Scripture, Logic, Language

1. Dharmakīrti, Āryadeva and Dharmapāla on Scriptural Authority

+ –


1: Dharmakīrti, Āryadeva and Dharmapāla on Scriptural Authority

IT IS IMPORTANT TO NOTE that the epistemological school of Dignāga and Dharmakīrti, in spite of its insistence on the two means of valid cognition (pramāṇa), viz., direct perception and inference, did recognize that there was a whole class of propositions which could not be directly justified by means of these two pramāṇas, but demanded recourse to scriptures (āgama) or treatises (śāstra).1 This tension between scripture and reason, which is a recurrent one amongst religious philosophers, was however approached in a novel way by the Buddhists, a way which allowed them to accept certain “propositions of faith” but nonetheless retain a rationalistic orientation and extreme parsimony with regard to acceptable means of knowledge.

The key elements in the epistemologists’ position are to be found in kārikā 5 of the Svārthānumāna chapter in Dignaga’s Pramāṇasamuccaya (i.e., in PS II, k. 5a) and are developed by Dharmakīrti in the Svārthānumāna and Parārthānumāna chapters of Pramāṇavārttika (i.e., PV I and PV IV, respectively). However, what is remarkable, as we shall see later on, is that Dharmakīrti’s presentation also bears important similarities to, and perhaps may have even been influenced by, some passages in chapter 12 of the Catuḥśataka (CS) of Āryadeva.

The Epistemological School’s Position

Let us begin with some of the relevant passages from Dignāga and Dharmakīrti:

PS II, k. 5a:2 Because authoritative words (āptavāda) are similar 28[to an inference] in not belying, they are [classified as] inference.

PV I, k. 215:3 A [treatise’s4] having no visaṃvāda (“lies”) [means that] there is no invalidation of its two [kinds of] propositions concerning empirical and unempirical things by direct perception or by the two sorts of inference either [viz., inference which functions by the force of [real] entities (vastubalapravṛtta) and inference which is based on scripture (āgamāśrita)5].

PV I, k. 216:6 As authoritative words are similar [to other inferences] in not belying, the understanding of their imperceptible (parokṣa) object is also termed an inference, for [otherwise] there would be no way [to know such objects7].

PV I, k. 217:8 Or, they do not belie with regard to the principal point [viz., the four noble truths9], for the nature of what is to be rejected and what is to be realized as well as the method is acknowledged. Therefore [the understanding arising from the Buddha’s words can properly] be an inference in the case of the other things [too, i.e., radically inaccessible objects10].

Now, first of all, the usual types of inferences which we associate with Dignāga and Dharmakīrti, such as those of sound’s impermanence and the like, are said to be vastubalapravṛttānumāna in that they derive their truth from the fact that the reason—being a product (kṛtakatva)—is in reality, or objectively, related with the property—impermanence—and qualifies the subject, sound. However, an important point which needs to be made clear is that in spite of the numerous passages in which these authors talk about one state of affairs proving another, or about natural connections (svabhāvapratibandha) between the terms in an inference, it is not the case that every inference functions by the force of [real] entities (vastubalapravṛtta).11 (Often, for convenience, we will adopt a less literal translation for this technical term, i.e., “objective inference.” The point here, very briefly, is that the usual or paradigmatic type of inference in Dharmakīrti is one which functions objectively, or “by the force of real entities,” in that it can and should be evalutated purely on the basis 29of facts and states of affairs, and not in any way because of belief, acceptance or faith in someone or his words.) Vastubalapravṛtta is certainly an unbending requirement for the normal or “straightforward” type of inferences with which we are familiar, but, as we see in PV I, k. 215, there are also inferences based on scripture; that is to say, there exist inferences in which a scriptural passage rather than a state of affairs is given as the reason. The questions then easily arises as to (a) which sorts of scriptural passages can be used in such inferences, and (b) how the admittance of scriptural proofs can be harmonized with the general tenor of Dignāga and Dharmakīrti’s thought which is, no doubt, oriented towards vastubalapravṛttānumāna.

