Download - The Philippine chalenge to Universal Grammar

Transcript

INSTITUT FR SPRACHWISSENSCHAFT UNIVERSITT ZU KLN ..... ARBEITS PAPIERNr.15(Neue Folge) THE PHILIPPINE CHALLENGE TO UNIVERSAL GRAMMAR Nikolaus P.Himmelmann Oktober 1991 ... ,.-._ .. . ".. HerausgeberderReihe:InstitutflirSprachwissenschaft derUniversittzuKln 0-5000Kln41 (c)beidenAutoren Acknowledgements ThispaperisbasedonmyunpublishedMAthesis(Mnchen 1983)andmyPhDthesis,writtenin1985andpublishedin 1987.IoweveryspecialthankstoJohnWolffwhocommented atIengthonthepublishedversionofthedissertationand encouragedmetoproceedalongthelinesdevelopedthere.I alsothankHans-JrgenSasseandFritzSerziskoforthe pertinentandhelpfulcommentsonthispaper.Last,butnot leastIhavetothankChrisSearlsandVictorFusilerofor brushingupmyEnglish. CommentsonthisworkingpaperversionarehighIywelcome! contents [ 1.Introduction 2.Tagalqgclausestructure 2.1.Preliminaries 2.2.angandtheproblemofsubjecthood 2.3.Thefunctionofang 3.Syntacticcategories 4.Orientation 4.1.Thenatureoforientationmarking PAGE 1 6 6 8 4.2.Thesemanticsoftheorientation f f i ~ e s15 16 25 27 34 4.3.Aspectjmoodinflectionandinherentorientation 36 4.4.Derivationandpredidtion41 5.Syntacticcategoriesandthegrammaticalization argumentstructure of 43 6.Theequationalhypothesisandtheorderof 7.Conclusion Abbreviations References constituents 49 52 55 56 - 1 -1_Introduction Grammaticalrelations- inparticulartherelation'subjectof' - andvoiceareofcentralconcerntoanytheoryofuniversal grammar.Withrespecttothesephenomenatheanalysisof Tagalog(andthePhilippinelanguagesingeneral)hasturned outtobeparticularlydifficultandcontinuestobeamatter ofdebate.Whattraditionallyhasbeencalledpassivevoicein theselanguages(forexamplebyBlake(1925),Bloomfield(1917) andWolfenden(1961appearstobesodifferentfromvoice phenomenainthemorefamiliarIndo-Europeanlanguagesthatthe term'focus'wasintroducedinthelate1950stounderscoreits 'exceptional'nature(cf.Llamzon(1973:168),MatsudaFrench (1988a.Thisterm,however,isamisnomersinceingeneral linguistics'focus'isusedtorefertothepragmatic phenomenonofhighlightingneworcontrastiveinformationand, asmostresearcherstodayagree,the'focus'affixesin Philippinelanguagesdonoth ~ v suchahighlightingfunction.1 ItisalsoquitegenerallyacceptedthatthePhilippine'focus' cannotbeanalyzedaspassive(cf.Shibatani(1988:89-96), DeWolf(1988:150-160),Foley(1991.Thecurrentdebate concernstheissueofwhetherthePhilippinelanguagesshould beconsideredergativelanguages(wherebytheconstruction traditionallycalledactiveistobeanalysedasantipassive). Thisproposalhasbeenmadewithintherespectiveframeworksof relational(Gerdts1988)andlexicasegrammar(deGuzman(1978, 1988),Starosta(1986,1991,asweIlasfromadiscourse-functionalperspective(cf.Payne(1982),Cooremanetal. (1984,1988.2TheergativeanalysisisrefutedbyShibatani (1988:96-115),DeWolf(1988:158-160)andFoley(1991)with basicallythesameargumentsthatmaybeadducedagainstthe passiveanalysis.Themajorpointisthis:whereastheactive andtheergativeconstructioncanbeshowntobetheunmarked constructionsinaccusativeandergativelanguages, 1Forlackofaconvenientalternative,however,mostauthorscontinueto usethetermfocus.ExceptionsareCumming(1986)andSchachter (1987:949ff)whousethetermtriggerinstead. 2Blake(1988)makesaprimarilymorphologicalargumentforergativityin Tagalog.Thiswillbecommenteduponinsect.4.3. - 2-respectively(bothintermsofmorphosyntacticpropertiesand discoursefunction3),thisisnotpossibleforthePhilippine 'active'and'passive'constructions(seesect.4). Furthermore,asalsopointedoutbyShibatani(1988:114)and DeWolf(1988:156f),aninflationaryusehasbeenmadeofthe term'ergative'inthelastdecade;itcanthusnolongerbe assumedthatit hasanunequivocalandspecificmeaningin typologizinglanguages,apartfromthetechnicaldefinitionit mightbegivenwithinaparticularframework.4 Butifthe Philippine'focus'constructionsareneitherpassivenor ergative,howelsecantheybeanalysed?Shibatani(1988)and DeWolf(1988),whobothrefutethepassiveasweilasthe ergativeanalysis,donotofferanalternativeproposal.In thispaperaeasewillbemadefortheclaimthat'focus' markingshouldbeanalysedintermsoforientation5,aconcept usedbyLehmann(1984:151f)forcapturingthedifference betweenEnglish(and,moregenerally,Indo-European)orientated nominalisationssuchasemploy- eroremploy-ee,and unorientatednominalisationssuchasemploy-ing.Thisapproach 3ForTagalog,thefigure sconcerningthediscoursedistributionof 'active'and'passive'(ar'ergative'and'antipassive')constructionsvary accordingtothetheoretica1positionoftheauthors,seeShibatani (1988:95f,111ff). 4Theproblemwasa1readypointedoutin1981byvanValin.Iprefer defining'ergative'asanominalcaseforrn,i.e.thecase - formoftheagent inatransitiveevent .Therei5noergativecase-forminPhilippine 1anguages(asopposedtootherAustronesianlanguagessuchasSamoanor Tongan)andthus ,inmyvi ew,theissueofergativitydoesnotarisefor these1anguages.Thisissueisfurtherdiscussedwithrespecttotheso-ca11edergative1anguagesofCentralandSouthSulawesiinHimmelmann (1991) . 5TheGermantermusedbyLehmannis'Ausrichtung'.Tomyknow1edge , 'orientation'isnotatechnica1termintheEng1ishliterature,a1though itoccasiona11yappearsindiscussionsofvoiceandergativity(cf.,for examp1e,Comrie(1981:69passim)andDeLancey(1982:167.Itis,however, atechnica1termintheUNITYPmodel( cf .intera1iaSeiler(1986)and Seiler& Premper(1991whereitisus edasacovertermforallphenomena re1atedtotheorientationordirectednessofeventexpress ions(Serzisko 1991),inc1udinginherentorientation(accusativevs .ergative1anguages) andmechanismsofreorientation(passive,antipassive,inverseinflexion, etc . ) .A1thoughPhi1ippine'focus'markingisalsocoveredbytheUNITYP usageoforientation ,notethatthistermisusedhereinmorespecific sense,i .e.itreferst oaderivationa1processthatisapplicabletoall kindsofexpr es s ions ,notjus teventexpress ions( sees ect.4.1). -3-impliesthat'foeus'markingisderivationalratherthan infleetionalasoftenpresumedintheliterature.Thisisto saythatwhatistypologieallyeonspieuousinTagalogisnot the'foeus'phenomenonperse,sineethisisverysimilarto orientatednominalisationsinmanyotherlanguages,butrather theveryprominentuseoforientatedformations(i.e., derivationalmorphology)inbasicelausestrueture. Beforepresentingtheanalysisoforientationaffixesin TagalogwewillbrieflysketchTagalogelausestrueturein seetion2whereit willbediseussedwhetherthereisa grammatiealrelation'subjeetof'inTagalogornot.The eontroversyregardingthistopiehasdevelopedalonglines similartothoseeoneerningthephenomenonof'voiee'in Philippinelanguages.Bloomfield(1917)andBlake(1925)use theterm'subjeet'withoutfurtherdiseussion.Thesame reasoningwhiehledtothereplaeementoftheeoneeptof 'voiee'bytheeoneeptof'foeus'alsoledtoreplaeing 'subjeet'by'topie,.6Theeurrenteontroversywasinitiatedby Sehaehter(1976),whoarguesthatthereisnosubjeetin Tagalogsincethepropertiesusuallyattributedtosubjects(as spelledoutinKeenan(1976aresharedbetweentwo partieipants(the'topic'andtheagent).7Foley& vanValin 6SeeMcKaughan(1973),whointhisartic1eexplicit1yreVlseshisposition andreturnstotheuseof'subject'again. 7Shibatani(1988:115 - 130)basical1yfol1owsSchachter'sana1ysis,but drawsadifferentconclusionsinceheoperateswithaprototypeapproachto subjects.Forhirn,actor-topicsareprototypica1subjects(exhibitingall ofthesubjectproperties),whi1enon-topicactorsandnon- actortopicsare non-prototypicalsubjects(exhibitingonlyapartofthesubject properties).Asaconsequenceofthisviewheisforcedtoanalysesome clausesascontainingtwo(non-prototypica1)subjects(1988:126ff).This, inmyview,restsonaseriousmisunderstandingoftheconceptofsubject, whichwasintendedtocapturethefactthatinmany1anguagesthereisone participantwhichismorecentraltotheexpressionofagivenstateof affairsthanothersandtheprimacyofwhichisref1ectedbothin morpho1ogica1markingandmorphosyntacticbehaviour(traditiona11y, 'subject'hasbeendefinedas'whatthesentenceisabout'or,more precisely,that'whichundergoesapredication';fordetails.seeSasse 1982,Foley& vanValin(1984:108ff.Bydefinition,therecanbenotwo primaryparticipants;i.e.iftherearetwoparticipantsthatpartia11y exhibitpropertiesofmorphosyntacticcentrality,neitherofthemisehe primaryorcentra1participant,thoughbothofthemmaybemorecentra1 thanotherparticipantsandthusbecoreparticipants(Fo1ey& vanValin - 4-(1984:134ff)considertheang-phrase(seesect.2)tobethe pragmaticpivotofaTagalogclause.Lateron(p143)thisis furthercharacterisedas'clauseinternaltopic'.Hoekstra (1986)andMcGinn(1988)workingwithinaGBframework- both considertheang-phrasetobethesubject;Carrier-Duncan (1985)andGerdts(1988),however,treattheagentasthe subjectinaltinstanceswithoutpresentinganexplicit argumentfortheiranalysis.InthispaperIprincipallyfollow thearguments . presentedbyDeWolf(1979:67-86,1988:144-150) thatasubjectrelationindeedexistsinTagalog .Themajor problem,t h o u ~ h doesnotseemtometobethedefinitionand identificationofthesubjectrelationinTagalog,butthe recognitionofthefactthatTagalogclausestructureis essentiallyequationalasarguedbyNaylor(1980)andDeWolf (loc.cit.).!nIndo-Europeanlanguagesverbalpredicates exhibitinherentmorphosyntacticrelationality.As ; aresult thereexisttwoaspectstothesubjectrelation:subjectas opposedtoobject(subject1 intermsofMatthews(1981:104ff)) andsubjectasopposedtopredicate(subject2).InTagalog, however,allpredicates(both'verbal'and'nominal')are morphosyntacticallynon-relational;thusonlyoneaspectofthe subjectrelationispresentinTagalog(thatis,theopposition betweensubjectandpredicate).Indeed,theessentially equationalnatureofclausestructureinTagaloghas repercussionsformanyaspectsofclausegrammar:ofcourse,no objectrelationexistsinTagalog.Itiscorrelatedwiththe prominenceoforientatedformationswhich- inasensetobe explainedinsections3and4- arethefunctionalequivalent ofargumentstructureinTagalog .8 Furthermore,it is (1984:77ff);seeDurie(1987)forananalysisofAcehnesealongthese lines).Thisisnotmeanttoimplythattheprototypeapproachmaynotbe usefullyappliedtosubjects.Asubjectlikeit init rainsiscertainlya 1essprototypicalsubjectthantheboyintheboyshothisgrandpa,butit stillexhibitsimportantpropertiesofEnglishsubjectssuchaspreverbal positionandtriggeringagreementonthepredicate.Tobeausefulconcept, however,evenanon -p rototypicalsubjecthastobemoresubject - likethan anythingelseinagivenclause. 8ThatargumentstructureinTagalogrequiresadistinctlydifferent treatmentisalsopointedoutbyFoley(1991)whooffersaproposalfor dealingwiththiswithintheframeworkofalexically-basedfunctional - 5-correlatedwiththefactthatthesyntacticcategoriesnounand verbaredifficult,ifnotimpossible,todistinguishin Tagalog(seesect_3)_ Thenatureofthedistinctionbetweennounsandverbswill thusbeofmajorconcerninthispaper_Itisimportanttokeep inmindthatthisdistinctionpertainstoatleasttwolevels: it pertains,ontheonehand,tolexicalsemanticclasses (wordsdenotingentitiesvs_wordsdenotingevents),andonthe other,tomajorsyntacticcategories,i.e.classesofwords exhibitingcommonmorphosyntacticproperties.9 Inthispaper thetermsnounandverbwillbereservedforthelatter,while lexicalsemanticclassesarereferredtobytermssuchas entity orevent.Entityisusedtoembracelexicalsemantic classesthataretypicallyexpressedbynounssuchaspersons, things,institutions,etc.,whereaseventisusedfortypically verbaldenotatasuchasstates,processes,andactions.In Tagalogitisdifficulttofindamorphosyntacticallyrelevant differencebetweeneventexpressionsandentityexpressions. Thus,thereisnoneedtomakeadistinctionbetweenthe syntacticcategoriesnounandverb.This,however,doesnot meanthatformaldifferencesdonotexistatall.Indeed,the mostpervasivedifferenceconsistsinthedifferentstress patterns10 foundincertainderivedformations,aquiteunusual phenomenonwhenseenfromatypologicalpointofview. unificationgrammar.Webrieflycomrnentonthisinsect.4.1. 