Can Inference Be Valid?

Dreyfus calls this “the greatest difficulty in Dharmakīrti’s system”, and Chapter 18 of dreyfus-recognizing-reality is given over to it (dreyfus-recognizing-reality, p. 320).

The Problem

Dharmakīrti’s ontology restricts reality to the specifically-characterised, momentary and causally effective individuals, while inference operates on the generally-characterised, which by his own account is unreal. Stated that baldly it looks like a mismatch of objects, and Dharmakīrti has an answer to that: concepts are produced by perception and in turn produce further perceptions, so they are causally in touch with reality.

The difficulty is that the causal link does not do the work asked of it. Dreyfus’s example: watch a mountain over time and notice minute changes. From that same valid experience one may conclude (a) the mountain is impermanent, since it changes, or (b) the mountain is permanent, since for the most part it has not changed. Both conclusions are caused by valid perception and both may lead on to further valid perceptions, yet (b) has no validity whatever on a Buddhist account. The causal story cannot tell them apart, so something normative has to be added — and norms do not exist in reality. They come from the conceptual framework, which is the very thing whose grip on reality is in question (dreyfus-recognizing-reality, pp. 319–320).

Responses

Dharmakīrti’s Own Response

Coordinate the two. Perception supplies contact with reality, conception supplies the norm, and between them truth can be told from falsity: in case (a) there is a fit between how things are and how they are conceptually determined, in case (b) there is not.

Dreyfus’s objection is that the two cannot be coordinated, because they do not apprehend the same thing. “All the cognitions generated from the sense bases have as their object individuations (viśeṣa, ཁྱད་པར་). It is impossible for words to relate to these individuations.” Thought bears only on unreal conceptual content; perception never apprehends an unreal universal. Positing a shared object of application only defers the question, since that object is apprehended differently by each — really by the one, and through an unreal universal by the other. What then guarantees that conceptions apply to real objects they never grasp? Dreyfus’s verdict is that we are in a vicious circle (dreyfus-recognizing-reality, p. 320).

Sa-paṇ puts the same problem in a phrase: how to coordinate blind conception and dumb perception. His own answer is self-awareness, which inheres in both and so can introduce one to the other.

The Gelug Realist Answer

The realist move is not primarily ontological. It is a reinterpretation of what the object of a thought is, using a typology that goes back to dharmottara and was worked out by cha-ba: the appearing object (སྣང་ཡུལ་) or held object (གཟུང་ཡུལ་) on one side, the object of application (འཇུག་ཡུལ་) on the other. My thought of a chair primarily apprehends the object universal chair, but it also applies to a real chair.

Antirealists accept that last sentence and read “applies” pragmatically. Cha-ba reads it intentionally: the thought of a chair applies to the real chair in that it apprehends it. Thought is then not sealed inside the conceptual after all; it apprehends reality through the mediation of constructs (dreyfus-recognizing-reality, pp. 322–323).

The cost is exegetical, and Dreyfus does not minimise it: Dharmakīrti insists that thought’s apprehension does not mirror reality, so a tradition committed to that has to explain how mistaken thought nonetheless apprehends what is real. kay-drup’s answer runs through a distinction between appearance (སྣང་བ་) and appearing object (སྣང་ཡུལ་) — “the held object of conceptual thought is that object that appears to a conceptual thought as if it existed in front of us … from the point of view of conception, the appearing and held objects are only objects [indicated by] words (སྒྲ་དོན་, śabdārtha).”

The Sakya Position

Inference is pragmatically valid and its objects stay conceptual. The gap between concept and reality is not a defect to be closed but a consequence of what conceptual cognition (རྟོག་པ་) is: mistaken, and known to be so.

gorampa gives this position its sharpest formulation in gorampa-pramanavartika. On PV 1.80–81 he argues that conceptual designations of intrinsic entities are “exclusion-of-others object-possessors” that are nevertheless “supports for finding the intrinsically existent entities of the engaged objects” — they are indirectly created by, and therefore indirectly track, the real particulars they fail to directly apprehend. On PV 1.58 he distinguishes the appearing object (mistaken) from circumstances of ascertainment (not knowing vs. knowing), so that validity does not require the appearing object to correspond to reality, only that the ascertainment stand in the right causal relation to its real target. The reason and predicate in a same-nature inference are of one entity but distinct isolates — so inference does track something real, via a conceptual route that cannot help but distort its object.

What is unresolved

Dreyfus leaves the difficulty standing rather than adjudicating it, and it is worth being clear about which part is unresolved. It is not whether inference works; both sides agree it does. It is whether either account explains why. The antirealist owes an account of what “pragmatically valid” comes to, given that the norm distinguishing (a) from (b) above is itself a conceptual construct. The realist owes an account of how a mistaken cognition apprehends a real object, and pays for it in fidelity to Dharmakīrti. Dreyfus notes the parallel with Kant on intuitions and concepts, which locates the problem without solving it.

The Bifurcation of Knowledge

A related difficulty: dignaga rejects pramāṇasamplava (mixing of epistemic instruments). When we infer fire from smoke and later perceive the fire, we do not achieve a second perspective on the same fire — the perceived fire (svalakṣaṇa) and the inferred fire (sāmānyalakṣaṇa) are radically distinct objects. The apoha theory was developed partly to bridge this gap between the manifest image and the reductionist vision (westerhoff-dignaga-dharmakirti, pp. 234–235).

Sources

• dreyfus-recognizing-reality — Chapter 18
• westerhoff-dignaga-dharmakirti — bifurcation of knowledge, fallibilism and externalism of inference (pp. 230, 234–235)
• gorampa-pramanavartika — the Sakya orthodox response: conceptual designations as “supports for finding intrinsic entities” (PV 1.80–81); appearing object vs. circumstances of ascertainment (PV 1.58); one entity, distinct isolates