Chapter 1 Dignāga: The Foundations of Buddhist Logic and Epistemology¶

The discussion in essays 5, 6, and 7 traced Indian Mahāyāna philosophy along two axes: ontology and the theory of practice. Madhyamaka declared everything empty. Yogācāra rebuilt epistemology on the premise of emptiness, while tathāgatagarbha thought established an affirmative basis in Buddha-nature from another direction. These three discourses were essentially developments within Indian Buddhism itself. Their main opponents in debate were the Buddhist schools themselves.

From the fifth and sixth centuries onward, another line of thought developed within Indian Buddhism: systematic epistemology and logic, or pramāṇa-śāstra and hetu-vidyā. Its central figure was Dignāga (c. 480–540). The system he founded changed philosophical debate within Buddhism and profoundly influenced the course of epistemological discussion throughout Indian intellectual history.

I. The Philosophical Setting for the Rise of Buddhist Epistemology¶

Why at this time?

By the fifth century, the Indian tradition of debate between schools (vāda) was already well developed. Vaiśeṣika and Nyāya had developed sophisticated theories of categories and logic. Mīmāṃsā had established an epistemology centered on explaining the authority of the Vedas; and Sāṃkhya and Vedānta each had their own metaphysical and epistemological frameworks.

In these debates, a philosophical school without a systematic epistemology could hardly maintain the rigor of its position in debate. Early Buddhist Abhidharma already offered detailed analyses of mind and cognition. Madhyamaka had a powerful method of reductio but no affirmative epistemology, as it established no position. Yogācāra addressed epistemological questions, but its emphasis fell more on psychological analysis than on logic.

Before Dignāga, the logic used in Indian Buddhism was largely borrowed from Nyāya's five-member argument: thesis, reason, example, application, and conclusion. This form had its imprecisions and had been developed within Nyāya's philosophical framework. Using it meant accepting, to some extent, Nyāya's philosophical presuppositions.

Dignāga sought to establish Buddhism's own epistemology and logic, independent of the Nyāya framework and directly challenging its philosophy.¹

The Yogācāra epistemological background

Dignāga's philosophical position belonged to Yogācāra, as Essay 6 noted (see the timeline in 6.3). He had a teacher–disciple relationship with Vasubandhu. Yogācāra's analysis of the structure of cognition, especially its distinction between the object aspect, the subject aspect, and the aspect of self-cognition (6.3), became a rigorous epistemological theory in his work. Yogācāra asked, "What is the internal structure of consciousness?" Dignāga asked, "What are the conditions of valid cognition?" Their guiding questions differed, but they shared Yogācāra's realist commitment to particular characteristics (svalakṣaṇa, individual realities).

II. Dignāga's Life and Works¶

Life

Dignāga, rendered in Chinese as Chenna (陈那) or Dayulong (大域龙), was active roughly from the late fifth to the early sixth century (c. 480–540). Tradition says that he was born into a Brahmin family in South India and later converted to Buddhism. According to this tradition, he first studied with Nyāya teachers and then with Vasubandhu, moving toward a synthesis of Yogācāra and epistemology. What can be established is that his works show a deep knowledge of the epistemologies of Nyāya, Mīmāṃsā, and Vaiśeṣika, and develop systematic critiques of these traditions.

Major works

The Pramāṇasamuccaya is Dignāga's most important work. Its Sanskrit original has been lost; modern scholarship reconstructs it mainly from the Tibetan translation and Dharmakīrti's later commentary, the Pramāṇavārttika.²

III. Two Means of Valid Cognition: Perception and Inference¶

Dignāga's first major epistemological innovation was to reduce the sources of valid cognition (pramāṇa) from Nyāya's four to two: perception (pratyakṣa, direct perception) and inference (anumāna, reasoning).

