Jump to ContentJump to Main Navigation
The Golden Age of Indian Buddhist Philosophy$

Jan Westerhoff

Print publication date: 2018

Print ISBN-13: 9780198732662

Published to Oxford Scholarship Online: June 2018

DOI: 10.1093/oso/9780198732662.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 17 September 2019



(p.84) 2 Madhyamaka
The Golden Age of Indian Buddhist Philosophy

Jan Westerhoff

Oxford University Press

Abstract and Keywords

The chapter begins by presenting a general overview of the rise of the Mahāyāna and its relation to the main schools of Buddhist philosophy associated with it, Madhyamaka and Yogācāra. This is followed by an introduction to the Madhyamaka school proper, focusing on the life and works of its founder, Nāgārjuna. The third section of the chapter examines the foundational sūtras of the Madhyamaka school, the Prajñāpāramitā or Perfection of Wisdom texts, focusing on their criticism of the Abhidharma philosophical project, their comprehensive illusionism, and their prima facie acceptance of contradictions. This is followed by an account of how these themes play a key role in the Madhyamaka system as set out by Nāgārjuna. The chapter then turns to examining the philosophical contributions of major Madhyamaka thinkers such as Buddhapālita, Bhāviveka, Candrakīrti, Śāntarakṣita, and Kamalaśīla. The chapter concludes by a discussion of the relation between Madhyamaka and Nyāya.

Keywords:   Madhyamaka, Mahāyāna, Nāgārjuna, Prajñāpāramitā, Buddhapālita, Bhāviveka, Candrakīrti, Śāntarakṣita, Kamalaśīla, Nyāya

1. The Rise of the Mahāyāna and Its Relation to Buddhist Philosophy

The most important development in Buddhism during the period we are considering here was the rise of the Mahāyāna. The Mahāyāna movement brought with it an enormous amount of new (or, as the Mahāyāna would put it: previously unknown) sūtras, a new spiritual ideal (that of the bodhisattva, considered as a superior aspiration than the quest for arhatship), and, it would appear, exciting new philosophical developments. A key distinction between the Abhidharma on the one hand and Madhyamaka and Yogācāra on the other is that they are commonly associated with different kinds of Buddhism: the last two are philosophical schools of the Mahāyāna, while the Abhidharma philosophy belongs to what is pejoratively called the ‘Hīnayāna’ (‘little vehicle’), and more neutrally, the Śrāvakayāna.

The history of the beginning of the Mahāyāna and the causes that led to its development are still quite unclear.1 There is some consensus, however, on what the Mahāyāna was not. It was not a lay movement2 that tried to shift the balance of power away from the monks and nuns, nor a group of stūpa worshippers,3 nor was it the result of a doctrinal schism between different Buddhist schools,4 along the lines of the split between Roman Catholicism and Protestantism. The majority of the Mahāyāna’s early supporters (when it achieved more widespread support towards the middle of the first millennium CE, several centuries after its inception) were monastics.5 Stūpa worship was not confined to the Mahāyāna,6 and the notion of a split of the monastic community on doctrinal grounds is quite alien to Buddhism; traditionally such splits took place because of a difference about which monastic rules a given community should follow. This is not too surprising, given that a difference about which rules to adopt can be highly disruptive to the functioning of a (p.85) monastic group, while the beliefs of an individual monk about what he is doing when he is practising meditation tend not to be.

The difficulty of connecting the early Mahāyāna with any kind of historical or archaeological evidence have led some to argue that it was a purely textual movement, with a focus on the exposition and transmission of the revealed Mahāyāna sūtras, without developing alternatives to the social and institutional framework of Buddhism at the time.7 This accounts well for the profusion of Mahāyāna sūtras we see in the development of Indian Buddhism, a profusion based on a kind of continuous revelation of the Buddha’s teaching, with the emergence of texts that are regarded as authoritative even though they may not have been taught by the Buddha during his life on earth.8 The surprising amount of textual documents the Mahāyāna produced may by itself seem to justify a conception of it as a ‘cult of the book’9 (or, more precisely, if less concisely) a collection of different cults of different books.

Insofar as it is possible to identify unifying conceptual features underlying the vast corpus of Mahāyāna sūtras, one prominent feature is a different vision of what the Buddha is. The Buddha is considered as not having completely disappeared after his parinirvāṇa, but as in some sense still present and helping beings to achieve enlightenment. This idea of the Buddha as an enlightened being perpetually acting out of his great compassion began to be considered as an ideal to be emulated, and as preferable to that of an arhat,10 and linked up with an interest in the previous existences of the Buddha as a bodhisattva, or Buddha-to-be. The previous lives are recorded in the jātaka stories, which describe the future Buddha as helping other beings out of his great compassion, often by giving up his own life. With this came an intention to follow the ideal of the bodhisattva to become, in due course, a compassionate enlightened (p.86) being like the Buddha himself.11 This changed conception of the Buddha might be considered as a source for the prominence of the bodhisattva ideal in Mahāyāna, the emphasis on the quality of compassion, and also the profusion of Mahāyāna texts. An important difference between Buddhas and arhats is that the former were taken to be omniscient, and as Buddhas-to be the bodhisattvas could therefore be expected to require more knowledge than the arhats. This additional knowledge was helpfully supplied by the newly emerging sūtras specifically aimed at the needs of bodhisattvas.

However, so far it is not clear what the connection between Mahāyāna and the philosophical developments of Madhyamaka and Yogācāra was. Even though it has been argued by some scholars that the connection is rather tenuous, questioning whether Nāgārjuna was a Mahāyānist at all, and pointing out that the difference between Mahāyāna and ‘Hīnayāna’ thought cannot have been that great if a school like that of Diṅnāga and Dharmakīrti could actually be described as combining both in a form of ‘Yogācāra-Sautrāntika’, the historical connection between Mahāyāna and the schools of Madhyamaka and Yogācāra is too obvious to deny. What is worthwhile investigating, though, is whether the connection is more than a historical accident. Was the Mahāyāna simply a religious development that became associated with specific thinkers and their schools, without having much of an influence on their philosophy,12 or is there a more fundamental connection between Mahāyāna ideas and those later developed in Madhyamaka and Yogācāra texts? Does the Mahāyāna have any specific philosophical ramifications apart from its religious, doctrinal, and soteriological consequences?

In fact, the changing view of the Buddha in Mahāyāna texts just mentioned is a particular case of a widening of the Buddhist vision of the world we find in these sūtras: a more comprehensive soteriological goal, more extensive cosmological accounts, including ‘celestial’ Buddhas residing in pure lands, a wider corpus of teachings to be considered as the Buddha’s word, describing sets of new, powerful practices.

This extended vision incorporated the pre-Mahāyāna view of the Buddhist path; in particular, it subsumed and endorsed the ideas of the śrāvaka and the pratyekabuddha.13 Yet in order to present the conceptions that preceded it as (p.87) special cases to be included in the wider compass of the Mahāyāna, it was necessary to de-ontologize them. Both the ordinary, unenlightened conception of the world, as well as the theory of dharmas that formed part of the Abhidharma and claimed to describe the ultimate truth about how things exist at the ultimate level, had to be considered as lacking fundamental reality, as fundamentally illusory, though of pragmatic and instrumental value, in order to be able to conceive of them not as conflicting with, but as forming a part of the Mahāyāna vision. The development of a more comprehensive view of the Buddhist world could not consider the more restricted and sometimes contradictory pre-Mahāyāna conception as a complete and ultimately true account, but could only incorporate it as true ‘in a manner of speaking’. The world as it appears and the world as early Buddhist dharma theory analysed it had to be regarded as a mere illusory reality, in order to be regarded as a reality at all.

It is this broadly illusionistic view of the world, I would argue, that forms the best point of connection between the Mahāyāna sūtras and the philosophical developments of Madhyamaka and Yogācāra.14 To be sure, early illusionistic views exist in Buddhism outside of Mahāyāna sūtras,15 and there is much more to the extremely complex philosophical systems of Madhyamaka and Yogācāra than simply the idea that the world is just like a magic show.16 Yet if we ask ourselves which ideas from the Mahāyāna sūtras these philosophical texts developed, and which they in turn referred to in order to back up their philosophical perspective by texts considered to be the Buddha’s words, the illusionistic view of the world occupies a prominent place. In addition to the possible conceptual reason for the arising of this view just mentioned, the illusionistic view may have a foundation in the meditative practices of early Mahāyānists. While the theory that the entire Mahāyāna arose as the reflection of meditative practices of contemplative ascetics is unlikely to be true,17 we have textual evidence for meditative exercises supposed to bring about the perception of the Buddha as present in this very world. The Pratyutpanna-buddha-saṃmukhāvasthita-samādhi-sūtra, a Mahāyāna text from the second century CE or earlier, teaches a form of meditation (samādhi) enabling the (p.88) practitioner to stand face-to-face (saṃmukhāvasthita) with the present (pratyutpanna) Buddha. After completing a set of meditative exercises, the meditating bodhisattva:

does not see the Tathāgata through obtaining the divine eye; he does not hear the true dharma through obtaining the range of the divine ear; nor does he go to that world-system in an instant through obtaining magical powers—Bhadrapāla, while remaining in this very world-system that bodhisattva sees the Lord, the Tathāgata Amitāyus, and conceiving himself to be in that world-system he also hears the dharma.18

The text makes clear that the sight of the Buddha in this case is not due to worldly epistemic super-powers such as the ‘divine eye’ (divyacakṣuḥ) allowing one to see things very far away, or through magical travel to a distant world to observe the Buddha there. It then becomes an intriguing question what the nature of these Buddhas—on the one hand present in this world, on the other, disappearing when the practitioner leaves meditative absorption—actually is. The sūtra clarifies this as follows:

Having thought: ‘Did these Tathāgatas come from somewhere? Did I go anywhere?’ he understands that those Tathāgatas did not come from anywhere. Having comprehended that his own body did not go anywhere either, he thinks: ‘These triple worlds are only mind. Why? Because however I mentally construct things, so they appear’.19

The illusionistic position that things are not as they appear (but, in this case, possess a very different, mentally constructed reality) appears to arise here in order to make sense of specific meditative experiences. In order to account for the meditator’s experiences of the Buddha as actually present in the world, it is necessary to regard the ordinary perception of the world, post-meditative experience, and even the meditatively trained perception of the world of the Abhidharma practititioner according to which the Buddha is not present in this world, as unable to undermine meditative experience, simply because these former two kinds of perception are not grounded in the way the world really is. These ways of viewing the world (which are the ways the world appears to most (p.89) people most of the time) are grounded in illusion and have no implications for what exists. The illusionistic worldview therefore coheres naturally with the way the world would be conceived by a practitioner of the kind of early Mahāyāna meditative exercises described in this sūtra.

Having now considered a possible conceptual connection point between the Mahāyāna and the subsequent developments in Buddhist philosophy in India, we are ready to turn to the first of its two main schools: Madhyamaka.

2. The Madhyamaka School

The Madhyamaka school is one of the most puzzling (and most intriguing) branches of Buddhist philosophy. On the basis of a casual acquaintance with Madhyamaka texts it is far from clear what precisely their doctrine amounts to. David Ruegg has put this well by pointing out that it

has been variously described as nihilism, monism, irrationalism, misology, agnosticism, skepticism, criticism, dialectic, mysticism, acosmism, absolutism, relativism, nominalism, and linguistic analysis with therapeutic value.20

Since the day these lines have been written numerous other -isms, such as deconstructivism, dialetheism, and ontological non-foundationalism, have enlarged the menu of interpretative options even further.

In fact our puzzlement with Madhyamaka is likely to begin already with the biography of its founder. We know that the school was founded by Nāgārjuna, an Indian monk and philosopher, and one of the two or three greatest thinkers that Indian intellectual history has produced. It is only exaggerating slightly to say that this is already where our certainties end. When it comes to Nāgārjuna we are unclear about when he lived, where in India he spent most of his time, what texts he composed, and even how many Nāgārjunas there were in the first place.

Nāgārjuna’s biography is transmitted to us in a variety of accounts that abound with hagiographical detail. But Nāgārjuna has entered the history of Buddhist thought even before we get to these biographies. If we follow traditional Buddhist accounts, the arising of the Madhyamaka school was no historical accident, but a development already predicted by the historical Buddha Śakyamuni. Nāgārjuna (referred to just as Nāga) is mentioned at various places in the Mahāyāna sūtras and tantras.21 The most famous of these is a prophecy in the Laṅkāvatārasūtra.22 Addressing the bodhisattva Mañjuśrī, the Buddha declares that:


In Vedalī, in the southern part, there will be a monk widely known as Śrīmān, who will be called Nāga. Destroying the positions of existence and non-existence he will teach my vehicle, the unsurpassed Mahāyāna to the world. He will attain the stage called muditā [‘joyful’, the first Bodhisattva ground] and will pass on to the pure realm of Sukhāvatī.23

Other scriptures add detail to this, such as that Nāgārjuna is going to be born 400 years after the death of the historical Buddha, or that he will live for 600 years, but what makes this prophecy particularly interesting is that it says Nāgārjuna will achieve the first Bodhisattva ground (bhūmi).24 This achievement requires realization of emptiness. It ensures that Nāgārjuna not only knows what he is talking about, but has realized it directly.

The majority of details of Nāgārjuna’s life are transmitted to us in a variety of colourful accounts from later writers such as Kumārajīva, Bu ston, and Tāranātha, accounts that exhibit surprisingly little agreement with each other. Jan Yün-Hua gives a succinct account of the common themes, and points out that:

he came from a Brahminical family, was well versed in magic power, and had a romantic life when he was young. After renouncing his worldly life and being initiated into the Buddhist Saṅgha, he studied Mahāyāna texts on the Snow Mountain, went to and obtained more important Mahāyāna scriptures from the palace of the Nāgas under the sea, and won the mind and support of the king of Sātavāhana dynasty. These sources also say that he settled in South India until the last days of his life. He had a long life, lasting several hundred years.25

Some of these points merit further comment. The ‘romantic life’ refers to a period in Nāgārjuna’s pre-monastic days reported in Kumārajīva’s account and connects with a dominant theme in his biographies, his mastery of magical powers. Nāgārjuna and his friends are said to have procured an invisibility potion and used it to enter into the royal harem unawares, to enjoy the company of the royal consorts. The king finds out about this and is not amused. He sets them a trap and observes the footprints the invisible men leave in the sand, then sends in his soldiers to aim their swords at where their heads would be. All are killed save for Nāgārjuna, who stands immediately behind the king, out of reach of the swords. After this brush with death, Nāgārjuna ‘conceives a dislike of the idea of desire’26 and becomes a monk.

The association with magical powers mentioned here plays an important part in Nāgārjuna’s biographies. He is, in fact, counted as one of the famous set (p.91) of eighty-four siddhas, Indian tantric masters renowned both for their spiritual accomplishments and for their displays of magical powers. Magical elements feature already at the very beginning of his life. Initially his life is predicted to be very short; his parents therefore send him to study at the famous monastery of Nālandā at the age of 7. There the abbot, Rāhulabhadra, is supposed to have taught him to prolong his life by means of the recitation of mantras. Nāgārjuna becomes highly proficient at tantric practice, achieving, amongst other things, the elixir of long life.27 According to Bu ston, Nāgārjuna managed to extend his life for 600 years. Even after this period he did not die a natural death, but allowed the son of a king to behead him—the king’s and Nāgārjuna’s lifespan having somehow become linked through their respective longevity practices, the prince was understandably concerned that he should never succeed his father on the throne.

Some biographies describe Nāgārjuna as finding an elixir to make gold, a feat he used in order provide food for the monastic community during a famine. According to other accounts, Nāgārjuna is carried through a river by a cowherd. He creates an illusion of crocodiles that seem to attack them, and when the cowherd has carried him across the river unperturbed, he grants him a boon. The cowherd wants to be a king, and so Nāgārjuna turns him into one, creating elephants, armies, and all kinds of other kingly possessions to go with it. The king, called Śālābhanda, later becomes his disciple, and Nāgārjuna composes the Ratnāvalī for him.28

The second important recurring motive in Nāgārjuna’s life-stories mentioned above is already evident in his name: his association with the nāgas. His name is a compound of two nouns, nāga and arjuna. The nāgas are mythological snake-like creatures29 who live in palatial aquatic abodes, in an underwater city called Bhogavatī (longs spyod can), under the earth or in mountain caves. Nāgas are often depicted as beings that are half-snake, half-man, with a human torso and a lower body in the form of a coiled snake, and are renowned for their great beauty. They are guardians of tremendous wealth (they are sometimes said to have a jewel embedded in their heads), wise, and powerful.

Arjuna, Bu ston informs us, refers to someone ‘who has procured worldly power’. According Kumārajīva’s account, however, the term is a name of a kind of a tree, Nāgārjuna’s mother having given birth to him under a tree.30 According to a third account, the second half of Nāgārjuna’s name refers to (p.92) the Pāṇḍava brother from the Mahābhārata known for his skills in archery,31 since Nāgārjuna is able to spread the Mahāyāna as securely as Arjuna shoots his arrows.32

Be this as it may, Nāgārjuna’s association with nāgas is of central importance in the story of his life. His biographies speak of two ladies from the retinue of the king of the nāgas listening to his teaching, filling the place with the scent of sandalwood.33 Nāgārjuna then travels to the palace of the nāga king under the sea, a place overflowing with a variety of gems and jewels. Amongst these valued possesions of the nāgas, the most important one is a rare treasure of Buddhist scriptures, the Perfection of Wisdom (prajñāpāramitā) sūtras. These are said to have been entrusted by the historical Buddha to the nāgas for safekeeping. Their content is supposed to be so subtle that it may be easily misunderstood. For this reason these scriptures must wait for their right interpreter who can correctly explain their meaning. This interpreter is, of course, Nāgārjuna, who brings the scriptures back to the human realm.

The Perfection of Wisdom sūtras are a family of highly influential Buddhist texts of varying lengths. Their length is usually indicated in their titles, so we have texts like the Perfection of Wisdom in 8,000 verses, the Perfection of Wisdom in 25,000 verses or even the Perfection of Wisdom in 100,000 verses. These kinds of titles might give rise to two misunderstandings. First, we might think the texts are in verse, even though they are generally in prose. Nevertheless, their length is measured by how many units of thirty-two syllables (verses, or ślokas) they contain. The Perfection of Wisdom in 8,000 verses, for example, is about 110,000 words long in English translation—about the length of the book you are reading just now. The second mistaken impression we might have is that the version in 100,000 verses, for example, contains four times as much information as the 25,000-verse version, because it is four times as long. In fact the longer versions of these sūtras differ from the shorter ones not so much by including more information, but by spelling out lists in full that are only given in part in the shorter versions.34 The Perfection of Wisdom in 100,000 verses spells out the claim that ‘x is emptiness, and emptiness itself is x’ by going through a long list of about 200 items one by one. The shorter versions provide abbreviated forms of these lists, and sometimes only mention their first and last elements.

(p.93) As their name makes clear, the Perfection of Wisdom texts are sūtras; they begin with the customary words ‘Thus I have heard’, and purport to give an account of discourses of the historical Buddha held in front of his disciples (such as Śāriputra) and an assembly of bodhisattvas. They describe the practices a bodhisattva should follow in order to achieve enlightenment. A key element of these practices is the development of the six perfections, generosity (dhāna), moral virtue (śīla), patience, (kṣānti), effort (vīrya), meditation (dhyāna), and wisdom (prajñā). Special prominence is given to the final perfection, the perfection of wisdom; it is sometimes considered to include all the other perfections within it.

The conceptual core of this final perfection is the realization of emptiness, the understanding of the insubstantiality of all phenomena. Despite the fact that the notion of emptiness is the dominant theme of the Perfection of Wisdom texts, they do not present a great number of arguments for the claim that everything is indeed empty, nor do they discuss potential objections. It was Nāgārjuna’s aim to provide a set of arguments in support of the claims of the Perfection of Wisdom sūtras, to explicate their contents, and to demonstrate their philosophical feasibility.

