Proofs of Idealism in Buddhist Epistemology : Dharmakīrti’s Refutation of External Objects

To appear in: Joerg Tuske (ed.), The Bloomsbury Research Handbook of Indian Epistemology and Metaphysics. (Bloomsbury Handbooks in Asian Philosophy). London u.a.: Bloomsbury 2017

This article is dedicated to Tosaki Hiromasa and Tilmann Vetter (†), without whose pathbreaking contributions to the study of Dharmakīrti’s theory of perception in the Pramāṇavārttika and the Pramāṇaviniścaya it could never have been written.

Medieval Indian Buddhist philosophers developed a variety of proofs of “mere-cognition” (vijñaptimātra(tā)), the signature concept of the Yogācāra school which stands for the principle that when cognition is aware of objects, there is just (mātra) that awareness, and no external object that would correspond to its content. Mere-cognition is not found in the earliest Yogācāra literature, and it only gradually rises to the prominent position that it occupies in pertinent works of Vasubandhu (probably between 350 and 430 CE),2 and in later philosophical literature predominantly of the Buddhist tradition of logic and epistemology (pramāṇa), initiated by Dignāga (ca. 480-540 CE or slightly earlier) and Dharmakīrti (active between mid-sixth and mid-seventh century CE).3

1 Research for this article was undertaken within the research group “Practices of Argumentation in Transcultural Perspective” (directed by Joachim Kurtz and Birgit Kellner) of the Cluster of Excellence “Asia and Europe in a Global Context – the Dynamics of Transculturality” at the University of Heidelberg. I am, as always, grateful to John Taber, whose readings of Buddhist philosophy have greatly enriched my approach (though he may not agree with my conclusions). 2 This date is conjectured by Deleanu 2006: vol. 2, 186-194. 3 Cf. Schmithausen 2001 for a succinct historical overview on vijñaptimātra(tā). Cf. Krasser 2012 for a recent proposal to move Dharmakīrti up from Frauwallner’s 600-660 CE and place his time of activity in the

There has been much debate in recent times whether mere-cognition philosophically represents a form of idealism.4 More specifically, objections were raised to earlier interpretations that consider mere-cognition to correspond to what in the history of western philosophy has been called subjective idealism: the view that objects cannot exist without being cognized, or, as it is sometimes put, that there is no “mind-independent world”. But an important methodological consideration seems to be largely missing in this debate. The positions that scholars take in the idealism debate are typically supported through studies of the semantics of key terms, of textual context, of doctrinal and intellectual background, and of individual arguments that individual Buddhist thinkers advance in support of mere-cognition. But when all that has been, rightly, taken into

consideration, the question remains whether the various individual proofs form part of a whole. Do they indicate a strategy, a method? More generally, is the attribution of a philosophical “position” to a historical author complete without an enquiry into a method, strategy or style of reasoning that she pursues? Such questions are all the more pertinent for Dignāga and Dharmakīrti, who engaged in the construction of rigorous theories of inference and proof. In addition to pursuing certain practices of argumentation that might follow more or less established patterns and conventions, they explicated standards of validity that were to serve as yardsticks for all argument. This, then, raises the question how individual areas of investigation (such as mere-cognition) are related to explicit logical theorizing. What first springs to eye in the different proofs of

mere-cognition found in the works of Vasubandhu, Dignāga and Dharmakīrti is that they advance through a primarily critical method that refutes external objects. This is not surprising, for arguing that external objects do not exist (or cannot be known) is an evident choice of method for establishing that whenever cognition is aware of something, there is just that awareness, and not an middle of the sixth century. Krasser argues that Bhāviveka (490/500-570 CE) referred to Dharmakīrtian ideas and arguments. The implications of this proposal, as well as its methodology, remain to be assessed in consideration of a wider context. 4 Some positions in this debate are reviewed in Kellner/Taber 2014.

external object. Vasubandhu argues along these lines in his Viṃśikā Vijñaptimātratāsiddhiḥ, the “Proof of Mere-Cognition in Twenty Stanzas” that comes with a prose autocommentary. Dignāga disproves external objects in his “Investigation of the Object-Support” (Ālambanaparīkṣā, short ĀP), a treatise in verse furnished with a prose auto-commentary, the Ālambanaparīkṣāvṛtti (ĀPV). But it is Dharmakīrti who presents by far the broadest array of arguments against external objects, in the chapters on perception (pratyakṣa) of his two comprehensive epistemological treatises, the “Commentary on the Means of Valid Cognition” (Pramāṇavārttika, short PV) in verse, and the “Ascertainment of the Means of Valid Cognition” (Pramāṇaviniścaya, short PVin) in verse and prose. To bring Dharmakīrti’s arguments into relief, it will be helpful to first review the arguments

advanced by Vasubandhu and Dignāga. Vasubandhu’s position has been a matter of some debate, and some have argued that it differs strongly from that of Dharmakīrti. Dharmakīrti, for his part, continues to some degree along lines of argument established by Dignāga. 1 The refutation of external objects in Vasubandhu’s Viṃśikā and Dignāga’s Ālambanaparīkṣā Idealist readings of Vasubandhu’s Viṃśikā (short Vś) tend to focus on a set of famous and widely quoted arguments offered in Vś 11-15. Vasubandhu here argues that objects of perception neither exist as continuous wholes (avayavin), nor as a multitude of minute and indivisible particles, nor as a single such particle, i.e., an atom. Like Dignāga and Dharmakīrti, Vasubandhu focuses his discussion on the perceptual awareness of objects; the term “cognition” in what follows should therefore be understood as referring primarily to sense perception. For Vasubandhu, the analysis of perceived objects into physical parts is necessary – for otherwise a host of anomalies would occur –, but the very idea of a single, indivisible atom turns out to be logically impossible. For one, atoms would have to be connected with each other to form macroscopic objects. But if an atom were simultaneously connected with other atoms on each of its six sides (the cardinal directions, above

and below), it would have to have six parts and could no longer be single.5 Non-idealist interpreters of the Viṃśikā construe this argument as showing that “reason” cannot postulate external objects as causes of experience, since they are logically incoherent,6 or as pursuing the more limited, phenomenalist claim that the objects we are experiencing are not physical objects.7 While these and other arguments against idealist readings of the Viṃśikā cannot be easily dismissed, we have recently argued8 that one can still make better sense of the text as a whole as pursuing subjective idealism. Our argument is extensive and cannot be reproduced here in full. But the main points are as

follows. In the ninth chapter of his Abhidharmakośabhāṣya, Vasubandhu argues for a negative thesis – that a self as substance over and above cognitions does not exist – through what effectively amounts to an extended argumentum ad ignorantiam: an ātman or pudgala does not exist because neither of the accredited means of valid cognition (pramāṇa) perception, inference or scripture offer any evidence for its existence. A closer look at the Viṃśikā led us to find the same strategy at work: external objects do not exist because there is no evidence for them through any pramāṇa. In this reading, the mereological arguments in Vś 11-15 simply support the larger argument that there is no scriptural evidence for external objects, for it clarifies how the Buddha’s teaching that the constituent factors are without essence (dharmanairātmya), epitomized in mere-cognition, is to be correctly interpreted: it does not proclaim the existence of objects of perception. It is true that Vasubandhu does not explicitly state the conclusion to the argument from ignorance – that objects do not exist –, but this can be accounted for in two mutually non-exclusive ways. On the one hand, it can be argued that he did not have to – what other conclusion is there to draw from the 5 Vś 12. See Kapstein 1988 for an insightful philosophical appreciation of this and other arguments from Vś 11-15, and Oetke 1992 for a rigorously argued different reading. 6 Hayes 1988: 100. 7 Oetke 1992. 8 Kellner/Taber 2014.

complete absence of evidence for external objects than their non-existence? It is not farfetched to believe that the non-existence of something is the best explanation for the utter lack of evidence for its existence.9 On the other hand, Vasubandhu’s reticence at stating his conclusion might have soteriological motivations, for in the closing stanza (Vś 22) of the Viṃśikā he stresses that mere-cognition in all its aspects is the domain of the Buddha – how dharmas truely are can only be realized in a higher meditative state, in transconceptual meditation or nirvikalpasamādhi. The point of departure for Dignāga’s argumentation in the ĀP(V) is that the “object-support” (*ālambana) of a perception has to fulfil two conditions: it must be a cause of perception and perception must have its appearance (*ābhāsa).10 Perception, in other words, is related to its

object through causation and resemblance. In Indian epistemological literature after Dignāga this ultimately representationalist theory of perception is taken to be characteristic for the Sautrāntika school of thought.11 With this definition as a yardstick, Dignāga develops an argument along the following lines. Some claim that individual atoms are the object-support because they cause perception, while others hold that a collection (Tib. ’dus pa, Skt. *samudāya) of atoms is the object-support because it appears in perception. What appears are, after all, spatially extended shapes and colours. Dignāga argues that the first proposal is wrong because perception does not carry the appearance of individual atoms, just like it does not carry the appearance of the sense-faculty (ĀP 1). 9 In Kellner/Taber 2014, we discuss arguments from ignorance in Indian philosophy more generally, and also address their logical limitations and the parameters that make some of them plausible abductive evidence (though they are not deductively valid). 10 The ĀP(V) are not extant in Sanskrit (save for a few quotations), but only preserved in Chinese and Tibetan translation. Asterisks indicate that Sanskrit terms of the ĀP(V) are tentatively reconstructed from the Tibetan. My analysis is based on the Tibetan translation, and in its general contours seems consistent with the Dharmakīrtian tradition in India. Interpretations in Tang China, and even the translations by Paramārtha and Xuanzang, seem to differ to some degree from my analysis, cf. Lin 2007 and 2014, and Lusthaus 2014. For a recent philosophical appraisal of the ĀP(V) see Siderits forthcoming a. 11 The Chinese exegete Xuanzang, to whom we owe much insight into the intricate doctrinal fabrics of medieval Indian Buddhist thought, attributes this definition to “Hīnayāna masters” in general, with exception of Saṃmitīyas (La Vallée Poussin 1928: 42f.).

