© SHESL/EDP Sciences

1 Introduction

Toward the end of the first millennium CE¹ grammarians throughout South Asia and beyond began to adopt the structure and theory of Sanskrit grammars to new languages including those of the Dravidian family.² Earlier examples are, of course, available for the Prakrit languages and Pāli, but the application of these models to Dravidian languages, structurally much further afield than the Sanskrit daughter languages, forced a variety of new theoretical choices in the description of the target languages as well as about their relation to Sanskrit. Understanding how these new grammars relate to their Sanskrit predecessors requires both a close reading of the grammars themselves and a fairly broad knowledge of the possible source texts. Work on the grammars of Dravidian languages has been rather uneven, and, excepting Tamil, only a relatively small number of studies exist on the grammars of Kannada, Malayalam, and Telugu. Even in the case of Tamil, for which there is more extensive secondary literature,³ there remains substantial work to be done on the numerous grammars of the second millennium before they can be properly situated in the history of grammatical literature in India.

One such Tamil grammar, the Vīracōḻiyam (VC), composed in the 11^th century by Puttamittiraṉ and commented on by Peruntēvaṉ in the 12^th,⁴ represents an attempt to incorporate Sanskritic elements into the grammatical and poetological description of Tamil even though a more suited grammar cum poetics for Tamil, the Tolkāppiyam (T), had long been in existence.⁵ As I and others have discussed elsewhere, the motivation for creating such a grammar at the time and place that Puttamittiraṉ did is complex and stems not only from the general milieu in which the vernacular languages looked to Sanskrit as a model of a literary language regulated by a grammar, but also from Puttamittiraṉ's personal interest to convey Sanskrit knowledge to a Tamil audience. In addition, the Buddhist literary community that Puttamittiraṉ belonged to clearly determined which texts he used as models. The present paper, a continuation of a preceding article,⁶ will focus on detailing the system of kārakas “factors of an action” in the VC and which texts served as its sources. As I have demonstrated in the aforementioned essay, this line of research helps to illuminate both how the source material has been appropriated for a different language and, conversely, which Sanskrit sources had importance for the literary circles during the period in question. As I will show, to understand the VC, we must look outside of the Pāṇinian school at works hardly studied nowadays, especially the Prayogamukha (PM), but ones that are paramount for understanding the development and spread of grammar in South Asia.

2 Kārakas and Case

2.1 Introduction to Kāraka Theory

The relation between an object, as denoted by a noun, and an action, as expressed by a verbal root, forms one of the core facets of Sanskrit grammar. To describe these relations, Pāṇini has defined in the Aṣṭādhyāyī (A) six kārakas “factors”⁷ which can be applied to both nominal stems involved in an action as well as to suffixes added to a verbal root. For example, if one has the three elements devadatta (nomen proprium), odana “rice”, and pac “to cook”, the speaker will label the two nouns according to the role they are to play in the syntactic unit. If Devadatta is independently responsible for the act of cooking, he will receive the designation kartṛ “agent”,⁸ and if the rice is the thing which one most wants to obtain through the act of cooking, it is labelled the karman “patient”.⁹ Similarly, the 3^rd pers. sing. ending -ti in pacati “cooks” is likewise said to express the agent, hence we know it is an active verb in the sentence devadatta odanaṃ pacati “Devadatta cooks rice”. Within the framework of Pāṇini's grammar these labels serve primarily to produce correct morphology. The simplest application is that a noun labelled with a particular kāraka will take a particular case. If, for example, a noun has been labelled the karaṇa “instrument” of an action, then it will receive the third or “instrumental” case by P. 2.3.18 kartṛkaraṇayos tṛtīyā “there is the third (instr.) case when (the noun) expresses the agent or instrument (provided neither are already expressed).”¹⁰ Similarly, if it is labelled karman, it takes the second or accusative case by P. 2.3.2 karmaṇi dvitīyā “the second (acc.) case occurs when the patient is expressed (provided it is not already expressed).” It is important to note that in the sūtra the word for the second case (dvitīyā) is in the nominative and the corresponding word for the patient kāraka (karmaṇi) is in the locative, since this syntax will be repeated in the VC.

Each kāraka does not, however, correspond to one case ending, and the addition of the overt case is contingent on several conditions, including the type of suffix added to the verbal root. At present the most important condition is that for rules like P. 2.3.2 and 18 to apply, a noun's kāraka must not be expressed elsewhere in the relevant syntactic unit. In effect, this means that if a suffix added to a verbal root expresses the same kāraka as a noun involved in the action expressed by that root, then the general rules for adding a case to that noun are barred.¹¹ Accordingly, when the speaker wishes the personal ending -ti in pacati to express the kartṛ “agent”, then we can no longer add the instr. to devadatta by P. 2.3.18, because the kartṛ is already expressed. In such instances, the nominative case will be used since it is not associated with any particular kāraka, only the meaning, etc. of the nominal stem.¹² Therefore, devadatta stands in the nominative in the sentence devadatta_nom. odanaṃ_acc. pacati “Devadatta_nom. cooks¹³ rice _acc”, and not the instrumental. On the other hand, if the personal ending added to pac should express the karman “patient”, then P. 2.3.2 is barred, and odana “rice” will take the nominative case. Since the personal ending does not express the agent in this case, devadatta will receive the instrumental case ending by P. 2.3.18, and the resulting sentence is: odano_nom. devadattena_instr. pacyate “Rice_nom. is cooked¹⁴ by Devadatta_instr.” Through the introduction of the condition anabhihite at the beginning of P. 2.3 and leaving the nominative case without a specific kāraka, Pāṇini is able to account for active and passive constructions in a quite ingenious manner. Although one could add much detail to my brief summary of the kārakas in the A, for our present purpose the above outline will suffice.

