Studies on Bhartṛhari and the Pratyabhijñā: The Case of svasaṃvedana

Abstract

The article addresses a critical problem in the history of South Asian philosophy, namely the nature of the ‘knowledge of knowledge’ (svasaṃvedana). In particular, it investigates how the Śaiva tantric school of the Pratyabhijñā (10th–11th c. CE) used the notion as an argument against the Buddhists’ ideas on the nature of the self. The paper then considers the possibility that the source of the Śaivas’ discussion was the work of the philosopher/grammarian Bhartṛhari (5th c. CE).

1. Introduction

This article is the first outcome of an ongoing project that is assessing the impact of the grammarian/philosopher Bhartṛhari (5th c. CE) on the thought of the Pratyabhijñā, in particular on the phase of that school as reflected in the works of Utpaladeva (925–975) and Abhinavagupta (975–1025).1

The fact that Bhartṛhari influenced this Śaiva tantric tradition is certainly not news. It was already noticed by the editors of the Kashmir Series of Texts and Studies at a time in which the grammarian’s masterpiece, the Vākyapadīya (VP) was mainly accessible through manuscripts.2 Since the textual and hermeneutical knowledge of both Bhartṛhari and the Pratyabhijñā has consolidated over the last decades, the interest of scholars in this particular issue has started to grow.3 Some affinities have been easily noted: both traditions, for instance, defend a non-dualistic metaphysics and emphasize the importance of language in humans’ understanding of reality. Yet, to get to the core of the relationship between the two traditions, a third element must be introduced, namely the role that Buddhist thought—in particular that of the Pramāṇavāda school—played in defining the aims and the arguments of the Pratyabhijñā.4 That many doctrines of the school derive from Utpaladeva’s innovative fusion of ideas more or less explicitly inspired by Bhartṛhari and arguments developed by the Pramāṇavādins (Pratyabhijñā’s main opponents but also a major source of philosophical inspiration) is a mainstay of the work that has been done by Raffaele Torella.5 In his 1994 edition and annotated translation of Utpaladeva’s magnum opus—the Īśvarapratyabhijñākārikās (ĪPK) accompanied by the author’s short commentary thereon, the Vṛtti (ĪPKVṛ)—Torella already suggested that Utpaladeva employed concepts that are traceable in Bhartṛhari’s texts with the intention to attack one of the most basic conceptions of Buddhism—if not the most basic—namely the idea that the self is a fictitious notion (nairātmyadarśana). In 2008, Torella returned to this question by focussing on the history of Bhartṛhari’s appropriation by the Pratyabhijñā. He dealt in particular with the radical change that occurred between Somānanda (900–950), the founder of the school, and his disciple Utpaladeva. The former’s main work, the Śivadṛṣṭi (ŚD), contains in fact a long tirade against Bhartṛhari’s conceptions, in which the author exhorts the grammarians to mind their own business and not worry about philosophical questions (Torella 2008, p. 512; Nemec 2011, pp. 59–67). In contrast, Utpaladeva’s stance towards Bhartṛhari was much more favourable. Possible reasons behind these diverging attitudes are that the two authors had a different readership in mind and were operating in different intellectual contexts. The most striking peculiarity of the ŚD is how it combines mystical insights with criticism of opposing philosophical views. This was the consequence, as John Nemec has it, of the work being “probably intended for a philosophically oriented audience, but one that was primarily made up of tantric initiates, or for potential initiates who would be predisposed to the scriptural tone and high, if mixed, register of the work” (Nemec 2011, p. 20). In contrast, Utpaladeva was less concerned with religious questions than with a rational justification of the Śaiva doctrines. He therefore considered it necessary to place the Pratyabhijñā into the broader field of non-sectarian debate, by discussing his own ideas in connection with those of other traditions, including Buddhist Pramāṇavāda, Bhartṛhari, Nyāya-Vaiśeṣika, Mīmāṃsā and Sāṃkhya. Why this happened precisely with him is difficult to say. Torella has hypothesized that the purpose of Utpaladeva’s version of the Pratyabhijñā was “to offer itself implicitly as an alternative to the dominant Śaiva Siddhānta, or at least to establish itself as a non-extraneous element” (Torella 2002, p. xiii); Alexis Sanderson, on the other hand, has stressed that the decision derived “from the nature of the commentators’ social milieu, which is one of Śaiva brahmins eager to consolidate their religion on the level of high culture” (Sanderson 2007, p. 241). Whatever the reason, Johannes Bronkhorst is certainly right in pointing out that Utpaladeva’s openness towards others testifies to the radical influence that opponents’ tenets had had on the Pratyabhijñā. In addition, implicitly, it also proves how remarkable the capacity of attraction of the Indian rationalist tradition was, if even tantric gurus were finally forced to prove their theses by turning to philosophical scrutiny (Bronkhorst 1996, p. 2). With regard to Bhartṛhari, Utpaladeva must have realized that in this new context, where the Pratyabhijñā was competing on a wider arena, he could count on the grammarian’s arguments to promote his own ideas and to challenge those of his rivals. This is why in the timespan of a generation, if one wants to keep Torella’s expression, Bhartṛhari abandoned the role of “the main adversary” by taking on that of “the main ally”.

