Reverberations of Dharmakirti’s Philosophy

by Birgit Kellner | 2020 | 264,305 words

This page relates ‘Prajnakaragupta’s Argument for Two Means of Valid Cognition’ of the study on the philosophy of Dharmakirti (6th century) and his predecessor Dignaga (5th century). This collection of articles reflects philosophical currents in India, China and Tibet during their time and investigates the Buddhist theories of Pramana (“instruments of trustworthy awareness”).

Prajñākaragupta’s Argument for Two Means of Valid Cognition

(By Patrick McAllister)

1. Introduction

The[1] Buddhist logico-epistemological tradition starting with Dignāga (ca. fifth to sixth cen-tury CE) accepts only two means of valid cognition (pramāṇa): perception and inference.[2] Whilst many aspects of pramāṇa theories have been studied carefully by modern scholars, the history of the arguments that are used to prove that only inference and perception are pramāṇas has not been investigated in great detail. The analysis usually offered is that there are, first, two means of valid cognition since there are two objects, but that, second, really only the particular is the object of a valid cognition.

For example, Franco and Notake (2014) characterize PV III 1–63 as follows, Franco and Notake 2014: 4:[3]

[The] argumentative structure is clear: There are two pramāṇas because there are two prameyas, and there are two prameyas because of the four criteria that distinguish between universals and particulars.

In their comments (Franco and Notake 2014: 30, n. 2) on PV III 1, they add that,

[…] in the final analysis, for Dharmakīrti and Prajñākaragupta it cannot be said that the fact that there are two kinds of object is the reason for there being two kinds of means of knowledge, but that two modes of cognition of the same thing are the reason for there being two kinds of objects of knowledge (prameya).

Such a paraphrase of the two arguments for inference and perception is in no way the result of misunderstanding what Dharmakīrti said. He does certainly say this. The question is, rather, how the two arguments cohere.

That there is a tension between the two statements can be seen by contrasting PV III 1, the opening verse of the pratyakṣa chapter, and PV III 53:


pramāṇaṃ dvividhaṃ meyadvaividhyāc chaktyaśakti-taḥ |
arthakriyāyāṃ keśādir nārtho’narthādhimokṣataḥ ||

The means of valid cognition is twofold because that to be cognized [by it] is twofold; for [that to be cognized by it] is [either] capable [or] incapable of fulfilling an aim;[4] [an illusory object like] hair and so on is not an object [that is to be cognized by it], because one does not apply [oneself to it] as an object.[5]

This corresponds exactly to the first part of the view of Franco and Notake (2014) quoted above: there are two means of valid cognition because there are two objects of valid cognition. This view is already endorsed by Dignāga.[6]

The verse corresponding to the second statement is this:

PV III 53d:

…meyaṃ tv ekaṃ svalakṣaṇam ||

There is, however, [only] one [thing] to be validly cognized, the particular.

So according to the first argument, there are two means of valid cognition because there are two objects of valid cognition. Here, however, Dharmakīrti maintains that there is only one object of valid cognition.

Dharmakīrti continues, giving the reason that there is only one object:

PV III 54:

tasmād arthakriyāsiddheḥ sadasattāvicāraṇāt |
tasya svapararūpābhyāṃ gater meyadvayaṃ matam ||

For one examines the existence and non-existence [of the particular], since the fulfilment of an aim is accomplished [only] due to this [particular]. [We] think[7] that there are two [objects] of valid cognition, because this [particular] is cognized through [its] own nature and through another nature.

So the difference of the objects of valid cognition is due to the same object being understood in two different ways: either it is understood through the characteristic particular to it, as it is apprehended by direct perception, or it is indirectly understood through another, general aspect, which is how a conceptual cognition comprehends its object. This argument broadly separates the field into two parts: it specifies perception and “something else”; this other is conceptual cognition, of which inference is only a subclass.[8]

Prajñākaragupta lets an opponent use the tension between the two arguments for an attack in the following passage, immediately after the citation of PV III 54ab:

PVABh2 213, 4–8: yadi svalakṣaṇam eva dvābhyām api viṣayīkriyate ekaviṣa-yatvād ekam eva mānaṃ prasaktam. athaikaviṣayatve’pi sāmagrībhedāt pra-māṇabhedaḥ. evaṃ sati prameyadvaividhyād iti virudhyate. uktaṃ cācāryeṇa yasmāl lakṣaṇadvayaṃ prameyam iti. sāmagrībhedena ca pramāṇabhede ca-kṣurādivijñānānām api bhedaḥ sāmagryā iti tāvanti pramāṇāni bhaveyuḥ.

[Opponent:] If only the particular is made an object by all two, [then], because there is [only] one object, only a single means follows.

[Proponent:] Now, even though there is only one object, there is a difference of the means of valid cognition because of a difference of the causal complex.

[Opponent:] If that is so, then [the statement] “because there are two [objects] to be validly cognized” (PV III 1) is contradicted, as well as what the teacher [Dignāga] said: “[There are two means of valid cognition] because the [object] to be validly cognized has two characteristics, [i.e., one particular to it and one that it has in common with other things].” (PS I 2bc)

But if there were a difference of the means of valid cognition because of the difference of the causal complex [generating a cognition], [then] also [different types of] cognitions such as visual [cognition] and so on would be differentiated [from each other] due to [their] causal complex. So there would be as many means of valid cognition [as there are different causal complexes].

In this passage, the opponent addresses the most important difficulties in maintaining both PV III 1 and PV III 53d–54ab. The core of the criticism can be explicated like this:

1) Given that there is only one object of valid cognition, the particular, and given the argument in PV III 1 that the number of objects is the reason for the number of means, it results that one means of valid cognition, perception, would be sufficient: Dharmakīrti would thus be contradicting the position that there are two means, expressed in PV III 1.

2) If the duality of the means of valid cognition is to be maintained, but for a reason other than the number of objects that are to be validly cognized, then Dharmakīrti is contradicting both the reason he himself gave in PV III 1, “because the object of valid cognition is twofold”,[9] as well as Dignāga’s reasoning in PS I 2bc.

3) In addition to this, if the new reason for the duality of the means of valid cognition should be that each valid cognition’s specific set of causes is different, then one will have to assume that there are as many means of valid cognition as there are variations in the causal complexes.[10]

The argument in PV III 53d–54ab is thus problematic in various respects: it contradicts Dharmakīrti’s previous statement, and it contradicts the tradition Dharmakīrti claims to uphold. Furthermore, if the actual reason for the differentiation of means of valid cognition lies in the difference of their causal complexes, then this reason does not prove what it is supposed to prove (according to this opponent, it could prove that there is either only one or that there are very many means of valid cognition).

In the following, I will focus on Prajñākaragupta’s strategy in answering this objection. The most relevant passages in this regard are his commentary on PV III 1–2 and on PV III 53d–58.[11]

2. Prajñākaragupta on PV III 1–2: a conventional criterion of validity

Prajñākaragupta’s interpretation of the argument in PV III 1, which is clearly influenced by the later argument found in PV III 53d–54, is as follows:

PVABhIn 36, 25–27:[12]

viṣayasya caikasyaiva dvaividhyaṃ pratipattiprakāra-sya dvaividhyāt | pratipattibhedaś ca pramāṇabhedaḥ | sa eva ca viṣayabhe-daḥ |

And the object, which is only one, is twofold, because the manner of [its] cognition is twofold. And the difference of means of valid cognition is [this] difference of cognition. And exactly this [difference of the means of valid cognition] is the difference of the object.

Prajñākaragupta is here interpreting PV III 1 in the light of the later argument: the object of valid cognition is only one, but we think it is twofold because it is apprehended in two different ways.

Before examining the later verses in the light of Prajñākaragupta’s commentary, it is necessary to understand a few of the programmatic points of his commentary on PV III 1–2: for it is here, in the opening section of the pratyakṣa chapter, that he describes the purpose and scope that he considers the chapter to have, and therefore these statements help in interpreting his later arguments.

The commentary on these two verses, up to PVABhIn 44, 27, is too long to be discussed here, and I will therefore limit myself to two issues that are central to Prajñākaragupta’s interpretation of Dharmakīrti’s argument: the role that self-awareness[13] and everyday activ-ity have in establishing that there are two means of valid cognition, and Prajñākaragupta’s characterization of the relation between inference and perception.

2.1 Self-awareness and everyday activity

Prajñākaragupta’s general answer to why one can say there are two objects of valid cognition is as follows:[14]

PVABhIn 39, 6–9: atrocyate | viṣayadvaividhyaṃ pratyakṣata eva siddham | sadṛśāsadṛśapratītir hi pratīter eva dharmaḥ | sa ca svasaṃvedanapratya-kṣasiddhaḥ | na ca pratītiḥ svarūpe bhrāntisaṅgatā | tatra bhrāntiśaṅkāyām avyavahāra eva bhaved anavatārahetur vā vādiprativādiprāśnikavacanasya |

To this [we] respond: That there are two [kinds of] object [for cognition] is established solely from perception. For the cognition of [something as] similar [to other things or as] dissimilar [to everything else] is a property only of cognition; and this [property] is established by the perception [that is cognition’s] awareness of itself. And cognition cannot be mistaken about [its] own form. If there is a suspicion about [the possibility of] an error with regard to that [form of cognition itself], there would be no everyday activity at all, or there would be no reason for the talk of [either] the proponent, opponent, or questioner to take place.

In other words, the fact that there are two objects of cognition is evident. A cognition of something can occur in two modes: it can be cognized as similar or dissimilar to something else—that is, it can be cognized as something that has something in common with other things, as a universal in the broadest sense of the term; or as being dissimilar from everything else, as a unique thing. These two modes are qualities of cognition, not of the object. As we will repeatedly see below,[15] it is the distinct (spaṣṭa) or indistinct (aspaṣṭa) appearance of the object that differentiates the two types of cognition. This appearance as such is directly perceived by the awareness that every cognition has of itself, and doubt about this aspect of cognition would end all conventional activity, as well as make any debate impossible.

In explaining this, Prajñākaragupta emphasizes that the duality of means of valid cognition is only conventional.

This becomes especially clear in the following passage, where he addresses the overall aim of the pratyakṣa chapter:

PVABhIn II 20–21, and 39, 20–24:

pratītibheda evāstu mānabhe-daḥ kathaṃ bhavet |
nanu prasiddhaṃ mānatvaṃ pūrvaṃ sāmānyalakṣaṇāt ||
tadbhedavyavahāro’yam idānīṃ sādhyatāṃ gataḥ |
savikalpakam adhyakṣam eṣo’gnir iti yo vadet ||

[...] vyavahārataḥ pravṛttinivṛttilakṣaṇāt prāmāṇyaṃ sāmānyalakṣaṇenaiva prathamapariccheda eva prasiddham | bhedavyavahāramātrakam evedānīṃ sādhyam āpannam | tatra savikalpakam ekam evedaṃ pratyakṣaṃ yad utāgnir ayam asmād abhipretārthakriyākārīti yo vadet taṃ prati dvitayam etad iti | pratītyākārabhedāt |

[Opponent:] There certainly may be a difference of cognitions. [But] why would there be a difference of means of valid cognition?