Let us begin with (b). The epistemological school solves this problem by introducing three sorts of objects: perceptible (pratyakṣa), imperceptible (parokṣa) and radically inaccessible (atyantaparokṣa). The first sort consists of those things such as form (rūpa), vases, etc. which are accessible to direct perception, while the second consists of things (such as impermanence, selflessness, etc.) which can be proven through the usual vastubala kind of inference. The third kind, however, are objects such as the different heavens (svarga) or the details of the operation of the law of karma, which are, of course, inaccessible to direct perception, but which also cannot be proven by citing some other state of affairs as a reason. In short, we might say that these objects are beyond the limits of ordinary rationality.12 A slight complication which should be cleared up at this point is that Dharmakīrti often uses parokṣa, a term which also has an extremely important place in PV III, in the sense of atyantaparokṣa. However we see in the commentaries that what is at stake in PV I, k. 216 is indeed atyantaparokṣa, and moreover, it is clear from certain passages elsewhere (in PV IV) that Dharmakīrti himself did explicitly accept this threefold division of objects.13

So Dharmakīrti limits the scope of scripturally based inferences to cases where the object is radically inaccessible (atyantaparokṣa), and hence beyond the range of ordinary ratiocination. By means of this strict delimitation, he can preserve his theory of inferences being objectively grounded, for this will be a requirement of logical reasoning which applies to pratyakṣa and parokṣa objects. He can also at the same time distance himself from the non-Buddhist schools’ use of scripture. In effect the error which a Mīmāṃsaka or Sāṃkhya makes in citing scriptural passages as a means of proof (sādhana) is that they apply scriptural arguments to propositions, such as sound’s impermanence, etc., which can and should 30be decided by vastubalapravṛttānumāna, and which are not at all outside the bounds of ordinary ratiocination.14

As for question (a), viz., the kinds of scriptural passages which can be used, Dharmakīrti introduces what Tibetan scholastics would come to call “the threefold analysis” (dpyad pa gsum) for testing as to whether scriptures (lung = āgama) are sound bases for inference or not.15 In particular, as PV I, k. 215 makes clear, such a scripture must be (i) unrefuted by direct perception, (ii) unrefuted by vastubalapravṛttānumāna, and (iii) free from contradiction with other propositions whose truth is scripturally inferred. Put in this way it might seem that what is being said is simply that the scripture cannot be refuted by any pramāṇa, or that it cannot come into conflict with any of the other three kinds of objects. However, the point at stake, as we find it elaborated in PV I, k. 216, Dharmakīrti’s Svavṛtti or Svopajñavṛtti (PVSV) and Karṇakagomin’s Ṭīkā, is more subtle, and is essentially an inductive argument: the scripture’s assertions concerning pratyakṣa and parokṣa are seen to be trustworthy, and so, similarly, its assertions about atyantaparokṣa, if not internally inconsistent, should also be judged trustworthy. The argument is given an alternative formulation in PV I, k. 217 when Dharmakīrti says that because the (Buddhist) scriptures are trustworthy concerning the principal points, viz., the four noble truths, they should also be trustworthy on radically inaccessible matters. The four noble truths are accessible to proof by vastubalapravṛttānumāna—as we see in the second chapter of PV—and thus, as these propositions in the Buddhist scriptures are trustworthy, so the others should be, too.

In short, scriptural argumentation—when applied to atyantaparokṣa, which is its only proper domain—is an inference: there is no need to postulate an additional pramāṇa such as the śābda (“testimony”) of certain Hindu schools. It is, however, a rather special, indirect case of inference, in that it turns on an inductive generalization which presupposes the use and correctness of direct perception and vastubalapravṛttānumāna.

Āryadeva and Dharmapāla

Now, a remarkable point in this connection is that the Tibetan writer Tsong kha pa blo bzang grags pa (1357–1419) in his Tshad ma’i brjed byang chen mo noticed that Dharmakīrti’s PV I, k. 217cd resembles k. 280 in chapter 12 of Āryadeva’s Catuḥśataka.16 Tsong kha pa was followed 31in this by rGyal tshab dar ma rin chen (1364–1432), who also remarked that CS XII, k.280 was the same reasoning as found in Dignāga and Dharmakīrti (phyogs glang yab sras). Subsequently, the Mongolian A lag sha ngag dbang bstan dar (1759–1840), in his sTon pa tshad ma’i skyes bur sgrub pa’i gtam, elaborated on the two verses, paraphrasing them into an identical formal argument (prayoga), and citing them in his proof that the Buddha is a “person of authority”’ (tshad ma’i skyes bu).17 While it seems impossible to definitively establish a direct lineage from Āryadeva to Dharmapāla to Dharmakīrti, the similarities between the verses in question do seem more then coincidental, and it is not at all impossible that Dharmakīrti was aware of Āryadeva’s thought, and that he made use of certain elements. Let us look at CS XII, k. 280 with Dharmapāla’s commentary.

