Chapter 5. Candrakīrti and Prāsaṅgika Madhyamaka¶

The preceding chapter clarified Buddhapālita's disagreement with Bhāvaviveka and identified its unresolved tension. At which level does the analysis of ultimate truth itself belong? Candrakīrti, around the seventh century, is the dispute's final adjudicator, at least in later Tibetan classifications. Siding with Buddhapālita, he offers the most systematic defense of reductio thus far and builds a complete philosophical system for Prāsaṅgika Madhyamaka. This chapter examines his arguments in detail.

1. Candrakīrti and his works¶

Biographical information remains scarce. Tibetan sources say that he was born in Samanta in southern India and served as abbot of Nālandā. Stories of his riding a cow and supplying the monastery with dairy products are naturally not historical facts. Only a rough outline is credible. He was active in the early to middle seventh century and roughly contemporary with Dharmakīrti. He knew the active Yogācāra and Buddhist logical traditions and wrote extensively against Bhāvaviveka and Yogācāra thinkers.¹

Two works are most important:

The Madhyamakāvatāra, or Entry into the Middle Way, systematically develops Madhyamaka philosophy within the framework of the bodhisattva's ten grounds. The first five receive brief treatment. The sixth, the manifest ground, occupies by far the longest portion in its arguments on emptiness. From the seventh onward, the account concludes with nondiscriminating wisdom and liberation. It is Candrakīrti's representative philosophical work and later becomes a central Madhyamaka textbook in Tibetan Buddhism.

The Prasannapadā, or Clear Words, is a commentary on the Mūlamadhyamakakārikā, the only such commentary surviving complete in Sanskrit. Candrakīrti criticizes Bhāvaviveka's interpretations chapter by chapter, defends Buddhapālita, and gives his clearest explanation of why Madhyamaka cannot offer independent arguments.²

He also wrote commentaries on the Sixty Verses on Reasoning, Seventy Verses on Emptiness, and Four Hundred Verses (only in Tibetan translation), all explanations of works by Nāgārjuna and Āryadeva.

2. The full logic of reductio: why independent arguments fail¶

The preceding chapter, 5.4, identified Candrakīrti's central criticism. Independent argument requires its subject and terms to be established for both sides by valid cognition. Madhyamaka, however, holds that everything lacks intrinsic nature and recognizes nothing as established through particular characteristics, even conventionally. Bhāvaviveka's concession of conventional characteristics for argument implicitly retains some realism. The argument's full structure can now be examined.

Bhāvaviveka invokes Dignāga's logical requirements. A complete argument has a thesis to be proved, a reason, and an example. Both parties must accept the terms of the thesis and reason as already established through perception or inference, the shared valid means of cognition. His argument for the absence of intrinsic nature roughly takes dependent origination as its reason, a pot as its example, and the claim that phenomena ultimately lack intrinsic nature as its thesis. The difficulty is that if the subject, phenomena, must be established by valid cognition, some cognition must confirm it as a shared object of argument. Under Dignāga's epistemology, that cognition must rest on some particular characteristic of the object.

Dignāga and his successor Dharmakīrti, around the seventh century, form the main line of Indian Buddhist epistemology. Philosophically they belong to Yogācāra and will be introduced in Essay 6. A table of their central epistemological terms makes Candrakīrti's target clear:

This is Candrakīrti's crucial objection. To invoke Dignāga's standards of valid cognition for independent argument, Bhāvaviveka must concede some established status to the particular characteristics perceived, even if only at the linguistic level. Candrakīrti sees the concession as a remnant of realism.

His reply is that this imposes an epistemological structure on Madhyamaka. Dignāga's system, including the requirement that cognition rest on an object's particular characteristic, is itself under criticism. Invoking its rules within Madhyamaka concedes in advance the very premise that requires examination. That premise cannot supply a common basis for argument.²

One detail matters. Bhāvaviveka does not claim particular characteristics are real in the Sarvāstivāda sense. He speaks of characteristics established through linguistic convention (vyāvahārika-svalakṣaṇa), conventionally established rather than ultimately real. Candrakīrti finds this far from sufficient. Even with the linguistic qualifier, admitting a conventional particular characteristic says that something conventionally makes itself what it is. That self-establishment is precisely what emptiness analysis denies. The qualifier softens the wording but cannot remove the underlying realist presupposition.³

Candrakīrti's logic of reductio thus has two levels.