Well, the state of being a means of valid cognition was well established earlier, from [its] general characteristic.

[Proponent:] Now, this everyday activity [of ours that treats] the means of valid cognition as different, has come to be what is to be established,

[as] someone might proclaim a conceptual perception, [like] “This is a fire.”

… On the basis of everyday activity, [that is], from the characteristic of engagement or nonengagement [with an object], what it is to be a means of valid cognition has been well established in the first chapter [the {?}\textit{?} pramāṇasiddhi chapter of the PV, but] only according to [its] general characteristic.

Now, [in this chapter, what it is to be a means of valid cognition] has become what is to be established only [inasfar as] it is commonly treated as differen-tiated [into two types]; [the explanation] “this [means of valid cognition] is twofold” [is given by Dharmakīrti] for that [person] who would say about this [means of valid cognition] that there is only a single conceptual perception, for example, “This is a fire, from it the attainment of a desired aim is brought about.”; [this is said to him] because the form of cognition is differentiated.

In these statements Prajñākargupta is contrasting the scope of the first (pramāṇasiddhi) chapter and the current chapter: in the first chapter, validity was established in its general form—as a definition applicable to all kinds of valid cognitions. Moreover, validity was established there conventionally—that is, in relation to the engagement in some activity (or abstention from an activity) that a means of valid cognition facilitates in a way that makes that activity successful.[16] On this background, Prajñākaragupta continues, an opponent might object: conceptual perception is compatible with this criterion of validity, and, in fact, it is the only (ekam eva) means of valid cognition that has to be assumed. It is compatible with this definition of validity because, if conceptual perception is possible, it could contain the ascertainment “This is a fire, it can fulfil my aim.” This is not acceptable to Dharmakīrtians. Therefore, a further argument is necessary to show that there is not only one combined means of valid cognition, a perception containing a conceptual cognition, but rather that there are different means of valid cognition.

In other words, the pramāṇasiddhi chapter does not answer the question of which means of valid cognition there are, but only of what a means of valid cognition is. The pratyakṣa chapter will however deal exactly with this question: what are the different means of valid cognition? In addition, Prajñākaragupta qualifies this with the statement that that which is to be established now is bhedavyavahāramātraka: it consists exclusively in the conventional talk of a division of the means of valid cognition as defined in the pramāṇasiddhi chapter.

The importance of noting this lies in the fact that Prajñākaragupta separates two ele-ments: the capacity for leading to successful activity, the general criterion of any cognition’s validity, and what cognitions are like (in particular, whether they are conceptual or per-ceptual), which is the basis for distinguishing the types of valid cognitions and is evident in any cognition’s awareness of itself. In a conventional sense, a cognition can be con-sidered ‘valid’ if it allows one to act successfully; it is not important for its validity what type of cognition it is. And vice versa, the general mark of validity is not important for distinguishing the types of these cognitions.

2.2 The relation of inference and perception

Prajñākaragupta then discusses which means of valid cognition can lead to successful activity, or, in other words, conforms to the conventional and general criterion of validity.

First, an opponent, apparently a Buddhist,[17] raises this objection:

PVABhIn 39, 30–32:

atha pratyakṣam eva pravartakaṃ nāparam |
tat tu mano’ntaraṃ
[18] bhavad api na pramāṇam |
na hi saha tena yāvad bhavati tāvat pramāṇam |
śarīrabhūtalādīnām api prāmāṇyaprasaṅgāt |

Now, perception alone is what causes activity, nothing else. That other cogni-tion, though it exists, is not a means of valid cognition. For, [something] is not a means of valid cognition to the extent that [it] exists together with [a means of valid cognition], because it would result that even the body, the earth, and so on [which might all co-exist with a means of valid cognition] would be means of valid cognition.

Prajñākaragupta answers that perception is not, in and of itself, capable of letting a person act:

PVABhIn II 23:

anvayavyatirekābhyām upayogītarasthitiḥ |
na ca kevalam adhyakṣaṃ tadabhāve pravartakam ||

[Something] is determined as assisting [something else] or not through the positive and negative concomitance [of that which assists and that which it assists].

But perception by itself is not, when that [which assists it, i.e., conceptual cognition], is absent, what causes activity.

Without the introductory passage explaining that the topic here is the conventional difference of perception and inference, this would of course be in stark contrast to Dharma-kīrti’s well-known position that, in activity following upon perception, only the perception is a means of valid cognition, but not the conceptual cognition that perception needs to be followed by in order to cause a person to act.[19]

But, given that the discussion has here been restricted to the question about the difference of means of valid cognition, the verse says that it is not possible that perception alone—only a cognition with an object that is distinct (spaṣṭa)—can cause activity. The statement thus emphasizes that the general and conventional characteristic of a means of valid cognition, that it prompts activity, cannot be upheld if one wishes to also bind it to only one type of cognition—the perceptual one. That would immediately land one in an untenable position. In other words, to take perception alone as a means of valid cognition in this conventional sense is impossible: it will only work if one grants that status also to a conceptual cognition (or at least some conceptual cognitions).

This is then elaborated in the following, where Prajñākaragupta examines the relation-ship between inference and an activity following upon perception. Even in the case of complete habituation—a state in which a being is able to act upon a perception without in-tervening conceptual cognition—conceptual cognition is involved: for, so Prajñākaragupta, the fact that there is a habituation presupposes an inferential cognition on whose basis the habit must initially have been formed.[20] To hold otherwise, so PVABhIn II 24, would be as clever as thinking, because one contracts a disease the first time one goes somewhere, it is unsafe to go there the first time, but safe thereafter. In this sense, inference can be said to be the main element (pradhāna) even in purely habitual activity, which the opponent would like to see as the main support for the claim that perception can promote activity without conceptual aid.[21] We can now investigate Prajñākaragupta’s comments on the later verses.

3. Prajñākaragupta on PV III 53d–54: one object known in two ways

In this section of the PVABh2, Prajñākaragupta is presenting a discussion in which, as we will see, the following argumentative aims are intertwined:[22]

1. He needs to resolve and explain the possible contradiction between PV III 1 and PV III 53d (described above).

2. He has to prepare for the proof that perception and inference are the only means of valid cognition.

3. He has to prove that perception and inference are different means of valid cognition. (Points 2 and 3 together make it possible to maintain that there are two, and only two, means of valid cognition.)

4. He has to uphold cognition’s awareness of itself as the central cognitive faculty, in the sense that in reality perception and inference are two forms of this, and can only be conventionally separated.

Prajñākaragupta starts his commentary on PV III 53d (see above, section 1) as follows:

PVABh2 212, 29–30:

na hi sāmānyaṃ nāma prameyam, yathākalpanam ayo-gāt | svalakṣaṇam eva paramārthataḥ prameyam | tasyaiva sadasattvenāva-bodhasya prayojanatvāt | kuta etat |

tasmād arthakriyāsiddheḥ sadasattāvicāraṇāt |

arthakriyākāriṇo hi padārthasya sattvāsattvābhyām arthitā prekṣāvatām | tadavabodhāya ca pramāṇam anviṣyate | anyathā pramāṇaparīkṣaṇam apre-kṣāpūrvakriyaiva bhavet | tasmād arthakriyākāripadārthabhāvābhāvaviṣayī-karaṇasamartham arthavat pramāṇam | tasmāt dvābhyām api pratyakṣānumā-nābhyāṃ svalakṣaṇam eva viṣayīkartavyam | anyathā pramāṇatvāyogāt |

For the so-called universal is not an object of valid cognition, since it is not possible as it is imagined. In reality, only the particular is [the object] to be validly cognized, because the knowledge of the existence or non-existence of it alone is of use. Why is that?

Because one examines the existence and non-existence [of the particular], since the fulfilment of an aim is accomplished [only] due to it. (PV III 54ab)

For, judicious beings are intent upon [whether] the object that produces the fulfilment of [their desired] aims exists [or] not. And for [the purpose of] know-ing that, [i.e., whether that objects exists or not], a means of valid cognition has to be sought for. Otherwise, the investigation by a means of valid cognition would be an entirely injudicious activity. Therefore, a means of valid cognition has an object [insofar as] it is capable of making the existence or non-existence of a thing that produces the fulfilment of an aim [its] object.

Therefore, by all two [means of valid cognition], perception and inference, only the particular is to be made the object, because otherwise it would be incoherent that [they should] be means of valid cognition.

So perception and inference are each a means of valid cognition only inasfar as they direct a person to the successful accomplishment of a purpose. In order to do this, they have to reliably let a person attain a particular, since only that is capable of causing the desired effect. From this perspective, the particular alone is the object that is to be validly cognized, the prameya. The opponent immediately raises the objection that, if there is only one object of valid cognition, then one means would be enough, too.

Prajñākaragupta just remarks that this is wrong, since it is due to the differences in the respective causal complexes that two types of valid cognition can be assumed—implying that it is not because there are two objects.[23]

It is after this statement that the opponent delivers the main objection, presented above in section 1, and it should now be clearer what the exact direction of that objection is. McCrea (2011: 322 f.) has characterized it well, along with the main defence that Prajñākaragupta uses:

… the opponent attempts to find some way to distinguish perception from inference as a more direct or immediate mode of awareness, so that he can still deny the validity of inference without similarly condemning perception. In each case, Prajñākaragupta demonstrates that the opponent’s purported distinction is spurious, and that perception and inference are similarly indirect, such that one could not accept one as a pramāṇa without accepting the other as well.

In the current context, the problem concerns the causal complex of the means of valid cognition. The opponent distinguishes the object (prameya), which is the main factor in the causal complex, and the other, secondary factors in his criticism:

PVABh2 II 229:

paramārthaprameyatve syād anantaprameyatā |
apekṣākṛtabhedatve paramārtho na lakṣaṇam ||

If the object of valid cognition were real, there would be infinite objects of valid cognition. If [the object] is differentiated in dependence [on some other factor], reality is not a characteristic [of the object so differentiated].

