Diachronic pertinacity of light verbs
Miriam Butt
a,* , Aditi Lahiri
baUniversityofKonstanz,Germany
bUniversityofOxford,UnitedKingdom
Abstract
WecontrastthehistoricaldatawithrespecttolightverbsandauxiliariesinIndo-Aryanandshowthatlightverbsarecomparatively stableandunlikelytobesubjecttoreanalysisorrestructuring.Weproposethatthereisaverytightconnectionbetweenalightverbandits correspondingmainverb,andthatthisconnectiondiffersmarkedlyfromtherelationshipanauxiliarybearstothemainverbitisderived from.Inparticular,wedepartfromthereceivedviewthattheexistenceofalightverbisduetoahistoricalprocessofsemanticbleaching.
Weinsteadproposethatsynchronicallythereisasingleunderlyinglexicalentrywhichtightlybindslightverbstotheircorrespondingmain verb.Thisproposalaccountsnotonlyforthesimultaneoussynchronicusesoflightandmainverbs,butalsoforthehistoricaldata.
Keywords: Urdu;Bengali;Auxiliaries;Historicalpertinacity;Lightverbs;Complexpredicates;Reanalysis;Grammaticalization
1. Introduction
Complexpredicate formation isan integral partof South Asianlanguages (see Chatterji,1926; Masica, 1976 on SouthAsia asalinguisticarea).WeconcentrateonV--Vcomplex predicatesofthetypein (1).1,2
(1) a. nadya=ne xɑt lıkh di-ya
Nadya.F=Erg letter.M.Nom write give-Perf.M.Sg
‘Nadyawrotealetter(completely).’ (Urdu)
b. nadya a gɑ-yi
Nadya.F.Nom come go-Perf.F.Sg
‘Nadyahasarrived.’ (Urdu)
*Correspondingauthorat:FBSprachwissenschaft,UniversitätKonstanz,Fach184,78457Konstanz,Germany.Tel.:þ497531885109;
fax:þ497531884865.
E-mailaddress:miriam.butt@uni-konstanz.de(M.Butt).
1Thelistof abbreviationsused in thispaper isasfollows: A= Atmanepadam,Acc = Accusative,Aug= Augment,Caus =Causative, Cl=Classifier,Compl=Complement,Dat=Dative,Dem=Demonstrative,Emph=Emphatic,Erg=Ergative,F=Feminine,Fut=Future,Gen= Genitive,Imp=Imperative,Impf=Imperfect,Ind=Indicative,Inst=Instrumental,M=Masculine,Nom=Nominative,Obl=Oblique,Obj= Objective,Opt=Optative,P=Parasmaipadam,PP=PastParticple,Perf=Perfect,Pl=Plural,Pres=Present,Pron=Pronoun,Ptcp=Partciple, Redup=Reduplication,Rel=Relative,Sg=Singular,Top=Topic,Voc=Vocative,W=Weak.A‘-’indicatesamorphemeboundary,a‘=’aclitic boundary.
2Notethatthisisarevisedandupdatedversionofapaperwhosepreviousincarnationswerecalled‘Historicalstabilityvs.historicalchange’ and‘VerbalPassepartouts’.
Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-266903
https://dx.doi.org/10.1016/j.lingua.2012.11.006
InUrdutheformofthefirstverbisidenticaltothestemformandnevercarriesanyinflectionintheV--Vconstruction.
Thisisthemainorfullverb. Thesecondverb carriestense/aspectmarkingandisinflectedaccordingtothestandard paradigmsgoverningsimpleverbinflection.Thisverbislightinthesensethatalthoughitisform-identicaltoamainverb (andhenceglossedwiththemainverbmeaningintheclosegloss),thepredicationalcontributionisnotthatofamainverb (cf. Jespersen, 1965, who first coined the term lightverb). Rather,it serves to modify the main verb semanticsby expressingsuchnotionsascompletion,inception,benefaction,forcefulness,suddennessorvolitionality(Hook,1974).
Theparametersofcompletionandinceptionareresponsibleforthepervasivenotionintheliteraturethattheselightverbs areaspectual (e.g.,Hook, 1991;Singh,1994).
WeconcentrateontheSouthAsianlanguagesUrdu/Hindi3andBengaliinparticular(seeZbavitel,1970;Hook,1974;
Ramchand,1990;Singh,1994;Butt,1995amongothers).4SomeBengaliexamplesareshownin(2).Theformofthefirst verb in the sequence carrieswhat has often been dubbed ‘‘perfective’’ morphology. We gloss this morpheme as a participle, whichreflectsitsOldIndo-Aryanorigin.5
(2) a. rɑm bɑgh-t
_ɑ-ke mer-e phel-l-o Ram.Nom tiger-Cl-Acc hit-Gd throw-Past-3
‘Ramkilled thetiger.’ (Bengali)
b. rɑm eʃ-e por-l-o Ram.Nom come-Gd fall-Past-3
‘Ramarrived.’ (Bengali)
Thefirstverbinthesequenceisagainthemainverb.Thesecondverbcarriestensemarkingandisinflectedaccording tothestandardparadigmsgoverningsimpleverbinflection.BengalilightverbsinV--VconstructionsfunctionliketheUrdu onesintermsofsyntax andsemantics.
Inthispaper,wearguethatthereisaverytightconnectionbetweenalightverbanditscorespondingmainverb,and thatthisconnectiondiffersmarkedlyfromtherelationshipanauxiliarybearstothemainverbitisderivedfrom.Wedepart fromthereceivedviewthattheexistenceofalightverbisduetoahistoricalprocessofsemanticbleaching(e.g.,Hopper andTraugott,1993;Hook,1991)andinsteadproposethatthereisasingleunderlyinglexicalentry,whichcanaccountnot onlyforthesimultaneoussynchronicusesoflightandmainverbs,butalsoforthehistoricaldata.
Section2firstrevisitstheevidencethatlightverbsformasyntacticclassthatisdistinctfromauxiliariesandmainverbs.
We establish that light verbsare characterized by the followingproperties: theyco-predicate witha mainverb in a syntacticallymonoclausalconstructionandtheyservetosemanticallymodulateorstruturetheeventpredicationofthe mainverb.Incontrasttoauxiliaries,theydonotsituatetheeventpredicationtemporallyoraspectually,i.e.,theycannotbe seenasinstancesofIorT,butmustbeanalyzedasasubclassofV.Wethusworkwithanarrower,butmoreprecise understanding oflightverbsthanisoftenassumedintheliterature.
Section3presentshistoricalevidenceshowingthatlightverbstendtobestableorpertinaciousindiachronicterms.
Thisisagainapropertythatsetslightverbsapartfromauxiliaries,whichincontrastareverymutableacrosstime.In section4,wediscussourproposalswithrespecttothesingleunderlyinglexicalentrythataccountsfortheform-identical light and mainverb uses synchronically and acrosstime. Section5 provides a comparative lookat data from other languagefamilies anddiscussesexistingviewswhichcontrastwithours.Section6concludesthepaper.
2. Syntacticstatus
Thereareseveralimportantpointstobemadeaboutthesyntacticstatusoflightverbs.Forone,lightverbsenterintoa copredicationaldomainwiththemainverb.Thatis,themainverbandthelightverbtogetherfunctionasthepredicateof theclause.Butt(1995)presentsdetailedargumentationinvolvinganaphoraresolution,control,andverbagreementwhich shows thattheV--Vconstructionsexaminedherearemonoclausalinthesensethattheclausecontainsonlyasingle subject,asingleobject,etc.Thiscontrastswithconstructionsasin(3),whicharesimilar,butconsistofamainverb(‘go’) and anembeddedclause(‘havingwrittenaletter’).