After Dharmapāla has argued that the doctrines of the non-Buddhist “Outsiders” (wai dao, image) contain various faults and untruths, his commentary then has the Outsider object:

In that case, the noble teaching in the Tripiṭaka of the Tathāgata [also] sometimes has statements which are scarcely believable, and so (shi ze, image = evaṃ ca kṛtvā (?)) all the Insiders’ and the Outsiders’ texts would be untrustworthy; thus a gross absurdity (tai guo shi, image = atiprasaṅga) would ensue. How so? [Because] in the Buddha’s sūtras are mentioned various miraculous transformations (shen bian, image = vikurvaṇa; ṛddhi18) which are unimaginable. Or [these sūtras] speak about objects which have extremely profound (shen shen, image) natures; no sentient beings can fathom [these things]. ...[The objector now goes on to describe the miraculous powers and qualities of the Buddha and various other difficult to comprehend facts. He then sums up the objection:] As things such as these are all hardly credible, we harbor deep reservations about them. [Reply:] Phenomena, if they merely existed, could indeed give rise to [such types of] doubts. But phenomena are also void. Hence [Āryadeva] states in the following verse:19

             When someone gives rise to doubt concerning the inaccessible [things] (parokṣa) taught by the Buddha, then he should develop conviction in these very things on account of voidness.20


The point is that correctness of the Buddha’s teaching on voidness, which is accessible to ordinary inferential understanding, should lead one to believe that his teachings on matters inaccessible to such inferences are also correct. It is interesting to note that the Sanskrit of Āryadeva’s verse employs the term parokṣa, which is translated into Chinese as shen or “profound” (see n. 20). In Dharmapāla’s commentary we see him using the term shen shen in this context, which would thus very likely be the equivalent of atyantaparokṣa, although such an equivalent is not to my knowledge attested elsewhere. (Usually shen shen = gambhīra.) At any rate, it is clear that the use of parokṣa/shen at stake in Āryadeva and Dharmapāla, just as in Dharmakīrti’s PV I, k. 216, does refer to propositions inaccessible to direct perception and ordinary inference.

The similarities between Dharmakīrti and Dharmapāla’s approaches become even more striking when we look at the argumentation in the subsequent kārikā in CS with Dharmapāla’s commentary. In this verse, Āryadeva gives a kind of contraposed version of the reasoning found in CS XII, k. 280, arguing that because the Outsiders are mistaken on objects which are accessible to inference, then they must also be mistaken on those which are not. Dharmapāla, at this point, launches into a long refutation of the Vaiśeṣika’s metaphysical categories (padārtha) and the Sāṃkhya’s theory of the primordial nature (prakṛti) and the three qualities (guṇa) to show that the Outsiders are indeed hopelessly mistaken in their accounts of rationally analyzable objects, and hence cannot be trusted in their accounts of what is unanalyzable and is essentially more difficult to comprehend. Now, not only is this completely consonant with Dharmakīrti’s approach in PV I, k. 215–16, but conspicuously, Dharmakīrti in the Svavṛtti to k. 215 explicitly mentions the three qualities and the Vaiśeṣika categories of substance, motion, universals. etc. as being prime examples of refutable objects.