This verse sums up the two sides of the opponent’s attack: if the particulars are the objects of valid cognition, then there would be infinitely many different objects. In other words, if the main factor in the causal complex generating a cognition were the real thing, there would be very many such cognitions, each a proper and separate means of valid cognition. So the Buddhist position expressed in PV III 53d, that there is only one object of valid cognition and that it is real, would be incoherent since this object’s reality implies its multiplicity. Alternatively, the opponent continues, the proponent might claim that the objects of valid cognition are differentiated in dependence on some other factor in the causal complex (as claimed in PV III 54cd); this, however, would mean that the difference maintained for the objects of valid cognition would not belong to the objects themselves, but would be external to them. Again, this consequence would violate what is endorsed in PV III 53d, namely that the object of valid cognition is real. In other words, one cannot maintain both that the object of valid cognition can be defined as real (paramārtho na lakṣaṇam) and that it is only one (anantaprameyatā), both of which are asserted, however, in PV III 53d: meyaṃ tv ekaṃ svalakṣaṇam.

The immediately following excursion leads up to PV III 55. For our purposes, a look at its conclusion is sufficient:

PVABh2 215, 4–9:

kathaṃ tarhi meyāntaram | tasyaiva pararūpeṇa pratīteḥ | tathāpratīyamānaṃ dṛṣṭāntasādhāraṇena rūpeṇa meyāntaram | pratipatti-bhedena tadrūpāropān na paramārthataḥ |[24] evaṃ tarhy apekṣākṛtatvān na paramārthatā | satyam avastu sāmānyam iti pratipāditam eva | nedam apū-rvam ucyate | vastusaṃvādadvāreṇa vyavahāribhir alakṣitanānātvair vastv iti vyavahriyate | tena tadapekṣayedam ucyate prameyadvaividhyam |[25]

How then is there another object of valid cognition? Because of a cognition of exactly that [same object of valid cognition] with another nature. [Inasfar as the particular is] being cognized in this way with a form that it has in common with the example [in an inference], there is another object of valid cognition; [in other words, there is another object] because that form [which the particular has in common with the example] is superimposed [on it] due to a difference in cognitions; [but there is] not [another object of valid cognition] in reality.

[Opponent:] In that way, then, because it is made in dependence [on something else, this other object of valid cognition] is not real.

[Answer:] That is true. Indeed it was taught that the universal is not a real thing.[26] [But] this is not said without precedent. [For], in virtue of [a univer-sal] cohering well with a real thing, [people] engaged in everyday activity commonly act [with regard to a universal] by [considering it] a real thing, [inasfar as] they do not take note of the fact that [the particulars they are acting towards] are different. Therefore, it is in dependence on this [real thing that appears in two forms in everyday activity] that the duality of the object of valid cognition is spoken of.

The heart of the passage rephrases a part of PV III 54, namely the cognition of a particular in an indirect way (literally, “with another nature”; tasya svapararūpābhyaṃ gater meyadvayaṃ matam). Prajñākaragupta is again making the notion of the duality of the object of valid cognition depend on an everyday understanding of the matter: normal people, when inferring something, think about the object of that cognition as “the real thing,” even though what they are actually cognizing—the proper object of the valid cognition—is a future real thing; and it is this which is not directly cognized, but only by means of a feature that it shares with the example. If a fire on the hill is being inferred from smoke, it, the actual fire that could cook our food if we reach the place on the hill soon enough, is being cognized, not as it is in itself, but as similar to a kitchen fire previously seen. This similarity cannot be a phenomenal one, of course: the fire on the hill does not appear.

With this, the scene is set for the discussion of how inference can be a means of valid cognition. The main points to bear in mind in reading this discussion, in addition to those previously stressed in Prajñākaragupta’s commentary on PV III 1, are as follows:

1. That there are two means of valid cognition, as well as two objects of valid cognition, is only a concession to everyday activity. This is, essentially, the justification for the tension between PV III 1 and PV III 53d–54.

2. Only the real thing is the object of valid cognition.

3. There are two ways in which the real thing is known: through its own nature, and through another nature.

4. Perception is dependent on inference because inference initiates what becomes habitual. Without habituation, and thus without inference, perception does not cause a person to act.

The main controversy discussed in the following passages of PVABh2 is about how to understand item 2. The question is whether the statement in PV III 53d is true conventionally or in reality. Depending on the answer, the consequences vary. When this statement is taken to express a convention, the opponent’s claims are very strong: the tension between, on the one hand, inferring “there are two means of valid cognition because there are two objects of valid cognition, a real thing and an unreal one”, and, on the other hand, maintaining that “there is only one object of valid cognition, the real thing” can hardly be contained if the term “real thing” is here taken in the same sense. If the latter statement is, however, taken as expressing a fact of reality (and the former one not), then the opponent’s attacks lose most of their power; but this forces us to reconsider in what sense the particular is the only object of valid cognition.[27]

4. Prajñākaragupta on PV III 55–58: saving inference

This section asks how inference can be considered a means of valid cognition, even though it is mistaken because it has an imagined universal, not a real thing, as its object. This problem is raised in PV III 55cd, and Dharmakīrti’s answer is given in PV III 56–58.[28]

Prajñākaragupta opens this exchange with the following objection:

PVABh2 215, 18–19:

kathaṃ tarhi paramārthasya viṣayīkaraṇād bhedaḥ | svarūpasākṣātkaraṇe hi pratyakṣataiva bhavet | tadasaṃsparśe kathaṃ tadvi-ṣayatā | How, then, is there a difference [between perception and inference], since [they both] make a real thing [their] object? For if [a cognition] directly presents the nature [of something], then [it] is just perception. [But] if [a cognition] does not touch that [nature], how is that [nature] the object [of that cognition]?

The opponent is here engaged in a variation on McCrea’s schema quoted above (in section 3).[29] Either one takes it seriously that there is only one object of valid cognition, and concludes that there is no difference between cognitions of that object, or one takes the position that one cognition has that object and the other does not. This latter position implies that the cognition is not correct (that it is an error or bhrānti); this means, in the eyes of the opponent, that it cannot have the status of a valid cognition.

Here, and in the following discussion, the opponent’s general position will be that the difference between validity and invalidity hinges on the contact of cognition with reality: if the cognition is in direct contact with a real thing, it is valid; if not, then it is invalid. In this way, we can read the debate as not only being about what differentiates inference from perception, but also as being about the implicit criterion that constitutes the validity of a cognition. As we will see, Prajñākaragupta maintains quite simply that the contact with the real thing is not what makes a cognition valid: neither perception nor inference are valid for this reason.

Prajñākaragupta first explains (PVABh2 215, 21–23) what it means that a cognition is erroneous in the sense that the opponent is using in his argument: it means that a cognition does not have the actual particular as its object, and can thus not be said to conceive of its object as it is.

The opponent immediately counters that then this cognition cannot be considered to have the status of a means of valid cognition, presenting the gist of PV III 55cd, as follows:

PVABh2 215, 26–27:

yadi pararūpeṇa gatiḥ kathaṃ tasya gatiḥ | pararūpasyai-vāsau gatiḥ | tatrānyasya prāptau bhrāntir eva | bhrānteś ca na pramāṇatā |[30]

If there is a cognition [of something] with another nature, then how is [this] a cognition of that [thing]? This is a cognition only of another nature. When one attains another [thing, there is] simply an error with regard to that [thing which was cognized]. And an error is not a valid means of cognition.

The argument of the opponent can be analyzed as follows:

1. There is a cognition of something (X) in a form other than its own (Y), e.g., the cognition of a particular fire in the form of the general concept ‘fire’.

2. So this “cognition of X” is, in fact, a cognition only of something else, Y.

3. If one acts upon this cognition of Y, one attains X.[31]

4. But the cognition of Y is an error with regard to X (because X is not cognized as it really is).

5. An error cannot be a means of valid cognition.

6. So the cognition of X in the form of Y cannot be a means of valid cognition.

Dharmakīrti answers this in PV III 56. Prajñākaragupta’s interpretation of this verse is indeed, as McCrea (2011: 321) has noted, striking in “… the overall similarity, indeed the virtual identity, of perception and inference” that the verse is said to imply.

The first passage runs as follows:

PVABh2 215, 30–216.2:

anyatrāpi yathārthābhiniveśaḥ katham avaganta-vyaḥ | abhiprāyāvisaṃvādād eva | sa cātrāstīti nāyathārthābhiniveśaḥ | idaṃ tu vāsanābalāj jñānaṃ pratibhāsabhedato bheda iti naivaṃ vyavahāriṇo vida-nti | vyākhyātṛpratītir eveyam |

Even in the other [case, that of perception], how is [this] “determination [of the object] as the object is” to be understood? Only on [the basis that determination] does not belie [a person’s] intentions. And this is the case here [in the case of inference as well]. So there [is, in fact], no determination that is not according to [the] object [in the case of inference either].

But [people] engaged in everyday activity do not know that this cognition [has, in fact, originated] in virtue of the impressions [left by previous experiences, and] that there is a difference [of perceptual and conceptual cognition only] due to the difference in the appearance [of an object to cognition]. This is only the insight of those [people] who explain [things].

Prajñākaragupta is here basing his argument on the expression ayathābhiniveśena of PV III 55a, by which Dharmakīrti means to say that inference does not determine its object correctly (lit., “as [it] is”). It is this element that Dharmakīrti lets the opponent criticize in PV III 55cd: if inference is not correct in that it does not determine its object correctly—Prajñākaragupta’s opponent here taking “correctly” as “how the object really is”—then how can it be a means of valid cognition? Prajñākaragupta turns the question around, and simultaneously gives the more precise interpretation of “determines [its object] as [that] object [really] is” (yathārthābhiniveśena) to Dharmakīrti’s formulation: what, he asks the opponent, is the meaning of the qualifier “determine something as it really is” in the case of perception?

Prajñākaragupta adds an important qualification: normal people do not understand that cognition—which I take here to include both perception and inference—is not due to different causes (it is always generated by mental imprints, not external objects), and that cognitions are different on account of how something appears in them, not, one must understand, essentially. ‘Philosophers’, on the other hand, do know that this is the case.

It is therefore justified to say that Prajñākaragupta, as in the interpretation of PV III 1–2 and 53d (section 2 and section 3), again distinguishes two levels on which one can differentiate perception and inference, in this passage and in the following ones: one respect in which they follow the everyday usage of normal people, who say that they directly perceive some things, and indirectly cognize others (correctly, when by inference); and one respect, corresponding to reality, in which the cognitions are distinguished only on account of how an object appears in them.

Now, there might be a problem: if Prajñākaragupta says that ayathārthābhiniveśa is not the case in inference, then PV III 55ab, ayathābhiniveśena dvitīyā bhrāntir iṣyate, would seem to be contradicted; there, Dharmakīrti obviously accepts that the second kind of cognition, inference, is an error because it does not conceive of its object as it really is.

