Skip to content
Coordinates of ThoughtHome of Philosophy
Join
Question · Explain

One source of knowledge, two, four, five or six? Why Indian schools argued over the count

Wanjiru Kamau Contributor
· 5 replies · 151 views
Knowledge & beliefBuddhistIndianIslamicCarvakaMimamsaNyayaDharmakīrtiKumārila BhaṭṭaPrabhākara#comparative-epistemology#postulation#pramana#sources-of-knowledge
Follow

I build credit-scoring models, and my working epistemology is Bayesian: there is one way to learn, which is to update on evidence. So I was surprised to find that classical Indian philosophers organised epistemology around a different question, how many sources of knowledge (pramāṇa) there are, and gave different answers.

As I understand the scoreboard: the Cārvākas accepted perception alone; Buddhist logicians accepted perception and inference; Nyāya accepted four, adding comparison and testimony; the Prābhākara Mīmāṃsakas five; the Bhāṭṭa Mīmāṃsakas six.

What I don't understand is what the extra ones are, and what the argument is really about. Is this just bookkeeping, different ways of carving up the same thing? Or is there a real disagreement about whether some kinds of knowledge can be reduced to others? And if it's the second, wouldn't a Bayesian simply say everything reduces to updating on evidence, and side with the Buddhists?

LikeAnswer
Follow5 answers

Members reply here. Reading is always free.

Sign in to reply

5 answers

Accepted answer

  1. Meera Kulkarni

    ✓ Accepted answer Fellow

    It is the second, Wanjiru: the argument is about irreducibility. A pramāṇa counts as separate only if the knowledge it gives cannot be obtained by the others.

    • Perception and inference everyone (nearly) accepts.
    • Comparison (upamāna), for Nyāya: you are told a wild ox (gavaya) is like a cow; in the forest you see a cow-like animal and learn that this is what the word names.
    • Testimony (śabda): knowledge from the word of a reliable speaker. Buddhist logicians like Dignāga and Dharmakīrti argue it is really a kind of inference, from the speaker's reliability to the truth of what is said.
    • Postulation (arthāpatti), for both Mīmāṃsā schools: fat Devadatta does not eat by day, so he must eat at night. A fact is required to reconcile two known facts.
    • Non-apprehension (anupalabdhi), for the Bhāṭṭas only: you know there is no jar on the floor because you would see one if it were there.

    So each extra source is a claim that some knowledge does not fit the patterns of perception or inference. Your Bayesian reduction is itself a position in this debate, and it is the Buddhist one: everything non-perceptual is inference. The Mīmāṃsakas would ask you whether postulation, which resolves a conflict between facts, really is the same as updating a probability.

    Helpful · 1

Other answers

  1. Arvind Poojary

    Fellow

    A footnote on the Cārvākas, since they're usually summarised as "perception only". The sources we have are mostly hostile reports, and some later Cārvākas seem to have accepted inference for things that could in principle be perceived, while rejecting inference to things beyond any possible perception, like rebirth or heaven. So their dispute was less "no inference" than "no inference beyond the senses".

    Helpful · 2
  2. Govinda Prasad Mishra

    Fellow

    Let me defend postulation against reduction. Inference in the Nyāya model needs an established general rule, an invariable concomitance, like smoke with fire. Postulation needs no prior rule; it starts from an apparent conflict between two things we already know and supplies the one fact that removes the conflict. Wanjiru's world would call that inference to the best explanation. Whether that reduces to inference is exactly the argument.

    Helpful · 3
  3. Layla Haddad

    Fellow

    The Islamic thinkers asked a related question differently. Their pair was ʿaql and naql, reason and transmitted report, and the kalām theologians held that a report transmitted by so many independent chains that collusion is unthinkable (mutawātir) yields certain knowledge. That is a theory of when testimony becomes knowledge, built around the structure of the chain rather than the reliability of one speaker.

  4. Wanjiru Kamau

    Contributor

    This is wonderful, thank you. I see now that "everything is updating" is not neutral; it's a reductive thesis like the Buddhists'. And postulation is uncomfortably close to what my fraud models do when two records cannot both be true: we don't update so much as postulate the hidden fact. I may have been a Mīmāṃsaka all along.

    Helpful · 1