeC behavior. We havefoundanothermixedsystem,thatof thedialectof Amisos,formerlyspokenin Turkey, which showsvariationbetweentypeA andB (thePonticvariety). Even in thesedialects,the clitics behaveuniformly in that all, regardlessof their grammaticalfeatures,showthe samesyntacticvariation. Our analysispre-dicts that they representdialectmixturedueto contactand/orto migration(andnotendogenouschangein progress).Seesection5.5.

7A noteon the terminologyof this paper:by CLITICS we meanelementswhichareaddedto words(asopposedto stems).By AFFIXES we meanelementswhich areaddedin the lexicon (ratherthanin the syntax). Thus, the categoriesof clitics andaffixesarenotmutuallyexclusive.Forexample,theclitics of StandardModernGreekareaffixes,accordingto whatwe proposebelow.

8For preverbalfocusin SMG seeTsimpli 1995. For emphaticnegativeelementsseeVeloudis1982,Giannakidou1998,andTsimpli & Roussou1996.

9See,e.g.,Anagnostopoulou1994onclitic doublingin SMG.10Like otherapproacheswhichhavemultipleIPprojectionsandwhichdonotposit

verbmovementpastTNS0, weoweanaccountof why thespecifierpositionof certainsuchprojectionsremainsunfilled. AlexiadouandAnagnostopoulou1998arguethatthesespecifierpositionsarenot licensedin null subjectlanguages,suchasGreek.

11In fact, our analysisis consistentwith an expansionof ΣP into FocP, MoodP,NegP, etc.,providedthatonly thespecifierpositionof FocPmaybefilled.

12This appearsto bea pervasivegeneralizationgoverningclitics in needof a the-oreticaljustification.

13This correspondsto thefirst clauseof Halpern’s formulation. Halpernalsoin-cludesa secondclausewhich dealswith nondirectionalclitics, absentin Greek,ac-cordingto ouranalysis.

14Thedistributionof clitics in typeA dialectsis similar to thatof Bulgarian.King1996hasproposedan analysisof the latter that makesuseof prosodicinversionaswell.

15Anotheraccountbasedonverbmovementis thatof Philippaki-Warburton1995.Clitics areadjoinedto TNSP,asin ouranalysis,but theverbmovesfrom TNS0 to theheadof MoodPin indicativeclauses.Thepreverbalappearanceof clitics in indicativeclauseswith a frontedfocus,and in wh-interrogatives,presentsempiricalproblemsfor thisanalysis.

16This factalsohasdiachronicimplications.Philippaki-Warburton1995hassug-gestedthat the shift from enclisisto proclisisin Greekis dueto the fact that modalparticleslike nabecomephonologicallydependentontheverb,andcannotsupportanencliticon theirown. TheCappadociandatashowthatthecausalconnectionis atanyratenotanecessaryone.

17A reviewerremindsus that the Oenountiakavarietyof Ponticshowsvariationwhich is similar to thetypeA pattern(Oikonomides1958:3,413).

18Miller 1992arguesfrom co-ordinationdatathatFrenchclitics arelexical. SeealsoAuger1995andMiller andSag1997,andLabelleandHirschbuhler2000for anopposingview.

19As Drettas1997:100putsit: “On voit quecesphenomenes,obligatoiresdansle cadred’un paradigmedonne (en l’occurrence,la conjugationd’un verbe),ne re-produisentpasforcementdescontraintesphonologiqueset que,par consequant,on

ne peutrendrecompteau moyend’une partie“r eglesphonologiques”de la langue;nousavonsaffaireadesfaitsmorphologiquesqui serontpresentesaveclesunitescon-cernees(parexample,l’article, l’objet verbal,etc.).” Drettasalsoarguesagainsttheclaim that faz ‘feeds’ plus -sen‘you’ is realizedas faisenin orderto avoid the pro-hibitedsequence*-zs-, on thegroundsthatonecouldachievethatby othermeans,forexample,by insertinge into thecluster. But clearlya processis not unmotivatedjustbecauseanotherprocessmighthaveachievedthesameend.Onthecontrary, therearealmostalwaysmultiple waysof avoidingconstraintviolations.For example,prohib-ited consonantclusterscanbeavoidedby epenthesis,deletion,lenition,assimilation,or metathesis.Indeed,a languagemay useseveralof thesedevicesunderdifferentconditions,dependingon therankingof its otherconstraints.But in anycase,avoid-anceof *-zs-sequencesis notevenat issueif faisenis derivedasproposedin thetext.Thus,theargumentbasedon avoidanceof *-zs- is bothfallaciousandirrelevant.WeconcludethatDrettashasnotmadeagoodcasefor theaffixal statusof Ponticclitics.