Nyāya recognized four means of valid cognition: perception, inference, comparison (upamāna), and authoritative testimony (śabda). Dignāga argued that comparison could be reduced to inference, since analogical reasoning is a kind of inference. The validity of authoritative testimony ultimately also depends on reasoning about whether an authority is trustworthy. There are therefore only two independent means of valid cognition.¹

The definition of perception (pratyakṣa)

Dignāga's definition of perception is a classic: cognition free from conceptual construction (kalpanāpoḍha, nonconceptual) and free from error (abhrānta, nonerroneous).

Both conditions are indispensable:

• Free from conceptual construction: perception must be nonconceptual. It is simply the direct presentation that occurs when a sense faculty encounters an object, without any superimposed concept. "Seeing something red" is perception; "seeing an apple" already introduces the concept "apple" and is no longer pure perception. In the strict sense, perception therefore cannot be expressed in language.
• Free from error: this excludes unreal perceptions such as hallucinations and dreams.

Dignāga further divided perception into four kinds:

The definition of inference (anumāna)

Inference is cognition that reaches new knowledge by reasoning from known facts. Dignāga's major contribution was to lay down strict conditions for inference: the "three characteristics of a reason" (trirūpa-hetu), discussed in Section VII.

IV. Particular and General Characteristics: Epistemology's Realist Basis¶

Dignāga's epistemology rests on an explicit ontological commitment: only particular characteristics (svalakṣaṇa) are real; general characteristics (sāmānyalakṣaṇa) are conceptual constructions.

Particular characteristics are the individual features of each concrete thing or event. An example is this apple's specific combination of exact weight, color, and taste at this moment, which cannot be completely identical to that of anything else. Particular characteristics are the objects of perception, what cognition encounters at its most immediate level.

General characteristics are common features abstracted from multiple particular characteristics, such as "applehood," "redness," and "roundness." A general characteristic is not an independently existing entity, contrary to Nyāya's realism about universals. The mind constructs it by conceptually combining multiple particular characteristics. General characteristics are the objects of inference.

This distinction has an important consequence. Language and propositions deal entirely with general characteristics, which are conceptual constructions, not immediate realities. Therefore, no linguistic proposition can directly capture the reality known through perception. Language can handle only abstract general characteristics, whereas perception knows specific particular characteristics. A fundamental gap separates the two.

This position closely fits Yogācāra's theory of manifestation by consciousness. What actually exists—particular characteristics—is the immediate content of consciousness. Language and concepts deal with general characteristics abstracted from that content. Such characteristics are constructions, not immediate reality.³

The following comparison concerns ideas, not historical influence. Dignāga here confronts the same question that arose in Western philosophy, the problem of universals. Nyāya's realism about universals is structurally close to Platonic realism. Universals such as "cowhood" and "roundness" are real, independently existing entities, and words work by positively referring to these universal properties. Dignāga's position is closer to nominalism. Only the concrete features of individual things—particular characteristics—are real; universals are concepts constructed by the mind from multiple particular characteristics and have no independent reality. This is structurally very similar to the position of Ockham in the fourteenth century. Both hold that only particulars are real and that universals are merely names or concepts. Both face the same further question: if universals are not real, how can a general word such as "cow" apply to every cow?

One qualification makes the comparison more precise. Unlike Plato, Aristotle did not hold that universals exist independently beyond particulars. His moderate realism placed universals within particulars (in re), between Platonism and nominalism. Dignāga's opponents, the Naiyāyikas, are thus closer to Plato, while Dignāga himself is closer to nominalism. Aristotle's approach has no directly corresponding figure in the Indian debates over valid cognition.

Dignāga's account of language offers a further innovation. Ockham and Western nominalists generally still accepted that language could function effectively, denying only that universals were independent entities. Dignāga went further and asked how language was possible. His answer, exclusion theory (apoha), was purely negative. Words delimit meaning by excluding what does not belong to a category, rather than by positive reference. This proposal has no direct counterpart in Western nominalism. It is closer to discussions of negative semantics after Frege, and offers a distinctive answer to a problem shared across cultures.

