Madhyamaka: Conventional Categories in Madhyamaka Philosophy

Jessica Frazier (ed): Categorisation in Indian Philosophy. Thinking Inside the Box, Ashgate, Farnham, Surrey, 2014, 115-131. Chapter 8 Madhyamaka: Conventional Categories in Madhyamaka Philosophy Preprint only. If you would like to cite this chapter please refer to the final published version. Jan Westerhoff The status of categories within Madhyamaka philosophy is a curious one. On the one hand there is a strong tendency to reject philosophically refined analyses of the constituents which make up the world, thereby rejecting systems of categories as well. The Mādhyamika, it seems, accepts whatever conventions the world accepts at the merely conventional level but does not propose any conventions of his own. In fact there appear to be good reasons for such a view. Given that the membership of an object in a category is generally taken to be a clear example of a property an object has intrinsically, and since the Mādhyamikas reject intrinsic properties (properties which exist by svabhāva), they should reject categories as well. On the other hand, however, Mādhyamikas make use of the very sophisticated and intricate categorial frameworks found in traditional Indian grammar and in the Abhidharma. Furthermore they also vehemently argue against the use of other frameworks, such as that of the Naiyāyikas. This chapter will explore ways of resolving this tension and investigate more generally what role categories play in the Madhyamaka system of philosophy. Madhyamakas and Ontological Categories When it comes to the study of ontological categories it may seem as if the Mādhyamika does not have much of interest to say. Candrakīrti points out that ‘What is admitted to exist in the world, I too admit that to exist. What is admitted to be nonexistent in the world, I too admit that to be nonexistent.’1 This assertion, made in the context of a discussion of the Buddha’s graded teaching suitable to different kinds of disciples, appears to reflect an extreme loko mayā sārdhaṃ vivadati nāhaṃ lokena sārdham vivadāmi / yal loke ‘sti saṃmataṃ tan mamāpy asti saṃmataṃ / yal loke nāsti saṃmataṃ mamāpi tan nāsti saṃmatam. In this passage of the Prasannapadā, commenting on MMK 18:8, Candrakīrti cites the first chapter, called Trisaṃvaranirdeśaparivarta, of the Ratnakūţa (de la Valleé Poussin 1903–13, 370.6–8). 1 Categorisation in Indian Philosophy 116 conventionalism. Whatever ontological commitments ‘the world’ makes are accepted for the purely pragmatic purpose of teaching those who hold these commitments. In the following we will refer to this as the common-sense principle (CS), a principle asserting that, for the Mādhyamika, the only systems of ontological categories asserted at the conventional level are those accepted by common sense. For thinkers accepting this principle there seems to be no place for the traditional philosophical enterprise of ontological theory-building: thinking about what our most fundamental conceptual and linguistic approaches to the world are and attempting to systematise them in a single coherent theory. Apart from its incompatibility with ontology as traditionally conceived, CS also leads to a rather unsatisfactory account of spelling out the notion of conventional truth (saṃvṛtisatya). If we equate any kind of truth with popular consensus we lose the ability to argue that the majority of people can ever be wrong about it, since being thought to be so is being so, at least if sufficiently many people do the thinking. This difficulty seems to be aggravated by the Madhyamaka claim that the conventions of the world (loka) are to be understood non-analytically (avicāratas). It is this nonanalytical mode of engagement of mundane cognitions (of all persons) that determines what Candrakīrti calls ‘worldly convention’ (lokasaṃvṛti). These conventions form the basis of conventional truth but do not include inquiry into the way things really are. We will refer to this claim as the no-analysis principle (NA). The principle rejects the application of analysis to the systems of categories licensed by CS in order to adopt one over another and to add theoretical refinements to the systems one ends up adopting. Adopting NA in an unqualified way entails that our conventional manner of epistemic engagement with the world does not require us to go beyond what our senses tell us in order to arrive at some more fundamental level of knowledge about the world. With respect to causation Candrakīrti points out that ‘The worldly person indeed, without starting to analyze whether it is from itself of from other and so forth, understands only this much: from a cause an effect arises. The Master also explained things in the same way.’2 There seems to be no room for a philosophically sophisticated theory of causation over and above pointing out what is conventionally accepted: rice comes from rice-seeds and barley from barley-seeds, sheep give rise to sheep, cows to cows and so forth. Candrakīrti regarded the analysis of ordinary notions, attempting to arrive at an understanding of them that goes beyond ordinary usage, as fundamentally mistaken (see Tillemans 2010). This emphasis on the conventional and commonly acknowledged later led Tibetan scholars to characterise the loko hi svataḥ parata ityevamādikaṁ vicāram anavatārya kāraṇātkāryam utpadyata ityetāvanmātraṁ pratipannaḥ | evam ācāryo’pi vyavasthāpayāmāsa || (de la Valleé Poussin 1903–13, 27:4–5). 2 Madhyamaka 117 Prāsaṅgika Mādhyamikas (i.e. Candrakīrti and his followers) as ‘jig rten grags sde pa, as ‘the school [based on] what is known by the world’ (Ruegg 2000, 58, note 124). Does the joint adoption of CS and NA mean we have to understand the Madhyamaka conception of conventional truth as ‘whatever it is that people think’, without any possibility of analysing whether what they think is actually true? Fortunately this is not the case. In fact later Madhyamaka writers are quite explicit that any attempt to take general opinion uncritically at face-value cannot be a philosophically viable position. Kamalaśīla observes: Suppose it were thought ‘Why should we analyze it, when the production of sprouts and the like being conditioned by seeds and so forth is just simply acknowledged by everyone from cowherds on up? Judicious people should not analyze in order to ascertain the natures of entities, because it would follow that there would be no end [to such analysis] and it would follow that it would not be judicious.’ This is not right. ... Concerning this, consider this example: it is acknowledged that perishing is something that has a cause, and though people acknowledge that matter and the like are external objects, this can be invalidated by a means of reliable cognition if it is subjected to analysis.3 In the same way here too, what people acknowledge could also turn out to be false, and hence one really should analyze it.4 How could we know that what people acknowledge turns out to be false? One way of doing so is by determining that they have based their opinion on unreliable epistemic instruments.5 Tsong kha pa points out that: Kamalaśīla here refers to the Sautrāntika view that the cessation of a momentary existent does not require a separate cause but already results from the existent itself and to the Yogācāra denial of external objects. Both are positions to be accepted, even though they contradict common sense. It is only by analysis of the convictions of common sense that their deficiency can be established. 4 gal te ‘di snyam du myu gu la sogs pa’i skye ba sa bon la sogs pa’i rkyen can gnag rdzi yan chad la grags pa kho na yin na de la dpyad par bya ci dgos / dngos po’i bdag nyid gtan la phab pa la ni rtog pa dang ldan pa rnams dpyad par rung ba ma yin te / thug pa med par thal bar ‘gyur ba’i phyir dang / rtog pa dang mi ltan pa nyid du thal bar ‘gyur ba’i phyir ro snyam du sems na / de ni rigs pa ma yin te / [...] / de la dper na ‘jig pa yang rgyu dang ldan pa nyid du grags la / gzugs la sogs pa yang phyi rol gyi don nyid du grags zin kyang / dpyad pa byas na tshad mas gnod pa srid pa de bzhin du ‘di la yang grags pa brdzun pa’i ngo bo yang srid pas dpyad par bya bar rung ba nyid do // Sarvadharmaniḥsvabhāvasiddhi (Peking: 312a–312b). 5 Tsong kha pa gives three conditions for something to exist at the level of conventional truth: 1. it appears to conventional consciousness, 2. no other epistemic instrument contradicts it, 3. ultimate analysis (see below) does not contradict it. (tha snyad pa’i shes pa la grags pa yin pa dang ji ltar grags pa’i don de la tha snyad pa’i tshad ma gzhan gyis gnod pa med pa dang | de kho na nyid la’ang rang bzhin yod med tshul bzhin du dpyod 3 118 Categorisation in Indian Philosophy The internal impairments of the sense faculties are such things as cataracts, jaundice, and such things as hallucinogenic drugs one has consumed. The external impairments of the sense faculties are such things as mirrors, the echoing of sounds in a cave, and the rays of the autumn sun falling on such things as white sand … The impairments of the mental faculty are, in addition to these, such things as erroneous philosophical views, fallacious arguments and sleep … Taking conventional objects grasped by such unimpaired and impaired cognitive faculties to be real or unreal, respectively, merely conforms to ordinary cognitive practice. This is because they actually exist as they appear or do not, according to whether or not they are undermined by ordinary cognition. This distinction is not drawn from the perspective of the Āryas. This is because just as such things as reflections do not exist as they appear, such things as blue, that appear to exist through their own characteristics to those who are afflicted by ignorance, do not actually exist as they appear.6 The point made here is that there are some beliefs a group of people may hold (such as that there is a snake in the dark room in front of them or water in a desert) which can be refuted by conventional means, that is by demonstrating that the epistemic instruments used for making these cognitions were in some way impaired (for example by lack of light or by light reflected in a particular way) and by replacing them by faculties which are not so impaired. It is therefore possible to draw a distinction between what is real (a snake, water) and what is unreal (a snake mistaken for a rope, a mirage) at the level of conventional reality. This does not change the fact, however, that all four have the same degree of unreality at the ultimate level, since they all lack an intrinsic nature. It is seems to be the case, therefore, that NA has to be modified to allow the Mādhyamika to carry out some analysis as long as he does not stray too far from pa’ i rigs pas gnod pa mi ‘bab pa zhig ni tha snyad du yod par ‘dod la | (1985, 627:15–18, see Cutler/Newland 2000–2004, vol. 3, 178). The following passage considers cases which violate the second condition. 6 Samten/Garfield 2006, 485. de la dbang po la gnod pa’i skyon nang na yod pa ni rab rib dang mig ser la sogs pa dang brag phug la sogs pa nas sgra brdzod pa dang sos ka’i nyi ma’i ‘od zer bye ma skya bo’i yul dang nye bar gyur pa la sogs pa ste | ... yid kyi dbang po’i gnod pa ni de dag dang grub mtha’ log pa dang gtan tshigs ltar snang dang gnyid kyi gnod pa sogs so | ... de lta bu’i gnod pa med pa’i shes pa drug gis bzung ba’i kun rdzob pa’i don dang | de las bzlog pa’i don la yang dag dang log par ‘jog pa ni ‘jig rten pa’i shes pa kho na la ltos nas yin te | de dag ji ltar snang ba ltar gyi don du yod pa la ‘jig rten pa’i shes pas gnod pa med pa dang yod pa’i phyir ro | ‘phags pa la ltos nas ni yang dag dang log pa gnyis su med do | ji ltar gzugs brnyan la sogs pa snang ba ltar gyi don du med pa bzhin du | ma rig pa dang ldan pa rnams la sngon po la sogs pa rang gi mtshan nyid kyis grub par snang ba yang snang snang ba ltar gyi don du med pa’i phyir (Dhondup 2006, 2: 287–8). Madhyamaka 119 ‘what is admitted in the world’ and thus licensed by CS.7 The Mādhyamika is not obliged to concur with any belief the people happen to hold but only with those which are the result of applying the right kinds of epistemic instruments in the right manner. Beliefs that do not pass this test can be dismissed as deficient. But does the possibility of analysis at the conventional level also allow us to carry out ontological analyses of the categorial structure of the world? Unfortunately this is