Partitive Case and Aspect Paul Kiparsky Stanford University 1 Partitive case as a syntactic and semantic problem Current theories make a distinction between two types of case, STRUCTURAL case and INHERENT (or LEXICAL) case (Chomsky 1981), similar to the older distinction between GRAMMATICAL and SEMANTIC case (Kurylowicz 1964). 1 Structural case is assumed to be assigned at S-structure in a purely configurational way , whereas inherent case is assigned at D-structure in possible dependence on the governing predicates’ s lexical properties. It is well known that not all cases fall cleanly into this typology . In particular, there is a class of cases that pattern syntactically with the structural cases, but are semantically conditioned. These cases however depend on different semantic conditions than inherent cases do: instead of being sensitive to the thematic relation that the NP bears to the verbal predicate, they are sensitive to the NP’ s definiteness, animacy , or quantificational properties, or to the aspectual character of the VP , or to some combination of these factors. The Finnish partitive is a particularly clear instance of this apparently hybrid category of semantically conditioned structural case. 2 The interest of partitive case goes beyond the fact that it poses an apparent puz- zle for case theory . The combination of conditioning factors governing the partitive raises semantic questions as well. In its aspectual function, partitive case is assigned to the objects of verbs which denote an unbounded event, in a sense to be made clear below . In its NP-related function, partitive case is assigned to quantitatively indetermi- nate NPs (including indefinite bare plurals and mass nouns), even if the verb denotes a bounded event. Moreover, in Slavic, the distinction between perfective and imperfec- tive aspect has the very same two functions (among others). How can these functions 1 This paper was inspired by C. Pi ˜ non’s dissertation. I am deeply grateful to him and to Cleo Condoravdi and Henriette de Swart for their searching comments on earlier versions, as well as to Mary Laughren and David Nash for generously answering my questions about W arlpiri. In addition, the paper has benefited from the scrutiny of two reviewers. 2 For evidence that partitive case is a structural case, see section 3, and V ainikka 1993, who argues that the partitive is the default case assigned to obligatory complements of verbs, prepositions, comparative ad- jectives, caseless numerals, and certain quantifiers. See also Nikanne 1990 and Maling 1993. The present paper is concerned only with the partitive as the case of the object of verbs, the only function in which it alternates with the accusative according to the semantic conditions investigated here. 1
38
Embed
Partitive Case and Aspect - user.phil.hhu.defilip/kiparsky.1998.pdf · Partitive Case and Aspect Paul Kiparsky Stanford University 1 Partitive case as a syntactic and semantic problem
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
PartitiveCaseandAspect
PaulKiparskyStanfordUniversity
1 Partitive caseasa syntacticand semanticproblem
Currenttheoriesmakea distinction betweentwo typesof case,STRUCTURAL caseand INHERENT (or LEXICAL ) case(Chomsky1981),similar to the older distinctionbetweenGRAMMATICAL andSEMANTIC case(Kuryłowicz 1964).1 Structuralcaseisassumedto beassignedatS-structurein apurelyconfigurationalway, whereasinherentcaseis assignedat D-structurein possibledependenceon the governingpredicates’slexical properties.It is well knownthatnot all casesfall cleanlyinto this typology. Inparticular,thereis a classof casesthatpatternsyntacticallywith thestructuralcases,but aresemanticallyconditioned.Thesecaseshoweverdependon differentsemanticconditionsthaninherentcasesdo: insteadof beingsensitiveto the thematicrelationthat the NP bearsto the verbalpredicate,they aresensitiveto the NP’s definiteness,animacy, or quantificationalproperties,or to the aspectualcharacterof the VP, or tosomecombinationof thesefactors.TheFinnishpartitiveis aparticularlyclearinstanceof thisapparentlyhybridcategoryof semanticallyconditionedstructuralcase.2
The interestof partitive casegoesbeyondthe fact that it posesan apparentpuz-zle for casetheory. The combinationof conditioningfactorsgoverningthe partitiveraisessemanticquestionsaswell. In its aspectualfunction,partitivecaseis assignedto theobjectsof verbswhich denoteanunboundedevent,in a senseto bemadeclearbelow. In its NP-relatedfunction,partitivecaseis assignedto quantitativelyindetermi-nateNPs(includingindefinitebarepluralsandmassnouns),evenif theverbdenotesaboundedevent.Moreover,in Slavic, thedistinctionbetweenperfectiveandimperfec-tive aspecthasthevery sametwo functions(amongothers).How canthesefunctions
1This paperwasinspiredby C. Pinon’sdissertation.I amdeeplygratefulto him andto CleoCondoravdiandHenriettedeSwartfor their searchingcommentson earlierversions,aswell asto Mary LaughrenandDavidNashfor generouslyansweringmy questionsaboutWarlpiri. In addition,thepaperhasbenefitedfromthescrutinyof two reviewers.
2For evidencethat partitive caseis a structuralcase,seesection3, andVainikka 1993,who arguesthatthepartitive is thedefaultcaseassignedto obligatorycomplementsof verbs,prepositions,comparativead-jectives,caselessnumerals,andcertainquantifiers.SeealsoNikanne1990andMaling 1993. Thepresentpaperis concernedonly with the partitive asthe caseof the objectof verbs,the only function in which italternateswith theaccusativeaccordingto thesemanticconditionsinvestigatedhere.
1
beunified?And how cancase,a nominalcategory, expressthesamemeaningsastheverbalcategoryof aspect?
Froma typologicalperspective,theaccusative/partitivecasealternationin Finnishis intriguingbecauseits aspectualsideresemblestheso-calledconativecasealternationbetweenaccusativeanddativeobjectsfoundin AustralianandPolynesianlanguages,amongothers(Hale1982,Laughren1988,Simpson1991).Thesemanticeffectof thetwo casealternations,andtheclassof verbsthatparticipatein them,arealsosimilar.And the conativedativealsoseemsto straddlethedistinctionbetweenstructuralandinherentcase. The dativeobject is syntacticallytransitivein manyrespects,yet theconativehasbeenarguedto bea kind of antipassiveconstruction,attributedby somelinguiststo NP-movementin thesyntax,by othersto theoperationof a lexical rule atthelevelof argumentstructure.
Finally, thereis a historicalaspectto the problem,which hasbeenthe subjectofmuchdebateamongFinno-Ugrists.Partitivecaseoriginatedasapurelyadverbiallocalcasewith the meaning“from”. It developedfirst its NP-relatedfunctions,in severalstages.Theaspectualfunctionwasthelast to emerge.How is this stepwiseextensionof thepartitive’susesrelatedto its shift from asemanticcaseto a structuralcase?
In this paperI put forwardananalysisof theFinnishpartitive which shedssomelight oneachof theseissues.Thefunctionof partitivecaseis to licenseunboundednessat the VP level. This unifies the aspectualandNP-relatedfunctionsof the partitiveat the level of theVP, andappropriatelygeneralizesto theconativeconstructions.Byseparatingtheconfigurationalassignmentof structuralcaseat S-structurefrom its se-manticconditioning,we preservethe spirit of the distinctionbetweenstructuralandsemanticcase.Theorigin of thestructuralpartitivecanthenbeseenasagrammatical-izationprocessof theclassictype.
2 The semanticsof partitive case
The Finnishpartitive hastwo functions,which canbe termedASPECTUAL andNP-RELATED. Theaspectualfunction,whichFinnishlinguistshavecharacterizedin termsof resultativity(Itkonen1976,HakulinenandKarlsson1979:183,Larjavaara1991)orboundedness(Ikola 1961,Heinamaki 1984,Leino1991),is illustratedin thesentencesof (1) with the verb meaning“to shoot”. It belongsto a classof verbswhich assigncaseto theirobjectsin two differentways,with adifferentaspectualinterpretation.3
(1) a. Ammu-i-nshoot-Pst-1Sg
karhu-abear-Part
//kah-tatwo-Part
karhu-abear-Part
//karhu-j-abear-Pl-Part
‘I shotat the(a)bear/ at (the)two bears/ at (the)bears’
3The relevantobjectcasemarkingsareglossedin boldface. Thus,in (1a) andin (4b), in kaksikarhua‘two bears’,thecaseof theobjectis accusative,markedon theheadkaksi, which in turn assignspartitive tokarhuainternally to theobject.
2
b. Ammu-i-nshoot–Pst-1Sg
karhu-nbear-Acc
//kaksitwo-Acc
karhu-abear-Part
//karhu-tbear-PlAcc
‘I shotthe(a)bear/ two bears/ thebears’
In theaspectuallyirresultative,unboundedversion(1a),theverbhasapartitiveobject,anddenotesanactivity (in Vendler’ssense),“to shootat”. In theaspectuallyresultative,boundedversion(1b), it hasanaccusativeobject,anddenotesanaccomplishment,“toshootdead”.4 In contrast,the unboundedversion(1a) is non-committalas to whathappenedto thebear,andits useconversationallyimplicatesthattheshot(s)missed.
In addition to verbslike “shoot”, Finnish hasverbswhich are intrinsically un-bounded,suchas“love”, “seek”,and“touch”, andverbswhichareintrinsicallybounded,suchas“kill”, “get”, “find”. Themajority of Finnishverbs,in fact, belongto oneofthesetwo classes.