V. Self-Cognition: Cognition's Grasp of Itself¶

Dignāga classifies self-cognition (svasaṃvedana) as a kind of perception, specifically inner perception. Here he directly inherits and develops Yogācāra epistemology.

The basic claim of self-cognition is that a cognitive act does more than cognize an external object, or an object at the level of the object aspect. It simultaneously cognizes itself. A cognitive act is "luminous" to itself and needs no further act of cognition to know it.

Within Yogācāra's theoretical framework (6.3), this addresses the problem of how cognition is known. If every cognition required another cognition to know it, an infinite regress would follow. Self-cognition cuts off this regress by holding that cognitive activity is self-illuminating.

Dignāga appeals to memory as evidence. We can recall a past experience: "I once perceived a tree." This memory contains both the cognitive content of that perception—the tree—and the perception itself, the act of perceiving. If the act of perception had not somehow been cognized and recorded, how could we later recall it? Perception must therefore have cognized itself in some way when it occurred. Otherwise, memory could not arise.²

Theravāda Abhidhamma explains the same phenomenon without self-cognition. Memory does not require perception to have known itself at the time. It requires only that the perception aggregate (saññā, Sanskrit saṃjñā, one of the mental factors) automatically "mark" the object when perception occurs, grasping its features and leaving their imprint. Later, mental consciousness (mano-viññāṇa), taking a mental object as its object, can take an object from a previous moment, or its imprint, as the content of present cognition. This is the mechanism of memory. In this system, a present cognitive act need not know itself as it occurs. It need only take a past mental event as its object, and the marking function of the perception aggregate ensures that the past event is available to be taken up.

The Theravāda does more than offer another route; it denies outright the premise Dignāga needs. The Kathāvatthu, Book V, Discourse 9, records the Theravādin's exchange: if there is knowledge of the present, does one know that knowledge by the same act of knowledge? The answer is "it should not be said so"; to admit it would be to say that one cuts a sword with that sword and touches a fingertip with that fingertip.⁴ Knowledge of the present can be known only by a later act of knowledge, and the English translators' running head for the page reads "Knowledge of the Present is Retrospective." The Visuddhimagga supplies the other half in its definition of perception: its characteristic is cognizing, its function is making a sign as the condition for recognizing again, as carpenters mark timber.⁵ At the time of cognition only a mark is made; recognition comes later. Nor is this the Theravāda's view alone. The Mahāvibhāṣā, fascicle 9, "in order to refute other schools," rejects the Mahāsāṃghika claim that "mind and mental factors can know their own nature, as a lamp illuminates itself and others," and lays down that "a nature does not know its own nature," first because "it is seen in the world that a fingertip does not touch itself, a blade does not cut itself, a pupil does not see itself," and second because the Buddha established that eye-consciousness arises from the two conditions of eye and form, which could not stand if consciousness knew itself.⁶ The simile of the knife that cannot cut itself was in use on both sides of the Abhidharma long before Candrakīrti; self-cognition descends from the Mahāsāṃghika and Sautrāntika line, not from an Abhidharma consensus.

The question is therefore no longer whether the Theravāda has an alternative but what Dignāga's regress argument assumes. The reasoning in the first chapter of the Pramāṇasamuccaya runs: if cognition, like form, were experienced by another cognition, that will not do, because experience by another cognition leads to infinite regress, since that other cognition will later be remembered too and so must itself be experienced by yet another.⁷ The reasoning assumes that every cognition must be known. The Abhidharma model does not accept the assumption: processes follow one another, and only the mind that a subsequent mind-door process takes as its object is known; the rest fall back into bhavaṅga, no one asks further, and the chain stops there. The regress is potential, not actual. Dignāga's first premise, that there is no memory of what was not experienced, is also empirically shaky: memory research files "remembering that one saw" under source monitoring, the attribution of a source to an experience being a later judgment that can go wrong; in one experiment six of twenty-four subjects, under suggestion, "remembered" a childhood episode of being lost that had never happened.⁸ If memory can be produced from nothing, it cannot prove that the cognition knew itself at the time. Higher-order theories in contemporary philosophy of consciousness give the same reply: what makes a mental state conscious is a higher-order thought that takes it as its object, and that higher-order thought is itself usually not conscious; only when one introspects again is there a third order, so there is no infinite regress.⁹