But while perception bears the appearance of the collection, the collection cannot be its cause, for it does not exist substantially, just like the second moon in the sky that is seen by persons suffering from floaters (*timira) (ĀP 2). Only atoms are causally efficacious, not collections. Even if Dignāga’s opponent were to modify his proposal, as he attempts to do, the fundamental problem remains that there is an incongruence between what is seen and what, in virtue of a given ontological theory, is believed to exist. What is seen is not real, and whatever might be real cannot be seen. Dignāga concludes that the objectsupport of perception is not outside, but inside consciousness (ĀPV 178,2f.). The objectsupport is the “cognisable form inside” (antar jñeyarūpam, ĀP 6) that merely appears as if it were external. As this appearance exists simultaneously with its perception, it seems to violate the requirement that cause and effect must exist in temporal succession. Dignāga consequently justifies that something simultaneous can be a c

ause, based on the principle of “logicians” (Tib. gtan tshigs pa) that “when the one is present or absent, the other is also present or absent” (*bhāvābhāvayoḥ tadvattvam, ĀP 7a with ĀPV).12 To preserve the notion of successive causality, one may alternatively consider that the form of an object produces a capacity (*śakti) – a seed (*bīja) – within the series of mental episodes that constitutes consciousness. This seed later gives rise to a perception with the same objectappearance (ĀP 7cd). Dignāga does not explicitly identify the object-support in this model; he does not say whether it is the earlier object-form or the seed. But the basic conclusion remains that the object of perception is internal (ĀP 8bcd). On the whole, Dignāga proceeds in the ĀP(V) by evaluating different candidates from a given ontology in terms of whether they can fulfil a given definition of the object of perception. This atomistic ontology is not subjected to any criticism comparable to that of Vasubandhu. It is simply determined as an unsuitable foundation of perceptual experi

12 Literally: “when [x] is present [or] absent, [y] has that (i.e., presence or absence).” For a different construal cf. Frauwallner 1930: 183. My construal is based on Uddyotakara’s explication of the same principle to explain the relationship between pramāṇa and pramāṇaphala at NV 6,11f. By “logicians”, Dignāga likely referred to Naiyāyikas (though not to Uddyotakara himself, who was active after him).

ence. Vasubandhu’s mereological arguments effectively demolish the notion of atoms – although this might not be his only and not even his main goal of proof –, whereas Dignāga’s arguments are logically consistent with the existence of a physical world made up of indivisible and minute material particles – a world, that, however, could not in any way causally impinge on human cognitive faculties. Both according to our novel reading of the Viṃśikā and according to earlier idealist interpretations, Vasubandhu argues that external objects do not exist (but in different ways). Dignāga’s arguments, on the other hand, only support the logically weaker, phenomenalist claim that external objects cannot be perceived, or that the objects of experience do not exist outside consciousness. Dan Arnold recently characterised these positions, respectively, as an “epistemic idealism”, expressed with the claim “that what we are immediately aware of must be understood in terms of the intrinsic properties of cognition”,13 and an ontologically committed “metaphysical idealism” that denies the existence of a physical world. At first sight, this distinction seems well-suited to capturing a significant philosophical difference between Vasubandhu and Dignāga. But as so often, bringing Dharmakīrti into the picture makes matters more complicated and offers opportunities for rethinking what initially appears to be obvious.

2 Dharmakīrti’s refutation of external objects

2.1 The framework: realism and idealism as different levels of analysis There is evidence to support that Dharmakīrti was generally committed to idealism. He wrote a treatise to prove the existence of other mental continua, the Santānāntarasiddhi, and there devoted considerable effort to averting the danger of solipsism. Who but a committed idealist would feel the need to counter solipsism?14 But a philosopher’s 13Arnold 2008: 15. Arnold here compares Dharmakīrti, not Dignāga, with Vasubandhu, but basically considers Dharmakīrti’s position to be the same that I have reconstructed for the ĀP(V). 14 I am grateful to Robert Sharf for bringing the relevance of this point to my attention.

commitment to a position need not mean that his arguments in favour of it are strong enough to support it; one cannot exclude that there are internal tensions, and it remains therefore necessary to attend to Dharmakīrti’s actual refutation of external objects. Moreover, Dharmakīrti’s attitude to a Yogācāra position is nuanced. His epistemological theory encompasses a theoretical hierarchy between a realist theory of cognition and its idealist counterpart – “idealist” at least in the sense of a theory that denies perceptual content to derive from, or correspond to, external objects that exist independently of consciousness.15 In the hierarchy of idealism and realism, idealism is superior because it provides the more accurate analysis of cognition, yet realism remains the default level of analysis in most areas of philosophy in which Dharmakīrti engages, notably in his theory of inference. This theoretical hierarchy is at the same time a soteriological one in that the idealist theory represents a level of analysis that

corresponds more closely to how beings that are further advanced on the Buddhist path to liberation are to experience reality. If the perceptual process is scrutinized, the view held “in the world” (loke) that perception is of external objects turns out to be misguided. As a matter of fact, perception only experiences – is aware of – an object’s form within itself; it does not experience an external object that differs from it in nature (parātman). This internal form is superimposed, projected outward, as it were, due to a “disturbance” (viplava, 431),16 that is, due to a systematic cognitive distortion issuing, no doubt, from ignorance (avidyā) as a fundamen15 See Dreyfus 1997, McClintock 2003 and Dunne 2004 for analyses of this hierarchy in terms of a “sliding” or “ascending” scale of analysis, and Kellner 2011a for a critical assessment of Dunne’s version of it on the basis of some of the material covered in the following. Aspects of the following section were also treated in Kellner 2009 (in German). In Kellner 2011a and 2011b I suggested to characterise the different positions as “externalist” and “internalist” (rather than “realist” and “idealist”), but in the present context the latter, more

established distinction is without harm, since the nature of the positions themselves is the topic of enquiry. 16 In the following three-digit numbers without any further qualification refer to stanzas in the PV’s chapter on perception. When variants are not explicitly discussed, the Sanskrit text corresponds to that given in the edition of Tosaki (1979 for stanzas 1-319, 1985 for stanzas 320-539). The verse numbering also corresponds to Tosaki’s. Tosaki did not have access to manuscript materials that became accessible more recently and permit more comprehensive text-critical work on the PV and its commentaries. Cf. Kellner 2010a for a comprehensive review of the PV’s editorial history and extant witnesses, and for first results of text-critical analysis, resulting in some emendations of Tosaki’s text and a better understanding of the mechanisms resulting in textual change.

tal conditio humana that on the Buddha’s diagnosis keeps unawakened human beings entrapped in saṃsāra.17 The appearance of cognition, by nature undifferentiated, as different in terms of subject and object, is also a “disturbance” (upaplava, 212, 214; viplava 331). It is out of an “inner disturbance” (antarupaplava) that cognition, one in nature, shows a multiplicity of false appearances, subject and object (361). Philosophically, distortions issuing from ignorance act as an error theory that allows to explain how ordinary beings come to think and experience their cognitive life in a way that does not represent how things really are.18 Scriptural teachings of the five psycho-physical aggregates of living beings (skandha) or of the twelve sense-spheres (āyatana) imply the existence of external objects, but such teachings are spread by the Buddha merely in

accordance with worldly understanding. In doing so they set aside non-duality (advaya) as that which is ultimately true (tattvārtha). In this the Buddha resembles an elephant who shuts one of its laterally set eyes and then only sees to one side (219; read with the commentary M1 184,10-14). 2.2 The transition from realism to idealism: defending Sautrāntika realism as a preparation for the proof of idealism The theoretical hierarchy of idealism and realism entails that one cannot speak of Dharmakīrti either as realist or as idealist. By default, he operates on the presupposition of external reality, suspending arguments that realism is flawed, but in certain contexts he explicitly argues for idealism and occasionally also addresses the relationship between these two levels of analysis. Of particular interest in this regard, and for the refutation of external objects as such, is an extended argument that makes a transition of realism to idealism.19 Here Dharmakīrti first defends a Sautrāntika-style theory of perception based on 17 For in-depth studies of ignorance in Dharmakīrti’s thought cf. Eltschinger 2009, 2010. 18 Cf. also Kellner 2009: 70f. 19 This argument occurs in the first part of his treatment of the means of valid cognition and its result (pramāṇa/pramāṇaphala), at PV 3.301-337 and the close parallel PVin 1 30,9-36,9, which takes Dignāga’s pithy remarks in PS(V) 1.8cd-10 for its point of departure. Dignāga employs an idealist analysis of percep

causation and resemblance against brahminical interlocutors (301-319).20 He then turns to questioning and ultimately rejecting the supposition, which underlies this theory, that perception is of external objects (320-337). Although the initial defense of the Sautrāntika model may seem to be of little relevance to proving idealism, it can be seen as the first step in Dharmakīrti’s proof of idealism, for it establishes a key ingredient of it and also prepares the ground for questioning realism in the first place. The point of departure for Dharmakīrti’s discussion is the connection of cognition with a particular object – or, in other words, cognition’s object-specificity. It is obvious