In the subsequent commentarial literature on the A, the kārakas generated erudite and philosophically deep discussions about the nature of linguistic expression and how the speaker wished to describe the external world, emanating, as always, out of discussions about Pāṇini's sūtras and how they can or cannot account for linguistic usage. Of lasting importance is the Sādhana-samuddeśa “lesson on factors” in the Vākyapadīya (VP) by Bhartṛhari (5^th cent.), where the theoretical discussions of the Mahābhāṣya (MBh) of Patañjali (2^nd cent. BCE?) are summarized and elaborated in verses called kārikās. One point of interest was the subdivision of the kārakas so that one could speak more precisely about an agent, patient, etc. in relation to a verbal action: the act of making a pot has a very different kind of effect on the patient (karman) than the act of looking at the sun. Membership to some of these subclasses also has formal linguistic implications.¹⁵ Although Bhartṛhari adduced several of these subvarieties from the MBh and introduced some of his own, later authors expanded and refined this list.¹⁶ At some point, these “sub-kārakas” became part of the basic curriculum in Sanskrit grammar and the number of subtypes were versified in the very popular introductory work variously known as the Prayogamukha (PM), Vārarucasaṃgraha or, in a shortened form, Kārakacakra (not to be confused with the work by Puruṣottamadeva), about which I will have much more to say shortly.¹⁷ This work along with a commentary attributed to Dharmakīrti clearly served as the basis for sections of the VC, a fact already noted with regard to the tokaip-paṭalam “Section on Compounds”.¹⁸ The theory of kārakas also disseminated into later non-Pāṇinian grammars in different ways; some remained relatively faithful to Pāṇini's definitions and overall plan,¹⁹ while others altered the terminology and simplified the scheme at the sacrifice of theoretical elegance and rigor.²⁰ For the VC, these non-Pāṇinian grammars were of as much importance as Pāṇini's A (if not more so), and their customary place in the backseat of scholarship on the history of Sanskrit grammar has resulted in a rather skewed view of the discipline's development. This neglect is, however, slowly being remedied.

2.2 Kārakas in the Vīracōḻiyam

The VC contains two sections that treat the description and assignment of case: the vēṟṟumaip-paṭalam “Section on Case” and the upakārakap-paṭalam “Section on the Sub-Kārakas.” As I have treated the former at some length elsewhere, I will here only mention a few relevant details from the former before moving onto my study of the kārakas. Puttamittiraṉ links the cases to the kārakas in VC 34 & 35 using much the same syntax as the Sanskrit grammarians, i.e., the case in the nom. and the kāraka in the loc. or oblique standing for a loc. Although the nominative is also reserved for expressing merely the meaning of a nominal stem (VC 33), we do not find any rule similar to P. 3.2.1 anabhihite much like in the Kā.²¹

In the upakāraka-paṭalam (VC 38–43), the kārakas that were initially listed in (VC 29) are now defined at the outset (VC 38), exemplified in a sentence (VC 39), and subdivided into 23 upakārakas “sub-factors” (VC 40–41). Thereafter comes a set of exceptions to the general correspondences between the cases and the kārakas set out in the vēṟṟumaip-paṭalam, some of which appear to be more descriptive of Sanskrit than Tamil usage (VC 42–43). This section of the VC provides us with some of the richest material for comparison with the Sanskrit sources because of the specificity necessary in naming and defining of the kārakas and their subdivisions. In the present section I will analyze the definitions in light of the Sanskrit sources and then turn to the upakārakas, a term that does not occur in Sanskrit in the same meaning, to the best of my knowledge, though its meaning is clear.²² Since I have already discussed the peculiarities of the Tamil names for the kārakas and their likely sources elsewhere, I will focus on the definitions themselves.

The six kārakas listed in VC 29 and defined in VC 38 are the following: 1) karuttā (= Skt. kartṛ) “agent”, 2) karumam (= Skt. karman) “patient”, 3) karaṇam (= Skt. karaṇa) “instrument”, 4) kōḷi “recipient”, 5) avati (= Skt. avadhi) “limit”, 6) ātāram (= Skt. ādhāra) “base”. The final three differ from their respective Pāṇinian counterparts, sampradāna “recipient”, apādāna “departure”, and adhikaraṇa “locus”. The reason for the last two is that Puttamittiraṉ follows the terminology used in the CV, namely avadhi in CV 2.1.81 and ādhāra in CV 2.1.88.²³ kōḷi, the only kāraka with a Tamil name – the remainder are direct borrowings from Sanskrit with the necessary phonological/orthographic adaptations – has been translated, I have argued, because according to some grammarians the Sanskrit term is expressly anvartha “etymologically significant”, whereas the other kārakas are not. This departure from the Pāṇinian model continues when we examine the definitions given for the kārakas and further demonstrates the importance of the non-Pāṇinian schools in the dissemination of Sanskrit grammar.