To illustrate the NP-relatedfunction of partitive case,consideran intrinsicallyboundedverbsuchassaada“get”. Theobjectsof suchverbsarepartitivewhentheyarequantitativelyindeterminate(in particular,whentheyareindefinitebarepluralsormassnouns),otherwiseaccusative:
4Barring specialcontextualinformation, (1b) is understoodto meanthat I killed the bear, its Englishcounterpartsimply that my bullet hit the mark, with possiblybut not necessarilyfatal results. Contextualinformationcanyield theEnglish-typeinterpretationin Finnish(Heinamaki 1984),andit seemsthat it canalsoyield theFinnish-typeinterpretationin English.
5But in theresultativemeaning,etsia cantakeaccusativeobjects;thedictionaryNykysuomenSanakirjacitestheexample(presumablyfrom a mathproblem)Etsi luvunalkutekijat (Acc.pl.) “Determinetheprimefactorsof thenumber”.
3
The propertyof verbssuchasampua“shoot, shootat” is that they shareboth ofthesepatterns.With sucha verb,a partitivebarepluralobjectthereforeyieldsa three-way ambiguity: ammuinkarhuja (see(1b)) canmean“I shotat thebears”(aspectualpartitivity), “I shotbears”(NP-relatedpartitivity), or “I shotatbears”(both).
In theexamplescitedso far, boundednesscoincideswith resultativity. Onemightbe led to the conclusionthat boundedpredicatesaresimply thosethat denotean ac-complishmentor achievement,andunboundedpredicatesarethosethatdenotestatesandactivities(processes).A closerlook showsthatresultativityis not thedecisivecri-terion. A classof boundedirresultativeverbs,suchasomistaa“own”, nahda “see”,assigncaselike resultativeverbs:6
(7) a. UnboundedNP predicates:A lot of bears,a lot of coffee.
b. BoundedNP predicates:#A lot of bear,#alot of thebear,#alot of a bear,#alot of two bears,#alot of manybears.
TheNPcontrastatstakedoesnotcorrespondexactlyto definitenessor to anyotherfamiliar determinerfeature.Formally indefinitebareplural or massnounsdo not al-waysget assignedpartitive casewith verbslike “get” or “seek”. Theydo so only ifthey havea quantitativelyindeterminatedenotation. The examplesin (8) illustratehow evenindefinitebarepluralsandmassnounsgetaccusativecaseif theydenoteaconventionallydelimitedset(of knownor unknowncardinality).
(8) a. Anu-llaAnu-Adess
onbe-3Sg
loistava-tbrilliant-PlAcc
oppilaa-tstudent-PlAcc
‘Anu hasbrilliant students’
b. Anu-llaAnu-Adess
onbe-3Sg
loistav-i-abrilliant-PlPart
oppila-i-tastudent-PlPart
‘Anu has(some)brilliant students’
c. Aki-llaAki-Adess
onhave-3Sg
iso-tbig-PlAcc
silma-teyes-PlAcc
//viikse-tmustache-PlAcc
‘Aki hasbig eyes/ a mustache’
7Theimaginativereaderwill find unboundedconstrualsof someof theseverbalpredicates.To just thatextent,theyshouldalsoallow degreeadverbsin English,andpartitiveobjectsin Finnish.
5
d. Aki-llaAki-Adess
onhave-3Sg
iso-j-abig-PlPart
silm-i-aeyes-PlPart
//viiksi-amustache-PlPart
‘Aki hasbig eyes/ mustachesin hispossession’
e. Vauva-llaBaby-Adess
onhave-3Sg
pitka-tlong-PlAcc
hiukse-thair-PlAcc
‘The/ababy’shair is long’
f. Vauva-llaBaby-Adess
onhave-3Sg
pitk-i-along-PlPart
hiuks-i-ahair-PlPart
‘The/ababyhas(some)longstrandsof hair’ (on its head,in its hand,etc.)
(8a)meansthatall of Anu’sstudentsarebrilliant, andimplicatesthateveryonein somerelevantcomparisonsethasstudents.(8b) is indeterminatein boththeserespects.8 In[8c], theuseof accusativecasein effectimpliesinalienablepossession:theeyesmustbeAki’ sown two eyesandthemustachehisownmustache(thelattera pluraletantumin Finnish).9 In contrast,[8d] suggestsanindeterminatenumberof alienablypossessedobjects,suchasglasseyesor anatomicalsamplesin a vat, falsemustaches(of whichtheremustnow beseveral),etc. (8e)refersonly to thetotality of thebaby’s own hair,(8f) couldbeaboutpartof thebaby’s own hair, or aboutan indeterminatenumberofloosestrandsit hasin its possession.
Conversely, a singularor definiteNP getsthe NP-relatedpartitive if it is generic.In (9), ‘this rose’is partitivebecauseit means“rosesof thisparticularkind”:
(9) Puutarhurigardener
istutt-iplant-Past3Sg
kaikkialleeverywhere
tatathis-Part
ruusu-a.rose-Part
‘The gardenerplantedthis roseeverywhere.’
Summarizing,we cansaythatanobjectis partitiveeitherif it governedby oneofa classof unboundedverbalpredicates(the aspectualcondition),or if it is quantita-tively indeterminate(theNP-relatedcondition). TherehasbeenmuchdisputeamongFinnish linguistsaboutthe questionwhetherit is possibleto unify theseconditions.Someclaim thattheaspectualandNP-relatedfunctionsof thepartitivearefundamen-tally distinct(Itkonen1976,Larjavaara1991).Othersarguethatunboundednessis thecommondenominatorof both(Heinamaki 1984,Leino1991).Thereareprobablytworeasonswhy thelatter,unifiedanalysishasnotbeenmoregenerallyaccepted:thelackof aprecisedefinitionof thecentralconceptof (un)boundedness,andthefact thatthereare languagesin which the partitive (or anotherobliquecase)haseither just the as-pectualfunctionor just the NP-relatedfunction. In this paperI attemptto overcomeboththeseobstacles,by providinga definitionof (un)boundednessthatcoversthefullrangeof accusative/partitivecontrastsin Finnish,andby showinghowthedifferenceintherole this featureplaysin Finnishandin theotherlanguagescanbesystematicallycharacterized.
9If we imaginethatAki hasmorethanonehead,then[8c] canof courserefercollectively to all thepairsof eyesandto all themustacheswhich hemayhaveon thoseheads.
6
It hasbeenknownfor a long time that thepartitivevs. accusativecasecontrastinFinnishsharessomeof its functionswith the aspectcontrastin RussianandsomeoftheotherSlaviclanguages.In fact, theconceptof boundedness(Russianpredel’nost’)is standardin Slavic aspectology. This parallelismhasbeenemphasizednotablybyDahl andKarlsson(1976)andby Dahl (1985). Theypoint out that if eitherthe verbis atelic (doesnot denotea completedevent),or theobjectis anindefinitebareplural,then Russianin generalrequiresimperfectiveaspect,and Finnish requirespartitivecase(see(10a)).Thus,in (10a)perfectiveaspect(in Russian)andaccusativecase(inFinnish)requireboth that the verb is telic, andthat the object is plural anddefinite.The samesentenceswith imperfectiveaspectand partitive case,respectively, are 3waysambiguous(see(10b)):
The contextsappendedin parenthesesto the translationsshowonecharacteristicuseof theaspectualopposition,thatof markingprogressiveversuscompletedevents.However,the oppositionmarkedby objectcasein Finnishandby aspectin Russianhasa muchwider rangeof interpretations,which againareoften parallel in the twolanguages.Considertheexample“He openedthewindow” in (11). Theatelic imper-fective/partitiveversionin (11b)canhave,in additionto its progressive-typemeaning“He wasopeningthewindow(. . .asJohnwalkedin)”, severalothersensesof temporaldelimitationor incompleteness:“He openedthewindowfor awhile” (it remainedopenfor a limited time only), or “He openedthewindowa bit” (it wasnot fully opened),or“He keptopeningthewindow” (iterative).10
(11) a. Russian:OnHe
otkry-l (Perf.)open-PstM3Sg
oknowindow-Acc
Finnish:HanHe
avas-iopen-Past3Sg
ikkuna-nwindow-Acc
10Not all thesesecondaryusesof the imperfectivein Russianhavecounterpartsin the Finnishpartitive,andvice versa;in section6 below I arguethat the verbalor nominalstatusof the categorylargely predictswhich onesoccur.
Notethat in English“He openedthewindow somemore” allowseachof thereadingscorrespondingto (2), (3), and(4).
What ties theseaspectualfunctionsof partitive caseand imperfectiveaspectto-gether?And how do theyjointly relateto theNP-relatedfunction?Beforeattemptingto answerthesequestionslet us take a look at two recenttheoriesof partitive casemarking.
3 DeHoop: partitive asweak structural case
In an influential thesis,de Hoop (1992)proposedto resolvethe puzzleof the statusof the partitive by enrichingthe typologyof structuralcase.In additionto structuralcaseof the traditionally recognizedtype, suchasaccusativecase,which shedubbedSTRONG structuralcase,de Hoop proposeda categoryof WEAK structuralcase,ofwhich theFinnishpartitiveis aninstantiation.Weakstructuralcasesaredefaultcases,configurationallylicensedat D-structure.But (unlikestrongstructuralcases),theyarenotconfigurationallylicensedby all transitiveverbsbutonly by asubclassof them.
