Candrakīrti ’ s theory of perception : A case for non-foundationalist epistemology in Madhyamaka

Some argue that Candrakīrti is committed to rejecting all theories of perception in virtue of the rejection of the foundationalisms of the Nyāya and the Pramāṇika. Others argue that Candrakīrti endorses the Nyāya theory of perception. In this paper, I will propose an alternative non-foundationalist theory of perception for Candrakīriti. I will show that Candrakrti’s works provide us sufficient evidence to defend a typical Prāsagika’s account of perception that, I argue, complements his core non-foundationalist ontology. Setting up the problem The current debate on Buddhist epistemology operates largely on the assumption that the theory of perception (pratyakṣa / mngon sum) of Dignāga-Dharmakīrti (from hereon Pramāṇika) is (perhaps) the only one found in Indian Buddhism. This view, if it is taken seriously and I believe people do take this seriously, has two major problematic implications for the Prsagika Madhyamaka of Candrakīrti: (1) to the extent one grants a theory of perception in the Prsagika Madhyamaka, the Prāsaṅgika must be seen as endorsing the epistemological project of the Pramāṇika (or else must be seen as endorsing1 the theory of perception of Brahmanical Nyāya and Mīmāṁsaka); (2) to the extent one rejects the theory of perception of Buddhist foundationalism and Brahmanical substantialism in the Prāsaṅgika philosophy, Candrakīrti must be read as rejecting2 all available theories of perception in Indian Buddhism. 1 Mark Siderits (1981, 157) and (2011, 167–81) claims that Candrakīrti endorses Nyāya epistemology rather than Dignāga’s because Nyāya epistemology is less open to the reductionist project than is Yogācāra-Sautrāntika epistemology. 2 Tibetan Mdhyamikas Go rampa Bsod nams Senge (1969, 375, 382), Stag tsang Lots ba (2001, 156–58), and Dge ‘dun chos ‘phel or Gendün Chöpel (Dge ‘dun chos ‘phel 1990, 161) argue that for the Prsagika Mdhyamika there can be no theory of perception. They argue that Candrakrti rejects perceptual cognitions as epistemic instrument unequivocally. Three key arguments are used to support their position: First, what is ontologically unreal and deceptive must also be epistemically flawed. Since all perceptions are ontologically deceptive and illusory in virtue of being causally conditioned, they must be epistemically flawed. Thus the so-called conventionally authoritative perceptual cognition must be rejected unequivocally. Second, perceptual cognitions all reify their ISSN 1648–2662. ACTA ORIENTALIA VILNENSIA 11 .1 (2010) : 93–124 94 S O N A M T H A K C H O E The former is problematic for it assumes Candrakīrti is being inconsistent. It presupposes that Candrakīrti synthesises foundationalist epistemology with nonfoundationalist ontology. So far we do not have any good evidence to support the claim that Candrakīrti blends the two irreconcilable positions, although it is clear enough from the works of Bhāvavevika, Śāntarakśita and Kamalaśla that the Svātantrika philosophers do synthesise the two systems. The latter is also problematic since it assumes that Candrakīrti is an epistemological sceptic, because it rules out the uses of any alternative theory of perception in the Prāsaṅgika. In this paper, I propose an alternative non-foundationalist theory of perception for Candrakīrti. I show that Candrakrti’s works provide us sufficient evidence to defend a typical Prsagika’s account of perception that, I argue, complements his core non-foundationalist ontology. The paper has two parts. In the first I present a brief summary of the theory of perception in the Pramṇika. This is only brief since my intention is only to highlight the basic principles underpinning the Pramāṇika’s foundationalist theory of perception, which enables us to properly assess the distinctive characteristics of the Prāsaṅgika’s theory of perception and easily distinguish it from the Pramāṇika’s foundationalist account. The second part of the paper contains three subsections and is where I present Candrakrti’s alternative account of perception; this forms the core of my paper. The pramāṇika’s pratyakṣa (perception) In the Pramṇasammuccaya (PS), Dignga defines perception (pratyaka) as follows: ‘Reliable cognition (pramṇa / tshad ma) constitutes perception (pratyaka) and inference (anumṇa). Reliable cognition has dual characteristics, for it is associated with (two) objects (prameyas / gzhal bya). There is no other reliable cognition [PS 1.2] ... Perception is free from the conception that weaves together name (nma / ming), class (jti / rigs), etc.’ (PS 1.3cd in Dignga 2003, 1).3 The Nyyapravea (NP = Tshad ma rigs ‘jug)4 adds more to this definition. ‘Valid knowledge (pratyyana / rab objects under the influence of primal ignorance. Thus they are all flawed and epistemically invalid. Third, no perceptual cognitions enable one to perceive ultimate truth directly, and hence all are invalid. 3 Mngon sum dang ni rjes su dpag // tshad ma’i mtshan nyid gnyis gzhal ba // de la rab sbyor phyir tshad ma // bzhan ni yod pa ma yin no //1.2//. Ming dang rus sogs su sbyor ba’i // rtog pa dang dral mngon sum mo //1.3cd//. 4 There is still an unsettled dispute between the two camps regarding the authorship of this text due to variance in the source of the evidence on which they each rely. The Tibetan school (represented by Satischandra Vidybhaṇa, Pandit Vidhushekhara Bhattacharya, and Keith) attributes the text to Dignga, while the Chinese school (represented by Ui, Sugiura, Tucci, Tubianski and Mironov, etc.) attributes it to Dignga’s disciple, akara Svmin. See Dhruva 1930, i–xxxvii for an excellent synopsis of the debate at issue. The text I cited here from the Nyyapravea only clarifies Dignga’s 95 C A N D R A K Ī R T I ’ S T H E O R Y O F P E R C E P T I O N tu rtogs pa) of the nature of objects entails only two reliable cognitions (pramṇas): perception (pratyaka) and inference (anumna). Perception is free of conception (kalpanapoha / rtog dral). It is the cognition of objects like forms without the conceptions of name and class. Perception is a faculty, one that is acquired through the faculty of sense’ (Dignāga 1987, 7).5 A recent Indian commentator on the Nyyapravea, Sempa Dorjee, explains that perception in Dignga’s definition has the sense of being cognition that is unconditioned by linguistic universals (sgra spyi) (thoughts and concepts associated with a name (nma / ming), such as ‘Devadatta’) and the object universal (don spyi) (thoughts and concepts associated with a class (jti / rigs), such as ‘Cowness’). Along this line, in Pramṇavinicaya (PV), chapter 1, Dharmakrti explains, conception is a linguistic cognition. Conception is a cognition to which the [object] appears fused with language. Perception is free from the conception because conception is impossible in the sensory cognition since it arises due to the force of objects. Because perception arises through the force of an object, it exclusively accords with the reality [of its object]. (Dharmakrti 2003c, 171)6 Similarly, ‘Jam dbyang Bzhad pa’i Rdo rje (2006, 410) explains conception is a cognition that conceives (zhen rig) objects and language interwovenly: it conceives a pot for instance as though the abstract property of potness pervades all homogenous forms (namely individual pots having similar characteristics of being a pot), and it conceives many heterogeneous forms clustered together as universals (tshogs spyi) such as [forest] for [clusters of individual] trees.7 definition of pratyaka offered in (PS 1.3cd in Dignga 2003, 1) and does not in any way represent a variance from the original position. So I will assume that it is authoritative to cite for the purpose of clarifying the definition of pratyaka at issue. 5 tmapratyyanrtha tu pratyakamanumna ca dve eva pramṇe // Tatra pratyaka kalpanpoha yajajnamarthe rpd nmajtydikalpanrahita / tadakamka prati vartata ite pratyaka //. Dorjee 1996, 179, 181: Bdag nyid kyis rab tu rtogs pa’i don la ni mngon sum dang / rjes su dpag pa ste / mtshad ma la gnyis nyid [du nges pa] yin no // de na mngon sum rtog pa dang dral ba ste / shes pa gang gzugs la sogs pa’i don la ming dang rigs la sogs pa’i rtog pa dang dral ba’o // de ni dbang po dang dbang po so so la yod a ni mngon sum zhes pa’o//. Grags pa rGyal mtshan’s Tibetan translation (1992, 163): Bdag nyid kyis rab tu rtogs par bya ba’i don la yang mngon sum dang / rjes su dpag pa’i mtshad ma nyid dag go // de la Mngon sum rtog pa dang dral zhes ba ste / gang gzugs la sogs pa’i don la ming dang rigs la sogs pa’i rtog pa dang dral ba de dbang po so so’i dbang po la yod a ni mngon sum zhes pa’o //. The latter version ignores the Sanskrit terms dve and jnam, obscuring the sense of the passage, whereas the former correctly reflect the Sanskrit version. 6 Rtog pa de yang gang zhi na // rtog pa mngon par brjod can gyi / shes pa / rtog pa ni brjod pa dang ‘drer rung ba snang ba’i shes pa ste / de dang dral ba’o // dbang po’i shes pa la ni de srid pa ma yin te / don gyi mthu yis ni / yang dag skye ba yin phyir ro // gang gi phyir de don gyi nus pa las skyes pa na de’i rang bzhin kho na’i rjes su byed pa’i phyir ro //. 7 Rtog pa gang yin zhes na / kho na ri // ming don ‘dres par ‘dzin pa dang bum pa lta bu rang gi rigs can kun la khyab par ‘dzin pa dang // rigs mi mthun pa’ shing sogs du ma ‘dus pa’i tshogs spyi ‘dzin pa’i zhen rig cig yin te //. 96 S O N A M T H A K C H O E In addition, Dharmakrti in his Pramṇavinicaya (PV) mentions several other aspects of perception. The PV 1 says perception and inference are the two forms of ‘correct cognitions’ (yang dag pa’i shes pa) and defines correct cognition as ‘nondeceptive with regards to its epistemic activities (don bya ba la slu ba med) because it engages with its objects by means of discrimination (yongs su bcad nas ‘jug)’ (Dharmakīrti 2003c, 167).8 Elsewhere the PV 1 says perception is a reliable cognition (pramāṇa) because it is


Setting up the problem
The current debate on Buddhist epistemology operates largely on the assumption that the theory of perception (pratyakṣa / mngon sum) of Dignāga-Dharmakīrti (from hereon Pramāṇika) is (perhaps) the only one found in Indian Buddhism.This view, if it is taken seriously and I believe people do take this seriously, has two major problematic implications for the Prsagika Madhyamaka of Candrakīrti: (1) to the extent one grants a theory of perception in the Prsagika Madhyamaka, the Prāsaṅgika must be seen as endorsing the epistemological project of the Pramāṇika (or else must be seen as endorsing 1 the theory of perception of Brahmanical Nyāya and Mīmāṁsaka); (2) to the extent one rejects the theory of perception of Buddhist foundationalism and Brahmanical substantialism in the Prāsaṅgika philosophy, Candrakīrti must be read as rejecting 2 all available theories of perception in Indian Buddhism.