20Thehaplologicalavoidanceof . . .C1VC1. . . sequencesis probablya contribut-ing factor (Oikonomides1937). Drettas105 cites evidencethat sucha haplologyprocessappliesproductivelyacrossword boundaries,e.g.,avuta ta pe�ıa→ avutape�ıa. We emphasize,however,thatour argumentfor theX0 statusof Ponticcliticsdependson thebarefact that/-sis/and/-zis/ sequencesaresystematicallyreducedto-is in Pontic,andnotonanyparticularanalysisof thatphonologicalreductionprocess.

21<H sunaloif� gÐnetai kaÈ ân sunekfora metaxÌ tou telikou fwn entoc thcprohgoumènhc lèxewc kaÈ tou �rktikou thc ápomènhc. [Vowel contractionalsooccursbetweenthe final vowel of oneword and the initial vowel of the followingword.] (Papadopoulos11).

22E.g.,amontoeksertspıson→ amonteksertspıson(Drettas78,103).

23Accordingto Horrocks1990,thechangefrom theancientlanguageto theme-dieval languageis that the domainof cliticization shrinksfrom the sententialto theverbaldomain: clitics in late MedievalGreekstill appearin secondposition,as inHomeric Greek,but now in the secondposition of the VP (or of the verbal com-plex) ratherthanS.Horrockslinks this changeto thecomplementizerhinabecomingthe particlena. Horrocks’ idea,while appealinglysimple, is dubious:asdescribedbelow, clitics appearpreverballyin the presenceof Wh-phrases,preposedfocusedconstituents,andnegation,evenif theyarefirst within theverbalcomplex.SeealsoPappas2001:135for somepertinentremarks.

24Negationand wh-elementsare the original V-to-C triggersin Germanic,andcontinueto triggerresidualsubject-auxinversionin English.

25Theclitic mouadjoinedto TNSP maybea possessiveclitic raisedfrom withinthe NP tin apasanikian or a benefactiveargumentof the verb. Either analysisiscompatiblewith ourproposal.

26Acrossword boundary, [-s] (aswell astheprecedingobstruentin clusterslike[-ks]) arevoicedbeforevoicedfricativesonly. SeeMargariti-Ronga1985:82,128for

a descriptionof thesevoicing assimilationprocessesin thecloselyrelateddialectofSiatista.

27A smallbutsuggestivepieceof evidenceis furnishedby anobservationof Pap-pas(2001:96)from 17thcenturyGreek. In theperiphrasticfuture,theclitic appearsbetween�eloandtheinfinitive, anddoesnot triggersecondarystressontheauxiliary,suggestingthat it is attachingto the following nonfiniteverb,asit is mutatismutan-dis in Kozani. In MedievalGreek,on theotherhand,asdiscussedby MackridgeandPappas,theclitic follows theexpecteddistributionwith respectto theauxiliary.

References

Alexiadou,Artemis.1994.Adverbsin modernGreek.Doctoraldissertation,Univer-sity of Potsdam.

Alexiadou,ArtemisandElenaAnagnostopoulou.1998. “ParametrizingAgr: wordorder,verbmovement,andEPPchecking.” Natural LanguageandLinguisticTheory16:491–539.

Anagnostopoulou,Eleni.1994.Clitic Dependenciesin ModernGreek.Doctoraldis-sertation,Universityof Salzburg.

Auger,Julie.1995. “Les clitiquespronominauxen francais parle informel: uneap-prochemorphologique.” Revuequebecoisedelinguistique24:21–60.