Syntactically, weakstructuralcasein deHoop’stheoryis configurationallylicensedatD-structure.Its specialsyntacticpropertyis thatit canonly belicensedin theirbasicD-structureposition.In consequence,NPsbearingweakstructuralcasedonotundergoscrambling(which de Hoopassumesto be A-movement).Semantically, weakstruc-tural casehasthe propertythat it inducesanexistentialreadingon the NP that bearsit. Accordingto deHoop,NPsthatbearweakstructuralcasearepredicatemodifiers,which areassumedto combinesemanticallywith the verbalpredicate,restrictingitsdenotationin the mannerof adverbialmodifiers. NPs that bearstrongcase,on theotherhand,arearguments,interpretedasgeneralizedquantifiers.
De Hoop’s accountconstitutesa pioneeringattemptto connectthe syntacticandsemanticpropertiesof partitivecasein a principledway. However,it doesnot go farenough.On the semanticside,it positsno intrinsic connectionbetweenverbclassesandNP properties,therebyleavingtherelationbetweencaseandaspectunexplained.And on the syntacticside, it glossesover the fundamentaldifferencesbetweenthe
8
Finnish partitive caseand the other putativeinstancesof “weak” case,suchas theTurkish bareaccusative,or the Inuit absolutive. Thesehaveto do with specificity— not quantitativeindeterminacy— andunlike the Finnishpartitive havethe samedistributionin singularandpluralNPs.Also, while (accordingto deHoop)weakcasein thoselanguagesblocksscrambling,in Finnishpartitiveobjectscanscrambleexactlylike accusativeobjects.For example,in (12) theobjectkarhuja “(the) bears”andtheadverbialsusein“often”, Lapissa“in Lapland”, andhaulikolla “with the/ashotgun”canbein anyof the24possiblemutualorders,regardlessof definitenessandtelicity:
Thus,althoughthe Finnishpartitive is supposedto be the prototypicalweakcase,itactuallydoesnot conformto theallegedsyndromeof weakcasepropertiesidentifiedby deHoop.
Otherthanthefalsepredictionaboutscrambling,it is notclearwhatconsequencesdeHoop’stheoryhasfor thesyntaxof partitivecasein Finnish.Eventhebasicdistribu-tion of thepartitiveseemsto bea problemfor it. As explainedin section1, partitiveisassignedby anytransitiveverbto quantitativelyindeterminateobjects(theNP-relatedpartitive),andby a propersubclassof transitiveverbsto anyobject(theaspectualpar-titive). De Hoop proposesto accountfor the formerby her principle (p. 90 ff.) thatobjectsbearingstrongandweakcasehavestrongandweakreadings,respectively(i.e.referentialvs. existentialreadings).But evenindefiniteNPswith accusativecasecanhaveexistentialreadings,so this cannotbe the relevantconditionfor NP-relatedpar-titive case. To accountfor the aspectualfunctionof the partitive,de Hoop proposesthat “a partitiveobjectcanbe regardedprimarily aspartof a predicateratherthanasanindependentargument”(p. 98), andthatobjectsof atelicverbsarereally predicatemodifiers,licensedby weakcaseat D-structure(p. 111). Shortof anaccountof howpredicatemodifiers,or NPswith weakcase,comeby their two interpretations(NP-relatedandaspectual),this is hardlymorethana restatementof thegeneralizationtobeexplained.
In anycase,theideathatpartitiveobjectsarepredicatemodifiersfalls afoulof thefact that theycanbe conjoinedwith accusativeobjects. In suchstructuresas(13), itis hardto seehow the object in the secondconjunct,the partitive ‘books’, could beanalyzedasa predicatemodifierwhich restrictsthedenotationof theverbalpredicate‘buy’, ratherthan as an argumentof it, parallel to the accusativeobjectof the firstconjunct,theaccusative‘newspaper’.
(13) Ost-i-nbuy-Pst-1Sg
lehde-nnewspaper-SgAcc
jaand
kirjo-j-a.book-Pl-Part
‘I boughtthe/anewspaperandbooks’
Thestatusof partitiveobjectsasargumentsis confirmedby animportantgeneral-izationof FinnishsyntaxknownasSiro’s Law, which statesthat a simpleclausecan
9
haveat mostoneobject.11 This generalizationholdsregardlessof whatthecaseof theobjectis. However,a clausecanhavean object(whetheraccusativeor partitive) to-getherwith apredicatemodifier,in factwith anynumberof predicatemodifiers.Thus,a partitive object is structurallyparallel to an accusativeobjectand not structurallyparallelto apredicatemodifier.
The last two grammaticalfactsconstituteevidencethat the Finnishpartitive is astructuralcase. de Hoop indeedrecognizesthe structuralstatusof the partitive onthebasisof theevidenceadducedby Vainikka1993.But thefact thatthepartitiveis astructuralcasewouldseemto undermineherclaimthatthepartitiveis apredicatemod-ifier. Structurally, partitiveobjectsarecompletelyanalogousto accusativeobjects,anddifferentfrom adverbialmodifiersor obliqueobjects.Thedistinctionbetween“weak”
and“strong” structuralcase,andtheassignmentof NPsbearing“weak” structuralcaseto thecategoryof predicatemodifiers,might beexpectedto havesomesyntacticcon-sequences.But asfar asI cantell, it hasnone.
I concludethatdeHoop’stheoryof casedoesnotaccountfor thedistributionor in-terpretationof partitivecasein Finnish.Theideathatthepartitiveis a“weak” structuralcasedoesnot shedanylight on its grammar. It hasneitherthesemanticpropertiesex-pectedonthatanalysis,noranysyntacticpropertiesthatwouldshowthatit is anythingmoreor lessthana directobject.
4 Krifka: a unified semanticsfor partitive case
A differentperspectiveon the Finnishpartitive is openedup by Krifka 1992. Whilehe doesnot addressthe syntacticaspectsof the problem,he formulates,for the firsttime, an explicit semanticanalysiswhich unifies the meaningsof partitive caseandimperfectiveaspect(his “progressive”),or at leastthe meaningsof someimportantusesof them.
Krifka showsthat, at the VP level, partitive objectmarking is equivalentto im-perfectiveV markingprovidedthat certainadditionalconditionshold. First, V mustdenoteapredicatewith DIVISIVE REFERENCE:12
(18) P hasDIVISIVE REFERENCEiff ∀x∀y[P (x) ∧ y v x→ P (y)]
Thatis,Pis closedunderthesubpartrelation.Forexample,write hasdivisivereferencebecauseanypartof aneventof writing is alsoaneventof writing.13
(19) a. UNIQUENESSOF OBJECTS: ∀e, x, x′[R(e, x) ∧R(e, x′)→ x = x′](therecanbeno two distinctobjectswhichbearR to thesameevent)
12Nominal predicatesaswell asverbscanhavedivisive referenceor not: for example,gold hasdivisivereference(for anypartof gold is alsogold), but thegold or a bookdoesnot havedivisive reference(for partof thegold is not thegold, andpartof abookis not abook).
13We setasideherethe so-calledminimal partsproblem,that an eventcannotin fact be subdividedadinfinitum.
11
b. UNIQUENESSOF EVENTS: ∀e, e′, x[R(e, x) ∧R(e′, x)→ e = e′](therecanbeno two distincteventswhichbearR to thesameobject)
c. MAPPING TO OBJECTS: ∀e, e′, x[R(e, x) ∧ e′ v e → ∃x′[x′ v x ∧R(e′, x′)]](if aneventbearsR to anobject,anysubpartof theeventbearsR to somesubpartof theobject)
d. MAPPING TO EVENTS: ∀e, x, x′[R(e, x) ∧ x′ v x → ∃e′[e′ v e ∧R(e′, x′)]](if aneventbearsR to anobject,anysubpartof theobjectbearsR to somesubpartof theevent)
In orderto get the partitive to yield the progressivereading,Krifka 1992:48sup-plementsthe two mappingproperties[19c,d]with anadditionalassumptionwhich hestatesasfollows:
(20) If aneventbearsR to anobject,thenthewholeobjectis eventuallysubjectedtotheevent.
This is in factaproblematicassumption,asKrifka himselfpointsout,since“Johnwaswriting a letter” (or its counterpartin Finnish) in no way implies that Johnactuallyfinishedwriting thewholeletter.
An exampleof a verb that hasthe propertiesin (19) is write, in the “authorial”sense.It satisfiestheuniquenessof objectsproperty[19a]: writing a letterandwritingthefirst line of theletteraredifferentevents.It satisfiestheuniquenessof eventsprop-erty [19b], providedwe assumethat you can’t authorthe sameletter twice (whereasyou canclearly “write” thesameletter twice in othersenses,for example,if you aresendingthe identicalmessageto two people,or if you area professionalcopyist).14
And it clearlysatisfiesthetwo mappingproperties[19c,d].Let usfurthermoreassumethatit alsosatisfiestheproperty[20] (settingasidetheabovementionedobjection).ByKrifka’s theoremit thenfollows thataneventof writing partof anobjectis partof aneventof writing the object,andconversely. Theequivalencein (10) is thusa conse-quenceof Krifka’s theory, at leastfor theverbwrite.