This alerts us to a further point in Prajñākaragupta’s interpretation of PV III 55ab: it stresses the provisional character of that statement. In order to avoid the contradiction to Dharmakīrti, the sentence governed by the verb iṣyate (“it is assumed”) must not be under-stood as expressing something endorsed by Dharmakīrti, but as a general statement of fact about what people normally take to be the case. Relying on this aspect of Prajñākaragupta’s interpretation, one might then want to translate this sentence as “The second [kind of cognition, inference], is [commonly] assumed to be an error since there is no determination [of an object] as it is.”[32] And the implied agent of the sentence then would be “by people engaged in everyday activity.” In other words, that the criterion ayathārthābhiniveśa, which makes a cognition an error, is applied to inference is something usually done by ‘normal people’. This is an important interpretative move, because now Prajñākaragupta has cre-ated enough room to say that it is, in fact, not the case that this assumption is correct. It leaves open three directions in which Dharmakīrti might have continued to argue: First, he could have tried to argue that inference is in fact not an error according to the criterion of ayathābhiniveśa (in its common sense), a path that was not chosen. Second, that inference is, unlike perception, an error, but that this does no harm to its status as a means of valid cognition; this is perhaps the simplest interpretation, upon which the statement “inference is assumed to be an error” is taken as implying that perception is not. Third, that not only is inference an error according to the criterion of ayathābhiniveśa, but also perception; on this reading, the statement “inference is assumed to be an error” entails an insufficiency: inference is assumed to be so by people, but they forget that perception is also like that. We will see that, with some caveats, this is the road that Prajñākaragupta presents Dharmakīrti as having chosen.

He then continues to explain the general equality of inference and perception in terms of their being means of valid cognition.

Having said that yathārthābhiniveśa, inasfar as it results only from abhiprāyāvisaṃvāda, is not different for them, he analyzes the next acknowledged criterion for being a means of valid cognition, the fact that it leads to (successful) activity:[33]

PVABh2 216, 2–7:

sa evābhiprāyaḥ katham anyadarśānād iti cet | na vika-lpānāṃ vastupratiniyamābhavāt | anādivāsanāsāmarthyam evaitat | tataḥ katham aparicchinnatattvas tatra pravartata iti na codyam etat | dṛṣṭe ca nā-nupapattisambhavaḥ | pratyakṣe’pi kathaṃ pravartate | tatrāpi naiva prāpta-vyarūpaparicchedaḥ sannihitamātrasya paricchedāt | pratyuta34 pratyakṣam evāpravartakaṃ sannihitamātrasya pariprāpteḥ | tatrāpi tadekatvādhyava-sāyād eva vṛttir bhāvini vastuni | tato’numāne’py evam eva vṛttiḥ | katham asamānatayekṣyate |

[If one asks:] How can there be exactly this intention from seeing something [completely] different?

[Then we answer:] It is not [because of seeing something that there is this intention], because conceptual cognitions are not restricted to real things. Such is simply the capacity of beginningless impressions. It must not be criticized how, because of this [capacity, someone] who has not discerned reality can act towards it. But if [this activity] is observed, [its] not being the case is impossible.

Also in [the case of] perception [one could ask the same question:] how does one act? Not only is there, in this [perception], no discerning of the nature that is to be attained, because only that which is [immediately] present is discerned, but perception as such [also] does not make [a person] act, because only that which is present is completely attained [by perception].

In that case [of perception] too [a person] acts towards a future real thing only because there is a determination [of it] as identical [with what is perceived]. Therefore activity is exactly the same also in the case of inference. [So] why [are they] regarded as not being the same [by you]?

The explanations here underpin the main point of the previous discussion: the opponent holds that inference cannot enable a person to act because it ‘does not discern reality’, i.e., it does not put the person in direct contact with the real thing that successful activity must be directed at—contrary, so the opponent still assumes, to perception. Prajñākaragupta’s answer to this is double sided: first, activity arising from inference can be observed, and can therefore not be impossible. Prajñākaragupta then turns the question around: how can one act in the case of perception? The opponent’s criticism of inference would be applicable there too: perception knows nothing of the object that is capable of fulfilling one’s aims (since what one perceives is a present, momentary object). And as such it can enable activity towards a future real thing only through a determination of the thing it has directly perceived and the future thing as identical; one must think that the thing’s existence is temporally extended. And in this respect, it would be just like inference.[34]

Again, the situation described by McCrea is found: Prajñākaragupta is applying the opponent’s statement about inference to perception, and the opponent cannot accept the consequences. If the opponent insists that the attainment of something that was not cognized is an error (item 4), then the exact same thing has to be said for perception: it too grasps only the present thing, yet acting upon it a person will attain a future thing; and it does, by itself, not actually make a person act.

A concise summary of this approach is found in PVABh2 II 234 and the following prose explanation:

PVABh2 II 234:

pravartako vikalpaś ced avastugrahaṇe katham | tathāpi vartayaty etad anumāne na kiṃ matam ||

PVABh2 216, 21–22:

yadi hi vikalpajananadvāreṇa pratyakṣaṃ pravartakam | āyātaṃ tarhi vikalpasya pravartakatvāt pramāṇatvam | tathā saty anumānasya prāmāṇyam avyāhatam eva |

If [you say that in the case of perception] a [following] conceptual cognition prompts [a person] to act, how [does it do that] without grasping a real thing?

If [you say that it] prompts [a person] to act nevertheless, then why is this not assumed in [the case of] inference?

For, if perception prompts activity by means of generating a conceptual cog-nition [that makes a person act], then one has arrived at [the position] that conceptual cognition, because it prompts activity, [would be] a means of valid cognition. [And] if it is so, it has not been rejected at all that inference is a means of valid cognition.

This passage is exemplary for Prajñākaragupta’s defence of the difference between inference and perception: given that perception itself does not prompt activity, it must be the following conceptual event. But that conceptual cognition, like inference, does not grasp anything real (being conceptual, it has a universal, not a particular, as its object). The alternatives are dire: if the opponent were to insist on his position, not only would the state of being a means of valid cognition be a quality of the conceptual cognition following perception, but also the fact that inference is a means of valid cognition—the point that the current debate is actually about—would not have been fended off. This would be a bad defeat for the opponent.[35]

The implied consequences are unacceptable to the opponent, who wishes to link what it is to be a means of valid cognition to this criterion of knowing a real thing as it is: he has to admit either that neither perception nor inference has a claim to being a means of valid cognition, or that both have.

The following discussion continues in much the same way. We can skip to the end of the discussion, and consider its summary:[36]

PVABh2 218, 2–8:

tasmād gatir api pratyakṣābhimatā ’nyathā dṛṣṭā para-rūpeṇaiva |38 na kācit pravṛttiviṣaye svarūpeṇa gatiḥ | anyatra tu vasturūpe svarūpe vā gatir ubhayor apīti bhāvivastuni ko viśeṣaḥ |

Therefore, also [that] cognition, [namely, the one] that is considered to be perception, is observed [to be] otherwise, [that is], only [to cognize an object] with another nature. As regards the object of activity, [that] is not cognized at all with its own nature; but concerning another, [be it] the nature of the real thing or the nature of cognition, all two, [perception and inference], cognize [it]. So what difference is there with regard to the future object?[37]

This shows in what sense Prajñākaragupta takes perception and inference to be parallel: neither apprehends the future object directly, obviously impossible; but both do apprehend something else directly: what this is, Prajñākaragupta can still leave open at this point in the discussion, content to call it either the form of an (external) real thing (vasturūpa) or the form of cognition itself (svarūpa).[38]

5. Jewels and keyholes

In the next section of the PVABh2, Prajñākaragupta comments on a famous example given by Dharmakīrti. According to Prajñākaragupta’s interpretation, the example is supposed to show that a cognition, though mistaken, can be a means of valid cognition; this is possible because even a mistaken cognition can lead to the attainment of an intended aim.[39]

Dharmakīrti’s illustration of this involves two very similar situations: two people, thinking that they have seen a jewel, proceed to get that jewel. What they have seen is, however, not in fact a jewel, but only the shine of something: one of them has seen the shine of an actual jewel (spilling through a keyhole, according to Prajñākaragupta), the other the shine of a lamp. For Prajñākaragupta’s interpretation of this example, the following points are important:

1. The cognition “A jewel!” is the same in both cases.

2. It is also wrong in both cases, since both persons only see the shine of something, never the jewel itself.

3. There is a difference in that one person will attain a jewel, and the other will not.

4. Based on this difference, one does not say that cognitions are mistaken in the same way.

In a first step, Prajñākaragupta explains how this example shows the possibility that inference, though mistaken, can a means of valid cognition:[40] it is actually not due to it being a cognition, but rather to factors external to the cognition. Though everything about the two people’s cognitions is the same, the circumstances in which they have them is not. And this is what differentiates an erroneous and non-erroneous cognition.

A further explanation of the varying degrees of erroneousness centers on whether an error can belong to the causes that give rise to a means of valid cognition.

It is in the context of this explanation that Prajñākaragupta again introduces everyday activity as an important factor in determining what a means of valid cognition is, and which means of valid cognition there are:

PVABh2 II 244–245:

vinānumānaṃ pratyakṣaṃ na pravartakam āditaḥ |
tathānumānaṃ pratyakṣaṃ vineti pratipāditam ||
viśeṣas tv ayam evātra kvacit pūrvaṃ kvacit param |
anumānāt paraṃ nākṣaṃ nākṣāt pūrvānumeṣyate ||

Without inference perception does not, at first, prompt [a person] to act; like-wise inference without perception. This was explained.

[There is], however, certainly this distinction here, [that] in some cases [one is] earlier, in some cases later. One does not assume a perception following upon inference, nor an inference prior to perception.