2 Tibetan Mdhyamikas Go rampa Bsod nams Senge (1969, 375, 382), Stag tsang Lots ba (2001,  156-58), and Dge 'dun chos 'phel or Gendün Chöpel (Dge 'dun chos 'phel 1990, 161) argue that for the Prsagika Mdhyamika there can be no theory of perception.They argue that Candrakrti rejects perceptual cognitions as epistemic instrument unequivocally.Three key arguments are used to support their position: First, what is ontologically unreal and deceptive must also be epistemically flawed.Since all perceptions are ontologically deceptive and illusory in virtue of being causally conditioned, they must be epistemically flawed.Thus the so-called conventionally authoritative perceptual cognition must be rejected unequivocally.Second, perceptual cognitions all reify their ISSN 1648-2662.ACTA ORIENTALIA VILNENSIA 11.1 (2010): 93-124 The former is problematic for it assumes Candrakīrti is being inconsistent.It presupposes that Candrakīrti synthesises foundationalist epistemology with nonfoundationalist ontology.So far we do not have any good evidence to support the claim that Candrakīrti blends the two irreconcilable positions, although it is clear enough from the works of Bhāvavevika, Śāntarakśita and Kamalaśla that the Svātantrika philosophers do synthesise the two systems.The latter is also problematic since it assumes that Candrakīrti is an epistemological sceptic, because it rules out the uses of any alternative theory of perception in the Prāsaṅgika.
In this paper, I propose an alternative non-foundationalist theory of perception for Candrakīrti.I show that Candrakrti's works provide us sufficient evidence to defend a typical Prsagika's account of perception that, I argue, complements his core non-foundationalist ontology.The paper has two parts.In the first I present a brief summary of the theory of perception in the Pramṇika.This is only brief since my intention is only to highlight the basic principles underpinning the Pramāṇika's foundationalist theory of perception, which enables us to properly assess the distinctive characteristics of the Prāsaṅgika's theory of perception and easily distinguish it from the Pramāṇika's foundationalist account.The second part of the paper contains three subsections and is where I present Candrakrti's alternative account of perception; this forms the core of my paper.

The pramāṇika's pratyakṣa (perception)
In the Pramṇasammuccaya (PS), Dignga defines perception (pratyaka) as follows: 'Reliable cognition (pramṇa / tshad ma) constitutes perception (pratyaka) and inference (anumṇa).Reliable cognition has dual characteristics, for it is associated with (two) objects (prameyas / gzhal bya).There is no other reliable cognition [PS 1.2] … Perception is free from the conception that weaves together name (nma / ming), class (jti / rigs), etc.' (PS 1.3cd in Dignga 2003, 1). 3 The Nyyapravea (NP = Tshad ma rigs 'jug) 4 adds more to this definition.'Valid knowledge (pratyyana / rab objects under the influence of primal ignorance.Thus they are all flawed and epistemically invalid.Third, no perceptual cognitions enable one to perceive ultimate truth directly, and hence all are invalid.
4 There is still an unsettled dispute between the two camps regarding the authorship of this text due to variance in the source of the evidence on which they each rely.The Tibetan school (represented by Satischandra Vidybhaṇa, Pandit Vidhushekhara Bhattacharya, and Keith) attributes the text to Dignga, while the Chinese school (represented by Ui, Sugiura, Tucci, Tubianski and Mironov,  etc.) attributes it to Dignga's disciple, akara Svmin.See Dhruva 1930, i-xxxvii for an excellent synopsis of the debate at issue.The text I cited here from the Nyyapravea only clarifies Dignga's tu rtogs pa) of the nature of objects entails only two reliable cognitions (pramṇas): perception (pratyaka) and inference (anumna).Perception is free of conception (kalpanapoha / rtog dral).It is the cognition of objects like forms without the conceptions of name and class.Perception is a faculty, one that is acquired through the faculty of sense' (Dignāga 1987, 7). 5 A recent Indian commentator on the Nyyapravea, Sempa Dorjee, explains that perception in Dignga's definition has the sense of being cognition that is unconditioned by linguistic universals (sgra spyi) (thoughts and concepts associated with a name (nma / ming), such as 'Devadatta') and the object universal (don spyi) (thoughts and concepts associated with a class (jti / rigs), such as 'Cowness').Along this line, in Pramṇavinicaya (PV), chapter 1, Dharmakrti explains, conception is a linguistic cognition.Conception is a cognition to which the [object] appears fused with language.Perception is free from the conception because conception is impossible in the sensory cognition since it arises due to the force of objects.Because perception arises through the force of an object, it exclusively accords with the reality [of its object].(Dharmakrti 2003c, 171)  6 Similarly, 'Jam dbyang Bzhad pa'i Rdo rje (2006, 410) explains conception is a cognition that conceives (zhen rig) objects and language interwovenly: it conceives a pot for instance as though the abstract property of potness pervades all homogenous forms (namely individual pots having similar characteristics of being a pot), and it conceives many heterogeneous forms clustered together as universals (tshogs spyi) such as [forest] for [clusters of individual] trees. 7finition of pratyaka offered in (PS 1.3cd in Dignga 2003, 1) and does not in any way represent a variance from the original position.So I will assume that it is authoritative to cite for the purpose of clarifying the definition of pratyaka at issue.
In addition, Dharmakrti in his Pramṇavinicaya (PV) mentions several other aspects of perception.The PV 1 says perception and inference are the two forms of 'correct cognitions' (yang dag pa'i shes pa) and defines correct cognition as 'nondeceptive with regards to its epistemic activities (don bya ba la slu ba med) because it engages with its objects by means of discrimination (yongs su bcad nas 'jug)' (Dharmakīrti 2003c, 167). 8Elsewhere the PV 1 says perception is a reliable cognition (pramāṇa) because it is non-deceptive (mi slu ba) with regards to the object.It is non-deceptive since it acquires its identity from its own [sensory faculty], whereas the sense of non-deceptivity is surely unreasonable when identity is acquired from another or else not acquired.(Dharmakīrti 2003c, 171) 9 Finally the PV 1 adds: 'perception is non-erroneous or incontrovertible (abhrnta / mi khrul ba); it does not produce errors such as errors of visual blurriness, speedy motion, being in a [moving] boat, etc. Perception is cognition without the conception' (Dharmakīrti 2003c, 171). 10f we examine the features provided in these texts, we get the Pramāṇika's definition of perception along these lines.
• Reliable cognition (pramṇa) is valid knowledge (pratyyana / rab tu rtogs pa) since it cognises the nature (svalakaṇa / svabhva) of objects.• Pratyaka is perception or a sensory cognition since it is acquired through the faculty of the senses.• Perception is one of the two reliable cognitions [the other being inference (anumna)], because there are only two kinds of objects (premayas)-unique reality or unique particulars (svalakṣaṇa) and common reality or universals (samaṇyalakṣaṇa)-to know.• Perception cognises its objects without weaving together name (nma / ming), class (jti / rigs) concepts, etc; hence perception is nonconceptual, (kalpanāpoha / rtog dral) because conception weaves together objects (artha / don) with their name (nma / ming), class (jti / rigs), etc. • Perceptual cognition is non-deceptive with regards to its epistemic activities (don bya ba la slu ba med); it engages with its objects by means of discrimination (yongs su bcad nas 'jug).• Perception is reliable cognition (pramāṇa) because it is non-deceptive (mi slu ba) with regards to the object, and it acquires its identity from its own object.
• Perception is non-erroneous and incontrovertible (abhrnta / mi khrul ba); it does not produce errors such as the errors of visual blurriness, speedy motion, being in a [moving] boat, etc.
Candrakīrti rejects all these features of the Pramāṇika's theory of perception.His critiques are found dispersed throughout most of his writings, and they are quite well known.I therefore do not intend to dwell on them in any detail here except to stress that Candrakīrti's critiques are intended to undermine both the Pramāṇika and Nyāya theories of perception; nevertheless I venture to mention the two primary sources of his critiques.
The first one is Catuatakak (CŚT), chapter 13 in particular.Here Candrakrti rejects the Pramāṇika's claim that perception is reliable cognition with regard to svalakaṇa / svabhva.Tom Tillemans ' (1990) Materials For The Study of ryadeva, Dharmapla And Candrakrti: The Catuataka Of ryadeva, Chapters XII And XIII, with The Commentaries of Dharmapla And Candrakrti provides us very useful notes on Candrakīrti's critique, along with translations of relevant chapters from Sanskrit, Tibetan and Chinese texts.The CŚT also raises objections against the theory of perception in the Skhya and Nyya-Vaieika, which Candrakīrti draws from Nāgārjuna's systematic critique of Nyāya epistemology in the Vigrahavyāvartanī (2005).Nāgārjuna's critique is closely studied and examined in Mark Siderits' (1980) excellent paper, 'The Madhyamaka Critique of Epistemology I'.
The second source of Candrakīrti's critique of foundationalist epistemology is chapter 1 of the Prasannapad (PP).Here Candrakīrti's target is the Pramāṇika.Dan Arnold's  (2005) paper 'Materials for a Madhyamika Critique of Foundationalism' provides us a good, up-to-date annotated translation of this section.Mark Siderits' (1981) 'The Madhyamaka Critique of Epistemology II' provides us a good survey of the objections Candrakīrti employs to undermine the Pramāṇika's account of epistemology.