I conclude then that the similarities between these authors do seem to represent a common approach, as the Tibetan scholastics maintain. This, of course, may have been because these ideas were more or less commonly familiar to various thinkers of that period of time, rather than due to any more direct relationship. If however we take the hypothesis that Dharmakīrti (ca. 600–660 C.E.) was familiar with, and even influenced by Āryadeva’s thought, then it may very well have been via the commentary of Dharmapāla (530–561 C.E.), which is after all the first commentary on the Catuḥśataka that we know of. Moreover, 33we know that Dharmapāla did write a commentary on Dignāga’s Ālambanaparīkṣā,21 and as M. Hattori maintains, he may very well have been a “grandpupil of Dignāga”22: in other words, on Dharmapāla’s side it is clear that he was, in spite of his commentaries on Mādhyamika texts, very close to the epistemological school. It remains to be investigated then as to what other significant points of similarity, or influences, can be found between Dharmapāla and that other illustrious member of Dignāga’s lineage, Dharmakīrti.23



The original publication of this article contains the following dedication: “The present article took shape from work which I did during 1983–85 at Hiroshima University and is offered as a gesture of gratitude to Prof. Atsushi Uno, of whom I retain very fond memories.”

  1   In what follows the difference between āgama and śāstra is of no consequence. Also, as Dharmakīrti argues in PV IV, k. 93–107 (Sāṅkṛtyāyana’s heading: āgamasvavacanayos tulyabalatā), there is no essential difference in the force or trustworthiness of one’s own words and those of scripture. The same evaluative procedures apply to both.

  2   āptavādāvisaṃvādasāmānyād anumānatā. Sanskrit preserved in PVSV: 108.1. For the interpretation of °sāmānya as “similarity,” cf. PSV: 29b7: yid ches pa’i tshig nyid gzung nas kyang mi slu bar mtshungs pa’i phyir de yang rjes su dpag pa nyid du brjod do /.

  3   pratyakṣenānumānena dvividhenāpy abādhanam /dṛṣṭādṛṣṭārthayor asyāvisaṃvādas tad arthayoḥ //.

  4   Cf. PVSVT: 392.15: asya śāstrasyāvisaṃvādaḥ.

  5   Cf. PVSVT: 392.14–15: ...anumānena ca dvividhena vastubalapravṛttenāgamāśritena ca /. Cf. also PVV ad PV I, k. 215.

  6   āptavādāvisaṃvādasāmānyād anumānatā / buddher agatyābhihitā parokṣe ’py* asya gocare // *PVV reads niṣiddhāpy. Cf. Tib. lkog gyur na’ang.


  7   Vibhūticandra comments on PVV’s (p.365) phrase agatyānumānatoktā: ato ’nyathā parokṣe pravṛttyasaṃbhavāt /.

  8   heyopādeyatattvasya sopāyasya prasiddhitaḥ /pradhānārthāvisaṃvādād anumānaṃ paratra vā //.

  9   Cf. PVSV: 109.15–16: heyopādeyatadupāyānāṃ tadupadiṣṭānām avaiparītyam avisaṃvādaḥ / yathā catūrṇāṃ āryasatyānāṃ vakṣyamāṇanītyā /.

  10   Cf. PVV ad PV I, k. 217: paratrātyantaparokṣe ’py arthe bhagavadvacanād utpannaṃ jñānam anumānaṃ yuktam iti vā pakṣāntaram /.

  11   Cf. for example PV IV, k. 15: arthād arthagateḥ, etc. For svabhāvapratibandha, classic sources are to be found in passages such as PVSV ad PV I, k. 14. Cf. Steinkellner (1971) and (1984).

  12   For the examples of the three kinds of objects, see n.13.

13 In PV I, k. 216, its Svavṛtti, and subsequently, Dharmakīrti does not himself use the term atyantaparokṣa, but his commentators Manorathanandin and Karṇakagomin do. (Cf. PVV and PVSVT ad k. 216.) However, it is clear from passages such as PV IV, k. 51 (tṛtīyasthānasaṃkrantau nyāyyaḥ śāstraparigrahaḥ) that Dharmakīrti does accept the threefold classification. Tṛtīyasthāna refers to atyantaparokṣa. Cf. also PV IV, k. 50 where Dharmakīrti speaks of the first two sorts of objects: tathā viśuddhe viṣayadvaye śāstraparigraham / cikīrṣo sa hi kālaḥ syāt tadā śāstreṇa bādhanam //.

