Indian epistemology: the pramanas

In short

A *pramana* is a means or instrument of valid knowledge (*prama*) - how we come to know something truly, and how a claim is justified. Indian epistemology is built around disputes over how many pramanas there are (one for the Charvaka, two for Buddhism, four for Nyaya, six for Bhatta Mimamsa and Advaita), how each works, and how error and validity are explained. It is examined heavily in BA/MA Indian Philosophy papers, UGC-NET Philosophy Paper II, the UPSC Civil Services Philosophy optional (Paper I), CUET-PG and university viva, and the inference theory in particular reappears as logical reasoning in CLAT, CAT and campus aptitude rounds. Outside exams the material feeds Indian Knowledge Systems (IKS) teaching mandated under NEP 2020, work on argumentation and knowledge representation, and the law of evidence, where oral testimony and presumptions raise the same questions Nyaya and Mimamsa debated.

Explained

The framework: knowledge, knower, object, means

• Prama / pramiti - valid cognition. Nyaya defines it as *yathartha anubhava*: an apprehension of an object as it is, which is also novel (not mere memory).
• Pramata - the knowing subject; prameya - the object known; pramana - the instrument through which the object is known. Nyaya treats pramana as the *karana* (the most effective instrument) of valid cognition.
• Four-fold scheme of any knowledge episode: pramata, pramana, prameya, pramiti. Exam questions often ask for this *catustaya*.
• Valid versus invalid cognition: valid = *prama*; invalid (*aprama*) covers *samsaya* (doubt), *bhrama/viparyaya* (error), *tarka* (hypothetical reasoning) and *smrti* (memory - true but not novel, so not a pramana's product).
• Intrinsic or extrinsic validity: *svatah-pramanya* (Mimamsa, and Advaita for sruti) holds that a cognition is self-certifying and is accepted unless defeated; *paratah-pramanya* (Nyaya) holds that validity is established externally, by *samvada* - corroboration and successful practice (*pravrtti-samarthya*). Buddhism makes validity pragmatic: a cognition is valid if it does not mislead action (*avisamvadi*).
• Why the number matters: each school admits only those pramanas it cannot reduce to another. Nyaya rejects *arthapatti* as a kind of inference, Vaisheshika rejects *upamana* and *sabda* as inference, and the Charvaka rejects everything except perception.

Pratyaksha (perception)

• Nyaya Sutra 1.1.4 defines it as knowledge arising from the contact of a sense organ with its object (*indriyartha-sannikarsa*), which is *avyapadesya* (not derived from words), *avyabhicari* (non-erratic) and *vyavasayatmaka* (definite, not doubtful).
• Nirvikalpaka (indeterminate) - bare apprehension of the object without naming or predication; savikalpaka (determinate) - "this is a cow", with name, class and quality. Nyaya accepts both; some add pratyabhijna (recognition: "this is that same man").
• Dignaga and Dharmakirti (Buddhist) accept perception as *nirvikalpaka alone* - *kalpanapodham*, free of conceptual construction - its object being the unique particular (*svalaksana*); all conceptualisation belongs to inference and is conventional.
• Laukika (ordinary) perception requires one of six modes of sense-object contact (*sannikarsa*): samyoga, samyukta-samavaya, samyukta-samaveta-samavaya, samavaya, samaveta-samavaya, and visesana-visesyabhava (for absence).
• Alaukika (extraordinary) perception, admitted by Nyaya in three forms: *samanyalaksana* (perception of a universal, which supports the grasp of vyapti), *jnanalaksana* (one sense evoking another's content - "the sandalwood looks fragrant"), *yogaja* (the yogi's intuitive perception).
• Six senses: five external plus *manas* (mind), the internal organ by which pleasure, pain and cognitions are perceived; the self is known by mental perception in Nyaya, inferred in some other accounts.
• Strength and limit: perception is the most direct and is accepted by every school; but it cannot reach the past, the distant, the subtle, or dharma and moksa - which is exactly why the other pramanas are argued for.

