Abstract

Fallibilism requires, at minimum, a witnessing consciousness capable of registering successive cognitions and recognising their incompatibility. This article argues that positing a merely conscious but fallible witness is insufficient, since it relocates rather than resolves the infinite regress problem. The sākṣī of Dvaita vedānta, whose infallibility pertains not to cognitive content but to the act of disclosure itself, provides the stable ground that fallibilism's own operations presuppose. Without the witness, the ad infinitum structure of fallibilism is rendered unstable since an endlessly revisable cognitive chain lacks the stable point from which error can be recognised. Accordingly, the coherence of fallibilism itself rests on an implicit presupposition of a stable ground of awareness (the sākṣī) whose illumination conditions the intelligibility of all cognitive operations and their continual revision.

Introduction

Fallibilism is an epistemic stance that denies the existence of any intrinsically authoritative sources of knowledge or infallible methods for establishing truth. It holds that all knowledge consists of conjectures that must remain open to rigorous criticism and testing. Explanations that withstand such scrutiny are accepted only provisionally valid until replaced by better explanations. Thus, fallibilism treats both knowledge and indeed the methods used to evaluate it as perpetually revisable, embracing an open-ended, ad infinitum process of improvement[1]. This position is most systematically developed in Karl Popper's account of knowledge growth through conjecture and refutation, where learning occurs not by verification but by the elimination of error[1]. Yet this process of revision itself presupposes something: a subject capable of registering one cognition as replacing another. The question of what grounds this registering capacity is one fallibilism does not address.[1]

The gap in fallibilism is not in its account of knowledge-episodes or their revision, but in its silence about the subject to whom revisions appear. Popper explicitly argues that the question "how do you know?" is the wrong question -- what matters is whether a claim can withstand criticism.[1] This refusal to ask about the knowing subject is methodologically deliberate: Popper later develops this into a full epistemology without a knowing subject, positing a World 3 of objective knowledge that exists independently of any mind.[2] But this displacement of the subject does not dissolve the problem. Criticism must be conducted by someone, and the recognition that a claim has failed to withstand criticism must occur to someone. For a correction to count as a correction, an earlier cognition must be held in relation to a later one, and their incompatibility must be registered by something. A purely computational chain of knowledge-episodes has no standpoint from which this comparison occurs. This requires, at minimum, a witness. This is what Dvaita vedānta calls the sākṣī.

Thus, the epistemology of Dvaita Vedānta diverges at a fundamental level by positing the sākṣī (the inner witness) as a non-conjectural, illuminative condition underlying every cognitive episode. While the mind generates, processes, and revises knowledge, these operations are manifest only insofar as they are revealed by the sākṣī. As the sākṣī neither infers nor evaluates, it is not susceptible to error; its role is solely to disclose cognitive episodes with unconditional immediacy.

The two epistemic systems agree that any established knowledge is revisable upon availability of contradictory facts. Popper himself acknowledges that no test occurs in a vacuum -- every refutation presupposes a body of background knowledge that is, at least temporarily, held fixed.[1] This is an implicit concession that the ad infinitum revisability of fallibilism requires a practical anchor. However, fallibilism does not account for the stable vantage point from which errors can be recognised. An endless chain of conjectures and refutations, taken alone, collapses into an infinite regress with nothing to anchor the process of revision. A witnessing consciousness is the minimum required to avoid this collapse. However, a merely conscious but fallible witness relocates rather than resolves the problem. If the witness itself can misregister cognitions, a second-order witness is needed to catch those errors, and the regress resurfaces at the level of witnessing itself. The sākṣī of Dvaita vedānta addresses this by being infallible not with respect to the content of cognitions (those remain revisable) but with respect to the act of disclosure. That a cognition is occurring cannot itself be doubted without generating a further cognition, which again requires disclosure.[3] This structural infallibility is what distinguishes the sākṣī from a merely conscious registering mechanism.

It should be noted that this argument is scoped specifically to the Popperian formulation of fallibilism. Other fallibilist traditions, such as the Peircean account of self-correcting community inquiry or Quine's holistic web-of-belief revision, handle the knowing subject differently and would require separate treatment.

Terminology

  1. Knowledge

Knowledge is the awareness of factual propositions that describe the real world.

  1. Means of knowledge

A means of knowledge refers to the mode of receiving data that can be processed to derive meaning. 

  1. Knowledge-episode

A knowledge-episode refers to the internal process by which data received through any means of knowledge is organised, interpreted, and mentally processed into a meaningful form. It includes:

  • the initial presentation of data from a means of knowledge (such as perception, inference, testimony),

  • the mind’s interpretive operations (comparison, synthesis, conceptualisation, creativity),

  • any background assumptions or logical steps implicitly used,

  • and the mental consolidation of this content.