What Dignāga's tradition really defends is a different level. Its reply to Candrakīrti is that self-cognition is not a relation in which a subject knows an object but the mode of presence that cognition itself has (see below). It is phenomenology that carries this level to the end: Brentano says that every mental act has itself as a secondary object; Sartre says that consciousness has a non-positional consciousness of itself, failing which there is a regress; Zahavi calls it the "for-me-ness" of the minimal self.¹⁰ The dispute was fought once already in the nineteenth century: Comte said that the thinker cannot divide himself in two and observe himself; Mill replied that a mental fact can be studied through the medium of memory, the moment after it is perceived, and that most of our best knowledge of our intellectual acts is acquired in just this way; Brentano's secondary object was framed precisely as an answer to Comte.¹¹ This level is not a claim about later moments, and a model of later moments cannot refute it. All the Theravāda can answer is one sentence: that presence which seems to come with the moment is the consciousness's discernment of its object together with the very rapid subsequent processes, experienced as simultaneous. The two sides have names today, higher-order and same-order theories, and one collection puts them side by side: Thompson treats the memory argument directly, Siderits defends the Abhidharma reduction of subjectivity, Dreyfus establishes self-cognition on Dharmakīrti's basis, and Zahavi establishes the minimal self.¹² This book's judgment is that the regress argument fails, and that the level of presence is left to the opponent.

Candrakīrti's critique of self-cognition (5.5) targets precisely this argument. He says that a knife cannot cut itself, and consciousness cannot cognize itself. Dignāga's tradition replies that "self-cognition" here is not a cognitive relation in which a subject knows an object. It is an intrinsic self-illumination of cognition, the way cognition itself exists. This debate appeared in the discussion of the Madhyamaka–Yogācāra dispute in 6.5.

VI. Exclusion Theory (Apoha): How Words Refer¶

Exclusion theory (apoha-vāda), one of Dignāga's most original philosophical contributions, addresses a central question in the philosophy of language: how do words refer to things?

The background to the problem

Both Nyāya and Mīmāṃsā accept realism about universals. A word can refer to a class of things—"cow" to all cows, for example—because the members of that class share a real common property, "cowhood" (gotva). The word directly refers to this common property.

Dignāga rejects realism about universals, holding that only particular characteristics are real. But if universals do not exist, how can words refer? If "cow" does not refer to a universal property called "cowhood," how can it apply to all cows and to nothing else?

Dignāga's answer

A word does not work by positively referring to a common property. It achieves reference by excluding (vyāvṛtti) what falls outside its category. The word "cow" does not refer to an entity called "cowhood." It delimits its range of application by excluding everything that is not a cow: non-cows. This is "exclusion," the establishment of meaning through exclusion rather than positive reference.¹³

A detail of cultural history deserves a brief mention. In Indian philosophy and linguistics since Pāṇini, the "cow" (go, gavī) has been almost the default noun in philosophical examples. "Cowhood" (gotva) is the preferred example in Vaiśeṣika and Nyāya discussions of universals, and Dignāga's exclusion theory likewise uses "cow/non-cow." This clearly relates to the cow's exceptional standing in Indian culture. Sacred in Hindu tradition, it is the civilization's most important animal. In medieval European debates over universals, "horseness" (equinitas) held a comparable place as the standard example. Porphyry used the horse when posing the problem of universals in the Isagoge; the Islamic philosopher Avicenna systematized this usage, which European scholasticism then took up. The horse was medieval Europe's most important animal, indispensable to warfare, farming, and transport, mirroring the cow's position in India. Faced with the same philosophical problem, each civilization chose its most prominent animal as its example. This is a small observation in cultural anthropology, appearing at the center of a shared philosophical difficulty.

In modern terms, this is very close to a negative theory of semantics. A word's meaning lies not in the positive entity it refers to but in the negative set it excludes. "Cow" means "not non-cow": the range delimited by excluding non-cows.

The theory has the advantage of explaining why words apply generally to things of the same kind while denying that universals exist.

Its difficulty is this: if a word's meaning is established by excluding non-cows, how is the concept "non-cow" established? Kumārila Bhaṭṭa of the Mīmāṃsā school raised a famous objection. Delimiting "non-cow" seems to presuppose that we already know what a cow is; otherwise, we cannot know what to exclude. The account is circular. Dharmakīrti, discussed in the next chapter, further revised and defended exclusion theory by introducing causal efficacy to explain the connection between words and things. Whether this revision escapes the circle or merely relocates it to the classification of effects will be addressed directly in Section VI of the next chapter.¹³

The following comparison concerns ideas, not historical influence. Exclusion theory has a striking structural parallel in an independent development in Western philosophy of language, Ferdinand de Saussure's Course in General Linguistics (1916). Saussure's central claim is that linguistic signs have no positive content, only differences: "The characteristics of any linguistic concept derive not from its positive content but from its relations to the other elements in the system." In his famous formulation, "In language there are only differences without positive terms." This claim that linguistic meaning is purely differential corresponds almost exactly in structure to exclusion theory's claim that meaning is established through exclusion rather than positive reference. Both reject the idea that words directly point to positive universal entities. Both ground meaning in relations to others—difference or exclusion—and oppose realism about universals. They differ in their starting points, Dignāga beginning with ontological nominalism and Saussure with linguistic structuralism. Kumārila's charge of circularity against Dignāga—delimiting non-cows already presupposes knowledge of cows—also arises in Saussure's framework. If all signs are defined through differences, how does this system of differences acquire its first anchor? This unresolved difficulty confronts both theories.

VII. The Three Characteristics of a Reason: Reforming Logic¶

Another of Dignāga's major contributions to Indian logic was to establish the "three characteristics of a reason" (trirūpa-hetu), setting strict formal conditions for the validity of inference.

Traditional Nyāya used a five-member argument: thesis, reason, example, application, and conclusion. Dignāga reduced it to three members and established three necessary conditions for the reason supporting an inference:

Take the classic example: we infer that "there is fire on the mountain" because "there is smoke on the mountain."

• First characteristic, being a property of the subject: smoke is a property of the mountain. This is the starting point of the inference.
• Second characteristic, definite presence in similar cases: smoke occurs in established similar instances with fire, such as kitchens. This does not require every fire to produce smoke.
• Third characteristic, universal absence in dissimilar cases: the reason must be absent from every established instance without fire. There must be no counterexample of smoke without fire.

Only a reason that meets all three conditions can support a valid inference.¹⁴

Using modern notation, let H denote the reason and S the property to be proved. The required direction of pervasion is H→S; it cannot be reversed into S→H. Similar instances provide positive evidence, while dissimilar instances exclude counterexamples. How these domains are defined, and how the relation applies to the subject under examination, require further argument. This modern explanation does not imply that the literal conditions of the three characteristics form an undisputed system of sufficient conditions in modern logic.¹⁴

Two parallels in the Western logical tradition invite comparison with the structure of the three characteristics. The comparison concerns ideas, not historical influence.

The first is Aristotle's syllogism, dating to roughly the fourth century BCE. Aristotle too formalized the conditions of valid inference. A typical form is: "All human beings are mortal; Socrates is a human being; therefore Socrates is mortal." A conclusion follows from an accepted universal proposition and a particular premise. Two questions must be distinguished here: what follows from an accepted universal relation, and what establishes that relation. Both arise when discussing an inference from smoke to fire. The distinction between deduction and induction alone cannot set the two traditions as wholes in opposition.