The first is methodological. Reductio (prasaṅga) borrows premises the opponent accepts and follows them to consequences the opponent cannot accept. Madhyamaka asserts no thesis in the process and needs no terms or particular characteristics accepted by both sides. The force comes from the failure of the opponent's premises. It does not come from Madhyamaka offering a better replacement.

The second is ontological. Madhyamaka can argue only through reductio because its position on emptiness thoroughly excludes anything established through particular characteristics. An independent argument would need a shared basis, and any such basis introduces a commitment that something establishes itself sufficiently to be shared. That commitment is incompatible with emptiness.

3. Establishment by designation: Candrakīrti's conventional truth¶

If reductio refutes without establishing anything, how do conventional things exist? Candrakīrti offers Prāsaṅgika Madhyamaka's version of conventional truth.

Conventional things exist through establishment by designation (prajñaptimātra), merely the assignment of conceptual labels, with no particular characteristics beneath them. A chariot's existence depends neither on a substance of chariotness nor on its presenting itself to the senses through particular characteristics. It depends on its hubs, body, and axle, and on people's customary use of the name chariot. Through these, it is established as something called a chariot.⁴

Candrakīrti systematizes this intuition in the sevenfold analysis (saptadhā yukti), again using a chariot. If it has intrinsic nature, its relation to its parts must be one of seven: the chariot is identical with its parts; different from them; dependent on them; that on which they depend; their possessor; their collection; or their shape. Each route is overturned. If identical, any part is the chariot, so a single wheel is a chariot. If different, removing all the parts should leave a separate chariot itself, but none can be found. If the parts depend on the chariot, it must exist independently before them, and again none can be found. The remaining routes are similarly untenable within a framework of intrinsic nature.

The analysis seeks exhaustive coverage, leaving no place to put a chariot with intrinsic nature. Its conclusion is that the chariot exists, but only through designation upon its parts within convention, without a chariot itself independent of those parts. Tsongkhapa later uses this extensively to explain attachment to a personal self. Substitute person for chariot and the five aggregates for parts. The same seven arguments leave no place for an intrinsically established I. Yet an I established through designation depends on the aggregates, and everyday speech continues normally.

Designation must also answer how causation survives. If everything is designated and lacks intrinsic nature, what does "A causes B" mean? Candrakīrti holds that causation too is established through designation. Calling A the cause of B conventionally labels a temporal sequence and stable covariation. It does not require A and B to have real intrinsic natures and then transmit some metaphysical substance between them. This fundamentally differs from Dharmakīrti, for whom reliable inference depends on an intrinsic connection (svabhāvapratibandha) between cause and effect, grounded in their particular characteristics. Candrakīrti brings causation back to convention without those characteristics. A philosophical resemblance to Hume may help, without implying historical influence. Hume locates causation in constant conjunction rather than logical necessity; Candrakīrti locates it in the conventions of designation rather than transmission of intrinsic nature. Both decline to give causation an ontological basis beyond convention, though their motives differ. The cost is the same as for conventional truth generally. Causal validity ultimately rests on conventional agreement, without metaphysical support.

A crucial distinction needs clarification:

Candrakīrti distinguishes correct convention (tathya-saṃvṛti) from mistaken convention (mithyā-saṃvṛti). Correct convention is cognition established in the world (loka-prasiddha). A healthy sense faculty encounters a normal object under commonly accepted conventions. Mistaken convention covers hallucinations, dreams, and erroneous perceptions that fail even conventional standards. A chariot is correct convention. People with normally functioning senses see it and call it a chariot. An illusory chariot is mistaken convention.⁵

The criterion for correct conventional cognition is convention accepted in the world, not particular characteristics presented by an object to the senses. This avoids Dignāga's premise that cognition rests on those characteristics. A simpler standard, that people conventionally recognize things this way, supports conventional knowledge.

The strategy has a cost. Reliance on worldly agreement makes conventional truth customary and variable across periods and communities, rather than true in a universal epistemological sense. Candrakīrti seems to accept the consequence. Convention is convention, valid through the stability of a community's practices rather than any guarantee beyond them. Ultimate truth occupies a sharply different level. Its emptiness analysis breaks every conventional presupposition that reifies; conventional agreement supplies the language and practices that keep life working once attachment is broken.