In his 'Madhyamaka Critique of Epistemology', Siderits goes one step farther than what is warranted.He claims that Candrakrti having a list of four epistemic instruments is an indication that the Prsagika endorses the Nyya theory of knowledge.In Siderits' words, 'It is clear that he takes their account of the four pramṇas as a model description of our epistemic practices ' (1981, 157).In Moonshadows (2010, 167-81) Siderits' claims that Candrakīrti endorses Nyāya epistemology rather than Dignāga's because Nyāya epistemology is less open to the reductionist project than is Yogācāra-Sautrāntika epistemology.I agree with Siderits that there exist some superficial similarities between Candrakīrti's epistemology and Naiyāyika's in that the number of epistemic instruments accepted is the same.Unlike Dignāga-Dharmakīrti tradition, both Candrakīrti and Naiyāyika propose the intermingling role of perceptual and inferential cognitions in that they can share a common epistemic object and that perceptual judgement must entail a determinate cognition rather than a purely indeterminate one as suggested by Dignāga.Do these superficial similarlies justify the claim that Candrakrti endorses the Nyya theory of epistemology?In my view, the evidence is less than convincing.Of course Candrakīrti and Naiyāyika epistemology have some shared features, at least on the surface.This is not suprising as these two traditions flourished side by side in India.The real question though is how far these similarities can take us.In my understanding not very far indeed!Take the case of perception for instance.For the Naiyāyikas, the definition of perception involves the senses (indriyas), their objects (artha), the contact of the senses with their objects (sannikara), consciousness produced by this contact (jnam), the contact of the self and mind (manas), and the contact of mind/intellect (manas) and the senses.
On the Naiyāyika's account of perception, all other conditions may be satisfied, but if the self is not present, perception would be impossible.The self controls the senses, synthesises their function, and confers a unity amongst the epistemic instruments.Consciousness is an integral property of the self.The self is the substratum of consciousness that does not need to be conscious always.Consciousness cannot exist apart from the self, even as light cannot exist apart from a flame.Consciousness is a quality of the self, produced in the waking state by the conjunction of the self with the mind (manas).Mind (manas) is seen as only the instrument by which the self thinks.Manas is atomic in size whereas the self is infinite.Mind, according to the Naiyyikas, does not have the ability to perceive; it is not a substance or the perceiver but rather a quality of the self that alone has the ability to perceive.Self is the perceiver of all things (sarvasya draṭ), the experiencer of all (bhokt), and the knower of all things (sarvnubhv) (Radhakrishnan 1998, 147-48).The Nyya-stra of Gautama (2.21) therefore rules out the possibility of perception without tman: 'Perception cannot arise unless there is conjunction of tma with mind' (Agrawal 2001, 16).
If Siderits' claim is right, we have to attribute to Candrakrti the Nyya's metaphysics of self, without which there would be no Nyāya epistemology, since the self is the backbone unifying all other epistemic instruments and the one that affords them cognitive life.In my view, attributing such a view to Candrakīrti is to totally defeat the whole purpose behind both Candrakīrti's and Nāgārjuna's painstaking critique of the foundationalist epistemologies of the Nyāya and the Pramāṇika.

Prāsaṅgika's theory of perception
We therefore need to look for an alternative solution to the problem, and this is precisely the aim of my paper.The paper explores Candrakīrti's own account of perception, which he defends in the CŚT and PP, although, shall we say, somewhat unsystematically.There exists, however, Candrakīrti's little known work, Pacaskandhaprakaraṇa (PSP Dbu ma ya 239b-266b), which provides us a more systematic outline supporting his theory of knowledge and perception.This text provides us a brief but useful discussion concerning Candrakīrti's treatment of the nature of sensory perceptions.Although his discussion is specifically on the epistemic issues surrounding the visual perceptual process-including the faculty of vision, visual consciousness, and form-it is clear from the text that the underlying principles can be applied to the epistemic practice of sensory perceptions.
I claim that Candrakīrti's theory of perception is a radical departure from the Pramāṇika's account.It is almost a complete reverse of the latter.As far as Candrakīrti is concerned: (1) Reliable cognition only makes sense with regards to the perceptibles (pratyaka) that are non-intrinsic (niḥsvabhāva) and dependently arisen, because everything is causally dependent.This rules out the possibility of the so-called reliable cognition (pramṇa) of intrinsic reality (svabhva) or unique particulars (svalakaṇa) of the Pramāṇika.(2) Perception can be nondeceptive about the perceptibles that constitute conventional reality and therefore can be defined as reliable cognition within the mundane context.But perception can never be regarded as non-deceptive (hence reliable cognition) with respect to the unique particulars (svalakaṇa), since all perceptibles lack the so-called unique particularity.
(3) The majority of perceptual cognitions, excluding the cognitive processes of noble beings in their meditative equipoise, are conceptual cognitive processes, and yet they may still be reliable cognitions since such cognitions could still satisfy the mundane epistemic standard.(4) Perception is defined as non-deceptive within the context of mundane epistemic practice.But it is not defined as invariably non-erroneous or incontrovertible (abhrnta / mi khrul ba) as the Pramāṇika does.If the Pramāṇika's definition were true, then perception would be intrinsically nonerroneous irrespective of any other considerations regarding the epistemic instruments.
It is beyond the scope of this paper to defend all these claims systematically.My attempt here in the remaining sections of the paper is to produce the Prāsaṅgika's defence for some aspects of these claims.I intend do this by drawing on Candrakīrti's arguments concerning the four central theses in his theory of perception: (1) Perceptual reliability (pramāṇa) is not an intrinsic (svabhāva) or unique (svalakṣaṇa) characteristic of the perceptions, (2) Perceptual reliability is determined by the perceptibles, (3) Perception is limited in what it comprehends, (4) Perceptual reliability can be associated with conceptuality, or that a conceptual cognition can be epistemically reliable.
The arguments presented to defend these theses, in my view, constitute the heart of Candrakīrti's distinctive theory of perception.So the primary task I propose to undertake in this paper is to analyse these four theses in turn and analyse the ways in which Candrakīrti's works provide us both the textual support and the philosophical arguments to set up the frameworks for his non-foundationalist theory of perception.

Perception redefined
We begin with the Prāsaṅgika's definition of perception.Candrakīrti defines reliable cognition (pramāṇa) in the Catuḥśatakatıkā (CŚT) as simply 'non-deceptive consciousness in the world' (Dbu ma ya 197b). 11The definition employs two key terms: (1) non-deceptive consciousness and (2) world.The former sets the standard criterion of reliable cognition.By 'reliable cognition' (pramāṇa), Candrakīrti means consciousness that is non-deceptive in the epistemic sense within a defined set of contexts.The use of the latter term, 'world', therefore provides the context predicating that the non-deceptive character of cognition is to be understood strictly within the epistemic bounds of the world or in the confines of the worldly convention.The term 'world' for Candrakīrti, as I have argued eleswhere (Thakchoe 2011, 41-3), is taken for granted by means of naïve common sense agreement.
Candrakīrti's use of the phrase 'world' (lokiya) in his definition of perception serves two important mutually entailing purposes: (1) it shows that the term nondeceptive in Candrakīrti's definition is an exclusive epistemic characterisation of the cognition in the uncritical mundane context; and therefore (2) it excludes the sense of non-deceptiveness discussed in the Madhyamaka's ontology, which asserts the thesis that everything-cognitions and objects-are deceptive and empty of any intrinsic reality.In the CŚT XIII.301,Candrakīrti therefore writes as follows: (1) The world regards non-deceptive (mi bslu ba) consciousness as being reliable cognition (pramāṇa).(2) Then the Transcendental Victor said that consciousness too, since it is a conditioned phenomenon ('dus byas), is unreal, deceptive and illusory.That which is unreal, deceptive and illusory cannot be non-deceptive because while such an entity exists in one way it appears in another.Thus it makes no sense to say that such a phenomenon is a correct cognition; otherwise it would follow absurdly that all consciousnesses are correct cognitions (Dbu ma ya 197b;Candrakīrti 1996, 334). 12 From the Pramāṇika's foundationalist perspective, this passage makes two contradictory claims: (1) non-deceptive (mi bslu ba) consciousness is reliable cognition (pramāṇa), but (2) there is no such thing as non-deceptive consciousness (and therefore there is no reliable cognition) since all consciousnesses are unreal, deceptive, and illusory.Candrakīrti argues that this contradiction is unavoidable in the Pramāṇika's theory of perception since it characterises non-deceptivity as the intrinsic nature of perception.
If this passage is viewed in the light of the context I have proposed, the two claims are not contradictory.When one reads the passage with the non-foundationalist lens of Candrakīrti, the two points he makes are rather complementary-the former is dependent on the latter.That is to say, for Candrakīrti, non-deceptive (mi bslu ba) consciousness is epistemically effective and reliable cognition (pramāṇa) by the mundane standard because it is ontologically deceptive since it is empty of any intrinsic reality.To put it differently, even though all consciousnesses are, according to Candrakīrti's ontology, unreal, deceptive and illusory-like, they are nevertheless efficient as reliable epistemic warrants.
So, what Candrakīrti proposes in the passage (the former claim) is the possibility of a mundane epistemic warrant without the need for any intrinsically real consciousness, and therefore what he denies (the latter claim) is precisely the possibility of epistemic warrant of intrinsically real consciousness.To put the point differently, Candrakīrti wants to propose the efficacy of the perceptual theory based on the non-foundational and dependently arisen ontology of consciousness, while rejecting the efficacy of foundational and intrinsically real consciousness.
Candrakīrti's contention is this: since there are only six types of sensory consciousnesses and only six kinds of sensory perceptions, they are only six faculties of sense and six corresponding sensory objects.This must follow because the production of each perception depends on the production of the respective consciousness, and the cessation of each perception is dependent on the cessation of the respective consciousness.Likewise the production of each consciousness is dependent on the existence of its respective sensory faculty and its objects; and the cessation of each sensory consciousness is dependent on the cessation of the respective sensory faculty and its object.'Therefore, in this context', Candrakīrti explains, 'It is well-known amongst all our systems that every consciousness is produced dependently, and they cease after every moment [i.e., they do not endure for two moments]' (Dbu ma ya 180a; Candrakrti 1996, 299-300). 15n the Pacaskandhaprakaraṇa (PSP), Candrakīrti further bolsters his dependency argument by demonstrating the need of having two supporting conditions for the consciousness.When the opponents asks, 'What is the faculty of visual sense?' Candrakīrti replies: '[The Prakaraṇa] states that it is a translucent 16 form that supports (āśraya / rten) visual consciousness.The support for visual consciousness is twofold: the faculty of visual sense, which arises simultaneously [with visual consciousness itself], and the antecedent mental [consciousness]...In just the same manner, one needs to understand the faculty of auditory sense, etc.' (PSP Dbu ma ya 240b-241a). 17he point made here is that sensory consciousness arises from two cognitive supports because (1) it co-arises and co-exists with a sensory faculty, and (2) it arises from the cessation of the antecedent mental consciousness.In the case of visual consciousness, for instance, the two cognitive supports are the faculty of visual sense, which co-arises and co-exists with visual consciousness itself, and the cessation of the antecedent mental consciousness.