9AsCroft(1991:37ff)pointsout,theformercanberegardedasthe ontologicalbasisforthelatterandisthususedbyhimasoneofthe externalparametersinhissupposedlyuniversaldefinitionofthemajor syntacticcategories(chapters2and3inCroft1991). 10StressinTagalogisunmarkedinstandardorthography.Itsanalysisis highlycontroversialsincesomeauthors(e.g.,Schachter& Otanes 1972:15-18)considervowellengththeprimaryphenomenon,whileothers considervowellengthanepiphenomenonofstress(cf.Bloomfield1917:141f; MatsudaFrench1988b:63f).Inthispaper,bothstressandvowellengthare indicated,but,sincestressassignmentinTagalogisnotfullyunderstood, allremarkspertainingtothisphenomenonmustberegardedwithcaution. NotethatmuchoftheliteratureonTagalogignoresstressaltogether. StressisphonemicinTagalog;comparebukas'tomorrow'andbukas'open', andplaysanimportantrleinaffixation.Primarystressonthepenultima willremainunmarked(thusbukasfor'tomorrow'),elsewhereitwillbe ma rkedbytheacuteaccent.Thegraveaccentmarksvowellength(along vowel,ofcourse,isalsoclearlystressed). -6-2_Tagalogclausestructure 2.1.Preliminaries Indiscussing ' Tagalogclausestructureit isusefultomakea distinctionbetweenfullwords.andfunctionwords(or particles;cf.Bloomfield1917:146).Functionwordsmaynotbe inflected;theymarkmorphosyntacticslotsorappearinclitic positions.Fullwordsmaybeextensivelyaffixedandoccurin thelimitedsetofmorphosyntacticslotsprovidedbythe functionwords.Therearesixmorphosyntacticslotsforfull wordsinaTagalogclause,fourofwhichcanbeillustratedby thefollowingexample : (1a)i-ni-ab6tng UGT-REAL(UG)-withinreachGEN mang-ga-gamot IRR_ACT-RDP2-medicine (Bl30/13) 11sasundaloangit16g, LOCsoldierREFegg 'Thephysicianhandedtheeggtothesoldier' Exceptforthefirstword,allfullwordsinthisclauseare precededbyfunctionwords(thefunctionmarkersang,ng,and sa).Theclauseinitialpositionisthepredicateposition, whichisunmarkedunlessanotherconstituentprecedes.Inthis case,the constituentismarkedbythepredicate marker(PM)ay,asin (1b)atangpareatsiyaynag-hinty andREF'priestand3 . SGPMREAL.ACT-wait ngsa-sabih-inngsundalo . '(loc.cit . ) GENRDP1-say-UGGENsoldier 'andthepriestandhewaitedforwhatthesoldierwould say. ' Fullwordsinpre-predicateposition,thefifthmorphosyntactic 11MostTagalogexamplesinthispaperarefromB1oomfield'sco11ectionof texts(=Bl) .Fortheseexamples,thepageandlinenumbersinBloomfield (1917)aregiven .AlthoughtheBloomfieldtextsareover70yearsoldand althoughtheyhavebeenproducedbyasinglenativespeaker,theystill representanadequateandreliabledatabaseforcontemporarystandard Tagalog(foranappreciation,seeWolff(1987)).Asfortheotherexamples. examp1e(20)isfromSchachter& Otanes(1972 : 163),examples(24)and(29) arefromTagalogletterstotheauthor,andtheremainingexampleshave beenelici ted . - 7-slot,maybeunmarkedortheymaybemarkedbythefunction markersangorsa_Fullwordsinpost-predicatepositionare alwaysprecededbyoneofthefunctionmarkers_Thesemarkers havethefollowingfunctions: saisagenerallocativepreposition(LOC)markingall kindsofobliqueparticipants.Itisthefinalconstituentof allofthemorespecificprepositionsinTagalog,suchas hanggangsa'until'orparasa'for'. ng[na9]marksgenitiveattributes(GEN) .Intheliterature itiscommontodifferentiatebetweenangmarkingnon-topic agents(e.g.ngmanggagamotin(1a)),onemarkingpatientsand themes,anotheroneforinstruments,adifferentoneformanner adverbials,andstillanotheroneforpossessors,etc.But,as convincinglyarguedbyNaylor(1980:37-42),ngitselfmarks nothingbuttheattributiverelation.Thisdoesnotmean,of course,thatthesemanticrolesmentionedcannotbe distinguishedineachcase(thelexicalsemanticclassofthe participantandtheorientationaffixofthepredicate expressionarethemostimportantcluesinthisrespect) .The claimissimplythatngisnotamulti-functionalcasemarker, andthefactthatneedstobeexplainediswhyparticipantsmay beconstructedasgenitiveattributesinTagalogirrespective oftheirsemanticrole. Beforediscussinganginthenextsecti o n,letusbriefly lookatthesixthmorphosyntacticslot,nots howninexample (1).Itisconstitutedbythelinker(orligature)na/_ng.12 Thisfunctionmarker'links',theelementsofamodifying constructionsuchasu161naungg6''fooli s hmonkey'butalso occursincompoundssuchaspuno-ng-saging'tree -LK-banana'and incomplementclauses(seeGonzales1971) .Theorderof modifyingconstructionsisnotfixedinTagalog;thusungg6'na u161isequallypossible.Thedifferencebetweenngandthe linkerpertainsto 'referentiality,i.e.ngi sacombinationof ang,markingreferentiality(seesect.2 . 3),andna(the 12-ng,asimplevelarnasal,istheallopmorphofthelinkerafterwords endinginavowel.Itisnottobeconfusedwiththegenitivemarker[na?] whichi sconventionallyspelledng. - 8-linker),markingattribution;comparethefollowing'minimal pair' : (2)a)bata-ngdalaga child-LKyoungwoman 'girl' b)bata'ngdalaga childGENyoungwoman 'childofth.eyoungwoman' Correspondingtothemarkersang,ngandsathereisa specialsetoffunctionmarkersforpropernouns,i.e.si,ni andkay,respecti vely.Similarly,thereareang-,ng- andsa-formsofthepersonalandthedeicticpronouns(cf.Schachter& Otanes(1972:88ff. 2 . 2.angand.theproblemofsubj ecthood angisgenerallycalledatopic(orsubject)markerinthe li terature.Asbrieflymentionedinsect.1,thediscussionhas mainlybeenconc.ernedwiththefactthatanon-agentang-phrase suchas-angit16gin(1 a)exhibitsonlypartofthesubj ect propertiesdi splayedbysubj ectsinIndo -Europ.eanlanguages. Thisisnosurprisegiventhefactthattheprototypical subjectinmodern'3Indo-Europeanlanguagesisacombinationof thepragmaticroletopicandthesemanticroleagent.Thereis nodoubtaboutthefactthattheang-phraseexhibitsmostof thetopic-relatedproperttesofsubjects.Thefactshavebeen , widelydiscussedintheliteratureandneednotberepeated here(seethereferencesinsect.1).Ofparticularimportance inthepresentcontextisthefactthattheorientationalaffix onthepredicate(i- in(lasignalsthesemanticroleofthe participantdenotedbytheang-phrase.Thismeansthatthereis aspecialrelation,apredicativerelation,betweentheang-phraseandthepredicate(asdistinctfromng- andsa-phrases). Inotherwords,theang-phrasedenotestheparticipant'the sentenceisabout',whichisthetraditionaldefinitionfor subjects(butalsofortopics,whichisamajorcauseforthe 13SeeSasse(1982)forasketchofthedevelopmentofthesubjectrelation inIndo -Europ.eanlanguages. - 9 -confusionsurroundingtheseterms}.Moreprecisely,this definitioncharacterizessubjectsasopposedtopredicates (subject2 intheterminologyofMatthews1981:104-113).I proposetousethetermpredicationbasetodenotethisaspect ofsubjecthoodsincetopic,whichhasalsobeenusedinthe sameway,hassomanyotherdifferentmeanings.Thuswemaysay thatthereisasubjectrelationinTagaloginthesensethat theang-phraserepresentsthepredicationbaseinclausessuch as(1a). Theinterrelationbetweentopicsandpredicationbasesis highlycomplexandcannotbespelledouthere.Clause-leve114 topicsrepresentpresupposedinformationandbasicallyhavea scene-settingfunction.Thisscene-settingfunctionmaybe furthersubdividedintothescene-settingfunctionproper(Ona lovelymorninginApril1965Clairesetoutto... ),the explicitrepresentationofwhatisunderdiscussion(asfor topics,Idonotbelievethatwewilleverbeabletocomeup withasatisfactorydefinition),andtherepresentationofthe discoursetopic(usuallybypronominalisationorzero anaphora).Althoughtopicsinthelattertwosensesoften coincidewithpredicationbases(seeforexampleang manggagamotatsiyain(1b}),thisisnotnecessarilythecase (foraTagalogexample,see(9)below).Indeed,topicsand predicationbasesarereltatedinsofaraspredicationbasesare grammaticalizedtopics(ofdiscussion),whichisalsothe reasonwhyidenticaldefinitionshavebeenproposedforbothof them.Inotherwords,predicationbasesaretopicswhichhave beentightlyintegretatedintocore-levelclausestructure(as evidencedbythewell-knownsubjectpropertiessuchas agreement,obligatoriness,etc.),andintheprocesstheyhave lostthepragmaticfunctionofscene-settingtovarying degrees.Toexemplifythelossofthepragmaticfunctionnote thatit wouldbenonsensetoattributet ~ functionofsetting 14Ihavenothingtosayhereaboutdiscourse-topicalityasdefinedin Giv6n(1983)_AsshownbyCooremanetal.(1984)ang-phrasesinTagalogdo notrepresentdiscoursetopics.Foley& vanValin(1984:143ff)consider relative-clauseconstructions'thecrucialnexus'intheinterrelation betweentopicandpredicationbase(=pivotintheirterminology). - 1 0-thesceneorprovidingthetopicofdiscussiontothemost grarnrnaticalizedpredicationbases,i_e.dummysubjectssuchas it init rains.InTagalogtheneedtodifferentiatebetween topicsandpredicationbasesisrelatedtothefactthatboth mayco-occurinthesameclause(seebelow). Intheprecedingdiscussionwehaveindiscriminatelydealt withtheang-phraseasapredicationbase,therebyimplying thatangmarkspredicationbases.This,however,isnot correct.Thefactthatangbyitselfmarksneitherpredication basesnortopicsisevidentfromclausescontainingtwoang-phrases: I (3)angmgabuh6klamangangp-in-u-putolngpata+im REFPLhaironlyREFRDP1-REAL(UG)-cutGENblade 'onlythehairwascutbytheblade'(Bl58/36) Infact,oneofthetwoang-phrasesappearinginthisclause hastobethepredicate.SinceTagalogpredicatesareusually inclause~ n i t i l positionandsincenoevidencecanbe presentedtothecontrary,angmgabuh6klamangmustbe analysedasthepredicateexpressioninthisclause.Amore literaltranslationwouldbe'thatbeingcutbytheblade(was) onlythehair'.Thus,predicateandpredicationbasemustbe definedconfigurationallyinTagalog:thepredicateexpression appearsinclauseinitialposition(orimmediatelyfollowing ay),whereasthepredicationbaseistheang-phrasefollowing thepredicateexpression.Topiesarealsodefinedbyposition, i.e .theyappearinpre-predicateposition,separatedfromthe predicatebyay.Topicsmaybeofvariedsemanticandformal make-up:wefind,forexample,unmarkedtemporalexpressions (4),sa-phrasesindifferingfunctions(5),ang-phrases(1b), complexconjunctions(6),gerundialconstructions(7)and eompleteclauses(8): (4)isa-ngkatanghali'anayma-tahimikangbayan one-LKnoonPMIRR.STAT-quietREFtown 'Onenoonthetownwasquiet.'(B1120/37) - 11-(5),sa106bngbaDngniy6gay LOCinsideGENshellGENcoconutPM i-ni-la-lagayangisa-ngpirasu-nglaman UGT-REAL(UG)-RDP1-positionREFone-LKpiece-LKmeat ngniy6g(Bl120/108) GENcoconut 'Insidethecoconutshellisplacedapieceofthemeatof thecoconut, (6)dahilditoayt-in-awagniyaangpare' causeDEM.LOCPMREAL(UG)-call3.SG.GENREFpriest 'Thereforehecalledthepriest... '(B164/25) (7)pag-ka-wika'niyanit6ayb-um-aliksiya GER-??