The composition (or at least the scripturalization) of the Perfection of Wisdom texts is characterized by a process of textual expansion followed by textual abbreviation. We can divide the development of the Prajñāpāramitā texts into four broad sections: the early phase, the phase of expansion, the phase of contraction (each lasting about two centuries), and a final, tantric phase.35 The earliest phase (about 100 BCE to 100 CE) sees the appearance of the earliest layer of the Prajñāpāramitā texts, a section of the Perfection of Wisdom in 8,000 verses (Aṣṭasahāsrikaprajñāpāramitā), which may go back as far as 100 BCE,36 though the process of composition of the text is likely to have extended over two centuries. This would date the writing down of the first Mahāyāna texts to the same time as (or possibly earlier than) the scripturalization of the Pāli canon.37

During the second phase (100 CE to 300 CE) the Perfection of Wisdom texts expanded, resulting in such works as the Perfection of Wisdom in 18,000, 25,000, or 100,00038 verses (there are even references to a version 125,000 verses (p.94) in length)—all substantially the same text, but differing in the extent to which repetitive lists are spelt out. During the same period commentarial works on the Perfection of Wisdom texts started to be composed, most importantly a gigantic commentary on the Perfection of Wisdom in 25,000 verses, the Mahāprajñāpāramitā-upadeśa-śāstra (大智度論‎), ascribed to Nāgārjuna, which was translated into Chinese by Kumārajīva. The Sanskrit original, now lost, is supposed to have been even longer (100,000 verses); Kumārajīva only translated the first chapter in full, and provided abstracts of the remaining eighty-nine.

In the third phase (300–500 CE) the expansive tendency of the Perfection of Wisdom literature is reversed. It is understandable that at this stage of its development the Perfection of Wisdom literature became very hard to read: the texts are difficult in themselves, and the enormous number of repetitions made it difficult to keep the main points in focus. If we are to believe the famous commentator Haribhadra, even scholars of the calibre of Asaṅga had difficulties dealing with them, finding that they ‘could no longer ascertain its meaning, because of the great number of repetitions, their inability to distinguish the different words and arguments, and its profundity’.39 It is therefore hardly surprising that various shorter versions made their appearance during this phase. Two of these are amongst the most famous Buddhist texts: one is the so-called ‘Heart Sūtra’ (Prajñāpāramitāhṛdayasūtra), a work that occupies an important role in virtually all Mahāyāna traditions, the other the ‘Diamond Sūtra’ (Vajracchedikāprajñāpāramitāsūtra). Both of these texts are quite short (the English translation of the former fits easily on one or two pages), but the abbreviation of the Prajñāpāramitā texts was taken to its extreme in the shortest of all versions, the Perfection of Wisdom in One Letter (Ekākṣarāprajñāpāramitāsūtra). Subtracting the usual preamble and conclusion it just consists of the letter A. This might appear a little bit less peculiar if we take into account that the sound A is not only the first sound of the Sanskrit syllabary, but can also be prefixed to nouns and adjectives to form their negations. Given the emphasis of the Perfection of Wisdom texts on negating various categories assumed by the Abhidharma, the idea of encapsulating the essence of the Perfection of Wisdom in the word ‘not’ is not entirely far-fetched.

Even though the Perfection of Wisdom texts have a special affinity with Madhyamaka, they are certainly not of exclusive interest to proponents of the Middle Way. Prajñāpāramitā texts continued to appear from about 100 BCE through the entire history of Indian Buddhism up to its demise in the twelfth (p.95) century, and have been studied, summarized, and commented upon by a variety of authors from different Indian Mahāyāna schools. The most famous of these is the Abhisamayālaṃkāra traditionally ascribed to Maitreyanātha, Asaṅga’s teacher. It is a 273-verse table of contents of the Perfection of Wisdom in 25,000 verses which has dominated the understanding of the text in India and Tibet. Hybrid versions that inserted the divisions of the Abhisamayālaṃkāra into the Prajñāpāramitā text itself appeared around the fifth and sixth centuries.

The Prajñāpāramitāsūtras form a thread that can be traced through the entire development of Buddhist philosophy in India (and beyond). It is no over-generalization to say that every Mahāyāna school of Buddhism in India has in some way sought to explicate the Perfection of Wisdom texts, attempting to show how their specific philosophical positions provide the best explanation of the theory of emptiness that these texts set out. All the great Yogācāra masters have composed commentaries on Perfection of Wisdom texts; apart from Maitreyanātha’s Abhisamayālaṃkāra, Asaṅga wrote a commentary on the Diamond Sūtra (Vajracchedikaprajñāpāramitāsūtra), Vasubandhu (at least according to the Tibetan tradition) composed a commentary on the Perfection of Wisdom in 100,000 verses, and Diṅnāga wrote a summary of the principal topics discussed in the Perfection of Wisdom in 8,000 verses, the Prajñāpāramitāpiṇḍārthasaṃgraha. Unlike other Mahāyāna sūtras that rose to prominence only within specific philosophical schools, the Perfection of Wisdom texts are of universal significance for the interpretation of any post-Abhidharma school of Buddhist thought in India.

In the last phase of the development of the Perfection of Wisdom texts (600 CE–1200 CE) various works clearly inspired by tantric modes of thought appeared. In these we find the attempt to reduce the essence of the Perfection of Wisdom to a single mantra or spell. Such as attempt can already be found in the Heart Sūtra, which encapsulates the text in the mantra of the Perfection of Wisdom (gate gate pāragate pārasaṃgate bodhi svāhā). In some of these texts we also find the Perfection of Wisdom personified, accompanied by specific rituals for worshipping her. She is depicted in female form, usually with four arms, the inner ones placed in the gesture of teaching the dharma, the outer ones holding a book (the text of the Perfection of Wisdom itself) and a rosary (for the repetition of her mantra).

After these brief remarks on the nature of the Perfection of Wisdom sūtras we can now return to Nāgārjuna’s biography. We may wonder whether its different elements do not pull in different directions. On the one hand there is the alchemist and magician, on the other the monk and philosopher who composes treatises to defend the Mahāyāna position. Modern Buddhologists have been wondering this too, and have suggested that we may be dealing with (p.96) distinct persons living at different times who have all been labelled with the same name.40

At least three Nāgārjunas have been distinguished in the literature. The first, and the Nāgārjuna most frequently referred to, is the philosopher who lived during the first and second centuries CE. He is also frequently connected with the famous Buddhist university of Nālandā, though this association faces certain difficulties.41 We do not have evidence for Nālandā as a major monastic establishment before the year 425 CE, a considerable time after the period during which Nāgārjuna is supposed to have lived. Moreover, neither of the Chinese pilgrims Xuanzang (玄奘‎), nor Yijing (義淨‎), both of whom spent some time at Nālandā, refer to Nāgārjuna as a famous alumnus.42

The second Nāgārjuna sometimes discussed is a tantric master who probably lived around 400 CE,43 and the third is an alchemist, probably to be dated around the seventh century.44 (We also occasionally find references to a fourth Nāgārjuna, an author of medical works.)45

This division also leads to a breaking up of the set of over a hundred works attributed to Nāgārjuna: all the philosophical works are considered to be composed by the first Nāgārjuna, while the tantric and alchemical works are taken to have been composed by the later ones. Traditional Buddhist narratives see no particular difficulty in accounting for the fact that different parts of this considerable number of works are likely to have been composed over the span of several centuries, since they argue that Nāgārjuna’s alchemistical experiments allowed him to extend his lifespan up to 600 years.

There are certain advantages to this theory of multiple Nāgārjunas, the chief one being that we can account for most of the motives in the various accounts of Nāgārjuna’s life without appealing to anything that would contradict the standard twenty-first-century naturalistic worldview. However, we should note the difficulties we see in traditional accounts of Nāgārjuna’s life (the reference to magical abilities, the long lifespan, the diversity of his literary output, the confusion of times and spaces associated with his life) are very much the product of a specific perspective chosen for looking at these accounts.

(p.97) As we have seen in the prophecy from the Laṅkāvatārasūtra cited above, Nāgārjuna is considered to have obtained the first bodhisattva ground. As the bodhisattva ascends the different grounds or levels of spiritual accomplishment, he acquires, in addition to the direct realization of emptiness that characterizes reaching the first bodhisattva ground, different sets of abilities. For the first ground these include the ability to live for 100 aeons, magically generate 100 versions of his body, and to teach 100 kinds of teaching.46 On the basis of this assumption it becomes clear that the traditional biographies of Nāgārjuna, including their descriptions of his various magical powers, being able control his lifespan or the place of his birth, being able to work miracles and so forth, are exactly the kinds of account one would expect.47 The fantastic, confused, or miraculous appearance of Nāgārjuna’s traditional biography only arises if we consider him to have been an ordinary human being, and assume that there is an objective set of truths out there about what happened during the life of that human being. As we noted before, the difficulty with this approach is that it does not cohere with several of the central claims of Buddhist philosophy. We can, I argue, achieve a more nuanced understanding of the complexities of the history of Buddhist thought by provisionally bracketing our contradicting assumptions, rather than attempting to ‘straighten out’ traditional accounts on the basis of contemporary historiography.

The number of works attributed to Nāgārjuna is large (more than 100 according to the Tibetan canon), but not all of them play the same important role. They include not only the highly theoretical works Nāgārjuna is famous for, but also contain some extremely practical texts: the Dhūpayogaratnamālā preserved in the Tibetan canon, for example, contains a recipe for making incense ascribed to Nāgārjuna.48

His single most important work is the Fundamental Verses on the Middle Way (Mūlamadhyamakakārikā), a set of 450 verses which is considered as intrinsically linked with Nāgārjuna as the Iliad is with Homer: what we speak about when refer to them are the authors of these respective texts.

The central works of Nāgārjuna can be divided into three broad categories: technical philosophical writings, letters, and hymns. The technical philosophical (p.98) works include six major texts,49 sometimes referred to as the yukti-corpus, or the ‘six texts on reasoning’ (rigs pa’i tshogs drug), and comprise, in addition to the Mūlamadhyamakakārikā, two shorter works on the notion of emptiness, the Sixty Verses on Reasoning (Yuktiṣāṣtikā) and the Seventy Stanzas on Emptiness (Śūnyatāsaptati), the Dispeller of Disputes (Vigrahyavyāvartanī), a discussion of more complex issues raised by the theory of emptiness in question-and-answer format, and the Vaidalyaprakaraṇa, a criticism of the sixteen ‘categories’ concerned with logic and debate discussed by the non-Buddhist Nyāya school. A sixth text, the *Vyavyhārasiddhi, seems to be no longer extant, apart from a few verses quoted by later authors.50

Nāgārjuna also composed two letters of advice to a king that contain some metaphysical discussion but are primarily concerned with ethical matters. These are the Jewelled Garland (Ratnāvalī) and the Friendly Letter (Suhṛllehka). These letters also offer a possibility of locating Nāgārjuna somewhat more firmly in space and time. The texts themselves do not give the name of the kings they were addressed to, though their Tibetan and Chinese translations provide us with the names bDe spyod (‘good conduct’) and Chantaka. Even though records of kings with these names have not come down to us, there is some possibility that the terms do not refer to individual kings, but to the Sātavāhana dynasty or one of its major sites.51 The Sātavāhana empire was based around Amaravati in today’s Andhra Pradesh, and lasted for four and a half centuries, from about 230 BCE to 220 CE. Based on a verse from the Ratnāvali, where Nāgārjuna mentions an image of the Buddha seated on a lotus,52 and the fact that images such as this were available only during the late part of the Sātavāhana dynasty in the Eastern Deccan, Joseph Walser has argued that Nāgārjuna composed the text during the reign of King Yajña Śrī Sātakarṇi (about 175 to 204 CE).53 The uncertainties inherent in such reasoning are apparent; nevertheless, it is valuable as constituting the best attempt so far at linking up Nāgārjuna’s philosophical activity with some dateable events in Indian history.

Nāgārjuna’s hymns, finally, are a group of short texts on the Buddha and his transcendent nature, interesting for their positive characterization of ultimate reality. In the Niraupamyastava, for example, we find characterizations of the ‘Dharma-body’ (dharmamayakāya) as blissful (śiva), stable (dhruva), and (p.99) permanent (nitya). There at least appears to be a certain tension between such characterizations and Nāgārjuna’s conception of ultimate reality as empty, yet they also provide an interesting link of Nāgārjuna’s thought with the theory of Buddha-nature (tathāgatagarbha) expounded by some Mahāyāna sūtras. What the precise relation between the notion of universal emptiness described by the Prajñāpāramitāsūtras and the notion of Buddha-nature found mainly in sūtras ascribed to the ‘third turning of the wheel of doctrine’ (see below, p. 186) amounts to is a question that has occupied the hermeneutical abilities of Buddhist commentators ever since. The main interpretative choice to be made is whether these notions are, despite their seeming tension, somehow compatible with each other and therefore to be assigned to the same level of truth, or whether they are inconsistent, so that one of the two has to be assigned to the category of provisional teachings (neyārtha) while only the other can be held as expressing the Buddha’s definite position (nītārtha).54

3. The Teachings of the Perfection of Wisdom

A good way of approaching Nāgārjuna’s Madhyamaka teachings is by looking at certain prominent topics within the Perfection of Wisdom literature and considering how they were philosophically developed in his works. The scope of the Prajñāpāramitā literature is vast, and its teachings are complex. Nevertheless, there are certain recurrent themes one can identify. Particular important amongst them are:

  1. 1. a criticism of the Abhidharma project;

  2. 2. the doctrine of illusionism;

  3. 3. an explicit acceptance of contradictions.

a. Criticism of the Abhidharma project

The characters that speak in the Perfection of Wisdom sūtras are the historical Buddha Śākyamuni, as well as usually various bodhisattvas, and the Buddha’s disciple Śāriputra. This ‘general of the doctrine’ (dhammasenāpati) is characterized in these texts as a ‘representative of an inferior kind of knowledge’,55 and this signifies the attitude towards earlier schools of Buddhism, and towards the Abhidharma in particular. The Perfection of Wisdom texts frequently criticize the ideals of realized practitioners of earlier Buddhism, the arhats and the pratyekabuddhas, and focus instead on the ideal of a (p.100) Bodhisattva. Here is how the Perfection of Wisdom in 8,000 verses56 characterizes their difference:

A Bodhisattva should not train in the same way in which persons belonging to the vehicle of the arhats and Pratyekabuddas are trained. How then are the arhats and Pratyekabuddhas trained? They make up their minds that ‘one single self we shall tame, one single self we shall pacify, one single self we shall lead to final nirvana’.…A Bodhisattva should certainly not in such a way train himself. On the contrary, he should train himself thus: ‘My own self I will place into Suchness, and, so that all the world might be helped, I will also place all beings into Suchness, and I will lead to nirvana the whole immeasurable world of beings’.

What is criticized here is not the validity of the realization of the arhats and pratyekabuddhas, but their limited scope. The Perfection of Wisdom in 25,000 verses compares the arhats and pratyekabuddhas to glow-worms, and the bodhisattvas to the sun.57 Both have kindled the flame of enlightenment, but the formers’ light only illuminates their own immediate surroundings, whereas that of the latter can potentially light up the whole world.

The Prajñāpāramitā texts also set out to reject the metaphysical doctrines of the Abhidharma, in particular its conception of fundamentally existent dharmas. A concise example is provided by the Heart Sūtra, which explains matters as follows:

O Śāriputra, any son or daughter of noble family who wishes to practice the conduct within the profound Perfection of Wisdom, should observe in this way:

He properly sees the five aggregates, and sees them as empty of intrinsic nature (svabhāva)…

Therefore, Śāriputra, in emptiness there is no matter, no feeling, no notion, no formations, no consciousness;

no eye, no ear, no nose, no tongue, no body, no mind, no form, no sound, no smell, no flavour, nothing to be touched, no dharmas;

there is no eye-sphere up to no mind-sphere, no sphere of dharmas, no sphere of mental consciousness,

no knowledge, no ignorance, no destruction, up to no destruction of old age and death

no suffering, no arising, no cessation, no path,

no cognition, no attainment, and no non-attainment either.

This passage is a negation of all the categories that form the heart of the Abhidharma’s ontological enterprise. The bodhisattva Avalokiteśvara, speaking through the inspiration of the Buddha, goes through the core categories of the Abhidharma system, beginning with the key dichotomy between nāma and (p.101) rūpa as represented by the five skandhas,58 through the twelve āyatanas, the twelve dhātus, the twelve links of dependent origination, up to the four noble truths and even enlightenment itself, and states that none of them exist in emptiness.

It is hard to overestimate how radical this step was. The theory of dharmas was the standard Buddhist account of how reality was constituted at the fundamental level, a theory that accounted both for what there is at the rock bottom, and what kind of phenomenology is based on this. If all this is rejected, the audience of the Prajñāpāramitā texts (who we have to imagine as well trained in the theories of the Abhidharma) might well have wondered what, if anything, was left.

b. The doctrine of illusionism

What seems to be left is a world that is not quite what it seems, a mere ephemeral creation similar to an illusion. In the Perfection of Wisdom in 8000 verses the gods question Subhūti, one of the disciples of the historical Buddha:


  • Beings are like a magical illusion, are they not just an illusion?
  • Subhūti:

  • Like a magical illusion are those beings, like a dream. For magical illusions and beings are not two different things, nor are dreams and beings. All dharmas also are like a magical illusion, like a dream. The various classes of Saints,—from Streamwinner to Buddhahood—also are like a magical illusion, like a dream.

  • A fully enlightened Buddha, also, you say, is like a magical illusion, is like a dream? Buddhahood also, you say, is like a magical illusion, is like a dream?

  • Even Nirvāṇa, I say, is like a magical illusion, is like a dream. How much more so anything else!

  • Even Nirvāṇa, holy Subhūti, you say is like an illusion, is like a dream?

  • Even if perchance there would be anything more distinguished, of that too I would say that it is like an illusion, like a dream. For illusion and Nirvāṇa are not two different things, nor are dreams and Nirvāṇa.59
  • (p.102)

    The illusionism propounded by the Perfection of Wisdom SŪTRAs is comprehensive. We are here not faced with a view of the world that relegates our everyday surroundings to the status of mere appearance in order to underline the truly real status of some religiously transcendent world. Rather, the entire round of existence including all beings, all dharmas, various degrees of realized practitioners, the Buddha, and even nirvāṇa are considered illusory in nature. Even the process of leading beings to liberation is compared to a magician dissolving a previously created illusion. Just as we would not want to say that the magician made an elephant vanish, because there was no elephant present in the first place, in the same way, the Perfection of Wisdom texts argue, there are no beings that are led to liberation.60

    This illusionistic doctrine runs through the entire corpus of the Perfection of Wisdom literature, but, rather surprisingly for such an unintuitive position, the texts do not in fact offer any arguments for why we should believe everything is illusory in the first place. Some scholars have suggested that the origin of this illusionistic doctrine is not the conclusion of a set of philosophical arguments, but a reflection of a particular mental state experienced in meditative absorption.61 The texts would then reflect the particular way the world appears to the meditator, thereby also acting as a guide by providing a description of the kind of state the associated practices are supposed to lead to. Sometimes the ‘attainment of cessation’ (nirodhasamāpatti) is mentioned in this context, an advanced meditative state in which all sensory perceptions and all mental activity are supposed to cease.62 Whether the world appears in any way to a meditator in this state where all mental activity has been suspended, and whether it could thus appear as a wholly illusory entity, is difficult to say. However, even if the illusionistic experience is not specifically connected with the ‘attainment of cessation’, the idea that at the core of the description of the world from the perspective of the Perfection of Wisdom texts lies a set of meditative experiences opens up fruitful ways of understanding the origin and the aim of the Prajñāpāramitā.

    One way of understanding the illusionism of the Perfection of Wisdom texts (as well as other instances where meditative practices appear to be a factor in shaping Buddhist philosophy) is as an ontologizing of meditative phenomenology. Because the world appears to the meditator in a specific way, and because meditative cognition is regarded as a particularly reliable route to knowledge, the world must also exist in the way in which the meditator experiences it. There is certainly some truth to this idea, but especially in the Madhyamaka context it is important to understand it in a sufficiently nuanced way. The idea seems to be that because of the intrinsic epistemic superiority of (p.103) meditative experience, and because of the soteriological efficacy of the corresponding meditative states which play a key part in progressing on the path to enlightenment, the meditator’s phenomenological claims should also be regarded as authoritative ontological claims. Yet the Mādhyamika would disagree with the some key premises of this argument. For them there are no epistemic instruments that by their very nature lead to knowledge of ultimate reality, and it is furthermore mistaken to believe that the efficacy of a theory (including its soteriological efficacy) must rest on the ultimate truth of that theory. As an insubstantial (niḥsvabhāva) chariot can fulfil its function of carrying wood, so an empty theory, a theory not grounded in ultimate reality, can lead to liberation. Instead of arguing that the salvific efficacy of specific meditative states and experiences shows that they correspond to the way the world works at the most fundamental level, their very efficacy is sufficient to argue why they, rather than other non-standard phenomenological states,63 should be cultivated, independent of any claims to ultimate truth. We should therefore be aware that the ‘ontologization’ of meditative phenomenology happens in Buddhist thought, and that it can explain a great deal about the development of Indian Buddhist philosophy, but that the Buddhist philosophers themselves (certainly the Mādhyamikas amongst them) finally move beyond it when spelling out the theory of the emptiness of emptiness.