This is a serious attempt, but the book judges that it fails to dissolve the tension between conventional and ultimate truth. It exposes the tension more precisely. If conventional truth rests on worldly custom, while emptiness analysis itself must proceed through conventional language and inference, the question of where each belongs has only been deferred. Emptiness reveals conventions' lack of intrinsic nature, but the revelation depends on those same conventions. Candrakīrti does not break this circle, nor does his framework intend to. He claims only that within it, worldly agreement is enough and nothing more is needed.

Part II addresses the circle by changing the basis. Essay 19, Chapter 7 says that argument rests on shared conventional tools: the use of "same" and "different", procedures of individuation, and the law of noncontradiction, rather than customs that vary by time and community. Essay 18's clause B governing levels inventories them. Realists and Mādhyamikas have always shared them, and their centuries of debate testify to their use. With shared tools as the common basis, Candrakīrti's prohibition on taking the substance of phenomena as the thesis's support can coexist with the positive inferential force Bhāvaviveka demands.⁶

Part II, Essay 18, Chapter 6 addresses the other half of the problem. Emptiness analysis must itself proceed through conventional language and inference. Qingmu's statement that "the ultimate depends entirely on speech; speech is conventional" grounds the judgment that ultimate discourse must begin conventionally. Metalevel linguistic usage and logic are conventional tools. This is the two-truth framework's design, not its defect.⁶

4. Refuting consciousness-only: consciousness also lacks intrinsic nature¶

Candrakīrti's criticism of Yogācāra is concentrated in Chapter 6 of the Madhyamakāvatāra.⁷

Yogācāra's basic position is that external objects, the material world, have no substantial reality; only consciousness (vijñāna) exists. Every supposed external object is a transformation of consciousness (vijñāna-pariṇāma). Yogācāra thereby seeks to resolve realism's difficulty. If external objects are real independently of consciousness, how can we know them with certainty? Identifying them as contents of consciousness directly removes the epistemological problem.

Candrakīrti considers Yogācāra realism redirected. Replacing intrinsic nature in external objects with intrinsic nature in consciousness leaves the presupposition of something intrinsically existent intact, merely relocating reality from objects to consciousness. Emptiness analysis targets establishment through intrinsic nature itself, regardless of whether the bearer is an external object or consciousness.⁷

His criticism attacks at several points:

First, consciousness cannot know itself. Yogācāra accepts reflexive awareness (svasaṃvedana). While knowing an external object, consciousness also illuminates itself. It thereby confirms its own existence without another consciousness knowing it. Candrakīrti replies that a knife cannot cut itself and a finger cannot touch itself; consciousness likewise cannot know itself. Knowing is a relation directed toward another. If the object of a cognition is that very cognition, the relation cannot be established. Reflexive awareness fails in its logical structure and cannot ground Yogācāra's self-confirmation.²

Second, consciousness established through particular characteristics is equally subject to emptiness analysis. Candrakīrti's logic is thoroughgoing. Anything claimed to have intrinsic nature, material or mental, is a target of reductio. Consciousness and external objects established through particular characteristics face exactly the same emptiness analysis. Yogācāra thinks it is safe after transferring substantial reality from objects to consciousness. For Candrakīrti, the substantial reality of consciousness is equally unable to withstand reductio.

Third, Yogācāra makes common sense harder to explain. Candrakīrti holds that saying "I see a cup, which is conditioned, designated, and conventionally established" is more straightforward than Yogācāra's explanation. The latter first says that what is seen is a cup's object-aspect manifested by consciousness, then that consciousness alone is real. It takes more turns away from common sense without greater explanatory power.⁴

The criticism shows that Candrakīrti's Prāsaṅgika is both a technical position on methods of argument and a thorough critique of ontology. Material, mental, or epistemological, anything asserted as established through intrinsic nature is subject to refutation. Bhāvaviveka allows conventional particular characteristics, leaving Yogācāra the reply that consciousness has only such a conventional characteristic. Candrakīrti disallows even this and brings that reply within his criticism. The book nevertheless does not judge the dispute settled. Yogācāra has deeper arguments for the reality of consciousness. Candrakīrti strikes one part but does not necessarily end the whole controversy. The treatment of Yogācāra in Essay 6 will develop it further.

5. The two truths revisited: Candrakīrti's response and the remaining tension¶

The preceding chapter asked whether ultimate analysis of phenomena belongs to conventional or ultimate truth. Candrakīrti gives a clear answer, but it does not entirely dissolve the tension.