(pratyakṣa) to everyone from their experience that awareness is restricted (prativedana), first and foremost insofar as it is limited to particular objects.21 Awareness is always awareness of something; awareness does not occur to us pure and simple. But what accounts for cognition’s object-specificity? Dharmakīrti argues that whatever differentiates cognition according to its object must belong to cognition itself, or to cognition’s nature (ātman). Cognition is differentiated according to its object by having the object’s form (ākāra, rūpa) within itself; this constitutes cognition’s “resemblance” (sārūpya) to the external object.22 Nothing extraneous to cognition, like one of its causes, can account for its object-specificity. One criterion for determining the factor that accounts for it is to examine whether cognition’s object changes when that factor changes. This, Dharmakīrti argues, is not true for the sense-faculty, one of the causes of perception. Whether the sensefaculty is sharp (paṭu) or dull (manda) certainly affects and in some way differentiates

tion in this context, but does not offer a refutation of external objects in the PS(V), which presents itself as the culmination of his life-work and was therefore most likely composed after the ĀP(V). 20 For this section cf. also the helpful brief summary in Taber 2005: 197f., n. 98, where correspondences in the pratyakṣapariccheda of Kumārila’s ŚV are pointed out. 21 The peculiar term prativedana, “restricted awareness”, occurs only at 320 (corresponding to PVin 1 33,11), but the object-specificity of perception constitutes the main explanandum throughout the entire section 301-337. The commentator Manorathanandin expands the compound prativedanam as pratiniyataṃ vedanam, and interprets it to also include cognition’s restriction to a particular mental series, not only to a particular object, cf. M1 215,9: nīlādyākāreṇa pratiniyataṃ vedanaṃ pratisantānaniyataṃ vā. 22 302-305, corresponding to PVin 1 31,4-32,3.

perception, but not in the relevant way of making it about a different object, for the “condition for the close connection [between perception and its object] is absent there” (pratyāsattinibandhanābhāvāt). But if one wished to argue that some specific property of experience, caused by the external object, but not tantamount to resemblance, accounted for object-specificity, then one would have to say what precisely this spurious feature might be. As Dharmakīrti ridicules his interlocutor in the PVin, if some indeterminate feature should determine of which object cognition is, things would be truly well determined!23 It is true that others have specified properties of cognition that they might claim to be caused by the object. But these are reducible to cognition’s having the object’s form, to its form-possession. What is the “(mere) seeing of an object” (arthālocana), used to define perception in a number of other schools,24 other than cognition’s having that object’s form? (310; PVin 1 32,15-33,1) Similarly, the cognition of

a qualifier (viśeṣaṇajñāna), such as the colour white, can only account for the cognition of what is qualified by it (viśeṣyajñāna), as e.g. a white cow, if both are distinguished by having the form of their respective object (313bcd; PVin 1 33,1-3). The sense-faculty, moreover, is common to every sense perception (312; PVin 1 32,15). Every visual perception is caused by the visual sense-faculty, hence the sense is not a suitable basis for distinguishing the seeing of blue from that of yellow. The “sense-object contact” (indriyārthasannikarṣa) that is according to Nyāya, Vaiśeṣika or Mīmāṃsā among the causes of perception applies to an object in its entirety. But one only sees objects

with some of their qualities, like colour, whereas others, like their being a mass of atoms, remain unseen (316; PVin 1 33,8-10).25 In sum, if 23PVin 1 31,11-32,2: na hi paṭumandādibhiḥ svabhedair bhedakam apīndriyārthenaitad ghaṭayati, tatra pratyāsattinibandhanābhāvat. asty anubhavaviśeṣo ’rthakṛtaḥ yata iyaṃ pratītiḥ, na sārūpyād iti cet. atha katham idānīṃ sato rūpaṃ na nirdiśyate? nedam idantayā śakyaṃ nirdeṣṭum. anirūpitena nāmāyam ātmanā bhāvān vyavasthāpayatīdam asyedaṃ neti suvyavasthitā bhāvāḥ. Cf. also 305cd. The term pratyāsatti for the close connection between a cognition and its object also occurs in 324, cf. below, p. ###. 24 Cf. Sāṅkhyakārikā 28, Praśastapādabhāṣya (Bronkhorst/Ramseier 1944: 44,11 and 45,11f.), and ŚV pratyakṣapariccheda 71-72a, 112-113 (Taber 2005). 25 The example is taken from M1 213,22. Dharmakīrti’s basic argument goes back to PSV 1 10,8-12 (cf. esp. sarvātmanā sannikarṣāt). The commentator Jinendrabuddhi traces this sannikarṣapramāṇavāda back to a Vaiśeṣika author named Śrāyasaka (PSṬ 1 118, 15). This author is not known from any other sources.

one holds that perception is of an external object, one has to grant that perception itself has that object’s form. No other factor involved in the causation of perception can account for its object-specificity. Whatever feature of perception itself might be invoked turns out to be reducible to “resemblance”, as specified in terms of form-possession, or is unsuitable, as the sense-object-contact. In order to appreciate Dharmakīrti’s point here as well as his further strategy, we should bear in mind that realists active prior to (or more or less simultaneous with) Dharmakīrti appear to have considered the object-specificity of cognition as evidence for realism.26 The Mīmāṃsaka Kumārila argued

that cognition’s object-specificity was only possible on the premise of external reality. His remarks from the beginning of the nirālambanavāda chapter of the Ślokavārttika even suggest an argument to this effect: if cognitions were devoid of external objects (nirālambana), scholastic distinctions like that between prior position and established position (pūrvapakṣa/siddhānta), and philosophical distinctions such as that between valid and invalid cognition would collapse – as would categories that form the foundation of ritual science.27 Much like Dharmakīrti, Kumārila elsewhere states that it is established for all living beings that something is apprehended having the form of blue or yellow, long or short.28 But for Kumārila, if one recognises the object-specificity of cognition, one also has to recognise external reality as its foundation. Vasubandhu counters a similar realist challenge at the beginning of the Viṃśikā by first pointing to dream experience: a spatial or temporal restriction (niyama) of cognitions is also known to occur in dreams, and dreams surely occur without external objects. Why should the same not also apply to waking cognitions?29 26 The relative chronology of Dharmakīrti and Kumārila is a complicated subject, but suffice it to say that they were most likely near-contemporaries, perhaps even contemporaries, so that interaction – each responds to the other – cannot be ruled out. 27 ŚV nirālambanavāda 1-3; cf. Taber 2010: 279f. 28 ŚV śūnyavāda 5 (cf. Taber 2010: 283): tatra tāvad idaṃ siddhaṃ sarvaprāṇabhṛtām api // grāhyataṃ nīlapītādidīrghādyākāravastunaḥ // 29 On our new reading of the Viṃśikā, Vasubandhu does not put forward a veritable “dreaming argument”

Considering a realist argument for external reality from cognition’s restricted nature as the backdrop of Dharmakīrti’s argumentation offers new perspectives. At this stage, Dharmakīrti believes to have brought his imaginary opponent to agree that a perception has its object's form. The opponent would thereby be driven to concede the general point that cognition is a type of entity that can have “form”. As is well-known, this is one of the main points of disagreement between Buddhist epistemologists and their Naiyāyika and Mīmāṃsaka opponents, as these opponents attribute “form” to external objects, and deny that cognition can have form; they are “advocates of formless cognition” (nirākārajñānavādin). By having established that cognition has form, Dharmakīrti already sets the stage for his refutation of external objects. If restricted awareness is nothing but cognition’s carrying the object’s form within itself, why should cognition be of an external object in the first place? What is the evidence that we cognise something external?30

2.3 Arguments against external objects Dharmakīrti’s actual arguments against external objects can be divided into two groups, although this division is not made explicit in the text. The first group of arguments point out defects of the Sautrāntika theory of perception, while the second group argues on the ground of the nature of cognition. It is striking that Kumārila distinguishes two types of arguments that others advanced against external objects along quite similar lines: those based on an examination of the object, and those based on (an examination of) the means of valid cognition, i.e. cognition as offering evidence for objects.31 for idealism, but rather uses this argument to dispute that there is inferential evidence for external objects (cf. Kellner/Taber 2014). 30 320c: tad (sc. prativedanam) arthavedanaṃ kena, corresponding to PVin 1 33,12. 31 ŚV nirālambanavāda 17: arthasya parīkṣaṇāt … pramāṇam āśritaḥ … Taber 2010: 280.