3TheSouthAsianlanguagesUrduandHindiarecloselyrelated.Bothareamongthe18officiallanguagesofIndiaandarespokenprimarilyin thenorthofIndia.UrduisthenationallanguageofPakistan.ThedatapresentedinthispaperaredrawnprimarilyfromthedialectofUrduspoken inLahore,Pakistan,aswellasfromexamplescitedintheliteratureonbothUrduandHindi.
4NotethatwhilecomplexpredicationexistsinDravidianlanguages,themorphosyntaxofthoseconstructionsdiffersquitemarkedlyfromthatof theIndo-Aryanlanguages.Itisnotpossibletoaddresstheseissuescompetentlywithinthescopeofthispaperandwethereforeconcentrate exclusivelyonexamplesfromIndo-Aryan.
5TheancestralconstructionfunctionedlikeaparticipialandwasdubbedagerundintheWesterntraditionofSanskritscholarship.
(3) nadya [xɑt lıkh kar] ga-yi
Nadya.F.Nom letter.M.Nom write having go-Perf.M.Sg
‘Nadya left,havingwrittenaletter.’ (Urdu)
In order to be maximally clear about the contrast between monoclausal vs. biclausal constructions, we here show the relevant differences in terms of an LFG (Lexical-Functional Grammar) analysis that is based on Butt (1995).Thef(unctional)-structureinFig.1correspondstothemonoclausalcomplexpredicatein(1),thef-structurein Fig. 2 to the biclausal sentence in (3). The precise details of theLFG analysisare notimportant here(for details see Butt, 1995). Important is that complex predicates correspond to a monoclausal functional structure, not a biclausal one.
Sincecomplexpredicationsarefunctionallymonoclausal,thereisatendencyintheliteraturetoanalyzelightverbsas beinganalogoustoauxiliaries.Thatis,asbeingfunctionalelements.However,thistendencyonlyservestoobscurean important syntacticdimension: itcanbe shownfor languageafterlanguagethat lightverbsformadistinct class that behavesquitedifferentlyfromthatofauxiliaries(Butt,2003,2010).Someoftherelevantevidenceisreviewedinthenext sections.
2.1. Lightverbsvs.auxiliaries
Analyses across the theoretical spectrum have tended to analyze light verbs as a type of auxiliary. In recent developmentswithintheMinimalProgram(MP),forexample,vhasbeenregardedasanaturalpositionforlightverbs (Adger,2003,134),particularlyasitisassociatedwithanotionofcauseandcanthusbeusedtoaccountforcausative formationinlanguageslikeRomance,wherethecausativeverb‘make’hasbeenanalyzedasalightverb(Rosen,1989;
Alsina, 1996). Butt and Ramchand, 2005 propose v as the relevant position for Urdu/Hindi light verbs of the type investigatedhereaswellas forpermissives.
Fig.2. F-structureofbiclausalsentence.
Fig.1. MonoclausalF-structureofcomplexpredicate.
However, v is not universally reserved forlight verbs. The position has, for example, also been associated with auxiliaries (Chomsky, 1957) and (root) modals (e.g., Roberts and Roussou, in press, 47). In Head-Driven Phrase StructureGrammar(HPSG)Abeilléetal.(1998)analyzebothtenseauxiliariesandcausativefaireconstructionsinFrench ascomplexpredicates,thusconflatingacausativelightverb(faire‘make’)withotherauxiliariesinthelanguage.6Within the more functionally oriented literature, Hook (1991), for example, sees light verbs as a stage in the on-going grammaticalization frommainverbstoauxiliaries.
Wewouldthereforeliketomakeclearthatlightverbsneedtoberecognizedasadistinctsyntacticclass.Thepointcan be illustrated quite straightforwardly with respect to Urdu, where auxiliaries and light verbs show distinct syntactic behaviorswithregardtocasemarking,wordorder,reduplicationandtopicalization.Inthispaper,werestrictourselvesto presentingevidence fromwordorderandreduplication(see Buttand Geuder,2001forfurtherevidence).
Urdu/HindiisanSOVlanguageswithfairlyfreewordorder.Theverbmayexpandintoarelativelylargeverbalcomplex, asshownin(14).7Wordorderwithintheverbalcomplexisveryrigidandauxiliariesandlightverbsdonotappearinthe samepositionalslot.Asshownin(4),lightverbsandauxiliariesarenotinterchangeable.
(4) MainVerb(Light Verb)(Passive)(AspectualAuxiliaries) (Tense Auxiliary)
(5) a. bɑʧʧa so ʤa rɑh-a he
child.M.Nom sleep go stay-Perf.M.Sg be.Pres.3.Sg
‘Thechildisgoingtosleep.’ (Urdu)
b. *bɑʧʧa so rɑh gɑ-ya he
child.M.Nom sleep stay go-Perf.M.Sg be.Pres.3.Sg
‘Thechildisgoingtosleep.’ (Urdu)
AreviewernotesthattheorderingrigidityexhibitedintheUrduverbalcomplexdoesnotbyitselfnecessarilyargue againstananalysisofthelightverbasanauxiliary.Thelightverbcouldinprinciplebeaspecialkindofauxiliarythatis restrictedtoacertainslot,justastheaspectualandtenseauxiliariesare(andasfoundintheEnglishauxiliarysystem).
Thisistrue,whichiswhythisisnottheonlyevidenceweadduce(seebelow).However,notethepositionofthepassive auxiliaryandalsonotethatUrduhasamorphologicalcausativewhichattachestotheverbstem(seeSaksena,1982for details).Thepartsoftheverbalcomplexwhicheffectchangesintheargumentstructureofthemainverb,i.e.,passives andcausatives,areclosertotheverbthantheaspectualandtenseauxiliaries.InMinimalistanalysesthiswouldfollow nicelyfromtheassumptionthatapectualauxiliariesaresituatedinAspPandtenseauxilariesinTP/IP,whichareboth consideredtobeabovethevP/VP,whichinturnwouldcontaintheargumentchangingelementssuchasthelightverb,the passiveauxiliaryandthecausativemorpheme(e.g.,seeJulien,2002foracrosslinguisticperspectiveandRamchand, 2008 forUrdu/Hindiin particular).Theposition ofthelightverb asoccuring closetothe verband beforethepassive auxiliarythussuggeststhatitisnotsituatingtheeventdescribedbytheverb,asthetenseandaspectauxiliariesdo,butis partofthevP/VPinwhich theoverallargument structureoftheverbisdetermined.
Another difference between light verbs and auxiliaries is found in the possibilities for reduplication. In complex predicates, eitherthe lightverb orthemainverborboth maybe reduplicated(see Fitzpatrick-Cole,1994,1996 fora detailedstudyonBengali,cf.Abbi,1991).Examples(6)and(7)illustratereduplicationofthemainandthelightverbin Urdu, respectively.Anauxiliarymaynot besimilarlyreduplicated,as (8)and(9)show.
(6) a. vo so-ti th-i
Pron.3.Sg.Nom sleep-Impf.F.Sg be.Past-Sg.F
‘Sheusedtosleep.’ (Urdu)
b. vo so-ti voti th-i
Pron.3.Sg.Nom sleep-Impf.F.Sg sleep.Redup be.Past-Sg.F
‘Sheusedtosleep(andsuchlikethings).’ (Urdu)
6MostoftheworkthatweareawareofwithinHPSGtreatslightverbsaseitherauxiliariesorasavariationonraising.OneexceptionisChoiand Wechsler’s(2001)analysisofKoreanN--Vformationswith‘do’.Here,thelightverbfollowsthegeneralschemaofaraisingverb,butspecial mechanismsareintroducedtoreflectthemonoclausalpropertiesoftheconstruction.