-language3.SG.GENDEM. GENPMREAL.ACT-return3.SG 'Havingsaidthishewentback , (Bl20/4Q)... ( 8 )nangd-um-atingsilasaisa-nggubatay whenREAL.ACT-arrive3.PLLOCone-LKforestPM in-iwansi langkanila-ngmagulang(Bl32/25) REAL(UG)-abandon3.PLGEN3.PL.DAT-LKparent 'Whentheycametoajungle,theirparentsleftthem, Thus,topicsareneitherlimitedtoang-phrasesnoristhe presenceofangsufficienttoidentifythepredicationbase. Thattopicandpredicationbasemustbedistinguishedin Tagalogandthatneitherissolelyidentifiedbyangisshown bythefollowingclause: (9)kanyaangmgabuntisnababayesabayangya6nay thereforeREFPLpregnantLKwomanLOCtown:LKDEMPM lalo-ngma-lakiangtakot(Bl36/16) surpassing-LKIRR.STAT-bigREFfear 'Thereforethepregnantwomeninthetownhadgreatfear.' Here,thefirstang-phraseinpre-predicateposition(angmga buntisnababayesabayangya6n)denotesthetopic,thesecond ang-phrase(angtakot) ,whichfollowsthepredicate,functions asthepredicationbase.Amoreliteraltranslationofthis clauseis'therefore,thepregnantwomeninthetown,very great(was)thefear'. Giventhisstateofaffairsthequestionarisesastohow clausessuchas(1b),whereanang-phraseoccursintopic positionandnoang-phrasefollowsthepredicateexpression, - 1 2-shouldbeanalysed.Isangmanggagamotatsiyathetopie,the invertedpredieationbase(asit iseommonlyassumedinthe literature,viz.thetermay-inversion(Sehaehter& Otanes , (1972:485ff,orboth?Therearetwofactswhiehstrongly suggestthatit issimplythetopie:First,inelausessuchas (6)and(8) ~ inversionanalysiswouldbehighlyartifieial sineetheeorrespondingnon-invertedeonstructionsdonot oeeur.In(4)and(7)thenon-invertedconstructionsare possiblealthoughextremelyrare.Thisstronglysuggeststhat thetopicpositionshouldbeeonsideredanoptionalconstituent ofbasicelausestructureinTagalogratherthantheoutputof someBortofmovementrule.Seeond,thepredicationbaseisnot , anobligatorypartofaTagalogclausebutmayremain unexpressedifit isreGoverablefromthecontext:15 (10)d-um-atingangaswang,um-akyatsaisa REAL.ACT-arriveREFvampireREAL.ACT-climbLOCone puno-ng-suha',atna-rinigniya-ng tree-LK-grapefruitandREAL.STAT-hear3.SG.GEN-LK p-um-itasngmarami-ngbunga(BI36/34) REAL.ACT-pickGENmany-LKfruit ' thevampirecameandclimbedonagrape-fruittree,and he[i.e.,apoliceman]heardit pickingmanyfruits.' Here,boththepredieatesumakyatandpumitas(moreprecisely, thisisacomplexpredicatenarinignapumi tas'beheard picking')lacktheexpressionofapredicationbase.The , vampire(angaswang) ,whichtheisthe'subject'ofthese predications,isnotanaphorieallyreferredtobyapronounin theseelauses.Anotherexample,involvinganentityexpression asapredicate,is: 15Ofcourse,therearec1ausetypesinTagalogwhichneverinve1ve predicationbases,forexample ,existentialclausessuchasmay-roonding ila - ngbahaynatabla'therewerealsosemeframehouses'(Bl34/37).Here weareonlyconcernedwithclauseswhereapredicationcanbeexpected . - 13-(11)anghitsurangkulamayhindi' REFappearanceGENPMNEG pare-pareho. RDP3-same Kung if minsanayisa-ngbat60isa-ngmanika(Bl40/13) oncePMone-LKstone . orone-LKdoll 'Theappearanceofthekulamisnotalwaysthesame. Sometimes(it)isastoneorasmalldoll... ' ThiskindofzeroanaphoraisquitecommoninTagalogtextsand providesindependentevidencefortheclaimthatthe predicationbaseinaTagalogclausemayoptionallyremain unexpressed.Giventhisfactandthefactthattopicsshould notbeanalysedasinvertedorextractedconstituents,themost simpleandeconomicalanalysisofclausessuchas(1b)liesin assumingthattheycontainatopicconstituentandthattheco- predicationbaseremainsunexpressed.16 Letussummarize: therearebothtopicsandpredicationbasesinTagalog clauses .Bothhavetobeconfigurationallydefined(in termsoforderingbe foreandafterthepredicate). neithertopicsnorpredicationbasesareobligatorily expressedineveryclause. thereisnoone-to-onecorrespondencebetween morphosyntacticfunctionandmorphologicalmarking .Though predicationbasesareobligatorilymarkedbyang,ang-phrasesmayaisofunctionastopics(1b,9)andpredicates (3) . Theanalysismaybeillustratedmoreschematicallywiththe helpofthefollowingformulas.17 Optionalconstituentsare giveninparentheses: 16McGinn(1988:278)proposesasomewhatsimilaranalysiswithintheGB frameworkwhichalsoinvolvesellipsis.Inhisanalysis,however,the functionsareassignedexactlyoppositetothewayitisdonehere(andhas beendoneinmostprecedinganalyses).ang-phrasesprecedingthepredicate areconsideredproperlygovernedsubjects(aybeingthegovernor),andang-phrasesfollowingthepredicateareconsideredinvertedsUbjects .Asfaras Icansee,examplesinvolvingtwoang-phrases(suchas(9cannotbe adequatelydealtwithinthisanalysis. 17Theseformulasarenottobemisunderstoodasaproposalconcerningthe formalizationofTagalogclausestructure! - 1 4-I.Tagalogclausestructure (TOPICay)PREDICATE(PREDICATIONBASE)(OBLIQUE) Theinternalstructureofeachoftheseconstituentsis basicallythesame.Inmostcasesthereisanobligatory functionmarker,followedbyafullword18 whichmayb ~modifiedfurtherbynon-referentialmodifiers(whichare linked),genitivesand/orobliques. II.Tagalogphrasestructure FUNCTIONMARKERFULLWORD(MODIFIER(S(GENITIVE(S(OBLIQUE(S Thisstructureofthephrase,ofcourse,alsoappliesto genitivesandobliques.Example(1a)thusconsistsofa predicatephraseandapredicaMonbase.Thepredicatephrase inturncons,istsofafullword t h ~ eventexpressioniniab6t) , agenitive(ngmanggagamot)andanoblique(sasundalo).There isnofunct {onmarkerforthispredicatephrase.The predicationbaseconsistsofanobligatoryfunctionmarker (ang)andafullword(i t16g) Thefollowingcorrespondencesholdbetweenmorphosyntactic functionsandfunctionmarkers(inthecaseseveralmarkersmay , beusedinagivenmorphosyntacticfunction,theyarelistedin theorderoft 'he'irrelativefrequency): III.Morphosyntacticfunctionsandfunctionmarkers: TOPIC: PREDICATE: PREDICATIONBASE : OBLIQUE: GENITIVE: MODIFIER: sa,ang,(Iif) ld',ang,sa ang sa ng na/-ng 18Theidentificationanddefinitionofphrasa1headsisextreme1y difficultinmOdifying(1inked)constructions.SinceIdonotwanttoe'nter intoalengthydiscussionofthisissueatthispoint,Irefrainfrom identifyingaphrasalheadinthisformula.Inthepresentanalysis,there is00needforadistinctionbetweennounphrasesandverbphrasesin Tagalogwhichisthereasonthetermfullwordisusedintheformula ratherthanthemorecommontermsnounorverb. - 15-2.3.Thefunctionofang Thefollowingtableliststhefunctionsoftheang-phrasesin theBloomfieldtexts: IV.FunctionsoEang-phrases>nBloomEieldtexts TOPlC PREDlCATlONBASE

REST TOTAL 458 693 31 37 1219 Thesenumbersclearlysupporttheclaimmadeabovethatang doesnotmarkamorphosyntacticfunction.Giventhisfact,the questionarisesastowhatthefunctionofangisoInthe literatureit isoftenclaimedthatangphrasesarenecessarily definite.This,however,isnottrue,asshownbythefollowing example: (12)a)do6nayna-kitanilangisa-ng DEM.LOePMREAL.STAT-see3.PL.GENREFone-LK ma-Iaki-nghigante(BI32/31) IRR.STAT-big-LKgiant 'Theretheysawagreatgiant Infact,ang-phrasesarealwaysreferential,buttheymay involveallkindsofreferentiality(definite,indefinite, generic:fordetailsseeAdams& Manaster-Ramer1988).Thus, notdefinitenessbutreferentialityistherelevantparameter here.Thisclaimcanbefurthersupportedbylookingbackto example(3)whereanang-phrasefunctionsasthepredicate.If angmarks thisclaus6shouldinvolvetwo referentialexpressions,andthepredicationshouldbeacase ofanidentificational20 predication(thethe-murderer-is-the-19Thisincludescaseswhereang'phrasesappearintitles(21times)which maybeconsideredaspecialcaseofthetopicfunction .Furthermore,it includestheuseofangintermsofaddresssuchasangiyopO'ngkamahalan 'YourMajesty'(BI26/23)andafewmorecomplexconstructions,cf.,for example,BI28/10or56/19andBI46/37. 20Theterminologyforclausesinvolvingnominaloradjectivalpredicates (e.g.,Johnisateacher,Thewallisgreen)isquitevariedandthus highlyconfusing(seeDeclerk(1988:lff)forabriefsurvey).Following Schachter& Otanes'(1972:61)usageforTagalog ,Iwilluseequational clause(ratherthancopularornominalclause)asacoverterm.Onemajor subgroupofequationalclausesareclausesassertingtheidentitybetween , . - 1 6-gardenertype)_Furthermore,itshouldbepossibletodropthe angofthepredicateexpression,andtheresultingclause shouldinvolveacharacterizationalpredication(thegardener isa murderer)_Thisis,infact,thecase:mgabuh6klamang angpinuputolngpatalimmeans'thatbeingcutbytheblade (was)onlyhair'_Whilein(3)angmgabuh6kreferstothe specificha.irofoneoftheprotagonistsofthestory,mga buh6kintheprecedingclausespecifiesthekindofobjectthat isbeingcut.Allinstancesofang-phrasesfunctioningas predicatesknowntomecanbeexplainedwiththehelpofthis analysis(for. furtherdiscussion,seedeGuzman1982).Tosum up,wemaystatethatthefunctionofangisbasicallysimilar tothatofanarticle(asassumedbytraditionalgrammar,cf. Blake1925: 205f).Thisisalso- shownbythefactthatangmay bemissingfromphrasesinwhichtheang-formofoneofthe demonstrative'pronounsprecedesotherfullwordsasin: (12)b)kungiyo-ngda-dalh-inito-ngsupotko if2.SG. DAT-LKRDP1-carry-UGDEM-LKbag1.SG.GEN ngkuwaltasaakingasawa(Bl34/4) GENmoneyLOe1.SG.DAT:LKspouse 'Ifyouwillcarrythisbagofmoneyofminetomy wife' Strictlyspeaking,however,i tisnotadefinitearticle,but, moregenerally ,areferencemarker(REF).Asopposedtongand sa,angthenisnotacasemarker. 3_Syntacticcategories Ithaslongbeenrecognisedthatthedistinctionofmajor syntacticcategories(orpartsofspeech)inPhilippine languagesisdifferentfromthatfoundinotherlanguages.The majorconcernhasbeenwiththedistinctionbetweennounsand tworeferentialexpressions.The searecalledidentificationalclauses(de Guzman1982andStarostaetal.1982 : 150usethesameterm,Declerkuses specificational).Theothermajorsubgroupofequationalclausesinwhich thereferentofthesubjectexpressionisfurthercharacterizedbythe predicateexpression(byascribingsomepropertytoitorspecifyingits classmembership')arecalledcharacterizational(Starostaetal.' s descriptive,Declerk'spredicational)clauses.Clausesinvolvingverbal predicates(callednarrationalbySchachter& Otanes(loc .cit.are simp1ytermedverbalclauses. - 1 7-verbs,21andtheclaimhasoftenbeenmadethatPhilippine verbsareactuallynouns .Traditionally,thisclaimhasbeen restrictedtoundergoer-orientatedeventexpressions(the various'passives') .Themajorevidenceadducedwasthefact thatintheseconstructionstheexpressionfortheagentis identicaltothatofapossessor(i.e.it appearsinang-phrase,seeexamples(1a),(6),(8)}. 22Schachter&Otanes (1972:62)claimthat'Tagalogverbsandverbphrasesare... muchmorenoun-likethantheirEnglishcounterparts' .This claimreferstoallkindsof'verbs'andisbasedon distributionalevidence :'thereisvirtuallynocontextin whichanounoccursinwhichit cannotbereplacedbyaverbor verbphrase'.Theyneverthelessdistinguishbetweennounsand verbsintheirTagaloggrammar,primarilybecause,intheir view,verbsareinflectedforaspect(1972:65).Starostaetal. (1982:146f)brieflydiscussandrefutesomeofthearguments againstthesupposedlynominalcharacterofPhilippine'verbs'. Intheirview,Philippine'verbs'historicallywerenounswhich werelaterreinterpretedasverbsbutwhichretainedastrong nominalcharacter(1982:158ff).Themainevidenceforsucha reinterpretationisthepositionofthesubject(predication base)wewillsaymoreaboutthisinsect.6.Themost radicalstandinthisdebateistakenbyLemarechal(1982)who claimsthatnodistinctionofmajorsyntacticcategoriesin Tagalogispossible(andnecessary);therearejustfullwords (whichhecallssuperpartiedudiscours).Again,hisevidence isdistributional,i.e .everyfullwordcanoccurinevery majormorphosyntacticfunction. 21Inthispaperwewillbeconcernedwiththedistinctionbetweenonly nounsandverbs.Butregardingtheothermajorsyntacticcategory , adjectives(andadverbs),thesameclaimcanbemadeandsubstantiated:the evidenceforsuchacategoryismarginal,thereisnone edtomakeuseof itinanalyzingTagalogmorphosyntax. 