Therefore in the PSP Candrakīrti defines the epistemic functions of each consciousness as follows: 'Visual consciousness is a specific cognition of forms dependent on the faculty of visual sense.… Auditory consciousness is a specific cognition of sounds dependent on the faculty of auditory sense.… Olafactory consciousness is a specific cognition of smell dependent on the faculty of olafactory sense.… Tactual consciousness is a specific cognition of tacticle objects dependent on the faculty of tactual sense.… Mental consciousness is a specific cognition of phenomena [i.e., mental contents] dependent on the faculty of mental sense' (PSP Dbu ma ya 266a). 18sual sense 'exists in the eyeball and resembles the size of a grape seed or a zera flower.Rather it exists in between or is covered beneath a transparent membrane.This is so', says Candrakīrti, 'since it is [inferentially] cognised from its effect-i.e., visual consciousness' (PSP Dbu ma ya 240b: Mig gi dbang po 'di ni mig gi rten gyi mig gi 'bras bu'i nang na rgun 'bru'i 'bras bu'i tshad tsam zi ra'i me tog lta bu ltar med pa pags pa dang bas gyogs pa dang bar gnas pa rang gi 'bras bu rnam par shes pa las rtogs pa ste … ).That is, the existence of the transparent faculty of visual sense is, at least for an ordinary being, not a domain of direct perception, but rather a domain of an inference since it needs to be inferred from its effect, namely the visual consciousness.'For this reason, although blind people with eyeballs appear to possess the cognitive support of the faculty of visual sense, we can understand that the sensory faculty is absent in them from the fact that visual consciousness does not arise in them' (PSP Dbu ma ya 240b: De yang 'di ltar ce re long rnams kyi mig gi rten 'dra bar gnas kyang rnam par shes pa mi 'byung bas dbang po med par rab tu shes so //).Bhāvavevika says that he adopts the Ābhidharmic theory of the sensory faculty being a transparent material form on the conventional level but argues that the sensory faculty is not ultimately a transparent material form for the reason that it is causally produced.See Tarkajvālā III.43, Dbu ma dza 67a: chos mngon pa las mig gi dbang po ni go snyod kyi me tog lta bu 'am zhog chu lta bu yul kha dog gzugs dang ba'o zhes de'i mtshan nyid bstan pa de ni kho bo cag kyang kun rdzob tu 'dod do // don dam par ni mig gi dbang po yul gyi kha dog gzugs dang ba ma yin te / byas pa nyid kyi phyir ro zhes bya ba la sogs pa'i gtan tsigs dag gis dpe yang pags pa dag la sogs pa bzhin no //. 17Mig gi dbang po gang zhe na / mig gi rnam par shes pa'i rten gzugs [241a] // dang po'i zhes 'byung ngo // mig gi rnam par shes pa 'di'i rten ni rnam pa gnyis te / lhan cig skyes pa'i mig dang 'das pa'i yid do // de la mig la mig gi rnam par shes pa'i rten du brjod na yid kyis kyang mi gces bya ba 'thob par 'gyur bas / de bas na gzugs dang ba zhes smos so // de bzhin du rna ba la sogs pa la yang shes par bya'o//.
18 Mig gi rnam par shes pa gang zhe na / mig gi dbang po la brten nas gzugs so sor rnam par rig pa'o // rna ba'i rnam par shes pa gang zhe na / rna ba'i dbang po la brten nas sgra so sor rnam par Based on Candrakīrti's definition, 'Jam dbyang Bzhad pa'i Rdo rje reconstructs the definition of reliable sensory perception, as 'a cognition that directly by the medium of the faculty of physical sense as its dominant condition (adhipati pratyaya) is nondeceptive with regard to the perceptible (pratyakṣa) as its apprehended object' ('Jam dbyang Bzhad pa'i Rdo rje 2006, 423). 19Dge shes Blo bzang Rgya mtsho concurs and defines it as '[a]n awareness that by the means of the faculty of physical sense as its unique dominant condition is directly and experientially aware of its object' (Dge shes Blo bzang Rgya mtsho 2002, 190). 20Applying this generic definition of reliable sensory perception, a reliable visual sensory perception, for instance, could be defined as a cognition that directly by means of the faculty of visual sense as its dominant condition is non-deceptive with regards to its apprehended object: forms, colour, etc. Likewise the same definition is applied to define other reliable perceptions.
On close observation, the Prāsaṅgika's definition of sensory perception stresses the sensory faculties as the specific or dominant condition (adhipati pratyaya) unique to the operational process of each individual sensory perception.The dominant or specific condition (adhipati pratyaya) that provides, according to the Prāsaṅgika, necessary epistemic differentiations or individuations between the three classes of perceptions (sensory, mental and yogic) or between individual perceptions within the same class.For instance the contrasts between visual sensory perception vis-àvis auditory sensory perception (although both instances of sensory perceptions) can be explained through the differences between the faculty of visual sense and faculty of auditory sense as the dominant condition.The one that has visual sense as its dominant condition is the visual sense perception, while the one with auditory sense as its dominant condition is the auditory sense perception.
The Ābhidharmikas and the Logicians claim that the eyes and other sensory faculties are intrinsically existent because we observe them from their effects-the sensory consciousnesses.In the CŚT XIII.312,Āryadeva rejects the possibility of proving the intrinsic existence of sensory faculties through consciousness on three counts: [1] 'Because the conditions (pratyaya / rkyen) would be incomplete, the consciousness could not exist before the sight.[2] But after [sight], the consciousness would be pointless.In the third case [viz.simultaneously], the instrument (karaṇa / rig pa'o // sna'i rnam par shes pa gang zhe na / sna'i dbang po la brten nas dri so sor rnam par rig pa'o // lce'i rnam par shes pa gang zhe na / lce'i dbang po la brten nas ro so sor rnam par rig pa'o // lus kyi rnam par shes pa gang zhe na / lus kyi dbang po la brten nas reg bya so sor rnam par rig pa'o // yid kyi rnam par shes pa gang zhe na / yid kyi dbang po la brten nas chos so sor rnam par rig pa'o //.
byed pa) would be pointless' (Dbu ma ya 202a; 1996, 343). 21Elaborating on Āryadeva's critique of the foundationalist position, Candrakīrti in the CŚT XIII.312 points out that the existence of consciousness itself remains unproven for the foundationalist on three grounds.
First, visual consciousness does not exist before the existence of the faculty of visual sense, since eyesight, perception, would lack its dominant condition (adhipatipratyaya / bdag po'i rkyen) even though the other three conditions-the causal condition (hetupratyaya), the objective condition (ālambanapratyaya), and the immediately preceding condition (samanantarapratyaya)-are present (Dbu ma ya 202a; Candrakrti 1996, 343). 22econd, if visual consciousness existed after the cessation of the faculty of visual sense, then the consciousness would be pointless.If the eye could see the visual form without there being visual consciousness, then there would be no point in assuming the reality of visual consciousness (Dbu ma ya 202a; Candrakrti 1996, 343). 23hird, if the visual faculty and visual consciousness are conceived as occurring simultaneously, it would be pointless to posit the faculty of visual sense as one of the instruments (karaṇa / byed pa) or conditions for the production of sight or visual consciousness.If they both existed simultaneously, they would be independent from each other.The visual consciousness that exists at the same time as the faculty of visual sense and is dependent upon the visual faculty could then not arise, just as the right and the left horns of an ox exist concurrently and independently from each other, and therefore they could not come into being if they were dependent on each other (Dbu ma ya 202ab; Candrakrti 1996, 343-45).24

Pratyakṣa as perceptibles
Candrakīrti's second thesis is the claim that perceptual reliability is determined by the perceptibles.This is in direct constrast to the Pramāṇika's claim that epistemic authority is an intrinsic (svabhāva) or unique characteristic (svalakṣaṇa) of perception because it is intrinsically non-deceptive and nonconceptual.For Candrakīrti to claim that perception is non-deceptive is to claim however that the cognition in question is epistemically reliable with respect to its principle perceptible object.In other words, the nondeceptivity of the cognition is not an intrinsic nature of the cognition itself.
To this effect, Candrakīrti proposes a different semantic valuation for the term pratyakṣa.The Sanskrit term pratyakṣa (mngon sum) and the meanings associated with it are at the centre of the argument.The term pratyakṣa, commonly rendered into English as 'perception', justaposed against 'perceptible', is not so straighforward as the term perception in English suggests.By definition, perception in English always refers to a type of cognition-the ability to see, hear, or become aware of things through the senses as derived from Latin perceptio(n), from the verb percipere, meaning 'sieze', 'understand'.It does not have the sense of perceptible objects, explicitly or otherwise.
