Candragomins Syntactic Rules Some Misconceptions

Author(s): Mahadev Deshpande
Publisher: Mahadev Deshpande
Catalog link: https://jainqq.org/explore/269495/1

Page #1
CANDRAGOMIN'S SYNTACTIC RULES, SOME MISCONCEPTIONS Madhav DESHPANDE The University of Michigan, Ann Arbor, Michigan 1. In their introduction to the translation of the Karakahnika section of Patanjali's Mahabhasya, S. D. Joshi and J. A. F. Roodbergen have presented a comparison of Panini's syntactic rules with those of Candragomin, the author of the Candra-Vyakarana. Useful as this introduction is to a student of Panini in particular and Sanskrit grammar in general, it contains certain misconceptions about Candragomin's syntactic rules and their operation. The purpose of this paper is to present Candragomin's syntactic conceptions as they are actually seen in his Candra-Vyakarana and his own commentary, Vrtti, on this grammar, and to set the record straight. 2., Joshi and Roodbergen rightly point out : "Panini has ... defined the syntactic meanings called apadana, etc. with the help of non-linguistic features, and that these features are indirectly connected with case-endings, namely, through the intermediary of the names, apa dana, etc. Candragomin, on the other hand, does not provide further definitions. He links syntactic meanings, and, in some instances, non-linguistic features directly with case-endings" (Karakahnika, Intro., p. xvi ). However, in regard to his Karaka section, they say: "In this section, ... the abhihita / anabhihita-device which ensures a uniform treatment of passive construction and the nominative endings is lost" (Karakahnika, Intro., p. xvi). As we shall see, this is a wrong statement concerning Candragomin's syntactic rules. They also claim that Candragomin leaves his entire application of Karaka rules subject to vivaksa "the wish of the speaker", instead of having principles such as the ekasamjna-rule (P. 1. 4. 1) in Panini's grammar. Candragomin's syntactic rules are actually by no means as loosely defined as they are made out to be by Joshi and Roodbergen. Below we shall go into some of the details of Candragomin's syntactic rules. 3. What is the abhihita / anabhihita principle in Panini ? Having defined various karakas such as kart, "agent", karman "object" etc., Panini proceeds to the assignment of various case endings. This particular section in Panini's grammar is headed by rule P. 2.3.1 (anabhihite): "If not expressed already )". This expression continues into the following rules. For instance, with the addition of anabhihite, P. 2. 3. 2. (karmani dviti ya ) comes to mean : " The second triplet of case endings is added to a nominal denoting the object, if the notion of objecthood has not already been

Page #2
expressed otherwise." Thus for instance, in the sentence ramah odanam pacati Rima cooks rice", the syntax of cases works as follows. The active voice affix -ti in the vero pacati denotes agent (karty) according to Panini (ref: P. 3. 4.69 ( karmani ca bhave cakarmakebhyah) and P. 1.3.3 Sesit kartari paramaipadam). Since the affix -ti denotes agent, the meaning "object" remains unexpressed. Thus by P.2.3.2 (karmani dvitiya) governed by the ansehihita "unexpressed" condition, we get the accusative case ending for the object odana. P. 2. 3. 46 (pratipadikartha.... prathama says, beside other matters, that the nominative case endings are added to a nominal merely to denote the meaning of the nominal itself, i. e. when no additional syntactic meaning such as "agent", "object" etc. needs to be or remains to be denoted. In the above given example, the affix - denotes the agent, and hence this particular meaning does not need to be nor remains to be denoted. Therefore, by P. 2. 3. 46, we get the residual nominative case for the agent rama. This briefly illustrates the functioning of the abhihita-anabihita "expressed / unexpressed" device in Panini's rules. 4. Joshi and Roodbergen claim that no such regulating device exists in the Candra-Vyakarana. As I shall show below, this is not true. The particular device exists in Candragomin's grammar as much as it exists in Panini's grammar. However, following the argument of Patanjali in the Mahabharya that such an explicit statement is not required and that one could simply get by with the generally accepted maxim uktarthanam aprayogah no linguistic item is used to denote something which is already otherwise denoted ", Candragomin does not state this principle in the form of an explicit rule. However, there is explicit evidence in his rules and in his commentary to show that this principle exists in his system. The maxim uktarthanam aprayogah is included in his Paribhasa-sutras.1 This maxim has the same regulating function which is accomplished by P.2.3.1 (anabhihite), except that this maxim has a much wider scope than the particular rule in Panini. Also see the Vitti on C. 1. 