For each of the six kārakas Pāṇini has given one, at times quite complex, primary definition and then several supplementary definitions under specific lexical or semantic conditions. Later Sanskrit grammarians of non-Pāṇinian schools do not uniformly follow these definitions and there is often much simplification in wording.²⁴ Puttamittiraṉ has likewise taken a simpler approach, one that mirrors what we find in the Kā., and its primary commentary, the vṛtti by Durgasiṃha (6^th – 8^th cent.?). To demonstrate this most effectively it is worth citing VC 38 in full along with parts of its commentary.

VC 38 mētaku nal toḻil ceyvāṉ karuttā. viyaṉ karuvi

tītil karaṇam. ceyappaṭṭatu ākum tiṟal karumam.

yātaṉiṉ nīṅkum avati atu ām. iṭam ātāram ām.

kōtu aṟu kōḷi maṉ koḷpavaṉ ākum koṭiyiṭai-ē.²⁵

The karuttā “agent” is the one who performs an eminent, good deed;

The faultless karaṇam “instrument” is the excellent means;

The strong karumam “patient” is what is done;

The avati “limit” is that from which one departs;

The ātāram “base” is the place;

The blemish free kōḷi “recipient” is the permanent one who receives,

O vine-waisted girl!

If we compare these definitions with the corresponding ones in the A, the rather introductory and pragmatic level of Puttamittiraṉ's text becomes clear. For example, in the A the karman is defined as kartur īpsitatam “what is most desired on the part of the agent” (P. 1.4.49); the agent is svatantraḥ “independent” (P. 1.4.54). The Tamil definitions will, however, sound slightly less shallow if we recall that, save kōḷi, the names of the kārakas are in Sanskrit whereas the definitions are in Tamil so that one also gets the feeling of an explanation of a foreign technical term. Still, the definition for kōḷi − the recipient is the one who receives − cannot but sound tautological. But leaving Pāṇini aside, we find parallels in the Kā. and its vṛtti. To demonstrate this, let us look at how two kārakas are defined in the respective texts.

1) karuttā / kartṛ

VC 38 mētaku nal toḻil ceyvāṉ karuttā.

The karuttā “agent” is the one who performs an eminent good deed.

Gloss: yāt' oru toḻilaic cevāṉ evaṉ, avaṉ karuttāk-kārakam ām.

The one who performs some action is the agent-kāraka.

Kā. 2.4.16 yaḥ karoti sa kartā.

The agent is the one who performs.

Gloss: yaḥ kriyāṃ karoti sa kartṛsaṃjño bhavati.

He who performs an action has the technical term “agent”.

2) avati / apādāna

VC 38 yātaṉiṉ nīṅkum avati atu ām.

The avati “limit” is that from which one departs.

Gloss: yātoṉṟiṉṟum oru poruḷ nīṅkuvatu aḵt' avatik-kārakam ām.

That from which an object departs is the limit-kāraka.

Kā. 2.4.8 yato 'paite bhayam ādatte vā tad apādānam.

The apādāna is that from which one departs, (of which) there is fear, or (from which) one receives.

Gloss: yasmād apaiti yamād bhayaṃ bhavati yasmād ādatte vā tat kārakam apādānasaṃjñaṃ bhavati.

That kāraka from which one departs, of which there is fear, or from which one receives has the technical term “departure”.

There are several striking parallels between the Sanskrit and Tamil texts, in particular Peruntēvaṉ's gloss. At the level of syntax, both make use of relative clauses,²⁶ a construction that was originally foreign to Tamil.²⁷ Although Puttamittiraṉ only uses the relative construction for avati, Peruntēvaṉ uses it in his gloss on all six kārakas. In fact, I have so far not been able to find any further occurrences of this relative clause construction in the VC and its commentary. Its frequency here must be attributed to the imitation of the Sanskrit definitions, in all likelihood those of the Kā.²⁸ With regard to lexical choice, Puttamittiraṉ and Peruntēvaṉ both have recourse to the root ceytal “to do, perform” in defining the agent just as in the Kā. and its gloss with the Sanskrit equivalent kṛ “to do, perform”. The same parallel also arises for the definition of karumam / karman²⁹ and to a somewhat lesser extent in the case of karaṇa.³⁰ The definition of the two remaining kārakas, ātāram and kōḷi, show even less similarity to what we find in the Kā. and its vṛtti besides the relative clause construction. This need not erase the significance of the other parallels since I do not wish to argue that Puttamittiraṉ and Peruntēvaṉ were translating the Kā. sūtras or its commentary into Tamil. Rather, the VC represents more of an attempt to import the general ideas about language from Sanskrit into Tamil, and I believe I have sufficiently shown that the general framework for the definitions of the kārakas derives from what we find in the Kā. school as opposed to any of the other schools current at the time.³¹ This fits well with the Buddhist environment in which the VC was produced, since the Kā., though generally popular in India, traveled with Buddhists into Tibet already in the first wave of translations.³²