The term pratyakṣa however is ambiguous. 25It has these senses: (1) 'perceptible' as an object and (2) 'perception' as a subject or cognition, and as Candrakīrti comments on it in the PP I.3, 'whether it is the subject of a characterisation (lakṣya) or a unique particular (svalakṣaṇa) or a universal (sāmānyalakṣaṇa)-if it exists in the world', according to Candrakīrti, 'it must be evident (aparokṣa) because it must always (sarvam eva) be an object of a direct perception.For that reason, along with its subject, which is the cognition apprehending it, a perceptible is also posited as pratyakṣa' (Candrakīrti 1960, 25). 26t the crux of this ambiguity is Candrakīrti's adjectival sense of pratyakṣa, 'perceptible', which is so central to his account of perception.It is this perceptibility of visible objects that drives home, Candrakīrti argues, the usage of the term that also denotes 'the cognition that has a perceptible object' (tadviṣayaṇa jānena saha).This being the case, the definition of reliable perceptual cognition for Candrakīrti is not one that privileges perception with an intrinsic epistemic authority.This is precisely the case in the Pramāṇika's account of perception since it defines it as intrinsically non-deceptive (pramāṇa) on the grounds of it being devoid of any conceptuality (kalpanāpoḍha). 25As we shall shortly see, this ambiguity is critical in the debate between the Buddhist foundationalist vis-à-vis the antifounationalist Prāsaṅgika. 26Tasmāl loke yadi lakṣyam / yadi vā svalakṣaṇaṃ sāmānyalakṣaṇaṃ vā / sarvam eva sākṣad upalabhyamānatvād aparokṣaṃ ata pratyakṣaṃ vyavasthāpyate tadviṣayeṇa saha //.(Dbu ma 'a 25b: De'i phyir gal te mtshan gzhi 'am rang gi mtshan nyid dam spyi'i mtshan nyid kyang rung ste / 'jig rten na yod na ni thams cad mngon sum du dmigs par bya ba yin pa'i phyir na lkog tu ma gyur pa yin te / de'i phyir de'i yul can gyi rnam par shes pa dang lhan cig tu mngon sum nyid du rnam par gzhag go //.)For Candrakīrti, perceptibles, both directly visible objects and conceptual abstractions or universals (sāmānyalakṣaṇas), determine the defining criterion of reliable perceptual cognition (cf.Arnold 2005, 461).Candrakīrti makes this point explicit in the PP I.3, stating Perceptible is an object that draws towards it the faculties of sense; hence the word pratyakṣa expresses the meaning evident (aparokṣa).From the statement 'an object that draws towards it the faculties of sense', visible objects like jars and colours and so forth are affirmed as perceptibles.A cognition that ascertains these [jars, colours, etc.] is designated as being a perception because the perceptible [objects] like straw or chaff-fire causes it.(Candrakīrti 1960, 24)  27 Therefore perception is only a provisional pratyakṣa, whereas perceptible is a pratyakṣa proper. 28It is perceptible objects that cause perceptual cognition to arise, since it arises when the perceptible objects draw towards them perceptual cognition even as the presence of chaff-fire causes the perceptual cognition that ascertains it to arise.Moreover, in PP I.3 we read: The word pratyakṣa [as being perceptible] is indeed well-known in the world.Whatever it is in the world is precisely what we explain [not by you Pramāṇika].But if your account [of pratyakṣa] undermines the ordinary categories as they are established, then it would undermine the very expression 'well-known' (prasiddhaabda).Therefore [your account of pratyakṣa] would not be what is [commonly] called pratyakṣa'.(Candrakrti 1960, 25) 29 Pratyakṣa as the perceptible is a well-known fact of ordinary discourse, and for a thing to be 'perceptible' is for that thing to be directly accessible to our ordinary senses as opposed to being a thing that is directly inaccessible.The Pramāṇika's theory of perception, which does not recognise pratyakṣa as perceptibles, therefore contradicts the mundane convention. 27Aparoṣa artha vācita vāta pratyakṣa abdasya sāṣāda bhimukho artha pratyakaṣa / pratigatam ṣamsminniti kṛatvā ghata nīlādīnām paroṣāṇāṁ pratyakṣatvaṁ sidhdaṁ bhavati / tatparicchedakasya jñānasya tṛuṇa tuṣānīgavat pratyakṣa kāraṇatvāt pratyakṣatvaṁ vyapadiayati /; (Dbu ma 'a 24b; Candrakrti 2003, 53): Mngon sum gyi sgra ni lkog tu ma gyur pa'i don gyi rjod par byed pa yin pa'i phyir / dbang po mngon du phyogs pa'i don ni mngon sum yin no // 'di la dbang po mngon du phyogs pas zhes bya ba nas bum pa dang sngon po la sogs pa lkog tu ma gyur pa rnams mngon sum nyid du grub par 'gyur la / de yongs su gcod par byed pa'i shes pa ni rtzva dang sog ma'i me bzhin du mngon sum gyi rgyu can yin pa'i phyir mngon sum nyid du rjod par byed do //.
28 This claim of Candrakīrti is radically different from the position held by all other Buddhist schools, which unanimously admit perception as the proper pratyakṣa and perceptible being the pseudo.
Candrakīrti applies the same principle even in his theory of perceptual error, according to which even the intentional objects of so-called perceptual errors are regarded as perceptibles and therefore pratyakṣa proper.This is for two reasons: (1) perceptual errors arise because of the representations of illusory objects such as double moons, and therefore (2) the intentional objects with which the erroneous perceptual cognitions are engaged do have the epistemic quality of being perceptible from the perspective of such cognitions.In the PP I.3, he writes: Double moons, etc., although they do not have the quality of being perceptible from the point view of the cognition without cataracts [i.e.normal cognitions], these objects indeed have the quality of being perceptible from the point of view of the cognition with cataracts, etc. (Candrakrti 1960, 25) Here Candrakīrti is not defending the epistemic authority of the defective perception per se without qualifying his statement.He is defending the claim that even defective perceptions are reliable in so far as they correctly perceive the intentional objects that appear to them, even though those objects really do not exist.This might sound implausible since it seems to contradict the mundane epistemic convention.Candrakīrti however insists that is not the case.'In this tradition, although the appearance of the double moons, the appearance of the falling hairs, and so forth are accepted as the perceptibles, this does not however contradict the mundane convention.This is because', as 'Jamyang bzhad pa explains it, 'there is no difference between [the mundane convention and the Prāsaṅgika's position] in terms of accepting what does and does not constitute the perceptibility (pratyakṣa), depending on whether or not the worldly sensory cognitions are with or without cataracts' ('Jam dbyang Bzhad pa'i Rdo rje 2006, 422). 31his follows, says 'Jamyang bzhad pa, on two grounds.(1) From the point of view of the cognition without the cataracts (i.e., mundane convention), the double moons, the falling hairs, etc., are not perceptibles, since these objects are neither directly perceived nor ascertained by normal cognitions.The cognitions apprehending the double moon, etc. are thus unreliable cognitions (apramāṇa) from the vantage point of the cognition without cataracts.(2) From the point of view of perceptual cognition with cataracts, however, the double moon, etc. are perceptibles.For there are such cognitions with cataracts, etc. that directly perceive and ascertain such objects.The 30 Dvi candrādīnaṁ tu ataimirika jānāpeṣya apratyakṣatvaṁ, taimirikādhyapeṣaya tu pratyakṣatvaṁ eva //.Dbu ma 'a 25b: Zla ba gnyis la sogs pa dag ni rab rib can ma yin pa'i shes pa la ltos nas mngon sum nyid ma yin la / rab rib can la sogs pa la ltos nas ni mngon sum nyid kho na'o //.
31 Des na lugs 'dir zla gnyis kyi snang ba dang nam mkha'i skra shad 'dzag snang sogs kyang 'di bas mngon sum par 'dod kyang 'jig rten pa dang mi 'gal te / 'jig rten pa'i dbang shes rab rib sogs med pa dang yod pa las mngon sum min pa dang yin par 'dod pa la khyad par med pa'i phyir //.cognition apprehending the double moon, etc. is thus a reliable cognition (pramāṇa) with respect to the perceptibles such as the double moons (ibid.). 32herefore for Candrakīrti, says 'Jam dbyang Bzhad pa, 'any dualistic cognition must necessarily be a reliable perceptual cognition with reference to the representational object that appears to it (snang yul)' (ibid., 421). 33That is to say that from the point of view of the mundane convention, and also from the point of view of the co-dependence of the subject and the object, perceptible is the actual pratyakṣa, whereas perception is the provisional pratyakṣa (ibid.). 34t this point the opponent of the Prāsaṅgika could raise two possible objections.First, they could object that a defective cognition could not be a reliable perceptual cognition on the grounds that its epistemic validity is undermined and contradicted by the worldly convention in which non-deceptive perceptions alone meet the standard of epistemical warrant.
The Prāsaṅgika replies to this by saying that there is no pervasion or necessary entailment in the opponent's argument.That is, even though the mundane convention undermines the epistemic status of the defective sensory cognition, it does not necessary follow that all defective epistemic states have no valid explanation of their own.Although the objects represented in such cognitions may not be real by the conventional standard, (and therefore they are regarded as utterly illusory, purely fictional), they nevertheless give rise to a conventionally real perceptual cognition.Fictional objects such as the double moons and falling hairs are unreal objects, but the fact remains that these fictional objects serve as the intentional objects of socalled defective perceptual cognitions, and these cognitions do conventionally exist, although the objects represented in these cognitions may not even be conventionally real.The intentional objects, i.e. the representational objects that appear to these erroneous cognitions, according to the Prāsaṅgika, give rise to the existence of these fallacious perceptions; hence it insists that defective perceptions are reliable perceptual cognitions in relation to their perceptibles.Therefore, for a Prāsaṅgika, a defective sensory cognition is nevertheless a conventionally real epistemic state, although the object it represents is not.
The opponent may raise another objection: Does this mean that the visual perception that apprehends a 'white cone shell as yellow' is also a reliable visual 32 De dang po la ltos nas mngon sum du mi snang bas mngon sum min pa des de la nges shes kyang 'dren mi thub pas tshad ma yang ma yin pa'i phyir dang // gnyis pa la ltos nas mngon sum du snang bas mngon sum yang yin / de yod pa la tshad ma gzhan la ma ltos par nges shes 'dren pas tshad ma yang yin pa'i phyir ro //.
perception since this perception also satisfies Candrakīrit's definition of a reliable sense perception?The Prāsaṅgika, according to Dge shes Rgya mtsho, replies to this objection by stating that this erroneous perception is not a case of a reliable visual perception; rather it is a case of mental perception since it is only with respect to the mental perception that the representation of the cone shell appears yellow (Dge shes Blo bzang Rgya mtsho 2002, 190-91).
There are numerous other examples of such perceptions, which while they are mental perceptions, appear to be sense perceptions.A visual perception apprehending a form or colour, for instance, is a sensory perceptual cognition with respect to the form or colour in question.With respect to the representations of the form or colour that appears to the mind, it is a mental perception.However, a conceptual cognition apprehending a form is conceptual with respect to the form in question, whereas it is a mental perception with respect to the representational appearance of the form.
According to Dge shes Rgyatso, this represents a distinctive presentation of the Prāsaṅgika's perceptual theory.Even though there is one recognised cognitive event depending on various objects, sense faculties, and consciousnesses involved in the process, it is possible separately to account for, or differentiate between, the mental and perceptual cognitions involved in the same cognitive process, without undermining or contradicting each other's operational or epistemic value (Dge shes Blo bzang Rgya mtsho 2002, 191). 35andrakīrti does not reject the efficacy of defective sensory perceptions, even though he knows that these cognitions do not satisfy the standard of mundane convention.This is because he defines them as epistemic warrants, not with respect to the mundane standard, but with respect to their intentional objects.