Conducting an argument about emptiness, including reductio, conceptual analysis, and stating that everything lacks intrinsic nature, is conventional activity. The philosopher conventionally uses language and concepts to lead the listener to see something. What is indicated, emptiness, belongs to the ultimate level; the act of expression remains conventional. Ultimate truth is neither another territory nor some higher content. It is more like a negative boundary condition. Whatever is established in terms of intrinsic nature is ultimately untenable. Conventional argument reveals this condition.³

The answer is coherent within Prāsaṅgika. It avoids making emptiness analysis itself discourse at the ultimate level, which would turn emptiness into a claim about ultimate substantial reality. It also avoids leaving emptiness wholly conventional. Emptiness is more than a custom and holds across all times and places.

Coherence does not, however, dissolve the tension. It remains here:

The concepts used in reductio, such as intrinsic nature, dependent origination, and establishment by designation, are themselves products of designation, as Candrakīrti acknowledges. Does analysis through designated concepts also apply to itself? If designation itself is designated, what validates its use to analyze whether everything lacks intrinsic nature? Candrakīrti answers that conventional agreement does. People use these concepts, and inferences made with them are conventionally valid. Yet this ultimately grounds the force of ultimate emptiness analysis in changeable, customary agreement. This is less an oversight than the cost of Prāsaṅgika's approach. The more thoroughly it preserves absence of intrinsic nature, the more it must ground every argument's validity in convention. Convention has no further metaphysical support; it simply operates as people agree that it does.

The book judges that Candrakīrti reaches the greatest coherence available for this tension within Prāsaṅgika's own framework. If asked what makes convention itself valid, he will give no metaphysical answer. He will say only that the question has passed beyond meaningful inquiry, because meaningful inquiry itself is a product of conventional agreement. This is both the strength and the limit of his position.

6. The Madhyamakāvatāra: emptiness and the bodhisattva path¶

The Madhyamakāvatāra combines emptiness philosophy with the bodhisattva's ten grounds. This structural arrangement has an internal logic; it is not decorative.

The ten grounds organize the work, with emptiness arguments placed at the sixth, the manifest ground (abhimukhī-bhūmi). Why the sixth? In traditional classifications, its defining feature is perfection of wisdom and direct perception of emptiness. Candrakīrti makes this his point of reference. The first five grounds present the preliminaries of giving, ethical conduct, patience, effort, and meditation. The sixth develops emptiness directly; the subsequent grounds are fruits of compassion and wisdom working together.

Chapter 6 concentrates the arguments on emptiness and is much longer than the others. Its approximate structure is:

The sequence shown below the Chinese diagram is: frame the question → refute mistaken attachments → establish the conventional → dismantle grasping at self → move toward liberation.

At the end of Chapter 6, Candrakīrti distinguishes two levels of no-self: the selflessness of persons (pudgala-nairātmya) and the selflessness of phenomena (dharma-nairātmya). He uses the distinction very differently from Yogācāra. Yogācāra makes it a hierarchy. Realizing personal selflessness is the first step, while the selflessness of phenomena is Mahāyāna's deeper meaning. Candrakīrti rejects this hierarchical presupposition. He says that śrāvakas and pratyekabuddhas also realize the selflessness of phenomena, though they do not cultivate it fully. The śrāvaka vehicle's scriptures already teach it, only briefly. Śrāvakas and pratyekabuddhas fully cultivate personal selflessness. They also see the selflessness of phenomena, although their cultivation of it remains incomplete; they are not wholly without that insight.⁸ The pair primarily distinguishes two aspects of analysis, with the difference lying in complete or incomplete cultivation rather than seeing or not seeing.¹

The book's terminology principle applies here. It does not use the Mahāyāna pair of personal and phenomenal selflessness when discussing early Buddhism. The pair presupposes a hierarchical classification imposed afterward by Mahāyāna. Early analyses of phenomena are described in their own terms. The pair is used here because Candrakīrti works within Mahāyāna, and his particular use is itself one of the chapter's subjects.

7. Nondiscriminating wisdom: how emptiness connects with liberation¶

Candrakīrti's account of liberation continues the account of the cessation of conceptual proliferation in the Nāgārjuna chapter, 5.1. He formulates it more precisely under Prāsaṅgika's epistemology.