The second is John Stuart Mill's inductive logic, set out in A System of Logic. His Method of Agreement requires the instances of a phenomenon under comparison to share only one relevant circumstance, from which its cause or effect is sought. His Method of Difference requires an instance where the phenomenon occurs and one where it does not to share all relevant circumstances except one. This identifies the cause, effect, or indispensable part of the cause.¹⁵ Both the three characteristics and these methods attend to positive and negative instances, which permits a limited comparison. Mill imposes further requirements for excluding or controlling other relevant circumstances, however. The two accounts cannot be treated as almost identical formal structures.

Dignāga also developed the "wheel of reasons" (hetucakra), a scheme that systematically arranges the possible distributions of a reason among similar and dissimilar instances, much like a truth table. Patterns 2 and 8 count as valid reasons. Pattern 8 permits the reason to occur in only some similar instances while being absent from all dissimilar ones. Definite presence in similar cases therefore does not require every instance of the property to be proved to possess the reason.¹

VIII. Challenging Non-Buddhist Epistemologies¶

From the outset, Dignāga's epistemology developed through competition with the Brahmanical schools.

The debate with Nyāya

Nyāya regarded logic as its particular strength. It accepted four means of valid cognition and grounded its epistemology in realism about universals. Dignāga's central criticisms concerned universals, means of valid cognition, and logical reasons. Universals are not independent realities; only particular characteristics are real. Nyāya's four means of valid cognition can be reduced to two, and its reasons fail to satisfy the genuine conditions of logical validity, the three characteristics.

Nyāya's principal respondent was Uddyotakara (c. 550–610), of the late sixth and early seventh centuries, who wrote the Nyāyavārttika to refute Dignāga point by point. This exchange, one of the most famous philosophical disputes between schools in Indian intellectual history, directly prompted Dharmakīrti's further developments, discussed in the following section.

The debate with Mīmāṃsā

Kumārila Bhaṭṭa (c. 600–700) of the Mīmāṃsā school was the principal critic of Dignāga's exclusion theory. He argued that it could not explain how we delimit "non-cow," since that delimitation seems to presuppose that we already know what a cow is. The account is therefore circular. Dharmakīrti's revision, introducing causal relations to explain the connection between words and things, responded precisely to criticisms of this kind.

IX. Historical Influence¶

In India

The epistemological tradition founded by Dignāga underwent continuous development within Indian Buddhism for more than two hundred years. His most important successor, Dharmakīrti (c. 600–660), built on his work to establish a more precise system of epistemology and logic, discussed in the following section. Commonly called the "Dignāga–Dharmakīrti tradition" or the "Buddhist epistemological tradition," it became a major scholarly school at Nālandā alongside Madhyamaka and Yogācāra.

In China

Xuanzang brought Dignāga's Nyāyamukha back from India and translated it. His disciple Kuiji wrote the Commentary on the Introduction to the Principles of Logic, laying the foundations of Buddhist logic in China. Its influence, however, was quite limited. It was studied as a specialized skill and did not enter into a deep philosophical synthesis with mainstream traditions such as Madhyamaka or Tiantai. Chinese Buddhism's main currents were tathāgatagarbha thought and Chan; the precise debates of Buddhist epistemology found no suitable place in this setting.

In Tibet

Dharmakīrti's major works, especially the Pramāṇavārttika, were systematically translated and studied, becoming a core subject in Tibetan monastic education. The Gelug tradition joined epistemology and logic to Madhyamaka as its view, treating them as work at different levels of a single path of practice. Logic addresses the conditions of valid conventional cognition; Madhyamaka analyzes emptiness at the ultimate level. This combination gave the Dignāga–Dharmakīrti tradition something like the status of a logical handbook in Tibetan Buddhism. Practitioners needed the tools of epistemology to defend their Madhyamaka positions in debate.

References¶