A similar line of argument is used for the reason not to undermine epistemic authority in mundane practice, even though scrutiny of the analytic cognition could easily undermine such epistemic authority.For no subject and object, according to Candrakīrti, defies reasoned analysis.The perceptibles that are established as real by reliable mundane cognitions are found to be unreal when those perceptibles are subjected to critical rational analysis.Therefore Candrakīrti writes in the PP.
Things like jars, colours of blue, etc. are not accepted as perceptible (pratyakṣa) from the standpoint of one who knows reality (tattvavidapekṣayā).Jars, etc. are accepted as perceptible exclusively according to the worldly convention.(Dbu ma 'a 24a) 36 As Āryadeva's CŚT states: When its colour is seen, the presence of the whole jar is not seen.But what knower of reality would say that a jar is [ultimately] perceptible?One supremely intelligent who is capable of this very same analysis appeases all of these-sweet fragrance, melodious sound, and softness.(Dbu ma ya 195ab, 198a) 37 As we can see, Candrakīrti cites the first two verses in Āryadeva's CŚT to corroborate his argument that when a Prāsaṅgika speaks of a perceptible jar and so forth, he is always speaking of it from a non-analytical conventional standpoint-the standpoint of ordinary mundane cognition, which he argues engages with its object uncritically.Hence only on this level of uncritical discourse does a Prāsaṅgika acknowledge a perceptible as conventionally real.This standpoint is contrasted in the above text with the critical analytical perspective of a 'knower of reality' (tattvavida), who comprehends that nothing is established as a perceptible when subjected to reasoned analysis.That is, from the analytic perspective, nothing is established as a perceptible, for everything is logically reducible, and therefore no perceptual cognition is said to arise ultimately.But this assertion does not undermine the conventional fact of perceptible things.Conventionally things do arise from their causes and conditions, and convention accepts things as they are conventionally without subjecting them to rigorous analytic tests.The analytic perspective therefore does not undermine the possibility of perceptual cognition arising conventionally.In the same way, Candrakīrti makes the point that the epistemic authority of cognition in mundane usage should not be exploited to undermine the epistemic efficacy of defective sensory cognitions.

Perception and epistemic limits
In direct contrast to Candrakīrti's position, which argues that it is the perceptibles that determine the efficacy of the perceptual process, the Pramāṇika argues that it is the intrinsic nature of the faculties of sense to function as the means by which we establish perceptible objects (pratyakṣa).Candrakīrti challenges this assertion and says if this were the case then it would be illogical for the Pramāṇika to attribute different cognitive functions to different faculties of sense in virtue of the different objects.
After all, if the Pramāṇika was right, the presence of the objects, however different they may be, should not limit the ways in which the faculties function if their operations are intrinsic to themselves.As Āryadeva's CŚT XIII.301 says, 'The eye sees only the visual form but not odours and other things, because they are different objects' (Dbu ma ya 196a; Candrakrti 1996, 331). 38Commenting on this in his CŚT XIII.310-11,Candrakīrti writes: If the sense faculties did have the [intrinsic] power to discriminate 39 (pariccheda / yong su gcod pa) [objects], then they would have visual form and so forth as their objects, but they do not have this [intrinsic power to discriminate].Why?Because the five [sense faculties] such as the eyes are (1) all derived from the elements (bhautika / 'byung 'gyur) and ( 2) their functions differ because of the different objects.Hence, the eyes, for instance, see only [visual] form but do not hear sound.The ears perceive sound strictly and do not see form.(Dbu ma ya 201ab;Candrakīrti 1996, 341-42) 40 Candrakīrti's argument does not deny the existence of the faculties of sense and their mundane epistemic functions.The following discussion in the CŚT XIII.311 makes this point clearer.The opponent objects, 'If the eyes and so on thus cannot exist, then how do you establish that faculties of sense such as the eyes, etc. are the effects of actions (karma)?' Candrakīrti replies, 'Did we deny that they are the effects?' Again the opponent objects, 'By having negated the existence of the eyes, etc. have you not denied [that they are karmic effects]'?Candrakīrti replies: It is because our analysis is primarily concerned with searching for the intrinsic natures (svabhāva) of the objects.Here, in this context, we are negating the things that are established by their intrinsic natures (svabhāva / rang gi ngo bo).We do not negate the eyes and so forth, which arise owing to the karmic effects of causal production and dependent co-arising.Thus, the eyes and the like do indeed exist, for we say that they are effects of karmic actions.(Dbu ma ya 201ab; Candrakīrti 1996, 342) 41 38 Mig gis ni gzugs gcig pu mthong gi dri la sogs pa dag ni ma yin te yul tha dad pa'i phyir ro //. 39Here Candrakīrti is refuting his Buddhist foundationalist opponent who claims the reality of the faculties of sense on the grounds that they have an intrinsic nature (svabhāva) to apprehend their objects.The claim goes in the CŚT as follows: 'Perceptible things (pratyakṣa / mngon sum) such as visual form and so forth do really exist, because the faculties of sense such as the eyes, etc. that apprehend them do exist.These faculties of sense that exist must necessarily engage with their own objects; objects like form, etc. with which these [sense faculties] engage are [thus] perceptible'.CŚT, Dbu ma ya 201b: yul gzugs la sogs pa mngon sum dag ni yod pa kho na ste de'i 'dzin par byed pa mig la sogs pa'i dbang po yod pa'i phyir ro // yod par gyur pa'i dbang po 'di rnams ni gdon mi za bar rang gi yul la 'jug par 'gyur dgos la / gang du de rnams 'jug pa srid pa gzugs la sogs pa don de dag ni mngon sum yin no / bshad par bya ste.
41 Gal te de ltar mig la sogs pa rnams mi srid na / de'i phyir ji ltar mig la sogs pa'i dbang po 'di rnams las kyi rnam par smin pa'i ngo bor rnam par gzhag ce na / ci kho bo cag gis 'di rnams kyi rnam par smin pa'i ngo bo nyid bkag gam / gal te mig la sogs pa rnams 'gog par sgrub pas de ji ltar ma bkag Here is one of the clearest articulations that Candrakīrti's critique of epistemological foundationalism entials the rejection of the intrinsic natures of the faculties of sense.But it does not ential the rejection of the faculties of sense per se.This passage therefore makes it clear that Candrakīrti does not reject the existence and epistemic efficacy of the faculties of sense.Therefore sensory perceptions, since they are causally conditioned, dependently arise Candrakīrti's critique of foundationalism also denies the epistemic validity of sensory perception with respect to ultimate reality.The Samādhirājasūtra reads: 'The eyes, ears and nose are not reliable cognitions.The tongue, body and mind are also not reliable cognitions.If these sensory faculties were reliable cognitions, of what purpose would the noble path serve to anyone?' (Mdo sde da 20b). 42Commenting on this sūtra in the Madhyamakāvatāra (MA 6.30), Candrakīrti writes: If ordinary cognitions were reliable cognitions (pramāṇas), then the mundane cognitions would see reality as it is.Then what necessity would there be for those other noble beings (āryas)?What purpose would the noble path serve?It makes no sense that fools are reliable cognitions.(Dbu ma 'a 205b; Candrakīrti 1996, 156) 43 The two passages are interpreted very differently by Tibetan Mādhyamikas such as Go rampa Bsod nams Senge (1969, 375, 382), 44 Stag tsang Lots ba (2001, 156-58), 45 Dge 'dun Chos 'phel (1990, 161), 46 and others who rule out the possibility of any account of conventional epistemic authority in the Prāsaṅgika.Careful reading of Candrakīrti's passage and the relevant sūtra literature does not, however, support this line of interpretation.On Candrakīrti's reading, these two texts reject the authority of ordinary uncritical perceptual cognitions as epistemic warrants only with reference ce na, kho bo cag gi rnam par dpyod pa don rang bzhin tsol ba lhur byed pa nyid kyi phyir ro // kho bo cag ni 'dir dngos po rnams rang gi ngo bos grub pa 'gog gi mig la sogs pa byas shing rten cing 'brel par 'byung ba'i las kyi rnam par smin pa nyid ni mi 'gog pa'o // de'i phyir de yod pas gang zhig rnam par smin pa nyid du bsnyad pas mig la sogs pa yod pa nyid do //.
46 Yod med blo yis bzhag pa'i shes bya dang / bden rdzun yul la ltos pa'i tshad ma gnyis / gcig gi rdzun khungs gcig la gtad mthong nas / tha snyad tshad grub 'job la blo ma bde // ma brtag ma dpyad 'jig rten rnam gzhag dang / brtags shing dpyad pa'i grub mtha'i gzhung lugs gnyis / gcig gi rtsa ba gcig la thug mthong na / tha snyad tshad grub 'jog la blo ma bde //. to ultimate truth or ultimate knowledge on the grounds of their inability to apprehend ultimate truth and the absurd consequence that would follow if they did.Candrakīrti makes this point even more explicit in the MA 6.31: Mundane consciousness is not authoritative [with regards to the ultimate] in all respects.Therefore, mundane consciousness does not undermine reality.If mundane objects that exist in virtue of being known by mundane cognition are contradicted by it, they are undermined by mundane cognition.(Dbu ma 'a 205b; Candrakrti 1996) 47 Glossing this verse in his commentary on 6.31, Candrakīrti makes the point even clearer: 'With reference to reality as it is, ordinary cognitions are not authoritative in any sense.Nor is it the case that reality as it is can be undermined by worldly cognitions' (1994, 114-15). 48Candrakīrti's stress is on the relation between mundane cognitions and reality as it is.'Mundane cognitions' are conventional cognitions, 'reality as it is' is ultimate truth, and their relation is exactly what he denies.
Here it is very clear that Candrakīrti rejects the authority of perception with its reference to ultimate truth, but not on account of its knowledge of conventional truth, and it is this knowledge that must be shown to be repudiated if Candrakīrti is to be interpreted as denying the epistemic authority of perception altogether.This must be so since in Candrakīrti's theory of perception the authority of perceptual cognition arises from its ability to ascertain conventional truths as its principal objects, but not by its ability to ascertain ultimate truth or even the svalakṣaṇas / svabhāvas.