not the case. When Candrakīrti points out that worldly convention does not engage in analysis he intends this to extend to ultimate analysis as well. Tsong kha pa observes that … the Master [Candrakīrti] repeatedly said that these conventionalities, such as shape and sound etc do exist. However, they are not in the least established by ultimate analysis (dpyod pa’i rigs pa), that is by reasoning that analyses whether they have intrinsic nature. Thus the scrutiny of reason is not applied to them.8 Even if Candrakīrti rejects intricate theories of conventional phenomena full of philosophical commitments he does not reject the usefulness of commonly accepted conventional concepts such as shape and sound. At the level of conventional truth, by means of conventional analysis, medium-sized dry goods such as tables and chairs are known to exist and mirages, hallucinations and other unreal objects are known not to exist. At the level of ultimate truth, however, tables and chairs do not exist. This is because what exists ultimately is established by ultimate analysis, which is an analysis trying to find an object’s substantial or intrinsic nature (svabhāva). Since this, the Mādhyamikas argue, does not exist, then tables and chairs do not exist from the perspective of ultimate truth. The important point now, is that the categorial analysis ontologists are concerned with is a variety of ultimate analysis and the truths it establishes are therefore ultimate truths. Of course this does not apply to any kind of categorial analysis. 7 Tsong kha pa points out (1985, 627: 18–628: 2) that conventional consciousness does not completely lack inquisitive powers: ‘In a sense, conventional consciousness operates in a non-inquisitive manner. It operates within the context of how a given phenomenon appears to it without asking, “Is this how the object actually exists, or does it just appear that way to my mind?” It is thus called non-analytical, but it is not the case that it is utterly non-inquisitive.’ (de la tha snyad pa’i shes pa ni cos gang la’ang ji ltar snang ba ltar gyi rjes su ‘jug pa tsam yin gyi | snang ba’i don de blo la de ltar snang ba tsam yin nam | ‘on te don gyi yin tshul la de ltar grub pa yin snyam di mi dpyod pa’i shes pa ma brtags par ‘so | de nyid la ma dpyad pa’i shes pa zhes bya yi | brtag pa ye me byed pa ni min no | See Cutler/ Newland 2000–2004, vol. 3, 178). For Tsong kha pa the process of conventional analysis may carry out specific analyses as long as these are not concerned with the way objects actually exist. 8 gzugs sgra la sogs pa kun rdzob pa ‘di rnams yod du chug kyang de kho na la dpyod pa’am rang bzhin yod med dpyod pa’i rigs pas gtan mi ‘grub pas de dag la rigs pa’i brtag pa mi ‘jug go zhes slob dpon ‘dis yang yang gsungs shing | (1985, 608: 10–13. See Cutler/ Newland 2000–2004, vol. 3, 157). Categorisation in Indian Philosophy 120 Classifying knives, forks and spoons into different categories of eating implement is perfectly fine as a categorial structure at the conventional level. Problems arise once we pass from the realm of lower level non-ontological categories to the higher realm of those categories ontologists are interested in.9 Nobody will want to claim that classifying some object as a spoon will say anything about that object’s intrinsic nature. It is classified as a spoon because there are certain conventions about which shapes of eating implements 21st-century humans distinguish and the object in question resembles other objects called ‘spoon’ more or less closely. It is the existence of these conventions which make it a spoon, not any kind of already-present spoon-ness which its classification as a spoon reflects. Classifying an object in terms of the ontological category it belongs to (such as ‘material object’ or ‘spatially extended object’), on the other hand, is to be understood as based on the object’s intrinsic nature, some property it has in and through itself, something the object could not even lose in any conceivable circumstances. (Try imagining a world in which this spoon, or its counterpart, is not a material object).10 For this reason a theory of categories in any philosophically important sense is not just a conventionally approved classificatory device but a theory of characteristics objects have by svabhāva. Since the Mādhyamika denies the existence of any such things, such a theory must necessarily be vacuous. Furthermore it cannot be any theory that is part of the conventional analysis of phenomena, since it investigates characteristics that are supposed to be intrinsic.11 It therefore appears that despite the fact that conventional analysis can very well save us from the strict reading of CS and NA that Tillemans calls the ‘dismal view’ of having to accept whatever a sufficiently large group of people believes as ultimate truth, a theory of ontological categories is incompatible with the main claim Madhyamaka thought sets out to establish, namely that there is no svabhāva. Madhyamaka and Grammatical Categories The story does not end here, however. If we look at Madhyamaka texts we realise that, despite the advocacy of an ordinary, cowherd-compatible view of the world, Mādhyamikas employ fairly sophisticated categorial schemas to make their points. 9 For a discussion of different between ontological and non-ontological categories see Westerhoff, 2004. The book-length treatment is in Westerhoff, 2005. 10 At least this is the way ontological categories are usually understood. I do not think this view stands up to scrutiny and have given my reasons for this elsewhere. It is, nevertheless, the dominant view of ontological categories. 