You read that the capital of Mongolia is Ulaanbaatar. You come to know it. By perception? You have not been there. By inference? No general rule entails it. By postulation? The world makes sense without you knowing this. You know it because a source told you, and the source was the kind that knows. The Indian schools call this cognitive route shabda - testimony - and treat it as a distinct pramana with its own validity conditions, not a degraded form of perception or inference.
The Nyaya analysis is precise. Valid testimony requires a reliable speaker - an apta, one who has true knowledge and the disposition to communicate it. It requires a coherent utterance: each word having a grasped meaning, the words combining grammatically, the combination interpretable. And it requires a competent hearer who receives the utterance correctly. When these three conditions hold, the hearer comes to know what the utterance says. When any one fails, the cognitive route fails - and the failure modes are structurally distinct.
A speaker who does not know the truth produces unreliable testimony - the failure is at the source. A speaker who knows but distorts produces lying testimony - the failure is in transmission. An utterance whose words do not combine produces incoherent testimony - the failure is in the message. A hearer who mis-parses produces misunderstood testimony - the failure is in reception. Each is a separate route to invalidity. The shabda frame keeps them straight. The English word “misinformation” collapses all four.
Where English Falls Short
Western analytic philosophy treated testimony as a second-class source of knowledge for most of the twentieth century - reductionist in its reflex, insisting testimony only works insofar as it can be backed by the hearer’s own inference. The Indian schools never had this prejudice. Shabda has been treated as a structured cognitive route, with its own validity conditions, for over two thousand years. Coady, Lackey, Goldberg, and Fricker arrived at similar territory only in recent decades - and even now, the Indian frame is more operationally developed.
The Western tradition asks the high-level question “is testimony reducible to other sources?” The Indian frame asks the working questions: is the speaker an apta on this topic? Is the utterance coherent? Is the hearer competent to receive it? Which of these is failing when the testimony fails? For practical work with AI output, expert claims, and information systems, the working questions are the useful ones.
Where it Shows Up
AI as a testimonial source. Much of what language models produce is testimonial in structure - they tell you things, and the underlying facts go unchecked. The shabda frame asks the right questions. Is the model an apta for what it is claiming to know? Is the output coherent and correctly parseable? Is the user positioned to receive it correctly? Different AI failures sit at different points in this chain, and diagnosing them requires knowing which point has failed.
Citation chains. “Source A says X, citing source B.” Validity propagates only if each link is reliable, coherent, and competently received. Citation chains break in characteristic ways: the original source did not say what is attributed to it (source failure); the citation distorts the original (transmission failure); the cited paper itself rested on bad testimony (recursive failure). The shabda frame makes each failure locatable.
Expert claims. “The expert says X.” The frame decomposes this: is this person an apta on this particular topic? Is what they said well-formed and interpretable? Was it received without distortion? Stakeholders often treat expert testimony as monolithic. The frame insists on locating which component to trust.
Research review. A paper reports findings based on interview data. The data is testimony from interviewees. The paper’s validity depends partly on whether the shabda conditions held at the interview level - were interviewees positioned to know? Did they communicate clearly? Did researchers interpret correctly? Each is a distinct link.
Onboarding and documentation. Documentation is testimony from past builders to future ones. The conditions for valid shabda - reliable speaker, coherent message, competent hearer - apply directly and are routinely violated in each direction. Most documentation failures trace to one of these three conditions.
Disinformation analysis. The shabda frame is operational at scale. A claim is invalid testimony if any condition fails. Most disinformation lives at the source-reliability layer (the speaker does not know or is lying). A non-trivial portion lives at the reception layer - the hearer mis-parses an utterance that was itself valid. These require different responses.
IKS Roots
The Sanskrit term is shabda (शब्द), literally “word” or “sound,” used technically for testimony as a pramana. Nyaya defines it at sutra 1.1.7: aptopadeshah shabdah - shabda is the instruction of a reliable person. The key term is apta (आप्त), the reliable one - one who has true knowledge and the disposition to communicate it accurately. The technical apparatus of shabda includes padartha (word-meaning), vakya (the sentence as a unit of meaning), and tatparya (the speaker’s intended import). The analysis of how words combine into meaningful utterances is a major branch of the tradition and overlaps with Vyakarana (see G3, G4). Mimamsa extends the treatment to apaurusheya utterance - testimony without a human author - which is the school’s specific doctrinal claim about the Vedas and does not bear on the operational use of shabda in general.
See N1 (the wider pramana frame), G3 (lakshana - secondary meaning, important when testimony is figurative rather than literal), and G4 (shaktigraha - how word-meaning is grasped, the reception side of shabda).
Further Reading
Bimal Krishna Matilal, Word and the World (Oxford), for the most accessible scholarly entry to the Indian theory of testimony. Arindam Chakrabarti’s essays on shabda in Knowing from Words (edited with Matilal). Jennifer Lackey’s Learning from Words for the Western testimony epistemology that independently arrived at similar territory. The Karl Potter Encyclopedia of Indian Philosophies volumes on Nyaya and Mimamsa for the canonical school positions.