22 ,The ' mostelaborateaccountofthisclaimisCapell(1964)wherethe earlierliteratureisbrieflyreviewed.Notethattheclaimhasalong traditioninAustronesianstudies,particularlyinearlyDutchworkon WesternAustronesianlanguages.Forabriefdiscussionandreferences,see Milner(1980). - 18-ThereisnoneedtoreviewallofLemarechal'sevidence here.Thefollowingobservationsmaysufficetosupporthis (and, ' somewhatmorecautiouslyphrased,Schachter& otanes') claimwithrespecttothedistributionoffullwords.Inthe precedingsections,sixbasicmorphosyntacticfunctionsina Tagalogclausewereidentified.Insection2.2thefunctions predicate,predicationbase,andtopicweredefined.Insection 2.1it wasbrieflymentionedthattherearethreefurtherbasic functions :genitive(orreferential)attributes(markedbyng), non-referentialmod'ifiers(markedbythethelinkernaj-ng) , andobliquecomplements(markedbysaandvariousmorecomplex prepositions.allof ' whichinvolvesa).Thefactthatentity expressions- words "whichpotentiallymaybeclassifiedas nouns- fnrtherderivation)inallsixfunctions isevidentfrom presentedthusfar(example(2) involvesthelinker"(11)exemplifiesthepredicatefunction). r The.factthatthere-.isnocopulainTagalogisofmajor importanceinthissincethepresenceofacopula distinguishesverbalfromnominalpredicatesinmany languages .23 In theevidenceforeventexpressionswe willlimitthediscussiontothesupposedlyhardestcase,that ..a...-- '*-.-is,aspectuallyinflectedformsappearinginpresumablynominal functions(forexample,asobliquecomplements;recallthat aspectualinflectionhasbeenproposedasadefiningfeature forverbsbySchachter& otanes).WithSchachter& otanes' (1972 : 66f)aspectterminologyinmind,notethatin(3)an IMPERFECTIVEform(pinuputol)appearsasthepredicationbase;in (1b)acontemplatedform(sasabihin)appearsasagenitive attribute .InthefollowingexamplewefindanIMPERFECTIVEform (pinatutunguhan)asalocativecomplement: 23Notethatthereisalsonodistinctionwithregardtonegation,ascan befound,forexample,inIndonesianwherebukanmayservetodistinguish nominalfromverbal(andadjectival)predicates(negatedbytidak). - 19-(13)na-ka-tanawsiyangbahaynama-ilaw REAL.STAT-??-visible3 . SGGENhouseLKIRR.STAT-light sap-in-a-tu-tunguh-anngkalabaw LOCCAUS-REAL(UG)-RDP1-gotowards-UGi GENwaterbuffalo 'hesawalightedhouseinthedirectiontowardwhichthe caribouwasgoing.'(Bl72/6) Althoughexamplessuchas(13)arerare,aspectuallyinflected formsinang-phrases(functioningaspredicationbasesoras topics)and,toalesserextent,ng-phrasesareverycommon. Intheliteratureit isoftenimpliedthatthefunction markersang,ng,andsahaveanominalisingfunctioninthe examplesinvolving(inflected)eventexpressions(cf.,for example,Schachter& Otanes(1972:150ff.Butthisisanad hocexplanationforwhichnoevidencecanbeadduced.Thereis nodifferencebetweenang-,ng- orsa-phrasescontaining (uninflected)entityexpressionsandthosecontainingevent expressions.Rather,thedistributionalfactsjustmentioned suggestthefollowinginterpretation:thecross-linguistically well-supportedobservationthatentityexpressionsinthe unmarkedcasehaveareferentialfunctionwhileevent expressionsintheunmarkedcaseareusedaspredicatesdoes notholdinTagalag.24 Fullwordsdonotdifferinmarkedness withrespecttomorphosyntacticfunction.Everyfullwordthat servesareferentialfunctionhastobemarkedbyoneofthe prepositivemarkersang,ngorsa.Thus,thesethreefunction markershave,notanominalising,butareferentializing function;inturneveryTagalogfullword,includingentity expressions,mustbemarkedbyoneofthesethreemarkersin ordertoserveasreferentialexpressions.Similarly,every fullwordmaybeusedwithoutfurthermarkingasa (characterizational)predicate.Therefore,Tagalogfullwords areneithertrulynounsnortrulyverbs.Theyseemtobe precategorial,asclaimedpyFoley(1991:5f).Helimitsthis 24Thepresentargumentisphrasedintheframeworkandtermin010gyof Croft(1991chapters2and3,cf.especia11yp#62ff;notethatinsteadof entityandeventheusesobjectandaction);theargumentdoesnot, however,dependonthisframeworkbutmaybeexpressedwi thinanyframework thatpropos esacross-1inguisticdefinitionofnounsandverbs. -20-claim,however,tobaseforms,whilehe retheclaimismadefor allfullwords,includingorientatedones.Notethatthis phrasingismorepreciseandlesspronetomisinterpretations thansuchsweepingstatementsas'allverbsinlanguageXare actuallynouns'.AspointedoutbyWalter(1981)andHopper& Thompson(1984),nounandverbarecorrelativeconcepts;the degreetowhichtheyaredistinguishedassuchmaybeplotted onacontinuumrangingfrompre- (ora-)categorialitytofully implementednoun/verbhood.Toputit differently,wherethere arenonounstherecanalsobenoverbs.Totalkaboutthe 'nominal'characterofverbsinagivenlanguageisa potentiallymisleadingshortcuttosayingthatexpressionsthat aresemanticallysimilartoverbsinIndo-Europeanlanguages displaymorphosyntacticproperties -cornrnontoIndo-European nouns.Whatisnecessaryinthissituationisnotthe impressionisticlabel'nominalcharacter'butacareful investigationofcategoriality. What,then,isimpliedbytheclaimthatTagalogfullwords areprecategorial?Basically,therearetwopropertiesofnouns andverbswhichTagalogfullwordslack:ontheonehand,they lacktheinherentreferentialitycharacteristicofnouns,hence thepervasiveuseofmarkersforreferentialitywhichisa majorcharacteristicofTagalogclauses.Ontheotherhand, theylackaprop.ertyinherenttoverbs,onehoweverwhichis somewllatmoredifficul ttodetermine.Themaj orfunctionof verbsistopredicate(cf.Croftloc.cit.).Tagalogfullwords maybeusedinthisfunctionwithoutrequiringfurther function-markingmorphologycharacterizational)predicatesin Tagalogareunmarked);thusit maybehypothesizedthatTagalog fullwordsareinherentlypredicative.Theyareunlikeverbal predicates,however,inthattheyseemtolackmorphosyntactic relationality(orvalency).Thus,Tagalogfullwords,including eV'entexpressions,donothaveaninherentargumentstructure (seeagainFoley(loc.cit)).Thishypothesisissupportedby thefOllowingfacts: - 21-'arguments'maybefreelyomittedinTagalog .25 Any Tagalogtextwillprovidenumerousexamplesforthisclaim. assumingTagalogeventexpressionswereinherently relational,it wouldbedifficulttoexplainwhytheymay occurunalteredinreferentialfunction.Englishverbal predicatessuchascuts,isjwasbeingcut,etc.maynotbe usedasnominals(*ajtheisbeingcut)whilethisis perfectlypossiblewiththecorrespondingTagalog expressions(cf.(3)above). Foley(1991:8f)notesthattheTagalogorientation morphologydoesnotallowdifferentiationofargumentsfrom non-arguments.Instead,orientationmaytakeplacewith regardtopracticallyanyargumentrole.Ifarguments cannotbedistinguishedfromnon-arguments,theassumption ofaninherentargumentstructuremakesnosense. thishypothesisprovidesanaturalexplanationforthe factthatallargumentsofaTagalogeventexpression (apartfromthepredicationbase)aremarkedaseither genitiveattributes(ng)orobliquecomplements(sa).In thisregardtheyresemblecertainEnglishnominalisations (viz.thegivingofanappletoMarybyPeterinmy backyard). NotethatthehypothesisthatTagalogeventexpressionslack inherentmorphosyntacticrelationalitydoesnotmeanthatthey arenotsemanticallyrelational.Ofcourse,alexemesuchas bigay'give'semanticallyevokesaframewhichpossibly includesanagentwhodoesthegiving,thethinggiven,the persontowhomsomethingisgiven,theplacewherethegiving happens,theinstrumentwithwhichsomethingisgiven,etc.-justasitsEnglishequivalent.ButtheEnglishequivalentis differentinthatsomeofthe ' argumentsarenotmerely semanticallyevoked;rather ctheslotsforagent,themeandgoal arepartoftheinherentargumentstructureoftheEnglish lexeme,whichisevidentfromthefactthatintheunmarked 25cf,McGinn(1988:276)whoalsopointsoutthatinthisrespect,Tagalog resemblesJapaneseratherthanapro-droplanguageoftheItaliantype (p291,endnote1) , -22-casetheyareobligatorilyexpressedandthattheyare expressedascorearguments(subject,directandindirect object)_Thedifferencebetweeninherentmorphosyntactic relationalityandsemanticrelationalitymayalsobe illustratedbyEnglishnominalisations:Asopposedtotheverb 'give'theactionnominalisation'giving'doesnotobligatorily requiretheexpressionofanargumentthough,ofcourse,it is possibletoexpressthem.Iftheyareexpressed;theyare expressedintheformofadnominalmodifiers(genitives)or prepositional.complements(Peter' Sgivingofanappleto Mary).26Thus"theframewhichissemanticallyevokedbythe verb'give'andthenominalisation'giving'isthesame,but thewayinwhichthesemanticallyevokedargumentsarerealized morphosyntacticallyisdistinctly-different.Nominalisationis, ,. ofcourse,acomplexissue(seeComrie& Thompson(1985)fora cross-linguisticreviewofthephenomenaandSpencer(1991 chapters6-8)forareviewofthetheoreticalissuesinvolved) . Noteinparticularthatwearenotconcernedherewiththe intricaciesofargumentinheritance.Whatisofinteresthere isonlythefactthatthemorphosyntacticstatusofthe argumentsofverbsisdifferentfromthemorphosyntacticstatus ofthe'arguments'ofnominalisations,andthatTagalogfull wordsmorecloselyresemblethelatterthantheformerinthis regard. Sineethedistinctionbetweeninherentmorphosyntactic relationalityandsemanticrelationality27willbeimportant throughoutthispaperletusintroducethefollowing 26Forthesakeofourargumentwewillabstractfromthefactthat n g ~ i s hgerundformationsalsoallowfora'verbalway'ofexpressingthearguments (PetergivinganappletoMary). 27Thisdistinctionissimilartothedistinctionbetweenargument structureandlexicalsemantic(orconceptual)structureproposedin variousgenerativeframeworks(cf.,amongothers,Jackendoff(1990), Grimshaw(1990),Lefebvre(1991:44f),Spencer(199l:342f.However,the conceptualstructuresintheseframeworksarefairlyclosetoargument structureandtheyinvolvecompositionalrepresentationsofthemeanings denotedbyaneventexpression .Semanticrelationality hereistobe understoodinabroadersense,i . e.astheeventframeorsceneevokedby aneventexpreBsionwhichincludesnotonlythosepart"icipantsa compositionalrepresentationprovidesslotsfor,butinsteadincludesany imaginableparticipantandtheplacewhereaneventhappens. -23-terminologicalconventions:Thetermargumentwillbelimited toargumentsthatfillslotsprovidedbyamorphosyntactically relationalexpression(i.e.thataregovernedbyaverb/are partoftheinherentargumentstructure) .'Arguments'thatare merelyevokedsemanticallyarecalledparticipants.Theconcept ofsemanticevocationneedstobefurtherrefined,since,as willbeseenbelow,notallparticipantsinTagalogdisplay exactlythesamemorphosyntacticproperties.Thatis,the expressionofparticipantsasgenitiveattributesorobliques isnotarbitrarybutislinkedtothesemanticrolestheyplay inthee ventexpressed(thesameholdsforparticipantsin actionnominalisations) .Asaconsequnceofthis,itis possibleinTagalogtodifferentiatebetweencentraland peripheral(orcoreandnon-core)participants.Thisinturn meansthatthedistinctionbetweenmorphosyntacticandsemantic relationalityreferredtointheprecedingparagraphsisnot absoluteandthatthereisnosuchthingaspuresemantic relationality.Putdifferently,thedifferencebetweenan EnglishverbandaTagalogeventexpressionis,strictly speaking,notthefactthattheformerismorphosyntactically relationalandthelatternot;rather,argumentstructureis grammaticalizedinEnglishtoahigherdegreethaninTagalog. Wewillreturntothispointinsections5and6 .Forthetime being,however,wewillcontinuetorefertotheoversimplified two-waydistinctionmorphosyntacticvs.semanticrelationality. Tosumup:Tagalogfullwordsareinherentlypredicative, butunlikeverbalpredicatestheylackaninherentargument structure.Thustheymostcloselyresemble(characterizational) nominalpredicates(isastonejathingbeingcut),whichare bothnon-referentialandpredicativebutmorphosyntactically non-relational .Thischaracterizationholdsforbothentityand eventexpressions(and,thoughnotdiscussedhere,property expressions)sincetheysharethesame properties . Severalpointshavetobeaddedtothisdiscussionof categorialityinTagalog : -24-First,notethatitisnotclaimedthatfurthermorphosyntactic