Anumana (inference)

• Definition: knowledge of an unperceived thing through a perceived mark (*linga/hetu*) known to be invariably connected with it. The instrumental cause is *linga-paramarsa* - the consideration "this hill has smoke, which is invariably accompanied by fire".
• Three terms: *paksha* (subject - the hill), *sadhya* (what is to be proved - fire), *hetu/sadhana* (the reason - smoke).
• Vyapti - invariable concomitance between hetu and sadhya. Established by repeated observation (*bhuyodarsana*), supported by *tarka* (reductio) and, in Nyaya, by *samanyalaksana* perception; it must be free of conditions (*anupadhika*), an *upadhi* being a hidden extra condition that makes the generalisation fail.
• Classification by purpose: *svartha* (for oneself) and *parartha* (for convincing another, stated as a formal argument).
• Nyaya Sutra classification: *purvavat* (cause to effect - clouds, so rain), *sesavat* (effect to cause - flooded river, so rain upstream), *samanyatodrsta* (by general correlation, neither being cause of the other).
• Classification by vyapti: *anvaya-vyatireki* (both positive and negative concomitance available), *kevalanvayi* (only positive - "whatever is knowable is nameable"), *kevala-vyatireki* (only negative, no positive instance available).
• Panchavayava - the five-membered demonstrative syllogism of Nyaya: (1) *pratijna* thesis, (2) *hetu* reason, (3) *udaharana* general rule with example, (4) *upanaya* application, (5) *nigamana* conclusion. Buddhist logic compresses it to three (thesis, reason, example), holding the rest redundant.
• Trairupya - Dharmakirti's three marks of a sound reason: *paksadharmata* (the reason belongs to the subject), *anvaya* (positive concomitance), *vyatireka* (negative concomitance).
• Hetvabhasa - five fallacies of the reason (Nyaya): *savyabhicara/anaikantika* (inconstant, too wide), *viruddha* (contradictory, proves the opposite), *satpratipaksa* (counterbalanced by an equal opposing reason), *asiddha/sadhyasama* (unproved reason), *badhita/kalatita* (the thesis is defeated by a stronger pramana).
• Debate concepts: *vada* (honest discussion), *jalpa* (wrangling), *vitanda* (destructive cavil), *chala* (quibbling on words), *jati* (futile rejoinder), *nigrahasthana* (grounds of defeat) - from Nyaya Sutra Book 5.
• The Charvaka attack: vyapti can never be certified, since no pramana covers all unobserved cases - perception reaches only observed instances, inference would beg the question. Nyaya's answer is *tarka* plus the perception of universals; the problem parallels Hume's problem of induction and is a favourite comparative question.

Upamana and sabda

• Upamana (comparison/analogy) - Nyaya: on being told "a gavaya is like a cow", a person who then sees such an animal in the forest knows that *this is what the word "gavaya" denotes*. Its result is therefore knowledge of the denotation of a word through similarity.
• Mimamsa's version differs: on seeing the gavaya one knows the remembered cow's similarity to it - the object known is the similarity, not the word-meaning relation. Vaisheshika and Buddhism reduce upamana to perception plus inference or testimony.
• Sabda (verbal testimony) - *aptavakya*, the statement of a trustworthy authority who knows the matter and reports it truthfully. Knowledge comes from understanding the sentence, not from inferring the speaker's reliability (which is why Nyaya keeps it separate from anumana).
• Four conditions of understanding a sentence: *akanksha* (syntactic expectancy of words for one another), *yogyata* (mutual compatibility of meanings - "he wets it with fire" fails), *asatti/sannidhi* (proximity, words uttered without undue break), *tatparya* (the intended meaning, fixing the sense of ambiguous words by context).
• Types: *vaidika* (scriptural) and *laukika* (secular); *drstartha* (verifiable content) and *adrstartha* (not verifiable, e.g. heaven, dharma).
• Mimamsa: the Veda is *apaurusheya* (authorless) and therefore free of human defect; its injunctions are self-valid and are the only pramana for dharma. Nyaya instead grounds the Veda's authority in Ishvara as a perfect speaker.
• Theories of sentence meaning: *abhihitanvayavada* (Kumarila Bhatta - words first yield their own meanings, which are then related) versus *anvitabhidhanavada* (Prabhakara - words denote meanings already as related); Bhartrhari's *sphota* holds the sentence meaning to be an indivisible whole.