    It is also worthwhile to note the somewhat different role that the Buddhist illusionistic worldview plays in the development of different Buddhist schools. We find early forms of this idea already in some of the canonical sūtras, when the Buddha says that ‘sensual pleasures are impermanent, hollow, false, deceptive; they are illusory, the prattle of fools’.64 The immediate aim of these teachings, that compare phenomena to foam,65 bubbles, mirages, the coreless trunk of a banana tree, and so on is to enable the practitioner to rid himself of attachment and aversion towards these insubstantial things. The Perfection of Wisdom texts added another turn of the screw by extending the illusionistic doctrine from the five psycho-physical constituents (skandhas) to nirvāṇa and even to the Buddha himself. The reason for the popularity of this comprehensive illusionism in the development of the Mahāyāna is not difficult to determine. One of its implications is the insubstantiality of the soteriological goal of the non-Mahāyāna Buddhist schools; another is the fundamental equivalence of cyclic existence and liberation. Both of these entail that for a practitioner (p.104) who has gained insight into the illusory nature of reality, the idea of leaving saṃsāra in order to obtain nirvāṇa must appear absurd. This view provides strong support for the ideal of the bodhisattva, a practitioner who, out of compassion for all beings, remains within cyclic existence until he has succeeded in liberating all of them as well. If there is no distinction between saṃsāra and nirvāṇa at the ultimate level, there is nothing the bodhisattva needs to escape from in order to obtain liberation. He can remain in saṃsāra in order to help sentient beings, increasing his insight by means of great compassion, and finally, once all beings have been liberated, can make the cognitive shift that transforms saṃsāra into nirvāṇa and the bodhisattva into a Buddha.

    At this stage we might wonder about the consistency of the Prajñāpāramitā’s worldview. On the one hand it rejects all the Abhidharma categories, and seems to come close to the view that there is nothing at all; on the other hand it does speak about bodhisattvas, about illusory appearances, Buddhas, and nirvāṇa, all of which it appears to take to exist in some way. How can these positions go together?

    c. An explicit acceptance of contradictions

    Contradictions (or, at least apparent contradictions) abound in the Perfection of Wisdom literature. How does the Perfection of Wisdom procure all-knowledge? We learn that, ‘In so far as it does not procure, to that extent it procures.’66 How do dharmas exist? ‘As they do not exist, so they exist.’67 What does the profundity of the Perfection of Wisdom consist in? ‘It cannot be developed by anything, nor by anyone, nor is there anything to be developed. For in perfect wisdom nothing at all has been brought to perfection.’68

    Edward Conze, one of the most important Western scholars of the Prajñāpāramitā texts, sums up this perplexing situation in a concise manner:

    The thousands of lines of the Prajñāpāramitā can be summed up in the following two sentences: 1. One should become a Bodhisattva…2. There is no such thing as a Bodhisattva, or as all-knowledge, or as ‘being’, or as the perfection of wisdom, or as attainment. To accept both these contradictory facts is to be perfect.69

    It is important to be aware that there is a sense in which the claim that ‘there is no such thing as a Bodhisattva’ is uncontroversially accepted by early Buddhism, namely the idea that each person (including a bodhisattva) is only a superimposition on a shifting coalition of psycho-physical aggregates, the skandhas. But the Perfection of Wisdom texts go further than this. As we saw above, their aim is the rejection of dharmas, not just the rejection of higher-level appearances based on dharmas. We frequently find the phrase ‘bodhisattva or (p.105) bodhisattva-dharma’ (bodhisattvaṃ vā bodhisattvadharmaṃ vā) as something to be negated, that is, not just the bodhisattva, in the sense of a person existing by svabhāva, but also the set of dharmas identified in the Abhidharma that can be collectively designated as a bodhisattva are rejected.70

    It is therefore evident that the Perfection of Wisdom texts left Buddhist philosophers with a formidable task: first, to determine what precisely they are saying, and second, to come up with a justification for why what they are saying is true, that is, a justification that does not simply rely on their authority as buddhavacana. Would it be possible to come up with arguments in support of the Prajñāpāramitā’s startling statements?

    The first Buddhist philosopher to develop the philosophical position of the Perfection of Wisdom texts in a systematic manner, developing arguments for their conclusions and considering replies to actual and potential objections, was Nāgārjuna. Before we can consider how the three themes just discussed feature in his texts, however, we must first consider a curious historical fact. If Nāgārjuna’s role is really to be understood as that of the recoverer (literally or metaphorically) and explicator of the Perfection of Wisdom texts, we would expect him to say so quite explicitly, and to quote the Prajñāpāramitā texts (and Mahāyāna texts in general) frequently in his works. In fact this is not the case at all. Of course matters depend to some extent here on what we consider Nāgārjuna’s authentic works to be. If we assume that the Mahāprajñāpāramitā-upadeśa-śāstra (大智度論‎) was written by Nāgārjuna, his association with the Perfection of Wisdom tradition is fairly obvious, and even other texts for which the attribution to Nāgārjuna seems more plausible contain a certain amount of Mahāyāna references. The Suhṛllekha, for example, encourages the king it addresses to emulate the Bodhisattva Avalokiteśvara and the transcendental Buddha Amitābha; the fourth chapter of the Ratnāvalī explicitly praises the virtues of the Great Vehicle. Nevertheless, this picture is slightly different if we concentrate on the other five works of the yukti-corpus, specifically on the Mūlamadhyamakakārikā. The only sūtra Nāgārjuna refers to by name here is the Kātyāyanāvavāda (the Sanskrit parallel to the Kaccānagottasutta);71 in addition a variety of other sūtras from the Tripiṭaka are quoted, but without explicitly giving their source.72 Based on this, some twentieth-century Buddhologists have argued that Nāgārjuna’s association with the Mahāyāna that forms part of traditional Buddhist history should be questioned.73 Far from being an explicator of the Prajñāpāramitā texts, or a (p.106) Mahāyāna proselytizer more generally, Nāgārjuna’s aim in the Mūlamadhyamakakārikā, it is argued, was merely to refute the excesses of scholastic overlay that the Abhidharma works had deposited on the structure of early Buddhism. Their philosophical positions, such as the view of the continuous existence of entities in the past, present, and future, or the notion of svabhāva (a term that is not mentioned in the sūtras of the Tripiṭaka), are something Nāgārjuna argues against, in order to recover the pure and unfalsified teaching of the historical Buddha.

    Both the idea of the rediscovery of historically unpolluted Buddhism,74 as well as the characterization of Nāgārjuna (and of the Buddha) as ‘empiricist philosophers’75 owe more to modern intellectual concerns than the originators of these ‘back to the basics’ calls might have thought. Nevertheless, there are certain aspects of this view that seem plausible. From the evidence available in Nāgārjuna’s works it is clear that he did not consider himself as an innovator or defender of the ‘new’ Mahāyāna creed against the benighted Abhidharma heretics.76 What Nāgārjuna set out to do in the Mūlamadhyamakakārikā and his other works is to explicate what he considered the true meaning of the Buddha’s words. He saw this meaning as expressed both in the Perfection of Wisdom texts and in the sūtras of the Tripiṭaka, and Nāgārjuna was striving to explain the unified philosophical vision of these texts, to supply arguments that are not present or not explicit in these texts, and to defend them against variant interpretations he regards as erroneous, such as those found in the Abhidharma. Still, we might ask ourselves, why does he then not explicitly quote Abhidharma and Prajñāpāramitā texts side by side? Why do all the quotations in the Mūlamadhyamakakārikā come from texts of early Buddhism?

    First, a fact that earlier authors overlooked, even though it had already been pointed out by Conze77 some time ago, is that the salutary verses (nāmaskāra) in praise of the Buddha by which Nāgārjuna begins his Mūlamadhyamakakārikā are clearly derived from a passage in the Perfection of Wisdom in 25,000 verses. Interestingly, though, this citation consists largely of phrases that occur individually in the early Buddhist sūtras, though not in the form in which they (p.107) are put together here.78 The nāmaskāra thereby fulfils a curious double function: on the one hand putting a section of a Prajñāpāramitā text at the very beginning of this key treatise conveys a clear signal about the intellectual lineage in which Nāgārjuna wants to position his work. On the other hand, nothing forces this reading. All the terms Nāgārjuna uses here can be found in the sūtras of the Tripiṭaka. The reason for this, and for the prevalence of non-Mahāyāna citations in the Mūlamadhyamakakārikā, as well as for the absence of explicit reference to Mahāyāna texts, Walser has argued79 lies in the specific historical situation the Mahāyāna movement found itself in during Nāgārjuna’s times. As a minority school, it was trying to promote the acceptance of its interpretation to a majority of Ābhidharmikas. It would obviously have been of very little use to try to do this by reference to Mahāyāna sūtras; even though these explicitly endorsed the superiority of the Mahāyāna interpretation, they were not accepted as authoritative by the Ābhidharmikas. Nāgārjuna’s strategy, therefore, was to argue for Mahāyāna conclusions by restricting himself to explicit references to texts that both the Ābhidharmikas and the Mahāyānists would consider as authoritative. We see here a milder form of the problem Buddhists later faced when debating with non-Buddhist opponents. Obviously such debates could not make any reference to scriptural authority, since the two parties did not acknowledge each other’s canon. But in the Buddhist case there was such a shared canon, and Nāgārjuna set out to demonstrate that the Tripiṭaka texts that the Ābhidharmikas regarded as authoritative could be given a Mahāyāna reading. This procedure might remind us of the cuckoo and her eggs, but there is no reason to believe that matters appeared in this way to a thinker like Nāgārjuna. His aim was to bring out what he considered to be the authentic meaning of the Buddha’s teachings, and to do so in a manner that would convince the largest possible number of his co-religionists.

    4. Key Themes of Nāgārjuna’s Thought

    After this brief discussion of the nature of Nāgārjuna’s relation with the Abhidharma, let us now get back to the question how the key themes of the Perfection of Wisdom literature that we identified above get taken up in Nāgārjuna’s works.

    a. Nāgārjuna and the criticism of the Abhidharma project

    There are two obvious areas of the Abhidharma project that Nāgārjuna criticizes: their idea of the goal of the Buddhist path as obtaining the stage of a śrāvaka or pratyekabuddha, which he replaces by the ideal of the bodhisattva, and the Abhidharma metaphysics of dharmas as ultimately real entities. In the (p.108) advice to a king Nāgārjuna gives in the Ratnāvalī he is very explicit in pointing out that the Mahāyāna (and the bodhisattva ideal it teaches) constitutes a set of teachings of the Buddha for beings with higher spiritual capacities than those for whom the sūtras of the Tripiṭaka are intended:

    Just as a master of grammar teaches even the alphabet to disciples,

    Even so the Buddha teaches his doctrine to those to be tamed as it is accessible to them.

    He taught his doctrine to some so that they turn away from evil deeds,

    To others so that they could accomplish meritorious deeds, to others [a teaching] based on duality.

    To some others [he taught a doctrine] beyond duality, deep, terrifying those who are afraid [of such teachings];

    The heart of compassion and emptiness, the means of obtaining enlightenment.

    Therefore the wise ones must destroy any feeling of aversion towards the Mahāyāna

    And generate special faith in order to attain to complete enlightenment.80

    This idea of a progression of doctrines (a common topos in Indian philosophy and doxography) suggests that the idea frequently found in the secondary literature, that Nāgārjuna’s aim was ‘to refute the Abhidharma’, needs to be seen in a more nuanced manner. In another passage in the Ratnāvalī Nāgārjuna advises the king to ‘definitely realize with vigour’ a list of fifty-seven ethical faults in order to avoid them.81 This list most plausibly derived from an Abhidharma text.82 A list of seven kinds of pride given in the same work83 also derives from Abhidharma sources, and is later included in Vasubandhu’s Abhidharmakośabhāṣya.84 Nāgārjuna’s attitude is therefore very far from a wholesale rejection of the teachings of the Abhidharma; in this context he explicitly recommends the close study of one of its topical lists (mātṛkā). In fact there would have been no reason for rejecting Abhidharma doctrines tout court. They are, after all, an attempt to systematize, explicate, and develop the teachings contained in the Buddha’s sūtras of early Buddhism, and the fact (p.109) that Nāgārjuna considered some of these explications (even very central ones) as mistaken does not mean that he would have rejected them all. Even though it is difficult to be definite about this, there may have been a political (or, to use a favourite Mahāyāna term, ‘skilful’ (upāya)) dimension to Nāgārjuna’s acceptance or endorsements of certain Abhidharma positions. We know that the different pre-Mahāyāna schools of Indian Buddhism had distinct but conceptually overlapping Abhidharma texts, even though only two, the Sarvāstivāda and the Theravāda, have come down to us in their entirety. It is not entirely implausible to assume that some of the Abhidharma positions Nāgārjuna takes on board were those defended by the Abhidharma of the monastery in which he was staying. Like the references to Tripiṭaka sutras, the references to this canon would have allowed Nāgārjuna to underline the legitimacy of the Mahāyāna outlook in the eyes of his non-Mahāyānist fellow monks.85

    The key metaphysical notion of the Abhidharma that Nāgārjuna attacks is that of svabhāva or intrinsic nature. His theory of emptiness means simply that all things are empty of intrinsic nature. In early Buddhist teaching the doctrine of emptiness was primarily spelt out as the selflessness of persons (pudgalanairātmya), arguing that there is no permanent, self-sufficient personality core that notions like ‘I’, ‘me’, or ‘mine’ picked out. With Madhyamaka the domain of emptiness expanded so that it covered all non-persons as well; in addition to the selflessness of persons, it gave a prominent position to the selflessness of all dharmas (dharmanairātmya). The idea of the emptiness of dharmas is not entirely straightforward. Whereas persons might be considered to possess a substantial soul or ātman, other things do not do so in any obvious sense. For this reason we need a more general notion of what all empty things, persons and non-persons, are empty of, and this is the notion of svabhāva. Nāgārjuna characterizes svabhāva by two important properties: it is not adventitious (akṛtrimaḥ) and not dependent on something else (nirapekṣaḥ paratra).86 If we recall the notion of svabhāva we find in the Abhidharma,87 we remember that an object has some property as its svabhāva if this property does not depend on other things, apart from the thing that brought the object into existence in the first place. Thus a chariot is no chariot by svabhāva because it borrows its nature from its simultaneously existent components. The great elements (mahābhūta) like water, fire, and so on do not derive their nature from anything else. They have their nature by svabhāva, yet the Abhidharma has no difficulties with accepting that they are causally produced. But we just saw Nāgārjuna say that dependent things cannot exist by svabhāva, and a causally produced dharma is of course dependent on the causes and conditions (p.110) that brought it into existence. It seems as if the Abhidharma and Nāgārjuna meant quite different things when they spoke of svabhāva, and if that is the case we might wonder whether Nāgārjuna’s rejection of svabhāva, setting out to explicate its rejection in the Perfection of Wisdom texts, is really a rejection of the svabhāva the Ābhidharmikas had in mind, or whether it is the rejection of something else.

    The reason why Nāgārjuna regards being causally produced as incompatible with having svabhāva, while the Ābhidharmikas did not see a conflict between these, lies in their different concepts of causation. In order to see why this is the case we need to look more closely at Nāgārjuna’s causal argument against svabhāva.

    Nāgārjuna’s argument is based on some important assumptions about time. The first is presentism. This is the idea that only the present moment, and neither the past nor the future, exists. The second is momentariness, the view that nothing has any temporal ‘thickness’, that all things, once arisen, last only for a moment and disappear immediately afterwards. Causal processes happen in a sequence: first there is a cause (say, a spark), and at a later time the effect arises (the explosion).88 But this entails that the causal relation is always missing one related object. When the cause exists the effect does not yet exist, and when the effect exists the cause has already passed out of existence. Since the joint premises of presentism and momentariness have squeezed the whole of reality into the present moment, and since the causal relation requires at least two successive moments, one relatum of this relation will always fail to exist. It seems as if this shows that causation does not exist, since there cannot be two-place relations without two relata. Yet the world appears to us as causal through and through, and the Mādhyamika has to find a way to account for this. He does so by suggesting that even though the effect cannot exist at the same time as the cause, the idea of the effect can. When the cause exists, the effect is supplied by anticipation (when the spark flashes, we expect the explosion); when the effect has arisen, the cause is supplied by memory (we remember the spark when the explosion happens). What this means is that the causal relation inevitably contains one relatum that is a conceptual construct, and if this is the case, causation cannot be an objective relation that obtains in a mind-independent manner. Moreover, causally produced things have the property of being causally produced essentially, they would not be what they are without being so produced. If the teacup in front of me had not come into being from a set of causes and conditions it would not be that cup. But this means that the essence of such things involves conceptual construction, and (p.111) anything that does so cannot exist by svabhāva, since it entails that the object is not what it is all by itself, but in dependence on other things, such as conceptualizing minds. For this reason all causally produced things must be empty.

    It is now easier for us to understand why Nāgārjuna considered having a cause as incompatible with having svabhāva, while the Ābhidharmikas saw no incompatibility. We recall that for the Sarvāstivādin, the property of having svabhāva applies to dharmas atemporally. Whether a dharma presently exists, has existed or is yet to be, its svabhāva is there; the only thing that changes is the dharma’s efficacy. But in this case causal production or destruction does not affect the existential status of svabhāva, and as such there seems to be little motivation for thinking that such causal relations detract from its self-sufficiency or ‘own-being’. Of course, as we have seen, not all schools of Abhidharma accepted this curious idea of continuous existence. The Sautrāntikas considered dharmas as momentary phenomena, with no persistent svabhāva in the background. Rather, they suggest that as soon as a dharma passes out of existence, it causes its similarly momentary successor-dharma. Instead of a fire-dharma with a continuous svabhāva that gains its efficacy only in the present moment, we have a string of fire-dharmas, each causing the next.

    The Abhidharma’s criteria for being a dharma (and hence for having svabhāva) were formulated in terms of analysis at one point in time. If something does not disappear once it has been taken apart, either physically or conceptually, it is a dharma. The fact that a particular fire-dharma is caused does not conflict with the fact that as long as it exists in cannot be broken down into more basic constituents in either of these two ways. It is also worthwhile to remember that the reason for denying objects with parts a svabhāva is that they are taken to borrow their nature from objects that do not have that nature. A chariot thus borrows its nature from the axle, wheel, and so forth, none of which is itself a chariot. We would not even be able to say that a momentary fire-dharma borrows its nature from its cause in this way, for the reason that its cause is a fire-dharma too. The fire-dharma does not seem to depend on anything that is not a fire-dharma. But once we replace the Abhidharma conception of causation with the one we find in Madhyamaka, entities that were formerly regarded as unproblematically possessing svabhāva suddenly do not do so any longer. Being causally produced now incorporates an element of conceptual construction, since, the Mādhyamika argues, without the mind’s handiwork there would be no causal relation in the first place. The Sautrāntika’s momentary dharmas are caused by their very nature, and if we properly understand what that means we realize that upon taking their nature apart there is, after all, an element there that, when taken away, will make the dharma disappear. If the fire-dharma is not causally produced, and thus ultimately dependent on our conceptualization, there would be no fire-dharma. The Ābhidharmika and the Mādhyamika do not operate with (p.112) different conceptions of svabhāva and are therefore not talking past each other. They do, however, have different concepts of causation, and this has repercussions on their views of the kinds of things to be considered as having svabhāva in the first place.

    The second main argument against svabhāva Nāgārjuna discusses is not based on the notion of causation, but on the problem of identity or difference of wholes and parts. The argumentative structure will be familiar to us from Abhidharma arguments setting out to show that wholes cannot be ultimately real objects (dravya) but must be conceptually constructed (prajñapti).89 Their reason was that if wholes were ultimately real they should stand in a clearly defined relation to their parts. But it turns out that both the assumption that the whole is identical with the parts and the assumption that it is a separate entity distinct from them leads to problems. From this the Ābhidharmikas infer that wholes are no entities in their own right, that they do not exist in the same sense as the parts, but are merely conceptual superimpositions on the parts.

