Version classiqueVersion mobile

Bilingual discourse and cross-cultural fertilisation: Sanskrit and Tamil in medieval India

 | 
Whitney Cox
, 
Vincenzo Vergiani

Section II. Regulating language: grammars and literary theories

The adoption of Bhartṛhari’s classification of the grammatical object in Cēṉāvaraiyar’s commentary on the Tolkāppiyam

Vincenzo Vergiani

Texte intégral

  • 1 I would like to express my gratitude to Jean-Luc Chevillard for the many stimulating conversations (...)

1In this article1 I will look at one episode in the long history of the interaction between the Sanskrit and Tamil grammatical traditions and draw from it some — largely tentative — conclusions in the hope that they may help to cast some light on larger processes at work.

2The episode in question is quite a clear-cut case of conceptual borrowing. It consists in the adoption of a semantic classification of the grammatical object, first formulated by the Sanskrit author Bhartṛhari (probably 5th century CE), at the hands of Cēṉāvaraiyar (13th–14th century CE), a mediaeval Tamil commentator of the Collatikāram (TC) of the Tolkāppiyam (T.). In order to assess the full significance and the wider implications of this borrowing, I will first have to situate both authors within their respective scholastic traditions. In particular, their views on the grammatical object will need to be seen in the broader context of each school’s treatment of cases. This is a vast, complex and a highly technical topic, and even more so in an inter-linguistic perspective like the one attempted here. I will therefore have to be content with providing an inevitably sketchy outline of the two systems before narrowing my focus down to the grammatical object, and with that focus in mind I will also briefly discuss other broadly coeval texts and traditions, in the hope that even a simple presentation of textual data will give a sense of the complexity of the underlying socio-cultural dynamics.

1. The grammatical object (karman) in the context of Pāṇini’s kāraka system

  • 2 Pāṇini’s date, like most dates in early South Asian history, is the object of an often heated deba (...)
  • 3 For example, a simple sentence such as yogī parvatād āgacchati “the ascetic comes from the mountai (...)

3Bhartṛhari, the celebrated grammarian and philosopher of language, is an exponent of the principal school of Sanskrit grammar initiated by Pāṇini (ca. 5th–4th century BCE2), who established a theoretical model that generally prevailed throughout the history of Sanskrit linguistic thought, even among the self-styled non-Pāṇinian schools. In brief, this is a highly formalistic derivational model, postulating abstract elements, namely bases — whether verbal or nominal — and affixes, for which it provides the rules of combination leading to the formation of correct Sanskrit words and sentences.3

  • 4 There is a vast secondary literature on kārakas. Among the most significant contributions see Card (...)
  • 5 I will deal in greater detail with the definitions of the object below.

4Within this model, a crucial function is performed by the kāraka system.4 Kāraka is the general name given by Pāṇini to the six semantic-syntactic categories corresponding to the roles that a nominal item may play in a Sanskrit sentence when it is directly related to the action: kartṛ “agent”, karman “object”, karaṇa “instrument”, adhikaraṇa “locus”, saṃpradāna “recipient” and apādāna “point of departure”. For each of these, Pānini gives a broad semantic definition in one of the sūtras of the section A. 1.4.24–55. Thus, for example the agent is said to be the entity which is “independent” (svatantra) in relation to the action and the instrument “that which is most effective” (sādhakatama) in bringing it about.5

5As pointed out by Cardona (1974: 279–80, n. 1), Pāṇini’s kārakas are not

purely notional, semantic, extra-linguistic categories, totally separated from their linguistic representations and uninfluenced […] by grammatical considerations. [They] are essentially syntactic categories for which Pāṇini gives semantic characterizations. […] The undeniable fact is that Pāṇini not only arrived at his kārakas through an analysis of Sanskrit syntax, but also took such syntactic features into consideration in [the relevant] rules.

6In other words, the so-called kārakas singled out by Pāṇini are semantic categories describing certain basic relations of things with actions as they are grammaticalised in Sanskrit. While Cardona’s remarks refer to the kāraka system in particular, they equally hold for any other linguistic feature — whether phonological, morphological, syntactical or sentential — taken into account in the A. Pāṇini’s grammar was neither meant to be universally valid nor to apply to any specific language other than Sanskrit, even though many of the theoretical assumptions on which it is based may certainly be generalised to language in general, so as to serve as the foundation for the description of some other language. Consequently, any such attempt based on the Pāṇinian model would have to calibrate it to the specific characteristics of the language in question. If there is great typological difference between this language and Sanskrit, one may reasonably expect not so much an adaptation of the original model as the borrowing of selected terms and concepts, the validity of which is amenable to be extended to other languages than Sanskrit. As will be shown below, this is precisely the kind of strategy the Tamil grammarians seem to have followed, though — significantly — to a different extent in different ages.

  • 6 As is well known, there are seven declensional endings in Sanskrit (the vocative being treated by (...)
  • 7 The actual morphemes are represented by small letters, while the capital letters are metalinguisti (...)
  • 8 These are of course the western terms. The Sanskrit tradition simply calls them with their ordinal (...)
  • 9 Thus, e.g., A. 7.1.9, ato bhisa ais, prescribes the substitution of the morpheme ais for the instr (...)

7In the initial stages of the Pāṇinian process of derivation, when a nominal item in a sentence corresponds to the description given in a rule of the kāraka section, it is assigned the corresponding kāraka designation. In its turn, this causes the introduction in the next stage of a particular declensional ending (vibhakti).6 These are all listed in what is assumed to be their “basic” form in A. 4.1.2: sV-au-Jas; am-auṬ-CHaṣ; Ṭā-bhyām-bhis; Ṅe-bhyām-bhyas; ṄasI-bhyām-bhyas; Ṅas-os-ām; Ṅy-os-suP.7 This list consists of triplets (separated here by a semi-colon), each of them including the singular, dual and plural morphemes (separated by a hyphen) for one of the seven Sanskrit morphological cases, in the following sequence: nominative, accusative, instrumental, dative, ablative, genitive and locative.8 Most of these are replaced by allomorphs enjoined by specific rules for certain classes of nouns, usually depending on the gender and the final sound of the stem.9

  • 10 The rule is required because with other verbs the prayojya maintains its original kāraka designati (...)

8As far as karman, “object”, is concerned, Pāṇini lists three major rules assigning this designation under different semantic conditions. A. 1.4.49, kartur īpsitatamaṃ karma, defines the object as “that which the agent most wishes to attain” (īpsitatama), implicitly assuming the voluntary act of a conscious agent as the metaphorical prototype of any relation between agent and object through an action. A. 1.4.50, tathāyuktaṃ cānīpsitam, extends this definition to an object that the agent does not wish to attain (anīpsita), but which is linked in the same way (tathāyuktam) to the action as the īpsitatama object, and A. 1.4.51, akathitaṃ ca, classifies as karman that which is “not designated (akathita) [otherwise]”, namely the secondary object in double-accusative constructions, which are quite common with certain Sanskrit verbs, such as brū- “to say”, prach- “to ask”, duh- “to milk”, nī- “to lead”, etc. A fourth rule, A. 1.4.52, gatibuddhipratyavasānārthaśabdakarmākarmakāṇām aṇi kartā sa ṇau, assigns the designation karman to the prayojya, the instigated agent or causee, in a causative construction with verbs of motion, knowledge, eating, those whose object is a sound and intransitives.10

  • 11 Namely, by means of either a verbal ending (as I explain below), or a primary or secondary suffix, (...)

9When an item is named karman, it will be followed in principle by one of the endings of the triplet called dvitīyā, i.e. the accusative, according to A. 2.3.2, karmaṇi dvitīyā. However, all the rules in the vibhakti section of the grammar, A. 2.3.2–73, are subject to the condition stated in sūtra A. 2.3.1, anabhihite, lit. “if unexpressed”, meaning that a certain ending is introduced only provided that its sense is not already expressed otherwise.11

  • 12 According to A. 3.4.69, laḥ karmaṇi ca bhāve cākarmakebhyaḥ, in which kartari is to be read by anu (...)
  • 13 According to A. 2.3.46, prātipadikārthaliṅgaparimāṇavacanamātre prathamā, which prescribes the nom (...)
  • 14 According to A. 2.3.2, karmaṇi dvitīyā, already quoted above.
  • 15 According to A. 2.3.18, kartṛkaraṇayos tṛtīyā, whereby the instrumental is the “elective” case for (...)

10In the case of the agent and the object, the condition anabhihite serves to regulate the alternation between active and passive constructions. According to Pāṇini, a finite verbal ending will in fact denote either the agent or the object, and, in the case of intransitives, either the agent or the action (bhāva) itself.12 If the verb has an active ending, which denotes the agent — as a semantic-syntactic function — the nominal item referring to the agent will be assigned the nominative (prathamā), for the syntactic function is already expressed (abhihita),13 while the object, being unexpressed (anabhihita), will be denoted by the accusative;14 vice versa, if the verb has a passive ending, which expresses the object, then the nominal item denoting the object will be followed by the syntactically “neutral” nominative, while the agent will be denoted by the instrumental.15 Thus, one obtains a basic pair of alternative derivational strings, which may be represented as follows in Pāṇinian terms:

bālaḥ odanaṃ khādati “the boy eats rice”
bāla-s nom odana-am acc khād-ti3rd sg act

bālena odanaḥ khādyate “rice is eaten by the boy”
bāla-āinst odana-snom khād-ta3rd sg pass

  • 16 It is worth recalling that the grammatical category of subject does not exist in Pāṇini’s grammar, (...)

11Most importantly, this implies that any item designated as karman is potentially amenable to appear in the nominative, that is, as the subject16 of a passive verb.

2. The TC on the grammatical object (ceyappaṭuporuḷ) and the second case (ai-vēṟṟumai)

  • 17 While the term vēṟṟumai (lit. “difference”) is generally used for both notions, another term, urup (...)

12A markedly different approach to the treatment of cases and case endings (vēṟṟumai17) is found in the TC. The “Chapter on Cases” (Vēṟṟumaiyiyal) opens with a sūtra (T. 2.2.1 = TC 62) stating that the cases are seven (as in Sanskrit), but T. 2.2.2 immediately specifies that they are in fact eight with the vocative. Then, T. 2.2.3 lists them as follows:

  • 18 The translation of all the TC sūtras in this paper is my English rendering of Chevillard’s French (...)

avai tām
peyar ai oṭu ku
iṉ atu kaṇ viḷi eṉṉum īṟṟa.
“These [are]: the noun [i.e. the nominative], ai, oṭu, ku,
iṉ, atu, kaṇ and, to conclude, the vocative”.18

  • 19 Clearly, there is nothing “natural” about this order (for instance, cf. Latin grammar, where the a (...)
  • 20 Elsewhere (e.g. T. 2.2.4) it is called eḻuvāy vēṟṟumai, lit. “case of the origin/source” (evidentl (...)

13The order — from nominative to locative — is the same as that found in the abovementioned A. 4.1.2,19 except that the latter has no separate mention of the vocative. As can be seen from the translation, most of these (ai, oṭu, etc.) are designated by their (most common) case morphemes. Unlike the other cases, the nominative, being unmarked in Tamil, is referred to here with a word that simply means “name”.20

  • 21 T. 2.2.4 (TC 65), avaṟṟ’uḷ eḻuvay vēṟṟumai peyar tōṉṟu’nilaiyē.
  • 22 T. 2.2.14 (TC 75), nāṉk’ākuvatuvē ku eṉap peyariya vēṟṟumaik kiḷaviy epporuḷ āyinuṅ koḷḷum atuvē.