subclassificationofTagalogfullwordsisimpossible;rather itisclaimedthataclassificationintomajorsyntactic categoriesisimpossible.Alow-levelclassificationisboth necessaryandpossibl,e.Twoexamplesmaysufficeas illustrations.Oneexamplearepropernounswhichobviously formasubclassoftheirownsincetheyrequireadifferentset ofprepositive,markers(si,ni,kayinsteadofang,ng,sa)and allowforaplMralformationthatisuniquebothformallyand semantically(seeSchachter& Otanes(1972:93f,113).Another exampleis groupofauxiliariesorpseudo-verbsmeaning 'like','want '"'can',etc.,whosedistributionandmorphology clearlysetthemapartfromallotherfullwords,eventhough theymayalsoappear- :rnthesixmajorslotsdiscussedabove (cf.Schachter& Otanes(1972:261ff)). Second,wehavediscussedcategorialityonlywithrespectto function(asreflectedindistribution).Thus thereisstillthepossibilitythatTagalogfullwordsmaybe classifiedintomajorsubclassesonpurelymorphological grounds(theresultbeingmorphologicalcategoriesratherthan syntacticones,thoughJhelabelswouldbethesame(noun, verb,etc.)).Thisisnottheplacetopursuethispoint further(seeHimmelmann(1987:78ff)formoredetails),though itseemsreasonablysafetopredictthattheresultwouldbe similartotha tobt'ainedhere. - Themajorparameteris compatibilitywiththeorientationalaffixes(towhichthe aspectualinflectionisinherentlylinked,seebelow).Ascan beeasilygleanedfromthemajorgrammars(Bloomfield1917, Blake1925,Schachter& Otanes1972)anddictionaries (Panganiban(1972),Santos(1983)),thereisnoorientational affixwhichiscompatiblewithalleventexpressions.Asfor entityexpressions,thereareonlyveryfewwhichare incompatiblewithorientationalaffixes.Theorientational possibilitiesofexpressionsforanimateandhumanbeingsare themostlimited,thoughevenhe rewefindformationssuchas langgam-in'infestedwithants'langgam'ant')orma-lalakih-an'subjugatedbysomeone'svigorousmanliness'ormag--25-lalaki'actlikeaman'lalaki'man,male'). Third,wehavenotclaimedthatTagalogfullwordsmaynotbe semanticallyclassifiedintomajorlexicalsemanticclasses (using,forexample,thecriteriaproposedbyCroft (1991:62ff.Giventheappropriatequalifications,thesemight alsobelabelednoun,verb,etc.,thoughinthispaperthe labelsentity,event,etc.arepreferredinordertounderscore thesemanticnatureoftheclassification.Thereisnodoubt thateventandentityexpressionsaresemanticallydistinctin Tagalog:bat6clearlydenotestheobject'stone',whilelakad clearlydenotestheactionofwalking.Moreimportantly,the distinctionbetweenthesetwokindsofexpressionsisoften formallymarked(bystress),i.e.it isgrammaticallyrelevant. Asanexamplenotethatmanggagamot'doctor'in(1a)is segmentallyidenticaltomanggagamot'willheal,heal habitually'.Botharederivedfromthebasegamot'medicine' andbotharesemanticallycloselyrelated(adoctoris obviouslyonewhohabituallyheals).Thesemanticdifference expressedbythedifferentstresspatternsshowsthatthe formerunambiguouslydenotesanobject(aperson)whilethe latterdenotesanaction .Thisandvariousothersimilar patternsarehighlyproductiveinTagalog .Theproblemof categorialityinTagalogisthereforesomewhatmorecomplex thanhasbeenpresentedintheprecedingparagraphs.The distinctionbetweeneventandentityexpressionsisnota purelysemanticonebuthassomeformal(suprasegmental) correlates.Note,however,thatitisstillvalidtoclaimthat thedistinctionisnotrelevantmorphosyntactically. 4.Orientation Tagalogeventexpressionsusuallydisplayanaffixthat indicatesthesemanticroleofoneoftheparticipantsinvolved inthestateofaffairsdenotedbytheeventexpression.Using thetermsforsemanticmacro-rolesintroducedbyFoley& van Valin(1984:27-32),theseaffixesareglossedasACTOR(ACT)or UNDERGOER(UG,plussubscriptswhichareexplainedbelow)inthe presentpaper.Thefollowingexamplesillustratethebasicfour - 26-affixes : ( 1 4)t-um-ang6'angungg6'(Bl16/6) ACT - nodREFmonkey 'themonkeynoddedinassent' (15)dikdik-insiyasalus6ng(Bl16/27) crush-UG3.SGLOCmortar '(that)he(i.e .theturtle)becrushedinamortar' (16)hulug-anmoak6!(Bl16/17) drop-UGi 2.SG.GEN1.SG 'dropme(sorne)!' (17)kungi-ta-tanimniyaangkaniya-ngka-parte ifUGT-RDP1-plant3.SG.GENREF3.SG.DAT-LKASSOC - part 'ifhewouldplanthispart(forhirn)'(Bl16/6) Intheseexamples,theaffixesindicatethesemanticroleof thereferentwhichappearsintheang: phrase(siyain(14)and ak6in(15)areang-formsofthepronoun).Thus,theinfix-um-in( 14)indicatesthatit isthemonkeywhodoesthenodding, andin(15)thesuffix-inindicatesthattheturtleistobe theUNDERGOERofthecrushing(ratherthantheACrOR). Asalreadyindicatedinsection1,thegrammarofthese affixesishighlycontroversial.Apartfromthecentraipoint whetherorientationmarkingisavoicephenomenonornot,the followinginterrelatedproblemsareinvolved: thenatureoftherelationbetweenthepredicationbase andtheorientationaffixesisunclear .Mostanalyses, however,agreeinassumingaspecialrelationbetweenthe predicationbaseandtheorientationaffix(see4.1) . areorientationaffixesinflectionalaffixesor derivationalones ?AlthoughweagreewithBybee(1985 chapt .4)thatthereisnoclearandabsoluteborder beb.eeninflectionandde.rivation,itisnevertheless importantfortheunderstandingofagivenaffixation processwhetheritislocatedclosertooneortheother endofthecontinuum.Theissuehashardlyeverbeen explicitlydiscussedwithrespecttoTagalogorientation affi xes.AnotableexceptionisdeGuzman(1978,seealso 1991),whowasthefirsttoquestionthetraditional -27-inflectionaltreatmentandtoproposeacleardistinction betweenwhatisinflectionalandwhatisderivationalin orientationmarking.Thepresentaccountradicalisesher positionbyclaimingthatallorientationmarkingis derivationalratherthaninflectional(forasimilar position,seeStarosta1986andendnote4indeGuzman (1991:46)28,Foley1991) . thenumberofaffixesordistinctly-codedsemanticroles variessubstantially:Thetraditionalapproachrecognized fourbasicformations(oneactiveandthreepassives,cf. Blake1906and1925 : 38ff;Bloomfield(1917 : 154);Scheerer (1924);Lopez(1937:46-48);Llamzon(1973: 169;1976:89). Wolfenden(1961: 14-16),Ramos(1971:21-23,56-69),and Foley(1976:105-113)claimthattherearefivebasic formations;McFarland(1976:16-24)proposesseven,de Guzman(1978eh .111)andRamos(1974:19-40)eight, Schachter& Otanes(1972:344)eleven,andfinallyDrossard (1983:87f,1984:86)twelve. 4.1.Thenatureoforientationmarking Thewayourexamples(14-17)havebeenpresentedsuggeststhat amajorrelationexistsbetweenorientationalaffixesandthe predicationbase.Manyproposalshavebeenofferedastothe natureofthisrelation:thattheorientationalaffix highlightsorfocusesonthepredicationbase,thatit determinesthesemanticrelationshipbetweenpredicateand predicationbase,thatitexpressesthecaseofthepredication base,etc.Thus,mostapproachesattributesomekindof relationalqualitytotheorientationalaffixes.Thatthese approachesallheadinthewrongdirectionisevidentfromthe followingfact:Althoughorientatedwordsoccurmostcommonly inpredicativefunction,theyarenotatallrestrictedtothis functionbutoccurinsteadinallthefunctionsavailableto fullwordsinTagalog(seesection3).Inexamplessuchasthe followingitisimpossibletoidentifyapredicationbasefor 28IagreewithdeGuzman ' spointthataspectinTagalogshouldbe consideredaninflectionalcategory. -28-nagsasabuy: (18)athulih-inangnag-sa-sabuysakanya andeateh-UGREFREAL.ACT:GER-RDP1-spatterLOC3.SG.DAT ngbuhangin(BI68/8) GENsand 'andeatehtheonewhowasthrowingsandonhirn Inthisexample,therearetwoorientatedeventexpressions (hulihinandnagsasabuy) ,butthereisapredieationbasefor onlyone(hulihin).Notethatthisisnotaeaseofellipsis; inprineiple,it isimpossibletointrodueeapredieationbase fornagsasabuy.Instead,nagsasabuyispartofthedefinite deseriptionthatfunetionsaspredieationbaseforhulihin.Let usemphasizeoneemorethatthiskindofeonstruetionisquite eommoninTagalogtexts;it isnotamarginalandhighlymarked construction.Therefore,anyattempttogivearelational analysisoforientationmarking,i.e.onethatreferstothe predieationbase,isdoomedtofail.Inexamplessuchas(14-17)thereisnospecialrelationbetweenthepredicationbase andtheorientationaffix.Thereisonlythepredicative relationbetween ' predicationbaseandpredicatethatholdsin anyTagalogclause.Thispredicativerelationisnotdependent ontheorientationaffixbutpertainstostructuralpositions inTagalog(seesection2.2) .Thus,it seemsmorepromisingto investigateorientationmarkinginmorelocalterms,i.e.in relationtothebasestowhichtheaffixesareapplied. It hasbeensuggestedthatorientationaffixesare functionallysimilartonominalisingaffixesinotherlanguages (cf.- amongmanyothers- Starostaetal.(1982:147f.Te callthemnominalisingaffixes,however,isnotveryrevealing aslengasit remainsunexplainedwhyinalanguagesuchas Tagalogtheoverwhelmingmajorityofpredicatesappearsina 'nominalised'form.Furthermore,itisnecessarytodelimit morepreciselythederivationalprocessinvolvedsincemany differentnominalisationstrategiesarefoundinthelanguages oftheworld.Inparticular,it isnecessarytodistinguish betweenthemorphosyntacticandthesemanticaspectsinvolved -29-innominalisations(seesect.3above) . Asformorphosyntax,thetermnominalisationimpliesa changewithregardtothepartofspeechofagivenitem:A verbisturnedintoanounandthismeansthatthemorpho-syntacticpropertiesoftheword(itsdistribution,affixation, etc.)change.Insection3it wasshownthatthereisno morphosyntacticallyrelevantdistinctionbetweennounsand verbsinTagalog.Thus,tocallorientationaffixes nominalisingaffixesseemsmisleadingsinceitcannotbeshown thatthereareanyverbstobeginwith. Thereis,however,asimilaritytothesemanticsideof certainnominalisingstrategies.Orientationaffixeschangethe orientationofagivenwordinsuchawaythatit maybeused torefertooneoftheparticipantsinvolvedinthestateof affairsdenotedbythebase-formoftheword.Forexample,-um-isanACT-orientatinginfixwhichisusedtoderivefromaroot suchastang6''nod,nodding~ assent'theexpressiontumang6' 'onewhonods,nodder'.Thisexpressionnolongerdenotesthe actionofnodding,butrathertheparticipantwhonods. Inthisregardtheorientationaffixesarefunctionally similartothoseaffixesinotherlanguages(includingIndo-European)whichareusedtoformagentivenouns(nomina agentis),objectivenouns(nominaacti(patientis,locative nouns(nominaloci),instrumentalnouns(nominainstrumenti), etc.Lehmann(1984:151f),whointroducedthetermorientation ('Ausrichtung')forthisprocess,characterizesitinthe followingway:Therearetwomajortypesofnominalisation strategiesinmanylanguagesoftheworld.29 Inoneofthese types,resultinginactionnominals(nominaactionis),thecore 29Comrie& Thompson(1985:349ff)proposethesamedistinctionbetween nominalisationsthatresultinthenameoftheactivit ydenotedbytheverb andthosethatrepres entoneoftheargumentsoftheunde rlyingverb.Note aga inthatthefollowingdiscussionofnominalisationisextremelybrief andsurfaceoriented.Itspurposeistopointoutthesimilaritybetween nominalisationinEnglishandorientationmarkinginTaga log.Althoughthe Tagalogdata,ifanalysedthewayithasbeendonehe re.areof considerableimportanceforthelivelydebateconcerningthenatureof nominalisationand,consequently,theinterfacebetweens yntaxand morphology,itisnotthepurposeofthispapertosupportoradvanceany particulartheoryinthisregard. -30-argumentsoftheformerverbmaystillbeaddedasadnominal modifiers,asinPeter'semployingof mybrother.Herethe nominalisedverb(employing)simplydenotesastateofaffairs andimpliesnoorientation.Intheothernominalisation strategyorientationisinvolved:theexpressionforthestate ofaffairsactuallydenotesoneoftheparticipantsinvolvedin theprocessofemploying,e.g.employerwhichinvolves