In short, a knowledge-episode is the full mental construction that results from processing received data. It is the structured content that constitutes “what is to be known,” including the internal logic or reasoning through which the mind processes the input. Thus, a knowledge-episode is a computational process. However, it does not yet include the conscious awareness “I know this.” It is the content that can be revealed to a subject.

  1. Cognition

A cognition is the moment of revelation in which the knowledge-episode becomes consciously manifest to a subject. It is the experiential event in which:

  • the content becomes illuminated,

  • the subject becomes aware of it,

  • and the notion “I know” arises.

Thus cognition is not the processing, but the disclosure of what has been processed.

A cognition is marked by:

  • immediacy,

  • self-presentation,

  • and the sense of acquaintance with the content.

  1. Witness (Sākṣī)

Witness is the subject of cognition. It is ‘within whom the cognition integrates as knowledge’.

Mechanism of Knowledge Generation

A means of knowledge provides data that can be processed into meaning. This data is organised and interpreted by the mind into a knowledge-episode, which contains the internal logic, assumptions, and structure through which the initial input is rendered intelligible. The knowledge-episode becomes knowledge only when it is revealed in cognition, which is the conscious disclosure of this processed content.

The sequence is:

  • Means of knowledge provide data that does not yet contain knowledge.

  • Knowledge-episode involves the mental construction of meaning from that data.

  • Cognition is the event in which this constructed content becomes consciously manifest to a subject.

Cognition yields the experience “I know this”, and it is within this moment of revelation that a knowledge-episode is integrated as knowledge by the witness. This structural distinction is important because the authority of knowledge derives not from the data itself or from the computational processes applied to it, but from its revelation within consciousness.

This leads directly to the role of the witness.

The Witnessing Consciousness as the Subject of Cognition

Dvaita vedānta identifies the witness (sākṣī) as the subject within whom cognitions arise and integrate as knowledge. With respect to our experience of the external world, the sākṣi functions as the jñānagrāhaka (the one that apprehends knowledge) and as the prāmāṇyagrāhaka (the one that judges its validity), scrutinizing it when doubts arise and otherwise accepting it immediately.

The witness does not perform the mental operations that constitute a knowledge-episode. Its function is to illuminate cognition when it occurs. It also illuminates the scrutiny when doubts arise. Without this illumination, the knowledge-episode remains an uninterpreted construction with no conscious significance. It is the repository of the qualitative aspects of knowledge. Concretely, it is defined as 

The witness is therefore the condition that enables:

  • the disclosure of a knowledge-episode,

  • the sense of acquaintance with its content,

  • and the unity in which multiple cognitions are related, compared, and revised.

This witnessing condition is not optional. It is presupposed in every instance where a cognitive event is experienced as mine, and where earlier and later cognitions can be apprehended in relation to each other. The structure of knowing requires a stable subject capable of receiving, integrating, and recognising cognitions as belonging to a single stream of awareness.

The witness is also the repository from which the qualitative nature of knowledge arises. It is defined as:

सुखादिविषयं स्वरूपभूतं चैतन्येन्द्रियं हि साक्षीत्युच्यते (Nyāya Sudhā)

While a knowledge-episode and cognition can ascertain the outdoor weather to be cold/hot, only the witness can attach it to the quality of “coldness” and “hotness”. The quality of any knowledge is non-conjectural. It is this innate property of the witness that gives rise to “subjective experiences”.

Fallibilism and the Need for a Witness

Fallibilism presents knowledge as an open-ended process of conjecture and correction. It asserts that every claim is revisable when contradicted by new information. While this account captures the dynamism of inquiry, it does not address the more fundamental question of how contradictions, revisions, or cognitive tensions become intelligible in the first place.

For fallibilism to function:

  • one cognition must be recognised as replacing another,

  • an earlier knowledge-episode must be grasped as defective,

  • a later one must be apprehended as corrective,

  • and both episodes must appear within a single field of awareness that allows comparison.

These operations presuppose the witness.

Popper replaces the justificationist regress with a critical regress -- no claim requires ultimate justification, only the capacity to survive testing.[1] But the critical regress has its own grounding problem. Each act of criticism produces a new conjecture about what counts as a refutation, which is itself open to further criticism.[1] Popper accepts this and considers it a virtue of his system. But he does not ask what anchors the recognition of each step in this regress. The witness is not a response to the justificationist's demand for certainty. It is a response to the more basic question of what makes any step in the regress intelligible as a step at all.

