Chapter 2 Dharmakīrti's Logic and Epistemology¶

Dignāga laid the foundations of Buddhist epistemology: two means of valid cognition, perception and inference; the distinction between particular and general characteristics; exclusion theory; and the three characteristics of a reason. Dharmakīrti (c. 600–660) took up this framework and refined every crucial part of it, sometimes defending it, sometimes revising it, and sometimes greatly extending it. He was the most influential thinker in the history of Indian Buddhist epistemology. His works occupied a central place in the monastic institutions of late Indian Buddhism and became standard textbooks of logic and epistemology in Tibetan Buddhism. Whereas 8.1 focused on Dignāga, this section examines where Dharmakīrti went further, the direction he took, and his fundamental disagreement with Candrakīrti over the foundations of epistemology.

I. Dharmakīrti's Life and Works¶

Life

Reliable historical evidence for Dharmakīrti's life is also scarce. Tradition says that he was born into a Brahmin family in South India, studied the Brahmanical teachings widely in his youth, and later converted to Buddhism. According to the same tradition, he studied at Nālandā under Dignāga's disciple Īśvarasena and then surpassed his teacher to become the most important developer of Dignāga's system.

This lineage is itself worth considering. Dharmakīrti studied under Dignāga's disciple, not Dignāga himself, yet his interpretation and criticism of Dignāga's writings were more searching and thorough than anyone else's. Traditional accounts place him roughly in the same period as Candrakīrti, and there is clear philosophical tension between their positions.¹

The seven works on valid cognition

Dharmakīrti's central writings are commonly called the "seven works on valid cognition" (saptaviṃśatikā or saptaprakaraṇa):

The Pramāṇavārttika's four chapters discuss the establishment of valid cognition (pramāṇasiddhi, arguing that the Buddha himself is a means of valid cognition), perception (pratyakṣa), inference for oneself (svārtha-anumāna), and inference for others (parārtha-anumāna). The first chapter argues logically that the Buddha is an exemplar of a source of knowledge (pramāṇabhūta). This unusual approach in the history of philosophy reconstructs the grounds for trusting the Buddha's teaching at the epistemological level, without directly appealing to premises of faith.¹

II. Inheritance and Innovation: From Dignāga to Dharmakīrti¶

Dharmakīrti accepted Dignāga's overall framework but treated every crucial question more precisely, and sometimes more radically.

What he accepted: the two means of valid cognition, perception and inference; the ontological distinction between particular and general characteristics; the three conditions of valid inference; and the basic direction of exclusion theory.

What he developed or revised:

First, Dignāga's explanation of why a reason supports an argument remained a formal description: the three conditions. Dharmakīrti asked a deeper question: under what conditions can these three characteristics hold? His answer was "natural connection" (svabhāvapratibandha). Inference must rest on an intrinsically necessary connection between its objects. This is his most central contribution, discussed in the following section.

Second, Dignāga's formulation of exclusion theory faced Kumārila's charge of circularity. Dharmakīrti introduced "causal efficacy" (arthakriyā) to address it.

Third, Dharmakīrti treated perception more systematically, especially the relation between momentary perception and conceptual judgment, and the epistemological status of yogic perception.

III. Natural Connection: The Basis of Reliable Inference¶

The problem

Dignāga's three conditions describe the formal structure of a valid inference: the reason must be present in the subject of the thesis, present in at least some established similar instances, and absent from all established dissimilar instances. This description is external and formal. It does not explain why the structure exists, or under what conditions we are entitled to regard it as actually holding.

For example, observing smoke together with fire does not establish pervasion simply by increasing the number of observations. We must also explain why smoke depends on fire for its production. Smoke is produced by fire; the two do not merely coincide by accident.

Dharmakīrti answers by introducing natural connection (svabhāvapratibandha, a connection grounded in nature). An inference is reliable if and only if there is an intrinsically necessary connection between the reason and the thesis-property, the property being inferred. This necessary connection has only two sources:²

First, a connection of nature (svabhāva): B is part of A's nature, or A is a form of B. For example: "This is a tree (A), so it is a plant (B)." A tree's nature includes the property of being a plant. The two are necessarily connected by nature, rather than accidentally coexisting.