2.3.1 Defects of the Sautrāntika theory Dharmakīrti’s focus on pointing out defects specifically of the Sautrāntika model of perception, and not of realism in general, might be construed as a weakness in argument, but it becomes more plausible if the preceding defense of Sautrāntika realism is regarded as the first step in an extended argument to prove idealism. For on that account, Dharmakīrti first strove to commit his brahminical interlocutors to a causation-cum-resemblance account of perception’s object-specificity, and now believes to have established that perception has the form of its object. If the Sautrāntika theory is the most rational realist account of perception, which Dharmakīrti believes to have established, then it is sufficient to point out the defects of this theory in order to refute realism as a whole. Dharmakīrti first (321cd) picks up the argument from incongruence sketched in Dignāga’s Ālambanaparīkṣā and -vṛtti discussed above. While a number of minute and subtle atoms are supposedly the cause of perceptual awareness, what appears in it is a single, “coarse” (sthūla) form. Yet, this coarse, spatially extended form is not a feature of what causes perception, the individual atom.32 The next stanza 322 continues by saying “Therefore (i.e. given the argument from incongruence), cognition does not have the form of the object; or, if it had, that [form-possession] would be deviating (vyabhicārin).”33 The nature of the “deviation” is an illustrative case to show how commentators read their own agendas into the text. According to the commentator Manorathanandin, probably active in the second

32 At 195-196, Dharmakīrti presents a refined version of the Sautrāntika theory: many “conglomerated” (sañcita) atoms are said to be the external object of cognition, for the single atoms have the distinctive feature (viśeṣa) – of being capable to producing a perception – only when they are together, not individually. This explains why many things are the object of (non-conceptual) perception, yet without being a universal (sāmānya), which according to Dignāga and Dharmakīrti has to be grasped by conceptual cognition and cannot be apprehended by sense perception. But even in this refined theory, the atoms lack a coarse appearance (211). Cf. Dunne 2004 for extended discussion of this section of the PV. In the Pramāṇaviniścaya, Dharmakīrti complements Dignāga’s argument by adding a refutation of the “whole” (avayavin) of the Vaiśeṣika, as an example for a singular, spatially extended object; cf. PVin 1 34,10-35,6, with parallels in PV 2.84-85; cf. Funayama 1990, based on the Tibetan translation of PVin. 33 322ab: tan nārtharūpatā tasya satyāṃ vā vyabhicāriṇī. Cf. also 320: tad arthavedanaṃ kena tādrūpyād vyabhicāri tat / “By what is that (i.e. restricted awareness) an awareness of an external object? On account of [cognition’s] having the form of that [[[object]]]? That is deviating.” 11/25/16 - 14 half of the eleventh century, the brief mention of a “deviation” hints at a new argument that appeals to perceptual illusion. If the Sautrāntika accounts for restricted awareness by stipulating a resemblance with an external object, then this is fallacious because there are cases of restricted awareness where there is no resemblance. In perceptual illusions like that of a mesh of hair or a double moon experienced by someone suffering from the timira-disease (i.e., floaters), one is aware of a specific object, yet there is no external object to which the appearance in cognition would resemble.34 The Sautrāntika, wishing to resist such a move, might point to causation as an additional condition: whatever is the object also serves as the cause for the cognition that bears its form. But there is no real double moon that causes said illusion; the illusion would turn out to be not a case that makes the independence of cognitive content from external reality plausible, but simply an aberration, caused by some defect in the sense-faculty or some other cause of cognition.35 The following stanza 323 then points out a further problem that arises from the Sautrāntika account. Unlike Manorathanandin, the earlier commentator Devendrabuddhi (ca. 630-690) does not take Dharmakīrti’s mention of a “deviation” as a reference to perceptual illusion, but rather directly connects it with the argument that follows in 323.36 Dharmakīrti himself also does not speak of any additional “deviation” in the section of the Pramāṇaviniścaya that otherwise closely parallels these stanzas in PV, nor does he invoke an analogy with illusion there. This is worth emphasizing because earlier Yogācāra sources use dream experiences to make the doctrine of mere-cognition plausible.37 As already noted, Vasubandhu also appeals to the appearance of unreal objects in dreams in the very beginning of the Viṃśikā. The Buddhist position is even presented with a formal 34 M1 216,9 on 322 (dvicandrajñānādiṣu), and also M1 215,14f on 320: dvicandrakeśoṇḍūkajñānādyākārasyārtham antareṇāpi bhāvāt. 35 M1 216,10-12, introducing 323: na kevalād arthasārūpyād arthasaṃvedanatvaṃ yena vyabhicāraḥ syāt, kiṃ tarhi sarūpyatadutpattibhyām. te ca dvicandrajñānādīnāṃ na staḥ, candradvayasyābhāvāt tadutpatter ayogāt. 36 Det D219b4=P257b3. Devendrabuddhi makes no reference to the double moon or other illusory objects in his commentary on 320 and 322. 3

“dreaming argument” (Taber 1992) by Kumārila, and Prajñākaragupta (ca. 750-810 CE) and Jñānaśrīmitra (ca. 980-1030 CE), two significant and original thinkers in Dharmakīrti’s tradition, also advanced formal dreaming arguments.38 Yet, Dharmakīrti nowhere makes such an argument, and apparently did not place great weight on appeals to perceptual illusion in his refutation of external objects. The argument in 323 questions that causation-cum-resemblance unequivocally establishes that perception has external objects. Assume a situation where a person has two perceptions with identical object-appearances, e.g. blue, in immediate sequence. This situation is less contrived than it might initially seem. Given that the Sautrāntika assumes objects as well as mental events to be of only momentary existence, any seemingly continuous perception would in fact just be a succession of perceptual events with identical appearances. And many of our perceptions, if not all, are seemingly continuous. In this situation, the earlier perception is a cause of the later one; in the technical terminology that Abhidharmic analysis developed to classify the causes of perception, it is the “immediately preceding homologous

condition” (samanantarapratyaya); hence we can dub this argument the samanantarapratyaya-argument. Both perceptions have the same form of blue. The preceding perception therefore fulfils both conditions for being an object – causation and resemblance –, and it could therefore just as well be considered the object of the later one! The Sautrāntika believes that his definition of the object of perception by causation and resemblance limits the role of the “object” to an external object, but this is inconclusive.39 The Sautrāntika responds by pointing to a subsequent determinative cognition (ad38 ŚV nirālambanavāda 23. Cf. Taber 1992: 219, Taber 1994, Kobayashi 2011 (arguing that Vś 1 does present a formal inference), and Kellner/Taber 2014: 736, n. 94 (arguing that this is a less charitable reading of the stanza). In our new reading of the Vś, we take the first part of the Viṃśikā (Vś. 1-7) to be directed against inferential evidence for external objects: there are no reasons to postulate external objects because a number of facts that are usually explained by them can also be accounted for through the doctrine of merecognition. 39 323: tatsārūpyatadutpattī yadi saṃvedyalakṣaṇam / saṃvedyaṃ syāt samānārthaṃ vijñānaṃ samanantaram // PVin 1 33,12-34,1: anantaraṃ tarhi vijñānaṃ tulyaviṣayaṃ viṣayaḥ prāpnoti. Cf. also Kellner 2011a: 295 (following an interpretation first given in Tosaki 1985: 7), and Arnold 2008: 10 (advocating a different interpretation).

hyavasāya): When a determinative cognition with the content “this was seen” or “this was heard” arises after a perception, it must have been preceded by an experience of that which was seen or heard, i.e., of the object. But such a determination simply does not occur with respect to an immediately preceding cognition, hence that cognition is not the object. We do not determine “this preceding cognition was seen”. Yet, Dharmakīrti insists, it is precisely the close connection (pratyāsatti) between perception and its object that is under scrutiny: only when such a connection exists can a subsequent determination arise. That connection remains to be accounted for.40 And, to complete Dharmakīrti’s argument, if it were to be accounted for by causation and resemblance, then there would be no reason why the determination should not just as well refer to the preceding and homologous condition; the initially raised problem remains. Although the argument is premised on an ontology of exclusively momentary events, it does not logically depend on it. All that is needed is a realist view that considers mental events to have other mental events among their causes. When this is granted, a sequence of two cognitions with the same mental image would trigger the problem that both the external object and the preceding cognition fulfil the definition of an object of perception, i.e., to cause a subsequent cognition that has the same form.