Boskovic, Zeljko. 2001. On the natureof the syntax-phonologyinterface. Amster-dam:Elsevier.

Cernyseva,T.N. 1958.Novogreceskijgovorselpromorskogo(Urzufa)i Jalty, Pervo-maiskogoraiona,StalinskojOblasti. Kiev.

Christidis,A.F. (ed.).1999. DialektikoÐ JÔlakoi thc <Ellhnikhc Glwssac / Di-alectEnclavesof theGreeklanguage. Athens:Centrefor theGreekLanguage.

Christopoulos,G. et. al. 1974. <IstorÐa tou <Ellhnikou ^Ejnouc. Tìmoc I'. [His-tory of theGreekNation,volume10.] >Ekdotik  >Ajhnwn.

Danguitsis,Constantin.1943. Etudedescriptivedu dialectede Demirdesi. Paris:Maisonneuve.[Dang]

Dawkins,RichardM. 1916. ModernGreekin Asia Minor. Cambridge:CambridgeUniversityPress.[D]

Dawkins,RichardM. 1931. “Folk talesfrom Sourmenaand the valley of Ophis.”>Arqeion Pontou 3: 79–122.

Dawkins,RichardM. 1940. “The dialectsof ModernGreek.” Transactionsof thePhilologicalSociety, 1–38.

Dawkins,RichardM. 1950. Forty-five stories from the Dodecanese.Cambridge:CambridgeUniversityPress.[DD]

Drachman,GaberellandKlidi, Sila.1992. “The propertreatmentof adverbialques-tions in Greek:Theextendedminimal structurehypothesis.” Studiesin GreekLinguistics13:371–390.

Deo,Ashwini. 2001.“The developmentof IP in Indo-Aryan.” Manuscript,StanfordUniversity.

Drettas,Georges.1997.Aspectspontiques.Paris:Arp. [Dr]

Fontana,JosepM. 1993. Phrasestructureand the syntaxof clitics in the history ofSpanish.Doctoraldissertation,Universityof Pennsylvania.

Fostiropoulou,Despina.1938.“ParamÔjia >Imèrac.” [“StoriesfromImera.”] >ArqeionPontou 8:181–202.

Garrett,Andrew. 1995. Unpublishedhandoutfrom NELS 24 Workshopon Indo-European.

Giannakidou,Anastasia.1998.Polarity sensitivityas(non)veridicaldependency. Phil-adelphia:Benjamins.

Halpern,Aaron.1995.On theplacementandmorphologyof clitics. Stanford:CSLIPublications.

Halpern,AaronandJosepM. Fontana.1994.“X 0 andXmax clitics.” WCCFL12:251–266.

Horrocks,GeoffreyC. 1990.“Clitics in Greek.A diachronicreview.” In M. RoussouandS.Panteli,eds.,GreekOutsideGreeceII, 35–52.Athens:DiasporaBooks.

Horrocks,Geoffrey. 1997.Greek: A History of its languageandits people. LondonandNewYork: Longman.

Inkelas,Sharon.1989. Prosodicconstituencyin the lexicon. Doctoraldissertation,StanfordUniversity.

Janse,Mark. 1998. “Cappadocianclitics andthe syntax-morphologyinterface.” InBrian D. Joseph,GeoffreyC. HorrocksandIrenePhilippaki-Warburton,eds.,Themesin GreekLinguisticsII , 257–281.Philadelphia:Benjamins.

Joseph,Brian.1988. “Pronominalaffixesin ModernGreek: thecaseagainstclisis.”CLS24,203–215.

Katsoyannou,Marianna.1999.“The idiom of Calabria.” In A.F. Christidis,ed.,113–119.

King, HollowayTracy. 1996. “Slavic clitics, long headmovement,andprosodicin-version.” Journalof SlavicLinguistics4:274–311.

Kiparsky, Paul.1995.“Indo-Europeanoriginsof Germanicsyntax.” In AdrianBattyeandIan Roberts,eds.,Clausestructureandlanguagechange. Oxford: OxfordUniversityPress.