Second, a causal connection (kārya-kāraṇa): B is an effect produced by A. For example: "There is smoke (B), so there is fire (A)." Smoke is an effect produced by fire. The inference from smoke to fire requires this relation of production, not that every fire must produce smoke. This causal relation grounds the necessary link between reason and thesis-property, beyond mere empirical regularity.

Pervasion has a reliable basis only when it rests on one of these two kinds of natural connection. Concomitance in a limited set of instances does not establish that all three characteristics have been fully demonstrated. If reason and thesis-property merely coincide accidentally, without a necessary connection, the inference is unreliable.

The philosophical implications are far-reaching. Every reliable inference requires an intrinsic, necessary connection in an ontological sense between the reason and the thesis-property. This presupposes that things have some real intrinsic nature (svabhāva) capable of supporting such a connection. Here lies the crucial target of Candrakīrti's criticism, discussed in Section VII.

IV. Kinds of Reasons: Nature, Effect, and Nonapprehension¶

We must first distinguish two questions often conflated. The first asks what grounds pervasion (vyāpti). Dharmakīrti's answer is that a necessary connection must link the reason and the property to be established, and this connection has only two bases: causal production (tadutpatti) and essential identity (tādātmya). The second asks how many kinds of valid reason there are. Here the answer is three: an effect reason, a nature reason, and a nonapprehension reason. The first question concerns the basis of a relation; the second concerns the kinds of reasons. They are not the same question.³

Starting from the two connections of essential identity and causal production, consider first the two corresponding kinds of reasons:

Nature reason (svabhāva-hetu, inference based on a connection of nature)

When the reason and the thesis-property are related through inclusion by nature—B is part of A's nature—the inference takes this form: - Thesis: this A is B (for example, this tree is a plant). - Reason: because it is A (because it is a tree). - Similar cases: all trees are plants.

This inference involves no causal relation. It rests purely on inclusion by nature. The nature of a "tree" includes the property of being a "plant," so once something is identified as a tree, one can infer that it is a plant without further empirical evidence.

Effect reason (kārya-hetu, inference based on a causal connection)

When the reason is an effect produced by the cause to be inferred—B is produced by A, as smoke is produced by fire—the inference takes this form: - Thesis: there is fire there (A). - Reason: because there is smoke there (B is the effect of A). - Similar cases: wherever there is smoke, there must be fire.

This inference depends on the necessity of a causal relation. B is a product of A, so observing B allows us to infer reliably that A is present, or has been present.

Nonapprehension reason (anupalabdhi-hetu, negative inference based on nonapprehension)

The third kind infers nonexistence from nonapprehension. For example: "If a pot were here, it would be perceptible under the present conditions of lighting, distance, and sensory capacity. It is not perceived, so we can judge that there is no pot here." This is restricted to an object that would be perceptible under the present conditions; it does not equate every failure to see something with its nonexistence.⁴ It is not a variant of the first two. It is classified separately because what it seeks to establish (sādhya) is an absence. The criterion concerns the nature of what is proved, not whether the reason itself is an absence. If "there is a fierce fire here" is used to prove "there is no lasting cold touch here," the reason remains a genuine negative reason even though it asserts a presence.

Inferences outside these three kinds, such as those based solely on empirical regularity, analogy, or authority, are not fully reliable within this epistemology.⁵

V. Further Development of the Theory of Perception¶

Building on Dignāga's four kinds of perception, Dharmakīrti analyzed perception's internal structure more precisely.

Momentariness (kṣaṇikatā) and perception

Dharmakīrti firmly upheld Buddhist momentariness. Everything arises and ceases in a moment, and nothing lasts for more than one moment. This has profound implications for perception. Genuine perception knows a momentary object: an instantaneous, unrepeatable, individual particular characteristic. When we think "I see a tree," the actual cognitive activity is a succession of momentary perceptions, each with a unique particular characteristic as its object. The concept "tree" is a subsequent conceptual superimposition.