2.3.1.1 Problems with resemblance In a later section of the Pramāṇavārttika, Dharmakīrti points out additional problems of the notion of resemblance that underlies Sautrāntika realism. If cognition resembled its putative external object in all respects, it would no longer be cognition; it would be an external object. But if it were only partially similar to the object, then “everything would be aware of everything else.”41 The cognition of a pot, Devendrabuddhi explains, would be 40 324-325: idaṃ dṛṣṭaṃ śrutaṃ vedam iti yatrāvasāyadhōḥ / sa tasyānubhavaḥ saiva pratyāsattir vicāryate // dṛśyadarśanayor yena tasya tad darśanaṃ matam / tayoḥ sambandham āśritya draṣṭur eṣa viniścayaḥ // (For vedam in 324, PrA’ reads cedam; for darśanaṃ in 325, attested in MA, PVZh, Rt, sādhanam is attested by PVt, PrA’, PrB.) Cf. also PVin 1 34,2–5; Kellner 2011a: 295. For a different interpretation of 324 see Arnold 2008: 10f. 41 434: sarvātmanā hi sārūpye jñānaṃ ajñānataṃ vrajet / sāmye kenacid aṃśena syāt sarvaṃ

that of potsherds, as the cognition does after all resemble potsherds in some respects because both potsherds and cognitions have the property of being cognisable (*jñeyatva). If one were to object that the cognition of the pot is not one of potsherds because it simply has the form “pot”, then the next problem arises: the cognition of one pot would be a cognition of all pots.42 Cognition’s possession of an object’s form should make it indexically linked to one specific object, to the one particular pot in front of my eyes. Alas, form-possession cannot do this work, for it is a generic notion.43 2.4 Arguments from the nature of cognition Dharmakīrti concludes from the argument from incongruence and from the samanantarapratyaya-argument that there is nothing other – no external object – to be experienced by cognition; cognition is only aware of itself. This is one of several claims that come to be condensely expressed in the notion of

svasaṃvedana, “reflexive awareness”. Now, that objects are obviously brought to awareness in a restricted fashion (pratyakṣaprativedyatva), the phenomenon to be accounted for, is also just cognition’s own nature. Cognition shines forth by itself (svayaṃ prakāśate); it is not of anything other, of an external object, and it is also, qua cognition, not brought to awareness by any further cognitive act.44 The sarvavedanam // The stanza is quoted in commentaries on ŚV śūnyavāda 20 with reversed halves, and with tu for hi (Kellner 2010a: 180, n. 54). The version cited here is preserved in PrA’, PrB and PVt. For the first half, Det, MA, Rt and PVZh read na ca sarvātmanā sāmyaṃ ajñānatvaprasaṅgataḥ, “And [[[cognition]]] is not completely the same [as its object], for then it would absurdly follow that it is not cognition.” For further textcritical discussion cf. Kellner 2010a: 200, esp. n. 111. 42Det D244b5f = P289b4-7. The second problem is also mentioned in M1 248,11. 43 Cf. King 2005 for similar arguments against

likeness-based theories of perception in medieval European philosophy. 44 326-327: ātmā sa tasyānubhavaḥ sa ca nānyasya kasyacit / pratyakṣaprativedyatvam api tasya tadātmatā // nānyo 'nubhāvyas tenāsti tasya nānubhavo 'paraḥ / tasyāpi tulyacodyatvāt tat svayaṃ tat prakāśate // Cf. Kellner 2010a: 196-199 for a discussion of variants in 327. Most significantly, for tat svayaṃ tat prakāśate (“Therefore, that [[[cognition]]] shines forth by itself”), which is attested in MA, PVZh, Rt and probably also Det, the witnesses associated with Prajñākaragupta’s commentary (including PVt) attest svayaṃ saiva prakāśate, which was also adopted by Tosaki and corresponds to the parallel stanza PVin 1.38 (“It is this very [[[cognition]], sc. buddhiḥ] that shines forth by itself”). As outlined in Kellner loc. cit., the PVin stanza selectively influenced the transmission of its counterpart in the PV. In place of tasyāpi tulyacodyatvāt in PV, PVin 1.38 has grāhyagrāhakavaidhuryāt.

second group of arguments against external objects that can be detected in Dharmakīrti’s works offers support for the first part of this two-part theorem: cognition is not of anything else. To the extent that it is aware of objects, it is only aware of an internal objectform contained within itself. Unpacking the two arguments in this group is not an easy task, for they are less explicit than the arguments in the first group. They have also, in traditional interpretations, been taken to serve additional purposes, over and above the refutation of external objects – and for one of them, the famous sahopalambhaniyama-inference, Dharmakīrti himself states that it can be used to establish that cognition has its own form and that of its object on the basis of realist presuppositions (397); the inference does not refute external objects on all interpretations he himself considers legitimate. As a result of these complexities, a full discussion would lead us too far; a limited outline with a focus on just how these arguments refute external objects if so understood will have to suffice. The first argument, referred to as the saṃvedana-argument by Takashi Iwata who provided a meticulous analysis of it,45

is presented in such cryptic form in the Pramāṇaviniścaya that a slightly elaborated paraphrase is more helpful than a literal translation: What is called ‘awareness’ (saṃvedana) is just appearing-in-a-certain-way (tathāprathana) because it has that nature (tādātmyāt). This awareness is not of anything else, just like the awareness of cognition itself is not of anything else. For this reason, too, it is not possible that awareness applies to a thing other than cognition itself.46 Cognition cannot be of some other, external thing; whatever it apprehends must be within itself. Object-awareness is like cognition’s reflexive awareness of itself 45 Cf. Iwata 1991: 9-15. 46 PVin 1 42,3-6: saṃvedanam ity api tasya tādātmyāt tathāprathanam, na tad anyasya kasyacid ātmasaṃvedanavat. tato 'pi na tad arthāntare yuktam. Cf. Iwata 1991: 9 (based on the Tibetan translation; the Sanskrit was not available at the time).

(ātmasaṃvedana), for cognition is by nature just an intransitive “appearing-in-a-certainway”, as opposed to a transitive apprehension of something else. Elsewhere, Dignāga and Dharmakīrti deny that cognition might perform any kind of activity (vyāpāra) directed at an external object.47 “To appear” (prathate)48 is an intransitive verb, for it makes no sense to ask the “what” question for an object: “What did you appear?” is a nonsensical question, as opposed to “what did you read?” or “what did you dream about?”49 To liken cognition with transitive activities such as cutting down a tree with an axe, as Dharmakīrti’s brahminical interlocutors as well as Sarvāstivāda Buddhists are prone to do, may seem in line with ordinary language use, but it is not analytically accurate. As Manorathanandin points out, when Dharmakīrti uses expressions like “cognition cognises itself” (dhīr ātmavedinī, 329), this is only metaphorical parlance. In reality, cognition simply arises with awareness for its nature, comparable to light that has luminosity for its

nature.50 The saṃvedana-argument professes that both dimensions of awareness that belong to cognition are ultimately one, and intransitive: its awareness of itself qua cognition, and its awareness of an object’s form that it contains within itself. Readers familiar with the more recent history of western philosophy may feel tempted to link this argument with so-called one-level accounts of consciousness as are characteristically advanced by philosophers in the modern phenomenological tradition, such as Jean-Paul Sartre.51 Conscious mental states are intrinsically and pre-reflectively conscious of themselves while being conscious of objects. This self-consciousness is not subsequently produced, or otherwise separate from object-consciousness. But the conclusion that Dharmakīrti draws 47 PS 1.8cd (Kellner 2010b: 219, with further discussion of background in Abhidharma and Yogācāra literature) and PSV ad PS 1.9d (op. cit., p. 223); for Dharmakīrti, cf. 308=PVin 1.37. 48 The verb form prathate is used in 349. 49 Cf. Legrand 2009 for a helpful clarification of the transitivity/intransitivity of consciousness. 50 M1 218,5-7. Cf. Watson 2014 for an illuminating (!) overview of different ways in which the light analogy is used in controversies between idealists in Dharmakīrti’s tradition and Naiyāyikas. 51 Zahavi 2005: 20ff.

here is one that phenomenologists shun: given a one-level account of consciousness, cognition cannot be of an external object. Dharmakīrti does not spell out the further implications himself, but the direction of his argument seems to be that if cognition cannot apply to anything outside itself because it has the nature of intransitive awareness, by implication whatever might be “out there” and does not partake in cognition’s luminous nature (which is the foundation of its intransitivity) cannot become manifest at all. As later authors in Dharmakīrti’s tradition put it, “insentient” (jaḍa) objects simply cannot appear because to appear means to have the nature of

“illumination” or “shining forth” (prakāśa), which is exclusive to consciousness.52 Even if insentient objects might be causes of cognitions, which the saṃvedana-argument when taken on its own does not strictly speaking rule out, they still cannot possibly appear within cognition. External reality is not perceptually accessible. This conclusion is one of the main obstacles for a wholesale assimilation of Buddhist epistemology to contemporary phenomenology, however instructive certain parallels might be.53 The second argument in this group is the famous sahopalambhaniyama-inference, the inference from the necessary joint cognition of an object like blue colour and its perception to their “non-difference” (abheda). This inference has a particularly rich tradition of interpretation. It is widely discussed among non-Buddhist philosophers and awaits full philosophical assessment.54 At its locus classicus in the Pramāṇaviniścaya, the focus of the inference is placed on the object’s non-difference from perception. Even though the form (rūpa) of blue appears as different from experience, it is not different from it because the two are necessarily cognised together, like the two moons seen by someone suffering