14In the following sūtras, for each case the TC first gives a broad general definition — thus, for example, the nominative is said to be “the state in which the noun manifests itself”21 while the dative signifies “the acceptance of some object”22 -, then a detailed characterisation of its functions, often in the form of a sūtra schematically enumerating some exemplary contexts in which the case occurs. Interestingly, most of these enumerative sūtras do not claim to be exhaustive, for they conclude the list of illustrations with the phrase aṉṉa piṟavum “and similar ones”, thus leaving the door open to other not explicitly sanctioned idioms.

15An identical pattern is found also in the case of the object. T. 2.2.10 (TC 71) gives the general definition:

iraṇṭ’ākuvatuvē
aiy eṉap peyariya vēṟṟumaik kiḷaviy
evvaḻi varinum viṉaiyē vinaik kuṟipp’
avviru mutaliṟṟ’oṉṟum atuvē.

  • 23 That is, according to Cēṉāvaraiyar, even when the ending ai is governed by a verbal noun, as in pu (...)
  • 24 kuṟippuviṉai, which Chevillard translates as “verbes idéels” (2008: 128, s.v., and Appendix C, 482 (...)
  • 25 Cf. the remark of Subrahmanya Sastri about TC 71 (1997: 220, n. 2 continued from 219): “… [Tolkāpp (...)

16“The second, the case designation that is called ai, wherever it is met,23 is that which appears as a primary cause of these two, verbs or notional24 verbs”. Unlike the definitions of the other cases, it is remarkably vague, as if the diversity of the grammatical object eluded any univocal description.25

17The next sūtra, T. 2.2.10 (TC 72), deals with the semantics of the object, illustrating it with a long, seemingly random enumeration of verbs that can have an object:

  • 26 “Du fait du garder, du ressembler, du chevaucher, du construire, du fait du chasser, du louer ou d (...)

kāppiṉ oppiṉ ūrtiyiṉ iḻaiyiṉ
ōppiṟ pukaḻiṟ paḻiyiṉ eṉṟā
peṟaliṉ iḻaviṟ kātaliṉ vekuḷiyiṭ
ceṟaliṉ uvattaliṟ kaṟpiṉ eṉṟāv
āṟuttaliṟ kuṟaittaliṟ ṟokuttaliṟ pirittali
niṟuttaliṉ aḷaviṉ eṇṇiṉ eṉṟāv
ākkaliṭ cārtaliṭ celaviṟ kaṉṟaliṉ
ṉokkaliṉ añcaliṭ citaippiṉ eṉṟā
aṉṉa piṟavum ammutaṟ poruḷav
eṉṉa kiḷaviyum ataṉ pālav eṉmaṉār.
“By guarding, resembling, riding, building,
by hunting, praising or blaming,
by obtaining or losing, loving or being angry at,
by hating, rejoicing or learning,
by cutting, decreasing, collecting or separating,
by weighing, measuring or counting,
by forming, concerning, going, craving,
by watching, fearing or ruining —
all such expressions, which are [nuances of] the sense of this
primary cause,
and other similar ones, are its varieties [of usage], they say”.26

18It is evident that, unlike the A., the T. assigns meanings directly to the case morphemes, without the intervening kāraka categories.

19However, an echo of Pāṇinian kārakas is perhaps to be seen (at least according to Cēṉāvaraiyar, as I will show below) in a sūtra found in the next section of the TC, namely the “Chapter on the blending of cases” (Vēṟṟumaimayaṅkiyal), which deals with instances where form and meaning do not match in the way one might expect on the basis of the previous chapter — a puzzling place for such a sūtra, a fact for which I cannot see any obvious explanation. The sūtra, T. 2.3.29 (= TC 112), reads:

viṉaiyē ceyvatu ceyappaṭuporuḷē
nilaṉē kālaṅ karuviy eṉṟāv
iṉṉataṟk’itu payaṉ āka veṉṉum
aṉṉa marapiṉ iraṇṭoṭ’ un tokaiiy
āy eṭṭ’eṉpa toḻiṉ mutaṉilaiyē.

20This is translated by Chevillard (1996: 218) as follows:

21“The action, that which does [it], and the object that undergoes the action, the place, the moment and the instrument, [these six] being added to [those] two of similar tradition, namely “for such a thing” [and] “this being the gain”, these eight — they say — [are] the antecedents of the act”.

  • 27 nimittabhāvo bhāvānām upakārārtham āśritaḥ “One resorts to the nature of (auxiliary) causes of thi (...)
  • 28 yat tu kriyāpadopāttāyāḥ kriyāyā nimittaṃ tat kārakam eva “But that which is a causal factor in an (...)
  • 29 Chevillard (2008: 219, n. 5) thinks that the term mutaṉilai “looks like the Tamil calque of a Sans (...)

22What are these “antecedents” or “preconditions” of the action? And what purpose, if any, do they serve in the T.? The term mutaniḻai, lit. “previous state or condition”, seems to be used here in a sense similar to that of the Sanskrit nimitta, e.g. in VP 3.7.14ab27 or PrPr ad VP 3.7.24,28 as a cause or condition that makes the action possible,29 the idea being that the factors of its accomplishment must necessarily exist — logically and ideationally — before the action comes into being. In this sense, both kārakas and mutaṉilais can be said to be “preconditions” of the action.

  • 30 Similar ideas are also known to the Sanskrit tradition, but I cannot discuss them here.
  • 31 The wording of the sūtra suggests that these two may not have belonged to the original scheme (but (...)

23The TC list opens with viṉai, which Chevillard (2008: 167, s.v. toḻil) thinks may stand for “action” in general, that is, a sort of “prototypical” or possibly notional action, while toḻil, at the end of the sūtra, would be the individual (expressed) act.30 The list continues with ceyvatu, “that which does it”, i.e. the agent, followed by ceyappaṭuporuḷ “the thing undergoing the action”, i.e. the object, and then the place, the time and the instrument, to which TC adds two more that may be identified with the recipient and the goal.31

  • 32 Resorting to Chevillard’s insightful concept of “FIRST-translation” (see Chevillard 2009a: 73, n. (...)
  • 33 For similar considerations on the relation between veṟṟumais and kārakas, cf. Chevillard (2000b: 1 (...)
  • 34 If expressed grammatically, that is by means of case endings, different temporal complements are e (...)
  • 35 This is hardly surprising considering that classical Tamil does not have an equivalent of the abla (...)
  • 36 For this sūtra and its relevance in this context, see below.
  • 37 Cf. the Sanskrit equivalent avadhi, which is sometimes found in the place or as a gloss of apādāna (...)

24Interestingly, the list contains no Sanskrit loan-words except kālam, which is not — strictly speaking — a technical term in Sanskrit grammar and, in any case, is not a member of the set of Pāṇini’s kārakas. On the other hand, it is tempting to take ceyvatu and ceyappaṭuporuḷ, both containing forms of the verb ceytal “to make”, if not as proper “translations” of the Sanskrit terms kartṛ and karman, at least as inspired by these, which are etymologically transparent forms of the verb kṛ- “to make”.32 Anyhow, there is clearly no one-to-one correspondence between the list in the TC (whether in its shorter, six-item version or with the additional two members) and the six Pāṇinian kārakas.33 Besides viṉai, kālam too has no equivalent in the Sanskrit list of kārakas.34 On the other hand, the Tamil list, even in its expanded form, does not include an equivalent of apādāna,35 although Cēṉāvaraiyar construes T. 2.2.17–18 (TC 77–78), which treat the ablative, and T. 2.6.37 (TC 234), which deals with the possible antecedents of a peyareccam (“adnominal participle”),36 in such a way as to embrace the notion of ellai “limit”.37

  • 38 One only finds karuvi in T. 2.2.12 (defining the third case, i.e. the instrumental) and the pair n (...)

25It is worth pointing out that while individual kāraka designations are used by Pāṇini to trigger certain grammatical operations, and in particular the introduction of appropriate case endings, the mutaniḻai names do not play this role in the TC. Certain sūtras of the Vēṟṟumaiyiyal do mention some of these terms38 among the possible senses of a case morpheme, but the approach is merely descriptive, not derivational.

  • 39 “toḻiṉ mutaṉilai” eṉṟatu toḻilatu kāraṇattai. kāriyattiṉ muṉṉiṟṟaliṉ “mutaṉilai” āyiṟṟu. kāraṇam e (...)
  • 40 T. 2.6.37: nilaṉum poruḷuṅ kālamuṅ karuviyum viṉaimutaṟ kiḷaviyum viṉaiyum uḷappaṭav avvaṟu poruṭk (...)

26Despite the obvious differences between kārakas and mutaṉilais, commenting on T. 2.3.29 Cēṉāvaraiyar states that the expression toḻiṉ mutaṉilai refers to the causes (kāraṇam) of the action and then goes on to say that kāraṇam is a synonym of kārakam.39 Chevillard (1996: 219, n. 5) remarks that kāraṇam is a loan-word which has been fully integrated into Tamil (“un emprunt acclimaté”), while kārakam is an occasional loan, a technical reference (“un mot d’emprunt occasionnel… une référence technique”). Is Cēṉāvaraiyar suggesting that the TC is appropriating or adapting Pāṇini’s kāraka system? This would clearly be quite problematic, as I have pointed out. However, I would tentatively suggest that Cēṉāvaraiyar’s intention is simply to recall the notion of kāraka — a theoretical construct that was presumably well known to his readers — and thus establish an analogy which could help to explain the concept of mutaṉilai, for both categories refer to a set of elements — both notional and grammatical — that are needed for an action to be cognised and expressed. According to Cēṉāvaraiyar, the significance of mutaṉilai within the TC is to be seen in relation to sūtra T. 2.6.37 (TC 234), which specifies the semantic-syntactic roles in which a peyareccam (“adnominal participle”) can occur by referring to the first six categories listed in T. 2.3.29.40

27To summarise, even a cursory survey shows that at the level of the foundational texts of either tradition there is a contrast which primarily originates from a different theoretical approach: while the A. is a derivational grammar providing the rules for the combination of bases and affixes through basic operations such as affixation, substitution, etc., the TC appears to be a purely descriptive/analytic grammar, even though — probably under the influence of Sanskrit grammar — it analyses words into their constituents and thus, for instance, recognises the existence of case morphemes (vēṟṟumai).

  • 41 The relatively few instances where this happens — such as the nominative singular of most stems en (...)
  • 42 Alternatively, the case of an unmarked noun may be conveyed by its position in the word order of t (...)

28As far as the treatment of cases is concerned, the divergent approaches may also reflect the intrinsic typological difference of the two languages: Sanskrit is an inflectional language, in which no nominal stem can occur without a case marker,41 and various morphemes may express the same case while simultaneously conveying differences in gender and number; in Classical Tamil, which is agglutinative, case markers are to a large extent optional,42 generally have only one form for all types of stems, and do not carry any other information than the case, while a different morpheme will for example convey number.

  • 43 For the use of passive in Tamil and its acknowledgment by the grammatical speculation (first occur (...)

29Moreover, one of the crucial functions of the designation karman in Sanskrit, namely that of allowing the derivation of passive sentences, is not needed in Tamil, for the use of passive — which later develops by means of periphrastic forms using the verb paṭutal, lit. “to come into existence, to appear”, as an auxiliary plus an absolutive of the ceya type — is extremely rare in the early period and receives no treatment in the TC.43

3. Bhartṛhari’s classification of the varieties of the grammatical object

  • 44 An example of this is the induced agent (prayojya) or causee in a causative construction with cert (...)

