How can we tell whether we know?

Confidence needs a connection to the world

Imagine you are at a harbor and someone says the last ferry has already left. You might see the empty berth for yourself, work out the departure from the timetable, or hear the news from a crew member. Each route can inform you, and each can go wrong: the boat may simply be out of sight, the timetable may have changed, or the speaker may be guessing.

The question is what makes a way of arriving at a belief a source of knowledge. Confidence alone cannot do that work. Nyāya philosophers look for the answer in the connections between things in the world and our cognition of them. “Cognition” here means an episode of awareness or thought, such as seeing a boat or understanding a report.

The Nyāya Sūtra, traditionally attributed to Gautama, belongs to an evolving tradition of inquiry and debate. Stephen Phillips places its root text around 200 CE, while noting that it may contain earlier and later layers. The text also connects the correction of error with release from suffering and rebirth, so its concern with knowledge belongs to a larger philosophical and religious project. [1] [2]

Four ways of coming to know

A pramāṇa is a means or source of knowledge: a process through which something is known correctly. The text names four, and each requires a connection with what is actually the case, which confidence alone cannot supply. [1]

• Perception: awareness arising through a sense’s contact with its object. The definition rules out error and unresolved uncertainty: seeing something hazy at the harbor is not yet seeing that it is a departing ferry.
• Inference: reaching knowledge through a sign whose connection with something else is established. Smoke, for instance, may indicate a fire that cannot be seen directly.
• Comparison: learning what a word applies to by connecting a description with a perceived likeness. In the traditional example, someone who has been told that a gavaya, a wild bovine, resembles a cow meets one and recognizes what the unfamiliar name denotes. Comparison in this sense is narrower than argument by analogy in general.
• Testimony: learning from the words of someone who knows and communicates reliably. A person’s title or certainty does not by itself make them such a source.

Nyāya does not demand that we investigate every perception or speaker before we can learn anything. Phillips describes an ordinary presumption of trust, with checking called for when a reasonable doubt arises. The crew member may tell us something we have no independent way of seeing; evidence that they were away all day would give us reason to examine the report. Trust and correction belong to the same account. [2]

The five steps of an inference

The text lists five parts of a demonstration: proposition, reason, example, application and conclusion. Here they are in a reconstruction of the traditional hill-and-fire case, which appears in the translator’s explanation of 1.1.32; Phillips uses the same case to explain the tradition. [1] [2]

1. Proposition: There is fire on that hill.
2. Reason: Smoke is rising from that hill.
3. Example with the general relation: Wherever there is smoke, there is fire, as in a kitchen hearth.
4. Application: This hill has smoke of that kind, so the relation applies here.
5. Conclusion: Therefore there is fire on that hill.

The first and last sentences say the same thing but do different jobs: the first states what needs establishing, and the last presents it as established by the reasons in between. The application does the connecting work, bringing a general relation to bear on the particular place under discussion.

The kitchen hearth does not prove the general relation just by being one convenient instance. Its job is to put the proposed connection where everyone in the discussion can examine it. Vyāpti, often translated as pervasion, is the relation on which the inference relies: the reason must not occur without the property being inferred. Here, smoke must be connected with fire in exactly that way. [2]

Each step can still be questioned. Is what we see actually smoke, or could it be mist? Is it coming from the hill? Have we really established the relation between smoke and fire? Five sentences in the right order can’t make up for a failure at any of these points, and Phillips stresses that our grounds for accepting the general relation can themselves be challenged and corrected. [2]

Recall lesson 2’s distinction between a conclusion following from premises and those premises being true. Nyāya is interested in inference that produces knowledge of the world, so it needs grounds for accepting both the reason and the general relation. A neat pattern is not enough.

Who checks the means of checking?

A serious opponent can press deeper. If perception is trustworthy because another cognition confirms it, what confirms that cognition? Repeating the procedure seems to launch an endless demand for further confirmation. Yet letting a means of knowledge vouch for itself looks like granting it a privilege that any disputed belief could claim.

The challenge, in other words, asks what could justify our trust in any means of checking. The Nyāya Sūtra takes it up at 2.1.16–19, where it considers means of knowledge becoming objects of knowledge and the demand for another means to establish them. The translator offers more than one interpretation of the reply, and we need not pretend that one short passage settles the whole dispute. [1]

Phillips explains one Nyāya response with the image of a weighing scale. A scale can establish the weight of gold; when the scale itself is in doubt, a known weight or another scale can test it. What served as an instrument becomes the object of inquiry, and its usual role does not exempt it from examination. [2]

The broader reply also rejects the requirement that every cognition be certified in advance. Inquiry normally begins with a specific reason for doubt, and it draws on whatever other sources are available to resolve it. On this view error remains possible, but we do not need an unchallengeable starting point before we can know anything.

An opponent can still ask whether successful checking in particular cases shows that the whole procedure is reliable. The scale example shows how checking can work; the opponent asks whether it goes far enough. At bottom, the two sides disagree about what justification must provide.

Pause and try

Someone argues, “There is fire on the hill, because something white is rising above it; kitchen fires produce white smoke.” Identify a weak step. Then explain why discovering that the plume is mist would not prove there is no fire anywhere on the hill.

Read an explanatory answer

The weak step is the reason, which identifies only a white plume: it may be mist or something else. The speaker has established neither that there is smoke on the hill nor that every white plume is dependably connected with fire. If the plume turns out to be mist, this particular inference loses its support, but there could still be a fire elsewhere on the hill, because the failure of an argument does not establish the opposite conclusion. What matters is what the available evidence warrants.

Carry this forward

Knowledge depends on how we arrive at a claim and on how we respond when its grounds are questioned. The next lesson asks what kind of knowledge could free a person from suffering, following selected Sāṃkhya and Yoga accounts of consciousness and the changing mind.

Sources and limits

[1] Gautama, The Nyāya Sūtras of Gotama, translated by Satis Chandra Vidyabhusana, The Sacred Books of the Hindus, volume VIII, Pāṇini Office, Allahabad, 1913. Selected sūtras and English explanations were read:

• The means of knowledge: 1.1.1–8, printed pp. 1–4.
• The five-part demonstration: 1.1.32–39, pp. 10–13.
• Knowing the means of knowledge: 2.1.16–19, pp. 27–28.

The five-part example on printed p. 10 was checked against the page image because its text extraction was empty. The reconstruction above follows that example rather than a different demonstration on p. 13, whose printed conclusion appears to contain an error.

The translation uses such terms as “means of right knowledge” and “word.” The lesson explains these as knowledge sources and testimony, and it does not adopt the edition’s “a priori” terminology for a kind of inference. The polemical note about Buddhism on p. 28 is the translator’s own account, neither a sūtra nor a fair summary of all Buddhist philosophy, and the lesson does not adopt it. The harbor is an original example; the hill exercise adapts the traditional case.

[2] Stephen Phillips, “Nyāya,” chapter 17 in History of Indian Philosophy, edited by Purushottama Bilimoria, Routledge, 2018, for the dating, the presumption of trust, pervasion and the scale. The chapter was read continuously from its opening through the discussion of knowledge sources, the five-part inference, testimony, comparison, the scale and responses to regress; later passages on ontology and theology were sampled. Its contemporary terms help explain the argument but are not exact equivalents of the Sanskrit terms. Phillips connects the skeptical challenge with Nāgārjuna; this lesson claims no separate reading of Nāgārjuna’s argument and no independent Sanskrit analysis.