3 The 23 Upakārakams

3.1 The Origins of the Upakārakams

In addition to a general definition for each kāraka, Pāṇini adds a number of additional sūtras that describe other conditions under which a noun may be labelled a particular kāraka. For example, the karman “patient” is not only what is most desired to be obtained on the part of the agent,³³ but also what is “similarly related but not desired” and what “is not denoted by any other kāraka”.³⁴ Kātyāyana, whom Patañjali follows, makes further semantic distinctions among the karman, classifying some as nirvartyamāṇa “being created” and others as vikriyamāṇa “being modified” in an effort to restrict the application of P. 3.2.1.³⁵ It is Bhartṛhari, however, who collects these subtypes into kārikās and presents them in the Sādhana-samuddeśa, the seventh section of the Padakāṇḍa of the VP, which deals with the kārakas. The subtypes of karman, which are spread out in the MBh, are presented together in VP 3.7.45–48, though not every kāraka receives such a clear-cut subdivision. Bhartṛhari was, therefore, responsible for systematizing and deepening the discussion about the kārakas and gave the impetus for later grammarians to simplify his presentation of the kāraka subtypes and modify them as was seen fit.

One text that lays out only the number of subdivisions for each kāraka, 23 in total, is a set of 26 verses on the basics of Sanskrit grammar³⁶ according, more or less, to the Pāṇinian system and which generally goes by the name Vārarucasaṃgraha. In order to know the names of each subtype, however, a commentary is necessary,³⁷ and based on the specific names of the sub-kārakas I wish to demonstrate that the commentary attributed to Dharmakīrti must have been known to Puttamittiraṉ and Peruntēvaṉ. In manuscripts this commentary along with the verses are sometimes referred to as the Prayogamukha, and I will also use this title out of convenience for both the main text and commentary.³⁸ Further information about the text's authorship and place of composition are unfortunately unknown, although the distribution of manuscripts containing the text makes its popularity across the subcontinent indisputable and its influence on the VC places it at least as far back as the beginning of the second millennium.³⁹ Furthermore, the PM must have had strong currency among Buddhist circles. In addition to influencing the VC, it was also, as I will show, used by Aggavaṃsa for his Pāli grammar, the Saddanīti, and travelled to Tibet in the first wave of Sanskrit grammatical works to be translated as part of the collection Bstan-'gyur “Translation(s) of the (exegetical) treatises (i.e. commentaries and compendia)” in the early 14^th century, under the title Rab-tu-sbyor-ba'i-sgo'i-'grel-pa or Prayogamukha-vṛtti.⁴⁰ Despite this popularity, other works of Sanskrit grammar do not follow the PM in listing the subtypes or even the same number of subtypes of kārakas when they are given. In commenting on the relevant sūtras in the A, grammarians are more likely to quote from the VP and repeat the divisions found there. In fact, I am not aware of any works within the Sanskrit grammatical tradition that presents the exact same scheme as in the PM.⁴¹ In Tamil, we find renewed interest in the topic some centuries later in the Pirayōkavivēkam by Cuppiramaṇiya Tīkṣitār (17^th cent.), which has long been known to be a translation of, or at least closely modeled on, the Vārarucasaṃgraha, or however the text was known at the time. I will return to this work at the end of the next section.

3.2 The Prayogamukha and the Upakārakams in the Vīracōḻiyam

Both the PM vv. 1–7 and the VC 40–41 divide the six kārakas into 23 subtypes.⁴² As noted above, for the precise names of these subdivisions we must rely on a commentary, and it is evident that Puttamittiraṉ was familiar with the list as Dharmakīrti gave it. This should come as no surprise given that both texts were written in Buddhist circles. The later commentary of Nārāyaṇa, by way of comparison, shows a number of divergences and follows more closely the Pāṇinian school. A full exposition of each kāraka along with its examples would far exceed the scope of this paper, so at present I will only discuss here 2 kārakas in detail, the kartṛ / karuttā “agent” and the sampradāna / kōḷi “recipient”. Nevertheless, this will be sufficient to prove the relation between the texts.

The agent is said to have five subtypes in PM 2a, which are then specified in the commentary on p. 151f.

Kartṛ “agent” − 5 subtypes:

1. svatantra- “independant-”,

3. karma- “patient-”,

2. hetu- “cause-”,

4. abhihita- “expressed”,

5. anabhihita-kartṛ “unexpressed-agent”.

All of these categories can be traced back to the A itself. The first two derive from P. 1.4.54 and 55, the two definitions of kartṛ given in the kāraka section of the A. The former defines the kartṛ as the one independent (svatantra) in effecting the action; the latter accounts for the agent of causation in causative constructions by also terming the instigator (prayojaka) of an independent agent kartṛ “agent”, whereby it also receives the designation hetu “cause”. Thus in sentences such as in “John makes Bill cook,” Bill is independent with respect to carrying out the act of cooking, but John instigates Bill to act. In Sanskrit such an agent is termed the hetu-kartṛ. The third type, the karma-kartṛ, refers to the agent in constructions like pacyata⁴³ odanaḥ svayam eva “The rice cooks by itself,” where the Sanskrit grammarians conceptualize the agent, in this case the rice, as also acting like the patient (karmavat) according to P. 3.1.87. The final two categories harken back to the heading sūtra P. 2.3.1 anabhihite discussed above and indicate where the agent-kāraka is expressed. When the agent is expressed (abhihita) — we must understand expressed by the personal ending on the associated verb — the agent stands in the nominative case, i.e., we have an active construction. When the agent is not expressed (anabhihita), the personal ending on the finite verb expresses the karman (or bhāva “the action itself”), and the agent is consequently put in the instrumental, i.e., we have a passive (or impersonal) construction. Furthermore, these last two categories apply concomitantly to a kartṛ that already has one of the first three labels.⁴⁴ E.g., a svatantra-kartṛ can be either abhihita or anabhihita depending on whether the main verb is active or passive. In essence, this fivefold classification does not add anything new to what we find already in the A nor does it add any particular insight into the nature of the agent. It is, however, a handy way for beginning students to identify agents according to the sūtra primarily responsible for their derivation and whether the sentence is active or passive. As yet, I am not aware of any other Sanskrit text that gives this exact list,⁴⁵ although the SN recognizes the exact same fivefold classification.⁴⁶