It might be objected that a fallible witnessing consciousness is sufficient for these operations. In the sense, that the witness need not be infallible, only present. But, a fallible witness generates its own regress, namely, errors in the witness's registration would themselves require a further witness to detect, and so on. The sākṣī avoids this not by making the content of cognitions certain, but by making the act of illumination itself structurally immune to error. Disclosure is not a claim since it cannot be false in the way a proposition can be false. A misperception is still genuinely illuminated as occurring. The error lies in the knowledge-episode's content, not in the fact of its revelation.

Without the witness, a sequence of knowledge-episodes remains a sequence of mental constructions with no standpoint from which they are recognised as episodes, no unity in which revision can occur, and no subject for whom error gains meaning. The recognition of error requires awareness of the previous cognition, awareness of the present cognition, and awareness of their incompatibility. This triad cannot be accounted for by the computational structure of knowledge-episodes alone. The witness provides the stable ground where all errors and corrections are illuminated.

Thus, fallibilism’s characteristic movement from conjecture to refutation depends on a witnessing consciousness that it does not explicitly acknowledge.

Intrinsic Validity and the Role of Cognition

In Dvaita vedānta, cognitions are considered intrinsically valid when they arise in the sākși. Knowledge is not certain on first apprehension. It is secured only through the sākṣi’s confirmation after scrutiny (when scrutiny is needed). This is noted as:

न च साक्षिणा विना बाह्यप्रत्यक्ष मात्रेण ऐन्द्रियार्थकं अपि सत्यता सिद्ध्यति इत्यतः साक्षीत्युक्तं (Nyāya Sudha)

This intrinsic validity is not a claim of infallibility but a structural feature of cognition: a revealed content presents itself and ‘some knowledge’ is established within the witness until a later cognition overturns it. The latter cognition is called bādhā (sublation), and it corrects the previously established knowledge. This avoids infinite regress, since each cognition does not require external validation to function as knowledge in the moment of its revelation.

The structure is now clear. Fallibilism correctly identifies that knowledge is perpetually revisable and that no source of knowledge is beyond criticism. What it does not explain is how the process of revision is itself possible. For revision to occur, successive cognitions must be compared, and their incompatibility must register somewhere. This requires a witnessing consciousness. But a witnessing consciousness that is itself fallible only pushes the problem back a level: its errors would require yet another witness to detect them, and the regress continues without resolution. The sākṣī resolves this not by placing any knowledge beyond criticism, but by distinguishing two entirely different kinds of infallibility. The content of a cognition remains revisable. The fact that it was illuminated does not. Disclosure is not a conjecture. It is the precondition for conjecture to mean anything at all.[1] This is precisely what Popper's epistemology without a knowing subject cannot account for: the logical structure of refutation does not itself refute anything.[2] A subject is required, and that subject must be structurally stable in the way the sākṣī is stable. Intrinsic validity, cognition, and the witness therefore form an interlocking structure, and it is only within this structure that epistemic correction becomes intelligible rather than merely assumed.

Conclusion

Fallibilism is one of the great insights in the theory of knowledge. The idea that all knowledge is conjectural, that no source is beyond criticism, and that error is not a failure but the engine of progress -- these are genuine advances in understanding how knowledge grows. But fallibilism is incomplete. It gives a compelling account of how knowledge changes and says nothing about what makes change intelligible in the first place.

For revision to occur, a prior cognition must be held against a later one, and its incompatibility must register with someone. This is not a minor technical detail. It is the precondition for the entire enterprise. A chain of knowledge-episodes with no subject to whom they appear is not knowledge in revision. It is just a sequence of unwitnessed computations.

Dvaita vedānta supplies what fallibilism omits. By distinguishing between the means of knowledge, the computational formation of knowledge-episodes, the revelatory event of cognition, and the witnessing consciousness that receives and integrates all of it, it identifies the structure that any theory of epistemic revision must presuppose. The sākṣī is not a retreat into dogmatism. It does not place any content beyond criticism. It simply identifies the condition without which criticism itself has no subject and correction has no meaning (the condition that Popper's epistemology deliberately set aside and never recovered.[1],[2]).

Fallibilism operates within this structure. It has always operated within this structure. It just has not said so.

References

  1. Popper, K. R. (1963). Conjectures and Refutations: The Growth of Scientific Knowledge. Routledge.

  2. Popper, K. R. (1972). Objective Knowledge: An Evolutionary Approach. Oxford University Press.

  3. Jayatīrtha. Nyāya Sudhā (Commentary on Madhva's Anuvyākhyāna). Standard Sanskrit edition.