In these two verses, Prajñākaragupta is stating that inference and perception are de-pendent on each other: perception would never allow a person to act if there were not—temporally prior to it—an inference that ascertains the connection between what is seen and what is acted towards. The important thing to note here is the qualifier āditaḥ, “at first” or “the first time”, which alludes back to the point that Prajñākaragupta made at the beginning of the chapter, namely, that habituation, the necessary prerequisite for perception’s directly leading to activity, cannot happen without inference.[41] And the same holds for inference: it does not prompt a person to act without perception, and could hence not be considered a means of valid cognition without perception.[42]

In PVABh2 II 245, Prajñākaragupta admits to a difference between the two different types of means of valid cognition: sometimes one is earlier, sometimes one is later;[43] and, as a kind of loose reason given for this difference, he puts forward what is not commonly held to be the case: that perception follows upon inference, or, in other words, that inference precedes perception. Implied, we must understand, is that the usually held view is correct, but has to be augmented by this other case. The commonly held difference of perception and inference is made with reference to cases where perception precedes inference (as in the usual case where a person sees smoke and infers ‘There’s a fire!’), and inference follows perception.[44]

6. The final position: self-awareness, perception, and inference

After the opponent states that, whereas inference does depend on perception, perception can occur without inference, Prajñākaragupta further clarifies his position:

PVABh2 II 246–248:

pratyakṣam anumānena vinā mānaṃ svave-dane |
vyavahāras tathā nāsti pramāṇatve’pi kiṃ bhavet ||
svasaṃvedanamātre ca pratyakṣe’rthāprasiddhitaḥ |
bhedasya ca na kiṃcit syād advaitam avaśiṣyate ||
tasmād arthasya bhedasya nādhyakṣasādhakaṃ vinā |
anumānaṃ tatas tasya pramātvaṃ nānumāṃ vinā ||

Perception without inference is a means of valid cognition concerning the awareness [any cognition has] of itself.[45] In that way, [however], there is no everyday activity. So even though it is a means of valid cognition, what should happen?

Furthermore, in a perception [that is] merely self-awareness, an [external] object is not established, nor a difference; so there would be nothing at all; [only] non-duality remains.

Therefore, perception does not establish an object [or] a difference without inference. Therefore, this [perception] is not a means of valid cognition without inference.

These three verses for the first time present the final position that Prajñākaragupta has had in view throughout the whole debate.[46] The two main points are:

1) Perception is a means of valid cognition in and of itself, but only as far as its own appearance to itself is concerned.

2) However, in order to count as a means of valid cognition in the conventional sense, i.e., as a cognition that allows a person to reliably attain a desired aim in interaction with external objects, it has to be supplemented by inference.

It is this distinction between perception in the sense of direct awareness of consciousness itself and the direct awareness of an external object that has been driving, and sometimes confusing, the discussion.

Prajñākaragupta is not prepared to admit that perception, taken as a cognition that reliably leads to activity, can be given the status of a means of valid cognition without also granting that status to inference. The reason for this is that perception, whilst in and of itself a means of valid cognition, does not let a person act. The only object that it can be considered to be a means of valid cognition for is its own appearance to itself. Any other object, or quality of an object, must be ascertained by another type of cognition. If one still insists, as the opponent does, that perception does reliably cause successful activity, then one must include a cognitive element which is not perception itself, and hence has lost the argument that only perception is a means of valid cognition.

To close, let us look at how this interpretation fits in with the example of the lamp and the jewel. Prajñākaragupta explain the errors that are involved in this situation:

PVABh2 221, 1–3:

maṇipratibhāsas tu maṇau maṇiprabhāyāṃ ca samāna eva | tatra kvacid deśabhrāntiḥ | kvacit svarūpabhrāntiḥ | kvacid ubhayam | kvacid anubhayam | maṇiprabhāyāṃ maṇijñānasya deśabhrāntir | maṇāv eva prāpyasvarūpabhrāntiḥ | sāmānyānumānasyobhayabhrāntiḥ | svasaṃvedana-sya nobhayathāpīti prakāraḥ |

The appearance of a jewel [to cognition], however, is exactly the same in [the case of] a jewel and in [the case of] the shine of a jewel. In this [situation], there is, in some respect, an error about the place; in some respect, an error about the proper nature; in some respect, both; in some respect, neither.

As regards the shine of a jewel, the cognition of a jewel is wrong about the place; for the actual jewel, there is an error about the nature of the [thing] to be obtained; the inference, [concerned with] a general property, is wrong about both [the place and the nature of the obtained thing]; [cognition’s] awareness of itself is wrong about neither. This is the manner [in which one has to understand this].

There are thus quite a few different respects in which the cognition of a jewel can be mistaken:

1) When one perceives not the jewel itself, but the shine of the jewel, the jewel cognition is wrong about the place of the thing that will be attained on account of this cognition.

2) When one is actually perceiving a jewel, the jewel cognition is still wrong about the essential nature of the object that will be attained (because the present jewel is not identical with the jewel that will be attained).[47]

3) An inference is wrong about both the place and the nature of the attained thing.[48]

4) The jewel cognition’s awareness of itself is wrong about neither the place nor the nature of what will be attained (though it is of little practical value that the gleam-of-a-jewel-appearance is of the same nature as the gleam-of-a-jewel-appearance that will be attained).

After having presented this analysis, Prajñākaragupta can make the first part of his closing statement:

PVABh2 221, 4–7:

tato yad uktaṃ yā gatiḥ sā svarūpeṇaiva yathā pratyakṣā gatiḥ | yat pramāṇaṃ tad abhrantaṃ yathā pratyakṣaṃ tad ayuktam | pra-tyakṣāpi gatir na svarūpeṇa | na cābhrāntaṃ pratyakṣam asti | svarūpe ca yathā pratyakṣam abhrāntaṃ tathā’numānam apy anye ca bhrāntābhimatāḥ pratyayā iti na tathā bhrāntatā pratipādanaṃ kvacid upayogi |

Therefore, what was said:[49] “That is a cognition which [cognizes something] according to [the thing’s] own nature, like perceptual cognition. What is a means of valid cognition is not erroneous, like perception.” is incorrect.

Even if it is perceptual, a cognition is not according to [the thing’s] nature. Nor is there a non-erroneous perception.

But as perception is non-erroneous with regard to [its] own nature, so also inference is [non-erroneous], as well as other cognitions which are assumed to be erroneous.[50] So an explanation as being erroneous in these ways is of no use in any case.

Prajñākaragupta is here succinctly restating the main points of his commentary on the example of the jewel and the lamp, which in turn exemplifies the points made in the section starting from PV III 53d: the whole section has been aimed at refuting an opponent who claims that ‘being erroneous’ is the criterion by which to differentiate what is a means of valid cognition from what is not.[51] And as Prajñākaragupta has made abundantly clear, this is not how means of valid cognition can be identified, at least not if one simultaneously wants to uphold that perception is such a means of valid cognition: one would have to admit that inference, and actually also other types of cognition which are generally held to be erroneous, are ‘erroneous’ and ‘non-erroneous’ in the exact same way as perception. All are wrong concerning the object to be attained (though they might be wrong in different ways),[52] and all are correct concerning their direct awareness of themselves. That they are wrong in some respect is a necessary condition for enabling everyday activity; that they are necessarily correct in terms of self-awareness might be a nice philosophical insight, but one that is not useful for normal people at this point.[53]

In the second part of the closing statement for this section, we can see that the direct awareness a cognition has of itself is essential for the differentiation of two, and only two, means of valid cognition, the other central point that was at issue:

PVABh2 221, 7–10:

evaṃ tarhi kathaṃ pratyakṣānumānayor lakṣaṇabhe-daḥ | uktam atra spaṣṭetarapratibhāsabhedād iti | sa eva pratibhāsabhedas tathā kuto bhavati | yasya svatantraṃ grahaṇaṃ tatra spaṣṭapratibhāsatā | yasyānyathā tad aspaṣṭapratibhāsam atīndriyaṃ parokṣam |

[Opponent:] If it is so, how [can there be] a different definition for perception and inference?

[Proponent:] To this [question] it was said that [it is] due to a difference in the appearance as distinct [or] not.[54]

[Opponent:] Exactly this different appearance in such a way, what does it come from?

[Proponent:] In the case of that which is grasped independently [from any form other than its own], there is a distinct appearance. That [which is grasped] in another way has an indistinct appearance, is beyond the senses, is remote.

With this, Prajñākaragupta has restated the main reason (as he saw it) for the differentia-tion of two types of valid cognitions: it is wholly ‘intrinsic’ to the content of the cognitions, and has nothing to do with the correctness or erroneousness of the respective cognitions.[55]

7. Conclusion

I have tried to show that Prajñākaragupta’s commentary on PV III 53d–57 is carefully composed and consistently argued in view of his commentary on PV III 1–2: the same central concerns underlie many arguments, and it can sometimes be misleading to focus on one particular segment of the text in isolation from its wider context.

The main points from Prajñākaragupta’s discussion, as I have analyzed it here, can be restated as follows:

1) To be a means of valid cognition is here being considered as making activity according to a cognizer’s expectations possible (pravartaka-tva is the main criterion for validity in these passages).

2) Under that definition, there are two such means of valid cognition, perception and another, based on the two ways in which objects appear in awareness: distinctly and indistinctly.

3) Both means are erroneous about the object that they enable a cognizer to obtain: that object does not appear in perception (which grasps a present thing), and the other type operates without particulars, one of which will be obtained.

4) Using any notion of “correspondence to an object” other than making successful activity possible, in order to distinguish what is and what is not a means of valid cognition, is doomed to failure, because to be a means of valid cognition implies non-correspondence to the very object to which it facilitates everyday activity; that object is never present to it.

Prajñākaragupta thus resolves the tension between PV III 1 and PV III 53d—that the two means of valid cognition result from two objects of valid cognition, but that there is in fact only one such object—on the basis of a complex argument that works on two levels: if everyday activity is to be possible, the duality of the means of valid cognition must be accepted; perception by itself cannot support this activity. That only the particular is sought for in this activity does not change this: perception and inference cannot be distinguished in their erroneousness about this future thing, and the main argument of PV III 1 has therefore to be understood to mean that there are two means of valid cognition because this particular, the future object of successful activity, is cognized in two ways, distinctly or indistinctly. Alternatively, it is possible to maintain that only perception is a means of valid cognition. Only the perception of cognition by itself exists in reality: this does, however, not support everyday activity, is not the topic of the pratyakṣa chapter, and is not acceptable to the opponent.

Quite a few factors make the interpretation of Prajñākaragupta’s work as difficult as it is interesting: its sheer extent and the unreliability of its edition are two major ones; they are compounded by Prajñākaragupta’s style of writing, nearly colloquial in its flow of arguments, and his skill in letting multiple threads run throughout long exchanges that shift in their main focus: the question of a cognition’s correctness, the general definition of a means of valid cognition, the distinction of the means of valid cognition acceptable to the participants in Prajñākaragupta’s debate, the relation between cognition’s accordance to an object and the definition of a means of valid cognition, and a number of assumptions that in the end turn out to entail unwanted consequences and have to be discarded.

I have here presented one example of such an exchange, taking care to weigh the arguments in respect to their dialectical context, in terms both of their content and of the commitment with which they are made. I believe this is the best way to come to a clearer understanding of the claims that Prajñākaragupta is actually making.