52 Cf. Watson 2014: 415, especially n. 37. One of the clearest articulations of this trope is TS 2000: vijñānaṃ jaḍarūpebhyo vyāvṛttam upajāyate / iyam evātmasaṃvittir asya yājaḍarūpatā // 53 Cf. Coseru 2012 for an illustrative attempt to work towards such an assimilation. 54 Taber 2010: 292f. gives some clues as to which direction such an assessment might take.

from floaters. Blue and its perception are cognised together because “there is no cognition of one of these two without a [simultaneous] cognition of the form of the other.”55 For good reasons, commentators have rejected an interpretation of saha (“together”, “joint”) as just indicating simultaneity, for there are after all many cases where ontologically different objects are cognised simultaneously. Simultaneous perception of objects simply does not equal identity.56 Rather, it must be necessary for perception to

be cognised when an object is perceived, and, conversely, that an object is perceived when its perception is cognised. The second claim seems less controversial. It seems reasonable that when I am aware of a perception that shows blue colour, blue colour is perceived. The first claim is more controversial, for it is not obvious that a perception of an object is not possible without a simultaneous awareness of that perception. In the Pramāṇaviniścaya, Dharmakīrti sets out to justify this claim by arguing that “for someone who does not perceive perception, the perception of the object is not established either.”57 Perception cannot be cognised after it grasps its object, for this would lead to an infinite regress since perception itself would have to be grasped by a further perception, and so on. It therefore has to be cognised simultaneously with its object. But whether Dharmakīrti anywhere establishes the premise of this argument, that all cognitions have to be cognised, remains to be studied; as presented in PVin, the argument seems questionbegging.58

55PVin 1.54cd: sahopalambhaniyamād abhedo nīlataddhiyoḥ. PVin 1 40,1-3: na hi bhinnāvabhāsitve 'py arthāntaram eva rūpaṃ nīlasyānubhāvāt tayoḥ sahopalambhaniyamād dvicandrādivat. na hy anayor ekākārānupalambhe 'nyopalambho 'sti. 56 Cf. Iwata 1991: 66-103 for a detailed overview of commentarial interpretations of saha in sahopalambhaniyama, including rejections of simple simultaneity. Dharmakīrti uses the word “simultaneous” (sakṛt) in 387, but also makes it clear in the immediately following stanza 388 that there is no necessity of a (joint) awareness for different objects like blue or yellow (saṃvittiniyamo nāsti bhinnayor nīlapītayoḥ). 57PVin 1.55ab: nāpratyakṣopalambhasyārthadṛṣṭiḥ prasidhyati. Cf. Kellner 2011b: 420, n.28 for a justifcation of this translation from the context, against one that takes the genitive apratyakṣopalambhasya as an absolute genitive, following the Tibetan translation dmigs na: “if perception is not perceived ...”. 58 In Kellner 2011b, I argued that Dharmakīrti does not establish this claim in the Pramāṇaviniścaya passage, but it is cannot be ruled out that some of the arguments presented in the second half of the

Commentators identify two further passages in the Pramāṇavārttika as stating the same sahopalambhaniyama-inference. Of these, the passage 387-390ab is closer to the Pramāṇaviniścaya version and contains the main ingredients of the inference: the conclusion is that there is no “separation” or “difference” (vivekitā, 389) of object and cognition, or, more focussed, that the object (artha) is not different from cognition, that it “does not extend beyond” (avyatirekitva, 390) cognition. The reason is that the object is necessarily brought to awareness simultaneously with cognition (sakṛt saṃvedyamānasya niyamena dhiyā saha viṣayasya …, 387; cf. also saṃvittiniyama in 388). In stanza 389, the necessity of the joint perception is explicated in two claims that anticipate the corresponding two-part explication in PVin: “no object is [observed being

experienced] without awareness; nor is awareness observed being experienced without an object” (nārtho 'saṃvedanaḥ kaścid anarthaṃ vāpi vedanam … saṃvedyamānaṃ dṛṣtaṃ); recall that the corresponding part in the version in PVin is “there is no cognition of one of these two without a [simultaneous] cognition of the form of the other”. In the passage 333-335, the realist first asks what would be wrong in assuming that an external object is experienced. Dharmakīrti answers: “nothing at all! Only this [remains to be asked:] Why would it be said that an external object is perceived?” At this point in the discussion it is established for both parties that cognition has form. The question remains whether this form originates from an external object or from imprints left by earlier experience in the mental series. Dharmakīrti answers in stanza 335: Because [something blue] is not apprehended without the additional qualifier (upādhi) of perception, [and] because [blue] is apprehended when this [qualifier of perception] is apprehended, perception has the appearance of blue. There is no external object by itself.59

Pramāṇavārttika’s chapter on perception, still largely unexplored, fulfl this task. For further philosophical reflection on the infinite regress cf. Siderits forthcoming b. 59 333-335: yadi bāhyo ’nubhūyeta ko doṣo naiva kaścana / idam eva kim uktaṃ syāt sa bāhyo ’rtho ’nubhūyate // yadi buddhis tadākārā sāsty ākāraviśeṣiṇī / sā bāhyād anyato veti vicāram idam arhati // darśanopādhi

The argument is very close to a sahopalambhaniyama-inference, if not fully identical with it: the conclusion is that there is no external object “by itself” (kevalaḥ), a conclusion that can plausibly be understood to mean that there is no external object that would be different of cognition, i.e., separate or independent from cognition. The reasoning to support this conclusion consists in a joint apprehension, expressed in two claims that structurally correspond to the ones from stanza 388. But there may be some significance to the characterisation of perception as an “additional qualifier” (upādhi) of the apprehended object. It is one thing to say that when blue is apprehended, it is always apprehended as qualified by its perception, but it is another thing to say that when blue is apprehended, its perception is also apprehended. Whenever I perceive blue, I

am aware of blue perceptually, but this does not have to mean I am aware of the perception of blue (or of perceiving blue). The argument presented in 335 may therefore be a weaker form of the sahopalambhaniyama-argument that does not yet involve the innate reflexive awareness of perception, svasaṃvedana, in quite the same way as the inference from PVin. But the conclusion, that there is no external object by itself – independent from cognition –, seems to be the same in all versions of this intriguing argument. Dharmakīrti continues, in 336, by stressing that cognition’s restricted nature does not depend on external objects; a mechanism of awakened imprints manages to account for it.60 Earlier experience leaves traces in the mental series that give rise to object-specific cognitions. Dignāga, as shown above, invoked a similar model at the end of the ĀP(V). Dharmakīrti thereby takes up the realist’s challenge that Vasubandhu’s opponent and Kumārila both raise: that restricted awareness demands external reality. Imprints in the

rahitasyāgrahāt tadgrahe grahāt / darśanaṃ nīlanirbhāsaṃ nārtho bāhyo ’sti kevalaḥ // (333d: for ’nubhūyate PrA’ reads ’nubhūyeta; 334: for -viśeṣiṇī PrA’ reads -niveśinī; 335: kevalaḥ PrB, and perhaps also PrA’ (difficult to read), PVZh, against kevalam MA. The Tibetan translation yan gar found in Det, PVt and Rt is inconclusive.) Cf. Taber 2010: 291 for translation and discussion of these stanzas. 60 336: kasyacit kiñcid evāntarvāsanāyāḥ prabodhakam / tato dhiyāṃ viniyamo na bāhyārthavyāpekṣayā //

mental series also help to explain the difference between valid and invalid cognition that according to Kumārila collapses without external objects: An invalid cognition is one arising from imprints left behind in the mental series by a disturbed cognition, and that therefore does not lead one to attain a desired goal. By contrast, a valid cognition is one arising from strong imprints (dṛḍhavāsanā) that has an uninterrupted connection with the desired goal and is reliable in action.61 Even the determination of the causal relation between seed and sprout or fire and smoke, and drawing inferences on its basis, is possible without assuming external objects (392-396), just on the basis of mental appearances; this is the doctrine of the wise (viduṣāṃ vādaḥ, 397). External objects are, in short, not required to explain any of the phenomena that realists explain by them. Dharmakīrti suggests an inference that nevertheless might be used to prove the existence of external objects: external objects could be proven “from absence” (vyatirekāt). When

all other causes for a perception are assembled, and perception still does not arise, its absence logically implies that an additional cause is needed – and that further cause might well be the external object, unless the idealist were to claim that that additional cause is a special material cause of the cognition, i.e. a preceding mental episode in the same mental series.62 It seems then, that this inference is a theoretical possibility, to be used if the (superior) idealist account is not adhered to, if it is suspended for the purpose of explaining things as they are in the world. 3 Conclusions: Dharmakīrti’s idealist position and aspects of logical method Arguments from Dharmakīrti’s rich and extensive, and still largely unexplored analysis of svasaṃvedana (425-539) might well have a further bearing on the refutation of external 61 PVin 1 43,14-44,6; cf. Krasser 2004: 143f. for text and translation. 62 PVin 1 58d: bāhyasiddhiḥ syād vyatirekataḥ, elaborated in PVin 1 43,10-12: satsu samartheṣu anyeṣu hetuṣu jñānakāryāniṣpattiḥ kāraṇāntaravaikalyaṃ sūcayati. sa bāhyo 'rthaḥ syāt, yady atra kaścid upādānaviśeṣābhāvakṛtaṃ kāryavyatirekaṃ na brūyāt. The basic argument is also offered in 390d-391ab (Krasser 2004: 142f.). Some traditional interpreters seem to take this inference as a response by the Sautrāntika to criticism by the Yogācāra, while others construe it as an expression of the Sautrāntika’s view that the external object is only inferred, and not perceived (see Kyūma 2011: 314, n. 28 for textual references).