7Indeed,(4)isasimplificationoftheactualsituationbecausetheslotoccupiedbythemainverbcouldalsobeoccupiedbyanN--VorAdj--V complexpredicate--thelightverbslotshownin(4)wouldstackontopofthese.Nounincorporationstructurescouldalsoappearinthemainverb slotandwehavenotsaidanythingabouttheinteractionwithothertypesofcomplexpredicatesliketheUrdupermissive(seeButt,1995;Buttand Ramchand,2005).Detailingalltheseinteractions,includingamorefine-grainedlookatthesetofavailableaspectualauxiliaries(ButtandRizvi, 2010)wouldtakeustoofarafieldfromthecentralpoint,whichisthatthedistributionofauxiliariesdoesnotoverlapwiththoseofthelightverbs.
(7) a. vo so ʤa-ti th-i
Pron.3.Sg.Nom sleep go-Impf.F.Sg be.Past-Sg.F
‘Shetousedtogotosleep.’ (Urdu)
b. vo so ʤa-ti vati th-i
Pron.3.Sg.Nom sleep go-Impf.F.Sg go.Redup be.Past-Sg.F
‘Sheusedtokeepgoingtosleep(atinopportunemoments).’ (Urdu)
(8) a. vo so rɑh-i th-i
Pron.3.Sg.Nom sleep stay-Perf.F.Sg be.Past-Sg.F
‘Shewassleeping.’ (Urdu)
b. *vo so rɑh-i vahi th-i
Pron.3.Sg.Nom sleep stay-Perf.F.Sg stay.Redup be.Past-Sg.F
‘Shewassleeping.’ (Urdu)
(9) a. vo so-ti th-i
Pron.3.Sg.Nom sleep-Impf.F.Sg be.Past-Sg.F
‘Sheusedtosleep.’ (Urdu)
b. *vo so-ti th-i si
Pron.3.Sg.Nom sleep-Impf.F.Sg be.Past-Sg.F be.Redup
‘Sheusedtosleep.’ (Urdu)
Finally,unlikeauxiliaries,lightverbsalwaysspantheentireverbalparadigm.Thatis,whileauxiliariestendtoappearin onlypartoftheverbalparadigm,lightverbsmayappearinanytense/aspect(cf.ButtandRizvi,2010).Forexample,the presentand futuretenseareexpressed by theauxiliary versionofho ‘be’ (cf.(5)).Thisforminflects fornumber and person. However,the pasttense isexpressed by a suppletiveform th, which inflects for number and gender (see examples above).Theverb rɑh-‘stay’forms thesimple progressiveby usingthe morphologythat usuallyexpresses perfect/past:rɑh-a/i/e.Therearegood historicalexplanationsfortheseseeminlyconfusing patternsandthehistorical explanations areentirelyin linewith whatisknown aboutthe grammaticalization ofverbsinto tense/aspectmarkers crosslinguistically (Bybee et al.,1994).However,defective paradigmsof idiosyncraticpatterns of verbalmarkingare neverfoundwithlightverbs.Lightverbsinflectjustliketheirmainverbcounterparts.Lightverbsandauxiliariesthushave averydifferentdistribution.
2.2. Lightverbsvs.mainverbs
Lightverbsalsocontrastwithmainverbsintermsofdistributionandsyntacticbehaviour.Inthissection,wediscuss somerepresentativedatatoshowthatlightverbscannotbeanalyzed asmainverbs.
RecallthatV--Vsequencesaresometimes ambiguous betweenacomplex predicateandan embeddedadverbial structure.Anovertrealizationofkar‘having’precludesanypotentialdisambiguity,asitisfelicitouswithanembedded adverbialstructure butnotacomplexpredicate,asshown in(10).8
(10) a. nadya=ne ɑbu=ko mɑkan
Nadya.F=Erg father.M.Sg=Dat house.M.Nom [bɑna *kɑr di-ya]
make having give-Perf.M.Sg
‘Nadyabuiltahouse(completely)for(her) father.’ (Urdu)
b. nadya=ne ɑbu=ko [mɑkan bɑna kɑr] di-ya
Nadya.F=Erg father.M.Sg=Dat house.M.Nom make having give-Perf.M.Sg
‘Havingbuiltahouse,Nadyagave(it)to(her)father.’ (Urdu) Itispossibletoinsertanotherdirectobjectintheadverbialconstruction,butnotinthecomplexpredicateone.Thisis because theadverbialconstructionin(11b)consistsoftwoindependent verbs,oneofwhich (bɑna‘make’)headsan
8Example(10)iswell-formed(intherightcontext)becauseUrduisapro-droplanguagewhichallowsanyandallargumentstobedroppedas longastheycanberecoveredfromthediscourse.Thistypeofpro-dropischaracteristicofSouthAsianlanguagesingeneral.
embeddedclause whosesubjectiscontrolledbythematrix subject.Thecomplexpredicate,on theotherhand,isan instanceofcopredicationwhereboththeverbscombinetoprovideasinglepredicationalheadandthelightverbcannot actasanindependentpredicator.Assuch,thereisnosubjectcontrol,noembeddedclause,andonlyonepossibleobject.
Thisisshownin (11).
(11) a. *nadya=ne ɑbu=ko mɑkan bɑna peısa di-ya
Nadya.F=Erg father.M.Sg=Dat house.M.Nom make money.M.Nom give-Perf.M.Sg
‘Nadyabuilt(her)fatherahouse(completely)money.’ (Urdu)
b. nadya=ne ɑbu=ko [mɑkan bɑna kɑr] peısa di-ya
Nadya.F=Erg father.M.Sg=Dat house.M.Nom make having money.M.Nom give-Perf.M.Sg
‘Havingbuiltahouse,Nadyagave(her) fathermoney.’ (Urdu)
Anotherdifferencebetweenmonoclausalcomplexpredicatesandbiclausalconstructionssuchasin(11b)isprosodic.
ComplexpredicatesareknowntohavespecialprosodicpropertiesinUrduandBengali(BayerandLahiri,1990;Hayesand Lahiri,1991;LahiriandFitzpatrick-Cole,1999;Fitzpatrick-Cole,1996)inthatthetwomembersofthejointpredicationare contained within one phonological phrase. Relevant contrasts from Urdu and Bengali are shown in (12) and (13), respectively.9
(12) a. nadya=ne xɑt (lıkh di-ya)ϕ
Nadya.F=Erg letter.M.Nom write give-Perf.M.Sg
‘Nadyawrote aletter(completely).’ (Urdu)
b. nadya=ne xɑt (lıkh (kɑr))ϕ (di-ya)ϕ Nadya.F=Erg letter.M.Nom write having give-Perf.M.Sg
‘Havingwrittenaletter,Nadyagave(it).’ (Urdu)
(13) a. ʃæmoli ɑrʃola-t
_a-ke (mer-e phel-e-ʧh-e)ϕ Shamoli.F.Nom coakroach-Cl-Acc hit-Gd throw-Perf-be-3
‘Shamolikilledthecoakroach(completely).’ (Bengali)
b. ʃæmoli ɑrʃola-t
_a-ke (mer-e)ϕ (phel-e-ʧh-e)ϕ Shamoli.F.Nom coakroach-Cl-Acc hit-Gd throw-Perf-be-3
‘Havingkilledthe coakroach,Shamolithrew it.’ (Bengali)
c. ʃæmoli ɑrʃola-t
_a-ke (mer-e)ϕ (phel-e di-e-ʧh-e)ϕ Shamoli.F.Nom coakroach-Cl-Acc hit-Gd throw-Gd give-Perf-be-3
‘Having killedthecoakroach,Shamolithrewit(away).’ (Bengali) Lightverbsthushavepropertieswhich differmarkedly fromthoseofmainverbs.10 Inthe nextsection, weturnto evidence whichshowsthat lightverbsmustalsobedifferentiatedfromserialverbs.
9NativeBengalispeakerswouldprefertodisambiguate(13b)moreclearlybyaddingalightverb,asin(13c).Urdu/Hindinativespeakersprefer theexplicituseofkar‘having’asin(12b)inthebiclausalcontext.