    Nāgārjuna now continues this argument on the level of the dharmas, that is, he applies it to the entities that came out as ultimately real according to the mereological argument of the Ābhidharmikas. Of course he cannot do so with reference to the first criterion for being a dharma (mereological simplicity), as the dharmas have no parts. Instead, he focuses on the second, the idea that ultimately real objects must also be simple not just in terms of material decomposition, but also when it comes to conceptual decomposition.

    The question Nāgārjuna is now asking is whether a dharma as a conceptual whole is identical with or distinct from its parts. If we consider one of the ‘four great elements’ (mahābhūta), such as the water-element, as a dharma, we will naturally want to characterize it as an individual (the dharma) that has a number of properties (wetness, stickiness, and so on). Is the dharma distinct from its properties, or are they one and the same thing? Suppose they are distinct. In this case we have an individual on the one hand, and various properties attached to it. What is the nature of the individual? If it really is distinct from all its properties, it is just what is left when all its properties have been abstracted away—what in common philosophical terminology is called a bare particular. Is this bare particular what it is by its intrinsic nature, that is, by svabhāva? If it is, a water-dharma would come with two svabhāvas or natures, that of being a bare particular and that which is the specific characteristic or svalakṣaṇa of water, namely being wet. But nothing can have two natures, since a thing’s nature makes it what it is in contrast with other things. So it would be safer to say that a bare particular is not what it is by its own nature. (p.113) This, however, just means that it borrows its nature from something else, and, far from having reached ontological rock-bottom with this peculiar idea of an object with no qualities, we now have to continue our analysis. Being a bare particular would in turn be a property of something else, an even barer particular to which this property attached in the same way as the property of being wet attached to the original bare particular. At this stage we can of course repeat the whole argument, and the unappealing prospect of an infinitely descending sequence of barer and barer particulars opens up.

    These kinds of problems might suggest that we took a wrong turn earlier on. We should have rather said that the dharma, the whole, is not distinct from its properties, the parts. Of course the literal identity of the two is hard to make sense of, as the dharma is one, but its properties are many, and nothing can be both one and many. But we could still spell out this idea by arguing that the entire concept of an individual as a metaphysical condensation nucleus to which properties attach is superfluous in the first place. Instead, we could say that the water-element is not a dharma itself, but that it is simply a collection of coexisting property particulars, sometimes referred to as tropes.90 These property-particulars would then be the real dharmas. The water-element is just a conceptual construction; while there is a wetness-trope and a stickiness-trope and so on, which congregate, the dharma as an individual is simply superimposed on these in the same way the chariot is superimposed on its parts.

    Nāgārjuna now raises the question of what makes all these different particularized properties distinct. On the face of it, the answer seems to be obvious: a particularized property is particularized because it is the appearance of a property at a specific space-time location. What distinguishes two wetness-tropes is that one is wetness here now, and the other wetness there then. But the matter is not a simple as it looks, for if particularized properties are really the only fundamental category we assume (and this is the case with the Abhidharma’s dharmas) then they should account for all properties, including spatial and temporal properties. That means that ‘being at a certain place’ and ‘being at a certain time’ are both tropes as well. They then lose their status as privileged individuators of other tropes.

    One way of fixing this problem is as follows. We do not really need special properties in order to individuate one trope from the next. In fact, we do not even need to know which trope is which in order to tell one from the other. Tropes do not exist in a lonely state, but in complexes with other tropes. We can then simply individuate each trope by determining which other tropes it (p.114) co-occurs with in a complex we conventionally label as an ‘invididual’. This will be different for every trope.

    Despite its elegance, this procedure has a crucial weakness. Tropes are no longer individuated according to inner own nature or properties, but with reference to other tropes they co-occur with. This means that each trope depends for being the kind of thing it is on other tropes; it is only through the existence of tropes other than the wetness-trope that not all wetness-tropes coalesce. If this is the case, tropes can no longer be considered as possessors of svabhāva.

    It thus appears that all ways of spelling out the relations of the parts of dharmas to the whole either lead to problematic conclusions or to positions that cannot ascribe svabhāva to them.

    The last of Nāgārjuna’s arguments against svabhāva I want to consider takes as its starting point the observation that things around us change. The world we observe is not static but characterized by things continuously changing their properties, coming into existence, and going out of existence. Nāgārjuna sees a conflict between this fact and the potential existence of svabhāva. He points out that:

    there is the lack of svabhāva of things due to the observation of change.…If svabhāva was found, what would change? Neither the change of a thing itself nor of something different is suitable: as a young man does not grow old, so an old man does not grow old either.91

    Let us consider the problem at the level of tropes. If water is hot now, and cold later, what could account for this change? One suggestion might be that the heat-tropes that inhere in this conglomeration of tropes that is the hot water turn into something else. This seems impossible, given that these tropes have heat as their nature. If this is the core of what they are, how could they ever turn into anything else? (This is the point of Nāgārjuna’s remark that an (intrinsically) young man could not grow old.) What would rather happen is that the heat-tropes go out of existence, and other, different tropes arise in their stead. This leaves open the problem of what causes these other tropes to arise. Obviously not tropes of the same kind (else there would be no change), nor does it seem satisfactory to assume that they have arisen without a cause. Yet it seems that one of these two possibilities has to obtain if there is to be change at all. If all the tropes only give rise to further tropes just like them, everything will always be the same: a man who is already old will not change into an old man.

    (p.115) A final possibility one might consider is that the tropes are actually permanent, do not come into existence or go out of existence, and that change is just to be explained by a local rearrangement of these tropes. When hot water cools down the heat-tropes neither change nor pass out of existence, they will just go somewhere else. The difficulty with this suggestion is that it does not explain why the permanent tropes arrange themselves in ever-changing combinations. It cannot be anything going on inside them, like the arising and ceasing of some repulsive force, since in this case the tropes would precisely not be permanent, but would be changing. Yet it is this notion of change that we want to explain in the first place, and a theory that involves entities existing by svabhāva seems to face considerable challenges in doing so.

    We have thus seen that Nāgārjuna underpins the Prajñāpāramitā’s criticism of the ontological part of the Abhidharma project, and of the central notion of dharmas that exist by svabhāva that this involves, with a variety of arguments involving such different concepts as causation, parthood, or change. These are certainly not all the arguments against svabhāva we find in Nāgārjuna’s texts, nor are they all we find in the Madhyamaka literature that follow him. But they constitute a representative sample of argumentative approaches to backing up a set of claims made in the Perfection of Wisdom sūtras that also have considerable systematic potential as philosophical arguments in their own rights.

    b. Illusionism in Nāgārjuna’s thought

    We noted above that the Perfection of Wisdom texts expound a thoroughgoing illusionistic theory; for more-or-less any concept that forms part of the Buddhist path we can find a section of a Prajñāpāramitā text that says that this concept is not real. However, the theory expounded there is not one that defends a kind of appearance/reality distinction such as we can find, for example, in Vedānta. Its aim is not to show that the world we find around us is all empty and hence illusory, while that there is another world separate from this one that is the only real one. The Abhidharma account, on the other hand, can be interpreted as endorsing this appearance/reality distinction: composite objects are the appearance, but the only real things are the individual dharmas.

    Nāgārjuna employ’s the Perfection of Wisdom’s illusionistic metaphors frequently in his own works. At the conclusion of chapter 17 of the Mūlamadhyamakakārikā he notes that:

    31. Just as the Teacher by his supernatural power fabricates a magical being that in turn fabricates yet another magical being,

    32. so with regard to the agent, which has the form of a magical being, and the action that is done by it, it is like the case where a second magical being is fabricated by a magical being.

    (p.116) 33. Defilements, actions, and bodies, agents, and fruits, are similar to the city of the gandharvas; they are like a mirage, a dream.92

    The example of one of the Buddha’s miraculous performances, in which he produced a phantom magician who then produced another phantom in turn, illustrates the claim that action is not fundamentally real, nor is the agent who brings it about. Things that lack svabhāva can be brought about by other things that lack svabhāva, and this point can be generalized: unlike in an appearance/reality scenario where appearances are finally grounded in something substantial, in this case it is insubstantial entities all the way down. The dharmas that serve as the basis of partite entities are not any more real than the partite entities themselves. It is perhaps not too surprising that both non-Buddhist and Buddhist critics quickly accused Madhyamaka of nihilism because of arguments such as these.93 Asaṅga, for example, accuses the Mādhyamikas of misunderstanding the meaning of the Mahāyāna sūtras. He claims in the Bodhisattvabhūmi that:

    Hence, some who have heard the sūtras connected with the Mahāyāna, which are difficult to understand, and associated with profound emptiness, manifest an indirect meaning, do not know the meaning of the description of reality as it really is. Inaccurately, they have views without cogency obtained through mere inference (tarka) and say: ‘All this is in reality just a designation. Who sees matters in this way sees them correctly.’

    For those, the designation does not exist in any way, because of the non-existence of a given thing (vastumātra) that is the basis of designation. But how will reality be mere designation? In this manner both reality and designation are rejected. Because of the rejection of designation and reality they are to be known as the most extreme kind of nihilist (pradhāna nāstika).94

    Yet the Mādhyamikas take great care to distance themselves from the nihilist position. As their name, ‘followers of the middle way’, indicates they emphasize the fact that they reject both the extreme of nihilism as well as the extreme of (p.117) postulating substantially existent entities. Nāgārjuna stresses the fact that even though things like chariots and pots are neither fundamentally real nor based on something fundamentally real, they can still perform various functions such as carrying wood or water. As a monetary economy does not need anything intrinsically valuable to serve as a guarantor of the value of currencies, but can function by relying on the beliefs and expectations of the participants in its economic exchanges, so things do not need to be grounded in something existing by svabhāva to do what they are supposed to do. As long as sufficiently many people participate in a process of joint designation the entities thus designated will continue to exist. This point about the functional efficacy of everyday entities is particularly important for the Mādhyamika, since a charge he sees himself confronted with frequently is that of moral nihilism. If, as the opponent claims, the Mādhyamika rejects the existence of all things, he will also reject the existence of karmic potentials. But if this is rejected it seems as if a main incentive for moral behaviour for ordinary people is gone. How can those not yet convinced of the virtues of altruism be held in check without arguing that their non-virtuous actions will, via their karmic fruits, undermine their own selfish desires? Add to this the fact that the Perfection of Wisdom texts the Madhyamaka is based on do not exempt bodhisattvas, the Buddha, and liberation from its all-encompassing view of emptiness, and we may be forgiven for asking to what extent we are still dealing with a form of Buddhism here. It is therefore essential that the Mādhyamika distinguishes existence without svabhāva (the way all things exist) from non-existence (as, for example, flowers in the sky and sons of barren women are non-existent), and underlines that the emptiness characterized by absence of svabhāva does not entail emptiness of functional efficacy. The fact that things are insubstantial does not mean that they cannot interact with one another, and for this reason cyclic existence, the Buddhist path, and its goal, liberation, have a firm place in the Madhyamaka worldview.

    c. Contradictions and Nāgārjuna’s thought

    A casual reading of Nāgārjuna works is likely to give the reader the impression that contradictory statements form an essential part of his philosophy. He frequently employs of a specific form of argument, the tetralemma or catuṣkoṭi, which lists four apparently exclusive and jointly exhaustive possibilities, all of which are then rejected. Consider the following example from the Mūlamadhyamakakārikā:

    ‘It is empty’ is not to be said, nor ‘It is non-empty,’ nor that it is both, nor that it is neither; it is said only for the sake of instruction.95

    (p.118) When we reject the view that things are empty, does that not mean that we say that they are non-empty? Denying both seems to be contradictory. But even if we have managed to do so, Nāgārjuna then points out that the denial of both is to be rejected as well. How are we supposed to make sense of passages such as these? Moreover, contradictions do not seem to be limited to tetralemma-style arguments. At the end of the twenty-fifth chapter of the Mūlamadhyamakakārikā Nāgārjuna points out that: ‘This halting of cognizing everything, the halting of hypostatizing, is blissful. No dharma whatever was ever taught by the Buddha to anyone.’96 After spending twenty-five densely argued chapters to explain the teachings of the Buddha (whom Nāgārjuna praises as the ‘best of teachers’ at the beginning of the work), we now learn that the Buddha never taught anything. We seem to find ourselves in the territory of the Perfection of Wisdom texts again, where one sentence asserts something which the next one then goes on to deny.

    Some modern interpreters have tried to address the puzzling occurrence of contradictions in Nāgārjuna’s arguments by suggesting that he may have adopted a non-classical logic that tolerates contradictions.97 While these interpretations do not lack a certain technical ingenuity, they appear to be more of a further development of certain Madhyamaka ideas than an attempt at a rational reconstruction of what Nāgārjuna might have had in mind when he composed these verses. Neither in Nāgārjuna’s own writings nor in those of his Indian commentators do we find clear evidence that they were trying to develop a non-classical logic, nor that he challenged the principle of the excluded contradiction as a logical law.98 To this extent we need to understand the contradictions proclaimed by the Perfection of Wisdom texts that the Mādhyamikas set out to explicate as merely apparent, but not as actual contradictions.

    There is, however, another hermeneutic device to help us understand what is going on with these apparently contradictory statements that has a clearly attested historical status. This is the theory of the two truths. This doctrine occurs in some form or other in most systems of Buddhist philosophy, though the Mādhyamikas are probably those that made the greatest use of it. It claims that we have to distinguish two different kinds of truth (or two different kinds of reality—the Sanskrit term satya can refer to either), a conventional truth of everyday reality (saṃvṛtisatya) and an ultimate truth (paramārthasatya). Note that we are dealing here with two kinds of truth, not with a truth and a falsity. (p.119) Both conventional and ultimate truth have their uses, though they differ in soteriological efficacy. Conventional truth allows us to achieve innerworldly aims (build an aeroplane, calculate the value of π‎), while ultimate truth is what we have to realize in order become liberated from saṃsāra; it describes the ultimate nature of reality.99

    We can then use this distinction to dispel the appearance of paradox in instances of the tetralemma such as the one given above, by arguing that not all its negations relate to the same truth. Rather, when properly understood, what the passage from Nāgārjuna says is that we should not assert that things are conventionally empty, since our daily interaction with them relies on the mistaken assumption that they exist with svabhāva. Nor should we say that they are ultimately non-empty, since when we apply Madhyamaka reasoning to them, reasoning which is designed to uncover ultimate truths about them, we do not find that their existence with svabhāva can be supported. If we do not assert these two, we obviously do not assert them both. What about rejecting them both? We cannot do so if we assume that this rejection then gets at the ultimate truth of what reality is. When Nāgārjuna says that these assertions about emptiness are made ‘only for the sake of instruction’, he rejects the notion that they are in the business of telling us anything about what the world is like at the most fundamental level. The Madhyamaka only asserts them in order to refute specific misunderstandings his opponent may have.

    Similarly, we can agree that ultimately ‘no dharma whatever was ever taught by the Buddha to anyone’, because at the level of ultimate truth there is no Buddha, no dharma, and nobody listening to it. These are all only superimpositions valid on the conventional level, and for this reason it is conventionally false to say that the Buddha did not teach.

    The distinction between the two truths therefore provides us with an effective means to make sense of the seemingly contradictory statements we find in Nāgārjuna’s writings, and also in the Perfection of Wisdom literature. The drawback with this interpretation is that we have to presuppose that the sūtras and commentaries in question are incomplete; whenever they say ‘there is no x’ what they really mean is that there is no x at the level of ultimate reality, even though we are still allowed to speak about x at the level of conventional reality. There is certainly justification for this view, and the later Tibetan scholar Tsong kha pa (1357–1419), who put great emphasis on this ‘interpolation procedure’, provides a passage of from the Laṅkāvatārasūtra to support it.100 The Buddha points out to the bodhisattva Mahāmati: (p.120) ‘Mahāmati, thinking that they are not produced intrinsically, I said that all phenomena are not produced.’101

    The difficulty that arises when this procedure is applied too widely is that the negations we find in the Prajñāpāramitā literature or in Madhyamaka works can easily look as if they only concerned a kind of scholastic epiphenomenon, and not any kind of entity we are familiar with from everyday experience. When the Heart Sūtra says that there is no matter, it means ultimately real matter; when the Perfection of Wisdom in 8,000 verses says that a bodhisattva cannot be found, it means an ultimately real bodhisattva; and when Nāgārjuna says in the Vigrahavyāvartanī that he has no thesis, he means an ultimately real thesis. We might then ask ourselves first of all what these strange ultimately real simulacra amount to (just what is the difference between holding an ultimately real thesis and just holding a thesis?) or why it would matter that there are no such things (if there is no ultimately real matter, can mere matter not equally give rise to the kind of attachment the Buddhist path tries to show us to transcend)?102

    If concerns such as these are raised the method of interpolation has probably been over-applied, not only dissolving seeming contradictions but domesticating various Madhyamaka denials by restricting them to the realm of the ultimate. But this in itself is, of course, no criticism of the distinction between the two truths as a key hermeneutic principle for understanding Nāgārjuna’s writings, as well as the Perfection of Wisdom texts they set out to explicate.

    5. The Commentators

    After considering some of the origins of Madhyamaka in the Prajñāpāramitā literature and tracing the continuity of a set of its key themes in Nāgārjuna’s works, we want to look at what happened to Indian Madhyamaka in the roughly one thousand years between the composition of the Mūlamadhyamakakārikā and the eventual disappearance of scholastic Buddhism in India. This development has produced a vast body of literature of dazzling philosophical complexity, and attempting to do justice to it in a part of a chapter might easily seem like a foolish undertaking. What we can do, however, is point out some conceptual cross-sections, like a trench cut through a field of archaeological excavation, in the hope that at least parts of the major sights come into view. One such possible cross-section is the sequence of commentaries on the school’s foundational text, the Mūlamadhyamakakārikā. In true Indian scholastic fashion, variant interpretations and divergent conceptions of how (p.121) Madhyamaka was to be understood were most frequently made in the form of commentaries that aimed to clarify what Nāgārjuna meant when composing his major work.

    Commensurate with its importance, the Mūlamadhyamakakārikā is a work that has been frequently commented on in ancient India. Of the commentaries we know of, only a single one, Candrakīrti’s Prasannapadā, is preserved in its Sanskrit original. An early commentary, the Akutobhayā (sometimes considered to be an auto-commentary by Nāgārjuna), Buddhapālita’s commentary, and Bhāviveka’s Prajñāpradīpa are still extant in Tibetan, and further commentaries by Piṅgala and Sthiramati are preserved in Chinese translations. Four further commentaries, by Devaśarman, Guṇaśrī, Guṇamati (Sthiramati’s teacher), and Rāhulabhadra are, apart from some occasional quotations, lost.103 It is particularly interesting to note that Yogācāra masters such as Guṇamati and Sthiramati composed commentaries on the Mūlamadhyamakakārikā. This indicates that Nāgārjuna’s main work was not primarily conceived of as a treatise with a specific sectarian orientation, but as a fundamental Mahāyāna text with relevance for thinkers with different basic orientations.

    In our present discussion we will focus on ‘the great triumvirate of Madhyamaka commentators’,104 three scholars who are particularly important for understanding the different kinds of interpretations Madhyamaka thought attracted during its development in India: Buddhapālita, Bhāvikeka, and Candrakīrti.

    a. Buddhapālita

    In order to discuss Buddhapālita and his commentary it is necessary to first go back to a somewhat enigmatic earlier text, the Akutobhayā. This is an influential work belonging to the earliest stratum of Madhyamaka after Nāgārjuna. The other prominent works from this period are those of Āryadeva, who is believed to have been Nāgārjuna’s direct disciple, but did not compose any commentaries on his master’s work. Already in fourth-century India there is a tradition considering the Akutobhayā as going back to Nāgārjuna himself, a tradition that would explain the high regard in which this text was held.105 The relationship between this text and Buddhapālita’s own commentary (which, unlike the later commentaries, has no specific title—it is simply called the (p.122) Buddhapālitavṛtti, ‘Buddhapālita’s commentary’) is interesting. On the one hand the Akutobhayā is only half the length of Buddhapālita’s commentary, and often only provides a straightforward paraphrase of Nāgārjuna’s arguments, while Buddhapālita often expands on the arguments and offers a greater amount of analysis. On the other hand Buddhapālita borrows extremely liberally from the Akutobhayā; in fact the final five chapters of both commentaries are virtually identical. On the whole, about a third Buddhapālita’s commentary comes straight from the Akutobhayā. What is peculiar about this is not so much the extent of the borrowing but the fact that Buddhapālita nowhere points out that he is citing large passages from an earlier commentary on Nāgārjuna’s Mūlamadhyamakakārikā. Huntington106 has made the intriguing suggestion that the reason for this was that the Akutobhayā was not considered to be a work properly separate from Nāgārjuna’s root text, but a set of explanatory notes transmitted in a somewhat fluid form together with that text, notes that may go back to oral explanations of the root verses by Nāgārjuna himself. If this was the case, it would not be so surprising that Buddhapālita does not acknowledge a previous commentator he is citing, since he is simply selecting from an elucidatory tradition considered to have co-originated with the root text itself.