Determinate judgment (vyavasāya) and perception

Dharmakīrti distinguishes two levels: pure perception, which is nonconceptual, and the subsequent determinate judgment (vyavasāya), which involves concepts. In everyday experience, perception always occurs together with an automatic judgment such as "This is a tree." Epistemological analysis must separate them. The former is genuine perception; the latter is conceptual construction (kalpanā).

The practical implication is that the boundary of language falls precisely between perception and conceptual judgment. Perception itself cannot be directly expressed in language, since concepts cannot fully capture particular characteristics. Language deals with conceptual judgments. "I see a tree" no longer describes pure perception; it describes perception with a concept superimposed on it.

Kinds of perceptual error (bhrānta)

Dharmakīrti systematically analyzed perceptual errors: illusions (māyā), mirages (marīci), dreams (svapna), seeing two moons (dvicandra-bhrānta, double vision caused by pressure on the eyeball), and others. These errors share a common feature: the senses operate outside their normal functional conditions, or conceptual judgment exceeds what is actually perceived. This analysis is both empirical and an attempt to delimit precisely the scope of perception's condition of being "free from error."

VI. Revising Exclusion Theory: Causal Efficacy¶

Dignāga's exclusion theory (apoha) faced Kumārila's charge of circularity: if "cow" means "excluding non-cows," how do we delimit "non-cow"? This presupposes that we already know what a cow is and thus forms a circle.

Dharmakīrti responded by introducing causal efficacy (arthakriyā, a thing's capacity to produce effects) as a firmer basis for exclusion theory:⁶

Different kinds of things differ systematically in their causal capacities. What a cow can do, such as produce milk or plow a field, is not what a horse or a tree can do. These differences in causal efficacy determine the objects to which we apply "cow" and those to which we do not. A word's application is ultimately grounded in actual causal differences between things, without requiring a common entity called "cowhood."

This revision means that exclusion theory is no longer purely negative, equating "cow" with "excluding non-cows." It also has a positive basis. A thing's specific causal capacities determine its cognitive classification. We call something a "cow" because it shares a specific profile of causal capacities with other "cows." This profile is real, at the level of particular characteristics, rather than purely conventional.

This also connects with Dharmakīrti's criterion of existence, discussed in the next section. What really exists is precisely what has causal efficacy; the classifications made by words track actual differences in that efficacy.

Does the causal account escape circularity? We should acknowledge that this revision did not end the debate. Mīmāṃsā criticisms reappeared at another level. To say that individual cows "share certain causal capacities"—all can produce milk or carry loads—already presupposes a classification of effects. What licenses treating instances of "producing milk" as the same kind of effect? If it is a universal property of "producing milk," realism about universals returns. If it is exclusion, with "producing milk" defined by excluding what does not produce milk, the circle has merely moved from words to effects. Dharmakīrti's response is to stop the regress at nonconceptual perception and habitual dispositions. Similar particular characteristics produce similar perceptual judgments in cognizers. This production of similar judgments is itself a causal fact and needs no further foundation in conceptual classification. Stable classification is ultimately sustained by both beginningless conceptual dispositions (vāsanā) and actual causal regularities. This response effectively concedes that exclusion theory has causal and psychological foundations, not logical ones. Whether it resolves the circularity or redescribes it as an acceptable naturalistic fact remains disputed in scholarship on Buddhist epistemology.⁶ A fair assessment, in this book's view, is that Dharmakīrti reduced a fatal logical objection to a theoretical cost one could live with. That is a substantive achievement in defending the theory, but it is not the end of the argument.

VII. The Criterion of Existence: Capacity to Produce Effects (Arthakriyāsāmarthya)¶

Dharmakīrti proposed an influential ontological criterion: what really exists is what can produce effects (arthakriyāsāmarthya).⁷

This criterion has several implications:

First, only particular characteristics can truly produce effects. A general characteristic is the mind's conceptual integration of multiple particular characteristics. As a product of that integration, it produces no effects itself. Concrete individual things, particular characteristics, produce effects; abstract common features do not. Ontologically, then, it is particular characteristics that truly exist, not general characteristics.