Arthapatti, anupalabdhi, and the full school-wise map

• Arthapatti (postulation) - assuming the only fact that would explain an otherwise inexplicable one: Devadatta is growing fat yet is never seen eating by day, therefore he must eat at night. Accepted by Mimamsa and Advaita; Nyaya reduces it to a *kevala-vyatireki* inference.
• Anupalabdhi / abhava-pramana (non-apprehension) - absence is known by the non-perception of what, if present, would have been perceived: "there is no jar on the floor". Kumarila Bhatta and Advaita accept it as a distinct pramana; Nyaya makes absence an object (*prameya*) perceived through *visesana-visesyabhava* contact, hence no new pramana.
• School-wise count (give this as a table in the exam): Charvaka 1; Buddhism and Vaisheshika 2; Samkhya, Yoga, Visistadvaita (Ramanuja) and Dvaita (Madhva) 3; Nyaya 4; Prabhakara Mimamsa 5; Bhatta Mimamsa and Advaita Vedanta 6; some Pauranika writers 8, adding *sambhava* (inclusion - knowing a hundred is in a thousand) and *aitihya* (tradition/hearsay), both of which Nyaya reduces to inference and testimony.
• Jainism classifies knowledge differently: *pratyaksha* (direct) and *paroksha* (indirect), with the five knowledges *mati, sruta, avadhi, manahparyaya, kevala*; its *anekantavada*, *syadvada* (sevenfold conditional predication) and *nayavada* (standpoints) make every judgement perspectival.
• Advaita's special use: the six pramanas operate only within *vyavaharika* (empirical) reality; Brahman is known by *sruti* alone, and on realisation the pramana-prameya distinction itself is sublated.

Error, illusion and the theories of knowledge-value

• Khyativada (theories of error) - standard comparative question: *anyathakhyati* (Nyaya - a real object elsewhere is misapprehended here, "rope-snake" is a real snake remembered and misplaced), *akhyati* (Prabhakara - error is only a non-discrimination of two valid cognitions, so no cognition is intrinsically false), *atmakhyati* (Yogacara - the content is only a projection of consciousness), *asatkhyati* (Madhyamika - the object is simply non-existent), *anirvacaniyakhyati* (Advaita - the illusory object is indeterminable, neither real nor unreal, created by *avidya*).
• Viparita-khyati and the Mimamsa claim that all knowledge is true as given: falsity is detected only later, by a contradicting cognition.
• Tests of truth used in practice: correspondence (Nyaya - cognition answers to the object), coherence and non-contradiction (*abadhita*), and pragmatic success (*arthakriya* - Buddhist; *samvada* - Nyaya).
• Comparison with Western epistemology: the pramana approach is reliabilist and causal - it asks which process produced the belief - rather than asking for an internal justification, which is why B. K. Matilal and J. N. Mohanty read Nyaya as anticipating responses to the Gettier problem; *prama* requires a non-defective instrument, not merely true belief with a reason.