10Atthispoint,onereviewerwondersabouttestsformonoclausality.Asmentionedinsection2,Butt(1995)adducesseveralargumentsforthe monoclausalityoftheseV--Vcomplexpredicatesbasedonanaphoraresolution,controlandagreement.Theargumentshavealreadybeen repeatedinseveralpapersandthereaderisreferredtoButt(1997,2010),ButtandRamchand(2005).Therevieweralsowonderswhether modificationwithtimeadverbialsornegationcanbeusedasdiagnostics.Thisraisesapointthatiswellworthmaking.Wehavefoundthattime adverbialsandnegationaretrickytouseasdiagnosticsforsyntacticmonoclausalityaswhattheypickoutare(sub)eventsofagivenpredication.
WithregardtotheV--Vcomplexpredicatesitisindeednotpossibletousetimeadverbialsinthesameclause(e.g.,*Nadyatodaybuild-gavea houseforherfathertomorrow.’).However,itismarginallypossibletousetwotimeadverbialsinUrdupermissives,whicharealsosyntactically monoclausalbythesyntactictests(agreement,control,anaphora)(e.g.,Nadyayesterdayallowedherfathertobuildahousenextyear.).The adverbialtestisthusnotconclusive.Negation,interestingly,canbeusedtopickoutjustthelightverbintheV--Vcomplexpredicates,thoughitis todatenotclearjustwhatisbeingnegatedsemantically.NegationofUrdu/HindiandBanglaV--Vcomplexpredicatesingeneralisactuallyvery restrictedandonlyholdsunderasetofconditionswhichareatpresentnotunderstood.NegationthusshowsthattheV--Vcomplexpredicatesare specialinsomesense,butthetestisagainnotconclusiveasthemonoclausalUrdupermissivesdoallownegationofeachsubevent(e.g.,Nadya didn’tallowherfathertobuildahouse.vs.Nadyaallowedherfathertonotbuildahouse.).Foragooddiscussionofsomeproblemswiththese diagnosticswithrespecttothemonoclausalvs.biclausalstatusofEnglishhave-constructions(asinNadyahadherfatherbuildahouse.),see Tantos(2008).
2.3. Lightverbsvs.serialverbs
ThecomplexpredicatesinvestigatedhereconsistofV--Vsequences. Acomparisoncouldthereforebe drawnto anotherwellknownmultiverbphenomenon:thatofserialverbs.Thissectiondrawsadistinctionbetweenprototypical complex predicatesandprototypicalserialverbs.AquickcomparisonofonetypicalsortofserialverbfromAkanin (14)withatypicalUrduV--Vcomplexpredicatein(15)illustratessomeconcretesyntacticandsemanticdifferences.
Serial verbstypicallystackseveraleventsin asingleclause. Thisisnotpossiblefor complexpredicates.The light verble‘take’in(15)merelycontributesadditionalinformationtotheextractingevent,butdoesnotcontributeanevent of itsown. In(14),ontheotherhand,therearethreedistinct events:agettingevent,abuilding eventandaselling event.
(14) Gyasiba nyá-à sika sí-ì dan tɔn-èè Gyasiba get-Compl money build-Compl house sell-Compl
‘Gyasibagotmoney,builtahouseandsoldit.’ (Akan)
(Osam,2003)
(15) ɑnjIm=ne pɑthɑr=ko bahar nıkal li-ya
Anjum.F.Sg=Erg stone.M=Acc out extract take-Perf.M.Sg
‘Anjumextractedthestone.’ (Urdu)
Syntactically,thekindofobjectsharingmadefamousbyBaker(1989)forserialverbsisnotacharacteristicofcomplex predicates. Another well-known property associated with serial verbs is the sharing of a single subject and the requirementthateachverbbeseparatelyanduniformlymarkedfortense/aspect.ThisistrueforAkan,forexample.Itis alsotrueforPaameseand(16)providesanexamplewhichmakesthepointverynicely:eachmemberoftheserialverbin Paamesedisplaysagreementfeatures whichhavetobe consonantwithoneanother.
(16) iire rehe-sooni vakilii rehe-haa
1Pl.Incl 1Pl.Incl-Distant.Throw canoe 1Pl.Incl-Distant.Go
‘Wewillgo, putting(throwing) ourcanoetosea.’ (Paamese)
(Crowley,1987:47)
Thispropertyisnotexhibitedbytypicalcomplexpredicates:onlyoneofthemembersoftheverbalcomplexcarries tense/aspect or agreement features. This in turn indicatesthat what isat stake isa difference atthe levelof event structure: complexpredicatesdescribeoneevent,whereasserialverbsdenoteacomplexevent.
WefollowButtandRamchand(2005)andButtandGeuder(2001)inanalyzingthefunctionoflightverbsasprovidinga modulation ormodificationofthe maineventsemantics.In contrast,as shownby Durie(1997),serialverbstypically appeartodenoteacomplexconceptualevent.Serialverbscanthusonlybeusedtodescribeeventswhichare‘‘normal’’
inagivenculturalcontext.Forexample,buytakefishisanexpectedsequenceofevents,whileselltakefishisnot.In Srananthe formerthereforeexists asaserialverb,while thelatterdoesnot.AnotherexamplecomesfromAlamblak (Bruce,1988),whereanactionwhichinvolvesclimbingatreeinordertolookforinsectsisareasonablecomplexevent, but anactionwhichinvolvesclimbing atreeinordertolookatthemoon isnot.
The literatureon serialverbsandverbal complexes isvastandthereare severalworkswhich attempta formal characterizationofthepropertiesofserialverbs(e.g.,Sebba,1987;Durie,1997;Osam,2003).However,muchmore needs to be done for a precise characterization of both serial verbs and complex predicates that can lead to a comprehensive theory about differing typesof verbal complexes. The work presented hereis intended as a step towardsamoreprecisecharacterizationofthepropertiesofcomplexpredicatesandlightverbs.Foradiscussionof serial verb constructions which proposes a typologythat is very much inline with theviews presented here, see Ameka (2006).11
11Ameka’scharacterizationincludesdiachronicargumentsinthatserialverbsareknowntogiverisetoprepositions,complementizersand aspectualmarkers(e.g.,Lord,1993).Theclaiminthispaperisthatlightverbsarehistoricallyinert.Areviewerhassuggestedthattheserialverb datainvalidatesthediachronicevidenceadducedhereforlightverbs.However,notethatwhiletheevidencepresentedinthispaperisdrawnfrom historicalsources,theevidenceforthegrammaticalizationofserialverbsisbasedprimarilyoncomparativesynchronicdata(sincenooronly fragmentarydiachronicevidenceisavailableformostoftheserializinglanguagesthathavebeeninvestigated)andanexpectationofthekindsof historicalprocessesthatshouldapply.
2.4. Lightverbsvs.compounds
TheV--Vconstructionsin(1)arealsosometimesanalyzedas compoundverbs(e.g.,Hook,1974)thusshiftingthe burdenofverbalcompositionintothelexicon.However,theseV--Vconstructionsexhibitnoneofthepropertiesusually associated with compounds. If the V--V constructions were compounded, one would expect a tight morphological relationship between the lightverb andthe mainverb. However,thisisnot so: as (17)shows, the lightverb canbe separatedfromthemainverbviaphenomenasuchastopicalizationandthetwoverbscanbeseparatedbyafocusclitic.
(17) a. bɑʧʧa (so)v (gɑ-yav) child.M.Nom sleep go-Perf.M.Sg
‘Thechildhasgone tosleep.’ (Urdu)
b. so to bɑʧʧa gɑ-ya
sleep Top child.M.Nom go-Perf.M.Sg
‘Thechildhasgone tosleep.’ (Urdu)
c. bɑʧʧa so hi gɑ-ya child.M.Nom sleep even go-Perf.M.Sg
‘Thechildhasevengonetosleep.’ (Urdu)
Onewouldalsoexpectthemainverbtobeinitsstemform(asisthecaseinnominalcompounding,cf.theLevel- OrderingHypothesis(Kiparsky,1982a)),however,astheexamplesfromBengaliin(2)showed,thisisalsonotthecase:
the mainverbcarriesperfectiveorgerundivemorphology.12
Weconclude that the twoverbsmusttherefore beconsidered to be separatesyntacticentities and thatthe term compoundverb isamisnomer.