    This textual history is interesting because it shows that with Buddhapālita’s commentary we can to a certain extent reach back to a relatively early stratum of commentaries on Nāgārjuna’s root text. Biographical information on Buddhapālita is scarce. We can approximately date him to c.470–540, and like many great Indian philosophers he seems to have been born in South India. The Tibetan tradition regards him as a direct disciple of Nāgārjuna, a claim that requires us either to accept an extraordinarily long lifespan for Nāgārjuna, or a fairly wide understanding of what ‘discipleship’ is taken to amount to. Tāranātha, for example, explicitly raises the possibility of Nāgārjuna taking up a vidyādhara (‘knowledge-holder’) form to teach disciples after having quit his earthly body.107 In this way they would still be considered direct disciples of Nāgārjuna, even though they did not meet in human form.

    The only extant work of Buddhapālita’s is his commentary on Nāgārjuna’s Mūlamadhyamakakārikā; the Tibetan tradition also considers him as the author of various other commentaries on the sūtras and tantras, but none of these appear to have come down to us.108

    In his commentary Buddhapālita analyses and expands on Nāgārjuna’s arguments, and he does so exclusively in terms of prasaṅga methodology, (p.123) that is, demonstrating how Nāgārjuna’s arguments identify contradictions in the opponent’s own assumptions. At the time of Buddhapālita we have already reached a stage where marked differences in interpretation arise—Bhāviveka, who would severely criticize Buddhapālita’s expository methodology, can be dated to about 500–570. But as we know that the textual relation between the Akutobhayā and Buddhapālita’s commentary is quite close, and since the former did not spell out the arguments of Nāgārjuna in anything but prasaṅga terms, we may feel more justified to understand Buddhapālita not just as an individual commentator but as representative of a tradition of understanding Madhyamaka arguments that considerably preceded him.109

    b. Bhāviveka

    Bhāviveka, another master of likely South Indian descent, most probably overlapped with Buddhapālita at the beginning of the sixth century, though we have no information on whether they ever met.110 Unlike in the case of Buddhapālita, who left us only one text, various of Bhāviveka’s works have been preserved.

    Of primary interest for us is his ‘Lamp of Wisdom’ (Prajñāpradīpa), a very detailed commentary on the Mūlamadhyamakakārikā, and one that seems to have attracted considerable attention in the Indian philosophical world. At least two sub-commentaries were written on it, Avalokitavrata’s massive work (the longest single work in the Tibetan collection of commentarial works, the bsTan gyur), and one by Guṇadatta, which is no longer extant.

    Bhāviveka’s commentary is best known for its criticism of Buddhapālita’s commentary, and for its introduction of new argumentative tools for the exposition of Madhyamaka. In explaining Nāgārjuna’s arguments, Buddhapālita presents the reader with prasaṅga arguments, arguments that start by provisionally adopting some of the opponent’s theses, in order to show that a contradiction can be derived from them. Note that this differs in at least one important respect from a reductio ad absurdum argument. In the case of a reductio we begin with a hypothetical premise (e.g. that there are only finitely many prime numbers) and derive a contradiction from this. As a result, we can (p.124) then not only reject the original statement but also take the negation of the hypothetical assumption (there are infinitely many prime numbers) as established. A prasaṅga argument takes the first step, but not the second: the contradiction-generating hypothesis must obviously be rejected, but the argument does not commit us to adopt its negation instead.

    Bhāviveka argues that we should not just explain Nāgārjuna’s arguments by showing how they allow us to derive contradictions from the opponent’s assertions. In addition to this destructive enterprise, which can be best conceived of as clearing the ground by removing erroneous views, the Mādhyamika also needs to construct a set of positions of his own, and should provide complete syllogistic proofs of these. The contrast between their different ideas of how Nāgārjuna’s thought should be explained can already be seen from their remarks on the very first verse of the Mūlamadhyamakakārikā, where Nāgārjuna says that:

    Not from themselves, not from another, not from both, nor from no cause,

    do any originated entities ever exist anywhere.111

    In commenting on Nāgārjuna’s rejection of the first kind of origination, origination from itself,112 Buddhapālita points out that:

    To begin with, entities do not originate from their own selves, because their origination would be pointless and because there would be no end to origination. For there is no purpose in the origination again of entities that exist by their own selves. If they do originate again even though they exist, never would they not be originating. That, too, is not accepted.113

    We can clearly see here how Buddhapālita backs up Nāgārjuna’s claim (who, in this verse, has not provided us with any reason why entities do not originate from themselves) by providing an argument in support. The argument is a prasaṅga: it draws out two contradictory consequences from two slightly different conceptions of self-causation. First, causation is a process by which an existent cause brings about a not-yet-existent effect. If, as self-causation presupposes, cause and effect are the same thing, the effect is already there when the cause is, so that no causal relation would in fact have to take place. Secondly, the opponent might think that things are actually more short-lived than they appear: they pass out of existence frequently, but are immediately replaced by near-identical copies caused by the thing that existed in the (p.125) previous moment. If this thing is the only cause of the copy that is produced in the next moment, this scenario could plausibly be described as one of self-causation. The difficulty in this case is that it is hard to explain how anything could ever go out of existence. If all that is needed as a cause is the previously existent thing, everything should be permanent, since the conditions for self-copying always obtain.

    Given that neither of these scenarios accurately describes how causation occurs to us (effects follow their causes, and we regularly see things passing out of existence), Buddhapālita can argue that these absurd consequences allow us to reject the idea of self-causation.

    Here is Bhāviveka’s response to Buddhapālita’s exposition:

    That is not right, because no reason and example are given, and because faults stated by the opponent are not answered. Because it is a prasaṅga argument, a property to be proved and a property which proves that are opposite in meaning become manifest by reversing the original meaning: Entities originate from another, because origination has a result and because origination has an end.114

    Bhāviveka manages here to condense a number of points into a few words, and it is worthwhile to unpack them carefully. First of all, he points out that reason and example are missing from Buddhapālita’s exposition. Bhāviveka thinks that Nāgārjuna’s arguments should be presented in a commentary in the form of three-membered syllogisms, which would include reason and example.115 Such a syllogism would include:

    1. 1. a thesis (pratijñā) which ascribes the inferred property (sādhya-dharma) to the subject of the argument (pakṣa);

    2. 2. a reason (hetu) which ascribes the inferring property (sādhana-dharma) to the subject of the argument (pakṣa); and

    3. 3. an example (dṛṣṭānta) of something that has both the inferred and the inferring property.

    The syllogism that Bhāviveka supplies116 in support of Nāgārjuna’s claim that entities do not originate from themselves is the following:

    1. 1. Thesis: The six sense-organs [the subject] do not originate from themselves [the inferred property];

    2. 2. Reason: because the sense-organs exist [the inferring property];

    3. 3. like consciousness.

    (p.126) Bhāviveka only presents us with a particular instance of a general thesis (Nāgārjuna is not specifically talking about the sense-organs), but he clearly presupposes that a similar syllogism can be produced for any potential candidate for self-production. The sense-organs do not cause themselves because they already exist, and anything that presently exists no longer requires causal production. With the example Bhāviveka mentions, consciousness (caitanya), he makes a reference to the Sāṃkhya satkāryavāda theory of the pre-existence of the effect in the cause. The Sāṃkhya consider caitanya to be another name of puruṣa, a pure consciousness that is eternal, non-arising, and therefore also not self-arising.117 However, one need not follow Sāṃkhya in order to accept this example. Mādhyamikas also accept the existence of consciousness at the conventional level, and things that are already there do not need to be produced.

    When Bhāviveka speaks about ‘reversing the original meaning’ he has in mind a reader of Buddhapālita’s commentary who, having been convinced by the prasaṅga argument that self-causation has the contradictory consequence of producing something that is already there and continuing to produce it perpetually, might then infer that for this reason objects must be caused by what is different from them, as we experience causation to produce what is not there, and not doing so incessantly. In other words, the reader would have misunderstood the prasaṅga argument as a reductio, where the rejection of one alternative entails the adoption of the other. Bhāviveka believes that this is a danger connected with employing the prasaṅga methodology. As it does not endorse any positive thesis, people might mistakenly adopt the negation of a rejected option, even though it is fundamentally as deficient as the original position.

    He underlines this point by referring to the difference between two kinds of negation, implicative (paryudāsa) and non-implicative negation (prasajya-pratiṣedha). Originally a grammatical distinction, it was first given a substantial philosophical role by Bhāviveka. An implicative negation automatically endorses one of the remaining alternatives, as when saying that a man is a non-Brahmin we assert that he is a member of one of the other castes. On the other hand, we can also formulate matters slightly differently by saying that a man is not a Brahmin, and mean by this that he does not belong to any of the other castes either (because the system of castes is not applicable to him, or because he lives at a time when there are no castes, and so on). The exact way in which implicative and non-implicative negation are expressed grammatically does not matter; what is important is that we mean different things by them. Now Bhāviveka argues that when Nāgārjuna negates a proposition, such as (p.127) that things are self-caused, he does so in a non-implicative manner, without committing himself to any of the other ways in which a thing could be causally produced. From a Madhyamaka perspective this makes sense, since non-implicative negations are usually employed when we want to reject a presupposition made both by a statement and its negation. When we deny that the number three is red, we do not want to say that it has another colour instead (that is, negate it in an implicative manner), but deny that numbers could have any colour at all. In the same way, the different alternatives rejected in Madhyamaka arguments have the shared property of presupposing the existence of objects with svabhāva in some form, and if we negate them individually by non-implicative negation that is motivated by our rejection of this underlying assumption.

    Bhāviveka’s insistence on spelling out Madhyamaka arguments in the form of syllogisms seems to conflict somewhat with the fact that Nāgārjuna himself did not provide these in the Mūlamadhyamakakārikā. If they are so important, why were they not provided by the Master himself? Bhāviveka’s explanation is that Nāgārjuna, as the author of a root text (sūtrakāra), obviously wants to condense a considerable amount of complex material into the shortest possible form. This is fine for those students who have the mental capacity to understand the text in this form, but for all the others more extensive forms of explanation have to be provided. It is the task of the commentator to draw out the reasoning implicit in the root text and to unwrap it as much as possible in order to make it comprehensible to its audience. Explaining arguments by rendering them explicitly in syllogistic structure makes their underlying machinery visible and thereby generates maximal perspicuity. This is the reason, Bhāviveka claims, why the commenators of Nāgārjuna should explain his arguments in this way, even though he did not do so himself.118

    Bhāviveka’s explication of Madhyamaka arguments in terms of syllogisms implies the ascription of a thesis (or a multitude of theses) to Nāgārjuna, since it is such theses that syllogisms set out to establish. For Bhāviveka, this has the immediate advantage of preventing the danger of a student getting lost amongst the profusion of prasaṅga arguments, where seemingly every proposition is negated. The student may then endorse the wrong kind of positive thesis, namely one that is simply the opposite of a negated thesis and that is rejected as well (as in the case of someone who thought that Nāgārjuna’s denial of self-origination implied origination from other things). Of course, (p.128) Bhāviveka has to defend himself against the charge that such an attempt, helpful as it may seem, goes against certain key assertions of Nāgārjuna’s, who seems to reject that he holds a philosophical thesis quite explicitly. One of the most famous passages in this respect is verse 29 of the Vigrahavyāvartanī, where Nāgārjuna asserts that:

    If I had any thesis, that fault would apply to me. But I do not have any thesis, so there is indeed no fault for me.119

    Quotations such as these can easily be multiplied.120 What Bhāviveka needs to argue (and there is a certain leeway for doing so) is that the type of ‘thesis’ (pratijñā) Nāgārjuna rejects here is a specific kind of thesis (one, we would think, existing in terms of svabhāva, and thereby based on an objective, mind-independent world–language link),121 and that the theses ascribed to Nāgārjuna in his explicative syllogisms are of an altogether harmless, svabhāva-free kind that is not affected by Madhyamaka criticism. Still, do the theses Bhāviveka ascribes to Nāgārjuna not concern the ultimate, and does Nāgārjuna not hold that the ultimate is beyond concepts?

    Bhāviveka points out that while it is indeed true that there is a sense in which the ultimate truth is free from conceptualization (niṣprapañca) and hence is inexpressible and cannot be captured by any thesis, there is also another sense, which he calls ‘purified worldly knowledge’(śuddha-laukika-jñāna), which is accessible to concepts.122 Bhāviveka therefore takes the position that not all mundane knowledge is equally bad, and that below the level of enlightened beings, who can access the ultimate truth in a non-conceptual sense, there is also room for improved worldly cognitions, cognitions that are philosophically more sophisticated than those of the common cowherd, while at the same time still firmly located in the realm of the conceptual.

    Apart for seeing them as ensuring that Nāgārjuna’s message would not be misunderstood because of its largely negative methodology, Bhāviveka had at least two other reasons to stress the ascription of specific theses to Nāgārjuna’s Madhyamaka enterprise. The first has to do specifically with debate. We will remember that debates were of tremendous importance in ancient India for the promotion of ideas, for the prestige of individual scholars and scholarly communities, and for the patronage and its benefits that came with this prestige. In order to ensure the intellectual standing of Buddhism and the worldly endowments of its institutions, it was essential for Buddhist scholars to participate in debates, and one of Bhāviveka’s aims in his works is to facilitate (p.129) such debate for the Mādhyamikas.123 Unfortunately, Madhyamaka looks like a philosophical school that does not seem to sit well with debating conventions. If the Mādhyamika wants to enter into a debate with non-Buddhist parties he would have to play by the rules. And the rules of debate, such as those found in the Nyāyasūtra, specify that opponent and proponent must have a thesis they each want to defend. Mere intellectual sniping is not allowed; the Naiyāyikas specify a particular violation of debate rules called vitaṇḍā in the case of a debater who only wants to refute his opponent, without having a position he himself defends.

    Bhāviveka is clear in pointing out that the Madhyamaka does not commit the vitaṇḍā-fault. In fact, according to him, Nāgārjuna himself came up with a thesis about the nature of reality he sets out to establish in verse 18:9 of the Mūlamadhyamakakārikā, where he says:

    Not to be obtained by means of another, pacified, free from hypostatization, without conceptualization, not having many separate meanings—that is the nature of reality.124

    Of course, it is evident that all the terms Nāgārjuna uses in this verse are negative. However, we have something here that is presented in the form of a thesis ‘to encourage those who are just beginning’.125 Even though reality is ultimately beyond words, such negative characterizations can provide the basis of an eventual non-conceptual realization of the nature of reality.126 From his commentary on this passage we can see clearly how Bhāviveka tries to bring together the need for a thesis to be defended on the one hand, and the Madhyamaka reluctance to make any pronouncements on the ultimate nature of reality on the other. Elsewhere, Bhāviveka is more explicit, formulating the basic thesis of Madhyamaka himself: all things are empty of intrinsic nature (svabhāva), and that is their nature.127

    By establishing that Madhyamaka has a thesis to defend, Bhāviveka could ensure that the thought of Nāgārjuna was able contend in the intellectual arena of ancient Indian debate, demanding responses and setting out to refute contending positions. Ascribing a thesis to Madhyamaka thus appeared not just to have benefits for the Buddhist account seen from the inside (ensuring that amongst all the negations the wrong affirmative statement was not embraced by mistake), but also from the outside perspective, making sure (p.130) that Madhyamaka philosophy could effectively debate with its Buddhist and non-Buddhist opponents.

    The final reason for Bhāviveka’s concern to elucidate the Madhyamaka thesis is connected with this interest in doxography. His Madhyamakahṛdayakārikā and auto-commentary constitute the first surviving example of the doxographical genre in which separate philosophical schools are discussed in individual chapters. In this work Bhāviveka deals with two Buddhist and four non-Buddhist schools, the Śrāvakas and Yogācāras, as well as with Sāṃkhya, Vaiśeṣika, Vedānta, and Mīmāṃsā. The chapters begin with the opponent stating his position, followed by Bhāvavikeka’s reply. This structure makes a detailed discussion of rival positions possible, and presents them with the opportunity to describe their positions in a connected way, rather than using the opposing viewpoints as a mine from which objections are drawn more or less at random in order to explain certain aspects of the main theory being discussed. In also allows one to show how Buddhist thought relates to various non-Buddhist schools, achieving a greater integration of the Buddhist debate within the larger intellectual context of the time. Furthermore, one key aim of such doxographic treatises was to establish a doxographic hierarchy,128 that is, to set out different schools in ascending order of truth. This idea mirrors the early Buddhist distinction between sūtras with interpretable meaning (neyārtha) and those that did not need to be interpreted but could be taken literally (nīthārtha). Applied to doxographies, this distinction entails that different doctrines are not described as a set of varying wrong views that differ from the one correct view the author wants to defend; instead they are arranged in a hierarchy with the view to be defended at the top. The remaining doctrines can then be arranged in a sequence of positions that succeed in approaching the final view more and more closely. It is then natural to assume that if Madhyamaka is to be included in these doxographies it is essential that it is described as having a set of views to defend, and if a Madhyamaka author wants to structure a doxography such that Madhyamaka comes out at the top, all other positions have to be described as more or less accurate approximations of this final view.

    If we look at Bhāviveka’s spelling out of Nāgārjuna’s arguments in terms of syllogisms, and the specification of Madhyamaka theses this implies, he seems to follow a very sensible expository strategy. He worries that a purely negative, prasaṅga-style exposition of Nāgārjuna that, ‘based on what their opponents (p.131) accept, evokes a consequence that is unacceptable to their opponents’,129 might be misunderstood by generating unwarranted affirmative statements. He is also concerned that Madhyamaka should find its rightful place in Indian intellectual life. To this end it was necessary to present Madhyamaka positions in a way that could be defended in debates and included in doxographical hierarchies. Why, then, would these apparently harmless and reasonable expository points generate such forceful criticisms by later Mādhyamika authors, and why does the Buddhist tradition consider Bhāviveka’s works to mark the beginning of a key division of Madhyamaka into two incompatible sub-schools, the Svātantrika and Prāsaṅgika? Before addressing this question we need to consider the other central figure in this dispute: Candrakīrti.

    c. Candrakīrti

    Candrakīrti lived during the first half of the seventh century; as with most Indian thinkers, very little is known about his life. According to Tibetan accounts he was born in southern Indian and was a monk at Nālandā.130 Like Nāgārjuna, Tibetan sources describe him as either very long lived or as having obtained a form of immortality.131 Traditional accounts mention his magical abilities, which include milking a picture of a cow drawn on a wall in order to supply the monks of Nālandā with milk and butter, and animating a stone lion to frighten away a hostile army threatening the monastery, which promptly flees in terror.132 There is a clear philosophical significance to these stories that make their connection with a Madhyamaka master particularly apt. The theory of emptiness entails that things exist without an intrinsic nature or substantial core (svabhāva), but that this does not keep them from fulfilling their function at a conventional level. In fact this is precisely the point Nāgārjuna’s opponent raises at the beginning of the Vigrahavyāvartanī, and Nāgārjuna responds by saying that empty things can still fulfil their functions, an empty chariot can carry wood, an empty blanket can warm, and so on.133 These stories about Candrakīrti underline the efficacious power of the merely conventional by ascribing additional powers to things that are nothing more than representations. A painted cow cannot give milk, and a stone lion cannot roar, but the story that they can (or the illusory appearance produced by Candrakīrti that they can, depending on how we interpret the story), when (p.132) read in a Madhyamaka context, stresses that all causal efficacy there is ever going to be flows from the merely conventional.

    Candrakīrti’s key works are, first, a comprehensive commentary on Nāgārjuna’s Mūlamadhyamakakārikā, the ‘Clear Words’ (Prasannapadā), and second, a major independent work in verse together with an auto-commentary, called ‘Introduction to Madhyamaka’ (Madhyamakāvatāra). This latter text subsequently became extremely influential in Tibet. Rather than Nāgārjuna’s own foundational work, Tibetan scholars regarded the ‘Introduction to Madhyamaka’ as the seminal text for the study of Madhyamaka. Scholars from all schools of Tibetan Buddhism wrote commentaries on this text, and it was included as one of the five ‘key texts’ into the curriculum of the dGe lugs pa school.