11 Dan Arnold notes the fact that theoretical re-framings of conventions, such as categorial analyses, do not themselves have any grounding in conventional practice and are therefore not conventional truths. ‘What is conventionally true, in other words, is just our conventions, and any technical redescription thereof is, ipso facto, not conventionally valid’ (Arnold 2005, 98–9). Madhyamaka 121 The schemes appear to go well past anything one would expect a principle like CS to license. A particularly clear example of this is Candrakīrti’s use of kāraka theory from traditional Indian grammar in explicating the second chapter of Nāgārjuna’s MMK.12 Kāraka theory deals with the various functions objects described by the different parts of a sentence can occupy relative to the action described by the verb; for example, being the agent, being the recipient, being the instrument and so forth. In order to point out these functions the relevant expressions are put in particular cases, such as nominative, dative, instrumental etc. In his commentary Candrakīrti notes that the verb gamyate ‘is moved’ requires a term denoting an agent (kartṛ) that is the mover (gantṛ).13 However, Nāgārjuna observes in MMK 2:5 that a moving mover has to be conceived of as endowed with two kinds of motion-properties: a constitutive property by which it is a moving object (yena tat gamyamānam) and an instantiated one by which it has the property of moving (yat … gamanam). The idea behind this seems to be that in our ordinary use of predication we usually single out one property as constitutive which then provides the metaphysical condensation-nucleus which can instantiate other properties. Something which has the properties of being shiny, blue and pot-shaped is conceptualised either as a shiny blue pot or as a blue, pot-shaped shiny thing or as a shiny, pot-shaped blue thing. Speaking about moving movers, however, the only property available has to be split into two and each has to play a different role in order to satisfy the familiar individual-property instantiation framework: one movement constitutes an individual and the other specifies what that individual does. But if the property of moving requires a mover, and if two different motions have to be in play – a constitutive and an instantiated one – then this means that we are also faced with two distinct agents, one for each motion, rather than just with one. There seems to be something seriously wrong with analysing statements like ‘the mover moves’ according to the individual-property framework. An obvious reply to this seems to be that the presence of many actions does not necessarily entail the presence of many agents. We can, even though we should not, drive a car, talk on the phone and read the newspaper at the same time. But this does not mean there are three people in the car, a driver, a talker and a reader. In order to address this issue we have to consider the traditional Indian account of kārakas a bit more closely. Candrakīrti draws on Bhartṛhari’s account when he points out that the kāraka is not a substance (dravya) but a power (śakti) (Bhattacharya 1980, 89). It is interesting to note that Bhāviveka also argues against Kāraka is set out by Pāṇini in the Aṣṭādhyāyi I.4.24–54; for a brief discussion see chapter 5 of Matilal 1990. 13 In his commentary on MMK 2:6 Candrakīrti observes that ‘Since an action (kriyā) necessarily depends on a means of accomplishing it (svasādhana) [which is] either the object (karman) or the subject (kartṛ) [of the action], the action of motion also involves an agent and therefore depends on an agent of motion.’ yasmād avaśyaṃ kriyā svasādhanam apekṣate karma kartāraṃ vā | gamikriyā caivaṃ kartary avasthitā ‘to gantāram apekṣate (PP 96: 8–9). Here sādhana is taken to be synonymous with kāraka. 12 Categorisation in Indian Philosophy 122 the conception of the agent as a substance when he points out that the grammarians consider the agent of the action of going to be the goer. He objects that the agent is merely an assembly of conditioned factors, meaning that he took the grammarians to regard the agent as something over and above the assembly of such factors, namely as a substance.14 The difficulties with regarding the kārakas as substances are apparent. After all, the very same object can fulfil different semantic functions in different contexts. In one sentence the same expression may denote an agent, in another a recipient. If the referent was substantially an agent, however, it could never play a different role in a different context. It is therefore more sensible to understand the kārakas as a set of powers denoting what the object can do in different contexts: it has the power to act as an agent in one context, as a recipient in another and so forth. The irresponsible driver mentioned earlier has the powers to drive, smoke and read and each power functions as an agent for the three different actions of driving, smoking, and reading. These powers are differentiated with reference to the actions of which they are agents; the power to smoke is not the same as the power to drive, because smoking is different from driving. But the constitutive and instantiated motion discussed by Nāgārjuna are actions of the same nature and therefore should have the same power as an agent. If they are really two motions as the analysis demands, however, then we also have to split up the power of the agent. Nāgārjuna and Candrakīrti use this result in order to argue that there is something fundamentally wrong with analysing the individual referred to in statements like ‘the mover moves’ in terms of constitutive and instantiated properties and, more generally, that such an analysis generates problems even when applied to other individuals. In the present discussion there is no need to follow the further ramifications of the argument. We want to note, however, the way in which Candrakīrti’s discussion utilises the kāraka theory. We are obviously dealing with an example of ultimate analysis here; what is at issue is whether objects have an intrinsic nature (svabhāva). Nevertheless, the framework used is part of the highly intricate system that is traditional Indian grammatical analysis. While it is a set of conventional truths it is certainly not at the level of cowherd-compatible truths such as ‘fire is hot’ and ‘rice-sprouts come from rice-seeds’. We would also not settle questions arising about the framework, such as whether a kāraka is a substance or a power, by trying to determine what is ‘admitted by the world’. People other than grammarians are unlikely to have any view on this matter and, even if they did, this would be unlikely to supply us with the answer we were looking for. We should also note that the piece of grammatical theory Candrakīrti uses is one that involves a substantial amount of ontological theorising, a classification of the different kinds of relata of verbs and an analysis of these in terms of other ontological categories, such as substances and powers. The use of kāraka theory therefore seems to suggest that Madhyamaka arguments can and do employ sophisticated systems of conventional truths for the purpose of ultimate analysis. 