30A few centuries after the composition of the A., Bhartṛhari presents a sevenfold classification of the grammatical object in VP 3.7.45–46: the īpsitatama object (A. 1.4.49) is divided into 1) nirvartya “to be produced”, 2) vikārya “to be modified”, and 3) prāpya “to be attained”; the anīpsita object (A. 1.4.50) is distinguished between 4) an indifferent (udāsīna) object and one that is 5) not wanted (anīpsita), sometimes also called dveṣya “loathed”; then there are 6) the akathita object (A. 1.4.51), and 7) the anya(saṃjñā) pūrvaka, namely, an object that had previously received a different kāraka designation:44

  • 45 Namely, any object that is either anīpsita or akathita.

nirvartyaṃ ca vikāryaṃ ca prāpyaṃ ceti tridhā matam |
tatrepsitatamaṃ karma caturdhānyat tu kalpitam ||
audāsīnyena yat prāpyaṃ yac ca kartur anīpsitam |
saṃjñāntarair anākhyātaṃ yad yac cāpy anyapūrvakam ||
Among these [kārakas], the object that [the agent] most desires to attain is considered to be threefold: to be produced, to be modified and to be attained. Any other [object than the īpsitatama45] is conceived of as being fourfold:
That which is attained unintentionally, that which the agent does not wish to attain, that which is not called with any other [kāraka] designation and also that which had a different [designation] before.

31The complexity of this classification reveals the difficulty, I think, confronting any attempt at finding a common semantic element underlying all varieties of grammatical objects, as noticed above.

  • 46 VP 3.7.47: satī vāvidyamānā vā prakṛtiḥ pariṇāminī | yasya nāśrīyate tasya nirvartyatvaṃ pracakṣat (...)
  • 47 VP 3.7.48ab: prakṛtes tu vivakṣāyāṃ vikāryaṃ.
  • 48 These examples are found in the Prakīrṇaprakāśa of Helārāja (ca. 10th century), a commentary on th (...)
  • 49 VP 3.7.51: kriyākṛtā viśeṣāṇāṃ siddhir yatra na gamyate | darśanād anumānād vā tat prāpyam iti kat (...)

32According to Bhartṛhari, the object “to be produced” (nirvartya) is one “whose primary cause — whether it exists or not [before the production] — is not invoked [in the utterance] as undergoing a transformation”, as in mṛdā ghaṭaṃ karoti “he makes a pot with clay”.46 On the other hand, the object to be modified (vikārya) is so called “when there is the intention to express the primary cause [as undergoing a transformation]”;47 it occurs either with verbs such as dah- “to burn”, lū- “to cut”, pac- “to cook”, etc., the sense of which implies a transformation of a pre-existing material cause, as in kāṣṭhāni dahati “he burns the logs”, kāṇḍāni lunāti “he cuts the stems” etc., or with the verb kṛ- “to make”, in a common type of construction where kṛ- governs two objects, in expressions such as mṛdaṃ ghaṭaṃ karoti. Here mṛd, “clay”, is in the accusative like ghaṭa “pot”, thus conveying the identity, or at least the continuity, between the original material and the product.48 Finally, the prāpya object is characterised by the fact of not being affected in any way by the action, at least as far as can be ascertained through ordinary means of cognition such as perception and inference.49 Typical examples of prāpya objects are ādityaṃ paśyati “he looks at the sun”, nagaram upasarpati “he approaches the city” and vedam adhīte “he studies the Veda”.

33What is the purpose of these labels and their definitions? It is important to stress that, in presenting this classification, Bhartṛhari does not intend to supply any additional rule to the existing body of the A. Rather, he is spelling out the underlying semantics of the Sanskrit grammatical structures described by Pāṇini, who typically formulates very terse meaning conditions for different varieties of karman.

  • 50 kartā is to be read here by anuvṛtti from a previous sūtra, A. 3.1.68, kartari śap.
  • 51 These expressions do not translate easily into English. In certain cases, the verb can be used in (...)

34Thus, objects of the nirvartya and vikārya types are capable of undergoing the operation traditionally known as karmavadbhāva, literally “treatment [of the agent] as the object”. This is regulated by A. 3.1.87, karmavat karmaṇā tulyakriyaḥ, according to which an agent50 whose action is similar to the activity it would carry out if it were an object will be treated as an object, that is, it will require passive verbal morphology. Such an agent is called karmakartṛ. Examples of karmavadbhāva are pacyate odanaḥ (svayam eva), lūyate kedāraḥ,51 etc. Conversely, the prāpya object, in which the object itself is not affected in any way by the action that resides entirely in the agent, is not amenable to this treatment.

35Furthermore, one type of vikārya object occurring with kṛ-, namely the product resulting from the transforming action, can undergo a kind of incorporation into the verb by means of the secondary (taddhita) suffix CvI (=-ī) according to A. 5.4.50, kṛbhvastiyoge sampadyakartari cviḥ. Thus, a sentence such as mṛdaṃ ghaṭam karoti has the derivational alternative mṛdaṃ ghaṭīkaroti.

36Similarly, the designation karman for the akathita object is needed to regulate double-object constructions, while the anyapūrvaka variety includes — among others — the important subtype of the prayojya, the instigated agent of verbs of motion, etc. in a causative construction, as mentioned above. It is worth reiterating that in all these cases assigning the name karman to an item opens the way to the passivisation of the sentence and the possibility for that item to appear in it as the grammatical subject.

4. Cēṉāvaraiyar’s classification of the grammatical object

37It is in the final portion of his commentary on T. 2.2.10 (TC 71) that Cēṉāvaraiyar proposes the classification which is the starting point of this paper.

38He states that the grammatical object (ceyappaṭuporuḷ) can be of three types: to be created (iyaṟṟappaṭuvatu), to be modified (vēṟupaṭukkappaṭuvatu) and to be attained (eytappaṭuvatu). He then goes on to give a short explanation of each type: “to create” is to bring into existence that which did not exist before (iyaṟṟutalāvatu muṉṉillataṉai uṇṭākkutal); “to modify” means to alter that which already existed before (vēṟupaṭuttalāvatu muṉṉuḷḷataṉait tirittal); “to be attained” is the fact of happening to be simply associated (uṟutal) with the result (payaṉ) of the action without there being either creation or modification (eytappaṭutalāvatu iyaṟṟutalum vēṟupaṭuttalum iṉṟit toḻiṟpayaṉ uṟuntuṇaiyāy niṟṟal).

  • 52 TCC 72.1: ceyppaṭu poruḷ mūṉṟaṉaiyum paṟṟivarum vāypāṭukaḷai virikkiṉṟār “he elaborates on the mod (...)
  • 53 TCC 72.6: iyaṟṟappaṭuvatu orutaṉmaittu ākaliṉ ataṟku oru vāypāṭē kūṟiṉār.

39Then, under T. 2.2.11 (TCC 72), Cēṉāvaraiyar introduces his gloss of the long list of verbs found there saying that it illustrates this threefold classification.52 According to him, there is just one verb, iḻaittal “to make, to build” — the fourth item in the list — representing the first type, namely the object to be created (iyaṟṟappaṭuvatu). For this he gives the example eyilaiy iḻaittān “he built a fortress” (TCC 72.5). Further down, after dealing with the other two subdivisions, he comments that creation has only one nature, which explains why TC mentions just one verb associated with this type.53

40The verbs ōpputal “to chase”, iḻattal “to lose”, aṟuttal “to sever”, kuṟaittal “to cut”, tokuttal “to gather”, pirittal “to split”, ākkutal “to effect, to practise” and citaittal “to destroy” are said to have objects of the second type, “to be modified” (vēṟupaṭukkappaṭuvatu). A larger group of verbs illustrate the third type, the object to be attained (eytappaṭuvatu): kāttal “to guard”, oppātal “to resemble”, ūrtal “to ride/drive”, pukaḻtal “to praise”, paḻittal “to blame”, peṟutal “to obtain”, cellutal “to go”, nōkkutal “to look”, etc. For all of these Cēṉāvaraiyar provides appropriate examples, and on the basis of the phrase aṉṉa piṟavum in the sūtra, he adds others, notably including the two kuṟippuviṉais uṭaiya, “to have”, and ila, “to have not”.

  • 54 Significantly, the Pāṇinīyas themselves express some reservation about its usefulness: for example (...)
  • 55 Note, in particular, that there is no provision for causative forms in the T. (cf. Subrahmanya Sas (...)

41It is important to note that these three categories, which are regarded by Bhartṛhari and later Pāṇinīyas as varieties of the īpsitatama object alone, are apparently applied by Cēṉāvaraiyar to the entire set of grammatical objects. This is understandable: the Pāṇinian distinction between īpsitatama and anīpsita is silently dropped,54 while the other two varieties — akathita and anyapūrvaka — are not relevant in Tamil, for the corresponding morpho-syntactical structures do not exist,55 nor is the designation ceyappaṭuporuḷ meant to facilitate passivisation. Therefore, the threefold distinction of the grammatical object ends up having a purely semantic value, for one type cannot be distinguished from another on the basis of morpho-syntactic features.

42As far as I can tell, Cēṉāvaraiyar is the first Tamil grammarian to mention this classification, and this seems to be confirmed by the fact that he does not refer to any earlier authority. The derivation of Cēṉāvaraiyar’s classification from Bhartṛhari’s is undeniable, I think. Nevertheless, there is a remarkable difference in approach and scope, which shows that Cēṉāvaraiyar appropriates the concept and adjusts it to the needs of Tamil grammar in general and, specifically, to the context of the work he comments upon, the TC.

43Before moving on to a consideration of the possible direct and indirect historical links between Bhartṛhari and Cēṉāvaraiyar, in the next section I will make a partial digression in order to examine how, dealing with a closely related topic, Cēṉāvaraiyar has recourse to a Sanskrit grammatical concept to explain a sūtra in the TC.

5. The case of the karmakartṛ, the “agent treated as an object”

44As mentioned above, the threefold classification of the īpsitatama object is morpho-syntactically relevant in Sanskrit also in the case of karmavadbhāva, the treatment of the agent as an object, which can apply to nirvartya and vikārya, but not prāpya objects.

45A construction semantically close to karmavadbhāva is described for Tamil in T. 2.6.49 (TC 246) in the Viṉaiyiyal, in the context of other sūtras on idiomatic usages:

ceyappaṭuporuḷaic ceytatu pōlat
toḻiṟpaṭak kiḷattalum vaḻakkiyaṉ marapē.

  • 56 “C’est une habitude qui est en force dans l’usage, que d’énoncer, mis en action, l’objet qui subit (...)

46“There is an idiom, proper to current usage, by which one refers to the object, which is involved in the action, as if [it were] the one who has done [it]”.56

  • 57 TCC 246.2: vaḻakkiyaṉ marap’eṉavē ilakkaṇam aṉṟ’eṉṟav āṟu ām “En disant ‘habitude qui est en force (...)

47According to Cēṉāvaraiyar’s interpretation, this sūtra concerns an irregular57 usage such as tiṇṇai meḻukiṟṟu, lit. “the verandah cleansed (itself) [with cow-dung]”, with a past active form of the verb meḻukutal, while the expected expression would rather be tiṇṇai meḻukap paṭṭatu “the verandah was cleansed [with cow-dung]”, with a periphrastic passive consisting of a past active form of paṭutal “to appear” plus the absolutive of meḻukutal. In either case, it should be noted, tiṇṇai bears no case marker.

48The expression toḻiṟpaṭa means — says Cēṉāvaraiyar — that here the object is referred to as if the action of the agent, which he calls viṉaimutal, pertained to it. And in order to clarify his view, he adds that this is similar to the expression arici tānēy aṭṭatu “rice has cooked (aṭṭatu) by itself (tānē)”, which is used “seeing that it has been cooked easily” (eḷitiṉ aṭap paṭutaṉōkki), and that this is called karumakaruttaṉ, a Tamilised form of the Sanskrit karmakartṛ.