References and abbreviations

Arnold 2012 Dan Arnold, Brains, Buddhas, and Believing: The Problem of Intentionality in Classical Buddhist and Cognitive-Scientific Philosophy of Mind. New York 2012.

Dunne 2004 John D. Dunne, Foundations of Dharmakīrti’s Philosophy. Boston 2004.

Franco 1997 Eli Franco, Dharmakīrti on Compassion and Rebirth. Vienna 1997.

Franco 2004 Eli Franco, Towards a Critical Edition and Translation of the Pramāṇavā-rttikālaṃkārabhāṣya: A Propos Two Recent Publications. Wiener Zeitschrift für die Kunde Südasiens 48 (2004) 151–169.

Franco 2012 Eli Franco, How to Distinguish between Non-existing Entities?: Dhar-makīrti and Prajñākaragupta on Universals as Objects of Knowledge. Rocznik Orientalistyczny 65.1 (2012) 51–62.

Franco and Notake 2014 Eli Franco, Miyako Notake, Dharmakīrti on the Duality of the Object: Pramāṇavārttika III 1–63. Berlin et al. 2014.

Hattori 1968 Masaaki Hattori, Dignāga, On Perception: Being the Pratyakṣapariccheda of Dignāga’s Pramāṇasamuccaya from the Sanskrit Fragments and the Tibetan Versions. Cambridge et al. 1968.

HBS Dharmakīrti, Hetubindu. In: Dharmakīrti’s Hetubinduḥ. Vol. 1: Tibetischer Text und rekonstruierter Sanskrit-Text, ed. Ernst Steinkellner. Vienna 1967, 29–99.

Inami et al. 2002 Masahiro Inami et al., Prajñākaragupta ni okeru nishu no taishō to ninshiki-shudan (“Prajñākaragupta’s View on the Two Kinds of Valid Cognitions and the Two Kinds of Objects”). Nanto Bukkyō 81 (2002) 1–53.

Kajiyama 1998 Yuichi Kajiyama, An Introduction to Buddhist Philosophy: An Annotated Translation of the Tarkabhāṣā of Mokṣākaragupta, ed. Ernst Steinkellner. Reprint with corrections in the author’s hand. Vienna 1998.

Katsura 1993 Shōryū Katsura, On Perceptual Judgement. In: Studies on Buddhism in Honour of Professor A.K. Warder, ed. Narendra K. Wagle, Fumimaro Watanabe. Toronto 1993, 66–75.

Kellner 2004a Birgit Kellner, First logic, then the Buddha? The Controversy about the Chapter Sequence of Dharmakīrti’s Pramāṇavārttika and the Soteriological Relevance of Inference. Hōrin 11 (2004) 147–167.

Kellner 2004b Birgit Kellner, Why Infer and not just Look? Dharmakīrti on the Psychol-ogy of Inferential Processes. In: The Role of the Example (dṛṣṭānta) in Classical Indian Logic, ed. Ernst Steinkellner, Shōryū Katsura. Vienna 2004, 1–51.

Kellner 2010 Birgit Kellner, Self-Awareness (svasaṃvedana) in Dignāga’s Pramāṇasa-muccaya and -vṛtti: A Close Reading. Journal of Indian Philosophy 38 (2010) 203–231.

Kellner et al. 2020 Birgit Kellner et al., eds., Reverberations of Dharmakīrti’s Philoso-phy: Proceedings of the Fifth International Dharmakīrti Conference Heidelberg, August 26 to 30, 2014. Vienna 2020.

Kobayashi 2010 Hisayasu Kobayashi, Self-Awareness and Mental Perception. Journal of Indian Philosophy 38.3 (June 2010) 233–245.

Patrick McAllister 249

Kobayashi 2011 Hisayasu Kobayashi, Prajñākaragupta’s Interpretation of svalakṣaṇa. Journal of Indian and Buddhist Studies 59.3 (Mar. 2011) 1256–1261.

Krasser 1991 Helmut Krasser, Dharmottaras kurze Untersuchung der Gültigkeit einer Erkenntnis Laghuprāmāṇyaparīkṣā: Materialien zur Definition gültiger Erkennt-nis in der Tradition Dharmakīrtis. Vol. 2: Übersetzung. Vienna 1991.

Krasser 1995 Helmut Krasser, Dharmottara’s Theory of Knowledge in his Laghuprāmā-ṇyaparīkṣā. Journal of Indian Philosophy 23.2 (1995) 247–271.

McCrea 2011 Lawrence J. McCrea, Prajñākaragupta on the pramāṇas and their objects. In: Religion and Logic in Buddhist Philosophical Analysis: Proceedings of the Fourth International Dharmakīrti Conference; Vienna, August 23–27, 2005, ed. Helmut Krasser et al. Vienna 2011, 319–328.

McCrea and Patil 2006 Lawrence J. McCrea, Parimal G. Patil, Traditionalism and In-novation: Philosophy, Exegesis, and Intellectual History in Jñānaśrīmitra’s Apo-haprakaraṇa. Journal of Indian Philosophy 34.4 (2006) 303–366. Nagatomi 1967–1968 Masatoshi Nagatomi, Arthakriyā. The Adyar Library Bulletin 31–32 (1967–1968) 52–72.

NBhūṣ Bhāsarvajña, Nyāyabhūṣaṇa. In: Śrīmadācārya-Bhāsarvajña-praṇītasya Nyāya-sārasya Svopajñaṃ Vyākhyāṇaṃ Nyāyabhūṣaṇam, ed. Svāmī Yogīndrānandaḥ. Vārāṇasī 1968.

Ono 2000 Motoi Ono, ed., Prajñākaraguptas Erklärung der Definition gültiger Erkennt-nis: Pramāṇavārttikālaṃkāra zu Pramāṇavārttika II 1–7. Vienna 2000.

PS I Dignāga, Pramāṇasamuccaya 1. Reconstruction in Steinkellner 2005.

PV III Dharmakīrti, Pramāṇavārttika: Chapter on perception. In: Bukkyō ninshikiron no kenkyū: Hōshō “Pramāṇavārttika” no genryōron. Bukkyō ninshikiron no kenkyū, ed. Hiromasa Tosaki. 2 vols. Tōkyō 1979–1985.

PVABh-msB Prajñākaragupta, Prajñākaragupta’s Pramāṇvārttikabhāṣyam: Manuscript

B. In: Sanskrit Manuscripts of Prajñākaragupta’s Pramāṇvārttikabhāṣyam: Fac-simile Edition, ed. Shigeaki Watanabe. Patna, Narita 1998, 1–81.

PVABh-msE Prajñākaragupta, Prajñākaragupta’s Pramāṇvārttikabhāṣyam: Manuscript

E. In: Sanskrit Manuscripts of Prajñākaragupta’s Pramāṇvārttikabhāṣyam: Fac-simile Edition, ed. Shigeaki Watanabe. Patna, Narita 1998, 83–101.

PVABh2 Prajñākaragupta, Pramāṇavārttikālaṅkārabhāṣya. In: Pramāṇavārtikabhāshyam or Vārtikālaṅkāraḥ of Prajñākaragupta (Being a Commentary on Dharmakīrti’s Pramāṇavārtikam), ed. Rāhula Sāṅkṛtyāyana. Patna 1953.

PVABhIn Prajñākaragupta, Pramāṇavārttikālaṅkāra ad PV III 1–2. In: Prajñākaragupta ni okeru nishu no taishō to ninshiki-shudan (“Prajñākaragupta’s View on the Two Kinds of Valid Cognitions and the Two Kinds of Objects”), ed. Masahiro Inami et al. 2002, 1–53.

PVAṬ-t Jayanta, tshad ma rnam’grel gyi rgyan gyi’grel bshad (Pramāṇavārttikālaṅkā-raṭīkā). In: dpe bsdur ma, ed. krung-go’i bod rig pa zhib’jug lte gnas kyi bka’ bstan dpe sdur khang gis dpe bsdur zhus. Vol. 100–101. Beijing 2009, 100.759–101.1592.

PVAṬSt Yamāri, tshad ma rnam’grel rgyan gyi’grel bshad shin tu yongs su dag pa (Pramāṇavārttikālaṅkāraṭīkā Supariśuddhā). In: dpe bsdur ma, ed. krung-go’i bod rig pa zhib’jug lte gnas kyi bka’ bstan dpe sdur khang gis dpe bsdur zhus. Trans. by Sumatikīrti, rngog blo ldan shes rab. Vol. 102–103. Beijing 2009, 102.471–103.932.

PVin 2 Dharmakīrti, Pramāṇaviniścaya 2. In: Dharmakīrti’s Pramāṇaviniścaya: Chap-ters 1 and 2, ed. Ernst Steinkellner. Beijing, Vienna 2007, 45–101.

PVin 3 Dharmakīrti, Pramāṇaviniścaya 3. In: Dharmakīrti’s Pramāṇaviniścaya: Chapter 3, ed. Pascale Hugon, Toru Tomabechi. Beijing, Vienna 2011.

PVM Dharmakīrti, Pramāṇavārttika-kārikā. In: Pramāṇavārttika-kārikā (Sanskrit and Tibetan), ed. Yūsho Miyasaka. Quotes of chapters are by number, in the following order: pramāṇasiddhi quoted as “2”; pratyakṣa as “3”; svārthānumāna as “1”; parārthānumāna as “4”. Narita 1971–1972, 1–206.

PVV Manorathanandin, Pramāṇavārttikavṛtti. In: Dharmakīrti’s Pramāṇavārttika with a Commentary by Manorathanandin, ed. Rāhula Sāṅkṛtyāyana. Patna 1938–1940.

Steinkellner 2005 Ernst Steinkellner, ed., Dignāga’s Pramāṇasamuccaya, Chapter 1: A hypothetical reconstruction of the Sanskrit text with the help of the two Tibetan translations on the basis of the hitherto known Sanskrit fragments and the linguistic materials gained from Jinendrabuddhi’s Ṭīkā. Apr. 2005. fileadmin/Institute/IKGA/PDF/forschung/buddhismuskunde/dignaga_PS_1.pdf (last visited 09-2020).

TBh2 Mokṣākaragupta, Tarkabhāṣā. In: Tarkabhāṣa and Vādasthāna of Mokṣākaragupta and Jitāripāda, ed. H.R. Rangaswami Iyengar. Mysore 1952, 1–71.

Footnotes and references:


The material in this article was first presented at the XVIIth Congress of the International Association of Buddhist Studies, 2014, in Vienna. The research was conducted in the project (226063163) “Systems of epistemology in classical Indian philosophy: Prajñākaragupta (ca. 750-810) on the number of instruments of knowledge (pramāṇa)”, sponsored by the DFG, German Research Foundation. I would like to thank Birgit Kellner for her thoughtful and critical appreciation of the arguments I make here, and Markus Viehbeck for helping me with many Tibetan issues.