14 In Prajñāpradīpa on MMK 2:6, see Ames 1995, 308. Madhyamaka 123 Madhyamaka and the Categories of the Abhidharma Madhyamaka philosophy can be regarded as a direct reaction to the philosophical theories set out in the Abhidharma. A particularly clear example of this is to be found in one of Nāgārjuna’s hymns, the Lokātītastava, where he goes through a variety of ontological categories familiar from Abhidharmic analysis – the five skandhas, the four great elements, agent and action, characteristic and characterised and so on – in order to argue that none of them exist by svabhāva. In this text Nāgārjuna, addressing the Buddha, says: You are convinced that apart from the mere skandhas no sentient being exists, and yet, great sage, you surrendered great pain for the sake of sentient beings. O Intelligent one, you declared to the Intelligent Ones that the skandhas are like an illusion, a mirage, a city of Gandharvas, and a dream.15 The realisation of the illusion-like nature of the skandhas (i.e. their emptiness) is achieved on the basis of splitting up the person into five groups, according to the theory of persons developed in the Abhidharma. Nāgārjuna rejects the Ābhidharmika’s understanding of categories such as these but uses them in his own discussion in order to establish the theory of universal emptiness. The Lokātītastava and a variety of other similar texts, such as the Heart Sūtra or Nāgārjuna’s Pratītyasamutpāda-hṛdaya-kārikā, make it clear that a grasp of the categories of the Abhidharma constitutes a precondition of the realisation of emptiness. These have to be understood first before before emptiness can be realised. This is why the Heart Sūtra points out that whoever ‘… wishes to practice the conduct within the profound Perfection of Wisdom should make the following observation: he properly sees the five skandhas, and properly sees them as empty.’16 According to this sūtra the Buddhist path does not therefore start out with the theory-free or unsophisticated conventionalities of the ordinary man on the street and proceed from them to a direct realisation of emptiness, as some readings of CS might lead one to think. The starting point is rather the intricate analytical frameworks the Abhidharma has projected onto the world, in this case the theory of the five skandhas that breaks up what is commonly referred to as a person into five clearly defined constituents. That the exposition of the theory of emptiness is based on the Abhidharmic categories is of course also evident from the discussion in Nāgārjuna’s MMK. Nagarjuniana 3–4: skandhamātravinirmukto no sattvo ‘stīti te matam / sattvārthaṃ ca paraṃ khedam agamas tvaṃ mahāmune // te ‘pi skandhās tvayā dhīman dhīmadbhyaḥ saṃprakāśitāḥ / māyāmarīcigandharvanagarasvapnasaṃnibhāḥ (Lindtner 1982, 128). 16 gambhīrāyāṃ prajñāpāramitāyāṃ caryāṃ cartukāṃāḥ tenaivaṃ vyavalokitavyam pañca skandhāṃstāṃśca svabhāvaśūnyān samanupaśyati sma (Vaidya 1961, vol. 1:98). 15 Categorisation in Indian Philosophy 124 A variety of categories the Abhidharma accepts, such as characteristic (lakṣaṇa) and characterised (lakṣya), cause (hetu) and condition (pratyaya), space (ākāśa) and so on are all analysed in order to show that none of them exists with svabhāva. Candrakīrti makes it clear that such concepts as the ‘immediately contiguous condition’ (samanatara-pratyaya) used by Nāgarjunā are indeed the same as employed by the Abhidharma.17 The analysis in the MMK clearly does not start from a basis of unsophisticated conventions in order to arrive at the theory of emptiness by analysing these. It rather speaks to someone familiar with the highly complex framework of Abhidharma categorisation of phenomena. In order to follow the arguments Nāgārjuna offers in the MMK it is necessary to have a thorough acquaintance with the system of Abhidharma in order to use it as a stepping-stone to proceed to an understanding of emptiness. But, one might argue, perhaps it does not matter which stepping-stone one uses. Candrakīrti does not think that there is anything wrong with the categorial analysis of the Abhidharma at the conventional level,18 even though it is ultimately mistaken by assuming that some phenomena exist by svabhāva. So perhaps at the level of conventions it does not really matter which theory of categories one comes up with. It may be one derived from the common sense of cowherds or it may stem from the common sense of a philosopher. This is because, for each possible theory one could propose, a variant of Nāgārjuna’s arguments can be developed. While there is something true about this reply it does not do justice to the complexity of the status of systems of categories in Madhyamaka thought. This is because the Mādyhamikas are perfectly willing to criticise philosophical theories at the conventional level, thereby appearing to contradict NA. Let us look at one specific example. Madhyamakas and the Nyaya Categories We will consider two passages which appear to conflict with a straightforward reading of NA. Both come from the Vaidalyaprakaraṇa, a treatise by Nāgārjuna in which he presents a comprehensive criticism of the 16 Nyāya categories. The first passage I shall examine in the context of this discussion occurs in verse 68 and considers the category of argumentative quibbles (chala). These constitute a kind of fallacious response made by re-interpreting the meaning of one of the 17 ‘In this context, the succeeding cessation of the cause, which is the condition for the arising of the effect, is the characteristic of the immediately contiguous condition’. tatra kāraṇasyānantaro nirodhaḥ kāryasyotpādapratyayaḥ samanantarapratyayalakṣaṇam (de la Vallée Poussin 1903–13, 86.4). See Dhammajoti 2004, 131–2 for a discussion of definitions of samanantara-pratyaya in different systems of Abhidharma. 