ThreemoresubjecthoodfeaturesinPanini’straditionArtemijKeidan,SapienzaUniversityofRome

artemij.keidan@uniroma1.it

Paniniandhisschool• Panini,around500BC(dateuncertain)

• Panini’sAṣṭādhyāyī:mostinEluentialgrammarinAncientIndia

– powerful,anticipatingandreEined

– dealingwithSanskrit(mostlymorphology)

– descriptivebutalsoprescriptiveandcontrastive2

Paniniandhisschool• DealingwiththeAṣṭādhyāyīwemustconsiderthreelanguages:

– objectlanguage=Sanskrit(perhapsLateVedic)

– descriptionlanguage=specialalgebraiccode

– audience’slanguage=mothertongueofthegrammar’susers

…perhapsaMiddleIndo-Aryanvariety

3

Paniniandhisschool• IsobjectlanguageunnaturalSanskrit?

– somecallitgrammarians’Sanskrit

• DidtheaudiencespeakSanskrit?

– onlyasL2(whencethegrammarisneeded)

– theirmothertongueisvisiblecontrastivelyinthegrammarandintheexamples

4

Paniniandhisschool• CommentatorsoftheAṣṭādhyāyī

– Katyayana(IIIc.BC):varttikas‘glosses’

– Patanjali(IIc.BC):bhāṣyas‘explanations’

– Katyayana+PatanjaliformtheMahābhāṣya

• LatercommentatorstoAṣṭādhyāyīorMahābhāṣya

– mostimportant:Bhartrhari(Vc.AD)

5

SubjectinSanskrit• Sanskrit,asmanyancientIElanguages,hadasubjectwithjustafewfeatures

• CommonopiniononPanini’sgrammar:

– thereisnosubjecthere

…becauseithadkārakas‘semanticroles’

…becausesubjectisnotverypivotalinSanskrit6

SubjectinSanskrit• J.S.Speijer,SanskritSyntax,1886

«Vernaculargrammarhasnotermtonamethesubjectofthesentenceorgrammaticalsubject»

• G.Cardona,“Panini’skarakas:agency,animation,andidentity”,J.Ind.Phil.,1974

«Panini’sgrammarischaracterizedbyanimportantabsence:thenotionofgrammaticalsubjectisabsent»

7

but• ScholarsdonotalwaysunderstandPanini

– nosemanticrolesintheWestuntilFillmore

• NogooddeEinitionofsubjectwasathand

– Speijerreferstotheloose“subject”ofthegrammarschool

– CardonareferstoChomsky’s“externalargument”8

Mysuggestion• Let’sseekforKeenan’sfeaturesin:

– thegrammaticalrulesoftheAṣṭādhyāyī

– thecommentators’innovations

– thelinguisticexamplesdiscussedbythem

• Let’sconsidertheaudience’slanguage,ratherthanSanskrit

9

Panini’ssyntax• Semanticrolesvs.morphologicalforms

• Semanticroles(kāraka):

10

– apādāna‘source’

– sampradāna‘goal’

– adhikaraṇa‘locus’

– karaṇa‘instrument’

– karman‘patient’

– kartṛ‘agent’

Panini’ssyntax• KārakasareexplicitlydeEinedinsixdeEinitionalsūtras

– etymologyofkārakatermsplaysnorole

– deEinitionsaresemantic,butmoreabstractandexplicit

11

Panini’ssyntax•Morphologicalrealisationsofkārakas:

– Einiteverbalendings

– caseendings

• Thetwooptionsaremutuallyexclusive

– noideaofagreement(inPanini)

12

Panini’ssyntax• Consideringthecase-formsexpressionofkārakas

– noone-to-onerelation,inbothways

– case-formsaresemanticallyblind

– onecanonicalrealisation+someoptionalones

13

Panini’ssyntax

14

karman‘patient’Accusative

canonically

Instrumentaloptionally

karaṇa‘instrument’kartṛ‘agent’

canonically pitrāsaṁjānīte

‘heacknowledgeshisfather’

Panini’ssyntax• Let’sconsiderkartṛ‘agent’

– etymologically‘thedoer’,butthisisignored

– nosemanticspecialization:macrorole?