2.5. Summary
Lightverbsformadistinctsyntacticcategorywithdistinctdistributionalproperties.WefollowButtandGeuder(2001b) andviewlightverbsasasubtypeofV,ratherthanasasubtypeofanauxiliary(e.g.,TorI).Thatis,weseelightverbsas primarilylexicalinnature(ratherthanfunctional).However,whilelightverbsareinthelexicaldomain,theyarenotmain verbs(orserializingverbs),butinsteadinteractwiththesemanticsofamainverbbymodifyingthemainevent(section4).
3. Historicalevidence:pertinacityvs.transience 3.1. Semanticbleaching
Thecrosslinguisticexistenceoflightverbsisgenerallyattributedtoaprocessofsemanticbleaching(e.g.,Hopperand Traugott,1993;Hook,1991),whichisassumedtobeconnectedtoahistoricalprocessofgrammaticalization,whereby contentverbsproceedalongtheclineshownin(18)ontheirwaytoareanalysisasfunctionwords/morphology.Theterm vectorverbisusedin onetraditionofliterature onSouthAsianlanguagestodesignate lightverbs.13
(18) full verb>(vectorverb)>auxiliary >clitic>affix
(Grammaticalization Cline,Hopperand Traugott,1993:108)
Clinesasin(18)aregenerallypositedaspartoftheGrammaticalizationapproach(Lehmann,1982,1985).14Thecline makesahistoricalpredictionthatlightverbsarederivedhistoricallyfrommainverbsandthatfromtheretheydevelop furtherinto auxiliariesandfinally intoverbalmorphology.
Inparticular,Hook(1991,1993)analyzestheexistenceoflightverbsinSouthAsianlanguagesintermsofahistorical processofaspectogenesis.Thelightverbsareseenasaclassofperfectiveaspectualauxiliarieswhichhavedeveloped from main verbs via grammaticalization. In a larger and later survey, Hook (2001) is more careful with the claims
12ThemainverbinUrdu/HindididaswelluntilitwaslostinOldUrdu/Hindi(seesection3.2.2).
13ThetermvectorverbisoriginallyduetoPray(1970).
14Itisnoteworthythatthisapproachistheonlyanalysiswithinhistoricallinguisticswearecurrentlyawareofwhichmakesacleardistinction betweenlightverbsandauxiliariesinthatlight(vector)verbsareseenasanoptionalprecursortoauxiliaryformation.
surroundingperfectivity,notingthatthenotion‘‘perfective’’isnotwelldefinedandthatlightverbsshowfunctionswhichgo beyondthepurelyaspectual.
In our view, the aspectogenesis hypothesis runs counter to the observable facts in that light verbs do not behave likeauxiliaries either synchronicallyordiachronically.Thenextsection provides evidenceforthehistorical pertinacity of light verbs in Indo-Aryan. This pertinacity is then contrasted with the comparative mutability of auxiliaries in section 3.3.
3.2. Diachronicpertinacityoflightverbs
It is generally agreed (e.g., Hook, 1991; Tikkanen, 1987; Hendriksen, 1944; Chatterji, 1926) that the ancestral constructionofthemodernV--VcomplexpredicateistheSanskrit‘‘gerund’’or‘‘absolutive’’in-tvā(ya),or-ya/yā.These suffixesservedasderivationalmorphemeswhichresultedinanindeclinableparticiple(e.g.,Whitney,1889:345--360).15 The useofthe tvāparticiples wasmanifoldandvaried.Tikkanen(1987:7)usesthe followingconstructed exampleto illustratethevariouspossibletranslationsfound intheliterature.
(19) indram ārabh-ya car-a Indra.ACC grasp-GD go-IMP.2SG a.‘HavingtakenholdofIndra,move!’ b.‘TakeholdofIndraand move!’ c. ‘TakeholdofIndrabeforemoving.’ (d.‘Movebytakingholdof Indra!’) (e.‘Go totakeholdofIndra!’)
f.‘KeepyourselftoIndra! (Sanskrit)
Notethatwhilethetranslationalpossibilitiesina--daretruth-conditionallyequivalent,thepossibilitiesinaandfcanbe contrasted withone another. Thesetwotranslationalpossibilitiesare reminiscent ofthe patternfound in the modern languages:theusageinfiscomparabletothemodernV--Vcomplexpredicate,whileaiscomparabletothebiclausal adverbialreadingfoundwiththesameV--Vstring (cf.(13)).
3.2.1. Clausechaining
TheSanskrittvāisalsosometimesreferredtoasaconjunctiveparticiple(CP),presumablytoreflecttheclausechaining effectundertheadverbialreading.ThesequentiallinkingofclausesviathetvāparticiplewasverycommoninSanskrit.An exampleisshownin(20).
(20) ... [up-tvā kes´as´mas´rūn
_i] [nakhāni ni-kr
_-tya] [ājyā{-} bhyaj-ya] ... shear-Gd hair-of-head-and-beard.Nom.Pl nail.Nom.Pl in-do-Gd clarified-butter.Nom.Pl smear-Gd
‘...havingshavedhishairandbeard,cut hisnails,smearedhiseyesandanointedhisbody,...’ (Sanskrit) (Jaiminīyabrāhman
_a,fromTikkanen(1987:187)
This‘‘clausechaining’’phenomenonisstillcommonin themodernlanguagesandemploysthe same(-e)thatis
usedincomplexpredicatesinBengali((13)),andwhichisdescendedfromtheoriginalparticipialaffix(section3.2.2).
Chatterji(1926:1011)citesanexamplewithasmanyas15subclausesforBengali.Justthreeofthese15areshown in (21).
(21) bhore ut _
h-e, ʃt
_eʃane põuch-e, t _ikit
_ kor-e, ... col-e jeo early rise-Gd station reach-Gd ticket buy-Gd walk-Gd go.Imp
‘getupearly,gettothestation,buyaticket...and leave’
[‘havinggotten upearly,havinggone tothestation,havingboughtaticket...,leave’]
(Bengali,Chatterji,1926:1011) Thefinal-eaffixon‘walk’doesnotsignifyyetanothermodificatoryadverbial,butmarksthemainverbofthecomplex predicate ‘walk go’(=‘leave’). This is the mainpredication of the sentence. Chatterji’s example nicely illustrates the
15Thisparticiplereplacedavariantwith-tvīandisthoughttobeafrozenformofanoldinstrumentalverbalnounintu(Speijer,1886,§379, Chatterji,1926:1008).
possibilities forinterpretative ambiguity between a verb embedded in a ‘having’-type clause and the mainverb of a complex predicate. This interpretive ambiguity has survived intact in the modern language, but can be resolved prosodically,as wasseenin section2.2.
3.2.2. Theancestralmorphology
Theancestryofthemodernformsinboththeclausechaining(adverbial)constructionaswellastheV--Vcomplex predicates has been reconstructed in some detail by Chatterji (1926:1006--1011), Kellogg (1893:341) and Beames (1872--1879,vol.3,§73).Welistthoseformswhichpertain directlytoBengaliandUrdu/Hindiin(22).16
(22) Sanskrit Early MIA LateMIA Old ModernColl.