Kiparsky, Paul.1996. “The shift to head-initialVP in Germanic.” In HoskuldurThrainsson,SamuelD. Epstein,andStevePeter,eds.,Studiesin ComparativeGermanicSyntax,Vol. 2.. Dordrecht:Kluwer.

Kiparsky, Paul.1997.“The riseof positionallicensingin Germanic.” In AnsvanKe-menadeandNigel Vincent,eds.,Parametersof morphosyntacticchange. Cam-bridge:CambridgeUniversityPress.

Kiparsky, Paul.To appear. “Finnish Noun Inflection.” In Diane Nelsonand SatuManninen,eds.,GenerativeApproachesto Finnic Linguistics. Stanford:CSLIPublications.

Kiparsky, Paul,CleoCondoravdi,AnnaPliatsiou,KaterinaPliatsiou,AchilleasTheo-phanos,FanisKatsanos,andAnastasiaKastanoulia.2001. “The missinglink:X0 procliticsin Kozani.” Manuscript.

Kontosopoulos,Nikolaos.1994. Di�lektoi kaÐ >Idi¸mata thc Nèac <Ellhnikhc.[Dialectsandlocal varietiesof ModernGreek.] Athens.

Kretschmer,Paul.1905.NeugriechischeDialektstudienI: derheutigelesbischeDia-lekt. Wien: AkademiederWissenschaften.[K]

Laka,Itziar. 1990. Negationin syntax: On the natureof functionalcategoriesandprojections. Doctoraldissertation,MIT.

Labelle,MarieandPaulHirschbuhler. 2000. “Les ‘clitiques’ argumentsen serbo-croateet dandl’histoire du francais.” In ClaudeMuller et al., eds.,Clitiqueset cliticization: actesducolloqueduBodeaux. Paris:HonoreChampion.

Lightfoot, David.1999.Thedevelopmentof language:acquisition,change,andevo-lution. Oxford: Blackwell.

Mackridge,Peter. 1987.“Greek-speakingMoslemsof north-eastTurkey: prolegom-enato a study of the Ophitic sub-dialectof Pontic.” Byzantineand ModernGreekStudies11:115–137.

Mackridge,Peter. 1993.“An editorialproblemin medievalGreek:thepositionof theobjectclitic pronounin theEscorialDigenesAkrites.” In NikolaosM. Panayio-takis, ed., >Arqèc thc neoellhnikhc logoteqnÐac / Origini della letteraturaneogreca,325–342.Venice.

Mackridge,Peter. 1999. “The Greekspokenin theregionof Pontus.” In A.F. Chris-tidis, ed.,101–105.

Mavroxalividis,G. andI.I. Kesisoglu.1960.Tì glwssikì ÊdÐwma thc ^Axou. [Thelocal dialectof Axos.] Athens:KentroMikrasiatikonSpoudon.[M & K]

Malikouti-Drachman,Angeliki andGaberellDrachman.1977.“Tupoi fwnologikwnnomwn kai boreia Êdi¸mata.” [“Typesof phonologicalrulesandtheNortherndialects.”] A' Sumpìsio GlwssologÐac tou Boreioelladikou Q¸rou. Jes-salonikh.

Margariti-Ronga,Marianna.1985.Fwnologikh �nalush tou Siatistinou Idiwmatoc.[Phonologicalanalysisof the local dialect of Siatista.] Doctoraldissertation,AristotelianUniversityof Thessaloniki.

Miller, Philip H. 1992.“Postlexicalcliticizationvs. affixation:coordinationcriteria.”CLS28:382-396.

Miller, Philip H. andIvanA. Sag1997. “Frenchclitic movementwithout clitics ormovement.” NaturalLanguageandLinguisticsTheory15:573–639.

Minas,K. 1970.T� >Idi¸mata thc Karp�jou. [Thelocal dialectsof Karpathos.]Athens:Klapakis.

Oikonomidis,D.E.1938.“Grammatik� thc pontikhc dialèktou.” [“On thegrammarof thePonticDialect.”] PontiakaPhylla2:200–206.