18 With reference to the ‘supporting condition’ (ālamba-pratyaya) Candrakīrti notes that it is merely conventional but not ultimate and therefore unproblematic (sāṃvṛtam etallakṣaṇam na pāramārthikamityadoṣaḥ (la Vallée Poussin 1903–13, 85.6)). Madhyamaka 125 discussion’s key terms. Vidyabhusana (1971, 65) gives the example of someone denying that a boy has a new blanket (nava-kambala) by arguing that he has only one blanket, on the basis of reading nava-kambala as ‘having nine blankets’.19 A reply based on a quibble can never be a successful move in an argument, independent of whether the point the quibbler raises in is fact true. This is because the quibbler changes the topic of debate: even if, as a matter of fact, the boy only has one blanket this does not have any relevance for deciding whether this blanket is new or old. Let us consider the passage in question.20 I have added some comments in italics. It is objected: Even though according to you all have been asserted as categories, they do not exist ultimately. There is therefore the fault of chala in the present debate since Nāgārjuna and his Nyāya opponent mean different things when they speak about categories. The Nayāyika thinks they are the fundamentally real things, the Mādhyamika that they have a merely conventional reality. But then such a discussion seems to be as pointless as one about the causes for the collapse of a bank, where one party is thinking about the collapse of a financial institution, and the other about the collapse of a riverbank. It is to be replied: That is not the case, because there would be this consequence for all replies. If this was really a problem then there could be never any discussion, philosophical or otherwise, where the two parties do not agree on the meaning of all the terms. But if agreement is to be presupposed, depending on how comprehensive we take the meaning of terms to be (if, for example, we assume that whether a term is taken to be instantiated is part of its meaning) it will then be unclear why one would have a discussion in the first place. If we agree on all facts about X, including whether or not X exists, what is there to discuss? This variety of quibble is called vāk-chala, quibble regarding a term. For other varieties see Vidyabhusana 1971, 65. 20 smras pa | khyod kyi thams cad tshig don du brjod pa yin gyi don dam par ma yin no zhe na brjod par bya ste ma yin te | lan thams cad la thal bar ‘gyur ba’i phyir ro | ‘di ni de lta bu ma yin te | gang cung zad lan du brjod pa de thams cad tshig don nyid du ‘gyur ba’i phyir ro | gang las she na | smra ba po thams cad ni ‘thad pas gdon mi za bar tshig nyams par byed pa yin no | de’i phyir lan du ‘thad pa yin no | tshig dor ba med pa’am | yang na tshul ‘di mi ‘dod na de’i phyir tshig dor ba med par ‘gyur ro | (Tola and Dragonetti 1995, 50). 19 Categorisation in Indian Philosophy 126 It is not like that, because whatever I addressed in a reply, all those are meant precisely as categories. This is an important point. Nāgārjuna’s criticism of the Nyāya categories does not concern their ontological status, i.e. the fact that the Naiyāyika believes them to be ultimate existents, while the Mādhyamika only grants them conventional existence. Nāgārjuna disputes the status of the Nyāya categories on the conventional level, that is as a system that describes a satisfactory systematization of the things which are in the world independent of any concerns about their ultimate nature. Why is that? All disputants, by agreeing, would undeniably impair their own assertions. If the two disputants agreed that the concept they debated about is not one, but rather two distinct concepts then their respective assumptions of having refuted the opinions of the other would be fallacious: all they have refuted is a straw man, since both have been talking about different concepts Therefore is what is said in reply not reasonable, rather than a fallacy? If the same concept is talked about, Nāgārjuna’s refutations have to be regarded as successful, unless they can be criticised on other grounds than that of chala. Otherwise, if one does not assert this principle then there is no fallacy. And if one does not assume that there are two distinct terms at issue then there is no risk of the fallacy of talking past each other in the first place. What is especially interesting about this passage is Nāgārjuna’s observation that his criticism of the Nyāya categories is not a general one, arguing that they mistakenly superimpose svabhāva onto things which really lack them and are therefore unsatisfactory from the perspective of ultimate truth. Nāgārjuna criticises them from the perspective of conventional truth, i.e. by arguing that the difficulties already arise at the level of producing a categorial theory of the world, not just when the resulting theory is interpreted in a particular way. This therefore seems to constitute a violation of NS which rejects analysis at the conventional level. The second passage to consider in our attempt to understand the precise impact of NA occurs at the end of his discussion of inconclusive reasons in verse 62 of the VP.21 Nāgārjuna here argues against the possibility of the Naiyāyika’s reason having any argumentative force, since the objects these 21 An inconclusive (savyabhicāra) reason is a reason which leads to several conclusions which might not be consistent with each other. See Vidyabhusana 1971, 63–4. Madhyamaka 127 reasons refer to are momentary (skad cig ma).22 The difficulty Nāgārjuna seems to have in mind here is that argumentative processes are spread out in time, whereas objects are not: they come into and go out of existence in an instance. As soon as the reason has established anything about the object it has already passed away, as a result of which central terms in the statement of the conclusion will not refer anymore. In the present context we are not primarily interested in the kind of response the Naiyāyika could give to this argument but rather in its presuppositions. Nāgārjuna’s argument rests on a premise which is universally accepted by Buddhists from the Abhidharma onwards, that of the momentariness of all existents.23 Universal momentariness is a conventional truth and it would be hard to argue that