Second, to exist is to be capable of producing effects (sat = arthakriyāsamartha). Conversely, if something cannot produce any effect at all, it does not exist or is merely a conceptual construction. This provides an ontological basis for Buddhist momentariness. Something that persists produces effects at every moment, but those effects differ from moment to moment. Otherwise there would be no change, contradicting the phenomenon of continuity. Different effects show that the "thing" producing them changes at each moment. Thus no "thing" truly persists beyond a single moment. What is permanent and unchanging can produce no effect, because "producing" itself entails change.

Third, this criterion connects epistemology and ontology. Perception knows precisely those individual things, particular characteristics, that have causal efficacy in the present moment. Inference grasps the connections between them through reasoning. Cognition's reliability ultimately derives from the causal reality of its objects.

The following analogy with modern philosophy serves only to aid understanding. Dharmakīrti's criterion of causal capacity is very close to pragmatism's theory of truth. Something is real if it "makes a difference," producing a distinctive result in some way. It is also close to one criterion in scientific realism, which treats an entity as real if we can interact causally with it. Dharmakīrti's purpose, however, is Buddhist. He seeks to join momentariness to epistemology, not to develop pragmatism.

This criterion does not conflict with the Yogācāra denial of external objects. A reader may object here. Dharmakīrti's school is usually placed under the Yogācāra name, and Yogācāra holds that nothing stands apart from consciousness. How can the same thinker make "real existence" his criterion? The answer lies in Dharmakīrti's method. His treatises speak on more than one level. Most of the time he adopts a Sautrāntika-style external realism, granting momentary particulars outside the mind and using them to explain perception and inference. In one section of the Pramāṇavārttika, the discussion of the result of valid cognition in the perception chapter, he steps back to a deeper level on which the object of awareness is internal and talk of external objects is only provisional. A lower level holds within its own range and is superseded when analysis is carried through. Dharmakīrti himself rarely stays at the deepest level. His aim is to defend Dignāga's theory of valid cognition, and he usually argues at the level of external realism.⁸ The criterion of causal capacity says nothing about "outside the mind." It says only that what really exists is what produces effects. Whether that producer is an external atom or a moment of consciousness depends on the level in play. At the level of external realism, what gives rise to perception is the external particular. At the Yogācāra level, it is the imprint within consciousness. The universal produces nothing at either level. The third point above, "the causal reality of its objects," therefore needs a qualification. That reality is reality relative to conceptual construction, not reality outside the mind. The layered method lets Dharmakīrti keep both ends, and it is also why later interpreters have assigned him to opposite camps.

VIII. Yogic Perception: Connecting the Epistemology of Practice with the Theory of Valid Cognition¶

Dignāga listed yogic perception (yogi-pratyakṣa) as a kind of perception without arguing in detail why it counts as genuine perception. Dharmakīrti offered a fairly detailed defense. This is among the most controversial and religiously significant parts of his work.

The problem: yogic perception claims to be direct insight into reality, such as the Four Truths, no-self, and emptiness, attained through deep meditation (dhyāna) and calm and insight (samādhi). But how does it qualify as "perception"? By Dharmakīrti's own criteria, perception must be nonconceptual and have particular characteristics—individual entities with causal efficacy—as its objects. Yet "suffering," "no-self," and the other things supposedly known in meditation are abstract. Can they be objects of perception?

Dharmakīrti's defense:⁹

Through prolonged meditative training, practitioners can turn what was originally conceptual knowledge—"impermanence is true"—into a vivid direct experience resembling sensory perception. This does not mean that impermanence becomes a color visible to the eyes. The point is that practice can allow a proposition's truth to be experienced through direct presentation, without inference. For comparison, someone may know intellectually that being burned is painful, yet "directly know" it only at the moment of actually being burned. Through meditation, a practitioner turns the propositional knowledge that "suffering is suffering" into a perceptual experience of directly feeling the nature of suffering.

The philosophical difficulty is whether this defense stretches the criteria of perception—being nonconceptual and knowing particular characteristics—beyond what the concept of a particular characteristic can cover. Dharmakīrti himself recognized this tension. His position is closer to treating yogic perception as a practitioner's direct, vividly clear perception of things in meditation. Its objects are concrete instances of suffering, impermanence, no-self, and so forth, rather than abstract propositions. This perception has simply reached a purity at which conceptual judgment no longer intervenes.