– svatantra‘independent’

• CanonicallyexpressedbytheInstrumental

15

Panini’ssyntax• Additionalsūtraswherekārakarolesareamended

– amendmentslesselegantthanthedeEinitions

– basedontheconfusionbetweencasesandroles

• MostscholarsconsiderthemtogetherwithdeEinitions

– resultingcategoriesareodd

16

Excursus1:exampleofakārakaamendment

• semanticrolenamedkaraṇa‘instrument’isdeEinedas“themosteffectivemeans”

• itscanonicalcase-formrealisationistheInstrumental

• withtheverbdiv‘toplaydice’theinstrumentiscodedwiththeAccusative

• herethe“mosteffectivemeans”correspondstokarman‘patient’

• therefore,karmanispatient+whatevergoesinAccusative17

Panini’ssyntax• ConsideringkārakadeEinitionsonly

– ismoreconsistent

– avoidspostulatingmixedcategories

• Goodevidencethattheamendmentsarespurious

– let’signorethem

18

Panini’ssyntax

19

the semantics side, he distinguishes six categories, named kārakas, which are quitesimilar to our semantic roles. On the syntactic side, he surveys all the nominal casecategories, named vibhaktis, and assigns a few different functions to each of them,among which there is also that of coding the kāraka roles. The latter can be expressedalso by other morphological means, such as derivative suffixes and, surprisingly, ver-bal endings (personal agreement markers). Moreover, the nominal coding of kārakascomes as the last option, after it is ascertained that the other possibilities have notbeen used (therefore, only one expression per kāraka is admitted). It is also to be notedthat the nominative case is not assigned to any kāraka. The kāraka role that resemblesour semantic role of agent is called kartṛ. Its canonical realization through vibhaktiis tṛtīyā ‘third case’, i.e. the instrumental (rather than the nominative, as we wouldexpect); optionally it can also be expressed by the genitive. Alternatively, kartṛ is ex-pressed by the active verbal endings or by some agentive suffixes. See the followinganalysis of a couple of typical Sanskrit sentences; grammatical glosses are provided,with the indication of the kāraka roles “expressed” by each word, if any.11 Besides kartṛ,karman is also mentioned, which corresponds to the undergoer or patient semantic(macro)role in the modern system.(2) a. pacaty

cook:3sg.actkartṛ

odanaṃrice:acckarman

DevadattaḥD.:nom{no kāraka}

‘Devadatta is cooking rice’b. odanaḥ

rice:nom{no kāraka}

pacyatecook:pass:3sg.midkarman

DevadattenaD.:instrkartṛ

‘Rice is being cooked by Devadatta’Obviously, the karaka/vibhakti device accounts very well for both active and passivesentences. As we can see from the functional labelling, while the semantic roles re-main unchanged, their morphological encoding changes. Three descriptive odditiescan be highlighted here.

i. Only single exponence is admitted: Pāṇini «(…) adopts the one-to-one corre-spondence between morphological elements and morphosyntactic features»(Kiparsky 2002: 45), i.e. there appears to be no idea of anaphora or agreement.

ii. One of the morphological means of expressing the arguments’ semantic rolesis the verbal endings, which is quite unusual — not to say inconsistent — withhow we normally describe the morphology of the ancient IE languages.

iii. No precedence is reserved for the active voice over the passive: both are just twoequiprobable distributions of kartṛ and karman within sentence morphology, inno anyway “derived” from each other, see Cardona (1974: 286, fn. 36).

same theory is provided in Kiparsky (2002). For a critical review of some interesting aspects of this systemsee also Keidan (2007).

11. Note that this is meant in the Indian sense of “expressing”: either the nominal case termination orthe verbal endings can express the kārakas. This explains the unusual placement of the kāraka labels in theexamples (2) to (4).

8

Panini’ssyntax• Thisexampleisinferablefromthegrammar

– wecanconsiderit“paninian”

• Nopreferenceforeitheractiveorpassivevoice

– bothconstructionsareequallylikely

– vivakṣā‘communicativeintention’istheguidingprinciplehere

20

Panini’ssyntax

21

definition does not refer to any other kāraka.16 In any case, the term svatantra ‘inde-pendent’ resembles the modern phrase “privileged argument”: subject is the only ar-gument capable of being qualified as independent, whatever idea of “independence”we may have. 17 The defense of kartṛ’s independence is made explicitly by Bhartṛhari,who lists a few qualities that characterize it, see Cardona’s (1974: 239) summary. Butespecially, this definition is highly abstract, i.e. detached from the semantics of con-crete verbs, which fits quite well with our understanding of subjecthood. Sanskrit cer-tainly lacked a strong notion of subjecthood, but grammarians’ mother tongue pos-sibly did have one. So, again, this definition could have been an attempt to reconcilethe official grammar with the linguistic feeling of the audience.