it was one acceptable according to CS, admitted by the world ‘by everyone from cowherds up’. There are therefore at least some cases in which the Mādhyamika criticises Nyāya arguments for conventional reasons, pitching two statements at the level of conventional truth against each other, one of which (that of universal momentariness) is regarded as superior to the other (that of temporally thick objects). Not only do we find that the Mādhyamikas use analytical tools in order to assess different systems of categories and to decide, there is also a marked difference to be observed in the way in which they deal with the categorial systems of the Abhidharma and those of Naiyāyikas. Whereas the former are regarded as acceptable as long as one does not take them seriously ontologically (that is, as long as one does not regard them as obtaining at the level of ultimate truth), the latter are deficient even at the level of conventional reality.24 Candrakīrti explicitly warns against incorporating such notions into one’s own system:25 It is not logically acceptable to include things postulated by the heretics, which are contrary to reason, into one’s own doctrine. gzhan yang | skad cig ma nyid kyi phyir yang ngo | don gang la bsgrub pa’am sun ‘byin par byed pa de la ‘jig pa yod pa yin na sgrub pa’am sun ‘byin par byed pa rigs pa ma yin no | gnyi ga gnas pa yin na gtan tshigs sgrub pa’am sun ‘byin par byed pa yin gyi de ‘jig pa na gtan tshigs de sgrub pa’am sun ‘byin par byed par rigs pa ma yin te | gnyi ga gnas pa med pa’i phyir ro| (Tola and Dragonetti 1995, 47–8). 23 How the Buddha’s doctrine of universal impermanence transformed into the doctrine of universal momentariness is a question which requires more research. For some discussion see von Rospatt 1995, 15–39, Ronkin 2005, 59–60. 24 Sthiramati, commenting on the final chapter of the Bodhicaryāvtāra agrees by noting that ‘what is misconstrued by the philosophical conclusions of the various heretics is false even from the point of view of a worldly person.’ (athyāsvaṃ tīrthikasiddhāntaparikalpitaṃ ca lokata eva mithyā, Vaidya 1988, 77: 15–16). 25 na ha tīrthikaparikalpitā yuktividhurāḥ padārthāḥ svasamaye’bhyupaganta.m nyāyāḥ (de la Vallée Poussin 1903–13, 67.1). 22 Categorisation in Indian Philosophy 128 He also points out the conventional falsity of non-Buddhist systems with specific reference to the Saṃkhya theory of the three guṇas:26 Like the intrinsic nature (bdag nyid) imputed (brtag pa) by the heretics completely carried away by the sleep of ignorance, what is imputed on illusions, mirages and so on, those also do not exist from a worldly perspective. These heretics desire to enter into reality and do not grasp the things which are unmistaken and certainly identified, known [by all] up to ignorant people, cowherds, women etc, [namely] arising, cessation etc. They desire for excellence [but] are like one climbing up a tree who, not having taken hold of the first branch, does not grasp the second; there is a great collapse by falling into the ravines of wrong views. Apart from perceiving the two truths there will be no obtaining of the result. Therefore their imputations, the three guṇas, etc. are certainly non-existent according to worldly conventions. This passage appears to be a clear endorsement of CS. Candrakīrti’s main worry with the Sāṃkhyas and related ‘heretic’ approaches is that they postulate metaphysical systems which are not compatible with the theories of conventional truth generally admitted by the world. In constructing a metaphysical theory it is necessary to base oneself on the truths allowed by CS, i.e. those commonly accepted by the world (the first branch) in order to reach the higher branches of absolute truth, a level at which coherence with what the world thinks is no longer required. In the context of this very critical attitude towards non-Buddhist philosophical systems it might strike us as surprising that Candrakīrti accepts the Nyāya quartet of epistemic instruments (pramāṇa), something which is rejected by Nāgārjuna both in the VV and the VP.27 But there need not be any difficulty with this as long as we keep two things in mind. First of all the Nyāya quartet can be conceived of mi shes gnyid kyis rab skyod mu stegs can | rnams kyis bdag nyid ji bzhin brtag pa dang | sgyu ma smig rgyu sogs la brtags pa gang | de dag ‘jig rten las kyang yod min nyid mu stegs can ‘di dag ni de kho na nyid la ‘jug par ‘dod pas skye bo ma byang ba gnag rdzi dang bud med la sogs pa yan cad la grags pa’i dngos po skye ba dang ‘jig pa la sogs pa phyin ci ma log par yang dag par nges par ma bzung bar | phul du byung bar ‘gro bar ‘dod pas shing la ‘dzegs pas yal ga snga ma btang nas yal ga phyi ma ma bzung ba ltar lhung ba chen pos lta ba ngan pa’i ri sul dag tu ltung bar ‘gyur zhing | bden pa gnyis mthong ba dang bral ba’i ‘bras bu mi thob par ‘gyur ro de’i | phyir ‘di dag gis gang brtags pa yon tan gsum la sogs pa de dag ni ‘jig rten gyi kun rdzob tu yod pa ma yin pa nyid do (de la Vallée Poussin 1907–12: 105.9–106.1). 27 ‘Thus the conventional understanding of things is established based on this fourfold scheme of epistemic instruments.’ tad evaṃ pramāṇacatuṣṭayāl lokasyārthadhigamo vyavasthāpyate (de la Vallée Poussin 1903–13, 75.9) PP 75.9. 26 Madhyamaka 129 without the Vaiśeṣika metaphysics later grafted on to it. To this extent Candrakīrti can regard it as a theoretically useful device without a great many philosophical implications. Secondly we should note that the property of the Nyāya quartet rejected by Nāgārjuna is their fundamental status, the existence by svabhāva of these instruments, while what Candrakīrti affirms is their conventional existence, the fact that these are pragmatically useful, not that they have a special ontological status. To this extent the Madhyamaka view of the categorial systems of Nyāya and Abhidharma appears similar: they are admissible on the conventional level as long as one does not take them seriously at the ultimate level. Of course ‘taking them seriously’ means different things in different cases: taking the dharmas of the Abhidharma seriously ontologically means claiming that they do not in turn depend on other things for their existence; taking