This defense is the crucial link connecting the epistemological tradition with Buddhist liberation theory. If yogic perception does not count as perception, insights gained through practice fall outside the epistemological system and become indefensible mysticism. By including it, Dharmakīrti seeks to give religious cognition, insight in practice, epistemological legitimacy.

IX. The Fundamental Disagreement with Candrakīrti¶

The discussions in 5.5 and 6.5 have presented the opposition from Candrakīrti's perspective. Here we add Dharmakīrti's side.

Dharmakīrti's natural connection (svabhāvapratibandha) appears to Candrakīrti as a "residue of realism." To say that things have intrinsically necessary connections implies that each has some fixed, unchanging intrinsic nature (svabhāva). Otherwise, what grounds the "necessary connection"? Candrakīrti's Prāsaṅgika Madhyamaka recognizes no intrinsic nature beyond linguistic designation and therefore no "necessary connection" in the full sense. It reduces causation to conventions established by designation (5.5).

A response can be inferred from Dharmakīrti's theoretical position, though he did not directly answer Candrakīrti. The two were roughly contemporary, and the history of the debate is unclear. Without some necessary connection grounded in nature, the reliability of inference cannot be guaranteed. "Smoke implies fire" holds not merely because we conventionally say so, but because smoke is actually produced by fire. This fact does not depend on convention; it reflects the actual causal structure of things. If that structure too is reduced to linguistic designation, how can we know that any inference is reliable rather than merely customary?

Ultimately, Dharmakīrti's epistemology needs some "causal reality of the objects of cognition" as its foundation; without it, epistemology itself cannot stand securely. Candrakīrti holds that conventions accepted by the world are already sufficient and need no metaphysical foundation beyond convention. Their disagreement gives the clearest expression within the epistemological tradition of the question raised in 8.1, extending the discussion of Dignāga: does valid cognition require a metaphysical foundation, or are conventional practices enough?¹

X. Historical Influence¶

In India

Dharmakīrti was a central figure in Nālandā's scholarly tradition. His seven epistemological works became standard texts for teaching the subject at Nālandā and later at Vikramaśīla. Important commentators followed him. Dharmottara (c. 750) wrote the Commentary on the Ascertainment of Valid Cognition, commenting on the Nyāyabindu. Prajñākaragupta (c. 800) wrote the Ornament of the Commentary on Valid Cognition, an extensive commentary on the Pramāṇavārttika. Śākyabuddhi also wrote a commentary on the Pramāṇavārttika. These commentators differed in detail and became one source of later Tibetan interpretive traditions.

Nyāya's challenges to Dharmakīrti also continued for a long time. Uddyotakara's successors, chiefly Maṇḍanamiśra and later the Vedānta school, maintained critical pressure on the epistemological tradition.

In Tibet

Dharmakīrti's reception differed from Dignāga's. Dignāga's writings were earlier, but from the eleventh century onward Dharmakīrti's seven epistemological works, especially the Pramāṇavārttika, became central to Tibetan scholarly education, receiving far more extensive reading and commentary than Dignāga's works. Tibet's main commentary traditions followed two lines:

One was Sakya. Its representative figure, Sakya Paṇḍita (1182–1251), wrote the Treasury of Reasoning on Valid Cognition (Tshad ma rigs pa'i gter), a foundational work of Tibetan Buddhist epistemology based chiefly on Dharmakīrti. The other was Gelug, which integrated Dharmakīrti's epistemology with Madhyamaka philosophy into a framework of "conventional valid cognition" (tha snyad pa'i tshad ma). Epistemology addresses the conditions of valid conventional cognition; Madhyamaka addresses ultimate emptiness. Each does its own work without conflict.¹

The first chapter of Dharmakīrti's Pramāṇavārttika, on the establishment of valid cognition (pramāṇasiddhi), argues that "the Buddha is the ground of knowledge." Tibetan Buddhist logic teaches this as an epistemological defense of the path of practice, a philosophical argument for trusting the Buddha's awakening. It directly connects the purely logical theory of valid cognition to the spiritual significance of practice. This is one reason for the lasting importance of Dharmakīrti's writings in Tibetan Buddhism.

References¶