The Buddha’s special role in the context of means of valid cognition is beyond the scope of this article. See Motoi Ono’s contribution to this volume for a discussion of that issue.


The pratyakṣa (perception) chapter is here counted as the third of the Pramāṇavārttika’s four chapters, in accordance with how Dharmakīrti can be determined to have arranged them (see Kellner 2004a), and referred to as PV III. Prajñākaragupta comments on the pratyakṣa chapter in the second chapter of his Pramāṇavārttikālaṅkārabhāṣya (PVABh2).


The term arthakriyā is here translated as referring to the usefulness something can have in conventional, human activity. It is also a technical term in Buddhist pramāṇa theory (cf. Nagatomi 1967–1968, Dunne 2004: 256–260). In the current instance it can be taken in both ways: a thing fulfils a person’s aim, or, in the technical sense, it is able to cause an effect. In this article, I will render it as a technical term (by “causal efficacy”, or similar), when I think it is referring to the property of a real thing independently of that thing’s use by another being.


For adhimokṣa, and also abhiniveśa (a term relevant below), see Kobayashi 2010: n. 23.


Cf. Hattori 1968: 24, and nn. 1.13–14.


Franco and Notake (2014: 140) take the subject of matam to be Dignāga, translating “It is held [by Dignāga] …”, referring to PVV 132, 8–9, where an opponent asks whether PV III 54d–55b does not contradict Dignāga’s statement that “there is no [object] to be validly cognized apart from the particular and the universal” (cf. PS I 1, 19 and Hattori 1968: 24). I agree, but think the impersonal matam here carries the additional notion that, at least in a revised way, this is still maintained to be the case. The “we” I have added here is therefore meant to include Dignāga and Dharmakīrti. Also Prajñākaragupta relates this to Dignāga, see below.


The difference between conceptual cognition in general and inference in particular is a separate argument and is not discussed here.


There is a a slight textual variation in the text as quoted here and PV III 1: (pra)meyadvaividhyāt.


The term cakṣurādivijñāna suggests that a visual and an auditory cognition would have to be taken as different types of means of valid cognition because at least one element in each causal complex, the sense organ involved, is different. This is contrary to the Buddhist classification of both as a single means of valid cognition, perception.


Kobayashi 2011 has presented a concise analysis of two elements that are important to Prajñākaragupta’s interpretation of Dharmakīrti explored in the following sections: first, Kobayashi 2011: 1257–1259 shows that, against Dharmottara, Prajñākaragupta maintains that the object of activity which any means of valid cognition directs a person towards is always a future particular (bhāvivastu), and that this is the single meya that Dharmakīrti is referring to; second, Kobayashi 2011: 1259–1260 illustrates that, according to Prajñākaragupta, both perception and inference can be said to be erroneous with regard to this future object.


Inami et al. (2002) provide a critical edition, Japanese translation, and analysis of PVABh2 ad PV III 1–2. I do not read Japanese, but benefited a lot from the edition.


The term “self-awareness”, which renders the Sanskrit svasaṃvedana, here “… refers to the idea that all mental states and the factors like passion or feelings that accompany them are aware of themselves.” (Kellner 2010: 204) I will below also use the phrase “cognition’s awareness of itself” for svasaṃvedana to make it clear that it means that here the object of cognition is cognition itself.


Franco and Notake (2014: n. 2, p. 31) point out that Manorathanandin makes a similar argument. Prajñākaragupta’s argument here is also quoted and refuted in NBhūṣ 382, 3–9.


Cf. section 4, n. 56, and n. 48.


For Prajñākaragupta’s general analysis of cognition’s validity and its relation to activity, see Ono 2000 and Franco 2004. The explicit equation of “being a means of valid cognition” with enabling or motivating successful activity derives from PVM II 1–5, and was already elucidated by Devendrabuddhi and Śākyabuddhi (see Dunne 2004: 253–256). We find it fully developed as one of a number of equal definitions of validity in Dharmottara’s work, cf. Krasser 1995: 247–248, Franco 1997: 52.


Cf. Yamāri, (PVAṬSt: 103, 37):’on te zhes bya ba la sogs pas ni rang gi phyogs pa’i rtsod pa slong ba’o/ Yamāri goes on to describe a position held by these others (explicitly excepting Bhaṭṭārcaṭa), on which a conceptual cognition that has an object that accomplishes some end is different from perception, and therefore not a means of valid cognition. Inami et al. (2002: 26, n. 35) give more details.


PVABh-msB 84a5 reads mano’ntaram, which I think is possible: it could be understood as ‘another cognition’, which corresponds to the Tibetan translation. Inami et al. (2002) correct to mānāntaram based on the Tibetan. Prof. Inami informs me that the tat tu is missing in his edition only due to an unfortunate misprint, and it is represented in his Japanese translation.


See Katsura 1993 for a study of this “perceptual judgement” in Dharmakīrti’s writings.


See section 5 for how this position is taken up again in showing how inference and perception depend on each other.


See the following passage: PVABhIn II 28: uktam atra vinābhyāsān na pratyakṣe pramāṇatā | tato numānam evātra pradhānam iti gamyatām || (It was explained that without habituation there is no means of valid cognition in a perception. Therefore it must be understood that inference alone is the main element here.)


McCrea (2011) has analyzed this section. Whilst he makes many valid points, he does not take the overall argumentative context that is carried over into this section from the beginning of the pratyakṣa chapter into due account. This results in attributing the following positions to Prajñākaragupta, which directly contradict Dharmakīrti: (1) that perception is not free from error (McCrea 2011: 327); (2) that it is not perception, but rather the following moment of conceptual awareness that is a means of valid cognition (McCrea 2011: 323)—I do not agree with these attributions. Point one, though indeed Prajñākaragupta does make it, cannot be taken out of its context (see section 6 and n. 54). Point two is a position that the opponent is being forced into and cannot accept (see section 4).—Another issue that McCrea (2011: 323) discusses is that the conceptual awareness following perception does not have the fault of gṛhītagrahaṇa, or that this fault is not a reason to not be a means of valid cognition. The discussion of this point is beyond the scope of this article (see the comment in n. 40).


This idea of differences in the causal complex already appears in the analysis of PV III 2, most explicitly in PVABhIn II 49, and the accompanying prose, especially this passage: PVABhIn 44, 16–18: na hi sarvadā pramāṇadvitayaṃ prameyadvitayāt sādhyate | api tu sāmagrīsambhavād iti vayam brūmaḥ | sāmagrīsambhavaṃ ca paścāt pratipādayiṣyāmaḥ | (“For that there are two means of valid cognition is not always established from there being two objects of valid cognition. Rather, we say that [two means are established] because [they each] arise from [a specific] causal complex. And we will later explain [this] arising from a causal complex.”)—The context there is an opponent’s argument that, if one takes the two objects of valid cognition and the two means of valid cognition as, respectively, cause and effect, the inference proving the two means of valid cognition would be an inference from cause to effect, and therefore not certain. Prajñākaragupta’s answer is that in this case the inference is possible, because the effect is inferred not from an individual cause, but from the causal complex, one factor of which is the object of valid cognition.


Read paramārthataḥ with PVABh-msB 106b2, PVABh-msE 210b5 against paramārthaḥ PVABh2.


The last line, vyavahāri° … prameyadvaividhyam, is repeated in PVABh2. This must be a printing error, it is not found in either PVABh-msB 106b2 or PVABh-msE 210b6.


Franco (2012) and Franco and Notake (2014: 4–17) discuss how unreal things can be the objects of valid cognition.


We will see in section 4 that no present particular is the object of valid cognition: it is the future particular, both for perception and inference. In other words, to take PV III 53d as a statement that accords with reality means that the object of valid cognition mentioned there is neither of the two objects in PV III 1–3, the present particular or the universal.


PV III 56 is: ayathābhiniveśena dvitīyā bhrāntir iṣyate | gatiś cet pararūpeṇa na ca bhrānteḥ pramāṇatā || abhiprāyāvisaṃvādād api bhrānteḥ pramāṇatā | gatir apy anyathā dṛṣṭā pakṣaś cāyaṃ kṛtottaraḥ || See Franco and Notake 2014: 141–142 for a translation.


It would be interesting to know who the opponent is, exactly. It seems that it is someone who—like Prajñākaragupta, and Dharmottara before him—accepts that ‘prompting activity’ (pravartaka) is a central criterion for being a means of valid cognition, but links this capacity to the fact that the cognition has for its object something real as it is. If one, reasonably, assumes that only the Cārvākas would argue against inference as a means of valid cognition, then one could infer that they recognized pravartakatva as an important criterion for being a means of valid cognition.


There are two textual difficulties here. First, for pararūpasyaiva: PVABh-msB reads pararūpasyavāsau, but this is not completely clear, and PVABh-msE reads pararūpasyāsau. Both seem possible, and the could easily have gotten lost. Since PVABh-msB is really not clearly legible here, I think one should follow Sāṅkṛtyāyana’s reading.—The second problem is with prāptau bhrāntir eva: PVABh-msB originally read prāntireva and was then corrected, but the marginal addition is now illegible. PVABh-msE supports prāptau bhrāntir eva. The printed bhrāntau bhrāntir eva (PVABh2 215, 26–27) must be due to either Sāṅkṛtyāyana misreading the correction here or a misprint. Both prāptir and prāptau seem possible. I prefer the latter, in accordance with PVABh-msE, and probably PVABh-msB post correctionem.


It could also be that one attains some altogether different element, Z. But the important thing is that Prajñākaragupta has to concede that one does not attain Y.


To understand iṣyate like this does not preclude that Dharmakīrti himself did indeed take inference to be an erroneous cognition, cf., e.g., PVin 2 1cd: …bhrāntir api sambandhataḥ pramā || The point is rather that to take inference like this has no bearing on the question of whether it is a pramāṇa or not; likewise, for perception, its not being erroneous is irrelevant for its status as a means of valid cognition.


A part of this passage is also translated and discussed in McCrea 2011: 321–322.


As McCrea (2011: 322) notes, already Dharmottara considered “this implication of the theory of momentariness for […] perception.” See Krasser 1991, 1995; McCrea 2011; McCrea and Patil 2006.


My interpretation here diverges from the interpretation of the prose part of this passage by McCrea (2011: 323), who takes it as expressing Prajñākaragupta’s own position. I think one has to read the passage as a prasaṅga, that is, a consequence unacceptable to the opponent (and to Prajñākaragupta, too), but not as a position that Prajñākaragupta would endorse.—Note that Kobayashi 2011: 1257–1258 identifies Dharmottara as the opponent in the statements that follow the passage just quoted and translated.