the epistemic instruments of the Nyāya seriously means assuming that they have their own nature as instruments independent of the epistemic objects. But there are also cases where the Mādhyamika criticise the Nyāya categories without just focusing on the ontological status the Naiyāyikas ascribe to them which the Mādhyamikas reject. It would be mistaken to assume that Buddhist authors simply favoured the categorial system of the Abhidharma because, unlike that of the Naiyāyikas, it formed an established part of the Buddhist doctrine. Such partisanship is not supported by what these authors say themselves. They do not claim that the problem with the Nyāya categories is that they come from a non-Buddhist framework but that there is something conceptually wrong with them. From the Madhyamaka perspective there are in fact good systematic reasons for favouring the Abhidharma framework. Apart from cases where the Nyāya framework clashes with important Buddhist theses about the nature of conventional truth (such as the thesis of universal momentariness) the primary reason is the fact that non-Buddhist systems of categories are generally built around the notion of a substantial self. The Nyāya-Vaiśeṣikas hold, for example, that cognition (buddhi) is a property of the self (ātman) (Chatterjee 1978, 11). But such a self is not accepted by the Madhyamaka even conventionally. As such they are unable to adopt a system of categories which at various points appeals to objects they regard as non-existent. Conclusion I hope that the preceding discussion has shown that the Madhyamaka view of conventional truth cannot be that of unreflected intuitions held by everybody ‘from cowherds on up’.28 The Madhyamaka discussion employs sophisticated 28 Compare Salvini 2008, 16: ‘It is worth pointing out that here [in Candrakirti’s commentary on MMK 24:8] loka must necessarily refer to the “common person” (pṛthagjana), in the technical sense of someone who has not entered the Path of Seeing: in other words, someone who is not yet a Noble One (ārya). Alternatively, it must be a Noble 130 Categorisation in Indian Philosophy philosophical systems of categories (such as those of grammar and the Abhidharma) in the exposition of its arguments. Not all systems of categories are deemed equally suitable, that of the Naiyāyikas is rejected as unsatisfactory. If we consider the way much of the Madhyamaka discussion is set up it is clear that both NA and CS have to be understood in very specific ways. When Candrakīrti argues that the worldly person understands, without analysis, that effects arise from causes, he might just want to point out that there is no need for a philosophical theory over and above the recognition of regularities in order to make conventionally correct statements about causal processes that obtain in the world (De la Valleé Poussin 1903–13, 27: 4–5). For making such statements no analysis is required but that does not mean that, in the process of analysing conventional truth in the context of Madhyamaka arguments, one account of causation may not be preferable to another one. Furthermore, the ‘common sense’ in CS must not be reduced to what is apparent to the least inquisitive and least well-informed. That this kind of common sense has no place in Madhyamaka arguments is apparent from the sophisticated theoretical frameworks they do indeed employ in their arguments. Rather, ‘common sense’ has to be understood as ‘informed common sense’, the best theory we have for making sense of the world. Understood in this way commonsensical theories can be very refined and are also subject to revision. How are we then to understand the Madhyamaka conception of categories? It is evident that they cannot be regarded as ultimately true theories; their use is merely transactional. As such the Mādhyamika disagrees fundamentally with the understanding of the ontological as the ultimate furniture of the world according to some final theory. But this does not mean that the Mādhyamika is not able to refer to ontological categories at all but that he has to understand them in a different way. That the Madhyamaka ascribes a merely pragmatic use to systems of categories fits well with the fact that, unlike other schools of Buddhist thought, they never developed such a system of their own but rather utilised those already in existence. Due to the specific philosophical views of the Madhyamaka it was not possible, however, to use such an ‘off the shelf’ system without further qualification. The systems in question rather had to be rephrased in such a way that they did not clash with the most fundamental of Madhyamaka tenets, that of the absence of svabhāva. When the Mādhyamika speaks of Abhidharma categories, such as the five skandhas, the four great elements, the four kinds of condition and so on, he is in fact referring to a suitably modified Abhidharma from which the claim that there are substantially existent objects (dharmas which are dravyasat) has been removed. One engaging with saṁvṛti, not in meditative absorption on emptiness … What this means is that, the loka of the lokasaṁvṛti need not be an entirely unrefined person: in fact, even an Abhidharma author could fit this category, as long as he would not have entered the Path of Seeing, or even after that, when engaging in post-meditation’. Madhyamaka 131 It may therefore be plausible to assume that when the Mādhyamika deliberates between different systems of categories at the conventional level he is considering two main criteria: a) its general philosophical usefulness and b) the extent to which the systems can be made ‘svabhāva-free’. This allows him to rank such systems and provides criteria to choose between systems, none of which have any claim to truth at the absolute level. It is evident that there is no fundamental tension between the Madhyamaka enterprise and the construction of systems of categories and that this analysis is also not hindered by CS and NA properly understood. The Mādhyamika is able to use any such system of categories for putting forth his arguments as long it can be made to work without the assumption that the categories in play constitute necessary and intrinsic properties of the objects categorised.