(P¯¯ali) Bengali Hindi Bengali Hindi tv ¯a tv ¯a
- - -tt¯a,-t¯a, -cc ¯a -cca,
-y ¯a/ya -ya,-iya -ia, -i i¯a,-iy¯a -i -e
Sanskritshowsevidenceforadistributionaldifferencebetweentvā,whichwascommonlydubbedtheAbsolutiveI,and ya,theAbsolutiveII.Theyawasgenerallyusedinconjunctionwithpreverbs(Whitney,1889,§989,alsoseeMacdonell, 1917forVedic).Giventhatthemodernlightverbscontributetothepredicationoftheeventinmuchthesameelusiveway aspreverbsdidinSanskrit,thismaybeasignificantfactorinthespreadofV--Vconstructions(section5).Itmightalso furnishanexplanationforwhyitwastheyaandnotthetvāformwhichwastheancestorofthemodernforms.Ontheother hand, Chatterji (1926:1008) and Whitney (1889, § 990) assert that a number of exceptions to this complementary distributioncanalreadybefoundinlaterSanskritandthatitappearstohavebeencompletelyignoredinMiddleIndo- Aryan.
The-yaforminearlyMiddleIndo-Aryanalreadyshowedsignsofassimilation(Chatterji,1926),withthe-i-ofthe-iya formcomingfromtheprecedingmaterial.Thisshortvowelcanbefoundinthewritingsofmedieval(Old)Hindi,buthas beenaltogetherlostinmodernUrdu/Hindi.ForBengali,ontheotherhand,Chatterji(1926:1010)assumesthattheshort vowelwasstrengthenedbyeithera‘‘definitive’’formā,orbytheāfromanextendedandrelatedformoftheparticiple:
iyān_ã,whichbecame-i ¯a˜.TheliteraryformofBengaliretainstheOldBengaliiā,-iyā,thecolloquialformexhibitstheshort vowel -e.
3.2.3. Ancestralusage
Wethus seethatthe ancestral ‘‘absolutive’’ orgerund in -yahaspersisted in the modernlanguages and that an unbroken chain of developments can be reconstructed for the form of the New Indo-Aryan affixes. Beyond this morphologicalcontinuity,ourcontentionisfurthermorethatthelanguagesalwaysallowedfortheuseoflightverbsand that ineverycase wherea lightverbreadingisprobable,thelightverb isin fact form-identicaltoa mainverbin the language.
WebeginwithclassicalSanskritandVedicasdiscussedbyTikkanen(1987),whonotesthattheconstructionsformed with -tvā/ya displayed somecurious syntacticand semantic properties.Syntactically, there isevidence for akind of
‘‘clauseunion’’inthesensethatnegationandwh-operatorsappeartobeabletotakescopeoveralltheverbsintheclause
(Tikkanen, 1987:16). Inaddition, while thedefaultinterpretation ofconstructions involvingthe tvā/ya isthe adverbial
‘havingdoneX’(cf.(20)),thisdefaultinterpretationisoften unsatisfactoryTikkanen (1987).
Oneexampleofanawkwardinterpretationunderthedefault‘having’readingistheVedicinstantiationin(23)ofthe constructed Sanskritexamplein(19).ThetranslationshownhereisfromTikkanen,whobaseshimselfonDelbrück’s (1888:406)Germanrendition:diewirwandern,nachdemwirdichergriffen,zumStützpunktgenommenhaben‘wewho wander, after wegrasped you,took youas ourbase’ or die wir unsimmer an dich halten‘thosewho alwayskeep themselvestoyou’.
(23) ime ta indra te
Pron.Dual Dem.Pron.3.Sg Indra.Voc.Sg Pron.2.Sg.Gen
vayam purus
_t _uta
Pron.1.Pl.Nom much.praise.PP.Voc.Sg
16Otherformsincludeformationswith-tvīor-tuandthelateradditionofa-na.
ye tvā-rabh-ya car-āmasi Rel.Pron Pron.2.Sg.Acc-grasp-Gd go-Pres.1.Pl
‘Wehereareyours,Oever-praisedIndra,whowanderabouthavingtakenholdofyou/who constantlykeep
ourselvestoyou.’ (Vedic)
(R_gvedaI.57.4; Tikkanen,1987:175)
Grasmann(1872:437)pointsoutforcar‘go,move’thatinconjunctionwithaparticipleitsmeaningisweakenedtosuch anextentthatitonlyseemstoexpressthedurationoftheactiondenotedbytheparticiple.Thisisthecasein(23),where anyanalysisintermsofanembeddedadverbialyieldsastrangeinterpretation.Ontheotherhand,ifoneanalyzesthe‘go’ asalightverbandthe‘grasp’asamainverb,themoreplausiblereading‘‘weholdontoyou(constantly)’’(=‘weconstantly keepourselvestoyou’)becomespossible.
However,weneedtoadmitatthispointthattheabovepiecesofdatacannotbeseenasunassailableproofofcomplex predication inSanskrit.Rather,theirritation ofscholarslikeDelbrückand thelengthyattemptstodealwiththerather puzzling rangeof interpretationsand structuralproperties associatedwiththe -tvā/ya participle in bothSanskrit (see Tikkanen,1987forasummary)andPāli(Hendriksen,1944),mustbetakenasindicationsthattheconstructionisnotwell understood.
ThissenseofprofoundpuzzlementisahallmarkofcomplexpredicatesinmanymodernSouthAsiangrammarsaswell (where theyarevariouslytreatedas compoundverbs,idioms, adverbials,etc.).Forthemodern languages,however, variouslinguistictestsareavailable(e.g.,AissenandPerlmutter,1983;Neeleman,1994; Mohanan,1994;Butt,1995;
Alsina, 1996; Alsina et al., 1997) so that complex predicate constructions can be distinguished from other V--V combinations.
TheSanskritandVedicformsexaminedabovearedrawnfromdifferenttimespans.Thereisnoprecisedatingofthe timeswhenthedifferentlanguageswerespoken,buttheoldestattestedformisthoughttogobackto1200BCE.Vedicis generallydateduntilabout600BCE.EpicandClassicalSanskritfallintothetimefrom600BCEto200CE.Togetherwith Vedic,thesearereferredtoas OldIndo-Aryan.MiddleIndo-AryanincludesPāli (mainlypreservedin Buddhisttexts), severalPrākritlanguages(whichincludenon-standarddialectsofSanskrit),Apabhrams´aandinscriptionsoftheEmperor As´oka(270--232BCE).TheMiddleIndo-Aryanperiodstretchesfromabout200BCEto1100CE.Thelanguagesofthe period fromthen on are commonly referredto as New Indo-Aryan. As of 1100 CE distinct ancestors of the modern languages suchasOldHindi, OldBengaliorOldMarathihaveemerged.
Complex or joint predication becomes much easier to identify in Middle Indo-Aryan (Hook 1993). The Pāli examplesin(24)bothinvolve theverb‘give’asafinite verbwhichcombineswiththeparticipleof ‘make’.Forboth these sentences it would be strange to assume that the meaning should be rendered as: ‘‘having led her to the hermitage,havingmadeafire,hegave(it)(toher)’’.Rather,thecomplexpredicatebenefactivereadingintheglosses seems moreappropriate.
(24) a. ... assamapadam
_ āne-tvā aggim
_ ka-tvā a-dāsi
hermitage.Acc lead-Gd fire.Acc.Sg make-Gd Aug-give.Impf.3.Sg
‘...broughthertohishermitageandmadeafireforher’ (Pāli) [‘havingbrought(her)tothe hermitage,madeafire(forher)’]
Jatāka TalesI.296.10, SriLanka(Hendriksen,1944:134) b. daruni āhari-tvā aggim_ ka-tvā dassati
sticks bring-Gd fire.Acc.Sg make-Gd give.Fut.3.Sg
‘Bringingwoodhe’llmakeafire(benefactiveuse).’ (Pāli) (Trenckner1879:77,citedbyHook,1993:97)
Hendriksen(1944)providesadetaileddiscussionoftheinterpretationaldifficultiesassociatedwiththegerundinPāli, someofwhichechoesthediscussionforSanskritfoundinTikkanen(1987).ComparethePālitothemodernUrduin(25), wherethefinitelightverb‘give’alsoallowsforabenefactiveinterpretation.
(25) ... apne ghɑr le kɑr rot
_i bɑna d-i
self.Obl house.M.Sg.Nom bring having bread.F.Sg.Nom make give-Perf.F.Sg
‘...havingbroughtherhome, (he/she)madefood(forher/him)’ (Urdu) InNewIndo-AryancomplexpredicatescanbeidentifiedevenmorereadilythaninMiddleIndo-Aryan.Weshowsome examples for OldBengaliin (26)from the Caryapad (950--1550CE), which consistsof 46complete songs and one incompletesongof6linesby24differentpoets.Herethefiniteverbs‘take’((26a))and‘give’((26b))cannotbetakenin
their mainverbsense,but mustratherbeinterpreted as lightverbswhich signalcompletion, muchas isdone in the modernlanguage.Alsonotethat‘give’hasabenefactivesense,while‘take’indicatesaself-directedaction,apattern which istypicalof lightverbscrosslinguisticallyand whichisfound inseveralof themodern SouthAsianlanguages.
Finally,notealsotheseparabilityofthemainandlightverbin(26b).Aswasseeninsection2.4,thispropertyremainsa feature inthemodernlanguages.
(26) a. caus
_athi kot _
ha gun
_-iā lehu
sixty-four rooms count-Gd take
‘countsixty-fourrooms (foryourself)’
(Caryapad12,Mojunder1973:248) (OldBengali)
b. bājule di-la moha-kakhu bhan _-iā Bajula.Obl give-Past.3.Sg roomsofillusion count-Gd
‘Bajulacountedtheroomsofillusion(forhisdisciple).’
(Caryapad35,Mojunder1973:248) (Old Bengali)
WeclosethissectionwithafewexamplesofOldHindiwhichdatetothemiddleages(TulsīDās’Rāmāyan
_aofabout 1575 CEcontainedV--Vformationstosuchanextentthatadissertationonthesubject waswarranted,seeMeißner, 1964).Example(26)andtheexamplesin(27)ofIndrajitofOrchā’sBrajBhās
_āprose(ca.1600CE)areveryclosetothe modernforms.McGregor(1968:209--213)explicitlynotesthattheywereusedmuchlikecomplexpredicatesinmodern Hindi.
(27) a. ... cor-i le-tu hai
steal-Gd take-Impf be.Pres.3.Sg
‘...(he)steals’ (OldHindi)
b. kād
_h-i le-i
pullout-Gd take-Perf
‘(he)pulledout(witheffort)’ (OldHindi)
c. samudrahim _ nās
_-i jā-ta haim
ocean.Obj cross-Gd go-Impf be.Pres.1/3.Pl_
‘(They/We)crossoceans(completely).’ (OldHindi)
Insummary,thissectionhasshownthattheuseoflightverbscanbetracedtoaveryearlystageofthelanguageand thatateverystage,thelightverbwasform-identicaltoamainverb.Thissuggeststousthatratherthanengaginginthe typeofreanalysiswhichleadstotheformationofauxilariesormodals,lightverbsarediachronicallyinertandhavebeen an integralpartofthelanguageateverystage.
3.3. Auxiliariesand diachronicchange
GiventhattheavailabilityanduseoflightverbsinV--Vcomplexpredicatescanbetracedbackoverthemillenia,one mightspeculatethatthediachronicpatternpresentedabovecouldbeduetoageneralstructuralinertnessinIndo-Aryan.
However,whiletheV--Vcomplexpredicateconstructionappearstohavebeenavailableinbasicallythesamestructural formas farbackas thewrittenrecordextends, otherstructuralpropertiesinIndo-Aryanhaveundergoneasignificant amountofchange.Arelevantcontrastisfurnishedbythetense/aspectsystem.TheheavilyinflectionalsystemofOld Indo-AryanhasbeenreplacedbyamixedperiphrasticandinflectionalsysteminNewIndo-Aryanwithmostoftheold inflectionalmorphologyhavingbeenlostoverthemillenia.
Inthissection,welookattwoverbs,‘be’and‘go’,whicharealsousedaslightverbsinV--Vcomplexpredicates.We showthatwhiletheseverbswereusedasmainverbsaswellaslightverbs,theyalsosimultaneouslygaverisetonew periphrasticauxiliaryconstructionsandwerethenfurtherreanalyzedasinflectionaltense/aspectmorphology.Despite theriseofauxiliation,themainandlightverbuseswerenotlost,aswouldbepredictedbythegrammaticalizationcline in (18).
Otherwise,thetypeofhistoricaldevelopmentfoundwiththeseverbsisnothingspecialincrosslinguisticterms:thepath ofchangeisatypeofreanalysisthathasalreadybeenwellestablishedcrosslinguistically(e.g.,HarrisandRamat,1987;
HarrisandCampbell, 1995).
3.3.1. Theverb‘be’
Thefollowingexamplesshow‘be’inBengaliandUrduinamainverb((28)),alightverb((29)),anauxiliaryuse((30)) and exampleswheretheoriginal‘be’isrealizedas aclitic((31a))and amorphologicalaffix((31b)).
(28) a. Main Verb ɑmi bhɑloɑʧhi I.Nomwell be.Pres.1
‘Iamwell.’ (Bengali)
b. Main Verb m~e t
_
hikhũ
I.Nomwellbe.Pres.1.Sg
‘Iamwell.’ (Urdu)
(29) a. Light Verb
amar mone ɑʧhe I.Gen mind.Loc be.Pres.3
‘Iremember.’ (Bengali)
b. LightVerb
mIjhə yad he
I.Obl.Dat memory.M.Sg.Nombe.Pres.3.Sg
‘Iremember.’ (Urdu)
(30) a. Auxiliary
nadya=ko xɑt mıl-e he
Nadya.F=Dat letter.M.Nom receive-Perf.M.Pl be.Pres.3.Sg
‘Nadyahasreceivedletters.’ (Urdu)
b. Auxiliary
nadya=ko xɑt mıl-t-e h~e
Nadya.F=Dat letter.M.Nom receive-Impf-M.Pl be.Pres-M.Pl
‘Nadyausedtoreceiveletters.’ (Urdu)
(31) a. Clitic
rɑm ʧithi pe-(y)e=ʧh-ilo
Ram.Nom letter.Nom receive-Perf=be-Past.3
‘Ramhadreceivedletters.’ (Bengali)
b. Morpheme
rɑm ʧithi pɑ-ʧʧh-ilo
Ram.Nom letter.Nom receive-be-Past.3
‘Ramwasreceivingletters.’ (Bengali)
IntheUrduexamples,the‘be’isimmediatelyrecognizablebecauseit isinstantiatedas alexicalitem thatisform- identicalwiththemainverbuseof‘be’.InBengali,ontheotherhand,itisnotimmediatelyobviousthattheformsglossed as‘be’in(31)arerelatedtothe‘be’of(28)and(29).Thisisbecausetheverbɑʧh‘be’hasgivenrisetoaninflectional morphemeviaastageinwhich‘be’formedaperiphrasticauxiliarywithaformerparticiple(Chatterji,1926;Lahiri,2000).
Bengaliisparticularlyinterestingsincethereisnotonlya‘be’auxiliaryandaninflectionalmorphemethatisderived froma‘be’auxiliary,butalsoacliticversion.TheperfectinBengaliisexpressedwiththecliticversionof‘be’(31a),the progressivewithamorphemederivedfrom‘be’(31b).Asdiscussedintheremainderofthissection,theevidenceforthis analysisisintricate,butsolid.
