Bhaaviveka and the early Maadhyamika theories of language

By Malcolm D. Eckel
Philosophy East and West
28:3 July 1978 p. 323-337


p. 323 In the last fifty years, Western interpreters of Maadhyamika Buddhist philosophy have worked diligently to devise a philosophical vocabulary in which the insights and techniques of the Maadhyamika, dialecticians can be accurately and intelligibly expressed to Western readers.(1) This is not an easy task, but it has sometimes been done quite effectively.(2) Even in the most successful studies, however, one element is often conspicuously lacking. Scholars have compared the work of early Maadhyamika philosophers with similar work in the West, but they have been reluctant for various reasons to compare the early Maadhyamika philosophers with each other.(3) This, of course, has led to a certain admirable simplicity in the results of their comparison, but it has sacrificed a degree of sophistication and philosophical accuracy that would enrich their results. In this article I would like to redress the balance in one small area by considering the development of the theory or theories of language in the works of Naagaarjuna, Bhaavaviveka, and Candrakiirti. By so doing, I hope to demonstrate that a sure way to promote conceptual accuracy in the comparative enterprise is to understand how individual philosophers in the Maadhyamika tradition chose to develop and differentiate themselves from the work of their predecessors. The comparison of early Maadhyamika philosophers with each other has been hindered in recent years by the relative scarcity of major texts translated into Western languages. We are fortunate to have translations of the basic works of Naagaarjuna and Candrakiirti, but we only have fragments of the works of other authors like Buddhapaalita and Bhaavaviveka, and original Tibetan works on Maadhyamika philosophy are almost unknown in Western languages.(4) This imbalance has led, perhaps inevitably, to the notion that early Maadhyamika philosophy was considerably more homogeneous than it actually was. What is necessary now to expand our understanding of this school is greater familiarity with the lesser known authors, like Bhaavaviveka, and with the great Tibetan scholars like Tso^n-kha-pa, who wrestled in their own works with the diversity of the early philosophy. Such familiarity would show that the homogeneity of the early tradition is merely apparent. In fact, Bhaavaviveka distinguished himself quite sharply from the earlier tradition on certain points, and Candrakiirti, in turn, distinguished himself from Bhaavaviveka. Tso^n-kha-pa and his successors recognized this and, in their own efforts to harmonize the differences, gave a very useful account of the ways in which the two disagreed. Theories of language play an important part in the Maadhyamika philosophy of the early period, not primarily because the individual philosophers were interested in constructing a positive semantic theory, although that interest did impinge somewhat on the works of Bhaavaviveka and Candrakiirti, but ---------------------- Malcolm D. Eckel is Instructor, Religion Department, Ohio Wesleyan University, Delaware, Ohio. Philosophy East and West 28, no. 3, July 1978. reserved. p. 334 because the disputes between Maadhyamika and rival Indian schools were often, at bottom, cast in terms of disagreements over the use of language. This is perhaps a natural consequence of the Maadhyamika critical method. Maadhyamika philosophers were interested more in devising a critical scheme for removing their opponents' misconceptions than they were in building their own positive theory. In the absence of shared metaphysical assumptions, their criticism often took the form of objections to certain uses of language. The Maadhyamika account of language is thus useful, in the first case, as a mirror of the relationship between Maadhyamika philosophers and their Indian opponents, but its importance is not limited just to this. The account of language is also closely related to a central Maadhyamika notion, the two levels of truth. Maadhyamika philosophers recognized this as a distinction between a level of nonconceptual, ultimate truth (paramaartha) and a level of truth that lay within the domain of concepts and words (vyavahaara). The two were distinct, but the second was understood to function in some way as a vehicle for the first.(5) The account of language given by any particular Maadhyamika philosopher necessarily affected his notion both of the exact nature of the distinction between the two truths and of the way one served indirectly to express the other. An understanding of the development of Maadhyamika accounts of language is thus useful to us in a number of ways, both in describing the Maadhyamika response to other Indian schools and in following internal differences on certain fundamental points. It also has the advantage, as will be evident later, of sharply delineating basic differences between Bhaavaviveka and Candrakiirti. Naagaarjuna laid the groundwork for later Maadhyamika accounts of language in the second century A.D., at a time when Indian philosophers were becoming conscious in a rudimentary way of the need to formulate rules for debate between opposing philosophical schools. Sanskrit was being used increasingly as a tool for learned discourse, and the Hindu logicians were attempting to develop a theory of semantics and syllogistic reasoning on which philosophical argument could be based. Naagaarjuna's position in this philosophical environment was necessarily rather ambiguous. He was committed, as a Sanskrit dialectician, to the process of discussion and debate facilitated by the developments in Hindu logic, but he could not, as a Buddhist, accept the ontology on which the theories were based. In particular, he could not accept the notion advanced by the Hindu logicians that the meaning of a term was the substantial entity to which it referred. He appeared, in fact, to assert exactly the opposite, namely, that all things are empty of substantial reality, and terms which refer to such things are equally empty of reality, since there is no real substance to which they refer. In the terminology of the Sanskrit philosophical schools this was expressed in the following words: "Nothing at all possesses intrinsic nature" (sarve.saa.m bhaavaanaa.m sarvatra na vidyate svabhaava.h).