  • 58 Cf. MBh II. 67.20: atrāpi yāsau sukaratā nāma tasyā nānyaḥ kartā. The example odanaḥ pacyate svaya (...)
  • 59 Compare, for example, the English verb “to break” in the following sentences: “he broke the glass” (...)
  • 60 Note that the Tamil sentence is not an exact calque of the Sanskrit, for the Tamil verb is active (...)

49These remarks are clearly reminiscent of several passages in Sanskrit works discussing the notion of karmakartṛ, a classical example of which is precisely pacyate odanaḥ svayam eva “rice is cooking by itself”, which is meant to convey the easiness (sukaratā) of the action of cooking.58 Apparently, Cēṉāvaraiyar resorts to the concept of karmakartṛ developed by Sanskrit grammarians in order to stress the semantic affinity of certain expressions in Tamil and Sanskrit respectively, even though the grammatical structures themselves are quite different: while Sanskrit karmavadbhāva requires passive verbal morphology, in fact, Tamil — like English — resorts to the active forms of certain verbal roots, used intransitively.59 Once again, this concept and the texts dealing with it were presumably quite well known to his readers. It is apparently sufficient for Cēṉāvaraiyar to give a standard example of karmakartṛ in Sanskrit grammar and suggest that, in the Tamil example, the tiṇṇai is karumakaruttaṉ, in order to explain what had perhaps become an obsolete — and therefore, obscure — usage in his time. Interestingly, he even goes to the extent of giving a functionally meta-linguistic rendering of the Sanskrit sentence odanaḥ pacyate svayam eva with the apparently artificial Tamil arici tānēy aṭṭatu60 (much in the same way as somebody teaching Sanskrit through English may provisionally render an expression such as tena supyate with the blatantly un-English “by him it is slept” in order to help the students grasp both the grammar and the sense of the original).

6. From Bhartṛhari to Cēṉāvaraiyar

50The long gap — at least seven or eight centuries — between Bhartṛhari and Cēṉāvaraiyar prompts us to look for possible intermediaries, both in the Sanskrit and the Tamil grammatical traditions.

  • 61 PV 12, auto-commentary (p. 65). The mention of Kaiyaṭa (possibly 11th century), the author of the (...)

51It is possible, of course, that, prior to Cēṉāvaraiyar, other Tamil authors, unknown to us, had adopted or at least discussed the threefold classification of the grammatical object. In this respect, some interesting clues are found in a late grammatical work, the Pirayōka Vivēkam (PV) of Cuppiramaṇiya Tīṭcitar (fl. 17th century) — exhibiting the traces of a massive Sanskritic influence already in the title — which explicitly recognises the link between Cēṉāvaraiyar’s tripartite scheme and the one found in the VP and in Kaiyaṭa:61

iṉi vākkiyapatīya ttuḷḷum kaiyaṭa ttuḷḷum nirvarttiyam vikāriyam pirāppiyam eṉak kūṟiya vaṇṇam, cēṉāvaraiyarmutalāyiṉār iyaṟṟappaṭuvatum vēṟupaṭukkappaṭuvatum eytappaṭuvatum eṉac ceyappaṭuporuḷai mūṉṟākkuvar.
Now, [grammarians] such as Cēṉāvaraiyar etc. divide the grammatical object (ceyappaṭuporuḷ) into three types, namely “to be produced” (iyaṟṟappaṭuvatu), “to be modified” (vēṟupaṭukkappaṭuvatu) and “to be attained” (eytappaṭuvatu), in the way it is said to be nirvartya, vikārya and prāpya in the Vākyapadīya and in Kaiyaṭa.

  • 62 The work is ascribed to Devanandī (also known as Pūjyapada), who was probably active in the 5th ce (...)

52Then, after quoting Cēṉāvaraiyar’s definitions and examples for each type, the PV goes on to hint at a different classification of the object found in the Naṉṉūl (N., early 13th c.), a grammatical treatise independent from the T. tradition, which it relates to the views of Cainēntiraṉ, that is Jainendra, presumably the Jainendravyākaraṇa,62 a non-Pāṇinian Sanskrit grammar:

naṉṉūlārum ākkal aḻittal aṭaital mutalākac cainēntiraṉ matam paṟṟip palavākkuvar “And the author of the Naṉṉūl spends many words on the view of Jainendra, namely that [the object] is [divided into varieties such as] ākkal aḻittal aṭaital etc.”.

53On the basis of these indications, we may now look back at the various strands in both the Sanskrit and the Tamil grammatical traditions and try to retrace the historical evolution that leads from Bhartṛhari to Cēṉāvaraiyar.

  • 63 This rule prescribes the primary (kṛt) suffix aṆ (-a) after a verbal base co-occuring with a noun (...)

54Within the Sanskrit tradition, later grammarians regularly refer to the threefold classification of the object when they discuss karman. Among the few works surviving from the second half of the first millennium CE to the early second millennium, the first post-Bhartṛhari Pāṇinian work known to us, the Kāśikā Vṛtti (KV, 7th century CE) makes no reference to the threefold distinction under A. 1.4.49, kartur īpsitatamaṃ karma, but commenting on A. 3.2.1, karmaṇy aṇ,63 it begins with the words trividhaṃ karma nirvartyaṃ vikāryaṃ prāpyaṃ ceti and gives appropriate examples for each type, even though it does not provide any definition. However, its subcommentaries, the Nyāsa (8th century) of Jinendrabuddhi and the somewhat later Padamañjarī of Haradatta, both treat the topic at some length with regard to the īpsitatama object of A. 1.4.49, bringing out the syntactical implications of the three categories.

  • 64 See Belvalkar (1915/1997: 32–33); Tripāṭhī (1981: xvii–xxi).

55The case of the latter is particularly interesting because Haradatta is commonly believed to have hailed from South India64 and therefore may have been active in śāstric circles that were conversant with both traditions. Under A. 1.4.49, he quotes the verses of the VP detailing the general classification of all types of karman into seven categories (VP 3.7.45–46), beginning with nirvartya, vikārya and prāpya, as well as some of those providing definitions of the latter three (VP 3.7.49–51).

  • 65 On the dates of these texts, and of the JV, mentioned in the following paragraphs, see Oberlies (1 (...)
  • 66 Literally, īpsita, the passive past participle of the desiderative stem īpsa-, from the root āp- (...)
  • 67 CVV and CV 2.1.43 (1, 138): odanaḥ pacyata ity odanaśabdād vyāpyatā na gamyate, kiṃ tarhi tiṅantāt
  • 68 Pace Deshpande (1979), who considers the CV and its vṛtti to be the work of the same author and th (...)

56However, references to either the threefold or the sevenfold classification of the object are also found in some non-Pāṇinian Sanskrit works, which may in their turn have affected the Tamil grammatical tradition, as indicated by the PV. The first faint echo is possibly found as early as the 6th century CE in the Cāndravṛtti, Dharmadāsa’s commentary on the Buddhist Candragomin’s Cāndravyākaraṇa (C., 5th century CE).65 Even though it is a derivational grammar along the Pāṇinian model, the latter does not resort to kāraka categories but — like the T. — it relates case morphemes directly to very broad semantic definitions. Thus, C. 2.1.43 kriyāpye dvitīyā prescribes the second case ending (i.e. the accusative) to denote “that which is to be attained through the action”, possibly in an effort to simplify Pāṇini’s minute classification and at the same time capture the semantic kernel of the grammatical object by simply emphasising its close relation with the action and, implicitly, its key role in the structure and the sense of the utterance. Commenting on it, the vṛtti (CVV) does not explicitly mention the threefold distinction of the grammatical object but gives three examples that virtually illustrate the three categories: kaṭaṃ karoti (nirvartya), odanaṃ pacati (vikārya) and ādityam paśyati (prāpya). Moreover, under the headings īpsita “desired”,66 anīpsita “undesired” and na vepsitaṃ nāpy anīpsitam “neither desired nor undesired”, it gives examples of “secondary” objects in double-accusative constructions, of unwanted objects such as snakes and poison, and of indifferent objects, respectively. All of them, it says, possess vyāpyatā, “object-hood”-apparently an attempt at accounting for the Pāṇinian categories in the light of Bhartṛhari’s observations. Interestingly, the CVV also remarks that in a passive sentence such as odanaḥ pacyate “rice is being cooked”, the object-hood of rice is only understood from the verbal ending, not from the word for “rice” itself, for the latter is in the nominative.67 Its author apparently feels that the absence of the abhihita/anabhihita device in the CV makes this statement necessary.68

57Similarly, the Jainendrayākaraṇa (JV, 5th century CE), which does resort to kārakas, defines karman in JV 1.2.120, kartrāpyam, merely as “that which is to be attained by the agent”, renouncing any attempt at a semantic characterisation but putting the agent back at centre stage. About two centuries later, however, its main commentary, the Mahāvṛtti (JVV) of Abhayanandin (7th century), quotes a verse of unknown origin which partly echoes Bhartṛhari’s threefold classification (no longer confined, though, to the īpsitatama object, for the Pāṇinian classification is discarded in the JV), while seemingly trying to provide more accurate categories pertaining to the agent’s will to attain the object:

prāpyaṃ viṣayabhūtaṃ ca nirvartyaṃ vikriyātmakam |
kartuś ca kriyayā vyāpyam īpsitānīpsitetarat ||

58In this way one arrives at a sevenfold classification of karman, for which the following examples (among others) are given:

  1. prāpya = grāmaṃ gacchati
  2. viṣayabhūta = jainendram adhīte
  3. nirvartya = ghaṭaṃ karoti, odanaṃ pacati
  4. vikriyātmaka (i.e. vikārya) = kāṣṭhāni dahati, ghaṭaṃ bhinatti
  5. īpsita = guḍaṃ bhakṣayati, odanaṃ bhuṅkte
  6. anīpsita = grāmaṃ gacchan vyāghraṃ paśyati, kaṇṭakān mṛdnāti
  7. anubhaya (corresponding to itarat in the verse) = grāmaṃ gacchan vṛkṣamūlāny upasarpati

59As the examples show, nirvartya and vikārya are the same as for Bhartṛhari, while prāpya is divided into two categories, one bearing the same name, the other called viṣayabhūta, lit. “that which has become a (cognitive) object or the scope” of the action. Judging from the examples, in the former case the object is physically attained, as in grāmaṃ gacchati “he goes to the village”, while in the latter it is not: jainendram adhīte “he studies/recites the work of Jinendra”. While these characterise the relation between the object and the action, the next three categories pertain instead to the agent’s attitude towards the object, which may therefore be either īpsita, as in guḍaṃ bhakṣayati “he eats molasses”, or anīpsita, as in grāmaṃ gacchan vyāghraṃ paśyati “going to the village, he sees a tiger”, or anubhaya “neither one nor the other”, that is “indifferent”, as in grāmaṃ gacchan vṛkṣamūlāny upasarpati “going to the village, he touches upon the roots of a tree”. It is evident that any object belonging to a category included in the former set is also inevitably classifiable according to a category of the latter such as īpsita, anīpsita or anubhaya, thus making the whole scheme fuzzy, despite its seeming quest for greater subtlety than that achieved by the Pāṇinian scheme. Nor is there any attempt at relating these semantic categories to morpho-syntactic ones.

60If we now turn to the Tamil tradition, some grammatical treatises outside the T. tradition, the Vīracōḻiyam (VC, possibly 11th century) and the Naṉṉūl (N., beginning of 13th century), mentioned in the PV, do contain classifications of the object, but different from those later found in the TCC. Analysing these in detail is beyond the scope of this paper. However, even a cursory look at their categories, which I list below offering tentative translations of their names, casts some light, I think, on the historical circumstances that may have led Cēṉāvaraiyar to adopt Bhartṛhari’s threefold scheme.

  • 69 VC 41ab: paṟṟoṭu viṭē irupuṟam tāṉ teri tāṉ teriyā naṟ karuttā tīpakam ām karumam. I wish to thank (...)