Webeginwithacloseinvestigationofthesurfaceformsoftheprogressiveandtheperfectin(31).Thisrevealsseveral differences.Forone,therootvowelchangesintheperfect(theperfectofpɑ‘receive’in(30a)ispe(y)eʧh),butnotinthe progressive.Additionally,theprogressivesuffixhastwoallomorphs:/ʧh/and/ʧʧh/.Thechoiceofallomorphdependson theshapeoftheverbroot.Forinstance,intheCVC-verbpɑr‘topickfromahigherlevel,tobeableto’,the/ʧh/doesnot geminate,whilethereisgeminationwiththeCV-verbpɑ(pɑr-ʧh-evs.pɑ-ʧʧh-e).
Anotherdifferenceisbroughtoutbythedistributionoftheemphaticclitico‘also’(Fitzpatrick-Cole,1996;Lahiriand Fitzpatrick-Cole,1999).Asshownin(32),thiscliticcanbeintroducedbetweenthestemandthe-tʃhintheperfect,but notintheprogressive.Thesuffix/-e/intheperfectisatruesuffix.ThisisthemorphophonologicalremainderoftheOld
Indo-Aryan pastparticiple -ta (Chatterji, 1926:660).The clitic o cannotintrude between a stemand an affix.It can, however,intrudebetweenthestemandthe/ʧh/usedintheperfect.Thisindicatesthatthe‘be’intheperfecthasthestatus ofacliticizedauxiliary.In(30b),ontheotherhand,the‘be’hasbeenreanalyzedasapieceofverbalinflection which signifiestheprogressive.
(32) Perfect Progressive pa pe-e-t∫h-e=o pa-t∫t∫h-e=o
‘receive’ pe-e=o-t∫h-e *pa=o-t∫t∫h-e
*pe=o-e-t∫h-e
par per-e-t∫h-e=o par-t∫h-e=o
‘beable’ per-e=o-t∫h-e *par=o-t∫h-e
*per=o-e-t∫h-e
Wepositthereanalysisin(33)fortheprogressiveandtheperfect.Sincetheprogressivebehaveslikeasuffixrather thanaclitic,therootþsuffixcombinationplusthepersonsuffixesmustbeanalyzedasasingleprosodicwordinmodern Bengali.Thecombinationof‘be’plusaperfectstem,ontheotherhand, isanalyzedastwoseparateprosodicwords, whicharethenrealizedasoneviatheincorporationoftheclitic(e.g.,ZecandInkelas,1990;BayerandLahiri,1990;Lahiri, 2000).
(33) Progressive:
[RootþSuffixPART]vþ [beþSuffixPERS/NUM]v
>[[RootþSuffixPART]þ[beþSuffixPERS/NUM]CLITIC]v
>[[RootþSuffixPART þbeAFF]þ SuffixPERS/NUM]v
>[[RootþbeAFF]þ SuffixPERS]v Perfect:
[RootþSuffixPERF]vþ [beþSuffixPERS/NUM]v
>[[RootþSuffixPERF]þ[beþSuffixPERS]CLITIC]v
Asshownin(33),theprogressivewasoriginallyformedviaaperiphrasticconstructionwhichconsistedofanauxiliary andaparticiple(attestedinMiddleBengali).Theparticiplewasformedviaasuffixiwhosepreciseoriginisunclear.17The perfectwasalsoformedviaaperiphrasticconstruction,butwith/-iyɑ/,thedescendentoftheOldIndo-Aryanpastparticiple -ta.This/-iyɑ/coalescedtogivemodernBengali-e.Bothparticiplestookaformof‘be’whichinflectedforpersonand number.NumbermarkingwaslostinthecourseofhistoricalchangeforunrelatedreasonsandsomodernBengalionly displayspersonagreement.Theabstractschemain(33)canbe renderedmoretransparentbyconsideringanactual examplewiththerootpɑ‘receive’.
(34) Progressive:
[pɑþi]vþ[/ɑʧh/þ SuffixPERS/NUM]v
>[[pɑþi]þ [ʧhþSuffixPERS/NUM]CLITIC]v
>[[pɑþiþ ʧhAFF]þSuffixPERS/NUM]v
>[[pɑþʧʧhAFF]þe]v Perfect:
[pɑþiyɑ]v þ[/ɑʧh/þSuffixPERS/NUM]v
>[pɑþe]vþ [/ɑʧh/þSuffixPERS/NUM]v
>[[pɑþe]þ[ʧhþ e]CLITIC]v
Theobservantreaderwillnotethattheformsin(34)donotquitereflecttheactualformsfoundin(32).Inparticular,the reasonforthegeminationintheprogressiveisnotclearfromwhathasbeensaidsofar.UndertheassumptionofLexical Phonology(Kiparsky,1982b,1985),asummaryoftheanalysisneededtoaccountforthemodernsurfaceformsisshown in (35).Afulldiscussion canbefound inLahiri(2000).
17InLiteraryBengali,thisparticipleisformedwithite,adescendentoftheSanskritpresentparticipleaffix-ant.
(35) Perfect Progressive
Root pa par pa par
Deriv. pa+iya par+iya pa+i par+i
Morph.
Coalescence pa+e par+e Harmony
V pe-e per-e
Clitic
Aux. (pe-e t∫he) (per-e t∫he) (pa+i t∫he) (par+it∫he) Reanalyis [pa+i+t∫h]+e [par+i+t∫h]+e Med.i Loss [pa+t∫h]+e [par+t∫h]+e
Gemination pa+t∫t∫h —
Morph.
Infl. pat∫t∫h+e part∫h+e
Ascanbeseen,participlesarefirstformedviaderivationalmorphology./ɑ/goesto/e/inBengaliviavowelharmonyin thesynchronicgrammar.ThisaccountsforthesurfaceformoftheBengaliperfect.Theauxiliarycliticizestotheperfect andnothingmoreneedtobesaidaboutthat.Intheprogressive,however,theclitic(withouttheperson/numberending) wasreanalyzedasatense/aspectaffix.Thiswasfollowedbythelossoftheoriginalparticiplemarkeri.Thelossofthis medialifreedamorain theCV roots,whichledtogemination.In theCVCrootsno moraisfree,so nogemination occurs.18
Thus,Bengalishowsevidenceofprogressivereanalysiswithrespecttotheverb‘be’,whichcanbeusedasamain verb,alightverbandasanauxiliary.Butthereisacrucialdifferencebetweenthelightverbuseandtheauxiliary.Whilethe lightverbremainsform-identicaltothemainverb,theauxiliaryhascliticizedintheperfectandhasbeenreanalyzedasa pieceofinflectionalmorphologyintheprogressive.Thisreanalysishastakenplaceinthespaceofafewcenturieswhile the lightverbshaveremainedform-identicaltothemainverb formillenia.
3.3.2. Theverb‘go’
Inthissection,wepresentanothercaseinwhichreanalysisfrommainverbtoauxiliarytoinflectionalmorphemetook placeinthespaceofafewcenturies.Justaswith‘be’,theverb‘go’canbeusedasamainverb(36)andalightverb(37) inBengaliandUrdu.Additionally,itisusedasapassiveauxiliaryinUrdu(38)andgaverisetotoday’sfuturemorphology -g-(39).
(36) a. MainVerb ɑmi ge-l-am I.Nom go-Past-1
‘Iwent.’ (Bengali)
b. MainVerb m~e gɑ-yi I.Nom go-Perf.F.Sg
‘Iwent.’ (Urdu)
(37) a. LightVerb
bɑʧʧa por-e gæ-ʧh-e child.M.Nom fall-Gd go-Perf-Sg
‘Thechildfell(down)’ (Bengali)
b. LightVerb
bɑʧʧa gır gɑ-ya
child.M.Nom fall go-Perf.M.Sg
‘Thechildfell(down).’ (Urdu)
18NotethatthereisfurtherevidencethatthesynchronicprogressivemarkerinmodernBengaliisthegeminatedform/-ʧʧh/(Lahiri,2000),and thattheCVCrootsaresynchronicallyformedbydegemination.