But asKrifka pointsout, his accountdoesnot generalizeto otherclassesof verbs(suchas(1), (4), (21)), which lack oneor moreof thepropertiesin (19). Thus,verbslike proveandbuy do not havedivisive reference:part of a provingor buyingeventis not necessarilyitself a provingor buyingevent. Kiss, touch,andshootat lack theuniquenessof objectsproperty:kissingMary andkissingMary’s lips canbethesameevent. And all theseverbs(aswell as suchverbsas want, love, read, andwrite inmanyof its senses)do not havethe uniquenessof eventsproperty:you cando these
14In truth, it is evenpossibleto author thesameletter twice, at leastbecausetwo textsthatyou composecouldaccidentallyturn out identical.Thustheuniquenesspropertiesareproblematic.However,this maybeunimportantbecausetheonly role theyplay in Krifka’s proof (1992:48)is to insurethat thesubpartrelationin (19c) is not trivially satisfiedby thecasex′ = x. But presumablytherequiredproperpart relationcouldbeobtainedby weakermeans.
12
thingstwice to the sameobject. Nevertheless,the equivalencebetweenpartitive andimperfectiveis valid for suchverbsaswell. We havealreadyseenit for “shoot” in (1).It holdsfor “write”, not just in theauthorialsense,but in all senses,asdocumentedin[10]. In general,partitivecasecanbeusedin Finnishfor objectsof numerousverbs,suchas“buy”, “eat”, “read”, andit mustbeusedfor others,suchas“kiss”, which failto meetoneor moreof the requisiteconditionsin [18]-[19], asthe examplesin (21)attest:
(21) a. MattiMatti-SgNom
ost-ibuy-Pst3Sg
maito-amilk-SgPart
(tunni-n)(hour-Acc)
‘Matti boughtmilk (for anhour)’
b. MattiMatti-SgNom
ost-ibuy-Pst3Sg
maido-nmilk-SgAcc
(tunni-ssa)(hour-Iness)
‘Matti boughtthemilk (in anhour)’
c. MattiMatti-SgNom
luk-iread-Pst3Sg
kirjo-j-abook-Pl-Part
(tunni-n)(hour-Acc)
‘Matti readbooks(for anhour)’
d. MattiMatti-SgNom
luk-iread-Pst3Sg
kirja-tbook-Pl-Part
(tunni-ssa)(hour-Iness)
‘Matti readthebooks(in anhour)’
e. AnuAnu
suutel-ikiss-Pst3Sg
Esa-aEsa-Part
//#Esa-n#Esa-Acc
(tunni-n(hour-Acc
//#tunni-ssa)hour-Iness)
‘Anu kissedEsa(for anhour/ #in anhour)’
Krifka (1992,48) suggeststhat partitiveslike thosein [21] areanalogicalexten-sionsof thecasesin (10). This couldmeanseveraldifferentthings,dependingonhowweunderstandtheambiguousterm“analogy”.WhatKrifka probablymeantis thattheyareidiosyncraticusageswhich havearisenhistoricallyby analogyto thesemanticallymotivatedcorecasesin [10]. Anotherview, moreconsonantwith theideathatanalogyis structuraloptimization,includingin particulartheprojectionof a grammaticalregu-larity to newcases,would bethat the innovationsaresystematicusageswithin a newgrammarof Finnishthathasarisenby ageneralizationin therulesof caseassignment.On this view, thecasesin [21] andthecasesin (10) areequallyregularusages,whichwould meanthatKrifka’s analysisis not correctfor the currentstateof the language(althoughit might havebeenfor an earlierstage).I believethe Finnishsystemis an“analogical” innovationonly in this latter sense,and will presentsomeevidenceinfavorof thatpositionin section8.
Krifka’s analysisrepresentsanidealizationwhich missessomesystematicaspectsof Finnishpartitivecaseassignment.Neverthelessit is quite illuminating,andindeedcomesremarkablycloseto characterizingthe partitive usageof anotherFinno-Ugriclanguage,whichrepresentsanancestralstagein theevolutionof partitivecase.In whatfollows I will first presentanalternativeaccountof Finnish,andthen(section8) sketchout thepathandcausesof its evolutionfrom theoriginalsystem.
13
5 Boundedness
We canunderstandhow thenominalcategoryof partitive casecanhaveanaspectualfunction,andhowtheverbalcategoryof aspectcanhaveanNP-relatedfunction,if wethink of thembothaslicensedby the unboundednessof the VP predicate.This ideais not new. In theFinnishtradition,Leino 1991hasemphasizedthatboundednessis apropertyof situationsandnot just of individual predicatesin isolation. Verkuyl 1972arguesexplicitly thatpropertiesof thedirectobjectinfluencetheaspectof the predi-cate,andthattheoppositionbetweenimperfectiveandperfectiveaspectis notamatterof the lexical categoryitself. Instead,aspectis constructedby thecombinationof theverb and its object (the VP) andsubsequentlyof the VP and its subject. Verkuyl’s(1989)categories“terminative”and“durative” areverysimilar to “bounded”and“un-bounded”usedhere.
As a first approximation,we can say that if a VP denotesan unboundedevent,then its object is partitive in Finnish,and its verb is imperfectivein Russian,and ifit denotesa boundedevent,they areaccusativeandperfective,respectively. In bothlanguages,theVP canhavethis semanticpropertyeitherin virtue of its headV or invirtueof anominaldependent.But themorphologicalmarkingassociatedwith thisse-manticpropertyis uniformly nominalin Finnish,anduniformly verbalin Russian.Forexample,by thecriterionof boundednessmentionedin thefirst section,thatonly un-boundedpredicatesaremodifiableby adverbsof degreesuchasa lot, theVP predicatesin [22] areunbounded,andthe VP predicatesin [23] arebounded.In (22a),the un-boundednesscomesfrom theverbalpredicatehate(inducingthe“aspectualpartitive”in Finnish),andin (22b),theunboundednesscomesfrom theNPpredicatebombs, withquantitativelyindeterminatereference(inducingthe“NP-related”partitivein Finnish).
(22) a. Theyhatedthebombsa lot.
b. Theydroppedbombsa lot.
In (23),boththeverbalpredicatenor theNP predicatearebounded,andthepartitive
(23) a. #Theydroppedthebombsa lot.15
b. #Theydroppedmanybombsa lot.
The majority of telic verbs,suchas thosein (24), are bounded,and assignac-cusativecaseto their objects(unless,of course,theseobjectsarequantitativelyinde-terminate,in whichcasetheygettheNP-relatedpartitive):
c. Continuousmotionor contact:heiluttaa ‘swing backandforth’, ravistaa‘shake’, keinuttaa‘rock’, nyokyttaa ‘nod’, suudella‘kiss’, hyvaill a ‘ca-ress’,koskettaa‘touch’, nussia‘fuck’, hieroa ‘massage’.
Verbs which take both partitive and accusativeobjectsdependingon the VP’sboundednessincludeverbsof creationanddestruction(suchasthosein (26a)),otherverbsdenotingeventswhoseprogressis mappedoutinto thepartsof theobject((26b)),andanumberof verbswith differentlexicalmeaningsdependingonthecaseof theob-ject ((26c)).
(26) a. syoda ‘eat’ (partitive: soi piirakkaa ‘ate pie, someof thepie’, accusative:soi piirakan ‘ate a/thepie’), leikata ‘cut’, kaivaa‘dig’, kirjoittaa ‘write’.
b. lukea ‘read’ (partitive’: ‘read (at leastpart of)’, accusative:‘read up tosomepoint’, usually‘finish reading’),tutkia ‘investigate’,siirtaa ‘move’,sekoittaa‘mix’.
c. lyoda ‘beat’ (partitive: ‘beator hit (at) someone’,accusative:‘beatsome-one at something’),nimittaa ‘name’ (partitive: ‘call (by a name)’, ac-cusative:‘nominate’),muistaa‘remember’(partitive: ‘commemorate’,‘re-membersomeonewith agift or greetingonaspecialoccasion’,accusative:‘recall’).
Temporalextent(durationor frequencyof iteration)sufficesto makean intrinsi-cally boundedpredicateunbounded;seereading(4) of example(11),or thefollowingcontrast:
(29) a. MattiMatti
lainas-iborrow-Past3Sg
kello-awatch-Part
‘Matti borroweda watch.’ [temporarily]
b. MattiMatti
lainas-iborrow-Past3Sg
kello-nwatch-Acc
‘Matti borroweda watch.’
To summarize:theverbsin [24] takepartitiveobjectsjustwhentheNPis quantita-tively indeterminate(the‘NP-related’partitive).Theverbsin [25] takepartitiveobjectsregardlessof thenatureof theobject(the‘aspectual’partitive).Theverbsin [26] workin bothwaysdependingonthemeaning,e.g.‘eat (at) thesausage’(partitive),‘eat (up)the sausage’(accusative).Whentelic, the verbsin [26] work like the verbsin [24],suchas ‘kill’ (their object is accusativeunlessit is an indefinitebareplural or massnoun).Whenatelic,theverbsin [26] work like theverbsin [25], suchas‘touch’ (theirobjectis partitiveregardlessof its inherentproperties).
Although telicity (or resultativity)by andlargecorrelateswith boundedness,it isnot exactlythe right semanticcriterion for characterizingtheconditionsunderwhichobjectsareaccusative.Thereis botha classof boundedatelic verbswhoseobjectisaccusative(unlessit is quantitativelyindeterminate),anda classof unboundedtelic(resultative)verbswhoseobjectis partitiveregardlessof theNP’snature.