Now, this paper is a first attempt to continue to follow the thread Torella and others have picked up.6 My central claim is easily formulated: I believe that Bhartṛhari’s ideas lie behind several aspects of the Śaivas’ theoretical construction. The Pratyabhijñā’s appropriation of these concepts was intentional and selective, in the sense that it was subordinated to the school’s philosophical agenda, which in turn had been largely defined by its stormy relationship with Buddhism. Among the various cases of acquisition and reworking of Bhartṛhari’s concepts one is particularly subtle and it has been largely overlooked. It concerns an epistemological theory that was widely debated by South Asian pramāṇa theorists, that is, the idea that a cognition is by definition able to reveal itself. The notion—technically expressed by terms such as svasaṃvedana, ātmasaṃvedana, svasaṃvitti etc.—comes down to the fact that when cognizing an entity one is also always aware of the cognition itself. For example, when seeing a blue lotus one is necessarily aware of the cognitive act directed at the lotus. This opinion is obviously questionable but it is nevertheless adopted by all the thinkers under analysis here—Bhartṛhari, the Buddhists Pramāṇavādins, and those of the Pratyabhijñā—although it is given a somewhat diverging connotation by each.

The main contention of this article is that Utpaladeva and Abhinavagupta offer a peculiar and restrictive interpretation of svasaṃvedana in order to mount an attack on the Buddhist idea of nairātmya. This idea has been inspired by several stanzas of Bhartṛhari’s VP, a couple of which are explicitly quoted by Abhinavagupta in the Īśvarapratyabhijñāvivṛtivimarśinī (ĪPVV). In what follows, I shall devote the first section to clarifying the theoretical premises of the debate and to summing up the Buddhist view of the question. The second will explain why self-reflexivity of cognitions played a pivotal role in the Pratyabhijñā’s perspective, and how its main thinkers presented the notion as an argument to defend the existence of a real self. Finally, I shall explore the possibility that Bhartṛhari was the source of the Pratyabhijñā’s analysis.

2. svasaṃvedana: the Theoretical Background and the Buddhist Position

The dispute on the nature of the knowledge of knowledge predates the establishment of the Pratyabhijñā, and by the time the school had reached its height it had long been a source of confrontation among South Asian philosophers. To disentangle the intricacies of the debate one can refer to a scheme developed by B.K. Matilal (Matilal 1986, pp. 141–79), who arranged the controversy in terms of contrasting propositions. The model is rather detailed but for our purposes a simplified version is sufficient. There are two main theses facing each other:

T¹:

a cognition C¹ grasps an object or an event and also itself.

T²:

a cognition C¹ grasps an object or an event, but in order to apprehend C¹ another cognition is necessary, C².

Generally speaking, T¹ is accepted by the Buddhists, the Pratyabhijñā, Bhartṛhari and the Prābhākara Mīmāṃsā. T² is endorsed by the Nyāya-Vaiśeṣika and the Bhāṭṭa Mīmāṃsā. This main dichotomy generates further alternatives, the following being those that most concern us:

T³:

once a cognition C¹ arises, it is necessarily aware of itself.

T⁴:

once a cognition C¹ arises, the fact that such a cognition is also self-aware depends on contingent factors. In other words, it is not true that any cognition is necessarily aware of itself. Some cognitions may arise without being cognized.

T³ is supported by both Mīmāṃsā denominations, the Buddhists, the Pratyabhijñā and the Advaita Vedānta. T⁴ is defended by the Nyāya-Vaiśeṣika and, in the peculiar way we are going to discuss below, by Bhartṛhari.

Despite the usual historical provisos, we are reasonably sure that the question was first formalized by the Buddhist Pramāṇavādins, specifically by Dignāga in the Pramāṇasamuccaya (PS).7 In PS 1.6_ab Dignāga affirms that cognition’s self-awareness has a non-conceptual nature, it being a form of perception. This means that svasaṃvedana is a rightful means to acquire knowledge, exactly in the same way standard, sense-based perceptions are.8 The text further adds that to have a non-conceptualized self-awareness of a conceptualized cognition is perfectly legitimate. In other words, a cognition may be conceptualized (vikalpaka) as far as its content is concerned, but its self-awareness is always perceptual.9 Dignāga’s most decisive contribution comes nevertheless in PS 1.11_ab, where he states that cognitions have a double nature (dvirūpatā), meaning that they cognize at the same time both their content and themselves.10 In the corresponding Vṛtti (PSV) the author puts forward three different reasons to support this claim. First, he argues that if a cognition were to possess just one nature—be it its own-one (which he calls the svābhāsa aspect) or the content-one (the viṣayābhāsa one)—one would encounter problems, because the distinction between the ‘cognition of the object’ and that of ‘the awareness of the cognition of the object’ would collapse.11 Second, if cognitions possessed just the svābhāsa nature, thus lacking the viṣaya one, a later cognition would be unable to illuminate the content of a former one, because the content of the earlier cognition will be gone when the later takes place.12 Third, memory itself is a proof that a cognition has two forms: when one remembers something, both the former cognition and its content are in fact recollected.13 This third consideration also proves that cognitions are self-aware, since memory applies only to what has already been experienced: if I remember that I perceived a blue lotus it means that the perception of the blue lotus has previously been cognised.14 It goes without saying that one could explain the awareness of this perception differently, that is, by arguing that the original perception is cognised by a second one (proposition T² of Matilal’s scheme). Nonetheless, Dignāga rules this possibility out immediately because it would lead to infinite regress: to negate that cognitions are self-aware requires the postulation of a second-order cognition to explain the first, then a third to explain the second and so on.15 Hence the conclusion, accepted by the whole ensuing Buddhist tradition, is that a cognition necessarily reveals both its content and itself.

The discussion is then carried on by Dharmakīrti, Dignāga’s most influential follower. His treatment of self-awareness has been less studied than his predecessor’s, but one can rely on a recent essay by Birgit Kellner that focusses on the Pramāṇaviniścaya (PVin), a study I am mostly following here (Kellner 2011, pp. 420–23). Differently from Dignāga—whose most compelling argument in favour of svasaṃvedana is founded on the impossibility to explain memory without resorting to second-order cognitions—Dharmakīrti’s defence of the concept is linked to one of his most crucial epistemological tenets. In PVin 1.54 he in fact claims that a cognition does not really differ from its object, since both are perceived together at the same time: if a cognition were not perceived, its content would not be perceived either.16 In the prose commentary thereupon Dharmakīrti goes into the details of this problem. His task is twofold: he must first explain why one never has the perception of an object without that of the cognition. Then, he has to describe how such a perception takes place. As for the first question, the arguments Dharmakīrti provides are not particularly compelling. In the end, one gets the impression that he has failed to explain why a cognition must always be cognised together with the object. As Kellner puts it: “the problem […] is that while one may grant[…] that perception needs to exist in order for the object to be perceived, this by no means entails that the perception needs to be known in order to exercise its function” (Kellner 2011, p. 421).17 However, what concerns us most is the discussion of the second issue—how actually a perception is known—since it is here that svasaṃvedana comes up again. Dharmakīrti in fact claims that a cognition can be cognised only by itself; to suppose that is revealed by another one is unsound for at least two reasons. First, one would end up in the rather bizarre situation in which the object is not established at the time of perception—since the perception cognising it is has not yet been established—but it is established at the moment of the subsequent cognition, when the object has clearly vanished.18 Second, one would again fall into the infinite regress trap: postulating the existence of a second-order cognition to account for the first requires a third-order one to explain the second and so on.19

That said, Dignāga and Dharmakīrti’s ‘epistemological’ interpretation of svasaṃvedana is not the only one Buddhist philosophy has offered over time. One has to mention at least one further version, endorsed in the works of slightly later thinkers such as Śāntarakṣita and Kamalaśīla (8th c. CE). Śāntarakṣita’s discussion of self-awareness largely takes place in a chapter of his Tattvasaṃgraha (TS) dedicated to the investigation of nature of external objects (bahirarthaparīkṣā). Here the author is defending the idea introduced by Dharmakīrti whereby the content of a cognition and the cognition itself do not differ at all.20 The conception is heavily criticized by other schools, in particular by the Mīmāṃsā, whose position Śāntarakṣita is presenting here by quoting extensively from Kumārila’s Ślokavārttikas (ŚV).21 The refutation of Kumārila’s positions is carried out by exploiting the epistemological argument and the memory argument based on infinite regress we have already seen at work in Dignāga’s and Dharmakīrti’s discussion. This is particularly evident in TS 2022–2023 (where Śāntarakṣita sums up the epistemological argument introduced by Dharmakīrti in PVin almost paraphrasing it)22 as well as in TS 2024, which clearly evokes Dignāga’s appeal to infinite regress.23 Still, Śāntarakṣita considers two further features of self-awareness that Dignāga and Dharmakīrti did not take into account explicitly: self-awareness is not determined by other cognitions,24 and it is the hallmark of the living.25 All of this has induced Paul Williams to define two fundamental ways Buddhist philosophers coped with the notion of svasaṃvedana. While Dignāga and Dharmakīrti adopted an ‘intentional’ conception of self-awareness (Williams 1998, p. 30)—in which a cognitive event is aware of itself as an object—Śāntarakṣita, both in the TS and in the Madhyamakālaṃkāra, backs also a further variety, which Williams calls the ‘reflexivity’ type of self-awareness. In this latter case, as Williams remarks, “consciousness is self-referring in a non-objectifying way, just as a lamp illuminates itself not as one object among others to be illuminated, but through the very act of being a lamp, an illuminator of others” (Williams 1998, p. 20). Whereas the ‘intentional’ type of self-awareness exemplified in Dignāga’s and Dharmakīrti’s works, “in some sense takes an object, and in some sense that object is itself” (Williams 1998, pp. 20–21), in this second case, svasaṃvedana is to be intended as pure reflexivity. As such, it does not require any philosophical inquiry, being perfectly evident even to the man in the street.26

Conceived as pure luminosity, svasaṃvedana is therefore for Śāntarakṣita ‘not objectified by knowledge’ as well as the factor that ‘distinguishes the living from what is inert’.27 As we are going to see below, these two points are pivotal also in the Pratyabhijñā’s understanding of svasaṃvedana. However, even more crucial is the opposition between the intentional and the reflexive type of self-awareness. Utpaladeva and Abhinavagupta consider these somewhat contradictory and use this as an argument against the Buddhists’ way of conceiving self-awareness and, more generally, against their understanding of reality.

