Dharmakīrti: a reason that establishes its conclusion

A light outside the kiln

Imagine a red lamp glowing beside a pottery kiln. Every time Anika has seen that light, the kiln has been hot. Today she arrives after everyone else has left. Can she know from the light that the kiln is hot now?

The lamp might respond to the heat. Or someone might switch it on whenever a firing begins, and occasionally forget to switch it off. Either arrangement could explain what she has seen. She recognizes the light, but what does it entitle her to conclude?

The question behind Anika’s situation is one that the Indian Buddhist philosopher Dharmakīrti pursued: what makes a sign a reason that yields knowledge? The opening of his Pramāṇavārttika, together with his own prose commentary, examines inference for oneself, that is, coming to know something through a sign. His concern is not confined to winning a debate; the work connects sound inference with distinguishing what is beneficial from what is harmful. [1]

This lesson follows one argument about the connection a reason needs and our grounds for recognizing it. How can we tell whether we know? and Nyāya: what warrants an inference? offer useful preparation, but you can also begin here; each term is introduced as it is needed.

The reason must belong to this case

Dharmakīrti begins with two requirements for a hetu, a reason or item of evidence in this inferential role. It must be a property of the subject under consideration, and it must be pervaded by the property to be established. Pervasion, vyāpti, means that the reason cannot occur without what it establishes. [1] [2]

For Anika, the subject is this kiln now; the proposed reason is its having a glowing indicator; the conclusion is that it is hot. A reliable indicator on the kiln next door would establish nothing about this one, because evidence must belong to the case the conclusion concerns. It need not be unique to that case, though: the same kind of evidence can occur elsewhere.

The second requirement governs the direction of the connection. Anika needs grounds for saying that this indicator cannot glow unless this kiln is hot. Knowing that the light always comes on when the kiln is hot would not be enough, since the light might also come on for another reason. The needed relation can be stated positively (whenever this indicator glows, the kiln is hot) or negatively (when the kiln is not hot, this indicator does not glow).

The two statements express the same restriction. Anika still has to establish that it holds.

Why accompanying is not enough

Dharmakīrti grounds a dependable connection in what the things are or how they arise. His term for such a connection, svabhāvapratibandha, is often rendered “natural relation.” Here “natural” does not mean familiar, usual or found outdoors; it points to a connection that makes the reason inseparable from what it establishes. John Dunne distinguishes this question from the further task of determining where such a relation holds. [2]

Two forms matter for affirmative inference. An effect can indicate the cause on which it depends, or what something is can warrant a further description of that same thing. In the text’s example, being a śiṃśapā, a kind of tree, is a reason for being a tree. This second form is commonly explained through identity of nature: the two descriptions differ, but they concern the same thing, not two objects that happen to occur together. Neither form lets us simply reverse the inference; being a tree, for instance, does not establish being a śiṃśapā. [3]

Our kiln case concerns the first form. A habit of switching on a lamp at firing time would explain Anika’s observations, but it would leave the light able to occur without heat. To use the light as an effect indicating heat, she needs grounds for excluding that possibility in the circumstances at issue. Calling the lamp an “indicator” does not supply them.

How much can an effect tell us?

Even a genuine causal connection tells us only so much. In his commentary on the second verse, Dharmakīrti asks which features of a cause can be established from its effect. The inference reaches only those features without which the relevant effect could not occur: knowing that something had a cause does not disclose everything about it. [3]

Suppose Anika establishes, for this imagined arrangement, that the indicator’s glow cannot occur unless the kiln is hot. She may then infer heat. She cannot infer that a particular potter lit the kiln, that blue bowls are inside or that the firing has finished, because the very same heat and light could accompany a different potter, different pottery or an unfinished firing. Each of those claims needs its own connection to the evidence.

The restriction also tells her how to look for stronger evidence. If she wants to infer a more specific feature of the firing, she needs a sign that depends on that feature; more confidence in the original observation cannot make it reveal more than its connection supports.

A cause produces an effect; evidence makes something known. These are different roles, even when a causal connection explains why an effect can serve as evidence. In our example, the kiln’s heat would help explain the light, while the observed light would give Anika grounds to infer the heat. The order of explanation and the order of discovery run in opposite directions.

A connection must be known

A relation might hold without Anika knowing that it does. That is why Dharmakīrti’s explanation of pervasion includes ascertainment: establishing, through means of knowledge, the reason’s presence in the subject and its positive and negative connections. The person making the inference needs grounds for using the sign in this way. [1]

This brings us back to Anika’s first observations. Seeing a red light beside a hot kiln fits several explanations, and even a long run of such sightings would leave her needing to tell a dependable connection from a shared routine. What would bear on that difference is evidence about how the indicator responds when the kiln is cold, what powers it and whether someone can operate it independently. These are questions for our invented case; they are not a procedure quoted from Dharmakīrti.

