Nagarjuna’s Madhyamaka

Nagarjuna’s Madhyamaka A Philosophical Investigation Jan Christoph Westerhoff Department of Philosophy University of Durham 50 Old Elvet Durham DH1 3HN United Kingdom westerhof f Q cantab.n et This dissertation is submitted for the degree Doctor of Philosophy at the School of Oriental and African Studies, University of London. O c t o b e r 2007 l ProQuest Number: 10731457

INFORMATION TO ALL USERS The quality of this reproduction is d e p e n d e n t upon the quality of the copy subm itted. In the unlikely e ve n t that the author did not send a c o m p le te manuscript and there are missing pages, these will be noted. Also, if m aterial had to be rem oved, a n o te will indicate the deletion. uest ProQuest 10731457 Published by ProQuest LLC(2017). C opyright of the Dissertation is held by the Author.

789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48 10 6- 1346

Abstract

This dissertation constitutes a discussion of Nagarjuna’s Madhyamaka as contained in his six main philosophical works. It presents a synoptic presentation of the main topics Nagarjuna investigates. Particular emphasis is put on an analysis of the philosophical content of Nagarjuna’s Madhyamaka. A part from discussing the soundness of Nagarjuna’s arguments for particular conclusions I also want to examine to which extent Nagarjuna’s philosophy forms a coherent philosophical system rather than a collection of individual ideas. The dissertation consists of four parts. In the first part (chapter 2) I discuss the central concept of Madhyamaka philosophy, the notion of svabhava. This is a notion of considerable complexity; for the purposes of understanding Nagarjuna’s arguments I argue that it is particularly important to distinguish two of its conceptual dimensions: an ontological and a cognitive one. The second part (chapters 3 and 4) discusses some properties of the form of Nagarjuna’s arguments, properties which are, however, also intricately connected with their contents attempting to establish the Madhyamaka theory of emptiness. The two topics investigated are the place of negation in Nagarjuna’s philosophical assertions and his use of the argumentative framework known as the catuskoti or tetralemma. An analysis of the background of these formal aspects is indispensable for an understanding of Nagarjuna’s

arguments presented in the following chapters. The third part (chapters 5 to 9) discusses Nagarjuna’s arguments dealing with particular topics, such as causation, motion, the self, epistemology, and language. Here Nagarjuna sets out to establish the absence of svabhava in areas which are particularly central to our cognitive interaction with the world. He investigates both the world around us (for the examples of causation and motion), the subjective world (the self) as well as the way in which the two are connected (by our epistemic faculties and by language). The final fourth part (chapter 10) attem pts to present a concise synoptic overview of Nagarjuna’s conclusions described in the preceding chapters and sets out to evaluate them from a systematic point of view. I also discuss how these various conclusions form a coherent philosophical whole and attem pt to evaluate some of them in the light of the contemporary philosophical discussion. A cknow ledgem ents I would like to thank my supervisors Ulrich Pagel and David Seyfort Ruegg for their advice and guidance in putting together this dissertation. Their knowledgeable comments and criticism provided many insights and helped me to avoid numerous errors. Thanks are also due to the following people who contributed in one way or another to the completion of this work: Naomi Appleton, Yoshihiko Arita, Jay Garfield, Yuka Kobayashi, Ralf Kramer, Jowita Kramer, Alexandra Leduc, Charles Ramble, Ulrike Roesler, Sherab Gyatso, M attia Salvini, and Tom Tillemans. Whilst working on this dissertation I received some support from the Spalding Trust and the Society for South Asian Studies which is gratefully acknowledged. J.C.W. 5 Contents Preface 2 A bstract 3 A cknow ledgem ents 5 A bbreviations 10 1 Introduction 13 1.1 Nagarjuna the p h ilo so ph er............................................................. 15 1.2 Nagarjuna’s works .......................................................................... 16 1.3 The philosophical study of Nagarjuna in the W e s t ................... 19 1.4 Overview.............................................................................................. 24 2 Interpretations of svabhava 34 2.1 The ontological dimension ............................................................. 35 2.1.1 Essence- svabhava ................................................................. 36 2.1.2 Substance- svabhava..................................................................40 Dependence relations ...............................................................45 The rejection of substance-svabhava................................. 50 The property argum ent....................................................... 53 The mereological argum ent................................................. 59

CONTENTS 7 The argument from change................................................. 62 2.1.3 Absolute s v a b h a v a ............................................................. 66 Ames’ solution .................................................................... 68 Tsong kha pa’s solution .................................................... 69 Absolute svabhava as essence-s v a b h a v a .......................... 71 2.2 The cognitive dime n s io n .................................................................... 75 3 N egation 84 3.1 Nyaya theory of negation................................................................ 85 3.2 Negation and non-denoting t e r m s ................................................ 88 3.3 Negation and temporal relatio n s..................................................... 100 4 T he catuskoti or tetralem m a 104 4.1 Two kinds of n e g a tio n ......................................................................106 4.2 Rejection of two alternatives............................................................ 109 4.3 Rejection of four alternatives............................................................ 113 4.3.1 Distinctness of the four alternatives...................................115 4.3.2 The status of the third alternative...................................... 124 4.3.3 Rejection of four alternatives: the case of relations . . . 133 4.4 Affirming four alternatives: the positive te tra le m m a .................. 137 5 C ausation 141 5.1 Causation: preliminary re m a rk s ......................................................143 5.2 Interdependence of cause and e f f e c t............................................... 146 5.3 The four ways of causal p ro d u c tio n ............................................... 153 5.3.1 S elf-causation.......................................................................... 154 Identity of cause and effect................................................... 154 Cause and effect related as part and w h o le ...................... 157 5.3.2 Causation by another o b je c t................................................ 160 CONTENTS 8 5.3.3 Causation by itself and another o b j e c t .......................... 168 5.3.4 Absence of c a u s a tio n .......................................................... 171 5.3.5 Identity, difference, and sv a b h a v a ....................................173 5.4 Temporal relations between cause and e f f e c t................................175 5.4.1 Cause and effect as successiv e............................................ 176 5.4.2 Cause and effect as o v erla p p in g ......................................... 182 5.4.3 Cause and effect as simultaneous ...................................... 184 5.4.4 Cause and effect as one e v e n t .............................................187 5.4.5 Temporal relations and s v a b h a v a ...................................... 188 5.5 Analysis of tim e ................................................................................... 191 6 M otion 195 6.1 Arguments concerning m o tio n .........................................................197 6.1.1 The property-absence a rg u m e n t......................................... 198 The ‘mathematical’ in terp retatio n ......................................201 6.1.2 The property-duplication a rg u m e n t...................................204 6.2 The beginning of m o tio n .................................................................. 212 6.3 The interdependence of mover and m o t io n .................................. 220 6.4 The second chapter of the MMK in its argumentative context 224 7 T he self 228 7.1 The self and its p a r t s ......................................................................... 230 7.2 The self and its p ro p e rtie s ............................................................... 236 7.3 Epistemology of the s e l f ...................................................................239 7.4 The Madhyamaka view of the s e l f ...................................................241 8 E pistem ology 245 8.1 Means of knowledge as self-estab lish ed ......................................... 248 8.1.1 Means of knowledge compared to f i r e ................................249 CONTENTS 9 8.1.2 The problem of the independence of means of knowledge from their objects255 8.2 Means of knowledge and their objects as mutually established 256 8.3 Temporal relations between means and objects of knowledge . 262 8.4 The aim of Nagarjuna’s arguments ............................................. 266 9 Language 270 9.1 Nagarjuna’s view of language and the ‘no-thesis’ v ie w ................ 270 9.2 VV 29 in c o n t e x t....................................................................... 272 9.3 The semantic in te rp re ta tio n .......................................................... 277 9.4 The specific role of verse 29 .......................................................... 286 10 C onclusion 293 10.1 M etaphysics........................................................................................293 10.2 Personal I d e n t i t y ..............................................................................306 10.3 E t h i c s ................................................................................................. 312 10.4 E p istem o lo g y .................................................................................... 318 10.5 Language and T r u t h ....................................................................... 322 B ib liograp h y 331 A bbreviations References to works in verse give the number of the chapter and verse or halfverse (e.g. MMK 24:18, RA 1:49a), apart from works not usually subdivided into chapters, such as the YS, SS, VV, and VP, where only the number of the verse is given. References to works in prose give the page and line of the editions cited below (e.g. VV(S) 82:3-7). W o r k s b y N a g a r j u n a MMK Mulamadhyamakakarika J.W. De Jong, Christian Lindtner (eds): Nagarjuna’s Mulamadhyamakakarika Prajha Nama. Adyar Library, Adyar, Chennai, 2004. YS Yuktisastika Cristina Anna Scherrer-Schaub: Yuktisastikavrtti: Commentaire d la soixantaine sur le raisonnement, ou, du vrai enseignement de la causalite. Institut beige des hautes etudes chinoises, Brussels, 1991. SS Sunyatasaptati ‘The Sunyatasaptati of Nagarjuna’, chapter 3 of Fernando Tola, Carmen Dragonetti: On Voidness. A Study of Buddhist Nihilism, Motilal Banarsidass, Delhi, 2002, 53-99. 10 CONTENTS 11 VV VigrahavyavartanT VV(S) Svavrtti on the VigrahavyavartanT Kamaleswar Bhattacharya, E.H. Johnston, Arnold Kunst (eds): The Dialectical Method of Nagarjuna. Motilal Banarsidass, Delhi, 1998. VP Vaidalyaprakarana VP(S) Svavrtti on the Vaidalyaprakarana Fernando Tola, Carmen Dragonetti: Nagarjuna !s Refutation of Logic. Motilal Banarsidass, Delhi, 1995. RA RatnavalT Michael Hahn: Nagarjuna’s Ratnavah. The Basic Texts (Sanskrit, Tibetan, Chinese). Indica et Tibetica, Bonn, 1982. W o r k s b y o t h e r a u t h o r s PP Candralurti’s Prasannapada Louis de la Vallee Poussin (ed.): Prasannapada Mulamadhyamakavrttih. Bibliotheca Buddhica IV, St Petersburg, 1903-1913. J.W. De Jong: ‘Textcritical notes on the Prasannapada\ Indo-Iranian Journal, 20, 1978, 25-59, 217-252. (Corrections of the above edition.) MA Candraklrti’s Madhyamakavatara MAB Candraldrti’s Madhyamakavatarabhasya Louis de 1a. Vallee Poussin (ed.): Madhyamakavatara par Candrakirti Bibliotheca Buddhica IX, St Petersburg, 1907-1912.

CONTENTS 12 CS Aryadeva’s Catuhsataka Karen Lang (ed.): Aryadeva’s Catuhsataka. On the Bodhisattva’s Cultivation of Merit and Knowledge. Akademisk Forlag, Copenhagen, 1986. BCA Santideva’s Bodhicary avatar a Louis de la Vallee Poussin (ed.): Bodhicarydvatdra of (fantideva, Asiatic Society of Bengal, Calcutta, 1907.

NS Nyayasutra Taranatha Nyaya-Tarkatirtha, Amarendramohan Tarlcatirtha (eds): Nyayadarsanam with Vatsyayana’s Bhasya, Uddyotakara’s Vartika, Vdcaspati M isra’s Tatparyatika and Visvanatha’s Vrtti, Munshiram Manoharlal, Delhi, 1985, Chapter 1

Introduction

The following pages contain an investigation of Nagarjuna’s philosophy from a systematic perspective, Considering Nagarjuna’s important place in Buddhist philosophy as well as in Indian thought more generally it is not surprising th at his works have given rise to an enormous number of commentaries, studies and analyses in Asia, and, more recently, also in the West. A large amount of these takes the form of commentaries on specific texts, following their structure and analysing individual passages in considerable detail. While the importance and usefulness of such commentaries is beyond dispute, the present work sets out to approach Nagarjuna’s philosophy from a different perspective. The idea is to present a synoptic overview of Nagarjuna’s arguments concerning different philosophical problems in order to present an account of the whole of his philosophy, showing how its individual parts fit together as elements of a single philosophical project. In order to achieve this it is not sufficient to give a mere paraphrase of Nagarjuna’s arguments (as is frequently found in the secondary literature). We will have to analyse their philosophical contents, examine actual as well as possible objections, determine whether the arguments can in fact be made to work, and, if so, what

CHAPTER 1. INTRODUCTION 14

kind of philosophical conclusion they support. Comparatively little work has been done in this direction. Since a great part of the contemporary Western studies of Nagarjuna are primarily interested in philological, historical or religious aspects of his works, genuinely philosophical studies have been rare. The aim of the present study is to help closing this gap. The following pages should be of interest both to philosophers looking for a systematic account of Nagarjuna’s philosophical position, as well as to Indologists and scholars of Buddhist studies interested primarily in the philosophical aspects of Nagarjuna’s works. To make this material as accessible as possible to readers with little or no background in Indian philosophy 1 generally use English equivalents of technical Indian philosophical terms (such as ‘object’ for dharma, ‘emptiness’ for sunyatd, ‘primary existence’ for dravyasat and so forth), providing the Sanskrit term in brackets if necessary. The only case where I have systematically violated this policy concerns the term svabhava. My reason for this is that there is no single term used in Western philosophy which covers the different aspects of its meaning in the Madhyamaka context in a satisfactory manner. But given that all of chapter 2 is dedicated to a discussion of how we are to understand the notion of svabhava the reader should have a sufficiently clear conception of its meaning when he encounters it again in later chapters. For the benefit of Indologists and Buddhist scholars the Sanskrit and Tibetan (and occasional Pali) of all quotations is given in the footnotes. Some material and references in the footnotes will be particularly relevant to philosophers, some are of more historical or philological interest. I have made no attem pt to differentiate the ‘philosophical’ and ‘Indological’ footnotes but trust in the reader’s discernment to find the material which interests him. Different kinds of readers might prefer different routes through the maCHAPTER 1. INTRODUCTION 15 terial presented here. Those interested in a step-by-step introduction to Nagarjuna’s philosophy should read the chapters in numerical order. Readers with previous acquaintance with Madhyamaka material interested in what I have to say on a particular Nagarjunian topic will prefer to go directly to the relevant chapter. For those wanting to get straight at the philosophical content I recommend finishing the introduction, followed by chapter 2, then immediately jumping to chapter 10. After this it is possible dip into any of chapters 3 to 9 for more specific discussion of topics one finds interesting.

I.1 N agarjuna the philosopher

Nagarjuna, one of the greatest thinkers in the history of Asian philosophy remains an enigma. Despite the existence of various legendary accounts of his life passed down in Buddhist literature1 contemporary scholars agree on hardly any details concerning him. It is unclear when he lived (although some time during the first three centuries A.D. is most likely),2 where he worked (almost all places in India have been suggested),3 what he wrote (the Tibetan canon attributes 116 different texts of very diverse content and quality to him),4 and even how many Nagarjunas there were in the first place (up to four different ones have been distinguished).5 1Walleser (1923), Dowman (1985). 2Mabbett (1998, 332). For an extensive list of references see Ruegg (1981, 4-6, note 11).

3Walser (2005, 67). 4A similarly high number of works in listed in the Chinese canon (Lindtner 1982, 9). 5The most common view distinguishes only three Nagarjunas: the Madhyamaka philosopher, the tantric adept, who possibly flourished around 400 A.D.(Lindtner 1982, II, note 12) and the alchemist, who might be placed in the seventh century (Walser 2005, 69, 75-79), (Eliade 1969, 415-416). For criticism of the thesis of multiple Nagarjunas see Hua (1970). CHAPTER 1. INTRODUCTION 16 Recent research by Joseph Walser suggests that Nagarjuna may have written the Ratnavali sometime between 170 and 200 A.D. in the area around present-day AmaravatT.6 This conclusion is based on two facts. Firstly there is a variety of evidence connecting Nagarjuna with the Satavahana dynasty.7 This is not very helpful on its own, as this dynasty spanned several centuries. However, in verse 232 of the Ratnavali Nagarjuna mentions a depiction of the Buddha sitting on a lotus (padmapitha). Given that such images were only available during the late part of the dynasty in the Eastern Deccan Walser comes to the tentative conclusion that Nagarjuna composed the text during the reign of king Yajna Sri Satakarni (about 175 to 204 A.D.).8 Of course none of this can be regarded as hard evidence, especially as the necessary detour via art history (in order to find the earliest date for the type of depictions of the Buddha Nagarjuna describes) introduces a whole new range of complexities and uncertainties. Nevertheless, given our present inability to find out the time and place of Nagarjuna in any other way determining them approximately based on the support of a variety of historical data such as suggested by Walser is surely to be preferred to not determining them at all. 1.2 N agarjuna’s works Assuming we resolve the uncertainty about Nagarjuna’s time and place by locating him in the second century A.D. in the Eastern Deccan, how do we deal with the multitude of works ascribed to him? Our present investigation will be based primarily on six of Nagarjuna’s works: 6Walser (2005, 61). 7Walser (2005, 293, note 26). 8Walser (2005, 86). CHAPTER 1. INTRODUCTION 17 1. the ‘Fundamental Verses on the Middle Way’ (Mulamadhyamakakarika, MMK) 2. the ‘Sixty Stanzas on Reasoning’ ( Yuktisastikd, YS) 3. the ‘Seventy Stanzas on Emptiness’ (Sunyatasaptati, SS) 4. the ‘Dispeller of Objections’ ( VigrahavyavartanT VV) 5. the ‘Treatise on Pulverization’ ( Vaidalyaprakarana, VP) 6. the ‘Precious Garland’ (Ratnavali, RA) This set, the so-called Yukti-corpus, is well-known in the Tibetan tradition where is it called the ‘collection of the six texts on reasoning’ (rigs p a ’i tshogs drug).9 We cannot be assured that all six texts were indeed composed by Nagarjuna; apart from the MMK, where Nagarjuna’s authorship is taken to be true by definition, the attribution of every other one has been questioned. 10 However, apart from the fact that all these texts were attributed to Nagarjuna by a variety of Indian11 and Tibetan Madhyamaka authors they also expound a single, coherent philosophical system. For the purposes of this discussion we will therefore identify Nagarjuna with the author of the Yukti-corpus.

9See e.g. Bu ston’s History of Buddhism (Obermiller 1931, I, 50-51). 10Warder (1973, 79) notes that the authorship of Nagarjuna for texts other than the MMK ‘has not been established beyond doubt and we ought not to assume it1. For comments on the authorship of the YS see Tola and Dragonetti (1995b, 19-20), for the SS Tola and Dragonetti (1995b, 54-55). The attribution of the VV is questioned in Tola and Dragonetti (1998) (but see Ruegg (2000, 115, note 10)), that of the VP in Tola and Dragonetti (1995a, 7-15) and Pind (2001). For the RA see Walser (2005, 271-278). ^Including Bhaviveka, CandrakTrti, Santaralrsita and KamalasTla. See Lindtner (1982, 10-11, note 9). CHAPTER 1. INTRODUCTION 18 The texts under consideration are all written in verse. In some cases they are accompanied by an autocommentary in prose; though the status of these autocommentaries is not always unproblematic.12 As the present inquiry is intended to be a study of Nagarjuna’s Madhyamaka the texts of the Yukticorpus constitute the basis of our discussion. It is nevertheless not possible to provide a philosophically satisfactory exposition of Nagarjuna’s thought based exclusively on these texts. This is because their versified form often leads to a very condensed expression of arguments which requires a variety of details to be filled in. In itself this is hardly surprising given th at Indian philosophical texts (unlike their Western counterparts) were generally not intended to provide the reader with a self-contained exposition of the author’s thoughts. Instead their versified form provided the structure of the argument to be memorized, which would then be elaborated by written commentaries and by a teacher’s oral explanations. The reader of Nagarjuna’s works will frequently encounter passages in which Nagarjuna asserts that a certain position is deficient and ought not to be accepted, without giving the reason why this is the case. Sometimes this difficulty can be solved by considering the remainder of Nagarjuna’s texts. In other cases it is necessary to fill in the missing details in order to arrive at a philosophically coherent account of Nagarjuna’s arguments. Such filling in (much like the filling in of missing parts when restoring a painting) always involves an element of conjecture. In these cases I consulted the Indian, Tibetan, and contemporary Western comment arial literature in order to reconstruct the missing step in the argument 12Despite substantial Indian support the auto commentary on the MMK, the Akutobhaya is not regarded as genuine by contemporary researchers (Lindtner 1982, 15-16, note 33). The auto commentary on the SS is regarded as authentic by Lindtner (1982, 31), Tola and Dragonetti (1995b, 57-58) disagree. There seems to be no dispute about the authenticity of the autocommentaries on the VV and VP. CHAPTER 1. INTRODUCTION 19 in a way which would be likely to be acceptable to a Madhyamika. My way of determining such acceptability is to consider both systematic and doctrinal coherence. To achieve systematic coherence I attem pted to fill in the missing details in the argument in such a way that it can be made to work. Doctrinal coherence is aimed at by avoiding any assumptions which disagree with the bulk of Madhyamaka literature. As Madhyamaka authors present different opinions when interpreting Nagarjuna (especially when dealing with the more subtle points) doctrinal coherence is to be regarded as a guiding principle, rather than as a goal which can be realized completely. In the cases where this principle did not provide sufficient guidance the reconstructions presented are those which I regard as philosophically most successful. As we have no way of determining the manner in which Nagarjuna himself intended to spell out the argument a filling-in based on the methodological principles just mentioned appears to be the most we can hope for in a philosophical exposition of Nagarjuna’s thought.

1.3 The philosophical study of Nagarjuna inthe West

Western interest in Nagarjuna as a philosopher is a comparatively recent phenomenon, going back little more than a century.13 In itself this constitutes only a part of Nagarjunian scholarship, a substantial portion of which concerns itself with problems of philology, textual history or the study of 13The earliest systematic Western treatment of Madhyamaka more generally is to be found in the works of the Jesuit missionary Ippolito Desideri (1684-1733). Desideri published a number of works in Tibetan in which he attempted a refutation of Tibetan Buddhism from the perspective of Roman Catholicism. See Desideri (1981-1989).

CHAPTER 1. INTRODUCTION 20

religion. A concise overview of the specifically philosophical investigation of Nagarjuna in the West has been presented by Andrew Tuck.14 Tuck argues th at its history can be divided into three phases, corresponding to three Western philosophical frameworks against which Nagarjuna used to be interpreted. First there is the Kantian phase, then the analytic phase, which was finally succeeded by a Post-Wittgensteinian one.15 A clear example from the first phase is Theodore Stcherbatsky’s The Conception of Buddhist Nirvana which was first published in 1927.16 Stcherbatsky’s interprets Nagarjuna as dividing the world into appearance and reality, the former corresponding to samsara, the realm of cyclic existence, the latter to nirvana, liberation. In his attem pt to defend Nagarjuna against the charge of nihilism, especially clear in the exposition given by La Vallee Poussin,17 Stcherbatsky ascribes to Nagarjuna the assumption of an absolute noumenal reality which underlies the constantly changing and ephemeral world of phenomena. The further development of this Kantianization of Nagarjuna is presented in what is still a basic texts of Buddhist studies, T.R.V. M urti’s 1955 The Central Philosophy of Buddhism.18 As M urti’s exposition of Nagarjuna is considerably more detailed than Stcherbatsky’s the fundamental difficulties of interpreting Nagarjuna according to a Kantian framework become more readily apparent. He observes th at ‘the relation between the two [i.e. the Absolute and the world of phenomena] is not made abundantly clear. This may be said to constitute a drawback in the Madhyamika conception of the Absolute.’19 On i4 (1990). 15(1990, 16-30). 16Stcherbatsky (1968). 17(1908, 101). 18<The position occupied by the Madhyamika in Indian philosophy is similar to that of Kant in modern European philosophy [...]’ (Murti 1955, 123). 19(1955, 237).

the other hand we might want to argue that the reason for this ‘drawback’ is not a defect in the Madhyamaka position but in the choice of interpretative framework. To conceive of Nagarjuna’s Madhyamaka as a theory where an indescribable Absolute grounds the world of appearances means reading assumptions into it which Nagarjuna does not share, thereby resulting in an unsatisfactory theory. The second, analytic phase of Western studies of Nagarjuna can be regarded as starting shortly after the publication of M urti’s book, with Richard Robinson’s 1957 article ‘Some Logical Aspects of Nagarjuna’s System’.20 He sets out to analyse some of Nagarjuna’s arguments using the resources of modern symbolic logic, the ultimate aim being ‘to transcribe the Kdrikas entirely, chapter by chapter, into logical notation, thus bringing to light formal features which do not appear from the consideration of examples taken out of context and listed topically’.21 The shift from the Kantian to the analytic reading of Nagarjuna which Robinson’s paper inaugurates brings with it a shift in the aspects of his thought receiving most attention. The focus is shifted from an investigation of the primarily metaphysical problem of the relation between samsara and nirvana to the logical aspects of Nagarjuna’s thought: his use of quantification and negation as well as the mechanics of the notorious form of argument known as the ‘tetralemma’ (catuskoti). If one considers the bigger picture, however, once again the limitations of the reading of Nagarjuna during the analytical phase become apparent. Many of his views, concerning, for example, the rejection of a foundationalist ontology or the difficulties of assuming a world conforming to the structure of the language we use to refer to it contradict assumptions of analytic phi20Robinson (1957). 21 (1957, 307).

losophy of the first half of the twentieth century. While the employment of certain tools dear to analytic philosophers could be seen as presenting Nagarjuna’s arguments more clearly it was also evident that Nagarjuna would have had little regard for the goals aimed at by analytic philosophers. Neither the attem pts to develop a logically perfect language for describing the world nor that or grounding our knowledge of the world on the supposedly secure foundation of sense-data could find much favour with Madhyamikas. Analytic philosophy with its specific set of philosophical assumptions was helpful in trying to understand Nagarjuna, but only up to a point. In fact by a rather curious turn in the third, Post-Wittgensteinian phase of interpreting Nagarjuna the paragons of analytic philosophy were now identified which Nagarjuna’s opponents, such as the Abhidharmikas and Naiyayikas.22 Works such as Frederick Streng’s Emptiness23 or Chris Gudmunsen’s Wittgenstein and Buddhism24 set out to stress the close similarities between Nagarjuna and in particular the later Wittgenstein and his criticism of analytic philosophy. While the relation between samsara and nirvana had been the chief concern for the Kantian readers of Nagarjuna, and that of the logical consistency of svabhava or substance for analytic interpreters such as Robinson25 the new key term of the Post-Wittgensteinian phase was pratltyasam utpada or dependent origination. This was regarded primarily as reflecting the underlying idea of a Wittgensteinian philosophy of language according to which language, and in particular the language of philosophical statements could not be regarded as independent of the interrelated nature 22Tuck (1990, 78). 23 (1967). 24'there is not nearly as much difference in the roles of Wittgenstein and Nagarjuna as one might imagine’ (1977, 68). 25Robinson (1967, 41).

of conceptual thought and conventional language. Words were not supposed to gain their meaning by referring to something outside of the system of language; the relation of words to their referents is not seen as being indicative of ontological status but is solely of practical value. Looking at the way in which the Western study of Nagarjuna was influenced by the philosophical fashions of the day we may be worried that work following the Post-Wittgensteinian phase will later appear to be a similar example of trying to shoe-horn Nagarjuna’s thought into a fundamentally alien framework. While it is certainly not possible (nor indeed desirable) to proceed with our investigation and to leave behind our specific interests, expectations and concerns there is no reason to panic. The fact that each interpretation takes place against a specific conceptual framework does not mean that successive interpretations might not lead to a deeper and more comprehensive understanding of Nagarjuna’s thought. In fact the literature published over the last decades suggests that the study of Nagarjuna is becoming more m ature.26 First of all most authors now try to treat Nagarjuna’s writings as expressing a single, unified system of thought, rather than as a quarry of cryptical verses from which individual isolated samples can be extracted to suit one’s idiosyncratic interpretation. There finally appears to be an agreement th at any interpretation of Nagarjuna should cohere with his assertions in all the works which can be plausibly ascribed to him. Secondly, and more importantly, it has become evident th at Nagarjuna is worthy of philosophical investigation in his own right. There is no more 26Amongst the most philosophically sophisticated contemporary commentaries on Nagarjuna’s texts the works by Kalupahana (1991), Tola and Dragonetti (1987), (1995a), (1995b), Garfield (1995), and Bugault (2001) have to be mentioned. Some of the best monographs are Huntington (1989) and Siderits (2003) (and, to a lesser extent, Wood (1994) and Burton (1999)). CHAPTER 1. INTRODUCTION 24 need to legitimate a study by setting out to show him to be a proto-Kant, proto-Wittgenstein or proto-Derrida. While such comparisons may be of hermeneutic use for those primarily acquainted with the Western tradition most writers on the topic now agree th at it is no more necessary to put on a Kantian lens to understand Nagarjuna than it is to wear a Nagarjunian lens to understand Kant. Therefore, even though we cannot interpret Nagarjuna free of the preconceptions and concerns of our own time we are justified in expecting th at the more mature study of his works will provide us with more accurate and stable knowledge of his philosophy than was previously possible.

1.4 Overview

Even a casual acquaintance with Madhyamaka literature makes it evident that the central philosophical concept discussed is that of emptiness (sunyata). The main difficulty in explaining what this concept means is th at it is a purely negative one: emptiness is the emptiness of something and indicates that something is not there. This absent something is what the Madhyamaka authors refer to by the term svabhava, sometimes translated as ‘inherent existence’ or ‘own being5. For this reason the term ‘emptiness’ is often glossed as ‘empty of inherent existence’ (svabhavasunya). A good way of understanding the Madhyamaka notion of emptiness is therefore to provide a clear conception of what they mean by svabhava. This is what chapter 2 sets out to do. Even if we restrict ourselves to Nagarjuna’s Madhyamaka, svabhava turns out to be a very complex concept. It unifies two very different aspects, an ontological and a cognitive one. The ontological aspect of svabhava is the

one discussed in most detail in the contemporary commentarial literature. The basic idea here is that an object has svabhava if it possesses its nature in an intrinsic manner. In order to spell out this still rather imprecise idea we have to differentiate three distinct ontological understandings of svabhava, all of which play some role in Nagarjuna’s arguments. The first is the understanding of svabhava as essence, as a property an object cannot lose without ceasing to be that very thing, the second an understanding as substance, as something which does not depend on anything else, and the third as what I have called absolute svabhava, as a property which is regarded as the true or final nature of things.

Even though I argue that understanding svabhava as substance occupies the most im portant place in Nagarjuna’s arguments we would be ill-advised to regard svabhava simply as some variant of the concept of substance found in the Western philosophical tradition. This is because svabhava has an important additional cognitive component which is completely absent from the concept of substance as it is usually conceived. The notion of svabhava is regarded as a conceptual superimposition, as something which is autom atically projected onto a world of objects which actually lack it. Unlike the notion of substance, svabhava is not just a theoretical concept of ontology, but a cognitive default, an addition the mind unwittingly makes when trying to make sense of the world. This cognitive understanding of svabhava makes clear why Madhyamaka, metaphysics (unlike metaphysics in the Western tradition) is not a purely theoretical enterprise but something which also has to be put into practice. If svabhava is an automatic mistaken superimposition we cannot just get rid of it by going through arguments attem pting to show that svabhava does not exist. We will also have to train ourselves out of the automatic habit of projecting svabhava onto a world which lacks it. CHAPTER 1. INTRODUCTION 26 This point can be illustrated by considering two different ways of studying higher-dimensional geometry. It is, for example, possible to prove various facts about a four-dimensional cube without having any idea of what such a cube would look like. We simply regard it as a theoretical entity which is defined in a certain way, and then proceed to prove further facts on the basis of this definition. On the other hand we could also try to develop a spatial intuition for the fourth dimension, that is try to get an idea of what such a cube would look like. We could, for example, imagine the ways in which a two-dimensional creature living on a plane could form the conception of a cube by extrapolating from a square and similarly try to extrapolate a four-dimensional cube from a three-dimensional one. It is evident th at this latter attem pt at enlarging our spatial intuition is not just about proving theorems, but requires certain exercises for enlarging our imagination.27 In the same way for the Madhyamaka the removal of the superimposition of svabhava is not just about working through philosophical arguments, but also requires certain exercises to effect a cognitive shift which keeps the mistaken projection of svabhava from occurring. A great part of Nagarjuna’s writings consists of the investigation of individual phenomena in order to argue that they do not exist with svabhava. Before we can turn to the examination of these arguments, however, it is necessary to deal first with some formal aspects of Nagarjuna’s arguments. I call these aspects ‘formal’ because they all have to do with negation, which is generally regarded as a formal notion. Nevertheless this adjective is also somewhat misleading if one regards as ‘formal’ those aspects of an argument which are independent of its content. For in the discussion of Nagarjuna 27The Victorian mathematician Charles Howard Hinton spent considerable time developing such exercises. See e.g. Hinton (1904).

the point is precisely that there are certain presuppositions made by the traditional Indian theory of negation which conflict with the contents of his philosophical conclusions. In order to formulate his philosophy Nagarjuna must therefore come up with an adapted conception of negation which counters these presuppositions. The main difficulty involved here (which is addressed in chapter 3) is the assumption of the Naiyayikas, who elaborated the standard Indian account of negation, th at the constituents of negative statements must always refer to real entities. A statement like 'there is no p o t’ is always to be understood along the lines of 'there is not pot at a particular place’. In this case both the pot and the place exist, it is only that the former does not occur at the latter. But Nagarjuna obviously cannot interpret his statement ‘there is no svabhava' along these lines, as he does not want to assert that svabhava is a real entity existing anywhere else. The second im portant formal issue, taken up in chapter 4, is the wellknown catuskoti or tetralemma. This is a rather puzzling form of argument, frequently employed by Nagarjuna, which consists of the rejection of four positions: a statement, its negation, their conjunction, and their disjunction. An important prerequisite for making sense of the tetralemma is to realize that the various negations occurring in it are not all of the same type. Some are implicational negations (paryudasapratisedha) which make an assertion about the object referred to ('the apple is not red’ implies that it is some other colour), while others are non-implicational (prasajyapratisedha) and do not make such an assertion (‘the force of gravity is not red’ does not imply that it is some other colour). Once the interrelations between these two kinds of negation have been taken into account it becomes clear th at Nagarjuna uses this form of argument in order to reject all the possible alternative

statements one can make about an entity on the assumption that it exists with svabhava. If all these alternatives turn out to be inapplicable we can conclude th at the initial assumption was wrong and that there is no svabhava to be found in th at entity. Having examined these formal aspects of Nagarjuna’s arguments (which are intricately connected with the contents of his philosophy) and taking into account the clarification of the different aspects of the notion of svabhava the reader will be able to understand Nagarjuna’s discussions of the emptiness of specific kinds of phenomena without too much difficulty. Chapters 5 to 9 deal with the main topics Nagarjuna analyses in order to demonstrate their emptiness, th at is, lack of svabhava. All of these play a major part in our cognitive interaction with the world and therefore constitute areas where we are particularly likely to mistakenly superimpose the existence of svabhava on phenomena which in fact lack it. Chapter 5 deals with the central notion of causation. Nagarjuna’s analysis concentrates on two aspects, that of the identity or difference between cause and effect, and th at of their temporal relation. Each can be spelt out in different ways. We can assume that cause and effect are fundamentally the same thing, or th at they are different, or th at are related as part and whole. Similarly the cause can be regarded as preceding the effect, as following it, or as being simultaneous with it. By rejecting all these different ways of conceiving of cause and effect Nagarjuna attempts to demonstrate that our underlying assumptions about causation are deficient. Causation is not a mind-independent, objective relation which connects objects ‘which are there anyway’. It is rather something which would not exist without a substantial mental contribution; it is a conceptually constructed relation which would not exist without the conceptualizing mind. This also entails th at the objects CHAPTER 1. INTRODUCTION 29 connected by such a relation cannot exist in a mind-independent way since their existential dependence on a cause holds via a relation which is not itself mind-independent. Chapter 6 deals with the concept of motion. This might strike one as hardly as central a notion as those discussed in the other chapters. We do, however, have to take into account that according to the Indian world-view motion is not just something which characterizes billiard-balls, chariots or the planets but also the person moving through successive rounds of rebirth. It is therefore essential to keep in mind that when Nagarjuna speaks of a ‘mover’ this can refer to a person crossing the street as well as to one crossing from this life to the next. To this extent this discussion is also connected with th at of a person, which will be discussed in the following chapter. On one level Nagarjuna’s arguments attem pt to establish the absence of an objective, mind-independent existence of the spatio-temporal location of motion. But in fact his conclusions are more far-reaching than that. When discussing the relation between mover and motion he regards his arguments as a template which can be employed in a variety of contexts. Motion constitutes an im portant illustration of Nagarjuna’s point, but his exposition is not confined to it. The point to be illustrated is concerned with the relation between individuals and the properties they instantiate. Nagarjuna uses the example of motion to argue th at the standard analysis of phenomena into independently existent individuals and properties (as encountered for example in the Nyaya-Vaisesika theory of dravya and guna) is deficient and should be rejected. It is to be replaced by a view which regards individuals and properties as linguistically or conceptually mediated projections of at best pragmatic importance, but not as objective features of a mind-independent reality.

In the examination of the self considered in chapter 7 Nagarjuna moves from the investigation of outer phenomena, such as causation or motion, to the most im portant example of a subjective entity. Following Buddhist tradition Nagarjuna rejects the view of a substance-self, an essentially unchanging unifier of our mental life distinct from our body on the one hand and our psychological states on the other, a self which is an agent whose decisions shape our life. A part from the familiar investigation into the relation of a substance-self and its parts well-known from Buddhist literature Nagarjuna also seeks to refute the substance-self by a different type of argument. This concerns the worry of the opponent that if there are properties of the self there also must be a self which is the bearer of such properties. Given Nagarjuna’s theory of individuals and properties there is, however, no necessity to draw this conclusion. Individuals and properties are seen as linguistic or conceptual artefacts, rather than as fundamental constituents of reality which exist in dependence on one another. Accepting that there are properties of the self does not force him to accept the existence of a substance (dravya) as their bearer on which they depend for their existence. The conception of self emerging as an alternative is that of a process-self, something which is a sequence of physical and psychological events but mistakenly assumes that it is no such sequence, but a substance-self. As in the case of emptiness of objects where the superimposition of svabhava on phenomena had to be overcome, correcting the mistaken self-awareness of the process-self cannot be based solely on working through arguments demonstrating the non-existence of the substance-self, that is a self existing with svabhava. Since viewing itself as a substance-self is the self’s cognitive default, establishing a correct self-awareness can only be achieved by continuous practice. The self and the world are connected in the theory of knowledge which is CH APTER 1. INTRODUCTION 31 the topic of chapter S. For Nagarjuna the discussion of epistemology entails both examining yet another kind of phenomena for existence by svabhava (by investigating whether the means we employ to acquire knowledge of objects are intrinsically such means, and whether the objects are intrinsically such objects) as well as establishing an epistemological framework to explain how emptiness can be known. In this discussion Nagarjuna’s intellectual interaction with the Naiyayikas is particularly pronounced. They provide the source of the idea that means and objects of knowledge can be established ‘from their own side’, a view which Nagarjuna understandably rejects. He concentrates on an examination of the different ways in which we could find out th at particular putative means of knowledge are indeed such means. Both the idea that these means are in some way self-established, or that the means and objects of knowledge mutually establish one another are rejected by Nagarjuna. His aim is to show th at there are no epistemic procedures which are intrinsically and essentially means of knowledge, and th at their objects are not independently existing reals. Essentialism about epistemic procedures is thus replaced by contextualism: procedures can give us knowledge in some contexts, but not in others without ceasing to be means of knowledge. This turns out to be exactly the epistemology the proponent of emptiness needs. For if everything is empty there is obviously nothing which is a means of knowledge intrinsically, by svabhava. But if that means that there are no means of knowledge at all this leads to the problem that emptiness could not be known, contrary to Nagarjuna’s assertion. It is therefore essential to come up with an account of epistemology like the contextualist one which allows for means of knowledge but does not assume that they exist intrinsically. A philosophical system which is concerned as much with the way in which our conceptual and linguistic conventions shape our view of the world as

Madhyamaka is will have something to say on the philosophy of language. Somewhat surprisingly discussions of language do not occupy a great part of Nagarjuna’s writings. The greatest part of the Madhyamaka philosophy of language is a later development. Nevertheless it is possible to extract some of Nagarjuna’s views on the more important issues in this area from the extant sources. This is the subject-matter of chapter 9. A central and well-known assertion in this context is Nagarjuna’s pronouncement that he (and, we may conclude, the Madhyamika in general) does not hold a thesis or position. The commentarial tradition has supplied us with a variety of ways of understanding this statement. I want to argue that the most coherent reading in the context of Nagarjuna’s philosophy as a whole is to understand it as a semantic pronouncement. W hat Nagarjuna means when he says th at he has no thesis is th at he has no thesis which should be interpreted by a particular semantic theory. This theory, which I call the ‘standard picture’, assumes th at the world of referents is endowed with a mind-independent structure and th at our language manages to latch onto the world not just by force of convention, but by the existence of some objectively existent structural similarity between language and world. Both of these assumptions, th at of a ‘ready-made’ world and of an objective reference relation, are incompatible with Nagarjuna’s theory of emptiness as each would entail the existence of entities with svabhava. Once again the interconnectedness of Nagarjuna’s philosophy becomes evident. The doctrine of emptiness, which arises primarily in the context of a metaphysical and cognitive discussion demands an adaptation not only of the standard view of epistemology, but also of the standard view of semantics. Ultimately the Madhyamika will have to explain both the structuring of the world as well as the reference relation in terms of conventions and speaker intentions in order to avoid re-introducing the

notion of svabhava by the back door. The aim of the final chapter 10 is not just to bring all the parts of the discussion together, but to step back from the texts to assess Nagarjuna’s thought systematically. The objective is to examine both the plausibility of Nagarjuna’s conclusions as well as to establish that his philosophy is not a disparate collection of arguments but a coherent philosophical project the different components of which are interconnected in intricate ways. In some instances I also set out to relate Nagarjuna’s conclusions to the contemporary discussion of the matters he deals with. Here my aim is not so much what is sometimes called ‘comparative philosophy’ but an investigation of the question which aspects of contemporary philosophy might be of interest to one aiming to enlarge Nagarjuna’s Madhyamaka from a theory of purely historical interest to a philosophical system with systematic appeal. The chapter is divided into sections dealing with the main problems Madhyamaka thinking investigates: metaphysics, personal identity, epistemology, and language and truth. I have also added a section on ethics, a topic which Nagarjuna treats at length but which, I argue, is discussed by him with less emphasis on problems arising from a specifically Madhyamika point of view than would justify a longer treatm ent in the context of this discussion.

Chapter 2

Interpretations of svabhava

The notion of svabhava provides the central conceptual point around which the greatest part of Nagarjuna’s Madhyamaka revolves. Although it is never used in the sutras and is rare in the Pali canon the term svabhava, often translated as ‘inherent existence’ or ‘own-being’ acquired a dominating role in the thought of the Madhyamika. Despite its centrality, its status is fundamentally negative: one, if not indeed the central concern of Madhyamaka argumentation is to demonstrate that, despite our intuitions to the contrary, svabhava does not exist. The notion of emptiness (sunyatd) denotes precisely the absence of svabhava. There are various difficulties to be faced when trying to get a clear idea of what svabhava as a philosophical concept entails. First of all, like many philosophically central terms svabhava is used in a variety of ways in different philosophical traditions. The early Buddhist Abhidharma metaphysics uses svabhava in a different way from the later Madhyamikas, their use is in turn different from DharmakTrti’s use of the concept, as well as from the Yogacara notion of the ‘three natures’ (trisvabhava). A second problem consists in presenting a clear explication of a concept

which is taken to be vacuous and in fact, if clearly examined, inconsistent. When looking at the Madhyamaka arguments it is often quite hard to attribute anything like a defensible philosophical theory to the proponents of svabhava at all, as these often appear to be conveniently set up straw men.1 A final difficulty is the fact that the concept of svabhava does not have any straightforward equivalent amongst the concepts discussed in the history of Western philosophy. This is not to say th at it is a fundamentally alien concept, but merely that it combines a number of features which we do not see thus combined in the Western context. In order to get a clear conception of svabhava it is essential to appreciate that it incorporates three important conceptual dimensions: an ontological dimension, a cognitive dimension, and a semantic dimension. This chapter will attem pt to spell out the first two of these three different aspects of svabhava. The semantic dimension will be taken up in the discussion of Nagarjuna’s view of language in chapter 9. Our focus will be on Nagarjuna’s use of the term svabhava though we will sometimes refer to later Madhyamaka authors. I do not claim that the above analysis will be adequate for the understanding of svabhava in other Buddhist schools of thought. By explaining how the three aspects of svabhava hang together I also hope to be able to address the second difficulty, that is, give a clear account of what a proponent of svabhava asserts and why this might be a philosophical position to be taken seriously , 2.1 The ontological dimension

Conceiving of svabhava as an ontological concept is no doubt the interpretation most commonly found in the contemporary commentarial literature, 1See Robinson (1972, 326).

and one which gave rise to translations using such metaphysical terms as essence,2 nature,3 substance4 or aseity5. In the Madhyamaka literature after Nagarjuna we find a useful distinction between three different senses of svabhava in Candralurti’s commentary on the MMK,6 a distinction which is already partly present in earlier Abhidharma literature. We will refer to the three senses distinguished by Candraklrti by the terms essence-svabhava, substance-svabhava, and absolute-svabhava, respectively.7 2.1.1 Essence- svabhava Already in the early Buddhist literature we encounter an understanding of svabhava as a specific characterizing property of an object. One characteristic passage from the Milindapahha (composed between 150 B.C.E. and 200 A.D.) asserts:8 Garfield (1995, 89), Komito (1987, 69). 3Napper (1989, 65). 4Lopez (1987, 445-446). 5Ruegg (1981, 9). 6This distinction is still alive in contemporary dGe lugs commentarial textbook literature. See the annotated translation of dKon mchogs ’jigs med dbang po’s Grup pa'i m tha’i m am par bzhag pa rin po che’i phreng ba given in Sopa and Hopkins (1976, 122). 7Sopa and Hopkins (1976, 122) refer to these as phenomena’s ‘conventionally existent nature’, their ‘true or independent existence’, and their ‘real and final nature’. Further attempts at differentiating the different usages of svabhava in Candraklrti can be found in Schayer (1931, xix, 55, note 41), who distinguishes four different senses, as well as in de Jong (1972, 3) and May (1959, 124, note 328), who distinguish two. Although there are obvious connections with the senses distinguished here the relations between the different senses discussed by the three authors and, in Schayer’s case, the distinctness of the four senses given by him are too unclear to make an attempt at comparison worthwhile. 8maranan ti kho maharaja etam aditthasaccanam tasaniyam thanam [...] maranass ’ eso maharaja sarasabhdvatejo tassa saras abhavatejena sakilesd satta maranassa tasanti bhayanti (Trenckner 1928, 149). For a translation see Davids (1890, 211).

Death, great king, is a condition which causes fear amongst those who have not seen the truth, [...] This, o king, is the power of the specific quality (sarasa-sabhdva) of death, because of which beings with defilements tremble at death and are afraid of it. Although at this early stage svabhava does not yet constitute a clearly defined piece of philosophical terminology it is apparent that it denotes a feature by which a particular phenomenon is to be individuated, thereby rendering it knowable and nameable. This understanding of svabhava is made more precise by the Sarvastivadins identification of svabhava and svalaksana,9 the specific quality which is unique to the object characterized and therefore allows us to distinguish it from other objects. Objects have specific qualities as their own (svabhava) because they are distinguished from the qualities of other objects (parabhava) .10 In this context svabhava is understood as an antonym to the common characteristics (samanyalaksana) which are instantiated by all phenomena.11 This understanding of svabhava as the specific quality of objects is further restricted by CandrakTrti’s identification of svabhava with the essential property of an object.12 Every essential property will be part of the specific 91 svabhava is precisely their own characteristic, a common characteristic, on the other hand, is the impermanence of compounded phenomena.’ svabhava evaisam svalaksanam samanyalaksanam tu anityata samskrtdnam. Bhasya on Vasubandhu (1970-1973, 6:14). For further references see Williams (1981, 243). 10svabhdvena parabhdvaviyogatah. (Vasubandhu 1970-1973, 1:18). 11According to the Buddhist interpretation these characteristics are being impermanent, unsatisfactory, and devoid of self (Ronkin 2005, 114-115). 12Note that this sense of svabhava is not to be equated with that of an haecceity or quiddity. An haecceity or ‘individual essence’ is a property only a single individual can have (the socratesness of Socrates is a stock example). But svabhava in the sense discussed here is shareable. The svabhava of fire is heat, a characteristic which cannot just be

quality of an object, but a specific quality need not be an essential property. The specific quality of an object is the unique combination of properties which distinguishes the object from all others. An essential property is something an object cannot lose without ceasing to be that very object. Nagarjuna observes th at ‘ svabhava [in the sense of essence] cannot be removed, like the heat of fire, the fluidity of water, the openness of space’.13 To consider a different example: assume for example that for some reason all existing samples of gold weighed more than 10 grams. In this case ‘weighing more than 10 grams’ is a part of the specific quality of gold, since we use this property together with others to distinguish samples of gold from other things. But even though we never come across a lighter piece of gold in this world, ‘weighing more than 10 grams’ is a property any particular sample of gold could lose without ceasing to be gold — cutting a piece of 10 grams in half does not transform it into another kind of metal. Therefore ‘weighing more than 10 grams’ would be part of the specific quality of gold, but not part of its essential nature. In interpreting svabhava as essence Candraklrti notes th a t14 For, in common usage, heat is called the svabhava of fire, because it is invariable in it. The same heat, when it is apprehended in water, is not svabhava, because it is contingent, since it has arisen instantiated by fire, but also for example by water {even though heat does not constitute the svabhava of water). 13 na hi svabhavah, sakyo vinivartayitum yathagner usnatvam apdrn dravatvam akasasya niravaranatvam. VV(S) 82:14-15. agner ausnyam hi loke tad avyabhicdritvat svabhava ity ucyate \ tad evausnyam apsupalabhyamdnam parapratyayasambhutatvdt krtrimatvan na svabhava iti. PP 241:8- 9. A similar characterization of solidity (khara) as the invariable specific quality and thus svabhava of earth is given in Sthiramati’s Madhyantavibhdgatikd (Williams 1981, 242-243),

from other causal conditions. Heat is a property which is always instantiated by fire (and, for Candraklrti at least, every instantiation of fire is also an instantiation of heat).15 Water, on the other hand, can be either hot or cold and requires some special conditions (apart from just being water) to heat it up. Although not stated explicitly, the notion of essence-svabhava also appears to include a modal element: if fire lost the property of heat it would no longer be fire. Water, however, can cool down and still remain water. This conception of svabhava therefore agrees very well with a common understanding of an essence or essential property in contemporary metaphysics which conceives of them as the properties an object cannot lose without ceasing to be that very object. The notion of essence-svabhava is not one Nagarjuna frequently employs in his arguments concerning svabhava.16 One of his rare references to this conception can be found in the Ekaslokasastra where he states17 because one, two and many each have its own bhava, therefore we call it svabhava. For example, earth, water, fire, and air are respectively hard, moist, hot, and moveable. Each has its own svabhava. And because the nature of every one of the things has Ames (1982, 170). Buddhapalita, on the other hand, clearly has the notion of essence-svabhava in mind when he claims that the aim of Nagarjuna was to teach the svabhava (ngo bo nyid) of dependent origination (Walleser 1913-1914b, 4:16-17). As dependent origination identified with emptiness is the exact opposite of svabhava this expression would constitute a contradictio in adiecto unless we realize that Buddhapalita wants to say that Nagarjuna teaches the specific quality of dependent origination. 17Iyengar (1927, 160). Another translation of this passage of the sastra can be found in Edkins (1893, 307-307). We might want to note, however, that Lindtner (1982, 16) classifies this text as ‘most probably not genuine’. CH APTER 2. IN TERPRETATIO N S OF SVABHAVA 40 its own specific quality (svalaksana) it is said that each has its svabhava. Here svabhava appears to be identified with a quality each of the four elements cannot lose without ceasing to be what it is. It furthermore plays the role of an object’s specific quality (svalaksana)18 which allows the observer to individuate the elements and therefore reflects their essential qualities, i.e. their svabhava.19

2.1.2 Substance-svabhava

The notion of essence-svabhava just discussed which equates svabhava with the specific qualities of an object and contrasts them with those qualities it shares with other objects serves mainly epistemological purposes. It provides a procedure for drawing a line between a variety of objects with shared qualities and thereby allows us to tell them apart. There is, however, a second understanding of svabhava which is of much greater importance in the Madhyamaka debate which considers svabhava to be a primarily ontological notion. Rather than seeing svabhava as the opposite of shared qualities (samanyalaksana) it is contrasted with conceptually constructed or secondary (prajhaptisat) existents and equated with the mark of the primary ones (dravyasat). The distinction between primary and secondary existents constitutes the most fundamental ontological distinction drawn by the Sarvastivadins.20 18Some information on the conceptual relationship between svabhava and svalaksana can be found in Ronkin (2005, 110). Nagarjuna might here have the Vaisesika conception of the five elements (bhuta) in mind, all of which are substances (dravya) and are taken to have peculiar qualities which distinguish them from the other elements. See Sharma (1960, 177). Williams (1981, 236-237).

Primary existents constitute the irreducible constituents of the empirical world, secondary existents, on the other hand, depend on linguistic and mental construction for their existence. For the Sarvastivadin primary existents encompass primarily partless moments of consciousness out of which secondary existents, as for example medium-sized dry goods such as tables and chairs would be constructed. Although both classes of objects were taken as existents (sat), only the primary ones were assumed to possess svabhava. On this understanding svabhava no longer denotes an individuating property of objects by which they can be told apart from other objects (as it did when conceived in terms of essence) but an indication of ontological status.21 To have svabhava means to exist in a primary manner, unconstructed and independent of anything else.22 This notion of svabhava, which we are going to call substance-svabhava is also the sense of svabhava most prominent in Nagarjuna’s arguments.23 The 21Ronlcin (2005, chapter 3) argues that there was a gradual move away from a basically epistemological understanding of svabhava as a characteristic mark to individuate different aspects of experience to an ontological understanding, where svabhava subsumes objects with a particular ontological status. 22 In the contemporary commentarial literature we find occasional reference to the notion of an intrinsic property to spell out the notion of svabhava (Tillemans (2001, 14, note 24), Siderits (2004, 117)). Intrinsic properties are those which ‘things have in virtue of the way they themselves are1, while extrinsic properties are had 'in virtue of their relations or lack of relations to other things’ (Lewis 1986b, 61). While it is clear that all properties constitutive of a primary existent must be intrinsic not all intrinsic properties characterize a primary existent. For example the property of being the tallest man in the room is extrinsic (since a man can only have it in relation to the other men in the room), while that of being a man is intrinsic. However, a man does not exist by svabhava since he is causally, mereologically, and conceptually dependent on a variety of factors. 23Hayes (1994, 311) distinguishes two senses of svabhava, svabhava in the sense of identity and svabhava in the sense of independence, The former expresses the understanding as

fifteenth chapter of the MMK, investigating the notion of svabhava begins by saying: Svabhava cannot result from causes and conditions, because if it was produced from conditions and causes it would be something artificially created. But how could svabhava be artificially created, as it is not artificially created and not dependent on anything else? 24 Substance-svabhava is therefore taken to be something which does not depend on anything else. Candraklrti in fact takes MMK 15:2b to constitute the definition of svabhava: This is the definition of it: Svabhava is not artificially created and not dependent on anything else.25 The notion of substance-svabhava as ‘something which does not depend on anything else’ is stronger than the one found in the Abhidharma literature, For the Abhidharmikas some objects which have primary existence svalaksana, the latter as dravya. Hayes then goes on to argue that Nagarjuna equivocates between these different readings, thereby rendering his arguments invalid (316). For some comments on this see Taber (1998), Tillemans (2001), Siderits (2004, 135, note i). 24na sambhavah, svabhavasya yuktah pratyayahetubhih / hetupratyayasambhutah svabhavah krtako bhavet / / svabhdvah, krtako nama bhavisyati punah katham / akrtrimah svabhavo hi nirapeksah paratra ca. MMK 15:1-2. 25tasya cedam laksanarn akrtrimah svabhavo hi nirapeksah paratra ca, PP 265:5-6. See also CandrakTrti’s commentary on Aryadeva’s CS 12:13: 'Here ‘self’ is a self-existent object which does not depend on other objects. The non-existence [of such an object] is selflessness.1 de la bdag ces by a ba ni gang zhig dngos po m am s kyi gzhan la rag ma las p a ’i ngo bo rang bzhin ste [ de myed pa ni bdgag myed p a ’o. (Candraklrti 1999, 321:1-2). See also Tillemans (1990, 126).

(dravyasat) can be dependently originated. A conditioned object (samskrta dharma) will have svabhava but is still dependent on causes and conditions. It would be wrong, however, to assert26 that the Abhidharmika’s only criterion for absence of svabhava is having parts, so that all other forms of dependence would be compatible with an object being a primary existent. Walser27 cites a passage from the Theravadin Puggalapahhatti Atthakathd where one example of objects existing through dependent designation (upddaya prajhapti, i.e. objects which do not exist in a primary way (dravyasat)) given is the measuring of time and space through the revolution of the sun and the moon.28 Now the division of time into days depending on the rising and setting of the sun is certainly no example of mereological dependence: the latter is not part of the former. It is rather the case that the concepts of day and night owe their existence to the rising and setting of the sun. This is why they are not primary existents but dependently designated, or, as we may also want to put it, conceptual constructs. For the Abhidharmika an object existing with svabhava does therefore not have to be independent of everything (in particular it can depend on its causes and conditions), on the other hand there are reasons distinct from having parts which explain why a thing is merely a secondary existent (prajhaptisat) and therefore lacking svabhava. A view of svabhava which is not explicitly formulated by Nagarjuna but is nevertheless prominent in the Indian and Tibetan commentarial literature is that of findability under analysis.29 Candraklrti observes that [...] worldly things exist without being analysed. When analysed, 26As done by Burton (1999, 90-92). 27 (2005, 242-243). 28For a discussion of tlie variety of ways in which the notion of prajnapti is understood in Abhidharma literature see Priestley (1999, chapter 9). 29See Tillemans (2001, 5-6). CH APTER 2. IN TERPRETATIO N S OF SVABHAVA 44 [however,] there is no self different from form and the other [four constituents] .30 The underlying idea is th at whatever is not ultimately real disappears under analysis so that what we are left with must be an ultimately real object existing by its own nature. The reason why composite objects like a chariot or the self (atman) are not ultimately real is because they do not withstand logical analysis (rigs pas dpyad bzod pa). Once the chariot or the self are conceptually dissolved into the parts which constitute them the objects have disappeared and all we are left with are parts collectively conceptualized as a chariot or a self.31 Findability under analysis and independence of other objects imply one another. For assume some objects x have been determined by analysis to be the ultimate constituents of some thing. If the existence of these xs in turn depended on the existence of some y only y ) but not x , could be a candidate for an ultimately real object. Conversely, if no ultimate constituent can be found under analysis (and if this is not due to practical limitations) it is because every potential candidate is again dependent on something else. There are some conceptions of substance in Western philosophy which exhibit a certain amount of similarity with the notion of substance-svabhava just described. Descartes’s view of substance as something which does not require another thing for its existence32 springs to mind, as well as W ittgenstein’s theory of logically simple objects presented in the Tractatus.33 These objects, which are supposed to constitute the substance of the world are sim- 30 [...] avicaratasca laukikapadarthcinam astitvat \ yathaiva hi rupadivyatirekena vicaryamana atma na sambhavati. PP 67:7-8; see Ruegg (2002, 117). 31Trenckner (1928, 27), Davids (1890, 44). 32Principes I, 51. 33Keyt (1963), Proops (2004). CH APTER 2. INTERPRETATIO N S OF SVABHAVA 45 pie, unchanging and exist independently of one another and constitute a notion which shares at least some important properties with the Sarvastivadin’s primary objects.

Dependence relations

In order to understand this ontological conception of substance-svabhava it is important to get a clear idea of what precisely is meant by the dependence of an object on another one. First of all it is important to distinguish two notions of ‘dependence1 which should not be confused. These are • existential dependence: An object a existentially depends on objects falling under the property F iff necessarily, if a exists there exists something falling under F. • notional dependence: Objects falling under the property F are notionally dependent on objects falling under the property G iff necessarily, if some object x falls under F there will be a distinct object y falling under G. Saying that a sprout depends existentially on its cause means th at necessarily, if a sprout exist there will be some objects falling under the property ‘causes of the sprout’, such as a seed, soil, water, sunlight, and so on. Similarly, if a complex physical object exists, so will all its parts; therefore the object existentially depends on its parts. Necessarily, if a book exists, so will each of its pages. Notional dependence, on the other hand, is a quite different case. Northern England depends on Southern England, but we would hardly want to say th at this dependence is existential. If due to some geological disaster all of Southern England were destroyed this would not affect the existence of the CH APTER 2. IN TERPRETATIO N S OF SVABHAVA 46 stretch of land now called Northern England. But it affects its description as Northern England, since now there would be nothing south of it which was also England. The concept Northern England therefore depends notionally on the concept Southern England, but the object in the world the concept ‘Northern England’ picks out does not depend existentially on the object ‘Southern England’ picks out.34 It is interesting to note that in the later dGe lugs commentarial tradition three varieties of existential dependence are distinguished: causal dependence, when an object depends for its existence on its causes and conditions, mereological dependence, when an object depends on its parts, and thirdly conceptual dependence, postulating the dependence of an object on a basis of designation, a designating mind and a term used to designate the object.35 These dependence relations are supposed to stand in a qualitative and doxographical hierarchy. Causal dependence is seen as the coarsest understanding of dependent arising and is associated with the Vaibhasikas or Sarvastivadins, the Sautrantikas and the Cittamatrins; mereological dependence is a bit more subtle; the Svatantrika Madhyamikas are assumed to understand emptiness in terms of both causal and mereological dependence. The most subtle understanding which incorporates all three forms of dependence is associated with the Prasangika Madhyamikas,30 34A detailed discussion of different kinds of dependence relations can be found in Simons (1987, chapter 8). Our notion of existential dependence is there called ‘generic dependence’. 35Ngag dbang dpal ldan (1797-?) refers to these three kinds of dependence in his (1964, 154:6-155:1) (which constitutes commentary on the Grub m tha’ chen mo of ’Jams dbyangs bzhad pa (1648-1721)) as ’phrad ba (prdpya), Itos ba (apeksya) and brten pa (pratitya), respectively. See Magee (1999, 56-57), Hopkins (1983, 166-177), Komito (1987, 190), Gyatso (2005, 20-21). 36Ngag dbang dpal ldan (1964, 154:6-155:1), Gyatso (2005, 33, 43-44, 59-62).

There are a variety of examples from Nagarjuna’s works which show that both the notions of existential and notional dependence are employed in his arguments. Verse 13 of the SS asserts:37 The father is not the son, the son is not the father, those two cannot exist one without the other, those two cannot be simultaneous, likewise the twelve links of dependent origination,38 When saying th at the son cannot exist without the father Nagarjuna obviously means that the son is existentially dependent on the father: if some person a exists there exists something falling under the property ‘father of a \ But claiming that the father cannot exist without the son cannot be a case of existential dependence as well. Abelard (that very man) could have existed without having ever fathered Astrolabius. But Abelard as a father depends notionally on Astrolabius: if nobody was subsumed by the concept ‘son of Abelard’ Abelard would not fall under the concept ‘father’. It is therefore evident that the ‘mutual dependence’ of father and son Nagarjuna postulates is based on two different dependence relations, the son depending existentially on the father, the father notionally on the son. For Nagarjuna’s argument, however, it is necessary that the two entities discussed are related by a symmetric dependence relation.39 37 pha bu ma yin bu pha min / de gnyis phan tshun med min la / de gnyis cig car yang min Itar / yan lag bcu gnyis de bzhin no. SS 13. 38The father-son example is also used in VV 49-50. See also MMK 8:12, 10:10. 390etke (1989, 11) claims that ‘the assumption of isomorphism or identity of logical and causal dependence relations [which correspond to our notional and existential dependence relations] explains a significant part of Nagarjuna’s arguments and simultaneously elucidates numerous apparent difficulties’.

