Alignment and Word Order in Old Japanese Author(s): Yuko Yanagida and John Whitman Source: Journal of East Asian Linguistics, Vol. 18, No. 2 (May, 2009), pp. 101-144 Published by: Springer Stable URL: http://www.jstor.org/stable/40345246 Accessed: 13-09-2015 04:35 UTC Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/ info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Springer is collaborating with JSTOR to digitize, preserve and extend access to Journal of East Asian Linguistics. http://www.jstor.org This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions JEast AsianLinguist(2009) 18:101-144 DOI 10.1007/sl0831-009-9043-2 Alignmentand word orderin Old Japanese Yuko Yanagida • JohnWhitman online:16 June2009 Received:2 November2008 /Accepted27 March2009/Published Science+BusinessMedia B.V. 2009 © Springer Abstract This paperarguesthatOld Japanese(eighthcentury)had splitalignin inmainclausesandactivealignment ment,withnominative-accusative alignment for active alignmentin nominalized nominalizedclauses. The main arguments of activesubjectsand thedistribution of twoverbal clausecomefromga-marking and sainactive for active /-for predicates(cf. Yanagida,In: predicates prefixes: in Japanese], shubun clause no gensho[Main phenomena Hasegawa(ed.) Nihongo and of non-accusative the treatment review We 2007b). alignment arguethatactive We as as a distinct should be type. proposea formalanalysisof analyzed alignment is clausesin Old Japanese.The externalargument in nominalized activealignment are as in in v. out situ inherent case, spelled Spec, Objectarguments ga, assigned licensedby severaldistinctmechanisms, (Yanagida,In: includingincorporation in and case MIT 2007a) assignment by Papers Linguistics, Working Miyamoto(ed.) O wo S ga V head above vP. The latteraccountsforthedistinctive a functional word orderof OJ nominalizedclauses noted by Yanagida (J. of East Asian of [nominal]v, as 2006). Inabilityto assignobjectcase is a property Linguistics, in and case marking Japanese.Syntaxand proposedby Miyagawa (Structure We discuss the diachronic Semantics,vol. 22, 1989). originsof the OJ active attested that it a out and exemplifies cross-linguistically point system alignment nominalizations. in that from clauses of non-accusative originate alignment pattern of S.-Y. Kuroda. This paperis dedicatedto thememory Y. Yanagida(13) of Tsukuba,Tsukuba305-8571,Japan of ModernLanguagesand Cultures, Institute University e-mail:[email protected] J.Whitman 203 MorrillHall, CornellUniversity, of Linguistics, Department Ithaca,NY 14853-4701,USA e-mail:[email protected] ^ This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Springer 102 Y. Yanagida,J.Whitman • Case • Keywords Activealignment• Ergativealignment• Splitintransitivity • • • NominalizationVerbalprefixes Cliticpronouns Nominalhierarchy 1 Introduction of thetwomajorclause typesin Old This paperdiscussesthesyntactic alignment whatwe label 'nominalized'clauses, 8th conclusive and (1) Japanese(OJ, century): in the adnominal examples (2). by represented (1) Conclusive: tamamokaru miy-u. Mi-watas-efcflamawotomyc-domo seaweed gatherappear-Conc (MY 17/3890)1 to be thescene,thefishermaidens 'When(I) surveyed gathering appeared seaweed.' look-cross-whenfishermaiden-Pi (2) Nominalized(adnominal): a. feSfcfct^tfcttM Saywopimyeno kwo ga t^iJ*^lJ9rSJ# pire puri-si (MY5/868) yama Gen child Agt scarf wave-Pst.Adnmountain Sayohime 'themountain whereSayohimewavedhercloth' b. *n«^*fi fjS#&{&HBft Wagimokwoga my.wife so £ft&i#&£ (MY 20/4357) swode mo sipoponi naki-si Agt sleeves even drenched cry-Pst.Adn fojmopayu. Foe long.for 'I longformywife,who criedso thatevenhersleevesweresopping.' c. ^*4g TftMJGIfc pisakwi0 opu-rw kiywoki kapara ni riverbankon catalpa grow-Adnclear 'on thebanksof theclearriverwherecatalpasgrow' (MY 6 /925) We arguethatwhileconclusiveclausesdisplaynominative-accusative alignment, also knownas activenominalized clauseshaveactivealignment. Inactivelanguages, verb stative(Klimov1974,1977;Mithun1991),thesole argument ofan intransitive showstwodistinct intransitive patterns: subjectspattern generally speaking, agentive 1 This andglossingconventions forOld Japanesein Frellesvig paperfollowsin generalthetranscription Ourdata andWhitman (2008); howeverwe glossinflectional endingsonlywhencrucialfortheargument. is takenfromthe Man'yoshu (My; compiledmid-eighth on Yoshimura's based primarily century), electronic textas well as theeditionsby Nakanishi(1978-1983),Kojimaet al. (1995) and Satakeet al. is attestedin phonocrucialforthe argument (2002). Examplesare citedonlywhenthe morpheme in simpletext. form(transcribed in italics);material is transcribed attested grammatic logogrammatically ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese alignment Syntactic 103 ones pattern withtransitive withtransitive subjects;non-agentive objects.We see in(2). In (2a-b) theexternal thatis,theagentofthetransitive sucha pattern argument, (2b) verbs,is marked (2a) andunergative bytheparticle ga. In(2c), thepatient subject verbin(2a): bothare verbbehavesliketheobjectofthetransitive oftheunaccusative bareandoccurimmediately adjacentto theverb. morphologically in therealis(izenkei) Othernominalizedclause typesincludeclauses inflected (3b), and nominalclausesin -(a)ku (3c). (3a), irrealis(mizenkei)conditionals (3) a. Realis (izenkei)conditional ura sipo mid ku. Wa ga wor-e-ba I Agt be-Rls-when bay tide be.full comes 'WhenI was presentthetidewas highin thebay.' b. Irrealis{mizenkei)conditional (MY 15/3707) masakikute imo ga ipap-a-ba wife Agt bless-Irs-if safely 'if youbless me godspeed' c. Y-aku Nominal form2 (MY 15/3583) ni tug-uraku wotome-ra ga ime maiden-Pi Agt dream in recount-Noml in mydream' 'whatthemaidensrecounted (MY 16/4011) of clausetypesin (3) sharetheactivealignment Each ofthenominalized properties withmarking of theexternalargument adnominalclausesin (2), beginning by ga. in As described clausesdisplayanother nominalized Transitive property. important detailby Yanagida(2006), whenthedirectobjectis markedwithaccusativewo, it in nominalized external clauses,as shownin (4): argument precedesthega-marked (4) #ts^ mm&$t Rnmufc tama nuku wotomye-ra ga pana tatibana wo thread-Adn maiden-s Agt bead orange blossom Obj made ni (MY 19/4166) until Loc 'untilthemaidensthreadtheorangeblossomson theirbeads' We develop an analysis of OJ nominalizedclauses that accountsfor the in (2) andthe[O wo S ga V] object oftheactivealignment co-occurrence properties inherent clauses Nominalized in 4. case, spelledoutas ga, to assign marking pattern in itsbase positionin Spec, vP. Followinga proposaldue to theexternalargument Miyagawa(1989), nominalizedverbalprojectionsfail to assignaccusativecase. 2 The term"nominal"fornominalized clausesin -aku followsWrona(2008). Wronashowsthat-aku clausesin EMJtexts. clauserolestakenon byadnominal nominalclausesfulfill manyofthesubordinate derivedfromtheadnominal, see Sect. 6. viewthatthenominalendingis historically For thestandard £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 104 Y. Yanagida,J.Whitman Two case licensingstrategies are availablefordirectobjects:theymaybe assigned abovevP,as in (4); ofa functional case,spelledoutas wo,inthespecifier projection if into the verb or, theyare non-branching, theymay undergoincorporation for also available is patientsubjects, (Yanagida2005,2007a,b).The secondstrategy in a limitedcontext,firstpointedout by Vovin(1997) and deas is wo-marking scribedin Sect. 5. Thepaperis organizedas follows.Section2 reviewspreviousanalyses.Section3 of active alignment, outlinesthe typologicalproperties emphasizingthatit has Section4 showsthat fromthebetterknownergativepattern. differences important in two specificdonominalizedclauses in OJ shareactivealignment properties of the and prefixalcross-referencing mains:case markingof subjectarguments on the verb.This sectionpresentsa formalanalysisof active subjectargument in OJ. Section5 focuseson objectmarkingin nominalizedclauses.In alignment in a broader Sect. 6 we discussthediachronicsourcesof theOJalignment pattern typologicalcontext.Section7 concludesthepaper. 2 Previousanalyses 2.1 Miyagawa(1989) and Miyagawaand Ekida (2003) Miyagawa(1989) proposesthatin OJandEarlyMiddleJapanese(EMJhenceforth), The adnominaland conclusiveclauses have distinctcase assigningmechanisms. case to theobjectin conclusiveformoftheverbis trulyverbalandassignsabstract andhas underlying objectpositionwhiletheadnominalformhas nominalproperties no case assigningability.In adnominalclauses,theobjectis assignedovertstructuralcase in the formof wo in orderto avoid a violationof the Case Filter. is statedin (3). Miyagawa's (1989) generalization (5) Miyagawa's generalization(1989, p. 206) case while AccusativeCase Assignment: The conclusiveformassignsabstract thecase assigningfeature oftheattributive (=adnominal)formmustbe manifestedovertly as wo. Giventhatovertobjectcase marking is normally requiredin modernJapanese, Miyagawa(1989) and Miyagawaand Ekida (2003) proposethatJapaneseunderwenta changefroman abstract to a morphological case marking languageandthat the drivingforceforthischangeis the increaseduse of the adnominalin main restricted clauses.In OJ,nominalized formsincludingtheadnominalare generally to embeddedenvironments and thecase forirrealisconditionals (thisis exclusively the -aku nominalizedform);the matrixuse of the attributive is predominantly limitedto thekakarimusubifocusconstruction (see Whitman1997 andreferences citedthere).The kakarimusubiconstruction, however,began to breakdown in EMJ(cf.Hendriks1998),withtheresultthatadnominalinflection cameto be used in mainclauses withouta kakari focusparticleand eventually replacedtheconclusivein mainclauses.Miyagawa(1989) and Miyagawaand Ekida (2003), based £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese Syntactic alignment 105 on an extensivesurveyofEMJliterary ofadnominal texts,arguethatthereanalysis formled to theincreaseduse ofobjectmarking as a mainclausepredicate withwo. 2.2 Kuroda(2007) difference betweenmodernandearlier Kuroda(2007) proposesthatthediachronic Parameter (cf.Kuroda1988):agreement Japaneseis accountedforbyan Agreement is forcedin earlierJapanesebutnotin themodernlanguage.In a forced-agreement is triggered and optionality features, by agreement-inducing language,movement does notcomeintoplay.KurodaproposesthatearlierJapanesewas a forced-agreeis obligatory, andbothsubjectand movement mentlanguage;heclaimsthatw/z/focus In modernJapanese,in contrast, case marking. whabstract objecttakeobligatory case marking is phrasesdo notmove,relativeclauseheadsneednotraise,andabstract ofOJarecompatible with characteristics optional.We willshowthatthealignment are forcedin OJnominalized theviewthatcertain,butnotall, typesof movement markedby wo obligatorily move out of VP. clauses. In particular, complements inOJ,is associated withw/z-movement movement, However,thisobligatory together in OJsyntax, clauses.The withthedomainofactivealignment namelynominalized and othertypesof optionality are allowedin wordorderflexibility characteristic conclusiveclauses,thedomainofaccusativealignment. 2.3 Previousnon-accusative analysesof Old Japanese 2.3.1 Vovin(1997) Vovin(1997) suggeststhatthesuffix -i, analyzedas a subjectcase marker by traand infactactivecase,marking ditionalgrammarians, subjectsoftransitive represents intransitive verbs.3Vovin verbsbutnotsubjectsof non-active of activeintransitive wo marksabsolutivecase,inthatwo appearsnot further arguesthatthecase marker verbsbutalso withthesubjectofstativepredicates, onlywiththeobjectoftransitive with-mi, calledby Vovin"qualitystativeverbs." in particular predicatessuffixed of -/and wo, VovinconcludesthatOJis a languagewith Based on thedistribution inmany ouranalysisofOJactivealignment activealignment. paperdiffers Although ofthehypothesis ofplaceas theoriginator fromhis,Vovindeservesprimacy respects is in important thatOJsyntactic respectsnon-accusative. alignment 2.3.2 Yanagida(2005, 2007a,b) byMiyagawa changedescribed Yanagida(2005,2007a,b)proposesthatthehistorical wellthecross-linguistically (1989) and Miyagawaand Ekida (2003) instantiates documented changefromsplitergativeto accusative.Yanagida's basicclaimis that thesplitoccurs Old Japaneseis an ergative-active languagewitha splitcase system; betweenmainandembeddedclauses,a typeidentified byDixon(1994,pp. 101-104). 