61The VC, composed by the Buddhist Puttamittiraṉ, enumerates seven types of objects,69 significantly called with the Sanskrit loan-word karumam, and gives an example for each type:

1) paṟṟuk karumam = object to be attained
poṉṉai ācaippaṭṭār vaṟiyōr
“Those who desired2 gold1 are fools3

2) vīṭṭuk karumam = object to be abandoned
oḻukkattai ikaḻvar tīyōr
“Those who scorn2 proper conduct1 are evil3

3) irupuṟak karumam = object that is neither (with regard to the previous ones)
māccōru uṇkiṉṟa ciṟukkaṉ ataṉkaṇ vīḻnta tūḷiyinait tiṉṟāṉ
“The boy3 eating2 a lot of rice1 gobbled up7 the dust6 that had fallen5 into it4

  • 70 For the parallel between categories no. 4 and 5 in this list and the Pāṇinian notions of abhihita/ (...)

4) tāṉ teri karumam = self-manifesting object (i.e. marked by the case-ending and, therefore, akin to the abhihita “expressed” object in the Pāṇinian sense70)
vīṭṭai eṭuttāṉ taccaṉ
“The carpenter3 built2 the house1”. (vīṭu “house” + case marker ai)

5) tāṉ teriyāk karumam = non self-manifesting object (i.e. devoid of case ending and, therefore, anabhihita “unexpressed” in the Pāṇinian sense)
vīṭu taccaṉ kaṭṭiṉāṉ
“The carpenter2 built3 the house1”. (vīṭu unmarked)

6) karuttā karumam = object that is an agent (corresponding to the prayojya “instigated” agent of certain Sanskrit causative constructions)
koṟṟaṉai ūrkkup pōkkiṉāṉ cāttaṉ
“Cāttaṉ4 made3 the mason1 go3 to the village2

  • 71 For this interpretation, cf. below. I know of no Sanskrit work where the phrase dīpaka karman is u (...)

7) tīpakak karumam = seemingly, a borrowing from Sanskrit (i.e. dīpaka karman, lit. “object with light(s)”?), although the sense and origin of the phrase are unclear; it refers to one of two objects (or possibly both) in a double-accusative construction71
pacuviṉaip pālaik kaṟantāṉ
“He milked3 milk2 (from) the cow1” (pacu “cow”+ augment iṉ + ai; pāl “milk”+ ai).

62Once again, we are confronted here with a seven-item list. The first three varieties are clearly reminiscent of those of īpsita, anīpsita and na vepsitaṃ nāpy anīpsitam/anubhaya in the CVV and the JVV. The following two seem inspired by Pāṇinian notions, but they are cleverly redefined to serve the description of a typically Tamil feature, namely the possibility of an unmarked object. The sixth category accounts for the induced agent turned into an object of a causative verb, while the last refer to a kind of morpho-syntactic construct whose relevance in Tamil grammar is rather dubious.

  • 72 N. 295 (according to the numbering found in the editions with Mayilainātar’s commentary), p. 145: (...)
  • 73 However, the list is concluded by āti (corresponding to Skt. ādi), thus allowing other values to b (...)

63On the other hand, the N.72 of the Jaina Pavaṇanti differentiates grammatical objects into just six types73 on the basis of the following semantic values, each of them identified through a “prototypical” verb governing it:

  1. ākkal, “to create”
  2. aḻittal, lit. “to destroy”, but possibly employed here in the sense of “to change, to modify”
  3. aṭaital, “to reach”
  4. nīttal, “to renounce”
  5. ottal, “to compare”
  6. uṭaimai, “possession”
  • 74 I am not sure I understand the reasons that may have led the author of the N. to make this a separ (...)

64In a way, the N. adopts the same enumerative strategy found in the TC, but mitigated by an effort of systematisation, which it is tempting to attribute to a reaction to the conceptual confusion of the classification advanced in the VC. The first three categories seem to correspond quite neatly to Bhartṛhari’s notions of nirvartya, vikārya (if aḻittal is indeed to be taken as meaning “to change”) and prāpya. They are accompanied by the object “to be renounced” (possibly an echo of anīpsita?), the object “to be compared”74 and, finally, the object “to be possessed”, a category that takes into account the existence in Classical Tamil of verbs having the sense of “to have” and “to have not” (unlike Sanskrit, where the same notion is expressed by means of the genitive of the possessor and a form of the verb “to be” agreeing with the possessed item).

65From this classification, with its clear focus on semantics, one gets the impression that Pavaṇanti was well acquainted with the ongoing debate on the classification of the object in the Sanskrit and Tamil grammatical traditions and made an effort to go beyond the inherited ideas in order to account for the semantic varieties of the object in Tamil.

7. Pāli grammars

  • 75 The inclusion of Aggavaṃsa’s work here is justified by the fact that Burmese Buddhism had close an (...)
  • 76 It is worth recalling here that the first reordered commentary on Pāṇini’s A. was the Rūpāvatāra o (...)

66Before I conclude, it is worth having a look — once again, inevitably superficial — at the way the classification of the grammatical object is treated in two 12th-century Pāli grammars, the Moggalānavyākaraṇa (MV) of Moggallāna, probably composed in Sri Lanka, and the Saddanīti (SN) of the Burmese Buddhist Aggavaṃsa.75 This was the third tradition that was thriving in the same centuries in the south of the subcontinent and presumably interacting with the other two.76

  • 77 MVP 2.2: nibbattivikatippattibhedena tividham kammam […] yassāsato jananaṃ karīyati sā nibbatti, y (...)
  • 78 See Kahrs (1992: 46 ff.).

67Both the MV and the SN know the threefold classification originally proposed by Bhartṛhari for the īpsitatama object. The MV does not mention it explicitly, but under 2.2, kamme dutiyā, it gives the three classical examples for each type: kaṭaṃ karoti, odanaṃ pacati, ādiccam passati. However, its auto-commentary, the Moggalānapañcikā (MVP), devotes some space to the threefold classification and for each type it provides brief definitions, which are clearly reminiscent of those we have seen above,77 before specifying that the first example in the MV is an instance of nibbattiya (i. e. nirvartya) object, the second of vikariya (i.e. vikārya) and the third of pattiya (i.e. prāpya). Similarly, the SN78 reports a threefold classification of the object (kamma) into nibattanīya, vikaraṇīya and pāpanīya. However, more interestingly, this text also knows a scheme comprising seven categories, namely icchita, anicchita, nevicchitānicchita, akathita, kattukamma, abhihita and anabhihita, which seem to reflect those found — in a different order — in the Buddhist Tamil grammar, the VC, and in their turn going back to Pāṇinian categories (though mediated, to some extent, through non-Pāṇinian systems such as CV and JV). The following table summarises this intricate network of exchanges:

8. Conclusions

68How to interpret the facts discussed so far? Admittedly, the textual material I have examined here is too limited to allow the extrapolation of any firm conclusion about the broader historical picture of the relations between Tamil and Sanskrit. Nevertheless, I think it may give us some indications about the larger processes of bilingual — or possibly, as I have shown, multilingual — “cross-cultural fertilisation” operating in the Tamil country between approximately 500 CE and the early second millennium.

  • 79 The T. itself is likely to be the culmination of a long period of earlier speculation (see Chevill (...)
  • 80 That is, the process by which a language is provided with a grammar, namely a formal description. (...)

69If the current dating of the T. in its present form to ca. 500 CE is correct,79 one can surmise that earlier indigenous speculations about language were catalysed by the exposure to linguistic ideas and texts coming from the North. According to Sylvain Auroux (1994), the process of grammatisation80 of a language can be either a spontaneous, entirely autochthonous process or one prompted — and to a variable degree influenced — by the earlier grammatisation of another language, used within either the same or a different community and invested with special significance because it is associated with a hegemonic or otherwise prestigious literary culture. While Auroux has focussed in particular on the case of Latin grammar, which from the Middle Ages served as the matrix of those of several other languages (initially, European ones), something very similar seems to have happened in South Asia in the wake of the precocious grammatisation of Sanskrit.

  • 81 For the concept of Sanskrit cosmopolis, see Pollock (1996; and 2006: in particular 39–280).
  • 82 Among them — in different ages — Kannada, Telugu, Malayalam (for which see Rich Freeman’s contribu (...)

70In the first millennium CE, as Sanskrit affirmed itself as the cosmopolitan language of an ever-increasing area,81 and Indic cultural traits spread from North India to the south of the subcontinent and then to South-East and Central Asia, the model embodied by the Sanskrit grammatical speculation appears to have fertilised virtually all the regional cultures that came into contact with it, favouring the development of grammatical traditions that engaged in the description of several languages belonging to different linguistic families.82

  • 83 On the origins of Tamil grammar, see Chevillard (2000 and 2009b).
  • 84 On the crucial role that writing has played, historically, in the development of linguistic knowle (...)
  • 85 See Mahadevan (2003: 90 ff.).
  • 86 See Chevillard (2000b: 174; and 2008: 29–36).

71Tamil seems to have been the first among these.83 Crucially, according to the patterns of grammatisation described by Auroux,84 it had already been committed to writing at the beginning of this period. The earliest inscriptions in Tamil Brahmī, which is an adaptation of the Northern script and testifies to sustained contacts between the two regions of the subcontinent, are in fact dated before the beginning of the Common Era.85 And various features of the Eḻuttatikāram, the first part of the T. devoted to the “letters” (eḻuttu) of the language, show a preoccupation with writing.86

72It may prove impossible to reconstruct the early stages of development of Tamil grammar — and its relationship with the Sanskrit grammatical tradition — with any degree of accuracy. However, the situation improves for later times, as there is greater abundance of surviving textual evidence, but several questions are still to be answered.

  • 87 Ormultilingual, if we also consider Pāli and possibly other languages (such as literary Prakrits, (...)
  • 88 Chevillard (2009a: 72, n. 2) mentions Campantar (ca. 8th century), one of the authors of the Tēvār (...)

73To confine ourselves to the domain of linguistic ideas, it is evident that the dialogue between the Sanskrit and the Tamil traditions continued for centuries, well beyond the initial fertilisation. However, relatively little is known about the individuals and institutions involved in this process. For example, it is not clear whether — and to what extent — both traditions were actively cultivated by the same scholarly circles or by different ones, which would then beg the question of what were their different constituencies. Should we imagine a milieu of fully bilingual87 intellectuals, who were conversant with both languages, or were there different degrees of specialisation, more like separate streams running in parallel and occasionally brushing against one another?88 And were these mainly Brahmanical circles, in terms as much of social rank as of religious affiliation, or were Buddhists and Jains also involved, as the textual material presented here suggests?

  • 89 On this important phenomenon, see the various contributions in Wilden (2009a).
  • 90 This was preceded or accompanied and, in a sense, prepared by the anthologisation of the Caṅkam co (...)

74If we go back to the passage that has been the starting point of my investigation, Cēṉāvaraiyar’s adoption of Bhartṛhari’s classification of the grammatical object may be seen as an instance of what Chevillard (2008: 17) calls “a second wave of Sanskritisation” in the Tamil grammatical tradition, which seems to coincide with the so-called “age of commentaries”, as some researchers have proposed to call it, namely the emergence of a massive production of exegetical treatises on a variety of subjects,89 from approximately the 8th century onwards.90

75The complexity of this scenario should not be underestimated. As Chevillard (2008: 13) points out, in the second half of the first millennium CE “[t]he world of [Tamil] grammarians had not yet been standardized, although the efforts for defining a classical norm were clearly underway, in a manner which remains to be studied precisely in its dynamics”. Some time later, clear traces of a massive Sanskritic influence can be detected in a work such as the VC, which appears to have drawn concepts and terms not only from the Pāṇinian system, but also from the Cāndra and the Jainendra schools, as I have shown above. Moreover, similar concepts and terms were also circulating in the contiguous and coeval Pāli tradition.