objects. As the discussion of the second group of arguments has shown, Dharmakīrti’s arguments are terse to the point of occasionally being cryptic, and only a close reading – invariably guided by pointers from commentaries – can advance our understanding of them. But, for now, let us suppose that the arguments discussed above are Dharmakīrti’s main ones against external objects, and indicative of his strategy, assuming that he has one. First of all, it is evident from them that Dharmakīrti nowhere produces straightforward arguments that the external world does not exist. Like Dignāga, he does not advance the kind of mereological arguments against external objects that Vasubandhu put forward in Viṃśikā 11-15. Nor does he advance an argument from ignorance that, as we argued, can be uncovered from the Viṃśikā. Dharmakīrti’s first group of arguments is quite involved,

for it points out defects in one specific realist theory, the Sautrāntika’s, that is in his view the most rational one. These arguments lead Dharmakīrti to conclude that cognition is not of anything else – and this conclusion is then supported by the second group of arguments that argue from the nature of cognition. Nothing clearly indicates how Dharmakīrti conceived of the relationship between the two groups of arguments. But when compared to the first group, the second group, while far more difficult to interpret, presents stronger and more independent evidence for the absolute imperceptibility of external objects. It is not just that the most rational realist theory of perception has defects, but even more: there is something to the very nature of cognition that makes it impossible for it to be of external objects. Given that Dharmakīrti in all these arguments reveals himself as being chiefly concerned with what we perceive, it seems prima facie reasonable to read him as pursuing the point that the objects we experience are not, and indeed cannot be, outside consciousness. This may initially reinforce the plausibility of Dan Arnold’s distinction between an “epistemic idealism” (Dignāga/Dharmakīrti) and an ontologically committed “metaphysical idealism”

(Vasubandhu).63 But one challenge to this distinction was recently raised by Isabelle Ratié. If Dharmakīrti indeed argues for the complete and utter imperceptibility of the external world, is this not tantamount to arguing for its non-existence in a different way? Is not epistemic idealism already ontologically committed because what is absolutely imperceptible can only be non-existent? After all, in Dharmakīrti’s ontology existence is defined as causal capacity, through the notion of arthakriyā. If an object is fully imperceptible and can never produce even the minimal effect of a perceptual awareness, is it not thereby non-existent?64 At this point, considerations of logical method and theory become pertinent. Dharmakīrti was active after Vasubandhu, and the relationship between their views and arguments has therefore historical dimensions that abstract

comparison cannot properly grasp. Dharmakīrti introduced significant innovations in logical theory, notably the restriction of acceptable logical reasons to three, and only three, types. We should not only expect that his various proofs are in accordance with his own theory or at least do not openly violate it (however difficult the interpretation of that theory may turn out to be), but we should also consider that features of logical theory determine the nature of claims that can be proven with its help. Might the refutation of external objects be such a case? And is there anything in Dharmakīrti’s work that signals how he might have, or could have, conceived of Vasubandhu’s arguments against external objects from the vantage point of his own position? Dharmakīrti himself does not explicitly address Vasubandhu’s arguments. But as Arnold (2008) and Ratié (2014) have noted, a rare passage in Manorathanandin’s Pramāṇavārttikavṛtti on 335 speaks to this difference; it deserves to be quoted in full: 63 Cf. above p. ###. 64 Ratié 2014: 361.

“[[[Sautrāntika]] opponent:] No! Even so, [i.e.] even if there is no argument proving an external object, which is [on your view?]65 beyond the reach of the senses, the non-existence of external objects is not [thereby] established. [[[Manorathanandin]]:] Because [we] establish what [we] intend only as far as [saying] that cognition appears, whereas the external object does not appear at all, we do not have any regard for negating the external object, which behaves like an [imperceptible] demon [and] is without a means of valid cognition that proves it.66 But if the opponent were to strongly insist on negating the [[[external object]]], he should be made to examine the master [[[Vasubandhu’s]]] negation of atoms according to whether one supposes that [the external object] has parts or is partless.”67 The Sautrāntika opponent points out that Dharmakīrti’s argument does not prove the non-existence of external objects, even if he might have managed to establish that there is

no evidence for their existence. This, now, is a classic objection to arguments from ignorance: Absence of evidence does not constitute evidence of absence. Manorathanandin does not dispute the objector’s point: indeed, Dharmakīrti did not prove the non-existence of external objects by showing that there is no evidence for them. Manorathanandin does not say this explicitly, but a look at Dharmakīrti’s theory of inference supports an 65 It is not evident from the text whether the qualifier that the object is beyond the reach of the senses expresses the Sautrāntika’s own position or that which he attributes to the Yogācāra. The former entails that the Sautrāntika here claims the external object can only be inferred, and not perceived, but this is not such a fixed position that it can be unproblematically supplied to Manorathanandin’s argument (cf. above n. 63###). But this point of uncertainty does not affect the main argument. 66 Here my translation follows Ratié 2014: 359, n. 22, against Arnold 2008: 16. 67 M1 220, 16-20 (MA 43a4-6): na, tathāpi parokṣasya bāhyasya sādhakasyābhāve ’pi nābhāvasthitir iti cet, pratibhāsamānaṃ jñānaṃ bāhyaṃ tu na pratibhāsata eveti tāvataivābhimatasiddheḥ, sādhakapramāṇarahitapiśācāyamānabahirarthaniṣedhe nāsmākam* ādaraḥ. yadi tu tanniṣedhanirbandho garīyān sāṃśatvānāṃśatvakalpanayā paramāṇupratiṣedha** ācāryīyaḥ paryeṣitavyaḥ. [* For niṣedhe nāsmākam, the edition prints niṣedhenāsmākam. The manuscript MA reads niṣedhesmākam, with -nā- added in the bottom margin with a correction sign in the actual text. ** For paramāṇupratiṣedha MA, the edition M1 prints paramāṇupratiṣedhe.] Cf. also Ratié 2014: 359, n. 23. My translation follows Ratié 2014: 358 (with minor and largely stylistic differences), against Arnold 2008: 16 (who only translates Manorathanandin’s response).

even stronger point: Dharmakīrti could not have proven the non-existence of external objects in this way. In Dharmakīrti’s logical framework, negation is proven through a special type of reason called “non-apprehension” (anupalabdhi), but the scope of this reason is very limited. One can only infer that objects which would be necessarily perceived in a certain situation if they existed, but which are not perceptually apprehended, are suitable for cognitive, linguistic and physical treatment as non-existent.68 This type of reason does not permit universal ontological denial, for it presupposes that the negated object is of a kind that will necessarily be perceived (given the presence of all additional causes for its perception) if it exists. For that to be established requires that the object is one that can be perceived in principle. Thus restricted,

inferences based on the non-apprehension of a perceptible object can only prove the occasional (situational) absence of objects, not the non-existence of an entire class of objects. Throughout his works, Dharmakīrti offers different reasonings for why general arguments from ignorance are flawed.69 The most straightforward of these is that remote objects (viprakṛṣṭa) – objects remote in time or place, or by their very nature, such as piśāca-demons70 – cannot be proven as absent on the ground that they are not apprehended through perception, inference and scripture because they lack the causal capacity to produce a cognition of themselves. They might therefore exist without giving rise to a pramāṇa, and the absence of a pramāṇa for them cannot prove their non-existence. Dharmakīrti, in short, effectively eliminates arguments from ignorance from his theory of inference. It is, then, in keeping with his logical theory that the possibility for proving the 68 Cf. esp. Kellner 1999 and 2003 for further elaboration of Dharmakīrti’s complicated theory of non-apprehension. 69 For a brief overview cf. Kellner/Taber 2014. 70 To be precise, piśācas are thought of as imperceptibly by their nature (svabhāva) for human beings, i.e., in relation to a particular type of cognising subject. Cf. Kellner 1999 and now especially the pertinent remarks in Ratié 2014.