Key points

• Pramana - an instrument/means of valid knowledge; prama - the valid cognition it yields.
• Catustaya - pramata (knower), pramana (means), prameya (object), pramiti (resulting knowledge).
• Nyaya Sutra 1.1.3 - lists exactly four pramanas: pratyaksha, anumana, upamana, sabda.
• Pratyaksha - perception from sense-object contact; definite, non-erratic, not word-derived.
• Nirvikalpaka / savikalpaka - indeterminate versus determinate perception; Buddhists allow only the first.
• Alaukika perception - samanyalaksana, jnanalaksana, yogaja (Nyaya only).
• Anumana - inference through a mark; its instrument is *linga-paramarsa*.
• Vyapti - invariable, unconditional concomitance of hetu and sadhya; the heart of inference.
• Panchavayava - pratijna, hetu, udaharana, upanaya, nigamana; Buddhist logic uses three.
• Trairupya - paksadharmata, anvaya, vyatireka: Dharmakirti's test of a valid reason.
• Five hetvabhasa - savyabhicara, viruddha, satpratipaksa, asiddha, badhita.
• Upamana - knowledge of a word's denotation (Nyaya) or of similarity (Mimamsa), through analogy.
• Sabda - testimony of an *apta*; needs akanksha, yogyata, asatti and tatparya to be understood.
• Arthapatti - postulating the only explanation of a known fact (Mimamsa, Advaita).
• Anupalabdhi - knowing absence by non-apprehension (Bhatta Mimamsa, Advaita).
• Svatah vs paratah pramanya - intrinsic (Mimamsa) versus extrinsic, corroboration-based (Nyaya) validity.
• Khyativada - anyathakhyati, akhyati, atmakhyati, asatkhyati, anirvacaniyakhyati: five theories of illusion.
• Charvaka's challenge - only perception is a pramana, because vyapti cannot be established without circularity.

Examples

Case 1 - Setting out a full Nyaya inference (parartha anumana). Claim: the hill has fire.

1. *Pratijna:* The hill has fire.
2. *Hetu:* Because it has smoke.
3. *Udaharana:* Whatever has smoke has fire, as in a kitchen (positive example); whatever has no fire has no smoke, as in a lake (negative example).
4. *Upanaya:* This hill has smoke that is invariably accompanied by fire.
5. *Nigamana:* Therefore the hill has fire.

Terms: paksha = hill; sadhya = fire; hetu = smoke; vyapti = smoke-fire concomitance; sapaksha = kitchen; vipaksha = lake. The same content as *svartha anumana* would be a single mental step - the paramarsa - without the five members.

Case 2 - Naming the fallacy. Identify the hetvabhasa in each:

Case 3 - Arthapatti versus anumana. Fact: Devadatta is alive but not at home. Postulation: he must be somewhere else, since being alive and not here is otherwise inexplicable. Mimamsa counts this a separate pramana because no previously known vyapti is used; Nyaya recasts it as *kevala-vyatireki* inference - "whatever is not elsewhere is not alive-and-absent-here" - with only a negative example available. Both reach the same conclusion; the dispute is purely about the number of pramanas.

Case 4 - Anupalabdhi in operation. I enter a room and do not see the jar on the floor. Steps: (i) the jar, if present, is of a kind normally perceptible (*yogya-anupalabdhi*); (ii) the conditions of perception - light, sense organs, attention - are intact; (iii) yet no jar is apprehended; (iv) therefore "there is no jar here". Bhatta Mimamsa: a sixth pramana. Nyaya: ordinary perception of a qualified absence, the floor being perceived as qualified by jar-absence. Note the condition: non-apprehension of an imperceptible thing (an atom, a ghost) proves nothing.

Case 5 - School-wise comparison of pramanas.

Where it is used

• Competitive and PG examinations: UGC-NET/JRF Philosophy Paper II, UPSC Civil Services Philosophy optional Paper I (Indian philosophy section), CUET-PG and state PSC philosophy and Sanskrit papers - pramana theory is a near-certain question every cycle.
• Teaching and Indian Knowledge Systems: NEP 2020 pushed IKS content into general education; AICTE/UGC IKS curricula use pramana theory as the Indian account of method and evidence, taught in central universities, Sanskrit universities (e.g. the campuses of Central Sanskrit University, formerly Rashtriya Sanskrit Sansthan) and IIT humanities departments.
• Law and evidence practice: direct versus hearsay evidence, the weight of an expert's statement, and presumptions of fact mirror sabda-pramana and its conditions; the Bharatiya Sakshya Adhiniyam, 2023 (replacing the Indian Evidence Act, 1872) retains the oral-evidence-must-be-direct rule and the expert-opinion provisions, and law entrance papers set analogy and inference questions directly.
• Logical reasoning in placement and entrance tests: syllogisms, "too wide a reason", counterexamples and analogy items in CAT, CLAT, bank PO and IT aptitude rounds are the same structures as vyapti and hetvabhasa.
• Research, editorial and policy roles: Indian Council of Philosophical Research projects, journal and translation work on Navya-Nyaya, and argument-mapping or fact-checking roles where the question is which source counts as *apta*.
• AI and knowledge representation: reliability of sources, defeasible inference and testimony-based belief in Indian AI/NLP research, often cited as a precedent for source-ranking and provenance in retrieval systems.

Interview questions

• Q: What exactly is a pramana? A: The instrument or means by which valid knowledge is produced - distinguished from the knower, the object, and the resulting cognition.
• Q: How many pramanas does Nyaya accept, and on what authority? A: Four - perception, inference, comparison and verbal testimony - listed in Nyaya Sutra 1.1.3.
• Q: Why does the Charvaka accept only perception? A: Because vyapti, the universal concomitance that inference needs, cannot be established without circularity, perception reaching only observed cases.
• Q: Distinguish nirvikalpaka from savikalpaka perception. A: The first is bare apprehension without name or predication; the second is determinate, as "this is a cow". Buddhists accept only the first as perception.
• Q: What is the instrumental cause of inference in Nyaya? A: *Linga-paramarsa* - the cognition of the mark in the subject together with its invariable connection to what is to be proved.
• Q: State the five members of the Nyaya syllogism and how Buddhists differ. A: Pratijna, hetu, udaharana, upanaya, nigamana; Buddhist logic keeps only thesis, reason and example, treating the others as repetition.
• Q: What is an *upadhi* and why does it matter? A: A hidden condition that makes a supposed vyapti fail; a genuine vyapti must be *anupadhika*, otherwise the inference commits savyabhicara.
• Q: How does upamana differ in Nyaya and Mimamsa? A: Nyaya makes its result the knowledge of what a word denotes; Mimamsa makes it the knowledge of similarity between the perceived and the remembered object.
• Q: What four conditions must a sentence satisfy to convey knowledge? A: Akanksha, yogyata, asatti and tatparya - expectancy, compatibility, proximity and intended sense.
• Q: Explain svatah-pramanya versus paratah-pramanya with the schools that hold them. A: Mimamsa holds validity intrinsic and self-certifying unless defeated; Nyaya holds it extrinsic, established by corroboration and successful action.
• Q: Give the Nyaya and the Advaita explanation of the rope-snake illusion. A: Nyaya's *anyathakhyati* says a really existing snake, remembered, is misapprehended in the rope; Advaita's *anirvacaniyakhyati* says an indeterminable object is projected by avidya.
• Q: Is memory a pramana? A: No - in Nyaya valid cognition must be a fresh apprehension of the object, and memory only reproduces an earlier one, though it is not thereby false.

Common mistakes

• Treating pramana as "knowledge" itself; it is the *means*, with prama as the result - the distinction is often what the marks are for.
• Attributing four pramanas to every orthodox school; Vaisheshika has two, Samkhya and Yoga three, Bhatta Mimamsa and Advaita six.
• Confusing the hetvabhasa: *savyabhicara* is a reason that is too wide, *asiddha* is an unproved reason, *badhita* is a thesis defeated by stronger evidence.
• Writing the five members out of order, or omitting *upanaya* - the step that applies the general rule to the present subject.
• Equating the Nyaya syllogism with the Aristotelian one; Nyaya requires an actual example (*drstanta*), so it is not formal validity alone.
• Mixing up the khyativadas, especially akhyati (Prabhakara) with anyathakhyati (Nyaya), and assigning asatkhyati to Yogacara instead of Madhyamika.
• Claiming that Mimamsa's self-validity means no error is ever possible - it means validity is presumed until a contradicting cognition defeats it.

Related topics

Orthodox schools: Nyaya, Vaisheshika, Samkhya, Yoga, Mimamsa and Vedanta · Heterodox schools: Charvaka, Buddhism and Jainism · Vedic and Upanishadic thought · Indian logic and theories of meaning · Western epistemology: knowledge, justification and scepticism · Law of evidence and legal reasoning