Causal Underpinnings of Coherence: Re-evaluating Śāntarakṣita's Epistemology — Epoche C2
In the chapter of the Tattvasaṅgraha that examines the Mīmāṃsaka doctrine of intrinsic validity, Śāntarakṣita denies that a cognition can certify itself and holds that its warrant is conferred from outside it, by whether acting on it succeeds. That is a strange position for a philosopher who is supposed to be a pure coherentist. The thesis of this essay is the one its earlier version advanced — that Śāntarakṣita does not reject causal explanation in epistemology but builds it into the assessment of belief — and the object of this expansion is to replace the earlier version's suggestion that the causality is subtle and implicit with the demonstration that it is explicit, structural, and inherited by design. Along the way three claims in that version have to be withdrawn, and the withdrawal of the first changes what the essay is arguing against. 1. What has to be withdrawn, and what the argument is really against The earlier version opened by attributing to Georges Dreyfus's Recognizing Reality (1997) the view that Śāntarakṣita is a proponent of purely non-causal coherentism, and described this as the prevailing scholarly consensus. I cannot substantiate the attribution. Dreyfus's book is a study of the seventh-century philosopher Dharmakīrti and of the divergent readings of him developed in Tibet — principally the dispute over whether universals are real, with the Sa skya tradition denying it and the dGe lugs tradition granting them a qualified reality. Śāntarakṣita is not its subject. More to the point, the account of Dharmakīrti's theory of the means of knowledge ( pramāṇa ) that runs through that literature is causal: a cognition counts as a means of knowledge in virtue of being non-belying, and non-belying is glossed in terms of the cognition's issuing in successful activity. The word 'coherentism' does circulate in this scholarship, chiefly in describing how the Tibetan realist reading makes conceptual cognition answerable to a system rather than to a bare particular. But the specific claim the earlier version fought — that a recognised authority reads Śāntarakṣita as a non-causal coherentist — I cannot verify, and an essay should not be built on an opponent it cannot produce. There is a real dispute in the vicinity, and it is worth stating because it is the one the essay's conclusion actually bears on. Dan Arnold's study of South Asian epistemology of religion (2005) sets Buddhist and Mīmāṃsaka positions against each other on precisely the axis the earlier version cared about. The Mīmāṃsaka Kumārila holds svataḥ prāmāṇya , that a cognition's validity is intrinsic: a cognition presents itself as veridical and stands unless defeated, so warrant needs no causal certification. Arnold reads this as a broadly justification-based and non-causal epistemology, and reads the Buddhist alternative as a causal and broadly reliabilist one. The interesting question about Śāntarakṣita is therefore not whether causal considerations appear in his epistemology — they appear in the first line of it — but at which level of analysis they operate and what becomes of them at the other. John Dunne's reconstruction of Dharmakīrti (2004) supplies the tool for asking that question: Dharmakīrti argues on a sliding scale, giving one analysis at the level of external realism, where objects cause cognitions, and a stricter one at the level on which external objects drop out, with the earlier level retained as provisionally serviceable rather than repudiated. Two smaller withdrawals. Śāntarakṣita (c. 725–788) was an Indian philosopher of the Nālandā milieu, not, as the earlier version called him, a figure of Tibetan Buddhist epistemology; he travelled to Tibet at the invitation of the emperor Khri srong lde btsan, was involved in the founding of the first Tibetan monastery at bSam yas and in the first Tibetan ordinations, and is therefore foundational for Tibetan Buddhism as an institution, but the works at issue are Indian works composed in Sanskrit for Indian opponents. And the earlier version's comparison of his argumentation to the nested tales of the Pañcatantra is dropped, because the Tattvasaṅgraha 's structure is the opposite of nesting. It is an adversarial survey: chapter by chapter it sets out a rival doctrine — the Sāṃkhya primordial nature, the Nyāya creator, the eternal self, the eternal Veda, intrinsic validity — in the opponent's own strongest formulation, then refutes it, with Kamalaśīla's Pañjikā supplying the commentary. Nothing is concealed inside anything; everything is stated in order to be attacked. 2. Why the formal apparatus cannot ground an inference by itself The earlier version's second step claimed that the necessary concomitance ( vyāpti ) between a reason and what it establishes is derived from repeated observations of co-occurrence and non-co-occurrence. That is half right and the missing half is the whole point, so it is worth deriving. Take the formal apparatus first, since it is the part that carries no causal load at all. On Dignāga's account, inherited by everyone downstream, a good reason has three marks: it is a property of the subject under discussion; it occurs in at least some similar instances, those already known to have the property being established; and it occurs in no dissimilar instance, none of those known to lack it. Dignāga's 'wheel of reasons' holds the first mark fixed as satisfied and enumerates the possibilities for the other two. The reason's distribution over the similar instances is trichotomised — it occurs in all of them, in none, or in some — and likewise over the dissimilar ones, the trichotomy being exhaustive because a class is either wholly included, wholly excluded, or partly both. So the enumeration has $3 \times 3 = 9$ cells, and it is 9 rather than 27 because the first mark is held constant rather than varied. Reason's occurrence in dissimilar instances → Occurrence in similar instances ↓ In all In none In some In all inconclusive valid inconclus