The investigation has two connected tasks: identify what would make the light dependable, and establish that this arrangement works that way. Describing the required relation guides the investigation; it does not establish that the relation holds here.

Hayes and Gillon press the second task as a difficulty for the theory, since knowledge of a causal connection needs support too. Dunne also distinguishes specifying a relation from establishing it, and notes disagreement over how Dharmakīrti’s procedure should be reconstructed. The question is what would establish the dependence, and raising it does not settle whether inference can yield knowledge. [4]

When absence is evidence

Anika’s investigation also raises a question about negative evidence: when does failing to find something show that it is absent? Dharmakīrti recognizes non-apprehension, or failure to perceive, as a kind of reason, but it warrants absence only when the relevant conditions for perceiving the thing are met. His example concerns a pot that would be perceptible at a specified place if it were there. [5]

Anika cannot establish that a shelf is empty by failing to see a bowl through a closed cupboard door. If she opens the door, lights the shelf and looks attentively, her failure to see a bowl means something different. The point is not that every unsuccessful search proves absence; it is that she needs grounds for thinking that the thing, if it were in the place searched, would have been found.

Likewise, “I have never noticed this lamp glowing beside a cold kiln” does not by itself show that the combination is impossible. Perhaps she has never entered the workshop after a canceled firing. A gap in observation and an established absence are different things.

Comparing the two accounts fairly

The Nyāya lesson examined Vātsyāyana’s account of remembering a known connection and perceiving the present sign, Uddyotakara’s explanation of reflection on that sign, and the five-part presentation for another person. Those accounts already require more than an orderly recital or repeated confidence: the relation must support the claim. [6]

Dharmakīrti’s opening puts the reason’s constraints in the foreground: it must belong to the subject, and a connection must make it inseparable from what it establishes. The two lessons reach a shared question by different routes: what makes the sign informative, and how do we know that it is? Both give experience a role and seek knowledge of what is true. What is worth comparing is how the selected accounts explain and establish the connection.

The previous lesson also introduced the much later Gaṅgeśa through Stephen Phillips’s scholarship. That later account is not an early commentator’s reply to Dharmakīrti. Our comparison places selected arguments beside one another for examination; it does not reproduce a historical conversation, and a fuller judgment about the traditions would require much more than these two lessons.

Study checkpoint: three questions about a reason

Return to the invented kiln. Anika argues: “Whenever this kiln’s indicator glows, the kiln is hot. Its indicator is glowing now. Therefore it is hot now.” Her evidence for the first statement is that light and heat occurred together on every firing she attended. She has not examined how the indicator works.

Begin with what follows. If both stated premises are true, can the conclusion be false? Explain how that question differs from asking whether Anika knows the kiln is hot.

Now suppose she discovers that the lamp has an independent switch and can glow while the kiln is cold. Which part of her support fails? Does the discovery prove that the kiln is cold today? And what would she need to establish to infer heat from a suitably restricted sign?

Finally, bring the two accounts together. How could the Nyāya presentation studied in the previous lesson help expose the problem? State Dharmakīrti’s demand concerning the connection, and offer a fair objection about how that demand is met, without claiming that either account simply trusts repeated observations or dismisses them altogether.

Read a checkpoint response

Following from the premises. If both premises are true, the conclusion must be true, provided we keep the same kiln, time and sense of “hot” throughout. The first premise rules out a glowing indicator without a hot kiln; the second says that the indicator is glowing now. Together they require the kiln to be hot now. This shows what follows from the premises, without establishing that either premise is true. Anika’s past observations may leave her without adequate grounds for the general claim, even if today’s conclusion happens to be true, and that is why knowing the kiln is hot is a separate question.

Testing the relation. What fails is the first premise, the claimed connection between the glow and the heat. Establishing that the light can occur beside a cold kiln defeats the asserted necessity of the connection, and an actual such case would also directly refute the unrestricted general statement. This does not show that the kiln is cold today: it may be hot while the light is on, just as Anika thinks. What she loses is the entitlement to infer heat from that light alone.

Restricting the sign. A repair needs evidence of heat that the independent switch cannot produce. Suppose investigation identified a mode in which the indicator could glow only in response to the kiln’s heat. Anika would then also have to establish that the indicator was operating in that mode now. The story has not established that such a mode exists. Adding “when the indicator is reliable” would merely rename the missing work. And even a warranted inference to heat would not identify the potter or prove the firing complete.