  • 91 His explicit mention of uṭaiya and ila among verbs governing an object to be attained may have bee (...)

76With his threefold classification, Cēṉāvaraiyar is likely to have been responding to what is found in the VC and the N. (presumably, in the context of a rivalry among different grammatical schools). His classification, based on the one devised by Bhartṛhari, is more economical than the one proposed in the N. (three categories instead of six), but it is equally exhaustive and, while it takes into consideration the specific features of the Tamil language,91 it respects the theoretical approach of his mutal nūl, the TC. And, perhaps, it also had the additional appeal of being invested with the authority and prestige of the Sanskrit intellectual tradition that had originally produced it.

77The fact that the Sanskrit authors, and in particular the sources of the threefold classification, are not explicitly mentioned, may well be a sign of the familiarity Cēṉāvaraiyar and his fellow grammarians had with that tradition and the ease with which they moved from one cultural universe to another. All this seems to suggest a historical picture — the exact outlines of which are still largely elusive — of very complex intellectual exchanges overcoming the boundaries between languages, countries and religious affiliations, which need to be investigated further.

Bibliographie

BIBLIOGRAPHY

I. Primary Sources

Abbreviations

A.: Aṣṭādhyāyī

CV: Cāndravyākaraṇa

CVV: Cāndravyākaraṇavṛtti

JV: Jainendravyākaraṇa

JVV: Mahāvṛtti ad Jainendravyākaraṇa

N.: Naṉṉūl

PrPr: Prakīrṇaprakāśa

PV: Pirayōka Vivēkam

T.: Tolkāppiyam

TC: Tolkāppiyam, Collatikāram

TCC: Cēṉāvaraiyar’s commentary on the TC

VC: Vīracōḻiyam

VP: Vākyapadīya

Aṣṭādhyāyī of Pāṇini. Roman Transliteration and English Translation by Sumitra M. Katre. University of Texas Press, Austin. (First Indian edition) Delhi: Motilal Banarsidass, 1989.

Cāndravyākaraṇa of Candragomin, 2 volumes. Edited by K.C. Chatterji. Poona: Deccan College, 1953.

Jainendravyākāraṇa, tasya ṭīkā Ācārya-Abhayanandipraṇītā Jainendramahāvr̥ttiḥ Pūjyapādadevanandiviracitam. Edited by Shambhu Nath Tripathi and Mahadeo Chaturvedi. Kāśī: Bhāratīya Jñānapīṭha, 1956 Kāśikāvṛtti of Vāmana-Jayāditya, with the commentaries Nyāsa or Pañcikā of Jinendrabuddhi and the Padamañjarī of Haradatta Miśra, 6 volumes. Edited by Dwārikā Dās Śāstri and Kālika Prasād Shukla. Varanasi: Sudhi Prakashan, 1983–85.

Mahābhāṣyapradīpa of Kaiyaṭa: see Vyākaraṇa Mahābhāṣya of Patañjali, Chaukhamba edition.

Moggallāna pañcikā suttavuttisametā, anurādhapure thūpārāmamahāv ihāramajjhāvutthena mahāsaddikena Sirimatā Moggallānamahāsāminā viracitā. Edited by Śrī Dharmānanda, P.A. Peries Appuhamy Wirahena. Colombo: Saccasamuccaya Press, 1931.

Naṉṉūl mūlamum mayilainātaruraiyum. Edited by U. Vē. Cāminātaiyar. Ceṉṉai (Chennai): U. Vē. Cāminātaiyar Nulnilaiyam, 1995.

Nyāsa of Jinendrabuddhi: see Kāśikāvṛtti.

Padamañjarī of Haradatta Miśra: see Kāśikāvṛtti.

Prakīrṇaprakāśa of Helārāja: see Vākyapadīya of Bhartṛhari with the Commentary of Helārāja, Subramania Iyer edition.

Pirayōka Vivēkam Cuppiramaṇiya Tīkkitar iyaṟṟiya, mūlamum uraiyum. Edited by T.V. Gopal Iyer [Ti. Vē. Kōpālaiyar]. Tanjore: Tañcai Caracuvati Makāl veḷiyīṭu 147, 1973.

Cēṉāvaraiyam. Edited by Āṟumuka Nāvalar. (Title page: Cēṉāvaraiyam, Ceṉṉapaṭṭaṉam, Citampara caivap pirakāka vittiyāt tarumaparipālakar Śrīmati Po. Pavaṉiyammai yavarkaḷāl Ceṉṉapaṭṭaṉam vittiyānupālaṉa yantiracālaiyil acciṟ patippikkap paṭṭatu, iraṇṭām patippu). Chidambaram, 1934.

Vākyapadīya of Bhartṛhari with the Commentary of Helārāja. Kāṇḍa III, Part I. Critically edited by K.A. Subramania Iyer. Poona: Deccan College, (Deccan College Monograph Series no 21), 1963.

Bhartṛharis Vākyapadīya. Die Mūlakārikās nach den Handschriften herausgegeben und mit einem Pāda-Index versehen. Critically edited by Wilhelm Rau. Wiesbaden: Franz Steiner, 1977.

Vīracōḻiyamum Peruntēvaṉār iyaṟṟiya uraiyum viḷakkaṅkaḷuṭaṉ. Edited by T.V. Gopal Iyer [Ti. Vē. Kōpālaiyar]. Śrīraṅkam: Śrīmat Āṇṭavaṉ Ācciramam, 2005.

Vyākaraṇa Mahābhāṣya of Patañjali with Kaiyaṭa‘s Pradīpa and Nāgeśa‘s Uddyota, 6 volumes. Edited by B.B. Joshi et al. Delhi: Chaukhamba Sanskrit Pratishthan, 1987–88.

Vyākaraṇa Mahābhāṣya of Patañjali. (Edited by Franz Kielhorn, 3 volumes. Bombay: Government Central Press, 1880–85). 3rd Edition by K.V. Abhyankar, Poona: Bhandarkar Oriental Research Institute, 1962–72.

II. Secondary Sources

Auroux, Sylvain (1994). La révolution technologique de la grammatisation. Liège: Mardaga.

Belvalkar, Shripad K. (1915). Systems of Sanskrit Grammar. (Reprint) Delhi: Bharatiya Book Corporation, 1997.

Cardona, George (1974). “Pāṇini’s kārakas: agency, animation and identity”. Journal of Indian Philosophy, 2, pp. 231–306.

Cardona, George (1976). Pāṇini. A Survey of Research. Delhi: Motilal Banarsidass.

Chevillard, Jean-Luc (1996). Le commentaire de Cēṉāvaraiyar sur le Collatikāram du Tolkāppiyam: sur la métalangue grammaticale des maîtres commentateurs tamouls médiévaux, Vol. 1. Pondichéry: Institut Français de Pondichéry (Publications du Département d’Indologie no 84.1).

Chévillard, Jean-Luc (2000a). “Les débuts de la tradition linguistique tamoule”. In S. Auroux et al. (Editors), History of the Language Sciences, vol. 1, pp. 191–94. Berlin: de Gruyter.

Chévillard, Jean-Luc (2000b). “Le Tolkāppiyam et le développement de la tradition linguistique tamoule”. In S. Auroux et al. (Editors), History of the Language Sciences, vol. 1, pp. 194–200. Berlin: de Gruyter.

Chevillard, Jean-Luc (2007). “Syntactic duality in Classical Tamil poems”. In Colin P. Masica (Editor), Old and New Perspectives on South Asian Languages: Grammar and Semantics, pp. 177–210. Delhi: Motilal Banarsidass (MLBD Series in Linguistics no 16).

Chevillard, Jean-Luc (2008). Companion Volume to the Cēṉāvaraiyam on Tamil Morphology and Syntax. Le commentaire de Cēṉāvaraiyar sur le Collatikāram du Tolkāppiyam, Vol. 2. Pondichéry: Institut Français de Pondichéry/École française d’Extrême-Orient (Collection Indologie no 84.2).

Chevillard, Jean-Luc (2009a). “The metagrammatical vocabulary inside the lists of 32 tantrayukti-s and its adaptation to Tamil: towards a Sanskrit-Tamil dictionary”. In E. Wilden (Editor), Between Preservation and Recreation: Tamil Traditions of Commentary. Proceedings of a Workshop in Honour of T.V. Gopal Iyer, pp. 71–132. Pondichéry: Institut Français de Pondichéry/École française d’Extrême-Orient (Collection Indologie no 109).

Chevillard, Jean-Luc (2009b). “The Pantheon of Tamil Grammarians: a Short History of the Myth of Agastya’s Twelve Disciples”. In G. Colas and G. Gerschheimer (Editors), Écrire et transmettre en Inde classique, pp. 243–268. Paris: École française d’Extrême-Orient (Études Thématiques no 23).

Dash, Prafulla Chandra (1986). A Comparative Study of the Pāṇinian and Cāndra Systems of Grammar (Kṛdanta portion). New Delhi: Ramanand Vidya Bhavan.

Deshpande, Madhav (1997). “Candragomin’s syntactic rules, some mis-conceptions”. Indian Linguistics, 40, pp. 133–145.

Hinüber, Oskar von (1989). Der Beginn der Schriftund frühe Schriftlichkeit in Indien. Wiesbaden: Steiner Verlag.

Kahrs, Eivind (1992). “Exploring the Saddanīti”. Journal of the Pāli Text Society, 17, pp. 1–212.

Mahadevan, Iravatham (2003). Early Tamil Epigraphy: from the Earliest Times to the Sixth Century A.D. Chennai, India: Cre-A — Cambridge, Mass.: Dept. of Sanskrit and Indian Studies, Harvard University (Harvard Oriental Series no 62).

Oberlies, Thomas (1992). “Verschiedene neu-entdeckte Texte des Cāndravyākaraṇa und ihre Verfasser (Studien zum Cāndravyākaraṇa II)”. Studien zur Indologie und Iranistik, 16/17, pp. 161–84.

Oberlies, Thomas (1996). “Das zeitliche und ideengeschichtliche Verhältnis der Cāndra-Vṛtti zu anderen V(ai)yākaraṇas (Studien zum Cāndravyākaraṇa III)”. Festschrift Paul Thieme, Studien zur Indologie und Iranistik, 20, pp. 265–317.

Pollock, Sheldon (2006). The Language of the Gods in the World of Men: Sanskrit, Culture, and Power in premodern India. Berkeley-Los Angeles-London: University of California Press.

Pollock, Sheldon (1996). “The Sanskrit cosmopolis, 300–1300: transculturation, vernacularization, and the question of ideology”. In Jan E.M. Houben (Editor), Ideology and Status of Sanskrit: Contributions to the History of the Sanskrit Language. Leiden: Brill.

Sharma, Rama Nath (1987). The Aṣṭādhyāyī of Pāṇini, Volume I: Introduction to the Aṣṭādhyāyī as a Grammatical Device. Delhi: Munshiram Manoharlal.

Subrahmanya Sastri, P.S. (1934). History of Grammatical Theories in Tamil. Chennai: The Kuppuswami Sastri Research Institute (Reprint 1997).

Tripāṭhī, Tīrtharāja (1981). Sìrīharadattācārya-praṇītāyāḥ Padamañjaryāḥ paryālocanam. Nava Dillī: Bhāvanā-Prakāśanam.

Vergiani, Vincenzo (2009). “A Quotation from the Mahābhāṣyadīpikā in the Kāśikāvṛtti”. In P. Haag and V. Vergiani (Editors), Studies in the Kāśikāvṛtti. The section on pratyāhāras. Critical edition, translation and other contributions, pp. 161–189. Florence-Delhi: Società Editrice Fiorentina — Manohar. (Reprint: London: Anthem Press, 2011).