(6) For Naagaarjuna this quasi-assertion posed a basic problem concerning the function of language. The problem is simply this: if all things are empty (`suunya) of intrinsic nature p. 325 (svabhaava), then terms that refer to them are similarly empty. In the semantic theory of the Hindu logicians, an empty term, that is, one with no reference, is meaningless. So to say "all things are empty of intrinsic nature" is to say that all terms are meaningless, including those in the assertion itself. The assertion is thus useless as a means of argument. Naagaarjuna formulated this problem for himself in the first part of the Vigrahavyaavartani. Here an objector says: [Vs. 1] If nothing at all possesses an intrinsic nature, then your statement [that nothing possesses an intrinsic nature] itself possesses no intrinsic nature, and it cannot refute intrinsic nature.(7) [Vs. 9] If there is no intrinsic nature. then even the word "no intrinsic nature" (ni.hsvabhaava) is impossible, because there can be no word without an object [to which it refers] (naama hi nirvastuka.m naasti).(8) In the second part of the work, Naagaarjuna formulates his reply largely in terms of examples. [Commentary on verse 223 You have not understood the emptiness of things. ...lf things existed by virtue of their own intrinsic nature. they would exist even without causes and conditions. But they do not. Therefore they have no intrinsic nature, and they are called empty. Similarly, because it is dependently produced, my statement has no intrinsic nature, and because it has no intrinsic nature, it is reasonable to call it empty. Now, things like a cart, a pot, or a cloth, though they are empty of intrinsic nature because they are dependently produced, serve their various functions. For example, they carry wood, grass, or dirt, they contain honey, water, or milk, or they protect from cold, wind, or heat. Similarly, my statement serves to establish the fact that things have no intrinsic nature, even though, because it is dependently produced, it has no intrinsic nature.(9) Naagaarjuna shows a number of the important characteristics of his method in these passages. The first point to note is that he works out his own account of words and their function primarily in response to the challenge of a Hindu logician, who wants to force him to say more than he is willing to say. The response he gives is largely negative. He refused to be pushed by the logician into admitting that either his words or the things to which they refer exist by virtue of their intrinsic nature (svabhaava). The second point has to do with the way words actually function, even though they are empty of intrinsic nature. In fact, Naagaarjuna does not present a positive theory of language to account for the effectiveness of his sentence: he simply makes an appeal to conventional usage, His words admittedly have no intrinsic nature, but they work conventionally as well as does a cart. The cart, when we examine it, has no nature which we can designate as its "cartness," but it still manages to carry out its function effectively. We cannot actually say that Naagaarjuna presents a coherent theory in these lines. In his appeal to ordinary usage, however, Naagaarjuna suggests the direction in which some future Maadhyamika philosopher might go in developing a theory based on pure convention. p. 326 If this were all Naagaarjuna had to say about his philosophical statements, our problem would be greatly simplified; but Naagaarjuna recognized that the statement "All things are empty of intrinsic nature" contains an added element of complexity. It purports to convey a general truth about the nature of things: all entities, without exception, are empty of intrinsic nature. If the statement functions this way, however, it raises a number of new difficulties. We might ask, in particular, whether the truth conveyed in this statement has, in Naagaarjuna's terminology, an intrinsic nature. If it does, it renders the statement itself false. If it does not, it is not clear what the statement is meant to convey or how it is meant to convey it. Naagaarjuna actually poses this question for himself in a somewhat different form. He asks whether the statement "All things are empty of intrinsic nature" asserts anything, and if not, what it is understood to do. His explanation is the following: [Vs. 29] If I made any assertion (pratij~naa), I would be in error. But I make no assertion, thus I am not in error, [Commentary] If I made any assertion, then the error you describe would be mine. But I make no assertion. How can there be any assertion when all things are empty, completely at peace and isolated by nature.(10) [Vs. 63] I do not negate anything, nor is there anything to be negated. Therefore you slander me when you say that I negate something. [Commentary] If I negated something, what you say would be correct. But I do not negate anything at all, for there is nothing to be negated. Therefore, when all things are empty and there are no negation and thing to be negated, your statement is slanderous. [Vs. 64] You may say that something that does not exist can be negated without words. But in this case [in our statement] speech simply makes known that it does not exist; it does not negate it. [Commentary] You may say, "Something that does not exist can be negated without words; then what point is there in your statement that all things lack intrinsic nature?" We reply that our statement that all things lack intrinsic nature does not cause all things to have no intrinsic nature; it simply makes known that things lack intrinsic nature. For example, when Devadatta is not in the house, someone might say, "Devadatta is in the house." Someone else might then say to him, "He is not." That statement does not create Devadatta's absence in the house, but only makes known his absence in the house. Similarly, the statement, "Things have no intrinsic nature." does not create the absence of intrinsic nature; it only makes known the absence of intrinsic nature.