PVV ad PV IV, k. 50: śāstropadarśite viṣayadvaye pratyakṣaparokṣe rūpanairātmyādau tadā pramāṇapravṛttyā viśuddhe nirṇīte sati paścād atyantaparokṣe svargādau śāstreṇa śāstrāśrayaṇenānumānaṃ cikīrṣoḥ sataḥ sa hi kālo ’bhyupagamasya* yadi śāstrabādho na bhavet / “Sup-pose that at that time the two [types of] objects taught in a treatise, viz., perceptible and imperceptible [objects], such as form and selflessness, etc., are ascertained by a pramāṇa as faultless. Then subsequently, when one wishes to make an inference concerning radically inaccessible [objects] such as heaven, etc. by means of a treatise, i.e., by recourse to a treatise, then if the treatise is not invalidated, this would indeed be the occasion to accept it.” *Sāṅkṛtyāyana: abhyupagamya. Cf. PVBh: 505.4: sa hi kālaḥ syād abhyupagamasya /.

  14   Cf., e.g., PV IV, k. 2 and our commentary in Tillemans (1986c).


15 Cf., e.g., rGyal tshab rje’s commentary (rNam ’grel thar lam gsal byed) to PV I, k. 215, (Sarnath ed., 177–78), where the three criteria in PV I, k. 215 are presented as a formal argument (sbyor ba = prayoga):

shin tu lkog gyur ston pa’i bcom ldan ’das kyi gsung chos can / rang gi bstan bya la mi slu ba yin te / dpyad pa gsum gyis dag pa’i lung yin pa’i phyir / “The speech of the Bhagavān which describes radically inaccessible [objects] is non-belying with regard to the [entities] described, because it is a scripture which is [judged] immaculate through the three [kinds of] analyses.”

This is more or less a standard version of what in rtags rigs literature is categorized as a “reason based on authority” (yid ches kyi rtags). Interestingly enough, this literature then goes on to treat such reasons along the same lines as other types of valid reasons, classifying them in terms of effect (kārya), essential property (svabhāva) and nonperception (anupalabdhi). Cf. Yongs ’dzin rtags rigs, p.46.

  16   See pp. 158–59 where Tsong kha pa cites these two verses together and says that they show the same way (tshul mtshungs pa) to prove radically inaccessible propositions.

  17   For rGyal tshab, see his bZhi brgya pa’i rnam bshad, p.5: spyi’i rnam gzhag ni phyogs glang yab sras kyis bshad pa dang / shing rta chen po rnams ’dra bar yod do /. Cf. also rNam ’grel thar lam gsal byed (Sarnath ed.: 179). For A lag sha ngag dbang bstan dar, see sTon pa tshad ma’i skyes bur sgrub pa’i gtam, pp. 43–44. This author obviously relies on Candrakīrti’s commentary to CS XII, k. 280, where the correctness of the Buddha’s teaching on voidness is said to be an example (dṛṣṭānta) on the basis of which we can infer his correctness in other matters. Finally, for the idea of the tshad ma’i skyes bu, see Steinkellner (1983) and the references therein. Cf. also Inami and Tillemans (1986) for the triple division of tshad ma (= pramāṇa) into shes pa (“consciousness”), ngag (“speech”) and skyes bu (“person”). [Editor’s note: this subject is extensively treated in Tillemans (1993a).]

  18   Shen bian image = shen tong image (ṛddhi). See H. Nakamura, Bukkyō-go daijiten. Tokyo, repr. 1983: 795.

  19   T. XXX 1571, 216c. image image image image image 36

  20   Translated according to the Sanskrit of CS: buddhokteṣu parokṣeṣu jāyate yasya saṃśayaḥ / ihaiva pratyayas tena kartavyaḥ śūnyatāṃ prati /. Note that the Tibetan interprets śūnyatāṃ prati as “on the basis of voidness” or “relying on voidness”: de yis stong pa nyid brten te // ’di nyid kho nar yid ches bya //. Cf. the Chinese trans. of CS: image “When someone gives rise to doubt concerning the profound things (shen shi, image) taught by the Buddha, then he can rely on the voidness which is free of all [defining] characters and [can thus] gain sure faith.”

  21   T. XXI 1625, pp. 889–92.

  22   Hattori (1968: 2).

  23   I shall further develop some of these points in a translation of CS XII and CS XIII with Dharmapāla’s and Candrakīrti’s commentaries which I am preparing. [Editor’s note: see Tillemans (1990).]

Join Wisdom

This content is only available to All-Access, and Plus members of the Wisdom Experience. Please log in, upgrade your membership, or join now.

Join Now
rotate left rotate right