Perception as conceptual
Finally, one of the most central aspects of the Pramāṇika's theory of perception is the claim that perception is invariably non-conceptual, a claim that affects its cognitive reliability.Candrakīrti rejects this claim and proposes the view that says perceptions are, by and large, conceptual, and still they can be reliable cognitions.
In the final section of this paper, we turn to Candrakīrti's defence of this crucial claim.First we will briefly consider Candrakīrti's objection against the Pramṇika's theory.Then we will turn to Candrakīrti's arguments.The objection consists of a twopronged approach: the first one provides reductio arguments to show that there is no valid reason to support the Pramṇika's theory.
It is makes no sense [for the Pramāṇika] to conceive perception (pratyaka) as a reliable cognition (pramāṇa), because then you accept that perception (pratyakṣa) is cognition that is devoid of conception.This is not how even the mundane convention works, and you 47 Rnam kun 'jig rten tshad min de yi phyir / De nyid skabs su 'jig rten gnod pa med / 'jig rten don ni 'jig rten grags nyid kyis / gal te sel na 'jig rten gyis gnod 'gyur //6.31//. 48de kho na nyid kyi skabs su 'jig rten rnam pa thams cad du tshad ma ma yin zhing / de kho na nyid kyi skabs su 'jig rten gyi gnod pa yang ma nyin no //.nevertheless aspire to explain reliable cognition (pramṇa) and perceptible (prameya) in the mundane context.(PP 1.3; Candrakīrti 2003, 55)  49 Candrakrti's reductio argument runs like this: because the Pramṇika only accepts pratyakṣa as a cognition that is devoid of conception, it makes no sense to define pratyakṣa as a reliable cognition (pramṇa), because mundane convention has objects, that is, things that are perceptible (mngon gyur) as real pratyaka (mngon sum dgos), whereas the subject, that is, sensory cognition that is conceptual in nature, is only considered pseudo pratyaka (mngon sum btags pa ba).The Pramṇika's claim that pratyaka is an exclusive cognition devoid of conception is therefore incompatible with mundane convention.
'Jam dbyang Bzhad pa's Tshigs gsal stong thun explains Candrakīrti's reductio argument as follows.The claim that only cognition that is devoid of conception is pratyaka, and the claim that pratyaka is pramṇa both contradict mundane epistemic convention.In the mundane convention, when the object is perceived by virtue of a direct experience (mngon sum du myong stobs), it necessarily entails a direct perception (mngon sum du rtogs pas khyab), but surely this does not require the cognition to be devoid of conception.There are, in fact, innumerable cases in which conception apprehends its objects by virtue of direct experience (myong stobs); as we say 'I directly experience joy'.'I directly experience suffering.''I directly apprehend the objects.'These are well-know reliable mundane cognitive facts, all of which are conceptual in nature.Therefore Candrakrti concludes: '[The Pramāṇika's claim] is not even the way in which mundane convention operates' ('Jam dbyang Bzhad pa 'i  Rdo rje 2006, 415). 50andrakrti's second objection of the Pramāṇika's account rejects the textual authority upon which the Pramāṇika bases its exposition.Dignga claims that pratyaka is cognition devoid of conception since it has authoritative basis, and cites PP 1.3: One with visual cognition is aware of blue but does not [conceive] that it is blue.(Dbu ma a 25b) 51 49 Rrtog pa dang dral ba'i shes pa nyid mngon sum nyid du khas blangs pa'i phyir dang / des kyang 'jig rten pa' tha snyad byed pa med pa'i phyir dang / 'jig rten tshad ma dang gzhal bya'i tha snyad bshad par 'dod pa'i phyir na mngon sum tshad mar rtog par ni don med pa nyid du 'gyur ro //.
51 Mig gi rnam par shes pa dang ldan pas sngon po shes kyi sngon po'o snyam du ni ma yin no shes bya (Cakurvijsamagi nla jnti no tu nlam iti) mig gi rnam par shes pa dang ldan pas sngon po shes kyi sngon po'i snyam du ni ma yin no /.Dan Arnold (2005, 459) reports that Dignga cites this passage in 1.4 of his commentary on the Pramṇasamuccaya.

Also the Tatvasaṅgrahapañjika of Kamalala claims:
The definition of pratyakṣa is devoid of error and conception because the Lord Buddha has said: 'One with visual cognition has the awareness of blue but does not [conceive] that it is blue.'By stating 'awareness of blue ' [Lord Buddha] shows the subject is nonerroneous, thus unmistaken.By stating 'but [it] does not [conceive] that it is blue' [the Buddha] says that this cognition is devoid of conception since it does not apprehend objects in association with names.(Tshad ma ze pa 143b) 52 In his response, Candrakrti says that the Pramāṇika's conclusion is not supported by the text: The context of this text (gama / lung) is not where the definition of pratyaka is explained.The context is that the naïvety of the five sensory cognitions is explained.This text indeed does not prove that pratyakṣa is only a cognition that is devoid of conception [which the opponent wanted to show].Therefore [the Pramṇika's claim that pratyaka is devoid of conception] makes no sense.(PP 1.3 in Candrakīrti 2003, 55)  53 Candrakrti rejects the Pramṇika's interpretation of the text on the grounds that this quotation is misappropriated, that it is taken out of context, and that it does not deal with the definition of pratyaka at all.The proper context of the text is that it explains the naïve mode of engagement of the five sensory cognitions, that they lack critical or analytical capacity.
According to Candrakīrti, a reliable cognition can be associated with conceptuality, and therefore a conceptual cognition can be non-deceptive.This is because a nondeceptive cognition is the definition of a reliable cognition, and this definition does not exclude conceptuality from being part of that cognitive process.
Underscoring the critical importance of this criterion, both 'Jam dbyang Bzhad pa and Dge shes Rgya mtsho's definitions of reliable perceptual cognition omit the mention of 'freedom of conceptuality' (kalpanāpoḍha / rtog dral), which forms the essential part of the Pramāṇika's definition of perception.Instead, the former asserts that a reliable perceptual cognition does not directly rely on any valid reasoning to apprehend the perceptible object, and it does not rule out the possibility of indirect conceptual involvement.
53 Mig gi rnam par shes pa dang ldan pas sngon po shes kyi sngon po'i snyam du ni ma yin no zhes bya ba'i lung yang mngon sum gyi mtshan nyid brjod pa'i don can gyi skabs ma yin pa nyid kyi phyir dang / dbang po'i rnam par shes ba lnga po rnams blun po nyid du ston par byed pa nyid yin pa'i phyir / lung las kyang rtog pa dang dral ba'i rnam par shes pa kho na mngon sum nyid ma yin pas 'de ni mi rigs so /.'cognition by means of which, without directly relying on any valid reasoning as its cognitive support, is non-deceptive with regards to the perceptible (pratyakṣa / mngon sum), determinable object it apprehends' ('Jam dbyang Bzhad pa 'i Rdo rje  2006, 421). 54This definition proposes a constitutive correlation between a reliable perceptual cognition and the perceptible-the object of a direct experience.As Geshe Blo bzang Rgya mtsho's definition has it: 'It is an awareness (rig pa) that by the power of direct experience is non-deceptive with regards to its principle object' (Dge shes Blo bzang Rgya mtsho 2002, 189). 55hus, 'Jam dbyang Bzhad pa's definition leaves open the possibility that an indirect conceptual cognition may be involved in a non-deceptive perceptual cognition.Geshe Blo bzang Rgya mtsho asserts that a non-deceptive cognition apprehends its principle object by the power of a direct experience.Again, he does leave open the possibility of the involvement of conceptuality in the process of direct experience.
This open-ended definition of reliable cognition is vital to extend its scope to a wider range of epistemic processes that, while non-deceptive with regard to their perceptible objects, may be associated with conceptuality.The Prāsaṅgika therefore claims that a conceptual cognition may be epistemically valid and non-deceptive.
The argument Candrakīrti employs to defend the conceptual involment in the perceptual process is that perception must necessarily entail the operations of both sensory consciousness and mental consciousness where the latter is conceptual in most part (excluding the exalted cognitive process in the meditative equipoise) and that there exists a time gap between mental cognition on the one hand and the sensory cognition and its object on the other.The mental cognition arises only after the sensory consciousness and its corresponding object have ceased, since only after having ceased does the visual consciousness and the object visually seen earlier externally arise, mental consciousness apprehending the mental representations of the object.Based on these cognitive processes we, according to Candrakīrti, conceptually construct perceptual judgements or notions about things we experience.This demonstrates why most perceptions are conceptual in nature.
Candrakīrti introduces this argument in the CŚT XIII.322.Here the opponent asks: So then there is utterly no judgment of objects whatsoever?In his reply Candrakīrti says: 'No, it is not nonexistent, for things exist without any intrinsic natures.[As Āryadeva puts it:] The object that is seen earlier and that the mind apprehends is like a mirage.This [cognition] is termed the aggregate of perception (saṃjāskandha) 56 for the evaluation of all dharmas' (Dbu ma ya 205b). 57This is because '[w]hen visual consciousness ceases, after having arisen dependent on the eye and form, it ceases along with the sense faculty and the objects.This [visual consciousness] having ceased, the very object that had been seen earlier is [then] apprehended by the mind' (Dbu ma ya 205b; Candrakīrti 1996, 350-51). 58andrakīrti develops his argument further in the CŚT XIII.322.Again the opponent objects: 'But how could you admit that something that is not even present is being apprehended?' 'Like a mirage', says Āryadeva.'Although there is not the slightest amount of water in a mirage, still, through the power of causes and conditions, a perception (saṃjā) that has the representation (ākāra) of water does in fact occur.Similarly', argues Candrakīrti in the CŚT, 'even though it has no intrinsic nature, like a mirage, a conceptual consciousness (vikalpaṃ vijānaṃ) arises with respect to the representation of the object that was [perceptually] apprehended earlier' (Dbu ma ya 205b; Candrakīrti 1996, 350). 59Since conceptual consciousness is the cause of the determination of all dharmas, it is therefore called the aggregate of perception (saṃjāskandha), for in it are the representations (ākāra / snam pa) associated (saṃprayoga) with it or corresponding to perceptual judgement.'It should be understood' says Candrakīrti's CT, 'that the determination of all dharmas are due to the power of the perceptual cognition, but it is not caused by the intrinsic nature (svabhāva) of the things, for intrinsic nature is impossible in all cases whatsoever' (Dbu ma ya 205b; Candrakīrti 1996, 350). 60andrakīrti's argument makes it clear that a perceptual cognition is a conceptual process that interlocks the epistemic practices of both sensory consciousness and saṃjā carries the process of making the judgements based on various factors such as conceptual evaluations, inner dispositions, and representations.Candrakīrti's emphasis in the current context is the representations appearing before the mental cognition that triggers the perceptual judgement to occur.