The difficulty disappears if we take into account that if some object x is essentially F, and if it also depends notionally on some y being G, then x will also depend existentially on j/’s being G, since x has to have F to exist at all (this is just what F being an essential property of x means). Therefore if we assumed that Abelard was the father of Astrolabius essentially Abelard would indeed depend existentially on his son, since having Astrolabius as a son would be a property Abelard could not lose without being th at very man. Of course we might wonder now why we should assume th at Abelard was essentially the father of Astrolabius. Even if we do not think th at this is reasonable (because we think that a childless Abelard would have been the very same man) it is important to note that Nagarjuna intends the fatherson example as an argumentative pattern in which different predicates can be substituted. For example we might think (as Nagarjuna’s opponent does) that cause and effect have their respective natures essentially. In this case it is then evident th at the existential dependence between the two must be symmetric: the effect depends existentially on the cause, but the cause also depends existentially on the effect. I hope this small example has convinced the reader of the importance of keeping the two different kinds of dependence relation apart when analyzing Nagarjuna’s thought.40 A more detailed discussion of how this distinction is 40The failure to distinguish between existential and notional dependence has resulted in considerable confusion in the contemporary commentarial literature, primarily in connection with the so-called principle of coexisting counterparts (Taber 1998, 216) (Ruegg (1977) calls it ‘the principle of the complementarity of binary concepts and terms’) which is supposed to say that ‘a thing cannot be a certain type unless its counterpart exists simultaneously with it’. Far from being ‘a blatant contradiction of common sense’ (Taber 1998, 238) it expresses an obvious truth about notional dependencies: something cannot be Northern England unless Southern England exists at the same time. It is only if we think that notional dependence is the very same thing as existential dependence that we can acCH

put to work in his arguments will be found in chapters 5 to 9. Returning to the understanding of svabhava in terms of substance we should note that for Candraklrti such substance-svabhava is qualified by its non-dependence on other objects, either existentially or notionally. This is evident from the examples Candraklrti gives for objects which are dependent on causes and conditions: the heat of water, the farther and nearer shore, long and short.41 While the heat of fire depends existentially on the causes which heat up the water, the concepts ‘farther shore’ and ‘long’ depend only notionally on the concepts ‘nearer shore’ and ‘short’. The farther shore would not cease to exist if the nearer shore did, nor would long objects decrease in length if short objects disappeared, but their descriptions as ‘farther shore’ or ‘long’ could no longer be employed. It is evident th at the notion of substance-svabhava is much stronger than th at of essence-svabhava. In particular we can assert the existence of the second without affirming that of the first. It could be the case that every object had some properties it could not lose without ceasing to be that very object (although in some cases it may be more difficult than in others to determine what these properties are) and therefore be endowed with essence-svabhava. But at the same time everything could in some way (either existentially or notionally) be dependent on something else so that substance-svabhava did not exist at all. cuse Nagarjuna of being unable to distinguish ‘between saying that a thing exists at all and saying that it exists under a given description’ (Hayes 1994, 315). For more (unfortunately not very clear) discussion of this matter see Taber (1998). 41PP 264:1.

T h e rejection ofsubstance -svabhava

It is important to note that the elaborate Madhyamika criticism of the notion of svabhava is directed against this stronger notion of substance-svabhava, rather than against essence-svabhava. As the common conception of svabhava was in terms of essential properties (a conception ‘well known5, as CandrakTrti charmingly puts it ‘to people, including cowherds and women5)42 CandrakTrti explicitly distinguishes it from his notion of substance-svabhava: even though it is an essential property the heat of fire is no more the svabhava of fire than it is the svabhava of water. Let it be recognized th at heat, also, is not the svabhava of fire, because of its artificiality. Here one apprehends that fire, which arises from the conjunction of a gem and fuel and the sun or from the friction of two sticks, etc., is purely dependent on causes and conditions, but heat does not occur apart from fire. Therefore, heat, too, is produced by causes and conditions, and therefore is artificial; and because of its being artificial, like the heat of water, it is clearly ascertained that it is not svabhava.43 CandrakTrti does not attem pt to refute the notion of essence-svabhava but asserts its existence in conventional terms (vyavahara). If something lacked the property of heat we would not call it fire.44 Candraklrti5s emphasis is on gopaldhganajanapvasiddham. PP 260:14. 43 yad etat ausnyam tad apy agneh svabhavo na bhavatiii grhyatam krtakatvat | iha maniindhana- dditya-samagamad arani-nidharsanad escagn&r hetupratyaya-sdpeksa taivopalabhyate | na cdgni-vyatiriktam ausnyam sambhavati \ tasmad ausnyam api hetupratyayajanitam \ tatasca krtakam \ krtakatvac capamausnyavat svabhavo na bhavatfbi sphutam avasiyate. PP 260:9-13. 44Schayer (1931, xix) argues that the Madhyamika denies the existence of essence-svabhava. As everything is causally produced ‘there is no property which could never be

establishing that essence-svabhava ‘does not deserve to be called svabhava,45 and is distinct from the notion of substance-svabhava Nagarjuna deals with. Unlike substance-svabhava, however, CandrakTrti has no difficulties in agreeing with the usefulness of essence-svabhava as a concept for everyday usage.46 For the Abhidharmikas substance-svabhava does exist; it is the intrinsic and essential quality of ultimately real objects (dravya). The justification for the assumption of such objects is evident if we consider the case of objects consisting of parts.47 A partite object cannot exist by svabhava since it exists only in dependence on its parts. For the same reason its parts cannot exist by svabhava either as long as they have parts in turn. For the defender of substances this regress must stop somewhere, because even though it might be possible to have a chain of explanations stretching back infinitely (if we explain the properties of the whole by the properties of the parts and then in turn provide an explanation of their properties in terms of their parts) a chain of dependency relations must terminate ultimately, th at is the hierarchy of dependency relation must be well-founded.48 The Abhidharmikas consider the entities which are the foundation of the mereological dependency relation to be ultimately real objects which have their properties essentially and intrinsically. These objects exist by substance-swab/mm. missing from a particular object’ (55, note 41). While the Madhyamika will accept this the important point to get is that saying some property is part of the essence-svabhava of an object of type X does not mean it could never be missing from that object, but that if it was missing we would not consider it to be of type X . 45nayam svabhavo bhavitum arhati. PP 260:15. 46Ames (1982, 170). 4TSiderits (2004, 118-119). 48See Burton (1999, 109-111), Walser (2005, 243-244).

The Indian and Tibetan Madhyamaka literature contains a variety of ways for classifying arguments against the existence of substance-svabhava.49 A five-fold classification distinguishes the following kinds: 1. the diamond slivers50 2. the refutation of the production from existent or non-existent 51 3. the refutation of the four kinds of production52 4. the argument from dependent origination53 5. the ‘neither one nor many’ argument54 1), the diamond slivers, so called because of the power ascribed to this argument in refuting substance-svabhava analyses four ways in which an object could be causally produced: by itself, by another object, by both, or without a cause.55 This argument will be discussed in detail in section 5.3 of chapter 5. 2), the refutation of the production from the existent or non-existent concerns the temporal relation between cause and effect.66 It will be discussed in section 5.4 of chapter 5. 49Most classifications distinguish either four or five kinds of arguments; there are also slight variations concerning which arguments are subsumed under which heading. For details see Tillemans (1984, 371-372, note 16). 50 vajrakana, rdo rje gzegs. 51 sadasatutpadapratisedha, yod med skye ’gog. 52 catuskotyutpadapratisedha, mu bzhi skye ’gog. pratityasamutpadahetu, rten cing ’brel ba’i gtan tshigs. 54 ekanekaviyogahetu, gcig du bral gyi gtan tshigs. Hopkins (1983, 132-150, 639-650). Tillemans (1984, 361). The temporal reading of this argument is not always so clear. Sometimes (1984, 361) it is argued that the diamond slivers and the refutation of the production from the existent or non-existent are to be distinguished by the fact that the first analyses the cause, the second the effect. This analysis then investigates whether a

3), the refutation of the four kinds of production is generally taken to refer to an argument that considers the numerical relations between cause and effect: many causes creating one effect, many causes creating many effects, one cause creating many effects, one cause creating one effect. It is the only one of the five argument which is does not have a textual basis in Nagarj una’s works; we therefore do not discuss it here any further.57 4), the argument from dependent origination considers the compatibility of substance-svabhdva with a variety of dependence relations such as those considered in the previous section.58 5), the ‘neither one nor many argument’ will be discussed below.59 In the following discussion I will also examine two arguments against the existence of substance- svabhava which we find in Nagarj una’s works but which are not included in the classification given above: the property argument and the argument from change.

The property argument

One problem with the assumption of primary existents endowed with substance- svabhava becomes evident once we analyse them in terms of the familiar distinction between individuals and properties. According to classical Buddhist ontology there are different kinds of primary existents (mahabhutas: earth, water, fire, wind) which are distinguished by different qualities.60 This list is sometimes enlarged to a list of six elements or dhatvs by adding space cause produces an existent, a non-existent, a both existent and non-existent or a neither existent nor non-existent effect. See Hopkins (1983, 151-154). 57Some discussion is in Hopkins (1983, 155-160). 58See Hopkins (1983, 161-173). 59See Hopkins (1983, 176-196). eoLa Vallee Poussin (1988-1990, 68-70), Dhammajoti (2004, 147-148).

and consciousness.61 It is this list of six Nagarjuna’s account in chapter five of the MMK is based on.62 The problem he discusses, however, is independent of our willingness to assume the existence of primary ‘fire-atoms’ and so forth. It arises whenever we assume that there are different categories of primary existents distinguished by different properties.63 We can easily conceive of ordinary individuals as lacking some qualities which they in fact possess; for example we can conceive of a red apple as lacking the property of redness and being green instead. In the case of primary existents, however, this is not possible. If we abstract the property of heat from a fire-atom there is nothing left, unless we believe in a propertyless ‘bare particular’ which could act as the individual instantiating the property of heat. Nagarjuna considers this possibility in the case of space:64 No space is evident prior to the characteristic (laksana) of space. If it existed prior to the characteristic it would follow that it was without the characteristic.65 Thus assuming that a ‘space-atom’ existed first without its characteristic and 61La Vallee Poussin (1988-1990, 88). 62MMK 5:7. 63See Siderits (2003, 120-123). 64The ontological status of space is a controversial issue in the Buddhist philosophical literature. Whilst not being one of the four mahabhutas (Dhammajoti 2004, 148-149) the Abhidharmakosabhasya nevertheless includes it together with these in a list of six elements ( dhatvs) (La Vallee Poussin 1988—1990, 88). Moreover, the

• Abhidharmamahahavibhasasastra argues that space can be a dominant condition (adhipatipratyaya)

for the mahabhutas and therefore possesses svabhava (Dhammajoti 2004, 384). Problems with properties of the mahabhutas will therefore equally apply to space. ®5nakasam vidyate kimcit purvam dkdsalaksanat / alaksanam prasajyeta sydt pui'vam yadi laksandt. MMK 5:1. CH APTER 2. IN TERPRETATIO N S OF SVABHAVA 55 only later acquired it, in the way in which an apple can exist without the property of redness which is only acquired once the apple is ripe commits us to the existence of a propertyless bare particular. This is due to the fact th at unlike ordinary objects such as apples primary existents have all their properties essentially. Since the only essential characteristic of space is its particular space-nature a space-atom without this characteristic it is like a knife without a handle which has lost its blade: there is nothing left. For Nagarjuna introducing bare particulars at this point is not an option; he claims that ‘an object without characteristics is not to be found anywhere’.66 Why does Nagarjuna reject the notion of a bare particular? Bare particulars do not appear to be straightforwardly contradictory entities, in fact their existence is postulated by metaphysicians claiming that individuals must be more than just bundles of properties.67 The problem seems to be this. Let us assume that there was indeed a bare particular left over once we abstracted the property of heat from a fireatom. Assume furthermore that this particular would have its nature (its bare-particular-ness) intrinsically and essentially. In this case heat could not be its svabhava as well, since something cannot have two different svabhdvas. Its further characterization by heat would therefore be superfluous for establishing its status as a primary existent. Alternatively we could assume that the bare particular did not have its nature intrinsically and essentially, but dependent on something else. We could then ask again whether this something else has its property essentially, and so on.68 In this case we get into a regress which the opponent of Nagarjuna has to term inate somewhere, since he wants to establish th at some G6alaksano na kascic ca bhavah, samvidyate kvacit. MMK 5:2a. 67Armstrong (1997, 109-110, 123-126). 68PP 130:5-13, Siderits (2003, 121).

objects (i.e. the true primary existents) exist by svabhava and are therefore not dependent on anything else, We therefore end up with the first possibility again, as the various properties which make up the supposed svabhava of the primary elements fire, water and so forth are superfluous in characterizing these foundational objects as primary existents, as these objects are already existent as such. This is what Nagarjuna means when he says that The occurrence of a characteristic is neither in the uncharacterized nor in the characterized. It does not proceed from something other than those with or without characteristics.69 If we regard the bare particular as characterized by its bare-particular-ness intrinsically and essentially, any further characteristic is superfluous for bestowing the status of a primary existent. If we do not regard it as so characterized, however, we end up in an infinite regress without establishing any primary existents at all. Since these possibilities are mutually exclusive the notion of a bare particular seems to be facing a problem. The proponent of bare particulars might now be inclined to say th at all this shows is th at the pluralist theory of the six primary elements was mistaken and th at we have to assume that there is only one kind of primary existent, namely bare particulars having their nature intrinsically and essentially. They constitute the ‘pure stuff’ of the world which is then ‘flavoured’ by such properties as heat, wetness etc. in order to form fire-, water- and other atoms. Bracketing the difficulty of how these different bare particulars are to be told apart the most im portant problem with this is that Nagarj una’s oppo69 nalaksane laksanasya pravrttir na salaksane / salaksanalaksanabhyam napy anyatra pravartate. MMK 5:3.

nent also wants to argue that the primary existents are mind-independent, th at they exist whether or not any conscious beings are around. But while this has a certain plausibility for objects which can be distinguished by their properties (such as the four mahabhutas or the fundamental particles of contemporary physics) a bare particular from which all characteristics have been abstracted away bears the mark of the mind’s handiwork. Bare particulars are nothing we are immediately (or even mediately) acquainted with — they are conceptual fictions, theoretical entities introduced in the course of constructing an ontological theory, but hardly anything we would supposed exists ‘from its own side’, independent of conscious minds. If Nagarj una’s opponent does not want to postulate the existence of bare particulars he might try to solve the problem of properties of primary existents by arguing that primary existents are property-particulars, rather than things characterized by properties. This is the dual of the bareparticular view, for we now assume properties without bearers, rather than bearers without properties. As a m atter of fact ontological theories based on property-particulars (also called tropes) have become relatively popular in the recent metaphysical discussion.70 The fundamental idea here is that the redness of an apple is not regarded as one thing inhering in different red objects. The redness of the apple, th at of a tomato and that of a postbox are rather regarded as three distinct property-particulars which are sufficiently similar to be classified under the common name ‘red’. Nagarjuna is clear on his rejection of property-particulars71 but unfortunately not very explicit on his reasons for doing so. A plausible reason for 70See Williams (1953) for an early example, Bacon (1995) for a more recent discussion. 71 In the absence of the characterized the characteristic does also not exist. laksyasydnupapattau ca laksanasyapy asambhavah. MMK 5:4b.

Nagarjuna’s rejection is provided by Marks Siderits.72 If we assume th at the different primary existents, such as fire- and water-atoms are just property particulars of heat, wetness and so forth we face the problem of how the different atoms are to be individuated.73 We obviously cannot say that two fire-atoms are different because the property of heat is instantiated in different bearers, as this would get us back to the scenario discussed earlier on. It seems that the best we can do is individuate clusters of property-particulars, as in saying that in one cluster heat is associated with wetness (as in hot water), in another with solidity (as in a red-hot iron ball) and th at in this way the two property-particulars of heat are individuated. However, now the problem is that the independence of primary existents is compromised, as we now have to rely on distinct property-particulars to tell them apart. Therefore their existence as distinct primary existents is not any quality they possess from their own side, but only something they have dependent on occurring on clusters with other property-particulars. It now becomes evident that whatever analysis of primary existents in terms of individuals and properties we propose seems to face fundamental difficulties. If we treat the primary existents and their properties as distinct and independent entities (as we do in the case of ordinary objects) we realize that the two cannot be independent at all, since we cannot conceive of a primary existent without its characteristic property. If, on the hand we subsume primary existents under one side of the individual-property divide, that is if we assume th at they are either bare particulars (individuals without properties) or tropes (properties without individuals) it becomes evident that neither of these can play the desired role of mind-independent foundational 72(2003, 122-123). 73 A discussion of different ways of individuating tropes is in Schaffer (2001).

objects existing from their own side. Given that Nagarjuna regards these options as exhaustive he considers the above difficulties as a reductio of the notion of a primary existent. For him the primary existents and the properties they instantiate have to be regarded as existentially dependent on one another. If the properties did not exist there would be no particulars to characterize, in the absence of the particulars there would be no characterizing properties. But in this case a fundamental property of primary existents is no longer fulfilled, namely that these existents should be independent of all other objects.75 Depending for their existence on the properties characterizing them the supposed primary existents cannot produce the foundation for a hierarchy of dependence relations. It therefore turns out that the only satisfactory way of understanding the relation between primary existents and their properties has to deny that there are primary existents in the first place. The m ereological argum ent In the same way in which applying the framework of individual and property to primary existents leads to problems, Nagarjuna argues, conceiving of them in terms of parts and wholes entails difficulties. Ordinary objects are either mereologically complex (i.e. they have proper parts) or they are mereologically simple, that is they are atomic, partless things. Prim ary existents, however, the Madhyamika argues, are neither complex nor simple. As everything must be either complex or simple, the notion of a primary existent is thereby reduced to absurdity. While this ‘neither one nor many argument’76 is mentioned by Nagarjuna 74MMK 5:3, 5. 75MMK 5:4-5. 76 ekanekaviyogaheiu, gcig du bral gyi gtan tshigs.

at several places77 it does not play a prominent role in his arguments and is not spelt out in great detail. The later Tibetan commentarial literature, on the other hand, contains quite an extensive development of this argument for refuting substance-sva&hava.78 We will use some of these materials in order to formulate the details of the argument which can at best be said to be implicitly present in Nagarjuna’s texts. It seems to be clear that we do not want to say th at primary existents are partite phenomena. This is because partite phenomena depend for their existence on their parts, and primary existents are not supposed to be existentially dependent on anything. For any object with parts it is at best possible that its parts are primary existents, but the composite object itself cannot be. This denial of the primary nature of partite entities is a well-entrenched Buddhist position which can be traced back to some of the earliest textual material79 and provides the reason for Nagarjuna’s denial that any candidate for a primary existent, whether causally produced or not, can regarded as having parts.80 The other alternative, that primary existents could be atomic is also denied by Nagarjuna. In RA 1:71 he claims: Because of having many parts it is not simple; there is nothing which is partless.81 77SS 32ab; RA 1:71. VP 33-39 also contains a mereological discussion which is of relevance in this context. Unfortunately this material is relatively opaque and also set in the specific context of the rejection of the notion of the parts of a syllogism. Its usefulness in explicating Nagarjuna’s understanding of the ‘neither one nor many argument’ is therefore limited.

79See Siderits (2003, chapter 1) for a discussion of this ‘Buddhist reductionism’. 80 'dus byas dang ni ’dus ma byas / du ma ma yin gcig ma yin. SS 32ab. 81 naiko ’anekapradesatvan napradesas ca kascana. rGyal tshab dar ma rin chen in his CH APTER 2. IN TERPRETATIO N S OF SVABHAVA 61 Unfortunately Nagarjuna does not give us an argument here for why he thinks th at nothing is partless. Considering the secondary literature we can come up with at least two different accounts of what the argument might have been. Their difference is mainly due to different understandings of the word pradesa (phyogs), here translated as ‘p art’. Tucci82 reads pradesa as denoting a straightforward mereological part or side of an object and assumes that Nagarjuna here refers to the well-known Yogacara argument against the existence of atoms.83 In a nutshell the idea is th at if we assume that atoms can conglomerate to form macroscopic objects, individual atoms must touch one another, like neighbouring mustard-seeds in a heap. But if we now regard the side (pradesa) of an atom where it touches its right neighbour as spatially coinciding with the side where it touches its left neighbour the entire conglomeration of atoms will collapse to a single spatial point. If we regard them as spatially distinct, on the other hand, the atom must be seen as spatially extended and therefore cannot be atomic. A more comprehensive understanding of the notion of part in the ‘neither one nor many argument’ which we find for example in Tsong kha pa subsumes under it not just mereological parts, but also temporal stages and aspects (m am pa).84 This allows us to run a more general argument than is possible on the purely mereological interpretation. The question investigated in this case is the relationship between the parts, stages or aspects of a primary existent and the primary existent itself.85 commentary explicitly regards this verse as a statement of the ‘neither one nor many argument’. See Hopkins (1998, 103). 32(193 4—1936 , 324). The argument is described in verses 11-14 of Vasubandhu’s Vimsatika (Tola and Dragonetti 2004, 127-129, 142-145). See also Kapstein (2001), Hopkins (1983, 373). Tillemans (1983, 308). Thurman (1984, 97).

For the sake of simplicity consider the case of a supposed primary existent, say an earth-atom, and two of its properties (say ‘being solid’ and ‘being heavy’). Now obviously the earth atom cannot be identical with both these ‘parts’, as one thing cannot be identical with two. So it must be distinct from them and should be conceived of as an individual which constitutes the bearer of the two properties. But this understanding of primary existents then leads us straight back to the problems we have encountered when discussing the property argument: we either end up with a bare particular (which is difficult to consider as existent in a mind-independent way) or with a trope (which cause problems with individuation). The view that primary existents exist without parts, stages or aspects should therefore be given up. The conclusion Nagarjuna wants to draw from this is that since everything either has parts or does not, and since neither option is possible for primary existents our difficulties stem from assuming such a notion in the first place. From a mereological point of view no consistent account of primary existents can be given. It is a notion we should do without. The argument from change Nagarjuna considers the existence of substance-svabhava to be incompatible with change:86 If svabhava existed the world would be without origination or cessation, it would be static and devoid of its manifold states. But given that we do perceive change in the world this provides us with an argument against substance-svabhava: 86ajdtam anivuddham ca kutastham ca bhavisyati / vicitrabhir avasthabhih svabhave rahitam jagat. MMK 24:38. Other passages dealing with the permanence of svabhava include 13:4, 21:17, 23:24, and 24:22-26.

By of the observation of change [we can infer] the lack of svabhava. of things. [...] If svabhava was found, what would change? Neither the change of a thing itself nor of something different is suitable: as a young man does not become old, so an old man does not become old either.87 No thing which we perceive to be changing can exist by substance-svabhava. This is because an object existing by substance-svabhava, i.e. a primary existent, constitutes an independent, irreducible and unconstructed fundamental constituent of reality. If the young man had its age as an essential and intrinsic property (i.e. if he was young by svabhava) he could never grow old. The obvious reply the advocate of substancersvabhava should make at this point is to say that both change and subst&nce-svabhava exist, though not at the same level. Things which we perceive as changing do not possess substance-svabhava, while those which do possess it do not change. There are at least two different ways in which we could spell this out. According to the annihilationist view an rc-atom existing by svabhava can never change into a y-atom. W hat can happen, however, is th at the reatom ceases to exist and is replaced by a y-atom. W hat we perceive as macroscopic change in the nature of entities (hot water cooling down, green leaves turning brown) is in fact nothing else but the microscopic arising and ceasing of entities the natures of which do not change.88 According to the permutationist view no entities existing by svabhava ever pass out of existence. The change we observe is merely a difference in 87bhdvdnam nihsvabhdvatvam anyathabhavadarsanat / [ •• •] kasya syad anyathabhdvah svabhavo yadi vidyate / / tasyaiva nanyathdbhavo napy anyasyaiva yujyate / yuvd na jvryate yasmad yasmdj jirno na jiryate. MMK 13:3a, 4b-5. 88See Siderits {2003, 124-125) for a description of this view.

arrangement of the eternally existing objects. When hot water cools down this is not because the fire-atoms in the water pass out of existence, but rather because the set of permanently existent atoms changes its arrangement so that fewer fire-atoms are now mixed amongst the water-atoms. There are two main difficulties for the annihilationist view. First of all it is not obvious to which extent the cessation of entities existing by svabhava is theoretically less problematic than a change in their nature. The annihilationist view is based on the assumption that if some object passes out of existence its svabhava is not changed, since the object does not exist any more. It did not lose one nature and acquire another one, as there is nothing left which could possibly acquire such a nature. Whether this in fact works depends on the interaction of the conception of svabhava with that of momentarily existent objects. This is an intricate issue89 which we fortunately do not have to settle here. There remains a second problem, namely answering the question what is responsible for the cessation and production of entities existing by svabhava. If they are dependent on causes and conditions for their production and annihilation then they cannot be ultimately real entities after all, as the whole point of postulating entities existing by svabhava was to have some objects which are not existentially dependent on any others.90 The permutationist does not have this problem. He still has to assume that the ultimately real objects congregate in certain ways dependent on causes and conditions, but this only concerns the existential dependence of the objects they thus constitute, objects which were not supposed to exist by svabhava in the first place. While the permutationist view thus seems more 89See von Rospatt (1995). 90Siderits (2003, 125).

attractive than the annihilationist one it has the curious consequence that the supposedly ultimately real objects existing by svabhava recede more and more. The idea of fire-atoms as ultimately real objects is obviously only of historical interest. It is far from clear, however, whether the conception of elementary particles of contemporary physics is much more attractive to the permutationist. While the mahabhutas had the advantage of being relatively close to objects of ordinary experience, such as fire, water, and so forth, various subatomic particles, quarks or strings are purely theoretical posits very different from anything we usually observe. Nobody has ever seen a string and nobody ever will, as these not accessible to sensory perceptions. Moreover, their assumed properties are very different from what we observe in the macroscopic world. As with all theoretical posits claims for their existence are based on the explanatory work this concept can do in a particular theory. It is therefore quite ironic that our best candidates for ultimately real entities existing independent of human conceptualization turn out to be objects which are so highly theory-dependent and the existence of which seems to be considerably less secure than that of the medium-sized dry goods with which we interact daily. It appears th at neither the annihilationist nor the permutationist view manage to give a satisfactory explanation of the existence of change in the presence of substance-svabhava. In the absence of any other explanations Nagarjuna thus concludes that our experience of change constitutes an argument against the existence of substance-suafr/mua. Considering the previous discussion it is evident that most of Nagarj una’s arguments, as well as those found in the work of his commentators such as CandrakTrti are concerned with the rejection of substancersvabhava^ while the CH APTER 2. IN TERPRETATIO N S OF SVABHAVA 66 examination of essence-svabhava appears to play at best a minor role. From certain passages in CandrakTrti’s works, however, we get the impression that a third conception of svabhava is in play in the Madhyamalca arguments as well. This third notion does not seem to share the marginal status of essencesvabhava and is also not the aim of attem pted refutations. We will call this conception absolute svabhava.

Absolute svabhava

Candrakirti describes absolute svabhava in the following way:91 Ultimate reality for the Buddhas is svabhava itself. That, moreover, because it is itself nondeceptive is the truth of ultimate reality. It must be known by each one for himself. While he stresses that substance-svabhava is a notion erroneously ascribed to objects which in fact lack it92 he also asserts that svabhava does not in any way appear to those having misknowledge.93 It therefore appears that svabhava is both a mistaken ascription made by beings with deficient cognitive capacities as well as something which does not appear to such beings. To make sense of this we have to assume that there are two different conceptions of svabhava in play here: substance-svabhava which the Madhyamaka arguments attem pt to show to be non-existent on the one hand, and, on the other hand, another kind of svabhava, which I call absolute svabhava, 91sangs rgyas m am s kyi don dam pa ni rang bzhin nyid yin zhing \ de yang bslu ba med pa nyid kyis don dam pa ’i bden pa yin la | de ni de m am s kyi so sor rang gis rig par bya ba yin no. MAB 108:16-19. 92PP 261:3-4. 93MAB 107: 15. See also 306. CH APTER 2. INTERPRETATIO N S OF SVABHAVA 67 which constitutes the true and intrinsic nature of phenomena.94 CandrakTrti explicitly characterizes this as changeless (avikaritva), not originated (sarvadanutpada) and not dependent on something else (paranirapeksa). 95 Based on this the later Tibetan commentarial literature conceives of svabhava as ‘triply characterized5.96 Tsong kha pa describes it as

1. not produced by causes and conditions97

2. unchangeable98

3. set forth without depending on another object99

The interesting problem arising at this point is th at both CandrakTrti5s attributes as well as Tsong kha p a’s triple characterization are supposed to be applicable both to substance-svabhava as well as to emptiness, i.e. the absence of substance-svabhava.100 But taking into account that substancesvabhava is argued not to exist, while emptiness does exist this view faces an 94Some of the synonyms for absolute svabhava CandrakTrti gives include ‘objecthood of objects’ ( dharmanam dharmata), ‘intrinsic nature’ (tatsvarupam), ‘original nature’ (prakrti), ‘emptiness’ (sunyata), ‘lack of svabhava’ (naihsvdbhdvyam), ‘thusness’ (tathata), ‘complete non-origination’ (sarvadd anutpada), and ‘being thus, changelessness, everabidingness’ (iathdbhdvo ‘vikaritvam sadaiva sthayita) PP 264:11-265:1. 95PP 265:1-2. 9Gkhyad par gsum dang Idan pa (Tsong kha pa Bio bzang grags pa 1985, 643:12), (Tsong kha pa Bio bzang grags pa 2000-2004, 3:194). This characterization follows Nagarjuna’s discussion of svabhava in MMK 15:2 and 8. See also Magee (1999, 87-88). 97rgyu dang rkyen gyis ma bskyed pa. (Tsong kha pa Bio bzang grags pa 1985, 643:12- 13), (Tsong kha pa Bio bzang grags pa 2000-2004, 3:194). g8gnas skabs gzhan du mi ’gyur ba. (Tsong kha pa Bio bzang grags pa 1985, 643:13), (Tsong kha pa Bio bzang grags pa 2000-2004, 3:194). 99rnam ‘jog gzhan la mi Itos pa. (Tsong kha pa Bio bzang grags pa 1985, 643:13), (Tsong kha pa Bio bzang grags pa 2000-2004, 3:194). 100This is the reason why Tsong kha pa does not regard the three characteristics as sufficient for identifying the object of negation (dgag bya). ’Jam dbyangs bshad pa asks

obvious difficulty. The lack of svabhava seems to have exactly the properties of substance-svabhava, so the absence of svabhava should both exist (since svabhava does not) and not exist (since it has the same properties as the nonexisting svabhava). Emptiness (that is, the absence of svabhava) appears to be a contradictory concept.101 A m es’ solution William Ames, in his analysis of CandrakTrti’s use of the concept svabhava tries to resolve this problem by arguing that substance-svabhava and lack of svabhava or emptiness do not collapse into one another since the latter differs from the former in two important ways:102 (1) Being purely negative, it does not satisfy the implicit condition that svabhava be a positive quality. (2) It is not a quality of things, but a fact about qualities of things, namely, that none of them are svabhava. It appears to me th at neither of these supposed differences can be made to carry much weight. The difference between ‘positive’ and ‘negative’ qualities seems to be purely an artefact of language. If our language did not contain the word ‘blunt’ we might describe a blunt knife as ‘not sharp’ and conclude th at sharpness is a positive quality while bluntness is not. If we did not have in the mChan bu bzhi: ’di stong nyid kyi khyad par yin pas dgag byar ga la rung ‘These [three characteristics] being characteristics of emptiness how could they be the object of negation?’ (Jam dbyangs bshad pa, Ba so chos kyi rgyal mthsan, sDe drug mkan chen and Bra sti dge shes rin chen don grub 1972, 387.6). 101 Most philosophers would regard this as a problem with the notion of emptiness. An exception is Graham Priest (2002, 249-270) who regards it as an indication of the fundamentally contradictory nature of reality. 102Ames (1982, 174).

the word ‘sharp’, the reverse would be the case. But we would not have to assume that this indicates any difference between the properties we refer to. Concerning the second point it does not seem to help much to observe that there is a fact about qualities of things which holds continuously, causelessly, and necessarily. All we have done is push up the location of svabhava to the level of second order properties: it is now not the property of heat (or any other first order property) which qualifies as the svabhava of fire, but one of its second order properties, i.e. the property that none of its first order properties is the object’s svabhava, But it is hardly satisfactory for the Madhyamika to avoid the above problem by saying that when he claims that no objects have svabhava what he really means is that none of an object’s first order properties are its svabhava. Tsong kha p a’s solution Another solution to this difficulty is suggested by Tsong kha pa Bio bzang grags pa (1357-1419). Though this Tibetan scholar is separated from Nagarjuna by a considerable temporal, geographical and linguistic distance both the ingenuity of his solution as well as the considerable influence of his thought on the later interpretation of Madhyamaka justify its inclusion in this discussion. Tsong kha pa attem pts to solve the above difficulty by arguing that substance-svabhava (i.e. the Madhyamika’s ‘object of negation’) is to be distinguished from emptiness by its having additional characteristics. Apart from being triply characterized substance-svabhava is also 4. established from its own side103 103rang gi ngo bos grub pa. (Tsong kha pa Bio bzang grags pa 1985, 648:5), (Tsong kha pa Bio bzang grags pa 2000-2004, 3:199).

5. a natural, not a learned notion.104 Concerning the first Tsong kha pa states that Ultimate tru th is established in this way as positing the nature of things (chos nyid) by svabhava (rang bzhin du), but what establishes it as svabhava is the fact that it is not fabricated and does not depend on other objects. It does not in the slightest exist by svabhava which is established from its own side.105 Here Tsong kha pa regards ‘establishment from its own side’ (rang gi ngo bos grub pa) as distinct from ‘independence from other objects’ (gzhan la mi Itos pa) in order to drive a wedge between the characterizations of substancesvabhava and emptiness or absolute svabhava. It should be noted, however, that this interpretation is not shared by all dGe lugs commentators, some of which read Candralurti’s nirapeksah as meaning ‘the establishment of an object from the perspective of its own entity’.106 Concerning the second point it should be noted that Tsong kha pa draws a distinction between conceptions of svabhava which are acquired misconceptions (kun brtags) and those which are innate (lhan skyes). Given the fundamental cognitive change the understanding of emptiness is supposed to bring about he regards the removal of the latter as considerably more important 104fcnn brtags. (Tsong kha pa Bio bzang grags pa 1985, 644:20), (Tsong kha pa Bio bzang grags pa 2000-2004, 3:196). 105 don dam p a ’i bden pa ni chos nyid la rang bzhin du bzhag pa der grub kyang rang bzhin der ['jog byed bcos ma min pa dang | gzhan la m i Itos pa ni rang gi ngo bos grub pa ’i rang bzhin der cung zad kyang med pa. (Tsong kha pa Bio bzang grags pa 1985, 648:3-5), (Tsong kha pa Bio bzang grags pa 2000-2004, 3:199). An alternative translation of this passage is in Magee (1999, 92-93). 10(5 rang gi ngo boH sgo nas yul gyi steng du grub pa. The relevant passage from Ngag dbang dpal ldan is cited in Magee (1999, 94-95).

than the former.107 Later commentaries108 classify the triply characterized svabhava as such an acquired misconception. The triply characterized svabhava is too wide a notion to capture the object of negation, which is therefore further specified as an innate, rather than an acquired misconception.109 A b so lu te svabhava as essence-svabhava Let us conclude by considering one final way of distinguishing substance-svabhava from absolute svabhava in order to solve the apparent contradiction inherent in this understanding of emptiness,110 The basic idea is that, whilst agreeing th at both substance-svabhava and absolute svabhava are characterized as a) not fabricated (akrtrimah), b) immutable (na anyathabhavah), and c) not dependent (nirapeksah), we assume that b) and c) are understood in very different ways for the two different notions of svabhava. But let us consider these three characterizations in turn. Absolute svabhava is described as not fabricated (akrtrimah) or as ‘complete non-origination’ (sarvasa anutpada) to make clear that it is not in any way produced together with an empty object and does not cease once the object is destroyed. It is therefore unlike the hole in a cup or a vase, which is dependent on the cup or vase for its existence and is destroyed if the cup or vase are broken. This point can be clarified by considering CandrakTrti’s assertion that 107(Tsong kha pa Bio bzang grags pa 1985, 644:18-645:1), (Tsong kha pa Bio bzang grags pa 2000-2004, 3:196). 108See Magee (1999, 96). 109 The problem of the differentiation between substance- svabhava and absolute svabhava was later further elaborated in the Tibetan tradition in the context of the debate over self-emptiness (rang stong) and other-emptiness (gzhan stong). For further details see Hookham (1991), Magee (1999, 103-115). 110I thank Mattia Salvini for helpful discussion of this point. CH APTER 2. INTERPRETATIO N S OF SVABHAVA 72 svabhava ‘neither exists, nor does not exist, by intrinsic nature’.111 It is evident th at since svabhava does not exist, it also does not exist by intrinsic nature. But why does it not fail to exist by intrinsic nature? In other words, why does emptiness not exist by substance-svabhava! After all for Nagarjuna phenomena do not just happen to lack svabhava, but could not have possibly had svabhava, no m atter what. W hat Candrakirti wants to say here is th at the property of lacking svabhava is dependent as well, since it depends on the erroneous ascription of svabhava in the first place. It is not a property phenomena have independently of everything else. If someone hallucinates white mice running across his desk then part of what it means that this is a hallucination is that there are in fact no white mice on his desk. But even someone with a rather promiscuous attitude towards existence-claims concerning properties would hesitate to say th at besides being brown, rectangular and more than two feet high the table also has the property of being free of white mice. If there is any distinction to be made between the properties an object has in itself and those which are merely ascribed to it by an observer, purely negative properties such as being not round or free of white mice seem to be the best candidates for being included in the latter category. CandrakTrti stresses this point in a passage dealing with a person suffering from vitreous floaters112 which cause the illusory appearance of hair-like objects in the visual field.113 An ordinary observer would not generally ascribe the property ‘free of hairs’ to an empty pot, as this is one of the countless things the empty pot is empty of. But in order to correct the impression 111 na tadasti na capi nasti svarupatah PP 264:3. The terms svarupa and svabhava are generally used interchangeably by CandrakTrti. 112 rab rib, timira. U3MA 6:29, MAB 106:10-110:3.

of the patient with the eye-disease the pot might indeed be described in this way. The property of hairlessness (like that of the absence of svabhava) is something ascribed to an object to correct a mistaken attribution of the property of being filled with hairs. It is not a property an object would have independently of such an attem pt to rectify a mistake. Emptiness as a correction of a mistaken belief in svabhava is therefore not anything objects have from their own side, nor is it something which is causally produced together with the object, like the empty space in a cup. It is also not something which is a necessary part of conceptualizing objects, since its only purpose is to dispel a certain erroneous conception of objects. In the same way as it is not necessary to conceive of tables as free of white mice in order to conceive of them as all, in the same way a mind not prone to ascribing substance-s^ab/iava to objects does not need conceive of objects as empty in order to conceive of them correctly. When absolute svabhava is interpreted as immutable (na anyathdbhavah), as changelessness (avikaritva) and ever-abidingness (sadaiva sthayitd) this does not mean the same as when for example the Sarvastivadin’s primary existents (dravya) are described in this way. Emptiness is not to be regarded as some unchanging, permanent, absolute reality. CandrakTrti does not mean that if some empty object like a pot or a flower is destroyed the p o t’s or flower’s emptiness somehow stays behind, as it is changeless and ever-abiding. If the pot or flower are destroyed there is no use in referring to their emptiness. The point is rather th at whatever phenomenon is conceptualized by ordinary beings will turn out to be empty, as they will ascribe substance-smfr/mwa to this phenomenon, and it is empty of such svabhava. In this sense emptiness is unchanging, since it is a property to be ascribed to all things ever considered, once they have been correctly analysed.

Finally, when we say that something is not dependent (nirapeksah) there are two different things we can mean. We might want to say th at it does not depend on any object whatsoever or th at it does not depend on some specific object. For example when saying that a mathematical theorem is independent we might make the claim that it does not depend on anything (human beings, minds, the world,...) for its existence, or me might mean something much weaker, namely that it does not depend on some particular thing (the person who proved the theorem, its inscription on a blackboard,...), i.e. th at it would exist if someone else had proved it, or if some inscription or other existed on some blackboard or other. These two meanings can also be employed when speaking about svabhava. We could say that if something exists by svabhava, it does not depend on anything whatsoever. This is the meaning of svabhava usually identified with substance-svabhava and corresponds to the Sarvastivadin’s primary existent. But we could also say some property exists by svabhava if as long as any objects are around they have that property. This, I would want to argue, is the best way to understand the assertion of emptiness being not dependent. It does not mean th at emptiness is some sort of primordial reality ante rem but rather that as long as objects exist, and are conceived of by beings with deluded minds more or less like ours these objects will be empty. The bottom line of this way of resolving the difficulty is the claim that for Nagarjuna there are not three different senses of svabhava, but only two. Absolute svabhava is equated with the essence-svabhava of all objects. In the same way as the property of heat constitutes the essence-svabhava of fire, emptiness, i.e. the absence of substance-svabhava constitutes the essencesvabhdva of all things. There are therefore only two different senses of svabhava to be distinguished, namely essence-svabhava and substance-svabhava\

what I have called ‘absolute svabhava’ turns out to be an instance of the former. Apart from resolving the above contradiction is also allows us to make sense of such characterizations of emptiness as the ‘objecthood of objects’ (dharmdnam dharmata), ‘thusness’ (tathata) ‘intrinsic nature’ (tatsvarupam), or ‘original nature’ (prakrti). These epithets do not equate emptiness with some objectively existent noumenal reality but simply indicate that emptiness is a property all objects could not lose without ceasing to be those very objects. 2.2 T he cognitive dim ension If we conceive of the Madhyamika. arguments about svabhava solely in ontological and semantic terms we are likely to miss one important dimension of the concept which occupies a central place in the Buddhist understanding of emptiness. This is the idea th at the purpose of determining the existence or non-existence of substance-svabhava is not just to arrive at a theoretically satisfactory understanding of the fundamental objects which make up the world, or of the relation between words and their referents, but is supposed to have far more comprehensive implications for how we interact with the world. Nagarjuna notes in the final verses of chapter 26 of the MMK th a t114 with the cessation of ignorance, formations will not arise. Moreover, the cessation of ignorance occurs through right understanding (jnana). Through the cessation of this and that [link of dependent origination] this and that [other link] will not come about. 114 avidyayam niruddhayam samskarandm asambhavah / avidyaya nirodhas tu jnanasyasyaiva bhavanat / / tasya tasya nirodhena tat tan nabhipravartate / duhkhaskandhah kevalo ’yam evam samyag nirudhyate. MMK 26: 11-12.

The entire mass of suffering thereby completely ceases. Nagarjuna claims here that with the realization of the non-existence of substance- svabhava the first link (ignorance) of the twelve links of dependent origination, which constitutes the fundamental Buddhist theory of the generation of the cognitive constitution of the human mind,115 will cease to exist. The first link being cut off, all consecutive links, beginning with formations, will no longer arise. W ith the cessation of the entire chain, Nagarjuna argues, suffering, which is the distinguishing mark of human existence, will cease as well. How exactly the twelve links of origination are to be interpreted, and how the cessation of ignorance brings them to a halt is a complex and much debated question within Buddhist philosophy. It is not one I want to focus on in this context, however. The main idea I want to highlight here is that the cessation of suffering is supposed to be brought about by a cognitive shift, which is constituted by the realization of the absence of svabhava. Candraklrti remarks in his commentary on the above passage th at ‘the one who sees dependent origination correctly does not perceive a substance (svarupa) even in subtle things’.110 Note that svabhava is here not regarded as a theoretical posit, as something an ontologist or semanticist might postulate when investigating the world or its representation in language. The underlying idea here is rather that seeing objects in terms of svabhava is a kind of cognitive default which is criticized by Madhyamaka arguments against svabhava, such as ones described above. It is important to realize that svabhava is seen here as playing a fundamental cognitive role insofar as 115See Wiliams and Tribe (2000, 62-72) for an overview. 116 yo hi pratiLyasamutpadam sam.yak pasyatUi sa suksmasyapi bhdvasya na svarupam upalabhate. PP 559:3-4.

objects are usually conceptualized in terms of svabhava. This conceptualization (which the Madhyamika tries to argue is also theoretically deficient) is taken to be the ultimate cause of suffering. According to this cognitive understanding svabhava is here regarded as a super imposition (samaropa) which the mind naturally projects onto objects when attem pting to conceptualize the world. The term samaropa is only mentioned once by Nagarjuna in the MMK,117 but acquires a more prominent role in CandrakTrti’s commentary. I think that agreeing with CandrakTrti about the presence of a notion of svabhava as superimposition in Nagarjuna1 s arguments allows us to give a theoretically coherent account of his view of svabhava,118 while it also helps us to understand why the establishment of the absence of substance- svabhava occupies such a central place in Madhyamaka thinking.

Candrakirti argues th at the understanding of svabhava in terms of a superimposition is of central importance for understanding the entire intellectual enterprise of the MMK: Thus, when it is said th at entities do not arise in this way first of all the initial chapter was written to counter the mistaken attribution (adhyaropa) of false intrinsic natures; the remaining chapters were w ritten to eliminate whatever distinctions are superimposed anywhere.119 117MMIC 16:10. See Tanji (2000, 352, 355). 118This does deny that establishing what is ‘really meant’ by Nagarjuna’s arguments is in many cases more difficult to establish than for other philosophers (Tillemans 2001, 17), (Griffiths 2000, 24). Internal coherence of the arguments presented and external coherence with the context of Nagarjuna’s thoughts is all I think can be reasonably expected from a presentation such as this. 119 tasmad anuppand bhdvd ity evam tavad vipanta-svarupa-adhyaropa-pratipaksena

It is important to note that CandrakTrti is not merely concerned with the refutation of a theory he assumes to be mistaken, but with something more fundamental: For one on the road of cyclic existence who pursues an inverted [view] due to ignorance a mistaken object such as the superimposition (sam,dropa) on the aggregates appears as real, but it does not appear to one who is close the view of the real nature of things.120 Independent of one’s particular theoretical position concerning the existence or non-existence of svabhava, svabhava is something which is superimposed on ordinary objects in the process of conceptualization. The five aggregates, for example, are seen as a single, permanent, independent self, due to the superimposition of svabhava on such a basis. The same happens when ordinary partite and causally produced material objects, linguistic items, and so forth are apprehended. It is because this cognitive default of the superimposition of svabhava is seen as the primary cause of suffering that the Madhyamika draws a distinction between the understanding of arguments establishing emptiness and its realization. Being convinced by some Madhyamaka argument that an object does not exist with svabhava does usually not entail that the object will not still appear to us as having svabhava. The elimination of this appearance is only achieved by the realization of emptiness. The ultimate aim of the prathamaprakarandrambhah \ idanim kvacid yah kascid viseso ’dhyaropitas tad visesdpakarandrtham sesaprakarandrabhyah. PP 58:10-11. 120samsarddhvani vartamandndm avidya-aviparydsa-anugamdn mrsdrlha eva skandhasamaropah satyatah pratibhasamdnah paddrbha-tattva-darsana-samvpasthdnam na pratibhdsate. PP 347:1-3.

Madhyamika project is therefore not just the establishment of a particular ontological or semantic theory, but the achievement of a cognitive change. The elimination of svabhava as a theoretical posit by means of arguments such as those presented above have to be followed by its elimination as an automatic cognitive superimposition by means of specific practices. But what kind of evidence is there that svabhava constitutes an automatic cognitive superimposition? I agree with Tillemans that for anyone trying to establish this ‘the Indian Madhyamika literature would offer very little evidence, apart from a number of quotations from scriptures and a lot of doctrinal talk about people being ignorant, under the influence of karma, etc.’.121 However, it might be possible to adduce some evidence from other sources which make this assumption at least plausible. Buddhist philosophy generally assumes that the superimposition of svabhava applies to two things: to the self and to other phenomena we encounter. This superimposition at least entails conceiving of the self as unitary and permanent, and also viewing objects as external or observer-independent, as well as permanent We will have more to say 011 the former when considering Nagarjuna’s analysis of the self later on, so let us at the moment just consider our perception of objects. I would like to suggest that there is a cognitive default which a) determines that other things being equal we conceive of a sequence of stimuli as corresponding to a single enduring (though changing) object, rather than to a sequence of different, momentary ones and b) makes it more likely th at we assume an external rather than internal objects as the source of the stimulus. Let me call these the principle of permanence and the principle of externality. 121Tillemans (2001, 18).

The principle of permanence ensures that we generally conceive of objects as enduring phenomena which may change over time, but still remain fundamentally the same object, rather than as unrelated momentarily arising and ceasing phenomena, each of which lasts only for an instant. It should be noted that this latter way of interpreting the information we get through the senses is not in any way logically deficient, it is just not the way we see the world. There are good reasons why we do not do so, primarily th at such a representation is vastly too complex to use in practice. Any mind who lived in such a world of kaleidoscopically flashing phenomena would presumably be at an evolutionary disadvantage when compared to one which represented a world of stable, enduring objects. The principle of externality makes us assume that the causes of sensory stimuli are objects lying outside of us, rather than the product of our own perceptive mechanism. We generally assume that our perception is evidence for things lying outside of ourselves and that we do not live in a hallucinatory world of our own devising. Again, such a principle makes evolutionary sense: running away from an imaginary tiger is not as detrimental to our chances of passing on our genes as is declaring a real tiger rushing towards us to be a figment of our imagination. Whether the principles of permanence and externality really determine our conceptualization of the world is of course an empirical question which can hardly be decided in a philosophical discussion such as this. W hat we can do, however, is to acquaint the reader with two simple empirical results which could serve as evidence that something like these two principles might play an im portant role in our cognitive access to the world.

The first is the so-called beta phenomenon which has been known to experimental psychologists for a long time.122 The subject of the experiment is shown two slides, the first of which contains a dot in the top left-hand comer, the other a dot in the in the bottom right-hand corner. W hat the subject perceives if these slides are shown in quick succession is not two stationary dots, but a single dot moving diagonally from the top left to the bottom right across the slides. W hat has happened here is that our brain has interpreted the sequence of two stationary dots as a single moving object which is first seen on the left and then on the right. Rather than interpreting this particular stimulus as one object appearing at one spot and immediately disappearing, which is followed by another object appearing at a different spot the principle of permanence causes us to see the two dots as indications of a single object changing its position in space. When offered the choice of either regarding some sequence of stimuli as corresponding to a series of momentarily arising and ceasing objects, or as an enduring object changing its attributes our brain seems to opt automatically for the latter.123 Some evidence for the principle of externality can be drawn from the psychological investigation of dreaming, in particular of the phenomenon of lucid dreaming.124 A lucid dream is a dream in which the dreamer is conscious of dreaming without waking up. Although lucid dreams happen spontaneously 122Tlie earliest description of the beta phenomenon is in Wertheimer (1912); further results and interpretations are given by Dennett (1991) (who erroneously refers to it as the phi phenomenon) and Hoffman (1998). 123The problem of ‘object permanence’, i.e. of the question when two distinct perceptions of an object are regarded as being caused by a single thing has been investigated extensively in developmental psychology. See Piaget (1937) and Spelke (1990) for two now classical discussions. An interesting related experiment is described in Subbotskii (1991). LaBerge, Levitan and Dement (1986).

to some people there are also a variety of techniques for inducing them.125 But the fact th at some special effort is required to have a lucid dream points to the fact that our natural reaction to perceptions in dreams is to regard them as caused by external objects, rather than by our own mind. So it seems that our view of sensory information both in the waking state and in the dream state is generally determined by the principle of externality: in both cases we regard the source of the information to be something which is both external to us and existing independently of us. It requires a particular cognitive effort to question in a dream whether the things one sees are indeed caused by external sources, an effort which appears to be essential in inducing lucid dreaming. If it is plausible to understand the Madhyamika’s notion of superimposition (samaropa) of substance-svabhava in terms of certain cognitive defaults (such as the principles of permanence and externality)126 which govern our representation of the world then it becomes clear why the Madhyamika draws a sharp line between intellectual understanding and realization. As familiarity with any optical illusion attests, neither merely understanding that it is an illusion, nor even understanding how it works, will make the illusion go away. Now if there was a way of training oneself out of perceiving a particular illusion we would have an apt example of the relation between understanding and insight as seen by the Madhyamika. First of all we have to gain an understanding of how the illusion (in this case the superimposition of substanc e-svabhava) works, and in a second step we have to follow some training which eventually makes even the appearance of the illusion go away. 125LaBerge and Rheingold (n.d.). 126 Further investigation of our perceptions of the self, of causality, or of mereological relations might provide other aspects which cohere with the view of svabhava as a superimposition.

But now this point also indicates the limitations of appealing to results from cognitive science for gaining a better understanding of svabhava. Even though such references are useful in giving us an idea of why the Madhyamika’s view of superimposition could be plausible they give us very little insight into how the removal of such superimpositions could be possible and what it would entail. The reason for this is obvious: according to the traditional Buddhist view those who have realized (as opposed to merely understood) the absence of svabhava and thereby emptiness are few and far between. Empirical research into the way such persons perceive the world is therefore naturally difficult. Fortunately this is not a task the present discussion has to achieve. For our purposes it is sufficient to point out that a mere understanding of svabhava as a theoretical posit (arrived at within an ontological or semantic theory) is not sufficient for understanding the central role it occupies in Buddhist thought. The notion of svabhava must also be something which plays a much larger part in the mental life of the majority of people who are after all neither ontologists nor semanticists. The cognitive understanding of svabhava provides us with an interpretation which achieves this.

The role of negation in Nagarjuna’s arguments

The preceding chapter has acquainted the reader with the main objective of Madhyamaka thought, that is the rejection of svabhava. Before we can discuss the further ramifications of this idea in Nagarj una’s philosophy it is necessary to discuss some formal aspects of his arguments which the reader primarily acquainted with Western philosophical literature might find puzzling. They arise mainly from specific logical and methodological considerations connected with the concept of negation which were widespread in Indian philosophy but are not always shared by the Western notion of negation derived primarily from formal logic. W hat makes the issue particularly intricate is the fact that there exists a tension between some presuppositions of the traditional Indian account of negation and the contents of Nagarjuna’s philosophical views, so th at Nagarjuna sometimes sees himself challenged to adapt these presuppositions in order to formulate his philosophical position. In the present chapter I will discuss Nagarjuna’s view of the standard view of negation as presented in the Nyaya system. The following chapter

will discuss a specific form of argument, the catuskotz or tetralemma, which is frequently employed in Nagarjuna’s writings and essentially involves single and iterated negations. Nagarjuna’s central argumentative aim is to develop a philosophical theory which does not have recourse to the notion of substance or svabhava. His main strategy is to examine all the possible ways in which a particular phenomenon (such as physical objects, causation, the self, language etc.) could be thought to exist with svabhava, and conclude that on close inspection none of these is satisfactory. It then remains to conclude that the phenomenon in question does not exist with svabhava. Since many of Nagarjuna’s conclusions are therefore negative ones it is essential to gain a clear understanding of the role of negation in his philosophical system. This is more difficult than it may sound initially, especially as Nagarjuna’s discussion of these matters, the greatest part of which is to be found in the VV and, to a lesser extent, in the VP is formulated against the background of the Nyaya theory of negation. This differs significantly from accounts of negation with which the reader primarily acquainted with Western philosophical discussions is likely to be familiar.

3.1 Nyaya theory of negation

The philosophical system known under the name ‘Nyaya’ incorporates a comprehensive theory of logic and epistemology which proved to be extremely important in Indian intellectual history, influencing not only different kinds of philosophical inquiry, but also such disciplines as linguistics, poetics, rhetoric,

and law.1 The system is based on the Nyayasutra attributed to G autam a (also known as Aksapada). There is little agreement on when the sutra was composed; the dates proposed range from the 6th century B.C. up to the 2nd century A.D.2 It is, however, relatively clear that they achieved the form in which they have been transm itted to us around the time of Nagarjuna and might even have in parts been composed as a reply to Nagarjuna’s arguments. 3 We must therefore keep in mind that when we refer to the relations between Nagarjuna’s Madhyamaka and Nyaya we are dealing with a very early phase of the latter. At Nagarjuna’s time none of the long sequence of works on Nyaya,4 including Vatsyayana’s Bhasya, had been w ritten yet.5 In order to understand the Nyaya theory of negation we have to note that in the underlying ontological system properties are seen as separate entities over and above the substrata in which they inhere.6 Now the Naiyayika regards the absence (abhava) of a property as a category in its own right 1Matilal (1968, 21), Potter (1970-2003, II, 1-3), Guha (1979, 1-2). A concise summary of the assumptions underlying the Nyaya system is given in Matilal (1986, 5-6). 2Potter (1970-2003, II, 4). Jha (1939, viii) even suggests a date of composition as late as the 6th century A.D. A comprehensive account of the history of the composition of the NS is given in Meuthrath (1996).

Bronkhorst (1985) argues the greatest part of the NS existed before Nagarjuna and was known to him, and that some parts were added later in response to Madhyamaka objections. Bronkhorst’s conclusions have been severely criticized by Oetke in his (1991) and (1997). For a reply see Bronkhorst (1993). 4Potter (1970-2003, II, 9-12) 5For more details on the relation between the VY and the NS see Meuthrath (1999). 6MatiIal (1968, 16). The Nyaya system of logic and epistemology usually relies on the Vaisesilca ontology. The association between the two systems is so close that one often refers to them jointly as Nyaya-Vaisesika. Opinions differ on how the two systems came to be associated, and even if we are to speak of two systems rather than one. See Potter (1970-2003, II, 12-13), Bronkhorst (1985, 123-124). CHAPTER 3. NEG ATIO N 87 (padartha)7, as something which can equally be possessed by a substratum .8 The referent of a negative statement like ‘there is no pot in the house’ is therefore regarded as the qualification of the house by an absence, namely the absence of a pot.9 Judgements, whether they are affirmative (‘There is a pot in the house5) or negative (‘There is no pot in the house5) involve a qualification or an attribution, which can be either an attribution of a presence or an attribution of an absence. As such the attribution can be construed as either affirmative or negative. Whether it is expressed by an affirmative or a negative statement is then a question about how the judgement of th at attribution is formulated in ordinary language, not a fact about the attribution itself.10 According to the Naiyayika account a judgement is correct if it combines some parts of reality in the way in which they are indeed combined (such as the house and the absence of the pot, if there is no pot in the house), and it is erroneous if it combines elements from reality in a way in which they are not combined (such as the house and the absence of the pot if there is a pot in the house).11 Error will therefore always arise from the way elements are 7Chatterjee (1939, 166-168), Sharma (1970, 11-12). 8We will not go into the reasons offered in defence of this assumption. Sharma (1970, 3-11) argues that the conception of liberation (moksa or apavarga) espoused by the Nyaya- Vaisesiltas made it ‘imperative for this system to posit Absence as an ultimately real entity (padarthay (6), so that this ontological point also acquired soteriological significance. 9Matilal (1968, 3). 10Matilal (1968, 92-93). 11Matilal (1970, 95). As such there is a close similarity between this view and the familiar correspondence account of truth we find for example in a semantics based on states of affairs. There the sentence ‘there is a pot in the house’ is regarded as true iff the referents of the constituents of the sentence (the pot, the house, the ‘inside o f’ relation) are arranged in structurally the same way in the state of affairs in which the constituents are arranged in the sentence. The main difference is of course that this view does not

combined in judgements, never from simple perceptions. For the Naiyayika ‘a simple, non-complex property can never be empty’.12 3.2 N egation and non-denoting term s This leaves us with a substantial difficulty if we want to deny the existence of certain entities, for the Naiyayika faces the very problem Quine observes at the beginning of ‘On what there is’:13 When I try to formulate our difference of opinion, I seem to be in a predicament. I cannot admit that there are some things which [my opponent] McX countenances and I do not, for in adm itting that there are such things I should be contradicting my own rejection of them. Suppose we want to say that Pegasus does not exist (or th at every place is characterized by an absence of Pegasus). Such a judgement would intuitively be regarded as true, but for the Naiyayika a true judgement has to combine elements of reality in the right way. But Pegasus is not an element of reality, since he does not exist, Matilal observes th a t14 the property of Pegasus-ness thus arrived at would be, according to Nyaya, unexampled or fictitious because it has no locus to occur in, th at is, no locus possesses this property. In such cases, Nyaya asserts that we cannot even say that such a property is absent or does not occur somewhere. [...] Thus, if a sentence is accord any ontological status to absences, as tlie Nyaya theory does. Matilal (1970, 96).3Quine (1953, 1). Matilal (1968, 154-155).

said to express an absence of such an unexampled property, it becomes nonsensical. Now of course this does not mean that adopting the Nyaya semantics we have no way of saying that a possible entity (like Pegasus) or an impossible one (like the round square) does not exist. If this was indeed the case it would not be a very satisfactory theory to begin with. W hat we have to do is to rephrase the statement to make evident that the source of the vacuity of such properties is the combination of more basic properties each of which exists in reality, but which are not thus combined in the world. So the Naiyayika can make the meaningful assertion that wingedness and horseness are never combined in the same animal, and th at roundness and squareness are never combined in the same figure.15 All the properties referred to now have referents in reality, so the Naiyayika1 s semantics is taken care of.16 There is an obvious similarity between this approach and the Russellian procedure of replacing a non-denoting term by a definite description.17 This view of non-denoting terms is just a reflection of the fact that for the Naiyayika language must hook up with the world at some fundamental level via a denotation relation. Even if there are non-denoting terms in our language they can only exist parasitic on denoting terms. Simple designators are therefore guaranteed to refer, whilst complex designators may or may not do so. This of course means th at according to Nyaya theory negative statements involving simple designators (statements of absence of some entity) can only ever be statements of a local absence and will always entail the 15See Uddyotakara’s commentary on the Nyayasutrabhasya 3.1.1 (Chakravarti 1982, 232-233). 16Matilal (1968, 9, 23). Matilal (1970, 85), Chakravarti (1982, 211-212),

presence of that entity somewhere else.18 In his Varttika on Vatsyayana’s Bhasya on 2.1.12 of the NS Uddyotakara notes th a t19 when the word ‘ja r’ is coordinated with the term ‘does not exist’ it does not convey the non-existence of the jar; all that it does is to deny, either the [spatial] connection of the jar with the house or its specification [as located at] a particular point in time. Since the statement ‘There is no pot in the house’ or ‘There is an absence of a pot in the house’ is only meaningful if the pot referred to does indeed exist it must be present somewhere else. It would be nonsensical if there were no pots at all, at least if ‘p o t’ is regarded as a simple designator. Vatsyayana raises this objection in his commentary on the passage from the NS 2.1.11, where he argues against the opponent’s attem pt to deny the existence of means of knowledge:20 If you want to deny the existence [of the means of knowledge, this denial] implies their existence, and the refutation of [means of knowledge like] perception and so forth is not accomplished. Phanibhusana’s subcommentary elucidates this point by adding th a t21 18Phanibhusana (Chattopadhyaya and Gangopadhyaya 1968,11:28) remarks that ‘there is not sense in denying the absolutely non-existent like a flower imagined to blossom in the sky. The denial of something can have sense only when its existence is admitted somewhere else, e.g. to say that there is no jar in the room means that it exists elsewhere.’ 19 na hy ayam ndsti nasamdnadhikarano ghatadisabdo ghatdbhavam pratipadayati api tu gehaghatasamyogam vd kalavisesam va [...] pratisedhati (Gautama 1887, 192: 20-22), (Jha 1984, II: 623). 20 tod yadi sambhavo nivartyate sati sambhave pratyaksadinam pratisedhanupapatih. (Nyaya-Tarkatirtha and Tarkatirtha 1985, 425: 2-3). 21 Chattopadhyaya and Gangopadhyaya (1968, 2: 26).

the very attempt to deny their existence presupposes the admission of their existence inasmuch as there is no sense in demolishing the possible existence of something which has no existence at all, just as it is impossible to smash with a stick the jar which does not exist. The idea that simple designators have to have a denotation seems to possess some intuitive plausibility. Suppose I present you with some simple, non-denoting name such as ‘Hopzik’. You could not be taught the meaning of TIopzik’ by ostension (since there are none), nor could I give you an analysis in terms of other properties (since it is a simple designator). But then the problem occurs of how you could make sense of any sentence containing the term ‘Hopzik’, including statements such as ‘Hopzik does not exist’. We do not have any idea what this negation means, since we do not have any positive notion of the entity being denied. Having taken account of this background it is understandable that the issue of non-denoting terms is raised by the Naiyayika opponent against the central philosophical thesis of Madhyamaka that there is no svabhava. He argues that if the Madhyamaka claim was true and if there was indeed no svabhava, then the claim would be nonsensical. For if a negative statement about svabhava was interpreted along the same lines as a negative statement about a pot in the house,22 we would have to hold that the existent svabhava stood in an absence relation to the world, in the same way as the pot stood in an absence relation to the house. But if we understand the statement in this way svabhava exists after all and so the Madhyamaka thesis must be false. The statement ‘there is no svabhava1 has to be either false or nonsensical, 22 VV 11. CHAPTER 3. NEGATION 92 since ‘there is no name without referent’.23 A later manifestation of the same difficulty can be found in certain problems connected with formal reasoning.24 According to the Nyaya theory a formally set out ‘inference for oneself’ (svaHhanumana) establishing that the subject (paksa) has the qualifying property (sadhya) must provide both an agreeing and a disagreeing example (udaharana) ,25 Thus, in order to establish the thesis th at all white things are coloured we need both an ‘agreeing’ example of a subject having the qualifying property (such as a conch shell, which is both white and coloured), as well as a ‘disagreeing’ example of the contraposed version (‘whatever is not coloured is not white’, space, which is neither, being a case in point).

But we realize th at this leads to a problem if we assert a universal thesis like ‘all things lack svabhava’ or ‘all things are momentary’. In this case the disagreeing example would have to be an instance of some object which has svabhava or is not momentary. Since we want to establish that there are no such things, the term has to be empty. But if the occurrence of an empty term renders a statement meaningless, as the Naiyayika asserts, it immediately follows that the two theses cannot be established. If, on the other hand, the two theses are meaningful (as they appear to be) then they must assert absences of existent things, and svabhava or the property of permanence must exist. So once again we are faced with the unenviable choice between falsity and nonsense. 2Znarna hi nirvastukam nasti. VV 9. 24This is discussed in Matilal’s analysis of part of the Atmatattvaviveka by Udayana (10-11th century AD) in (1970). 25For a concise summary of the Nyaya theory of inference see (Potter 1970-2003, II, 179-208).

Nagarjuna suggests a variety of possible replies to the Nyaya difficulty of non-denoting terms. First of all he remarks th a t26 To one who says that the name is sadbhuta you would have to reply: ‘There is svabhava?. [ ...] [However,] since things have no svabhava, that name also lacks svabhava. Because of this it is empty, and, being empty, it is not sadbhuta. Bhattacharya27 here translates sadbhuta as ‘existent’. This does not strike me as a very fortunate rendering, primarily because Nagarjuna does not want to claim th at emptiness entails a lack of existence. It is evident th at what Nagarjuna wants to say in the first sentence is that if the Nyaya account was indeed correct (and each simple term in a negative statement had to denote an existent object) svabhava would exist. It therefore seems plausible that saying th at a name is sadbhuta is not supposed to mean th at the name exists (something which neither Nagarjuna nor his opponent denies) but that it functions in accordance with Nyaya theory: that each non-complex term is hooked up with a designated object in the real world. But this is exactly what the proponent of emptiness denies. For the Madhyamika a satisfactory semantics cannot consist of an objectively existent reference relation which links the terms of our language to an objectively existent world. He will argue that both the question how the world is sliced up into individual entities and the question of how these entities link up with the parts of language are to be settled by convention. There is no ‘ready-made world1 of simples out there which could provide the semantic 26 yo namatra sadbhutam bruydt sasvabhava iti evam bhavatd prativaktavyah syat | [...] tad api hi bhdvasvabhavasya abhdvan ndma nihsvabhavam tasmdc chhunyam sunyatvad asadbhutam. VV(S) 76:16-77:2. 27Bhattacharya, Johnston and Kunst (1978, 128). CHAPTER 3. NEGATION 94 foundation for the simple terms of our language.28 Nagarjuna therefore argues in this passage that the Naiyayika criticism is justified only if one is antecedently convinced of the Nyaya picture of semantics. This, however, is something which the Madhyamika does not want to share. We might wonder at this point why a relatively obvious reply to the Naiyayika worries about statements like ‘there are no objects existing with svabhava1 is not made by the Madhyamika. This reply consists in arguing that svabhava is not a simple designator. As we saw above it is possible to assert th at there are no unicorns even on Nyaya terms. This is done by rephrasing the statement as an assertion about real entities (such as wingedness and horseness) which do not occur together. We could now similarly break up the property ‘existing with svabhava’ into its simpler components (such as not depending causally on other phenomena, not depending notionally on other phenomena and so forth) and argue that since all these dependence relations exist, all we are asserting by a statement denying svabhava is that there is no object which is qualified by the absence of all these dependence relations at the same time. This reply is certainly adequate for answering the Naiyayika worry, but it is hardly a position a Madhyamika would want to adopt without further qualification. For, according to the Naiyayika’s interpretation of this answer, there is still a world of objectively existent simple properties which the simple terms in our language refer to in a way which is independent of linguistic conventions. While the statement ‘there is no svabhava1 is thus at least rendered comprehensible to the Naiyayika, this comes at the price of accepting a view of semantics fundamentally at odds with the one the Madhyamika is 28See chapter 9 for further discussion of this point.

arguing for. For this reason trying to establish that svabhava is a complex designator is not a very satisfactory response by Madhyamaka standards. Within the context of discussing the problem of negating a non-denoting term Nagarjuna also discusses a somewhat curious objection raised in W 12. The opponent asks about the point of negating a non-existent object since 'the negation of a non-existent, such as the coolness of fire or the burning of water is established without words’.29 This passage entails th at there are two kinds of negation for the opponent: those established without words, and those established with words. Looking at the choice of examples used it seems that members of the former group include negations of an essential property of an object, such as the heat of fire. W hat kinds of negation are included in the latter group is not quite clear. A reasonable assumption is to include negative contingent statements in here (such as 'There is no pot in the house’). Now the opponent does not want to say th at the absence of the pot in the house is brought about by the assertion 'there is no pot in the house’: just saying it is so does not make it so. There are two different ways of understanding what the opponent could mean here. Firstly, note th at the Naiyayika does not draw any fundamental distinction between the judgements expressed by affirmative and negative statements. As Matilal observes30 all determinate cognitions or judgemental cognitions (savikalpa jnana), which can be very well regarded as the counterpart of statements, involve a qualification or attribution, and such an attribution cannot be construed as either affirmative or negative. 23asato hi vacanadvina siddhah pratisedhah tadyathagneh saityasya apamausnyasya. W (S ) 51:6-7. Matilal (1968, 92).

The affirmative-negative distinction is not one the Naiyayika regards as ontologically fundamental. Whether a particular statement is affirmative or negative just mirrors the way the particular qualification or attribution is expressed in ordinary language. W hat is meant therefore by saying that a negation like ‘There is no pot in the house’ is established with words is that it is only by the force of language that a negative meaning is expressed. The world itself contains no negations, only presences and absences of different kinds. It is only the words which bring negations into existence. The absence of a non-existent entity (such as the absence of the opposite of an essential quality, like the absence of coolness in fire) can be regarded as a ‘negation established without words’ (vacanad vina siddhah pratisedhah) since it cannot rely on language for its expression as a negation, as the Nyaya restrictions on non-denoting terms render the phrase ‘absence of coolness of fire’ meaningless. It is of course the case that there is no coolness in fire, but, for the sake of argument assuming th at ‘coolness of fire’ is a simple designator, this is not something which can be expressed in language, nor it is something which indeed need to be expressed. We might, after all think there is a pot in the house when there is none, but who would think there is coolness to be found in fire? Secondly we can understand the claim that some negations are established with words, while some are established without not as referring to the negation itself, but rather to the corresponding cognition of the negation created in our mind.31 To establish a cognition like the one th at there is no pot in the house we generally need language (as we are not able to inspect all the parts of a house at once). Other cognitions, however, like those that 31 As argued below on page 100 this is also the most satisfactory way to understand the opponent’s worry about the temporal relations between negation and negated object.

fire is not cool and water is not burning are established without linguistic mediation. Our acquaintance with fire or water directly acquaints us with these essential properties of them. There is therefore no need to assert the absence of properties opposed to the essential nature of things in language. As the claim concerning the absence of svabhava is of the very same nature it is therefore not obvious what the point of asserting this negation is supposed to be. Nagarjuna replies by saying that the purpose of a statement like ‘there is no svabhava5 is to point out the absence of svabhava, not to create the absence of something which is in fact there.32 He uses the example of saying ‘Devadatta is not in the house5 when he is in fact not there — this statement obviously just reports the absence of Devadatta and does not cause him not to be in the house.33 We noted above that it is quite unlikely that the opponent should hold such a peculiar position; unless what is created is supposed to be the cognition of the absence, rather than the absence itself. In this case Nagarjuna could be interpreted as saying that both kinds of negation, those involving essential and non-essential properties have to be linguistically mediated to be cognized by us. W hat would be replied to our first interpretation of the opponent’s worry, 1.e. the position that the absence of something necessarily non-existent (such as the coolness of fire, or the svabhava of things) cannot and need not be 32VV 64. 33A similar point is made by Uddyotakara in his Varttika on Vatsyayana’s Bhasya on NS 2, 1, 11: ‘negation does not have the power to make an existing thing otherwise [i.e. nonexistent]. Because it makes something known it does not cause the existence of something to cease; [therefore] this negation [too] makes something known and does not cause the existence of something to cease.’ na ca pratisedhasyaitat samarthyam yad vidyamanam padartham anyathd kuryat jndpakalvdc ca na sambhavanivrttih jndpako ’yam pratisedho na sambhavanivarlaka iti. (Gautama 1887, 191: 13-15), (Jha 1984, II: 619).

expressed? We have already seen above that the impossibility of expressing such negations depends on the peculiar nature of the Nyaya semantics which the Madhyamaka will not want to accept. And concerning the need to express this negation, the Madhyamaka will argue that whilst nobody in his right mind will think fire to be cool (and could therefore benefit from having this pointed out to him) the belief in the svabhava of things is extremely widespread, and realizing its falsity is one of the essential preconditions of liberation. Thinking that there is svabhava in things is like thinking there to be real water in a mirage. Deceived people who are likely to run towards the mirage to quench their thirst will benefit from getting to know th at there is no real water there, as ordinary people will benefit from learning that things exist without svabhava.34 Nevertheless we might think that when we negate water perceived in a mirage there is still the person perceiving the mirage, its perception, and the perceived object (i.e. the mirage), as well as the person doing the negating (us), the negation, and the object negated (namely real water in the mirage). 35 But if this is the case, the Naiyayika will argue we are facing the familiar problem again: if the thing we negate (the mirage, svabhava) does exist after all the statement asserting its negation is obviously a falsity. If, on the other hand none of them exist, if there is no perceiver, perception, and perceived object, no negator, negation and object to be negated36 then the Madhyamaka argument vanishes altogether and the existence of svabhava is established by default.37 34 VV 13. 35VV 14. 360 f course what the opponent must mean here is that the object of negation (i.e. svabhava) does not exist as an object of negation (because there is no negation), not that it does not exist at all. 37VV 15-16.

W hat we have to say here (and what Nagarjuna in fact does say)38 is that all these things can exist without existing in quite the way the Naiyayika supposes. As we have seen above for the Naiyayika the different entities referred to by the simple designators in a negative statement (a statement of absence) each have to be real for the statement to be meaningful. The Madhyamika, however, thinks that an unreal entity, such as the water in a mirage, or the appearance of svabhava can be very well be the object of an (erroneous) cognitive state and also able to be referred to in a true sentence. Prom the fact th at ‘something has become the ‘object5 (visaya) of a cognitive state it does not follow that it must have been causally related to the production of th at cognitive state.539 This is due to the fact th at for the Madhyamika the source of error is not exclusively located in the erroneous combination of individually existing properties, as the Naiyayika assumes.40 The example of the mirage presents us with the case of a simple yet erroneous perception. As long as we assume that the object of perception and the object of negation are all dependently arisen objects, rather than entities existing in their own right we can deny their existence without antecedently having to regard them as real.41 38VV 65-G7. 39Matilal (1970, 94). 40Matilal (1970, 96). 41VP 16. This point is also underlined by CandrakTrti in commenting on MMK 15:11: ‘A healthy person does not perceive any of the hairs which appear to one afflicted by vitreous floaters. When he says ‘these [hairs] are not’ he does not say that they are an existing entity the existence of which was denied because the object of negation is not real. In the same sense we say that ‘all things are not’ in order to remove clinging to an error in those who see things in the wrong way like those afflicted by vitreous floaters.’ yas tu tairnirikopalabdha kesesviva vitaimiriko na kimcid upalabhate sa nasti iti bruvan kimcin nasti iti bruydt pratisedhyabhdvdt \ viparyastanam tu mithyd-abhinivesa-nivrtty-artham ataimirika iva vayam brurno na santi sarvbhdva iti. PP 273:14-274:3.

Even though the term ‘the water in the mirage’ is non-denoting, since there is no water in the mirage, there is still something created by the interplay of our senses, light, and heat on which the presence of water is superimposed, which we can subsequently deny. Similarly our language and general cognitive habits can, the Madhyamaka argues, create the unreal superimposition (samaropa) of svabhava which Nagarjuna’s arguments set out to refute. 3.3 N egation and tem poral relations Apart from worries about negative statements involving non-denoting terms the second main difficulty to do with negation raised by Nagarjuna’s opponent concerns the possible temporal relation between a negation and the object negated.42 This is a form of argument which we encounter frequently in Nagarjuna’s works, the most prominent examples being in the discussion of the relation between means of knowledge and object known, and between cause and effect. The worry of the opponent is that there is a general difficulty with negative statements (including the Madliyamika’s assertion that there is no object with svabhava). He argues that the negation can neither exist before, after, or at the same time as the object of negation, and therefore cannot possibly exist. Now this might strike us as a strange position to maintain. As we usually regard negation as a logical relation temporal considerations seem to be wholly irrelevant, in the same way as there is no point in asking which numbers in a mathematical equation are there first. In order to see the point at issue here we have to note th at for the 42VP 20, VP 13-15. The same problem is raised in NS 2.1.12 (NS 425-426). CHAPTER 3. NEGATION 101 Naiyayika the negation is the instrument making known a particular absence of a quality in some substratum. This making known is obviously a causal process,43 so th at it is clear th at the Naiyayika worries here just stem from an application of Nagarjuna’s criticism of causation44 to epistemology. The Naiyayika will argue that if causes and effects cannot exist standing in any of the three temporal relations (as the Madhyamika sets out to show) then this must also apply to epistemic causes and effects,45 and thus also to negations, which constitute one particular kind of epistemic cause. Therefore, if we take the Madhyamaka view of causation seriously we have a problem with establishing the negative thesis th at there are no objects with svabhava. The argument itself proceeds in the expected manner. The opponent argues that the negation cannot exist before the object to be negated, because then there would be a negation without anything negated. More worryingly, if the negated object does not (yet) exist, what is the point of negating it? Nor could the negation exist after the object of negation, for what is the point of negating something existing? We also have to note th at it is only the causal interpretation of negation in this context which allows us to make sense of the opponent’s criticism of the simultaneous existence of the negation and its object. We might think that this was indeed a satisfactory way of thinking of the two (in the same way in which we might think th at all the numbers in a mathematical equation exist at the same time). But considering negation in causal terms we face the problem that ‘the negation is not the cause of the object known by negation, nor is the object known by negation the cause of the negation’.46 As illustrated by the familiar analogy of the two 43See Jha (1984, 621). 44 See the discussion in chapter 5. 45VP 12. 4Gna pratisedhah pratisedhasyarthasya karanam pratisedhyo na pratisedhasya ca. VV(S).

horns of a cow47 which do not cause each other, in the case of simultaneously existing cause and effect we would have a problem in establishing which is which, since the conceptual distinction between cause and effect is drawn in terms of temporal priority. There are various ways in which one can respond to this problem. In VV 69 Nagarjuna tries to turn the tables on his opponent.48 As we saw above for the Naiyayika the existence of a negation is equivalent to the existence of an object of negation, i.e. of an object the absence of which in a particular substratum could be asserted. But if there is no negation, as the Naiyayika has just been trying to argue, there is also no object the absence of which can be asserted and therefore it follows on the Naiyayika’s own terms that there is no svabhava.49 A more general way of replying is to point out that in the same way as we can still talk about causal relations, even if the realist’s picture of causality turns out to be unsatisfactory, in the same way the fact th at some epistemic process cannot be made sense of in Nyaya terms does not mean 54:13-14. This translation appears to me more satisfactory than Bhattacharya’s, who just has ‘object of negation’ for pratisedhasya drtha (Bhattacharya et al. 1978, 106). It would seem very peculiar to ascribe to Nagarjuna’s Naiyayika opponent the view that the object of negation is causally brought about by the negation (or the other way round). 47See Bhattacharya et al. (1978, 106, note 1) for a list of references to this example. 48‘By virtue of your statement a negation is not possible in the three times, and, like the negation, the thing to be negated also [does not exist].’ tatha hi tvadvacanena pratisedhastraikalye ’nupapanna pratisedhavat sa pratisedhyo ‘pi. VV(S) 83:17-18. 49In VP 14-15 we find a different reply. Here the opponent argues that once the existence of the means and objects of knowledge is denied in the three times this then also infects the possibility of negation of the means and objects of knowledge, which also has to exist in the three times. Nagarjuna then replies that it is not feasible to first accept the negation of some object and then use this very acceptance to argue for its existence. See the commentary in Tola and Dragonetti (1995a, 108-110).

it could not be made sense of at all. After all what Nagarjuna criticizes in his analysis of causation is the conception of causes and effects as mutually independent, objectively existing entities. Similarly, in his treatment of epistemology, he sets out to refute the conception that being a means of knowledge is an essential property of some cognitive processes. If we do not make this presupposition, however, there is nothing intrinsically problematic with the existence of causes and effects in general, and also not with causes and effects in epistemic processes.

The catuskoti or tetralem ma • »

The second major formal aspects of Nagarjuna’s arguments we have to discuss is the catuskoti or tetralemma. Even though Nagarjuna employs this argumentative figure frequently and at important points in his arguments it is by no means restricted to his writings. In fact the tetralemma is likely to be familiar to any reader of Buddhist philosophical literature. Roughly speaking it consists of the enumeration of four alternatives: th at some propositions holds, th at it fails to hold, th at it both holds and fails to hold, that it neither holds nor fails to hold. The tetralemma also constitutes one of the more puzzling features of Buddhist philosophy as the use to which it is put in arguments is not immediately obvious and certainly not uniform: sometimes one of the four possibilities is selected as ‘the right one’, sometimes all four are rejected, sometimes all four are affirmed. It seems that this confusion is only exacerbated by the plethora of treatments we find in the modern commentarial literature, many of which try to analyse the tetralemma by recourse to notions of modern logic. There is no agreement about whether the four al104

ternatives are to be understood as quantified1 or unquantified propositions,2 whether any quantification is to be understood substitutionally or referentially, 3 whether the Law of the Excluded Middle holds for them,4 or whether they should be formalized in classical,5 intuitionist6 or paraconsistent logic.7 Despite some im portant work done during the last decades8 a comprehensive study of the origin and development of the catuskoti from its use in the earliest Buddhist literature up to its later employment in the Buddhist philosophical works of Tibet, China, and Japan remains yet to be written. The present chapter obviously does not intend to fill this gap, but has the more modest and more specific objective of giving an interpretation of Nagarjuna’s employment of the tetralemma which makes both logical sense and sheds most light on Nagarjuna’s philosophical position.9 The discussion in this chapter will be divided into four main parts. Firstly I will discuss the Indian distinction between two kinds of negation which will be of central importance for understanding the interrelations of the nested negations found in the tetralemma. In the second section I consider what might be taken to be a simplified case of the catuskoti, namely Nagarjuna’s rejection of two alternatives, of a position and its negation. Once the resources for understanding the argumentative role of this argument-schema 1 Robinson (1967, 57-58). 2Schayer (1933, 93). 3Tillemans (1990, 75). 4Murti (1955, 146), Staal (1975, 46-47), Napper (1989, 672-673, note 83). 5Robinson (1957). 6Chi (1969, 162-3). 7Priest and Garfield (2002). 8One of the most thorough treatments pertaining to its usage in the Madhyamika context is given by Ruegg (1977). 9To this extent the present discussion differs importantly from treatments like Robinson’s (1975) who explicitly restricted his investigation to the formal aspects of Nagarjuna’s arguments (295).

are in place we can move on to the third section, in which Nagarjuna’s use of the tetralemma proper as the negation of four alternatives is considered. In this section I concentrate on three main questions: whether the four alternatives are logically independent, what the status of third, ‘contradictory’ alternative is, and how instances of the catuskoti applied to properties are to be related to those applied to relations. In the fourth and final section I conclude my discussion with an account of the positive tetralemma, in which all four alternatives are affirmed. 4.1 Tw o kinds of negation The Indian philosophical tradition distinguishes two kinds of negation which are referred to as prasajya and paryudcisa, respectively. The origin of this distinction is grammatical; in prasajya-negation the negative particle connects with a verb (as in brahm,ana ndsti, ‘This is not a brahmin’), in paryuddsanegation it connects with a noun (as in abrahmana asti, ‘This is a nonbrahmin’). 10 This grammatical distinction corresponds to an important semantic distinction. If we refer to somebody as a non-brahmin we negate the term ‘brahmin’ and simultaneously affirm that he is a member of one of the three other castes. If, however, we simply say ‘This is not a brahmin’ we negate a proposition (i.e. ‘This is a brahmin’), rather than a term (‘brahmin’) and do not imply th at we speak about a person belonging to one of the three lower castes; in fact we do not have to speak about a person at all.11 In the Indian philosophical discussion (and particularly in the Madhyamaka context) it is this semantic distinction between implicational term-negation 10Oberhammer, Prets and Prandstetter (1991—, 11:163). n Renou (1942, II: 11), Cardona (1967, 40), Kajiyama (1973, 167-174). CHAPTER 4. THE CATUSKOTI OR TETRALEM M A 107 and non-implicational propositional negation which the terms paryudasa and prasajya are supposed to mark. It is therefore not necessarily the case that e.g. non-implicational propositional prasafya-negation is expressed as verbally bound.12 In fact this distinction is very familiar to contemporary philosophers. Not only can the grammatical distinction from Sanskrit be easily replicated in English, but the semantic distinction between the two types of negation also features prominently in the current discussion, particularly concerning the notion of a category mistake. Given that numbers are abstract objects it is clear th at claiming ‘The number seven is green’ is a category mistake. But what about ‘The number seven is not green’ ? This depends on how we take negation to operate in this case. It has been argued by a variety of authors13 that we have to distinguish two kinds of negation, called choice negation and exclusion negation. A choice negation presupposes th at an object falls under a property or its opposite. Presupposing that the apple on the table has some colour or other it must either be red or non-red. If we negate one alternative we affirm the other. Exclusion negation, on the other hand, ‘is supposed to reject merely what is denied, without making any presuppositions as to the fulfillment of sortal specifications’.14 Thus if we deny that the apple on the table is divisible by three we do not presuppose that it is the kind of thing which could be divided by three, but still (correctly) assert that it does not fall under the property ‘divisible by three’. It is then evident that ‘The number seven is not green’ is a category mistake only if the negation employed is taken to be a choice negation, not if it is an exclusion negation. 12Ruegg (1977, 5), (2002, 20-21). 13Mannoury (1947), Pap (1960), Routley (1969), Sommers (1965). 14Thomason (1972, 242).

While the distinction between choice negation and exclusion negation gives us a good model for understanding the distinction between paryudasa and prasajya negations, it should certainly not be identified with it, as there is no textual evidence th at Indian thinkers connected the distinction between the two kinds of negation specifically with categorial considerations. Rather, the difference between choice and exclusion negation should be considered as one example of the difference the pair paryudasa and prasajya indicates. This is the difference between negations carrying with them the presuppositions implied by the propositions they negate, and those which deny these presuppositions. Thus reading the ‘not’ in ‘The number seven is not green’ as a choice negation carries with it a presupposition ‘The number seven is green’ makes, namely that seven is a thing which could be green. This assumption is denied if the ‘not’ is read as an exclusion negation. Examples of these different kinds of negation which do not rely on sortal considerations are not hard to come by. There are two ways of negating the assertion th at the present King of France is bald, one making the negation true, the other false or meaningless,15 similarly there are two ways of negating the accusation of continuing to be an alcoholic, one asserting th at one has stopped drinking now, the other also denying the implication th at one ever was a heavy drinker.16 15The first being ‘It is not the case that there is somebody who is both the King of Prance and bald’, the other ‘The present Kind of France is not bald (i.e. has a full head of hair)’. The difference between the two is drawn in terms of the scope of the negation operator, that is, put formally, as the difference between ->(3\)(Kx A Bx) and (3!)(Ka; A ~>Bx). 16Shaw (1978, 63-64) notes the interesting idea of representing the proposition a sentence expresses as an ordered set, the last member of which is the sentence itself, the preceding one expressing the presuppositions that sentence makes, the one preceding this its presuppositions in turn and so on. A paryudasa-negeAion can then be understood as negating the final member of the set only, whereas a prasajya-negation negates both it and

As I will argue below the best way of interpreting Nagarjuna’s arguments is based on understanding the concepts paryudasa and prasajya in this particular manner. T hat is, pan/wdasa-negations will be regarded as negations which continue to endorse the presuppositions made by the proposition they negate, while the purpose of prasajya-negations is to be able to formulate negations which explicitly reject some of these presuppositions. 4.2 R ejection of tw o alternatives The distinction between the two kinds of negation helps to understand an im portant methodological tool which is used extensively throughout Nagarju na’s writings. Consider MMK 18:10 which sets out claiming th at whatever comes into being dependent on some object is not identical with th at object, nor is it different from that object.17 If we ascribe to Nagarjuna anything like the standard conception of identity (i.e. th at identity is the relation everything bears to itself, and nothing bears to any other object) it seems hard to make sense of it, at least if we want to stay within the domain of classical logic. Nagarjuna considers the property ‘being identical with the object it depends on for coming about’ (which we will abbreviate to ‘being identical with a ’) and denies th at it applies to any object, and also denies th at it fails to apply to any object. Expressed semi-formally this gives 1. For all x which come into being depending on some particular object, not (identical-with-a(x) or not identical-with-a(x)). some (possibly all) of its predecessors. 17prattiya yad yad bhavati na hi tavat tad eva tat / na canyad api [...].

But applying the familiar laws of logic (in this case DeMorgan’s law and Double Negation Elimination), this can easily be seen to be equivalent to 2. For all x which come into being depending on some particular object (not identical-with-a(x) and identical-with-a(x)), which is a contradiction. How can this interpretation be avoided? The key lies in the distinction between the two kinds of negation. We have to assume that the two occurrences of ‘not1 in 1. do not in fact refer to the same concept of negation, but rather th at the first is a prasajya-negation, a presupposition-cancelling negation,18 the second a paujudasa-negsktion^ understood as a presupposition-preserving negation. Taking the first negation as prasajya is also suggested by M atilal,19 who claims th at on this interpretation ‘the apparent contradiction of the joint negation1 disappears.20 Staal claims that such attem pts to avoid inconsis18 That the first instance of negation is supposed to be prasajya is stated both by CandrakTrti in PP 13:5 (see Ruegg (2002, 19) for a translation and commentary), as well as earlier by Bhaviveka in the Prajnapradipa (Walleser 1914a, 10:8). CandrakTrti does not explicitly say what kind of negation the second negation is. It is evident, however, that for him it cannot be prasajya-negation too. Considering the first two alternatives of the tetralemma CandrakTrti argues against the claim that the negation of the first alternative (j4) logically implies (prdpnoti) the second alternative (not A). If the ‘not’ in this ‘not A' was indeed taken to be prasajya it would be obviously entailed by the prasajya-negation of A. Since it is not so entailed, however, it cannot be a prasojya-negation as well. It is therefore plausible to regard the second kind of negation as paryudasa, an assumption which, as we shall see, also makes a good deal of exegetical sense. 19Matilal (1971, 164). 20Matilal refers both to the negation of two and of four alternatives (the catuskoti). As we will see below the interpretation of the latter involves additional complications Matilal does not seem to be aware of.

tency are unsuccessful, as ‘it is not true that contradictions do not arise between p ras ajy a-neg&tions’ .21 However, this disregards the fact th at Matilal’s point was th at only the outer negation of the two alternatives is to be regarded as prasajya, while the negations employed within the statement of the alternatives are supposed to be pan/udasa-negations. It is therefore the entire set of two mutually exclusive alternatives which is negated, and on this interpretation there is indeed nothing inconsistent about it. In order to see the motivation for this employment of two kinds of negation we have to understand that one of Nagarjuna’s main aims in the MMK as well as elsewhere is to demonstrate the deficiency of some key concepts of our conceptual scheme (such as causation, motion, identity and so forth).22 Their deficiency is taken to be due to a presupposition failure: in the same way th at we spot a deficiency in calling the number seven yellow (because the presupposition th at numbers are things which could possibly have a colour is not fulfilled), Nagarjuna regards commonsense concepts like causation to be deficient because they presuppose the existence of svabhava, the independent existence of objects, which, Nagarjuna argues, is a presupposition which is not fulfilled.23 It then becomes easy to see th at 1. should be interpreted along the lines of 3. For all numbers x, not (yellow(x) or not yellow(x)). 21See Staal (1975, 46). He also claims that the principle of contradiction only holds for prasajya- and not for paryizdasa-negations, a claim which he backs up by reference to the MTmamsa concept of two kinds of paryudasa (Staal 1962, 60-61). But this can only serve to show that the Madhyamaka concept of paryudasa is quite different as it is manifestly taken to be subject to the principle of contradiction. Compare the characterization of paryudasa by Avalokitavrata given in Kajiyama (1973, 169-172). 22Ganeri (2001, 45-47). 23Ruegg (1977, 51).

If the outer negation is taken to be exclusion negation and the second to be choice negation we cannot just read this as implying the contradictory statement th at all numbers are both yellow and not yellow. Rather we will read it as denying (in a prasajya-manner) that the property yellowness and its (paryudasa) opposite (which would imply that numbers were of some other colour) fail to be applicable to numbers.24 Interpreted in this way, since the outer negation is read as exclusion negation, 3. does also not presuppose that any other property is in fact applicable to numbers. In the context of 3. this neutrality is not particulary important, as we usually would want to claim that there are other (mathematical) properties which are applicable to numbers. It is, however, important for 1. since Nagarjuna wants to extend his arguments to all other st>a6/iat>a-presupposing concepts (which, according to him, are all the concepts we usually operate with). If we therefore read the first ‘not’ in Nagarjuna’s statement as prasajya and the second as paryudasa, the following interpretation emerges: 4. It is denied th at either the concept ‘identical-with-a’ or its choice negation ‘different-from-a’ can be ascribed to any object x which comes into being depending on some particular object, without assuming that there is any pair of a concept and its choice negation one of which can be applied to such an object. In order to demonstrate the deficiency of a concept Nagarjuna has then to examine both the concept and its paryudasa-negation and show th at both are not applicable to the objects under discussion, in the same way in which we argue that the concept ‘yellow’ is not applicable to numbers because 24See Galloway (1989, note 13, 29-30).( x is yellow’ and (x is not yellow’ are contraries when the referent of x is sortally incorrect (since they are both false). If the referent is sortally correct they are contradictories. See Raju (1954, 710-711).

numbers, not being material objects, cannot have a property like yellowness (which is exclusively had by material objects), nor can they have any other colour (the paryudasa-negation of the concept ‘yellow’).25 4.3 R ejection of four alternatives As will be obvious to any reader of Nagarjuna’s writings far more common than the case just discussed, where two alternatives (a concept and its pan/udasa-negation) are both rejected is the rejection of four alternatives: the rejection of the application of a concept, of the application of its negation, of the application of both the concept and its negation, and finally of the application of neither the concept nor its negation. For example we read in MMK 22:11 ‘Em pty’ should not be asserted, ‘Non-empty’ should not be asserted, both or neither should not be asserted, as these are only said for the purpose of designation.26 The same argumentative pattern of the rejection of four alternatives is also applied to ‘permanence’ and ‘finitude’ concerning the Buddha,27 to the 25Raju (1954, 701-702) employs this argumentative procedure to show that neither the concept ‘positive’ nor its pan/udasa-negation ‘negative’ is applicable to the number zero (sunya in Sanskrit) and claims that similarly for the Madhyamilta no concept is applicable to emptiness (sunyata). We should note, however, that there is no evidence in the Madhyamaka literature of an explicit connection between the mathematical concept sunya and the metaphysical concept sunyata having ever been made. See Ruegg (1977, 69, note 154), (1978), Galloway (1989, 27-28, note 7). 26sunyam iti na vaktavyam asunyam iti va bhavet / ubhayam nobhayam ceti prajnaptyartham tu kathyate. 27MMK 22:12.

existence of nirvana,28 to the existence of persons in the past,29 to their permanence,30 and to the finitude of the world.31 The employment of the tetralemma can be traced back to the earliest Buddhist scriptures. In the Kandaraka Sutta the four alternatives are employed as a classificatory tool for distinguishing four classes of ascetics, those which torment themselves, which torment others, which torment both and which torment neither.32 In this case the fourth alternative is explicitly recommended by the Buddha as the ideal to be emulated. A case of the rejection of the four alternatives by the Buddha concerning the question whether the Tathagata exists after death33 can be found in the Aggivacchagotta Sutta34 and the Culamalunkya Sutta.35 Although the relationship between the use of the tetralemma in early Buddhism36 and its employment by later Madhyamaka authors is complex and will not be investigated here it is nevertheless important to note at least that two different motivations can be discerned in the Buddha’s rejection of the four alternatives. One motivation is pragmatic; deciding which of the four position holds regarding specific questions (such as whether the Tathagata exists after death, whether the world is finite etc.) is seen to be irrelevant for the attainment of liberation. Buddha therefore wants to set these questions aside, a is illustrated in the well-know simile of the poisoned arrow.37 The other moti28MMK 25. 29MMK 27:13. 30MMK 27:15-18. 31MMK 27:25-28.

32Trencker (1888, 1:341), Bikkhu Nalamoli and Bikkhu Bodhi (2001, 445). 33Nagarjuna considers the same question in MMK 22:12, 34Trencker (1888, 1:484-485), Bikkhu Nalamoli and Bikkhu Bodhi (2001, 591). 35Trencker (1888, 1:431), Bikkhu Nalamoli and Bikkhu Bodhi (2001, 536). 36For some material on this see Gunaratne (1980). 37Trencker (1888, 429), Bikkhu Nalamoli and Bikkhu Bodhi (2001, 534-535).

vation is systematic; Buddha argues that the predicates applied in the four alternatives under consideration are in fact not applicable to their respective subjects, in the same way as any specification of spatial co-ordinates is not applicable in reply to the question where the extinguished flame of a candle went.38 All members of an exhaustive set of applications of such predicates (which the four alternatives are taken to be) therefore have to be rejected. 4.3.1 D istinctness of the four alternatives There are a variety of prima facie difficulties in interpreting the four statements of the tetralemma. The first difficulty concerns the distinctness of the four alternatives. It is fairly common in the Western commentarial literature to express the tetralemma in propositional form, so th at in MMK 22:11 cited above (letting A stand for the proposition “Em pty1 should be asserted1) Nagarjuna is taken to say that all of the following propositions are to be rejected:39 38Trencker (1888, 486-487), Bikkhu Nalamoli and Bikkhu Bodhi (2001, 593). 39See e.g. Schayer (1933, 93), Galloway (1989, 16), Ng (1993, 93), Tillemans (1999, 134). In some cases the equivalent form (—«j4 A -i-A ) is given for the fourth alternative. The reader might wonder why we expressed the fourth alternative as ‘not (A or not A )’ rather than ‘not (A and not A )’, i.e. as the negation of the third alternative, which would be equivalent to ‘A or not A ’. If we look at the way the fourth alternative is formulated in the MMK we realize that there is a considerable amount of variation which seems to allow both formalizations. We sometimes find it formulated as na ubhayam ‘not both [the first and second alternative]’ (22:11, 25:17, 25:23, 27:13) which supports the reading as ‘not (A and not A )’ and sometimes as naiva . .. naiva ... ‘not even . .., not even . . . ’ (18:8, 25:15-16) or na . . . na . .. ca ‘not . .. and not . . . ’ (25:22) which seems to support the reading as ‘not (A or not A). The reason for this variation is not that Nagarjuna had problems distinguishing ‘and’ and ‘or’ but rather that the context makes it clear that ‘not (A or not A )’ is intended. If we read the fourth alternative as ‘not (A and not A )’ this CHAPTER 4. THE CATUSKOTI OR TETRALEM M A 116 1. A 2 . - 1A 3. A A ~iA 4. -i(A V ->A)

It is, however, easy to see that on this understanding the final two alternatives come out as logically equivalent.40 Given the prominent place which the tetralemma occupies in Madhyamaka literature we would have to charge both Nagarjuna as well as later Madhyamaka authors with remarkable logical naivety for not realizing th at instead of considering four possibilities, they were in fact only dealing with three. In order to see how to solve this difficulty it is important to realize that once the fourth alternative is rejected, we are dealing with a statement with three nested negations, namely -i-i(A V -<A).41 If we read the negationsymbols just as straight truth-functional negation both this as well as the negation of the third alternative turn out to be equivalent to A V —'A, and it leaves us with three possible ways in which it could be true: either A obtains and not-A does not, A does not obtain and not-A obtains, or A does not obtain and not-A does not obtain either. Given that the first two possibilities would be inconsistent with the rejections of the first two alternatives earlier in the argument we are left with the third possibility, which just says the same as ‘not (A or not A )1.

Applying DeMorgan’s law to the fourth alternative, - ‘(A V ->A), we get (-M A ->-iA), which, by Double Negation Elimination, is equivalent to Af\->A, i.e. the third alternative. Robinson (1967, 57) is one of the surprisingly few authors to have picked up on this very problematic issue. 41 After rearranging the relevant parts of MMK 22:11 it is straightforward to see the three stacked occurrences of negation it contains (here highlighted in bold): n a vaktavyam na ubhayam sunyam asunyam. See also 27:13. CHAPTER 4. THE CATUSKOTI OR TETRALEM M A 117 is obvious that this is not the conclusion Nagarjuna wants to draw.42 I have now argued above that the two instances of negation in such statements as MMK 18:10 should be regarded as different kinds of negation, namely that the outer had to be taken as prasajf/a-negation and the inner as paryudasanegation. Since it is evident that the negation involved in the rejection of the four alternatives is meant to be prasajya-negation43 the rejection of the fourth alternative would then have to be read as prasajya--i prasajya-^A V paryudasa--'A). 42It is interesting to note that the Tibetan commentarial tradition tried to avoid this difficulty by plugging in various modifiers, such as ‘ultimately’ (don dam par) or ‘conventionally’ (tha snyad du). Abbreviating these by U and C respectively the tetralemma is taken to assert that all of the following should be rejected: 1. U A 2. C-iA 3. UA A C~>A 4. -i UA A ->C->A

It it thereby denied that A obtains ultimately, that it conventionally fails to obtain, that it both ultimately obtains and conventionally fails to obtain, and finally that it neither ultimately obtains nor conventionally fails to obtain. Tillemans (1999, 134-137) gives an example (slightly more intricate than the above) of such an interpolation procedure from Se ra rje btsun chos kyi rgyal mthsan’s sKabs dang po ’i spyi. It is evident that on this account the third and fourth alternative are not in turn equivalent to the Law of the Excluded Middle, without requiring us to assume that negation behaves non-classically. While the dGe lugs interpolation procedure here (as well as in other contexts) provides a very interesting interpretation of the Madhyamaka arguments there seems to be no textual evidence that Nagarjuna expected qualifications of the above kind to be supplied when interpreting the tetralemma. 43As stressed in PP 13:5.

If we now assumed that prasajya-negation obeyed Double Negation Elimination, i.e. that an even number of such negations cancelled each other out this would mean the rejection of the fourth alternative entailed the assertion of either A or its paryudasa-negation, which is clearly not what Nagarjuna wants to say. I therefore want to argue that this assumption is indeed not justified, i.e. that —<—>>4. = A does not hold when the negation is taken to be prasajya- negation. It is sometimes remarked in contemporary commentarial literature that the notion of negation at work in Madhyamaka arguments should be understood along the lines of intuitionist negation, which famously does not accept the equivalence -i-iA = A .44 It has to be kept in mind, however, th at the intuitionist rejection of -<-A. = A, which went hand in hand with a negation of the Law of the Excluded Middle was motivated by very specific mathematical reasons. Since the negation symbol was interpreted as expressing our ability to give a reductio ad absurdum of the mathematical proposition to be negated, while the assertion of an unnegated proposition was taken to imply our ability to provide a proof of that proposition, — iA could not entail A, as a demonstration that we cannot disproof a proposition does not amount to a proof of th at proposition.45 Moreover, given the existence of undecided mathematical sentences the intuitionist is unwilling to accept that we are able to provide either a proof or a refutation of each mathematical proposition, which is what A V ->A means for him.46 It is obvious that these problems in the ontology of mathematics were not problems Nagarjuna was 44For an exposition of intuitionist logic see Heyting (1971). The intuitionist reading was considered by Chi (1969, 162-3) and Staal (1975, 47). 45See Heyting (1971, 17-18) for an example. 46Heyting (1971, 99-100).

concerned with.47 W hat speaks furthermore against the intuitionist interpretation of Madhyamaka negation is the fact that while it is sensible to argue that prasajya-negation does not obey - 1- 1A = A in order to make sense of the tetralemma, I do not think Nagarjuna also rejected the Law of the Excluded Middle for it.48 For even if some property (or indeed all properties) should turn out to be inapplicable to an object, this means that the prasajyanegation of the ascription of the property to the object should be affirmed. And given Nagarjuna does not express any doubts about our ability to check whether properties are in fact applicable to objects in general, it appears to be unproblematic to affirm th at A V ~>A holds for prasajya-ilegation, that is to assume for any property and any object, that either this property is applicable to the object or it is not.49 47In (1974, 297) Richard Chi agrees with this point, calling his earlier intuitionist analysis of the tetralemma a ‘mistake’: ‘Despite the superficial resemblance, it is incredible that Nagarjuna and Brouwer could possibly think in the same way. Dialectics and pure mathematics are, after all, two different disciplines. The agreement of the two systems is a sheer coincidence; they reach the same result for different reasons’. 48Nor did Tsong kha pa. See Napper (1989, 61). 49That Nagarjuna accepts the Law of the Excluded Middle is also argued by Ruegg (1977, 48-49). His argument there is, however, based on the erroneous presupposition (also made by Staal (1975, 47)) that the intuitionist has to assume the existence of a third truth-value (see Dummett (1998, 178), (2000, 11)).

There does not seem to be any direct textual evidence in Indian Madhyamaka literature stating that prasajya-negation does not obey ->-■ A ~ A 50 If we consider one example of a presupposition-cancelling prasuji/a-negation discussed above, namely the case of exclusion negation it seems plausible th at -i-A = A does not hold for it. Remember th at when we use choice negation to negate a statement like ‘The apple is red’ we are merely saying of the apple th at it has some other colour. To use set-theoretic terminology we assert (within the domain of coloured things) that the apple belongs to the complement of the set of red things. Now if we use choice negation twice, saying ‘The apple is not not red’ we are just saying that the apple belongs to the complement of the complement of the set of red things, which is of course the set of red things itself. In brief, we just say that the apple is red. Exclusion negation, on the other hand, would be used to negate a statement like ‘The number seven is yellow’, thereby claiming that yellowness not just fails to be true of the number seven, but is indeed not applicable to it. If we then iterate this exclusion negation we say that it is not applicable to assert of the number seven that the property of yellowness is not applicable to it — and whatever this means, it seems quite distinct from saying that the number seven is yellow. Be this as it may, I think there is a more elegant way to dissolve the above difficulty of iterated negations. This involves the notion of illocutionary 50 An interesting case of a Tibetan rejection of this principle is provided by the Sa skya pa scholar Go rams ba bsod nams seng ge. As Tillemans (1999, 137) has argued, according to the mainstream dGe lugs approach the Tibetan analogues of prasajya- and paryudasar negation, med dgag and ma yin dgag, were assumed to obey -i-iA — A. (See e.g. Tsong kha pa Bio bzang grags pa (1973, 43-44).) Go rams pa bSod nams seng ge (1988, 51-52), however, rejects this principle, precisely to make sense of the tetralemma without the dGe lugs-style interpolations.

negation?1 The underlying idea is that propositions expressing a content can be prefixed by illocutionary operators forming assertions, commands, requests, promises and so on. Thus ascribing the property of being open to the window produces the assertion ‘The window is open’ when prefixed by the assertion operator, the command ‘Open the window!’ when prefixed by the command operator and so on. It is now important for our purposes to note that when one of these results is negated it makes a difference whether or not the negation operator is within the scope of the illocutionary force operator, i.e. whether we say ‘I promise not to open the window’ or ‘I do not promise to open the window’. Similarly there is a distinction between ‘I assert that the window is not open’ and ‘I do not assert that the window is open’ — the first involves familiar propositional negation, the second illocutionary negation. There are various reasons why someone may employ illocutionary negation. One example is obviously when the proposition to be negated carries an unwelcome presupposition which propositional negation would preserve. Thus we will be happy to say ‘I do not assert that the number seven is yellow’ (presumably together with ‘I do not assert that the number seven is not yellow’), but not ‘I assert that the number seven is not yellow’. In other words, one motivation for using illocutionary negation is the desire to employ a prasajya- rather than paryudasa-negation because we want to reject a particular presupposition made by the sentence to be negated. Note, however, that this is not the only reason why we might use illocutionary negation. Another obvious candidate is lack of evidence. We might say ‘I do not assert that the continuum hypothesis is true’ in order to indicate th at we have no good 51Searle (1969, 31-33); the distinction of illocutionary force from content goes back to Frege. The relevance of illocutionary negation to this problem was suggested by Jayatilleke (1963, 346, 475), (1967, 81), Chakravarti (1980), Ruegg (1983, 238) and Matilal (1986, 66-67, 88-90).

evidence either way; in this case the presupposition-cancelling consideration involved when discussing the colour of the number seven does not come into play. We do not want to say that the continuum hypothesis is not the kind of thing which could be true or false. A third case in which we might want to apply illocutionary negation to a proposition A is when A is not part of our language and we have no way of translating it. In this case we would not want to assert A because we do not know which situation would make it true, and which would make it false. It is therefore evident that illocutionary negation is a more general notion than presupposition-cancelling prasajy a-negation: it incorporates these, but it subsumes other considerations as well. It is now tempting to interpret the tetralemma as asserting that illocutionary negation52 should be applied to the following positions: 1. A 2 . ~ iA 3. A and ->A 4. I do not assert (A or -A.) 52It is interesting to note that in the Pali sources we sometimes find the four alternatives denied by the phrase na h ’ idam, ‘it is not so’ (for example in the Ahguttara Nikaya, (Morris 1888, 2:163)), and sometimes by the phrase ma h 7 evam, ‘do not say so’ (Samyutta Nikaya^ (Feer 1888, 2:19-20)). Some have argued that there is a semantic distinction between the two uses and that ‘it is not so’ is employed when the predicate in question is applicable to the situation discussed, but giving an affirmative answer to any one alternative would be misleading, while ‘do not say so’ is used where the predicate is not applicable to the situation (Jayatilleke (1963, 346), Gunaratne (1980, 231-231), Bharadwaja (1984, 312-313)). This second use corresponds to the illocutionary negation just introduced; interestingly enough this is employed in the passage from the Samyutta Nikaya just cited to reject the four alternatives claiming that suffering is produced by oneself, others, both or neither.

Here the negation-operator -i is to be read again as parywdasa-negation. The focus of our attention is of course the negation of the fourth alternative, which now features two illocutionary negations in a row, i.e. I do not assert that I do not assert that (A or -u4). The advantage of replacing the prasajya-negations in this way by illocutionary negations is that it allows us to see straightaway th a t the two negations do not reduce to an unnegated proposition, that is that - 1- 1A = A does not hold. Declining to assert a proposition which in turn asserts that we decline to assert a third proposition does not amount to an assertion of this third proposition. Tillemans has argued that the illocutionary reading of prasajya-negation in the context of the tetralemma has the ‘serious philosophical drawback’ th at it gives the impression of the Madhyamika’s refusing to adopt either a positive or negative position on some subject-matter. This impression would be misleading, however, since it is apparent that Nagarjuna and his commentators wanted to assert ‘some form of a negated proposition’53 when setting out the arguments for rejecting the different parts of the tetralemma. Fortunately this problem can be easily dissolved. While the application of illocutionary negation to some proposition entails that we want to be ‘uncommitted to the truth or falsity of it’54 it also means th at we want to assert a negative proposition when speaking about the proposition concerned. For example we might want to deny that there is enough evidence available for deciding it, or that we can translate it into our language, or that it carries with it a presupposition we want to assert. It is of course this last justification for using illocutionary negation the Madhyamika wants to adopt, as he wants 53Tillemans (1990, 74). 54Tillemans (1990, 74).

to deny the existence of svabhava presupposed by the four positions in the tetralemma. It is therefore unproblematic to assert th at the Madhyamika declines to assert any of the four positions, while still ‘asserting some form of negated proposition’. A further objection one might make at this point is that the interpretation in terms of illocutionary negation is not able to account for one important feature we would want to ascribe to the tetralemma, namely th at the four alternatives are logically disjoint. It is evident that if I refuse to assert some proposition A (that is, negate it illocutionarily) this will entail th at I also refuse to assert its conjunction with some other proposition. It could not be the case that I refused to assert the continuum hypothesis but would be happy to assert both the continuum hypothesis and Riemann’s hypothesis. But in this case the illocutionary negation of the first alternative will imply th at of the third, so that any distinct argument for rejecting the third possibility would be superfluous. This argument of course depends on the assumption that the ‘and’ in the formulation of the third alternative behaves like the truth-functional operator of conjunction, so that the third alternative entails the first. We will argue shortly that this is not generally the case. To do this, however, we must first have a closer look at the status of the third alternative itself. 4.3.2 The status of the third alternative An im portant problem in interpreting the tetralemma is connected with the rejection of the third alternative, which asserts the applicability of a property and its paryudasa-negation. Why, we might well ask, does Nagarjuna think we have to consider this contradictory option as well, as if this constituted a CHAPTER 4. THE CATUSKOTI OR TETRALEM M A 125 real possibility?55 Robinson suggests th at a way of dealing with this problem is to interpret the four alternatives not in a propositional, but in a quantificational way.56 If F is the property under consideration the four alternatives to be rejected become: 1. Everything is F. 2. Everything is not F. 3. Something is F and something is not F. 4. N o t: Something is F or something is not F. Here all negations are paryudasa, apart from the one in the fourth alternative set in boldface, which is a prasajya-negation. It is evident that when formulated in this way the third alternative is ambiguous, depending on whether we take the two occurrences of ‘something’ to refer to the same object. If we take them to refer to different objects, the third alternative is not any more problematic than saying that chesspieces are both white and not white, where this is to mean that some are white, and some are not white. This interpretation, however, does not fit well with the employment of the tetralemma by Nagarjuna. His aim is to investigate the applicability of various concepts (such as emptiness, permanence, finitude etc.) to objects. If the third alternative was taken to mean ‘the concept under discussion is applicable to some objects, and not to others’ this would not 55There are clear cases of pan/udasa-negation in the MMK where Nagarjuna assumes the Law of Non-contradiction (e.g. 7:30 and 8:7). The third contradictory alternative should therefore not constitute a genuine possibility. See also Robinson (1967, 50-52), Ruegg (1977, 48-49), Galloway (1989, 19-22). 56Robinson (1967, 57-58).

be an argumentationally interesting option for Nagarjuna, as the application of the concept to some objects, and its non-application to the others would then have to be investigated individually in any case. On this interpretation the third alternative would merely present a complex statement of two argumentative options which Nagarjuna will want to investigate separately. Richard Robinson remarks:57 It is a striking feature of the Stanzas that all predicates seem to be asserted totally of the whole subject. Existential quantifications are denied because the discussion is concerned, not with the denial or affirmation of commonsense assertions such as ‘Some fuel is burning, and some is not1, but with the concepts of own-being and essence. W hat pertains to part of an essence must of course pertain to the whole essence. Put briefly, given th at Nagarjuna wants to inquire into the applicability of particular concepts to objects tout court we should also consider the four alternatives as giving alternative ways of the application of particular concepts to objects tout court, rather than as implying their application to some objects, but not to others. We therefore have to interpret the two occurrences of ‘something’ as pertaining to the same object, i.e. the third alternative claims that ‘something is F and the same something is not F \ Whether this is contradictory depends on how we understand the application of the properties F and not F. For example, it is straightforward to assert that a chess board is black and not black, if we mean by this that some parts of it are black, and others not black. On this reading the contradiction is avoided by relativising the two properties involved to different mereological parts. The same result can be achieved 57Robinson (1967, 54). See also Gunaratne (1986, 225-226).

by relativizing to different respects or perspectives under which the object is considered, for example if we assign different utilities to an alternative in a decision problem under different descriptions.58 To see that these kinds of relativizing interpretations are present in Nagarjuna it is instructive to look at the reasons by which the third alternative is generally rejected. Here we can distinguish two varieties. In the first case Nagarjuna rejects it because its claim is as contradictory as asserting of a single object that it is wholly black and not black. For example we read in MMK 25:14:59 How could nirvana exist and not exist? Like light and darkness these two [i.e. existence and non-existence] cannot be at the same place.60 In the second case Nagarjuna rejects the third alternative since it would combine the difficulties facing the first and second alternative (which have already been rejected earlier in the argument). This point is clearly made by CandrakTrti:61 Things do not originate from both [themselves and from other things]. This is because the problems stated for both positions 58The Digha Nikdya (1:31) asserts that the world is both not existent and not nonexistent; the former because it ceases, the latter because it arises. See Jayatilleke (1967, 79), Robinson (1969, 75), Gunaratne (1980, 221). 59Further examples can be found in 7:30, 8:7 and 27:28. ^bhaved abhavo bhavas ca nii'vanam ubhayam katham / tayor abhdvo hy ekatra prakasatamasor iva. 61 dvabhyamapi nopajiiayante bhdvdh ubhaya-paksa-abhihita-dosa-prasangdt pratyekam utpdddsamarthydc ca. PP 38:1-2, (Ruegg 2002, 73). CandrakTrti makes the same point when commenting on MMK 12:9. See Schayer (1931, 20).

[i.e. the first and second alternative] will arise together one by one. It is clear from this way of rejecting the third alternative that it is here not understood to be contradictory, but that CandrakTrti takes it to be perfectly possible that something could be caused partly by itself and partly by other things. (One straightforward account of this consists in conceiving of an effect as a potential in a cause which is only actualized given the right background conditions).62 This is rejected because the presence of these two ways of causing would imply the difficulties of both causation from itself, and of causation from other things, both of which Nagarjuna has rejected as unsatisfactory earlier.63 We therefore have to conclude that Nagarjuna applies the argumentative figure of the tetralemma both to cases where he takes a concept and its paryudasa-negation (i.e. the conjuncts of the third alternative) to be contradictory, as in the first case just mentioned, as well as where he considers it 62Garfield (1995, 106-107). 63These two ways of rejecting the third alternative are also distinguished in Ghose (1987, 296-297). He also mentions a third way where the third alternative is rejected because ‘it attributes to the conjunction some properties which are common to both the conjuncts’. As an example Ghose discusses verse 25:12 from the MMK where Nagarjuna claims that ‘if nirvana was both existent and non-existent, it would not be non-dependent, as it would depend on both’. Nagarjuna here does not refute this alternative by saying that it is contradictory for something to be both existent and non-existent, but by arguing that since existence and non-existence both presuppose dependence, nirvana would be dependent, which it is not. However, it is evident that this is just an example of the second way of rejecting the third alternative too. Nagarjuna has already rejected (in verses 6 and 8) that nirvana is either existent or non-existent, because it would be dependent in each case. The third alternative is thus rejected because it implies the difficulties of both the first and second alternative, which happen to be the same difficulty in this particular case.

to be possible that both can be applied to an object, as in the second case. Obviously only in the second case recourse to the tetralemma would have been strictly necessary, as in the first case a consideration of two alternatives (of the concept and its paryuddsa-negation) would have been sufficient, given th at both of them together are regarded as contradictory anyway. We might perhaps explain the fact that Nagarjuna uses the four alternatives nevertheless on rhetorical, rather than on logical grounds. If it was assumed that all four alternatives of the tetralemma applied to a particular notion were positions actually propounded by some school of thought64 it would be heuristically useful, if not logically necessary, to go through all of them individually, even if this included an alternative which the Madhyamika regarded as logically contradictory. But if we thus regard the second case as the domain of the tetralemma proper (and the first only as a rhetorical expansion of the rejection of two alternatives) it is clear th at in the tetralemma proper the third alternative does not entail the first. Consider the case of the tetralemma applied to causation. Here the first alternative claims that things are exclusively caused by themselves, the second that they are exclusively caused by others. The third alternative constitutes a compromise between the first and second: it says that things are partly self-caused, and partly caused by other objects. But this obviously does not imply the first alternative, 110 more than saying that a chess-board is partly black and partly white implies that it is black all over. For this reason the illocutionary negation of the first alternative also does not imply th at of the third, since the third is not a truth-functional conjunction of the first alternative and something else. We should also note th at according to the quantificational reading given 64See note 77 to this chapter.

above the third and fourth alternative are logically distinct, as 3. says that some objects instantiate both the property F and its complement, whereas 4. says th at neither is in fact instantiated. Finally, as we conceive of the initial two negations in the negated fourth alternative as illocutionary negations, so th at they do not cancel each other out the rejection of the fourth alternative is not equivalent to ‘Something is F or something is not F \ It therefore becomes evident that what Nagarjuna wants to say in MMK 22:11 is th at the following four alternatives should all be rejected:65 1. ‘Em pty’ should be asserted of all objects. 2. ‘Em pty’ should be denied (in a paryudasa fashion) of all objects. 3. ‘Em pty’ should be asserted of some objects, and should be paryudasadenied of the same objects. 4. N o t: ‘Em pty’ should be asserted of some objects, or ‘Em pty’ should be paryudasa-denied the same objects. Nagarjuna’s usual argumentative procedure (as we will see below) is to argue th at each of the four alternatives leads to an absurd consequence, so th at the whole set is to be rejected. In this case, however, he does not discuss the four alternatives individually but dispatches them with a single argument, namely by saying that all assertions listed in the four alternatives ‘are only names’.66 Nagarjuna is therefore making a semantic point: while it 65The boldface ‘not’ indicates illocutionary negation. 66Since Nagarjuna does not give reasons for the rejection of the four alternatives individually we cannot say whether he would have wanted to reject the third alternative because he considers it to be contradictory (‘nothing can be empty and not empty at the same tim e’) or because it would combine the difficulties inherent in the first two alternatives. It is certainly conceivable that someone might adopt the third alternative by arguing

is of course true for the Madhyamika that every right-minded person should assert the emptiness of all objects, this should not be done by assuming that there are some objectively existent objects out there, referred to by a similarly objective reference relation, and th at these objects have the property of emptiness.67 Statements of emptiness should not be understood according to the standard semantic theory.68 W ith such a theory in mind it is neither correct to say that all things are empty, or th at they all lack emptiness, or th at some are both empty and not empty, or that the predicate ‘empty’ is not applicable to objects at all, in the same way as the predicate ‘yellow’ is not applicable to numbers. Before leaving the subject of the quantificational interpretation of the tetralemma it might be useful to have a brief look at the analysis presented by Tillemans in an appendix to (1990). There the four alternatives are formalized as

1. -.(3x)(Fx) 2. -i(3a;)(-iF x ) 3. -i(3x)(F x A->Fx) 4. —i(3x){—iFxA~'~>Fx).

that phenomena are empty in some respects but not in others, e.g. one might claim that they are empty insofar as they are causally produced, but not empty insofar as they exist independently of us. 67It should therefore be noted that the last three alternatives are in a way more deficient than the first one. For a Madhyamika the first assertion would be true if interpreted according to the right semantics, whereas the final three would still have to be rejected, because even with the right semantics they would be false. 68Garfield (1995, 280).

Tillemans argues that it is straightforward to make sense of the simultaneous rejection of all four positions if we assume that there is no x , i.e. if the domain of quantification is empty.69 While this reading makes the distinction between different kinds of negation in the tetralemma superfluous it also has a number of problems. On the one hand there is the familiar difficulty th at the third and fourth possibility come out as logically equivalent. On the other hand (as rioted by Tillemans himself) this interpretation implies that the Madhyamika would also have to accept all the four positions of the tetralemma, since the corresponding universal statements are also true in the empty domain. But there is no textual evidence in Madhyamaka literature that the four positions of the tetralemma are simultaneously to be rejected and accepted.70 Tillemans continues to argue th at quantified statements accepted by the Madhyamika are generally to be interpreted substitutionally, rather than referentially. Interpreted referentially the statement ‘All x are F ’ means that there is some set of objects such that every single one of them is F. Interpreted substitutionally it means that for every name substituted for lx ’ in ‘Fa;’ we get a true statement. The Madhyamika can therefore ‘use the world’s language to communicate about whichever day-to-day affairs the world concerns itself with: his sunyavada, however, dictates that he never accepts a referential interpretation of such language’.71 The difficulty with this interpretation is that the difference between referential and substitutional quantification is simply that between quantify69Tillemans (1990, 75). 70 As we will see in the final section of this chapter there are cases in which all four positions are affirmed (the so-called positive tetralemma). This, however, serves a very different purpose from the negative tetralemma. 71 Tillemans (1990, 75).

ing over objects in the world and quantifying over pieces of language. The Madhyamika distinction between the two truths, however, which Tillemans wants to spell out in this way, is concerned with two different ways of interpreting the ontological status of objects, or, to put it differently, with two different accounts of what it means for a statement to be true. At the conventional level a statement is true if what it says is indeed the case, i.e. if there are objects taken to exist with svabhdva which are related in the necessary ways. At the absolute level, however, the notion of svabhdva is to be found to be deficient and svabhdva is seen to be non-existent. But both the referential and the substitutional interpretation of a statement can be read either way: the objects quantified over can be seen as either existing with svabhdva or being empty, similarly the truth of the sentences featuring in the substitutional interpretation can be regarded as being made true by situations regarded at the level of conventional truth, or by situations regarded at the level of absolute truth, which are then seen as empty. It seems to be th at what is im portant from the Madhyamaka perspective is not so much whether a quantified statement is read referentially or substitutionally, but the way in which the notions of ‘object’ and ‘true statem ent’ contained in these readings are spelt out. 4.3.3 Rejection of four alternatives: the case of relations All of the examples of the rejection of four alternatives discussed concerned the rejection of one-place properties, such as emptiness,72 permanence, or 72Garfield (1996, 6) is of course correct in pointing out that ‘empty of’ denotes a relation. But what Nagarjuna has in mind is clearly emptiness of inherent existence, which is a oneplace property.

finitude. Nevertheless, some of Nagarjuna’s most famous arguments in fact involve the rejection of four alternatives concerning relations. A very clear example of the employment of the tetralemma in this way can be found in the twelfth chapter of the MMK. Nagarjuna starts out by listing the four possibilities available when applying the concept of causation to suffering: Some say that suffering is caused by itself, or by something else, or from both, or that it arises in an uncaused way.73 Now we could interpret this along the lines of the tetralemma concerning properties by just regarding it as about the property of self-causation, rather than about the relation of causation. The above verse would then amount to a rejection of the following four alternatives:

1. Everything is self-caused.

2. Everything is not self-caused (i.e. caused by others).

3. Something is self-caused and (the same) something is not self-caused.

4. N ot: Something is self-caused or (the same) something is not selfcaused. While this move allows us to treat the forms of the tetralemma dealing with properties and relations as exactly parallel, I think a more natural way of reading the above argument would run as follows.

The essential difference between a property and a relation is th at a property (such as yellowness) will divide the set of objects it is applicable to (material objects) into two subsets, those which have the property (like lemons, 73 svayamkrtam parakrtam dvabhyam krtam ahetukam / duhkham ity eka icchanti [...].

bananas, curry powder and so on) and those which lack it (like strawberries, apples, chili powder and so on). A relation,74 however, divides the set of objects it is applicable to into pairs of objects from the set which are related by the relation. There are various ways in which this set of pairs can be made up; it can consist

1. exclusively of pairs containing the same object twice, or

2. exclusively of pairs containing two different objects, or

3. of both pairs of identical and distinct objects, or finally 4. it can consist of nothing at all, i.e. it can be completely empty.

Which of these possibilities obtains determines the way in which the objects in the set are related by the relation. If, for example, we consider the ‘loves’ relation and a set of human beings then in the case of 1. we are dealing with a set of egoists, where people only ever love themselves, in 2. we deal with a set of altruists, where people only ever love other people, in 3. we have the (normal) situation of some people loving both themselves as well as others, and in 4. we have an emotional vacuum: nobody loves anybody, neither themselves nor others.75

If we thus wanted to argue for the deficiency of the concept of a particular relation along the lines of the above argument, we would consider the four possibilities of th at relation relating an object to itself, relating an object to something which is not itself (where the notion of negation involved is again 74For the sake of simplicity we will confine ourselves here to two-place relations. 75 It is important not to confuse this fourth case with the inapplicability of a relation to a set: in a set of people nobody may stand in the ‘loves’ relation, and nobody will stand in the ‘is the square root o f’ relation. But it is at least possible that people could stand in the former relation, whereas it is impossible that they stand in the latter.

of the paryudasa-kind) relating an object both to itself and to other objects, and relating it neither to itself nor to other objects, that is, relating it to nothing at all. If we succeed in showing all four possibilities to be unsatisfactory we can then deny all four alternatives by a prasajya-negation and thus apply illocutionary negation to them. In this way we demonstrate the inapplicability of the concept of the relation to the objects under consideration. We can therefore read the first verse from the twelfth chapter of the MMK as arguing th at if it makes sense to use the concept of causation when talking about suffering at all, it would either have to be the case th at causation related suffering to itself (i.e. that it was self-caused) or to another thing, or both, or that suffering was not causally related to anything. As is hardly surprising, Nagarjuna sets out to argue th at the concept of causation is not applicable in this context, and ends the verse by stating the conclusion to be established: To consider [suffering] as produced is not appropriate.76 In the remainder of the chapter Nagarjuna then sets out to refute each of these possibilities. Verse 2 attem pts to refute suffering’s self-production, verses 3 to 8 production from another, and verse 9 the final two possibilities.77 76 [...] tac ca karyam, na yujyate. 77It is sometimes argued (e.g. in Wayman (1977, 11-12)) that the four possibilities concerning causation mentioned in MMK 12:1, and, more generally, in 1:1 represent the views of four different Indian schools of philosophy. Self-causation is ascribed to the Samkhyas (Murti 1955, 168-169), causation by others to the theory divine causation expounded in the Vedas and Brahinanas (see Kalupahana (1975, 5) for some other examples of what he calls ‘external causation’), causation by itself and by others to the Naiyayikas and Vaisesikas (Dasgupta 1942, 1:320), (King 1999, 208) and finally absence of causation to the Lokayatas (Kalupahana 1975, 25). This last identification is denied by Schayer (1931, note 16, 20-21) who argues that the view of the Carvakas only denies causality in the conCHAPTER

As this chapter is primarily concerned with the argumentational mechanics of the tetralemma we can disregard the precise contents of these arguments. It is, however, im portant to note the generalization stated in the final verse of the chapter: Not only does suffering not exist in any of the four possible ways described, but no other external entity exists in these ways either. 78 Thus, apart from being a specific argument about the suitability of using the concept of causation to talk about suffering, Nagarjuna takes the contents of this chapter also the be an argument-schema, that is as a framework which can be employed to demonstrate the deficiency of other concepts when referring to external entities.79

4.4 Affirming four alternatives: t e positive tetralem m a

As we saw above the tetralemma is usually employed in Madhyamaka argumentation to provide an enumeration of four exclusive and exhaustive logical alternatives, all of which are then shown to be deficient and thus rejected. There is, however, one notorious exception in Nagarjuna1 s writings, in verse text of karma, but not all causal determinations, as they assert that things are determined by their intrinsic nature (svabhdva).

78 na kevalam hi duhkasya cdiuvvidhyarri na vidyatc / bahydndm api bhavandm caturvidhyam na vidyate. 79 As Robinson (1967, 50) points out, Nagarjuna frequently indicates that his arguments function as patterns into which other terms can be substituted. For examples from the MMK see 3:8, 16:7, 19:4, and 10:15.

18:8 of the MMK. There Nagarjuna seems to affirm all four alternatives by claiming that All is so, or all is not so, both so and not so, neither so nor not so. This is the Buddha’s teaching.80 In the commentarial tradition following CandrakTrti this verse is generally understood as indicating the graded nature of Buddha’s teaching (anus as ana),81 The idea is th at ‘all is so’ is taught to ordinary disciples, in order to convince them of Buddha’s insight into the nature of phenomena. ‘All is not so’ is taught subsequently to inform them about the impermanence and momentariness of all phenomena. ‘All is both so and not so’ is taught to show that what appears to be genuine and substantial from an ordinary perspective might not do so from the perspective of a Buddha’s disciple. Finally, ‘All is neither so nor not so’ is taught to show th at neither of these terms is applicable to reality in ultimate terms, in the same way, CandrakTrti observes, as the adjectives ‘pale’ or ‘dark-skinned’ are not applicable to the son of a barren woman.82 Neither of the four alternatives is therefore to be rejected in this context. They rather form an ascending series of views of increasing conceptual sophistication, each suitable for the purposes of a specific audience.83 80sarvam tathyam na va tathyam tathyam catathyam eva ca / naivdtathyam naiva tathyam etad buddhanusdsanam. 81See Ruegg (1977, 5-7). Further references to graded teaching by Nagarjuna are in RA 3:94-96, YS 30. 82PP 371:11-12. 83Robinson (1967, 56-57), Ng (1993, 94-99). Ruegg (1977, 6-7, 63-64, note 71) argues that since each alternative improves on the preceding one and even the fourth alternative is only intended for the ‘scarcely obscured’ all four alternatives should nevertheless be rejected. (This interpretation is criticized by Wood (1994, 140-146)). Even if we accept

Garfield offers a different interpretation based on the dGe lugs interpolation procedure already discussed above.84 Here the conflict between the four alternatives is dissolved not by relativizing them to different perspectives, as CandrakTrti does, but by adding the modifiers ‘ultimately’ and ‘conventionally’. The passage is thus interpreted as saying that 1. Everything is conventionally real. 2. Nothing is ultimately real. 3. Everything is both conventionally real and ultimately unreal. 4. Nothing is either conventionally unreal or ultimately real. While Garfield does not deny that the conception of graded teaching is something ‘with which Nagarjuna would agree’, he argues th at such a discussion seems ‘out of place’ in the argumentative context of chapter 18. The reason for this is not quite clear. After all the sixth verse asserts that Buddha taught the teachings of self, non-self, and neither self nor non-self, and Garfield himself asserts that these three were meant to counteract specific wrong conceptions of the self in the mind of the listeners,85 On the whole the reading of the positive tetralemma in terms of graded teaching seems to be more satisfactory, as it does not commit us to making any additions to the text itself.86 Ruegg’s position it is clear that the four alternatives as given in 18:8 are quite distinct from all the other uses in Nagarjuna’s writings, as in all other instances all four alternatives are negated and are not even assigned a heuristic value. 84(1995, 250-251). 85(1995, 249). 86See Tillemans (1990, 73).

I hope the above remarks have made it plausible that to understand the catuskoti it is essential to keep apart the different kinds of nested negations involved. In this way it is possible to see that the four alternatives of the tetralemma are logically independent, as well as to understand how the rejection of the four alternatives (as illocutionary negations based on a presupposition failure) fits in with Nagarjuna’s general philosophical attem pt to demonstrate the non-existence of svabhdva. Compared to some accounts in the contemporary commentarial literature the interpretation presented above is logically very conservative. It does not involve anything beyond the resources found in classical logic and in particular gets by without rejecting the Law of the Excluded Middle or adopting a paraconsistent logic. While I think there are some aspects of Nagarjuna’s works (for example the notoriously complex issue of the emptiness of emptiness) which can perhaps be fruitfully interpreted by reference to some variety of dialetheism, such as the one put forward by Priest and Garfield87 this does not apply to the methodological foundations of Nagarjuna’s arguments in the catuskoti These can be explained entirely within the framework of classical logic. 87(2002).

Chapter 5 Causation

Having dealt with some important formal aspects of Nagarjuna’s arguments chiefly connected with the notion of negation we are now equipped to explore further ramifications of the Madhyamika’s rejection of svabhdva outlined in chapter 2. A part from Nagarjuna’s general arguments against the existence of svabhdva presented there we also find in his writings specific investigations of phenomena arguing that these in particular lack svabhdva. In fact a large part of Nagarjuna’s writings can be best understood as an examination of various classes of things with the aim of establishing their emptiness. The phenomena examined are those which constitute a particularly important part of our view of the world, and which are therefore the most likely places where the mistaken ascription of svabhdva could arise. In the following five chapters we will investigate Nagarjuna’s arguments concerning five such phenomena: causation, motion, the self, epistemology, and language. The fundamentality of causation for making sense of our experiences need hardly be stressed. The notions of cause and effect provide us with one of the most fundamental set of tools we use to gain cognitive access to the world. Motion might strike us as a somewhat less central topic, but it is important to keep in mind

its importance in the Indian mindset in which this discussion originates. Motion, not so much in the literal sense but in the sense of moving from one life to the next (of which the succession of one mental state by the following one provides a small-scale example), the traversing of samsara in an infinite succession of existences, is an essential feature of the Indian (and more specifically Buddhist) view of the world. The self appears to be the entity which does this traversing, and provides us with what seems to be the most important notion of them all. Viewing ourselves as a self, a subject, a responsible agent appears to be indispensable for our conception of what we are. It provides the focal point of our cognitive life, the place where the disconcertingly diverse array of different experiences comes together and is unified in a view of the world from a unique perspective. This self interacts with the world around it in a variety of ways. The description of its direct cognitive interaction with its surroundings is the province of epistemology. This analyses how parts of the world can become parts of our mind, in short, how knowledge is acquired. However, our interaction with the world is not completely solitary. We do not just read off information from the world, but also conceptualize it in a variety of ways in order to share it with others. Our primary tool for framing and sharing bits of information is language. Arguing for the non-existence of svabhdva by examining different kinds of things one-by-one is of course beset with a fundamental problem. As there might be infinitely many or at least an indefinite number of things we are unlikely ever to conclude our investigation and to establish the thesis of universal emptiness. The Madhyamaka tradition does of course offer arguments which are proposed to work as general arguments for emptiness (such as the five types of arguments discussed in chapter 2). However, we could equally argue that the absence of any master-argument for emptiness might

constitute not a difficulty, but an inescapable consequence of Madhyamaka epistemology. Since the Madhyamika regards nothing as being intrinsically a means of knowledge, what establishes emptiness in one context might fail to do so in another.1 In any case there is hardly any doubt about the centrality of the five phenomena just introduced. Regardless of our view of its consequences for establishing the general thesis of emptiness, a demonstration of the absence of svabhdva in each of them would have fundamental consequences for our view of the world, of ourselves, and the relation between the two. 5.1 C ausation: prelim inary remarks Nagarjuna’s treatm ent of causation is an examination of different attem pts to analyse the relation between cause and effect which employ a variety of familiar and (and least prima facie unproblematic) conceptual frameworks. Causation being a two-place relation we will want to investigate the identity relation between its relata: are cause and effect the same or different, or some identical and some different? Are cause and effect related as part and whole? Or are there perhaps no items instantiating the causal relation at all? Causation taking place in time, we will want to investigate the temporal relation between cause and effect: are they successive events, or simultaneous, or overlapping, or are cause and effect just two aspects of a single unified event?

The interesting fact about Nagarjuna’s discussion of these analyses is that he sets out to show that they are all equally unsatisfactory:2 cause and effect 1Siderits (2003, 147), see also the discussion in chapter 8 below. 2The reader familiar with Greek philosophy will realize that many of Nagarjuna’s arguments concerning causation bear strong similarities to classical sceptical arguments as CHAPTER 5. CAUSATION 144 are argued to be neither identical nor different nor related as part and whole, they are neither successive, nor simultaneous, nor overlapping and so forth. In itself such a discussion would show little more than that the philosopher in question had failed to come up with a satisfactory theory of causation. It would at best be interesting for the criticism levelled against other positions, but would hardly constitute a unified philosophical outlook on its own. In order to see the point of Nagarjuna’s arguments, however, it is essential to understand th at he takes the possible analyses of causation examined to be exhaustive: any view one could possibly hold concerning the identity relation or the temporal relation between cause and effect can be subsumed under one of the alternatives considered. If Nagarjuna is successful in showing that all the alternatives are deficient there is only one conclusion to be drawn from this. The conceptual frameworks which these different analyses employ must be based on a mistaken presupposition. Suppose there was some peculiar theory which tried to find out what shape the colour red was. This could proceed by listing all the possible shapes the colour red could be, and then examining them one by one, until the right one was found. Unsurprisingly we would find th at the colour red is neither circular, nor triangular, nor rectangular, nor any other shape, since while there are doubtlessly examples of such shapes which are coloured red, the colour itself, being a property, does not have any of these shapes. The explanation for our inability of coming up with a satisfactory answer to the e.g. presented in the third book of Sextus Empiricus’ Outlines of Scepticism. Since the present discussion is not an attempt at comparative philosophy I will not discuss these resemblances in the following material. The interested reader is referred to McEvilley (1982), who even addresses the question whether Nagarjuna’s arguments might be derived from Greek works (28). For a broader discussion of possible Greco-Indian philosophical interactions see McEvilley (2002).

question ‘W hat is the shape of the colour red?’ is that it is built on the mistaken presupposition that the colour red has a shape at all. In the same way Nagarjuna wants to argue that our inability to come up with a satisfactory answer to the question whether cause and effect are identical or different is due to another faulty presupposition.3 This is the presupposition th at cause and effect exist with their own svabhdva, i.e. that they are independent and self-sufficient entities.4 Cause and effect existing with their own simbhava first of all means that cause and effect are qualitatively and therefore also quantitatively distinct objects. They do not require one another: first the seed exists without any need for the tree it will later produce; later, after the seed has produced the tree the seed has stopped to exist, and the tree will exist without any need for the seed to be still around. Secondly, the existence of cause and effect as ‘independent objects’ or as ‘existing from their own side’ does not just refer to their mutually independent existence, but also to their independence of a cognizing subject. According to such an objectivist understanding of causation the interlocking chain of causes and effects is something which exists in the world independent of any observers.5 It might be a transfer of energy from cause to effect, the cause’s raising the objective chance of the effect happening, or perhaps an unanalyzable, primitive causal relation, but in each case it is something 3See Ronkin (2005, 198). 4See Garfield (1994, 220), (2001, 509) 5Such an objectivist understanding of causation is entailed by the Abhidharmika’s claim that primary existents ( dravya) can be dependent on causes and conditions. For if the existence of these primary existents is mind-independent the existence of one such object, a conditioned object (samskrta dharma) which is an effect cannot depend on another one by a relation which is itself mind-dependent. This point is elaborated in Siderits (2004, 410-413).

which remains independent of human expectations and conceptualizations. Nagarjuna argues th at this commonsensical view of cause and effect constitutes the basis of the conceptual framework we employ in order to analyse causation. Should it now turn out, as Nagarjuna sets out to demonstrate, th at there is something problematic with all the analyses usually encountered this would provide an argument for questioning the commonsensical view of cause and effect underlying all of them. Before investigating different accounts of causation, however, we have to consider further what Nagarjuna means by the lack of independence and therefore the interdependence of cause and effect. 5.2 Interdependence o f cause and effect In order to get a clear conception of Nagarjuna’s view of the interdependence of cause and effect it is necessary to understand that his analysis of causation does not just distinguish two notions, cause and effect, but three. This is due to the fact that a cause does not bring about an effect on its own, but only against a background of supporting condition. A spark does not cause an explosion on its own, but only in the presence of oxygen, fuel, a suitable temperature, and so on. The Madhyamaka analysis of causation therefore includes reference to a collection of background conditions. Nagarjuna refers to the cause together with the background conditions as a ‘complete collection’ (.samagrt); we shall employ the term causal field for this.6 Nagarjuna asserts 6I<alupahana (1991, 61). It is interesting to note that the distinction between cause (hetu) and supporting conditions (pratyaya) is not found in early Buddhist texts (Ronkin 2005, 222) where the two terms are often used interchangeably. Kalupahana (1975, 59) notes that ‘while recognizing several factors that are necessary to produce an effect, it [i.e. early Buddhism] does not select one from a set of jointly sufficient conditions and

in a variety of places that cause and effect are interdependent, and can only be conceived of in such an interdependent fashion.7 It is apparent, however, that the dependence of an effect on a cause must be very different from the dependence of a cause on an effect. An effect depends for its existence on its cause: had the cause not existed, the effect would not have existed either.8 A cause, on the other hand, can exist without causing any effect, it would just not be described as a cause in this case. The fact that some particular acorn does not produce an oak tree does not mean the acorn does not exist, only that we do not refer to it as ‘the cause of an oak tree’. This label is attached to it not because of some internal property, but simply because it stands in a particular relation to another object, namely the oak tree. In the absence of the oak tree there would be 110 relation to that oak tree, so that the label would be inapplicable.9 presents it as the cause of the effect. [...] Thus, although there are several factors, all of them constitute one system or event and therefore are referred to in the singular.’ See also Ronkin (2005, 206). The distinction between cause and condition as two different elements involved in the causal relation is due to the Sarvastivadins (Frauwallner (1995, 199-201), Ronkin (2005, 221-232)) and coheres well with their distinction between primary existents or substances ( dravya) and secondary existents (prajnapti) (Williams 1981, 237). The cause of a particular secondary existent would be regarded as the primary existents it is based on, while its conditions could be seen as whatever causes its properties as a secondary existent. See Kalupahana (1975, 60-66). 7MMK 8:12, 10:8, 20:22, 24; see also SS 13ab, BCA 9:13-15. 8Nagarjuna makes the additional claim that everything is existentially dependent on its cause, since ‘no object whatsoever exists without being caused’ (na casty arthah kascid aketukah kvacit MMK 4:2). 9In MMK 1:5 Nagarjuna states that ‘something is called “condition” because in dependence on it something else arises. But as long as the second something does not arise, why do we not refer to the first something as a ‘non-condition’?’ utpadyate pratityeman itime pratyaydh kila / yavan notpadyata ime tdvan ndpratyayah katharn. In 10:9 he argues that if the dependence of cause and effect only held in one way, i.e. if fire (the effect) depended

Taking into account the distinction between existential and notional dependence described in chapter 2 it is clear th at some of the dependence relations Nagarjuna asserts to hold between cause and effect are quite straightforward. Cause and effect are notionally dependent on one another. If anything falls under the concept ‘cause’ something will fall under the concept ‘effect’, and vice versa. The effect also depends existentially on its cause, since in the absence of the cause, the effect would not exist. The main problem for understanding the supposed symmetric dependence between cause and effect lies in the assumption of the existential dependence of the cause on the effect. After all, as we have just seen in the case of the acorn (the cause) and oak tree (the effect), the latter failing to exist does not obliterate the former, so that we can at best speak of notional dependence in this context. There are three distinct ways in which we can make sense of the existential dependence of the cause on the effect. Firstly we can argue that if Nagarjuna is arguing against an opponent who holds that a cause has its property of being a cause essentially its notional dependence on the effect will entail its existential dependence. For something being a cause essentially means that this is a property it could not lose without ceasing to be that very object. But since the presence of this property depends on existence of the effect, the existence of the cause as that very object also depends on the existence of the effect.

A second interpretation which does not have to assume th at causes are essentially causes argues that Nagarjuna does not intend to refer to the exon fuel (the cause) but not vice versa this would imply the absurd consequence that the cause could exist as a cause without the effect existing, yad mdhanam apeksyagnir [...] evarn satindhanam capi bhavisyati niragnikam. CHAPTER 5. CAUSATION 149 istentia.1 dependence of some particular cause on its effect but rather to the existential dependence of the property of being a cause on the property of being an effect. If an object a falling under property F notionally depends on something falling under G this means th at the property F existentially depends on the property G, as F can only exist if some object falls under the property ‘identical with the property G’, i.e. if the property G exists. The property ‘Northern England’ depends existentially on the property ‘Southern England’, even though the objects falling under each do not existentially depend on one another. This is due to the fact th at one property could not exist without the other one, but the objects falling under them could. Nagarjuna refers to this existential dependence of properties on one another in RA 1:49:10 When there is no ‘short’ there is no ‘long’, they are without substance. When there is no lamp, there is no light. The existential dependence of the effect (the light) on the cause (the lamp) is here equated with the way the properties ‘long’ and ‘short’ depend on one another — in each case the latter could not exist if the former did not exist. According to this interpretation we would therefore conclude th at Nagarjuna means to say that the properties ‘being a cause’ and ‘being an effect’ w hrasve ’sati punar dirgham na bhavaty asvabhavatah / pradipasyapi anutpddat prabhayd apy asambhavah, Kalupahana (1975, 97) interprets this statement of Nagarjuna as ca rare interpretation of the causal principle’. This seems to be getting the order of concepts in the development of Nagarjuna’s thought the wrong way round. For him the notion of a dependence relation between objects is the more general concept, which can take a number of specific forms (such as mereological, causal, and cognitive dependence). Dependence is not a specific interpretation of causality, but causality a specific interpretation of dependence.

depend existentially on one another, even though the existential dependence of objects falling under them is not symmetric: the effect depends existentially on the cause, but the cause does not need the effect for its existence. The third, stronger reading claims that while Nagarjuna undoubtedly also wanted to assert the existential dependence of the properties ‘being a cause’ and ‘being an effect’ he moreover made the claim that not only the particular object which is the effect needs the cause for its existence, but the cause also needs the effect. Such a reading can be supported by considering an entire causal field rather than just particular causes. A causal field is a cognitive artefact, a collection of objects assembled with the sole purpose of explaining why a particular effect came about. Divorced from this explanatory role there is no reason for introducing the concept at all.11 We might therefore want to argue th at the causal field also depends for its existence on the effect it produces. This is of course not to say that every member of the causal field existentially depended on the effect they jointly bring about: the spark, petrol, and so forth would still exist, even if they for some reason did not manage to bring about an explosion. But the collection only exists if there is some effect it causes. Whether we want to argue th at a causal field depends for its existence on the effect it brings about is intimately connected with our view of the existence of collections. We might either think that whenever the are some objects there is the collection of those objects. Or we might deny that every arbitrary assembly of objects constitutes a collection. We would then argue that for some objects to form a collection there must be something which makes them hold together as a collection, for example that they all 11‘Because the effect is absent, where would conditions or non-conditions come from?’ phalabhavat pratyayapratyayah kutah. MMK 1:14b.

exemplify a property, or that they were put together for a specific purpose. If we adopt the first view of collections then clearly a causal field will only depend nominally on its effect, since ‘being a causal field bringing about that effect’ is only one way in which we can refer to the pre-existent collection which contains all the elements of the causal field, but not anything which brings it about. Adopting the second conception, however, it may be the case th at the only thing which binds all the members of the causal field together is th at they are considered to be the things which jointly bring about a particular effect. In the absence of this effect the collection disintegrates and ceases to exist. Nagarjuna seems to favour the second interpretation when he asserts that a cause could not exist without an effect.12 It is significant that what is denied here is not just the ascription of the label ‘cause’ to some object because it is related to some other object, the effect,13 but the existence of the cause in the absence of the effect. Nagarjuna endorses not just the uncontroversial notional dependence of the cause on its effect, but its existential dependence as well. Applied to the discussion of causal fields this implies that a causal field can only exist if the effect it brings about does, and for this reason cannot be taken to exist whenever all of its members do. If we adopt this third, stronger reading then we have to conclude that for Nagarjuna causes and effects are both notionally and existentially dependent 011 one another. They therefore cannot exist from their own side, irrespective of the existence of one another. Moreover, they also depend for their existence on us, because it is our cognitive act of cutting up the world of phenomena in the first place which creates the particular assembly of objects which constitutes a causal field, which then in turn gives rise to the notions of cause 12 nasty akaryam ca kdranam. MMK4:3. ‘How indeed can there be a causal field in the absence of an effect?’ asti pratyayasamagri kuta eva phalam vina. 20:24b. 13This is discussed in MMK 1:5.

and effect. This entails that the causal field, cause and effect are empty of svabhava. It is evident th at unlike the Abhidharmikas, Nagarjuna regards an object’s not being empty, i.e. having svabhava as incompatible with the causal production of th at object.14 A causally produced object depends on its cause for its existence and could therefore not stand outside of any dependence relation with other objects. Furthermore, if an object either existed or failed to exist by svabhava, it would always do so, since such substantial properties cannot change. But then an existent object cannot be caused, since it will always have existed, a non-existent object cannot be caused either, as it will never become existent. Therefore, in the presence of svabhava (and thus the absence of emptiness) there can be neither causation nor change.15 ft is equally clear that an object cannot depend notionally on some other object for having some property and yet have this property by svabhava. For this property is then obviously not had by the object from its own side, but only via its relation to some other object. Neither the existential nor the notional interdependence of cause and effect is thus compatible with ascribing svabhava to them. It has also to be noted that Nagarjuna asserts, somewhat puzzlingly, that the absence of svabhava, i.e. emptiness is not compatible with causation either. 16 Causation in this context has to be understood as an objectively obtaining relation which links objects and events independent of human con14MMK 15:2ab, YS 19. See also Siderits (2004, 399). 15SS 5: ‘What was born will not be born, what was not born will also not be born. The being born will not be born either, because it was both born and not born.’ skyes ba bskyed par bya ba min / ma skyes pa yang bskyed bya min / skye ba’i tshe yang bskyed bya min / skyes dang ma skyes pa yi phyir. See also MMI< 20:17, 21a. 16MMK 20:18, 21b.

ceptualizations. As the objects our theory relates do not exist ‘from their own side’ the same has to hold for any relation of causation linking them. If the objects in our everyday world owe their existence to a partly habitual, partly deliberate process of cutting up the complex flow of phenomena into cognitively manageable bits the causal relations linking them cannot exist independently of us, since their relata do not do so either.17 5.3 T he four ways of causal production Having investigated what Nagarjuna means by claiming th at cause and effect are interdependent we can now discuss his analysis of the different manners in which causal production could come about. Applying the four alternatives of the catuskoti or tetralemma to the relation of causation Nagarjuna distinguishes four ways in which a thing could be causally produced:18 • it could be caused by itself • it could be caused by something else • it could be caused by both

CandrakIrti’s commentary on 20:18 in the PP which makes this point is peculiar. In explaining Nagarjuna’s assertion that empty objects cannot arise or cease (’How can the empty arise, how can be empty be removed? It follows that the empty is not ceased and not arisen.’ katham utpatsyate sunyam katham sunyam nirotsyate/ sunyam apy aniruddham tad anutpannam prasajyate) he refers to the example of space (akaso), according to Abhidharma metaphysics an unconditioned phenomenon which neither arises nor ceases (Dhammajoti 2004, 383-384). But this ignores the fact that Nagarjuna here (as in the preceding verse) attempts to make a universal statement. As it is obviously not assumed that all empty phenomena are like space in lacking arising and ceasing this reading of the verse appears to be rather misleading. 1SMMK 1:1; 12:1.

• it could be caused by nothing at all 5.3.1 Self-causation The Buddhist discussion of self-causation subsumes two very different views of causal production which have to be separated clearly. The first is the view th at cause and effect are the very same object, so that at least in some cases the causal relation can relate an object to itself. The second does not assert th at cause and effect are identical, but that the effect is contained in, and forms part of, the cause. In the context of the present discussion this should be taken to mean that the effect is already contained in the causal field. Identity of cause and effect It is evident th at most instances of causation we encounter in everyday life are not cases of self-causation in the first sense. The spark which causes the explosion is not identical with the explosion. The causal field (the seed together with water, light, warmth etc.) is what is supposed to bring about the effect (the tree), but is not yet the effect itself — the collection of the seed, water, light and so on is not the same thing as a tree but something on which the existence of the tree depends.19 If we talk about one billiardball colliding with another billiard-ball which it thereby causes to move we talk about two balls, not just one.20 Nagarjuna’s rejection of causation as a reflexive relation21 does therefore not appear to be particularly controversial. We might think, however, that self-causation is what explains the persistence of objects through time. ‘W hat causes the existence of the present billiardl9MMI( 12:2. 20MMK 20:20a. 21MMK 4:6a; 20:19. CH APTER 5. CAUSATION 155 ball? Its existence at the preceding moment. Therefore the billiard-ball is the cause of its own persistence’. But this would be mistaken. If we assume that the billiard-ball has temporal parts (that is, if we are endurantists) the present temporal part of the billiard-ball being caused by a past temporal part is not an instance of self-causation, but an instance of one part of an object causing another, distinct part. Being perdurantists and assuming th at an object is wholly present during each moment of its existence seems to be a more satisfactory option. Here the object is not split up into temporal stages and being a cause turns out to be part of the svabhava of every perduring object. As long as the object exists it must be its own cause, moreover, its causal role is independent of any cognizing subject. We do not have to conceive of an object as its own cause for it to continue to exists as its own cause. The problem with this view is that is conflicts with our deeply ingrained assumption th at the cause temporally precedes the effect. No object, however, can temporally precede itself. A more successful example to which the notion of self-causation could be applied would be that of a creator god. Since most theists would not want to hold that the existence of such a god is contingent on something else they might want to argue that the creator god exists as causa sui, thereby only causally depending on himself.22 Such a theist would obviously not be very impressed by Buddhapalita’s argument against self-causation which Candrakirti mentions in his commentary on MMK 1:1, namely th at a self22This idea goes back at least to Plotinus (Hadot 1971), (Narbonne 1993). Everything which exists as causa sui obviously has its causal power as part of its svabhava. Another way of dealing with this issue consists in asserting the creator god’s aseity, i.e. his causal independence of everything, including himself.

caused entity would continue causing itself and would therefore be eternal.23 Eternal existence is one of the welcome consequences of regarding the creator god as his own cause. A more substantial difficulty with the notion of selfcausation appears to lie in the justification of using the term ‘causation’ at all. An object which causes itself cannot exist in time, at least if we assume that a cause must temporally precede its effect. Given that temporal priority is an irreflexive relation, a potential self-causer could not be a temporal object.24 This is not problematic if the god is taken to exist outside time. But then it is not clear why such a non-temporal object would have to produce itself all over again, given that it already fully exists when it exists as a cause. As such a renewed production cannot be required for its continued existence in the next moment in time, its production would be both without meaning (artha) and without purpose (prayoja).25 If the causal relation has any essential properties its role as a transm itter of change is surely one of them. Causation relates a state of affairs which is a cause, that is a state of affairs in which the effect is not yet present to one in which it is present, and which has thereby changed.26 But a self-caused object could not change. Since its entire cause 23‘If the existent were to be born it would never not be born.’ atha sannapi jayeta na kada cinna jayeta PP 14:2-3. See MMK 10:1-4. 24See Hart (1987, 183). 25‘Things are not produced from themselves because of the pointlessness of that production. [...] There is no purpose in the repeated production of existent things from themselves.’ na svata utpadyante bhdvdh tadutpadavaiyarthyat [...] no hi svatmana vidyamandndm paddrtkanam punarutpade prayojanam asti PP 14:1-2. See also MA 6:8cd. 26Nagarjuna relies on this fact in his argument against suffering being self-caused (MMK 12:4, 8). If some person created his own suffering this means that the effect (the person with suffering) was preceded by a cause which is different from the effect (the person without suffering), so that the transition from cause to effect could bring about the necessary change. But given the Buddhist identification of the notion of a person (pudgala), its constituents (skandha), and suffering (duhkha) (see Schayer (1931, 12-19 )) such a person

lies within the very object which is both cause and effect there is no room for variation: such an object remains eternally the same. It thus appears th at when the theist speaks of self-causation he chooses a rather misleading way of talking about an object he considers to be atemporal, acausal, and changeless. Of course Nagarjuna and his commentators do not think that such an object exists, but this is not the point at issue here. Nagarjuna does not attem pt to argue that there just happen to be no self-causing objects, but that the very notion of something causing itself is problematic.27 Even if the theist’s creator god existed we would not want to refer to him as self-caused. Cause and effect related as part and whole The second view of self-causation generally discussed asserts th at the effect is in some way already part of the cause or, more specifically, part of the causal field. A predecessor of this view can be found already in the Purusa Sukta of the Rg Veda which describes the creation of the world as the dismembering of the cosmic giant. The different parts of the world to be created (the four castes, earth, sky, sun, and moon) are already present in the body of the cosmic giant as parts, and only have to be separated from one another in order to be brought into existence.28 Nagarjuna gives two reasons for rejecting this mereological view of selfcausation. First of all this would mean that the effect would not have to be produced, since it is already present within the causal field. The causal field, however, is supposed to be that which brings the effect about in the first could not exist; the self-causation of suffering is thereby ruled out. See also Garfield (1995, 203-204). 27Ganeri (2001, 52). 2810.90. For a translation see O’Flalrerty (1981, 28-32). CH APTER 5. CAUSATION 158 place.29 Secondly, the effect would have to be something which we should be able to conceive of (grhyeta) within the causal field.30 Acquaintance with all the elements of the causal field (the spark, the fuel, the oxygen present etc.) does not acquaint us with the effect (the explosion). Of course this does not mean that given a complete scientific description of all the elements of the causal field we might not be able (at least in some cases) to infer the effect. But this is not what Nagarjuna has in mind here, he is concerned with the literal presence of the effect within the causal field, not with its relation to the field via an inferential relation, which could perfectly well obtain without the effect being part of the causal field. It is evident that according to this view of self-causation cause and effect do not exist independently of one another. Since an object depends on its parts for its existence, if the effect is one of these the object will existentially depend on its effect, However, cause and effect can be considered to exist independently of a cognizing subject, at least as long as we assume th at which parts an object has does not in any way depend on us but is an independent fact about the constitution of that object. If the mereological constituents of objects exist from their own side and causes and effects are special kinds of mereological constituents, causes and effects will also exist from their own side. The historical precedent of the view of causation in terms of part and whole is the satkaryavada theory of causation defended by the Samkhyas.31 29MMK 20:1, 16b; SS 4a; VP 24. 30MMK 20:3.

Prauwallner (1973, 1:303-306), Sharma (1960, 151-152). Parallels have been drawn between the satkaryavada theory and the Sarvastivadin assumption that not only present, but also past and future phenomena exist. The idea of a future effect being already real, though not yet present is very close to the notion of an effect existing in a latent but unmanifested form in the cause. See Stcherbatsky (1962, 1:111), Kalupahana (1975, 150- C H A P TE R S. CAUSATION 159 This account is somewhat more sophisticated than the postulation of the mereological containment of the effect in the cause. The Samkhya theory assumes th at the effect (karya) is already existent (sat) within the cause, although only in a yet unmanifested form. Everything we need for the production of a pot is already there in the causal field containing the lump of clay, the potter, and so on. Bringing about a cause is a transformation of the causal field, an unveiling of the previously hidden.32 For the Samkhya theorists the world is nothing but a sequence of transformations of primordial matter (prakrti). In a similar manner we might want to conceive of the universe as an arrangement of atomic particles, and of each future stage of the universe as a re-arrangement of these particles in some new way. On this conception causation would indeed never bring anything new into the universe, since all the combinatorial resources for the causal production of future stages are already there. All changes concern the way the individual particles are related to one another, but not what kinds of particles there are. The Samkhya doctrine has the advantage of being able to account for the fact that specific causes are related to specific effects.33 Curd can be made from milk but not from oil because only the milk-, not the oil-particles allow for a curd-re-arrangement, in the same way as ‘sator’ anagrams into ‘rotas’ but not into ‘horas’. However, even if the underlying idea of a fixed stock of primordial m atter should prove to be correct we should be as reluctant 152), von Rospatt (1995, 39, note 72). Samghabhadra in the *Abhidharmanyayanusara, Taisho 1562, 635a argues against this identification. See Dhammajoti (2004, 100). Murti (1955, 172) and Siderits (2004, 404) subsume the Sarvastivadin position under asatkaryavada theories of causation. 32Shaw (2002, 215). 33Siderits (2004, 404). CH APTER 5. CAUSATION 160 to accept the Samkhya conception of the presence of the effect in the cause as we would be to purchase a single volume at the price of an entire library, on the grounds that all the other books are ‘already present’ in this one, and just have to be brought from the unmanifested state to their manifested one by a transformation of the arrangement of letters.34 The presence of the m atter of the effect in the causal field is not sufficient to convince us that the effect is also present there, in the same way in which the presence of the letters making up a particular text is not the same as having the text in front of us.

5.3.2 Causation by another object

The idea th at cause and effect are distinct phenomena and that therefore an effect is caused by something distinct from it certainly constitutes the most natural way of understanding causation. The spark is an event distinct from the explosion it causes, the movement of the first billiard-ball distinct from the motion of the second one it brings about. The naturalness of this view of causation is underlined by the fact that in the Madhyamaka discussion and rejection of the four alternative accounts of causation the greatest part of the argument is usually devoted to the examination of the second alternative, i.e. to the claim th at objects are caused by something different from them. The initial plausibility of causation by other objects means that its refutation has to be worked out particularly well. There existed a number of contemporary theories which Buddhist thinkers were likely to subsume under the label ‘causation by another’. An obvious example is the theory of divine causation present in the Vedas and Brahmanas 34Compare Brunnholzl (2004, 768).

which the Buddhists argued against.35 The view of the Carvakas, who regarded change as a rearrangement of material particles according to fixed laws of nature was also classified in the Nikayas as paramkatam presumably because it leaves no room for a human agent.36 Determinist theories like those of the AjTvikas37 which denied free will were included amongst accounts of ‘causation by another’ as well.38 We can identify two main arguments put forward by Nagarjuna to refute causation from other objects. The first claims that causation from other objects entails an infinite regress.39 There are different ways in which we 35See Kalupahana (1975, 5) for some other examples of what he calls ‘external causation’. We also find interpretations which regard causation by Brahman as a mere transformation of the cause into the effect, which would subsume them under the theories of self-causation already discussed (Dasgupta 1942, 1:52-53). dPa’ bo rin po che’s commentary on the BCA specifically considers causation by the creator god Isvara (dbang phyug) as an example of causation from another (Brunnholzl 2004, 758-762). 36Ronkin (2005, 196-197). In the Tibetan tradition we find this position ascribed to the MTmamsakas (dpyod pa pa) (Brunnholzl 2004, 762). For more details on the Carvakas’ philosophical position see also Frauwallner (2003, II: 196-200), Steinlcellner (1986, 8-12).

Frauwallner (1973, 1:213). 38The justification for this classification becomes more transparent when considering that in the Nikayas the Buddha specifically considers causality by discussing the causal production of suffering (Ronkin 2005, 195). A view which denies the existence of human free will is one which has to explain the existence of suffering by reference to some cause other than the agent experiencing the suffering. The Buddha’s ethical discussion of causation concerning the origin of suffering was later generalized to incorporate a metaphysical discussion as well (see Ronkin (2005, 198)). 39MMK 7.19. A third, distinct argument is given in verses 5-8 of chapter 12, dealing with the existence of suffering. Nagarjuna examines the possibility that the suffering of a human being is caused by another, namely by an earlier stage of that person, for example by that person in an earlier life. Since on the Buddhist conception suffering (duhkha) and the person experiencing the suffering (pudgala) are taken to be necessarily coextensive,

could spell out the argument Nagarjuna has in mind here. Assume that some object x was caused by some distinct object x — 1. Now the cause of x — 1, which we will call x — 2 cannot be the same as x — 1, for then x — 1 would be self-caused, an alternative which has already been refuted. But it can also not be identical to x , since this would result in a causal loop, where first x causes x — 1 which then in turn causes x. Apart from the fact th at this leads to problems with the temporal ordering of causation (if the cause must precede the effect x will precede itself) causal loops also entail self-causation, even though an object will here not cause itself directly, but only via an intermediate chain of other objects.40 It therefore follows that x % x — 1, and x — 2 are three distinct objects. Given that the choice of these three was arbitrary, causation by another object thus entails the existence of infinitely many objects. Ganeri41 argues th at this result is sufficient to rule out causation from another thing because fit cannot [...] be a m atter of logical necessity that the world of objects is infinite’.42 Furthermore such a view of causation creates problems for the notion of causal explanation because ‘one never reaches the explanans\ A3 The problem with this interpretation is th at while we would agree th at e.g. a logicist like Russell found it difficult to establish the existence of infinitely this move does not help a lot. After all the earlier stage of a person must have been experiencing suffering too, so that we now have to explain where this suffering came from, and so forth. But it is evident that this argument does not generalize to show the nonexistence of causation from other objects tout court since the cause does not always share the property it is supposed to bring about, as in the case of the various stages of the person: the firewood, which is the cause of the fire, is not already blazing. 40For an argument that causal loops are logically possible see Lewis (1986a, 75). 41 Ganeri (2001, 52-55), 42Ganeri (2001, 52). 43Ganeri (2001, 55).

many objects as a truth of logic44 it does not seem at all problematic that our concept of causation might logically entail that the world was infinite (whether the world is in fact infinite is of course a distinct question). In particular I do not see why this should be a problem for Nagarjuna, as I am not aware of any claim of his to the effect that the infinity of the world cannot established by conceptual analysis alone, or indeed that the world is finite. Nor does it seem much more plausible to assume that a chain of causes stretching infinitely far backwards vitiates the concept of causal explanation. After all this only means that for every causal explanation we can come up with another one to explain the fact referred to in the explanation in turn. But this is a property of explanations more generally: we can always demand (and in most cases also provide) an explanation of an explanation given. In fact it is hard to come up with an example of an explanation where the explanans itself is unexplainable. Even if there are some, explanantia in general are not like this. In fact the only piece of textual evidence we have in connection with this refutation of causation from another does not claim that there is any problem with a world of infinitely many objects, but speaks specifically of the infinite regress produced by the assumption that objects are caused by objects distinct from them.45 It is therefore more satisfactory to understand Nagarjuna here as addressing a criticism of the following form: ‘Of course it is not possible to assume that an object both has svabhava, i.e. exists independently, and is causally produced, due to its dependence on the cause which brings it about. However, we can circumvent this difficulty by incorporating the causes of an object into our conception of the object. Instead of speaking of 44Potter (2000, 151-152). i5 anya utpadayaty enam yady utpado ’navasthitih. MMK 7:19.

the object which is a sprout we refer to the whole causal complex of which the sprout is the final result: the seed together with water, moisture and so on. It is such causal complexes (rather than individual objects) we want to ascribe independent existence to .’ As Nagarjuna points out, this procedure is only viable if you can justify drawing the line at the inclusion of the sprout, water, moisture etc., but without including anything else in the complex. This would be possible if the things included were indeed self-caused, an argument which is unfortunately not available to us.46 Thus enlarging the conception of an object by including its distinct causes will always allow us to include more things, obliterating the distinction between objects altogether and thereby defeating the point of the exercise. For in order to arrive at a plausible candidate for the ascription of svabhava we have to identify some things which could act as starting-points of the chain of causes. But these things would either have to be their own causes or arisen without a cause. Each of these possibilities is rejected by Nagarjuna, as our discussion in sections 5.3.1 and 5.3.4 shows.47 46This point is stressed in MMK 12:7-8: ‘If we cannot show that it has been caused by itself, how could suffering have been created by another? If another one made the suffering, this one would have to have caused his own suffering. Suffering is not self-caused insofar as nothing is self-caused. If the other is not self-caused, how can suffering be caused by another?’svayamkrtasyaprasiddher duhkham parakrtam kutah / paro hi duhkham yat kurydt tat tasya sydt svayamkrtam / na tavat svakrtam duhkham na hi tenaiva tat krtam / paro natmakrtas cet sydd duhkham parakrtam, katham. The point made here is that we can only speak of causal complexes as independent existents if we stop going back in the chain of causes at some point. The only systematic reason for stopping at a particular cause would be the fact that tracing down its cause would bring us back to the object itself, i.e. the object was self-caused. But the possibility of self-causation has already been ruled out earlier. 47It should be noted that some of the modern commentarial literature also ascribes a different argument from an infinite regress to Nagarjuna’ attempts of refuting causation by

The second Madhyamaka argument against the distinctness of cause and effect is not based on any difficulties with an infinite series but tries to show that if cause and effect were different objects, the complete absence of causation would be entailed, so that we could not draw any distinction between those objects which are causally related and those which are not.48 This might strike us as a particularly strange argument. W hat is wrong with assuming that the spark causes the explosion, though the spark is not the explosion, nor the explosion the spark? The important point to realize is that two distinct objects which are causally related could not exist independently, i.e. each having its own svabhava. This is due to the fact one will existentially depend on the other, which is its cause, while the other will other objects (Garfield 1995, 113-114), (Siderits 2004, 405-406). The defender of causation by other objects must explain why only particular pairs of objects, but not others, are related to one another as causes and effects. The obvious answer in this case it to say that only some, but not all objects are linked by the causal relation. Now the obtaining of the causal relation itself either does or does not rely on conditions in turn. If it does not it is unclear how much explanatory gain results from the postulation of a causal relation. If all we can say to explain that joghurt comes from milk but not from oil is that milk and joghurt stand in a particular primitive relation it seems as if we have not explained much. After all what we want to know is why certain pairs of things but not others stand in this relation. If, on the other hand, the obtaining of the causal relation in turn depends on conditions, we now have to explain what links the relation and its relata. And for whatever provides that linkage we can ask what finks it in turn to the things it links and so on, following the familiar Bradleyan regress. Of course we can just reply to this that a relation no more needs another relation to link it to its relata than glue needs superglue to make the glue stick to the things it is supposed to glue — the glue sticks all by itself, and relations are self-linking in the same way. (This is in fact the Naiyayika reply to Bradley’s problem. See Siderits (2004, 417, note 27)). There thus do not appear to be great problems for the defender of causation by other objects to seize the second horn of the dilemma, so that the ‘infinite regress’ argument fails on this particular interpretation. 48prthaktve phalahetvoh syat tulyo hetur ahctuna. MMK 20:20. See also 10:1a.

depend at least notionally on the one, because that is what makes it possible to describe it as a cause. When the Madhyamika speaks of causation by distinct objects it is this kind of distinctness he has in mind: cause and effect are supposed to exist independently, it is not sufficient to assume that they merely differ by having some different properties. But if we have a collection of objects such that each exists independently of each other it is very hard to see how we could make a principled distinction between those objects in the collection which are causes and those which are effects. Any division of the collection into causes and effects seems to be as justifiable as any other, so th at there is no distinction between the relation of causation and any other arbitrarily chosen relation defined on the set. Moreover, if we observe a collection of objects which all exist independently of one another over a period of time we realize that nothing happens in this collection at all. The different objects will just sit there, without influencing or changing each other. If we conceive of causation as bringing something about which was not there at an earlier moment, no causation will be found in such a world.49

It is im portant to realize th at this argument is not based on any assumption claiming that ‘the effect must resemble or pre-exist in its cause5, so that two distinct objects (which ex hypothesi did not resemble one another in such a way) could not be causally related.50 As we have seen above the idea of the pre-existence of the effect in the cause is explicitly denied by Nagarjuna.51 49‘[Fuel] which is different [from fire] is not reached; unreached it does not burn, moreover, not burning it will not be blown out. Not blown out it will continue to blaze, like something having a property essentially.’ anyo na prdpsyate ’prapto na dhaksyaty adahan punah / na nirvasyaty anirvdnah sthdsyate vd svaliiigavdn. MMK 10:5. 50This interpretation is given by Ganeri (2001, 54). 51See also MMK 20:1-3.

The point is rather that independently existent objects (whether resembling one another or not) could not be the relata of a causal relation. Given that one exists whether or not the other exists, how could one be the cause bringing the other one about as an effect? If, however, we still want to talk about such objects in causal terms we have to ask ourselves what the point of this is supposed to be. After all, there is no more justification for claiming that one object is the cause, the other the effect, rather than the other way round, or that the members of some specific pair of objects, rather than those of another one, stand in the relation of cause and effect. This, however, would only mean that whatever relation we are talking about here is not the causal relation, since the causal relation is not subject to this kind of arbitrariness. Assuming causation from another is a straightforward way of arguing for cause and effect existing by svabhava. As they exist independently they do not require one another. But trying to establish causal relations (which are dependence relations) between such independent objects then leads to all sorts of problems. First of all, to escape the obvious contradiction of an independent (causally) dependent object we can try to revise our conception of ‘object’, including everything an object causally depends on in this new conception of object and regarding this collection as the bearer of svabhava. As we have seen above this implies the difficulty that the things included bring with them what they causally depend on, which in turn bring other objects along and so forth. There is no mind-independent criterion for deciding where to draw the line between which objects to include and which to leave out. Secondly, a set of independently existent objects does not give us any indication of how the causal relations between them should be established. Since the existence of any object does not influence the existence of any other

object it appears to be completely arbitrary which way round we consider the causal relations between the objects to hold. For these reasons the assumption of distinct, independently existent objects does not support the view that cause and effect exist with their own svabhava. 5.3.3 Causation by itself and another object The third alternative to consider is that an object is caused both by itself and other objects. This possibility is usually dismissed very briefly in the Madhyamaka literature by claiming th at since self-causation and causation by other objects have already been refuted individually, there is no need to refute both of them together.52 To illustrate this point Candraldrti gives the rather unfortunate example of two conditions, neither of which is fatal, arguing that this demonstrates th at both together could not be fatal either.53 That this implication does not hold can be seen from the example of binary poisons which consist of two chemicals which are non-toxic individually, but poisonous when combined. I do not think that Nagarjuna here argues by relying on the (faulty)54 principle th at if neither of two entities has a property, both of them put together will not have the property either. W hat he wants to show in this context is that if we have disproved each of a set of two propositions we do not need a further argument to disprove their conjunction as this is entailed by 52MMK 12:9a states the contrapositive: ‘If suffering was caused by both [itself and others] it could be caused by each individually.’ syad ubhdbhydrn krtam duhkham syad ekaikakrtarn yadi. S3,It is not the case that who is not killed by each individually is killed by both [together] na caikaikena prandtipate ’krte dvabhyam krta. PP 233:6. 54 As even limited exposure to mixing paint or cooking demonstrates.

the individual refutations.55 It should be noted, however, that it is possible to interpret the third possibility in such a way that the simple refutation of self-causation and of causation by another object will not be sufficient. This is the case if we take the third possibility to be the position th a t the cause already contains the effect as a potentiality which is only brought out in the presence of particular supporting circumstances.56 In this way the effect, such as an explosion, is neither wholly caused by objects different from it (since it was already present as a potentiality in the spark which caused it) nor is it just produced from itself as the potentiality present in the cause (because without the auxiliary conditions, such as the temperature required or the presence of oxygen the explosion would not have happened). Therefore57 the happy compromise doctrine that emerges is the doctrine of causation-by-both: Effects are the result of the joint operation of the effect itself in potentia and the external conditions necessary to raise the effect’s mode of existence from potentiality to actuality. It is apparent th at on this understanding of the third possibility it is not enough just to point at the refutations of the first and second possibility. For in this scenario there is neither a perpetually self-reproducing object due to the necessity of auxiliary conditions, nor do we have to suppose th at there is a causal relation holding between existentially independent objects, given that cause and effect are connected by the latter being present as a potentiality 55See Ganeri (2001, 52). 56This view of causation has been ascribed to the Jains (Perrett (1998, 2:267- 257), Ronkin (2005, 197-198)). It coheres well with their multiperspectivalist outlook ( anekantavada) to argue that the effect is already present in the cause qua its potentiality (sakti) but not qua its fully developed form. 57Garfield (1995, 107)

in the former. It seems as if the ‘happy compromise5 manages to avoid the difficulties Nagarjuna attributes to both self-causation and causation from other objects. A slightly different compromise solution which gets by without the potential existence of the effect in the cause can be found in the theories of the Naiyayikas and Vaisesikas.58 Here causation is understood to proceed by means of two internal causes and one external cause. The two internal causes are the inherent cause (samavayikcirana) and the noninherent cause (asamavayikarana), the external one is the instrumental cause (nimittakdrana). If we consider the way a marble statue is produced we can identify at least three different components: the marble out of which the statue is made, the various properties of the marble (such as colour, hardness, density etc.) and the actions of the sculptor. The marble constitutes the inherent cause, the material basis out of which the effect is made and can be compared to the Aristotelian causa materialis. The non-inherent causes (the marble's properties) do not cause the statue, but rather cause properties of the statue. Unlike the inherent cause, changes in the non-inherent cause do not change the kind of effect produced. Whether the marble is white or red, we still end up with a marble statue. Had the marble been clay, however, we would not have done so. These two internal causes are then combined with the instrumental cause (comparable to the causa efficiens) to bring forth the effect. The sculptor chipping off pieces from the block frees the statue locked within. On this account the marble statue is obviously not completely self-produced: the block of marble and its properties will just continue to sit there until the sculptor comes along. On the other hand it is not produced from objects which are completely different either, as any part of the statue is also part of the block of marble which is its cause. Dasgupta (1942, 1:320), King (1999, 208).

The reason why this account of causation is not particularly attractive for Nagarjuna’s opponent is that the ascription of being an inherent, noninherent or instrumental cause as a property something has by virtue of its svabhava is hardly satisfactory. Nothing can be a causa materialis in itself; it depends on the existence of a causa efficiens to turn it into an effect. The causa efficiens (the sculptor working on the marble) could not exist without the inherent or material cause (the block of marble). The non-inherent cause (the properties of the marble) needs the inherent cause as something to inhere in. It is apparent th at the three kinds of cause distinguished here both notionally and existentially depend on one another, as well as on the effect they jointly bring about. Their various causal properties are therefore nothing the respective objects have from their own side, independently of other objects. For this reason this account of causation is unable to defend the claim that the causality of the cause and the effect-ness of the effect are properties cause and effect have in virtue of their svabhava,

5.3.4 Absence of causation

The final possibility to discuss is causation which is neither self-causation nor causation by another object.59 This is generally regarded as the absence of causation altogether, given th at the first three possibilities are taken to exhaust the ways in which the relata of causation can be related. If these are ruled out, the only remaining option is that objects do not exist as causal relata since there is no causal relation.60 We might think th at there are some 59 MMK 12:9b. 60It is not quite clear who the original proponents of this view of fortuitous origination ( adhiccasamuppada, see Ronkin (2005, 198)) actually were. Murti (1955, 135, note 3, 167) identifies it as the view of the svabhavavadin, which is usually identified with that of

entities which may plausibly be taken to exist outside the causal nexus, such as mathematical objects and other abstract entities. Nagarjuna does not talk about the metaphysical status of mathematical entities, so any account of what he would have said about these is by necessity highly speculative.61 Moreover, he is interested in analyzing general accounts of how causation works. As we would hardly want to describe all phenomena as abstract there still remains the question of how to understand the working of causality for phenomena unlike numbers and so forth.62 Two distinct problems with the absence of causation are distinguished in the commentarial literature. First of all it is not clear how any facts about the world could be grasped (grhyate)> given that our main route of epistemic the Carvakas (see also Kalupahana (1975, 25)). Namai (1996, 561) agrees and mentions svabhavavada, i.e. the view that ‘phenomena are spontaneously diverse, there being no intervention of destiny or divine will* (Tillemans 2000, 58, note 210) as one of the positions held by the Carvakas. Schayer, on the other hand, denies the ascription of this view, since the Carvakas, he argues, only denied causality in the context of karma, but did not deny all causal determinations, as they specifically assert that things are determined by their intrinsic nature (svabhava) (1931, note 16, 20-21). Given the limited amount of information about the Carvaka system we presently possess, whether their view of causation should be regarded as one implying the complete absence of causation, causation by another object (as claimed on page 161) or even self-causation remains a moot point. 61 One might want to argue Nagarjuna’s assertions that there is no effect without a cause (na casty arihah kascid ahetukah kvacit. MMK 4:2) and nothing which is not dependently originated (apratUyasamutpanno dharmah kascin na vidyate. 24:19) rule out the existence of abstract objects altogether. Nevertheless one might read the first statement as saying that something not causally produced could not be referred to as an effect and remark concerning the second that dependent origination does not just mean causal origination. The existence of some mathematical structure might, for example, be logically entailed by certain concepts we have and therefore be dependent on these concepts for its existence. This, however, does not mean that it is causally produced. Ganeri (2001, 52-53).

access to the world is causal. Assuming that we are not causally connected to phenomena in the everyday world would make their epistemology as problematic as those of objects outside of the causal nexus.63 Secondly, and more importantly, a world without causation would be phenomenologically very much unlike the world we experience. That we experience the world as regular largely means that we experience it as causally ordered. Certain effects proceed from certain causes, but not from others: blood will flow if we strike a man, not if we strike a stone. A world with no causation is a world in which more or less anything can follow anything else.64 Without claiming th at the assumption of the absence of causation ‘falls to the ground through sheer inanity’65 it is sufficient to note that such a world is not the world we experience, and therefore the fourth alternative is no satisfactory explication of our concept of causality. 5.3.5 Identity, difference, and svabhava If we imagine a set of points on a plane connected to one another by lines we can imagine various connective possibilities for a particular point. It can be connected to itself via a looping line, or to another point, or both to another point and itself, or not connected to any point at all, or it can be connected not to itself, but to a proper part of itself. The conclusion Nagarjuna wants 63‘If the world was empty of a cause it could not be grasped, like the hue and scent of an [imaginary] lotus in the sky’, grhyate naiva ca jagad yadi hetusunyam syad yadvad eva gagana-utpala-varna-gandhau. PP 38:7-8. This is a quotation from MA 6:100. 64As pointed out by Buddhapalita as quoted in PP 38:10-11: ’It is not the case that things arise without cause because of the difficulty of everything arising always and everywhere.’ ahetuto notpadyante bhavdh sada ca sarvatas ca sarvasambhavaprasangat. For the Tibetan see Walleser (1913—1914b, 11-12). The same point is made in MA 6:99. 65Murti (1955, 135).

to draw from the preceding arguments is that if we apply this conceptual structure to causation, by letting the points stand for objects or events, and the lines for the causal relation, none of the possibilities could obtain. Given that the set of connective possibilities is exhaustive the reason for this must lie in the fact th at we make a basic assumption about points connected by lines which is not justified when speaking about events connected by causal relations. This assumption is that the points exist with svabhava, th at is th at they are distinct objects existing from their own side, independently of one another as well as of the cognizing subject which attem pts to connect them by lines. Nagarjuna argues, however, that cause and effect do not exist in this way. They are both notionally as well as existentially dependent on each other, as well as dependent on us as a cognitive subject which orders a chaotic mass of diverse experiences into causal fields and the effects they bring about. If cause and effect were identical it could obviously not be the case that the effect was dependent on the cause for its existence at a later time. Their mutual dependence would also not be compatible with cause and effect being two distinct objects in the same way as two points are distinct objects, where none brings the other into existence, nor would such dependence allow that an effect depended causally both on itself as well as on another objects. If the effect was part of the cause, the existential dependence of the effect on the cause at a later time would also not be possible. Finally, if no lines connected the points, the points would still be points, but cause and effect could not exist in the absence of causal relations. The conclusion of Nagarjuna’s examination of cause and effect in terms of identity, difference and parthood is therefore that such a conceptualization fails because it makes presuppositions about the existence of the objects thus

related (about their notional and existential independence and existence with svabhava) which are not applicable to cause and effect.66 Cause and effect have to be conceived of as both mutually dependent, as well as dependent on the cognizing subject, and therefore as empty of svabhava. 5.4 Tem poral relations betw een cause and effect Causes and effects are events which occur against the background of a particular causal field. One metal sphere moving (the causal event) brings about the moving of another one (the event which is the effect) by colliding with it 011 a plane surface, in the presence of the laws of gravity, in the absence of strong currents of air, magnetic fields, and so on (the background conditions), If we consider two arbitrary events, such as Peter reading the paper, and Paul drinking tea it is clear th at there are three ways in which they can be temporally related. They can be successive (Paul drinks his tea one hour after Peter reads the paper), they can overlap (Paul starts drinking his tea after Peter is halfway trough the paper), or they can be simultaneous (they start and finish at exactly the same times). Assuming that causes and effects are events like any others it should be straightforward to classify their temporal relations in the same way. Interestingly enough, this is more difficult than expected.

One fairly clear point is that if two events are related as cause and effect there is a restriction on their temporal ordering: the effect cannot precede the cause. After all, the cause is what is supposed to bring the effect about, something it obviously cannot do if it only arrives on the scene after the 66Napper (1989, 63-64). CHAPTER 5. CAUSATION 176 effect. Unless we assume that there is another event before the cause which also begins before the effect comes into existence, the effect would come about without any cause at all. In SS 6ab Nagarjuna stresses this point: ‘If the effect existed [before the cause then what about] the cause which has this effect? [Ex hypothesi it does not exist yet.] If that [cause] does not exist, [however, the effect] would be similar [to something with] no cause’.67 Nagarjuna lists nine different ways in which cause and effect can be temporally related: the cause can be either past, present, or future, and for each of these the effect produced can also be past, present, or future.68 Assuming th at an effect cannot precede its cause we can immediately rule out three possibilities. A past effect could not have been brought about by a cause that is either present or future, nor could a present effect be brought about by a future cause. The remaining six possibilities fall into two large groups, those in which cause and effect exist (wholly or partly) at different times, and those in which they exist at the same time. 5.4.1 Cause and effect as successive While the effect cannot begin before the cause, could we assume th at the effect succeeds the cause? The idea is that some causal event (my pressing the switch) begins, endures and then stops. After this the effect (the light going on) begins. Unless we postulate another event which stretches right 67 ’bras yod ’bras dang Idan p a ’i rgyu / de med na ni rgyu min mtshungs. As is evident from the interpolations in my translation the interpretation of this verse is not entirely clear, to say the least. This unclarity is further exacerbated by the existence of a number of variant readings. For some different translations see Tola and Dragonetti (1987, 25), Komito (1987, 107-108), Erb (1997, 75).

up to the time when the light goes on (thus raising the whole question of temporal relations anew, now for the effect and this event) the cause will have stopped to exist before the effect begins. Nagarjuna considers it to be problematic that something which has already ceased could be regarded as a causal condition.69 This problem arises specifically for the Abhidharma view of phenomena as minimally extended space-time points.70 Within this theory of moments (ksanavada) it is difficult to see how the existence of a phenomenon limited to an atomic temporal point could be compatible with its causal efficacy, as when the cause exists, the effect is still inexistent, and when the effect arises the cause will already have ceased to be.71 It thus appears that being brought about by a cause which is past is no better than having no cause at all.72 The difficulty Nagarjuna raises with his criticism is the problem presentist theories of time face when accounting for causal relation. For the presentist only the present, but not the past or the future qualify as real. He therefore faces a problem when trying to explain that the present is the way it is due to causal influences from the past, as he will have to postulate th at of the two relata of the causal relation only one (namely the present) is part of reality.73 69 [...] niruddhe pratyayas ca hah. MMK 1:9. See also 20:10a . Stcherbatsky (1923, 37-38), Kalupahana (1975, 67—73). 71 Since a two-place relation needs two arguments the Madhyainika will therefore argue that one of the two is a mere conceptual construction (or, as a Humean would put it, a reification of our expectations). Causation therefore does not connect things existing from their own side. See Siderits (2004, 408-409). 72‘If the cause ceased without having passed on its causal power to the effect, that effect which is born when the cause has ceased would be without cause.1 heturn phalasyadattva ca yadi hetur nirudhyate / hetau niruddhe jatam tat phalam dhetukam bhavet. MMK 20:6. The same difficulty was also pointed out by Sankara in his bhasya on Brahmasutra 2 . 2 . 2 0 . 73See be Poidevin (2003, 139).

Amongst the Abidharmikas this presentist difficulty is faced by the Theravada account according to which only the present moment, consisting of the origination, endurance, and dissolution of a phenomenon exist and has svabhava, while past and future moments are devoid of svabhava74 This theory will then have to explain how something which is past and therefore nonexistent can nevertheless assert its causal influence on the present. It should be noted, however, that this problem does not arise for the Sarvastivadins, whose metaphysical theories generally constitute the main target of Nagarjuna’s philosophical criticism. For the Sarvastivadin past, present, and future all exist, they all have svabhava.75 Such a theory obviously needs some way of accounting for the privileged status of the present. Different Sarvastivada thinkers have proposed different ways of doing this.76 The most popular account, due to Vasumitra,77 argues that the special nature of the present is due to the fact that only present phenomena manifest causal activity (karitra).78 A phenomenon is past if it has already discharged its activity, it is future if it has yet to discharge it. Despite being devoid of causal activity a past phenomenon nevertheless continues to exist and remains able to cause presently existing phenomena.79 The obvious way of avoiding this problem of a succession of cause and effect within a theory of the momentary existence of phenomena in which 74Ronkin (2005, 66, 119-120). 75Dhammajoti (2004, 35, 39). 76*Abhidharmamahavibhasasastra) Taisho 1545, 396a-b, Dhammajoti (2004, 82-83). See also Pradhan (1975, 296-297), Frauwallner (1995, 185-208). 77Dhammajoti (2004, 82-83). Samghabhadra’s account of the nature of karitra is given in his *Abhidharvianyayanusara, Taisho 1562, 631c-633b. A summary of the discussion can be found in Dhammajoti (2004, 89-92). 78Williams (1981, 241-242), von Rospatt (1995, 39), Siderits (2003,136, note m), Ronkin (2005, 110, 227). 79Cox (1995, 93), Dhammajoti (2004, 94).

only the present is regarded as real is to assume that cause and effect are temporally contiguous. This conception of a contiguous cause (samanatarapratyaija) was adopted by a variety of different schools of the Abhidharma, by the Sarvastivadins as well as by the Sautrantikas and the Theravadins.80 According to this theory whenever one phenomenon follows another one without a pause the latter may be regarded as the cause of the former. In fact Nagarjuna raises the above problem specifically with regard to the notion of immediately preceding conditions (anantara), which is one of the four types of conditions distinguished by the Abhidharmikas81 discussed by him in the second verse of this chapter. Immediately preceding conditions are ‘the countless intermediary phenomena that emerge upon the analysis of a causal chain’82 and which happen between a particular causal event and its effect. The main advantage of the theory of temporal contiguity of cause and effect is th at it eliminates the existence of a temporal gap between the two, a gap during which the cause does no longer exist, because it has just ceased, and during which the effect does not yet exist because it is just about to begin. Such a gap would make it hard to explain how any causal efficacy can be passed on from cause to effect, as they are divided by the insulation of a causal vacuum, a gap in which no causation takes place. If the temporal moments of cause and effect are regarded as ‘directly touching’, however, this difficulty is avoided, even if only the present moment is regarded as existent. The past moment does not have to exist in the present in order to be causally efficacious; all th at is required is that it passes 011 its causal power when the moment of the cause and the directly successive moment of the effect meet. 80Kalupahana (1975, 72-73, 166-167). 81 Dhammajoti (2004, 131-132). Gavfield (1995, 109).

Whether this account is able to provide us with a satisfactory theory of causation, however, in turn depends on the conception of time the view of causal contiguity is based on. If we combine the notion of the contiguity of cause and effect with a view of time which does not view moments as atomic but as divisible in turn (as the Theravadins did) 83 this does not seem to help us much in addressing the problem of the temporal relation between cause and effect. For in this case no m atter how close we regard the last moment of the cause and the first, moment of the effect to be in time, there will always be a moment between them which is different from either, thereby perpetuating the problem of the temporal gap. If, like the Sarvastivadins84 and Sautrantikas85 we regard time as discrete and composed of atomic, indivisible moments,86 however, the situation we consider looks like the one depicted in figure 1. Here there is obviously 83von Rospatt (1995, 60), Ronkin (2005, 62-63). 84Ronkin (2005, 61-62), Dhammajoti (2004, 153). 85von Rospatt (1995, 60-63). 86There is an obvious tension between an atomic conception of time and the assumption that each moment is characterized by the three distinct characteristics of conditioned objects (samskrtalaksana) of origination, endurance, and dissolution, which can hardly be regarded as simultaneous. For a discussion of how the Sarvastivadins dealt with this issue see von Rospatt (1995, 49-59), for the Sautrantika account see 60-63.

no temporal gap between the end of the cause and the beginning of the effect, as there is no further moment of time between these two moments. Nevertheless it does not seem to be the case that this is really a view of the temporal relation between cause and effect the opponent of Nagarjuna would want to defend. Obviously, the causal event does not become a cause until the first moment of the effect is present, for if the effect had never existed, the cause would not have been a cause (and, if we adopt the Nagarjunian line of the existential dependence of the cause on the effect, in this case the cause would not have existed at all). When the first moment of the effect is present, however, the cause is already past, so that during no moment when the causal event was present it was actually a cause — the term ‘cause5 is only applied to it post festum. The cause or, more precisely, the causal field, only comes into existence as such (as opposed to a random collection of objects) after the effect has been produced. Only after the sprout has been produced I can identify all the various elements which led to its production as its cause. But this then implies th at being a cause cannot be an essential property of the underlying event, since the event can have the property at one time (after the effect has been produced) and lack it at another time (before the effect was produced). The above conception will therefore not support the idea that the cause exists as a cause from its own side, and that being a cause is part of its svabhava, a position which Nagarjuna’s opponent will want to defend. For the Sarvastivadin a phenomenon’s causal efficacy, like its svabhava, does not undergo any change through time, it is only the phenomenon’s activity (karitra) which becomes existent in the present moment and then ceases as the phenomenon vanishes into the past. But, as Nagarjuna argues, we do not call anything a cause unless it actually produces an effect; its being a cause depends on something’s being the effect it produces. In the presence of such

a dependence, however, being a cause cannot be part of the phenomenon’s svabhava, since this very notion demands its being able to exist as a cause from its own side, independent of any other object

5.4.2 Cause and effect as overlapping

Suppose there are two events related in the way depicted in figure 2. The causal event (my pressing the lightswitch) begins at time to and continues until t2. The effect, (the light going on) begins at t\ (while the cause still continues) and goes on until t3, after the cause has ceased. On the face of it this seems to be a reasonable understanding of the temporal relation of the cause and effect involved in this case. I press the switch (£0), after some time the light goes on (ti), then I stop pressing (£2); the light continues to be on until some future time £3 . There are two main difficulties with this picture. The first, and most general problem is that seeing cause and effect as overlapping - in contrast with all other accounts of their temporal relation - implies that cause and effect are temporally extended. But given that temporally extended things have temporal parts, and that partite things cannot be ultimately real this seems to defeat the point of the argument. Nagarjuna’s opponent is interCHAPTER

ested in establishing causal relations between ultimate existents, i.e. between objects which are (unlike temporally extended objects) not conceptually constructed. But conceiving of cause and effect as overlapping (whether we think that there is a clear cut-off line between cause and effect as indicated in figure 2 , or whether we take this to be a gradual process) entails th at we do just that, since we have to conceive of them as spread out across tim e.87 The second problem is that according to the situation depicted in figure 2 the cause would be taken to cease at t 2, ke. when the lightswitch-pressing event comes to an end. This, however, means that we have to regard the period between ti (when the light went on) and t2 (when I stopped pressing) as part of the cause. But my pressing after the light went on was of no consequence for the effect in any way, so it appears to be rather strange that we take a causally inert part of the event to be part of the cause. Nagarjuna argues th at in this case what we take to be the cause would have a double nature (atmadvaya): its causal power which is employed in bringing about the effect and its causal inertness, after the causing has been done.88 While having two incompatible properties at different times is of course not a problem in itself (an apple can be green now and red later), assuming that part of the nature of a cause is causally inert seems distinctly odd — if a cause has any nature at all it seems to consist in being able to bring about an effect. It is therefore undesirable for anyone who wants to argue th at being a cause is part of some object’s svabhava to assume that this object has a causally inert temporal part. 87See Siderits (2004, 406-408). 88‘If the cause ceased to exist after having passed on the causal power of [bringing about] the fruit, the cause would have a double nature: the given [causal power] and the ceased [nature after having passed on the causal power].’ heturn phalasya dattva, ca yadi hetur nirudhyate / yad dattarn yan niruddham ca hetor atmadvayam bhavet. MMK 20:5.

If we now try to avoid this problem by ‘cutting off’ the causally inert part of the cause ranging from ti to t2 and only regard the event ranging from tQ to ti as the cause we end up with a scenario in which cause and effect are temporally contiguous, the difficulties of which have been discussed above. 5.4.3 Cause and effect as simultaneous The third possibility to discuss is that of cause and effect being simultaneous, i.e. their coming into existence and ceasing at the same moment. The concept of simultaneous causation (sahabhuhetu) is of central importance in the Sarvastivada theory of causality. 89 The principal Sarvastivada example of simultaneous causation is a thought (citta) and its specific concomitants (caitasika) 90 which mutually depend on one another ‘like the poles of a tripod’. 91 A thought cannot arise earlier than its concomitant factors, nor can such factors be earlier than the specific thought they accompany. The concept of simultaneous causation is essential of for the Sarvastivada theory of the existence of past and future, as well as present objects. The main argument for this thesis of universal existence (sarvastitva) is th at since consciousness needs an existent object, and since there is consciousness of past and present phenomena, these phenomena must be existent objects. But of course this only establishes the existence of past, present, and future objects Dhammajoti (2004, 116-117) gives a selection of passages from the Sarvastivada Abhidharma literature dealing with simultaneous causation. See also Tanaka (1985), Burton (1999, 193), Dhammajoti (2003), Ronldn (2005, 217). The idea of a sahabhuhetu also becomes important in Yogacara literature where the dlayavijhana and bija are regarded as standing in a simultaneous causal relationship (Dhammajoti 2004, 121—123). "Dham m ajoti (2004, 162). 91Hopkins (1983, 339). Another example used is the simultaneous existence of a lamp and its light (Dhammajoti 2004, 120).

if the object of consciousness (the cause) and the consciousness of th at object (the effect) exist simultaneously. If a past object was able to cause a consciousness of it which is present, the past objects could be non-existent, even though the present consciousness of such objects does exist.92 Nagarjuna is primarily interested in the possibility of cause and effect coming into existence at the same time.93 An immediate problem with this idea is th at the cause is generally taken to be what brings the previously nonexistent effect about, and so as something which exists while the effect does not yet exist. We distinguish the effect from the causal field by observing that the causal field (the wires, the bulb, pressing the switch) is there first, without the effect (the light going on) which appears subsequently. A second difficulty is evident from a problem Nagarjuna discusses in a slightly different context.94 Speaking of cause and effect as simultaneous, we intend this to mean th at there are two distinct events beginning at the same time as one another, rather than one event referred to by two different names.95 Being distinct, we should be able to imagine one without the other; it should be logically possible that one of the two distinct entities exists while the other does not. At least this is the sense of distinctness the opponent of Nagarjuna will have in mind, who will attem pt to base the distinctness on the svabhava of the respective events. Considering the interdependence of cause and effect discussed above, however, such distinctness does not obtain, so that a claim for simultaneity cannot be made. The reader of Candrakirti’s commentary on the above passages will notice that he attempts to elucidates Nagarjuna’s assertion that cause and effect Dhammajoti (2004, 38, 125). 93MMK 20:7. 94MMI< 6:3-9. Garfield (1995, 156).

cannot exist together observing that simultaneously existing objects like the left and right hand,96 or the left and right horn of an ox97 are never seen to stand in a causal relationship. This seems to be incorrect, as there are in fact a multitude of prima facie examples where cause and effect come into being at the same time: the cause of the effect of the left-hand-side of a pair of scales going up (namely its right-hand-side going down) begins at the same moment in time as the effect,98 as does the cause of the cart’s moving (namely the motion of the horse). If we place a ball of lead on a soft cushion, the cause (putting down the ball) and the effect (the indentation in the cushion) equally arise at the same time.99 Nevertheless, we have to note that these are hardly the examples the critic of Nagarjuna who wants to establish cause and effect as independent, self-sufficient entities could be looking for. The motion of the horse and the motion of the cart are simultaneous, but clearly not distinct in the strong sense defined above: it is certainly physically (and presumably also logically) impossible that the cart moves forward, and the horse stays where it is. This example is therefore not able to establish the simultaneity of two independently existing events, 100 96PP 395: 9-10. 97PP 139:14, see also 224:4. 98This example is discussed in MA 6:18-19. See also Shaw (2002, 230), Siderits (2004, 408). "For the origin of the last example see Kant (1993, A 203, B 248); some commentary is in Rosenberg (1998) and Le Poidevin (1988). See also Bugault (2001, 252). 100 A different interesting argument against the simultaneity of cause and effect, attempting to show that it would undermine the existence of any succession of causes is given in Hume’s Treatise of Human Nature (Hume 1896, I, III, II: 76). See also Munsat (1971).) Suppose a cause and its effect existed at the same time, t\. Suppose further that there was another cause of the effect which existed at t_i, a short instance of time before this. Given that the simultaneous cause produces the effect immediately, but the earlier cause only after some delay (i.e. the time which passed between t_ i and t\) we would only want to

5.4.4 Cause and effect as one event A final possibility Nagarjuna raises in MMK 20:20 is the suggestion th at when speaking of causes and effects we are not dealing with two events at all but in fact only with one single event which in the course of its history transforms itself from a causal event into an effect-event. Now the question to consider is whether the cause retains its causal nature after the transformation. If it does not, it will not be the same any longer since for Nagarjuna’s opponent being cause and effect are not just intrinsic features of events, but also essential ones, i.e. features an object cannot lose without ceasing to be that very object. As Candraklrti points out, for the defender of the ‘transformation’ theory the difference between cause and effect cannot just be one in terms of description (samjhamatrabheda) , but must be one in terms of essence (dravyabheda) . 101 If the event retains its causal nature, however, then, after the cause has ceased an event with the very same svabhava will arise once more. This is problematic for the same reasons th at the presence of the effect in the causal field is problematic, as was discussed in 2 0 :1 : if the cause already existed, there would be no need to produce it once more. A further problem arises from the fact th at it is hard to see how this kind of self-causation regard the simultaneous cause as the true cause (the earlier cause is at best an earlier part of the effect’s causal history), Therefore, given the possibility of a simultaneous cause, no earlier event will be a plausible candidate for being a cause. Given that there are always events simultaneous with a another event, we will always prefer to regard these as causes, rather than some preceding event. But if we thus assume that for any effect its cause is simultaneous we are faced with the problem that simultaneity is transitive. If cause c is simultaneous with effect ei, and e\ is in turn simultaneous with whatever effect it may have (call this e2 ) then c and e2 are also simultaneous. Therefore all causes and effects would happen at the same time, and thereby there would be no such thing as a causal succession in time. 101PP 397:7.

should ever stop. Since there were no factors present apart from the causal event which triggered its transformation into the effect (which is the verysame event as the cause and is both cause and effect essentially) there is no absence of factors which could stop such a transformation. The cause would be self-perpetuating and exist for ever. 5.4.5 Temporal relations and svabhava Considering any two events it is evident that they either appear in succession (being contiguous or separated by a temporal gap), that they are temporally overlapping, that they happen at the same time, or that they are successive stages of one single event. As we have seen, Nagarjuna denies that any of these possible temporal relations holds of two events which are related to one another as cause and effect. The reason for this is the same we observed in the case of the relation between cause and effect in terms of identity, difference, and parthood. The above set of temporal relations only presents us with an exhaustive classification of all possibilities if we are dealing with distinct and mutually independent events which exist from their own side independent of a cognizing subject, th at is with events which exist by their own svabhava, Cause and effect, however, do not exist in such a way, as one brings about the other. Given this dependence relation cause and effect cannot be separated by a temporal gap, as in this case the effect would have to depend on a non-existent object, as the cause does not exist any more. Assuming them to be overlapping entails that the cause has a causally active and a causally inert part, which conflicts with the assumption th at being a cause is part of its own nature. If cause and effect come about at the same time, it is hard to see how the effect could rely existentially on the previously existing cause, while taking them to be stages of one event again means that

the cause cannot be a cause as part of its own nature, as this is a property the cause loses when turning into the effect. The consideration of the temporal relations between cause and effect therefore demonstrates once more the inapplicability of conceptual schemes suitable for discussing mutually independent and observer-independent objects to causes and effects. Being empty of svabhava they cannot be conceived of using the conceptual resources intended for discussing phenomena which exist both independently of one another and independently of a cognizing subject. Having investigated a variety of commonsensical theories of the identity relations between cause and effect, as well as theories of their temporal relations Nagarjuna concludes that the commonsensical view of objects underlying these theories is unsatisfactory, since it conflicts with each of the ways in which the relation between cause and effect can possibly be conceived of. According to the view Nagarjuna wants to establish instead of the commonsensical one cause and effect do not exist independently of one another: they require each other both notionally and existentially. There is no point in using the concept of a cause without that of an effect, or vice versa. More importantly, while an effect cannot exist without being brought about by a cause (or, more precisely, by a causal field) the effect brings about the causal field as well. This is because without the effect there would be no indication of which phenomena are to be included in some causal field, and which are to be left out. The causal field is not something found ready-made out there in the world, waiting to be discovered by the inquiring mind. It is a cognitive artefact brought about whenever the mind organizes its experiences. To this extent the causal field does not just depend on the effect which provided the justification for certain objects rather than others to be included in it, but

also on the mind which does the including. Cause and effect are therefore not just mutually interdependent, but also mind-dependent. This is what Nagarjuna means by saying th at causes and effects do not exist from their own side, that is th at they are empty of svabhava. A general worry we might want to raise about Nagarjuna’s attem pt to establish emptiness on the basis of causation is the following. It is clear that showing th at some object causally depends on another one shows th at it is empty in some way. For example, if we define an object to be non-empty iff it is mereologically, causally and conceptually independent, demonstrating that some object is causally produced is obviously enough to show that it is empty. But this might not be good enough.102 After all it is the notion of emptiness in terms of conceptual dependence or imputation which is generally regarded as the most subtle understanding of emptiness. But it is clear that establishing some object is causally produced does not entail th at this object is also conceptually constructed.

Siderits argues that this difficulty can be solved by attributing to the Madhyamika the principle that if some object essentially involves a property which is conceptually constructed, the object is conceptually constructed too. 104 If we replace ‘conceptually constructed’ by ‘fictional’ the tru th of the principle is immediately evident. Consider a violin performance by Sherlock Holmes as a simple example. This object essentially involves the relation ‘being performed by Sherlock Holmes’, which is a fictional property. For this reason the performance, which incorporates this property as an essential part 102It is useful to remember in this context that for the Abhidharmikas conditioned objects (samskrta dharma) were both regarded as having svabhava and as being dependently originated. 103See Burton (1999, 115). 104Siderits (2004, 411).

is fictional too. In the same way, the Madhyamika will argue, each material object essentially involves reference to causality, since all these objects are causally produced. But if it is now demonstrated, as Nagarjuna set out to do above, that the causal relation does not exist from is own side, is conceptually constructed and thus empty it follows that each material object must be conceptually constructed and therefore empty in the most subtle sense as well. In this way the discussion of causality is not just able to establish the comparatively crude emptiness in terms of causal interdependence, but also the more subtle one in terms of conceptual construction.

5.5 Analysis of time

The analysis of causation is intricately bound up with that of time as it refers to temporal relations at various crucial places in the argument. Nagarjuna discussed time in the shortest chapter of the MMK which consists just of six verses. He notes that first of all th at the past, present, and future existentially depend on one another. 105 Not only could there be no present if there was no past or future, there could also be no present if it was not related to past and future in exactly the way it is, that is sandwiched between them. Time is a fundamentally relational phenomenon with an intrinsic ordering. Assuming such an interdependence between the three times Nagarjuna claims that if the present and the future depended on the past then the present and future would exist in the past. 106 On the face of it this seems to be the now familiar 105<In turn the past is not found established independent of the two [present and future].5 anapeksya punah siddhir ndtitam vidyate tayoh. MMK 19:3a. 106 pratyutpanno ’nagatas ca yady atttam apeksya hi / pratyutpanno ’nagatas ca kale ’tile bhavisyatah. MMK 19:1.

point that when a dependence relation holds between two entities, both of them must exist. If, for example, we claim a pot to exist in dependence on its parts both the pot and its part must exist. But on the (not implausible) ascription of a presentist view of time to Nagarjuna according to which only the present is real, neither the present nor the future can depend on the past since the past does not exist any more. We would thus have a case of a dependence relation with only one term, as one of the relata failed to exist. We can identify a different problem here if we assume that dependence relations only exist between objects located in time. 107 For then past, present, and future would have to be located in another time, which would in turn give rise to temporal relations located in a yet another time and so one. In the end we would need an infinitely extended hierarchy of times to make sense of the dependence between the three times. Given these difficulties of analysing time in terms of dependently related phenomena Nagarjuna concludes that none of the three times can be found.108 The difficulty with this reading, however, is that Nagarjuna immediately goes on to assert that ‘by precisely the same method’ concepts such as highest, lowest, and middle should be elucidated.109 The problem is that, pace Garfield, 110 the above argument cannot be generalized to cover spatial relations as well. The higher, middle, and lower part of a building can perfectly well be described as depending on one another. The second floor notionally (and architecturally) depends on the first floor, since it is only in relation to the first floor th at in can be called ‘second floor’, and it is also the first 107As done by Garfield (1995, 255). A similar point, is made in rGyal tshab’s commentary on RA 3:56 (Hopkins 1998, 141). 108pratyutpanno ’nagatas ca tasmdt kdlo na vidyate. MMK 19:3b. Verse 4 makes clear that this assertion is also meant to apply to the past. 109 etenaiva [...] kramena [...] uttama-adhama-madhyadin [...] laksayet. MMK 19:4. 110Garfield (1995, 256).

floor which keeps the second floor up. But as all these parts of a building exist simultaneously all the relata of the dependence relation do. Therefore neither the argument from the non-existent relatum nor indeed the regress argument111 get off the ground in the spatial case. We could assume that what Nagarjuna means here is just the general fact that distinctions like high, middle, and low are essentially relational, 112 so th at no object could for example be regarded as essentially ‘middle’ since this property depended on it spatial relation to objects other than itself. But this seems at odds with Nagarjuna’s earlier attem pt to find fault with the very idea of establishing the three times as dependent entities. Alternatively we could understand the argument as claiming th at if any present or future entity depended on a past entity, this entity would have to have existed in the past. W hat the argument rejects on this interpretation is that each object has a ‘hard core’ persisting through the three times. That today’s cup is the same at yesterday’s cup would be taken to mean th at there is one thing (perhaps the substance of the cup) which was here yesterday and is here today, and is characterized by different accidental properties at the different times. This notion of a substantial core has been criticized by Nagarjuna in a variety of ways in earlier parts of the MMK. On this interpretation we could then read the statement in MMK 19:4 as claiming that in the same way in which it makes no sense to speak of m It makes much more sense to interpret MMK 19:5a as noting the difficulty of a temporal regress. Here Nagarjuna argues that one could conceive of neither a static nor of a nonstatic time. (ndstito grhyate kalah sthitah kdlo na vidyate) The reason for this is that in order to conceptualize time as static or non-static we have to locate it in another time relative to which it changes or does not change. 112This is the interpretation given by Tsong kha pa (2006, 397). See also Weber-Brosamer and Back (1997, 70-71).

a persisting substance across time there is no spatially persisting substance either. In a house which has an upper, middle, and lower part there is not one piece of m atter which runs through the entire house and is characterized by the attributes upper, middle, and lower. It is rather th at different parts of the house are designated in this way in relation to one another, and that they are regarded as parts of the same house by their spatial contiguity, not because of some sort of material backbone running through all of them. Similarly the different temporal parts of the cup can be conceptualized as belonging to one individual by their temporal closeness and the sharing of a significant number of properties without the need for a persistent cross-temporal unifier. In fact this interpretation of time which does not refer to the persistence of an underlying substance is the only way for the Madhyamika to affirm the existence of time. 113 As Nagarjuna described in detail the reasons for the rejection of such a substance he would then also be forced to deny the existence of time. The Madhyamika therefore has to explain how we can account for an object changing and persisting through time without having to assume that there is some unchanging aspect of the object which underlies all change. Nagarjuna claims that this can indeed be done. Understanding how this can be the case becomes particularly important in the context of the Buddhist conception of the self when the temporal continuity of persons has to be explained without reference to the concept of a persisting subjective core (atman). 113‘If time existed dependent on an object from where should it come without this object? As there is not any object, from where does time come?’ bhavarn pratUya kalas cet halo bhavad rte kutah / na ca kascana bhavo ’sti kutah kdlo bhavisyati MMK 19:6. Note that ‘object’ (bhava) in this context is to be understood as ‘object with svabhava\

Chapter 6 Motion

At a prominent place at the beginning of the MMK Nagarjuna analyses the concept of motion. This discussion, which takes up the whole of the second chapter is primarily concerned with the investigation of two questions: firstly ‘Where is the locus of motion?’, i.e. where is motion taking place,1 and secondly ‘W hat is the object of motionV, i.e. what is it that has the property of moving? 2 Imagine a car driving down a road, turning right at an intersection, then driving on. Where is it moving? We obviously do not want to locate motion anywhere where the car has just been, say twenty seconds ago, as this is not where motion is presently happening. Nor is a place where it has not been at all (say, turning left at the intersection) any better — not only is 110 motion presently taking place there, it has also not taken place there in the past. Neither the places the car has driven through in the past, nor those it has not are plausible candidates for locating its motion in the present moment.3 XMMK 2:1. 2MMK 2:8. 3‘As far as the place moved over does not move the place not moved over does not move either.’ gatam na gamyate tavad agatarn naiva gamyate. MMK 2:1a. 195

The car is obviously moving in the space which it presently traverses, which constantly changes as what is present changes: for each moment the car is presently moving where it is moving when that moment is the present moment. The locus of motion must be the space which is presently being traversed.4 Secondly, what moves? Not the car which is parked nearby and is stationary (agantr) but the one being presently driven. It is only the mover that moves.5 Now it appears that one of the main aims of Nagarjuna in this chapter is to analyse both these commonsensical answers, that present motion happens in the presently traversed space, and that it is the mover which moves, in order to demonstrate that they are more problematic than they might initially seem. While this impression is largely correct, as we shall see m atters are in fact a bit more complicated. The second chapter of the MMK has attracted considerable attention in the contemporary commentarial literature, not least amongst scholars interested in a certain kind of cross-cultural comparison, setting out to compare Nagarjuna’s arguments with Zeno’s paradoxes. The ways in which the chapter has been understood are very diverse and it does not seem as if an interpretative consensus has yet been reached. This is hardly surprising, given th at this chapter in particular brings out the difficulty of doing two things at the same time: understanding the internal structure of Nagarjuna’s arguments and placing them in the argumentative context of his philosophical enterprise. After all these arguments were not intended as particularized dialectical curiosities but occupy a central point in the structure of the MMK. 4gamyamane gatis. MMK 2:2b. bganta gacchati. MMK 2:10.

6.1 A rgum ents concerning m otion The arguments presented in the second chapter of the MMK can be best understood if we divide its 25 verses into three groups. The first group (verses 1-6, 8-11, 15-16, 22-25) investigates the locus and the object of motion by two arguments which I call the property-absence argument and the propertyreduplication argument.6 As I will argue later on these arguments are not specifically about motion. Nagarjuna rather uses the example of motion to give an example of a form of argument which can be applied to a variety of subject-matters and is indeed referred to again and again in different contexts within the MMK. The second group of verses (12-14, 17) discusses the interdependence of the concepts ‘beginning of motion’ and ‘end of motion’ and the division of the triple division of the space and time where motion takes place. This is a 6There is also a further argument supposed to show that there can be no motion in the space presently traversed. This is the so-called ‘foot argument’ given by Candraklrti in his commentary on verse 1. Candraklrti presents this as a refutation of the opponent’s claim (supposedly implicit in verse 1} that motion takes place in the space presently traversed. This is slightly peculiar, as the opponent will explicitly make this claim in the following verse. The argument attempts to show that the foot cannot be at the place presently traversed, since the foot is made up of atoms. But a place behind some atom at the front of the foot is already moved over, while some atom in front of some atom at the back is not yet moved over. There is some debate about how to interpret this argument (see for example Siderits and O’Brien (1976, 289) and Galloway (1987, 81-85) for diverging accounts). Fortunately we do not have to settle this issue here, as this specific argument belongs more properly to the thought of Candraklrti than to that of Nagarjuna. I share Bhattacharya’s concern (1985, 8) about the mathematical gloss CandrakTrti’s commentary imposes on the reading of the first four verses of chapter two (see also Mabbett (1984, 409-410)). For more discussion of the ‘mathematical’ and the ‘conceptual’ interpretation see section 6.1.1.

division of space into a space not yet traversed, a space presently traversed and a space to be traversed, and a division of time into the times of past, present and future motion. Nagarjuna’s aim in these verses is to establish th at the concepts of beginning and end of motion and the triple division cannot exist independently of one another. The third group (7, 18-21) considers the relation between mover and motion and sets out to establish that these two mutually depend on one another.

6.1.1 The property-absence argument

In the property-absence argument Nagarjuna seems to assert that some individual can only be said to have a property if it is at least conceivable that it lacks that property. An apple can have the property ‘red’ because it is conceivable that it lacks redness and has some other property instead, such as being green. However, how suitable is it to attribute motion to the space presently traversed, as far as attributing non-motion to it is not suitable? For whom motion is attributed to the space presently traversed, there should be such a space without motion — but ‘presently traversed space1 means ‘movement takes place there’ .7 7 gamyamanasya gamanam katham namopapatsyate / gamyamanam vigamanarn yada naivopapadyate / gamyamanasya gamanam yasya tasya prasajyate / rte gater gamyamanam gamyamanam hi gamyate. MMK 2:3-4. The reading vigamanarn (nonmotion) in verse 3 follows May (1959, 55, note 19). PP 94:7 has dvigamanam (double motion), Inada (1970, 44) has hy agamanam. See also de Jong (1978, 36). For some discussion of the varying philosophical interpretations suggested by these different readings see Siderits and O’Brien (1976, 290-291).

How suitable is it to say ‘a mover moves*, as far as a mover without motion is certainly not suitable? For the one who holds the position that a mover moves, and who is looking for the motion of the mover there is a mover without motion.8 Nagarjuna’s claim seems to imply that there is a problem with analytic statements9 and the ascription of essential properties. In an analytic statement the property attributed could not be conceived of as failing to be instantiated by the object under consideration. Because there are no married bachelors, it would be ‘unsuitable’ to say of a bachelor th at he is unmarried. Essential properties are such that an object cannot lose them without ceasing to be that very object, it is therefore not feasible to speak of the object in question as lacking an essential property. Because a block of ice would stop being ice if warmed up to more than 30°C, by the above principle it would also be problematic to assert of a block of ice that its tem perature is not above 30°C. 8ganta tavad gacchattti katham evopapatsyate / gamanena vind ganta yada naivopapadyate / pakso ganta gacchatUi yasya tasya prasajyate / gamanena vind ganta gantur gamanam icchatah. MMK 2:9-10. In order to understand the structure of the argument it is important to realize that 2:9-10 spell out the assertion made in 2:8, namely that neither the mover nor the non-mover moves. In fact Nagarjuna only considers the first alternative, he does not specify why the non-mover does not move. But we can infer what he would say from 2:16, which elucidates 2:15, being just the mirror-image of 2:8. 2:15 claims that neither the mover nor the non-mover is not moving. Here Nagarjuna only considers the alternative of the mover not moving, a presupposition which is contradictory and therefore to be discarded. 9Both the statement ‘present motion happens in the presently traversed space’ and ‘the mover moves’ can plausibly be regarded as analytic as their negations can be reduced to logical contradictions. The problem Nagarjuna discusses here, however, is not restricted to analytic statements.

We might wonder at this point what precisely is ‘unsuitable’ about saying th at a bachelor is unmarried. After all it is not only true but necessarily true. As long as expressions ‘bachelor’ and ‘married’ mean what they mean this assertion could not possibly be false. The difficulty Nagarjuna has in mind here is that of analysing the referents of statements like the above in terms of an ontology of mutually independent objects. If we consider the referent of a statement like ‘the apple is red’ it makes sense to regard the constituents of the state of affairs this refers to (namely the individual apple and the property red) as distinct objects. After all there are apples which are not red, and red things which are not apples. We rely here on the Humean principle that for things to be distinct we must be able to conceive of them independently of one another.

For statements like ‘the mover moves’ or ‘bachelors are unmarried’, however, this does not hold: there are no movers which are stationary, nor moving objects which are at rest; there are no married bachelors, nor unmarried non-bachelors. In each case the individual depends11 on the property it in10‘ We have observed, that whatever objects are different are distinguishable, and that whatever objects are distinguishable are separable by the thought and imagination. And we may here add, that these propositions are equally true in the inverse, and that whatever objects are separable are also distinguishable, and that whatever objects are distinguishable, are also different. [...] [A] 11 ideas, which are different, are separable. For it follows from thence, that if the figure be different from the body, their ideas must be separable as well as distinguishable: if they be not different, their ideas can neither be separable nor distinguishable.’ (Hume 1896, I, I, VII: 18, 24-25). 11 The dependence of the individual on the property it instantiates may be notional or existential, depending on whether the individual has the property in question essentially. As nobody is essentially a bachelor a bachelor who marries would still continue to exist, but would no longer be described as a bachelor. But since ice is essentially frozen, when we heat up a block of ice to more than 30°C it is not just that we would not describe the result as a block of ice any more, the block of ice will have ceased to exist.

stantiates. We therefore cannot analyse the referent of propositions like ‘the mover moves’ in the same way as that of ‘the apple is red’. Such an analysis would assume the existence of two distinct entities, a property and an individual, existing independently of one another (and therefore, as Nagarjuna’s opponent would put it, each existing by its own svabhava) which come together in a state of affairs where one instantiates the other. T he ‘m athem atical’ interpretation

We should note th at there is a different interpretation of the two passages under discussion which, unlike the interpretation just presented, regards them as an argument specifically concerned with motion, rather than more generally with the instantiation of properties. 12 This interpretation sees Nagarjuna as concerned with refuting a particular conception of space and time by demonstrating that motion would not possible under such presuppositions. These presuppositions are that space is infinitely divisible, but that time is not, and that it consists of a succession of temporal atoms of minimal duration. The argument then runs as follows. Let there be a moving object and consider the portion of space traversed by this object during one temporal atom. Even if this portion is very small, since space is infinitely divisible we can break it up into further portions of space. Now take some point within this portion of space. The moving object cannot have passed it during the course of its motion, since the time it would take to reach it would be a fraction of the temporal atom, and since atoms are indivisible no duration that short exists. So motion cannot happen in the space presently traversed, since all th at happens is that the moving object is at the beginning of the portion of space before the temporal atom, and is at its end afterwards, without 12Siderits and O’Brien (1976, 291).

having moved through any of the points in between. We are thus dealing not with motion, but rather with a succession of rests. Because of this we (unsuitably) have to attribute non-motion to the space presently traversed. The same interpretation can be given to verses 9 and 10 of this chapter: a mover moving in an infinitely divisible space during a temporal atom would be a mover without motion (gamanena vina ganta), because he does not traverse any of the infinitely many spatial points between the beginning and the end of the space traversed. Such a mover would be a mover at rest. This reading allows us to give a consistent interpretation of the individual verses discussed, of course always presupposing that Nagarjuna really made these particular assumptions about the divisibility of space and time. Nevertheless I think th at the ‘conceptual5 interpretation given above accords better with an understanding of the place of chapter two in the whole of the MMK than the ‘mathematical’ interpretation proposed by Siderits and O’Brien. 13 My main reason for thinking this is that the various references to the arguments in chapter two throughout the MMK14 are very hard to make sense of on the mathematical interpretation. For example Nagarjuna remarks in the discussion of fire and fuel in chapter 1 0 that the remaining points concerning these have been discussed in the treatment of the presently moving object, the moved and the non-moved. 15 If we follow Candraklrti’s interpretation th at this means we can substitute ‘what has burnt’ (dagdha) for ‘what has moved’ (gata), ‘what has not burnt’ (adagdha) for ‘what has 13These terms are theirs (1976, 289). This criticism is also shared by Mabbett (1984, 412). 14In the dedication, as well as in 3:3, 7:14, 10:13, and 16:7. 15‘In the place of ’fire’ all the other cases can be expressed by ‘what is presently moving’, ‘what has moved’, ‘what has not moved’.’ atrendhane sesamuktam gamyamdna-gataagataih. MMK 10:13b. Ye (2006b, 163-164) reads athendanam instead of atrendhane.

not moved’ (agata), and ‘what is presently burning’ (dahyamdna) for ‘what is presently moving’ (gamyamana) throughout the second chapter16 it becomes evident that this makes much more sense if we adopt the conceptual rather than the mathematical interpretation. For example, by substituting in MMK 2 : 3 we get something like the following: How suitable is it to attribute burning to the presently burning fire, as far as attributing non-burning to it is not suitable? For whom burning is attributed to the presently burning fire, there should be such a fire without burning — but ‘burning fire’ means ‘burning takes place there’.

If we interpret this statement according to the conceptual interpretation we can see that Nagarjuna makes the point that the fire and its property (i.e. burning) cannot be conceived of as mutually independent objects, like an apple and its redness, which come together in a state of affairs. For whereas it is possible for the apple and the property of redness to exist one without the other (if the apple is green, and redness is instantiated elsewhere) there cannot be an individual which is a fire and also not burning, nor can the property of burning be instantiated by something which is not a fire.17 The widespread use of the discussion of the mover, the non-mover, and the presently moving object throughout the MMK suggests, I think, (and this will become more evident in the following discussion) that this section of the second chapter was not meant to be a specific investigation of the problem of motion and the various structural properties of time and space. Rather it uses the discussion of motion as an example to illustrate an argumentative 16PP 211:8-12. 17See also Cheng (1980, 233-234).

template which can be used in a variety of different contexts. 18 I would want to argue th at main issue addressed here is th at of instantiation. The point Nagarjuna wants to establish by investigating the notion of a mover and its motion in MMK 2-3 and 9-10 is th at the standard analysis of instantiation in terms of independently existent individuals and properties is not universally applicable since a variety of predications (such as ‘the mover moves’, ‘the fire burns’ and so forth) cannot be analysed by it. The use of the example of motion for the illustration of this template is explained by its centrality in the Buddhist world view. After all the term ‘mover’ (gati, ’gro ba) does not just denote moving objects in the everyday sense of the term, but more specifically beings in transmigratory existence. In analysing the mistaken presuppositions behind statements like ‘the mover moves’ Nagarjuna thereby attem pts to clear away misconceptions likely to arise at the very core of the Buddhist view of human existence.

6.1.2 The property-duplication argument

The property-reduplication argument raises another difficulty with the statements ‘present motion happens in the presently traversed space’ and ‘a mover moves’. If motion is ascribed to the presently traversed space or to the mover 18This is fact is also noticed by Schayer (1929-1930, 44, note 26): ‘It has to be stressed that the critique of the gati bears no direct relationship to the problem of motion. ‘Going’ is only used as an example to demonstrate the general impossibility of action (kriyu). ’, Walser (1998, 204): ‘Nagarjuna’s root text indicates that there is something about the form of the argument in chapter 2 which should serve as a model or pattern for any subsequent argument.’, and, interestingly enough, by Siderits and O’Brien themselves (at least concerning some verses of the second chapter): ‘The attack is not against motion per se but against a certain attitude towards language, and so its basic point will have effect wherever noncritical metaphysics is practiced.’ (1976, 294).

we suddenly end up with two motions, rather than just one. If there is motion in the presently traversed space this eventuates two motions, that by which it is a presently traversed space; and also the motion itself. 19 Also, if the mover moves, two motions would be implied: that in virtue of which it is manifested as a mover, and, it being a mover, that [motion] with respect to which it moves.20 To understand this argument it is essential to note that Nagarjuna regards both the presently moving object (gamyamana) and the mover (gantr) as thin individuals. For an example of what I mean by a thin individual, consider the case of some object which is green, cubical, and heavy. When referring to such an object in language we will generally form the nominalization of one of the predicates denoting its properties, which we then take to denote the object which instantiates the other two properties. Calling the object a ‘green, heavy cube1 we have turned the predicate cubical into the common noun cube of which green and heavy are then predicated. According to the standard ontological interpretation of this expression we are thereby refer19gamyamdnasya gamane prasaktam gamanadvayam / yena tad gamyamanam ca yac catra gamanam punah. MMK 2:5. 20 gamane dve prasajyete ganta yady uta gacchati / ganteti cdyate yena ganta san yac ca gacchati MMK 2:11. La Vallee Poussin’s edition has the beginning of lib as ganteti cocyate, ‘in virtue of which it is called a mover’ (99:6). This is one of several places (such as 99:7, 105:15, 106:11) in the second chapter of this edition where the root vac (‘to say’) instead of ahj (‘to cause to appear’, ‘to manifest’) is used. The Tibetan translation as mngon pa supports the latter reading (see de Jong (1978, 37-38), May (1959, 62, note 46)). The philosophical content of these passages is largely unaffected by this, apart from the fact that the reading with aiij places less emphasis on the role of language in conceptualizing the mover as a mover.

ring to an individual with two distinct monadic properties. Let us call the property which we turned into an individual by nominalizing the predicate the constitutive property, since it brings about or constitutes the individual referred to (in our example this is being cubical), and the other two instantiated properties, since they are instantiated by the individual thus constituted (being green, being heavy). Which properties we regard as constitutive and which as instantiated depends on our choice. With equal justification we could have chosen to speak of a ‘heavy, cubical green thing’ (so th at being green is constitutive, being heavy and being cubical instantiated) or a ‘green, cubical heavy thing’ (so th at being heavy is constitutive, being green and being cubical instantiated). In each case we would have referred to a different individual with different properties.

Nagarjuna distinguishes explicitly between constitutive and instantiated properties. The constitutive property of the presently moving object is that ‘by which th at is a presently moving object’ (yena tat gamyamanam)21; the constitutive property of a mover is that ‘in virtue of which it is manifested as a mover’ (ganta iti cacyate)22 or ‘the motion by which the mover is manifested’ (gatya yayajyate ganta)23 An instantiated property of a presently moving object is ‘motion itself’ (yat [...] gamanam)24; an instantiated property of the mover that ‘[motion] with respect to which it moves, it being a mover.’ (ganta san yac ca gacchati) 25 In the example of the green heavy cube we are dealing with a case where constitutive and instantiated properties are distinct; the cube is therefore a thick individual A thin individual, on the other hand, is an object whose only 21MMK 2:5b. 22MMK 2:11b. 23MMK 2:22a, 23a. 24MMK 2:5b. 25MMK 2:11b.

instantiating properties are its constitutive property or properties entailed by its constitutive property.26 A good example of a thin individual is a clap of thunder. A clap of thunder is a particular sound-event caused by rapidly expanding air along an electric discharge know as lightning. The particular sound made is the constitutive property of the thunder-clap; it is what makes a thunder-clap a thunder-clap. Of course a clap of thunder does not just have the property of making the sound it makes, it also has a certain volume, goes on for a certain length of time, can only be heard in a particular area and so forth. But all of these properties are entailed by the thundering’s constitutive property of making the thundering sound. A clap of thunder does not have any other properties apart from these. Nagarjuna argues that in the case of thin individuals the familiar analysis in terms of objects instantiating properties no longer works.27 This is evident when we compare a statement about a thin individual, such as ‘The thunder roars5 with one about a thick one, such as ‘Farinelli sings’. In the case of the latter it is clear that Farinelli existed before he started to sing, and at that time there was a silent Farinelli. But it would make little sense to apply this to the roaring thunder. There was no silent thunder present before it began to roar; it is the roaring as its constitutive property which brings the thunder about. We are therefore faced with essentially the same problem we already encountered when discussing the property-absence argument. As the thunder and the sound it makes are mutually dependent on one another for 26This concept of a thin individual should not be confused with the concept of a thin particular familiar from the contemporary metaphysical discussion. This concept denotes the object which is left when all the non-relational properties are abstracted away. See Armstrong (1997, 109-110, 123-126). 27Compare Bhaviveka’s commentary on MMK 2:22 (Ames 1995, 330).

their existence we cannot analyse states of affairs in which they feature in the same way in which we analyse those involving a thick individual, namely as being constructed of various independently existing entities, like the cube, the property of being green and the property of being heavy. If, however, we insist on conceiving of a thin object in the way in which we usually conceive of thick objects we will end up with a duplication of properties.28 A thick individual has some properties which are logically independent of one another (in the case of Farinelli, for example, being a singer and having dark hair), and one of this can be used to constitute an object of which the other is then predicated as an instantiated property. But in the case of a thin object there is only the constitutive property and the properties this entails. If we think that every object is to be analysed like a thick object, th at is by regarding it as a collection of at least two distinct properties, one of which is regarded an individual to provide the metaphysical condensation nucleus which can instantiate the other property, we end up with having to split up the single property into two: one of which does the work of a constitutive property, the other that of an instantiated property.29 Such a split is ontologically hard to make sense of, as we seem to be only dealing with one property seen in two different ways, and not with two distinct properties. The fundamental problem is that the conceptualization of some situation in terms of an individual instantiating a property is purely a result of cognitive convenience. We conceptualize something which is green, heavy, and cubical as a green heavy cube if cubes are what most interests us in the present context. But it is mistaken to rest an ontological distinction on such an intrinsically pragmatic fact by assuming that our conceptualiza28MMK 2:5, 6, 23. 29Compare Siderits and O’Brien (1976, 292-294).

tion corresponds to the way reality itself if carved up, namely as consisting of an individual (the cube) instantiating some properties (greenness, heaviness). The examples of thin individuals and the resulting multiplication of properties show us where the problem lies. But it is important to realize that Nagarjuna’s arguments are not just directed against specific problems arising only for thin individuals. It is rather that these present a particularly extreme case indicating difficulties with the assumption of a ready-made world sliced up into individuals and properties in general. The same problem of property duplication also arises when we consider this argument against the background of the classical Indian theory of grammar going back to Panini. As Nagarjuna makes clear a duplication of the action of movement requires a duplication of its agent, and therefore two movers.30 Candraklrti’s commentary on this presupposes Panini’s theory of karakas, a theory of the semantic relations between noun and verb.31 The underlying idea is th at the various participants of an event described in a sentence occupy various participatory roles relative to the action denoted by the verb, roles which are generally marked by different grammatical cases. Consider the following sample sentence: In the palace the prince brings presents from the king to the queen on an elephant. The event described here is one of bringing, as indicated by the verb, in which various entities participate: The prince is the agent (kartr, generally marked by the nominative case in Sanskrit), the presents are the object (karman, in the accusative), the queen is the recipient (sampradanam, in the dative), the king is the point of departure (apadana, in the ablative), the 30dvau gantarau prasajyete prasakte gamanadvaye. MMK 2:6a. 31 Astadhydyi 1,4.24-54; see Ganeri (1999, 51-63).

elephant is an instrument (karanam, in the instrumental) and the palace is the location or ‘support’ (adhikaranam, in the locative case). The theory of karakas provides us with a general account of how the different thematic roles the participants in an action might occupy can be expressed in Sanskrit by the various vibhaktis or cases. CandrakTrti observes in his commentary on MMK 2:6 that the karaka required by the verb gamy ate ‘is moved’ is an agent (kartr) which is the mover (gantr).32 If the property of moving thus requires a mover, given the reduplication of motion discussed above we are faced with two distinct agents (one for each motion) rather than just one. We might argue at this place that on the face of it there seems to be no problem for a single agent being the means of bringing about two actions simultaneously, as for example in the case of someone simultaneously smoking and typing. This does not mean that there are in fact two persons sitting at the desk, a smoker and a typer, rather than a single one, a smoking typer. Multiplicity of actions does not always entail multiplicity of agents.33 To see where the problem lies here we have to have closer look at the various conceptions of the nature of the karakas or participants of an event in Indian grammatical theory.34 In his commentary CandrakTrti refers to Bhartrhari’s account when he claims that a karaka is not to be understood 32‘Since an action (kriya) necessarily depends on a means of accomplishing it (svasadhana) [which is] either the object (karman) or the subject (kartr) [of the action], the action of motion also involves an agent and therefore depends on an agent of motion.1 yasmad avasyam kriya svasadhanam apeksate karma kartdram vd | garnikriyd caivam kartary avasthitd ‘to gantaram apeksate. PP 96:8-9 Here sddhana is taken to be synonymous with karaka. 33Ganeri (1999, 58, note 12). 34Bhattacharya (1977, 269-270), ?. See also Renou (1942) s. vv. karaka, sakti, sadhaka, sddhana; Chalcravarti (1930, 225).

as a substance (dravya) but as a power or capacity (sakti) . 35 The reason for this is that if the karaka denoted a substance, the same object could not function in different ways in different contexts, as an agent in one, an object in the next, or as an instrument in the third.36 The karaka therefore refers to the powers of an object to fill specific roles in different contexts. The number of powers is diversified by the actions; the actions are not seen as properties of a single agent. For each action, such as smoking and typing we therefore assume a separate power which serves as its agent. The problem now arises if we assume th at the different powers are differentiated due to the different natures of the actions performed, such as typing and smoking. The two motions, however, are actions of the same nature, and should therefore be regarded as being brought about by the same power as an agent.37 Since the splitting of a single motion into two thus commits us to the unsuitable assumption of two different powers as agents of motions the splitting must be seen to rest on a deficient analysis of the situation at hand. We therefore have to conclude that thin individuals cannot be analysed in the same way as thick individuals if we want to escape the methodologically distasteful consequences of splitting up a single property and a single agent into two, thereby multiplying entities beyond necessity. 35Bhattacharya (1980, 89). See also Bhattacharya (1977, 269-270, note 21). That karaka is a dravya is also denied by Patahjali in his Mahabhasya (Kielhorn 1880-1885, I, 442:23-26). Note that Bhaviveka in the Prajnapradfpa, commenting on MMK 2:6 has the opponent assert that according to the grammarians (sgra pa dag, sabdika) the agent (byed pa po, kartr) of the action of going is the goer. The opponent must conceive of the goer as some sort of substance, as Bhavavivelca objects to this by pointing out that the goer is a mere collection of conditioned factors ( ’du byed, samskara). See Ames (1995, 308). 36Bhattacharya (1980, 89). 37Bhattacharya (1980-1981, 38).

i ’PAce a j> 1 \ i t # « • * * M m 6 » • ♦ • J . • • • « ✓ • • ®a . F IG U R E 1 6.2 T he beginning of m otion

In verses 12-14 of the second chapter Nagarjuna is concerned with the location of the point where motion begins (gamanasya arambha). His argument can be best illustrated by considering the diagram in figure 1 . For the sake of simplicity we consider both space and time to be discrete. There is a box which occupies different spatial points in succession: it starts off at point s<2 at times ti and 12 until it reaches point S4 at t\. The diagram thus depicts the motion of a box from the left to the right. If we ask where the motion of the box begins, the answer is obvious: the box commences its move to the right at point S2 ■ To begin a motion at some point an object must first be stationary at this point (so that there are at least two successive moments of time in which the box remains at the same point of space), and

at the immediately following moment it must be located at an adjacent point of space. At t 2 the box is located at point s2> at £3 at point S3 . So point s2 satisfies the condition for being the place where motion begins. Given th at there seems to be nothing inherently problematic about this, why does Nagarjuna claim that the place where motion begins ‘is nowhere perceived’ (adrsyamana sarvatha;)? Nagarjuna divides the space where motion takes place into three jointly exhaustive and mutually exclusive parts: the part already moved over (gata), the part presently traversed (gam.yamana), and the part to be moved over in the future (agata). To make things a bit more precise we can say that a place i is presently traversed by some moving object if the object is at a spatially adjacent place i — 1 at the preceding moment t — 1 , is at place i at t (which is the present moment), and at i + 1 at £ + 1. Similarly i is a place already moved over if £ is some moment in the past, and a place to be moved over if £ is in the future. Now assume th at the place where motion begins (let us call this b) is one of the places already moved over. In this case the moving object would have to have come from some adjacent place b — 1 at a moment before t (where t is in the past), reached 6 at £ and moved to bT 1 at t T 1. But it is obvious th at this cannot be the case, since if b is the place where motion begins the moving object cannot have got there from another place at the immediately preceding moment, because then b would just be one of the places moved over. Since a place already moved over must have been occupied by an object coming from the immediate vicinity at the immediately preceding moment, b cannot be one of these places. If b was a place already moved over the moving object would have come from the immediately preceding point of space. So there would have to be motion from b — 1 to b, i.e. motion before the beginning of motion. But

b — 1 can neither be taken to be a place already moved over, nor a presently traversed space, nor one yet to be moved over,38 since all of these are to be found after the beginning of motion. Therefore b cannot be a place already moved over. Analogous arguments show that b also cannot be a presently traversed place or a place to be moved over in the future. It is evident that the same argument can be run concerning the place where motion stops; for the reason just given it cannot be located in any of the three parts of the space where motion takes place.39 A simple numerical model illustrates this point: if we define a set of numbers such that for every number in it that number’s direct predecessor and direct successor must also be in the set it is clear that this set cannot have a smallest (or largest) element. For suppose x was this smallest element, then a;’s predecessor would also have to be in the set, but this is smaller than a;, so £ cannot be the smallest. A different interpretation of Nagarjuna’s argument for the unfindability of the beginning of motion is given by Siderits and O’Brien.40 Their interpretation is based on the presupposition th at Nagarjuna assumes the infinite divisibility of time as a background to this argument in verses 12 to 13. Suppose some temporal interval consisted of some object first at rest, then later starting to move. Now take the moment of time t dividing the rest from motion. Now m atter how short t is, it can always be divided further, subsuming 38<Neither the presently traversed [space] nor [the space] already moved over are before the beginning of movement. Where would motion begin? How could movement [begin] in the [space] yet to be moved over?’ prdg asti gamanarambhad gamyamdnam na vd gatarn / yatrdrabhyeta gamanam agate gamanam kutah. MMK 2:13. 39‘0 n e does not stop after the presently traversed [space], after the [space] moved over, also not after the [space] to be traversed.’ na tisthati gamyamdnan na gatdn nagatad api. MMK 2:17a. 40(1976, 295-296).

its initial sub-moments under ‘rest’, its later sub-moments under ‘motion’. Adding together ‘rest’ and ‘motion’ will then cover the entire duration of the temporal interval, without any place for t. On this ‘knife-edge’ view of t there is no moment where motion can begin, since t is just the dividing-line between rest and motion. It is not a temporal duration where anything can happen. We might want to note that matters don’t improve if we assume that time consists of discrete indivisible atoms. If we regarded t as an atomic moment between the last moment of rest t — 1 and the first moment of motion t + 1 we are again faced with the problem of where to locate t in the exhaustive division of the temporal duration into past, present, and future motion. Since the beginning of motion cannot be in the past or future, our best bet is the present motion. But then since t is atomic it cannot be the moment of present motion, as nothing moves during t: there can be no changes during an atomic moment of time.41 A third argument for the impossibility of locating the beginning of motion in the present motion is suggested by Candraklrti’s commentary on verse 1 2 . CandrakTrti claims that the beginning of motion ‘is also not in the present motion, since that does not exist and because it [absurdly] eventuates two actions and two agents’.42 This obviously is a reference to the propertyduplication argument mentioned in verses 5-6 and 11. In analogy with our interpretation of this argument given above we could here read CandrakTrti as trying to establish the impossibility of conceiving of the beginning of motion and its property of spatial location as independently existent objects. This point can then be generalized to apply to different examples of change and 41 Compare Galloway (1987, 81-82) who regards this argument as implicit in CandrakTrti's commentary on verse 1. A2napi gamyamane tadabhavat kriyddvayaprasangdt kartrdvayaprasaiigac ca. PP 100:8.

their respective locations.43 Unfortunately Nagarjuna’s verses do not allow us to decide which (if any) of the three arguments he had in mind. The enterprise of rational reconstruction can here only suggest plausible alternative arguments which the Madhyamika might want to put forward. We cannot tell what the argumentum ipsissimum of Nagarjuna might have been. We shall therefore continue the discussion by assuming th at it has been established by one of these arguments th at the beginning of motion is not to be found anywhere within the three parts of the space where motion takes place. Nagarjuna now points out th at this entails a problem for the discrimination of these three parts of space.44 That portion of the way which has been moved over in the past is just the collection of all the spatial points each of which is gata, th at is for each of these points the moving object must have been located at this point at some past time t, and it must have been at a preceding point at t — 1 and at a succeeding point at t + 1. But in order to know how many points to include in the collection we have to know where the motion begins. Otherwise we would not be able to distinguish those spatial points which have not been moved over from those which have. The same problem arises when ascertaining which collection of points forms the part of space yet to be moved over. Here we have to determine which point is the end of motion in order to distinguish the part which is yet to be moved over from that which is not. It is now clear th at in verses 12-14 Nagarjuna is arguing for two con- 43See Mabbett (1984, 414-415) for a defence of this interpretation. 44‘How are the [space] moved over, the presently traversed and the one yet to be moved over differentiated when the beginning of motion is indeed nowhere perceived?’ gatam kim gamyamdnam kim agatam kirn vikalpyate / adrsyamdna drarnbha gamanasyaiva sarvathd. MMK 2; 14.

elusions. Firstly, given the conceptual resources of the triple division of the space where motion takes place into the space which has been moved over (i.e. a collection of spatial points each of which is gata), the presently traversed space (the point which is gamyamana) and the space yet to be moved over (the points which are agata) it is not possible to define the spatial point where motion begins. This is due to the fact that to be in one of the three divisions a point must have had the moving object move to it at the preceding moment, whereas to be the beginning of motion a point cannot be such that something has just moved there, since it would then just be one of the many points across which motion takes place. But this passage is not just about the definition of concepts. Since the triple division of the space where motion takes place is seen to be exhaustive, and since the above argument shows that none of the points in the triple division can be the beginning of motion this implies th at the beginning of motion cannot be anywhere within the space where motion takes place.45 But this seems peculiar, since the beginning of motion is where motion takes place, not where it does not.

Secondly it is not possible to define two of the three divisions, namely the space which has been moved over and th at which is yet to be moved over without reference to the point where motion begins and its dual, the point where motion stops. These two are essential cognitive resources for our 45It is not the case that Nagarjuna just ‘falsely assumes that what is characteristic of individuals must be characteristic of the group containing those individuals’, as argued by Cheng (1980, 237). The argument should rather be understood as a proof by cases: if the place where motion begins is to be found anywhere then it should either be in the portion of space traversed in the past, in the one presently traversed or in the one not yet traversed. As each of these can be eliminated, the place where motion begins is nowhere to be found.

understanding of motion. They must provide the dividing line between the space which has already been moved over and that which has not, as well as th at between th at which is yet to be moved over and that where no motion is going to take place in the future. We are therefore faced with a paradox. The triple division of space where motion takes place presupposes the beginning of motion. The beginning of motion in turn presupposes the triple division of space in which this beginning is located. But the beginning is nowhere to be found within the space triply divided, nor would it make sense to say th at it exists outside of th a t space. The beginning of motion therefore must both exist (since it is conceptually necessary given the triple division of space) and cannot exist (since we can demonstrate that it cannot exist at any location within this division) . 46 The paradox can be resolved by rejecting the assumption that the beginning of motion is findable. We thereby deny that it can be picked out by a set of properties it has independent of us, who conceive of the space where movement takes place, for example by saying that some point b qualifies as the beginning of motion if the moving object occupies it at time t but did not occupy the directly preceding spatial point at the immediately preceding moment, or by trying to squeeze b into the infinitesimal temporal moment between rest and motion. Since on such an understanding b turns out to be unfindable we must come up with another conception of b. The idea here is to deny that any point qualifies as the beginning of motion independently of us, but th at it is rather our decision to regard it as such a point which makes it the beginning of motion. This does not mean that we could pick absolutely any point and take it to be the beginning of motion, but as long as certain boundary conditions are observed (e.g. that the beginning of motion 46Compare Siderits and O’Brien (1976, 295-296).

must be temporally and spatially before the place presently traversed) we can pick an arbitrary point and declare it to be the beginning of motion. W hat this means is that we regard that part of an event which begins with the presence of the moving object at a given point and stretches up to its being located at the space presently traversed as a single event, regardless of whether the moving object occupied an adjacent point at the immediately preceding moment. By deciding to regard some moment as the beginning of motion we split up the flow of events according to our cognitive needs and regard everything between this and the similarly imposed end of motion as part of a single event of motion. In this way the paradox disappears. We can still have the triple division of the space where motion takes place and have the beginning (and end) as boundaries of this. These two points are no longer unfindable since according to the present interpretation they are just where we draw the line between one event and another; they do not have to fulfill any additional conditions like the ones given above. As Nagarjuna argued earlier on the triple division of space is conceptually dependent on the notion of the beginning of motion. This, however, does not mean that the beginning of motion has to have any existence apart from the cognizing subject; in fact it is precisely this assumption which leads to the problems described by Nagarjuna in verses 12 to 14, The beginning of motion (as well as the beginning of events in general) is not something found out there in a ready-made world, but a boundary drawn by the mind in accordance with one’s particular interests and needs. On the basis of such an imposed boundary we can then establish the triple division of space and time into where and when an event had already taken place, where and when it is presently taking place and where and when it will take place. C

6.3 T he interdependence of mover and motion

Nagarjuna observes th at the concepts mover (gantr) and movement (gamanam, gati) are existentially dependent on one another. The concept of a moving object requires that of a movement this object carries out, the concept of movement must be the movement of something, i.e. of the moving object.47 This mutual dependence implies for Nagarjuna th at mover and movement can neither be regarded as identical nor as distinct objects.48 To regard mover and motion as identical would imply that agent and action are considered to be one object (ekibhava). This would mean that no agent could ever perform two distinct actions, since for this he would have to be identical with two distinct things. The agent must therefore vary with the action, for example by not being regarded as a substance (dravya) but as a power (sakti) to carry out a certain action, as done by CandrakTrti in his commentary on verse 6 . As these powers come into and go out of existence this entails the problem th at there would be no continuous existence of a single agent performing a sequence of actions over time. Since this point generalizes to individuals and their properties as a whole a theory which regarded these two as identical would have difficulties in explaining how we could ever regard such a sequence of distinct individuals and properties as a single unified temporally extended object 49 A more specifically Buddhist difficulty would arise in connection with the concept of karma. If for every action there is a 47MMK 2:7. 48MMK 2:18-21. 49This is a familiar problem for theories which equate objects with sets of properties. As two sets are identical iff they have the same members an object could never lose a property and yet remain the same object. Compare Armstrong (1978, 37-38).

distinct agent, what reason is there for the karmic consequence of an action to apply to one agent rather than another? As ex hypothesi none of the later agents are identical with the original one there seems to be no justification for the karmic result to be reaped by one rather than another.50 Mover and motion also cannot be regarded as distinct. It is im portant to note th at here, as well as in other contexts, Nagarjuna uses the word ‘distinct’ to mean ‘independently existent’. If mover and motion exist independently, like a piece of cloth and a pot, as CandrakTrti puts it,51 it would be possible for a stationary mover to exist, or for a movement to exist which was not the movement of any object. But since the two are existentially dependent on one another, neither of these is in fact possible. While the identification of mover and motion led to problems with the continuity of an individual over time, as we have just seen, regarding them as existentially independent generates a different problem. Even if we adopt the more sensible position of interpreting independence here as meaning that motion could be instantiated in a different object from the one it is in fact instantiated in, and that the moving object could instantiate a different motion from the one it in fact instantiates we end up with having to postulate a thin particular, a substratum which remains once all the properties have been abstracted away. For if any property could just leave the individual and go instantiating somewhere else how are we to characterize the individual? Since any property can exist in principle without it it must be something 50Kalupahana (1991, 128) interprets the identity of agent and action as the position of the Sarvastivadins, who assume ‘identity (sa eva) on the basis on an eternal substance (svabhava), thereby rendering the attribute (laksana) an ephimeral (sic) or impermanent come-and-go entity’. 51PP 105:5.

which could in principle exist without any of its properties.52 The difficulty resulting from treating mover and motion as independently existent objects is also what is behind Nagarjuna’s assertion in the final verses of chapter 2 , saying that neither an existent, nor a non-existent, nor a both existent and non-existent mover can carry out a triple movement.53 It is easiest to fit this verse into the argumentative context of chapter 2 by regarding the Triple movement’ as not referring to movement in the past, present, and future, but by following CandrakTrti1 s commentary.55 According to this interpretation ‘existent mover1 here means one in which the activity of moving (gamikriya) inheres, a non-existent mover is one in which it does not inhere, while a mover which is both is an entity in which it both inheres and does not inhere. Saying th at the movement, which is to be understood as the space gone over (gamyata) 56 is ‘triple’ equally means that either the activity of motion inheres in it, fails to inhere in it, or both. On this interpretation we are left with nine distinct possibilities: th at a mover in which the activity of motion inheres moves at a place in which this activity inheres also, that a mover in which the activity of motion inheres moves at a place in which this activity does not inhere; and so forth for the remaining possibilities. The philosophical idea behind this is straightforward. 52See Armstrong (1997, 123-126). 5Zsadbhuto gamanam ganta triprakaram na gacchati / nasadbuto ’pi gamanam triprakaram sa gacchati / gamanam sadasadbhutah, triprakaram na gacchati. MMK 2:24- 25a. Garfield (1995, 133). He also translates the Tibetan of MMK 25a (yin dang ma yin gyur pa yang / ’gro m am gsum du ’gro mi byed) by ‘Neither an entity nor an nonentity moves in any of the three ways1, rather than as ‘An object which both exists and does not exist does not carry out a movement in any of the three ways1. 55PP 107:9-14. CandrakTrti explicitly refers to MMK 8 for the interpretation of 2:24-25, which indeed gives a more detailed version of the argument Nagarjuna has in mind here. 56tatra gamyata iti gamanamihocyate. PP 107:9.

We should not assert that a mover in which the activity of motion inheres moves at a place in which this activity inheres also, if this is supposed to mean that the activity of motion inhering in the mover is independent of its inhering in the mover. This is due to the fact that one depends on the other: motion can only inhere in a mover if it moves at some place, a place can only be the locus of motion if something moves at it. Furthermore, it cannot be the case that a mover in which the activity of motion inheres moves at a place in which this activity does not inhere, since it would then not be a space gone over. It is obvious that the remaining examples are to be treated in a similar way. The ‘contradictory’ third alternative of both inherence and non-inherence seems to be given by Nagarjuna merely for the sake of completeness, as supposing that some property both inheres and fails to inhere in some object is inconsistent.57 The bottom line58 of the above arguments concerning the interdependence of mover and motion is that while the concepts of mover and motion (and, more generally, agent and action and individual and property) have to be regarded as non-identical, neither of them can be regarded as self-sufficient or existing from its own side, since the existence of each requires that of the other. It is therefore somewhat misleading to take Nagarjuna as arguing that mover and motion are not real.59 While it is certainly correct to say that Nagarjuna thinks that mover and motion are illusory to the extent to which the way they appear (namely as independently existent entities) is not the 57As Nagarjuna asserts in MMK 8:7. See page 129. 58MMK 2:21. 59As claimed by Murti (1955, 183), see also pages 137 and 307. Note that Jacques May translates CandrakTrti concluding his commentary of MMK 2:21 in PP 105:11 with the words nasti gantrgamanayoh siddhir ity abhiprayah rather misleadingly as ‘L’idee est que le mouvement et son agent sont depourvus de realite’ (1959, 71).

way they really are, their lack of reality is quite different from th at of other non-existent objects, such as hare’s horns and present kings of Prance, which do not exist even at the level of conventional reality (samvrtisat). 6.4 T he second chapter of the M M K in its argum entative context The second chapter of the MMK must be understood as playing a double role in Nagarjuna’s philosophical enterprise. On the one hand it is part of the discussion of a variety of different entities (such as agent and action, suffering, time, nirvana and so forth) attempting to show that none of them exists substantially, th at is, by svabhava. In this context the examination of motion deserves a particularly prominent place because of its centrality in the Buddhist world view. Cyclic existence or samsara is after all nothing but the moving about (samsr) in the various realms of rebirth. When Nagarjuna argues th at mover, motion and so forth are empty of svabhava he uses the terms both in their everyday and in their soteriological sense, where the mover (gati) is the subject to be reborn and motion is the move from one life to the next. It is in the context of this discussion that Nagarjuna’s arguments about the beginning of motion and the identity and difference of mover and motion have to be understood. If we accept Nagarjuna’s conclusion th at the beginning (and end) of motion are nothing to be found ‘out there’ in the world, but rather a boundary established by the mind this also entails th at the beginning and end of a particular motion in samsara, that is a particular birth and a particular death have no objective existence either, but are merely conventional ways of cutting up the flow of cyclic existence into conceptually CHAPTER 6. MOTION 225 convenient bits. Seen the other way round the concepts of past, present, and future lives only arise once we have decided to mark particular places in the continuity of consciousness as ‘birth’ and ‘death’. Read in this soteriological way Nagarjuna’s arguments in this section of chapter 2 of the MMK aim to establish that such central concepts like birth and death, past, present, and future lives are no objective features of reality but merely conventionally real boundaries drawn by the human mind. This is made more explicit by Nagarjuna in chapter 11 of the MMK where he notes that Where the earlier, the later, and the simultaneous do not appear, how [is there] a proliferation [of the concepts] ‘birth’, ‘ageing’ and ‘death’ ? 60 Given the cyclical nature of samsara what is earlier and what is later is very much dependent on where we identify the starting-point. The hands of a clock will reach ‘3’ before ‘5’ if we start at ‘2’, but they will reach ‘5’ before ‘3’ if we start at ‘4’. As Nagarjuna has argued that the starting-point is not something ‘out there’, but a boundary drawn by us in order to accord with our specific cognitive concerns, it becomes evident that we cannot ascribe any objectively existing referents to such concepts as ‘earlier’ or ‘later’, ‘birth’ or ’death’, and ‘past life’ and ‘future life’.61 60 yaira na prabhavanty tie purua-apara-saha-kramah / prapancayanti tam jatim taj jaramaranam ca him. MMK 11:G. 61 Jay Garfield observes that ‘to see particular entities as having determinate, nonconventional beginnings of existence and determinate, nonconventional termini and, hence, that there are distinct times at which there is a clear fact of the matter about whether or not they exist, independent of conventions for their individuation, is to see those entities as having necessary and sufficient characteristics for their identity, that is, as having essences [i.e. svabhava ]. [...] Once we see the world from the standpoint of emptiness of inherent existence, the history of any conventionally designated entity is but an arbitrary

The discussion of the identity and difference of mover and motion addresses another crucial issue which will be taken up again by Nagarjuna,62 namely the question of the status of the subject transmigrating through a succession of rebirths. Clearly the mover (the person in cyclic existence) cannot be identical with each different rebirth, since it would then be identical with a number of things which are taken to be distinct at the conventional level. But it can also not be distinct from them as anything resembling an a£mrm-like transmigrating substance is ruled out in the Buddhist view of persons. There is therefore something fundamentally mistaken with the view which sees the transmigrating person and his rebirths as two entities which could be related by identity and difference. While it thus appears that the arguments in the second (12-14, 17) and third (7, 18-21) group of verses of the second chapter are concerned with the investigation of the existence of svabhava in various entities connected with motion in both the everyday as well as in the soteriological sense, the first group (1-6, 8-11, 15-16, 22-25) is intended to play a more general role. It is not just that the concepts of mover and motion have to be understood in more than one sense, but rather that they serve as placeholders for which a variety of other concepts denoting an individual and a property could be substituted. Nagarjuna’s aim in these verses is therefore primarily to establish an ontological conclusion about the relation between individuals and their properties. By considering predications involving thin individuals (such as ‘the mover moves’ or ‘the fire burns’) Nagarjuna establishes th a t the standard analysis of predication into individuals and properties, which conceives of them as mutually independent entities combined in a state of affairs is not stage carved out of a vast continuum of interdependent phenomena.1 (1995, 199). 02In chapter 27 of the MMK.

satisfactory as a general analysis. Statements referring to thin individuals cannot be analysed in this way. Furthermore, Nagarjuna wants to argue that this problem generalizes to analyses involving thick individuals as well. Once we have accepted that talk of individuals and properties in the case of such statements as ‘the mover moves’ is nothing more than the projection of forms of language which are mistakenly given ontological weight we will be much more reluctant to take this analysis ontologically seriously in other contexts. We should rather conceive of this analysis as a reflection of what is cognitively convenient for us, rather than as a structure of the world mirrored in our language.

Chapter 7 The self

After considering Nagarj una’s arguments for the selflessness of external phenomena such as causation and motion we can now turn towards an assessment of the most im portant example of a subjective phenomenon, namely the self. Of all the discussions of the emptiness of various entities which Nagarjuna examines in his works that of the emptiness of the self occupies a special position. He notes th a t1 Where something prior to, simultaneous with or after seeing and so forth [which could be regarded as a self] is not evident, there conceptions ‘it exists’, ‘it does not exist’ [with svabhava ], have ceased. Nagarjuna claims here that once the emptiness or lack of svabhava in the self has been realized it will be comparatively easy to understand the emptiness of other phenomena. This is because the view of a substantial self is particularly natural and tends to assert itself in an especially convincing m anner.2 1prdk ca yo darsanadibhyah sdmpratam cordhvam eva ca / na vidyate ’sti ndstiti nivrttas tatra kalpand. MMK 9:12. 2Garfield (1995, 188). 228

Having seen through this fundamental illusion, Nagarjuna wants of argue, all other mistaken ascriptions of svabhava can be unmasked in a relatively straightforward manner. In order to appreciate Nagarj una’s arguments for the emptiness of the self it is essential to have a clear idea of what he argues against, that is what a self with svabhava would amount to. Such a self can be characterized by four core properties.3 It is an entity distinct from both our body and our psychological states. The self is not the same as the body, but is what has the body, similarly the self is what has sensations, thoughts, beliefs, desires and so forth. It is essentially unchanging. Whether or not we think that our selves survive the death of the bodies they have we still want to claim that it is the same self present in the elderly general now and in the schoolboy he was sixty years ago. This permanence of the self also serves as the foundation for the ascription of moral responsibility, as we are dealing with a single entity unified over time. Thirdly the self is a unifier:4 it integrates diverse sensory information, beliefs and desires in such a way as to allow us to make decisions and to act on the basis of them. Fourthly, the self is an agent, it is the permanent core which makes the decisions which shape our lives. The results of these decisions may then in turn influence the self, but there is little doubt that it is the self, not the decisions which occupy the driving-seat. This substantialist conception of the self appears to be a relatively accurate description of our intuitive, everyday belief of what we are.5 3Compare Gowans (2003, 70). 4Gautama makes this point in NS 3, 1, 1. See Chakravarti (1982, 222-223), Siderits (2003, 22-23). 5The extent to which the notion of a self denied by Nagarjuna (and the early Buddhists for that matter) was influenced by the Samkhya and Vaisesika concept of atman is difficult to determine. See Conze (1967, 38), Bliattacharya (1973), Harvey (1995, 33-34). CHAPTER 7. THE SELF 230 As is to be expected the aim of Nagarjuna’s examination of the self is to show th at this intuitively plausible view of the self is fundamentally mistaken. The substantialist view of the self has to be replaced by a different one. We can divide Nagarjuna’s discussion into two main parts. The first deals with the relation of the self to its synchronic parts at a single time and to its diachronic parts across time. The second investigates the relation between the self and its properties.

7.1 T he self and its parts

The Buddhist tradition divides the person into five main constituents (skandha): matter or the physical body (rEpa), sensation (vedana), perception (samjnd), intellect (samskara), and consciousness (vijnana) . 6 W hat is important from a philosophical perspective is not so much the precise nature of these constituents and the merits and demerits of the resulting psychological theory for explanatory purposes, but primarily the fact that the human person or self are conceptualized as composite.7 In addition it is essential to keep in mind th at the analysis of the self into different components is meant to be exhaustive. It is not just supposed to illustrate various aspects or properties of a person, but to list all the aspects it consists of. Bearing this in mind will keep us from interpreting a denial of any of the five constituents being the self as an assertion that something else is.8 Once an exhaustive analysis of the self into a fixed number of constituents 6These can in turn be subdivided further (RA 1:81). See Nyanatiloka (1950, s.v. khanda, 73-76) for an overview of the standard Abhidharma analysis. A detailed exposition is in chapter 14 of the Visuddhimagga (Buddhaghosa 1991). 7Garfield (1995, 142, 245). 8Collins (1982, 7-10, 98)

is in place the question concerning the relation between these constituents and the self they comprise naturally arises. Nagarjuna mentions four different ways in which the self and the constituents could be related.9 The self could be identical with the constituents (either with a subset or with all of them together), it could exist as a separate entity distinct from them, the self could contain the constituents as a part or finally it could itself be part of the constituents. Nagarjuna observes th at identifying the self with a particular constituent, such as the body, or consciousness entails the difficulty that the individual constituents are constantly changing. 10 Neither the body, nor consciousness, nor any of the other constituents remains as it is over time. Such an identification would therefore not do justice to the view of the self as essentially unchanging. This is a familiar argument for the absence of a substantial self and is frequently encountered in the Pali Suttas. 11 The Anattalakkhana Sutta, for example, describes the Buddha as investigating each of the constituents by asking: ‘Is what is impermanent, suffering, and subject to change fit to be regarded thus: ‘This is mine, this I am, this is my self’ ? ’12 If we cannot identify the self with a single constituent we might consider equating it with some or all of the constituents across a stretch of time. We would then for example regard as our self not just our body as it is now, but a sequence of bodies which incorporate the past as well as the future 9MMK 22:1, See also RA 1:82. A detailed discussion of these possibilities can also be found in MA 6: 126-165. The Buddhist commentarial literature often illustrates the possible relations between the self and its constituents by a series of similes, like the relation between a flower and its scent, a tree and its shadow etc. See Conze (1967, 38), Collins (1982) for references. 10MMK 18:1a, 27:3. 11 Collins (1982, 98). Samyutta Nikaya 22.59 (Bikkhu Bodhi 2000, I: 901-903).

stages of our body. This solves the problem of the self disappearing from one moment to the next (as each single constituent is only in existence for a short duration) but entails other problems. First of all, if we have to include constituents at future times (such as our body tomorrow) into the entity we regard as our self in order to explain our concern for our future self we face the difficulty that these future constituents do not yet exist. We then could not claim that our self as it existed today was in fact the entire self. We (that is, our selves) could never be wholly present at the present time. 13 Secondly our candidate for a self is now no unified whole any more, but rather a series of ever-changing parts. In this series there will not be any one thing which remains constant and only changes its accidental properties. Such an account would therefore be hardly satisfactory for an advocate of a substantial self. Given that the identification of the self with constituents at one time or across a stretch of time does not appear to be satisfactory the other alternative for the defender of a substantial self is to assume th at the self is an entity distinct from the various constituents of a person. 14 The self would then be regarded as the owner of the body, the experience!' of the sensations, the perceiving subject and so forth. However, as Nagarjuna points out, such a self could not bear the marks of the constituents (bhaved askandhalaksanah), th at is it could not be characterized as the owner of the body, the experiencer of the sensations, the one undergoing change etc.15 This is because such a self would be completely unknown to us. Once we have abstracted from all the constituents of the person there seems to be nothing left which could qualify as a self. 16 The familiar Humean observation that introspection 13See Garfield (1995, 345). 14MMK 18:1b. 1SMMK 13:5. 10RA 2:1.

shows us all sorts of inner psychological events, but never acquaints us with any object which has the characteristics of the substantial self17 means th at a self existing apart from the five constituents of a person (or any other set of constituents we might come up with) could not be one we are directly acquainted with. But it would then be distinctly odd to assume th at such a self would be the one we cared about. For all that we know it might be an entity we have never even come across. Another difficulty with this position is that the assumption of the self as an entity distinct from the constituents also implies that it would be possible for the self to exist without any of them, since it does not existentially depend on them . 18 There could be something which we would be justified in regarding as our self, even though it was not connected with our body, shared none of our memories, desires or preferences and would not even have to stand in any specific temporal or causal relation to these. Even if such a thing could exist it would be questionable with what motivation we could call it a self, given that it is devoid of all the connections which we usual regard as crucially important for our selves. More worryingly its independence makes it difficult to conceive of this self as an agent shaping our lives with its decisions. If there is no essential causal connection between the self and our various cognitive faculties, how does it enter into the formation of beliefs, the making of decisions, and the bringing about of actions? Such a self would be devoid of action (akarmakah) 19 and for this reason could not be regarded as 17‘For my part, if I enter most intimately into what I call viyself I always stumble on some particular perception or other, of heat or cold, light or shade, of love or hatred, pain or pleasure. I never can catch myself at any time without a perception, and never can observe any thing but the perceptions.’ (Hume 1896,1, IY, VI: 252). See also Parfit (1984, 223). 18MMK 10:1b. 19MMK 10:2.

an agent. The final two possibilities considered by Nagarjuna, th at the self contains the constituents as a part or is itself part of the constituents can be seen to be equally unsatisfactory in the light of the arguments presented above. The former would lead once again to the problem of the unity of the self, the latter would entail the difficulty of how constantly changing entities like the five constituents of a person could have any permanent parts at all. The picture of the self thus emerging might strike us as very close to Hume’s bundle theory. While this parallel is illuminating to some extent it is important to be aware that for Hume his view of the self is the corollary of an epistemological theory which is not seen as having any practical implications. For Nagarjuna, however, as for Buddhist thinkers in general the emptiness of the self constitutes a central philosophical position with major practical and soteriological implications.20 Its realization, th at is not just the intellectual understanding of the absence of svabhava in the self but the cognitive shift accompanying the ability to stop conceiving of oneself as a substantial self21 is taken to be an essential step on the road to liberation.22 The inability to come up with a satisfactory account of the relation between the self and its parts might now lead us to thinking that there is no temporally extended self at all. Thus what be take to be our self existing now would not be in any way the same as what we regarded as our self yesterday, since there is no continuity between them .23 There would just be mutually independent entities each of which would be regarded as a self at a time, 20MMK 18:4-5, Siderits (2003, 29-31). 21 Collins (1982, 94). 22Further cautionary remarks about drawing parallels with Hume’s bundle theory can be found in Conze (1963, 113-115). 23MMIC 27:9a.

but there would be nothing which could be interpreted as an overarching, essentially unchanging self. A minor difficulty with this view is that if the various selves were indeed independent and each existed by svabhava the existence of one would not depend on the non-existence of another. But then it would be conceivable that yesterday’s self just continued existing while today’s self came into being, so we would end up with two mutually independent selves at the same time, and thereby fail to account for the self’s unity.24 The main difficulty with this view of episodic selves is th at it transforms most of the relations we regard as intrapersonal into interpersonal ones. An obvious example is memory. Given that the self which had an experience and the one which later remembers it are independently existent objects the transmission of memory turns out to be on the same level as the transmission of information between two persons; perhaps even more problematic. Since any causal influence of the former on the later self would imply a dependence relation between the two it is hard to see how anything could be transm itted between them at all.25 In any case it would be impossible to distinguish between true and false recollections, since a criterion of the former is exactly the connection between selves which the episodic theory denies. An even greater difficulty is presented by the issue of moral responsibility.26 For apportioning praise and blame, but also for making sense of the Buddhist concept of karma we need to be able to postulate some sort of dependence relation between different stages of a self, because only in this way we could explain why today’s self should be responsible for yesterday’s actions, and 24MMK 27:10. 25 This then also entails the problem how the different stages of the self could arise at all, given that they are not causally produced. See MMK 27:12. “ MMK 27:11.

how a being reborn in the form of a god at the present time could have been a human being in a previous life.27 7.2 T he self and its properties Most of the above arguments for the emptiness of the self based on an investigation of its relation to its synchronic and diachronic parts are familiar to us from pre-Madhyamaka Buddhist literature. However, the MMK also presents us with a set of arguments against the substantial conception of the self which has a distinctly Nagarjunian slant. Nagarjuna’s opponent wonders ‘if there was no self, where would the self’s properties come from? ’28 Similarly we might want to ask ‘how can seeing and so forth belong to something which is not found? Therefore there is an independently existing thing (bhavo vyavasthita) which is earlier than those [sensory faculties] . ’29 The worry behind these questions is that the undeniable fact that there are properties of the self — since the Madhyamaka does not want to deny that seeing, feeling, tasting and so forth takes place — implies that there must be a bearer of such properties, i.e. a self. Since properties depend existentially on something that instantiates them a self must be postulated as the instantiator of all the mental properties we observe. Vatsyayana argues th a t30 desires are qualities, and qualities inhere in a substance, so th at 27See MMK 27:15-17. 28atmany asati cdtmvyam kuta eva bhavisyati. MMK 18:2a. 2Qkaiham hy avidyamanasya darsanadi bhavisyati / bhdvasya tasmat prag ebhyah so ’sti bhavo vyavasthitah. MMK 9:2. 30In the Bhasya to NS 1, 1, 5 (Nyaya-Tarkatirtha and Tarkatirtha 1985, 156:3-157:2): icchddayo gunah, gunasca dravyasamslhandh tadyadesam. sthanam sa atmeti.

which they inhere in is the self. Summarizing the Nyaya criticism voiced by Udayana in the 11th century Matilal observes th a t31 A sort of robust realism dictates that the substance or the substratum must be distinguished from the features, properties or qualities it holds. This would require a substratum for the so called mental episodes and dispositions, awareness, desires, preferences, etc., and the body, because of its continuously changing nature, cannot be regarded as adequate for such a substratumhood. However, if we take into account Nagarjuna’s distinction between constitutive and instantiating properties described on page 206 it seems possible to dissolve this worry. Nagarjuna differentiates between the property we see as constituting an individual (such as roundness in the case of a circle, treeness in the case of a tree etc.) and those properties which the individual is then taken to instantiate (such as redness in the case of the circle, and greenness in the case of the tree). As became evident in the discussion of motion where Nagarjuna introduces this distinction the difference between constitutive and instantiating properties is not regarded as bearing any ontological weight. It is rather a reflection of our epistemic priorities and practical concerns that we describe an object as a tree which is green, rather that as a green object which has the property of treeness.32 There is therefore no fundamental ontological difference between a substratum (dravya) and the qualities (guna) 31Matilal (1989, 76). See also Chakravarti (1982, 214-217, 227), Siderits (2003, 32, note b). 32Compare Siderits (2003, 26).

which inhere in it, contrary to what is assumed by the Naiyayika. When we speak of an individual having a property we nominalize the predicate expressing the property we take to be constitutive and ascribe the instantiating properties to the individual thus created. There would, however, be no deep ontological reason why we could not change our view of what the constitutive and what the instantiating properties are, and thereby describe the very same situation in terms of different individuals and properties. But if we accept this picture of ontology it is evident that we are not obliged to infer the existence of a substratum or underlying individual from the existence of a quality. Of course the Madhyamika does not deny th at there are a variety of sensory and mental events which happen in close temporal and causal connection. But our ascription of these to a single self does not commit us to the existence of such a self at the ontological level, any more than the ascription of redness to a circle commits us to the existence of an individual — the circle — and the redness it instantiates. In the same way in which we select one property, such as circularity, as constitutive and then group all the other properties around this new-found ‘individual’, in the same way we select certain properties of a causal nexus of sensory and mental events, some ‘shifting coalition of psychophysical elements’33 and group the remainder of the properties around this new-found ‘self’.34 To speak of the self and its properties in terms of substratum and quality is perfectly acceptable, as long as we do not assume that such talk is based on a distinction with an ontological grounding. 33Siderits (2003, 27). 34Dennett (1991, 228). For details of how this construction of a self might be carried out see the discussion in Siderits (2003, 43-51).

7.3 E p istem ology of th e self 239 In MMK 9:3 Nagarjuna raises the question of the origin of our knowledge of the self. He asks35 The independent thing which is earlier than seeing, hearing and so forth, and [also earlier than] feeling and so forth [i.e. the self], by which means is it known? This investigation of the epistemology of the self might strike us as curious. After all there seem to be few things which are more epistemically obvious than our own self, which seems to accompany us all the time. While there might be all sorts of problematic issues connected with the parts and properties of the self, surely the way in which we get to know it is unproblematic? The difficulty arises once we note th at in its role as a unifier of our cognitive life a substantial self is the subject of all experiences, but at the same time given the distinctness of such a self from our body and all parts of our mental life it must also be distinct from all experiences. So in order to have epistemic access to our self it must be able to function as a cognitive object. Since we assume, however, that it is not only a cognitive subject but also essentially a cognitive subject it cannot ever occupy this role — at least if we make the plausible assumption that being an object and being a subject are mutually incompatible properties.36 Now given that we do not seem to be able to acquire knowledge of the self by directing the self’s attention at it (i.e. by introspection) nor by empirical observation (due to the private 35 darsanasravanadibhyo vedanadibhya eva ca / yah prdg vyavasthito bhavah, ken a prajhapyate 'tha sah. 36Nagarjuna observes that the same epistemic difficulty of self-perception holds for vision: svam atmanam darsanam hi tat tam eva na pasyati MMK 3:2a.

nature of mental states) it appears that the only cognitive route left open to us is inference. We have to establish by a (hopefully sound) argument that the self exists. This is by no means an absurd position to hold (in fact it is just what the Naiyayika sets out to do) but it might still strike us as slightly curious th at what seems to be the most intimate object of our acquaintance has to be known by a most indirect route. We might also consider it as somewhat epistemically implausible to assume that everybody’s belief in a self is arrived at by a process of drawing inferences from a set of clues.37 One Nyaya argument for the existence of the self based on the supposed existential dependence of qualities (guna) on a substratum (dravya) has already been discussed above.38 The Madhyamika will be reluctant to accept it since he does not agree with the Nyaya ontology of individuals and properties it presupposes. Other arguments would obviously have to be dealt with on a case-by-case basis. But the Madhyamika will argue here th at in fact no such argument is needed, since it is perfectly possible to account for our self-awareness, as long as we give up the conception of a substantial self. If we conceive of the self as a temporally stretched-out compound of psychophysical events then there is no fundamental difficulty th at the same type of event turns up on the cognizing subject side on one occasion, and on the cognized object side on another. Given that there is no unified substratum constituting the self there is also no necessity for something to be essentially a subject of experience. As different parts can play different roles at different times our self-knowledge can be just explained by a momentary identification with a mental event which presently functions as a cognizing 37Siderits (2003, 20). 38Weil-known Western arguments to this effect are the Cartesian cogito and the Kantian demand for a self to unify mental events spread out in time as belonging to a single subject. For a Madhyamaka response to these two see Siderits (2003, 21-31).

subject.

7.4 The Madhyamaka view of t e self

Given that Nagarjuna rejects the picture of a substantial self described above we have to consider which alternative picture we are presented with instead. The self is obviously seen as depending on the five constituents, which rules out the assumption th at any independently existent substance could be regarded as a self.39 The emerging view of the self is characterized by two main properties. Firstly it is to be regarded as a sequence of events which stand in close temporal and causal relations. Physical processes cause sensory events, which are then framed by concepts, used as the basis of decisions, which give rise to actions, which in turn set physical processes in motion which cause new sensory events and so forth. The self is not seen as a cognitive nucleus which stays constant amidst the stream of changing sensory impressions and mental deliberations, but rather as the entire set of such sensory and mental events which are interconnected in complicated ways. In order to stress this point Nagarjuna compares the collection of constituents of the self to a lamp.40 The light of the lamp is not a persisting thing, but a process, a sequence of events one following the next which arise on the basis of the interaction of a complex set of causes, such as fuel, a wick, the presence of oxygen and so forth.41 It is for this reason th at Na- 39MMK 22:2. 40MMK 27:22. 41 In the Samyutta Nikaya 4.196-198 (Bikkhu Bodhi 2000, II: 1254) the same point is illustrated by the example of the sound of a lute, which is a process based on the parts of the lute and the skill of the player but no part to be found anywhere amongst them. See Collins (1982, 101).

garjuna claims that his analysis of fire and fuel given in chapter 10 of the MMK also explains the relation between the self and its constituents.42 This example adds the additional complexity of the mutual dependence of the self and its constituents.43 Not only does the self depend for its existence on the constituents, but the constituents only acquire their existence as distinct parts of the stream of mental and physical events by being associated with a single self, which, regarded as a constitutive property produces the basis for postulating the individual in which the various properties of the self inhere. It is precisely this reason which keeps the Madhyamika from regarding the constituents as ultimate existents (dravya) and the self as merely imputed (prajnapti), For the Madhyamika not only is there no substantial self, there is also no substantial basis on which a non-substantial self could be built. Secondly, the self is characterized by a mistaken self-awareness. This means that the self which is essentially a sequence of events does not regard itself in this way, but considers itself to be a substantial self, th at is an essentially unchanging unified agent distinct from its physical and mental properties.44 To this extent it is deluded about its real nature. Nagarjuna therefore compares the agent to an illusion (nirmita) created in a magical performance which in turn brings about another illusion.45 This construc42’ With [the investigation] of fire and fuel the way [for the solution of the problem of] the self and of grasping is completely described.1 Agnindhanabhyam vyakhydtd dtmopadanayoh kramah / sarvo niravasesena sardham. MMK 10:15. The example of the lamp is wellknown in the Buddhist philosophical tradition. It can already be found in the earliest Buddhist literature ((Trencher 1888, 486-487), (Bikkhu Nalamoli and Bikkhu Bodhi 2001, 593)) and lends itself to a further metaphorical explanation of transmigration (lighting one flame by another) and the notion of nirvana in terms of its literal meaning as the ‘blowing out of a flame1. See Collins (1982, 186-187); also compare Siderits (2003, 25-26). 43[...] indhanam apeksyagnir apeksydgnim [...] indhanam. MMK 10:8a. See also 10:9b. 44Gowans (2003, 71). 45‘As a magician creates a magical illusion by the force of magic, and the illusion proCHAPTER

tion allows Nagarjuna to reconcile his rejection of a substantial self as an essentially unchanging unifier of our mental live distinct from both its physical and mental attributes with the acceptance of the self as an agent who will experience the results of his actions, an assumption which could not be relinquished within the Buddhist world view. This is a very important point, since the identification of the self with a causally interlinked set of events might tem pt us to throw out all prudential considerations for our future selves, as well as those for other selves.46 Since none of these have any ultimate existence we might think th at all actions referring to them in some way (that is all our conscious actions) are all equally insubstantial too, so that in the ultimate analysis it does not make any difference how we act. Nagarjuna counters this view by distinguishing the view from the inside of an illusion from th at from the outside. When we are dreaming and are not aware we are doing so we understandably prefer to leave a building by using the stairs, rather than jumping out of the window. For somebody who is not dreaming, however (and also for our later, waking selves) it does not make any difference whether we jump or not since at the ultimate level (from the point of view of the awakened one) there is no fundamental difference between the two actions. This does not imply that whilst we are still under the thrall of the illusion we should leave all prudential and moral considerations behind. On the contrary, as long as we are under the influence of the illusion we have duces another illusion, in the same way the agent is a magical illusion and the action done is the illusion created by another illusion.’ yathd nirmitakam sasta niunirmtarddhisarnpada / nirrnito nirrnimitdnyam sa ca nirmitakah punah / / tatha nirmitakdkdrah kartd yat karma tat krtam / tadyatha nirmitendnyo nirrnito nirmitas tatha. MMK 17:32. This metaphor is encountered frequently both in Nagarjuna’s works as well as in later Madhyamaka literature. See YS 16-17; SS 66, VV 23, 27; RA 1:52-56; CS 7:24, BCA 9:150. 46Garfield (1995, 243-244). For a detailed discussion see Siderits (2003, chapters 3, 5, 9).

to act in accordance with its laws, even if we might suspect th at it is an illusion. Unlike in the case of dreaming, where the mere wondering whether we are dreaming sometimes allows us to see through the nature of the dream the mere suspicion th at there is no substantial self is (unfortunately) not yet a realization of the emptiness of the self. Chapter 8 E pistem ology An account of the theory of knowledge is important for Nagarjuna’s investigation for at least two reasons. First of all objects of knowledge and means of accessing them form an essential part of our conceptualization of the world and our place in it. The means of knowledge are instruments used by the self in order to apprehend objects of knowledge which connect our inner world with that of a world of outside objects. Given the centrality of these key notions of epistemology it is obviously important to investigate whether any of these could be regarded as existing with svabhava. Secondly Nagarjuna’s account of epistemology also has to provide the foundations of his own project. Given that the knowledge of the theory of universal emptiness is what Nagarjuna wants to establish it is important for him to describe the epistemology on the basis of which such knowledge is to be gained. The Indian philosophical tradition distinguishes a variety of means of knowledge (pramana) by which objects of knowledge (prameya) are epistemically accessed. Which means of knowledge were accepted and how their function is understood differs amongst different philosophical theories.1 In ISee Potter (1970-2003, II, 154-178) for a summary; also Chatterjee (1939, 53-74). 245

his discussion of epistemology Nagarjuna lists four such means of knowledge: perception {pratyaksa) , inference (anumdna), recognition of likeness (upamana), and testimony (dgama).2 Nagarjuna’s primary concern is not a discussion of the nature and interrelation of these different means of knowledge, but the question of how to establish any particular set of means of knowledge, whether it is the one just indicated or a different one. Once we have agreed th at the existence of objects of knowledge is established by the means of knowledge (as for example the existence of the desk in front of me is established by my perceptual abilities, in this case primarily non-defective vision) we then have to address the further question of how to establish the means of knowledge.3 How do we know that these means of knowledge are good guides to the objects out there in the world? There are three different ways in which we could try to establish the means of knowledge. First of all we could regard them as established by mutual coherence: perception is an adequate means of knowledge of the desk because its accuracy is established by other means of knowledge indicating its presence as well. Having the receipt of the delivery of the desk allows me to infer th at there must be a desk in my room (since if I have such a receipt the item in question must have been delivered), my perceptual recognition of the desk is in important ways like the perception of other medium-sized dry goods, such as tables and chairs, and finally I can rely on the testimony of other people who also see the desk in my room. In a similar way we could then argue for the establishment of inference by the fact that the conclusions inferred are supported by perception, likeness, and testimony, and so on for 2VV 5, VS(S) 46:15-16, 72:6-17. 3Matilal (1986, 49).

all the other means of knowledge. Secondly we conld assume th at the means of knowledge justify themselves. We do not have to go beyond perception to realize that perception usually delivers an accurate picture of the world, but perception itself presents a faithful representation of the world and of its own validity. A popular example illustrating this point is that of the lamp which illuminates other objects at the same time as illuminating itself. We do not need another lamp in order to illuminate the lamp. Finally one could regard the means of knowledge and their objects as mutually establishing each other. The means of knowledge establish an object of knowledge by giving us epistemic access to it. But we could also argue th at the object in turn establishes the means of knowledge. Given th at we manage to interact with the objects of knowledge more or less successfully (as confirmed by the evolutionary success of our species) there must be something amongst our cognitive means which gives us a relatively accurate account of the way things are. In this way epistemic success allows us to establish the means of knowledge via the objects successfully cognized. Nagarjuna does not devote a great deal of discussion of to the first alternative, the establishment of the means of knowledge by mutual coherence.4 This seems sensible, for even if the argument succeeds the kind of establishment of the means of knowledge which can be derived from it is not exactly what Nagarjuna’s Naiyayika opponent is looking for.5 He is trying to argue that the means of knowledge provide us with information about the nature of independently existing reals. But it is clear th at the mutual establishment of means of knowledge can do no such thing. All it can do is establish the 4m [,..] prasiddhih [...] parasparatah [...] bhavati [,..] pramandndm. VY 51. 5For the relationship between Nagarj ima’s Madhyamaka and Nyaya with a special focus on epistemology see Oberhammer (1963), Bhattacharya (1977), Bronlchorst (1985).

CHAPTER 8. EPISTEMOLOGY

coherence of statements arrived at by different means of knowledge. But the mere coherence of some set of statements is not sufficient for showing that there is anything with an independent existential status which they describe. There are, after all, coherent fairy-tales.6 Let us therefore now consider the remaining two possible ways of establishing the means of knowledge.

8.1 Means of knowledge as self-established

Regarding the means of knowledge as self-established has the immediate advantage of avoiding two difficulties. Firstly we get around the infinite regress of establishing the means of knowledge by other means of knowledge, which then in turn need yet other means of knowledge to establish them, and so forth.8 Unlike other forms of infinite regress which Nagarjuna accepts (such as an infinitely extended chain of causes and conditions) this regress is vicious, since the burden of proof is transferred in its entirety to the preceding stage, as a preceding means of knowledge would have to establish all the succeeding ones.9 Secondly the self-establishment of the means of knowledge allows the op6I do not think that the problem of an infinite regress (as argued by Matilal (1986, 56)) is the main difficulty with the establishment of means of knowledge by mutual coherence. 7na [...] svatah prasiddhih [...] bhavati [...] pramdndndm. VV 51. See Matilal (1986, 51-53). 8VV 32, 51, VP 5, Siderits (1980, 310-312), Matilal (1986, 50). Compare NS 2, 1, 17. 9To argue that the regress could just be stopped after a finite number of steps, after which the correctness of the means of knowledge is established as ‘highly probable’ (as is done by Burton (1999, 159), following the arguments presented in NS 2, 1, 8-20) confuses the pragmatic question of how our epistemic enterprise should proceed with the philosophical question of its justification. See also Siderits (1980, 331), Siderits (2003, 141).

ponent to hold on to the assumption that everything lmowable is established by the means of knowledge.10 It might be attractive to give up this assumption in order to escape the vicious regress, but this then makes it necessary to give a special reason (visesahetu) explaining why ordinary objects are established by means of knowledge, but not the means of knowledge themselves.11 8.1.1 Means of knowledge compared to fire In support of the self-establishment of the means of knowledge we are presented with the following example:12 Fire illuminates itself as well as other objects. In the same way the means of knowledge establish themselves as well as other objects. This argument is based on an recognition of likeness (upamana).13 Because the means of knowledge are like the fire, to the extent to which fire illuminates objects in the dark and thus brings them to our attention, in the same way the means of knowledge retrieve objects from the darkness of ignorance.14 10 don thams cad tshad mas bsgrub par bya ba yin no. VP(S) 23:15; see also VV(S) 63:6-7 31, 64:11-13. 11See NS 2, 1, 18. 12 dyotayati svdtmanam yatha hutdsas tatha pardtmdnam / svapardtmanavevam prasadhayanti pramanani VV(S) 64:18-19. See also VP 6, MMK 7:8-12, NS 2, 1, ID. 13Chatterjee (1939, 325-342), Potter (1970-2003,11:174-176), Matilal (1986, 57-58). 14Some references supporting the close connection between illumination and cognition in Indian thought are given by Burton (1999, 163-164). He also offers a different reading of this argument, claiming that like an illuminated object manifests the existence of light, the existence of a known object manifests the existence of a means of knowledge (161). This, however, appears to conflate this argument with the establishment of means of knowledge by their objects discussed separately below. This reading is also not very satisfactory from a hermeneutical perspective, given that it lets all of Nagarjuna’s arguments dealing with self-illumination come out as very problematic (as Burton sets out to argue on pages

Now it would be absurd to suggest that there is a vicious regress involved in the illumination of the fire, that is if someone argued as follows: ‘Because we can see the object, it must be illuminated by something. It is illuminated by the fire. But we can see the fire too. So something must illuminate it. So there must be a second fire, which is either invisible or visible. But how can it be invisible, since it illuminates a visible object (namely the fire)? So it must be visible. But then we need a third fire to illuminate the second fire, and so forth.’ It is clear th at the error occurs by assuming that there must be a different fire illuminating the fire: a fire can illuminate both itself and other things.Therefore, given the similarity of fire and means of knowledge and thereby of the relations of illumination and establishment the means of knowledge can both establish themselves as well as other things.15 Nagarjuna tries to counter the use of the example of fire to demonstrate the means of knowledge as self-established by arguing for two claims: • Fire does not illuminate other objects. • Fire does not illuminate itself. Note th at the establishment of either of these theses is sufficient for refuting the opponent, as each one would demonstrate that an argument by 165-172). 15We might want to note the similarity of this argument with the ‘glue’ objection to Bradley’s regress. This regress occurs once we think that what unifies two constituents of a state of affairs (such as an individual and a property) is the instantiation relation holding between them, and that this instantiation relation has a distinct ontological status. For then we need a further relation to connect the instantiation relation with the individual and the property and so forth. To see what goes wrong here we can argue that when glueing two things together we do not require superglue to first glue the glue to the objects, and then super-superglue to glue the superglue to the glue and so forth. It is the glue itself which can connect to the objects, as well as connecting the objects themselves.

the recognition of likeness between the means of knowledge and fire cannot be used. Nagarjuna sets out to establish both claims. In order to argue for the first claim he observes th at in order to illuminate an object, a fire has to illuminate the darkness concealing that object. In order to do so, fire and darkness must come into causal contact, so th at one can remove the other. This, however is impossible: A lamp cannot illuminate when it is connected with darkness since their connection does not exist. Why are the lamp and darkness not connected? Because they are opposed. Where the lamp is, darkness is not. How can the lamp remove or illuminate darkness?16 The point Nagarjuna wants to make here is that darkness and light cannot be understood as two independently existent objects one of which acts on the other when they come into contact.17 The illumination of darkness by light is not analogous to the dissolution of salt by water. This is because darkness is the mere absence of light,18 it is wherever light is not. For this reason the two can never come into spatial contact. 16re zhig m ar me ni mun pa dang phrad nas gsal bar byed par mi ’gyur te phrad pa med p a ’i phyir ro \ gang gi phyir mar me dang mun pa ni phrad pa yod pa ma yin te \ ’gal ba'i phyir ro \ gang na mar me yod pa de na mun pa med na ji Itar mar me ’di mun pa sel bar byed p a ’am gsal bar byed par ’gyur. VP(S) 24:2-8. See also VV 38, MMK 7:10. 17Another example where Nagarjuna rejects this assumption (even though we are here dealing with mutually dependent rather than with mutually exclusive entities) is the case of fire and fuel. Nagarjuna argues that we cannot conceive of these as two distinct entities which produce an effect (heat and illumination) when put together, in the same way in which the union of man and woman produces an effect, i.e. a child (MMK 10: 1, 6. See also Garfield (1995, 191—192)). This is because fire is existentially dependent on fuel, while fuel is at least notionally dependent on fire. 18 ’od med pa ni mun p a ’o. VP(S) 25:10. See also Burton (1999, 71-72).

Nagarjuna’s opponent argues in VV 38 that light and darkness might coexist during the process of origination (utpadyamana) of light. So light and darkness would be both present at the same time for a single moment, and then the light started to act on the darkness in order to remove it. But this leaves us then with the problem of explaining what causes light to remove the darkness in its second moment of existence but not in the first. If light does not have the causal power to remove darkness in the moment of its origination, how could it have this power later on?19 On the other hand it would be highly unsatisfactory to regard light and darkness as independently existing objects which interact without coming into contact. For if light could act on darkness at a distance without spatial contiguity (as the planets were seen to act on human beings in ancient India,20 and as we now know such forces as gravity and magnetism to work) without influencing it causally it is difficult to explain why for example a certain lamp can only dispel a certain darkness (namely the one in this room) but not other ones (such as the darkness in the room next door).21 The refutation of the second claim (that fire illuminates itself) proceeds by analysis of the notion of illumination. For something to be illuminated it must first exist hidden in darkness and subsequently made visible by light shining on it.22 But it is obvious that this is not true of the fire: it does not first exist hidden away in darkness, like a pot in a dark room, and is then made visible by shining its own light onto itself.23 If we conceive of 19See Siderits (1980, 314). 20VP 8, compare the discussion in Burton (1999, 178-179, note 27). See also Aryadeva’s Satasastra (Tucci 1929, 9). 21MMK 7:11, VP 8, VV 39, Tucci (1929, 9). See also Burton (1999, 169-171). 22VV(S)65:3-4. See also Burton (1999, 166). 23Siderits (1980, 313).

illumination as the prevention of darkness (tamasah pratigatah)24 we would have to assume th at darkness is somewhere in the fire or encapsulating it to hide it from view.25 But this would mean that we are again thinking of darkness as an independently existing substance preventing illumination, something like a thick lampshade which prevents the light from reaching our eyes. But as Nagarjuna has argued above this view of darkness as a substance is thoroughly unsatisfactory, since darkness is nothing but the absence of light. Nagarjuna also adds a second argument against the self-illumination of fire. He first claims th at since the illumination by fire proceeds by the consumption of fuel, self-illumination would entail self-consumption, that is, fire would burn itself as well as its fuel.26 Of course we cannot argue th at if one quality of an objects acts on itself, any other of its qualities will do so as well.27 An oven may heat a piece of wax as well as itself, and melt the wax, but not melt itself. Nagarjuna must therefore regard the fire’s burning of fuel and its illumination of objects as the very same process. But we do not have to say that the fire burns down because there is less fuel (it being gradually consumed by the fire) and less fire (because it gradually consumes itself). Fire is just the burning of the fuel rather than some distinct entity acting both on the fuel and also on itself.28 Therefore, if there is no self-consumption of fire there should also be no self-illumination.29 24VV(S) 66:10. 25VP 10. See also BCA 9:18. 26VV 35. 27As pointed out by Burton (1999, 167). 28MMK 10:1. 29It is interesting to note that the Madhyamaka argument against the Yogacara view of the mind as self-illuminating (svaprakasa) given by Santideva in the BCA is another l'eason by recognition of likeness (upamana): that of a sword which cannot cut itself (9:17). The Yogacara view is also criticized by Sankara in the UpadesasahasrT 16:13, the BrahmasutraCHAPTER

Secondly, if fire illuminated itself, and perception was to be understood as structurally similar to illumination, the different modes of perception should also be able to perceive themselves: seeing should be able to see itself, hearing hear itself, and so on.30 Vision would be possible in the absence of any distinct object, as vision could act as its own object of sight. Visual perception would then not be regarded as an intrinsically relational phenomenon, but as solely based on some essentially perceptive quality of vision. As such a non-relational understanding of visual perception (and of any other kind of perception as well) is unsatisfactory Nagarjuna argues that if the parallel between perception and illumination is maintained, fire cannot be seen as self-illuminating. Thirdly, since darkness is the opposite of light, if light illuminates itself, darkness should conceal itself.31 But then we should not be able to perceive darkness, as we are not able to perceive a pot concealed in darkness. Since darkness can be perceived, however, it is not self-obscuring and therefore, Nagarjuna argues, fire cannot be self-illuminating either.32 bhasya 2:2:28 and in his commentary on the Brhadaranyaka Upanisad 4:3:7. For further references to this ‘anti-reflexivity principle’ see Siderits (2003, 32, note a). It is important to note, however, that even though the example of the lamp or fire is used in the discussion of both, the question of whether the mind is self-illuminating and the question whether the means of knowledge are self-established are distinct. See Siderits (1980, 334-335, note 4), Burton (1999, 155-156). 30Nagarjuna only mentions the case of seeing seeing itself (MMK 3:2). This reading of the verse is supported by a variety of commentaries, such as PP 114:1-5, Tsong kha pa Bio bzang grags pa (2006, 130), Garfield (1995, 137-138) and Weber-Brosamer and Back (1997, 15). For a different reading see Kalupahana (1991, 133-134). See also the further references given in May (1959, 79, note 135). 31MMK 7:12, VV 36, VP 11. 32Some more discussion of this argument is in Burton (1999, 167-168).

8.1.2 The problem of the independence of means of knowledge from their objects Apart from attem pting to find fault with the opponent’s example based on the supposed self-illumination of fire Nagarjuna also presents us with positive arguments against the self-establishment of the means of knowledge. First of all, if a means of knowledge, such as visual perception, was self-established it should be able to exist independently of the existence of an object of vision.33 But if we then assume that it is an essential property of visual perception to see34 visual perception must be able to function as its own object, as otherwise there might be no other object to be seen. This, Nagarjuna claims, then leads to the same problem encountered in the analysis of motion.35 As the mover and the place being moved over cannot exist simultaneously, since motion takes time, vision cannot see something which exists simultaneously with it (such as itself), since vision takes time too.36 Secondly, observing again that if the means of knowledge are self-established they will be established independently of the objects known37 Nagarjuna argues as follows. Assume that we wanted to chose those amongst all the different means of cognitive access to the world which deliver accurate knowledge of the nature of the objects known, that is, which qualify as means of knowledge. We would select all those which have a specific internal quality (such as the cognitive equivalent to self-illumination). The possession of this quality would then guarantee that its possessor delivered accurate information about the nature of the objects cognized. But how is the connection between the 33MMK 3:2. 34As stated e.g. in Vimalamitra’s Vibhdsa-prabhdvrtti (Kalupahana 1991, 133). 35MMK 3:3, 36See Garfield (1995, 138-139). 37VV 40. See Siderits (1980, 314-15).

specific internal quality and the correct representation of the object justified? After all there are all sorts of properties our means of cognitive access to the world can have, so how do we know that a specific one is a guide to accurate representation? Suppose we are presented with a set of fancy mechanical devices and are asked to select the five best tin openers from these. No detailed study of the internal properties of each will allow us to accomplish that task. We have to analyse each in relation to a tin and try to determine the way in which it might open it. Only then would we be able to conclude which particular properties of the mechanisms are correlated with good tin-opening abilities. In the same way we can only regard an internal quality of a way of accessing the world as a characteristic of a means of knowledge once we have assessed it in relation to the objects cognized. Only then can we conclude that this particular property really leads us to the knowledge of the nature of the object, rather than doing something else. But in this case the establishment of the means of knowledge can no longer be regarded as self-establishment since it incorporates reference to other objects (namely the objects known) at an essential place.

8.2 Means of know ledge and their objects asmutually established

If the argument for the self-establishment of the means of knowledge is not successful, the remaining option is to argue that the means of knowledge and the objects of knowledge mutually establish one another.38 Assume th at I see an apple on the table. The existence of the apple, the object of knowledge 38VV 42-51.

is established by the means of knowledge which is perception. But we could equally argue the other way round: that the object known establishes the means of knowledge. This would invite the immediate objection th at we then need prior cognitive access to the object known, and if we have this we must already have established the means of knowledge.39 We are therefore pointlessly establishing it twice.40 But if we somehow gain this access without relying on the means of knowledge the whole project of establishing these means seems futile, since it is precisely the justification of our means of gaining knowledge of the world which we have set out to scrutinize.41 We will therefore need a different argumentative strategy to argue for the mutual establishment of objects known and means of knowledge, and in particular for the establishment of the latter by the former. One way of going about this (which does not commit us to the viciously circular mutual establishment criticized by Nagarjuna)42 is to argue that because the object of knowledge is perceived, there must be something bringing about such a perception, and this is the means of knowledge. In this case the apple establishes the existence of the means of knowledge by which it is known 43 An essential prerequisite for this latter direction of establishment is of course success. Because we successfully apprehend an apple our means of apprehension is regarded as a means of knowledge.44 If we were susceptible to frequent apple-hallucinations which disappeared once we tried to touch them we would not regard perception as a reliable apple-detector, that is, as 39Matilal (1986, 56-57), Burton (1999, 183). 40This is the fallacy of ‘proving the proven’ (siddhasya sadhanam). See VV 42, 41VV 44. 42VV 46-48. 43This way of establishing the means of knowledge is what Nagarjuna criticizes in VV 51 by saying prasiddhih [...] bhavati na ca prameyaih [...] pramdnandm,. 44VP 17. See the commentary in Tola and Dragonetti (1995a, 111-112).

a means of knowledge. But since we are generally successful in our cognitive interactions with the world and normally only perceive the existence of apples which are indeed there the very fact th at we successfully apprehend a world of outside objects serves as an argument for regarding the successful means of apprehension as means of knowledge.45 An immediate difficulty with this procedure is that we also need a means of knowledge for establishing the success of our cognitive actions,46 that is to ascertain whether we really are perceiving the apple or just an applehallucination. Nothing seems to rule out that some means of knowledge first deceives us about what we see, and later deceives us about the outcome of whatever procedure we use to establish whether the first cognition was successful.47 But this need not rule out any attem pts of mutually establishing the means and the objects of knowledge if we do not use epistemically suspect procedures (which we know to have lead to unsuccessful cognitions in the past) to establish the success of our cognitive actions.48 A more worrying question is whether the mutual establishment of means and objects of knowledge ~~ if successful - actually delivers the account of means of knowledge Nagarjuna’s opponent wants to defend. In order to see this we have to note first that the notion of ‘successful cognitive apprehension’ referred to above cannot just be an act of cognition which leads to a successful action, as many of our cognitions (and many of the beliefs subsequently acquired) are never acted upon. We therefore also have to include coherence with other cognitions or beliefs as a criterion for the success of some means 45 For more discussion of the Nyaya criterion of pragmatic success as an indication of valid knowledge see Chatterjee (1939, 81-89), D ’Almeida (1973, 46-62), Bijalwan (1977, 53-60) Matilal (1986, 160-179). 46Siderits (2003, 140-141). 47Burton (1999, 184-185). 48Siderits (1980, 317).

of knowledge as well. Our cognition of the apple on the table might therefore be deemed successful if it either leads to a successful action (we reach out, grasp the apple, and eat it) or if it coheres with other means of knowledge (for example with my memory of buying a bag of apples and putting them on the table). However, the difficulty with employing coherence in this way is that we have to select a certain set of cognitions or beliefs which we hold fixed, so that we can then evaluate the status of other cognitions relative to them. One problem now is of course how to assure the accuracy of this selected set: if they are not accurate themselves coherence with them has very little weight. But let us assume for the sake of argument th at they are indeed accurate and constitute an epistemological fixed point relative to which other means of knowledge could be justified. The remaining worry is that Nagarjuna’s Naiyayika opponent wants to establish the means of knowledge as something which gives us cognitive access to a world of independently existing reals. But as long as we do not know the initial set to be accurate (even though it may be) it is hard to see why coherence with the selected set should provide us with such access. Since we cannot establish the accuracy of the initial set without circularly assuming th at we have already established some means of knowledge the possibility remains that a different selected set would have provided us with a different set of means of knowledge coherent with it. As Mark Siderits observes: ‘since at no point in our proof is there appeal to any facts other than those concerning logical relations among cognitions, we cannot legitimately include in the resultant theory of the pramanas the claim that they yield direct knowledge of their objects’.49 There is no explanation why coherence with the selected set should assure us that the means of knowledge indeed ‘reach 49(1980, 318),

out’ to provide us with knowledge of an independently existent world of objects as long as we do not have an independent way of establishing the accuracy of the selected set.50 But if this could be done the entire attem pt of mutually establishing means and objects of cognition would be superfluous. The argument against the mutual establishment of means and objects of knowledge just presented is very much a rational extrapolation: it is what Nagarjuna (as well as a Madhyamika more generally) should say in response, even though we do not find such a detailed argument in Nagarj una’s texts. The argument in the VV fundamentally boils down to the observation that the mutual establishment of the means and objects of knowledge excludes the possibility of either existing by substance-svabhava, i.e. independently of any other object. If father and son were mutually established in the same way as means of knowledge and their objects, Nagarjuna argues, father and son would not be distinguished by any substantial difference.51 This is a familiar argumentative move we encounter repeatedly in Nagarjuna’s works. A different and more unusual response to the proposal of mutual establishment is given in the VP. Nagarjuna observes that:52 Potness is perceptible, but the pot is not. That which is the object of the sense-faculty (dbang po la rab tu phyogs pa’i don), that is perceived. If we apply the term (brda) ‘visual sense-faculty’ then the object [of that faculty] is perceptible and depends on 50See Siderits (1989, 237-238). 51VV 49-50. Compare BCA 9:12-14. 52bum pa nyid mngon sum yin gyi bum pa ni ma yin te dbang po la rab tu phyogs p a ’i don gang yin pa de ni mngon sum yin par byas nas [ mig gi dbang po la dbang p o ’i brda by as la de la rab tu phyogs p a ’i don gang yin pa de ni mngon sum yin zhing de yang snang ba la sogs pa ’i rkyen la Itos pa ste \ de phyir bum pa la sogs pa nyid mngon sum du yongs su grub pa yin. VP(S) 29:6-13.

conditions like light and so forth. Thus pot and so forth are established as perceptible. The idea behind this argument is to deny the establishment of the means of knowledge by the object known by denying that the object of knowledge (the prameya) is an external, independently existent real. The Naiyayika opponent wants to argue that since our cognitions are generally successful, and since this success serves as an indication of the existence of an external, mind-independent reality the means of knowledge are just those things which allow us access to this reality. Nagarjuna replies that in fact the object perceived by our sense-faculties is not the external object, but a mental representation of the object. Nagarjuna here embraces a representative theory of perception which does not assert (as a naive realist would) th at we have direct and unmediated cognitive access to the objects of the external world.53 W hat is directly perceived is the sense-datum (what Nagarjuna calls the object of the sense-faculty (dbang po la rab tu phyogs pa’i don)), on the basis of which information about the external object is inferred. If, for example, we look at a white disc under red light at an angle, our sense-datum will be a red ellipse, though, knowing some basic facts about vision and about the peculiar &3Burton (1999, 192) ‘suspects that Nagarjuna actually means that [...] the knowledgeepisode itself is constitutive of the object known’. This is indeed the case. Burton is also correct in claiming that there is not much of an argument for this position in Nagarjuna’s texts. But given the generally elliptic nature of the VP this philological gap does not necessarily entail the existence of a gap in Nagarjuna’s philosophical argumentation as well. Tola and Dragonetti (1995a, 12) suggest that the Yogacara flavour of VP 18 could be interpreted as evidence for locating the composition of the VP after the appearance of the Yogacara school about 350 A.D., which would speak against its attribution to Nagarjuna. Given the somewhat isolated nature of VP 18 in Nagarjuna’s argument this suggestion does not seem to me to be able to bear much weight.

lighting conditions, we perceive a white disc. Nagarjuna stresses the contribution of inference to knowledge gained by perception.54 He notes that in the same way in which the inferential process is based on some perception in order to arrive at some piece of non-perceptual knowledge, for example when we infer fire from smoke55 there is an inferential process at work in ordinary perception as well, which, on the basis of the object of direct acquaintance (the sense-datum) subsequently produces the object of perception.56 But it is evident that for the proponent of a representative theory of perception an object of knowledge cannot serve as establishing a means of knowledge in the way the Naiyayika requires. Since all we are directly acquainted with is the sense-datum we cannot use this acquaintance to support the view that there are epistemic processes which give us direct access to a world of external, independently existent reals.

8.3 Temporal relations betw een eans and objects of knowledge

Apart from the question of how the means of knowledge are established Nagarjuna also investigates the question of how they are related to the objects of knowledge. The two stand in a causal relation. In the case of perception, 54See Matilal (1986, ch 8). 55‘In the same way [in the case of inferential knowledge] based on the connection between fire and smoke there is an inference preceded by a perception.’ de bzhin du me dang du ba la ‘brel pa las mngon sum snon du ’gro ba can rjes su dpag pa yin no. VP(S) 29:14-16. 56The role of inference (or, as contemporary cognitive science would prefer to put it: the implicit reliance on biologically hard-wired rules in interpreting perceptual input) in the formation of visual perception in particular is now well-supported by empirical research. See Hoffman (1998) for an accessible summary.

for example the contact between the sensory faculty and the object perceived bring about the object of knowledge which for Nagarjuna is not an external object, but a sense-datum.57 It is therefore hardly surprising th at Nagarjuna sees the relation between means of knowledge and object of knowledge as facing the same difficulties as those of other causes and effects. He concentrates specifically on the problematic nature of their temporal relation.58 If the means of knowledge exists before its object, there is no justification for calling it a means of knowledge, as Nagarjuna argued earlier th at being such a means cannot be a purely intrinsic feature of some cognitive way of accessing the world.59 A different interpretation of this difficulty is found in Vatsyayana’s Bhasya on the NS 2.1.9.60 Here the means of knowledge is not regarded as the perceptual faculty, but as a specific act of perception, and the object of knowledge as the object in the world which, when in contact with the sense-organ produces the perception.61 The prior existence of the means of knowledge would then be equivalent to the existence of the effect before the cause.62 57See VV(S) 70:17-18 where the means of knowledge is described as the cause (karana) of the object of knowledge. 58VP 12. See also the discussion in Burton (1999, 191-199), Siderits (2003, 146). 59See also Burton (1999, 172-174). 60 NS 443-445. Bronlchorst (1985, 107-111) argues that the discussion of the arguments concerning the temporal relations between means and objects of knowledge in NS 2, 1, 8-15 is evidence of Sarvastivadin influence rather than an anticipation of Madhyamaka arguments, as claimed by Phanibhusana (Chattopadhyaya and Gangopadhyaya 1968, 21- 22 ). 61In accordance with the definition of perception in NS 1, 1, 4, 62 While it is instructive to compare the explanation of the difficulties connected with the temporal relations between means of knowledge and their objects in the commentaries on NS 2, 1, 8-15 and Madhyamaka treatises it is important to be aware of the significance of their respective background assumptions. The Nyaya theory of means of knowledge, based on Vaisesika metaphysics, regards a particular lcnowledge-episode, such as an instance of

If it existed after the knowable object there is no justification for calling the object of knowledge an object of knowledge, as there is not anything yet by which it is known.63 Nagarjuna also identifies another difficulty when he argues that an arisen and a non-arisen thing cannot abide together.64 If two things are such th at one exists only now, the other only at a later moment it cannot be the case that the second has any effect on the first, such as making it known. The final possibility is that the means of knowledge and its object exist simultaneously.65 The difficulty is here that for two simultaneously existing things (such as the two horns of a cow, which Nagarjuna gives as an example, but also e.g. the two ends of a see-saw moving in opposite directions) it becomes problematic to establish which is the cause and which is the effect. For Nagarjuna the means of knowledge brings about the object of knowledge, perception, as caused by the combination of an externally existing object of knowledge and the respective sense-faculty. For Nagarjuna, on the other hand, the object of knowledge (prameya) is no external but an internal object, a sense-datum. If we now regard the means of knowledge (pramana) as bringing about the internal representation which is the prameya we realize that the causal relation between pramana and prameya is seen in different ways by the Naiyayika and by Nagarjuna. For the former the prameya is causally prior, for the latter the pramana. 63For the Naiyayika this dependence is purely notional (Jha 1984, 609, note *), whereas for Nagarjuna it is both notional and existential, as argued in our discussion of causation on page 151. 64ma skyes pa dang skyes pa dag lhan cig mi gnas pa. VP(S) 28:24-25. 65In his commentary on NS 2, 1, 11 (NS 421—424)Vatsyayana argues that the problem is that if means and object of knowledge existed at the same time there could be no sequence of cognitions. The point seems to be that if, for example, we hold a pot in our hand and thereby have both a visual and a tactile perception of it the optical and haptic properties of the pot exist at the same time — and so should the corresponding perceptions. In our consciousness, however, they occur as successive, and indeed this non-simultaneity occupies an important place in the Nyaya theory of mind (see NS 1, 1, 16).

the sense-datum. As such the means of knowledge is a cause, the object of knowledge its effect. But in this case they would better not exist simultaneously, as it is the temporal ordering which characterizes one item as a cause, the other as an effect.66 One potential way of dealing with the difficulty of the temporal relation between means and object of knowledge is outlined by Vatsyayana in his commentary on NS 2, 1, l l . 67He claims that the term pramana is to be applied to something which has been the cause of apprehension of an object in the past, is so at present, or will be in the future. Similarly, the term prameya is to be used for an object th at was apprehended in the past, is presently being apprehended or will be apprehended in the future. Only in this way can we make sense of somebody saying ‘bring the cook to do the cooking’, as he has not cooked yet — else the term cook would fail to refer. In this case we use ‘cook’ just as a synonym for ‘whoever will do the cooking’. The problem with this reply is that in this case it is obvious that being a cook is not regarded as an essential property of the person referred to. As statements about the future are contingent the expression ‘whoever will do the cooking’ must be able to refer even if it turns out that person designated does not cook in the end. But now it is evident that someone who takes ‘cause’ and ‘effect’ to denote essential properties of things — as Nagarjuna’s Naiyayika opponent does — cannot help himself to this reply. As a statement referring to a cause or a means of knowledge might really be talking about what this thing is likely to do in the future it must be possible th at the thing 66I disagree with Burton’s claim (1999, 193) that Nagarjuna ‘provides no justification [...] for his unusual and far from self-evident assertion’, as this matter is extensively discussed in those passages where Nagarjuna deals specifically with causation (rather than with causal issues in epistemology). See chapter 5, section 5.4.3. 67NS 421-424.

picked out does not cause anything, or does not give us access to an object of knowledge — else its intelligibility now would depend on what happens in the future. Since this is not the case (we know what ‘what is going to be a means of knowledge in the future’ etc. refers to, and we do not know what is going to happen in the future) being a cause, or a means of knowledge cannot be an essential property of the thing, as it could lose this property whilst remaining that very thing.

8.4 The aim of Nagarjuna’s arguments

Nagarjuna’s aim in his discussion of epistemology is not to argue th at means of knowledge and their objects do not exist at all.68 Such an argument immediately generates a paradox, for the non-existence of the means and objects of knowledge is itself an object of knowledge arrived at by some means. But then there must be at least one means and object of knowledge, which was previously denied.69 W hat Nagarjuna sets out to do is refuting the existence of either means or objects of knowledge with substance-st/ab/iaya.70 He thus wants to show th at there are no procedures which are intrinsically and essentially means of knowledge, nor are there independently existent reals they give us knowledge of. Means of knowledge and their objects are notionally interdependent: without its ability to give us epistemic access to some thing we would not label a cognitive procedure a means of knowledge. Similarly something cannot be called an object of knowledge unless there exist a means which allows us to know it. We can also argue th at the two are existentially dependent 68As claimed by Burton (1999, 194, 198). 69VP 13. 70MMK 3:5-6, VP 3-4.

on one another. The existence of certain objects of knowledge allows us to divide off certain types of our cognitive procedures and label them as means of knowledge; it is not the case that this division proceeds along certain fissures which had been there all along. Since our consciousness is a continuous flow without ready-made fissures the objects of knowledge can be regarded as bringing the means of knowledge into existence. Similarly if the object of knowledge is to be identified with a mental representation (as Nagarjuna argued above) it is evident that the means of knowledge brings this object about by unifying information received through the different sensory modalities. Means of knowledge and their objects therefore cannot be regarded as distinct substances. This point is also stressed by Nagarjuna’s argument th at each can assume another’s role: a means of knowledge can be an object of knowledge and vice versa. There are two different ways to argue for this position. First of all,71 building on the familiar Nyaya definition th at a means of knowledge is what produces the knowledge of something72 we can argue that since the object of knowledge is what brings the means of knowledge about, it is an essential part of what produces the knowledge of something, and therefore an essential part of a means of knowledge. A similar argument can be applied to means of knowledge.73 Alternatively we could argue that a means of knowledge at one time can be an object of knowledge at another time, and vice versa. For example, when we establish a certain cognition as correct this means of knowledge is an object of knowledge. And what is thus known to be correct and is an object of knowledge can in turn be a means of knowledge for knowing other 71VP 2. 72See Vatsyayana’s Bhasya on NS 2, 1, 11 (NS 421-424). 73See Burton (1999, 177).

objects of knowledge later on. Vatsyayana illustrates this with the example of a piece of gold which can be both the object tested (if we want to determine how heavy it is) or a testing object itself (if we want to check the accuracy of a pair of scales).74 In fact, as Nagarjuna points out,75 the Naiyayikas themselves count ideas (buddhi, bio) as both a means of knowledge and as and object of knowledge.76 In either way it becomes evident that means of knowledge and their objects cannot exist as essentially different entities. It is important to realize th at Nagarjuna’s rejection of the essentially existing means and objects of knowledge is presented within the discussion of the knowability of emptiness. His opponent argues th at if all things are empty, means and objects of knowledge are empty too.77 But if ultimately there are no objects of knowledge, emptiness cannot be an object of knowledge. And if there are ultimately no means of knowledge, if nothing has the intrinsic nature which is characteristic of a means of knowledge, emptiness, even if it obtained, could never be known. There appears thus to be a fundamental inconsistency in the Madhyamaka project of establishing the truth of emptiness.78 Nagarjuna’s response to this difficulty is to investigate the realist’s way of accounting for means and objects of knowledge with characteristic natures as means and objects, in order to demonstrate that all possible ways of establishing them fail. This sequence of arguments, which has been discussed above is not so much to be seen as a tu quoque move on the side of Nagarjuna79 but as an assessment of the realist’s epistemological position which 74In the commentary on NS 2, 1, 16 (NS 433-440). See also (Bhattacharya 1977, 268). 75VP 20. 76NS 1, 1, 9. 77VV 5-6. 78Siderits (2003, 140) refers to this as the ‘self-stultification objection’. 79Siderits (2003, 147).

lets emptiness come out as unknowable. For the realist means and objects of knowledge have intrinsic characteristics, and there are invariant relations of epistemic priority, i.e. cognitive procedures which are means of knowledge in all possible contexts. On this account of epistemology it is indeed impossible to establish emptiness. But as we have seen earlier in this chapter Nagarjuna tries to argue th at this is not the right account of epistemology. Even though there are no means of knowledge which are intrinsically such, delivering knowledge in every context, there are still cognitive procedures which function as means of knowledge in the specific context in which they are employed, regimented by certain background constraints and other pragmatic features. By using these procedures (which, Nagarjuna argues are all the means of knowledge there are anyway) we can achieve knowledge of emptiness even though ultimately there are neither means of knowledge nor objects of knowledge. Certain procedures can still count conventionally as means of knowledge within the framework of certain aims and directions of inquiry. Nagarjuna’s arguments about epistemology have therefore to be seen as fulfilling two purposes. Firstly they continue his general project of examining different types of objects one by one and arguing that none have substancesvabhava by considering means of knowledge and their objects. Secondly, and more specifically, they establish the necessary background epistemology needed for understanding how emptiness could in fact be known. It is this second aim which is particulary interesting, since it provides us with the outlines of a specific Madhyamaka theory of knowledge.80 S0For a good assessment of this and its relation to the debate about anti-realism see Siderits (2003, ch. 7).

Chapter 9 Language

As opposed to topics like causation, motion, the self, or the theory of knowledge, language is not given much explicit discussion in Nagarjuna’s works. This does not mean that such matters were not important to Nagarjuna but merely that his extant writings do not contain an extended connected discussion of the impact of his theory of emptiness on our view of language. Nevertheless it is possible to extract some of Nagarjuna’s views on this philosophically highly interesting issue from remarks found at different places in his works.

9.1 Nagarjuna’s view of language and te ‘no the sis’ view

A good starting-point for the discussion of Nagarjuna’s conception of how the theory of emptiness affects our view of language is his so-called ‘nothesis’ view. This is without a doubt one of the most immediately puzzling philosophical features of Nagarjuna’s thought and also largely responsible for ascribing to him either sceptical or mystical leanings (or indeed both). The

locus classicus for this view is found in verse 29 of the VV: If I had some thesis the defect [just mentioned] would as a consequence attach to me. But I have no thesis, so this defect is not applicable to me.1 That this absence of a thesis is to be regarded as a positive feature is stressed in YS 50, where Nagarjuna remarks about the Buddhas: For these great beings there is no position, no dispute. How could there be another’s [opposing] position for those who have no position?2 Now it is important to observe that when considered in isolation it is very hard to make any coherent sense of these passages. For even if we assume th at the Buddhas do not hold any philosophical position anymore (having perhaps passed beyond all conceptual thinking), how are we to make sense of the first quotation which, in the middle of a work full of philosophical theses claims that there is no such thesis asserted at all? In fact this first statement is even more difficult to interpret than the famous last sentence of Wittgenstein’s Tractatus, which is preceded by the equally famous ladder-metaphor.3 Although Wittgenstein here denies that 1yadi kdcana pratijna sydn me tata esa me bhaved dosah / ndsti ca mama pratijna tasman naivasti me dosah. 2che ba’i bdag nyid can de dag / m am s la phyogs med rtsod pa med / gang m am s la ni phyogs med pa / de la gzhan phyogs ga la yod. 36.54 My propositions serve as elucidations in the following way: anyone who understands me eventually realizes them as nonsensical, when he has used them — as steps — to climb up beyond them. (He must, so to speak, throw away the ladder after he has climbed up it.)[...]

7. Whereof one cannot speak, thereof one must be silent.

his preceding statements are of anything but instrumental value, that they turn out to be nonsensical after they have fulfilled their instrumental role and that there is something outside of the grasp of these statements, at least he does not deny making any statements at all!

9.2 V V 29 in context

In order to get a clearer understanding of what these passages might mean, it is important to consider them in the argumentative context in which they occur. The VV, which contains the first passage given above is a work of seventy verses, accompanied by Nagarjuna’s autocommentary. As its title, which translates as ‘The Dispeller of Objections’ suggests its main aim is to answer objections which had been advanced concerning Nagarjuna’s theses. Being of a rather technical and specific nature makes it is plausible to assume that the VV was written later than his main work, the MMK, and was meant to deal with particular problems arising from the arguments set out in there.4 The first twenty verses and their commentaries contain criticisms of Nagarjuna’s position, which are answered in the remaining verses and their commentaries. Verse 29 given above specifically addresses the problem raised by the opponent in verse 4. The principal point the opponent makes at the beginning of the VV concerns the status of Nagarjuna’s claim of universal emptiness. The opponent argues that Nagarjuna faces a dilemma the horns of which are inconsistency and impotence. If he assumes his claim not to be empty he has contradicted his own thesis of universal emptiness (because there is now at least one thing which is not empty). If, on the other hand Nagarjuna takes his own claim 4Mabbett (1996, 306-307), Bhattacharya (1999, 124).

to be empty too, the opponent argues, this claim is then unable to deny the existence of independently existing phenomena the opponent asserts. As becomes clear later in verse 22, Nagarjuna accepts the second horn of the dilemma: everything is empty, and his claim that everything is empty is empty too. As he stresses in the following verse this, however, does not entail that the claim could not carry out its philosophical function. A key can open a door in a film even though it is only a key in the film, not a real key.5 Verse 4 now considers a specific comeback Nagarjuna could make in reply to the difficulty arising from accepting this second alternative, i.e. the charge of the argumentational impotence of his claim of universal emptiness. Nagarjuna could argue that if universal emptiness renders his own claim impotent, the opponent’s claims, being also subsumed under the universal statement of everything being empty are similarly impotent, and therefore cannot act as a refutation of Nagarjuna’s claim either. But as the opponent is quick to point out, this involves a blatant petitio principii: only if we already accept that everything is empty will the opponent’s arguments be rendered empty and impotent. But this is exactly the thesis the opponent denies. For him at least some things are not empty, and in particular his own statements are not subject to Nagarjuna’s claim of universal emptiness. The difficulty the opponent raised is a difficulty which arises because of the specific character (laksana) of Nagarjuna’s system, namely the claim that everything is empty. It does not apply to someone who does not make that assumption. Verse 29 then is made in reply to this supposed counter-argument and its rejection as a petitio. There Nagarjuna claims that the particular defect (of 5As Nagarjuna points out in MMK 1:10 this is in fact a necessary condition for its being able to perform its function: only a cinematic key could open a cinematic door, a real key could not. See also the commentary on this verse in Garfield (1995, 119).

his thesis of universal emptiness rendering his own philosophical assertions impotent) would indeed apply if he had any position. But given th at he has no position, the difficulty therefore does not apply to him. Now it will strike the reader th at this is a rather curious reply to make. It is evident th at the opponent’s criticisms formulated in verse 4 as well as in the preceding verses rest on a misunderstanding of the central term ‘emptiness’. W hat exactly this misunderstanding amounts to is less clear. In fact the above set of arguments would make sense if we assumed that the opponent understood ‘empty’ to mean ‘false’, or ‘meaningless’, or even ‘non-existent’.6 But as a reply to a criticism based on misunderstanding of this kind Nagarjuna’s reply in verse 29 seems a little extreme, given that it would have been perfectly sufficient and far less controversial for him to point out th at emptiness entailed neither falsity, nor meaninglessness, nor non-existence and th at he thereby could both assert that his claims are empty and simultaneously able to refute the opponent’s objections (in fact he makes exactly these points in verse 21 and 22). Even if we agree with M abbett that it may be the case that the objection addressed by a given verse has already been essentially refuted, but in turning to each new 6Indeed we might think that the argumentative context makes it most likely that the opponent misunderstands ‘empty’ as ‘non-existent’. In this case the problem that nonexistent statements cannot really refute anything seems to be most pressing. But in the case of the other two alternatives other problems become more serious. If Nagarjuna meant ‘meaningless’ when he said ‘empty’ his claim that everything is empty would obviously just be false, given that we perfectly well understand the claim he makes (in the same way as somebody saying ‘all statements, including this one, are not grammatically well-formed’ would utter a falsehood). If, however, ‘empty’ meant ‘false’ Nagarjuna’s thesis of universal emptiness would reduce to the liar paradox and there is no good textual evidence that this is the problem the opponent had in mind. On this last point compare also the discussion in Mabbett (1996) and Sagal (1992).

objection Nagarjuna seeks to make a fresh rebuttal in order to administer the coup de grace7 Nagarjuna here seems to use a sledgehammer to crack a nut. Why deny to hold any proposition whatsoever if it would have been perfectly sufficient to point out that since 'empty1 does not mean ‘non-existent’ it is completely unproblematic to claim that one’s own position is as empty as everything else? We can distinguish at least three different ways in which Nagarj una’s crucial statement that he has no position can be interpreted. I will refer to these as the semantic, argumentational, and transcendent interpretations. According to the semantic interpretation Nagarjuna does not claim to hold no thesis whatsoever, but only claims to accept no statements which are taken to have a particular semantics. If we follow the argumentational interpretation Nagarjuna makes a claim about how one should proceed in debates, namely by always refuting opponents via reductio arguments, without ever adopting any thesis oneself. The transcendent interpretation finally reads Nagarj una’s statement as the assertion the existence of an inexpressible reality beyond concepts and language. All of these three interpretations have historical predecessors in the commentarial tradition. The semantic and argumentational interpretation can be found in works of the dGe lugs tradition, in particular those of Tsong kha pa8 and mKhas grub rje,9 while a variety of views which can all be regarded as some kind of transcendental interpretation can be found in the writings 7(1996, 307). sTsongkhapa Bio bzang grags pa (2000-2004,111:230, 236-249). 9Ruegg (2000, 173-187).

of scholars like rNgog bio ldan sh.es rab,10 Go rams pa11 and dGe ’dun chos ’phel.12 In the following I will restrict myself to an exposition of the semantic interpretation. This is primarily due to the fact that this appears to give us the clearest understanding of the role of verse 29 in the context of Nagarjuna’s arguments. The argumentational and transcendent interpretations tend to use Nagarjuna’s denial of a thesis as a textual peg on which to hang an argument concerned with quite different matters from those dealt with in the VV. Tsong kha pa, for example, refers to this verse in the context of expounding the distinction between Svatantrikas and Prasangikas; Sa skya Pandita offers the transcendent interpretation in the context of a debating manual (advising the reader on how to debate with somebody who does not put forward a position); dGe ’dun chos ’phel’s work is, despite its title, not a study of Nagarj una’s thought in particular, but is mainly concerned with criticizing the then prevalent dGe lugs interpretation of Madhyamika philosophy more generally. This is of course not to say that the argumentational and transcendent interpretations are for this reason deficient or lacking in interest within the context in which they are presented. However, it is im portant to be aware th at these contexts were not Nagarj una’s context. There is certainly no reason for suspicion towards later Indian or indeed Non-Indian works as not giving a valid interpretation of Nagarjuna’s thoughts. Nevertheless, the most 10See Ruegg (2000, 32-33, note 59). 11 See Ruegg (2000, 194-195, note 135). 12See Lopez (1994). A translation of dGe ’dun chos ’phel’s Klu grub dgongs rgyan is in Lopez (2005). 13In his mKhas pa rnams ’jug pa ’i sgo, III: 37-39. See Jackson (1987, volume 1, 271) for the Tibetan text, volume 2, 341-342 for an English translation. A summary with comments is in Ruegg (2000, 169-171).

interesting of these for the present purpose of a philosophical analysis of Nagarjuna’s thoughts are those which allow us to understand passages from his works in their argumentative context, rather than using them as a starting points for presenting their own ideas on a particular topic, 9.3 T he sem antic interpretation If we consider the major dGe lugs pa commentaries on verse 29 it becomes evident th at these usually regard Nagarj una’s statement as elliptical. W hat Nagarjuna really means when saying that he has no position, these commentaries claim is th at he has no positions which are non-empty.14 The key to understanding the point made in these commentaries lies of course in a precise understanding of what it means for a position or statement to be empty. An object is empty if it does not exist from its own side and is therefore dependent on other objects, so th at its existence is not grounded in its ‘own-nature’ (svabhava, rang bzhin). The Buddhist commentarial tradition considers a variety of dependence relations in which objects stand and which prevent them from existing in a non-empty way. These dependence relations include causal dependence, dependence of a whole on its parts, as well as dependence on a cognizing subject.15 While in the case of certain objects their independent existence seems at least a prima facie plausibility which the Madhyamika then attem pts to refute by appropriate arguments, the emptiness of statements appears to be entirely uncontroversial. Material objects might be considered to exist in causal and mereological dependence, 14,It is not being said that the Madhyamika has no theses; he mei'ely has not theses that inherently exist.’ Hopkins (1983, 471). The same point is made in mKhas grub rje’s commentary on this passage; see Ruegg (2000, 179). 15See e.g. Gyatso (2005, 66-69).

but independent of a cognizing subject; abstract objects, platonistically conceived, will be assumed to be independent in all three ways. Statements, however, can hardly be taken to ‘exist from their own side’ in any of the three senses. As even Nagarj una’s opponent affirms in VV 1 token16 utterances are events, which arise in dependence on causes and conditions like all other events. When considering utterances as types it is equally clear that, assuming a compositional semantics, these are mereologically dependent on their parts, since the meaning of the sentence type is a function of the meanings of its constituents or parts. Finally, considering a constituent like the expression ‘red’, we realize that its referring to the colour red is no property the word ‘red’ has independently of everything else: the connection of this particular phonetic or typographic object with the property is a convention which holds for speakers of English; for speakers of French the same property is connected (by a different set of conventions) with ‘rouge’, for speakers of Tibetan with ‘dMar po’, and so forth. That ‘red’ refers to the colour red depends on a complex framework of conventions connecting a community of cognizing subjects which share a language. Unless we mistakenly consider ‘empty’ to mean ‘false’ or ‘meaningless’ or ‘non-existent’ the claim that utterances conceived of as either tokens or types are not empty seems to be a position it is hard to make sense of. Despite the prima facie strangeness of their claims theories of the nonemptiness of language have found their defenders. Perhaps the most extreme example is the view of language defended by the MTmamsakas.17 A primary 16A particular utterance of a sentence is a sentence-token, what is expressed by several such utterances which say the same thing is a sentence-type. 17The basic text of this school is J aim ini’s Purva Mimams a Sutra. For the Mlmamsa theory of language see especially the first adhyaya, first pada of this text (Jaimini 1916,

motivation of the Mlmamsa theory of language was to provide a justification of the authoritative status of the Vedas. As opposed to the Naiyayikas, who justify the Vedas by their divine authorship the Mlmamsakas regard them as authorless (apauruseya). The elements of the Vedic language are assumed to exist eternally, without the necessity of a speaker. Any particular human utterance of course depends on a phonetic or typographical instantiation of a piece of language, but the types thus instantiated exist ante rem, without depending on the tokens instantiating them. The referents of expressions, which the Mlmamsakas take to be eternal and unchanging universals are related to these expressions via a set of objective and necessary relations.18 While the Mlmamsa view of language attracted plenty of criticism from the Buddhist side (centred around Dignagas apoha theory)19 there is no good evidence that this is the view Nagarj una’s opponent in the W wants to defend.20 There is, however, some interesting evidence that at least some of Nagarjuna’s Indian commentators saw him as opposed to similar conceptions of language. When commenting on MMK 2:8 in his Prajnapradipa Bhaviveka raises the question why the verbal root gam, ‘to go’ is used in its atmanepada form ‘gacchate' rather than conjugated in the usual paraismapada manner as a ‘gacchatV21 Bhaviveka lists a varieties of quotations from Indian grammar1- 22). See also D ’Sa (1980, 80-82, 113-140). ls Sharma (1960, 220-222). 19See Dreyfus (1997, 213-215). 20That Nagarjuna’s opponent was a Naiyayilca (as claimed by Bhattacharya (1977, 265) and Bhattacharya et al. (1978, 1)) is supported by the close connection between the VV and the NS (Meuthrath 1999). See Bhattacharya (1999, 124) for further references. 21 Ames (1995, 309). This form is not found in any version of Candraklrti’s Prasannapada from which the Sanskrit text of the MMK is usually extracted. Here we just read lgacchati, (PP 97:14, see also de Jong (1978)). Only recently some fragments of older manuscripts of the MMK independent of the Prasannapada have been discovered (see Ye (2006a),

ians illustrating the perils of wrong grammar. When the god Tvastr created a serpent to destroy Indra he exclaimed indrasatrur vardhasva, intending to say ‘Go, destroyer of Indra!*. As he intended the compound to be a tatpurusa it should have been stressed on the ultimate syllable. Unfortunately Tvastr stressed it on the first syllable, turning it into a bahuvrihi meaning ‘having Indra as a destroyer*. The words did what they meant, rather than what Tvastr intended them to mean, and Indra destroyed the snake, not the other way round.22 Bhaviveka then continues to observe that Nagarjuna’s irregular use of lgacchat& was not only intentional, but served a philosophical purpose. By demonstrating th a t no disaster would strike from an irregular use of grammar Nagarjuna was aiming to convince his opponents to give up their attachment to mere words, together with the assumption that there was a substantial nature (svarupa) of words which determined that they could only appear in certain grammatical forms.23 Nevertheless, for the purposes of interpreting the VV it makes better systematic sense to ascribe a different (and less extreme) theory than th at to Nagarjuna’s opponent. According to this theory wether a statement is empty or not does not depend on the mind-independent existence of language in some Platonic heaven but focuses on the semantics we employ when interpreting the statement. Even if we accept that the link between ‘red’ and the property of redness is conventional, this does not imply th at we also think th at the property of redness only has conventional existence as well, It can still be a property which exists in the world independent on human conventions and intentions. Moreover, even if the linkage of particular words (2006b)). Unfortunately the verse in question is not amongst them. 22Ames (1995, 342, note 65). Notes 64-70 provide very useful information for identifying some of the authors Bhaviveka quotes. 23Ames (1995, 310).

to their referents should prove to be conventional, the linkage of entire sentences to the world might not be. For example we might suppose that the statement ‘The apple is red’ is linked to the state of affairs it refers to by a relation of structural similarity, by their sharing of a common logical form, which is in turn is not a product of convention. Once we have linked up the simple signs of our language with the simple objects in the world we then do not need a further set of conventions to link up the complex signs (the sentences) with the complex objects (facts or states of affairs), in the same way as once we have settled by convention how the different chessmen are to move we do not have to bring in further conventions to decide whether a particular distribution of chessmen on the board will allow white to mate in five moves. This can be decided just by reference to the initial conventions, and in the same way the truth-conditions of a sentence like ‘the apple is red’ can be worked out by considering the simple signs it is made up of and how these are put together in the sentence. In fact both the assumptions behind this picture of the non-emptiness of statements, that there is a ‘ready-made world’, to borrow a phrase of Putnam ’s, and the assumption of a structural link between language and the world are extremely widespread, so widespread indeed th at we might refer to them jointly as the ‘standard picture’. This standard picture provides us with good idea of what is meant by the notion of svabhava in the context of language, as opposed to an ontological understanding of svabhava in terms of substance, or a cognitive understanding in terms of a superimposition (samaropa) which conceptualizes objects as permanent and observer-independent. The standard picture therefore represents the third, the semantic dimension of svabhava mentioned in chapter 2. It is evident that the standard picture does not sit well with the thesis of

universal emptiness. Neither the existence of a world sliced up ‘at the joints’ into particulars and properties nor the existence of an objective structural similarity between sentences and the world would be acceptable for the Madhyamika. A Madhyamika-compatible semantics would deny the existence of a world differentiated objectively into different logical parts, and would try to replace the structure-based picture of the language-world link by a different one, perhaps by a theory built on speaker conventions. There is good historical evidence th at the standard picture is indeed what Nagarjuna’s opponent presupposes if we take into account how closely many of the arguments in the VV engage with the NS. Garfield24 points out that in the Nyaya-influenced logico-semantic context in which these debates [in the W ] originate the dominant view of meaningful assertion (the one that Nagarjuna calls into question) is one that from our vantagepoint can best be characterized as a version of Fregean realism: meaningful assertions are meaningful because they denote or express independently existent properties. A proposition is the pervasion of an individual entity or groups or entities by a real universal or sequence of universals.25 On this understanding of the emptiness of statements we can read the opponent as claiming in VV 1 th at because of Nagarjuna’s thesis of universal emptiness the Madhyamika cannot accept the standard semantic interpretation for his utterances. For Nagarjuna both questions of ontology (how the world is sliced up) and of semantics (how language and the world are linked) 24Garfield (1996, 12). 25 On the relation between the VV and the Nyaya school see Oberhammer (1963) and Bhattacharya (1977). For some remarks on the realist background of the Navy a-Nyaya see Ingalls (1951, 1, 33-35).

must be settled by appeal to conventions. The opponent, on the other hand, can assume th at there is a ‘ready-made world’, as well as an objective, structural way of linking this to our language.26 Now the opponent argues th at on this picture Nagarjuna never gets out of his system of conventions to connect his claims with the things — and that is the reason why his claims are unable to refute the opponent’s claims, which manage to connect with the things. Nagarjuna’s arguments can no more refute the opponent than the rain in a meteorological simulation can moisten real soil.27 Nagarjuna’s opponent thus considers the interesting case of a language where we have two kinds of statements: some are interpreted according to the standard semantics (referring via an objective reference relation to objects which exists independently of us), some are interpreted according to Nagarjuna’s semantics (which does not make these assumptions). The opponent argues that statements of the second kind could not possibly influence the first kind. To see this consider a similarly structured case. Assume we recognize two kinds of norms, norms which are real, objective, ‘out there’ and norms which are the product of human convention. (Moral realists take certain ethical norms to be of the first kind, rules for the regulation of traffic are generally considered to be of the second kind). Now it is clear that although the two kinds of norms could be in conflict, a norm of the second kind could never override one of the first 26 Another manifestation of the Naiyayika opponent’s conception of a harmonious wordworld link is the view that the simple terms of our language cannot fail to refer (as discussed in chapter 3). 27In VV(S) 43:2-6 the opponent claims that ‘a fire that does not exist cannot burn, a weapon that does not exist cannot cut, water that does not exist cannot moisten; similarly a statement that does not exist cannot deny the intrinsic nature of all things.’ na hy asatagnind sakyam dagdhum \ na hy asatd sastrena sakyam chettum j na hy asatibhir adbhih sakyam kledayitum \ evam asatd vacanena na sakyah sarva-bhavasvabhava- pratisedhah kartum.

kind, since the former are part of the objective normative framework of the world, while the latter are only a supplement of human design. Although he does not explicitly say so Nagarjuna’s arguments seem to imply th at he agrees this situation would indeed be problematic. If there are two kinds of statements the latter would be as impotent compared to the former as a film would be to reality: we could not escape the burning cinema by entering the scene projected onto the wall. Nagarjuna counters the charge of impotence by denying th at there are two kinds of statements, which differ like film and reality. All statements are to be interpreted in the same way, so that their interaction is not ontologically any more problematic than the interaction of different characters in a film.28 Understanding the emptiness of statements as their interpretation according to a non-standard semantics we can also give a more interesting rendering of the argument in VV 4. Remember that there the opponent claims that Nagarjuna might want to say According to this very method, a negation of negation is also impossible; so your negation of the statement negating the intrinsic nature of all things is impossible.29 The opponent has just claimed th at because Nagarjuna’s theory entailed a non-standard semantics his assertions did not manage to connect with the world and were therefore meaningless. But if the opponent then sets out to refute the thesis of universal emptiness, this either means that he takes it to be meaningful after all (and therefore deserving refutation) or th at the statement he wants to defend (which is the negation of Nagarjuna’s claim) 28See particularly VV 23, 27; MMK 17:31-33. 29 pratisedhapratisedho ’py anenaiva ka Ip en dnupap arm a h tatra yadbhavan sarva-bhavasvabhava- pratisedha-vacanam pratisedhayati tad anupapannam iti. VV(S) 45:16-18.

is meaningless as well, since plugging in the word ‘not’ will not help to turn nonsense into sense. The opponent could reply to this charge by pointing out the difference between internal and external negation. While it is plausible to assume that the internal negation of a nonsensical statement is nonsensical too (‘the number seven is not yellow (but rather some other colour)’ is as problematic as ‘the number seven is yellow’) this is not the case for an external negation (‘it is not the case that the number seven is yellow’ is not just meaningful but also generally taken to be necessarily true). Nagarjuna’s opponent could then claim that his negation of the claim of universal emptiness is external only and therefore not affected by the lack of meaning in the claim it negates.30 It is possible th at the opponent had argued like this, as a distinction between the different scopes of negation, as well as the accompanying presuppositional and non-presuppositional readings was made in the philosophical literature of the time.31 It has to be noted, however, th at the passage in question fails to make any direct reference to different kinds of negation being involved.32 A more abstract way of employing the distinction between the two kinds of negation in the opponent’s reply consists in rejecting Nagarjuna’s peculiar semantics. Here the opponent points out that he does not have to accept Nagarjuna’s semantics as it is a particular characteristic (laksana) of Nagarjuna’s system, but nothing the opponent would be forced to take 30Garfield (1996, 12) reads the argument in this way and argues that the opponent just wants to negate Nagarjuna’s position, without asserting the contrary. 31 For present purposes we can assume a (simplifying) identification of paryudasapratisedha with internal negation and of prasajyapratisedha with external negation. For further differentiation see Ruegg (2002, note 6, 19-24) as well as the discussion in chapter 4, section 4.1. 32Compare also the discussion in Ruegg (2000, 117).

on board.33 The opponent does not just negate Nagarjuna’s claim of universal emptiness, but the entire non-standard semantics which comes with it. If prasajya-negation is seen as a presupposition-cancelling negation which negates not just a proposition but also that proposition’s presuppositions,34 and if the semantics according to which a speaker wants the set of his utterances interpreted is included amongst these presuppositions denying a claim together with the semantics it comes with can be regarded as an example of p ra s aj y a- n e gat i on. 9.4 T he specific role of verse 29 It is interesting to note that verse 29, which is meant to be a reply to the opponent’s argument given in verse 4 does not attem pt a comeback in trying to argue that the opponent’s negation of Nagarjuna’s claim of universal emptiness is somehow impossible after all. Instead Nagarjuna addresses a difficulty (dosa) arising from the ‘specific character’ of his system which the opponent raises at the end of verse 4. In mKhas grub rje’s s Thong thun chen mo, an influential dGelugs commentary which deals with the interpretation of this passage35 this difficulty is taken to be inconsistency. If Nagarjuna assumed that his thesis of universal emptiness was non-empty itself (rang bzhin gyis yod pa) and, on our interpretation, would therefore have to be supplied with a semantics accord33‘ The objection applies only to the specific character of your proposition, not to that of mine. It is you who say all things are void, not I. The initial thesis is not mine.’ tava hi pratijhalaksanapraptam na mama \ bhavan braviti sunyah sarvabhava iti naham \ purnakah pakso na mama. VV(S) 45:19-66:2. 34As e.g. in Shaw (1978, 63-64). 35See Ruegg (2000, 173-187) for a summary and analysis of the relevant part of the commentary.

ing to the standard picture his position would be inconsistent (at least until he proposed a special reason why this statement should be excepted, which Nagarjuna does not do). But, mKhas grub rje argues, since none of Nagarjuna’s claims of universal emptiness are taken to be non-empty, the difficulty of inconsistency does not arise.36 The same point is made by Tsong kha pa:37 Therefore, the issue as to having or not having theses is not an argument about whether [Nagarjuna] has them in general It is an argument as to whether the words of the thesis ‘all things lack intrinsic nature’ have intrinsic nature. [Therefore the meaning of the lines from the VV is this:] If I accepted that the words of such a thesis had an intrinsic nature, then I could be faulted for contradicting the thesis that all things lack intrinsic nature, but because I do not accept that I cannot be faulted. W hat is unsatisfactory about this interpretation is th at Nagarjuna has already made the point ascribed to him here in verse 22. There he claims th at his claim of universal emptiness is also empty, and gives reasons why he thinks it can still have argumentative force, thus avoiding the charge of impotence. Unless we assume Nagarjuna to be unnecessarily repetitive it is not clear why we should assume that he makes the very same point once again a couple of verses later, and also formulates it in a much more obscure manner than the first time. 36150al-3, Ruegg (2000, 179). 37des na dam bca’ yod med ni spyir yod med rtsod pa ma yin gyi \ dngos po thams cad la rang bzhin med do zhes dam bcas p a ’i tshig la rang bzhin yod med rtsod pa yin pas | de ’dra ba’i dam bcas p a ’i tshig de la rang bzhin yod par khas blangs na dngos po thams cad rang bzhin med par dam bcas pa dang ’gal ba’i skyon nged la yod n a’ng \ nged de Itar mi ’dod pas skyon de nga la med (1985, 687:13-17), (2000-2004, 111:241).

It is important to note th at verses 21-28, which deal with the objections raised in the first three verses of the W are primarily concerned with solving the dilemma of inconsistency and impotence which is faced by Nagarjuna’s claim of universal emptiness, Verse 29, however, (pace mKhas grub and Tsong kha pa) is not again concerned with the thesis of universal emptiness, Nagarjuna realizes that the twin problem of inconsistency and impotence is not just a problem for his thesis of universal emptiness, but for any other claim he holds as well. Any other claim will either face the problem of being a counterexample to Nagarjuna’s assertion that all claims should be given a non-standard semantics, or it will fail to connect up with the world in the way sentences with the standard semantics do, and will therefore be meaningless. I want to argue th at this is the difficulty arising from the ‘specific character’ of Nagarjuna’s system the opponent refers to in verse 4 and which Nagarjuna takes up again at the beginning of verse 29. He is not interested in defending the claim (attributed to him by the opponent in verse 4) th at his thesis of universal emptiness could not possibly be negated. Instead he takes up the opponent’s more important point that apart from defending his claim of universal emptiness from the twin problems of inconsistency and impotence, he should better say something about the status of his other assertions as well. This is why he says in verse 29 that none of his other assertions should be regarded as propositions with standard semantics (pratijha) either.38 The plausibility of this interpretation rests on there being two meaning of ‘thesis’ (pratijha) in play here, one referring to theses with standard semantics (which Nagarjuna rejects) and one which refers to theses with non380etke (2003, 468-471) reconstruction of Nagarjuna’s argument results in a different reading of verse 29; he argues that here Nagarjuna claims that for the Madhyamika there is no thesis to be made at the absolute level (paramartha) — a reading entirely consistent with Nagarjuna’s other statements (e.g. MMK 18:9).

standard semantics (which Nagarjuna does not reject). In fact there appears to be good textual evidence th at the notion of ‘thesis’ is indeed used in two different ways in Madhyamaka literature. CandrakTrti’s commentary on Nagarjuna uses one sense of thesis (pratijha) to refer to statements with clearly unproblematic status, indeed some utterances by Nagarjuna himself are regarded as theses in this way,39 while theses in another sense are firmly rejected. We might want to refer to the first kind of theses as propositions, and to the second as views. How are we to understand the distinction between them? It has been claimed th at views are theses with philosophical or metaphysical commitments40 and, more specifically, that they postulate an independently existing entity (bhdva).41 Propositions, on the other hand, do not make such commitments and are therefore philosophically unproblematic. It is important to note at this point, however, th at what distinguishes a view from a proposition is not just that the former asserts the existence of objects existing by svabhava while the latter does not. On this understanding the statement ‘Object x does not depend in any way on any other object’ would be a view concerning x } while ‘Object x stands in a variety of dependence relations with other objects’ would not be. Ontological commitment only comes into play at the level of semantics. Whether someone asserting th at the average man has 2.4 children is committed to an object which acts as the reference of the expression ‘the average m an’ depends on the semantics given. If we interpret the statement 39For example MMK 1:1 in PP 13:3. See Ruegg (1983, 213-214) for further examples. Oetke (2003, 458-459), however, argues that the distinction between two senses of pratijna only arises in the later Prasangika literature and should not be read back into Nagarjuna’s works. 40Sagal (1992, 83). 41 Ruegg (1983, 213).

in the way statements like ‘Paul has two children’ are usually interpreted, such commitment to a strange man with partial children ensues; if, on the other hand we read it (more plausibly) as a statement about ratios between the number of men and children in a certain set there is no such commitment. It therefore seems to be plausible to take the distinction between views and propositions and between theses with standard and non-standard semantics as coinciding. The views the Madhyamika rejects are theses which are interpreted by referring to a ready-made world and a structural link between this world and our language. The propositions he takes to be unproblematic, and some of which he holds himself, are theses which are given a semantics which makes neither of these two assumptions. Some support for this semantic interpretation of the difference between the two senses of ‘thesis’ can be gained from MMK 13:8: The Victorious Ones have announced that emptiness is the relinquishing of all views. Those who in turn hold emptiness as a view were said to be incurable.42 Although Nagarjuna does not use the word pratijna for ‘view’ but rather talks of drsti, it seems to be sensible to treat the two terms as synonymous in this context.43 If the difference between propositions and views just depended on what the statement asserted, statements asserting the emptiness of some phenomenon such as ‘each spatio-temporal object depends causally on some other object’ ex hypothesi could not be views, contrary to what Nagarjuna says in the verse just cited. If, however, we treat ‘view’ as denoting 42sunyata sarvadrstmam proktd nihsaranam jinaih / yesdrn tu sunyatadrstis tan asadhydn babhasire. 43As done by Ruegg (1986, 232-233) and Mabbett (1996, 301). For more details on the relation between the two terms pratijna and drsti see Ruegg (2000, 129-136).

a statement together with the standard semantics, this is indeed possible. For if we read ‘each spatio-temporal object depends causally on some other object’ as asserting the existence of various objectively existing individuals in the world, linked by a relation of causation, about which we speak by exploiting an objectively obtaining structural similarity between language and the world it would indeed be turned into a view. T hat the point at issue here is a specific (and, as Nagarjuna sees it, inappropriate) conception of semantics is supported by CandrakTrti’s commentary on this verse. Candraklrti argues that one taking emptiness to be a view is like one who, when being told by a shopkeeper th at he has nothing to sell, asks the shopkeeper to sell him that nothing, The customer (like the White King in Alice through the looking-glass) treats ‘nothing’ like a proper name and therefore expects it to denote a particular object, as proper names do. But though justified by the surface grammar of the sentence concerned, this does not lead to an understanding of what the merchant wants to say. Similarly, giving a standard semantical interpretation of statements asserting emptiness does not lead to an understanding of what Nagarjuna wants to say.44 The semantic interpretation outlined above provides a good way of making sense of verse 29 within the argumentative structure of the VV and also provides us with an idea of what a Madhyamaka theory of language would look like. W hat Nagarjuna means when he says that he ‘has no thesis’ is that none of his theoretical statements (including the claim of universal empti44This interpretation does of course not imply that one could hold ‘any position at all’ as long as one gives it the required non-standard semantics, as Galloway (1989, note 5, 27) asserts. A statement like ‘things arise from what is other that themselves’ will be regarded as false by Nagarjuna, independent of whether it is interpreted according to the standard or the non-standard semantics. CHAPTER 9. LANGUAGE 292 ness) are to be interpreted according to a semantics based on the standard picture. For the Madhyamika no assertion is to be taken to refer to a readymade world of mind-independent objects, nor can he assume that there is a structural similarity linking word and world which is independent of human conceptual activity.

Chapter 10 Conclusion: Nagarjuna’s philosophical project

This chapter is to serve three purposes. First of all it will summarize the main philosophical conclusions Nagarjuna argues for. The arguments in support of these have been analysed in detail in the preceding chapters, I will therefore confine myself to a concise statement of the conclusions themselves. Secondly I set out to show that these are not just isolated philosophical statements but fit together as a unified philosophical theory which is Nagarjuna’s Madhyamaka. Finally I want to assess some systematic aspects of the emerging theory, its theoretical appeal as well as some connections with contemporary philosophical debates.

10.1 M etaphysics

Nagarjuna’s central metaphysical thesis is the denial of any kind of substance whatsoever. Here substance, or, more precisely, svabhava when understood as substaiice-svabhava is taken to be any object which exist objectively, the

existence and qualities of which are independent of other objects, human concepts or interests, something which is, to use a later Tibetan turn of phrase ‘established from its own side’.1 To appreciate how radical this thesis is we just have to remind ourselves to which extent many of the ways of investigating the world are concerned with identifying such substances. W hether it is the physicist searching for fundamental particles or the philosopher setting up a system of the most fundamental ontological categories, in each case we are looking for a firm foundation of the world of appearances, the end-points in the chain on existential dependencies, the objects on which all else depends but which do not themselves depend on anything. We might think that any such analysis which follows existential dependence relations all the way down must eventually hit rock-bottom. As Burton2 notes, ‘the wooden table may only exist in dependence upon the human mind (for tables only exist in the context of human conventions) but the wood at least (without its ‘tableness’) has a mind-independent existence.’ According to this view there is thus a single true description of the world in terms of its fundamental constituents, whether these are pieces of wood, property particulars, fundamental particles or something else entirely. In theory at least we can describe - and hopefully also explain - the make-up of the world by starting with these constituents and account for everything else in terms of complexes of them. The core of Nagarjuna’s rejection of substance is an analysis which sets out to demonstrate a variety of problems with this notion. The three most important areas Nagarjuna focuses on are causal relations between substances, change, and the relation between substances and their properties. 1rang ngos nas grub pa. 2Burton (1999, 115). Compare Siderits (2004, 395).

Causation

Supposing there were such things as substances Nagarjuna argues th at they could not stand in the relation of cause and effect. The first, and simplest argument to this effect is of course to point out that if one substance caused another, the latter being an effect would be existentially dependent on the first and therefore could not be a substance, as it is an essential property of substances not to be dependent in such a way. But Nagarjuna also employs a different argument which can be used against an opponent who does not want to rule out all dependence relations between substances. Like the Abhidhannilcas he might want to assert th at substances can form a causal network, but still claim that they are independent in im portant other respects (for example not being dependent on their parts or on human interests or concerns).3 This argument proceeds as follows. Nagarjuna agrees with the Abhidharma analysis that any substance, any object which exists by its own intrinsic nature has to be atomic. For if it consisted of parts it would be existentially dependent on them. But as the Abhidharrnika’s mereological argument aims to demonstrate a partite entity cannot be regarded as ultimately real. It is rather to be conceived of as a conceptual construction from (or as a superimposition 011) its parts, which may be ultimately real, presupposing that they do not depend for their existence on anything else. It is obvious that the demand for the atomicity of substance cannot just be restricted to mereological atomicity but must include temporal atomicity as well. For in the same way in which we can argue that a house is conceptually constructed from its proper mereological parts, such as bricks, beams, tiles and so on, in the same way we have to 3See Siderits (2004, 396), Burton (1999, 90), Walser (2005, 241-243).

regard it as being constructed as a collection of temporal moments, namely from its temporal part which exists now, from that in the next second, from the one after th at and so forth. We now already see the problem for the causal relation between substances on the horizon, as causation is a relation which is necessarily located in time. We will recall that Nagarjuna considered the different possible temporal relations between cause and effect and rejected them all as problematic when dealing with substances. Clearly it makes no sense to assume that the cause exists only after the effect, for once the effect exists there is no further necessity for the cause to bring it about. It might initially seem more plausible to assume that the cause exists first, and is succeeded by the effect. But after a moment’s thought it becomes apparent th at within the present ist framework in which the argument is set up this means that only one relatum of the causal relation will exist at one time. For while the cause exists the effect, being future, does not yet exist, and after it has come about the cause, being past, exists no longer. In this case it should be apparent that we cannot deal with a relation between two substances or ultimately real things, for one of the relata is only provided by our expectation4 (or our memory in the retrospective case). On this understanding causation cannot be regarded as a relation between items ‘which are there anyway’ but has to be seen as essentially dependent on the mind which supplies it with the missing relatum. The final possibility, that cause and effect are simultaneous does not fare much better, as discussions of causation commonly accept that it makes not much sense to speak of two simultaneous events (such as the rising and falling of the opposite ends of a perfectly rigid see-saw) in terms 4As noted by Bhaviveka in his commentary on MMK 1:3 in the Prajn dp ? u dip a (Pandeya 1988-1989, 26).

of cause and effect, for each argument we could produce for one event being the cause can equally well be used to claim that it is the effect. Even if we could get around this difficulty somehow we would certainly not want to say th at simultaneity of cause and effect is something to be found in all instances of causation. At this place in the argument we might want to accuse Nagarjuna of having neglected a crucial case: that of the cause and effect overlapping. After all this is the most straightforward model of causation we have. When we see a potter make a pot we see the pot during the process of its production, i.e. at a time at which both the cause (the pot-making potter) and the effect (the pot being produced) exist. So cause and effect appear to be able to exist as temporally overlapping events. But the problem is that this commonsense model of causal production cannot be used to account for the causation between substances. Because substances are temporally atomic they cannot undergo a temporally thick process of gradual emergence we observe in objects which are not part of the fundamental furniture of the world (such as pots). At any particular moment either there is such a substance or there is not. For this reason the three temporal relations examined by Nagarjuna are really all there is for substances to which the concept of gradual emergence does not apply.5 Assuming we accept Nagarjuna’s argument that causation is 5Siderits (2003, 131), (2004, 418, note 30) notes that during a process of gradual production such as that of a pot we find that there will always be a stage at which it is vague whether or not the pot has been produced yet. According to some views of pots the produced entity is already a pot, but not according to others. This is unproblematic, as we accept that there are different ways of conceptualizing pots. But we would not countenance a similar vagueness in the case of substances, for here it should precisely not depend on our conceptualizing whether a particular substance exists or not. This should be something settled by the world on its own. Arguing in this way provides us with another reason why we would not want to accept that the causal relation between

conceptually constructed it is clear how this can be employed against the Abhidharmika’s view of primary existents. For even though the Abhidharmika can accept that primary existents possessing svabhava can be dependent on causes and conditions they should not be dependent on the human mind. But then the existence of one primary existent should not be dependent on another one by a relation which is mind-dependent. So assuming these objects are connected by causation they cannot be primary (dravya) but must be secondary existents (prajnapti).

Change

It is an obvious fact th at the world around us is always changing. Now a substance, an object which has its properties intrinsically could not change with regard to these properties. For in this case the existence of the properties would rely on the causes and conditions which brought them about so these properties would turn out to be dependent after all. For the same reason substances could not come into existence or go out of existence; besides being changeless they also have to be eternal.6 For the defender of substances it is therefore necessary to regard all the change we observe as a mere difference in rearrangement of the most fundamental constituents of the world. Successive states of the world are just permutations of what is there all the time, An immediate consequence of this is th at the fundamental particles physics studies do not qualify as substances. For suppose such a particle is destroyed in a collision close to the speed of light. A the same time a burst of energy is detected in the close vicinity. Now we either say that substances could be conceived of in terms of a temporal overlap, °This presupposes that the annihilation of a substance does not imply any change regarding its properties, since both the substance and its properties cease to exist.

the particle went out of existence and the energy burst came into existence more or less at the same time, in which case neither can be a substance for the reasons just noted, or we say that the particle changed into the energy burst, in which case we have to explain this change in terms of some yet more fundamental elements the rearrangements of which could appear either as a fundamental particle or as a burst of energy. The fundamental constituents of the physical world thereby recede further and further. We might want to argue that we are acquainted with some eternal, unchanging entities, such as mathematical objects. Of course this depends heavily on our ontology of mathematics, and looking at the contemporary discussion Platonism does not seem to be a position attracting the most convincing defences. But even if we assume Platonism is true, we would have to argue in addition that all the objects of our experience can be reduced to abstract objects if we follow the downwards chain of dependencies for a long enough time. How a complex arrangement of objects without spatiotemporal location could turn out to have such a location in itself would be only one of the startling questions such a theory would have to answer. It therefore appears that the permanent entities we are acquainted with are not quite the right kind of thing for playing the role of fundamental parts of reality, while those which seem to be the right kind of thing (like the fundamental particles of physics) cannot be regarded as fundamental as long as they are subject to change as well. Once more the notion of objects existing with svabhava seems to have slipped our conceptual grasp. Substances and their properties Another difficulty arising if we assume 'there are substances is the relationship between such substances and their properties. We cannot just conceive of

some substance as an individual instantiating properties. For the sake of illustration (and using an Indian example) suppose that water-atoms are substances and that their only intrinsic property is wetness. Now what is the individual in which wetness inheres? As it is not characterized by any other properties it must be some kind of propertyless bare particular. W hat makes it a bare particular? Given that we are dealing with substances here it had better not depend on some other object. But if it is a bare particular by svabhava and being a bare particular is therefore its intrinsic nature we are in the same situation as we were with the water-atoms and their wetness. For now we can ask what the individual is in which being a bare particular inheres, and then we are well on our way into an infinite regress. Note that this problem does not got away if we feel uneasy about the property ‘being a bare particular’ and do not want to admit it. For we have to assume that the individual has some determinate nature due to which it is a bearer of its properties and the difficulty will just reappear with whatever we take such a nature to be. It does not help much if we conceive of substances as particularized properties or tropes instead. For then it is unclear how we can individuate one wetness-trope from another. We cannot differentiate them according to the individuals in which they inhere because we have just rejected the existence of individuals at the level of substances. We cannot say th at this wetness-trope is different from that because they turn up in different samples of water, since the samples of water are just collections of tropes. Of course we could try to tell apart the various trope-substances by the collections in which they occur (or, more precisely, by which other tropes they are related to via a higher-order compresence-trope). The difficulty for this is th at it introduces dependence-relations via the back door, for every trope will existentially deCH

pend on being connected to just these other tropes via a compresence-trope — we cannot take a trope and ‘move’ it to another collection. As we want to conceive of substances as entities which are not existentially dependent on one another this approach inevitably introduces a certain tension into our system. It thus becomes apparent that once more a conceptual scheme which can be more or less straightforwardly applied to non-substances breaks down once we attempt to analyse the supposedly foundational objects of our world in terms of it. This happened in the case of causation and can be observed once again in the case of individuals and properties. While there is no difficulty in analyzing the relation between a potter and a pot in terms of cause and effect, various problems ensue if we try to transpose this to an analysis of the relation of substances. Similarly while the analysis of a red apple into an individual and the property it instantiates is at least on the face of it unproblematic, the same analysis cannot be carried out when dealing with ultimate existents. After the criticism of the distinction between individuals and properties in Nagarjuna’s discussion of motion this should not be too much of a surprise. There he attempted to show that the distinction between individuals and properties is not one which exists independent of our conceptualizations. As the talk about the ‘property’ instantiated by a thin individual (such as a clap of thunder) had to be explained in terms of a single feature seen in two different ways - as constitutive and as instantiating - in the same way talk of the properties of ordinary thick individuals (such as Farinelli) could be seen to be equally a reflection of the division of their features into constitutive and instantiating properties, something which is just a reflection of our pragmatic concerns in conceptualizing the individual in question, but not a reflection

of its intrinsic nature. We could now imagine that somebody would argue as follows. Nagarjuna has shown that we run into difficulties if we attem pt to analyse the fundamental constituents of reality, objects which have an intrinsic nature in terms of such notions as cause and effect, change, or individual and property. Such objects cannot stand in causal relationships, they do not change, they cannot be thought of as bearers of properties. The most fundamental bits of reality therefore fail to be grasped by the familiar conceptual schemes we employ in our everyday lives in order to get around in the world. We therefore have to assume that substances are acausal, atemporal (since for them there is neither beginning, end, nor any change in between) entities which cannot be regarded as objects having properties. Given the fundamentality of the above conceptual schemes to our cognitive lives this seems to be nearly as good as saying that the nature of substances transcends conceptualization. Since the fundamental constituents of reality cannot be grasped by concepts which are our nearest and dearest they constitute an ineffable reality to which we can have no cognitive access. This interpretation was favoured by some Indian commentators. A particularly well-known example can be found in Dharmapala’s debate with Bhaviveka.7 There the Yogacarin Dharmapala takes Bhaviveka to task for asserting that even though things exist at the conventional level, nothing exists at the ultimate level. Dharmapala argues that it would be mistaken to assume that 7See Hoornaert (2004), Tillemans (1990). For modern defenders of the ineffability thesis which do not presuppose a Yogacara background see e.g. Murti (1955), Inada (1970, 24-26), Matilal (1990, 149), Mohanty (1992, 278). Interpreters who accept the ineffability thesis may or may not accept that there is a non-conceptual form of cognition through which substances can nevertheless be known.

at the ultimate level there was no svabhava at all, leaving the knowledge of ultimate reality without an object, like knowledge of a non-existent flower in the sky.8 It should rather be understood in terms of the Yogacara theory of the ‘three natures’ (trisvabhava) by claiming th at only the imagined nature (parakalpita svabhava), the mistaken projection of things as sustantial, was completely non-existent. Both the dependent nature (paratantra svabhava), the appearances, as well as the absolute nature (parinispanna svabhava), the fact th at the dependent nature is empty of the imagined nature, do, however, exist.9 Moreover, the true nature of things is completely beyond concepts.10 The difficulty with this interpretation is that if we regard the true nature of things as ineffable we still assume th at there are objects with a mindindependent11 intrinsic nature, namely th at of ineffability. This position assumes th a t there is a way things are from their own side, by svabhava, which is not in any way affected by us. The key difference between this kind of realism and the more familiar kind of realism which provides the background to much contemporary philosophy (and much of our everyday life) is an epistemological one. Common-or-garden realism asserts there is a way the world is which is independent of all description, and th at we can know at least a substantial part of it. Its less plain cousin agrees with the first 8Hoornaert (2004, 132-133). 9Hoornaert (2004, 139-140). 10ra£> tu phye ba tshig dang bral ba (Hoornaert 2004, 141, note 60). n We might think that because ineffability contains a reference to the conceptual frameworks we use it is in fact a mind-dependent quality. But its dependence is only notional, not existential. Compare it with the property ‘is so long that it cannot be measured by any measuring-rod on earth’. If any object had this property it would obviously not make sense to describe it in this way if our planet ceased to exist. But the annihilation of the earth would not affect its length. In the same way the ineffable nature of a substance remains the same whether or not there is anyone around trying to eff it.

part of the statement, but holds that this state of things forever eludes our grasp. But it is clear that for Nagarjuna neither form of realism is acceptable. The doctrine of emptiness tries to establish that there are no objects with intrinsic natures, whether they are knowable or not. The view of substances as ineffable which introduces entities with svabhava through the back door is therefore to be firmly resisted. A key element of the rejection of the view of ineffable substances is denying that it makes any sense to speak of objects lying beyond our conceptual frameworks, or, as Dharmapala put it, as ‘inaccessible to differentiating consciousness and words’ and ‘suspending all speech’.12 These frameworks are all we have and if we can show that some notion is not to be subsumed under them we must not conclude that it therefore has some shadowy existence outside of the framework. To this extent our conceptual framework is to be thought of not so much as a map of a country, but as a set of rules for a game. If a traveler brings us news from a city in some far-off land which we cannot find on our map we conclude (if we regard the traveler to be truthful) th at it must be located somewhere outside of the area covered by our map. But if somebody told us he had found a new opening gambit in chess, but th at this could not be w ritten down using the familiar notation be would be justified in being puzzled. After all the notation allows us to describe all the legal moves of chess (as well as some illegal ones), so how could something th at is part of a game not be constructed in accordance with the rules of the game? In this case we would not conclude th at because of the limited nature of the expressive resources of chess notation this gambit was beyond its grasp, but rather that there was no such gambit. It is not that there are some objects within the grasp of our cognitive capacities as well as some beyond them, 12Hoornaert (2004, 148-149). CH APTER 10. CONCLUSION 305 but rather th at the very concept of an object is something established by these capacities. It is not that parts of the world might not correspond to our linguistic and conceptual frameworks but that the idea of a structure of reality independent of these practices is incoherent.13 Our ability to grasp the world by concepts is acquired by our knowledge of language (or, as some might argue, is the very same thing as that knowledge). Language is a public phenomenon, an ability we display in interaction with other speakers. We would therefore want to claim that we can only be taken to have understood the meaning of a word or to have mastered some concept if we can give a public display of its use or application. A concept for which we could not give the application conditions even in principle, where we could not even tell in the abstract what kinds of objects would fall under it is not a concept at all. But this seems to be exactly the situation with the concept of substance when seen as ineffable. Because what falls under this concept is understood to transcend all our conceptual resources we would be necessarily unable to apply this concept to anything. It is for this reason that the Madhyamika claims that the concept of an ineffable substance is necessarily empty. And once this concept is ruled out the only remaining conclusion to draw from Nagarjuna’s criticism of substance is that there is no such thing, not even an ineffable one. The metaphysical anti-realism defended by Nagarjuna is not just of historical but also of considerable systematic interest. One reason is its wide scope. While most of the discussion of anti-realism we find in the contemporary literature is concerned with particular local phenomena (such as mathematical 13Candrakirti makes this point criticizing the Yogacara view of the mind as selfilluminating (svaprafcaso), See Siderits (1989, 243). This, however, does not imply that these frameworks would not be susceptible to criticism or change. See Tillemans (2003, 123, note 47).

objects or moral values), the Madhyamika’s anti-realism takes the form of a general anti-foundationalism which does not just deny the objective, intrinsic and mind-independent existence of some class of objects, but rejects such existence for any kinds of objects which we could regard as the most fundamental building-blocks of the world. A second interesting point is the fact that Nagarjuna does not regard his metaphysical theory to imply that anything is up for grabs. That there are no substantially existent entities does not entail th at there are no selves responsible for their actions, no distinction between the moral worth of different actions, no difference between true and false theories. The Madhyamika therefore has to come up with an account of convention which is both solid enough to ground our ethical, epistemic and semantic practices but not so rigid as to re-introduce some sort of realism regarding any of these.

10.2 Personal Identity

Nagarjuna’s rejection of entities existing by svabhava is not restricted to the study of the external world around us. At least as im portant as refuting the existence of fundamental substances which provide the basis for a world independent of human interests and concerns is the refutation of a substantial self, which constitutes the fixed point around which our internal world revolves. Such a substantial self is an essentially unchanging entity, distinct from our physical body and psychological states which unifies our sensory input and mental life and acts as a foundation of our agenthood in the world. Nagarjuna wants to replace this prima facie plausible and compelling view of a self which, however, he claims to be mistaken by a conception of the self as a set of causally interconnected physical and psychological events. He sets

out to account for the fact th at we normally do not see ourselves in this way by arguing that this set of events is usually under the misapprehension of its own properties, i.e. it sees itself as a substantial self, even though it is not. It is interesting to note that this alternative view of the self presented here (which, to be sure, is not a Madhyamaka speciality but widely shared between different Buddhist traditions) despite its intuitive implausibility finds a surprising amount of support in recent research on cognitive science. Of particular interest in this context is the so-called narrative view of the self, a theory which has been explored in detail by Daniel Dennett,14 who also present supporting evidence from our current knowledge of how the brain works. One of Dennett’s central observations is that the processing of neurophysiologically encoded information is spread out across the entire brain. There is no place in the brain where ‘it all comes together’, no ‘Cartesian theatre where the stream of sensory information is unified into mental content and presented to consciousness. He argues that not only is there no neurophysiological analogue to the self anywhere in the spatial organization of the brain, the temporal sequence of events in the brain also cannot be used as a foundation of a continuous self. Dennett shows that in certain cases the order of events as they appear in our consciousness does not line up with the temporal order of their underlying neurophysiological bases.15 The view of our selves as continuous, temporally extended entities therefore cannot be seen as a mere reflection of a series of events in the brain, but requires a significant deal of conceptual construction. Our subjective feelings of spatial and temporal location cannot be grounded on the spatially and temporally spread out, discontinuous series of events in the brain in a straightforward 14Most famously in Dennett (1991). 10Dennett (1991, 134-138).

manner. Onr view of the self as an essentially unchanging unifier and agent cannot be based on our biological makeup in the same way as our view of the nature of the centre of gravity (another conceptual construction) of some object cannot be based on the structure of the piece of m atter which occupies the space where we locate the centre of gravity. Dennett argues instead th at the self is a product of our linguistic capacities. The capacity to use language is hard-wired into our brain, and once we start using language we tell stories, including stories about ourselves which continuously create th at very self. The self emerging on this theory is not the author, but the authored. Dennett notes that ‘our tales are spun, but for the most part we don’t spin them; they spin us. The human consciousness, and our narrative selfhood, is the product, not their source.’16 For this reason there is no fundamental difference between the self created by our own narrative and the selves created in works of fiction.17 It is not the case th at the former are intrinsically more real than the latter, in fact they belong fundamentally to the same class of things (even though the fictional selves, unlike our own narrative selves, are usually not open-ended). Both are conceptual constructs produced by regarding a narrative, our own or that in some text, as revolving around a single fixed point. Assuming we accept the view of the self as a conceptual construction superimposed on a collection of physico-psychological events we might still ask ourselves what the point of all this constructing is. Why do we spin these narratives which in turn cause us to misapprehend the nature of the self, thinking th at there is a substantial self where in fact there is only a set of intersecting narratives? Some cognitive scientists have proposed evolutionary 16Dennett (1991, 418). 17Dennett (1991, 410-411), Dennett (1992, 105-111).

reasons for this phenomenon. Thomas Metzinger suggests th at we should look at the human self-model as a neurocomputational weapon, a certain data structure th at the brain can activate from time to time such as when you have to wake up in the morning and integrate your sensory perceptions with your motor behaviour. The ego machine just turns on its phenomenal self, and th at is the moment when you come to be. To have a good self-model means to be successful in a certain environment. It starts with simple properties: you need to know how far you can jump, what your body can do, how big you are, what your boundaries are, so th at you don't start to eat your own legs, as some primitive animals may actually do, or as some psychiatrically disturbed people do.18 According to this interpretation our intuitive view of ourselves as substanceselves is to be understood as a pragmatically successful self-deception. A self-model along the lines of a substance-self allows us to respond to many tasks more quickly than a more cumbersome one based on the notion of a process would. This is the reason why this sense of the self has spread so widely, as it provides the minds who hold it with an evolutionary advantage. It thus became the dominant and most natural way to see ourselves.19 Despite its popularity with cognitive scientists the reductionist view of the self as nothing but a causally connected chain of physical and psychological events faces some obvious philosophical problems. The first problem concerns the unity of a person. It is not clear whether reductionism can ac- 18Blackmore (2005, 153). 19An interesting account of the psychological consequences of the loss of such a selfmodel from the first-person (!) perspective is given in Segal (1998).

tually account for the boundaries between different persons in the right way. For suppose I decide to make a sound. This means that there is a casual sequence involving a psychological event (the decision) and a physical event (my making the sound). You hear th at sound and later remember it. Given th at all these events are causally connected to one another, how do we draw the line between the causally connected chain which constitutes ‘me’ and that which constitutes ‘you’ without already presupposing the concept of a person? If we think of causally connected chains of physical and psychological events it seems as if there is only one big network of these, without any obvious ways of dividing them into persons. A second problem arises with the rationality of prudential concern for ourselves. We usually think that it is rational to show concern for future stages of our selves, so that e.g. we buy an umbrella today so that we don’t get wet tomorrow. Similarly arguments built around the notion of karma presuppose in the same way that we should care about what happens to us in the future. But if a person is nothing but a logical fiction built on a succession of momentary psycho-physical events, how could such a fiction exhibit prudential concern? After all the person is never present at any particular temporal stage to function as a potential subject of such concern. Addressing these difficulties would obviously require a long and careful discussion of the implications of a reductionist theory of persons. This will not be attempted here, for two main reasons. First of all there exists now a philosophical literature of considerable depth and sophistication dealing with issues arising from a reductionist view of persons. In Western philosophy this developed as a reaction to Derek Parfit’s influential monograph Reasons and Persons. This describes a reductionist view of persons which the author

regards as fundamentally the same as the one found in Buddhist texts.20 It would neither be possible nor desirable to repeat the resulting discussion here. Secondly, the theory of persons described above is no position specific to Madhyamikas, but something shared by all main Buddhist schools. A prolonged discussion of these matters would therefore take us away from the main Madhyamaka focus of this study. At this place I would just like to sketch briefly two key concepts one could use to address the problems of unity and prudential concern, based on the discussion in Siderits (2003). In dealing with the first problem it is useful to establish the concept of a maximally causally connected set of psychophysical events.21 This is a set which we make as large as possible whilst maximizing the causal connectedness of the set. For example we will only include events connected with specific body-parts (such as the vocal chords and the ear) if they stand in continuous causal connection over time. While your ear-event might be causally connected to my vocal-chord event (because you hear me) and could therefore be included in the set doing so would reduce the overall causal connectivity of the set, since over time there will be fewer causal connections between my voice and your ear than between my voice and my ear. Such sets could then be made more and more comprehensive and could be regarded as reductionist substitutes for the notion of a person. One avenue to explore for the establishment of prudential concern is the idea that the concept of a person as a trans-temporal, non-momentary entity is a convention accepted because of its consequences. Given that there are several conventions which we could have adopted (such as the convention of momentary beings without temporal extension, or the convention of one 20Parfit (1984, 273, 280). More recent discussions of the issues involved here can be found in Siderits (2003), particularly in chapters 2 and 3, as well as in Albahari (2006). 21Siderits (2003, 45-46).

overarching mind of which everything else forms a part) the one we chose in the end must have something to recommend it. This something might either be cashed out in evolutionary terms, arguing that this person-convention just provided its bearers with the greatest survival value,22 or it could be given a normative justification. Here the idea is th at conceptualizing a causal sequence of psycho-physical events using the concept of a person rather than some other one maximizes utility. For example, under such a conceptualization it is much less likely that minor immediate pleasures will be traded in for major future pains. Moreover, the idea of future pain could be used as a deterrent, whereas under the conception of a person as a momentary entity this idea of punishment would not have much force.23 This utilitarian defence of the concept of a person of course presupposes that the concept of pain (which is what is to be minimized) does not bring in persons again through the back door. But assuming this could be done this seems to leave us with a sensible way of accounting for prudential concern against a reductionist background.

10.3 Ethics

Very little has been said in the previous pages on Nagarjuna’s ethical theory. Apart from the RA and some verses in the MMK most of his remarks on ethics are found in such works as the Suhrllekha2i and the Shes rab sdong 22Siderits (2003, 43). 23Siderits (2003, 39). 24Nagarjuna (2002), Nagarjuna (1979), Lindtner (1982, 218-224).

bu.25 The former text, which enjoyed considerable popularity in Tibet,26 presents the reader with concrete ethical advice for the layman; it stresses the importance of compassion and describes karmic consequences of various kinds of behaviour. The Shes rab dong bu or ‘Tree of Wisdom’ is a collection of aphorisms dealing with maxims for ethical behaviour, drawn from the M ahabharata, the Pancatantra and the Puranas,27 reminiscent of Sa skya Pandita’s Legs bshad rin po che’i gter28 who may have borrowed from it.29 While these works contain a considerable amount of discussion of ethical topics I have chosen not to analyse them in detail, but to confine my discussion to the remarks in this section. To see why we have to note th at we can distinguish three different kinds of ethical statements in Madhyamaka texts. First of all there are the ethical pronouncements made by a Madhyamika such as Nagarjuna as part of an exposition of the Buddhist doctrine which he, as a Buddhist, incorporates into his teachings. These will include remarks about the relative consequences of meritorious and non-meritorious actions, attachment as the primary cause of suffering, the importance of compassion and so on. Secondly we find an ethical discussion in connection with the concept of emptiness, in particular with the emptiness of persons. If there is no substantial self we might wonder who the agent of an action or the experiencer of a result, or the subject and object of compassion really are. The Buddhist propounder of a non-self has to give a re-interpretation of these notions 25Nagarjuna (1919). Lindtner (1982, note 29, page 15) regards this text as ‘dubious’ but ‘perhaps authentic’. Pathak (1974, 38) disagrees and ascribes this, as well as some other niti texts not to Nagarjuna the Madhyamika but to a pre-8th century compiler of the same name. See also Ruegg (1981, 27). 26Nagarjuna (2002, 1-2). 27Pathak (1974, 38). 28Pathak (1974, 78). 29Nagarjuna (1919, iv).

without the tacit presupposition of a substantial self underlying all of them. We should note th at this is not a challenge which applies specifically to the Madhyamika, but to any theorist holding a non-self view in conjunction with the common Buddhist ethical view. It is thus equally a task for a Vaibhasika, Theravadin or Sautrantika to give a theoretical account of this issue. The third kind of statements deal with the specific ethical consequences of Madhyamaka views. For example, the particularly Nagarjunian view th at there is no ultimate difference between samsara and nirvana demands an explanation of why we should engage in meritorious rather than non-meritorious actions, or Buddhist practices in general: if there is no difference between the liberated and the non-liberated state in any case, why bother? Most of Nagarjuna’s ethical remarks fall into the first, and some also into the second class. Those in the first class, though interesting in the context of Nagarjuna’s Buddhist worldview shed relatively little light on philosophical problems, especially concerning Nagarjuna’s most original thoughts, i.e. the metaphysical and epistemological considerations which form the heart of Madhyamaka philosophy. In the context of remarks in the second class, dealing with the relationship between ethical issues and the concept of emptiness Nagarjuna extends his analysis of phenomena also to such key ethical notions as pleasure and pain and notes that these too do not exist by substance~svabhava. In the RA he observes th a t30 Physical feelings of pleasure are only a lessening of pain. Perceptions [and the pleasures they produce] are made of thought, they are created only by conceptuality. 30 duhkapratikriyamatram sarvram vedanasukham / samjnam ayam manasam tu kevalam kalpanakrtam. RA 3:47.