With verbsof thefirst class,thepartitiveshowsonly its NP-relatedside,in spiteoftheirateliccharacter. Itkonen1976callssuchverbs“quasi-resultative”:
Leino 1991:166ff. proposesthegeneralizationthatverbsdenotingstatestakeac-cusativeobjects. This coversthe quasi-resultativesin (30), which includeverbsde-noting relationslike “own”, “contain” andverbsof knowing andperceiving. But itwrongly extendsto verbsof emotion,which takepartitive objects(see(25)). Leinosuggeststhat Finnish treatsemotionsas atelic activities rather than as states. Thiscouldnot bea peculiarityof Finnish,for on thestandardtests(e.g.Dowty 1979,Ch.2) theseverbsbehaveexactlyasdo their Englishcounterparts.18 The distinction ishoweverconsonantwith gradability, in thatonecanhate,admire,andfearsomethingmoreor less,whereasonenormallyeitherowns,contains,knows,or seessomethingornot.19
(33) a. TrumplikesNewYork a lot.
b. FredadmiresMary verymuch.
(34) a. #TrumpownsNew York a lot.
b. #JohnknowsMary verymuch.
To capturethe notion of unboundednessfor verbal and nominal predicates,weusethepropertiesof DIVISIVENESS andCUMULATIVITY , redefinedasin (36a,b),andDIVERSITY:
(35) a. P is DIVISIVE iff ∀x[P (x) ∧ ¬atom(x)→ ∃y[y @ x ∧ P (y)]]
b. P is CUMULATIVE iff ∀x[P (x) ∧ ¬sup(x, P )→ ∃y[x @ y ∧ P (y)]]
c. P is DIVERSE iff ∀x∀y[P (x) ∧ P (y) ∧ x 6= y → ¬x @ y ∧ ¬y @ x]
17It passedthe standardaccomplismentdiagnostics:John punishedthe child in an hour, Johnfinishedpunishingthechild.
18For example,verbslike “hate”, “admire”, “fear” do not felicitously combinewith agentiveadverbialssuchas“carefully”, or embedunderverbslike “force” in Finnishanymorethantheydo in English.
(36) A predicateP is UNBOUNDED iff it is divisiveandcumulativeandnotdiverse.
In (36a,b)the requirementthat x not be the maximalelement(the supremum)or aminimal (atomic)elementensuresthat,for example,aneventof paintinga housemaybe unbounded(with the objecthousein the partitive) evenif the whole houseor thesmallestpart of it was in fact painted.20 The diversity condition (36c) ensuresthatpredicatesthathaveonlyatomicelements(suchasFred), andpredicatesnoneof whoseelementsarerelatedto eachotherby thesubpartrelationarebounded(eventhoughtheysatisfy(36a,b)).
Thedefinition in (36) characterizesquantitativelyindeterminatepluralsandmassnounsasunbounded.Quantitativelyindeterminatecountnounsandindefinitenounswith a cardinalitypredicatearebounded.Certainmorphologicalelementsaffect theboundednessof verbalpredicatesby changingtheir quantitativedetermination.In En-glish, the particleat makesunboundedpredicates,andFinnishsuffixessuchas -ais-andRussianprefixessuchaspo-makeboundedpredicates.
(37) a. Unboundedpredicates:bombs,food,shootat, look for, touch,love,want,contemplate,doubt,use,expect,protect
b. Boundedpredicates:
a. fewbombs,a little food,Russianpoplacet′ ‘to cry for awhile’, pocitat′
20Becausethe accusativeobjectwould imply that the whole housewaspainted,the useof the partitivecarriesadefeasibleimplicaturethatonly partof thehousewaspainted.
18
b. Theydropped(+B) bombs(–B).
BoundedVP predicateshaveneither:
(40) a. Theydropped(+B) thebombs(+B).
b. Theydropped(+B) manybombs(+B).
In (39a),theVP predicateis unboundedbecauseits headV hatedis unbounded,andin(39b)theVP predicateis unboundedbecauseits objectNP is unbounded.In (40),ontheotherhand,neithertheheadnor theinternalargumentareunbounded,sotheVP isnotunbounded.
We cannow unify theaspectualandNP-relatedfunctionsof theFinnishpartitiveby thefollowing descriptivegeneralization:
(41) Theobjectof anunboundedVP is obligatorilypartitive.
[38] and[41] togetheraccountfor thetwo mainconditionsfor partitivecasediscussedabove,namelythatanobjectis eitheraquantitativelyindeterminateNP(anunboundednominalpredicate),or governedby anatelicverb(anunboundedverbalpredicate).Ifthe verb is unbounded,its objectmustbe partitive whetherthe object is unboundedor bounded.E.g. touchedthe/aflower (part.sg.),touched(the) flowers(part.pl.). Apartitive object is ungrammaticalif the VP is bounded,and an accusativeobject isungrammaticalif theVP is unbounded.A ruleequivalentto (41)applies(togetherwithothers)to licenseimperfectiveaspectin Russian.
Verbswhich areaspectuallyambiguouscanbe treatedasunspecifiedfor bound-edness;oncetheir boundednessis fixed they are treatedin exactly the sameway asthe aspectuallyunambiguousverbsin (22a)and(22b); see(1) and(10). Table [42]illustratesthis for theFinnishandRussianexamplesin (10),with theprogressiveagainservingasan (inadequate)shorthandglossfor theentirerangeof atelic verbalmean-ings:
The specialcasesin [8] also follow from the rule statedin (41); the Finnishac-cusative/partitivecontrastherereflectsa semanticdistinction betweenboundedandunboundedreferencewhich in this caseis not overtly expressedin English. Supposethat “eyes” in the inalienablypossessedsense(e.g.the eyesin my face)is implicitlyquantified,perhapsmeaningsomethinglike “pair of eyes”,eventhoughit is formallya bareplural. Thenthe expressionin this senseis bounded,andreceivesaccusativecase.In thealienablesense(anatomicalsamples,glasseyes)theexpressionis anormal
19
bareplural, andthereforeunbounded.(Notethatonly in the lattersenseis it possibleto sayJohnhassomemoreeyes, cf. (5) ff.) This showshow thedistributionof parti-tive versusaccusativecaseon theobjectdependson boundedness,ratherthanon themorphologicalfeaturesof definitenessandnumber,or onthepresenceor absenceof anovertquantifier.
The samepoint canbe madefor imperfectiveaspect,oneof whosefunctionsisto mark unboundednessat the level of the VP. Considerthe abovementionedverbalpredicatessuchaspoplacet′ ‘to cry for a while’, pocitat′ ‘to reada bit of’, nacitat′ ‘toreada lot of’ probolet′ ‘to beill for a certainamountof time’ (see(37)). Theseverbsare[+B] themselves,andtheyimposeaquantifiedreadingontheircomplement,whichmakesit [+B], evenif it is morphologicallyan indefinitebareplural. The resultingVP is therefore[+B] andrequiresperfectivemorphology. For example,in theRussiansentence
(43) nabrali(+B)Pref-gather-PAST-3Pl (Perf)
grybov(+B)mushrooms-GenPl
‘they pickeda lot of mushrooms’(Forsyth1970:22)
theverbalprefixna-combinessemanticallywith thebarepluralobjectgrybov“mush-rooms”([–B] by itself) to makea quantified(andtherefore[+B]) expression“a lot ofmushrooms”.
Thiscasevariationcanbeobtainedby assigningcaseeitherat thelevelof thelowerVP or at the level of thehigherVP. Thelower VP ‘kill thebear’ is bounded(cf. *killthebearsomemore), if thecaserule (41) appliesto it theobjectgetsaccusativecase;thehigherVP ‘try to kill thebear’is bounded(cf. try somemoreto kill thebear), if thecaseruleappliesto it is assignspartitivecaseto theobject.
Negatedverbsassignpartitivecaseto theirobjectsobligatorily(see(45a)),andop-tionally to certainmeasurephrases(adverbialsof timeandextent)whichareotherwiseaccusative(see(45b,c)).
(45) a. MattiMatti-SgNom
e-inot-3Sg
myy-nytsell-PstPart
talo-ahouse-SgPart
(#talo-n).(house-SgAcc)
‘Matti didn’t sell the/ahouse’
20
b. MattiMatti-SgNom
odott-iwait-Pst-3Sg
#tunti-ahour-SgPart
(tunni-n)(hour-SgAcc)
‘Matti waitedanhour’
c. MattiMatti-SgNom
e-inot-3Sg
odotta-nutwait-PstPrtc
tunti-ahour-SgPart
(tunni-n)(hour-SgAcc)
‘Matti didn’t wait anhour’
Thecasechange“goesdown” arbitrarily far into nonfinitecomplements:
It is not only overt negationthat selectspartitive case. Like a negativepolarityitem,partitivecasecanappearin implicitly negativecontexts.For example,a speakerexpectinga negativeanswer,or trying to bepolite,mightprefer(47b)to (47a).
(47) a. On-koBe-Q
sinu-llayou-Adess
kyna?pencil-Nom
‘Do youhaveapencil?’
b. On-koBe-Q
sinu-llayou-Adess
kyna-a?pencil-Part
‘Do youhaveapencil?’
6 Coercion
Both partitive and imperfectivemorphologycanmark different semanticvariantsofunboundedness.Although theserun parallel in Finnish and Russianin many cases(recallthediscussionaround(11)) this is not alwaysthecase.Differencesin how lan-guagesinterpretunboundednessresultfrom differentcoercionof boundedexpressionsinto unboundedexpressionsandviceversa.