4. 50., Vol. I., p. 139. 5. Candragomin's rale C.2.1.93 (artha-matre prathama) says that the nominative case endings are added to a word when only its own meaning, and nothing more, is to be denoted. The implication of this rule is that if a syntactic meaning such as "agent" has not been otherwise denoted and thus remains to be and needs to be denoted, one cannot use the nominative case. This implication is explicitly supported by Candragomin's own statements on C. 2. 1. 43 (krivipye dvitiya). This rule says that the accusative case is used to denote that which is desired to be encompassed by the particular action. Candragomin's examples on this rule are: katam karoti (He) makes a mat", odanam pacati "(He) cooks rice", and adityam pasyati "(He) sees the sun ". In all these cases, the affix -ti, according to

Page #3
CANDRAGOMIN'S SYNTACTIC RULES 135 Candragomin's own conceptions, denotes "agent", and hence to denote the unexpressed object, one can have the accusative case endings for the words kata, odana and aditya. Candragomin, like Panini, uses the symbol L for verbal inflections in general, and says clearly in several places that certain verbal inflections, substitutes for the original L, denote "agent". (2) (1) yada tu kartari lakaras tada, akrta katah svayam eva/"When the verbal inflection (i. e. L) denotes "agent", we have the usage: akrta katah svayam eva The mat made itself'." Vrtti on C. 1. 1. 78, Vol. I., p. 34. atapta tapas tapasa iti kartary eva lakarah/" In the usage atapta tapas tapasah, the verbal ending (i. e. Z) denotes only" agent "." Vrtti on C. 1. 1.78, Vol. I., p. 34. (3) C. 1. 1.82 (kartari sap) says that the affix Sap is introduced after a verb root, if that verb root is followed by an active verbal inflection (tiN) or an affix marked with S denoting agent (kartari). This rule does not mean that, for Candragomin, Sap denotes agent. 6. There is a passage in Candragomin's Vrtti which absolutely shows that he does have the abhihita | anabhihita principle governing his case syntax. On C. 2. 1. 43 (kriyapye dviti ya), Candragomin gives a usage where the word denoting object takes the nominative case ending, because the verb is passive. Candragomin says: odanah pacyate ity odana-sabdad vyapyata na gamyate | kim tarhi ? tinantat | Vrtti on C. 2. 1. 43, Vol. I., p. 168. "In the usage odanah pacyate rice is cooked', the object-ness of rice is not understood from the word odanah. What then? (it is understood) from the word (pacyate) which ends in the verbal inflection (tiN, in this case -te)." Thus Candragomin very clearly uses the abhihita | anabhihita principle to determine assignment of case endings in his grammatical system. Here, in the above discussion, the point made by Candragomin is that the affix -te in the passive verb pacyate denotes the object, and hence by C. 2. 1.93 (artha-matre prathama) we get the nominative case ending for the word odanah to denote "merely the meaning of the nominal stem", and no other syntactic meaning. Thus Joshi and Roodbergen have missed the inner working of Candragomin's syntactic rules, which once understood as shown above, is not terribly different from Panini's system. 7. Joshi and Roodbergen say: "Candragomin provides for passive constructions by C.1.1.80, which introduces the suffix yak in the meanings

Page #4
ings in the same meanings. The trick is simply to replace the word karman in Panini's rules (P. 3. 1. 67 and P. 1. 3. 13 ) by a pya" (Karakahnika, Intro., p. xix, fn. 57). This statement which talks about Candragomin's " tricks" itself contains several problems of its own. Candragomin's rule C. 1. 4.47 (bhava pyayoh, tan-and from C.1.4.46 ) says that middle endings (tan) and the affix ina denote bhava "action" and apya "object". Rule C. 1.1.80 tin- Siti yag a-lid-asirlini), with the continuation of the word bhava pyayoh from C. 1.1.78, has been totally misunderstood by Joshi and Roodbergen. They claim that this rule prescribes the use of ya k to denote bhava "action" and apya " object". This simply is not the case. If yak were to denote these two meanings, then in forms such as pa c-ya-te, there would be two elements ya and te denoting exactly the same meaning. Could. this be a "trick" played by Candragomin? That does not seem to be the case. Let us take a closer look at C. 1. 1. 80. This rule is an exact equivalent of P. 3. 1. 67 (sarvadha tuke yak). The term sarvadha tuka is defined by Panini as referring to in affixes and affixes marked by S(ref : P 3. 4. 