Wilden, Eva (2002). “Towards an Internal Chronology of Old Tamil Caṅkam Literature or How to Trace the Laws of a Poetic Universe”. Wiener Zeitschriftfür die Kunde Südostasiens, 46, pp. 105–133.

Wilden, Eva (Editor) (2009a). Between Preservation and Recreation: Tamil Traditions of Commentary. Proceedings of a Workshop in Honour of T.V. Gopal Iyer. Pondichéry: Institut Français de Pondichéry/École française d’Extrême-Orient (Collection Indologie no 109).

Wilden, Eva (2009b). “Canonisation of Classical Tamil Texts in the Mirror of the Poetological Commentaries”. In E. Wilden (Editor), Between Preservation and Recreation: Tamil Traditions of Commentary. Proceedings of a Workshop in Honour of T.V. Gopal Iyer, pp. 145–165. Pondichéry: Institut Français de Pondichéry/École française d’Extrême-Orient (Collection Indologie no 109).

Zvelebil, Kamil V. (1995). Lexicon of Tamil Literature. Leiden: Brill.

Notes

1 I would like to express my gratitude to Jean-Luc Chevillard for the many stimulating conversations that have helped me to shape some of the ideas I present here. I would also like to thank my student Alastair Gornall for drawing my attention to the Moggallānavyākaraṇa and its commentary.

2 Pāṇini’s date, like most dates in early South Asian history, is the object of an often heated debate, which I cannot even begin to summarise here. The time span I am giving includes the dates favoured by the majority of modern scholars who have directly addressed this issue: see Cardona (1976: 261–62), von Hinüber (1989: 34).

3 For example, a simple sentence such as yogī parvatād āgacchati “the ascetic comes from the mountain” is derived from a posited initial string < yogin-s parvata-as. ā-gam-ti > through an ordered sequence of operations of affixation and substitution.

4 There is a vast secondary literature on kārakas. Among the most significant contributions see Cardona (1974) and R.N. Sharma (1987: 141–164); for further references see Cardona (1976 and 1999), covering the scholarship until the end of the last century.

5 I will deal in greater detail with the definitions of the object below.

6 As is well known, there are seven declensional endings in Sanskrit (the vocative being treated by Pāṇini as a subtype of the nominative), but only five of them are specifically linked to one of the kārakas. The 1st vibhakti (= nominative) never directly signifies a relation with the action, while the 6th vibhakti (= genitive) is primarily expressive of non-kāraka relations.

7 The actual morphemes are represented by small letters, while the capital letters are metalinguistic markers. Here, I cannot go into a detailed explanation of the technicalities of Pāṇinian grammar.

8 These are of course the western terms. The Sanskrit tradition simply calls them with their ordinal adjective: prathamā “first” for the nominative, dvitīyā “second” for the accusative, and so on.

9 Thus, e.g., A. 7.1.9, ato bhisa ais, prescribes the substitution of the morpheme ais for the instrumental plural ending bhis after nominal stems ending in short a.

10 The rule is required because with other verbs the prayojya maintains its original kāraka designation of agent (kartṛ).

11 Namely, by means of either a verbal ending (as I explain below), or a primary or secondary suffix, or nominal composition.

12 According to A. 3.4.69, laḥ karmaṇi ca bhāve cākarmakebhyaḥ, in which kartari is to be read by anuvṛtti from a previous rule.

13 According to A. 2.3.46, prātipadikārthaliṅgaparimāṇavacanamātre prathamā, which prescribes the nominative to denote only the sense of the nominal base, gender, number and measure, but no syntactical function such as case.

14 According to A. 2.3.2, karmaṇi dvitīyā, already quoted above.

15 According to A. 2.3.18, kartṛkaraṇayos tṛtīyā, whereby the instrumental is the “elective” case for the agent and the instrument.

16 It is worth recalling that the grammatical category of subject does not exist in Pāṇini’s grammar, which operates exclusively with agents and objects. One reason for this may be that, when an intransitive verb is passivised, which is not uncommon in Sanskrit, its agent will be in the instrumental, and the sentence may therefore be formally regarded as “subject-less”.

17 While the term vēṟṟumai (lit. “difference”) is generally used for both notions, another term, urupu, borrowed from Sanskrit (rūpa, lit. “form”) is sometimes used specifically in the sense of “case ending”, and the two are also found together in the phrase vēṟṟumaiy urupu (see Chevillard 2008: 260 s.v.). Intriguingly, rūpa is used by Pāṇini (cf. A. 1.1.68) and his followers in the general sense of sonic form of a word (the “signifier”) as opposed to artha, its meaning, but not, as far as I know, in the sense of “case ending”.

18 The translation of all the TC sūtras in this paper is my English rendering of Chevillard’s French translation (Chevillard 1996); the reference TC is followed by a figure following the numeration found there.

19 Clearly, there is nothing “natural” about this order (for instance, cf. Latin grammar, where the accusative is fourth), although it is hardly surprising that the nominative should appear in the first position.

20 Elsewhere (e.g. T. 2.2.4) it is called eḻuvāy vēṟṟumai, lit. “case of the origin/source” (evidently, the source of the action). The same holds for the vocative, viḷi, lit. “call”.

21 T. 2.2.4 (TC 65), avaṟṟ’uḷ eḻuvay vēṟṟumai peyar tōṉṟu’nilaiyē.

22 T. 2.2.14 (TC 75), nāṉk’ākuvatuvē ku eṉap peyariya vēṟṟumaik kiḷaviy epporuḷ āyinuṅ koḷḷum atuvē.

23 That is, according to Cēṉāvaraiyar, even when the ending ai is governed by a verbal noun, as in pukaḻai niṟuttāṉ “one who has established [his] glory” or pukaḻaiy uṭaimai “the possession of glory” (examples given in TCC 71).

24 kuṟippuviṉai, which Chevillard translates as “verbes idéels” (2008: 128, s.v., and Appendix C, 482–484): for a definition of the latter, cf. TC 220. These are defective verbs that do not carry tense markers.

25 Cf. the remark of Subrahmanya Sastri about TC 71 (1997: 220, n. 2 continued from 219): “… [Tolkāppiyaṉār] directly mentions the meaning of all the cases other than the second in the sūtras dealing with all the other cases and the object is leftout…”.

26 “Du fait du garder, du ressembler, du chevaucher, du construire, du fait du chasser, du louer ou du blâmer, du fait de l’obtenir ou du perdre, de l’amour ou de la colère, du fait du haïr, du se réjouir, ou de l’apprendre, du fait du couper, du diminuer, du rassembler ou du séparer, du fait du peser, du mesurer ou de l’énumérer, du fait de constituer, de l’être attenant à, de l’aller, de l’être passionné, du fait de l’observer, du craindre ou du ruiner, toutes les expressions telles et d’autres semblables, qui sont [nuances de] valeur de cette cause première, sont ses types [d’emploi], dit-on” (Tr. Chevillard 1996: 157).

27 nimittabhāvo bhāvānām upakārārtham āśritaḥ “One resorts to the nature of (auxiliary) causes of things so that they can assist [in the accomplishment of actions]”.

28 yat tu kriyāpadopāttāyāḥ kriyāyā nimittaṃ tat kārakam eva “But that which is a causal factor in an action denoted by a verbal form is nothing but kāraka” (VP 3.i.251.2–3).

29 Chevillard (2008: 219, n. 5) thinks that the term mutaṉilai “looks like the Tamil calque of a Sanskrit term beginning with pūrva and ending with a suffix”. I cannot think of any such term that is commonly found in this context in Sanskrit grammatical works.

30 Similar ideas are also known to the Sanskrit tradition, but I cannot discuss them here.

31 The wording of the sūtra suggests that these two may not have belonged to the original scheme (but cf. T. 2.6.37, quoted below, for an alternative explanation of their being “apart”). Unfortunately, the sense of the phrase aṉṉa marapiṉ “of similar tradition”, which may cast light on this, is far from being clear.

32 Resorting to Chevillard’s insightful concept of “FIRST-translation” (see Chevillard 2009a: 73, n. 4: “A ‘FIRST-translation’ is not a translation. It is an attempt at translating [emphasis in the text]. It can fail or succeed, partially or fully. It has to create its target [neologism]. Ulterior translations follow an already open inter-linguistic path [un chemin déjà frayé], based on consensus”) and his treatment of its strategies, this would be a case of what he calls Strategy D, the use of a Tamil paraphrase of the original Sanskrit term (cf. Chevillard 2009a: 76).

33 For similar considerations on the relation between veṟṟumais and kārakas, cf. Chevillard (2000b: 197).

34 If expressed grammatically, that is by means of case endings, different temporal complements are either silently assimilated to the locus — rātrau “at night”, etc. — or treated by Pāṇini in the vibhakti section as if they were indifferently adverbial or adnominal: see A. 2.3.5–7 and KV thereon.

35 This is hardly surprising considering that classical Tamil does not have an equivalent of the ablative of place from which, but expresses it periphrastically by means of a nonfinite form of a verb meaning “to be, to stay, to stand” preceded by the word denoting the place, which may or may not be marked with the locative, for ex.: Tēvāram 4.17.4, viṇ niṉṟu / iḻittaṉar kaṅkaiyai “[Śiva] made Gaṅgā descend from [lit. “who had stood (in)”] heaven” (unmarked locative); I am grateful to Eva Wilden who has provided me with this example. (Cf. modern Tamil vīṭṭiliruntu “from the house”, where the “frozen” phrase-iliruntu is comprised of the locative marker -il and iruntu “having been”). Thus, the absence of the ablative in the TC list is appropriate in terms of linguistic description.

36 For this sūtra and its relevance in this context, see below.

37 Cf. the Sanskrit equivalent avadhi, which is sometimes found in the place or as a gloss of apādāna, for ex. in Cāndravyākaraṇa (CV) 2.1.81, avadheḥ pañcamī, or, in the Pāṇinian tradition, Kāśikāvṛtti ad 1.4.24, dhruvam yad apāyayuktam apāye sādhye yad avadhi bhūtaṃ tat kārakam apādānasaṃjñaṃ bhavati; see also Bhartṛhari’s use of the term avadhi in several verses of the Apādānādhikāra of VP 3.7 (e.g. VP 3.7.143ab, gatir vinā tv avadhinā nāpāya iti gamyate).

38 One only finds karuvi in T. 2.2.12 (defining the third case, i.e. the instrumental) and the pair nilam/kālam in 2.2.20 (defining the seventh case, i.e. the locative).

39 “toḻiṉ mutaṉilai” eṉṟatu toḻilatu kāraṇattai. kāriyattiṉ muṉṉiṟṟaliṉ “mutaṉilai” āyiṟṟu. kāraṇam eṉiṉum kārakam eṉiṉum okkum (Tr. Chevillard 1996: 218–19: “L’expression employée… ‘antécédents de l’ acte’ [désigne les] causes (kāraṇam) de l’action. Comme [la cause] précède (muṉṉiṟṟal) la conséquence (kāriyam), elle a été [appellée] ‘mutaṉilai’ (litt. ‘état initial’). Que l’on dise ‘cause’ (kāraṇam) ou que l’on dise kārakam ‘facteur causal’ (Skt.), cela revient au même (ottal)”).