orientationtowardstheACTOR(nomenagentis).Asaresult,the ACToR-argument.oftheverbemploycannolongerbeaddedasan adnominalmodifier,i.e.Peter'semployercannotmeanthat Peterwasthe. agentoftheemploying.Inordertoexpressthis (withthenominalisedform),onewouldhavetousean equationalconstructionsuchasPeteristheemployerofmy , brothe.r.Notethatthesameconstructionisimpossiblewi th nominaactionis(*peteristheemployingof mybrother) .Thus, thetwonominalisationstrategiesdifferinthewaytheydeal withtheargumentslotsoftheunderlyingverb.Theformer basicallyleavesthemuntouched,whilethelatterallowsoneof theargumentslotstobefilledbytheorientationaffixandno longerbyanadnominalexpression.3D Notethatthisdifference pertainstothesemanticrelationalityoftheitemsinvolved; morphosyntactically- andthispertainstobothstrategies-thenominalisedformsnolongerhaveargumentslotsthathave tobefilledobligatorily.Bakingisnotoneofmyfavorite activitiesisawell-formedexpression,whilebakesthatbananil cakeis,asitstands,incomplete. InTagalogwefindtwoderivationalstrategieswhich displayexactlythesameproperties: 1)Orientationaffixesareusedtoderiveorientated expressionsfromagivenbase.Theparticipanttowardswhich theexpressionisorientatedcannotbeexpressedasagenitive 30OneofthemanymoreformalwaystoexpressthisisDiSciu1loand Williams'(1987:4Df)notionof'controlofanargumentbyanaffix'. Theconceptoforientationmayberelatedtosemantical1ybasedaccountsof nominalisationssuchtheonebyBooij(seeSpencer1991: 342f)whoargues thatorientatedderivat ionsinvo1vethebindingofvariablesinlexico-conceptualstructureratherthanoperationsonthepredicate-argument structureofagivenverb. - 31-orobliquecomplement.Thus,in(16)thepronounmo(2.SG.GEN) canneverrefertotheUNDERGOERofthedroppingsincehulugan isanUNDERGOER-orientatedexpression.Theonlypossible morphosyntacticrelationbetweentheorientatedexpressionand anexpressionfortheparticipanttowardswhichitis orientatedisthatofanequationalpredication.Amoreliteral translationof(16)wouldthusbe'I(be)theplaceofyour dropping/yourdroppery'.Similarly,(14)is'nodd-erinassent (was)themonkey" ,(15)'he(be)crush - eeinthemortar',and (17)'ifhisplant-ee(wouldbelhispart'(cf.DeWolf 1988:157f).Thissupportsourclaimproposedabovethatthere isnospecialmorphosyntacticrelationbetweenanorientation affixandthepredicationbase.AllTagalogpredicationswhich involveapredicationbasearesimplyequational,irrespective ofwhetherthepredicateexpressionisorientatedornot. 2)Anotherveryproductivederivation,commonlycalledgerund formation(prefixpag-,fordetailsseeSchachter& Otanes (1972:159ff)),isusedtoderiveeventexpressionswhichare notorientated.AsintheEnglishcase,allparticipantsmaybe expressedasgenitiveandobliquecomplements : (19)atpag-ka-sabiniyanit6 andGER-??-say3.SG.GENDEM.GEN 'andwhenhehadsaidthis.. .' (BI80/1) Gerundsgenerallymaynotbepredicatedaboutapredication base,thus*pagkasabisiyanit6or*pagkasabiniyait6are ungrammatical.Gerundsmayfunctionasequationalpredicates onlyinthefollowingkindofconstruction: (20)pag-Iu-Iuto'ngpagkainangtrabaho GER-RDP2-cookGENfoodREFwork 'His/herjobiscookingfood.' niya. 3.SG.GEN Tagalogorientationmarkingandgerundformationthusshow considerablesimilaritytothetwonominalisationstrategies consideredabove. Ofcourse,manydifferencesexistbetweennominalising affixesin,forexample,Indo-EuropeanlanguagesandTagalog orientationaffixeswithregardtotheirproductivityand -32-semantics_Orientationaffixesmaybeappliedtobasesdenoting actionsasweilastothosedenoting:things(e.g.,bat6 'stone '-)batuh-in'throwstonesat(x)'),masses(e.g.,tubig 'water'-)tubig-an'addwaterto(x)'or'ricepaddy'),states (e . g . ,bago' new'-)baguh-in'change(x)'ori - bago'move(x) toanotherposition'),oranimatebeings(e.g.,baboy'pig'-) babuy- in' make(x)dirty ' ) .Orientationisthusamuchmore prevalentprocessinTagalogthanorientatednominalisationsin Indo-Europeanlanguages .Butapartfromthefundamental differencethatTagalogorientationaffixesarenot nominalisingintermsofmorphosyntax,theoverallsimilarity inthefunctionoftheseaffixesisconspicuous. Incidentally,thederivationalpossibilitiesjustmentioned showthatinsteadofcallingtheorientationaffixes nominalising,acasecouldbemadetoanalysethemas verbalisingaffixes .ThisisinfactFoley's(1991)approach, whichistosomeextentareversalofthenominalisation analysis .Accordingtohisapproachthebaseformsare precategorial,andorientationmarkingisusedtoderiveverbs fromthesebases .Inthisprocesstheverbsreceivetheir argumentst+,uctureandatthesametimethe'topic'function (predicationbase)isassignedtooneofthesearguments .This analysisisclearlypreferabletomostotheranalysesthathave beenproposedtodate .Ourpointofcontentionshouldbe obviousfromtheprecedingdiscussion(seealsosect.3):.Where istheevidenceforasyntacticcategoryverband, consequently,forargumentstructure? Theanalysisproposedherehasseveraladvantagesasweil asrepercussionsformanyareasofTagaloggrammar.Tomention justtwo :Firstandforemost,theabilityoforientated expressionstofunctionasreferentialexpressions(cf.(1b), (3),(13) ,(18))isaccountedfornaturally.Second,another oddityofTagaloggrammariseasilyresolved .Abrieflookat thelistofaffixesinSchachter& Otanesfornouns (1972:97 - 106),adjectives(1972 : 198f;216-229),andverbs (1972:344-355)immediatelyrevealsthatbasicallythesame affixesareinvolved(bothformallyandsemantically).The -33-differencesthatdoexistonlypertaintostressand,in correlationwithstress,tothewayaconceptisdenoted(as event,entity,orproperty).Theanalysisproposedhereallows forconsiderablesimplificationinstatingtheregularitiesof affixationsincetheseonlyhavetobestatedonce.The differencesthatexistbetweenentityandeventexpressionsare statedwithreferencetothelinguisticleveltowhichthey pertain,i . e.stressassignment.Togivejusttwoexamplesof formationswhereso-callednominalandverbalderivations clearlyoverlap:1)mag- tplusunstressedreduplication)occurs informationsdenotingprofessionals;thusfromnakaw'steal' mag-na-nakaw'thief'maybederived.Thisformdiffersonly withrespecttostressfromtheactiondenotingformation mag-na-nakaw'willsteal'(cf.sect.3andBloomfield 1917:242f;Schachter& Otanes1972:103).Bothformationsare clearlyACTOR-orientated.Thereisnoevidencetosupportthe distinctionbetweenmag- asanominalisingaffixandmag- asa verbalaffix.2)Thesuffix-inmaydenotetheentities undergoingtheactiondenotedbytheroot,e . g.,aral'study' -)aralin'lesson'.Again,itisthestressthat(often) differentiatesactionandthing,cf.aralin'study(x)'.Other examplesarekumpuni'repair'-)kumpunihin'thingstorepair' vs.kumpunihin'repair(x)';kain'eat'-)kanin'boiledrice' or'eat(x)'(nodifferencewhatsoever,comparealsokakanin 'sweets');in6m'drink'-)inumin'drinkingwater'or'drink (x)',cf.alsoinumin'beverage' (seeBloomfield1917:247; Schachter& Otanes1972:99f).Again,alltheseformationsare UNDERGOER-orientatedexpressions.Thedistinctionbetweenevent andentity,ifitisformallyexpressedatall,pertainsto stressandnottothesuffix.31 31Informationsinvolvingthesuffix-anstressalso(often)distinguishes aneventandanentityexpression .Butwiththissuffixthematteris furthercomplicatedbythefactthatathirdmeaningmaybedistinguished, i.e.collectiveaction(cf.Bloomfield1917:250-262) .Ther eareno formationswith-um- ori - whichmaydenoteentities. -34-4_2.Thesemanticsoftheorientationaffixes Thediscussionintheprecedingsectionshowedthatorientation affixesaretobeconsideredneitherassomekindof inflectionalmarkernorasmarkersofachangeinsyntactic category.Instead,theresultsoftheprecedingsection stronglysuggestaderivationalapproachtotheseaffixesand, morespecifically,aderivationalapproachbasedon compositionalsemantics.Therefore,thehypothesiswillbe advancedherethatorientationalaffixationmaybeanalysed withthehelpofderivationalrules,oneswhichreferbothto thesemanticsoftherootandthesemanticsoftheaffixes.The semanticsofthebasewillbediscussedinthenextsection.In thissection,wewillbrieflypresentourviewonthesemantics oftheaffixes. Inourview,boththeformalandthesemanticevidence stronglysupportsthetraditionalclaimthatthereare basicallyfourorientationsinTagalog:onetowardstheAGTOR, andthreekindsofUNDERGOER-orientation.Theunificationofthe latterthreeorientationsundertheheadingofUNDERGOER-orientationissupportedbythefactthattheydisplaythesame modalinflection(theREALISinfix-in-,see4.3).Infact,the existenceofanaspectual/modalinflectionalparadigmforall fourofthebasicorientationsisthemajorformalevidencefor combiningfourotherwiseformallyverydifferentformations undertheheadingoforientation.Thatis,apartfromthefact thatorientatedformsallowforthesameaspectual/modal inflections,theyarefarfromuniform,neithersemantically norformally. AGTOR-orientationismarkedbyaninfix(-um-)andis relativelystronglygrammaticalizedtotheextentthatit denotesnotonlyAGTORSthatcontrolanaction(asin(14))but alsoparticipantsinvolvedinaprocess,asinp-um-ula'become red'orl-um-u-lutang'befloating'.Furthermore,it occursin expressionsfornaturaleventssuchasum-ulan'rain'or l-um-ind6l'earthquake'.Thisinfixisnottheonlywayto signalACT-orientation.Theprefixmag- (REALISnag-)also indicatesthisorientation,cf.examples(1b)and(18). -35-FollowingdeGuzman(1978Chapt_3),thisprefixisanalysed hereasconsistingoftheprefixpag- usedingerundformation (seeabove),ACTOR-orientation(andmood)beingsignalIedby consonantalternation_32 Thedifferencebetweenthese formationspertainstothekindofactiondenotedandis furthercommentedoninHimmelmann(1987:178ff) _ UNDERGOER-orientationisexpressedbytwosuffixesandone prefix: ( 2 1 ) -inindicatesadirectlyaffectedUG,suchastheturtle in(15),thatwhichisbeingcut(inexample(3,the persontobecaught(inexample(18,orit6ininum-inmo it6'drinkthis'_1tistheunmarkedandthemoststrongly grammaticalizedmemberofthethreeUNDERGOER-orientations sincethereisnosuffixintheREALISmood(see4 _3)and sinceit isusedinallcaseswhichdonotclearlypertain totheothertwoUNDERGOER-orientations(formore discussion,seeHimmelmann(1987:107ff). -anindicatesthelocationtowardswhichanactionis orientated,suchasthe1. SGin(16)(calIeddirectional focusintheliterature)orthelocationwhereanaction takespI ace(theso-calledlocativefocus) ,e.g.: ni-Iakar-anko REAL(UG)-walk-UGi 1.SG. GEN 'Iwalkedonastonyroad' angma-bat6-ngkalye REF1RR.STAT-stone-LKstreet IpreferthetermINDIRECTUNDERGOER(UGi)becauseof examplessuchasbuks-anmoangpint6''openthedoor', wheretheparticipantdenotedisneitherlocativenor 32Theargumentforthisanalysis,aspresentedbydeGuzman,maybe summarizedasfo1lows:A numberofprefixpsdisplayregularalternationof theinitialconsonant:theI p/ -initialformisthebasicform(usedas gerundorimperative),I mlmarksIRREALISandI nlREALIS,compare pag-Imag- I nag-,pang' l mang-Inang-,paki -Imaki -Inaki -.Thealternation betweenthetwonasalformsalsooccurswiththeSTATIVEprefix(ma - Ina - ) , althoughforthesenoIp/-initia1basicformexists .Furthermore ,asma1l numberofI p/ -initialroots(containingafossilizedprefix)exhibitsthis alternation,e . g.pakinigmakinignakinig'listen',pano6dmano6dnano6d 'watch'.Thisalternationprobab1yd e v e 1 ~ e d frominfixedformations,with 10ssofthefirstsy1lable.Thus,mag- (pumag- andnag- (*pinag- (note, however,thatpinag- isaproductiveformation,i.e . ,REALIS(UG)of pag-derivedbas es. -36-directionalbut,moregenerally,anindirectlyaffected undergoer.Furthermore,thereisaclearopposition between-inand-anpertainingtodirectness,whichis evidentfrommanycontrastivepairs(compare,forexample, inum-inmoit6abovewithinum-anmoit6'drinkfrom/some ofthis'. i- indicatesanUGthatismoved(thematicUG(UGT) ,cf. Kroeger(1988:231-33)whousestheterm'translative focus'),suchastheeggin(1a)orthehalfofthetreein (17) .Theanalysisofthisprefixissomewhatmore controversial.Itisoftenanalysedasamarkerforthe instrumentalrolebecauseofexamplessuchas (22)angitkayi-p-in-utolkongsaging REFboloPMUGT-REAL(UG)-cut1.SG.GENGENbanana 'Icutbananaswiththebolo.' Thisusage,however,islessprominentandmayeasilybe accountedforasUGT (cf.Himmelmann(1987:139f.Buti-isalsousedtoindicatethebeneficiaryofanactionwith afewroots,asini-bili'buyfor(x)',whichis difficulttoaccountforunderanyanalysisproposedso far.33 Notethati- is outstanding butisalsoformallyhighlyconspicuoussinceit isthe onlyprefixamongtheorientationalaffixesaswellasthe mostirregularoftheTagalogprefixes(unlikeother prefixes,it isneverstressednorreduplicated). 