Topicality. The last, and most important, feature to mention is the fact that kartṛ isthe target of a set of transformations corresponding to what modern linguistics callsactancy derivation and voice. These phenomena are not mentioned directly by Pāṇiniand are only known from the commentators, starting from Patañjali, who introducethem as a problem: there are some sentences that are perceived as a challenge forPāṇini’s definition of the kartṛ, and then a solution is suggested. Let us start fromanalyzing the sentences in question:

(3) a. asināaxe:instrkaraṇa

chinatticut:3sg.actkartṛ

devadattaḥD.:nom

‘Devadatta is cutting [stuff] with an axe’b. asiś

axe:nom{no kāraka}

chinatticut:3sg.actkartṛ

‘The axe cuts [by itself]’

(4) a. devadattaḥD.:nom{no kāraka}

sthālyāṃpot:locadhikaraṇa

pacaticook:3sg.actkartṛ

‘Devadatta is cooking in a pot’b. sthālī

pot:nom{no kāraka}

pacaticook:3sg.actkartṛ

‘The pot cooks’

16. All the other kāraka definitions either refer to the kartṛ explicitly or are commented upon by thecommentators with reference to it. For example, the definition of karman ‘patient’ is kartur īpsitatamam ‘themost desired by the kartṛ’. Another possible interpretation puts the rule defining the kartṛ in comparisonwith the next one, where hetu ‘causative agent’ is introduced, from which the main agent is, in some way,“dependent”, see Freschi & Pontillo (2013: 47).

17. Interestingly, also the European philosophers and grammarians of the Middle Ages mentioned a verysimilar phrase per se stans ‘standing by its own’ while defining such notions as subject, substantive and thelike, see Alfieri (2014). It almost literally translates Sanskrit svatantra.

11

Panini’ssyntax• ThisexampleisaddedbyPatanjali

– let’sconsiderit“post-paninian”

• Unmarkedactancyderivation

– derivedconstructionislesslikely

– vivakṣā‘communicativeintention’isstillrelevant

22

Iskartṛasubject?• Modernscholarshaveconsideredkārakastobe

– equivalenttocases(Whitney)

– equivalenttosemanticroles(afterFillmore)

– some“syntacto-semantic”categories(Cardona)

23

Iskartṛasubject?• Commentators:everykārakacan“becomekartṛ”

– literally:everysemanticrolecanbecomeagent

– betterinterpretation:everysemanticrolecanberaisedtothesubjectposition

• Therefore,kartṛisthetargetoftopicality-driventransformations

24

Iskartṛasubject?• Anotherconsequence:kartṛisalwayspresent

– alwaysexpressed,eitherbyacase-formorbyaEiniteverbtermination

• Therefore,kartṛisanobligatoryargument

– thisisanothersubjecthoodfeature

25

Excursus2:middleterminations• Activeterminationsexpressthekartṛ

• Middleterminationsexpressthekarman

• Butinnon-passiveverbsmiddleterminationsexpresskartṛinstead

– karmavatkartṛ‘patient-likeagent’

– thissavestheobligatorinessofkartṛ26

Conclusions• Kartṛhasatleastthreesubjecthoodfeatures:

– semanticallynonspeciEic(macrorole?)

– obligatorilypresentineverysentence

– targetoftopic-driventransformations

• Shouldweconcludethatkartṛissubject?

27

Conclusions• Kartṛissubjectincaseweconsider:

– post-paninianevolutionofthegrammar

– deEinitionswithoutamendments

– audience’slanguage,ratherthanobjectlanguage

– examplesentences,besidesgrammaticalrules

28

References•Al-George,S.1958.“LesujetgrammaticalchezPanini”.StudiaetActaOrientalia1:39–47.• Bhate,S.1996.“Grammarian’sLanguage”.InAmṛtamandākinī.G.B.Palsulefelicitationvolume,ed.byS.Bhateetal.,90–97.• Cardona,G.1974.“Panini’skārakas:agency,animation,andidentity”.JournalofIndianPhilosophy2:231–306.• Cardona,G.1976b.“SubjectinSanskrit”.InTheNotionofSubjectinSouthAsianLanguages,ed.byM.K.Verma,1–38.•Deshpande,M.M.1980.EvolutionofSyntacticTheoryinSanskritGrammar:SyntaxoftheSanskritInUinitive-tumUN.AnnArbor:KaromaPub.•Hook,P.E.1980.“Aṣṭādhyāyī3.3.158andthenotionofsubjectinPanini”.RevueRoumainedeLinguistique25:79–87.•Houben,J.E.M.1999.“‘Meaningstatements’inPanini'sgrammar”.StudienzurIndologieundIranistik22:23–54.•Keidan,A.2015.“Form,functionandinterpretation:acasestudyinthetextualcriticismofPanini’sAṣṭādhyāyī”.Bulletind’ÉtudesIndiennes32:171–203.

29