    Given his exalted status in the Tibetan intellectual world, we might think that Candrakīrti’s thought was also very influential in India. Surprisingly, this is far from the case. Even though Candrakīrti’s works were preserved (and in fact his Prasannapadā is the only one of all the commentaries on Nāgārjuna’s main work that is still extant in Sanskrit), Candrakīrti’s place in the Indian intellectual landscape was—to put it mildly—inconspicuous during several centuries after he composed his works. We know of only one commentator on Candrakīrti’s works, Jayānanda, who lived during the twelfth century and wrote a commentary on his Madhyamakāvatāra. What is even more surprising is that Candrakīrti did not leave more of a mark in interactions with views he criticized. Given his sustained criticism of Bhāviveka’s commentary on Nāgārjuna, we might expect that Avalokitavrata (about 700 CE), the author of the massive subcommentary on Bhāviveka’s commentary, would devote a significant amount of space to defend the text he is commenting upon against Candrakīrti’s attacks. In fact all Avalokitavrata does is mention Candrakīrti as one of the eight authors who wrote commentaries on Nāgārjuna’s Mūlamadhyamakakārikā, without discussing Candrakīrti’s sustained criticism. While this fact does not allow us to infer too much about Avalokitravrata’s own view of Candrakīrti’s arguments (he might have thought that they were so deficient as to be unworthy of response, or so devastating that he did not know what to say), it does provide good evidence that Avalokitavrata thought that it was possible to write a commentary on Bhāviveka’s exposition of Nāgārjuna without giving an account of Candrakīrti’s criticism, and nobody would think he was leaving out anything obvious or declining to discuss trenchant criticism.

    Even highly influential later Madhyamaka authors, such as Śāntarakṣita (8th century) and Kamalaśīla (c.740–95), whose views do not at all cohere well with Candrakīrti, fail respond to his arguments or even mention him. This neglect of Candrakīrti’s writings continued with the early transmission of Indian Buddhism to Tibet, from its first introduction up to about 1000 CE. While practically all of the key Indian Madhyamaka writers were translated (p.133) into Tibetan, Candrakīrti’s main texts were not translated until the eleventh century.134

    Candrakīrti strongly disagreed with Bhāviveka’s exposition of Nāgārjuna’s arguments and tried to defend Buddhapālita’s interpretative stance against Bhāviveka’s exposition. His main point is the observation that, due to the curious nature of Madhyamaka thought, apparently procedural or exegetical matters, such as what format to present them in (in terms of prasaṅga-arguments or as syllogisms), could in fact not be confined to the merely formal, but have direct implications for the status of conventional reality. We can clearly see this in connection with Bhāviveka’s emphasis on spelling out Madhyamaka arguments in terms of syllogisms. Such syllogisms bring with them a thesis that the syllogism is an argument for, a thesis that the Mādhyamika holds independent (svatantra) of whatever their opponent accepts. As we have noted before, there seems to be a strong current in Nāgārjuna’s works to reject the acceptance of exactly such theses. We are no longer concerned here with the purely topic-neutral question of the best framework for explaining Madhyamaka teachings, because certain frameworks bring assumptions with them that have relevance at the level of the contents of the theory.135

    That syllogisms bring theses with them is not the only reason Candrakīrti distrusts them as expository devices for Madhyamaka. If we accept a syllogism we also have to be acquainted with its different parts, the subject, the inferred property, and so on, as well as with the various formal properties their relations must exemplify for the syllogism to be valid. According to the traditional Indian conception, all such knowledge is based on epistemic instruments (pramāṇa), such as perception or inference. Yet early Madhyamaka is very critical of these epistemological notions; Nāgārjuna discusses them at length in the Vigrahavyāvartanī, and Candrakīrti spells out his reasons for rejecting them in a variety of places in his works. In the Madhyamakāvatāra he points out that:

    If ordinary cognitions were epistemic instruments (pramāṇa), then mundane cognitions would see reality as it is. Then what necessity would there be for those other noble beings? What purpose would the noble path serve?136

    (p.134) Yet it is a common topos of Buddhist thought, going back to the very beginning of Buddhism, that the naive, unreflected, untrained cognition of the world get its nature thoroughly wrong. This is the reason why we need ‘noble beings’ such as the Buddha with sufficiently purified cognitions to show us a path we can follow so that we ourselves can see reality as it is. Ordinary cognition is so shot through with mistaken superimpositions resulting from ignorance that we cannot rely on it to give us a dependable account of conventional truth. Candrakīrti fears that Bhāviveka’s appeal to syllogisms will bring in its wake the traditional Indian epistemological picture of things that are by their inner nature instruments we can use to obtain knowledge of the world (pramāṇas), and others that by their nature are mind-independent objects this knowledge is knowledge of (prameya). This picture cannot be accepted by the Mādhyamika.137

    Candrakīrti also sees problems with Bhāviveka’s emphasis on the importance of Madhyamaka entering into a debate with rival systems. First, he holds that Nāgārjuna did not teach the arguments in his various works ‘out of fondness for debate’,138 that is, to defeat the opponent’s position and to establish his own view. In his Madhyamakāvatāra he makes it quite clear that attachment to one’s own view is something to be given up:

    Attachment to one’s own view, and likewise anger at the view of others, are mere conceptions.

    Therefore, those who eliminate attachment and anger and analyse correctly swiftly attain liberation.139

    Yet it seems as if engagement in debate in this way brings about precisely the kind of attachment to one’s own position that the Mādhyamika wants to avoid.

    Second, debates must obviously begin from a common ground; assumptions that are not common ground cannot be expected to have any probative force for both parties (a Buddhist and a Naiyāyika will not, for example, regard each other’s foundational sūtras as authoritative). But since the Buddhist’s opponent will conceptualize certain parts of the world as existing by svabhāva, the Buddhist will have to accept these claims at least at the level of conventional reality. This, Candrakīrti argues, reduces Madhyamaka to a kind of crypto-realism. (p.135) We are no longer dealing with a system that rejects entities existing by svabhāva, but rather with one that sides with the Mādhyamika’s opponent in accepting such entities, the only difference being that he follower of Bhāviveka will relegate the existence of svabhāva to the conventional, not the ultimate level. The denial of intrinsic natures characteristic of Madhyamaka seems to be no longer on the table, instead the argument now appears to be about what kind of reality should be assigned to them.140 Candrakīrti believes that something has gone radically wrong here. First of all, Nāgārjuna spends a considerable amount of time in his foundational work to demonstrate the contradictory nature of entities existing by svabhāva. How could we then believe that the Buddhist should incorporate them into their own theory of what reality is like at the level of everyday interaction? Secondly, we might well wonder whether the conventionally real intrinsic natures we have now introduced will not be able to function as objects of clinging just as well as the previous ultimately existent ones. If the key quality that keeps us locked in cyclic existence is grasping, why does it matter whether we develop unwholesome emotional attitudes towards our existent self, a self that exists with svabhāva conventionally (or other conventionally existent objects)? The overall result seems to be the same. We now see that what originally looked like a methodological disagreement about how to spell out Nāgārjuna’s arguments has at this stage been transformed into an ontological debate about what kind of things exist (this dimension of the dispute between Candrakīrti and Bhāviveka was later stressed by the fifteenth-century Tibetan scholar Tsong kha pa).

    A final point of Bhāviveka’s exposition that might create problems for the Mādhyamika is his fondness for doxography. If we arrange various systems of thought in a doxographic map, and additionally consider this to be hierarchically organized, with the systems discussed earlier being regarded as less accurate than the later ones, we have to assume that there are different ways of organizing conventional reality. Obviously none of the systems discussed (apart from possibly the final one, Madhyamaka) is able to provide us with an account of ultimate reality, so what they must do is provide us with gradually better accounts of the conventional reality, that is, the world we all live in. But Candrakīrti makes it clear that he does not accept such a stratification of the conventional that is based on its philosophical analysis. According to him, conventional truth is to be identified with ‘what even people like cowherds and women recognize’ (gopālāṅganājanaprasiddha), that is, with a view of the world that treats the regularities of the conventional world at face-value, without trying to come up with a series of theories of what is going on at the underlying metaphysical level.

    (p.136) Candrakīrti is therefore convinced that Bhāviveka’s commentarial technique is not just to be assessed as an expository device one may or may not find useful, but that it brings with it a host of philosophical assumptions that a Mādhyamika should be wary of accepting. It comes with the idea that Nāgārjuna held particular theses he sets out to establish (rather than just refuting those who, according to the Mādhyamika view, see the world incorrectly), that there are reliable epistemic instruments, that entities with intrinsic nature (svabhāva) exist at the level of conventional reality, and that a philosophical analysis of the conventional can supply us with better and better (though no best) theories of the nuts and bolts that underlie the working of the world. As we have just seen, Candrakīrti believes that all of these contradict key Madhyamaka ideas and is for this reason highly critical of Bhāviveka’s interpretation.

    The later Buddhist tradition saw Candrakīrti’s criticism of Bhāvikeka as the decisive point where the Madhyamaka tradition broke up into two distinct sub-schools: the Prāsaṅgika-Madhyamaka, who follow Candrakīrti and his defence of Buddhapālita against Bhāviveka, restrict themselves to prasaṅga arguments and do not endorse any of the consequences that Bhāviveka’s style of exposition brings with it; and the Svātantrikas, who follow Bhāviveka, and accept theses, epistemic instruments, and svabhāva at the conventional level, defending the utility of the philosophical analysis of the conventional. Understood in this way, the split into two sub-schools is certainly a doxographic fiction. Even though Bhāviveka was highly critical of Buddhapālita, his commentator Avalokitavrata does not appear to draw a distinction between the understanding of the two truths the two authors held.141 If Bhāviveka had perceived a great rift between his interpretation of Nāgārjuna and that of Buddhapālita, it is certainly peculiar that he mentions them all in the same breath. More straightforwardly, even the terms Prāsaṅgika and Svātantrika were not used by Indian Mādhyamikas to describe their own positions, but are retranslations from terms later Tibetan doxographers used in order to provide a systematic description of the different views Madhyamaka authors held. Apart from matters of terminology, it is also highly doubtful that Candrakīrti would have considered himself to be defending one sub-school of Madhyamaka, while his opponent Bhāviveka defended another one. Even though Candrakīrti does not refer to Bhāviveka by name, it is clear that he refers to him when saying:

    Even though a logician may take the side of the Madhyamaka school out of a desire to parade the extent of his own dialectical skill, it is evident that the presentation of svatantra reasoning becomes, for him, an enormous reservoir where faults pile up one after another.142

    (p.137) For Candrakīrti, Bhāviveka is not a Mādhyamika at all, but a logician (tārkika) who only takes the side of Madhyamaka in order to show off his argumentative abilities. From Candrakīrti’s perspective we do not have two possible interpretation of Madhyamaka, including one that countenances syllogisms and conventionally real intrinsic natures, but only one. According to him, Bhāviveka’s system is not a form of Madhyamaka, but is simply wrong.

    It is difficult to determine how Candrakīrti rose from a relatively obscure Indian philosopher to becoming a highly influential thinker and chief expositor of Madhyamaka. One intriguing suggestion143 is that Candrakīrti’s sudden prominence is intimately connected with the growing popularity of tantric scriptures. We find the triad of Madhyamaka authors Nāgārjuna, his direct disciple Āryadeva, and Candrakīrti repeated later in Indian Buddhism as a triad of tantric authors bearing the same names, living somewhere between 850 and 1000. They are key figures in the so-called Noble Lineage (‘phags lugs) of transmission of a central tantric work, the Guhyasamājatantra. The relationship between these two triads is complex. Traditional accounts believe these to be the same authors, while modern Buddhologists often consider such identification as either the result of fraud (later tantric authors claiming their works were penned by the Madhyamaka luminaries) or confusion (even though the later authors did not claim to be identical with the earlier ones, the tradition has mixed them up because they share the same name). Neither of these claims is particularly helpful in trying to understand what is going on between these two triads, considering the traditional accounts as rather more gullible than they really are. Traditional Buddhist historians like Tāranātha are very explicit in asserting that the tantric works of these authors were not spread when Nāgārjuna and Āryadeva ‘were actually residing in this world’,144 and that some of their works were not even composed then.145 Yet they are also, as Tāranātha stresses, ‘uncontroversially composed by the father [Nāgārjuna] and son [Āryadeva]’.146 To reconcile these claims we have to make assumptions that contemporary Buddhist studies are reluctant to accept, such as extraordinarily long lifespans of the Madhyamaka masters, or assume that the tantric works were written in the second century but then hidden and only circulated towards the end of the first millennium, or that they were composed or taught by the Madhyamaka masters at that later time in some form other than their earthly body.

    (p.138) Our knowledge of the connection between Madhyamaka as a philosophical school and tantra is very limited, and much more work is necessary to obtain a clear picture of the relation between the two systems of thought.147 There are, however, some noteworthy facts we can point out. One is that certain conceptual distinctions in tantra were explained according to distinctions in Madhyamaka, sometimes even mirroring the phrasing of Madhyamaka texts.148 Others are specific claims made in Madhyamaka works (especially by Candrakīrti) that seem to cohere well with specific tantric claims, such as the denial of established epistemic instruments, or a conception of ultimate reality that is not within the purview of human cognitive activities.149 There are also examples of tantric writers critical of Yogācāra background assumptions,150 a position that again resonates with Candrakīrti’s criticism of the Yogācāra school. Vose151 suggests that the rise in popularity of Candrakīrti’s philosophy is a result of an increased popularity of the Noble Lineage. This would mean that the increased interest in the Guhyasamājatantra and its commentaries towards the end of the first millennium led to an increased interest in the other, non-tantric works of Candrakīrti, since the Buddhist tradition draws no boundaries between the Madhyamaka Candrakīrti and the tantric Candrakīrti. This may be in part a backwards reflection of popularity, where the later works of an author lead to an increased appreciation of the earlier ones, and in part a result of the fact that there is some systematic affinity between some of the positions the Madhyamaka works defend and interpretations of the tantric texts the Noble Lineage set out to propagate. This suggestion is somewhat speculative and needs to be investigated in much more detail. Yet if further research supports this idea, it would supply us with a good example of the complexity of interactions of texts, authors, schools, and styles of thinking that characterize Indian philosophy, and of the necessity to keep these various interacting factors in mind in order to achieve a nuanced understanding of the way Buddhist philosophy developed in ancient India.

    With the works of Candrakīrti our account has reached the middle of the seventh century. Buddhist philosophy (and with it Madhyamaka philosophy) had about another 500 years of activity to look forward to before the decline of Indian Buddhist scholastic tradition at the beginning of the thirteenth century. What characterized this subsequent phase of Madhyamaka thought?

    (p.139) 6. The Great Synthesizers: Śāntarakṣita and Kamalaśīla

    Two key thinkers in the further development of Madhyamaka were Śāntarakṣita (725–88) and his disciple Kamalaśīla (c.740–95). Their importance has a systematic and a historical dimension. They produced an interesting synthesis of Madhyamaka and Yogācāra, and were two of the Indian teachers playing a crucial role in the transmission of Indian Buddhism (and of Madhyamaka thought more specifically) to Tibet in the eighth century. Thanks to this transmission, even after its decline in the country of its origin Buddhist philosophy of clearly Indian appearance would continue to develop in Tibet up to the present day.

    Śāntarakṣita’s philosophical importance is most clearly demonstrated by two of his works, the Madhyamakālaṃkāra and the Tattvasaṃgraha. The first is a major source for an attempt to bring together the distinct philosophical schools of Mahāyana thought, Madhyamaka and Yogācāra. We will have more to say on this project in the following chapter.152 The second is a long work of over 3,000 verses, which is preserved in its original Sanskrit together with Kamalaśīla’s commentary. The Tattvasaṃgraha is of particular interest as a doxographical work. It discusses a wide range of philosophical concepts, such as primordial matter (prakṛti), the creator god (īśvara), words and their referents (śabdārtha), perception (pratyakṣa), inference (anumāna), the existence of past, present, and future (traikālya), and the authoritativeness of scripture (śruti), and gives specific attention to the views of different philosophical schools on these concepts. There are, for example, detailed discussions of the account of the self (ātman) from the perspective of Sāṃkhya, Nyāya, Vaiśeṣika, Mīmāmsā, Advaita Vedānta, and Jainism. The Tattvasaṃgraha is a polemical work to the extent that it endeavours to show the mistakes inherent in all these non-Buddhist views (and some Buddhist ones, such as those of the Vātsīputrīyas) in order to demonstrate the truth of Śāntarakṣita’s Buddhist position. Yet independent of its success in this respect, it offers us a fascinating inside view of the state of philosophical debate in eighth-century India, a time that may be considered (both in terms of the variety of theoretical options explored, and in terms of their depth of conceptual penetration) as the peak of its development. Śāntarakṣita’s encyclopedic work demonstrates the extent to which Buddhist thought during this period was not an intellectually insulated enterprise, but interacted argumentatively with all the main philosophical currents of the time.

    (p.140) Śāntarakṣita was also a philosopher with considerable historical significance. Khri srong lde btsan, the second of the ‘three great dharma kings’ (chos rgyal), invited him to Tibet, where he became the first abott (upādhyāya) of the newly founded monastery of bSam yas, the first Tibetan monastic centre from which the establishment of Tibetan Buddhism began. The details of the transmission of Indian Buddhism from Nālandā monastery to the remote ranges of the Himalayan plateau lie beyond the scope of this history, but it is important just to recall at this point to what extent our knowledge of Indian Buddhist philosophy is indebted to the wholesale adoption of Indian Buddhist intellectual culture by the Tibetans, thereby preserving an immense amount of philosophical works long after their disappearance in their country of origin.

    Śāntarakṣita’s disciple Kamalaśīla played an equally important historical role, primarily by determining the direction the development of Buddhism took in Tibet. He is traditionally believed to have played a decisive role at the council of bSam yas, which took place around 797 CE. There Kamalaśīla is supposed to have defended the Indian model of a gradual approach to enlightenment, which saw enlightenment as the culmination of a process of purification based on ethical behaviour (śīla), meditation (samādhi), and insight (prajñā), against that of the Chinese model of sudden enlightenment defended by his opponent Heshang Moheyan (a Chinese rendering of ‘Mahāyāna’). Heshang Moheyan’s account saw the mind as intrinsically pure, and so all that was required for achieving enlightenment was a direct, non-conceptual insight into its nature. According to the Tibetan sources, the king judged Kamalaśīla to be the winner of the debate and declared his exposition of Madhyamaka as the official philosophical approach to be followed. Kamalaśīla is said to have been assassinated soon afterwards by jealous members of the defeated Chinese party.

    Most of the claims made about the council of bSam yas are contestable and remain contested: whether there really was a face-to-face debate, or merely a succession of texts written in response to the opponents’ views; whether there was a clear outcome of the debate; the precise nature of Heshang Moheyan’s views;153 and even the very historicity of the Chinese monk himself.

    Some facts, however, are uncontroversial. First, after this debate Tibet would import all of its Buddhism from India (which, unlike China, was not a military rival at the time), and Chinese Buddhism would not exert any noticeable influence on the further development of Buddhism in Tibet. Second, there is no doubt about Kamalaśīla’s view of enlightenment as a result obtained after following a stepwise procedure of training. One of his most important works, a text in three parts called Bhāvanākrama (‘Stages of Meditation’) describes how (p.141) the progression of the path to Buddhahood proceeds in stages that build on one another, such as the kinds of insight resulting from a succession of study, reflection, and meditation (śrūtamayī-prajñā, cintamayī-prajñā, and bhāvanāmayī-prajñā), different successive forms of meditative concentration (samādhi), calm abiding (śamatha) and insight (vipaśyana) meditation, a sequence of meditative objects (ālambanavastu), and so on.