40 T. 2.6.37: nilaṉum poruḷuṅ kālamuṅ karuviyum viṉaimutaṟ kiḷaviyum viṉaiyum uḷappaṭav avvaṟu poruṭkum ōraṉṉav urimaiya ceyyuñ ceytav eṉṉuñ collē “The locus, the object, the time and the instrument, including also the word for the agent and the action: words such as ceyyum and ceyta have a scope (urimai, lit. “right”) that is equivalent to these six types”. Note that the term mutaṉilai itself does not appear in the sūtra and the terms for “object” and, especially, “agent” are not quite the same as in T. 2.3.29: viṉaimutal and poruḷ against ceyvatu and ceyappaṭuporuḷ (on this sūtra and its relevance for peyareccam constructions, see Chevillard 1996: 362–65; 2007: especially 196–97).

41 The relatively few instances where this happens — such as the nominative singular of most stems ending in a consonant — are treated by Pāṇini as if the basic suffix was replaced by a “zero” suffix.

42 Alternatively, the case of an unmarked noun may be conveyed by its position in the word order of the sentence.

43 For the use of passive in Tamil and its acknowledgment by the grammatical speculation (first occurring in the Vīracōḻiyam), see Subrahmanya Sastri 1934 (1997: 175–76).

44 An example of this is the induced agent (prayojya) or causee in a causative construction with certain verbs, according to A. 1.4.52 (see above).

45 Namely, any object that is either anīpsita or akathita.

46 VP 3.7.47: satī vāvidyamānā vā prakṛtiḥ pariṇāminī | yasya nāśrīyate tasya nirvartyatvaṃ pracakṣate.

47 VP 3.7.48ab: prakṛtes tu vivakṣāyāṃ vikāryaṃ.

48 These examples are found in the Prakīrṇaprakāśa of Helārāja (ca. 10th century), a commentary on the third book of the VP, on VP 3.7.50 (VP 3.i. 260.l4–15), which contains a very clear and concise overview of the topic.

49 VP 3.7.51: kriyākṛtā viśeṣāṇāṃ siddhir yatra na gamyate | darśanād anumānād vā tat prāpyam iti kathyate.

50 kartā is to be read here by anuvṛtti from a previous sūtra, A. 3.1.68, kartari śap.

51 These expressions do not translate easily into English. In certain cases, the verb can be used in antitransitive modality (e.g. “rice is cooking”) to convey the exceptional easiness of the action, which happens as if by itself in the absence of the real agent, but in other cases, e.g. with “to cut” (Skt. lū-), one is obliged to resort to a passive construction. Thus, the translation of the second example would be “the meadow is (easily) cut”.

52 TCC 72.1: ceyppaṭu poruḷ mūṉṟaṉaiyum paṟṟivarum vāypāṭukaḷai virikkiṉṟār “he elaborates on the models that occur based on the three (types of) objects”.

53 TCC 72.6: iyaṟṟappaṭuvatu orutaṉmaittu ākaliṉ ataṟku oru vāypāṭē kūṟiṉār.

54 Significantly, the Pāṇinīyas themselves express some reservation about its usefulness: for example, see VP 3.7.80 and Helārāja’s commentary upon it.

55 Note, in particular, that there is no provision for causative forms in the T. (cf. Subrahmanya Sastri 1997: 146 ff.).

56 “C’est une habitude qui est en force dans l’usage, que d’énoncer, mis en action, l’objet qui subit [le faire], comme [s’il était] ce qui a fait” (tr. Chevillard 1996: 381).

57 TCC 246.2: vaḻakkiyaṉ marap’eṉavē ilakkaṇam aṉṟ’eṉṟav āṟu ām “En disant ‘habitude qui est en force dans l’ usage’, cela veut dire que ce n’est pas [selon] les règles” (tr. Chevillard 1996: 381).

58 Cf. MBh II. 67.20: atrāpi yāsau sukaratā nāma tasyā nānyaḥ kartā. The example odanaḥ pacyate svayam eva is not found as such in the MBh, but it is clearly implied at least once (MBh ad A. 3.1.67, II. 59.4). As far as I can tell, it is first spelled out in KV ad A. 1.3.78 and, later, in its commentaries.

59 Compare, for example, the English verb “to break” in the following sentences: “he broke the glass”, “the glass broke”.

60 Note that the Tamil sentence is not an exact calque of the Sanskrit, for the Tamil verb is active while the Sanskrit is passive. This conjures up a situation in which grammarians such as Cēṉāvaraiyar and his readers were familiar with the theoretical literature in both languages and the languages themselves. The translated example simply serves as a reminder.

61 PV 12, auto-commentary (p. 65). The mention of Kaiyaṭa (possibly 11th century), the author of the oldest surviving complete commentary on Patañjali’s MBh, is somewhat puzzling, for Kaiyaṭa explicitly refers to the threefold classification of karman under A. 3.2.1, where he quotes some of the relevant verses from the VP, and hints are found elsewhere in his work, but he does not dwell on it. However, this is probably explained by the fact that by the 17th century his Mahābhāṣyapradīpa had long established itself as the standard commentary on the MBh.

62 The work is ascribed to Devanandī (also known as Pūjyapada), who was probably active in the 5th century CE. According to a Jaina legend, the work was composed by the Jina in order to teach the god Indra, hence the name Jainendra (see Belvalkar 1915/1997: 52–53).

63 This rule prescribes the primary (kṛt) suffix aṆ (-a) after a verbal base co-occuring with a noun denoting the object of the action, to derive compounds such as kumbhakāra “pot-maker”, etc.

64 See Belvalkar (1915/1997: 32–33); Tripāṭhī (1981: xvii–xxi).

65 On the dates of these texts, and of the JV, mentioned in the following paragraphs, see Oberlies (1996: 269–75).

66 Literally, īpsita, the passive past participle of the desiderative stem īpsa-, from the root āp- “to obtain”, means “that which [the agent] wishes to obtain”.

67 CVV and CV 2.1.43 (1, 138): odanaḥ pacyata ity odanaśabdād vyāpyatā na gamyate, kiṃ tarhi tiṅantāt.

68 Pace Deshpande (1979), who considers the CV and its vṛtti to be the work of the same author and the Cāndra-Paribhāṣasūtra to go back to Candragomin; against these views, see Oberlies (1992: 162 ff.), and Vergiani (2009: 183 ff.).

69 VC 41ab: paṟṟoṭu viṭē irupuṟam tāṉ teri tāṉ teriyā naṟ karuttā tīpakam ām karumam. I wish to thank Whitney Cox for his help with the translation of the examples given below.

70 For the parallel between categories no. 4 and 5 in this list and the Pāṇinian notions of abhihita/anabhihita, cf. the section on Pāli grammars below.

71 For this interpretation, cf. below. I know of no Sanskrit work where the phrase dīpaka karman is used. Whitney Cox has suggested to me that this might be a terminological borrowing from Alaṃkāraśāstra, namely from the figure dīpaka, “illuminator” or zeugma, where one word must be multiply construed with other parts of the sentence (personal communication, December 2010). In this sense, the akathita object may be seen as serving as a kind of link between the action and the īpsitatama object. Note that T.V. Gopal Iyer, in his kuṟippurai to VC edition, calls it tuvikarumam (dvi-or dvaya-karman) and remarks that the author of VC had used the name tīpakak karumam, but he does not explain the phrase or the reason for Puttamittiraṉ’s choice of a different label (VC, p. 185).

72 N. 295 (according to the numbering found in the editions with Mayilainātar’s commentary), p. 145: iraṇṭāvataṉ urup’ aiyē ataṉ poruḷ ākkal aḻittal aṭaital nīttal ottal uṭaimai āti ākum.

73 However, the list is concluded by āti (corresponding to Skt. ādi), thus allowing other values to be added.

74 I am not sure I understand the reasons that may have led the author of the N. to make this a separate category, while it may probably be accommodated under aṭaital.

75 The inclusion of Aggavaṃsa’s work here is justified by the fact that Burmese Buddhism had close and constant links with Sri Lankan Buddhism.

76 It is worth recalling here that the first reordered commentary on Pāṇini’s A. was the Rūpāvatāra of the Sinhalese Buddhist Dharmakīrti, who probably flourished in the late 10th century (see Haraprasad Sastri, 1931: xcv–xcvi, and Cardona, 1976: 285).

77 MVP 2.2: nibbattivikatippattibhedena tividham kammam […] yassāsato jananaṃ karīyati sā nibbatti, yam pana vikarīyati santam evāvatthantaram āpādīyati sā vikati, yatra tu pattivyatirekena kriyākatā visesā na vibhāvīyante sā patti (pp. 37–38).

78 See Kahrs (1992: 46 ff.).

79 The T. itself is likely to be the culmination of a long period of earlier speculation (see Chevillard 2000a: 193). Nowadays, most scholars agree that it is probably a composite text, the various layers of which were composed at different times by more than one author. For a survey of the arguments on the date of the T. see Zvelebil (1995: 705–708).

80 That is, the process by which a language is provided with a grammar, namely a formal description. I borrow the term from Auroux (1994), who considers grammatisation a “technology” — in the sense, as I understand it, that it empowers a particular variety (one language) of that natural human faculty that is speech, providing it with a stability and transmissibility that potentially transcend local and temporal variation. I believe this concept may prove invaluable in understanding the history of South Asian culture, of which grammar is, as is well known, a key element.

81 For the concept of Sanskrit cosmopolis, see Pollock (1996; and 2006: in particular 39–280).

82 Among them — in different ages — Kannada, Telugu, Malayalam (for which see Rich Freeman’s contribution to the present volume), Tibetan, Old Javanese and Persian.

83 On the origins of Tamil grammar, see Chevillard (2000 and 2009b).

84 On the crucial role that writing has played, historically, in the development of linguistic knowledge, see Auroux (1994: 47 ff.). Although his argument is convincing, it raises important historiographical problems when applied to the Indian context, considering that the earliest attestations of the use of writing in ancient India are relatively late (3rd century BCE), while Pāṇini is generally dated not later than the 4th century CE.

85 See Mahadevan (2003: 90 ff.).

86 See Chevillard (2000b: 174; and 2008: 29–36).

87 Ormultilingual, if we also consider Pāli and possibly other languages (such as literary Prakrits, other Dravidian languages and Sinhala).

88 Chevillard (2009a: 72, n. 2) mentions Campantar (ca. 8th century), one of the authors of the Tēvāram, as “an example of ‘Sanskrit-Tamil’ tri-lingual person”. On the Sanskrit side, several important authors are believed to be from the Dravidian-speaking regions, from Nāgārjuna to Śaṅkara, Daṇḍin and, among the grammarians, Haradatta Miśra, as I mentioned above, but we do not know if they ever used their mother tongues in their intellectual pursuits.

89 On this important phenomenon, see the various contributions in Wilden (2009a).

90 This was preceded or accompanied and, in a sense, prepared by the anthologisation of the Caṅkam corpus, which Wilden sees as the “constitution of a Tamil literary past via a process of selection and evaluation” (Wilden 2009b: 145; on this process, see also Wilden 2002).

91 His explicit mention of uṭaiya and ila among verbs governing an object to be attained may have been meant to account for the category uṭaimai in the N.

Table des illustrations

URL http://books.openedition.org/ifp/docannexe/image/2888/img-1.jpg
Fichier image/jpeg, 235k

Auteur

Lecturer in Sanskrit at the Faculty of Asian and Middle Eastern Studies, University of Cambridge. His main areas of research are the Sanskrit grammatical tradition and the history of linguistic ideas in ancient South Asia. He is the director of the project “The intellectual and religious traditions of South Asia as seen through the Sanskrit manuscript collections of the University Library, Cambridge” (http://sanskrit.lib.cam.ac.uk/), funded by the UK Arts and Humanities Research Council. He has co-edited Studies in the Kāśikāvṛtti. The section on pratyāhāras. Critical edition, translation and other contributions (2009).

© Institut Français de Pondichéry, 2013

Conditions d’utilisation : http://www.openedition.org/6540

Acheter

Volume papier

amazon.fr
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search