4.3.Aspect/moodinflectionandinherentorientation Oneofthemostconspicuouscharacteristicsoforientation markinginTagalogisthefactthatthebase-formsdonotseem tohaveaninherentorientation;bothACT- aswellasUG-orientationinvolvemorphologicalmarkingwhile,forexample, inEnglishonlyUG-oriehtation(passive)isexplicitlymarked. 33Onepossibilityistoanalysethisasacaseofhomophony,i.e.,there aretwoUNDERGOERprefixes,athematiconeandabenefactiveone(which wouldimplythattherearefivebasicorientationsinTagalog).Themajor argumentagainstthisapproachisthefactthatbenefactiveparticipants mayaIsobeindicatedby-an ,thatis,asINDIRECTUNDERGOERs(whichmakes moresensesemantically) .Furthermore,benefactivei- ismarginalbothin termsoftypeandtokenfrequency(seeHimmelmann(1987:141f . -37-Thequestiorrwhetherthereisanyevidencefororientation inherenttothebaseformshas,tomyknowledge,neverbeen posedbefore.Theissueisquitecomplexandonlysome preliminaryremarksarepossiblehere.Wewillstartby pointingouttwoasymmetriesintheaspect/moodparadigmwhich maybeusedinformulatingahypothesis. FororientatedformsTagalogdistinguishesbetweentwo aspects(PERFECTIVEandIMPERFECTIVE)andtwomoods(REALISand IRREALIS).34IMPERFECTIVEaspectisindicatedbystressed reduplication(RDP1);PERFECTIVEaspectisunmarked.REALISmood isindicatedbytheinfix-in- orconsonantalternation(/m/-) , /n/,cf.fn32),IRREALISagainbeingunmarked.Ininteraction withorientationmarking,thefollowingparadigmscanbe establishedfortherootbili'buy',oneofthefewrootswhich iscompatiblewithallorientationalaffixes: IRR/pRF IRR/IMPF REAL/PRF REAL/IMPF IRR/PRF IRR/IMPF REAL/PRF REAL/IMPF IRR/PRF IRR/IMPF REAL/PRF REAL/IMPF ACT 'buy' b-um-ili bibi li b-um-ili b-um-ibili UG bilh-in bibilh-in b-in-ili b-in-ibili UG i-bili i-bibili i-b-in-ili i-b-in-ibili ACT 'seIl' mag-bili mag-bibili nag-bili nag-bibili UC bi1.h-an bibilh-an b-in-ilh-an b-in-ibilh-an Asmaybeimmediatelyobservabletherearetwoasymmetriesin theseparadigms(seetheIRRoftheum-infixedformsandthe 34ThisistheanalysisproposedbyB1oomfie1d,who,however,usesan over1yidiosyncratictermino1ogy(1917:217).Schachter& Otanes(1972:66ff, 361ffproposeasomewhatdifferentanalysisinvo1vingabasicform (=IRR/IMPF)andthreebasicaspects:CONTEMPLATED(=IRR/PRF),PERFECTIVE (REAL/PRF),andIMPERFECTIVE(REAL/IMPF).Theformalevidencesupports Bloomfield'sanalysis.Furthermore,thereisanotheraspectualformation, RecentPERFECTIVE(withprefixka'),whichisnotorientated.Thiswillnot beconsideredfurtherhere. -38-REALoftheUG-forms).Nevertheless,it seemsjlistifiedto establishtheseparadigmaticarrangementsandtospeakof aspectualandmodalinflection,sincetheseformationsare highlyregular(theyexistforeveryorientatedformandare alsopossibleforotheraffixcombinations,e.g.maki-, makiki-,naki-,nakiki-)andthecontextstheyareusedinare alsoidentical.Thus,IRR/PRFisusedinhypotheticaland complementclausesandincommands(cf.(15)and(16));IRR/IMPF isusedforfutureevents(cf.(17)),whereasREAL/PRFand REAL/IMPFareusedforpastandpresentevents. Theasymmetriesarerelevantindiscussingtheinherent orientationofTagalogroots.Strictlyspeaking,REALISUG-orientatedformsdonothaveamarkerfororientation. Similarly,thereisnomarkerforACT-orientationinthe IRR/IMPFoftheum-paradigm.Thismaybeinterpretedasevidence fortheclaimthatbaseformsofeventexpressionshave,notan inherentorientation,butratheradefaultorientation(which remainssegmentallyunexpressed)dependingonthecontextin whichaneventexpressionisused.InIRR/IMPFcontextsthe defaul torientationistowardstheACTOR;inREALIScontexts (bothPERFECTIVEandIMPERFECTIVE) towardstheUNDERGOER.Such acorrelationisnotuncommoncross-linguistically(notethe variouscasemarkingsplitsbasedontense/aspectsplits (Tsunoda1981,Mallinson& Blake1981:59ff)).Blake(1988:79f) eventakestheasymmetriesasevidencefortheclaimthat Tagalogisamorphologicallyergativelanguage,moreprecisely, asplitergativelanguage.Theproblemwithhisargument(and alloftheotherergativeproposalsmentionedinsect.1)is thefactthatBlakeisforcedtoclaimthatACTOR-orientation isakindofantipassive,i.e.ACTOR-orientatedconstructions areclaimedtobeintransitive(loc.cit.81).Thisclaim, however,isimpossibletosupportsinceACTOR-orientated predicationsarepossiblewithmorethanoneparticipant: .(23) nag-da-dal REAL.ACT-RDP1-bring -39-silngsarilenil-ng 3.PLGENown3.SG.GEN-LK (BI48/33)banda band 'they ngmusika GENmusic bringtheirownband' Thematteriscomplicatedbythefactthatthereisacertain prominenceofUNDERGOER-orientationinTagalog:Whenevera clauseinvolvesadefiniteundergoerthisisusuallymadethe predicationbase .Thisistakenbysomeauthorsasevidencefor theintransitivestatusofACTOR-orientatedconstructionssince inmanylanguageseventsinvolvingnon-specificandspecific-indefiniteundergoersareexpressedbyintransitiveclauses (cf.Hopper& Thompson1980) .AsforTagalog,however,thedata concerningthisphenomenonaremisrepresentedinthe literature.Itisoftenassumedthatinsteadofbeingavery generaltendencytheprominentuseofUNDERGOER-orientationisa syntacticruleofTagalog,whichitisnot,asshownbyexample (23).35Furthermore,thistendencyisnotsufficienttoproove theintransitivestatusofACTOR-orientatedexpressions,andI cannotseehowthiscanbedone.Onthecontrary,ifthe analysispresentedheresofarisbasicallycorrect,thewhole approachseemstorestonthewrongassumptionthattransitive andintransitiveverbsmaybedistinguishedinTagalog;this cannotbethecase,however,sincethereisnosyntactic categoryverbtobeginwith.Further,wherethereisno transitivitydistinctionwithrespecttoverbs,noergativity ispossible. Returningtotheasymmetrieswhichexistintheaspect/mood paradigmof a g a l o ~ orientatedexpressions,notethatthereis furtherevidencefortheclaimproposedherethatbaseslackan inherentorientation.Althoughthisishardlyeveracknowledged 35SeeAdams& Manaster'Ramer(1988)formorediscussionoftheissues involved,hutnotethatthoughtheiraccountisprobablythemostdetailed andconsiderateaccountavai1abletodateitstillhassomeflaws.In accordancewithalloftherelevantliteraturetheyclaim(1988:92ff)that apartfrompossessedundergoersasin(23)- whichintheiranalysisare not(!)definite- definiteundergoersmayneveroccurasgenitivesin Tagalog .Thisisnottrue;compareexamp1e(19)whereadefiniteundergoer ismarkedasgenitive(nit6). -40-intheliterature(themajorexceptionbeingBloomfield (1917:218ff),itispossibletouseunaffixedrootsdenoting eventsinpredicatepositionwhichareUG-orientatedinaREALlS context(oftenwitharesultativeconnotation): (24)antaykoangsag6tmo wait1.SG.GENREFanswer2.SG.GEN 'Iwaitforjexpectyouranswer' ACTOR-orientationwithunaffixedrootsinREALIScontextsis impossible: (25)a) b) *uwi'siyasabahay return3.SGLOChouse 'Hereturnedhorne.' *dalangman6kangkuya bringGENchickenREFolderbrother *'mybrotherbroughtachicken' ?.'thechickenbroughtmybrother' ko 1 . SG. GEN Theinformantsclaimedthat(25b)wasacceptableinthesense of'thechickenbroughtmybrother'(thoughthisis,ofcourse, peculiarpragmatically).Butunaffixedrootsmayoccurwith ACT-orientationinimperatives(i.e.,anIRREALIScontexts), e.g .,hintayka'youwait!'.Furthermore,theymaybeusedto denoteaneventwithoutorientation(andthusresemblethe gerundformationsmentionedabove)inbasicallythesame contextsinwhichUNDERGOER - orientationispossible: (26)iyak cry 'His angsag6tniyasaakin REFanswer3.SG.GENLOC1.SG.DAT answertomewascryingjtocry' Theconditionsapplicabletotheuseofunaffixedformsandthe natureofpossiblesemanticdifferencesbetweenaffixedand unaffixedformsarestillunclear(formorediscussion,see Himmelmann1987:157ff) .Thedatainvestigatedsofar,however, supportthehypothesisthateventexpressionswhichlack orientationaffixesalsolackaninherentorientation(andare thusneitheraccusativenorergative). - 41-4_4 _Derivationandprediction Ouranalysisoftheorientationaffixesultimatelydependson thehypothesisthatorientationmarkingisaderivational processratherthananinflectionalone,asisoftenassumedin theliterature.Inthissectionwewillpresentsomefurther observationstosupporttheclaimthatorientationmarkingin Tagalogisderivationalor,atleast,morederivationalthan inflectional. Synchronically,thederivationalcharacteroforientation markingisevidentfromthefactthatthereisnobasic orientatedformfromwhichotherformationsmaybepredicted . Thatis,ifoneknows,forexample,thatagivenrootallows fortheinfix-um- thereisnowaytopredictwhichother orientationsarepossible.McFarland,whohasinvestigatedthe orientationalpossibilitiesfor332rootsintexts,concludes: 'ThefocusinflectionsinTagalogaresubjecttoagreat amountofidiosyncraticbehavior .Thedegreeof predictabilityfromonefocusformtoanotherisvery 1 ow.'( 1 976 : 32) Thisisparticularlynoteworthybecauseofthefactthatother formationsinTagalogaremuchmoreregularandcanbe predictedonthebasisofbasicforms.McFarlandwrites: 'Asidefromthefocusinflections,theverbalinflectional systemexhibitsahighdegreeofregularity.Theformation ofgerunds,aptativeforms,causativeforms,etc.follow highlypredictablepatterns.'(loc.ci t. ) Furthermore,therehavebeenseveralattemptstodealwiththe idiosyncraticbehaviourofTagalogorientationmarkinginterms ofverbclasses:Blake(1925 : 38f)postulated17classes; Schachter& Otanes(1972:295ff)operatewith43classes;Cruz (1975)recognizes38classes;McFarland(1976:101ff)has53 classes;Ramos(1974,1975)postulates15classes;andde Guzman(1978)recognizes7mainclasseswith48subclassesfor 'primaryverbsterns'and14mainclasseswith32subclassesfor 'secondaryverbsterns'.Althoughthedifferencesamongthese authorsareinpartduetodifferencesintheirrespective frameworks,thisenormousrangeofproposedclasses,inmy -42-opinion,supportstheclaimthatanapproachintermsof inflectionalclassesissimplyinappropriate.Thederivational approachproposedhe repredictsthatvariousidiosyncracies willoccurinorientationmarkingandthusseemstobeclearly preferable(forapreliminarysketchofsuchaderivational approach,s"eeHimmelmann1987:129ff). Intheprecedingdiscussiontheheterogeneityofthe orientationmarkingaffixeswasalreadypointedoutseveral times.Althoughthisinitselfdoesnotproovethatthese affixationsarederivational,it isclearthatsuchavariety offormalexpressionsforonecategoryseldomarisesin inflectionalparadigms.Theheterogeneityclearlyshowsthatwe aredealingwithformationswhichhavedevelopedfromquite differenthistoricalsourcesandareprobablyofavaried historicalage .Itisbasicallytheaspectualjmodalinflection commontoallorientationaffixeswhichallowsustounitethis ratherheterogeneoussetofaffixesunder "thelabel orientation.The'holes'intheaspectualjmodalparadigm pOintedoutintheprecedingsectionsuggestthatweare dealingwithalanguagewhereatacertainstagethe derivationalmorphology'hasgonewild'.Thatis,atanearlier stageeventexpressionsprobablywereinflectedonlyforaspect (reduplication)andmood(infix_in_).36Lateron,theuseof orientatedforms- whichwereformerlyusedinasimilarwayas orientatednominalisationswereinIndo-Europeanlanguages-wasgeneralizedand,asaconsequence,theequationalpattern becamethedominantpatterninTagalogclausestructure.In thislineofevents,orientatedformsalsoacquiredregular aspectjmood-inflection.Onlythemostcommonformations 'survived',namelyreduplicationforACTOR-orientated IRREALISjIMPERFECTIVEand- in- infigationforUNDERGOER- orienta ted 36ThepresentdistributionoftheREALISinfix(itoccursonlyin UNDERGOER-orientation)clearlyisarecentdevelopment.Oldersourceson TagalogsuchasMller(1882:140)andBlake(1906and1925:41)adducethe followingREALIS-allomorphforthe-um-infix:-ungm-,e.g.,s-ungm-ulat 'wrote 'ands-ungm-u-sulat'is/waswriting'.Furthermore,inseveral Philippine-typelanguagestheACTOR-orientating-um- isstillcompatible withtheRELAIS-infix(see,forexample,Tondano(Sneddon1975:211)). .