Oikonomidis,D.E.1958.Grammatik  thc <Ellhnikhc dialèktou tou Pìntou. [Gram-marof theGreekdialectof Pontos.] Athens.

Ouspensky, Th. andV. Benechevitch.1927.ActesdeVazelon/ VazelonskieAkty. Lenin-grad:IzdanieGosudarstvennojPublicnojBiblioteki, SerijaV: Orientalia,No. 2.

Papadopoulos,Anthimos.1955. <Istorik  grammatik  thc pontikhc dialèktou.[Historical grammarof thePonticdialect.] Athens.[P]

Pappas,Panayiotis.2001. Weak object pronounplacementin Later MedievalandEarly ModernGreek.Doctoraldissertation,TheOhioStateUniversity.

Pappou-Zouravliova,Ekaterini.1999.“The Greeksof theAzov regionandtheir lan-guage.” In A.F. Christidis,ed.,129–137.

Philippaki-Warburton,Irene.1985. “Word orderin ModernGreek.” TransactionsofthePhilologicalSociety, 113–143.

Philippaki-Warburton,Irene.1995.“Diaqronik  je¸rhsh thc jèshc twn klitikwnmèsa st n prìtash.” [“A diachronicview of thepositioningof clitics within asentence.”] Melètec gi� t n <Ellhnik  Gl¸ssa / Studiesin GreekLinguistics,123–134.Thessaloniki:AristotelianUniversity.

Philippaki-Warburton,IreneandVassiliosSpyropoulos.1999.“On theboundariesofinflectionandsyntax:Greekpronominalclitics andparticles.” In GeertBooijand Jaapvan Marle, eds., Yearbookof Morphology1998, 45–72. Boston:Kluwer AcademicPublishers.

Pinon,Christopher. 1993.“SigmaPandHungarian.” WCCFL11:388–404.

Profili, Olga.1999. “The revival of Grico in the Greekcommunityof Salento.” InA.F. Christidis,ed.,121–128.

Taylor,Ann. 1994. “The changefrom SOV to SVO in ancientGreek.” LanguageVariation andChange6:1–38.

Terzi,Arhonto.1999. “Cypriot Greekclitics andtheir positioningrestrictions.” InArtemisAlexiadou,GeoffreyC. HorrocksandMelita Stavrou,eds.,StudiesinGreekSyntax, 227–240.Dordrecht:Kluwer.

Tomic,Olga.2000.“On clitic sites.” In Frits BeukemaandMarceldenDikken,eds.,Clitic phenomenain Europeanlanguages. Amsterdam:Benjamins.

Tsimpli, IanthiMaria.1995. “Focusingin ModernGreek.” In Katalin E. Kiss, ed.,Discourseconfigurationallanguages, 176–204. Oxford: Oxford UniversityPress.

Tsimpli, IanthiMaria,andAnnaRoussou.1996.“Negationandpolarityitemsin Greek.”TheLinguisticReview13: 49–81.

Veloudis,Ioannis.1982. Negationin modernGreek. Doctoraldissertation,Univer-sity of Reading.

Sta metagenestera Mesaiwnika Ellhnika opwc kai se diaforec sÔgqronec dia-lektouc oi proswpikec klitikec antwnumiec topojetountai se jesh pros�rthshcproc kapoio eswteriko kombo thc frashc IP.Se allec dialektouc oi idiec antwnu-miec topojetountai se jesh pros�rthshc proc to rhma enw stic pio kainotomoucdialektouc (sumperilambanomenhc kai thc koinhc Neoellhnikhc) eqoun exeliqjeise rhmatika projemata. H parousa ergasÐa dieurunei thn tupologia twn kli-tikwn stoiqeiwn pou eqei protajei apo touc Fontanakai Halpernkai parousiazeimia sugkekrimenh poreia thc istorikhc exelixhc twn klitikwn tou tupou Xmax serhmatika projemata, proteinontac wc endiameso stajmo klitika tou tupou X0.H istorikh exelixh twn klitikwn fanerwnei epishc thn anaptuxh enoc sÔnjetoukombou, tou ΣP.