(11) Naagaarjuna makes it quite clear here that his statement should not be understood either as an assertion (pratij~naa) or negation of any positive entity. When pressed to give a positive account of the function of his words, he again appeals to a conventional example to show that, while they do not assert anything, they still have significant effect. The account of the function of language presented in these passages is, of course only a small part of Naagaarjuna's philosophy, but it can serve to call attention to some of the basic features of his method. First, we have noted that he proceeds only in response to claims made by his opponents, and he p. 327 refuses to be drawn by their arguments into making positive assertions. In particular, he refuses to accept the notion that the statement "All things are empty of intrinsic nature" functions as an assertion of any positive entity. Second, on the positive side, he argues from conventional usage that the refusal to accept either the intrinsic nature or the assertive value of the statement in no way impairs its ability to function effectively. His words do not assert anything, but they do make known the absence of intrinsic nature. In this way, Naagaarjuna drew the outline of a Maadhyamika account of the function of language. We will see that his successors had great difficulty staying within its limits. In the four centuries that intervened between Naagaarjuna (circa 150 A.D.) and the next Maadhyamika philosopher we will consider, Bhaavaviveka (500-570 A.D.) , Indian philosophy underwent a remarkable expansion. The basic texts of the Hindu schools were settled and provided with commentaries, and the earlier dominance of Maadhyamika among the Mahayana Buddhist schools came to be challenged by a school of Buddhist idealists and logicians. In the face of this widening doctrinal diversity, Bhaavaviveka seems to have been by temperament and training particularly prey to the attraction of other philosophical opinions. He was apparently a brahman and retained a fondness for the diversity of brahmanical learning, from alchemy and palmistry to Advaita Vedaanta, long after his conversion to Buddhism and the philosophical method of Naagaarjuna.(12) A basic motivating impulse in his philosophy, in fact, seems to be the need he felt to reestablish Maadhyamika philosophy in a form that would allow room for the variety of conventional learning. Apart from matters of temperament, however, there were good logical reasons to reassess Naagaarjuna's handling of some basic questions. The rules of logical debate recorded in the Nyaaya-suutras seem to have evolved after Naagaarjuna and partly in response to his methods. This seems particularly evident in the definition of an unacceptable form of reasoning known as vita.n.daa or "cavilling." Nyaaya-suutra 1.2.3 defines this as "that [sophistry (jaati) ] which lacks the establishment of a counter-position."(13) Naagaarjuna, as we have seen, avoided making a positive assertion of anything and did not seem to be concerned that this would violate the rules of debate. Bhaavaviveka, on the other hand, makes a conscious effort in at least two places in his work to meet the objection that he is guilty of vita.n.daa and show that it does not apply.(14) In chapter 3 of the Tarkajvaalaa, for instance, he raises it as an objection. [Objection:] Because you do not establish your own position (svapak.sa) , but only refute your opponent's position (parapak.sa), are you not guilty of vita.n.daa? [Reply:] Our position is "emptiness of intrinsic nature" (svabhaava-`suunyataa); since this is the nature of things, we are not guilty of vita.n.daa.(15) Bhaavaviveka manages to deal with the objection but only at serious cost to the integrity of Naagaarjuna's method. He is now willing to admit something Naagarjuna fought hard to resist: he accepts "emptiness of intrinsic nature" as p. 328 a positive philosophical assertion. This change has formidable significance for the development of the Maadhyamika accounts of language. Bhaavaviveka's reasons for making this move deserve careful scrutiny. Before we consider Bhaavaviveka's reasons, however, we need to look again at another aspect of Naagaarjuna's argument. We saw earlier that Naagaarjuna took some pains to account for the way the words of the sentence "All things are empty of intrinsic nature" could express a significant truth about the nature of things, even though the words themselves were empty and the truth was not the object of an assertion. Those familiar with Maadhyamika philosophy will recognize this distinction as the verbal form of the distinction between two levels of truth. Naagaarjuna says more about this distinction in his commentary on verse 28 of the Vigrahavyaavartani. We do not say, "All things are empty," without resorting to conventional truth (vyavahaara-satya) or by rejecting conventional truth. For it is impossible to teach the Dharma without recourse to conventional truth. As we said [in the Maadhyamakakaarikaas]: "Ultimate truth (paramaartha) cannot be taught without resorting to conventional expressions (vyavahaara) ; nirvana cannot be reached without recourse to ultimate truth."(16) The distinction between ultimate and conventional truth has many implications for Naagaarjuna, particularly in the realm of practical behavior (as might be inferred from the conventional orientation of a work like Naagaarjuna's Ratnaavali).