non-existence of external objects through arguments from ignorance is not even considered in Dharmakīrti’s works (as far as I can tell), and that he does not conclude from the absence of evidence for external objects that they do not exist. Dharmakīrti’s proofs that external objects are absolutely imperceptible may entail their non-existence, along the lines suggested by Ratié. The difference between epistemic and metaphysical idealism may thereby become minimized and lose its significance as a heuristic device for detecting different varieties of mere-cognition in Indian Buddhist thought. But his logical theory prevents Dharmakīrti from proving the non-existence of external objects through arguments that merely demonstrate their imperceptibility. This theory opens up an evidential gap, in a manner of speaking, between imperceptibility and non-existence. A

Dharmakīrtian might claim that totally imperceptible objects are as good as non-existent, but when further pushed to prove it would have to resort to other arguments than those used by Dharmakīrti to refute external objects, one where absence of evidence does not become evidence for absence. Such arguments could, of course, be precisely those that Vasubandhu put forward in Viṃśikā 11-15. According to Manorathanandin, Vasubandhu’s mereological arguments merit consideration only if the opponent were to stubbornly insist71 on negating the external object (proving its non-existence) after Dharmakīrti’s proofs of its imperceptibility have been pointed out to him. Vasubandhu’s arguments are not declared to be invalid; there is nothing formally wrong with them. It is only that they are irrelevant because they prove something that is not of value, since the epistemic inaccessibility of the external world is all there really needs to be shown. Metaphysical idealism, to use Arnold’s term, is for those who insist on pursuing irrelevant questions.72 71 Interpreting Manorathanandin’s nirbandha as “insistence” (Ratié: “obstinacy”) is one significant departure from Arnold’s interpretation, where the word is taken (oddly) to refer to a neutral desire. 72 Ratié (op. cit.) entertains both possibilities in her interpretation of Manorathanandin’s passage: that the ontological question becomes irrelevant or that epistemic idealism already entails an ontological position. Considering Dharmakīrti’s elimination of arguments from ignorance, the second possibility seems a less likely account of the systematic implications within Manorathanandin’s rich passage – even though, as

4 References

4.1 Primary Sources in Sanskrit and Canonical Tibetan Translations

ĀP Dignāga’s Ālambanaparīkṣā. See Frauwallner 1930. ĀPV Dignāga’s Ālambanaparīkṣāvṛtti. See Frauwallner 1930. ASBh Abhidharmasamuccaya-bhāṣyam. Deciphered and edited by Nathamal Tatia. 2nd ed. Patna: K.P. Jayaswal Research Institute 2005. Det Devendrabuddhi’s Pramāṇavārttikapañjikā. Tibetan translation. Tshad ma rnam ’grel gyi dka’ ’grel, translated by Subhūtiśrī(śānti) and (Rma) Dge ba’i blo gros. D 4217 Che 1-326b4, P 5717b Che 1-390a8. M1 Dharmakīrti’s Pramāṇavārttika with a Commentary by Manorathanandin. Ed. by Rāhula Sāṅkṛityāyana. Appendix to Journal of the Bihar and Orissa Research Society 24-26 (1938-1940). MA Sanskrit manuscript (on paper) of Manorathanandin’s Pramāṇavārttikavṛtti, written by Vibhūticandra. In: Shigeaki Watanabe (ed.): A Sanskrit Manuscript of Manorathanandin’s Pramāṇavārttikavṛttiḥ. Facsimile Edition. Patna/Narita: Bihar Research Society/Naritasan Institute for Buddhist Studies 1998. MSg Asaṅga’s Mahāyānasaṅgraha. La somme du Grand Véhicule d'Asaṅga (Mahāyānasaṃgraha) [édité et traduit]

par Étienne Lamotte. Louvain-la-Neuve, Université de Louvain, Institut Orientaliste 1973. NV Nyāyabhāṣyavārttika of Bhāradvāja Uddyotakara. Edited by Anantalal Thakur. New Delhi: Indian Council of Philosophical Research 1997. PrA’ A modern transcript of an incomplete paper ms. of Prajñākaragupta’s Pramāṇavārttikālaṅkārabhāṣya, written by Vibhūticandra, extending from the commentary on PV 3.302 to the end of the work. Microfilm of the Nepal-German Manuscript Preservation Project in Kathmandu (reel no. A1219/26) (cf. Kellner 2010a: 167 and 168). PrB Palm-leaf manuscript of of Prajñākaragupta’s Pramāṇavārttikālaṅkārabhāṣya. In: Shigeaki Watanabe (ed.): Sanskrit Manuscripts of Prajñākaragupta’s Pramāṇavārttikabhāṣyam. Facsimile Edition. Patna/Narita: Bihar Research Ratié then expounds, later Śaiva authors exploited it.

Society/Naritasan Institute for Buddhist Studies 1998. PS Dignāga’s Pramāṇasamuccaya. Steinkellner, Ernst. 2005. 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. http://www.ikga.oeaw.ac.at/Mat/dignaga_PS_1.pdf, with revisions of 2014: http://www.ikga.oeaw.ac.at/mediawiki/images/f/f3/Dignaga_PS_1_revision.pdf (last accessed 25 June 2015). PSV Dignāga’s Pramāṇasamuccayavṛtti. See PS. PSṬ 1 Jinendrabuddhi’s Pramāṇasamuccayaṭīkā. Ernst Steinkellner, Helmut Krasser, Horst Lasic: Jinendrabuddhi’s Viśālāmalavatī Pramāṇasamuccayaṭīkā, Chapter 1. Part I: Critical Edition, Part II: Diplomatic Edition. [[[Sanskrit]] Texts from the Tibetan Autonomous Region 1]. Beijing – Vienna: China Tibetology Publishing House – Austrian Academy of Sciences Press 2005. PV 3 Dharmakīrti’s Pramāṇavārttika, chapter 3 (pratyakṣa). See Tosaki 1979

(stanzas 1319) and 1985 (stanzas 320-539). PV 2 Dharmakīrti’s Pramāṇavārttika, chapter 3 (pramāṇasiddhi). See Miyasaka 19711972. PVt Tibetan translation of Dharmakīrti’s Pramāṇavārttika. Tshad ma rnam ’grel gyi tshig le’ur byas pa, translated by Subhūtiśrīśānti and (Rma) Dge ba’i blo gros, revised by *Bhavyarāja (Skal ldan rgyal po) and (Rṅog) Blo ldan śes rab, retranslated or revised by Śākyaśrībhadra and Sa skya paṇḍita. D 4210 Ce 94b1151a7, P 5709 Ce 190a4-250b6. PVZh Zha lu ri phug manuscript of PV, readings reported in: Pramāṇavārttikam by Ācārya Dharmakīrti. Edited by Rāhula Saṅkṛtyāyana. Appendix to Journal of the Bihar and Orissa Research Society 24 (1938). PVin 1 Dharmakīrti’s Pramāṇaviniścaya. Chapters 1 and 2. Ed. by Ernst Steinkellner. [[[Sanskrit]] Texts from the Tibetan Autonomous Region 2]. Beijing – Vienna: China Tibetology Publishing House – Austrian Academy of Sciences Press, 2007. For Corrigenda, cf. Wiener Zeitschrift für die Kunde Südasiens 51 (2007- 2008) 207-208, as well as http://ikga.oeaw.ac.at/Mat/steinkellner07_ corrigenda.pdf (last visited 26 June 2015). PVSV Dharmakīrti’s Pramāṇavārttikasvavṛtti. The Pramāṇavārttikam of Dharmakīrti: the First Chapter with the Autocommentary. Edited by Raniero Gnoli. Serie Orientale Roma 23. Rome: Istituto Italiano per il Medio ed Estremo Oriente, 1960.

Rt Tibetan translation of Ravigupta’s Pramāṇavārttikavṛtti. Tshad ma rnam ’grel gyi ’grel pa las le’u gsum pa. D 4225, Phe 1-174a7; P 5722, Phe 1-208a7. Translators unknown. Sāṅkhyakārikā Sāṅkhyakārikā. Yuktidīpikā. The Most Significant Commentary on the Sāṅkhyakārikā. Critically edited by Albrecht Wezler and Shujun Motegi. Stuttgart: Franz Steiner Verlag 1998. ŚV Svāmī Dvārikadāsa Śāstrī, Ślokavārttika of Śrī Kumārila Bhaṭṭa, with the commentary Nyāyaratnākara of Pārthasārathi Miśra. Varanasi: Tara Publications 1978. TS Dvarikadas Shastri: Tattvasaṅgraha of Ācārya Shāntarakṣita with the Commentary ‘Pañjikā’ of Shri Kamalashīla. 2 Vols. Varanasi: Bauddha Bharati 1968; reprinted 1981. TSP Kamalaśīla’s Tattvasaṅgrahapañjikā. See TS. Vś Vasubandhu’s Viṃśikā. Vijñaptimātratāsiddhi: Deux Traités de Vasubandhu: Viṃśatikā et Triṃśikā. Edited by Silvain Lévi. Paris: Libraire Ancienne Honoré Champion 1925.