    The parts of the Bhāvanākrama partly overlap and repeat discussions of topics. This suggests that they may have not been supposed to form part of one single treatise, but that they were related writings dealing with a common topic composed on separate occasions. It is not implausible to suggest that they might have formed part of the debate at the council of bSam yas, which may have been conducted by the exchange of texts, or may have at least involved such an exchange of texts. In the third part of the Bhāvanākrama Kamalaśīla presents an explicit refutation of the view of:

    those who think that beings, under the influence of virtuous and non-virtuous acts, born from mental conceptions, wander in cyclic existence after enjoying heaven etc. as the result of their actions, but those who do not think of anything or do not do anything hope to be liberated from cyclic existence without needing to think anything or do any virtuous actions, believing that conduct such as giving has been taught only for ignorant fools. In this way the entire Mahāyāna becomes negated.154

    Absence of mental activity (amanasikāra) is, according to Kamalaśīla, not sufficient for achieving liberation. To escape saṃsāra it is not enough to put oneself into a quietistic state in which conceptual thought no longer arises, but both skilful means (upāya) (crucially involving the practice of moral perfections such as generosity) and insight (prajñā) need to be cultivated in order to achieve enlightenment.

    Whether the approach of Heshang Moheyan was simply to ‘switch off your mind’, as the above passage suggests, may reasonably be doubted.155 Even though in later Tibetan scholastic literature the antinomian view of Heshang became a prime exemplar of a position beyond the pale, it is important to note that some of Heshang’s position were not suggested for the first time in the history of Buddhist thought, and that as Kamalaśīla could quote numerous sūtra passages in his refutation, so could Heshang in support of his own position. Once again, a more nuanced view of the matter than simply asking ‘who has got it right’ can be achieved by asking ourselves what the different terms and concepts (p.142) already present in the Buddhist teachings were that could be used as a basis to develop such different views as the positions of Kamalaśīla and Heshang.

    We do not have the space here to describe further details of the last phase of the development of Madhyamaka in India, or of its relation to the development of tantrism for which it, together with Yogācāra, provided the conceptual basis.156 Let me mention two related points, however. First, the Yogācāra-Madhyamka synthesis developed by Śāntarakṣita and Kamalaśīla does not constitute the conceptual endpoint of Indian Madhyamaka. Later, the tide began to turn again in favour of ‘pure Madhyamaka’, and when Dīpaṃkaraśrījñāna (982–1054), better known as Atiśa, added a list giving the lineage of Madhyamaka teachers in the auto-commentary to his best-known work, the Bodhipathapradīpa, neither Śāntarakṣita, nor Kamalaśīla, nor any other defenders of the synthetic approach are included.157 And despite the initial introduction of their system into Tibet, the approach to Madhyamaka taken there was not the Yogācāra-Madhyamaka synthesis of Śāntarakṣita and Kamalaśīla, but from the twelfth century onwards the interpretation of Madhyamaka proposed by Candrakīrti, which is very critical of the Yogācāra approach, became the dominant reading.

    The second point to note is that by the beginning of the second millennium Madhyamaka is no longer an exclusively Indian enterprise. One way of making this point is by considering that at this time important Madhyamaka works were composed at places quite far away from the Indian subcontinent. One of Atiśa’s teachers, Dharmakīrti (not to be confused with the Dharmakīrti of the logico-epistemological school discussed in Chapter 4), composed his major work, a commentary on the Abhisamayālaṃkāra, in Suvarṇadvīpa (modern-day Sumatra or Java), and Jayānanda, a scholar from Kashmir active between 1050 and 1100, authored the only Indian commentary still extant on Candrakīrti’s Madhyamakāvatāra close to Wutai shan, in China’s Shanxi province.158 In the nine centuries since its beginning in India, Madhyamaka thought developed into a philosophical school that was studied and developed in Central, East, and South-East Asia.

    7. Madhyamaka and Nyāya

    We will have an opportunity to look in greater detail at the relation between Madhyamaka and the other schools of Mahāyāna thought in the following (p.143) chapters;159 I would like to conclude this chapter by considering the relationship between Madhyamaka and the non-Buddhist school of Nyāya. The fundamental text of the Nyāya system are the Nyāyasūtras, said to have been compiled by Akṣapāda Gautama in the second century CE. Nyāya and Madhyamaka thus seem to have appeared in close historical proximity, and the two schools have debated with each other from the very beginning of their existence. Two of Nāgārjuna’s shorter works discuss Nyāya arguments at great length. In the Vigrahavyāvartanī a Naiyāyika, together with an Ābhidharmika, are the main interlocutors;160 the Vaidalyaprakaraṇa is explicitly formulated as a refutation of the sixteen Nyāya categories.161 An extensive commentary on the Nyāyasūtras, the Nyāyasūtrabhāṣya, was composed in the fourth century by Vātsyāyana and later became the target of Diṅnāga’s attacks; in the seventh century Uddyotakara came to defend Nyāya against Buddhist criticism in another elaborate commentary, the Nyāyavārttika.

    We are here most interested in the early Buddhist interaction with Nyāya. When Nāgārjuna criticized the Nyāya system he interacted with a very early form of it, a system which cannot be uncritically equated with the one that has come down to us in the Nyāyasūtras as we know it today and is expounded in its main commentaries.162 It is likely that various of the objections the Naiyāyika’s opponent raises in these texts, and which are then subsequently provided with a Nyāya reponse, have their source in Madhyamaka.163 The two systems shaped one another, and their texts exhibit traces of their mutual influence.

    Nāgārjuna engages mostly with the epistemological and logical parts of the Nyāya system. His criticism of these Nyāya positions is particularly interesting, since these are not just topics on which Nyāya happens to have views that differ from the Buddhist ones, but because they concern subject matters of sufficient generality to be of relevance to all philosophical discussion, Buddhist or not. It is important to be aware that Nyāya was not just one philosophical system among many, one of the six darśanas of classical Indian philosophy, but in its later development, particularly in a form that came to be known as Navya-Nyāya (‘new Nyāya’)164 extended its influence far into other fields, beyond what we might identify as the philosophical claims of the Nyāya(-Vaiśeṣika) system. Navya-Nyāya created a technical philosophical language that was not (p.144) just employed in the context of the discussion of the Nyāya itself, but also in the discussion and development of other philosophical schools, such as Advaita Vedānta, Sāṃkhya, and Mīmāṃsā, as well as in non-philosophical contexts such as grammar, poetics, and law.165

    Of course Navya-Nyāya is removed from Nāgārjuna by a distance of nearly a thousand years. Nevertheless, his engagement with the early Nyāya system in the Vigrahavyāvartanī and the Vaidalyaprakaraṇa shows that the dispute is not simply about which of two philosophical systems should be regarded as superior, but also about the right methodological framework in which to pursue philosophy.

    Yet if Nāgārjuna criticizes not only some of the Nyāya positions, but also the logical and epistemological techniques it recommends for use in philosophical debates, this raises two immediate questions. First, why do the Mādhyamikas not consider it possible to adopt the fairly sophisticated logical and epistemological frameworks Nyāya describes, even if they disagree with the content of some of their assertions? And second, what do they propose to put in their place?

    One clear example where some of the Nyāya logical machinery gets in the way of Madhyamaka arguments is provided by their understanding of negation. Nāgārjuna has the Nyāya opponent object to the Mādhyamika that: ‘to the extent to which the negation “there is no pot in the house” is precisely a negation of an existent, your negation is a negation of an existing substance.’166 The issue here is that, according to the Nyāya understanding, absences can only ever be local absences. If we negate the existence of the pot in the house, we are committed to the existence of pots elsewhere (in the garden, say, or in the potter’s workshop), for if the negated thing was not anywhere, how would we even know what we talk about when we negate it? How could we have ever had any epistemic contact with it?

    Yet this way of viewing negations leads to problems if we apply it to the Madhyamaka understanding of emptiness, which is simply the negation of intrincally real entities (svabhāva) anywhere. The Madhyamaka could not even state his theory in a framework in which the assertion of the emptiness of a chariot (that is, the assertion of the absence of svabhāva in it) entailed the presence of svabhāva elsewhere.

    A related worry arises when trying to express Madhyamaka inferences in the Nyāya framework of the five-membered syllogism.167 Such a syllogism needs to incorporate two types of examples, a concordant example (sādharmya) and a (p.145) discordant (vaidharmya) one. If we consider the stock example of such a syllogism, ‘there is fire on the hill because there is smoke there’ (as whenever there is smoke, there is fire), a concordant example is ‘as in the kitchen’, while a discordant one is ‘as in a lake’. The concordant example is a case where the property to be established (fire) and the mark helping in establishing it (smoke) are co-instantiated in one object, while the discordant example is an object that instantiates neither.168 This seemingly unproblematic demand for concordant and discordant examples faces difficulties if we try to formulate a common Madhyamaka inference such as ‘all things are empty because they are produced dependent on causes and conditions’. As emptiness is considered by the Mādhyamika to be a universal property of all things, any object could be introduced as a concordant example, though he would not be able to provide a discordant example, as there are no things that are not empty.169

    This issue is of considerable importance for Mādhyamikas (and for Buddhists more generally), since the reference to entities that seem to exist (composite wholes, persons, entities existing with svabhāva) but in fact do not is an integral part of the Buddhist theory of the world. Composite wholes and so forth are conceptual imputations on other entities (some of which are ultimately real for the Ābhidharmika but not the Mādhyamika) that have no more reality than horns of rabbits or sons of barren women: they are mere words (or mere concepts), but without anything behind them that they refer to. Yet the Naiyāyika finds it very difficult to find a place for such objects in his semantics, epistemology, or ontology.170

    It is therefore apparent that the Naiyāyika’s logical machinery171 cannot simply be used as a philosophical framework for formulating the Madhyamaka theory. What, however, should an alternative framework be? If the Mādhyamika wants to compete with the schools of classical Indian philosophy in debate, he cannot simply reject all of the logical and epistemological standards according to which such a debate is conducted. One cannot take part in playing the game without accepting the rules.

    (p.146) This, the Mādhyamika would agree, is true, but there is also no need to accept them as something more than rules. Rather than rejecting the opponent’s logical and epistemological standards altogether, the Mādhyamika wants to accept them in a form that is compatible with emptiness. One way of doing so is exemplified in the prasaṅga methodology. If the Mādhyamika does not endeavour to establish any position, but only tries to reduce the opponent’s svabhāva-involving position to absurdity, all he needs to do is demonstrate this according to the logical standards the opponent takes to be required for the demonstration of absurdity; there is no need for the Mādhyamika himself to adopt these standards. Similary, the Mādhyamikas accept epistemological standards from their opponents, as long as they are sufficiently modified to exclude appeal to svabhāva.172 In this way, dialectical exchange can take place, but in a way that does not already presuppose metaphysical assumptions that are in fact being debated.


    (1) For a survey of recent scholarship on the matter see Drewes 2010, 55–65, 66–74.

    (2) This position is particularly associated with Przyluski 1934, Lamotte 1954, and Hirakawa 1990.

    (3) Hirakawa 1963.

    (4) Silk: 2002: 355–405.

    (5) Schopen 1997: 31–2.

    (6) Sasaki 1999: 191–3; Schopen 2005.

    (7) Drewes 2007: 101–43.

    (8) The way the status of these texts as buddhavacana is assured differs. Some of these are said to have been taught by the Buddha during his earthly life, but given to a group of bodhisattvas for safekeeping, who, after a stay in some divine realm, brought the texts back to earth (see Harrison 1990). Sometimes the sūtra will present itself has having been taught in such a divine realm in the first instance (see Powers 2004: 106). Sometimes Buddhas will appear to the practitioner during meditative absorption; when emerging from meditation he will then propagate and expound those teachings (Harrison 1978: 43, 52–4).

    (9) Schopen 2005.

    (10) From a relatively early stage in the development of Buddhism a distinction is drawn between the enlightenment of the Buddha and that of his disciples, the arhats. A Buddha is described as having specific powers that an arhat lacks, such as omniscience (Weber 1994; Jaini 1992); in addition, the Mahāyāna holds that the achievement of the arhat falls short of that of a Buddha insofar as the former has only overcome the afflictive obstructions (kleśāvaraṇa) but not the more subtle cognitive obstructions (jñeyāvaraṇa) connected with the fundamental misapprehension of the nature of reality. The Mahāyāna did not reject the ideal of the arhat but presented itself as a swift path to a loftier goal, Buddhahood, though this path is one that could also be travelled by those pursuing the more limited goal of an arhat (see Harrison 1987).

    (11) Williams 2009: 20.

    (12) Snellgrove (1987: 90) believes there to be no systematic connection between the theory of emptiness, the Mahāyāna bodhisattva ideal and the Mahāyāna emphasis on compassion: ‘However, the combination of these two teachings, the Bodhisattva ideal and the emptiness of all concepts, has probably come about in these texts quite fortuitously without any immediate awareness of the effect that so extreme a philosophical view might have upon what is probably the highest of moral aspiration to be found anywhere in this imperfect world.’ Bronkhorst (2009: 118) too claims that the main conceptual innovations behind Madhyamaka and Yogācāra ‘had nothing to do with its [i.e. Mahāyāna’s] main aspirations’.

    (13) Śrāvakas and pratyekabuddhas are distinguished by the way they reach the goal of arhatship (the śrāvakas by relying on a teacher, the pratyekabuddhas without doing so in their final lifetime, instead contemplating the principle of dependent origination). Morevover, pratyekabuddhas do not teach other beings about their attainments, whence their name ‘individually enlightened ones’.

    (14) A very similar point has already been made by Bronkhorst (2009: 122–3).

    (15) See e.g. p. 46 above.

    (16) It is certainly the case that the bodhisattva ideal, the development of compassion, and so on are not the first things that come to mind when considering what the central new ideas of Madhyamaka and Yogācāra are. The point suggested above, however, is that the illusionistic view of the world, like the ethical views revolving around the bodhisattva ideal, can be understood as resulting from the enlarged vision of what the Buddha is, while also playing a crucial role in the philosophical visions developed by Madhyamaka and Yogācāra.

    (17) Drewes 2010: 61–2.

    (18) lha’i mig thob pas de bzhin gshegs pa mthong ba yang ma yin | lha’i rna ba’i khams thob pas dam pa’i chos nyan pa yang ma yin | rdzu ‘phrul gyi stobs thob pas ‘jig rten gyi khams der skad cig tu ‘gro ba yang ma yin gyi | bzang skyong | byang chub sems dpa’ de ‘jig rten gyi khams ‘di nyid na gnas bzhin du | bcom ldan ‘das de bzhin gshegs pa tshe dpag med de mthong zhing bdag nyid ‘jig rten kyi khams de na ‘dug ba snyam du shes la | chos kyang nyan to, Harrison 1978: 43. Note that even though this passage refers the Buddha Amitāyus, a form of the ‘celestial’ Buddha Amitābha, other passages of the same sūtra make it clear that any Buddha can be the object of this type of meditative exercise.

    (19) de ‘di snyam du | de bzhin gshegs pa ‘di ga zhig nas byon tam || bdag ga zhing tu song tam | snyam pa las des de bzhin gshegs pa de gang nas kyang ma byon par rab tu shes so || bdag gi lus kyang gang du yang ma song bar rab tu ‘du shes nas | de ‘di snyam du | khams gsum pa ‘di dag ni sems tsam mo || de ci’i phyir zhe na | ‘di ltar bdag ji lta ji ltar nram par rtog pa de lta de ltar snang ngo, Harrison 1978: 46.

    (20) Ruegg 1981: 2.

    (21) Walser 2005: 66, 71–3.

    (22) Though it is not contained in the earliest version of this sūtra: see Walser 2005, n. 29, p. 293.

    (23) dakṣiṇāpathavedalyāṃ bhikṣuḥ śrīmān mahāyaśāḥ/

    nāgāhvayaḥ sa nāmnā tu sadasatpakṣadārakaḥ//

    prakāśya loke madyānaṃ mahāyānamanuttaram/

    āsādya bhūmiṃ muditāṃ yāsyate ‘sau sukhāvatīm// 10:165–6, Vaidya 1963: 118, Suzuki 1932: 239–40.

    (24) See also MacDonald 2015: 11–12, nn. 34–5.

    (25) Yün-Hua 1970: 140–1.

    (26) Walleser 1990: 28.

    (27) Walleser 1990: 9.

    (28) Dowman 1985: 115.

    (29) A group of tribes from eastern Assam is also collectively referred to as nāgas.

    (30) Walleser 1990: 30. As was Buddha Śākyamuni. The similarity of this ‘second Buddha’ to the historical one is frequently stressed in traditional accounts. According to Tāranātha, his body is adorned with the 32 auspicious signs that characterized the Buddha’s physical body (Lama Chimpa 1970: 110–11). Nāgārjuna is also one of the few figures in Tibetan iconography (together with Asaṅga, dGa’ rab rdo rje, and Guru shakya seng ge, a manifestation of Padmasambhava) who is depicted with the Buddha’s protuberance (uṣṇīṣa) on the top of his head.

    (31) MacDonald 2015: 2. 7, n. 13.

    (32) Tsonawa 1985: 4.

    (33) Walleser 1990: 10.

    (34) Despite being very critical of Abhidharma ideas, the Perfection of Wisdom texts share its fondness of lists (mātṛkā). For the understanding of later developments of the Perfection of Wisdom literature it is worth keeping in mind that mātṛkā both means ‘mother’ and can also, according to Monier-Williams, denote ‘an epithet of certain diagrams written in characters to which magical power is ascribed’ (Conze 1978: 5–6).

    (35) This division follows Conze 1978: 1–16.

    (36) Conze 1994. Even though there are no extant manuscripts of the Prajñāpāramitā texts dating back as far as this, a recently discovered manuscript of the Perfection of Wisdom in 8,000 verses written on birch-bark in the Gāndhārī language can be dated to about 47–147 CE (Falk and Karashima 2011–12; Karashima 2012–13). This manuscript itself appears to be a copy of an earlier text, lending additional plausibility to Conze’s assumption that the early Perfection of Wisdom texts pre-date the beginning of the Common Era.

    (37) According to traditional accounts, the Pāli canon was scripturalized during the reign of the Sri Lankan king Vaṭṭagāmaṇī between 32 and 35 BCE (Gómez 2002: 59).

    (38) This latter version was given to the nāgas for safekeeping, even though this is merely an abbreviated version when compared to versions kept—according to Bu ston—in other realms: a 10,000,000-verse version in the realm of the king of the gods, and 1,000,000,000-verse version in the realm of the king of the gandharvas (Conze 1978: 18, n. 1).

    (39) Conze 1955: 13.

    (40) At this point it is important to distinguish two claims. One is uncontroversial, namely that over the course of Indian history many authors have answered to the name ‘Nāgārjuna’. This does not mean that they were all operating under the pseudonym of the Madhyamaka master, or that their works claimed to be authored by him (see Walser 2005: 69). The other, more controversial claim states that the different facets we find attributed to the Madhyamaka author in traditional biographies (the philosophical, alchemical, medical, and tantric aspects) have to be understood as applying to different persons, not just to one.

    (41) Note, however, the interesting connection with the nāgas: both the Buddhist universities Nālandā and Takṣaśīla are supposedly named after nāgas, the former after Nanda, the second after Takṣaka. Walker 1968: 2. 107.

    (42) Walser 2005: 78.

    (43) Lindtner 1982: 11, n. 12.

    (44) Walser 2005: 69, 75–9, Eliade 1969: 415–16.

    (45) Winternitz 1968: 3. 547, 552–3.

    (46) The higher bodhisattva grounds are characterized by an increase of these numbers. There is nothing in the basis of the realization that differentiates the grounds from each other. See MacDonald 2015: 2. 356.

    (47) Yün-hua 1970: 151–2.

    (48) McHugh 2012: 267, n. 6, Laufer 1896. If we are to believe Yijing, Nāgārjuna also provided advice on dental hygiene. According to him, the Indian monks he encountered used to chew ‘the rough root of the Northern Burrweed…It hardens the teeth, scents the mouth, helps to digest food or relieves heart-burning…This is the means of securing a long life adopted by Bodhisattva Nāgārjuna’ (Yün-hua 1970: 28). See also Takakusu 1896: 34–5.

    (49) In addition, a set of shorter, but nevertheless interesting works are frequently ascribed to Nāgārjuna. See Ruegg 1981a: 26–30 for a concise discussion.

    (50) Lindtner 1982: 94–9.

    (51) Walser 2005: 63–5. See also Ruegg 1981a: 26–7, n. 59.

    (52) ‘Please construct from all precious substances images of the Buddha with fine proportions, well designed and sitting on lotuses, adorned with all precious substances’, rin chen kun las bgyis pa yi | sang rgyas sku gzugs dbyibs mdzes shing | legs par bris pa padma la | bzugs pa dag kyang bgyid do stsol, Walser 2005: 80, Hahn 1982a: 78, Hopkins 1998: 124–5.

    (53) Walser 2005: 61, 86.