113) (tin-sit sarvadha tukam). P. 3.4.114 (ardhadha tukan sesah) says that rest of the affixes prescribed after verb roots are ardhadhatuk a affixes. But P. 3.4.115 ( lit ca) and P. 3.4.116 (lin 8 fisi) say that past perfect affixes (lit) and benedictive optative affixes (asir-lin), though they are substitutes for tiN affixes, are ardhadhatuka affixes, and are not sarvadha tuka. Thus if we look at the resulting scope of the Paninian term sarvadha tuka, it is identical with Candragomin's tin-sit a-lid-asirlin. Just as in the case of P. 3.1.67 (sarvadha tuke yak), there is an anuyrtti" continuation" of the words bhava-karmanoh from P. 3.1.66 (cin bhavakarmanoh ), in the same manner, in the case of c. 1.1.80 (tin-siti yag a-lid-asirlini ), there is the continuation of bhavapyayoh from C. 1.1.78. Thus despite the difference of terminology, Candragomin's rules are exactly identical with those of Panini. P. 3.1.67 (sarvadhatuke yak), says that the affix yaK is added after a verb root, if the verb root is followed by a sarvadhatuka affix denoting either bhava " action" or karman "object". This rule does not mean, " add the affix yaK after a verb root to denote bhiiva or karman, if that verb root is followed by a sarvadha tuka affix". The same is true of Candragomin's rules, and we cannot claim that, for him, the affix yaK denotes these meanings. 8. In fact, the abhihita / anabhihita principle plays a very significant role in Candragomin's grammar. In particular I shall discuss Candragomin's explanation of an infinitive usage. I have discussed the evolution of the syntactic theory in the works of the Sanskrit grammarians, particularly as it concerns the infinitive -tum, in my forthcoming monograph Syntax of the Sanskrit Infinitive. Here I shall only deal with a particular explanation in Candragomin's Vitti.

Page #5
CANDRAGOMIN'S SYNTACTIC RULES 137 samartho bhoktum ... iti bhavateh sarvatra sambhavat siddham ... uccaranam tu vaktur ayattam), Vrtti on C. 1. 3. 150, Vol. I., p. 130. "The usage samartho bhoktum is established because the verb "to be" (bhavati ) occurs everywhere. Its actual pronunciation depends on the speaker. " This simple looking statement requires the full use of the abhihita / anabhthita device. If c. 2.1.93 (artha-matre prathama ) were an uncontrolled rulo prescribing the use of the nominative case, without regard to the abhihita/ anabhihita principle, there would be no problem in this example. However, Candragomin rightly sees a problem, and provides a workable solution. What is the problem? In samartho gantum " ( He is able to go ", the syntactic meaning of the infinitive affix -tum is bhava "action" according to Candragomin's rule C. 1. 3. 6 ( tumun bhave kriyayai tadarthayam). Since the affix -tum denotes " action ", the meanings " agent" etc. remain unexpressed. The word samartha " capable" is an adjective and refers to the agent of gantum. Since the meaning " agent" has not been denoted by -tum, it would have to be denoted by the case ending to be added to the word samartha. In this case, we would have to have the instrumental case by C. 2.1.62 (kartari trtiya ), and we cannot get the desired nominative by C. 2. 1. 93 (artha-ma tre prathama ), since the meaning that needs to be denoted includes " agent", and hence it is not merely the meaning of the nominal stem (artha-matra ). Thus, we would get the undesired sentence *samarthena gantum, and cannot explain samartho gantum. Given this problem, Candragomin solves it by assuming a deleted finite verb bhavati "is " in the sentence, making the underlying sentence samartho gantum bhavati. This suggestion is based on Katyayana's varttika ( astir bhavanti parah prathama-purusah aprayujyamano'py asti, vt. 8 on P. 2.3.1 ). Once we accept this deleted verb bhavati, the syntactic problems are solved, since the affix -ti denotes agent, and hence for this otherwise denoted agent we can have a nominative case ending. 9. The above discussion not only tells us explicitly that Candragomin used the abhihita / anabhihita principle, but a deeper analysis of this explanation implies a syntactic principle, which is unknown to Katyayana and Patanjali, and appears explicitly only in the work of Candragomin's grand-disciple. BhartThari, for the first time. We may briefly discuss this question here. I have discussed this fully in my forthcoming monograph Syntax of the Sanskrit Infinitive. 10. How to deal with the case-syntax of an item which is simultaneously linked to two different actions? For example, in the sentence bhoktur

Page #6
138 MADHAV DESHPANDE samartho bhavati, the referent of the word samartha is simultaneously linked to both gantum and bhavati. While the affix -tum for Candragomin denotes bhava "action", the affix -ti denotes "agent". Thus the syntactic meaning 66 agent " is left unexpressed by -tum, but it is expressed by -ti: Thus it is simultaneously both abhihita "expressed" and anabhihita "un-expressed". Given such a situation, how should one deal with the syntax of samartha ? The principle used in such cases by Katyayana is that as long as a karaka is anabhihita "unexpressed" by one affix, the particular vibhakti "case ending " rule based on this factor will apply, and the fact that the same karaka is abhihita" expressed" by another affix at the same time does not matter. 2 This principle is accepted by Patanjali to be valid3, though it finally lands him into problems from which he has been unable to come out. Joshi and Roodbergen discuss this principle in their Anabhihitahnika volume, but do not evaluate it, nor point out that it ultimately fails. For instance, this principle cannot explain the syntax of gantum samartho bhavati. Since the meaning "agent" is left unexpressed by -tum, even though it is expressed by -ti, we would have to have the instrumental case for the word samartha by P. 2.3.18 (karty-karanayos trtiya) or C. 2.1.62 (kartari trtiya). This would result in the undesired sentence * samarthena gantum bhavati. However, Candragomin says that the assumption of the deleted verb bhavati solves the problems in this example. (Note Candragomin's words: bhavateh sarvatra sambhavat siddham, Vrtti, Vol. I., p. 130.) 11. What could be the syntactic explanation which would solve this problem? The best solution is to say that when one karaka is shared by two hierarchically related actions, it is the main action that determines the surface syntax of that shared karaka. This is precisely the principle enunciated by Bhartrhari in his Vakyapadi ya (III. 7. 81-2). I have attempted to show in my forthcoming Syntax of the Sanskrit Infinitive that this principle was implicitly present in Panini's own rules, but that it was unknown to both Katyayana and Patanjali. Where did Bhartrhari derive this principle from? It is possible that he thought of it himself, but it is perhaps quite likely that he learned it from his teacher Vasurata who was Candragomin's disciple. From what we know about the history of the transmission of the Mahabhasya, it appears that Candragomin revived the study of this work which had fallen into disuse." Since this particular principle of syntax was unknown to Patanjali, and since the study of his work was almost discontinued by the time of Candragomin, the circumstantial evidence may point to Candragomin as the source of this principle. Thus we must revise the negative evaluation of Candragomin's syntactic rules given by Joshi and Roodbergen, and should recognize Candragomin's contribution as possessing greater merit than has been accorded to it by its critics. 12. We may also briefly look at Joshi and Rodbergen's criticism of Candragomin's grammar concerning the fact that it does not contain rules

Page #7
CANDRAGOMIN'S SYNTACTIC RULES such as P. 1.4.1 (a kadarad eka samjna) and P. 1.4.2 (vipratisedhe param karyam). They remark, "Moreover, since Candragomin leaves the syntactic meanings undefined, and relates the use of case-endings to vivaksa: the wish of the speaker (to present items in language as he likes)', it relieves him of problems such as rule-ordering, conflict-procedure, and adjustments due to the definitions. In fact, part of the grammatical burden has been shifted to vivaksa" (Karakahnika, Intro., pp. xvi-xvii). In a footnote to this statement, they state: "One effect of the ekasamjna-rule (P. 1.4.1) in the Astadhyayi, as far as the karaka-designations are concerned, is that it checks the vivaksa, and that standard usage is imposed " (Ibid., p. xvii). This criticism is also unfair to Candragomin, and as we shall see, there are principles such as rule-ordering, conflict-procedure etc. in the Candra-Vyakarana, and that the notion of vivaksa " speaker's desire " does not play the kind of rampant role it has been ascribed by these two scholars. 13. In the first place, we must note very clearly that Candragomin does indeed have rule-ordering and a conflict-procedure in his grammar. Rule C. 1.1.16 (vipratisedhe), read along with Candragomin's Vrtti (dvayoh savakasayor ekatra prasange yat param tad bhavati) says that of two conflicting rules, which have otherwise non-conflicting independent domains, the latter rule prevails. In contrast with P. 1.4.2 (vipratisedhe param karyam), which is historically speaking a rule limited to some sections, Candragomin's rule is applicable throughout his grammar. Here Candragomin is clearly following Patanjali's interpretation of P. 1.4.2. Thus this rule applies to Candragomin's rules regarding case endings, compounds, as well as to other parts of his grammar10. Once we recognize this factor, it would appear necessary that there should be a specific pattern of rule-ordering in his grammar, without which this principle would not make sense. A close look at Candragomin's case ending rules shows this significant rule-ordering. Let us look at the order of some of the major rules of Candragomin. C. 2.1.43 (kriyapye dritiya) (kartari trtiya) C. 2.1.62 C. 2.1.63 (karane) 139 (sampradane caturthi ) (avadheh pancami) (saptamy adhare) (artha-matre prathama) C. 2.1.73 C. 2.1.81 C. 2.1.88 C. 2:1.93 C. 2.1.95 (sasthi sambandhe) Joshi and Roodbergen have quoted and translated these rules in this order Karakahnika, Intro., p. xvii) without realizing the function of the order

Page #8
140 MADHAV DESHPANDE of these rules. There is no ekasamjna rule in Candragomin to be sure, and yet it is the order of the above rules which takes care of the resulting problems. For instance, in the case of causative usages such as devadatto yajnadattan gramam gamayati " Devadatta makes Yajnadatta go to the village", according to Candragomin's statements, Yajnadatta is both the prayojya-kartr "agent, being instigated " as well as the vyapya "object" of the causative action (prayojaka vyapara ).1 This being the case, both C. 2.1.43 (kriya pye dviti ya ) and C. 2.1.62 (kartari tyti ya) would conflict in this case, since Yajnadatta is both kriy&pya and karts. Since there is a conflict, by C. 1.1.16 ( vipratisedhe ), the latter rule would have to apply and this would create the unacceptable sentence : * devadatto yainadattena gramai gamayati. To avoid this kind of instrumental case in the case of particular verbs, Candragomin makes the following exception rule : C. 2.1.44 (gati-bodhdhdra-sabdarthana pyana prayojye). This rule, as an exception rule, lays down that the accusative case is used after a nominal to denote the agent, being instigated, of verbs meaning going, knowing, eating etc. However, in a sentence such as devadatto yajnadattena odanam pa cayati " Devadatta makes Yajnadatta cook rice" Yajnadatta is both the agent of the action of cooking and the object of the causative action, and here, according to Candragomin's system, we get the desired instrumental case by C. 2.1.62 (kartari trtiya ), which being a later rule supersedes the earlier rule C. 2.1.43 (kriyapye dviti ya). Thus it is the rule-ordering that is crucial in Candragomin's system in enabling us in deriving the proper forms. This is done by Candragomin without taking recourse to ekasamina "only one designation at a time" as is done by Panini. 14. We will briefly contrast Panini's handling of such usages as given above. Panini's rules assigning karaka designations are headed by P. 1.4.1 (a kadarad eka sa injna ) and P. 1.4.2 (vipratisedhe param karyam). While the first rule says that in the following sections, an item can have only one designation at a time, the latter rule says that in the case of a conflict, the latter of the two conflicting rules prevails. Thus if two designations obtain for one item, then the designation obtained by a later rule will apply, and not the designation obtained by a previous rule. Given this framework, we can see what happens in the case of a causative sentence. Let us look at the example : devadattah yajnadattena odanam pa cavali "Devadatta makes Yajuadatta cook rice ". Here Yajnadatta is karty" agent" of the action of cooking by P. 1.4.54 (svatantrah karta ), and is the karman "object" of the action of instigating denoted by the causative affix by P. 1.4.49 (kartur ipsitatamam karma ). However, both of these designations cannot be operative simultaneously due to P. 1.4.1., and hence by P. 1.4.2. the latter designation, i. e. kartr "agent" is retained. Since the affix -ti in pacayair denotes the agent of the causative action, the agent of the instigated action

Page #9
141 CANDRAGOMIN'S SYNTACTIC RULES remains unexpressed (anabhihita). Therefore, by P. 2. 3. 18 (karty-karanayos trtiya) we get the desired instrumental case for the unexpressed agent Yajnadatta. In the example devadattah yajnadattam gramam gamayati" Devadatta makes Yajnadatta go to the village", the same procedure applies first, and we obtain the designation karty "agent" for Yajnadatta, which in turn would lead to an instrumental case ending for the word yajnadatta. However, in this instance the correct usage requires an accusative affix for the word yajnadatta. This is achieved by Panini by rule P. 1. 4. 52 (gati-buddhipratyavasanartha-sabda-karma karmakanam ani karta sa nau, karma). This exception rule says: "That which is the agent of verbs of going etc. in the pre-causative (ani) becomes the object (karman) in the causative (nau)." Thus, in the case of the particular verbs, this exception rule in effect reverses the operation of P. 1. 4. 2 (vipratisedhe param karyam). Since Yajnadatta now holds the only designation, i. e. "object", we get the accusative case ending for the, word yajnadatta by P. 2.3.2 (karmani dvitiya) to denote its otherwise unexpressed objecthood. The above comparison of Panini's procedure with that of Candragomin makes it quite clear that Candragomin does have the principle of ruleordering and devices to resolve conflicts as much as Papini. However, as shown above, Panini takes care of the conflicting possibilities at the stage of applying the karaka designations, by means of P. 1. 4. 1. and P. 1. 4. 2 (i.e. ekasamjnadhikara and vipratisedha). Candragomin, on the other hand, has no independent karaka designations, and hence he takes care of resolving conflicting possibilities at the only stage available to him, i. e. his vibhakti "case ending" rules. Thus, rule-ordering is more significant for Panini's rules assigning the karaka designations, while it becomes more significant for Candragomin's vibhakti-rules. 15. Thus we must clearly reject the notion upheld by Joshi and Roodbergen that Candragomin's syntactic rules are mainly guided by vivaksa "speaker's desire", and that there are no internal mechanisms such as the abhihita/anabhihita device, rule-ordering and procedures to resolve conflicts. Candragomin has all the above features, though some of these operate somewhat differently from Panini's procedures. Candragomin does use the concept of vivaksa "speaker's desire" where there are, in his view, true options. Here his interpretation of options is often different from that of Panini, but that would simply mean that the particular usages mean something different to Candragomin than what they possibly meant to Panini. For instance, P. 2. 3. 7 (saptami pancamyau karaka-madhye) says that when there is an item denoting time or space which occurs between two actions of the same karaka, that word may take either a locative or an ablative case ending. For instance: adya bhuktva devadatto dvyahe | dvyahad va bhokta "Having eaten to day, Devadatta will eat after two days "12 What this rule implies is that despite the difference of case endings in dvyahe |

Page #10
CANDRAGOMIN'S SYNTACTIC RULES 143 that a verbal inflection may express agent, object or action (ref : P. 3.4.02 who is to decide whether the grammar should produce ramo gacchari o ramena gamyate? The fact is that in Panini's grammar certain cperatives are involuntary and do not require any input from the "user", while there at a great deal of operations which tacitly require the voluntary decision are choice by the "user". For instance, one can either say agnir voksn "Fire burns the tree" or aham agnind vrksam daha mi "I burn the tree with fire". In the first sentence, agni "Fire" is the agent, while in the second sertence it is the instrument. There is nothing in Panini's grammar, like the devise or "raising" in modern generative grammar, which would provide us ery systematic linking of these two sentences. The only explanation is rirak "speaker's desire". Patanjali uses the notion of vivaksd in this sense, Candragomin's usage is in large measure derived from that of Patanja u Thus the well-known grammatical maxim vivaksatah karakani bhararti "the karakas are dependent on the speaker's desire" is not meant to replace tre automatic operations in Panini's grammar, but to indicate the source of input for the voluntary operations. The use of vivaksa "speaker's desire" in Cando gomin's grammar is not different from its use in Panini's grammar in isciple. but only in details. It is not a stylistic concept, but a grammatical corcept referring to necessary voluntary input without which the involuntat operations of a grammar cannot function. A great deal of this rivakse, thongs not all of it, can be formalized in terms of pragmatic constraints and dis course structure and strategies, but the Sanskrit grammarians relezzied susta considerations to the concept of vivaksd. In doing this, Panini and Canzgomin do not differ substantially from each other. BIBLIOGRAPHY Candra-Vyakarana, with the Svopaj na-Vitti, by Candragomin, two vo's, e kc. Chatterji, Doccan College, Poona, 1953, 1961. Cardona, George (1978) Still Again on the History of the Mahabharys, Die Jac Volume, Annals of the Bhandarkar Oriental Research Institute, Poonais. 2. Deshpande. Madhav (Forthcoming-A) A monograph: Syntax of the Soz: S . Evolution of Syntactic Theories in Sanskrit Grammar. - (Forthcoming-B) Review of Vya karana-Mahabhasya, Ancb i ca Introduction, Text, Translation and Notes, by S. D. Joshi 2 J.AF. Reaches University of Poona, Poona, 1976. (appearing in Orientalistische Limeira (Forthcoming-C) Review of Vyakarana-Mahabhasya, Baia-Dandia, Text, Translation and Notes, by J. A. F. Roodbergen, University of Ponen 194 (appearing in Orientalistische Literaturzeitung) Mahabha sya, by Patanjali, with the commentaries Pradipa by Karyata a 40 Nagesabhafta, three vols., published by Motilal Banarasidass, Dei. 156. Vakyapadjaya, by Bhartshari, edited by K. V. Abhyankar and V. P. Leat. Gre of Poona Sanskrit and Prakrit Series, Volume II, Poona, 1965.

Page #11
144 MADHAV DESHPANDE Vyakarana-Mohabhasya, Karakahnika, (P.1.4.23-1.4.55), with Introduction, Translation and Notes, by S. D. Joshi and J. A. F. Roodbergen, Publications of the Centre of Ad. vanced Study in Sankrsit, Class C, No. 10, University of Poona, Poona, 1975. (Referred to : Karakalnika) Karano-Mahabha sya, Anabhihira hnika (P.2.3.1. - 2.3.17), with Introduction, Text, Translation and Notes, by S. D. Joshi, and J.A.F. Roodborgen, Publications of the Cerere of Advanced Study in Sanskrit, Class C, No. 11, University of Poona, Poona, 1975. (Referred to as: Anabhihitahnika) NOTES Mahabhasya on P.2.3.1., Vol.I., Pt. II., pp.476-482. Also see: Anabhihitahnika, Intro., pp. XXXV-xxxvi. Pataujali discusses two possibilities. The caso endings may be said to depote either (A) karakas, or (B) number. Joshi and Roodbergen rightly say: "In view (A), P.2.3.1. is not required, because we can manage by the principle uktarthanam aprayogah", Anabhihitahnika, Intro., P. xxxvi. Candragomin's Vrati on C.2.1.1. (Vol. L, p.161) says: ete ca (sva dayah) sa marthyad ekatva dimad-artha-va cinah tabdat pare bhavanti. This means to say that case endings (singular etc. in each triplet) occur after words which denote entities possessed of singularity etc. Thus, it appears that, according to Candragomin, the case ending itself does not denote number, but that it denotes the syntactic meanings such as agent and object. This is also clear from the Vrtti on C.2.1.87. This provides us the reason why Candragomin did not explicitly state a rule such as P.2.3.1. (anabhihite). The maxim uktartha nam aprayogah is included in Candragomin's Paribhasa sutras, see: Candra-Vyakarana, Vol. II., P. 396. This is the most direct evidence against the criticisms levelled by Joshi and Roodbergen. Also: Paribhasasarngraha, ed. by K.V. Abhyankar, Bhandarkar Oriental Research Institute, Poona, 1968, p. 47. 2. Varttikas 56 on P.2.3.1.. (dvayoh kriyayoh karake anyatarena bhihite vibhakty ablano-prasarigah, na va anyatarena nabhidhanat, anabhihite hi vidhanam). Also see: Anabhikita krika, Intro., pp. xxxvii and 37ff. 3. Mahabhasya, Vol. I, pt. II. pp. 483-4; Vol. II., pp. 251-253. 4. Mahabha spa on P.3.4.26, Vol. II., pp. 251-3. Also see: Madhav Deshpande (Forth coming-A). 5. Anabhihitahnika, Intro., pp. xxxvii-xxxviii, and pp. 37ff. Also see: Madhav Deshpande (Forthcoming-B). Punyaraja's cmmentary on Vakyapadiya, II. 484. (nya ya-prasthana ... etc.) I must note here that my inferred principle is not the only possible way of explaining Candragomin's handling of the sentence gantur samartho bhavati. There are a few other possible caxplanations, though not necessarily better ones. I can think of two possible alternatives. [A] Unliko Katyayana's assumption, ono may assume that if a koraka related to two actions is at the samo time expressed (abhihita, ukta) by one affix and not expressed (anabhihita, anukta) by another affix, it would be considered expresseul

Page #12
CANDRAGOMIN'S SYNTACTIC RULES 145 (abhihita, ukta). This would allow the use of the nominative case ending for the Word samartha. [B] One may take recourse to the notion of vipratisedha as defined by C.1.1.16. Since gantum does not express agent, we obtain the instrumental case by C.2.1.62 (kartari trtiya). However, the affix-ti in bhavati denotes agent, and hence we obtain the nominative case by C.2.1.93 (artha-ma tre prathama). Since C.2.1.93 is the later (para) rule, it prevails, and this would allow the nominative case for the word samartha. Easier and attractive as these alternatives may seem, they create problems, or rather do not solve problems in many other complex sentences. For instance, both of these alternatives fail to explain the syntax of the word devadattam in a sentence such as: devadattar gacchantam pasyati yajnadattah, while the principle suggested by me above explains its accusative case. It will be unfair to Candragomin to ascribe to him an assumption which causes more problems, particularly when he explicitly says that the assumption of the form bhavati solves problems in samartho gantum. 7. Vakyapadiya, II. 479-483; George Cardona (1978). . 8. Historically speaking, P.1.4.2. extends only up to P.2.3.38 (kadarah karmadharaye). For a historical discussion, see: George Cardona, "Some Principles of Papioi's Grammar", Journal of Indian Philosophy, Vol. I, 1970, pp. 40-74. Mahabhasya on P.1.4.2., Vol. I., Pt. II., pp. 204ff. Once P. 1.4.2. is extended to cover the whole of the grammar, occasionally the previous rule needs to apply in preference to a later rule. To get over this problem, Patanjali proposes to interpret the word para in this rule to mean ista "desirable". Thus the rule is, occasionally, interpreted to mean: Of two conflicting rules, the desired rule supersedes the other. Mahabha $ya, Vol. I., pt. I., p. 123, and pt. II., p. 207. Candragomin also accepts this interpretation of the word para. See the Vrtti on C.1.1.16, Vol. I, p. 13. 10. Vrtti on C.1.1.16, Vol. I., p.13. There are indeed a number of paribha $as "maxims" in Candragomin's system dealing with various conflict-procedures. See: CandraVyakarana, Vol. II, pp. 396-398, and Paribhasa sarngraha, ed. by K. V. Abhyankar, Bhandarkar Oriental Research Institute, Poona, 1968, pp. 47-48. 11. Vrtti on C.2.1.44., Vol. I., p. 169. 12. For these examples and various interpretations, see: Mahabhd sya, Vol. I, pt. II., p. 492. 13. Vrtti on C.2.1.51., Vol. I., p. 172. 14. Vrtti on C.1.3.150., Vol. I, p. 130. 15. Mahabh8sya, Vol. I., pt. II, pp. 241 and 249. [Received 26 February 1979]

Page #13
142 MADHAV DESHPANDE dvyahid there was no semantic or syntactic difference according to Panini. This reflects Panini's interpretation of such usages. According to Candragomin's own perception and interpretation, these usages meant different things. The use of the form dvyahe meant that the particular time was the location (adhara), while dvyahad meant something like " since two days ".13 Panini and Caudragomin differ on the meaning of various infinitive constructions. For instance, a construction such as ramah bhoktum odanam pacati " Rama cooks rice to eat ( it)" is derived by P. 3. 3. 10 ( tumun-nvulau kriyayam krijartha yam) which says that the affixes tumUN and Nvul may be used after a verb root, if the action denoted by that verb root is the future purpose of another action denoted by a verb used in the same construction. However, Panini feels that constructions like ramah gantum saknoti " Rama is able to go" do not express the same kind of semantic relationship between two actions, and hence he makes a separate rule to derive these kinds of constructions, i. e. P. 3. 4. 65 ( saka-dhrsa .... tumun). However, Candragomin's semantic perception is different from that of Panini and he explicitly says that both the above types of usages have one action for the sake of another action. Thus the notion of vivaksa "speaker's desire" 25 used by Candragomin has indeed a more substantive significance, than what is recognized by Joshi and Roodbergen. 16. Explaining their conception of vivaksa "speaker's desire", Joshi and Roodbergen say: "We could say that by making vivaksa responsible for che variety of syntactic construction Candragomin has introduced the bol CONSULT THE SPEAKER in the program of his machine. That is to say, the mechanical progress of the machine is interfered with every now 23d then. Panini, on the other hand, by means of his definitions has specified wonditions under which an item is supposed to be vyapya, or adhara, or sambandha in Candragomin's terminology. Therefore his machine, being rovided with a more explicit program, works better. Or, to put it differently, rocksa is not a grammatical concept, but a stylistic one. It merely says that ju of a number of modes of expression the speaker may select any particular 72" (Karakahnika, Intro., p. xviii). I find it difficult to agree with this formu tion. The analogy of Panini's grammar to a machine, derived from early wrceptions of Noam Chomsky, has been carried too far by Joshi and Randbergen. In my review of Roodbergen's Bahuvrihi-Dvandvaehnika (borearing in Orientalistische Literaturzeitung), I have pointed out to what at this analogy is misleading. Here I shall only deal with the concept avvaks&. In the first place, as I have shown, vivakso "speaker's desire" is not a selance principle in Candragomid's system. Secondly, even Panini's grammar issupposes a certain concept of vivaks& "speaker's desire", and it is not an tematon which can produce a text on its own. Who should decide whether je grammar should produce gantum icchati or jigamisati? Given the rule