Pleasure and pain therefore cannot be treated as basic reals on which our system of ethics could be based (for example in the form of some sort of utilitarian calculus aimed at maximizing pleasure), since they exist interdependently. There could be no pleasure in the absence of pain, or vice versa. But this then implies that neither could exist by substance-svabhava. Furthermore, if the extent to which a certain situation is regarded as painful or pleasurable depends on the way it is conceptualized, being painful or pleasurable is shown to be no intrinsic property of a part of the world out there, but something arising from the interaction between a conceptualizing subject and a conceptualized object.31 The question of the compatibility of the emptiness of the self and the notion of karma is raised in the MMK. Nagarjuna notes th a t32 If an action were uncreated fear would arise of encountering something not done. [...] It would be impossible to draw a distinction between merit and demerit. The worry here is that if there was no substantial self creating actions there would be no way of ascribing individual actions to individual selves, since there are no such selves, but only complexes of psychophysical events. But then it may happen that one experiences the consequences of a deed which one did not do, i.e. which arises from a different such complex. This would then not give us any way of differentiating actions into wholesome actions as those which have pleasant consequences, and unwholesome ones, which have unpleasant consequences. Moreover, on the analysis in terms of psychophysical complexes the entire set of distinctions into action, agent, consequence 31 See also RA 3:50. 32 akrta-abhyagama-bhayam sydt karakrtakam yadi / [ . . . ] / punyapapakrtor naiva pravibhagas ca yujyate. MMK 17:23a, 24b.

and experiencer of the consequence,33 which are of central importance for the entire system of Buddhist ethics disappears. As already noted in chapter 7 this worry may be answered by the example of the illusion created within an illusion. The fundamental mistake of an insubstantial self to regard itself as substantial creates the concepts of agent, action, consequent and experiencer which then in turn bring with them the whole system of karmic interrelations. Unfortunately this belief is so fundamental that mere intellectual understanding of the non-existence of substantial selves does not stop such selves from appearing to us. In the same way the understanding that some phenomenon is an optical illusion generally does not alter the way it appears to us, but at best how seriously we take this appearance. W hat is needed for the disappearance of such concepts as agent, action, consequence and experience is the realization of the non-substantiality of the self, that is the attainm ent of a cognitive shift which keeps the mistaken notion of the substantial self from arising. Remarks dealing with the ethical repercussions of emptiness like those just discussed are relatively rare in the works of Nagarjuna, Analyses belonging to the third class, i.e. those dealing with the specific ethical consequences of Madhyamaka thought are virtually absent. A major issue presenting itself at this point is the question to which extent there is fundamental relation between the Madhyamaka theory of emptiness and the ethical theory of compassion centred on the ideal of the bodhisattva. After all it seems th at a case for such an ethical theory can already be made from the perspective of the ‘lower schools’. If there are no substantial selves and therefore no psychological states — such as pains — essentially attached to selves all these unpersonal pains can be regarded as equally bad, irrespective of their location. But in 33 MMK 17:29-30.

this case my reason for removing my own pain is not more pressing than th at of removing the pain of other beings, in fact it is considerably less pressing, since the pains of other beings outnumber my own. We thus seem to be able to get relatively close to Mahayana ethics on the basis of Hlnayana metaphysics. So what is the distinctive advantage of the Madhyamaka theory of emptiness for establishing the ethical ideal of a bodhisattval There is much to be said about this question as well as others arising in the same context34, but the basis of such answers in Nagarjuna’s writings on Madhyamaka is at best implicit. The examination of these issues becomes considerably more interesting when we take into account later Madhyamaka texts which address questions dealing with the distinctive Madhyamaka consequences for ethics explicitly and in greater detail.35 We can imagine a variety of possible reasons why we find so little discussion of these m atters in Nagarjuna’s works. One obvious possibility is th at the respective text or texts were lost relatively early in the tradition. Alternatively it might have been the case that Nagarjuna’s focus of interest when developing the Madhyamaka approach was a set of metaphysical and epistemological questions, and that its ethical dimensions were only explored in detail by later writers. A final possibility is that discussions of the point where the perfections of wisdom and compassion join were regarded as too advanced to be put down in writing and were only transm itted orally. Whatever the explanation may be, the fact remains that the investigation of Madhyamaka ethics will find a more extensive set of data in later writers than in what is preserved in Nagarjuna’s Madhyamaka. 34For example Siderits (2003, chapter 9). 35An obvious source in this context is the BCA. See Williams (1998).

10.4 Epistemology

Nagarjuna’s account of epistemology is supposed to fulfill a purpose both at the object- as well as at the meta-level. At the object-level means of knowledge and their objects are just another set of central concepts which have to be investigated as entities potentially existing with svabhava. At the meta-level Nagarjuna’s theory of epistemology is supposed to present the theoretical background of his own account of emptiness. As the theory of emptiness is something we are supposed acquire knowledge of it is essential to get clear about the means by which we are supposed to do so, and indeed about what our object of knowledge consists of in this case. These two projects are inherently interconnected. For according to the standard Nyaya theory of epistemology Nagarjuna encountered knowledge is acquired by using a set of procedures (such as perception or inference) the nature of which it is to produce knowledge and which convey information about a set of objective, mind-independent set of individuals which are the bearers of specific qualities. But a theory which thus presupposed the existence of objects of knowledge with distinct natures which the means of knowledge could adequately represent could hardly be used as a basis for knowing emptiness, for it presupposes exactly what the theory of emptiness denies. A substantial part of Nagarjuna’s epistemological discussion is therefore dedicated to a criticism of the standard Nyaya theory of knowledge, Nagarjuna sets out to establish th at nothing can be regarded as intrinsically a means or object of knowledge. Means of knowledge and their objects have to be mutually established: the means of knowledge establishes the object by giving us cognitive access to it, our successful interaction with the object establishes the means of knowledge as a trustworthy route to the object.

Something will therefore be classified as a means or object of knowledge not because this is a reflection of its intrinsic nature, but because it is regarded as such once a reflective equilibrium has been reached. We use beliefs about the nature of the object in order to test our hypotheses concerning the means of acquiring such beliefs, these hypotheses are then in turn used to assess our view of the nature of the object. The reason why this could not lead to an establishment of the means of knowledge in the way the Naiyayika wants is that a different initial set of beliefs could have led to a different reflective equilibrium as a result. But given th at each would have led to a different view of reliable means of cognition and of the objects known we would not be able to determine which of the two faithfully reflects the nature of the means and objects of cognition. Assuming that establishing a reflective equilibrium is the only way of arriving at an account of the means and objects of cognition it is therefore impossible to establish the true nature of either. This criticism of the Nyaya position has been considered to be unsatisfactory by Mark Siderits in recent work.36 Siderits* main point is th at it relies on an internalist conception of knowledge according to which the justification for a knowledge-claim also has to be known to the subject. For the internalist it is not just sufficient to respond to a sceptical threat by demonstrating th at the subject is justified, it also has to know to be justified. It is thus not sufficient that the method of the reflective equilibrium might as a m atter of fact supply us with the right account of means of knowledge and their objects, but we also have to know th at it is the right account. But given the fact th at there can be several such equilibria the internalist fails to be justified. Siderits argues that the Nyaya view of veridical cognition as the 36(20 00 , 227).

product of a reliable causal process cannot be subsumed under such internalism, 37 indeed it is a typical externalist position where the justification for knowing something is located outside of the body of the subject’s knowledge. But if this is the case then the Madhyamika’s criticism loses its force, for it is now based on an assumption - namely epistemic internalism - which its opponent does not share. The force of this criticism is undermined to a certain extent by the fact th at the identification of the opponent Nagarjuna criticizes in individual arguments is notoriously difficult to establish by any but systematic reasons. While the heavy influence of Nyaya thought on the epistemological discussion in the VV is obvious this does not necessarily entail th at the opponent addressed in the passages dealing with the mutual establishment of means and objects of knowledge38 is a Naiyayika as well. We might equally use Siderits’ argument to claim th at the implicit internalist position entails that he cannot have been one, as the criticism would not have applied to him otherwise. While such questions are impossible to decide on the basis of the textual evidence available to us it is interesting to consider what kind of argument Nagarjuna could have used in his criticism of an externalist position. Siderits39 makes the interesting suggestion of employing Nagarjuna’s analysis of causation, which does not feature much in the VV and is not put to any epistemological use in the MMK. As the reader will recall Nagarjuna argues th at causation, far from being a mind-independent relation which establishes objective connections between phenomena is itself intricately bound up with conceptualization. As was argued above we cannot conceive of a causal re37( 2000, 223). 38VV 46-48. 39(2000, 229).

lation between two entities without a substantial mental contribution. If we plausibly assume th at the cause precedes the effect then, at the time of the existence of the cause, when the effect does not yet exist, our mind will have to supply the missing relatum in our expectation. Moreover, it is not just the causal relation which cannot be regarded as mind-independent, as it also plays an essential part in the construction of objects. This is because the very establishment of miscellaneous collections of entities as unified items called ‘cause’ (or ‘causal field’) and ‘effect’ could not proceed without the notion of causation in the first place. Causation cannot be regarded as a relation connecting items which are ‘there anyway’ in a mind-independent way. But now it is clear that if this criticism goes through then causation cannot be made to bear the epistemological weight the Naiyayika wants it to bear. For the externalist regards causation as a guarantor transm itting features of the object to perception in a reliable way because the nature of the causing object will determine the nature of the perception which is the effect. As the Madhyamika has argued, however, causation is itself conceptually constructed. A reliable means of cognition which incorporates causation as a central element therefore has to be conceptually constructed in the same way. There is therefore no way in which we can regard it as providing us with accurate knowledge of an objectively existent world independent of human conceptual practices. Nagarjuna therefore argues for an epistemology in which nothing is intrinsically a means or an object of knowledge. And if nothing is intrinsically a means of knowledge there is also nothing which could function as such a means in any context; it is only against a specific background th at it could fulfill such a role. Such an epistemology is able to provide a background for the knowledge of emptiness as means and objects of knowledge are no longer

regarded as being means and objects of knowledge intrinsically. Means of knowledge are such means only in specific contexts, and they are not supposed to adequately reflect the properties objects have from their own side, but provide the basis of successful interaction with them. The theory of emptiness therefore no longer contradicts the epistemology on the basis of which it is to be known. 10.5 Language and Truth As was mentioned earlier there exists no fully-formed Madhyamaka theory of language or tru th in Nagarjuna’s extant writings. This does not mean, however, th at his works do not give us a fairly good indication of what his views on some of the key questions within this area were. First of all it is apparent that the Madhyamaka theory of emptiness is not compatible with a the idea of a ‘ready-made world’, th at is of a world which exists independently of human interests and concerns and already shows a particular kind of structuring which our structured language could then set out to reflect. If nothing exists with substance-svabhava nothing in the world could exist from its own side and nothing could bear a structure which is intrinsic to it, rather than something ascribed to it from the outside. Moreover, the Madhyamika will reject the classic correspondence account of truth, according to which the tru th of a statement is grounded in a similarity of structure between a statement and the bit of the world it refers to. This also entails a rejection of the corresponding view of how language works, namely that our sentences manage to connect up with the world via a set of objectively existent structural similarities. The main reason for this rejection is that the Madhyamika cannot find any sufficiently substantial reCH

lation which would allow us to bind together world and word at the most fundamental level. The most plausible candidate for linking words and their referents is the causal relation, for example by using it to construct a causal chain from an ‘initial baptism 5 to our present use of the term. But as Nagarjuna has argued in detail the causal relation itself is conceptually constructed. But if causation cannot be regarded as a relation which functions objectively, independent of the concepts we employ then it can hardly be regarded as a mind-independent way of founding the relationship between language and the world.40 An alternative account which the Madhyamika might want to adopt is one which does not conceive of truth in terms of correspondence with an exterior reality but rather in terms of assertability conditions. In this case a statement is regarded as true if conditions obtain which warrant our asserting the statement. W hat makes the statement th at water is wet true is not a structural correspondence between it and a fact about water, but th at we have something which justifies us in making this statement. W hat this justification consists in depends on the further details of our theory of truth; it might be based on facts about empirical observation, about coherence with other beliefs, about pragmatic success and so forth. This view of course implies that there could not be any truths which are in principle beyond our ability to verify them. This is because we could never have a warrant for asserting such statements, and the existence of such a warrant is precisely what we consider the tru th of the statements to consist in. Such statements would have to be regarded as lacking a truth-value. This kind of denial of verificationtranscendent truths in turn agrees very well with Nagarjuna’s contextualist epistemology. For if nothing is intrinsically a means of knowledge nothing 40See Siderits (2003, 166).

can be intrinsically beyond the grasp of such means of knowledge either. As what constitutes a means of knowledge is context-dependent, th at a certain truth cannot be accessed by some means of knowledge is context-dependent too. There is no context-independent concept of knowledge we could use to form the idea of a truth which lies beyond all epistemic contexts. According to the Madhyamaka view of truth there can be no such thing as ultimate truth, a theory describing how things really are, independent of our interests and conceptual resources employed in describing it. All one is left with is conventional truth, truth with consists in agreement with commonly accepted practices and conventions. These are the truths which are arrived at when viewing the world through our linguistically formed conceptual framework. But we should be wary of denigrating these conventions as a distorting device which incorporates our specific interests and concerns. The very notion of ‘distortion’ presupposes that there is a world untainted by conceptuality out there (even if our minds can never reach it) which is crooked and bent to fit our cognitive grasp. But the very notion of such a ‘way things really are’ is argued by the Madhyamika to be incoherent. There is no way of investigating the world apart from our linguistic and conceptual practices, if only because these practices generate the notion of the ‘world’ and of the ‘objects’ in it in the first place. To speak of conventional reality as distorted is therefore highly misleading, unless all we want to say is that our way of investigating the world is inextricably bound up with the linguistic and conceptual framework we happen to employ. There are two worries one might have with the rejection of the notion of an ultimate truth. First of all one might think that progress in human inquiry requires th at we question what we now believe to be truths and perhaps replace them by other beliefs. Even a cursory acquaintance with the

history of science will show that we are where we are now only through a persistent process of replacing beliefs we once held to be true, but no longer think so. But it seems hard to explain what our justification for this is if it is not trying to bring our beliefs into greater accordance with the way things are. All we are ever seem to dealing with according to the Madhyamaka view is a purely immanent notion of tru th where the only kind of tru th we have access to is a reflection of conventional human practices and agreements. In response to this the Madhyamika might want to make the point that it is at least sometimes advantageous to treat truths as if they had a more than conventional grounding, th at is as if they were not just the product of agreement with commonly accepted practices and conventions. This is precisely because such practices need improvement from time to time, and since a spirit of inquiry is facilitated more by the idea that there is a mind independent truth waiting to be discovered.41 The Madhyamika could thus argue that for pragmatic reasons we should conceive of truths as reflections of an objective, external reality even though we do not think that there are any such truths in fact. We might object at this point that if the notion of the existence of at least some verification-transcendent truths is pragmatically useful, whoever believes in truth as warranted assertability then has to believe that some truths are not conventional, since asserting this is now supported by a warrant. But this will not just turn the anti-realist into a realist against his will, since his embracing of non-conventional truths is dictated by purely practical concerns: we are considerably better off if we build our inquiries 011 the convenient fiction of non-conventional truths. But they remain just that, namely conventional fictions; the anti-realist does not think, as the realist does, that the existence of such truths is in any way grounded by the way 41Siderits (2003, 183-184).

the world is independent of our interests and concerns. Another worry with the Madhyamika’s rejection of an ultimate truth is th at emptiness cannot then be regarded as the ultimate truth either. But surely, one will argue, for the Madhyamika emptiness is the end-product of the correct analysis of phenomena, and thereby indicative of the way things really are? As was argued in section 2.1.3 of chapter 2 the theory of emptiness is not to be seen as a description of reality as it is independent of human conceptual conventions, as its main purpose it to combat the wrong ascription of svabhava to things. The absence of svabhava is nothing phenomena have within themselves, but only something which is projected onto them from the outside in an attempt to rectify a mistaken cognition. Therefore the theory of emptiness is not to be regarded as an ultimately true theory either. Such a theory would describe things as they are independent of human interests and concerns. But the theory of emptiness is intricately bound up with such interests and concerns: if there were no human minds who mistakenly read the existence of svabhava into phenomena which lack it there would be no point in having a theory to correct this. It is only due to our erroneous view of things that the theory of emptiness is required as a corrective. A final problem with the Madhyamaka focus on conventional truth might be the fact that it entails a form of relativism we find unacceptable.42 If we This is a problem which despite its considerable complexity has not been explicitly discussed in Indian sources. It is tempting to speculate why relativism does not seem to have been a prominent problem for Indian Madhyamaka authors. Siderits (1989, 240) suggest that ‘the hegemonistic strategy which Brahmanic culture used to subsume the other cultures of the Indian subcontinent’ might have been responsible for this. It is true that if diverging perspectives are incorporated into one’s own view (perhaps as displaying a restricted understanding of some aspect of this view) rather than seen as independent and incommensurable views of the same subject-matter the problem of relativism loses much of its force.

regard truth as being a matter of warranted assertibility and not at something settled by ‘what the facts are’ we will have to agree that as human practices change, so do standards of warranted assertibility. But then it might be the case that what one culture regards as true, or rationally acceptable, or for th at matter as ethically acceptable, is very different from what we regard as true or acceptable. We will not then be in any way justified in criticizing their practices since there is no objective reality according to which they could be regarded as wrong. If some culture’s standards of warranted assertibility lead it to believe that the earth is hollow, that counter-induction is the best methodology for natural science and that female infanticide is morally commendable there is nothing for us to do save observing that these practices differ from ours.

W hat the Madhyamika should want to argue at this point is th at any culture with which we can interact at all, that is one which shares a form of life with us is one which shares with us at least some evaluative standards. If it did not we would not be able to ascribe to it anything like rational forms of belief formation or ethical norms, so that the whole idea of rational or ethical divergence and rational or ethical criticism would lose its point.43 The Madhyamika could then argue that even though different cultures can have different standards none of which can be regarded as ultimately true (since there is no such thing as ultimate truth) still some standards can be seen to be better than others, for example in terms of overall coherence with our practices (which are also a part of conventional truth) or in terms of their ability to reduce pain. Siderits44 discusses the interesting example of the conception of the self by Prasangika and Svatantrika writers. While 43This is essentially the point argued in Davidson (1973-1974, 19): ‘whether we like it or not, if we want to understand others, we must count them right in most matters’. 44Siderits (1989, 241-243). See also (2003, 206).

for the Prasahgikas the self is a mere label superimposed on the group of elements the Svatantrika regards it as a continuous series of inner moments of consciousness which take their inner states as objects. Of course the latter do not think that the self has any degree of ultimate reality but that amongst the variety of aggregates which make up the self we can identify one candidate (i.e. that part of our mind is aware of its own psychological states) which best coheres with our cognitive practices. So while there is no ‘best’ candidate amongst the entities which me might potentially identify with the self, because ultimately there is no such thing, according to the Svatantrika reading at least some candidates may be better than others. Tillemans mentions the interesting example of the wine-taster in illustration of this point. It is generally agreed t at secondary properties, and particularly olfactory and gustatory properties do not have a mindindependent existence. If there were no human beings around there would not be the properties of tasting sour or smelling sweet, since these are not properties existing in the objects themselves, but are only produced by interaction of the objects and our sensory faculties. Nevertheless, despite this mind-dependence we might want to argue that some ascriptions of secondary properties have more than a subjective validity. A wine-taster describing a wine as tasting acidic may be wrong, even though ultimately there is no property of tasting acidic which the wine has or lacks. Even within the realm of conventional truth we therefore do not ‘make it all up! but there are ways of ranking different conventional statements in terms of better or worse, even though there is no best, i.e. ultimately true account. While this is an enlightening example the case the critic worried about relativism is concerned with is probably less like the case of wine, and more 45 (2003, 110-111).

like th at of phenolthiourea. For genetic reasons this substance tastes bitter to about three-quarters of the population, while it is tasteless to the rest. Now if we separated the two populations we seem to end up with a case where one group has no reason to criticize the other’s taste-judgement as incorrect, because ultimately there is no fact to the m atter what phenolthiourea tastes like. And in this case it seems impossible to rank one taste-judgement as conventionally better or worse than another one. The only thing we could bring forward in response to this point seems to be the familiar Davidsonian observation th at if such disagreement between two cultures was widespread, i.e. if it did not just concern simple taste sensations but also more complex factual and moral judgements there would just be no basis for the two cultures to interact at all. Since their standards of rational justification or morality would be so different from ours the whole notion of factual and ethical criticism would lose its meaning. We would therefore have to rely on the assumption th at no two cultures which can interact would differ as radically in their conventions as illustrated by the example of phenolthiourea. In order to understand Nagarj una’s project as a philosophically coherent enterprise it is useful take into account the ethical and soteriological implications of different standards one of which might be better than another, but none of which can be best in the sense of corresponding to the way reality really is. For Nagarjuna the conception of truth supported by the way things really are presents a subtle object of clinging and thereby ultimately a source of suffering. Such clinging is not is not as coarse as clinging to possessions, one’s body, or one’s self, but it still generates a kind of attachm ent which in turn supports a sense of selfhood as a subject who has realized the way things really are. For the Madhyamika, in order to become truly selfless on has to give up the view th at we can obtain anything more than conventional truths,

some of which might be evaluated as better than others, but none of which can constitute the last word. The resulting epistemic humility is therefore a product of considerations of selfhood and ethics seen as interlinked with considerations of tru th and reality. Bibliography Albahari, M.: 2006, Analytical Buddhism. The Two-Tiered Illusion of the Self, Palgrave Macmillan, Houndmills; New York. Ames, W,: 1982, The notion of svabhava in the thought of CandrakTrti., Journal of Indian Philosophy 10, 161-177. Ames, W.: 1995, Bhavaviveka’s Prajhapradvpa. A translation of chapter two., Journal of Indian Philosophy 23, 295-365. Armstrong, D. M.: 1978, Nominalism & Realism.) Cambridge University Press, Cambridge. Armstrong, D. M.: 1997, A World of States of Affairs., Cambridge University Press, Cambridge. Bacon, J.: 1995, Universals and Property Instances. The Alphabet of Being., Blackwell, Oxford, Cambridge MA. Bharadwaja, V.: 1984, Rationality, argumentation and embarassment: A study of the four logical alternatives (catuskoti) in Buddhist logic., Philosophy East and West 34(3), 303-319. Bhattacharya, K.: 1973, L ’atman-brahman dans le bouddhisme ancien., Ecole frangaise d 1 Extreme-Orient, Paris.

BIBLIO G RAPH Y 332

Bhattacharya, K.: 1977, On the relationship between the Vigrahavyavartani and the Nyayasutras., Journal of Indo-European Studies 5(2-3), 265- 273. Bhattacharya, K.: 1980, Nagarjuna’s arguments against motion: their grammatical basis., A Corpus of Indian Studies. Essays in Honour of Professor Gaurinath Sastrf Sanskrit Pustak Bhandar, Calcutta, pp. 85-95. Bhattacharya, K.: 1980-1981, The grammatical basis of Nagarjuna’s arguments: some further considerations., Indologica Taurinensia 8 , 35-43. Bhattacharya, K.: 1985, Nagarjuna’s arguments against motion., Journal of the International Association of Buddhist Studies 8 , 7-15. Bhattacharya, K.: 1999, Nagarjuna’s Vigrahavyavartani, in K. Potter (ed.), Encyclopedia of Indian Philosophies, Vol. 8 , Motilal Banarsidass, Delhi, pp. 124-133. Bhattacharya, K., Johnston, E. and Kunst, A.: 1978, The Dialectical Method of Ndgdrjuna., Motilal Banarsidass, Delhi. Bijalwan, C.: 1977, Indian Theory of Knowledge based upon Jayanta’s Nyayamanjan., Heritage Publishers, New Delhi. Bikkhu Bodhi (ed.): 2000, The Connected Discourses of the Buddha., Wisdom, Boston. Bikkhu Nalamoli and Bikkhu Bodhi (eds): 2001, The Middle Length Discourses of the Buddha., Wisdom, Boston. Blackmore, S.: 2005, Conversations on Consciousness., Oxford University Press, Oxford. B IBLIO G RAPH Y 333 Bronkhorst, J.: 1985, Nagarjuna and the Naiyayikas., Journal of Indian Philosophy 13, 107-132. Bronkhorst, J.: 1993, On the method of interpreting philosophical Sanskrit texts., Asiatische Studien 47(3), 501-511. Brunnholzl, K.: 2004, The Center of the Sunlit Sky: Madhyamaka in the Kagyii Tradition., Ithaca, Snow Lion. Buddhagliosa: 1991, The Path of Purification., Buddhist Publication Society, Kandy, Sri Lanka. Translated by Bikkhu Nanamoli. Bugault, G.: 2001, Stances du millieu par excellence., Gallimard, Paris. Burton, D.: 1999, Emptiness Appraised: A Critical Study of Nagdrjuna’s Philosophy., Surrey. CandrakTrti: 1999, Dbu ma bzhi brgya pa’i rgya cher 1grel pa, Cultural and Welfare Society, Leh. Cardona, G.: 1967, Negations in Paninian rules., Language 43(1), 34-56. Chakravarti, A.: 1982, The Nyaya proofs for the existence of the soul., Journal of Indian Philosophy 10, 211-238. Chakravarti, P.: 1930, The Philosophy of Sanskrit Grammar., University of Calcutta, Calcutta. Chakravarti, S.: 1980, The Madhyamika catuskoti or tetralemma., Journal of Indian Philosophy 8 , 303-306. Chatterjee, S.: 1939, The Nyaya Theory of Knowledge., University of Calcutta Press, Calcutta. BIBLIO G RAPH Y 334 Chattopadhyaya, D. and Gangopadhyaya, M.: 1968, Nyaya Philosophy. Literal translation of Gautama’s Nyaya-sutra & Vatsydy ana’s Bhasya along with a free and abridged translation of the Elucidation by Mahamahopadhyaya Phanibhusana Tarkavdgisa., Indian Studies, Calcutta. Cheng, H.: 1980, Motion and rest in the Middle Treatise., Journal of Chinese Philosophy 7, 229-244. Chi, R. S.: 1969, Buddhist Formal Logic., Luzac and Co., London. Chi, R. S.: 1974, Topics on being and logical reasoning., Philosophy East and West 24(3), 293-300. Collins, S.: 1982, Selfless Persons. Imagery and Thought in Theravada Buddhism., Cambridge University Press, Cambridge. Conze, E.: 1963, Spurious parallels to Buddhist philosophy, Philosophy East and West 13, 105-115. Conze, E.: 1967, Buddhist Thought in India., University of Michigan Press. Cox, C.: 1995, Disputed Dharmas: Early Buddhist Theories on Existence., International Institute for Buddhist Studies, Tokyo. D’Almeida, A.: 1973, Nyaya Philosophy. Nature and Validity of Knowledge., Pontifical Institute of Theology and Philosophy, Alwaye. Dasgupta, S.: 1942, A History of Indian Philosophy., Cambridge University Press, Cambridge. Davids, T. R. (ed.): 1890, The Questions of King Milinda., Clarendon Press, Oxford. BIBLIO G RAPH Y 335 Davidson, D.: 1973-1974, On the very idea of a conceptual scheme, Proceedings and Addresses of the American Philosophical Association 47, 5-20. de Jong, J.: 1972, The problem of the absolute in the Madhyamika school., Journal of Indian Philosophy 2, 1-6. de Jong, J.: 1978, Textcritical notes on the Prasannapada., Indo-Iranian Journal 2 0 , 25-59, 217-252. Dennett, D.: 1991, Consciousness Explained., Little, Brown and Co., Boston, London. Dennett, D.: 1992, The self as a center of narrative gravity, Self and Consciousness: Multiple Perspectives, Erlbaum, pp. 103-115. Desideri, I.: 1981-1989, Opere Tibetane di Ippolito Desideri, ISMEO, Rome. Edited by Giuseppe Toscano. Dhammajoti, K.: 2003, The Sarvastivada doctrine of simultaneous causality., Journal of Buddhist Studies 1, 17-54. Dhammajoti, K.: 2004, Sarvastivada Abhidharma., Centre for Buddhist Studies, Sri Lanka. Dowman, K.: 1985, Masters of Mahamudra. Songs and Histories of the Eighty-four Buddhist Siddhas., State University of New York Press, Albany. Dreyfus, G.: 1997, Recognizing Reality. Dharmakirti’s Philosophy and its Tibetan interpretations,, State University of New York Press, Albany. D’Sa, F. X.: 1980, Sabdaprdmanyam in Sahara and Kumdrila. Towards a study of the Mimamsd experience of language., Gerold & Co, Vienna. B IBLIO G RAPH Y 336 Dummett, M.: 1998, Philosophy of mathematics,, in A. Grayling (ed,), Philosophy 2. Further Through the Subject, Oxford University Press, Oxford, pp. 122-196. Dummett, M.: 2000, Elements of Intuitionism., Oxford University Press, Oxford. Edkins, J.: 1893, Chinese Buddhism. A volume of sketches, historical, descriptive, and critical, Kegan Paul, Trench, Triibner, & Co., London. Eliade, M.: 1969, Yoga. Immortality and Freedom., Princeton University Press, Princeton, NJ. Erb, F.: 1997, Sunyatdsaptativrtti: Candrakirti’s Kommentar zu den ‘Siebzig Versen iiber die LeerheitJ des Nagarjuna (Karikas 1-14)., Steiner, Stuttgart. Feer, L. (ed.): 1888, Samutta Nikaya., Pali Text Society, London. Frauwallner, E.: 1973, History of Indian Philosophy., Motilal Banarsidass, Delhi. Frauwallner, E.: 1995, Studies in Abhidharma Literature and the Origins of Buddhist Philosophical Systems., State University of New York Press, New York. Frauwallner, E.: 2003, Geschichte der Indischen Philosophie., Shaker, Aachen. Edited by Andreas Pohlus. Galloway, B.: 1987, Notes on Nagarjuna and Zeno on motion., Journal of the International Association of Buddhist Studies 10, 81-87. BIBLIO G RAPH Y 337 Galloway, B.: 1989, Some logical issues in Madhyamika thought., Journal of Indian Philosophy 17, 1-35. Ganeri, J.: 1999, Semantic Powers. Meaning and the Means of Knowing in Classical Indian Philosophy., Clarendon, Oxford. Ganeri, J.: 2001, Philosophy in Classical India: The Proper Work of Reason., Routledge, London. Garfield, J.: 1994, Dependent co-origination and the emptiness of emptiness: why did Nagarjuna begin with causation?, Philosophy East and West (44), 219-250. Garfield, J.: 1995, The Fundamental Wisdom of the Middle Way. Translation and Commentary of Nagarjuna’s Mulamadhyam,akakarikd., Oxford University Press, Oxford. Garfield, J.: 1996, Emtpiness and positionlessness: do the Madhyamika relinquish all views?, Journal of Indian Philosophy and Religion (1), 1-34. Garfield, J.: 2001, Nagarjuna’s theory of causation: implications sacred and profane., Philosophy East and West 51(4), 507-524. Gautama: 1887, Nyaya-vartikum., Bibliotheca Indica, N.S. 625, Asiatic Society, Calcutta. Edited by Pandit VindhyesvarT Prasad Dube. Ghose, R. N.: 1987, The modality of Nagarjuna’s dialectics., Journal of Indian Philosophy 15, 285-309. Go rams pa bSod nams seng ge: 1988, ITa ba’i shan byed theg mtshog gnad kyi zla zer., Sakya Student’s Union, Sarnath. Gowans, C.: 2003, Philosophy of the Buddha., Routledge, London. BIBLIO G RAPH Y 338 Griffiths, P.: 2000, Review of David Burton’s Emptiness Appraised., Journal of Buddhist Ethics 7, 22-25. Gudmunsen, C.: 1977, Wittgenstein and Buddhism., Harper «Sc Row, New York. Guha, D. C.: 1979, Navya Nyaya System of Logic. Basic Theories & Techniques., Motilal Banarsidass, Delhi. Gunaratne, R.: 1980, The logical form of the catuskoti: A new solution., Philosophy East and West 30(2), 211-238. Gunaratne, R.: 1986, Understanding Nagarjuna’s catuskoti, Philosophy East and West 36(3), 213-234. Gyatso, L.: 2005, The Harmony of Emptiness and Dependent-Arising., second, revised edition edn, Library of Tibetan Works and Archives, Dharamsala. Hadot, P.: 1971, Causa sui, in J. Ritter (ed.), Historisches Worterbuch der Philosophic, Vol. 1, Wissenschaftliche Buchgesellschaft, Darmstadt, pp. 976-977. Hart, W.: 1987, Causation and self-reference., in S. Batlett and P. Suber (eds), Self-reference, Matinus Nijhoff, Dordrecht, Boston, Lancaster, pp. 179-192. Harvey, P.: 1995, The Selfless Mind. Personality, Consciousness and Nirvana in Early Buddhism., Curzon Press, Richmond. Hayes, R. P.: 1994, Nagarjuna’s appeal., Journal of Indian Philosophy 2 2 , 299-378. BIBLIO G RAPH Y 339 Heyting, A.: 1971, Intuitionism. An Introduction., North Holland, Amsterdam. Hinton, C. H.: 1904, The Fourth Dimension., Sonnenschein, London. Hoffman, D. D.: 1998, Visual Intelligence. How We Create What We See., W.W. Norton & Company, New York. Hookham, S.: 1991, The Buddha Within. Tathagatagarbha Doctrine According to the Shentong Interpretation of the Ratnagotravibhaga., State University of New York Press, Albany, NY. Hoornaert, P.: 2004, The Dliarmapala-Bhavaviveka debate as presented in Dharmapala’s commentary to Catuhsataka XVI.23., Kanazawa daigaku bungakubu ronshu kodo kagaku tetsugaku hen 24, 119-149. Hopkins, J.: 1983, Meditation on Emptiness., Wisdom Publications, London. Hopkins, J.: 1998, Buddhist Advice for Living and Liberation; Nagarjuna}s Precious Garland., Snow Lion, Ithaca. Hua, J.-Y.: 1970, Nagarjuna, one or more? A new interpretation of Buddhist hagiography., History of Religions 1 0 , 139-153. Hume, D.: 1896, A Treatise of Human Nature., Clarendon Press, Oxford. Huntington, C,: 1989, The Emptiness of Emptiness. A Study of Early Indian Madhyamika., University of Hawaii Press, Honolulu. Inada, K. K.: 1970, Nagarjuna. A Translation of his Mulam,adhyamakakdrika with an Introductory Essay., Hokuseido Press, Tokyo. Ingalls, D.: 1951, Materials for the Study of Navya-Nydya Logic., Harvard University Press, Cambridge, MA. BIBLIO G RAPH Y 340 Iyengar, H.: 1927, The Ekaslokasastra of Nagarjuna Bodhisattva., Journal of Mysore University 1, 158-162. Jackson, D.: 1987, The Entrance Gate for the Wise (Section III): Sa-skya Pandita on Indian and Tibetan traditions of Pram,ana and Philosophical Debate, Vol. 17 of Wiener Studien zur Tibetologie und Buddhismuskunde, Vienna. Jaimini: 1916, The Purva Mvmdmsa Sutra of Jaimini. Chapters 1-3., Sacred Books of the Hindus, vol 10, Sudhldranatha Vasu, Allahabad. Jam dbyangs bshad pa, Ba so chos kyi rgyal mthsan, sDe drug mkan chen, N. and Bra sti dge shes rin chen don grub: 1972, mNyarn med rje btsun tsong kha pa chen pos mdzad pa}i byang chub lam rim chen m o ’s dka’ ba’i gnad m am s mchan bu bzhi’i sgo nas legs bar bshad par bshad pa theg chen lam gyi gsal sgron, Chos ’phel legs ldan, New Delhi. Jayatilleke, K.: 1963, Early Buddhist Theory of Knowledge., George Allen & Unwin, London. Jayatilleke, K,: 1967, The logic of the four alternatives., Philosophy East and West 17(1), 69-83. Jha, G.: 1939, Gautama’s Nyayasutra., Oriental Book Agency, Poona. Jha, G.: 1984, Nyayadarsanam with Vatsyayana’s Bhasya, Uddyotakara’s Vartika, Vacaspati Misra’s Tatparyatikd and Visvanatha’s Vrtti., Motilal Banarsidass, Delhi. Kajiyama, Y.: 1973, Three kinds of affirmation and two kinds of negation in Buddhist Philosophy., Wiener Zeitschrift fur die Kunde Siid- und Ostasiens und Archiv fur indische Philosophie 17, 161-175. BIBLIO G RAPH Y 341 Kalupahana, D.: 1975, Causality. The Central Philosophy of Buddhism., University Press of Hawaii, Honolulu. Kalupahana, D. J.: 1991, Mumamadhyamakakarika of Nagarjuna., Motilal Banarsidass, Delhi. Kant, I.: 1993, Kritik der reinen Vem unft., Felix Meiner, Hamburg. Kapstein, M.: 2001, Mereological considerations in Vasubandhu’s ‘Proof of Idealism5., Reason’s Traces, Wisdom, Boston, pp. 181-2004. Keyt, D.: 1963, W ittgenstein's notion of an object, Philosophical Quarterly 13, 13-25. Kielhorn, F. (ed.): 1880-1885, The Vydkarana-Mahdbhasya of Patahjali., Government Central Book Depot, Bombay. King, R.: 1999, Indian Philosophy. An Introduction to Hindu and Buddhist Thought., Edinburgh University Press, Edinburgh. Komito, D. R.: 1987, Ndgdrjuna’s Seventy Stanzas: A Buddhist Psychology of Emptiness., Snow Lion, Ithaca, NY. La Vallee Poussin, L. d.: 1908, Bouddhisme: Opinions sur I’histoire de la dogmatique., Gabriel Beauchesne, Paris. La Vallee Poussin, L. d.: 1988-1990, Abhidharmakosabhdsyam., Asian Humanities Press, Berkeley, CA. English translation by Leo M. Pruden. LaBerge, S., Levitan, L. and Dement, W.: 1986, Lucid dreaming: physiological correlates of consciousness during REM sleep., Journal of Mind and Behavior 7, 251-258. BIBLIO G RAPH Y 342 LaBerge, S, and Rheingold, H.: n.d,, Exploring the World of Lucid Dreaming, Ballantine, New York. Le Poidevin, R.: 1988, The principle of reciprocity and a proof of the nonsimultaneity of cause and effect., Ratio 1, 152-162. Le Poidevin, R.: 2003, Travels in Four Dimensions. The Enigmas of Space and Time., Oxford University Press, Oxford. Lewis, D.: 1986a, The paradoxes of time travel., Philosophical Papers, Vol. 2, Oxford University Press, New York, Oxford, pp. 67-80. Lewis, D. K.: 1986b, On the Plurality of Worlds, Blackwell, Oxford. Lindtner, C.: 1982, Nagarjuniana. Studies in the Writings and Philosophy of Nagarjuna., Akademisk Forlag, Copenhagen. Lopez, D.: 1987, A Study of Svatantrika., Snow Lion, Ithaca, NY. Lopez, D.: 1994, dGe ’dun Chos ’phel’s position on Vigrahavyavartani 29., in T. Skorupski and U. Pagel (eds), Buddhist Forum, Vol. 3, School of Oriental and African Studies, London, pp. 161-185. Lopez, D.: 2005, The Madman’s Middle Way : Reflections on Reality of the Tibetan Monk Gendun Chopel, Chicago University Press, Chicago. M abbett, I.: 1984, Nagarjuna and Zeno on motion., Philosophy East and West 34(4), 401-420. M abbett, I.: 1996, Is there a Devadatta in the house? Nagarjuna’s Vigrahavyavartani and the liar paradox., Journal of Indian Philosophy (24), 295-320. BIBLIO G RAPH Y 343 M abbett, I.: 1998, The problem of the historical Nagarjuna revisited, Journal of the American Oriental Society 118(3), 332-346. Magee, W.: 1999, The Nature of Things. Emptiness and Essence in the Geluk World., Snow Lion, Ithaca, NY. Mannoury, G.: 1947, Les fondements psycho-linguistique des mathemathiques, Editions du Griffon, Neuchatel. Matilal, B.: 1968, The Navya-Nyaya doctrine of negation., Harvard University Press, Cambridge, MA. Matilal, B.: 1970, Reference and existence in Nyaya and Buddhist logic., Journal of Indian Philosophy 1, 83-110. Matilal, B.: 1971, Epistemology, Logic and Grammar in Indian Philosophical Analysis., Mouton, The Hague. Matilal, B.: 1986, Perception. A n Essay on Classical Indian Theories of Knowledge., Clarendon, Oxford. Matilal, B.: 1989, Nyaya critique of the Buddhist doctrine of non-soul, Journal of Indian Philosophy 17, 61-79. Matilal, B.: 1990, The Word and the World: India’s Contribution to the Study of Language., Oxford University Press, Delhi. May, J.: 1959, Prasannapada Madhyamakavrtti : Douze chapitres traduits du Sanscrit et du tib e ta in Adrien-Maisonneuve, Paris. McEvilley, T.: 1982, Pyrrhonism and Madhyamika.., Philosophy East and West 32(1), 3-35. BIBLIO G RAPH Y 344 McEvilley, T.: 2002, The Shape of Ancient Thought. Comparative Studies in Greek and Indian Philosophies., Allworth, New York. M euthrath, A.: 1996, Untersuchungen zur Kompositionsgeschichte der Nyayasutras., Oros, Wurzburg. M euthrath, A.: 1999, Die Nagarjuna zugeschriebene Vigrahavyavartani und die Nyayasutras. Eine Untersuchung des Verhaltnisses beider Texte zueinander., Verlag fur Orientalische Fachpublikationen, Reinbek. Mohanty, J.: 1992, Reason and Tradition in Indian Thought., Oxford University Press, Oxford. Morris, R. (ed.): 1888, Ahguttara N ik a y a Pali Text Society, London. Munsat, S.: 1971, Hume’s argument that causes must precede their effects., Philosophical Studies 22(1-2), 24-26. Murti, T. V. R.: 1955, Central Philosophy of Buddhism. A Study of Madhyamika System, George Allen and Unwin, London. Nagarjuna: 1919, The tree of wisdom : being the Tibetan text, with English translation of Nagarjuna’s gnomic verse treatise called the Prajhadanda., Calcutta University Press, Calcutta. Nagarjuna: 1979, Ndgarjuna’s letter : Ndgarjuna’s ‘Letter to a friend’, with a commentary by the Venerable Rendawa, Zhon-nu L o - d r o Library of Tibetan Works and Archives, Dharamsala. Translated by Geshe Lobsang Tharchin and Artemus B. Engle. Nagarjuna: 2002, Suhrllekha of Acarya Nagarjuna and Vyaktapadatikd of Acarya Mahamati. Sanskrit restoration and critically edited Tibetan text., Central Institute of Higher Tibetan Studies, Sarnath, Varanasi, BIBLIO G RAPH Y 345 Namai, C.: 1996, Rinne no Ronsho. Bukkyo ronrigakuha ni yoru yuibutsuron hihan., Toho shuppan, Osaka. Napper, E.: 1989, Dependent-Arising and Emptiness., Wisdom, London. Narbonne, J.-M.: 1993, Plotin, Descartes et la notion de ‘causa sui’., Archives de Philosophie 56(2), 177-195. Ng, Y.-k.: 1993, T ’ie n -T ’ai Buddhism and early Madhyamika., Tendai Institute, Hawaii. Ngag dbang dpal ldan: 1964, Grub m tha’ chen m o ’i mchan ’grel dka’ gnad mdud grol bio gsal gees nor., Pleasure of Elegant Sayings Printing Press, Sarnath. Nyanatiloka: 1950, Buddhist Dictionary. Manual of Buddhist Terms and Doctrines., Frewin &; Co, Colombo. Nyaya-Tarkatirtha, T. and Tarkatirtha, A. (eds): 1985, Nyayadarsanam with Vdtsyay ana’s Bhasya, Uddyotakara’s Vartika, Vdcaspati Misra’s Tatparyatika and Visvanatha’s Vrtti., Munshiram Manoharlal, Delhi. Oberhammer, G.: 1963, Ein Beitrag zu den Vada-Traditionen Indiens., Wiener Zeitschrift fur die Kunde Sud- und Ostasiens und Archiv fur indische Philosophie 7, 63-103. Oberhammer, G., Prets, E. and Prandstetter, J.: 1991—, Terminologie der friihen philosophischen Scholastik in Indien., Verlag der Osterreichischen Akademie der Wissenschaften, Vienna. Obermiller, E.: 1931, Bu ston’s ‘History of Buddhism’ (chos ’byung)., Harrassowitz, Heidelberg. BIBLIO G RAPH Y 346 Oetke, C.: 1989, Rationalismus und Mystik in der Philosophie Nagarjunas., Studien zur Indologie und Iranistik 15, 1-39. Oetke, C.: 1991, Zur Methode und Analyse philosophischer Sutratexte. Die Pramunapassagen der Nyayasutren., Dr. Inge Wezler Verlag fiir Orientalistische Fachpublikationen, Reinbek. Oetke, C.: 1997, Pragmatic principles and Bronkhorst’s maxims of interpretation., Studien zur Indologie und Iranistik 2 1 , 133-152. Oetke, C.: 2003, Some remarks on theses and philosophical positions in early Madhyamaka., Journal of Indian Philosophy 31, 449-478. O’Flaherty, W. D.: 1981, The Rig Veda. An Anthology., Penguin, London. Pandeya, R.: 1988-1989, Madhyamakasastram, Motilal Banarsidass, Delhi. Pap, A.: 1960, Types and meaninglessness., Mind 69, 41-54. Parfit, D.: 1984, Reasons and Persons., Oxford University Press, Oxford. Pathak, S. K.: 1974, The Indian Nitisastras in Tibet., Motilal Banarsidass, Delhi. Perrett, R. W.: 1998, Indian theories of causation., in E. Craig (ed.), Routledge Encyclopedia of Philosophy, Vol. 2, Routledge, London, New York, pp. 251-257. Piaget, J.: 1937, La construction du reel chez Venfant., Delacheux et Niestle, Paris. Pind, O.: 2001, Why the Vaidalyaprakarana cannot be an authentic work of Nagarjuna., Wiener Zeitschrift fur die Kunde Siidasiens 45, 149-172. BIBLIO G RAPH Y 347 Potter, K.: 1970-2003, Encyclopedia of Indian Philosophies, Motilal Banarsidass, Delhi. Potter, M.: 2000, Reason’s Nearest Kin. Philosophies of Arithmetic from Kant to Carnap., Oxford University Press, Oxford. Pradhan, P.: 1975, Abhidharmakosabhasya of V a su b a n d h u K.P. Jayaswal Research Institute, Patna. Priest, G. and Garfield, J.: 2002, Nagarjuna and the limits of thought., Beyond the Limits of Thought, Clarendon Press, Oxford, pp. 249-270. Priestley, L.: 1999, Pudgalavada Buddhism: The Reality of the Indeterminate Self., Centre for South Asian Studies, University of Toronto, Toronto. Proops, I.: 2004, W ittgenstein on the substance of the world, European Journal of Philosophy 12(1), 106-126. Quine, W. V. O.: 1953, On what there is, From a Logical Point of View, Harvard University Press, Cambridge, MA; London, pp. 1-19. Raju, P.: 1954, The principle of four-cornered negation in Indian philosophy., Review of Metaphysics 7(4), 694-713. Renou, L.: 1942, T e rm in o lo g ie g ra m m a tic a le du S a n sk rit., Champion, Paris. Robinson, R.: 1957, Some logical aspects of Nagarjuna’s system., Philosophy East and West 6 , 291-308. Robinson, R.: 1967, Early Madhyamika in India and China., University of Wisconsin Press, Madison, Milwaukee, London. Robinson, R.: 1969, Review of ‘Early Buddhist Theory of Knowledge’., Philosophy East and West 19(1), 69-81. BIBLIO G RAPH Y 348 Robinson, R.: 1972, Did Nagarjuna really refute all philosophical views?, Philosophy East and West 22(3), 325-331. Ronkin, N.: 2005, Early Buddhist Metaphysics: The Making of a Philosophical Tradition., Routledge, London. Rosenberg, J.: 1998, Kant and the problem of simultaneous causation., International Journal of Philosophical Studies 6 , 167-188. Routley, R.: 1969, The need for nonsense., Australasian Journal of Philosophy 47, 367-384. Ruegg, D. S.: 1977, The use of the four positions of the catuskoti and the problem of the description of reality in Mahayana Buddhism., Journal of Indian Philosophy (5), 1-171. Ruegg, D. S.: 1978, Mathematical and linguistic models in Indian thought: the case of zero and sunyata., Wiener Zeitschrift fur die Kunde Siidasiens und Archiv fur Indische Philosophie 2 2 , 1171-181. Ruegg, D. S.: 1981, The Literature of the Madhyamaka School of Philosophy in India., Harrassowitz, Wiesbaden. Ruegg, D. S.: 1983, On the thesis and assertion in Madhyamaka/dBu ma., in E. Steinkellner and H. Tauscher (eds), Contributions on Tibetan and Buddhist Religion and Philosophy, Arbeitskreis fur Tibetische und Buddhistische Studien, Vienna, pp. 205-241. Ruegg, D. S.: 1986, Does the Madhyamika have a thesis and philosophical position?, in B. K. Matilal (ed.), Buddhist Logic and Epistemology, D. Reidel, Dordrecht, pp. 229-237. BIBLIO G RAPH Y 349 Ruegg, D. S.: 2000, Three Studies in the History of Indian and Tibetan Madhyamaka Philosophy., Arbeitskreis fur Tibetische und Buddhistische Studien, Vienna. Ruegg, D. S.: 2002, Two prolegomena to Madhyamaka Philosophy., Arbeitskreis fur Tibetische und Buddhistische Studien, Vienna. Sagal, P. T.: 1992, Nagarjuna’s paradox., American Philosophical Quarterly 29(1), 79-85. Schaffer, J.: 2001, The individuation of tropes., Australasian Journal of Philosophy 79(2), 247-257. Schayer, S.: 1929-1930, Feuer und Brennstoff. Ein Kapitel aus dem Madhyamika-Sastra des Nagarjuna mit der V rtti des Candraklrti., Rocznik Orjentalistyczny 7, 26-52, Schayer, S.: 1931, Ausgewdhlte Kapitel aus der Prasannapada ., Naldadem Polskiej Akademji Umietjetnosci, Krakow. Schayer, S.: 1933, Altindische Antizipationen der Aussagenlogik., Bulletin de’l Academic Polonaise pp. 90-96. Searle, J.: 1969, Speech Acts. A n Essay in the Philosophy of Language., Cambridge University Press, Cambridge. Segal, S.: 1998, Collision with the Infinite. A Life Beyond the Personal Self., Blue Dove Press, San Diego. Shanna, C.: 1960, A critical survey of Indian Philosophy., Motilal Banarsidass, Delhi.

Sharma, D.: 1970, The Negative Dialectics of India. A Study of the Negative Dialecticism in Indian Philosophy., n.p. Shaw, J.: 1978, Negation and the Buddhist theory of meaning,, Journal of Indian Philosophy 6 , 59-77. Shaw, J.: 2 0 0 2 , Causality: Samkhya, Bauddha and Nyaya., Journal of Indian Philosophy 30, 213-270. Siderits, M.: 1980, The Madhyamaka critique of epistemology., Journal of Indian Philosophy 8 , 307-335. Siderits, M.: 1989, Thinking on empty: Madhyamaka anti-realism and canons of rationality., in S. Biderman and B.-A. Scharfenstein (eds), Rationality in Question, E.J. Brill, Leiden, pp. 231-249. Siderits, M.: 2000, Nyaya realism, Buddhist critique., in B. G upta (ed.), The Empirical and the Transcendental, Rowman & Littlefield, Lanham, pp. 219-231. Siderits, M.: 2003, Personal Identity and Buddhist Philosophy., Ashgate, Aldershot. Siderits, M.: 2004, Causation and emptiness in early Madhyamika., Journal of Indian Philosophy 32, 393-419. Siderits, M. and O ’Brien, J. D.: 1976, Zeno and Nagarjuna on motion., Philosophy East and West 26(3), 281-299. Simons, P.: 1987, Parts. A Study in Ontology., Clarendon, Oxford. Sommers, F.: 1965, Predicability, in M. Black (ed.), Philosophy in America, George Allan & Unwin, pp. 262-281. BIBLIO G RAPH Y 351 Sopa, G. L. and Hopkins, J.: 1976, Practice and Theory of Tibetan Buddhism., Grove, New York. Spelke, E.: 1990, Principles of object perception, Cognitive Science 14, 29- 56. Staal, F.: 1962, Negation and the law of contradiction in Indian thought: A comparative study., Bulletin of the School of Oriental and African Studies 25, 52-71. Staal, F.: 1975, Exploring Mysticism., Penguin, London. Stcherbatsky, T.: 1923, The Central Conception of Buddhism and the Meaning of the Word ‘Dharma}., Royal Asiatic Society, London. Stcherbatsky, T.: 1962, Buddhist Logic., Dover, New York. Stcherbatsky, T.: 1968, The Conception of Buddhist Nirvana., Motilal Banarsidass, Delhi. Steinkellner, E.: 1986, Dharmottaras Paralokasiddhi : Nachweis der Wiedergeburt, zugleich eine Widerlegung materialistischer Thesen zur Natur der Geistigkeit., Arbeitskreis fur Tibetische und Buddhistische Studien, Vienna. Streng, F.: 1967, Emptiness: A Study in Religious Meaning., Abingdon Press, Nashville. Subbotskii, E.: 1991, Existence as a psychological problem: Object permanence in adults and preschool children., International Journal of Behavioral Development 14(1), 67-82. BIBLIO G RAPH Y 352 Taber, J. A.: 1998, On Nagarjuna’s so-called fallacies: A comparative approach., Indo-Iranian Journal 41, 213-244. Tanaka, K.: 1985, Simultaneous relation (Sahabhu-hetu): A study in Buddhist theory of causation, Journal of the International Association of Buddhist Studies 8(1), 91-111. Tanji, T.: 2000, On Samaropa, in J. Silk (ed.), Wisdom, Compassion, and the Search for Understanding, University of Hawaii Press, Honolulu, pp. 347-368. Thomason, R.: 1972, A semantic theory of sortal incorrectness., Journal of Philosophical Logic 1, 209-258. Thurman, R.: 1984, The Central Philosophy of Tibet., Princeton Uninversity Press, Princeton, NJ. Tillemans, T.: 1983, The ‘neither one nor many’ argument for sunyata and its Tibetan interpretations., in E. Steinkellner and H. Tauscher (eds), Contributions on Tibetan and Buddhist Religion and Philosophy, Arbeitskreis fur Tibetische und Buddhistische Studien, Universitat Wien, Vienna, pp. 305-320. Tillemans, T.: 1984, Two Tibetan texts on the ‘neither one nor many’ argument for sunyata., Journal of Indian Philosophy 12, 357-388. Tillemans, T.: 2001, Trying to be fair to Madhyamika Buddhism. The Num ata Yehan Lecture in Buddhism, University of Calgary, Canada. Tillemans, T.: 2003, Metaphysics for Madhyamikas., in G. Dreyfus and S. McClintock (eds), The Svatantrika-Prasahgika Distinction: What Difference does a Difference make?, Wisdom, Boston, pp. 93-123.

Tillemans, T. J.: 1990, Materials for the Study of Aryadeva, Dharmapala and Candrakirti., Universitat Wien, Wien. Tillemans, T. J.: 1999, Formal and semantic aspects of Tibetan Buddhist debate logic., Scripture, Logic, Language. Essays on Dharmakvrti and his Tibetan Successors, Wisdom. Tillemans, T. J.: 2000, Dharmakirti’s Pramdnavdrttika : A n annotated translation of the fourth chapter (pararthdnumdna)., Verlag der Osterreichischen Akademie der Wissenschaften, Vienna. Tola, F. and Dragonetti, C.: 1987, Sunyatasaptati. The Seventy Karikas on Voidness (according to the Svavrtti) of Nagarjuna., Journal of Indian Philosophy 15, 1-55. Tola, F. and Dragonetti, C.: 1995a, Nagarjuna’s Refutation of Logic., Motilal Banarsidass, Delhi. Tola, F. and Dragonetti, C.: 1995b, On Voidness. A Study on Buddhist Nihilism., Motilal Banarsidass, Delhi. Tola, F. and Dragonetti, C.: 1998, Against the attribution of the Vigrahavyavartanl to Nagarjuna., Wiener Zeitschrift fur die Kunde Siidasiens und Archiv fu r Indische Philosophic 42, 151-166. Tola, F. and Dragonetti, C.: 2004, Being as Consciousness. Yogacdra Philosophy of Buddhism., Motilal Banarsidass, Delhi. Trencker, V. (ed.): 1888, Majjhima-Nikaya., Pali Text Society, London. Trenckner, V. (ed.): 1928, The Milindapanho., Royal Asiatic Society, London. Tsong kha pa Bio bzang grags pa: 1973, rTsa shes tik chen, Sarnath.

Tsong kha pa Bio bzang grags pa: 1985, Lam rim chen mo, Tso Ngon (Qinghai) People’s Press, Qinghai. Tsong kha pa Bio bzang grags pa: 2000-2004, The Great Treatise on the Stages of the Path to Enlightenment., Snow Lion, Ithaca, NY. Tsong kha pa Bio bzang grags pa: 2006, Ocean of Reasoning., Oxford University Press, Oxford. Tucci, G.: 1929, Pre-Dihnaga Buddhist Texts on Logic from Chinese Sources., Gaekwad’s Oriental Series, Baroda. Tucci, G.: 1934-1936, The RatnavalT of Nagarjuna., Journal of the Royal Asiatic Society of Great Britain and Ireland . Tuck, A. P.: 1990, Comparative Philosophy and the Philosophy of Scholarship: on the Western Interpretation of Nagarjuna., Oxford University Press, Oxford. Vasubandhu: 1970-1973, Abhidharmakosa and Bhasya of Acdrya Vasubandhu with the Sphutartha commentary of Acdrya Yasomitra., Bauddha Bharati, Varanasi. von Rospatt, A.: 1995, The Buddhist Doctrine of Momentariness., Franz Steiner, Stuttgart. Walleser, M.: 1923, The Life of Nagarjuna from. Tibetan and Chinese Sources., Probsthain, London. Walleser, M. (ed.): 1913~1914b, Buddhapalita’s Mulamadhyamakavrtti., Vol. 16 of Bibliotheca Buddhica, St Petersburg. BIBLIO G RAPH Y 355 Walleser, M. (ed.): 1914a, Bhavaviveka’s Prajhapradipa., Vol. 226 of Bibliotheca Indica, Calcutta. Walser, J.: 1998, On the formal arguments of the Akutobhaya., Journal of Indian Philosophy 26, 189-232. Walser, J.: 2005, Nagarjuna in Context. Mahay ana Buddhism and Early Indian Culture., Columbia University Press, New York. Warder, A.: 1973, Is Nagarjuna a Mahayanist?, in M. Sprung (ed.), The Problem of Two Truths in Buddhism and Vedanta, Reidel, Boston, pp. 78-88. Wayman, A.: 1977, Who understands the four alternatives of the Buddhist texts?, Philosophy East and West 27(1), 3-21. Weber-Brosamer, B. and Back, D. M.: 1997, Die Philosophie der Leere., Harrassowitz, Wiesbaden. Wertheimer, M.: 1912, Experimentelle Studien fiber das Sehen von Bewegung., Zeitschrift fur Psychologic 61, 161-265. Wiliams, P. and Tribe, A.: 2000, Buddhist Thought., Routledge, London. Williams, D.: 1953, The elements of being., Review of Metaphysics 7, 3-18, 171-192. Williams, P.: 1981, On the Abhidhaxma ontology., Journal of Indian Philosophy 9, 227-257. Williams, P.: 1998, Altruism and Reality. Studies in the Philosophy of the Bodhicaryavatara., Curzon, Richmond, Surrey. BIBLIO G RAPH Y 356 Wood, T. E.: 1994, Nagarjunian Disputations. A Philosophical Journey through an Indian L ooking-glassUniversity of Hawaii Press, Honolulu. Ye, S.: 2006a, The Miilamadhyamakakdrikd and Buddhapalita’s commentary (1): Romanized text based on the newly identified Sanskrit manuscripts from Tibet., Annual Report of the International Research Institute for Advanced Buddhology 10, 117-147. Ye, S.: 2006b, A re-examination of the Mulamadhyamakakarika on the basis of the newly identified Sanskrit manuscripts from Tibet., Annual Report of the International Research Institute for Advanced Buddhology 10, 149-170. U N I V E R S I T Y OF L O N D O N SENATE HOUSE. MALET STREET, LONDON, WCIE 7HU

REPRODUCTION OF THESES A thesis which is accepted by the University for the award o f a Research Degree is placed in the Library o f the College and in the University of London Library. The copyright o f the thesis is retained by the author. As you are about to subm it a thesis for a Research Degree, you are required to sign the declaration below. This declaration is separate from any wh ich may be made under arrangem ents with the College at which you have p u rsu ed your course (for internal candidates only). The declaration will be destroyed if your thesis is not approved by the examiners, being either rejected or referred for revision. Academ ic Registrar To be completed by the candidate N A M E IN FULL (Block Capitals) 1 4 V C ttfU'TTOPH W ^ .P T E tgH 0 t=t=________________ t i t l e o f t h e s i s A / A < n A R ^ vaJA'S1 M A y t f y A M A f < A ■ A P W L O f o p H i C A L

1. I authorise that the thesis presented by me in * [ l 0 0f-] for examination for the M Phil/PhD D egree o f the University o f London shall, if a degree is awarded, be deposited in the library o f the appropriate College and in the University o f London Library and that, subject to the conditions set out below, my thesis be made available for public reference, inter-library loan and copying. 2. I authorise the College or University authorities as appropriate to supply a copy o f the abstract o f my thesis for inclusion in any published list o f theses offered for higher degrees in British universities or in any supplem ent thereto, or for consultation in any central file o f abstracts o f such theses. 3. I authorise the College and the University o f London Libraries, or their designated agents, to m ake a microform or digital copy of my thesis for the purposes o f inter-library loan and the supply o f copies. 4. I understand that before my thesis is m ade available for public reference, inter-library loan and copying, the following statement will have been included at the beginning o f my thesis: The copyright o f this thesis rests with the author and no quotation from it or information derived from it may be published w ithout the prior written consent o f the author. 5. i authorise the College and/or the University o f London to make a m icroform or digital copy o f my thesis in due course as the archival copy for perm anent retention in substitution for the original copy. 6. I warrant that this authorisation does not, to the best o f my belief, infringe the rights o f any third party. 7. I understand that in the event o f my thesis being not approved by the exam iners, this dclaration w ould become void.

• Please state year.

DATE O Q ? - Note: The University’s Ordinances m ake provision for restriction o f access to an M Phil/PhD thesis and/or the abstract but only in certain specified circum stances ana for a m axium um period o f two years. If you wish to apply for such restriction, please enquire at your College about the conditions and procedures. External Students should enquire at the Research Degree Examinations Office, Room 261, Senate House. SIGNATURE 3 0 NOV 2007 THIS DECLARATION MUST BE COMPLETED AND RETURNED WITH THE EXAMINATION ENTRY FORM