We haveseenthatexpressionsdenotingboundednonpunctualevents(accomplish-mentssuchasshootthebear) haveunboundedcounterpartsthatdenoteprocessescon-stitutingsuchevents(suchasshootat thebear). Punctualevents(achievements,suchasdrop theball) areatomic(i.e., not constitutedby processes,Pinon1995,p. 91), sotheirunboundedcounterpartscannotdenotesuchprocesses.Theycan,however,denoteprocessesthatarecomposedof punctualevents.With imperfectiveaspector partitive
Similarly, an intrinsically boundedpredicatecanget a secondarydurativeinterpreta-tion, asanongoingactivity, againwith partitivecaseon theobject:
(49) a. Tapo-i-nkill-Pst1Sg
juurijust
karhu-a.bear-Part
‘I wasjust killing thebear.’
b. MattiMatti
ost-ibuy-Pst3Sg
(juuri)(just)
auto-a,car-SgPart,
(kun.. . )(when.. . )
‘Matti was(just)buyinga car,(when.. . )’
In additionto the iterativeanddurativeinterpretations,basicallyboundedpredi-catescanget a rangeof otherunboundedinterpretations,dependingon whetherthelanguageexpressesVP unboundednessby aspectualmorphologyon the verb or bycasemorphologyon theobject.Thegeneralizationseemsto beasfollows:
(50) a. Aspectcancoerceshiftsin thelexicalmeaningsof verbs.
b. Casecancoerceshiftsin thelexicalmeaningsof nouns.
An exampleof aspectualcoercionin the lexical meaningof verbsis that theRussianimperfectivevyigryvat’, like theEnglishprogressivebewinning), candenote(in addi-tion to asequenceof winningevents,i.e. theiterativemeaning)aprogressivestateof aprocessthatprecedesthepunctualeventof winning, i.e. “be ahead”.PartitivecaseinFinnishdoesnot inducethis meaning:in a VP headedby thecorrespondingverbvoit-taa “win”, a partitiveobjectcannotcoercethemeaning“be ahead”;Finnishrequirestheprogressiveverbconstructionin (51b)“be (in theprocessof) winning” to conveyit.
‘He wastrying to givememoney, but I refused.’ (Leinonen1984)
22
b. JaI
vasyou
obman-yva-l(Impf.) ,deceive-Pst3SgM,
nobut
mogcould-Pst3SgM
liQ
obman-u-t’deceive-Inf
(Perf)?
‘I tried to deceiveyou,butcouldI deceiveyou?’ (Unbegaun1969)
Theconativeshadesinto a varietyof otherrelatednuancesof incompleteaction:
(53) a. CtoWhat
zeso
delaldid-IpfPstM
Bel’tovB.
vin
prodolzeniecourse
etixthese
desjatiten
let?years?
VseEverything,
ilior
poctialmost
vse.everything.
CtoWhat
onhe
sdelal?did-PrfPstM?
Nicego,Nothing,
ilior
poctialmost
nicego.nothing
‘So what did B. do in the courseof theseten years? Everything,or al-mosteverything.What did he accomplish?Nothing,or almostnothing.’(Forsyth1970:71)
b. KolumbColumbus
by-lbe-PstM
scastlivhappy
nenot
togda,then,
kogdawhen
otkry-l Prfdiscover-Impf-PstM
Amerik-u,America-Acc,
abut
kogdawhen
otkry-va-l(Impf )discover-PstM
eeit-Acc
‘Columbuswashappynot whenhediscoveredAmerica,but whenhewasaboutto discoverit (in theprocessof discoveringit)’ (Unbegaun1969)
Further,in the pasttensethe Russianimperfectivecanalsobe usedto asserttheoccurrenceof a pastevent,with currently relevantconsequences.For example,Onpisa-l pis’m-a(imperfective)couldhaveroughlythesenseof English“He haswrittenthe letter (already)”. This is the “simple denotation”or “statementof fact” meaning(Forsyth1970:82,Smith1991:312).
Suchaspectual/temporalmeaningshiftsinducedonverbsby imperfectiveaspectinRussianhaveno counterpartsin theaspectualuseof theFinnishpartitive. In Finnish,on the otherhand,the accusative/partitiveoppositionis exploitedto yield a rangeofspecialNP-relatedinterpretations,which in turncannotbereplicatedby aspectin Rus-sian.
A pervasivephenomenonis the coercionof a telic meaning. BoundednessinFinnishcanbelicensedby resultativitythroughexplicit or implicit locativeor resulta-tive predication.For example,potkaista“kick” takesa partitiveobjectin themeaning
21Larjavaaranotesthat thecorrespondingusageis impossiblewith theLatviangenitive,which otherwisehasfunctionssimilar to that of the Finnishpartitive (bareplural andmassnounobjectsandobjectsundernegation),or with theFrenchde-partitive construction.TheRussianpartitive (secondgenitive)alsocannotbeusedin this way.
23
“kick at”, andan accusativeobjectwhena directionalcomplementis added,or evenimplied. In (55b),theverb“rub”, atelicin (55a),takesanaccusativeobjectlicensedbythe(explicit or implicit) resultativepredicate.A particularlycreativeusageof thiskindare(55j), wheretheatelicintransitive“love” is construedasresultative.22
(55) a. Hiero-i-nrub-Pst-1Sg
si-tait-Part
‘I rubbedit’
b. Hiero-i-nrub-Pst-1Sg
senit-Acc
pehmea-ksisoft-SgTrnsl
‘I kneadedit soft’
c. Ravist-i-nshake-Pst-1Sg
mato-ncarpet-SgAcc
(#kade-n)(hand-SgAcc)
‘I shook(out) thecarpet(my hand)’
d. Aitimother-SgN
makas-ilie-Pst-3Sg
lapse-nsachild-SgAcc-3Sg
kuoliaa-ksidead-SgTransl
‘The motherlay herchild dead(i.e.killed it by lying ontopof it)’
e. JussiJussi(NOM)
maalas-ipaint-Pst3Sg)
talo-nhouse-Acc
(punaise-ksi)(red-Transl)
‘Jussipaintedthe(a) [whole] house(red)’
f. JussiJussi(NOM)
maalas-ipaint-PAST(3SG)
talo-ahouse-Part
(punaise-ksi)(red-Transl)
‘Jussiwaspaintingthe(a)house(red)’
g. Luinread-PAST-1SG
kirja-nbook-Acc
(loppu-un)(end-Illat)
//(repale-i-ksi)(shred-Pl-Transl)
‘I readthebook(to theend)/ (to shreds)’
h. Luinread-Pst-1sg
kirja-abook-Part
‘I wasreadingthebook.’
i. Rakast-i-nlove-Pst-1Sg
tei-tayou-PlPart
‘I lovedyou’
j. Rakast-i-nlove-Pst-1Sg
te-i-da-tyou-PlAcc
rappio-lleruin-Adess
‘I lovedyou into ruin’ (EinoLeino)
The upshotis that the aspectualuseof partitive casein Finnish,while certainlyforming oneof its two corefunctions,is not asrichly elaboratedasthatof theverbalcategoryof aspectin Russian.Conversely, theNP-relateduseof aspectin Russianisnot asrichly elaboratedasthatof thenominalcategoryof casein Finnish.Thegener-alizationseemsto bethata morphologicalfeatureinducesextendedinterpretationsonthecategoryonwhich it is marked.
22It is aquotefrom thepoetEinoLeino. Example[55c] is from Leino(1991).(55d)is astandardexamplewhich appearsin Finnishgrammarsfrom Setala 1884onwards.
24
If the languagepossessescorrespondingovert morphologicaldistinctions,theseextendedinterpretationswill not develop,presumablyin consequenceof the princi-ple thatspecificmorphologicalelementsblock generalmorphologicalelementsin theoverlappingdomain.Forexample,in modernGreek,imperfectiveaspectseemsto haveno NP-relatedfunctionat all, thedistinctionsin questionbeingexpresseddirectly bymeansof thechoiceof thedefiniteor indefinitearticle,whichRussianlacks.Thus,cor-respondingto thetwo sentencesin [10], of which thesecondis threewaysambiguous,Greekhasthefour unambiguoussentencesin (56).
Morphologicaleconomy(No VacuousAffixation, Marantz1984)would precluderedundantaspectualmarkingby partitive case,andredundantmarkingof NP-relatedfunctionsby aspect. In fact, when [–B] is morphologicallymarkedon the verb inFinnish,theobjectneednot bepartitive.Thus,frequentativeverbscantakeaccusativeobjectsin Finnish. Eventhoughthe VP is [–B] in virtue of the frequentativeverb,that featureis licensedby the verbalaffix andthereforeneednot be licensedby thepartitive. In this context,thepartitive/accusativecase-markingdistinctionrevertsto itspurelyNP-relatedfunction:
(57) Lue-skel-i-nread-freq-Past-1Sg
senthat-Acc
kirja-nbook-Acc
(loppu-un)(end-Ill)
‘I readthatbookoff andon (through).’
By thesametoken,we would expectthat in caseswhere[–B] is morphologicallymarkedonthenoun,theRussianperfective/imperfectivedistinctionwould revertto itspurelyaspectualfunction. Russianconfirmsthis prediction,but with a curioustwist.Theevidencecomesfrom thequantificationalgenitive,a casemostlyusedwith massnouns,which in certainparadigmsis morphologicallydistinct from the regulargen-itive. This so-called“secondgenitive” casehasbeenarguedto be a partitive case(Jakobson1936,Neidle1988,Franks1995). ThusRussianhas,undercertaincondi-tions, two morphologicalmeansof marking[–B] at its disposal:imperfectiveaspectfor aspectualfunctions,andpartitivecasefor NP-relatedfunctions.And indeed,in justthosecaseswherethenominal[–B] featurecanbemarkedby partitivecase,theverb
25
appearsin the perfectiveaspect,ratherthanin the imperfective,as it would with anaccusativeobject(Klenin 1978).Thepattern,asI havebeenableto reconstructit fromtheciteddiscussions,is shownin (58),with thetabulardisplayin (59); theseshouldberespectivelycontrastedwith (10)andwith (42),illustratingthepatternwithoutpartitivecase.