\ I -43-REALIS.Theirdistributionwhichprobablyusedtoberestricted tothepredicativefunctionchangedinsuchawaythatthe earliermorphosyntacticdistinctionbetweennounsandverbs becameobsolete.Thiswasduetotheirinteractionwiththe (formerlynominal)orientatedforms.Whatwethen synchronicallyanalyseasaspect/moodparadigmforTagalog eventexpressionshistoricallyconsistsofnominalizedand verbalforms.Ishallnotgointothisdiachroniescenarioin detail.37 It maysuffice,however,toindicatethatitis possibletosketchaplausiblediachroniescenarioastohow theTagalogstateofaffairsmighthavearisen.Notethat withintheAustronesianfamilyonlyTagalog(and,tovarying degreestheotherPhilippine-typelanguages)haveevolvedthe peculiarmorphosyntaxweareinvestigatinghere. 5.Syntacticcategoriesandthegrarnrnaticalizationofargument structure Theprecedingdiscussionwasbasedonthehypothesisthatthere isnodistinctionbetweenthesyntacticcategoriesnounand verbinTagalog.Therelateddistinctionbetweeneventand entityexpressions,however,doesexistandis(sometimes) formallymarkedbystress.Thusoneofthetraditional definitionsfordistinguishingbetweennounsandverbs,that nounsandverbsaredifferentwaysofdenotingconcepts(modus significandi),isapplicabletoTagalogfullwords. Furthermore,therecanbenodoubtthatTagalogevent expressionsaresemanticallyrelational,asopposedto(most) entityexpressions.Themajorreasonforthehypothesisedlack ofadistinctionofsyntacticcategorieswasthefactthatthe differenceinsemanticrelationalitybetweenentityandevent , expressionsisnotclearlyreflectedintheirmorphosyntax.In otherwords,bothentityandeventexpressionsallowforthe samekindofcomplements(genitiveattributesorobliques),and neitherobligatorilydemandstheexpressionofacomplement . 37ThemajorevidenceisprovidedbyFormosanlanguageswherenounsand verbsaremoreclearlydistinguishedthaninTagalogandthefactthatin PANtherewereprobablytwoaffixes(* - iand*-aken)whichcouldbeused onlywithverbs(forsomediscussion,seeStarostaetal.(1982. -44-Therefore,thehypothesishasbeenputforwardthatallTagalog fullwordslackinherentmorphosyntacticrelationality(and inherentreferentialityaswell,seesect.3).Thishypothesis allowsforaconsistent,surface' -orientedandrelativelysimple statementofthecoreofTagalogmorphosyntax.Thefollowing characteristicsarethuseasilyaccountedfor: thedistributionoffullwords theuseofthemajorfunctionmarkersang,ng,andsa thebasicequationalcharacteroftheTagalogclause theveryexistenceandpervasivenessoforientation marking Thepurposeofthissectionistotestfurtherandrefine thehypothesilithatTagalogfull _wordslackinherent morphosyntacticrelationality.Itisnotthecasethatthere arenom?rphosyntacticdifferencesatallbetweenevent expressionsandentityexpressionsinTagalog;rather,the hypothesisproposedhereisthatthedifferencesthatdoexist aredistinctfromthoseencounteredinotherlanguagestosuch adegreethatananalysisinterms ,ofsyntacticcategories (nounsandverbs)isinappropriate.Putdifferently,syntactic categoriesarebutweaklygrammaITdlizedinTagalog.Letus illustratethiswithanexample: (27)a) b) mang-ga-gamot IRR.ACT-RDP2-medicine 'Heisadoctor.' mang-ga-gamot IRR.ACT-RDP,-medicine 'Hewillheal.' siy 3.SG siy 3.SG manggagamotin(27a)clearlydenotesaperson,while manggagamotin(27b)clearlydenotesanaction'.Onepertinent differencebetweenthetwoisthatthelatterdoesallowfor aspectual/modalinflection(manggamot,nanggagamot,nanggamot) whiletheformerdoesnot.Thishasoftenbeenusedasthe majorcriteriontodistinguishnounsandverbsinTagalog,but sincetheinflectedformsmayappearinallmorphosyntactic environments(functions),Idonotconsiderthisevidencefora -45-distinctioninsyntacticcategory(seesect.3) .Unliketheir Englishequivalents,theTagalogentityexpression , (manggagamot)mayfunctionaspredicatewithoutfurther morphosyntacticfunctionmarking.Furthe rmore,complementsmay beaddedinthesameway :mang-ga-gamotngbata'siyameans'he isadoctorof(for)children',mang-ga - gamotngbata'siyais 'hewillhealchildren' .Infact,thepervasiveuseofthe genitive(ng-phrase)inTagalogisthemajorfactorinthelack ofamorphosyntacticdistinctionbetweennounsandverbs .The twocasesavailableinTagalog,genitiveandoblique,arecases which,viewedfromcross-linguisticpersepctive,aregenerally usedforeitheradnominalcomplementsorperipheral participants.Thefactthatcomplementsofeventexpressions havetobeexpressedinthiswayinTagalogdeprivesthemof thecentralmorphosyntacticfeatureofverbhood,thegoverning ofcorearguments .Putdifferently,obliquecomplementsare equallypossiblewithbothnounsandverbs ,whereasgenitive complementsaretypicalfornounsinmost,ifnotall languages .Tagalogeventexpressionsthuslacktheexpression ofcomplementstypicalforverbs;insteadtheyexhibita 'nominal'charactersinceparticipantsareexpressedas genitivesandobliques. This,however,isnotthewholestorysincethereare variouscomplexitiespertainingtodefiniteundergoers.In example(27a)it ispossibletoaddadefinitecomplement markedasgenitive : (28)a)siyayangmang-ga-gamotngmga 3.SGPMREFIRR . ACT-RDP2-medicineGENPL 'heistheonewhoisthedoctorofthese bata-ngito child-LKDEM children' . Thisisnotpossiblein(27b) .Instead,thedefiniteundergoer hastobeexpressedasanoblique : (28)b)siyayangmang-ga-gamotsamgabata-ng 3.SGPMREFIRR.ACT-RDP1-medicineLOCPLchild-LK ' heistheonewhowillhe althesechildren'. ito DEM Herethecomplementofanentityexpressionandtheundergoer ofaneventexpressionareclearlymarkedinadifferentway; -46-thedifferenceinsemanticrelationalitybetweenentityand eventexpressionsisreflectedinthemorphosyntax.This,in fact,maybeinterpretedasevidencethatatleastaminor differenceinsyntacticcategorybetweeneventandentity expressionsexistsinTagalog. Ashintedatabove(sect.3),thedistinctionbetween semanticandmorphosyntacticrelationalityisnotanabsolute one.Instead,morphosyntacticrelationalityistheresultof thegrammaticalizationofsemanticrelationality:certain aspectsoftheframeevokedbyarelationalconceptare reflectedinitsmorphosyntacticexpression.Themajorfactor inthegradualformationofthesyntacticcategoriesnounand verbisthefactthatthewayinwhichtheparticipantsevoked byasemanticallyrelationalaremorphosyntactically linkedtothisexpressiondiffersatleastinsomedetailfrom thewayattributesarelinkedtonon-relationalexpressions (i.e.themajorityoftheentityexpressions).Asiswell known,themajormorphosyntacticreflectionsofsemantic relationalityareagreementandgovernment(cf.,forexample, Lehmann1982,1985),andalthoughthereisnoclearevidence foreitherinTagalog,someevidenceforthegrammaticalization == ofsemanticrelationalitycanbediscerned.Thatdefinite undergoersofeventexpressionsmustbemarkedasobliquesisa caseinpoint.Furtherfactspointinginthesamedirection includethefollowing: Participantrolesaremappedontoeitherthegenitiveor theobliquefunctionsinanon-arbitraryway.Wemaynotethe followingregularities(theseregularitiesaretobeunderstood asincludingthequalificationthatanyparticipantmayalsobe chosenaspredicationbase):agentsandexperiencersarealways expressedasgenitives;instrumentseitherasgenitivesor markedwithacomplexpreposition,butneverwithsimplesa; locativesalwaysasobliques;patients,themesandgoalseither asgenitivesorasobliques.Asforgoals,thesearegenerally markedasobliques,butoccasionallyagenitiveconstruction alsooccurs.Compare: (29) -47-b-um-aliknauli'ak6 REAL.ACT-returnLKagain1.SG 'Ihavereturnedoncemoreto ng GEN the probinsya province province. ' Theregularitiesforpatientsandthemesarequitecomplexand interactwiththefactorsinvolvedinchoosingaparticular orientation.Themostimportantparameterinthisregardisthe specificityofundergoers;thatis,specific,especially definite,undergoersareusuallymadethepredicationbase. Otherwise,specificundergoersaregenerallymarkedasobliques (butseefn35)andnon-specificundergoersasgenitives(for morediscussionseeNaylor1975and1986,Adams& Manaster Ramer1988).Asalreadyhintedatinthediscussionofexample (23)above,thefactsaremorecomplexthanthissimplified statementsuggests.Butthesewillnotbefurtherinvestigated here.Inthepresentdiscussionthefollowingobservationisof particularinterest:themorphosyntacticcodingofparticipant rolesinpartallowsforthedistinctionofmorecentralfrom lesscentralparticipants.Porexample,agentsareclearly centralparticipantsbecausetheyarealwaysmarkedas genitives;locativesaremoreperipheral,fortheyarealways markedasobliques.Thedistinctionbetweencentralandless centralparticipantsisobviouslyanecessarystageinthe grammaticalizationofargumentstructure.Notethatthe evidenceregardingthecentralityoftheundergoerrolesisnot decisive. Theasymmetriesintheaspectualjmodalparadigmsof orientatedformspointedoutinsect.4.3couldbeadducedas evidencefortheclaimthatactoranddirectundergoer (basicallypatient)arethemostcentralparticipantsin Tagalogeventexpressions,sincetheorientationdoesnothave tobeexpressedsegmentallyinsomecontexts(REALlSfor undergoers,IRREALISjIMPERFECTIVEforactors) . Withregardtosomeeventexpressionsthederivational morphologymayalsobeusedtodeterminethecentralityofa givenparticipantroleinthattheorientationtowardsone participantroleismorphologicallymorecomplexthanthatfor others.Porexample,thereisagroupofrootswhichallows -48-ACT-orientationonlywithmag- (thatis,firstagerundhasto bederivedbyprefixingpag- whichinturnisorientatedby consonantalternation,cf.fn32),whileuG-orientationis possible apriorderivation.Fromluto''cook'onemay notderive*lumuto'butratheronlymagluto',whiletheUG-orientatedformissimplyiluto'orlutu'in.Otherroots belongingtothisgroup,whichdoesnotexhibitacommon semanticdenominator,aredasal'prayer',hugas'wash',punas 'wipeoff',libing'burial',bayad'payment',kahoy'wood', hubad'naked',etc.(cf.Himmelmann1987:179f).Thusforthese rootstheclaimcouldbemadethattheundergoersaremore centralparticipantsthantheactorsinceorientationtowards themrequireslessmorphologicalmarking. Furthermore,thereisonemorpheyntacticcontextinwhich onlyentityexpressions,butnoteventexpressions,mayoccur: unaffixedrootsmaybeusedasimperativesonlyiftheydenote events(see4.3).Thereisno*lalakikat'beaman!'or*bat6 kasakaniya!'throwstonesathirn!'Thatis,atleastinthis somewhatmarginalcontextthereisamorphosyntacticdifference betweeneventandentityexpressionwhichcouldbeadducedas foradistinctionofthesyntacticcategoriesnounand :..-:=;.:0-" verbinTagalog.Butthenthisfactmayeasilyandnaturallybe statedintermsoflexicalsemanticclassesandthus,inmy viewdoesnotconstitutecompellingevidencet6alterthe analysisofsyntacticcategoriesproposedaboveinsect.3. Allofthefactsmentionedintheprecedingparagraphs doub