That the following passage contains a summary was first suggested by McCrea (2011: 325). Apart from clearly rephrasing PV III 56c, gatir apy anyathā dṛṣṭā, this is also explicitly stated by Yamāri, PVAṬSt: 103, 279–280: de ltar rnam pa gzhan gyis kyang rtogs pa mthong zhes bya ba’i rgya cher bshad nas mjug sdud pa ni de’i phyir zhes bya ba’o/ (“Having explained [the verse] gatir … dṛṣṭā in detail, [Prajñākaragupta] states the summary [of this explanation] with [the word] tasmāt.”)


See Kobayashi 2011: 1259–1260 for another translation and short discussion of this passage, alongside two others in which Prajñākaragupta presents alternative interpretations of Dharmakīrti’s phrase gatir apy anyathā dṛṣṭā. I follow Kobayashi 2011: 1259–1260 here as far as the future object is concerned.


McCrea (2011: 323, n. 13) notes in this context that Prajñākaragupta does not think that a conceptual cognition following perception grasps what is already grasped, and therefore accepts it as a means of valid cognition. In this claim, Prajñākaragupta would be in direct contradiction to Dharmakīrti and Dharmottara. (For Dharmakīrti, see the summary in Kellner 2004b: 9–10, referring to these two passages: HBS 2, 18 ff., and a prose passage to PVin 3 k. 48. For Dharmottara, see Krasser 1995: 248–249.)—To understand Prajñākaragupta’s argument here, one has to take into account his discussion of the second pramāṇa definition in PVM II 5c, ajñātārthaprakāśo vā, where ajñātārthaprakāśa excludes gṛhītagrahaṇa. Franco (1997: 50) translates an important passage of it (corresponding to Ono 2000: 79, 15–19): “This, (i.e., illumination of an unapprehended object) is the definition of the means of knowledge relating to absolute reality, whereas the previous one (i.e., a cognition that does not belie) is [a definition] of [the means of knowledge] relating to the conventional (sāṃvyavahārika).” It would thus not be surprising if, in the current context of discussing the conventional difference of means of valid cognition, the problem of gṛhītagrahaṇa were to play a secondary role. A full analysis is, however, beyond the scope of the current article.


Cf. the introduction of the passage by Prajñākaragupta, PVABh2 218, 28: avisamvādāt pramāṇatve’pi bhrāntatāṃ darśayati | (“[Dharmakīrti] explains that [a cognition], even though [it] is a means of valid cognition because it does not belie [a person’s intentions], is erroneous.”)


McCrea (2011: 321) stresses the point that Prajñākaragupta takes the example of a jewel’s and a lamp’s shine as exemplification of both inference and perception. This is true, especially if one reads the later passages, but the initial analysis of these verses by Prajñākaragupta does talk only about inference (and what is not really an inference), PVABh2 219, 2: tadvad anumānatadābhāsayor api tata eva pramāṇetarate | (“Like that [example of the jewel and the lamp], also inference and what [only] seems to be an [inference] are, just because of that, a means of valid cognition and something else.” The phrase ‘just because of that’ refers to what makes the difference in the case of the jewel and the lamp: the difference in attainment of what is desired.) Prajñākaragupta does, then, at least start off with what can be taken as Dharmakīrti’s intent in these verses. The “abrupt reversal of this emphasis” (McCrea 2011: 321), that is, of this example being only for inference, is actually not very abrupt: Prajñākaragupta analyzes the example of the jewel and the lamp first in terms of inference, then (as we will see shortly) makes the point that inference and perception are interdependent, and only then goes on to defend Dharmakīrti’s position, that inference and pseudo-inference have the same relation as the cognitions of the people in the example, by showing that, if one were not to accept this argument about inference, also perception would be vulnerable to the same problem. Rather than seeing Prajñākaragupta break sharply with the intent of Dharmakīrti’s example, we could thus equally well speak of his careful explication of it, perhaps with a view to opponents that had been trying to use the example against the Buddhist epistemologists.


Cf. n. 21, and the paraphrase of PVABhIn II 24 above, section 2.2.


I am not yet quite sure how exactly to understand this point.


Jayanta explains concisely, PVAṬ-t: 101, 97: la lar dang por zhes bya ba ni mngon sum mam rjes su dpag par yang ngo/ (“In some [cases] earlier, [that is], in [the case of] perception or also in [the case of] inference.”)


One of the standard uses of inference is to clear up wrong notions that arise after a perception, cf. Kellner 2004b: 6–9.


I have taken the locative svavedane here as signifying the object of perception, or what perception applies to. The locative could also state the condition for the main clause, “In [the case of] the [percep-tual] awareness [that every cognition has] of itself, perception is a means of valid cognition without inference.” This does not commit Prajñākaragupta to a particularly difficult position. But in the light of Prajñākaragupta’s commentary on PV III 63 (PVABh2 223, 19–20, discussed in n. 48), I prefer to take the locative as expressing the object of perception.


That this is his accepted position can be seen from his commentary on PV III 63, which marks the end of the discussion of how two objects are the reason for two means of valid cognition (see Franco and Notake 2014: 23):—PVABh2 223, 19–21: prameyaṃ pramāṇena sidhyati | pramāṇasvarūpaṃ tu svasaṃve-danākārasiddheḥ | jñānākāra eva ca svasaṃvedanaḥ svasāmānyatayopalabhyamānaḥ pratyakṣānumānaviṣaya ity uktam | viṣayadvaividhyād ākāradvaividhyād ity arthaḥ | svākāradvayasaṃvedane hi naikam iti yuktam ||—The object of valid cognition is established by the means of valid cognition. The own nature of the means of valid cognition, however, [is established] on [the basis of] the establishment of the form of [cognition’s] awareness of itself. And it was stated that [this] awareness [that cognition has] of itself, the very form of cognition [itself], is the object of perception [or] inference, [depending on whether] it is being apprehended as being [only] itself or common [also to other things]. [To say] “because there are two [types of] object” means “because there are two [types of] form [of cognition].” For, given [cognition’s] awareness of two forms of itself, it is incoherent [to say:] “There is one [object of valid cognition].”—This passage is also discussed and partially translated in Franco and Notake 2014: 149–150 (note that they emend °ākārasiddheḥ to °ākārasiddham).—I was unable to find an exact quote of the explanation referred to in this passage, but the differentiation made here is closely parallel to that made repeatedly for the distinct and indistinct forms of cognition, see n. 15, PVABhIn 39, 6–9 (discussed section 2.1), and PVABhIn 36, 25–27 (discussed section 2). The viṣayadvaividhyāt rephrases the expression “prameyadvitvena” in tasmāt prameyadvitvena pramāṇadvi-tvam iṣyate (PV III 63cd), which in turn might echo the expression “meyadvaividhyāt” of PV III 1 (cf. Franco and Notake 2014: 23).


If my interpretation is correct, this error would also apply to the situation described in the previous point. It would further imply that inference has (at least) the same two errors as that perception of the shine of a jewel: it is mistaken about the place of the inferred thing, and also about the nature of what will be obtained. This would actually suit Dharmakīrti’s example (an inference likened to a ‘slightly wrong’ perception). But Prajñākaragupta’s formulation of these different errors is not quite clear on this point: is each error specific to each case (but not the only one that is applicable), or is it the only error in each case? If the latter, it would contradict my understanding.


It is wrong about the nature because the thing to be attained is not its object; an inference does not have particulars as its object at all, but only general concepts. It is wrong concerning the place because an inference cannot say exactly where the inferred thing is placed: cf. PVABh2 II 249, where it is explained that one can infer only the general existence of a fire behind a wall, but not its specific location.


I was not able to identify a statement that corresponds to this literally, but the same argument was made in the context of Prajñākaragupta’s comment on PV III 55cd, PVABh2 215, 26–27, discussed above, section 4: there too, an opponent makes non-erroneousness a criterion of validity, and maintains that cognizing a thing with a form other than its own cannot be valid.


Cf. the quotation in TBh2 19, 7: svarūpe sarvam abhrāntaṃ pararūpe viparyayaḥ. (“Every cognition is non-erroneous regarding its own nature, but wrong concerning another nature.”) Kajiyama (1998: 52, n. 118) notes that the quote is not identified, and refers to PVABh2 331, 13–14 which expresses the same idea. I thank Birgit Kellner for alerting me to this passage.


Yamāri points out that the opponent here is “Cārvāka and so on”, PVAṬSt: 103, 295:’dis ni tshu rol mdzes pa la sogs pa gsal ba ma yin no zhes mjug sdud pa ni/ des na zhes bya ba’o/


From the list given above (section 6), it follows that the erroneousness of perception consists in not grasping the nature of what is to be obtained by it (prāpyasvarūpabhrānti, PVABh2 221, 2–3). The observation by McCrea (2011: 327)—that “[…] Prajñākara is […] flatly contradicting one of the most fundamental claims that Dharmakīrti himself made about perception” by saying that “perception is not free from error”—has to be modified: even according to Prajñākaragupta it is non-erroneous with regard to the present thing; but it is erroneous with regard to what is to be obtained by the activity that it could lead to. Taken in this sense, the contradiction to Dharmakīrti vanishes. Prajñākaragupta is only clarifying that pratyakṣa is not a means of valid cognition because it is non-erroneous.


In a soteriological context, however, cognition’s awareness of itself is a central issue, especially in Prajñākaragupta’s discussion of PVM II 1–6, edited in Ono 2000. To illustrate this with just one passage, cf. Franco 2004: 168: “Means of knowledge is only self-awareness, the single [type of] perception; there is no other, because the multiplicity (prapañca) [of the means of knowledge taught by the Buddha merely] follows [the needs of] the people to be trained.” See also Arnold 2012: 187–88.


There are various passages this could be referring to. The most likely candidate, in my opinion, is the last statement before the example of the jewel and the lamp is introduced, PVABh2 218, 26: tasmāt parokṣaviṣayapravartakatve’pi spaṣṭāspaṣṭabhedāt pramāṇadvitayam eva | See also n. 15.


Note also that, whereas this is a sufficient criterion for the distinction of perception and non-perception, it is certainly not a sufficient one for the distinction of inference and conceptual cognition. But Prajñā-karagupta is here trying to show only that there must be perception and at least one other type of valid cognition. For this, the argument is sufficient.

Help me keep this site Ad-Free

For over a decade, this site has never bothered you with ads. I want to keep it that way. But I humbly request your help to keep doing what I do best: provide the world with unbiased truth, wisdom and knowledge.

Let's make the world a better place together!

Like what you read? Consider supporting this website: