Nyāya: what warrants an inference?

The step beyond what we see

Smoke is rising from a hillside, but the fire is hidden. You have seen smoke coming from fires before. What entitles you to move from those occasions to a claim about this hillside?

A familiar association may suggest an answer, but to give us knowledge, the inference must connect what we see with what is actually there. Warrant means the grounds that entitle us to make that move. The Nyāya-sūtra, traditionally attributed to Gautama, and its commentators ask how such grounds work: how an inference arises in a person, and how it can be set out for someone else to examine. [1] [2]

How can we tell whether we know? introduces this tradition and its sources of knowledge. This lesson follows a single problem: what establishes the connection between a sign and what it is supposed to show? Each term is explained when we need it.

What the five parts accomplish

In sūtras 1.1.32–39, the text sets out five parts of an argumentative presentation. A proposition states what is to be established; a reason supplies the sign; an illustration exhibits the connection in a supporting case; an application connects the present subject with that support; and a conclusion gathers the case together. Vātsyāyana’s commentary explains how the parts depend on one another: each helps a hearer understand how the others bear on the claim. [2]

In the familiar smoke example, the hillside is the subject, smoke is the sign, and fire is what we seek to establish. A kitchen hearth supplies a case in which smoke and fire are already known together, and applying that connection to the smoky hillside leads to the conclusion that fire is there. Dasti and Phillips use this example to introduce the five parts; Vātsyāyana’s translated discussion chiefly uses a different argument, that sound is impermanent because it is produced.

Each part gives the listener something to examine. Is the sign present here? Does the example support the claimed connection? Does the present case fall under it? Reciting all five parts leaves those questions open, and an orderly argument can still lack the support its conclusion needs.

This Nyāya account does not prescribe a form for every Indian philosopher. It also asks more of an argument than the modern test of deductive validity does. That test asks whether an argument’s premises could be true while its conclusion is false; here we must also ask how the premises and their connection are known. [4]

A connection with a direction

The connection is called vyāpti, usually translated as pervasion. For the smoke inference, fire pervades smoke: wherever the relevant smoke occurs, fire must be present as its source. Smoke is the narrower sign; fire is what that sign warrants us in inferring. The claim reaches beyond the hearths we have already inspected. [3]

Notice the direction. If smoke warrants fire, it does not follow that fire warrants smoke. Finding fire without smoke would challenge that second, reversed relation and leave the first untouched; to challenge the first, we would need smoke without the fire it is supposed to indicate. The two relations call for different countercases.

We must also keep the sign steady. Begin with actual smoke and finish with anything that looks smoky, and we have changed the claim. Smoke carried from elsewhere, too, would send us to locate its source rather than place a fire wherever the smoke happens to drift. The question is what this sign establishes when it is correctly identified and situated.

Seeing the sign as a reason

Sūtra 1.1.5 connects inference with previous perception. Vātsyāyana explains that two perceptions contribute: an earlier perception of the connection between a sign and what it indicates, and a perception of the sign now. Remembering the connection lets the present sign disclose something that is not itself seen. [1]

Uddyotakara develops this account through liṅga-parāmarśa, reflection on the sign. Seeing smoke brings the remembered connection into play, and the person recognizes this smoke as indicating fire. As Dasti and Phillips explain, present perception and remembered experience must work together: knowing a connection and seeing a sign help us infer only when we bring them together in this case.

This explains how a known connection can guide us in a new case. But we can also faithfully remember a connection we were wrong to accept. What established it in the first place?

How far can experience take us?

Vātsyāyana’s discussion of the illustration relies on experienced connections, while Uddyotakara distinguishes inferences supported by positive and negative cases. For smoke and fire, positive cases show the two together, and negative cases show the absence of smoke where fire is absent. That negative evidence differs from finding fire without smoke; here too, the direction of the proposed relation matters. [1] [2]

The hearth gives us a familiar connection to examine, but why does that connection extend to the hillside? More observations can strengthen the case, yet a growing tally does not by itself tell us what makes the connection hold. Some further condition might have been present every time, unnoticed. Until we identify what matters, we may carry the example beyond what it supports.

Stephen Phillips describes how the much later Nyāya philosopher Gaṅgeśa approaches this problem: through wide experience, through positive and negative comparisons, and through tarka, suppositional reasoning. Tarka tests a competing supposition by drawing out its consequences; it supports inquiry without being an independent source of knowledge.

Gaṅgeśa challenges someone who doubts a causal connection while regularly relying on it in action, for instance by using fire to obtain smoke. The challenge asks whether that doubt is coherent; it does not supply an observation of every possible case. Phillips also stresses that repeated observations do not guarantee knowledge of a pervasion, since a connection accepted as known may meet a counterexample later. [5]

Gaṅgeśa’s later account helps us separate two tasks: establishing a relation, and answering a reason to doubt it. It should be kept distinct from the early commentators’ explanations. Phillips’s broader treatment also emphasizes our fallibility in taking a pervasion to be known. If an apparent inference misleads us, its result was not knowledge after all; what was mistaken was our confidence that we had knowledge. [4]

What a countercase reveals

Now consider the reversed claim: wherever there is fire, there is smoke. Dasti and Phillips explain a traditional objection to it through a hot iron ball, understood in Nyāya as bearing fire but giving off no smoke. The example belongs to that account of heat and fire; it is not a claim that modern chemistry classifies hot iron as burning. [3]

The editors distinguish this counterexample from an upādhi, an additional condition that the proposed relation overlooks; in their example, that condition is wet fuel. The iron ball shows the unrestricted claim failing, while the missing condition explains why fire alone was not enough for the inference to smoke. The countercase and the condition do different work.

This is why piling up successful cases may leave the defect untouched. If every fire we observed burned the relevant fuel, every observation could appear to confirm the unrestricted claim. It takes a different case to reveal that we credited fire alone with what depended on something more. A repair must then state the narrower relation and establish that its condition is met in the present case; merely adding another sentence to the presentation does neither.

Pause and try

Imagine that Lucia finds the courtyard paving wet and reasons: “It rained here, because the paving is wet. Wherever the paving is wet, it has rained, as on the ten mornings I checked. This is wet paving of that sort. Therefore it rained.” She has never checked the courtyard on a dry-weather cleaning day.

Identify the subject, the sign and the conclusion, and state the direction of the relation Lucia needs. Suppose you found that a hose had wetted the paving on a morning without rain: what would that establish? Then explain what remains to be investigated about today, and why a five-part presentation cannot settle it.

Read an explanatory answer

The subject is this courtyard on the occasion in question; wet paving is the sign; rain here is what Lucia wants to establish. She needs a connection that runs from wet paving to rain: this wetness could not have occurred without rain here. Knowing that rain wets the paving runs the other way. And ten mornings on which both occurred do not by themselves establish the connection she needs.

A dry-weather occasion on which a hose wetted the paving would supply wet paving without rain, and so refute the unrestricted relation Lucia needs. It would not show that it did not rain today: rain, a hose or both might account for today’s wet paving. To decide that, we need evidence about the present occasion.

Lucia could narrow her claim to circumstances in which rain is the only possible source of this wetness, but she would then need evidence that those circumstances apply today; simply failing to see anyone use a hose would not establish that no other source was involved. Her five-part presentation helps us locate the unsupported step, but its order supplies no evidence to repair it. Presenting a reason, in short, is different from establishing its warrant.

Connections and return

Return to How can we tell whether we know? with three questions: what sign is present, what relation would make it informative, and what establishes that relation here? Argument: reasons and what they support offers a short refresher on the vocabulary, and Testing a claim with a counterexample provides further practice in telling a failed reason from a false conclusion.

The next reading, Dharmakīrti: a reason that establishes its conclusion, takes up the relation itself: what could make a sign inseparable from what it indicates, and how could we know that it is? Keep this lesson’s distinction in view. Presenting a reason makes its support available for examination; warrant depends on what that examination can establish.

Sources and limits

[1] Gautama and early commentators, The Nyāya-sūtra: Selections with Early Commentaries, translated with introduction and explanatory notes by Matthew Dasti and Stephen Phillips, Hackett, 2017, for inference and earlier perception: sūtra 1.1.5 and all its selected commentary and explanation, pp. 28–33, including notes 4–5. The passages are attributed to Vātsyāyana and Uddyotakara, whose alternative interpretations of the sūtra are not a single, undifferentiated account.

[2] Dasti and Phillips, chapter 9, pp. 180–186, for the five parts: the introduction to them, 1.1.32–39, and the complete selected Vātsyāyana commentary, with note 1.

That note records the translators’ adoption of an alternative Sanskrit reading at 1.1.32. The edition’s bracketed numbers identify pages and lines in Anantalal Thakur’s Sanskrit editions, not pages of this translation. No independent Sanskrit or manuscript comparison is claimed.

[3] Dasti and Phillips, for pervasion and the additional condition: the explanatory discussions of pervasion at p. 127 and of the additional condition and hot iron at p. 132, and the glossary entries upādhi and vyāpti.

The explanation at p. 132 accompanies a passage reporting an opponent of Vācaspatimiśra. The fire, wet-fuel and iron example used here is the modern editors’ explanatory account of a traditional example. This lesson does not reconstruct the theological debate surrounding that passage. The volume is a selection, and only the specified passages and bounded introductory material were read.

[4] Stephen Phillips, “Nyāya,” chapter 17 in History of Indian Philosophy, edited by Purushottama Bilimoria, Routledge, 2018, for the wider Nyāya account and its emphasis on fallibility. It was read continuously from the opening through the discussion ending with tarka as distinct from an independent knowledge source. This is specialist interpretation, not a fresh reading of every author it discusses.

[5] Stephen Phillips, “Gaṅgeśa,” Stanford Encyclopedia of Philosophy, revision of 4 June 2024, §3.1 and §§5.1–5.3 (the symbol § marks a numbered section), for the bounded account of later inquiry into pervasion. Gaṅgeśa’s Sanskrit work and the whole encyclopedia entry were not read. All source passages are paraphrased; the courtyard case and its answer are original teaching material.