The unexpectedwrinkle, however,is that partitive casenormally appearsjust onobjectsof perfectiveverbs,not on objectsof imperfectiveverbs(Dahl andKarlsson1976:44,Klenin 1978).23 We might haveexpectedthepartitive/accusativedistinctionto be fully exploitedto yield a four-wayparadigmwhereunboundedreferencein thenominalandverbaldomainareseparatelymarked.
The generalizationin (50) seemsto be confirmedby aspectuallyinterpretedcasealternationsof Warlpiri andseveralPolynesianlanguages.
In Warlpiri, a classof verbshaveabsolutive(i.e. nominative)objectswhentheydenoteatelic event,anddativeobjectsotherwise,in whathasbeendubbedtheconativeconstruction(Hale1982,Laughren1988,Simpson1991):
(60) a. Ngarrka-ngkuka marlu luwa-rniman-Erg Pres kangaroo(Nom)shoot-Nonpast‘The manis shootingthekangaroo’
b. Ngarrka-ngkuka-rla-jinta marlu-ku luwa-rniman-Erg Pres-rla-jintakangaroo-Datshoot-Nonpast‘The manis shootingat thekangaroo’
Theconativealternationis foundwith verbsof impact(“strike”, “chop”, “carve”,“cut”,“”dig”, “pierce”, “grind”) andwith perceptionverbs(“see” vs. “look for”). Thereisalso a classof verbswith an inherentErgative-Dativecaseframe and with the ex-pectedatelic meaning(“seek,dig for”). This indicatesthat the meaningshouldbe
associatedwith thecaseframeitself, ratherthanwith a putativederivationalprocess.It seemsthat this casealternationshouldbe analyzedalongthe lines of the Finnishpartitive/accusativealternation,ratherthanasa lexical rule24 (Laughren1988)or by atransformationalprocessin thesyntax(BittnerandHale1993).25
Like theWarlpiri conative,theTonganmiddle is transitive,andthedativeobjectsin this constructionaresyntacticallyfull-fledgeddirectobjects,undergoingprocesseswhich are restrictedto direct objects(incorporationand passivization),or to directarguments(quantifierfloat),andthatanumberof constraintsvalid for transitiveclausesapplyto middleclausesaswell (Chung,p. 216-234).Chungargueson this basisthat
24Laughrenproposesthat the conativeform of the verb (the atelic one,which takesa dative object) isderivedby thefollowing rule:(61) (x produceeffect on y) by ((entity comeinto contactwith y) by (x manipulateentity))→ ((x ma-
nipulateentity) in order that (entity comeinto contactwith y)) in order that (x produceeffect ony).
This rule could becontestedon severalgrounds.First, the conativealternationis not associatedwith anaffix on theverb. (Conceivablyit couldbeonethesuffixeson theauxiliary. However,thatseemsdoubtful,asneitherof themseemsto bespecialto theconativeconstruction.-rla “registers”thepresenceof anydativeobject whatever,while -jinta marksa third singulardative object (Hale 1982).) We would either havetoassumea “spontaneous”lexical derivationalprocess,or postulatea triggering morphemewhich is alwaysnull (cf. Pesetsky1995). Secondly,by erasing/respecifyinginformation in lexical conceptualstructure,thederivationalprocessin [61] violatesa basicpropertyof monotonicitythat canotherwisebe maintainedforlexical processes.In general,derivationalruleseitheraddpredicatesor featuresto thebase(e.g.causatives,diminutives),or specifytheway theargumentsaresyntacticallyexpressed(e.g.passives,antipassives).Theproposedlexical rule would not do either of thesethings. Third, the rule doesnot expressany intrinsicrelationbetweentheconativemeaningof the“derived” verbandthedativecasethat it governs.
25Bittner andHaletaketheconativeconstruction[60b] to beanantipassive,i.e.aninstanceof theprocessusually analyzedasdemotionof the direct object. The absenceof a morphologicaltrigger is if anythingevenmoreof a problemfor theantipassiveaccountthanit is for thelexical rule account,because“relation-changing”processesalmostalwaysdependon diathesis-changingverbalaffixes, indeed,as far asI knowthereareno casesof systematic“spontaneous”antipassivization.Moreover,theWarlpiri conativeconstruc-tion is transitive.(Theevidenceincludesthefollowing: (1) thesubjectbearsergativecase,(2) thereis objectagreementin theAUX (see[60b]), and(3) thedativeobjectcancontrol -kurra clauses,which musthaveanobjectcontroller.) But transitiveantipassivesareotherwiseunattestedin languages;their absencefollowsfrom a principledtheoryof valencychange(demotionandadditionof argumentsastheonly mechanisms).
The translationalequivalentsof all Warlpiri conativeverbsmentionedby Laugh-ren,Hale,andSimpson,of all Tonganmiddleverbslisted by Chung(e.g.[62]), takepartitiveobjectsin Finnish(with theexceptionthatperceptionverbslike ‘see’ takeac-cusativeobjectsin Finnish). Theconverse,of course,is not true: partitiveobjectsareusedmuchmorewidely in Finnish.
It seemsclearthat theWarlpiri conativeandTonganmiddlecasealternationsareessentiallyaspectual,andakinto theFinnishpartitive/accusativealternation.However,no analogto the Finnish NP-relatedfunction of partitive caseis reportedfor theselanguages,seeminglycontradicting(50b).
But in both languages,theNP-relatedfunctionof thecasealternationis arguablyblockedby explicit morphology, asdiscussedabove.This is clearestin Tongan,wheredefinitenessis morphologicallymarkedby thearticlehe. ThusTonganwould be liketheGreekcasementionedaboveunder(56). As for Warlpiri, it hasa specialdefiniteplural suffix -patu (Nash1980:167).This suffix would presumablymakethedefinite-nessdistinctioncorrespondingto (56a,c)vs. (56b,d).26 Nevertheless,it is not clearto me why the conativecaseoppositionshouldnot distinguishbetweendefiniteandindefinitereadingsof singularmassnounsin Warlpiri.27
Thematerialdiscussedin thissectionsuggeststhatcoercionis in mostcasesalocalaccommodationof aV or NPto therespectivemorphologythatit bears(generalization(50)). In thiswayit differsfrom thebasicdeterminationof boundedness,whichis doneat theVP levelasdiscussedin earliersections.
7 Partitive subjects
In additionto partitiveobjects,Finnishalsohaswhataretraditionallycalledpartitivesubjects.Theyshowonly theNP-relatedfunctionof partitivecase,nevertheaspectualfunction.Thus,only bareindefinitepluralsandmassnounscanbepartitivesubjects:28
(63) a. Karhubear-SgNom
kuol-idie-Pst-3Sg
‘The beardied’
b. #Karhu-abear-SgPart
kuol-idie-Pst-3Sg
26Thanksto Mary LaughrenandDavid Nashfor pointingout this crucial fact.27TheScottishGaelicsystemstudiedby Ramchand(1993)is alsointerestingin this respect.Ramchand
Syntactically, partitivesubjects(andpresentationalsubjectsin general)differ fromregularsubjectsin their word order. Theyappeareitherpostverbally((67a,b),contrast(68a,b)), or preverballyin Spec-VPposition((67c,d)).If theyareplacedpostverbally,andthesentencebeginswith alocativeadverbial,theSpec-VPpositioncanbeoccupiedby aclitic pronounsita “it” in thepartitivecase((67e),contrast(68e)).
(67) Presentationalintransitives:
a. Nytnow
onbe-3Sg
synty-nytborn-PP
lapsichild-SgNom
(lapsi-a)(child-PlPart)
‘Now achild hasbeenborn’
b. Sytty-ibreakout-Pst-3Sg
sotawar-SgNom
‘War brokeout’ (Vilkuna1989,165)
c. Lapsichild-SgNom
(lapsi-a)(child-PlPart)
onnow
nytbe-3Sg
synty-nytborn-PP
‘The/achild (children)has(have)beenbornnow’
d. Sotawar-SgNom
sytty-ibreakout-Pst-3Sg
‘War brokeout’
e. Siellathere
si-tathat-Part
synty-ybeborn-Pst-3Sg
(lapsi-a)child-PlPart
‘Therechildrenarebeingborn(all thetime)’
Non-presentationalintransitives:
30
(68) a. #Hymyil-ismile-Pst-3Sg
lapsichild-SgNom
(lapsi-a)(child-PlPart)
‘The/achild (children)smiled’
b. #Voitt-i#win-Pst-3Sg
ruotsalainenSwede-SgNom
(ruotsalaisia)(Swede-PlPart)
c. Lapsichild-SgNom
(#lapsi-a)(child-PlPart)
hymyil-ismile-Pst-3Sg
‘The/achild (#children)smiled’
d. RuotsalainenSwede-SgNom
(#ruotsalais-i-a)(#Swede-PlPart)
voitt-iwin-Pst-3Sg
‘The/aSwede(#Swedes)won’
e. #SiellaThere
si-tathat-Part
hymyile-esmile-3Sg
laps-i-achild-SgNom(child-PlPart)
‘Therechildrenaresmiling (away)’
Thesedatacan be accountedfor on the following assumptions.First, partitivecaseis assignedonly inside VP; NPs in Spec-Inflget nominativecaseobligatorily.By the basicconstraintof Finnishsyntaxthat a VP cancontainonly a singledirectinternalNP argument(Siro’s Law, seep. 10 above),VP-internalsubjectsoccuronlywith intransitiveverbs.Therefore,partitivesubjectsarerestrictedto intransitiveverbs.