Turning now to the VC, we find the following five upakārakams “sub-kārakas” listed for the karuttā:

Karuttā “agent”, 5 upakārakams:

1. kāraṇak-karuttā “cause-agent”, 

2. tāṉ teri karuttā “self evident agent”, 

3. tāṉ teriyāk karuttā “non self evident agent”,

4. karumak-karuttā “patient-agent”,

5. talaimaik-karuttā “head-agent”.

In analyzing these five terms, we may look at both the method by which the Sanskrit words were adopted at the lexical level into the Tamil language as well as the semantics of each term within the context of Tamil grammar.⁴⁷ In sum, Puttamittiraṉ has employed two basic methods at the lexical level: direct borrowing with only phonological/orthographic adaptations and translations/calques, which attempt to mimic the structure and meaning of the Sanskrit original. There is, however, some interesting grey area that requires more explanation. With regard to the Tamilized term's function in the grammar there is both continuity as well as nuanced modifications to better fit the peculiarities of Tamil and the grammatical system developed by Puttamittiraṉ and Peruntēvaṉ. I must emphasize, however, that we are almost entirely dependent on the commentator for understanding the details of these technical terms and in my analysis below I rely entirely on his explanations and examples.

Among the five technical terms, only the term karumak-karuttā “patient-agent” is a direct borrowing. It is simply the Sanskrit word karma-kartṛ with the necessary phonological/orthographic adaptations for Tamil, such as splitting the consonant clusters rm and rt with a u. Peruntāvaṉ also understands the term to have a similar meaning as in Sanskrit, i.e., the subject of a sentence which acts as both the patient and the agent, but in application there are nuanced differences on account of the Tamil verb system. Let us turn to our first example:

Comm. ad VC 40 p. 42: naṉmai_nom. tāṉē veḷippaṭum viḻumiyōr pakkal eṉpuḻi, naṉmaiyaic_acc. cāttaṉ veḷippaṭuttiṉāṉ eṉṟāṟ pōla naṉmai karumamāyk karuttāp piṟit' oṉṟāy nillātu karumamuṅ karuttāvun tāṉēy āy niṟṟalāl, karumak-karuttāv āyiṟṟu.

When one says “Goodness_nom. arises (veḷippaṭum) by itself (tāṉ-ē) in excellent people,” because (the word) “goodness”, just as in (the sentence) “Cāttaṉ made goodness_acc. arise (veḷippaṭuttiṉāṉ),” (but) being the patient without any other agent, occurs by itself as both the patient and the agent, it has become the patient-agent.

In Tamil the verbal root under discussion is veḷippaṭu “to appear, arise”, which can be further analyzed as a combination of veḷi “outside” and the verbalizer paṭu “to experience, undergo”. The root is intransitive when conjugated as a so-called “weak” verb, i.e., veḷippaṭutal, as it is in veḷippaṭum in the example. When, however, the root is conjugated as a “strong” verb, i.e., veḷippaṭuttutalal, it is causative and hence transitive “to make appear, arise, to reveal”, as it is in veḷippaṭuttiṉāṉ in the example. Peruntēvaṉ sees the parallel between the two conjugations of veḷippaṭu, but does not analyze them in the way I have just done. Instead, he sees veḷippaṭu as inherently taking two arguments, an agent and a patient. When both are present, we have the causative/transitive conjugation, but when the agent is missing and only the patient is present, we have the intransitive conjugation, and the patient also acts as the agent. Implicit in Peruntēvaṉ remarks is that the morphology is dependent on the type of karuttā.

If we return now to the Sanskrit example pacyata odanaḥ svayam eva “the rice cooks by itself,” we find a number of similarities. To begin with Peruntēvaṉ imitates svayam eva with tāṉ-ē, a satisfactory Tamil translation. Additionally one can also see in the Tamil example the notion of sukaratā “being easily performed”, one of the semantic nuances in using the karma-kartṛ in Sanskrit,⁴⁸ for surely goodness arises very easily in excellent people. With regard to morphology, however, the two constructions are not exactly parallel. pacyate is the 3^rd pers. sing. pres. pass. of the root pac. By the time of the VC, Tamil also had a passive, described in VC 81, and formed with an infinitive plus paṭutal, as in aṭappaṭṭatu “it was cooked.” Interestingly, Peruntēvaṉ does not try to imitate the Sanskrit morphology, which would not have been idiomatic, but instead found another pair of verbal forms whose morphology is triggered by the type of karuttā. The other example composed by Peruntēvaṉ also follows this pattern: aṉpu keṭum tīyōr pakkal “Love perishes in sinful people.” The verb keṭu can be conjugated as either weak (keṭutal) “to perish” or strong (keṭuttal) “to cause to perish, to destroy” with the same intransitive/transitive distinction noted for veḷippaṭu. The application of the category of karumak-karuttā to Tamil does not parallel what we find in Sanskrit, above all because the construction in Sanskrit is relatively rare and requires passive morphology, but Peruntēvaṉ has still adopted the category to describe a prominent morphological phenomenon in Tamil verbal conjugation with some similarities to the Sanskrit counterpart.⁴⁹

The kāraṇak-karuttā is the equivalent of hetu-kartṛ “the causal agent”, but hetu “cause” has been replaced with the synonym kāraṇam,⁵⁰ a seemingly unmotivated change given that hetu was well established in Tamil as ētu by Puttamittiraṉ's time. This alternation could be chalked up to metri causa or simply the fancy of Puttamittiraṉ, but one could also see the choice as being based on an attempt to bring the causal agent in a closer etymological relation to the name for causative verbs, namely, kāritam in VC 65, a technical term likely borrowed from the Kā.⁵¹ Both kāraṇa and kārita are derivatives from kāri, the causative root of kṛ “to do”. With regard to function, the kāraṇak-karuttā is identical to its Sanskrit counterpart: it describes the instigator of another agent in a causal construction, as the examples make clear.⁵²

The three remaining terms are attempts to translate into Tamil Sanskrit technical vocabulary. The translation of svatantra as talaimai “headship, superiority” fits well with the standard meaning associated with svatantra in the commentaries, which usually include pradhāna “principal, most important thing”.⁵³ Based on the examples and Peruntēvaṉ's explanation the talaimaik-karuttā functions the same as its Sanskrit equivalent.

The last two terms are perhaps the most complex in so far that their source, (an-)abhihita, has a very technical meaning within the Pāṇinian system of grammar (explained above on p. 2f.) and the general concept is not explicitly adopted in the VC. As a result, Peruntēvaṉ understands the two terms to indicate whether or not a kāraka is unambiguously expressed at the level of morphology, not whether the kāraka is expressed by the personal ending on the verb. A tāṉ teriyāk karuttā “non-self evident agent” describes the agent-kāraka that is denoted by the nominative case. It is not “self evident” because one must first check the finite verb to determine which kāraka the noun in the nominative has. If the verb is active, the nominative is an agent; if it is passive, the patient. Hence, a tāṉ teriyāk karuttā occurs in active sentences in the nominative. A tāṉ teri karuttā “self evident agent” refers to an agent in the reverse situation where the morphology on a noun makes it clear that it is the agent without recourse to the verb, i.e., passive sentences in which the agent is expressed by the instrumental.⁵⁴ In short, the tāṉ teriyāk karuttā is found in active sentences in the nom., and the tāṉ teri karuttā is found in passive sentences in the instr. We can therefore equate the tāṉ teri karuttā with the anabhihita-kartṛ (both in passive constructions) and the tāṉ teriyāk karuttā with the abhihita-kartṛ (both in active constructions), an equation that becomes undoubtable when we look at the following examples along with Peruntēvaṉ's explanation.

Commentary ad VC 40 p. 42: koṟṟaṉāl_instr. koḷḷappaṭṭatu vīṭu eṉpuḻik karuttā iṭan terintu niṟṟalāl, tāṉ terikaruttāv āyiṟṟu.

When one says “the house was purchased by Koṟṟaṉ_instr.,” because the agent occurs with its (syntactic) place⁵⁵ known, it has become the self evident agent.

PM p. 153: anabhihitakartā. yathā: ...pacyata odanaḥ sūpakeṇa.⁵⁶

The unexpressed agent. For example: “The rice is cooked by the cook.”

Commentary ad VC 40 p. 42: tēvatattaṉ cōṟṟai aṭukiṉṟāṉ, eṉpuḻit tēvatattaṉ eṉṉuñ col tāṉē karuttā eṉṉum iṭan terintu nillāmaiyāṉum, cōṟṟai eṉṉuṅ kārakapatattāṉum aṭukiṉṟāṉ eṉṉuṅ kiriyāpatattāṉuṅ karuttā eṉṟu aṟiyappaṭutalāṉum, tāṉ teriyākkaruttāv āyiṟṟu.

When one says “Tēvatattaṉ is cooking rice_acc.,” the word tēvadattaṉ has become the non-self evident agent because the syntactic place “agent” does not occur as known, and because it is understood as the agent by means of the kāraka-word⁵⁷ “rice_acc.” and by means of the action word (i.e., verb) “is cooking”.

PM p. 152f.: abhihitakartā. yathā: ...odanaṃ pacati sūpakāraḥ.

The expressed agent. For example: The cook cooks rice.

Peruntēvaṉ clarifies that the iṭam, perhaps something like “syntactic place”, i.e., kāraka, is known in the case of the tāṉ teri karuttā. We can supply from context and the following example, that it is known immediately by the case ending. For the tāṉ teriyāk karuttā, its syntactic place is only inferable by reference to the other words in the sentence. The examples from the PM confirm the functional relation between the Sanskrit and Tamil terms described above.

Now that the function and meaning of the terms are clear, we can return to the specifics of the translation. The phrases tāṉ teri and tāṉ teriyā are both built off the verb terital “to be known, evident” with the 3^rd pers. sing. nom. pronoun tāṉ, often used as an emphatic particle like “itself” in English. Cf. tāṉ-ē in the examples for the karumak-karuttā above. teri is simply the bare root used in place of the relative participle teriyum; teriyā, on the other hand, is the negative relative participle. The pair nicely reflects abhihita and anabhihita which are also without and with negation, respectively. As I have demonstrated, the positive Tamil form tāṉ teri does not functionally correspond to the positive abhihita, nor the negated tāṉ teriyā to anabhihita. The reason for this is given by Peruntēvaṉ in the passage quoted above. One possible explanation for this oddity is that Peruntēvaṉ took abhihita to refer to a kāraka that is clearly expressed by the case ending on the word itself, not, as it is in Sanskrit, to a kāraka expressed by the personal endings on the main verb (inter alia). On account of this mismatch, which we can only attribute to Peruntēvaṉ with certainty, we have the rather counterintuitive correspondence between the Tamil and Sanskrit terms.

As a brief aside, I note that these same two terms do not apply to the karumam in exactly the same manner and hence no longer exactly relate to the (an-)abhihita karmans in Sanskrit. The tāṉ teri karumam is, as we expect, with an overt accusative ending and corresponds to the anabhihita-karman in the PM as can be seen from the corresponding examples:

Comm. ad VC 41 p. 44: vīṭṭai_acc. eṭuttāṉ taccaṉ.

“The carpenter built the house_acc..”

PM p. 155 anabhihitaṃ karma yathā kaṭaṃ_acc. karoti naraḥ.

“The unexpressed patient. For example: The man makes a mat_acc..”

The other pair, however, does not correspond in the same manner as the Sanskrit and Tamil agents did above. The abhihita karman refers to a patient that takes the nominative, the karman kāraka being expressed by the personal ending on the finite verb. This is the case in passive constructions as the example for abhihita karman in PM p. 155 shows: kaṭaḥ kriyate devadattena “A mat is made by Devadatta.”⁵⁸ In the VC the tāṉ teriyāk karumam still refers to a patient in a sentence where the karumam is expressed by ambiguous morphology, but, based on the examples given by Peruntēvaṉ, not to the expected passive construction.⁵⁹ Rather, it refers to instances where the patient simply lacks an overt case ending in an active construction, a common feature of Tamil.⁶⁰ Indeed, the VC expressly accounts for such unmarked accusatives by permitting the acc. suffix to sometimes be elided.⁶¹ Accordingly, we have an example very similar to the one just given but without the acc. case ending -ai: vīṭu taccaṉ kaṭṭiṉāṉ “The carpenter constructed a house.” Given this reasoning, in the VC the terms tāṉ teri and tāṉ teriyā have come to refer to the presence or absence of an unambiguous case ending, the instrumental for the agent and the accusative for patient. For the agent there is a direct parallel with Sanskrit anabhihita- and abhihita-kartṛ, but because of a peculiarity of Tamil grammar, tāṉ teriyāk-karumam also refers to the unmarked accusative (according to some versions of Peruntēvaṉ's commentary), and this has no parallel in Sanskrit.⁶²

The other kāraka I would like to discuss is the sampradāna / kōḷi “recipient”. What sets this one apart from the kartṛ / karuttā and the karman / karumam is that their subtypes have no direct link to any categories found in the A or MBh. For the three types of sampradāna we must look to Bhartṛhari. In adopting these categories into Tamil Puttamittiraṉ was rather creative in finding satisfactory Tamil equivalents.

The sampradāna or kōḷi “recipient” has three subdivisions in the PM and the VC:⁶³

PM p. 156: katamat trividhaṃ sampradānam? prerakam ānumantrikam anirākartṛkaṃ ca.

What are the three types of recipient? The instigator, pertaining to one who consents,⁶⁴ and pertaining to one who does not reject.

VC 40d: cīr aṇaṅku ārvam kiṭappu irappu ām kōḷi, tēmōḻiy-ē.

The recipient is affectionate, joined with goodness, circumstantial, (and) requesting, o girl with honey(-sweet) words!⁶⁵

I would equate the terms as follows: ārvam ∼ ānumantrika, kiṭappu ∼ anirākaraṇa, and preraṇa ∼ irappu.

All three of these are based on the causes for an object to gain the status of a sampradāna listed in VP 3.7.129:⁶⁶ anirākaraṇa “not rejecting”, preraṇa “instigation”, and anumati “consent”. The idea is that someone can be termed “recipient” because the person does not reject, i.e., is indifferent to, the donation, instigates the giving of an object, or consents to receiving it. The three corresponding Tamil terms have all, to some degree, added semantic nuances that are not inherently in the Sanskrit, although I believe that Puttamittiraṉ tried to reflect some of the contextual meaning that comes out in the examples associated with the each type of sampradāna.

Of the three subtypes of kōḷi, the irappuk-kōḷi “requesting recipient” has the clearest link with its corresponding Sanskrit term, preraka “instigator” and may be considered as a simple translation. Examples for this subtype involve donating to Brahmans⁶⁷ and giving alms to beggars,⁶⁸ because such people first ask for what they receive. The kiṭappuk-kōḷi “circumstantial recipient” appears, at first blush, to not fit well with anirākaraṇa “non-rejection”. Neither the privative prefix a- nor the semantics of the Sanskrit original (“rejection”) are replicated in the Tamil. Nevertheless, the meaning of kiṭappu, a nomen actionis from the verbal root kiṭattal “to lie, sleep, be inactive”, actually reflects the gist of the anirākaraṇa rather well. A kiṭappu-recipient is simply there, not doing anything, not rejecting the offering, just standing around hence, “circumstantial”. These are recipients who do not need the given object and so have not compelled the donor to give it. In the examples we find deities and memorials as the circumstantial recipients of flowers, something they don't need or ask for.⁶⁹

The ārvak-kōḷi “affectionate recipient” is perhaps the furthest from its Sanskrit equivalent anumati, but the examples and explanation still reflect a similar conceptualization, even if expressed in different terms. Peruntēvaṉ gives two examples for this upakārakam: giving food to ascetics and giving a place to guests.⁷⁰ The PM has a similar example involving the proper gifts for a guest.⁷¹ Peruntēvaṉ then explains how ārvam “affection” is connected: aruntavar ārvattōṭu koṇṭamaiyāṉum karuttā ārvattōṭu koṭuttamaiyāṉum ārvak-kōḷiy āyiṟṟu. “Because the ascetics receive with affection and because the agent gives with affection, it has become the affectionate recipient.” In contradistinction to the other two types of recipient, the acts of both giving and receiving involving an ārvak-kōḷi are performed willingly by the participants. Although I have not found any strict equivalent to this passage in Sanskrit texts, Peruntēvaṉ has still captured the basic idea expressed in the PM, that the ānumantrika recipient occurs when two conditions are fulfilled: the giver is not spurred on by the recipient, i.e., the gift is willingly given, and the recipient actively accepts the given object, i.e., the gift is graciously accepted. As the PM shows,⁷² the ānumantrika recipient stands in contrast to the two other subtypes, the preraka recipient who instigates the act of giving and the anirākartrika recipient who does not actively accept the gift. Puttamittiraṉ's choice to reframe this type of giving through ārvam is intriguing and may be tied to the meritorious status accorded to donations in Buddhism, although it is certainly not restricted to Buddhism. I have also not been able to turn up any passages in the extant Buddhist literature where ārvam serves as a key term, but further research might offer more clues.

4 Conclusions

The kārakas and 23 upakārakams in the VC provide rich material for studying the transmission and extension of grammatical concepts in South Asia and beyond. To understand the evolution of their number and names in Tamil, one must begin with the A itself and follow the treatment of the kārakas in the hands of the latter commentators and grammarians both within and without the Pāṇinian tradition. Although one must have familiarity with the great works of Patañjali and Bhartṛhari, I have shown that the influence of the non-Pāṇinian schools, such as the CV and Kā. also played an important role in how the kārakas are defined and named in the VC. Furthermore, the little studied PM with the commentary of one Dharmakīrti undoubtedly served as the basis for both the number and names of the upakārakams in the VC and likewise the kāraka subtypes in SN. Given the additional translation of the PM with a commentary attributed to Dharmakīrti into Tibetan in the 14^th century, one can safely conclude the PM travelled in Buddhist circles (but not to the exclusion of others) and had a great impact on Buddhist grammarians when they wrote grammars for other languages. That a relatively simple and concise work, like the PM, should have replaced the complex arguments and theorization found in the MBh and VP will come as no surprise for those familiar with these texts; it was surely because of its brief but nevertheless complete presentation of the kārakas and their subvarieties that the work was chosen as the basis for transmitting this topic into other languages. Further research, especially into the manuscripts of the PM and VC will undoubtedly reveal a more complex picture than the one I have presented here relying primarily on the printed editions.

The VC also provides us with further material for how Sanskrit terminology was conceptualized and adopted into Tamil, if even by a small community that was destined to disappear. For the names of the upakārakams I have discussed several techniques of Tamilizing Sanskrit: simple borrowing (karumak-karuttā), borrowing with some modification (kāraṇak-karuttā), literal translation (talaimaik-karuttā), and various types of adaptations that express the basic concept of the Sanskrit original but where the Tamil lexemes have a different meaning (tāṉ teri karuttā, ārvak-kōḷi, etc.). The translation of the technical vocabulary is evidence for a need to naturalize the language of grammar for its Tamil readership, an urge that was lost by the time of the PV in the 17^th cent., a work in which much more Sanskrit terminology is simply borrowed with the necessary phonological changes.

Finally, I emphasize that I have only begun to scratch the surface of a potentially much larger project that would ideally involve a complete and systematic analysis of all 23 upakārakas, their counterparts in the both the PM and the SD, as well as the PV. Such work must, however, be founded on a better understanding of these texts' manuscript history and take into account the variability found therein. This task is reserved for a later point in time.

Abbreviations

A: Aṣṭādhyāyī

CV: Cāndravyākaraṇa

JV: Jainendravyākaraṇa

Kā.: Kātantravyākaraṇa

KV: Kāśikāvṛtti

MBh: Mahābhāṣya

PM: Prayogamukha

PV: Pirayōkavivēkam

SN: Saddanīti

T: Tolkāppiyam

VC: Vīracōḻiyam

VP: Vākyapadīya

Primary Sources

Secondary Sources