    (54) See Ruegg 2010: 176, n. 32. For some discussion of the later Tibetan debate around this question see Brunnhölzl 2007: 43–55.

    (55) Conze 1978: 6.

    (56) Conze 1994: 163.

    (57) Conze 1955: 33.

    (58) In tracing the ancestry of the illusionism of the Prajñāpāramitā texts to the pre-Mahāyāna level it is interesting to note that in the Saṃyuttanikāya (22: 95(3), Bhikkhu Bodhi 2000: 951–2) the Buddha elucidated the five skandhas by five illusionistic similes, comparing matter to a lump of foam, feeling to a water bubble, perception to a mirage, volitional formations to the trunk of a banana tree, and consciousness to a magical illusion. See also Bhikkhu Ñāṇananda 1974: 5–7. Verses 12–13 of the Bodhicittavivaraṇa ascribed to Nāgārjuna cite this comparison, see Lindtner 1982: 188–9, 259–60.

    (59) Conze 1994: 98–9.

    (60) Conze 1994: 90.

    (61) Schmithausen 1973: 181.

    (62) Poussin 1937: 191; Frauwallner 1956: 353–4; Staal 1975: 88.

    (63) For an interesting comparison between the states created by the meditative practitioners and by schizophrenics see Beyer 1988: 84: ‘The yogin consciously bases his magical power upon his understanding and hence upon his control of himself and his reality; the schizophrenic’s power is based not on control but on chaos.’

    (64) Āneñjasappāyasutta, Majjhimanikāya 106: annicā bhikkhave kāmā tucchā musā mosadhammā māyākatam etaṃ bhikkhave bālalāpanam, Bhikkhu Bodhi 2001: 869.

    (65) Phenasutta, Saṃyuttanikāya 22:95, Bhikkhu Bodhi 2000: 951–5.

    (66) Conze 1994: 136.

    (67) Conze 1994: 87.

    (68) Conze 1994: 191.

    (69) Conze: 1978: 7–8.

    (70) On this point see Schmithausen 1977: 45.

    (71) In Mūlamadhyamakakārikā 15:7. Saṃyutta Nikāya 12.15 (Bikkhu Bodhi 2000: 544).

    (72) Warder 1973: 79–81.

    (73) See Warder 1973, Kalupahana 1991: 5–8. Ruegg 1981a: 6–7, Lindtner 1982: 21, n. 67, and Bronkhorst 2009: 136 remain unconvinced.

    (74) Schayer 1931: ix rightly criticizes this ‘protestant’ conception of Buddhism: ‘Daß die Pāli-Philologie so auffallend wenig zur Aufhellung der philosophischen Grundlagen des Hīnayāna beigetragen hat, damit hat es seine eigene Bewandnis. Die falsche Suggestion, daß in der Geschichte einer Religion nur das Ursprüngliche echt, alles Jüngere dagegen mehr oder weniger eine “Entartung” sei, hat von Anfang an den Gang der Studien, ihre Richtung und ihre Methode beeinflußt.’

    (75) Warder 1973: 85, 87.

    (76) The concept of a Mahāyāna group identity that could have formed the basis of such a self-identification took a long time to develop. The earliest use of the term ‘Mahāyāna’ is only found in Indian inscriptions dating from several centuries after the appearance of the first Mahāyāna texts. Williams (2009: 28) notes that for ‘a monk in the first or second century CE the Mahāyāna as a visible institution was scarcely evident’.

    (77) Conze 1975: 595, n. 11.

    (78) Walser 2005: 170–83.

    (79) Walser 2005: ch. 5.

    (80) yathaiva vaiyākaraṇo mātṛkām api pāṭhayet |

    buddho ’vadat tathā dharmaṃ vineyānāṃ yathākṣamaṃ ||

    keṣāṃcid avadad dharmaṃ pāpebhyo vinivṛttaye |

    keṣāṃcit puṇyasiddhyarthaṃ keṣāṃcid dvayaniḥśritam ||

    dvayāniśritam ekeṣāṃ gāmbhīraṃ bhīrubhīṣaṇaṃ |

    śūnyatākaruṇāgarbham ekeṣāṃ bodhisādhanam||

    iti sadbhir mahāyāne kartavyaḥ pratighakṣayaḥ |

    prasādaś cādhikaḥ kāryaḥ samyaksaṃbodhisiddhaye ||

    Ratnāvalī 4: 94–7, Hahn 1982a: 128–31, Hopkins 1998: 147, verses 394–7.

    (81) Hopkins 1998: 149.

    (82) Walser 2005: 226–7.

    (83) Ratnāvalī 5: 7–12, Hahn 1982a: 134–7, Hopkins 1998: 150–1, verses 407–12.

    (84) bhāṣya on Abhidharmakośa 5.10 (Pradhan 1975: 284–5, Poussin and Pruden 1988–90: 3. 784–5).

    (85) On this see further Walser 2005: 225–63. He rightly points out that ‘Nāgārjuna’s arguments should be examined in terms of the alliances they forge instead of merely whom they attack’ (226).

    (86) Mūlamadhyamakakārikā 15:2.

    (87) See pp. 70–1 above.

    (88) It is occasionally suggested that cause and effect are sometimes simultaneous, e.g. when Jill’s going down on the seesaw causes Jack’s going up. However, such cases give rise to various complexities we do not have the space to go into in the present discussion.

    (89) See above, pp. 71–2.

    (90) See above, p. 71.

    (91) Mūlamadhyamakakārikā 13: 3a, 4b–5 bhāvānāṃ niḥsvabhāvatvam anyathābhāvadarśanāt | […]

    kasya syād anyathābhāvaḥ svabhāvo yadi vidyate ||

    tasyaiva nānyathābhāvo nāpy anyasyaiva yujyate |

    yuvā na jīryate yasmād yasmāj jīrṇo na jīryate ||

    (92) yathā nirmitakaṃ śāstā nirmimītārddhisaṃpadā | nirmito nirmimītānyaṃ sa ca nirmitakaḥ punaḥ || tathā nirmitakākāraḥ kartā yat karma tatkṛtam | tadyathā nirmitenānyo nirmito nirmitas tathā || kleśāḥ karmāṇi dehāś ca kartāraś ca phalāni ca | gandharvanagarākārā marīcisvapnasaṃnibhāḥ || Siderits and Katsura 2013: 191. The ‘city of the gandharvas’ is a popular Indian example of an unreal appearance: a city seen in the sky that is not really there.

    (93) For further discussion of this point see Westerhoff 2016a.

    (94) ato ya ekatyā durvijñeyān sūtrāntānmahāyānapratisaṃyuktāṃ gambhīrāṃ śūnyatāpratisaṃyuktānābhiprāyikārthanirūpitāṃ śrutvā yathābhūtaṃ bhāṣitasyārthamavijñāyāyoniśo vikalpyāyogavihitena tarkamātrakeṇa ivaṃ dṛṣṭayo bhavanty evaṃ vādinaḥ | prajñaptimātram eva sarvam etat tattvam | yaś ca ivaṃ paśyati sa samyak paśyatīti | teṣāṃ prajñaptyadhiṣṭhānasya vastumātrasyābhāvātsaiva prajñaptiḥ sarveṇa sarvaṃ na bhavati | kutaḥ punaḥ prajñaptimātraṃ tattvaṃ bhaviṣyatīti | tad anena paryāyeṇa tais tattvam api prajñaptir api tad ubhayam apy apavāditaṃ bhavati | prajñaptitattvāpavādāc ca pradhāno nāstiko veditavyaḥ, Wogihara 19306: 1. 46. See Willis 1979: 161, Engle 2016: 81–2.

    (95) 22:11 śūnyam iti na vaktavyam aśūnyam iti vā bhavet | ubhayaṃ nobhayaṃ ceti prajñaptyarthaṃ tu kathyate, Siderits and Katsura 2013: 247.

    (96) 25:24 sarvopalambhopaśamaḥ prapañcopaśamaḥ śivaḥ |

    na kva cit kasyacit kaścid dharmo buddhena deśitaḥ, Siderits and Katsura 2013: 304.

    (97) See Priest and Garfield 2002, Garfield and Priest 2009, Priest and Routley 1989. See also Tillemans (2009) for a discussion of dialetheism in view of the fact that ‘contradictions were anathema…for later Mādhyamikas’ (96).

    (98) Ruegg 1969: 384.

    (99) Whether these two characterizations of ultimate truth coincide is a complex question in the interpretation of Madhyamaka. See Siderits 2007: 200–4 for some discussion.

    (100) Tsong kha pa 2002: 3: 188, see also 215–23.

    (101) svabhāvānutpattiṃ saṃdhāya mahāmate mayā sarvadharmā anutpannā ity uktāḥ, quoted in Candrakīrti’s Prasannapadā, Poussin 1913: 504: 5–6.

    (102) For a modern Tibetan criticism of this ‘interpolation procedure’ see Lopez 2006: 58–60.

    (103) In addition there is also a partial Chinese version of a commentary ascribed to Asaṅga.

    (104) Huntington 1986: 17.

    (105) The ascription of the Akutobhayā to Nāgārjuna has been repeatedly questioned, mainly because it quotes a stanza from Āryadeva. It is not clear, however, how conclusive this point is. First, the text of the Akutobhayā appears to have been somewhat fluid in the way it has been transmitted, so we cannot rule out that this quotation is a later interpolation. In addition it is not unthinkable that Nāgārjuna may have cited a work of one of his disciples, especially as we have renditions of the text in which Āryadeva is not referred to by the term ācārya (‘master’), but by the more modest bhadanta (‘venerable’, a term used to address Buddhist monks).

    (106) Huntingdon 1986: 149.

    (107) rig pa ‘dzin pa’i lus nyid kyis […] skal ldan rnams la bstan ba yang yin srid de, Wedemeyer 2007: 20.

    (108) Tsonawa 1985: 14.

    (109) This is not necessarily to be understood as an argument for the greater philosophical accuracy of Buddhapālita’s interpretation over later ones. That a specific commentarial tradition is earlier does not necessarily mean that it more accurately expresses the author’s intent, or has a greater claim to systematic validity. But it is important to be aware of the histories of these different interpretative approaches in order to provide a nuanced picture of their later interaction.

    (110) We have not much biographical information on Bhāviveka, but one noteworthy fact about his afterlife is that the Tibetan tradition considered him to have been later reborn as the Panchen Lama (the lineage of the Panchen Lamas contains various Indian sages before the first Panchen Lama (1385–1438)). Bhāviveka is also sometimes referred to as ‘Bhāvaviveka’ or ‘Bhavya’. The choice ‘Bhāviveka’ seems to be supported by the majority of evidence currently available, though the matter has not been settled definitely. See Ames 2009.

    (111) na svato nāpi parato na dvābhyāṃ nāpy ahetutaḥ | utpannā jātu vidyante bhāvāḥ kvacana kecana, Siderits and Katsura 2013: 18.

    (112) It is likely that one of the positions Nāgārjuna had in mind here is that of the pre-existence of the effect in the cause (satkāryavāda) defended by Sāṃkhya. For further discussion of self-causation see Westerhoff 2009: 99–104.

    (113) Ames 2003: 46.

    (114) Ames 2003: 46–7.

    (115) This form of the syllogism was introduced by Diṅnāga (c.400–80) and constitutes a streamlined form of the five-membered syllogism familiar from Nyāya.

    (116) Ames 2003: 50.

    (117) Ames 2003: 51.

    (118) Candrakīrti later replies that Nāgārjuna, when writing his auto-commentary to the Vigrahavyāvartanī, did not spell out his arguments in syllogistic form, so why should we expect later commentators like Buddhapālita to do so? (Ruegg: 2002: 42–3). Bhāviveka might respond by pointing out that such syllogistic elaborations are specifically for the benefit of later students who can no longer understand the original meaning in its full complexity, so that there is no reason to suppose that the Master himself would have composed a commentary in this way.

    (119) yadi kācana pratijñā tatra syān na me tat eṣa me bhaved doṣaḥ | nāsti ca mama pratijñā tasmān naivāsti me doṣaḥ; Westerhoff 2010: 63, Williams-Wyant 2017.

    (120) Huntington 2003: 71–4.

    (121) For more on this idea see Westerhoff 2009: 17–18, 194–8; 2010: 63–5.

    (122) See Eckel 2008: 50, 210–11.

    (123) See Bouthillette 2017.

    (124) aparapratyayaṃ śāntaṃ prapañcair aprapañcitam | nirvikalpam anānārtham etat tattvasya lakṣaṇam, Siderits and Katsura 2013: 202.

    (125) skye bo las dang po dag yang dag par dbugs dbyung ba’i phyir, D 3853, dbu ma, tsha 190a2; Eckel 2008: 52.

    (126) Eckel 2008: 52.

    (127) kho bo cag gi phyogs la ni ngo bo nyid stong pa nyid yin te | chos rnams kyi ngo bo nyid de yin, Eckel 2008: 52–3, Ames 2003: 46.

    (128) See Ames 2003: 75. The construction of such hierarchies is very widespread in Indian philosophical texts. Already in the Chāndogya Upaniṣad we find the sage Prajāpati instructing Indra about the nature of the real self by guiding him through a series of gradually more sophisticated views (8: 7–15, Radhakrishnan 1969: 501–12, Olivelle 1996: 171–6). It is interesting to note that Prajāpati does not simply tell Indra that all the lower views are deficient, but lets him work out the limitations of each view for himself before introducing a more sophisticated one.

    (129) This is Jayānanda’s definition (Cabezón 2003: 310).

    (130) Once again contemporary Buddhist scholarship distinguishes various authors that were called Candrakīrti. Of particular importance is a commentator on the Guhyasamājatantra, sometimes referred to as Candrakīrti II or the ‘tantric Candrakīrti’.

    (131) Chimpa and Chattopadhyaya 1970: 199. In contradiction to contemporary accounts Candrakīrti is also sometimes described as having lived during the later part of Nāgārjuna’s life (Tsonawa 1985: 16.). This is of course less strange than it sounds if one accepts the extraordinary long lifespan traditionally attributed to Nāgārjuna.

    (132) Tsonawa 1985: 17–18.

    (133) 22, Yonezawa 2008: 218: 8–11; Westerhoff 2010: 27.

    (134) Two exceptions are his commentaries on Nāgārjuna’s Yuktiṣaṣṭikā and Śūnyatāsaptati, presumably because they were the only Indian commentaries on these texts. See Vose 2009: 19–20.

    (135) The closest example of this difficulty in the contemporary context may be the question of the kinds of proof that an intuitionist mathematician can appeal to. Because the intuitionist does not assume that all mathematical facts are eternally settled in some Platonic realm, he cannot, for example, prove a statement by use of a reductio. Such proofs presuppose that since either A or not A obtains, showing that A leads to a contradiction allows us to prove not A. But this presupposition is something that the intuitionist cannot accept. The introduction of some supposedly topic-neutral machinery in terms of allowed proof techniques has thus been shown to reintroduce substantial philosophical assumptions through the back door.

    (136) 6:30, lokaḥ pramāṇaṃ yadi tattvadarśī syāl loka evety aparaiḥ kim āryaiḥ | kim āryamārgeṇa bhavec ca kāryam mūdhaḥ pramāṇaṃ na hi nāma yuktaḥ, Li 2015.

    (137) Both Nāgārjuna and Candrakīrti do, however, set out to develop a conception of epistemic instruments and objects that does not appeal to intrinsic natures. They argue that even though we can profitably employ both these concepts, it is essential that we realize that they mutually depend on each other for their existence: there cannot be epistemic instruments without objects they cognize, nor can there be epistemic objects without instruments. For further discussion see Westerhoff 2010, Siderits 2011a.

    (138) Madhyamakāvatāra 6: 118, na vādalobhād vihito vicāras tattvam tu śāstre kathitaṃ vimuktyai, Li 2015, Huntington 2003: 77.

    (139) Madhyamakāvatāra 6:119, svadṛṣṭirāgo ‘pi hi kalpanaiva tathānyadṛṣṭāv api yaś ca roṣaḥ | vidhūya rāgaṃ pratighaṃ ca tasmād vicārayan kṣipram upaiti muktim, Li 2015.

    (140) See Tillemans 2003: 108.

    (141) Lopez 1987: 57–8.

    (142) Huntington 2003: 82.

    (143) Raised by Kevin Vose (2009: 27–36).

    (144) ‘phags pa yab sras ‘dzam bu gling du dngos su bzhugs pa’i dus, Wedemeyer 2007: 18–19.

    (145) mdzad […] ma yin te, Wedemeyer 2007: 19.

    (146) yab sras de rnams kyi mdzad par rtsod pa med la, Wedemeyer 2007: 20.

    (147) It is usually assumed that the more natural philosophical background for tantra is Yogācāra; we will discuss this point further in Chapter 3.

    (148) An example is the adaptation of Nāgārjuna’s verse on the two truths in 24:8 of the Mūlamadhyamakakārikā by the Guhyasamājottaratantra, noted by Isaacson (Wedemeyer 2007: 40, n. 83).

    (149) Vose 2009: 28–9.

    (150) Vose 2009: 33–4.

    (151) Vose 2009: 31.

    (152) See below pp. 205–12.

    (153) For a detailed account see van Schaik 2015.

    (154) yastu manyate cittavikalpasamutthāpitaśubhāśubhakarmavaśena sattvāḥ svargādikarma phalamanubhavantaḥ saṃsāre saṃsaranti | ye punarna kiñciccintayanti nāpi kiñcit karma kurvanti te parimucyante saṃsārāt | tasmānna kiñciccintayitavyam | nāpi dānādikuśalacaryā kartavyā | kevalaṃ mūrkhajanamadhikṛtya dānādikuśalacaryā nirdiṣṭeti | tena sakalamahāyānaṃ pratikṣiptaṃ bhavet, Namdol 1984: 232.

    (155) van Schaik 2015: 133.

    (156) See Ruegg 1981a: 104–8 for some remarks.

    (157) Sherburne 2000: 237–41. Śantarakṣita is cited only once in the text (235, see also 272, n. 24).

    (158) Ruegg 1981a: 109–10, 113–14. A significant portion of the Mahāprajñāpāramitopadeśa (see p. 94) might also be of Central Asian or Serindian origin (Ruegg 1981a: 32–3).

    (159) We consider the relationship with Yogācāra in section 5 of Chapter 3, and discuss the relationship of Madhyamaka with the schools of Diṅāga and Dharmakīrti in section 7 of Chapter 4.

    (160) Westerhoff 2010, Meuthrath 1999.

    (161) Westerhoff 2018.

    (162) It is generally assumed that the first and fifth chapters or adhyāyas constitute the oldest part of the text (Meuthrath 1996: x). Some form of the material covered in these chapters is most likely what Nāgārjuna was familiar with and directed his criticism against.

    (163) Oberhammer 19634: 68, 70; Bronkhorst 1985.

    (164) Udayana (c.1050) and Gaṅgeśa (c.1200) are generally regarded as the founders of the Navya-Nyāya tradition.

    (165) Bhattacharya 2001: 102; Ganeri 2008: 109–24.

    (166) 11: sata eva pratiṣedho nāsti ghaṭo geha ity ayaṃ yasmāt | dṛṣṭaḥ pratiṣedho yaṃ sataḥ svabhāvasya te tasmāt, see Westerhoff 2010: 109–10.

    (167) Compare Vaidalyaprakaraṇa 28–9, Westerhoff 2018.

    (168) The counterexample of the lake may have been picked because a lake might sometimes look as if there was smoke present in it (when there is mist rising from the lake), though when properly examined we realize that there is neither smoke nor fire.

    (169) Matilal 1970, 83–110.

    (170) Matilal (1970: 91) notes that the Naiyāyika ‘wants to exclude from logical discourses any sentence which will ascribe some property (positive or negative) to a fictitious entity. Vācaspati remarks that we can neither affirm nor deny anything of the fictitious entity, the rabbit’s horn.’ As a consequence, Nyāya ‘does not admit that a totally ficititious entity can be the “object” of any cognitive state, even of an error.…[C]orresponding to each fundamental element of thought or cognitive state there is a fundamental element of reality. The so-called fiction is always constructed out of real elements’ (95).

    (171) The same, Nāgārjuna sets out to argue, is true of the Nyāya epistemology.

    (172) I argue in Westerhoff 2018 that Nāgārjuna’s main aim in Vaidalyaprakaraṇa is to develop a desubstantialized account of the Nyāya categories, that is, an understanding of them that retains much of their methodological usefulness but dispenses with the idea that any of them exist with svabhāva.