59 Yang ji ltar nye ba ma yin pa la 'dzin pa srid ce na / smig rgyu ltar zhes bya ba smros te / smig rgyu la chu bag tsam kyang med mod kyi / 'on kyang rgyu dang rkyen gyi dbang gis chu'i rnam pa can gyi 'du shes 'byung ba nyid yin pa de bzhin du smig rgyu la bya ba ltar yod pa ma yin pa'i rang bzhin can gyi dngos po gzung zin pa la yang gang rnam par rtog pa can gyi rnam par shes pa skye ba de ni chos thams cad rnam par 'jog pa'i rgyu yin no //. 60Chos thams [206a] cad rnam par 'jog pa'i rgyu nyid yin pa'i phyir na de nyid la 'du shes phung po zhes bsnyad de / rnam pa de lta bu'i 'du shes kyi khyad par dang mtsungs par ldan pa'i phyir ro // chos thams cad kyi rnam par gzhag pa yang 'du shes kyi dbang gis shes par bya ba'i dngos po rang gi ngo bo'i rgyu can ni ma yin te rang bzhin rnam pa thams cad du mi rung ba'i phyir ro //.mental consciousness.The opponent's objection (cited in the text above) assumes that the object in question must be intrinsically real and temporally durable, because an unreal and impermanent object could not act as the object of mental consciousness since it would have long ceased before it had been apprehended.
Candrakrti illustrates his response by the example of a mirage.In a mirage there is not the slightest bit of water.Still it is reasonable, argues Candrakīrti, to form a cognition based on the representation or the image appearing to be water due to its causes and conditions; hence the conception of 'mirage' exists.Similarly, argues Candrakīrti, the object that was previously visually apprehended is unreal since it lacks intrinsic nature and is momentary since it is no longer present before the mental consciousness.The object has ceased along with the visual perception.Nevertheless the representation of the object is able to appear in the mental cognition as its intentional object.This is because both the object and the visual cognition are unreal and momentary instants and because all necessary conditions are satisfied.It consequently gives rise to a conceptual cognition (vikalpa vijānaṃ), and it is this conceptual consciousness that is responsible for the conceptual determinations or evaluations of all dharmas.
There is another aspect without which the perceptual cogniton cannot operate, namely consciousness.When the opponent asks, 'If that were the case, the aggregate of perception would exist by its intrinsic nature, for if it did not exist, then one could not determine any dharmas', Candrakīrti replies: The perception is itself also associated with the consciousness and thus does not exist without the consciousness.Consciousness, in turn, does not exist by its intrinsic nature, since it is not established without the perception.This is so because, [as Āryadeva states,] '[t]he mind arises dependent on the eye and form, like an illusion (māyā)'.(Dbu ma ya 206a) 61 Indeed, there is, as Candrakīrti's CŚT explains, no consciousness that functions as the basis of the action of arising prior to its arising.When conditions such as the eyes and forms are available, consciousness, since it has no intrinsic nature, can nevertheless arise 62 because the action of arising cannot be set in motion if it is intrinsic.Candrakīrti therefore says that what we can ascertain from this is that 61 Gal te de lta na ni 'o na 'du shes kyi phung po rang bzhin gyis yod pa yin te de med na ni chos thams cad rnam par gzhag par mi nus so zhe na / brjod par bya ste / 'du shes de yang rnam par shes pa dang mtsungs par ldan pa'i phyir rnam par shes pa de ma gtogs par med la / rnam par shes pa de yang 'du shes de ma gtogs par ma grub pa'i phyir rang gi ngo bos med do // 'di las kyang yin te / gang gi phyir / mig dang gzugs la brten nas yid // sgyu ma bzhin du skye bar 'gyur //323ab//. 62Tillemans has translated this passage thus: 'Now, this consciousness does [nonetheless] arise '  (1990, 194).This conclusion seems to follow from the argument advanced, but the text does say otherwise.
consciousness has the quality of an illusion (Dbu ma ya 206a). 63When one examines consciousness in just the way in which it is perceived, then it can be ascertained that consciousness resembles a young girl created by magical illusions in that it too does not have any intrinsic nature.So, in Candrakīrti's view, it is completely accurate to assert that '[l]ike an illusion cognition arises dependent on the eye and form'.If, however, the cognition did have an intrinsic nature, then as Āryadeva's CŚT XIII.322 states: 'What has a real existence, could not be said to be an illusion' is true (Dbu ma  ya 206a; Candrakīrti 1996, 351). 64tating the arguments in brief: Candrakīrti argues for two things: First, he argues that perception is empty of intrinsic nature and that it may be conceptual because it arises dependent on the culmination of a wide range of conditions, including nonconceptual sensory and conceptual mental consciousnesses.Perception is empty of intrinsic nature on the grounds that it depends on mental consciousness, without which the former cannot exist since without consciousness there cannot be so-called perceptual cognition.
Second, he argues that consciousness is also empty of intrinsic nature because it in turn depends on sensory perception and objects, etc. Candrakīrti therefore proposes that consciousness is illusory since it arises just like the emergence of a young illusory girl created by a magician.There is no girl to be found in any of the conditions out of which the magician creates an illusory young girl; nevertheless, an illusion of a young girl does exist due to the force of its conditions.Similarly, says Candrakīrti, there is no consciousness to be found anywhere in the conditions that gave rise to its existence-neither in the sensory perceptions nor in the objects nor in their representations; still, consciousness with its cognising ability does come into existence when all its conditions are present.
Candrakīrti also cites the Sayutta Nikāya in the CŚT (Dbu ma ya 206ab), saying that when an illusionist at the crossroads creates various magical things such as elephants, a person endowed with vision sees the magical tricks being displayed, but upon critical reflection the elephant appears to him as nonexistent (asat), empty (rikta), insignificant (tuccha), and coreless (asāra).In the same way, whatever consciousness one takes, be it a temporal or spatial, outer or inner, faraway or nearby, 63 Gang zhig skye bar bya ba'i rten nyid du 'byung bar 'gyur ba rnam par shes pa de skye ba'i snga rol na ni yod pa ma yin te / mig la sogs pa'i rkyen rnams yod kyang rnam par shes pa'i rang gi ngo bo med par skye ba'i bya ba mi 'jug pa'I phyir skye ba mi rigs na / rnam par shes pa 'di ni skye ba ma yin no // de'i phyir sgyu ma'i chos nyid las ma gtogs par ci zhig nges par bya bar nus … 64 Ji ltar dmigs pa de ltar rnam par dpyad pa la rang gi ngo bos med pa'i phyir rnam par shes pa sgyu ma'i na chung dang 'dra'o zhes bya bar nges par nus so // de'i phyir / mig dang gzugs la brten nas yid // sgyu ma bzhin du skye bar 'gyur // zhes bya ba'i legs par gsungs pa yin no // yang gal te 'di'i rang gi ngo bo zhig du 'gyur na ni de'I tshe rang gi ngo bos / gang la yod pa nyid yod de // sgyu ma zhes byar mi rigs so //.upon closer reflection and analysis also appears to be nonexistent, insignificant, coreless, impermanent, void (śūnya), and selfless (anātman).

Conclusion
Having developed the arguments and defended the Prāsaṅgika's theory of perception without the need of any foundationalism, Candrakīrti closes chapter XIII of CŚT with this dramatic conclusion.
[Opponent]: It is astonishing that on the one hand faculties of sense can in no way apprehend objects and that on the other hand consciousness is produced dependent on the eye and various forms.[Candrakīrti's reply]: Is this the only astonishing thing that you have observed?Aren't the following astonishing?A sprout cannot reasonably arise from a seed that has ceased to exist or from one that has not ceased to exist, and yet the sprout does [indeed] arise dependent on the seed.Similarly, a volitional action (karma) that has been performed and accumulated cannot abide anywhere once it has ceased, but nevertheless from a volitional action that ceased to exist hundreds of thousands of eons ago there does manifestly arise an effect.Furthermore, vases and such [objects], if examined by the fivefold analysis as to whether they are identical with or different from their causes, cannot possibly exist, but due to dependent designation (upādāya prajapti) they are nevertheless suitable for performing actions such as containing and scooping honey, water, and other such [liquids].So therefore when there is nothing astonishing on earth for the wise, then what is so amazing about the apprehension of the faculties of sense?(Dbu ma ya 207a) 65 As we can see in the text, the conclusion Candrakīrti arrives at, after considering his arguments, astonishes his opponent, but not Candrakīrti himself.The contrast in their conclusions signifies the divide between the epistemological projects of the Prāsaṅgika's non-foundationalism and the Pramāṇika's foundationalism.The Prāsaṅgika's opponent advances arguments in an attempt to prove that sensory perceptions effectively function as reliable cognitions because of the intrinsic nature of the cognitions or of the objects.The opponent believes that only such a privileged and robust ontological foundation would allow perceptions to become a reliable epistemic cognition.This is so for the opponent, because only when the intrinsic reality of perceptual cognitions and the objects is established would the force of mutual dependency be established between the two: the senses and their objects. 65Ci nas kyang dbang po rnams yul 'dzin pa mi srid pa mig dang gzugs la brten nas mig gi rnam par shes pa yang skye'o zhes bya ba 'di ni ngo mtshar ba zhig go // brjod par bya ste / ci 'di kho na ngo mtshar gyi gang 'gags pa dang 'gag bzhin pa'i sa bon las myu gu 'byung bar mi rigs la sa bon la /.
Candrakīrti however advances his arguments by insisting that perceptions effectively function as reliable epistemic resources strictly on the grounds that they are unreal and lack any intrinsic nature, for only such an exclusive ontological reason allows the perceptual faculties to become reliable epistemic resources.This is so for Candrakīrti because only when the sensory perceptions and objects are proven empty of any intrinsic reality would we able to justify their existence and epistemic efficacy through the force of dependent co-arising.Candrakīrti therefore categorically refuses to attribute any intrinsic reality to things and cognitions and consistently argues that even the slightest reification of the faculties of sense, consciousness, and objects would render them causally ineffective and hence would rob them of their ability to perform any epistemic function.