(17) But in the philosophical works, like his Vigrahavyaavartani, Naagaarjuna develops the distinction in only a limited way. We might understand ultimate and conventional truth here simply as two sides of the same verbal strategy. Ultimate truth (paramartha) might be understood as that which the statement "All things are empty," acting as a verbal expedient, is meant to convey; but we must remember, of course, that Naagaarjuna resisted any formulation that would turn ultimate truth into the object of a positive assertion. This account of the relation between ultimate and conventional truth is simple and seems to stay close to Naagaarjuna's didactic intent, which was to call attention to the inadequacies and misconceptions hidden in conventional expressions and use them as a vehicle to realize the emptiness of things. Bhaavaviveka took a somewhat more complicated view of the matter. As we have seen, Bhaavaviveka was inclined temperamentally to include a large variety of contemporary views and practices into `his Maadhyamika system; he also wanted to make room for the possibility of positive philosophical assertions. Both these goals would have been hard to realize if he had closely followed Naagaarjuna's method, with its concentration on conventional truth only as a vehicle for the expression of ultimate truth. Bhaavaviveka needed a new form of interpretation, and he found it in a new grammatical analysis of the term paramaartha, "ultimate truth." He interpreted paramaartha not as ultimate truth itself but as knowledge of ultimate truth.(18) In is way he was able to change paramaartha from the content of teaching, which Naagaarjuna p. 329 discussed purely in linguistic terms, to a realm of experience that could be severed from vyavahaara, conventional truth. Paramaartha and vyavahaara could thus be separated into two realms of existence, each of which had practices and doctrines appropriate only to it. The two were still connected, but less in the lingustic way that Naagaarjuna outlined in the Vigrahavyaavartanii than in a temporal and causal way, representing a slow progression from one level to another along the stages of the bodhisattva path. Bhaavaviveka explains this process in the third chapter of the Tarkajvaalaa. [Vss. 10-11] Ultimate wisdom effects the complete negation of the network of conceptual thought and is motionless moving in the clear sky of ultimate truth (tattva), which is peaceful, directly experienced, without concepts or letters, and free from unity and diversity. [Vss. 12-13] It is impossible to mount the pinnacle of the palace of truth without the ladder of conventional truth. For this reason, the mind, isolated in conventional truth, should become clear about the particular and general characteristics of things.(19) Bhaavaviveka did not consider the process of climbing through conventional truth to be either easy or quick, as he says in his commentary on these verses. "It is impossible to climb this palace suddenly. For without ascending the ladder of conventional knowledge for seven countless eons, the completion of the perfections, powers, and super-knowledges is impossible."(20) Progress along the path could be quite leisurely and there was much time along the way to enjoy the subtleties of the conventional world.(21) By separating paramaartha and vyavahaara into two different realms of experience in this way, linked only by a gradual progress along the path to perfection, Bhaavaviveka succeeded in the first part of his program. He created a realm of experience in which he could concentrate on the subjects of the conventional world that caught his interest, without having to worry at every moment about applying Naagaarjuna's critique. That part of the Maadhyamika method belonged to the realm of ultimate truth and could be postponed indefinitely while one considered problems in the mundane realm. Bhaavaviveka's fascination with this realm had important consequences for historians of Indian philosophy; his diligence in collecting the details of other philosophical systems provided important evidence for the development of some of the Indian schools.(22) But what was the cost of this rehabilitation of conventional truth? Naagaarjuna achieved great power and simplicity in his philosophical method by treating every question rigorously, as if it were an ultimate question. In doing this he showed that there were no areas of existence that were not subject to the corrosive effect of his critique. By separating a particular realm in which this critique, for practical purposes, did not apply, Bhaavaviveka appears to have damaged the unity of Naagaarjuna's method and engaged in a subtle absolutizing process in which conventional truths are again established in their own right, The full consequences of this process will not be seen until p. 330 we consider the efforts by Tso^n-kha-pa and Candrakiirti to explore its im- plications, but at least one problem will be apparent when we examine Bh~vavi-veka's treatment of his second philosophical concern, the fashioning of positive philosophical assertions. We saw earlier that Bhaavaviveka was troubled by the accusation that he, as a Maadhyamika philosopher, was guilty of vita.n.daa. Now that we have seen Bhaavaviveka's method for separating the two levels of truth, we are in a position to examine the justification for his peculiar response to this charge. As we saw, Bhaavaviveka responded by saying that he, in fact, did maintain a positive position of his own, namely, the emptiness of all things. In the scheme of Bhaavaviveka's separation of the two levels of truth, this claim would be quite reasonable if confined only to the first level; for it was on the conventional level that Bhaavaviveka permitted himself the liberty of investigation into the maze of worldly knowledge. The difficulty is, however, that Bhaavaviveka eventually must bring himself, as a Maadhyamika philosopher, to discuss ultimate truth. This presents him with a dilemma. Does he continue to make his positive assertions into the realm of ultimate, nonconceptual knowledge, or does he confine his assertions to the conventional realm and again risk the charge of being guilty of vita.n.daa--this time on the matters of greatest importance to his philosophical school? If we look in both of Bhaavaviveka's major works, the Tarkajvaalaa and Praj~naapradiipa, we find that he takes a rather ambivalent position on this question. In the Praj~naapradiipa, he was writing a commentary on Naagaarjuna's root text, and this fact alone seems to have restrained Bhaavaviveka in his treatment of assertions at the ultimate level. The particular passage of interest on this point is the commentary on verse 18: 9 in which Naagaarjuna purports to give a "definition" of ultimate truth. Bhaavaviveka uses this as an opportunity to deal again with the question of vita.n.daa. [Objection:] If you think that ultimate truth (tattva) can be realized by completely rejecting the intrinsic nature of things which others conceptually construct, then you must state a definition of it. Otherwise you are refuting someone's position without establishing your own; and that is vita.n.daa. [Reply:] If the definition of ultimate truth can be expressed, it should be expressed. But it is not an object to be expressed (abhidheya). However, in order to give confidence to those who are just beginning, the following is said in terms of conceptual, discriminative knowledge. [Vs. 18: 9] Not caused by anything else, peaceful, not expressed by verbal diversity, non-conceptual, not diverse in meaning--this is the definition of ultimate truth (tattva). [Commentary] Since it is non-conceptual, it is not expressed by verbal diversity. Since it not expressed by verbal diversity, it is in the sphere of non-conceptual knowledge. Since it is in the sphere of non-conceptual knowledge, it is not known by means of anything else. Words do not apply to something that is not known by means of anything else. For this reason, the nature of ultimate truth completely surpasses words. It cannot be an object to be expressed, but the statement which negates both the intrinsic nature and the specific character- p. 331 istics of all things can make known the nature of ultimate truth. It [the statement] is produced by a superimposition of syllables which conform to the nonconceptual knowledge produced by the method of non-production. Therefore, since ultimate truth. which is actually directly known (svasamvedya), is taught here in an expedient way way (upaaya-dvaare.na) , we do, in fact, express a definition of ultimate truth. Thus we are not guilty of vita.n.daa, and your criticism does not apply.(23) Here, with some equivocation, Bhaavaviveka manages to stay close to Naagaarjuna. He admits that ultimate truth (tattva) cannot be directly expressed, but he says that one can, as an expedient, appear to give a definition of it. This, in his opinion, is enough to rebut the charge of vita.n.daa. In the third chapter of the Tarkajvaalaa, where he lays out his own independent philosophical position, Bhaavaviveka allows himself more liberty with Naagaarjuna's method. In this chapter he formulates some of the more characteristic elements of his own technique by considering a series of objections to a syllogism of the type (svarantra-anuma~na) for which his school of Maadhyamika Svaatantrika is named. The example he uses is a syllogism denying the intrinsic nature of the gross elements. We can formulate the syllogism in four steps: (1) earth, and so on. (2) do not have the intrinsic nature of elements, from the point of view of ultimate truth (paramaarthata.h), (3) because they are produced, (4) like consciousness. Steps 1 and 2 constitute the assertion (pratij~na) which Bhaavaviveka uses to deal with the accusation that he is guilty of vita.n.da. Step 2, however, raises another difficult question. As the syllogism is formulated in this example, the con- clusion belongs not to the realm of conventional truth, where words and concepts are appropriate, but to the realm of ultimate truth. How, then, can Bhaavaviveka allow himself to carry on conceptual thought in the ultimate realm? He deals with this problem in the following surprising way: [Objection: ] Paramaartha transcends all [conceptual] thought. and a negation of the intrinsic nature of things is in the domain of language. For this reason your negation fails. [Reply:] Paramaartha occurs in two forms. One of them is free from volition, transcendent, pure, and free from verbal diversity. The other is volitional, accords with the accumulation of knowledge and merit, clear, and possessed of the verbal diversity known as "worldly knowledge."24 Bhaavaviveka's interpretation of the word paramaartha allows him to do something that he did not permit himself with the word tattva in the passage just quoted from the Praj~naapradiipa. He interprets paramaartha as a compound meaning "knowledge of ultimate truth." This allowed him earlier to separate it from the experience of conventional truth; here it allows him to separate it into two different levels of experience of the same thing. One level is free from verbal diversity; the other is not. In this way, Bhaavaviveka can maintain p. 332 that, while ultimate truth (tattva) is one, the knowledge of ultimate truth (paramaartha) is not. The resulting distinction in levels of experience allows him to carry on positive philosophical activity at the "ultimate" level without being concerned about the fact that such activity involves words and concepts. Bhaavaviveka's analysis of the word paramaartha is thus a powerful tool in allowing him to carry out his philosophical program. He can maintain nominal adherence to the written text of Naagaarjuna's Kaarikaas and still permit himself to make positive philosophical assertions up to and within the realm of ultimate truth. Bhaavaviveka can claim a certain amount of success in adapting Naagaarjuna's method to the requirements of his own philosophical milieu. He gave up Naagaarjuna's prohibition against positive assertions, but he might well claim that this was a minor sacrifice made to keep the rest of Naagaarjuna's critique intact. A more damaging charge against Bhaavaviveka. however, might be that he violated the serious and fundamental prohibition against attributing intrinsic nature either to the words of a statement or to the things to which they refer. Bhaavaviveka did not discuss this point explicitly in either the Tarkajvaalaa or the Praj~naapradiipa on anything other than the ultimate level; perhaps he was not aware that it would be an issue. In any case, it is sufficient for our purposes in tracing the development of the early theories of language to know that a substantial portion of the Maadhyamika tradition did consider Bhaavaviveka guilty of this more serious charge. Candrakiirti and Tso^n-kha-pa both felt that, by establishing conventional truth as an independent realm, Bhaavaviveka had violated the most fundamental point of Naagaarjuna's method -- he had refuted intrinsic nature on the ultimate level, only to let it back into his account of language on the conventional level. In sorting out Candrakiirti's arguments against Bhaavaviveka's view of language, we are critically dependent on Tso^n-kha-pa's analysis of the issues between them. In the key passage in the Prasannapadaa, where he attacks the use of svalak.sa.na or "intrinsic identity" on the conventional level, Candrakiirti fails to identify his opponent. This has led Western interpreters to the quite reasonable supposition that the opponent Candrakiirti had in mind was not Bhaavaviveka, who does not explicitly develop svalak.sa.na as a substratum for his use of language, but the Buddhist logicians who do.(25) In the Legs-b`sad-s~ni^n-po, however, Tso^n-kha-pa argues on the basis of certain passages in the Praj~naapradiipa that Bhaavaviveka made a tacit assumption of svalak.sa.na on the conventional leve1.(26) Candrakiirti actually attacks the common idea that language requires a realistic basis in the world to function effectively. It would not be unreasonable to assume that the argument is meant to oppose any Buddhist who thinks he can analyze language on a conventional level and find any more substantial reality behind it than he does at the ultimate level. That way of thinking is normally associated in Buddhist philosophy with schools like the Vaibhaa.sikas or logicians who attempted in varying degrees p. 333 to give a realistic account of language, but Bhaavaviveka evidently slipped into this pattern as well when he distinguished so sharply between the two levels of truth. On the ultimate level, he maintained the emptiness of all things --although, as we have seen, he still permitted himself some freedom in making ultimate assertions--but on the conventional level he allowed language to function in the manner accepted by the other schools. Tso^n-kha-pa explains how widespread he thought this pattern to be. What way of thinking assumes that things are established by intrinsic nature (svalak.sana)? First, let us speak of the method of the philosophers. In an expression like, "Ths person performed an action and experienced the result," they investigate the meaning of the term "person" (pudgala) by asking whether this person is the same as his aggregates (skandhas) or different. In either case, whether he is the same or different, they can establish the person as the accumulator of karma, and so forth. If he cannot be established as either one, then they are not content with "person" as a mere term. In ths way, when the person is established by an investigation into that to which the term refers, the person is established by intrinsic identity. All Buddhist philosophers from Vaibhaa.sikas to Svaatantrikas hold this view.(27) The issue between Candrakiirti and Bhaavaviveka, then, as Tso^n-kha-pa sees it, is whether it is necessary to reintroduce an element of semantic realism into the conventional realm to anchor the use of language. Candrakiirti contends that this is impossible. The argument between Candrakiirti and the semantic realists, among whom Tso^n-kha-pa includes Bhaavaviveka, hinges on whether it is acceptable in conventional usage to make statements involving a svalak.sa.na or "intrinsic identity" underlying the ordinary use of language. Candrakiirti starts by having the opponent cite examples in an attempt to claim conventional justification for statements of this sort. The opponent says that a statement of the form, "Hardness is the intrinsic identity of earth," is acceptable, even though the intrinsic identity is identical with the earth, because similar expressions are part of conventional usage. For example, one can say, "the body of a statue" or "the head of Raahu," when the body is no different from the statue, and Raahu, a demon who has no body, is no different from his head. Candrakiirti puts the opponent's argument this way: Even so, even though there is no qualifier (vi`se.sa.na) apart from the body and head [which are qualified] in the cases, "body of a statue" or "head of Raahu," there is still a relationship of qualifier and qualified. Similarly, we can say "intrinsic identity of earth" even though there is no earth apart from its intrinsic identity. Candrakiirti replies: This is not so, because the cases are not similar. When the words "body" and "head" normally occur in grammatical connection with companion entities like "hand" or "mind," the thought produced on the basis of the words "body" p. 334 and "head" alone carries an expectation of the companion entities in the form, "Whose body?" and "Whose head?" Then it is reasonable for someone else, who wants to rule out a connection with a qualifier, to deny such an expectation by using the qualifiers "statue" and "Raahu" in a conventional way. But when earth is impossible apart from hardness, a relationship of qualifier and qualified is impossible.... Furthermore, the terms "statue" and "Raahu," which are the qualifiers, actually exist as part of conventional usage, and are accepted without analysis, as in the conventional designation "person." Therefore your example is incorrect.(28) There are actually two arguments here, one of which is somewhat stronger than the other. Candrakiirti says first that it is acceptable to qualify one word with a word that refers to the same thing only when there is some possibility of doubt about the fact that they are identical. There can be no doubt, however, that the intrinsic identity is the same as the earth; they are the same by definition. To say "intrinsic identity of earth," then, is to violate conventional usage by using the two terms in a grammatical connection that is appropriate only if there is the possibility that they refer to different things. The second argument is stronger. In this Candrakiirti says simply that "intrinsic identity" itself is an unacceptable term. It is a technical term masquerading as an ordinary word, and it is thus unacceptable as part of conventional usage in any grammatical connection at all. Candrakiirti's argument is most interesting, perhaps, when he gives his own account of the way language functions on the conventional level. He insists, as Naagaarjuna did, that he does not attempt to destroy the structure of ordinary language, as a semantic realist might suppose, but only to reestablish language on the proper grounds, that is, on pure convention. Candrakiirti outlines this theory in a reply to the further objection that, if his argument is correct, the expression "head of Raahu" can actually be no more acceptable than "intrinsic identity of earth," since, in both cases, when the objects referred are analyzed, they are found to be identical. His reply is that the matter of "analysis" (vicaara) is precisely what distinguishes ultimate from conventional truth. Conventional truth is only established unanalytically. When conventional terms are examined to find their true reference, they are no longer used conventionally. He explains this in the following passage: [Objection:] The examples are correct, because only the body [of the statue] and the head [of Raahu] are actually cognized, since no other entities exist apart from them. [Reply:] This is not the case; for such analysis is not carried out in conventional usage, and without such analysis, conventional entities exist. When the aatman is analyzed as to whether it is different from [or the same as] matter, etc., it is not possible. But it does exist conventionally (loka-sa.mv.rtyaa) with reference to the skandhas. The same is true of Raahu and the statue: Thus the example is not established. Likewise, in the case of earth, etc., after analysis, there is nothing to be qualified that is different from hardness, etc. [which are the qualifiers], and a qualifier without something to be qualified is groundless. Even so, the masters have maintained that they exist in mutual dependence p. 335 (parasparaapek.saa siddhi.h) as purely conventional. This must be accepted in just this way. Otherwise, the conventional could not reasonably be distinguished. It would become ultimate truth (tattva), not conventional.(29) The last few lines of this passage show what, in Tso^n-kha-pa's view, was the real point of dispute between Candrakiirti and Bhaavaviveka. On one level, it is a dispute over language, but one does not have to go far beneath the surface to find the troublesome issue of two levels of truth. Candrakiirti argues here that the attempt to establish an independent conventional realm, in which it is possible to carry out constructive philosophical reasoning, involves a misunderstanding of the distinction between the two levels. To give substantial reality to the conventional level, even with the laudable intention of promoting philosophical debate, was to transform it into a false ultimate. It was also to misunderstand the point of the discussion about language. To Candrakiirti it was unnecessary to find some substantial reality to which words could refer to acquire their meaning. It was necessary only that they be used the way they are. Whatever meaning they had was acquired by a process of mutual dependence (parasparaapek.saa siddhi), with one word depending for its meaning on the network of those that were used before it. In Candrakiirti's view, the move that Bhaavaviveka made on the conventional level was the one that led other Indian philosophers into trouble. It was an attempt to make the technical terms of philosophy into more than conceptual constructions. The development of the different accounts of language in early Maadhyamika philosophy is a complicated process, involving steps over which individual philosophers often strongly disagreed. Rather than be deterred by this diversity, however, we should accept it as a challenge to greater efforts of understanding. There are still formidable problems to be solved by both historians and comparativists. The peculiar methods of Bhaavaviveka, for instance, need much more thorough study before we can accurately assess his relation to Candrakiirti and the Praasa^ngika school that has so dominated Western interpretations of Maadhyamika. Maadhyamika philosophy was not monolithic. A greater historical sophistication in understanding the differences between philosophers is an essential element, not only in understanding the philosophers themselves, but in developing true conceptual precision in the act of comparison. NOTES 1. This article is the revised version of a paper first presented at a workshop of the Society for Asian and Comparative Philosophy of the American Philosophical Association, Boston, December 28, 1976. 2. Two particularly useful examples of this kind of study are: Jacques May, "La philosophie bouddhique de la vacuite, " Studia Philosophica 18 (1958):123-137; Jan de Jong, "The Problem of the Absolute in the Madhyamaka School." Journal of Indian Philosophy 2 (1972): 1-6. p. 336 3. My use of the phrase "early Maadhyamika" is, of course, somewhat arbitrary. In this article, I will use it to refer to the period in which the major differences between competing subschools were first formulated. This covers the period from the time of Naagaarjuna in the second century A.D. to Candrakiirti in the late sixth or early seventh. 4. The bibliography of Western translations of the works of Naagaarjuna and Candrakiirti is well known and need not be recited here. The available works of Buddhapaalita and Bhaavaviveka are perhaps less well known. A partial translation of chapter 2 of Buddhapaalita's commentary on Naagaarjuna's Madhyamakakaarikaas is available in, Musashi Tachikawa, "A Study of Buddhapaalita's Muulamaadhyamakv.rtti, (1) ." Bulletin of the Faculty of Letters of Nagoya University 63 (1974) : 1-19. Chapter 1 of Bhaavaviveka's Praj~naapradiipa is available in, Yuichi Kajiyama, "Bhaavaviveka's Praj~naapradiipa. (1. Kapitel) , " Wiener Zeitschrift fur die Kunde Sud- und Ostasiens 7 (1963):37-62 and 8 (1964): 100-130. Portions of Bhaavaviveka's Maadhyamakah.rydayakaarikaas (verses) and Tarkajvaalaa (commentary) have been translated in, V. V. Gokhale, "The Second Chapter of Bhavya's Maadhyamakah.rdaya (Taking the Vow of an Ascetic)," Indo-Iranian Journal 14 (1972): 40-45: V. V. Gokhale, "Masters of Buddhism Adore the Brahman Through Nonadoration' (Bhavya, Madhyamakah.rdaya, III)," Indo-lranian Journal 5 (1961-1962) : 271-275; V. V. Gokhale, "The Vedaanta Philosophy Described by Bhavya in His Madhyamakah.rdaya, " Indo-Iranian Journal 2 (1958): 165-180; Andre Bareau, "Trois traites sur les sectes bouddhiques, He partie." Journal Asiatique 244 (1950):167-199; Shotaro Iida, An Introduction to Svaatantrika Maadhyamika, University of Wisconsin Ph.D. dissertation (Ann Arbor, Mich.: University Microfilms, 1968); hereafter cited as Iida, dissertation. 5. See particularly Naagaarjuna's Maadhyamakakaarikaas 24: 8-101 found in Louis de la Vallee Poussin, ed., Muulamaadhyamakakaarikaas (Madhyamaka Suutras) de Naagaarjuna avec la Prasannapadaa Commentaire de Candrakiirti, Bibliotheca Buddhica, vol. 4 (St. Petersburg, 1913) , pp. 492-494; hereafter cited as Prasannapadaa. 6. E. H. Johnston and Arnold Kunst, eds., "The Vigrahavyaavartanii of Naagaarjuna with the Author's Commentary, " Melanges chinois et bouddhique 9 (1948-1952):108; hereafter cited as Vigrahavyaavartani. 7. Vigrahavyaavartanii, p. 108 (translations are mine unless otherwise noted). 8. Vigrahavyaavartanii, p. 115. 9. Vigrahavyaavartanii, p. 122. 10. Vigrahavyaavartanii, p. 127. 11. Vigrahavyaavartanii, pp. 145-147. 12. Iida, dissertation, p. 86. 13. Nyaaya-suutra 1.2.3: sa pratipak.sa-.sthaapanaa- hino vita.n.daa. The text of this suutra with commentaries can be found in Anantalal Thakur, ed., Nyaayadar`sana of Gautama, Mithila Institute Series, Ancient Text No. 20, vol. 1 (Darbhanga, 1967), p. 628. There has been disagreement over whether the Maadhyamikas were actually accused of vita.n.daa. In a recent article ("Maadhyamika et Vaita.n.dika," Journal Asiatique 263 [1975]: 99-102) , Kamaleswar Bhattacharya argues that Maadhyamikas were not guilty of vita.n.daa according to the strict definition of the term given by the Nyaaya commentators. Uddyotakaara, for instance, explains that vita.n.daa means the absence of proof (sthaapanaa) of a counterposition rather than absence of the counterposition itself. In his replies to the accusation of vita.n.daa, however, Bhaavaviveka does not differentiate between sthaapanaa and pratipak.sa in the subtle manner favored by the later commentators. He thought the charge sufficiently applicable to his Maadhyamika method to require serious refutation. 14. See notes 15 and 23 herein. 15. Tibetan text in Iida, dissertation, pp. 109-110 (translation mine unless otherwise noted). 16. Vigrahavyaavartani, p. 127. 17. Giuseppe Tucci, "The Ratnaavalii of Naagaarjuna," Journal of the Royal Asiatic Society, (1934). 18. Bhaavaviveka gives his grammatical analysis of the term paramaartha in both the Tarkajvaalaa and the Praj~naapradiipa. The appropriate passage from the Tarkajvaalaa is available in Iida, dissertation, pp. 101-102. The following is a translation of a similar analysis from chapter 24 of the Praj~naapradiipa. "Paramaartha is the 'supreme object' [karmadhaaraya compound] because it is both supreme (parama) and an object (artha). Or it is the `object of the supreme' [tatpuru.sa compound] because it is the object of supreme, non-conceptual knowledge. It is defined as `not realizable p. 337 through anything else.' Because paramaartha is true, it is `ultimate truth' (paramaartha-satya), and it always, in every way, remains the same. Non-conceptual knowledge, which possesses that as its object by the method of having no object (vi.sayaabhaavanayena) , is also paramaartha because it is `that whose object is ultimate' [bahuvriihi compound]." Tibetan text in Daisetz T. Suzuki, ed., The Tibetan Tripitaka: Peking Edition (Tokyo-Kyoto: Tibetan Tripitaka Research Institute, 1957), vol. 95, p. 246 (folio Tsha 286a-b); hereafter cited as Peking Tripitaka. 19. Sanskrit text in Iida, dissertation, pp. 82-83. 20. Tibetan text in Iida, dissertation, p. 84. 21. As Bhavaviveka explains in the Tarkajvaalaa, these subtleties included: "grammar (ak.sara`saastra), palmistry (mudraa), alchemy (?), medical science (cikitsaa) , arithmetic (ga.nanaa) , charms (mantra) , spells (vidyaa) , etc." (Iida's translation, dissertation, p. 86). 22. This is particularly true for Vedaanta, where little other evidence is available from this early period, and for the eighteen schools of Nikaaya Buddhism. See V. V. Gokhale, "The Vedaanta Philosophy Described by Bhavya in His Madhyamakah.rdaya, " Indo-Iranian Journal 2 (1958): 165-180; Andre Bareau, "Trois traites sur les sectes bouddhiques, IIe partie, " Journal Asiatique 244 (1950):167-199. 23. Peking Tripitaka, vol. 95, p. 227 (folio Tsha 237a-b). 24. Tibetan text in Iida, dissertation, pp. 107-108. 25. Th. Stcherbatsky, The Conception of Buddhist Nirvaa.na (Leningrad, 1927), pp. 149-156. 26. Tso^n-kha-pa, Dra^n-^nes-legs-b`sad-s~ni^n-po (Varanasi: Gelugpa Students' Welfare Committee, 1973). In all my observations about Tso^n-kha-pa, I am indebted to Acarya T. T. Doboom Tulku, who first read this text with me, and to Professor Robert A. F. Thurman, who kindly made available his unpublished translation. 27. Legs-b`sad-s~ni^n-po, p. 143. 28. Prasannapadaa, p. 66. 29. Prasannapadaa, p. 67.