The comparison. In the selected Nyāya account, stating a proposition, reason, illustration, application and conclusion makes the roles of the sign, the supporting case and the present application available for examination. A familiar firing could illustrate Anika’s claim, but it cannot make a faulty relation sound. Dharmakīrti likewise requires the reason to belong to the subject and be inseparable from what it establishes. For an effect used as a reason, a natural connection through its production must warrant the move to the relevant cause. The presentation helps locate the demand; the claimed connection must still meet it.

A fair objection. A critic can ask what establishes the claimed dependence rather than mere co-occurrence. If the only answer is the same collection of unexplained observations, invoking causation has not yet answered the challenge. A defender owes grounds for distinguishing the dependable connection from the independent-switch possibility. This objection identifies what remains to be established. It does not prove that all causal knowledge is impossible or that Dharmakīrti had no further account. The checkpoint is a constructed comparison, and its lamp, investigation and proposed repair are original teaching examples.

Carry this back to an argument

A reason can fail by missing its subject, by lacking the required connection, or by being used without grounds for thinking that connection holds. Keeping those failures apart helps us criticize an inference without rushing to deny its conclusion.

Return to How can we tell whether we know?, or revisit the Nyāya lesson with the distinction between a relation and its ascertainment in mind. Argument: reasons and what they support offers a short refresher, and Testing a claim with a counterexample gives further optional practice.

Sources and limits

[1] Richard P. Hayes and Brendan S. Gillon, “Introduction to Dharmakīrti’s Theory of Inference as Presented in Pramāṇavārttika Svopajñavṛtti 1–10”, Journal of Indian Philosophy 19 (1991), 1–73, for Dharmakīrti’s translated text and its explanation.

Division A translates the first ten verses and Dharmakīrti’s own commentary; Division B is the translators’ explanation. The opening, A.1–A.1.5 and B.2.1–B.2.2.1.2 support the account of the subject, pervasion and ascertainment. The letters and decimal numbers identify sections of this article; the verse numbers follow Gnoli’s edition. All of its translation, explanation and five notes were read. No independent Sanskrit or Tibetan collation is claimed.

[2] John D. Dunne, Foundations of Dharmakīrti’s Philosophy, Wisdom Publications, 2004:

• Chapter 1: “Shared Notions Concerning Inference” through “A Restatement,” with notes 27–45.
• Chapter 3: the opening and the first part of §3.1 before “The Two Senses of Svabhāva,” with notes 1–19.

Dunne’s account distinguishes inferential knowledge from its verbal presentation and specifies the two tasks of describing and determining the required relation. These are selected passages, not a whole-book reading.

[3] Hayes and Gillon, A.1.6–A.2.2, pp. 4–6; B.2.4–B.3.2, pp. 39–56; notes 3–5, pp. 70–71.

Their explanation draws on later commentators, especially Karṇakagomin and Manorathanandin; it is not all Dharmakīrti’s own wording. Compare Dunne, chapter 3, §3.1 and note 18 on the second verse. “Natural relation” and “identity of nature” are explanatory English renderings whose interpretation is debated. The lesson uses only the distinctions needed for this inferential problem.

[4] Hayes and Gillon, B.3.3, pp. 56–59; Dunne, chapter 3, “The Determination of the Production-mode,” with notes 70–72. Dunne records a challenge to Gillon’s reconstruction of the causal procedure. The lesson keeps the question about ascertainment open without treating one reconstruction or criticism as a final scholarly verdict. It does not reconstruct the larger argument of Pramāṇavārttika I.11–38.

[5] Hayes and Gillon, A.1.6 and A.3–A.4.6; B.2.4.3 and B.4.1–B.4.4. The pot example is from the translated source. The cupboard, kiln and checkpoint are original examples; they illustrate the need to establish the relevant conditions, without reproducing the full theory of non-apprehension.

[6] Gautama and early commentators, The Nyāya-sūtra: Selections with Early Commentaries, translated and explained by Matthew Dasti and Stephen Phillips, Hackett, 2017: 1.1.5 and its selected Vātsyāyana/Uddyotakara commentary, pp. 28–33, notes 4–5; 1.1.32–39 with selected Vātsyāyana commentary and the editors’ explanation, pp. 180–186, note 1. These selections were reread for the comparison. The preceding Nyāya lesson supplies the separately attributed later Gaṅgeśa discussion and its bibliography. This lesson paraphrases its sources and does not present Dharmakīrti as a spokesman for all Buddhists.