Ordinary cognition brings a conceptual framework. It classifies an object as a chariot, assigns a property such as goodness, and produces attraction or avoidance, such as wanting the chariot. It treats the object as an entity established through particular characteristics. Nondiscriminating wisdom (nirvikalpa-jñāna) does not mean knowing nothing. Its point is the absence of the act that frames an object as an X established through particular characteristics. Cognition continues without reifying its object through intrinsic nature. Cessation of conceptual proliferation (prapañca) means stopping that automatic reification.⁴

Candrakīrti connects this cessation with direct perception of emptiness at the sixth ground. When practitioners directly see emptiness rather than merely understand it conceptually, their seeing includes no act of establishing the object through particular characteristics. What matters is the manner of seeing, without a reifying framework, rather than a special content called emptiness.

This is why the Madhyamakāvatāra places direct perception of emptiness at the sixth ground rather than earlier. Practice in the first five prepares the conditions, but practitioners still work with emptiness as a concept in some way. The sixth ground's direct perception is the moment when that conceptual framework itself is relinquished.

A tension nevertheless deserves attention. If nondiscriminating wisdom means encountering objects without establishing them through particular characteristics, would newborn infants and animals without abstract conceptual capacities naturally occupy this state? They too lack conscious reifying frameworks and do not actively establish an object as an intrinsically characterized X. Candrakīrti's liberation theory must explain how its state differs. The problem is substantial. It concerns whether nondiscriminating wisdom purifies a cognitive capacity or suspends it. Volume 2 will return to this tension.

8. Historical influence and position¶

Candrakīrti's influence in India was not especially prominent during his life or the following century or two. The mainstream of middle-period Indian Buddhism, from the seventh to eleventh centuries, was Dharmakīrti's logic and epistemology and the synthesis of Svātantrika with Yogācāra Madhyamaka represented by Śāntarakṣita and Kamalaśīla. Candrakīrti's Prāsaṅgika approach, traditionally associated with Nālandā, was an important voice but not dominant.⁹

His principal historical influence lies in Tibet, discussed in Essay 11. After the eleventh century, his works were translated into Tibetan on a large scale, and the Madhyamakāvatāra became central to Madhyamaka education. When Tsongkhapa systematized Gelug thought in the fourteenth and fifteenth centuries, Candrakīrti's Prāsaṅgika was promoted as the highest Madhyamaka view, above every Svātantrika approach. The classification was so influential that Prāsaṅgika is now often simply equated with Candrakīrti's Madhyamaka. The equation has grounds but obscures India's more complex Buddhist landscape.

One historical detail matters. Dharmakīrti was among Candrakīrti's Indian opponents. He brought Indian Buddhist epistemology to a comprehensive form in the Pramāṇavārttika, constructing a rigorous system based on perception and inference. Philosophically Yogācāra, he accepts reflexive awareness and consciousness established through particular characteristics. Candrakīrti's criticism of reflexive awareness in Section 4 directly targets a key structure of this epistemology, while his rejection of conventional particular characteristics directly challenges Dharmakīrti's basis for perception.

Dharmakīrti's tradition broadly replies that reflexive awareness has direct empirical support, whereas the knife that cannot cut itself is only a comparison, insufficient as proof. Knowing and cutting need not share a logical structure. A deeper challenge concerns memory. How can memory be explained without reflexive awareness? Remembering yesterday's happiness requires that the past happiness left some mark of having been known at the time; reflexive awareness supplies that mark. Without it, the account has a gap. Candrakīrti responds that memory too can be understood through designation. Remembered happiness is a linguistic connection within a sequence of conditioned events and needs no reflexive awareness beneath it. Critics find this too brief and less precise than Dharmakīrti's account of memory.

The Indian dispute had no clear winner. Both traditions continued with their own followers. In Tibet, the Gelug school incorporated Dharmakīrti's epistemology into a Candrakīrti-style Madhyamaka framework. Tsongkhapa himself wrote little on epistemology. The synthesis was chiefly the work of his disciples Gyaltsab Je and Khedrup Je and their successors and will be discussed in Essay 11.¹⁰

The book places Candrakīrti as the Madhyamaka thinker who carried refutation without a position to its philosophical limit. Reductio, designation, criticism of reflexive awareness, and replacing every metaphysical foundation with conventional agreement form a system of considerable internal coherence. Its limit has been identified. The more thorough it becomes, the more it grounds every argument in conventional agreement without being able to question that grounding further. This is less Candrakīrti's failure than the inevitable cost of Prāsaṅgika's thorough absence of intrinsic nature. The next essay turns to Yogācāra because some thinkers found this cost too high and sought a path that could positively explain the structures of cognition and practice.

References¶