3 Vovin thatOJ-i maybe a loanfromKorean.We (2005,pp. 111-116) revisesthisanalysis,suggesting to thispointin ourdiscussionof EMJin Sect. 6.3. return £l Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 106 Y. Yanagida,J.Whitman Main predicatesin the conclusive(shiishikei)formshow accusativealignment. conPredicatesin thenon-conclusive forms,i.e., irrealis(mizenkei)conditionals, tinuative(renyokei),adnominal(rentaikei),and realis (izenkei), show ergativeformwas a vestigial activealignment. Yanagida(2007a) proposesthattheadnominal thedemotedobjectoftheantipassive andwo wasanobliquecase marking antipassive was lostin Old Japbutwas reanalyzedas an accusativecase whentheantipassive counteranese.4A mainobjectiveofYanagida(2007a) is toexplaincertainapparent (5). Yanagida(2007b) showsthatin examplesto Miyagawa's(1989) generalization clauseswhosesubjectis markedby theMan'yoshu,thereare90 tokensoftransitive tokensoccurwith unmarked. no orga butwhoseobjectis morphologically Fifty-five adnominalpredicates, as in (2a), repeatedas (6) below. (6) mkimnnm vammm &m kwo ga pire puri-si yama Saywopimye no Gen child Act scarf wave-PAdn hill Sayohime 'thenameof thehillwhereSayohimewavedherscarf' (MY 5/868) to Miyagawa'sgeneralization. counter-examples Exampleslikethisareapparent However,Yanagidashowsthatwhilebareobjectsdo occurwithadnominalpredthebareobjectsare almost to thecounter-examples: icates,thereis a clearpattern N°s. Based on thesedistributional without facts,Yanagida exceptionnon-branching on the proposesthatbareobjectsin nominalizedclauseslike (6) are incorporated, model of languageslike Chukchee(Spencer 1999). Chukcheehas two typesof markedantipassiveand the constructions: a morphologically derivedintransitive in strategy. objectincorporation YanagidaarguesthatOJused objectincorporation nouns a similarway.Followingthebasic approachof Baker(1988), non-branching intotheverb,and immediately adjacentto an adnominalpredicateare incorporated satisfiesthe case requirements of the incorporee.This preserves incorporation is notassigned thattheobjectoftheadnominal Miyagawa'sgeneralization predicate is a abstractcase in its base position.5Note importantly thatobjectincorporation salientfeatureof languageswithactivealignment as observedby Klimov(1977, pp. 125-126); cf. also Sapir(1911). Inthispaper,weretainYanagida's (2007a,b) analysisofOJas a languageinvolving andin particular herincorporation splitalignment analysisofexampleslike(6). The incorporation analysisis dicussedin Sec. 4. Howeverwe do notretainthehypothesis thattheadnominal suffix is a vestigialantipassive, forthefollowing reasons. of antipassivein ergativelanguagesis to make the clause First,the function inassignment But ofabsolutivecase totheexternal [-transitive], resulting argument. are commonin strictly Antipassives ergativelanguages:transitive subjectsare markedby absolutive case andobjectsbyobliquecase. It is widelyclaimedthatthehistorical shiftfromergativeto accusative as accusativetransitives and Hale 1996). languagesresultsfromreanalysisof antipassives (e.g., Bittner 5 to thegeneralization thatdirectobjects 7) recognizesfourcounter-examples Miyagawa(1989,footnote in adnominalclausesare uniformly markedwithwo in OJ. Miyagawasuggeststhattheexamplesare noun-verb In fact,thenumber ofbareN° + verbexamplesin theMan'yoshiiis muchlarger, compounds. andthequantity andlexicalvariety oftheseexamplesindicatea productive processofnounincorporation ratherthanlexicalizedcompounds. ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese Syntactic alignment 107 does nothave thiseffectin OJ:theexternalargument adnominalmarking in transitiveadnominalclausesis markedwithergative(active)case (ga), notabsolutive. Second,if adnominalmarkingin relativeclauses was a kindof antipassive,we wouldexpectit to be associatedonlywithsubjectrelatives.This is becauseof the Restriction on A-Barextraction so-calledAbsolutive (Aldridge2004),whichallows toundergorelativization (recallthatthetransitive onlyabsolutivearguments subject ButtheOJadnominalis usedto formrelatives becomesabsolutivein antipassives). of all types(cf (2a), an adjunctrelative).Crucially,theadnominalis also used to deriveobjectrelatives: (7) miuz #a#s kwo [taratineno papa kap-u] ga Mk Gen motherAct breed-Adnsilkworm bredby mymother' 'thesilkworms (MY 12/2991) if theadnominalwas an antipassivemorpheme. This is unexpected The thirdproblemwith the antipassivehypothesisis morphological.Any foranalyzingthe adnominalas a vestigialantipassiveholds forthe arguments othernominalizedclause types as well since they all occur withg<z-marked agentivesubjectsand wo-markedobjectsin the orderOSV. But each of these affix:realis-e-/-ure-, irrealiscondidifferent clause typesinvolvesa completely tional-a-, nominal-aku. This meansthattherewould be no specificmorpheme markingthe antipassive.The fourthand finalproblemis semantic.In ergative languages,the object argumentin antipassives(if realized at all) is typically indefinite or non-referential (Dixon 1994, p. 148). But wo-marked objectsin OJ are typicallydefinite(Motohashi1989). Again, this fact is unexpectedif wo as theobliquecase markerforobjectsin antipassives. originated thatan activelanguageshouldhave an antipassiveis a In fact,theexpectation This view has is a subtypeof ergativealignment. that active the view of product literature sinceat leastDixon (1979, pp. 80-85), beenpervasivein thetypological in particular, Klimov(1974, butitis byno meanstheonlyview.Othertypologists, 1977,pp. 29^-3) have claimedthatthetwoalignment typesare distinct(see also Wichmann2008 and thepaperscollectedin Wichmannand Donohue2008). We arguein thenextsection,on empiricaland formalgrounds,thatthelatterview is thefeature[itransitive],whichplays a crucialrole in the correct.In particular, of assignment ergativecase and in the derivationof the antipassivein ergative languages,does notplaythisrolein activelanguages. of In the followingsection,we outlinethe typologicaland formalproperties to OJ. beforegoingon to showhow theseproperties activealignment apply 3 Activealignment S refersto thesubjectof an intransitive In Dixon's (1979) basic terminology, verb, verb. A to thesubjectof a transitive verb,and O to theobjectof a transitive £l Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Y. Yanagida,J.Whitman 108 (8) A Nominative ^ Ergative A _J S S Patient Absolutive Accusative O Agent O In nominative/accusative languages,A and S receive the same case marking while O is distinct (nominative) (accusative).In ergative/absolutive languages,S and O receivethe same marking(absolutive)while A is distinct(ergative).In subthemorphological encodingof intransitive languageswithactivealignment, Certain and their of on the semantic arguments.6 properties predicates jectsdepends intransitive subjectswhileothers subjectspatternwithA, thatis, withtransitive theseas SA and So. withobjects.Dixon (1979, p. 80) distinguishes pattern 3.1 Splitintransitivity verbsintoactiveandinactive.The exactlexical Activelanguagesdivideintransitive verbsare but the two classes of intransitive divisiondifferscross-linguistically, theagent activeintransitive subjects(SA, typically distinguished by case marking: have the same markingas transitive of unergatives) subjectswhereas argument of inactiveintransitive subjects(So, typicallythepatientargument unaccusatives) have thesame markingas transitive objects.In Hindi,forexample,verbsin per-ne appearson thesubjectof fectiveaspectshowan activepattern: thecase marker butnoton thesubjectof unaccusatives: transitives and unergatives (9) Hindi(Mahajan(1990)) a. Raam-ne kelaa khaayaa. Ram-Erg banana Ate 'Ram ate a banana.' b. Kutte bhONke. (ne) dogs (Erg) barked The dogsbarked.' c. Siitaa (*ne) aayii. Sita (*Erg) arrived 'Sita arrived.' Lotha (Tibeto-Burman) also showsactivealignment(Dahlstrom1983); the case maker-na marksSA, -co So. 6 the termshave been used to describeactivealignment. Van Valin (1990) introduces Manydifferent termsplitintransitivity; on Sapir's(1917) originalactive-stative, othersincludevariations agent-patient a wide S (Dixon 1979, 1994). Klimov(1977) correlates (cf. Klimov1977; Mithun1991),fluid-S/split traitswithactivealignment. varietyof lexicaland syntactic £l Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese alignment Syntactic (10) Lotha(Dahlstrom(1983)) cho. echo a. John-na firo ci hit Perf Det John-Subj dog 'Johnhitthedog.' oki na b. Mpo-na hapoi ci from outside Det house he-Subj 'He wentoutside(fromthehouse).' a wopan ciag-co c. Nkolo co I family Det-Subj ago long 'Long ago myfamilylivedin Wokka.' 109 yi go cho. Perf Wokka-e van cho. Wokka-Loc live Perf in themorphological case markedon nounsas can be manifested Activealignment head we have seen,but manyactivelanguagesare strictly marking:theymark a- cross-references on theverb.Thusin Gurarani withNP arguments SA; agreement se- cross-references So: (11) Guarani(Mithun1991) 'I am sick.' Se-rasi. A-xa. 'I go.' Se-ropehii. 'I am sleepy.' A-pua. 'I gotup.' themnow.' A-gweruaina. 'I am bringing 3.2 The nominalhierarchy has tendedto classifyactiveas a literature We notedin Sect.2 thatthetypological One argument againstthisview is thatactivelansubtypeof ergativealignment. cruciallyfromergativelanguageswithrespectto howergativesplits guagesdiffer withSilverstein's interact (1976) nominalhierarchy: (12) The NominalHierarchy (Silverstein1976) > propernouns > commonnouns pronouns human> animate> inanimate 1st>2nd>3rdperson showsplits,that Dixon(1979) emphasizesthatlanguagestermed ergativeinvariably featuresin certaincontexts.This interactswith the is, nominative/accusative thehierarchy to "roughly Dixon (1979, pp. 86-87) interprets nominalhierarchy. indicatethe overall 'agencypotential'of any givenNP" and observesthat"a numberof languageshave 'split' case markingexactlyon this principle:an end up to somepointin the 'ergative'case is used withNPs fromtheright-hand and an 'accusative'case fromthatpointon, overto the middleof thehierarchy extremeleft of the hierarchy."This is exemplifiedby ThulungRai (Tibeto-ka marksA whenwhenit is loweron Burman),an ergativelanguage.The suffix Lahaussois cited thehierarchy 2003). (Allen 1975, by (13) ThulungRai (Lahaussois2003) basi. s.l-mu a. Gui pe-pa.hal dish wash-Nom.infObi lpl eat-Npst.Prt 'We mustwashthedishes.' £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 110 Y. Yanagida,J.Whitman b. Gatsi mam-lai kr.m-.a l.-mu basi. 2d mother-Datvisit-Purp go-Nom.InfObi 'You twomustgo visitmother.' c. Gumimim-ka kam be-mri. work do-3p/3s.Pst 3p-Erg do work.' They d. I-lwak-ka khl.i. i-mam-lai 2Poss-mother-Dat 2Poss-y. sibling-Erg help.3s/3s 'Your youngersiblinghelpsyourmother.' In ThulungRai, first case whilethird and secondpersonA appearswithnominative and common NP A follows an person ergativepattern. A splitbetweenpronounsand nounsis also typicalof languageswithactive NPs in the butcrucially, thenominalhierarchy alignment, appliesto theargument direction as first and Dahlstrom First second (1983). opposite person, suggested by whichareat thetopofthehierarchy, whilecommonNPs are showactivemarking, less likelyto be marked. (14) Lakhota(Dahlstrom1983) a. Wa-lowa. 'I sing.' lsg.Ag-sing b. Ma-haska. 'I am tall.' lsg.Pat-betall c. Ma-ya-gnaya-pl. 'You pl. trickedme.' lsg.Pat-2Ag-trick-Pl (15) a. Lowa-pl. They sing.' sing-Pi b. Haska-pl. They(anim.)are tall.' be tall-Pl c. Ma-gnaya-pl. They trickedme.' lsg.Pat-trick-Pl d. Wicha-wa-gnaya. 'I trickedthem.' anim.3Pl.Acc-lsgAG -trick In Lakhota,the firstand secondpersonpronounswa and ma displayan active butthirdpersonpluralpi has a nominative-accusative distribution. Indepattern, withmorphological As Mithun cases noradpositions. pendentNPs appearneither restricted to (1991) pointsout, case systemsbased on agency are frequently nominalsreferring tohumanbeings.7 ThusKoasatishowsagentivecase marking on pronominalprefixeswithinverbs but accusativecase markingon nouns.The active systemin Batsbi (Tsova-Tush)is limitedto firstand second persons. 7 Mithun thesemanticbasis of theactivemarkingof variousWestHemisphere lan(1991) identifies and diachronically. guages,bothsynchronically ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Syntacticalignmentin old Japanese 11 1 to humansonly.The CentralPornohas an active systemin nominalsreferring to humanbeings.The Yuki systemis reGeorgianactive systemis restricted the implication strictedto animates.From these cross-linguistic observations, side to therightis used withNPs fromtheleft-hand followsthatactivemarking thatis, if a languagehas agentmarkingin handside of the nominalhierarchy; thirdperson,it also has agentmarkingin firstand secondperson.This is exactly theoppositeof theright-to-left proposedby Dixon for applicationof thehierarchy betweenactive markingand the nominal ergativelanguages.The relationship hierachyis statedin (16): (16) The ActiveMarkingHierarchy In activelanguages,if activemarking appliesto an NP typea, it applies to everyNP typeto theleftof oton thenominalhierarchy. of activecase is dependent not discussionshowsthatassignment The preceding on on the of S the nominal the verb but role thematic place assignedby juston the Klimov (1974, 1979) emphasizesthis point,stressingthatin active hierarchy. languagesthesemanticsof boththepredicateand thesubjectNP governthedisof activecase. tribution Dixon (1979, pp. 80-83) dividesactivelanguagesintotwo groups;"splitS" and "fluidS" langaugessuchas Batsbi.In split-S languagessuchas Tupi-Guarani and whether verbshave fixedmembership, of intransitive thetwoclasses systems, on their class is based inactive or meaning.In prototypical theybelongto theactive of each on the divided fluidS systems,verbsare meaning particular depending has controlovertheactivity, token.The activepattern appearswhentheS argument andtheinactivepattern appearswhencontrolis lacking.ConsiderBatsbi,a fluidS languagecitedby (Comrie1978,p. 366). (17) Caucasian Batsbi:Northeast naizdrax Txo a. qitra. fell to-theground we-Abs the fell to 'We ground(unintentionally).' naizdrax b. Atxo qitra. fell to-theground we-Erg the 'We fellto ground(intentionally).' and thesubjectis markedabsolutivewhile is unintentional, In (17a) theactivity and thesubjectis markedergative/active. in (18b) theactivityinvolvesintention, of active of SA markingcan varyalong three the distribution Summarizing, theprototypical dimensions: meaningof theverb(whetherit is agentiveor nonandtheplaceofS on thedegreeofcontrolassociatedwiththeS argument, agentive), that thenominalhierarchy. Legate(2008) providesa framework can capturethese the In Legate'sframework, activefromergativesystems. anddistinguish properties externalargumentin ergativelanguagesreceivesinherentergativecase in its vPs. [Theanalysisof ergativeas of [+transitive] positionin thespecifier underlying £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 112 Y. Yanagida,J.Whitman inherent withWoolford case assignedto theexternalargument in situ originates as Aldridge(2004)]. In activelanguages, (1997) and is sharedby suchresearchers Activecase is assignedto theexternal transitivity playsno role:inherent argument We proposethatother in Spec, vP regardless or notv is [+transitive]. of whether of inherent activecase, including featuresmayalso play a role in theassignment personfeaturesand semanticfeaturessuch as [ianimate]. This allows us to of active account for languages where NP type determinesthe distribution case. 4 Evidenceforactivealignmentin Old Japanese in Old JapanesenomiIn thissection,we presentevidenceforactivealignment and verbalpreflxation. nalizedclauses,focusingon subjectcase marking 4.1 Agentmarking withga becauseitmarks case marker In modernTokyoJapanese, ga is clearlya nominative of intransitives, boththeexternalargument of transitives andtheinternal argument as in (18). (18) a. Taroo ga naita Nom Taroo cried Taroo cried.' b. Hana ga saita flower Nom bloomed 'Flowersbloomed.' c. Taroo ga hon o katta Taroo Nom book Ace bought Taroo boughta book.' The distribution of ga in OJ differssignificantly frompresent-day Japanese.Ga in OJ is one of two genitivemarkers; theotheris no, whichretainsthisstatusin modernJapanese.In additionto markingpossessorsof NP insideDP, bothga and no also markthe subjectsof nominalizedclauses. Ga is restricted to personal nounswhosereferent is someoneclose to thespeaker,suchas imo 'sister, wife,lover',or a pronounwitha specifichumanreferent. No, on theotherhand, is used withnonspecific animatenouns,such as pito 'otherpeople', and with inanimatenouns.8The use of ga dependsnotonlyon thesemanticsof theDP it marksbut also on the semanticsof the predicate.In nominalizedclauses,ga is 8 Thereare a few nounssuchas pi 'thesun' or animalsof examplesin whichspecificbutnonhuman suchas tadu 'crane'andsiwa 'snipe'aremarkedwithga. Thesearealmostcertainly specialsignificance a prominent rhetorical devicein theMan'ydshu. examplesof personification, ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese Syntactic alignment 113 and transitives used withactiveintransitives (19) whileno is used withinactive intransitives (20).9'10 (19) a. tk^sjg^s^ pt&mmm PitO'dumakoro wo iki ni waga sum. Obj longfor I.Actdo-Adn personwife 'I longforanotherperson'swife.' (MY 14/3539) kimi ga yuk-u miti lord Act go-Adn road 'theroadthatmylordtravels' (MY 15/3724) b. SaSA.it c. femmmmA itn^mm «*... yama Saywopimye no kwo ga pire puri-si Gen child Act scarf wave-Pst.Adnmountain Sayohime (MY 5/868) whereSayohimewavedherscarf 'themountain (20) a. JRA75 k»§JLffi no Yoki pito yosi to yoku mite well looking Gen Comp good good people Yosino (MY 1/27) ipisi yosi to Yoshino good Comp say-Pst.Adn 'Yoshino,whichgood peopletooka good look at and called good, said was good' b. izm fexm tukwi no saku pana flower Gen bloom month 'themonthwhenflowersare in bloom' (MY 18/4066) of the The firstand secondpronounswa and na are weakpronominal counterparts have the weak These and ware nare, pronouns respectively. strongpronouns marked with and are of clitics: adjacent appearstrictly ga they invariably properties 9 Stativeverbssuchas wori 'be sit' andunaccusative verbssuchas ku 'come' appearwithga when at, whichare rankedhigheston thenominalhierarchy. or secondpersonpronominal, thesubjectis a first (i) a. *skift £Hfi ftJftfe^ medurasiki kimiga ki-mas-aba Charming lordAct come-Hon-if lordcomes' 'if mycharming b. *n»¥*L» *&ZU (MY 18/4050) mw X miti ku wa ga wor-eba ura sipo be.full comes Actbe-when bay tide I 'WhenI am therethetidewill be highin thebay.' 10 Notethat (19a) is an examplefromEasternOld Japanese. £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 114 Y. Yanagida,J.Whitman to theverb.11This contrasts withfullNP subjectsmarkedby ga, whichallow an adverborevena clauseto freelyappearbetweenthesubjectandtheverb,as shown in (21). (21) a. ^SM&ft P^hMM^X wotomye-ra ga ime ni tugur-aku maiden-Pi Act dream-in tell-Nomnl 'The maidenstoldme in a dream.' b. *n®€&3£ wagimokwo my.wife ga Act m£&^&\£Z mat-amu wait-will to Comp 'whenmywifesaid thatshe wouldwait' (MY 17/4011) ^ft toki ipi-si say-Pst.Adn time (MY 15/3701,3713) are Given the strictadjacencycondition,we assumethatthe weak pronominals cliticsadjoineddirectly to theverb. We seefromthesemantic differences betweenga andno thatga marksNPs higher on thenominalhierarchy in (12) whereasno is usedwithNPs locatedloweron the This generalization hierarchy. appliesbothto ga and no as possessorsof NP and between ofsubjects(A andS) in nominalized clauses.Notethatsyncretism markers we For in common non-accusative and case is languages. clarity, genitive agentive and as it a nominalized clause as when marks the of Gen(itive) SA glossga Act(ive) whenitmarksthepossessorin DP. No is glossedas Gen(itive)throughout. Now considertheOJexamplesbelow: (22) a. mn&X i!75^«^ Kimi ga yuk-u miti no nagate Lord Act go-And road Gen length 'thelengthof theroadmylordtravels' b. hjb^m (MY 15/3724) mm¥^5L wo paya-ra/ (MY 11/2713) Asuka-gapa0 yuk-u se Asukariver shallows go-Adn Obj fast-Mi 'since theshallowswheretheAsukaRiverflowsare fast' In (22a), thepredicateyuku means 'go.' Its subjectis humanand volitionaland markedwithga. In (22b) yuku means 'flow.'Its inanimatesubjectAsuka-gapa 'Asukariver'is morphologically unmarked. The choiceof subjectmarking depends on whether theeventdenotedby theverbinvolvescontrolor intention: onlythe humanparticipant exercisescontrol.ThusOJnominalclausesare a fluid-Ssystem, in Dixon's sensedescribedin Sect. 3.1. 11 120 occurrencesof subject(w)a-ga are foundin the Man'ydshii(based on the Approximately Yoshimura'selectronic text);all are immediately adjacentto theverb.(Data citedhereincludepersonal withphonographs butnotthefreestanding n",whichcan be read withor pronounswritten ideograph without a case particle.) £l Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese Syntactic alignment 115 4.2 Ga/nomarking and nominalized clause types as Subjectsmarkedwithga andno appearin theclausetypeswe havecharacterized nominalized:adnominal(rentaikei),realis (izenkei), irrealis(mizenkei) condibut theyneverappearwithpredicatesin the tionals,and -aku nominalizations, Let us look morecloselyat theevidencethattheseclausetypes conclusiveform.12 at theOJperiod.First,as describedabove, statussynchronically havenominalized theirsubjectsappearwiththegenitivecase particlesga and no. The semanticdisthepossessorinDPs is paralleltotheirdistribution tribution ofga andno inmarking clauses:possessorsloweron thenominalhierarchy in nominalized appearwithno suchas personalpronouns, whileNPs higheron thehierarchy, appearwithga: (23) a. *g75^W75 feflUH Nara no miyakwo no Sapo Gen Saho Nara Gen capital 'theriverSahokawain Nara.' b. OTS3P8K kawa river ®%$L%^M tamoto tur-u imo ga wakayu young.sweetfishangle-Adn girl Gen wrist 'thearmsof mygirlwho fishesforyoungsweetfish' (MY 1/79) (MY 5/857) clauses: Second,thefourclausetypesappearin positionstypicalof nominalized and modifier embeddedcomplement including positionsand focusconstructions, in East and questions.Whitman(1997) shows thatit is common,particularly tobe realizedwithnominalizing Southeast Asia,forfocusandinterrogative patterns on thepredicate.Adnominalclausesappearas focusand interrogative morphology oras thesubjectofa clause.Realisclauses as theobjectofa postposition questions, withtheparticlekoso andas presupposed appearas questionsas focusconstructions conditionalstypicallyfollowedby the particles-ba 'as/since'and -do 'even (though)'.Irrealisconditionalsappearwiththe same two particles.Nominalized occur -aku clausesare analyzedas nominalizations byWrona(2008) andtypically above are embedded,all comin complement position.All of the environments of NP, exceptforthefocusand question of a verbor particleor modifier plements associatedwiththeadnominaland realis. constructions involvesa [nominal]feature In thispaperwe adopttheviewthatnominalization toT, theheadoftheextendedverbal associatedwiththelexicalverbandpercolated 12 Sasaki in whichhe claimsthatga appearswitha (1996) citessevenexamplesfromtheMan'yoshxx Z. sincethischaracter predicatein theconclusiveform.(We excludeexamplesinvolvingthecharacter canbe readeitheras thecase marker ga /noorthefocusadverbialsi.) (21b) aboveis one oftheexamples of his data revealsthatin six out of Sasaki's sevenexamples,as in citedby Sasaki; closerinspection (21b), thesubjectis notin theconclusiveto-clausebutin thehigherclause whosepredicateis in the ofthesesixclauses,as in(21b),is [Subject-gaj form.The structure adnominal [proi...Vconc]Vadn], where nullpro coindexedwiththega-markedsubjectin thehigher theembeddedsubjectis a phonologically adnominalclause. £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 116 Y. Yanagida,J.Whitman 13 in OJis TP is shownbythe projection. The factthatthedomainofnominalization of the and conditional irrealis adnominal,realis, endingsto selecttense, ability and modals as -mu. In such presumptive (2a-b), (20a), and (21), for negation, form of the we see examplesoftheadnominal instance, spelled pasttenseauxiliary, in out as -si. The 'high' locus of nominalization OJ bringsto mindlanguages is at the clausal level.14The like Turkish(Kornfllt 2003), wherenominalization concreterepresentation we proposeforOJ nominalizedclauses withgtf-marked to (21b). subjectsis givenin (24), corresponding TP (24) (=21b) /\ vp wagimokwogav' Act /\ my.wife VP T [nominal] Past v [nominal] ipi say in activecase (spelledoutas ga) to external [Nominal]v assignsinherent arguments as and unergatives its specifier.Ga thusappearson the subjectsof transitives describedin Sect. 4.1. In addition,inherent ga is subjectto additionalfeatural in Sect. 3.2, such as the as discussed restrictions of active typical languages restriction DP be [animate]. thattheactive-marked In contrast, case, assignedby D in DPs suchas (23a). genitiveno is a structural We assumethatD is also responsible forassigning no tothesubjectsofnominalized clausessuchas thosein (20), muchas in modernJapanese(see Miyagawa1993for an analysisofD as thelicenserof/20-marked genitivesubjectsin modernJapanese). The mechanism of subject/to-marking is discussedin greaterdetailin Sect.4.4. In thissectionwe havedescribeda dependent in OJcharaceristic pattern marking In thenext of activealignment: of A in nominalized clauses. and SA ga marking of sectionwe show thatOJ also displayedhead markingpatterns characteristic activealignment. 13 This contrasts withapproachesthatposita category-fixing head n thatselectsan acategorialroot inOJ,butthecategory-fixing (Marantz1997)orvP (Alexiadou2001). Suchan approachis notimpossible head wouldhave to selectT. is higherthanin OJ:Kornfilt However,in Turkishthelocusof nominalization (2003) placesit in the is spelledoutabove theTense-AspectModal projections. Agror Finitehead wheresubjectagreement in thisposition,noris thereanyovertFiniteorC OJ,likemodernJ,has no overtagreement morphology above theTAM (Tense/Modal/Aspect) etc. auxiliaries.Insteadwhatwe findare adnominal, morpheme of theseauxiliaries.For thisreasonwe interpret forms theadnominaland othernominalized allomorphs as thespelloutof [nominal]T. ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese alignment Syntactic 117 4.3 Active/inactive prefixes unnoticed A heretofore ofOJ completely pieceofevidencefortheactivealignment /-andsa-. Japanesetraditional havefailed comesfromtheverbalprefixes linguists function fortheseprefixes. a consistent semanticor syntactic Butcareful to identify analysisshowsthat/-is attachedto activeverbs,and sa- to inactiveverbs.These in nominalized, as wellas infinitive, clauses. twoprefixes appearalmostexclusively 4.3.1 i- on activeverbs The prefixi- is richlyattestedin theMan'yoshu,as in (25a-b). (25)a. mntMn fewwfc fi*sm kawa ni i-yuki itarite River-Loc i-go reaching (MY 1/79) 'I reachedtheRiverSahokawain Nara.' Nara no miyakwo no Sapo Gen Saho Nara Gen capital Kume no wakugwo ga i-pure-kyem-u Act i-touch-PConj-Adn Kume Gen youth iswo no kusa no ne rock Gen grass Gen root (MY 1/435) 'therootof thegrassthattheyouthof Kumewouldhave touched.' of of /-are foundin theMan'yoshu. The distribution A totalof 74 occurrences /-parallelsthatof the case markerga: bothappearin nominalizedclauses,i.e., irrealis(mizenkei) conditionalsand -aku nominal,adnominal(rentaikei),and realis(izenkei)clauses. dataforprefixi-15 (26) Quantitative Total Conclusive Imperative Irrealis Realis AdnominalInfinitive (Mizen) (Izen) (Rentai) (Renyo) (Shushi) (Meirei) 3 5 19 44 (2) (1) 74 The prefix/-attachesto activeverbs(all tokensofi- in theMan'yoshuarecitedin to theunerYanagida2007b). Thereare a numberof cases in which/-is prefixed to theunaccusative gativeverbyuku 'go' butno examplesin which/-is prefixed towhatappearto verbkuru'come'.Therearea fewexamplesinwhichi- is prefixed be nonagentive verbs,suchas (27). 15 The on thetotalsforconclusiveand imperative examplesof i- indicatethatall threeof parentheses theseexamplesare subjectto alternative analyses,as discussedbelow. £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 118 Y. Yanagida,J.Whitman (27) 3ft75 tlj. . . &&ffi\h7b ihm Miwa no yama. . . Nara no yama no yamano ma ni Miwa Gen mountainNara Gen mountainGen mountain among OTlTTft itP§ PfeWjftfc (MY 1/17) i-kakur-u made miti no ni kuma i-tumor-u made Loc i-hide-Adnuntil road Gen bend i-amass-Adn until 'Mt. Miwa. . . untilyouhideyourself of the mountains Nara, among untilyou loomin thebendsof theroad' wemight as inactivesincethesubjectMiwa-no Although expect(27) tobe interpreted is 4Mt Miwa' as personificathe inanimate, clause is interpreted yama superficially tionalby all Japanesecommentators.16 The use of /-herethusfitswithourcharacterizationof OJ as a fluid-S language in the previous section: ostensibly verbsmayappearwithactivemarking whentheyhavehuman(or pernonagentive sonified)subjects. Unlikeactivemarking ga, i- also appearsin infinitive (renyokei)clauses.Butin infinitive clauses too, theprefixedverbis unfailingly activein all of the clearly to a verbin theinfinitive, interpretable examples.Of the44 examplesof /-prefixed 18 involvetheunergative verbyuk-'go'. The overwhelming ofz+infinitive majority clauseshave agentiveempty(pro) subjects.17 In additionto beingrestricted to activeverbs,we see fromthetablein (23) that /-occursalmostexclusively withtheclausetypeswe haveidentified as nominalized or in infinitive clauses withagentivepro subjects.None of the threeepotential to thisgeneralization withphonograms. are written counter-examples Kojimaet al. vol. the 3, p. 369) interpret singlepotential (1995, imperative example(MYS 3169) as not involvingprefixal/-but ratherthe honorific verbof displacement imasIn both of the conclusive 'go/come(Honorific)'.18 fact, potential examples,MYS 1916 and 3885, are open to this same interpretation, as bothinvolvehonorific in Chinesecharacters. If this subjectsand a verbwiththe meaning'go' written is there are no of /with or interpretationcorrect, examples occurring imperative conclusivepredicates. 16 Wrona as a counter-example to the (2006) citesthesecondclause of (27) mitino kuma i-tumor-u that/-appearsonlyon activeverbs,interpreting thisclauseas 'bendsoftheroadpile up'. generalization Thisinterpretation is also followedbyKojimaetal. (1995) andSatakeetal. (2002). ButNakanishi(1978/ 'Mt. Miwa' as thesubjectof bothclauses.Because thispreserves the 2004) interprets personificational evidentparallelismof thetwoclauses,we have followedNakanishi' s interpretation here. 17 To be precise,40 of the44 infinitive exampleshave humanagentivepro subjects.Two have personificational subjects,shirakum(w)omo 'whiteclouds too' (MY 317) and amakum(w)o mo 'skycloudstoo' (MY 319), bothoccurring withunergative i-YUKI 7+going\BothNPs aremarkedwiththe subduedfocusmarker mo 'too/even'suggesting thattheyareexternalto theinfinitive clause.Onlytwo exampleshavepossibleclause-internal non-agentive subjects,butbothof these(MY 2145 and3409) are ofinterpretation. MY 2145 is particularly instructive. problematic Kojimaetal. (1995,vol.3, p. Ill) note thattheinfinitive clausein questionsa-wosikano kowe i-tukii-tuki'thevoiceofthebuck/-continuing, mustbe interpreted in contextas an ellipticalrealis(izenkei)conditional: 'whenpro hears /-continuing' thevoice of thebuck'. 18 Satakeet al. (2002, vol .3, p. 205) also acknowledgethisinterpretation. £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese Syntactic alignment 119 to activeverbs.It occursonly Summingup,theOJverbalprefix/-is restricted - thedomainwe have associatedwithergative-active withnominalizedpredicates - and infinitives withagentivepro subjects. alignment 4.3.2 sa- on Inactiveverbs The prefixsa- differs cruciallyfromi- in thatit appearsonlyon inactiveverbs, as in (28). (28) a. &&&9m ^i^MMM^ taka pa nak-ye-mu to (MY 17/4011) sa-narap-yer-u falcon Top cry-Pst-Presum sa-be.tamed-Perf-And Comp 'thatthetamedfalconswouldhave cried' b. &fezmMfc mmn\ALi& ni pa ide-tati sa-nQ-si tumaya ni asita sa-sleep-Pst.Adnbedroom in morningin Top leaving (MY 3/481) sinopi remembering leavingthebedroomwhere(I) slept' 'remembering, c. mnmm s*i opo kimi sa-niturap-u wa ga Gen great lord sa-shine-Adn I 'mygreatlordwho shines' d. mm* (MY 3/420) &¥mmm sugwi no nwo ni sa-wodor-u kigisi cedar Gen field in sa-dance-Adn pheasant field' 'thepheasantthatdancesin thecedar-covered e. mWiKtiL ¥&T'hM (MY 19/4148) kwo sa-basir-i ni pa ayu kapa se river shallow in Top sweetfish fry sa-run-Inf in therivershallows' 'theyoungsweetfish running (MY3/475) Thereare 30 tokensof theprefixsa- on verbs,includingneru 'sleep', niturapu 'shine',pasiru '(fish)run',wodoru'(birds)dance',wataru'(toads)cross',nebapu 'spreadroots',narabu '(birds)line up', kumoru 'get cloudy',nituku'get redandall havenon-agentive dened'.All theverbsareintransitive, subjects(asidefrom ne- 'sleep', all are nonhuman). dataforprefixsa(29) Quantitative Realis -aku AdnominalInfinitive ConclusiveImperativeTotal (Shushi) (Meirei) (Rentai) (Renyo) (Izen) 2 2 14 7 3 2 30 ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 120 Y. Yanagida,J.Whitman Like /-,the prefixsa- is used overwhelmingly (25/30 tokens)in nominalized 19 as in sa-yo 'night' clauses. Sa- also occursin theMan'yoshu as a nounprefix, while/-does not.This parallelsexactlythedistribution of agreement prefixesin inactive activelanguagessuch as Satere-Mawe (Meira 2006): prefixesoccuron verbs on occur active nounsand inactiveverbswhileactiveprefixes only.20'21 AlexanderVovin (p.c.) pointsout to us thatone verbin OJ,wataru 'cross', appearswitheither/-orsa-. Therearefourexamplesofi-watar-in theMan'yoshu (MY 1742,2081, 4101, and 4126), and six examplesof sa-watar-(MY 800, 971, 1960, 1976,2450, and 2804). The S of i-watar-is [+human]and volitionalin all fourexamples:'theyoungwoman,''Tanabata' (Vega, theweaverstar),'thefisherfolk,'and 'Vega and Altair.'The S of sa-watar-is [-human]in all six examples: 'toads' (800, 971), 'a cuckoo' (1960, 1976), 'themoon,''a teal'. Typicalexamples of each pattern are givenin (30).22 &^ff^1tt&& (30) a. £jfrt&ft& ama sky gtfcf&S# mo watasera-ba sono pe no gawa past yu from too over that Gen river bridge span-if Pffl£&fe^-?- (MY 18/4126) wo i-watar-as-am-u Conj /-cross-Hon-Prop-Adn 'thoughif one puta bridgeacrosstheMilkyWay,(they=Vegaand overon that' Altair)would/-cross b. mWiji #EttJ!75 J5HB*S> kumo ma ywori sa-wataru tukwi no opoposiku cloud among from s^-cross moon Gen faintly fflKT ^ (MY 15 /2450) misi kwo api saw child join likethemoon^-crossingfromamongtheclouds' 'thegirlI saw faintly I-watar-'cross (overthebridge)'is agentivevolitional,and telic,a stereotypical actionbutthe nota completed activeverb.Sa-watar-is non-agentive anddesignates 19 Threeof thefive involvene- 'sleep' withhumansubjects:conclusive(MY 2782) counter-examples and two withimperative (MY 636, 2629). Since sa-ne 'sfl+sleeping'also occursas a noun,these conclusiveMY 859 and4 156, The remaining twocounter-examples, examplesmaybe backformations. bothinvolvethecollocationayu sa-basiru'thesweetfish sfl-runs'. 20 Satere-Mawe on theverb(cf. (Tupian)has an activesystemmarkedbytwoseriesofpersonalprefixes similarto (possessed) Mithun1991). Meira(2006) showsthatin Mawe nonactiveverbsare strikingly do not nouns:thesame set of personalprefixesappearson nounsand nonactiveverbs;theseprefixes selectactiveverbs. 21 On bothnounsand verbssaon the rendaku(realizedin OJas prenasalization) (butnot/-)triggers sourceof theshape*saN(V). *Sa maybe voicelessobstruent. This suggestsan etymological following relatedtothemesialpronouns sa 'thus',so 'that',andsi 's/heit' while*N(V) appearsrelatedtogenitive/ inactiveno. 22 wedescribehere on(30b),Kojimaetal. ( 1995,vol.3,p. 19 1) observeexactlythedistinction Commenting sa-watar-is betweeni-watarandsa-watar-.Theynotethatwhilei-wataroccursonlywithhumansubjects, to otherverbs,however. restricted tononhuman subjects.Theyfailtoextendthisdistinction £i Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese alignment Syntactic 12 1 moonpassingbeforethe speaker'sview,in otherwords,a stereotypical inactive predicate. We have shownin thissectionthatOJ nominalizedclauses show notjust an activecase markingsystembut also at least thevestigesof activeprefixalmorbetweenactiveand inactiveverbs phology.Thisis evidencedby a strictopposition in bothdependent and head marking of activeS by strategies. Dependentmarking thecase markerga is in oppositionto markingof inactiveS by no (and, as we of activepredicatesby theprefix discussin thenextsection,zero). Head marking /-is in oppositionto theprefixsa- on inactivepredicates.23 4.4 Markingof InactiveSubjects surfacesin nominalizedclauses but notin As discussedabove, activealignment mainclauseswhosepredicatetakesconclusiveform.A main/nominalized splitis of also attested clauses,theinternal argument subjects.In nominalized byunmarked unaccusative verbscan be unmarked while,as we saw in Sect .3.3, morphologically verbsis typically markedby Active of transitive/unergative theexternalargument ga. In almostall cases, barethemesubjectsof nominalizedclauses appearimmedata,see Yanagida2007b). Examples diatelyadjacentto theverb(forquantitative are givenin (31). (3D a. x^n mnw.fc ni kiywoki kapara pisakwi0 opu-ra on clear riverbank grow-Adn catalpa 'on thebanksof theclearriverwherecatalpasgrow' (MY 6 /925) pisakata no pana 0 tir-u waga sono ni ume no I Gen garden in plum Gen blossom fall-Adn Epithet Gen ante ywori from sky 'in mygardenplumblossomsfallfromthesky' (MY 5/822) verbsare limitedto nonIn Sect. 5 we show thatobjectsadjacentto transitive The N°s and are thusanalyzableas havingundergoneincorporation. branching 23 An sais foundin MY 804, wherega and prefixal to thisgeneralization apparentcounter-example appearto surfacein thesameclause: (i) il^Mfa te2PttH?£4 ^SSrtt&ft wotomye-raga sa-nasu itado wo osipirak-i maiden-Pi Act sa-sleep door Obj pushopen-Inf 'pushingopenthedoorwherethemaidenssleep.' ' possessorofitado wotomyera interpret ga 'maidensGA' as thegenitive Kojimaetal. (1972),however, (wooden) door', a metonymic expressionfor 'bedroom';the entireNP thenhas the interpretation in (ii): 'pushingopenthemaiden's(bedroom)doorwheretheysleep' and thestructure wo sa-nasu ] itado]] osipirak-i (ii) [NPwotomyera ga [[ pro Gen maidens pushopen-Cont stf-sleep door Obj On thisinterpretation wotmyera ga is nottheclausematesubjectofsa-nasu 'sra-sleep1. £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 122 Y. Yanagida,J.Whitman themesubjects,suchas incorporation analysiscan be extendedto non-branching themesubjectsalso occurin thispopisakwi 'catalpa' in (31a). But branching sition,as shownby ume no pana 'blossomof theplum' in (31b). This indicates thatunaccusativesubjectshave a licensingoptionnot available for transitive objects. in Recall thatin Sect. 4.2 we describedtwogenitivesubjectmarking strategies nominal foragentivesubjectshighon the nominalizedclauses:activegfl-marking licensedbyD, muchas in adnominalclausesin andgenitiveJio-marking hierarchy; modernJapanese(Harada 1971; Miyagawa1993). A thirdoption,exemplified by thatremainin VP. Notethatthis (31b),is availableforbaresubjectsofunacusatives thirdoptioncannotinvolvean 'absolutive'case because absolutiveshouldbe availableforbothSo (unaccusativesubjects)and O (transitive objects),but,as mentionedabove and describedin moredetail in Sect. 5, branchingtransitive bare position.Notealso thattheVP-internal objectsdo notappearintheVP-internal (catalpas,plumblossoms)whilesubjectsmarked subjectsin (31) are nonspecific, as in (30b) (the or specific, as in (20b) (flowers), withno maybe eithernonspecific, moon). These factssuggestthatexampleslike (31) involvean impersonalconwiththebare themesubjectlicensedin situ insidetheVP. Impersonal struction, constructions requirea mechanismforassigningnominativecase to the theme clausemaybeara case feature subjectin situ.We proposethatT in OJnominalized whenT selects'defective'v, thatis, v circumstances: butonlyin veryrestricted ofitsown(Chomsky2001). On thisview,the anda case feature lackinga specifier barethemesubjectin (31b) is assignedcase by T in situ. the threecase markingstrategiesfor subjectsof nominalized Summarizing, clausesare shownin (32).24 a. (32) v VP DP0 V TP /\ v b. T c. vP /\ DPga v VP v DP /\ TP D ...DP no... V themesubjectsin situare assignedcase by T selectinga defectivevP Nonspecific Inherent (32a). ga is assignedto activesubjectsin Spec, vP (32b). Genitiveno is D to thatspecificthemesubjects assignedby subjectselsewhere.On theassumption out of VP move the (Diesing1992),thisexplainswhyspecificthemesubjectssuch 24 forsubjectNPs pattern Miyamotoet al. (1999) reportthatJapanesechildrenshowa case marking of whatwe have describedforOJ. They observethatchildrencommonlyomit highlyreminiscent nominative verbswhileconsistently usingga forsubjectsof transitives ga forsubjectsof unaccusative and unergatives. (ACDH) (Borerand Wexler1987) TheyproposethattheA-chainDeficitHypothesis Froma and unergatives. accountsfor why childrentreatunaccusativesdifferently fromtransitives it may be worthpursuinga unifiedaccountfor this parallelbetweenthe learnability perspective, acquisitionand syntactic change. ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Syntacticalignmentin old Japanese 123 25Underthis thanbaremarking. as 'themoon'in (30b) receiveno rather analysis, is licensedbya higherhead,outsidetheactivealignment Aio-marking systemof OJ clauses. nominalized subjectsoccurfreelybothas the Turningnow to conclusiveclauses,unmarked of an unaccusative, of a transitive and theinternal as externalargument argument shownin (33). (33) a. Mmmtii& fmfeimm&ffl Ume no pana 0 ima sakari nar-i. Plum Gen blossom now at.peak be-Conc The plumblossomsare now at theirpeak.' b. Rgtm ^m^m-k^^ ^jgginMjis karu Miwatase-fetf amawotomye-domo 0 tama mo 0 Look cross-when fishermaiden-Pi pearl seaweed gather (MY 17/3890) miy-u. appear-Conc 'WhenI look around,thefishermaidens appearto be gathering pearly seaweed. to nominalized markedby ga and restricted Whilecliticpronounsare uniformly clauses as shownin Sect. 3.3, strongpronounsin subjectpositionare unmarked and neverappearwithga. morphologically (34) a. ^^W ware 0 kusa 0 tor-er-i weed take-Perf-Conc I ' I am pickingweeds.' b. £#3Pftit& nT^^fctiS* (MY 10/1943) £*Uft£fi£ no kamwi wo kopitutu are 0 Ametuti I Ace pray heaven.earthGen god mat-am-u. wait-Presum-Conc (MY 15/3682) I will wait.' and of heaven the to earth, gods 'Praying Bare objectsas in (33b) in conclusiveclausescan be analyzedas receivingstructuralaccusativecase fromv, as proposedby Miyagawa(1989). Bare subjectsas in case fromT, as in modernJapanese (33a) can be analyzedas receivingnominative and accusative While the exponenceofnominative (Takezawa1987). phonological in OJconclusiveclausesare forcase marking mechanisms thesyntactic case differs, between thesameas in modernJapanese.However,thesharpdifference essentially thatOJwas a language clausesconfirms in conclusiveandnominalized thepatterns with split alignment:while conclusiveclauses follow a nominative-accusative S/A and O, nominalizedclauses show an activepattern, distinguishing pattern, and So and groupingSA withA. In thenextsectionwe turnto SA distinguishing in clauses. nominalized objectmarking 25 We leave the precise landing site of no-marked unaccusative subjects in OJ for futureresearch. £i Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 124 Y. Yanagida,J.Whitman 5 Object markingin nominalizedclauses This sectiondiscussesthetwopatterns in nominalizedclauses. of objectmarking butthesame We focuson adnominalclauses because of theirhigherfrequency, holdforthethreeothertypesof nominalizedclauses. generalizations 5.1 Bare objects As we saw in Sect. 2.1, Miyagawa(1989) arguesthatin EarlyMiddle and Old Japanese,adnominalpredicatesfailto assignaccusativecase, and hencean object mustbe licensedby morphological case in the formof wo in orderto avoid a violationoftheCase Filter.Thereare,however, a number ofOJexamplesin which an adnominal a takes an case,whicharethus predicate objectlacking morphological to s (1989) Yanagida(2007b) apparent counter-examplesMiyagawa' generalization. showsthatin theMan'yoshuthereare90 tokensoftransitive clauseswhosesubject is markedby no or ga and whoseobjectis morphologically unmarked. Fifty-five occurwithattributive as in (35). predicates, 05) a. femmnZik® tmmm Saywopimye no kwo ga pire puri-si Gen child Act scarf wave-Past.Adn Sayohime yama no na hill Gen name 'thenameof thehillwhereSayohimewavedherscarf b. ;tin75fi7KgP:t (MY 5/868) Ji±M«S Sika-no ama no keburi (MY 7/1246) sipo yak-u ShikaGen fishermenGen salt burn-Adn smoke 'thesmokyhaze raisingwhenfishermen of Shikaburnsalt' thereis a clear However,whilebareobjectsdo occurwithadnominalpredicates, to the The bare are almost without pattern counter-examples. exceptionnonobjects N°s. Of the 55 of this one has a branching examples type,only phrasalobject,MY 2639 in (36):26 (36) m&mz ^zz^m Tanome ya kimi ga waga na (MY 11/2639) norikyem-u? Trust Foc/Q lord Act my name state-PConj-Adn 'Is it because(he) trustsin me,thatmyLordhas statedmyname?' In relativeclauses,wherethe predicateis realizedin the adnominalform,bare as in (37) (see Appendix1 forpossible objectsare systematically non-branching, exceptions). 26 It shouldalso be notedthaton theclitic analysisof pronoun+ ga (4.1, Yanagida2007b),thebare objectwa ga na 'myname' in (36) is not,strictly speaking,phrasal. £l Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Syntacticalignmentin old Japanese (37) 5^*j 125 m^M-k^ [tamamokar-u] amawotome-domo seaweedcut-Adn fisherwoman-Pl 'thefisherwomen who are gathering seaweed' (MY 6/936) Based on thesedistributional facts,we proposethatbare objectsin nominalized clauses like (35) are to be analyzedon analogywith incorporated objects in Chukchee(Spencer1999): (38) Chukchee(Spencer1999) a. Muri kimit?-e. myt-ine-rety-rkyn we-Abs we-AP-carry-Pres/n load-Instr 'We are carrying theload.' b. Ytlyg-yn qaa-tym-g?e. father-Abs deer-killed-3sG The father killeda deer.' Chukcheeis a splitergativelanguagethathas two typesof derivedintransitive withsemantically constructions transitive verbs(Spencer1999). One is theantiin where is O marked with passive (38a), oblique case and A is absolutive;the N° is the restricted to in (38b). The second, objects, objectincorporation strategy for is also in attested American incorporation strategy objects widely languages activealignment (Sapir 1911). displaying We proposethat OJ uses the incorporation strategyfor bare objects,like Chuckchee.Followingthebasic approachof Baker(1988), non-branching nouns intotheverb,and immediately adjacentto an adnominalpredicateareincorporated satisfiesthe case requirements of the incorporee.This preserves incorporation that the of the adnominal is notassigned object Miyagawa'sgeneralization predicate abstract case in itsbase position. 5.2 Wo-marked objects thehypothesis thatOJis an activelanguage,proposesthat Vovin(1997),developing becauseit marksnotonlytheobjectsof transitive wo is an absolutivecase marker verbsbutalso thesubjectsofnon-active intransitives, primarily adjectives.Mostof theseoccurin a pattern involvingtheadjectivalstemplusthesuffix-mi: (39) x&mx& zm^w&zM tmnm kusa makura tabi wo kurusi-mi kopi wor-eba (MY 15/3674) grass pillow travelObj painful-milong.forbe-when 'as I am longingfor(mywife)travelbeingpainful' Therearealso someexamplesin whicha subjectmarkedbywo occurswithinactive followedby thecomplementizer to (40): predicates ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 126 Y. Yanagida,J.Whitman (40) a. mw ^zmuitzm to to yononaka wo u-si yasa-si world shameful-Cone Comp Obj dreary-Cone Comp (MY 5/893) omop-e-domo think-Rls-though and shameful' 'thoughI feeltheworldunpleasant b. mi¥ prnmtt^ mtz^& kwo-ra ki-mas-am-u to ware wo itu topi-si I Ace when come-Hon-ConcComp ask-P.Adnchild-Affec (MY20/4436) 'thatdeargirl,who askedwhenI wouldcome back' Miyagawaand Ekida (2003) proposethatexampleslike (40a-b) are instancesof exceptionalcase marking(ECM), on the assumptionthatwo is the spelloutof abstractcase assignedby thematrixverb.Yanagida(2006) arguesthatOJhas no withfo-clausesas in ECM construction and thatwo-marked subjectsin thepattern has a nominalized we as verb of the are observe, (which, (40a-b) higher arguments either Under realis and adnominal,respectively).27 inflection, analysis,wo is transitive the matrix but the intransitive not embedded predicate by assigned by to whichwe verb.Notethatthisleavesopenthestatusof theNP wo... -mipattern, return at theend of thissection. isthatthe vemarker Thestrongest argument againsttheanalysisofwo as anabsoluti realisor in wo is never marked non-active intransitive verb of a adnominal, by subject unmarked or-aku nominalclauses.Itis eithermorphologically irrealisconditionals, ormarkedbygenitiveno, as discussedin section4 andexemplified by(41-42). (41) a. ££P£ftft JU^n^min^ Tatuta yama mi-ma0 tikaduk-a-ba Irr-if Tatsuta mountain Hon-horsecomes.near'If yourhorsedrawsneartheTatsutamountain' b. m& RKtJtfEft (MY 5/877) Sft* sono ni ware yuk-amu. Ume no pana 0 saki tir-u Plum Gen blossom bloom fall-Adn gardento I go-will (MY 10/1900) 'I will go to thegardenwhereplumblossomsbloomand fall.' (42) a. &m mz&tiLTh «t*# Kono yupupyetumo no 50-yeda no nagare k-o-ba this evening mulberryGen branch Gen flow come-Irr-if (MY 3/386) 'if thiseveninga mulberry branchcomesflowingdown' 27 See in modern Japanese. Hoji (1991) fora similaranalysisofso-calledRaisingtoObjectconstructions S Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Syntacticalignmentin old Japanese b. ^mttm 127 fexm u no sak-u tukwi no pana blossom Gen bloom-Adn month Gen utugi 'themonthwhentheutsugiblossomis in bloom.' (MY 18/4066) we wouldhave no explanation If wo was an absolutivecase marker, forwhythe in in wo the contexts is never marked by given (41-42). subject fromitsdescendant o in modernJapanesein The particlewo differs significantly thatitmarksnotonlydirectobjectsbutall kindsofVP-internal arguments including and in (cf.Motohashi1989). In (43a), wo marksthegoal argument, quasi-adjuncts In and time wo it marks co-occurs witha source,locative, (43b-e) adjuncts. (43f-g) locativeadjunctmarkedby ni, 'in/at'. (43) a. fl/hM¥ ffM^ no tame wogapa wo yuki-te mi-m-u Gen stream Obj go-ing see-Presum-Adnpurpose (MY 3/332) 'in orderto go and see theKisa stream.' Kisa Kisa b. %&¥immi Nara wo k-i panar-e Nara Obj come-Inf leave-Inf 'comingawayfromNara.' (MY 17/4008) C. JHiI^#M ^ALfem d. M#$£^ B£A e. &jh,75 femm mmmywo asake to kapabe wo parusame ni ware tati nuru stand get.drenchedComp riverside Obj springrain in I (MY 9/1696) drenchedin thespringrainon theriverside.' 'thatI am standinggetting MW4^ Ame no puru ywo wo pototogisunaki-te yuk-u nari. Rain Gen fall night Obj cuckoo cry-ing go-Adn is (MY 9/1756) 'Throughthenightwhentherainfalls,a cuckoofliescrying.' samuki aki kaze no Gen cold autumn wind Sanu no woka morning Obj Sanu Gen hill kimi (MY 3/361) kwoyu-ram-u Adn lord cross-Pr.Conj'mylord,who wouldbe crossingovertheSanu hillin thecold wind.' morning f.%m$gm z^&^&mm ni wo ki-mas-e. Aga koromo sita underneathLoc Obj wear-Hon-Imp my robe 'Wearthisrobeof mineunderneath.' £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 128 Y. Yanagida,J.Whitman g. s^feMi^it VEX^mx aip-fejft^ Adisawi no yapye saku gotoku yatuyo Gen eightgenerations hydrangeas eight-layerbloom as Witt ni wo imase. (MY 20/4448) Loc Obj live-Imp so may(mylord)live 'As hydrangeas have eightfold flowers, foreightgenerations.' to analyzewo merelyas thespelloutof VP-internal These factsmakeit difficult case wouldbe structural case. In (43f-g),forexample,it is unclearwhystructural ni. for PPs headed the locative required by postposition aboutthewordorderofOJclauses Thisproperty correlates withthegeneralization 1. As shown wo mentioned in section by Yanagida(2006), therelative containing thatthelatter is such of and the wo-marked the alwaysprecedes object position subject theformer, as shownin(4) and(19a).28Additional examplesaregivenforadnominal (44a), -aku (44b),andrealis(44c) clausesbelow.(See Appendix2 fora preliminary in theMan'yoshu.) correlation ofconjugational formswithwo-marking (44) a. »*L3Mfe*fti& #€ft£#fPillS|c £«£&£ Ware wo yami ni ya imo ga kwop-i-tutu aru ram-ul I Obj dark in Q wife Act longing.for be PConj-Adn (MY 15/3669) 'Wouldmywifebe longingforme in thedark? b. im^ %$tm^x* kimi wo a ga mat-ana-ku ni lord Obj I Act wait-not-InfLoc 'withoutme waitingforyou' (MY 17/3960) wo yupugure ni Kusaka no yama Kusaka Gen mountainAce twilight in wa ga kwoye ku-re-ba (MY 8/1428) I Act cross come-Rls-Cond 'whenI crossoverKusakamountain in thetwilight' the lists XP 65 of wo Yanagida(2007a) examples precedingsubjects.In contrast, Man 'yoshucontains the order one as only exampleinterpretedinvolving Agentga ... XPwo.29 28 Kinsui (2001) also observesthisgeneralization. 29The exampleis: Yama no na to ipi tug-ye to kamo Saywopimyega MountainGen nameComp say tell-ImpCompQ Sayohime Act kono yamano pe ni (MY 5/872) pire wo puri-kyem-u this mountaintop on scarfObj wave-PPresum-Adn 'MightSayohimehave wavedherscarfon thetopof thismountain, (saying)"Pass it on! name!" This mountain's £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese alignment Syntactic 129 In additionto thisregularity abouttheirwordorder,it has been observedby In factthegeneralMotohashi(1989) thatiw-markedphrasestendto be definite. broader:wo-marked izationis slightly phrasesare specific.This can be shownby can be markedby wo, butwhentheyare,theyreceivea thefactthatw/i-pronouns in contrast tobarew/z-pronouns. Thisis shownin thecontrast specificinterpretation betweenthefollowingtwoexamples. (45) M^nupf wm ®gibm**g itatwo wo osi piraki siweya ide ko-ne Maki no wood Gen door Obj push open damn out come-Des nani se-m-u? noti pa (MY 11/2519) after Top what do-Presum-Adn 'Pushingopenthedoor(I say) "Come out,dammit!"Thenwhatwill (I) do?' (46) $3^±# ^mm w&. tamamo kari tum-ye ipye no imwo ga sipo pwi-na-ba house Gen wife Act seaweed cut gather-Imp tide recede-Perf-if (MY 3/360) pamaduto kop-aba nani wo simyesa-m-u? shore.giftwant-if what Obj proffer-Presum-Adn 'If thetidehas gone out,cutand gatherthepreciousseaweed!If mywife at homeasks forgiftsfromtheshore,which(other)shallI offerher?' in In (45), the universeof thingsthe speakermightdo is completelyundefined thesetof itemsthatthespeakermightoffer previousdiscourse.In (46), in contrast, as pamadutwo'giftsfromtheshore'.In thiscase nani wo 'what/ hiswifeis defined whichObj' picksout specificitemsfromthatset. of adnominalclausesas in (47): Yanagida(2006) analyzestheproperties (47) realizationin OJ Case and argument position (i) Gfl-marked subjectsstayin thebase externalargument (Spec, vP). intotheverb, (ii) Bare objectsare incorporated moveto theouterSpec of vP, (iii) Wo-marked objectsobligatorily feature. to checktheir[definite] Herewerevisethisanalysistotakeintoaccountthenewdatareviewedinthissection. clausesgenerally. in (47) applyto nominalized Second,moveFirst,theproperties featuresincewo-marked mentof wo-marked by a definite phrasesis nottriggered 29 continued footnote whereSapywopimega is takenas the However,thisexampleis opento at leastone otherinterpretation, of 'thismountain', i.e., 'On this,Sayohime'smountain, mightshe have wavedher possessivemodifier butthisis ofYanagida(2006) also citesMY 18/4036as a counter-example, scarf(saying). . .' A reviewer of thisexample. basedon a misinterpretation £l Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Y. Yanagida, J. Whitman 130 XPs mayincludew/z-phrases andPPs,amongotheritems.Instead,we wouldlikethe of wo-marked oftheirmovement. specificity phrasesto be a by-product ofMiyagawa tothebasicinsight We cancapture thesegeneralizations byreturning case. in nominalized clausesdoes notassignstructural (1989): theverbalprojection We saw in Sect.4 thatundertheactiveanalysisthegfl-marked agentin nominaized case there.FollowingMiyagawa's clausesremainsin Spec,vP andreceivesinherent that[nominal]v does notbearan accusativecase originalproposal,we hypothesize inthecase Thisleavestwooptionsforcase licensingobjects:incorporation, feature. ofnonbranching bya headabovevP. objects,orcase assignment that As we haveseen,wo is realizedtotheleftoftheexternal indicating argument, itis indeedassignedby a headabove vP. Thereare twopossiblecandidatesforthe ofthishead.One is T; thiswouldbringOJintolinewithanalysesofcertain identity ergativelanguageswhereabsolutivecase is assignedbyT (Aldridge(2004), Legate (2008)). The drawbackofthisapproachis thatOJwo, as discussedabove,does not absolutivecase: itdoes notappearon thesubjects showthedistribution ofa standard is to verb.A possiblewayaroundthisdifficulty ofinactive(oranykindofintransitive) case assignedby T in nominalizedclausespreexpandtheanalysisof nominative sentedin Sect.4.4. We proposedtherethatbarethemesubjectsare assignednominativecase by T selectinga defectivevP in nominalizedclauses. It could be that thatnominative case is also assignedbyT selectingnondefective, hypothesized is spelledoutas wo. inthisinstance orunergative vP,butthatnominative is,transitive so An analysisalongtheselinesseemspossible,butithas theflavorofa stipulation, here. we willnotpursueitfurther headbetweenv andT. The secondoptionis thatwo is assignedby a functional of AspectP.Washio(2004) We proposethatwo-marked DPs residein thespecifier thatAsshowsthataspectselectionin OJwas sensitiveto transitivity, suggesting that We hypothesize thanvP,was thelocusfora [itransitive]feature. pectP,rather that [-ftransitive] Principle)feature Aspectin OJbearsan EPP (ExtendedProjection to in theverbalprojection attracts thehighestnon-inherently case markedargument its specifier, as shownin (48). AspectP (48) /\ DP wo Asp' /\ vP Aspect[+transitive] /\ DPagen, V /\ VP v [nominal] /\ tDPV £) Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Syntacticalignmentin old Japanese 131 thatbareNPs extracted fromthenuclearscope of the Diesing's(1992) hypothesis ofwo marked clausereceivea specificinterpretation explainsthe[specific] property phrases. and the positionof wo-marked Note thatthisanalysiscorrelateswo-marking withthepresenceof [+transitive] AspectPin nominalizedclauses.We arguments in theothermajorclausetypes:conclusive, leave openthequestionof wo-marking and imperative. We showin Appendix2 thatwo-marking also occursin infinitive, theseclauses, but as observedby Miyagawa(1989) and Miyagawaand Ekida In conclusiveclauses,it is largelyrestricted to prop(2003), it is morerestricted. ositional attitudeverbs of thinkingor saying while in imperativeclauses seemsto have been a mid-eighth centuryinnovation, probablytrigwo-marking and imperative endings. geredby thephonologicalmergerof certaininfinitive to Vovin'scharacterization of thewo ... The AspectPanalysisextendsnaturally as thehead of participial-type nominalin (39). AspectPis identified -mi pattern izationin analysessuch as Embick(2004) and Alexiadouand Anagnostopoulou in (40), thewo-marked (2008). UnliketheobjectEqui wo-marking subject pattern to a matrixobject(or ECM) analysisbecause in wo... -miclausesis notsusceptible reasonorcause.We analyzethe wo...-mi clausesareadjuncts, typically expressing as adjunctAspPs,analagousto Acc-inggerundssuchas 'travel wo...-mi pattern beingpainful'in English: (49) [AspP tabiwo [Vpkurusi]mi] kofiwor-eba be-when travelAce painful-mi long.for 'as I longformywife,travelbeingpainful' On thisanalysis,-miis thespelloutof thehead of [+transitive] AspP. The stipuas one etymology lationthat-mi is [+transitive] motivation, mayhavea diachronic verbmi- 'see'. Wo... -mi of thetransitive for-mi derivesit fromtheinfinitive withour hypothesis thatwo is clauses do notcontaintense,whichis consistent head lowerthanT. assignedby a functional the resultsof this section,we have shown thatOJ had two Summarizing andwoclauses:incorporation forcase marking mechanisms objectsinnominalized to clauses accusative of v in nominalized The above vP. assign inability marking More generally, case is a directextensionMiyagawa's (1989) originalhypothesis. of OJ nomias we discussin detailin Sect. 6.2, the activealignment properties nalized clauses fitinto the cross-linguistic by "nominalist" patternidentified suchas Johns(1992) and Kaufman(2007). alignment analysesof non-accusative oftheobject.Objectsmusttherefore [Nominal]v is unabletocheckthecase feature of 'absolutive'case byT (Aldridge be case licensedbyotherstrategies: assignment or, in the 2004; Legate 2008), defaultabsolutive(Legate 2008), incorporation, above vP. head a functional instanceof OJ,case assigment by 6 Alignmentand nominalizationin diachronicand typologicalperspective to theclause typeswe have is restricted We have seen thatOJ activealignment in -aku, and called 'nominalized':adnominal(rentaikei),nominalcomplements ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 132 Y. Yanagida,J.Whitman realis (izenkei) and irrealis(mizenkei) conditionals.We have shownhow the nominalizedproperties of these clauses are intimately linkedwiththeiractive properties:nominalizedclauses assign inherentagentivega in Spec, vP, and to a positionabove the [transitive] Aspectin theseclauses attractscomplements In Sect. 6.1 we discussthe externalargument, wheretheyreceivewo-marking. in OJ.In Sect. 6.2 we show diachronic sourcesforthenominalized clausepatterns thatnominalizations sourcefornon-accusative area widelyattested cross-linguistic We focuson a specificcase, Caribanlanguagesas analyzedby Gildea alignment. (1998, 2000), and pointout thatit suggestsa possiblesourceforthewo-marking patternin OJ. Section6.3 discusseschangespossiblyalreadyunderwayin OJ, markerno. involvingthegenitive/subject 6.1 The nominalizing endings originsof theadnominaland irrealisconditional or Konoshima(1962) seemsto have been thefirstto arguethatthenominalizing anditsNP oftheadnominal function endingswasprimary, juntaigen'quasi-nominal' of[+N] Adnominal clausesinOJhavethedistribution function modifying secondary. LikeNPs,theymayserveas subjector i.e.,NPs anduninflected categories, adjectives. of function The NP modifying objectoftheclauseandbe followedbycase markers. adnominalclausesis parallelto uninflected adjectives,whichwereable to directly NP inOJ.30As we sawinSect.2.1,Miyagawa(1989) alsoanalyzesOJandMJ modify adnominalclausesas [nominal]. twoareheld Oftheremaining threeclausetypesthatwe havelabelednominalized, in-aku are tobe diachronically derivedfromtheadnominal. Nominalcomplements derivedbyOhno(1953) fromtheadnominal formoftheverbplusa noun*aku,e.g., + aku > kuraku yuk-u'go-Adn'+ aku > yukaku'going;'kuru'come-Adn' -urefor Whitman (2004) derivestheirrealisendings(-e forquadrigrade, 'coming.'31 otherconjugations) fromtheproto-Japanese formoftheadnominalending*-or. The irrealis(mizenkei)base in OJ is shownby Ohno to be of heterogeneous origin.It resultsfromreanalysisof the initialvowel in variousauxiliariesand suffixes as theendingof theirrealis(mizenkei)base. In the case of the irrealis the endingwas *-a, probablyrelatedto the nominalizing suffix*-a conditional, Sakakura This is (1966, 286-303). hypothesized by pp. ending preservedin such noun-verb as tuk'build : tuka 'mound;' pairs mur(e)- 'gather': mura 'group, up' The irrealis conditional village.' appearsproductively onlybeforetheconditional + which Ohno derives from locative ni -ba, particle pa. If Ohno's topicmarker is it confirms the function of *-<z as we expect correct, analysis originalnominalizing locativeni to selecta [nominal]form. 30 It is in pre-OJ(Omodakaet al. late innovation widelyheldthatinflected adjectivesare a relatively thenoun,e.g., 1967,p. 4). It is easy to findOJ examplesof uninflected adjectivesdirectly modifying kanasi imwo 'dearbelovedgirl.' 31 Ohno's nounof theform analysisis generally accepted,butitsweakpointis thatthereis no attested *aku. An alternative thecore of Ohno's insightwouldderiveaku fromtheexistential verb preserving roota- 'exist' plus theadjectivalcontinuative of a-ku. It has longbeen pointedout thatderivatives inflections, partakein "stative"-type e.g.,conclusive-/ in ar-i 'exists.'Nominalsin -(a)ku wouldthen derivefromV-adnom+ exist-coNT, 'beingV.' ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese alignment Syntactic 133 in OJall Summingup,thefourclausetypesthatshowactive-ergative alignment theadnominal, derivefromnominalizations: from nominal,and realisconditional suffix*-or,and theirrealisconditionalfroma nominalizing thepJnominalizing suffix*-a. 6.2 Nominalizations as sourcesforalignment structures as thediachronic A numberof linguistshave proposednominalization forlanguagesthatshowsyncresourcefornon-accusative particularly alignment, tismof agentand genitivemarking.Proposalsof thissortare made forMayan (Starostaet al. 1982; Kaufman2007), and Cariban (Bricker1981), Austronesian accountof Inuktitut (Gildea 1998, 2000). Johns(1992) develops a synchronic The starting based on nominalization.32 pointforthese'nominalist'acergativity is similarto Miyagawa'ssynchronic treatment countsof non-accusative alignment of adnominalclauses in OJ: nominalizedclauses are unableto assignstructural T is present,nomiaccusativecase. Dependingon thefeaturesof T, or whether nominative.The nonnalized clauses may also be unable to assign structural can be seen in familiarexamof nominalizations accusativealignment properties Thusin thecity'sdestruction by the ples,suchas Englishderivednominalizations. accusativenornominative case; barbarians,thenominalprojection assignsneither is is licensedbythepreposition theexternal argument by,andtheinternal argument assignedgenitivecase by D. holdsthatnon-accuthenominalist Froma diachronicperspective, hypothesis clausesare reanalyzedas mainclauses. resultswhennominalized sativealignment richrangeof alignmentand word order Gildea (1998) discussesa particularly in Cariban.We focushereon fromreanalyzednominalizations resulting patterns whatGildea(1998, pp. 190-196,2000, pp. 85-88), citingFranchetto (1990), calls the'De-ergative'systemin theCaribanlanguagesPanareand Kuikuro.The source to Gildea,was an objectnominalization forthissysteminearlierCariban,according withinthe theagentremained selectedbythematrix copula.In thissourcestructure, of thenominalizedverbcornominalizedVP whilethenotionalobjectargument look at a modernPanare respondedto thesubjectof thematrixcopula.Let us first (50).33 pattern exampleof thede-ergative (50) ni-a'kama-piti-hpemen pi yu-noh it dead Agt.Foc-tell-Iter-Pst 1-grandmother toldit overand over.' 'My late grandmother (Gildea (2000, p. 86)) thatGildea(2000: p. 88) positsforthede-ergative (51) showsthesourcestructure pattern: 32 See in thebroadsense,have thatcertaintypesofergativity, Manning(1996) fortheglobalhypothesis to Dan Kaufmanforpointingus in the We are grateful theirdiachronicoriginsin nominalizations. tradition ofstudieson ergativity, hisown. direction ofthisandotherresearchinthe'nominalist' including 33 We of Gildea's Panaredata in (50) for morphophonemic representation provideonlythesegmented clarityof presentation. ^ This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Springer Y. Yanagida,J.Whitman 134 Clause Independent (51) /\ Predicate Subject /\ NP Copula /\ Source: Poss n-V-Nmlzer 1 Result: A i I DErg-V-T/A (Auxiliary) 1 O sourcefora sentencelike (50), forexample,wouldinvolvean The reconstructed toldoverand over withthemeaning'whatmygrandmother nominalization object the matrix of this nominalization 'It' is copula.The complete by predicated again.' reconstructed sourcefor(50) wouldthushavetheform'It is [whatmygrandmother is reanalyzedas told over and over again].' In Panare,this source structure thenommonoclausal,as shownin (51): thecopulais reanalyzedas an auxiliary, external verb and the the inalizationas a verbalprojection (A) argument containing in itsbase position,and theoriginalsubjectas theobject(O). Gildea(2000,p. 98) is construction cites constituency testsshowingthatO in thePanarede-ergative A the verb. and externalto theminimalprojection containing can result structure Gildea's accountshowshow reanalysisof a nominalization marked in a also but not just in non-accusative cross-linguistically alignment of is exactlythesameas theproperty ofthisstructure structure. The basic property in 5.2: Sect. and discussed wo-marked clausesin OJidentified byYanagida(2006) theexternalargument theobjectsurfacesousidetheminimalprojection containing and theverb. ofa clearlydefined languagefamilyofgreattimedepth Japaneseis nota member like Cariban,so in reconstructing earlierJapanesesyntaxwe are confinedto the reconstruction. Howeverwe wouldliketo concludethissectechniqueof internal similarto (51) was thediachronic tionbysuggesting thepossibility thata reanalysis sourceof theobject in OJ.The etymological sourceof theO-wo S-genV pattern marker wo has longbeendebated,butonefactthathas notbeenobservedis thatthe verbrootwo-,whichappearsin shapeof thisparticleis identicalto theexistential theOJ verbswor- 'exist,sit' and wi- 'be at, sit.'34We suggestthatthatwo originatesas the copularverb in a construction parallelto the Caribande-ergative pattern. 34 Some particlewo, whichis linguists,such as Tokieda (1954), relatewo to the OJ sentence-final claimedto be emphatic;Tokieda thusclaims thatthe originalfunctionof object markingwo was itcannotmean,forexample,focus,sincewo Whateveris meantby emphaticin thiscontext, emphatic. definite and presupposed. markedobjectsin OJare typically ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Syntacticalignmentin old Japanese 135 CopP (52) /\ DP /\ Cop vP wo /\ DPagem v /\ VP Source: i I [AspP O wo [vPA VP]] I 1 Result: v [nominal] is reanalyzed In (52), as in (51), a sourcestructure involving copula+nominalization is as AspP, the The projection as monoclausal. containing originalcopula reanalyzed similarto thereanalysisof thecopula as an auxiliaryin Cariban.The restof the remainsthesame.The mostcontroversial structure aspectof (52) is likelyto be the have had a earlier that right-branching copula.Butevenat Japanesemay hypothesis functional such for evidence theOJstage,Japanese projections, right-branching gives as theaspectualauxiliaryari 'be', themodalauxiliarye 'be able', andthenegative 2005). na, all ofwhichappearto theleftofthelexicalverb(Whitman imperative thatboththe evidencesuggesting In thissectionwe have presented typological of OJnominalized clauses structure constituent and transitive properties alignment the clauses Nominalized fitintoa largercross-linguistic mayprovide diapattern. wherethe for a and chronicsourcefornon-accusative pattern synchronic alignment and theverb. the minimal the outside containing subject projection objectappears 6.3 AfterOJ:thereanalysisof no as nominative is As we haveseen,inOJthesubjectcanbe markedbyga orno,buttheirdistribution verbs but not of external marks the Ga different. argument transitive/unergative quite like a of an unaccusative.No, on theotherhand,patterns theinternalargument of a transitives inthatitcanmarkboththeexternal case marker nominative argument as we saw in Sect.4.1: ofintransitives, andtheinternal argument (53) a. 3&C»Mft|£ fcJfW&IUSgA. gotoku kop-u ga midorigwo no ti Gen breast desire-AdnAgt like infant 'like a childdesiresitsmother'smilk' b. %J^7hiL jfl0j4^ naka makwi no tatu ara yama Gen stand rough mountain inside tree coveredwithtrees' 'in theroughmountains (MY 18/4122) (MY 3/241) £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Y. Yanagida,J.Whitman 136 In (53a) no marksthesubjectof thetransitive, and in (53b) it marksthesubject at the of the unaccusative.The two particlesga and no behave differently level as well. Yanagida (2006) showsthatwhenthe subjectand object syntactic theobjectalwaysprecedesthe subject,as we saw in Sect. 4.2. are case-marked, But thereare a few cases of /to-marked subjectswhichviolatethisgeneralizain (54). (OtherMan'yoshu examplesare 196, 2772, tion.These are illustrated 2831, 3689.) (54) a. #M75 S^ft HtKWR nuras-aku Adn drench(MY 9/1697) '(that)thespringrain,however(I) tryto avoid it,drenchesme.' parusame springrain no Gen b. t^f^JP At Soko That mo ka too Q ware wo yokure-do Obj avoid-although I pito people &¥m&!& no wa wo Gen I-Obj 'Do peoplesay thatof me too?' koto nas-am-u? things say-Conj-Adn (MY 512, 1329,1376) is rarein thisconfiguration wo. Although In (54) no appearsin a positionpreceding The in EMJ. in kuntenglosseswritten theMan'yoshu,itbecomeswidespread early followingexamplesare takenfromthe Konkomyo Saisho Okyo The Sutraof are based on Kasuga 1969). GoldenLight'{kuntentextca. 830; interpretations (55) a. yoki womina no ... wotoko yoki sinkyau no Gen Gen/Nom reverent woman men good good nasamu kokoro wo (K 3-5:46) mind Ace produce mind' '(that)good menand good women...mightproducea reverent dou b. yoki wotoko yoki womina no ... Sanzyou man woman Gen/Nom Triyana way good good wo (K 3-5:50) syusemu Ace practice '(that)good menand good womenmightmastertheTriyanadoctrine' Anothersignificant revealedin thisearlyEMJ textis thatsubject development marking ga is used onlyin theformof aga 'I.Subj' and naga 'you.Subj';nominal subjectsare nevermarkedbyga?5 In theMan'yoshuthepronominal subjectsaga 35 Thereis one intheKonkomyoSaisho Okyo wherea nonpronominal subjecttakesga counter-example and theobjecttakeswo, in (i). (i) 5c #> ¥k% wotofito ga mi wo tutete... waga myGen brother Agt oneself Ace sacrifice sacrificed himself.' 'My brother (K 10-26:192) <£}Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese alignment Syntactic 137 andnaga arestrictly adjacentto theverb,butin theKonkomyoSaisho Okyo they can occurin sentenceinitialposition,precedinga wo-marked object,as shownin (56). (56) toki... (K 8-15:144) a. wa ga katari wo idas-are-mu I Agt words Ace produce-Pass-Presumtime 'thetimewhenI (was able to) producewords' kono Myaugyau-wau wo rufu-si... b. na ga yoku you Agt successfully this Myaugyau-wau Ace propagate (K 8-15:146) thisMyo-kyosuccessfully...' 'you propagating intheformalsensewe have Itappearsfromthisdatathatga as an activecase marker, ofga tofirst and inthispaper,was lostearlyintheEMJperiod.Thelimitation defined a steponthewayinthisprocess.(Recallthatin wasprobably secondpersonpronouns to first andsecondpersonalpronouns.) manylanguages,activecase is restricted activeprefix /-in OJ.In Finally,in Sect.4.3.1 we saw evidencefora pronominal the Konkomyo Saisho Okyo, however,thereare manyinstancesof the postof transitive/unergative nominalparticle/ markingtheagentarguments verbs,as shownin (57): (57) ti Ware i ... ga sakai wo tuutatus-eri. (K 8-19:155) Nom wisdom Ga border Ace pass-Rslt I theborderof wisdom.' 'I passedthrough As we saw in Sect.2.3.1,Vovin(1997) analyzesthepostnominal particle/in OJas thefactthatearlykuntentexts an active(agentive)marker.However,considering wereheavilyinfluenced by KoreanglossingforChineseBuddhisttexts,it is pos/.This marker oftheKoreannominative i is a borrowing siblethatsubjectmarking withthe view of Vovin (2005) thatthe muchrarersubject wouldbe consistent likeno, rather marker/foundin OJis a Koreanloan. In anycase, EMJ/patterns thanOJga, in thatit appearsin a positionprecedingwo. The wordorderfactssuggestthatJapanesewas well on theway to developing even in nominalizedclausesby theearlyEMJperiod.Tranaccusativealignment wereassigned of thephoneticformof theircase marker, sitivesubjects,regardless case in a positionhigherthanwo-marked objects.This case was spelledout as ga 35 continued footnote ofga as a nominative The eventualreanalysis particlein Tokyoand KansaiJapaneseis a topicthatfar of ga by exceeds the scope of thispaper.Yamada (2000) examinesthe increasein the frequency comparingthe originaltextof the Tale of Heike, whichis believedto reflectthe languageof the textofHeike,knownas theAmakusa Heike,publishedin withtheJesuitromanized fourteenth century, 1592.He observesthatmanyoftheunmarked subjectsin theTale of Heike cametobe markedbyga in The case in thelate 16thcentury. as a nominative AmakusaHeike. Thismayindicatethatga reappeared as reported andhonorific is confounded matter functions, bytheassociationofga andno withpejorative in theearlyseventeenth Interestingly, accordingYamada's observation, century. byJesuitgrammarians verbsandadjectivesbutrarelyon thesubjectsof ga in thisperiodis usedon thesubjectsofunaccusative of ga in theMan'yoshu. withthedistribution verb.This contrasts a transitive/unergative directly £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 138 Y. Yanagida, J. Whitman (on firstand secondpersonpronouns),/ (on some agentivesubjects),or no (the elsewherecase). 7 Conclusion In thispaper,we haveproposedthatJapaneseat itsoldestattested stage,thecentral dialectOJoftheeighthcentury, OJdisplaysa showsevidenceof activealignment. Main in case nominalized clauses. beween main and (conclusive) split marking in conclusiveclausesare accusative.Nominalizedclausesdisplayactivealignment case markingand cross-referencing of arguments on theverbthrough prefixation. The distinctive O wo S V pattern of transitive nominalizedclauses identified by wherethe witha formalanalysisof activealignment Yanagida(2006) is consistent external receivesinherent activecase in situ.Cross-linguistic comparison argument as is a widespreadsourcefornon-accusative showsthatnominalization alignment, well as fora patternwherethe surfacepositionof the objectis higherthanthe subject. Appendix1: Bare objectsin relativeclauses The possible counter-examples in (5) are given to Miyagawa's generalization below.Phrasalobjectscan appearwithadnominalrelativeclause predicateswhen theyoccurin mainclause initialposition,as shownin (58). (58) a. [Tomo-noUra-noiso-nomurwo-no kwi-01 Tomo Bay Gen beach Gen cypressGen tree [[promimu-goto-ni api-misi]imo]pa wasura-ye-me yamo (MY 3/447) see wheneversee aux Q maid-Top forget'WheneverI see thecypress-tree by thebeachbesideTomo Bay, would I everforgetmydearwho stoodby me and saw it?' b. [Kadusikano Mama no iriyeniutinabiku tamamo-0] KazushikaGen Mama Gen inletin wave seaweed si (MY 3/433) [[prokari-kyemu]tekwona] omopo-yu cut-Aux maidfoc remembee-Aux 'I remember themaidwho gathered thewavingseaweedin theinlet of Mama of theKazushikaregion.' The phrasalobjectsin clause-initial positionin (58) are leftdislocatedto theleft peripheral topicposition.A leftdislocatedtopicmay serveas whatis knownas rhetorical deviceused throughout the stanza',an important jokotoba 'introductory stanzaprefacesthecontents of thesucceedingverse. Man'yoshu.The introductory Notethata bareobliqueNP can also be leftdislocatedoutof a relativeclausewith an intransitive as in (59). predicate, & Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Syntacticalignmentin old Japanese (59) 139 tukwi]matugotosi. [Yupuyami-nokonopa 0] [[prokomoreru] hide moon wait like Gen tree-leave evening-dark (MY 11/2666) 'I feelas if I werewaitingforthemoonhiddenbehindan eveningtree.' verbbutalso thelocativeadjunctof an In OJ,notonlytheobjectof a transitive bare.The bareNP in (59) is nota counterverbcan be morphologically intransitive on twocounts:first, itis an adjunctandthus exampleto Miyagawa'sgeneralization is notassignedstructural case; second,it occursin mainclause initialpositionand as leftdislocated. thusmaybe interpreted is at work. In othercases ofbareNP adjunctswe knowthatthefirst explanation The two bare NP adjunctsama no sita 'the land' (lit. 'undertheheavens') and tamapoko-nomiti 'road' can appearinsidean adnominalclause: thereare four in (60). of ama-no-sitaand six of tamapoko-nomiti,as illustrated occurrences (For otherexamples,see MY 162,207, 230, 1738,3276,4006, 4098, 4465.) (60) ni ama no a. [Opotu no miya Palace to heaven Gen Gen Otsu sita 0 (MY 1/29) sumyeroki sirasime-kyem-u] under governed emperor 'theemperorwho came to thePalace of Otsuto ruletheland.' tori no kowe mo b. naku kikoye-zu hear-not sound Foe bird Gen sing miti no (MY 2/207) yuku] pito [tamapoko Gen street go people Epithet nothearingthesongsof birds' 'peoplewho pass on thestreet, Ama no sita 'land' and tamapoko-nomiti 'road' are analyzableas compound The wordtamapokobenounsand thusare probablynottruecounter-examples. which makura kotoba so-called of to the 'pillowwords'orepithets, category longs thatfollowsthem.Theyare conventionalized of thereferent are fixeddesignators listedbelow. Thereare twogenuinecount-erexamples, and oftenunintelligible. (61) Sanu no Sanu Gen kasa masiwo lend-Aux-Excl (MY 3/161) whentheautumnwindis cold,I wishto lendmy 'In theearlymorning dressto thelordwho is goingoverthehillof Sano.' no swode tuke b. [Warafa ga mi ni pa yupipata young Gen me at Top color-printedGen sleeve with warewo (MY 16/3791) ki-si] goromo0 wear-P.Adn I-Excl clothes 'At a youngage, I was like a childwearingclotheswithsleeves in colors.' printed wo a. [Akikaze no samuki asaake fall wind Gen cold earlymorning Obj oka 0 kwoyu-ramu] kimi ni kinu lord to dress cross-AUx hill £} Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Y. Yanagida,J.Whitman 140 and conclusiveclauses Appendix2: Wo in imperative threeinstancesof wo-marked objectsin theMan'yoshuarelistedin Eighthundred wo used formakurakoKoji (1988). (These do notincludewo-miconstructions, as such withotherparticles and occurrences toba epithets, wo-ba.)36Mostof the in the that 803 tokensappearin nominalized clauses,as expected; is, thepredicates adnominal continuative irrealis(mizenkei) conditionals, (rentaikei), (renyokei), realis(izenkei),and -aku nominalclauses.Thereare,however,someexceptions: form intheimperative ofwo-marked Six instances objectsappearwiththepredicate clauses and purposive (62), and 34 appearwiththeconclusivepredicatein main markedby to (63-64). (62) (6 tokens) Imperative: Aki no (MY 20/4444) yupupye wa warewo sinopas-e. autumnGen evening Top me Obj remember-Imp me.' '(On) autumnevenings,remember (63) Main clause conclusive:(23 tokens) a. Putarisite musubisi pimowo pitorisite are pa I Top cordObj alone together tied mi-zi. toki (MY 12/2919) untie try-Neg.Conj.Conc I will surelynotuntiealone.' 'The sash cordsthatwe tiedtogether, b. Ware-wo wasuras-u na. (MY 14/3457) me-Obj forget-ConcNeg.Imp 'Don't forgetme.' (64) clauses:(11 tokens) Conclusivein fo-marked tama wo pirop-u to jewel Obj find-ConeComp 'in orderto findthejewel' (MY 7/1220) Thereare 119 additionalinstancesof wo-marked objectsthatprecedeto-marked ofelliptedhigherverbssuchas omopu clauses,buttheymaybe takenas arguments of non'think'or ipu 'say'. In theseexamples,womarkedobjectsare arguments in (65). conclusive,non-imperative clauses,as illustrated (65) Kimi woj [proj sakiku are to] (omopite) lord safe be Comp (thinking) Obj sue-t-u. (MY 17/3927) ipapibe vessel place-Perf-Conc offering vessel.' '(Thinking)of mylord,'May he be safe,'I placedtheoffering 36 wo in theManyoshu. to Koji (1988), therearea totalof 1,557instancesofobjectmarking According ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese alignment Syntactic 141 As discussedin Sect. 5.2, we assumethatOJhas no ECM constructions; thewoofthehigherverbomopu markedsubjectoftheinactiveverbis in facttheargument as underlyingly 'think',whichhas beendeletedin (3927) butstandardly interpreted presentin thisand similarexamples.Thereare also in Koji (1988) at leastthree clauses; thesehowever examplesof wo-markedobjectsthatprecedeto-marked shouldbe takenas dislocatedtopics(MY 2/148,MY 10/2228,MY 20/4300).They of embeddedto-marked clauses but serveas jokotoba 'introare notarguments of wo-marked stanzas',as discussedabove.Therearealso fouroccurrences ductory objectsembeddedin -rasi clauses,as in (66) (MY 13, 3145,4111,4427). (66) ras-i. sinop-u Wagimokosi a wo Emphme Obj think-Conc seem-Conc my.girl of me.' 