The secondassumptionto be madeis that VP-internalsubjectsare licensedbylocativearguments,whichmaybeexplicit or implicit. Verbswhichhavesuchlocativeargumentsconstitutethe classof “presentationalverbs”. Therefore,VP-internalsub-jects,andpartitivesubjectsin particular,arerestrictedto presentationalverbs,andcanalwaysco-occurwith locatives.
Thesubjectof a presentationalverbis partitiveif it is unbounded:
(69) a. PoikaBoy (Nom)
saapu-iarrive-Pst3Sg
‘The/aboy ([+B)] arrived([+B]).’
b. Poik-i-aBoy-Pl-Part
saapu-iarrive-Pst3Sg
‘Boys ([–B]) arrived([+B]).’
Partitivesubjectsdiffer from partitiveobjectsin oneimportantrespect:theydonotmark unboundednessat the VP level. Eventhoughthe VP in (70) is unbounded,invirtueof theverbalpredicate,thisdoesnot licensea partitivesubject:
(70) a. KarhuBear
lymyile-e(Nom) lurk-3Sg
‘A bear([+B]) is lurking ([–B]).’
b. *Karhu-aBear-Part
lymyile-elurk-3Sg
‘A bear([+B]) is lurking ([–B]).’
31
On subjects,partitivecasemarkstheunboundednessof theNP itself. In otherwords,subjectsof all presentationalintransitivespatternlike objectsof boundedverbalpredi-cates,in thatpartitivecaseonthemhasonly theNP-relatedfunction.
8 The evolution of the aspectualpartitive
Thepartitivecasewasoriginally a spatialcasewith separative(“from”) meaning.Itsunboundedness-markingfunctionsdevelopedwithin Balto-Finnic. The original nu-cleusof the innovationhasbeenthoughtto bea classof intensionalverbsthatgovern“quirky” lexical partitivein Mordvinianandin certaindialectsof Lappish(E. Itkonen1972,1973,T. Itkonen1976). It hasalso beenarguedthat the usesof the partitiveareborrowedfrom the Baltic genitive(Larsson1983,1984). Larjavaara(1991)hasproposeda hypotheticalscenarioof theevolutionfrom separative(“from”) case,via aquantificationalmeaning(similar to Frenchdeasin mangerdupain ‘to eatbread’),totheaspectualfunctionof Finnishandits closerelatives.
Outsideof theBalto-Finnicsubfamily, limited aspectualfunctionsof partitivecaseareattestedin theearliestrecordsof thenow extinctSoutherndialectof Lappishoncespokenin Sweden(E. Itkonen1972,1973).30 But by far themostimportantevidenceevidencefor theoriginal systemcomesfrom Mordvinian(Itkonen1972:166,Larsson1983:124).
31Accusativecaseis morphologicallyidenticalwith thegenitive. In somedialectsa possessedobjectcanalsobealsonominativeif it is inanimateandthepossessoris third person(Itkonen1972:167).
33Mordviniangrammarscall it “ablative” case,thoughto avoidconfusionwith theFinnishablativeI willnotusethat termhere.
33
d. vanalook
kitwho
salcitsteals
monmy
praka-do-npie-PartSg-Indef
‘Look who’sstealingmy pies’
Verbsof thetype“eat”, “drink”, and“smoke”havethepropertiesin (18)and(19).They arethe prototypicalmembersof the verb classwhich is privilegedby Krifka’stheoryof thepartitive(section4). We canin fact saythattheMordvinianpartitivehasthemeaning“part of” asdefinedin (18a).Accordingly, it hasa muchmorerestricteddistributionthat the Finnishpartitive,andin particularis not triggeredby mereirre-sultativity (“shotat a bear”getsaccusative).As a complementto verbsof therelevantclass,it yieldsanindefinitebareplural or massobjector atelic interpretation(processor iteration).
Thedistributionof inessive(sometimescalledinessive-instrumental)objectscon-tributesanotherpieceto the puzzle. Inessiveobjectsaremorphologicallyindefinite,anddo not triggerobjectagreement,but theyaresemanticallyinterpretedasdefinite.Theyareassignedby the atelic versionof verbsthat areintrinsically unspecifiedfortelicity, suchas “carry”, “hit”, “chew”, “seek”, “read”, “wait for”, “send”, “chide”,“suckle”, “bring”:
(75) a. sov-nichide-3Sg
Vana-soIvan-InessSg
‘He chidesIvan’
b. davajSo
cavmobeat
esnendehe-InessSg
‘So, hebeganto beathim’
In the telic version, the sameverbsassignaccusative(or nominative)caseto theirobjects.Thecorrespondingverbsin Finnishalternatebetweenpartitiveandaccusativeobjects,dependingon telicity. This is in effect the Mordvinian instantiationof theconativealternation,involving theaspectualfunctionof objectcasemarking.
Thefact thatpartitiveandinessiveobjectsneveragreewith theverbsuggeststhatunlike the Finnishpartitive, which is syntacticallya structuralcase,the Mordvinianpartitiveandinessivearesyntacticallyinherent(lexical,oblique)cases.Theyalsocan-notbesubjectcasesin Mordvinian(unlikewhatis thecasein Finnish).
Thenewcompositeelativeandablativecasestook over the local functionsof theoldpartitive-ta. It retainedonly themeaning“part of”, thusbecominga strictly partitivecase(ateapples,ateporridge,ateof theporridge).
Partitivecasemusthavethendisplacedinessivecaseasamarkerof definiteobjectsof atelic(unbounded)verbsin theconativeconstructionshownfor Mordvinianin (75).With this change,the old functionof partitivecaseof markingNPswith unboundedreferencebecamegeneralizedto markingVPs with unboundedreference(carriedthefirewood(aways),hit at theman,chewedon thebone).
Now partitivecasecouldbeextendedto definiteandcountNPsof atelic verbsingeneral(loved the woman,neededa knife, tendedthe reindeer). Thus the partitivecomesto simplymarkunboundednessat theVP level.
Why did the aspectualfunctionof partitivecaseemergeonly in the Balto-Finnicbranch?(Elsewherein Finno-Ugrictheconativeconstructionremainedunchangedorwassimply lost.) AnotheruniqueBalto-Finnicdevelopmentwasthegrammaticaliza-tion of partitive into a structuralcase.34 It is temptingto makea causalconnectionbetweenthesetwo changes.Recall that theFinnishpartitive licensesunboundednessat the VP level, irrespectiveof whetherthe unboundednesscomesfrom the headVor from the object. Let us supposethat a lexical casemustbe interpretedin a local,compositionalfashion.If thatis thecase,thentheBalto-Finnicfunctionof thepartitivecouldnothaveemergedin theotherFinno-Ugriclanguages.Thegrammaticalizationofpartitivecasewouldbeapreconditionto its developmentasamarkerof unboundednessat theVP level.
9 Conclusion
The commonfactor of the aspectualand NP-relatedfunctionsof partitive caseandimperfectiveaspectis markingaVP’sunboundedness.A VP hasthispropertyin virtueof havingeitheranunboundedheadoranunboundedargument.Historical-comparativeevidencesuggeststhat thepartitive’s emergenceasa structuralcaseis a preconditionfor theriseof its aspectualfunction.
34Presumablythis developmentis relatedto the fact that only the Balto-Finnicbranchhaspartitive sub-jects;seesection7.
35
Bibliography
BITTNER, MARIA AND KEN HALE. 1993.“Ergativity: towardsa theoryof a hetero-geneousclass.” Ms.,RutgersUniversityandMassachusettsInstituteof Technol-ogy, Cambridge,Massachusetts.
JAKOBSON, ROMAN. 1936/1962. “Beitrag zur allgemeinenKasuslehre.” Selectedwritings,2. TheHague:Mouton.
KLENIN, EMILY. 1978. “Quantification,partitivity, and the genitiveof negationinRussian.” In BernardComrie (ed.) Classificationof grammaticalcategories.Edmonton:LinguisticResearch.
KORHONEN, MIKKO . 1981.Johdatuslapin kielenhistoriaan.Helsinki: SuomalaisenKirjallisuudenSeura.
36
KRIFKA, MANFRED. 1989.“Nominal reference,temporalconstitution,andquantifi-cation in eventsemantics.” In RenateBartsch,Johanvan Benthem,andPetervanEmdeBoas(edd.)Semanticsandcontextualexpressions. Foris:Dordrecht.
KRIFKA, MANFRED. 1992. “Thematicrelationsaslinks betweennominalreferenceandtemporalconstitution.” In Ivan A. SagandAnnaSzabolcsi(edd.) Lexicalmatters. Stanford:CSLI.
LARSSON, LARS-GUNNAR. 1984.“The role of Baltic influencein theaspectualsys-temof Finnish.” In CasperdeGrootandHannuTommola,(ed.) Aspectbound.Foris: Dordrecht.