'My girlseemsto be thinking (MY 12/3145) had spreadbeyondnominalized clausesin theOJ Thesedatashowthatwo-marking Most instancesof wo-marking appearsquiterestricted. period,butitsdistribution involve'that'clausessuchas (64-65), and manyof the withconclusivepredicates conclusivemainclause examplessuch as (64a) involvetopicalizedobjects.This to conclusiveclauses.Topicalized suggestsa scenarioforthespreadof wo marking attitude verbsselectinga to-clausecomplewo-marked objectswithpropositional fromtheto-clause,especially mentmayhavebeenreanalyzedas extracted directly in contextslike (65), wherethehigherverbis dropped. in imperatives. Supportforthisscenariocomesfromtherarityof wo-marking Koji (1988) citesonlysixexamples.Fourofthese(MY 4009,4444,4177,and4179) are fromsongsin volumes17-19 of the Man'yoshii,composedby Ootomono Yakamochior his associatesin the periodaround750. Of the remainingtwo examples,MY 3764 was composedaroundthe same periodby Nakatomono Yakamoriwhilewo in MY 2352 is analyzableas a vocativeparticle.This suggests It mayhavearisen lateinnovation. was a relatively in imperatives thatwo-marking clauses,triggered by a phonological by analogywithwo- markingin infinitive predicates merger.Fourof thefiveclearexamplescitedabove involveimperative -e had mergedafter withrootfinal-s; imperative -ye and lowerbigradeinfinitive /s/well beforetheeighthcentury. to ShigeruMiyagawa,KunioNishiyama, AcknowledgementsWe wishtoexpressourdeepappreciation The authorsare of course JEAL reviewerfortheircomments. AlexanderVovin,and an anonymous forscientific researchfrom Workon thispaperwas supported foranyerrors. bygrants-in-aid responsible of Science (JSPS) to Yanagidaand a JapanFoundationgrantto theJapanSocietyforthePromotion Whitman. Texts (Primarysources) Kasuga,Masaji. 1969. KonkomyoSaisho Okyo Koten no KokugogakutekiKenkyu.[A linguistic of theNishidaiji-bon Konkyosaisho6-kyo].Tokyo:Benseisha. studyof theKuntenannotation intheMany particles Koji,Kazuteru.1988.Man'yoshuJoshino Kenkyu.[Researchon thegrammatical oshu].Tokyo:Kasama Shoin. & Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions Y. Yanagida,J.Whitman 142 Tono. 1971-1975.Man'yoshu(\-4), NihonKoten Masatake,Kinosita,andHaruyuki, Kojima,Noriyuki, Bungaku Zenshu. Tokyo:Shogakkan. in 1978-2005). Nakanishi,Susumu.1978-1983.Man'yoshu.Tokyo:KodanshaBunko(reprinted Satake,Asahiro,Hideo, Yamada, Masao, Ohtani,Yoshiyuki,Yamazaki.1999-2003.Man'yoshu Shin of classical Japaneseliterature]. Nihon Koten Bungaku Taikei, (vol. 1-4). [New compendium Tokyo:Iwanami. Electronictexts ofVirginiaLibrary, Electronic TextCenter, TextInitiative http://etext.lib.virginia.edu/ University Japanese japanese/jti.texts.euc.html NationalInstitute of JapaneseLiterature. http://www.nijl.ac.jp/contents/d_library/index.html Makoto.http://yoshi01 Yoshimura, .kokugo.edu.yamaguchi-u.ac.jp/manyou/manyou.html References and word orderin Austronesian languages.Doctoraldissertation, Aldridge,Edith.2004. Ergativity CornellUniversity. Alexiadou, Artemis.2001. Functional structurein nominals: Nominalization and ergativity. Amsterdam: JohnBenjamins. 2008. Structuring Alexiadou,Artemis,and Elena Anagnostopoulou. participles.In Proceedingsof WCCFL 26, ed. NatashaAbnerand JasonBishop,33-41. Somerville:CascadillaPress. East Asia Papers, vol. 6. Allen,NicholasJ. 1975. Sketchof Thulunggrammar.Cornell University Ithaca:CornellUniversity. of ChicagoPress. Baker,Mark,C. 1988.Incorporation.Chicago:The University of case and agreement. determination Maria,and Ken Hale. 1996. The structural Bittner, Linguistic Inquiry27: 1-68. Victoria.1981.The sourceoftheergativesplitin YukatekMaya.JournalofMayan Linguistics Bricker, 2(2): 83-127. ed. ThomasRoeper ofsyntax.In Parametersetting, Borer,Hagit,andKenWexler.1987.The maturation Reidel. and EdwinWilliams,123-172.Dordrecht: In Step by step: Essays on minimalist inquiries:The framework. Chomksy,Noam. 2001. Minimalist 89David MichaelsandJuanUriagereka, syntaxin honorof Howard Lasnik,ed. RogerMartin, 156. Cambridge:MIT. In Syntactictypology,ed. WinfredLehmann,329-394. Austin: Comrie,Bernard.1978. Ergativity. of Texas Press. University In Proceedingsof the Dahlstrom, systems. languagesand splitcase marking Amy.1983.Agent-patient andclaudia 9thAnnual Meetingof theBerkeleyLinguisticsSociety,BLS 9, ed. AmyDahlstrom 37-46. Berkeley:BerkeleyLinguistics Brugman, Society. Diesing,Molly. 1992.Indefinites. Cambridge:MIT. Dixon,R.M.W. 1979.Ergativity. Language 55: 59-138. Press. Dixon,R.M.W. 1994. Ergativity. Cambridge:CambridgeUniversity inEnglish.LinguisticInquiry35.3: 355ofresultative Embick,David.2004.On thestructure participles 392. in kuikuroand otherCariblanguages.In AmaBruna.1990. Ergativity and nominativity Franchetto, Studiesin lowlandSouthAmericanlanguages,ed. DorisL. Payne,407-427. zonian linguistics: Austin:Universitv of Texas Press. JohnBenjamins. Amsterdam: 2008. Proto-Japanese. Frellesvig, Bjarke,and JohnWhitman. Press. Gildea,Spike. 1998. On reconstructing grammar.Oxford:OxfordUniversity Gildea, Spike. 2000. On the genesisof the verb phrasein Caribanlanguages.In Reconstructing ed. Spike Gildea,65-106. Amstergrammar:Comparativelinguisticsand grammaticalization, dam:JohnBenjamins. Harada,Shin-Ichi.1971.Ga-no conversionand ideolectalvariationsin Japanese.Gengo kenkyu60: 25-38. formsin the and attributive Peter.1998. Kakari particlesand themergerof thepredicative Hendriks, Korean Linguistics7: 197-210. Japaneseverbalsystem.Japanese/ ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions in old Japanese Syntactic alignment 143 at the ECM, and themajorobjectin Japanese.Paperpresented Hoji, Hajime. 1991. Raising-to-object, of Rochester. JapaneseSyntaxWorkshop. University Alana. 1992.Derivingergativity. Johns, LinguisticInquiry23.1: 57-87. in Austronesian. Daniel. 2007. The nominalist Kaufman, Papergivenat ZAS Berlin,August hypothesis 14,2007. fromthe ofJapanesegrammar Kinsui,Satoshi.2001. Joshikaramitanihongobunpouno rekisi[A history viewpointof particles]Lecturehandout.Tokyouniversity. or languagesor activetypology. Klimov,ueorgijA. 19/4.Un thecharacter Linguisticsui: l l-ZD. Klimov,GeorgijA. 1977. Tipologijajazykov aktivnogostroja [Typologyof languagesof theactive type].Moscow:Nauka. Konoshima,Tadatoshi.1962. Chukogoni okeruyogenrentaikeino yoho [The use of the participial adjectivein medievalJapanese].Kokugogaku48: 102-107. Jaklin.2003. Subject case in Turkishnominalizedclauses. In Syntacticstructuresand Kornfilt, and Luka Szusich,130-214.Berlin:Moutonde ed. Uwe Junghanns morphologicalinformation, Gruvter. changein earlierJapanese]. Kuginuki,Torn. 1996. Kodai Nihongo no keitaihenka [Morphological Osaka: IzumiShoin. of Englishand Japanese. we agreeor not:a comparative Kuroda,Sige-Yuki.1988. Whether grammar 12: 2-44. LinguisticaeInvestigationes Kuroda,Shige-Yuki.2007. On thesyntaxofOld Japanese.In Currentissuesin thehistoryand structure 263-318.Tokyo:Kurosio. J.C.Smith,andMasayoshiShibatani, ofJapanese,ed. BjarkeFrellesvig, in ThulungRai: A shiftin the positionof pronominalsplit. Lahaussois,Aimee. 2003. Ergativity Language variation:Paperson variationand changein theSinosphereand in theIndospherein honour of James A. Matisoff.Pacific Linguistics,ed. D. Bardleyet al. 101-112. Canberra: Nationaluniversity. Australian and abstractcase. LinguisticInquiry39.1, 55-101. Legate,Julie.2008. Morphological MIT. and movementtheory.Doctoraldissertation, Mahajan,AnoopKumar.1990.TheA/Aydistinction and grammaticalrelations.Stanford: D. 1996. Ergativity: Argumentstructure Manning,Christopher CSLI Publications. Alec. 1997.No escapefromsyntax:Don't trymorphological Marantz, analysisintheprivacyofyourown lexicon.In University of PennsylvaniaWorkingPapers in Linguistics,vol. 4.2, ed. Alexis Club. PennLinguistics et al., 201-225. Philadelphia: Dimitriadis andtheTupianfamily.In What'sin a verb: Meira,Sergio.2006. Stativeverbsvs. nounsin Satere-Mawe Studiesin theverbalmorphologyof thelanguagesof theAmericas,Lot OccasionalSeries,ed. GraduateSchool of LinGraz'ynaJ. Rowickaand EithneB. Carlin,pp. 189-214. Netherlands guistics.Netherlands. and itsmotivations. case marking Marianne.1991.Active/agentive Mithun, Language 67.3: 510-546. and case markingin Japanese.Syntaxand Semantics,vol. 22. Miyagawa,Shigeru.1989. Structure New York:AcademicPress. Miyagawa,Shigeru.1993. Case checkingand theMinimalLinkCondition.In Case and agreementII, MIT WorkingPapers in Linguistics,vol. 19,ed. Colin Phillips,213-254. Cambridge:MIT. in of theaccusativecase marking Miyagawa,Shigeruand Fusae Ekida. 2003. Historicaldevelopment texts.Journalof JapaneseLinguistics19: 1-95. Japaneseas seen in classicalliterary Edson T., KennethWexler,TakakoAikawa,and ShigeruMiyagawa.1999. Case-dropping Miyamoto, et al., 443-452. in Japaneseacquisition.In BUCLD 23, ed. Annabelgreenhill and unaccusatives Somerville:CascadillaPress. of theJapaneselanguage.Doctoraldissertation, Tatsushi.1989.Case theoryand thehistory Motohashi, of Arizona. University no kigenni tuite[On theoriginof theJapanese Ohno,Susumu.1953.Nihongono doshino katsuyokei verbconjugations]. Kokugo to kokubunsaku350: 47-56. of Japaneseby Omodaka,Histakaet al. 1967. Jidaibetsukokugo daijiten,jodai-hen [A dictionary period,Old Japaneseedition].Tokyo:Sanseido. Tokyo: Kadokawa Sakakura,Atsuyoshi.1966. Gokosei no kenkyu[Researchon WordFormation]. shoten. of Old and orthography Sasaki,Takashi. 1996. Jodaigo no kobun to hydki[The sentencestructure Japanese].Tokyo:HituziShobo. in Americanlanguages.AmericanAnthroSapir,Edward.1911. The problemof nounincorporation pologist13: 250-282. £i Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions 144 Y. Yanagida,J.Whitman Sapir,Edward.1917. Reviewof C.C Uhlenbeck:Het PassieveKarakter...In The collectedworksof Edward Sapir V (1990), ed. W. Bright, 69-74. Berlinand New York:Moutonde Gruyter. In Grammaticalcategoriesin Silverstein,Michael. 1976. Hierarchyof featuresand ergativity. Institute of Aboriginal Australianlanguages,ed. R.M.W. Dixon, 112-171.Canberra:Australian Studies. Spencer,Andrew.1999. Chukcheehomepage.http://privatewww.essex.ac.uk/~spena/Chukchee/CHUKCHEE HOMEPAGE.html In Starosta, Stanley,AndrewPawley,andLawrenceReid. 1982.The evolutionoffocusin Austronesian. Conferenceon AustronesianLinguistics,Vol. 2: Tracking PapersfromtheThirdInternational 145-170. thetravelers, PacificLinguisticsC-75,ed. S.A. Wurmand L. Carrington. in Japanese.Doctoraldissertation. Takezawa,Koichi.1987.A configurational approachto case-marking of Washington University Tokieda,Motoki. 1954. Nihon bunpo kogo-hen [Japanesegrammar, colloquial languageedition]. Tokyo:IwanamiShoten. of splitintransitivity. Van Valin,RobertD. 1990. Semanticparameters Language 66.2: 221-260. of Old Japanese.Journalof East Asian Linguistics Vovin,Alexander.1997.On thesyntactic typology 6: 273-290. Vovin,Alexander.2005. A descriptiveand compartivegrammarof WesternOld Japanese,vol. 1. Folkestone:Global Oriental. selectionin theEast.Journalof East Asian Linguistics13: 197-256. Washio,Ryuichi.2004. Auxiliary in Old Japanese.In Syntacticeffects Watanabe,Akira.2002. The loss of overtw/z-movement of morPress. 179-195.Oxford:OxfordUniversity phologicalchange,ed. David W. Lightfoot, InJapanese/Korean John.1997.Kakarimusubifroma Comparative Whitman, Linguistics, Perspective. CSLI. vol. 6, Ho-minSohnand JohnHaig, 161-178.Stanford: at of therentaikei and originsof theizenkei.Paperpresented John.2004. The reconstruction Whitman, the.OxfordKnhft Svmnnsinm nnthe.HistnrvnfTflnanese. Whitman,John.2005. Preverbalelementsin Korean and Japanese.In The Oxford handbook of comparativesyntax,ed. GuglielmoCinque and RichardKayne,880-902. Oxford:OxfordUniversitvPress. Wichmann,S0ren.2008. The studyof semanticalignment:Retrospectand stateof the art.In The typologyof semanticalignment,ed. S0renWichmannand Mark Donohue,3-23. Cambridge: Press. CambridgeUniversity Wichmann,S0ren,and Mark Donohue. 2008. The typologyof semanticalignment.Cambridge: Press. CambridgeUniversity Ellen. 1997.Four-way case systems: nominative, Woolford, objective,andaccusative.Natural Ergative, Language and LinguisticTheory15: 181-227. Wrona,Janick.2006. WhyOld Japanesewas notan ergativelanguage,ms.,KyotoUniversity. Wrona,Janick.2008. A studyof Old Japanesesyntax:Synchronicand diachronicaspectsof the complementsystem.London:Global Oriental. kakudaino yoso [Theexpansionoftheuse ofthe Yamada,Masahiro.2000. Shugohyqjiga no seiryoku betweentheoriginalTextoftheTale ofHeike andAmakusaban subjectdenotor ga: A comparison Heikel Kokugogaku51-1: 1-14. atWorkshop andbarenominalsinEarlyOld Japanese.PaperPresented Yanagida,Yuko.2005.Ergativity on TheoreticalEast AsianLinguistics, HarvardUniversity. in EarlyOld Japanese.Journalof East Asian Yanagida,Yuko. 2006. Wordorderand clause structure Linguistics15: 37-68. Yanagida,Yuko. 2007a. Miyagawa's(1989) exceptions:An ergativeanalysis.MIT WorkingPapers in 265-276. Cambridge:MIT. Linguistics,vol. 55, ed. Yoichi Miyamoto, in Old Japanese].In Nihongono Yanagida,Yuko. 2007b.Jodaigono nokakuseini tsuite[On ergativity shubungensho [Main clause phenomenain Japanese],ed. NobukoHasegawa,147-188.Tokyo: HituziShobo. ^ Springer This content downloaded from 132.236.27.111 on Sun, 13 Sep 2015 04:35:55 UTC All use subject to JSTOR Terms and Conditions
© Copyright 2026 Paperzz