3. The Pratyabhijñā: svasaṃvedana as an Argument

Most of Utpaladeva’s epistemological claims are put forth in the Jñānādhikāra—the first of the four chapters that make up the ĪPK—where the philosopher generally aims at establishing the existence of a self by stressing his capacity of being a knowing subject. In a broader perspective, this self is then regarded as the manifestation of an all-pervading consciousness (saṃvid), which on a theological level is equated with Śiva. The Jñānādhikāra develops this main theme through the typical pūrvapakṣa/siddhānta pattern and discusses several tangential issues, but its central principles can be condensed in few points:

• Utpaladeva examines knowledge by focussing first on a particular case: memory. If one keeps in mind the author’s purposes it is easy to see why. Memory is a cognitive event that is by nature extended over time. It is, therefore, prone to be explained through the presence of a permanent knowing self. The Buddhists take a stand against this notion and claim that recollection can actually be justified on the grounds of cognitions alone, more specifically by the traces the original perception leaves in the mental continuum of the “knower”.28 In Utpaladeva’s view, conversely, memory can only be explained by accepting that an element of self-awareness (which he calls vimarśa) is already present at the moment of the original cognition (called prakāśa). The two elements, both being manifestations of a unitary consciousness, have the same nature and ultimately point to the existence of a real self.29
• In his second step these results are extended to all types of cognition. How? If one presupposes that vimarśa is a permanent feature in any instance of memory (let us call it M¹)—thus belonging both to the original cognition x and to the subsequent recollection y—one must concede that vimarśa is also present in a later cognition z directed at y (hence a second instance of memory, M²) and so on, endlessly. This allows the conclusion that vimarśa is present in any cognition and is therefore the defining feature of all knowledge.30
• The third step consists in specifying that the most intimate nature of vimarśa is language.31
• Finally—and this is a plain apologetic step—even though vimarśa is language-informed, it is not a mental fabrication or a form of conceptualized knowledge as the Buddhists would be ready to point out.32

Now the first two points have one thing in common: their justification is based on the assumption that a cognition is always self-revealing and never the content of another. Such an idea is really one of the most crucial arguments of the Pratyabhijñā, possibly the most crucial. Its application is so pervasive and strict that one can safely regard it as a fundamental axiom in the epistemology of the school. To put it explicitly, the gist of the Śaivas’ reasoning is that if the Buddhists really want to argue for the self-reflexive nature of cognitions, they must by all means avoid the possibility that a cognition ends up being objectified by another. However, the only way to do that—and at the same time to provide a satisfactory explanation of reality—is to accept the existence of a knowing subject. Since the Buddhists do not recognise such a subject, their conception of svasaṃvedana is flawed.

The most explicit application of this principle comes at the very beginning of the third section (āhnika) of the Jñānādhikāra, where Utpaladeva discusses and rejects the Buddhist explanation of memory. Right at the outset, he makes it clear that memory can never access an original perception since, as the Vṛtti says, it “does not penetrate the former direct perception”.33 The original cognition is in fact absolutely “restricted to itself”.34 In this regard, Abhinavagupta adds that explaining memory by means of mental traces (saṃskāras) is incorrect, since in the end saṃskāras are cognitions too and therefore they cannot be objectified.35 The Śaivas argue that memory can be explained only by accepting the existence of a knowing subject who is able to guarantee a temporal continuity between a past perception and the subsequent recollection.

However, the svasaṃvedana-argument is by no means restricted to the discussion of memory. It is sufficient here to consider two further cases. The first concerns the analysis of the so-called ‘invalidated-invalidating’ relation (bādhyabādhakabhāva) that takes place between two cognitions in the case of wrong judgements, when a previous false cognition is invalidated by a later, correct one. Standard examples are those of the perception of silver instead of mother-of-pearl, or of a snake in place of a rope. The issue at stake is to explain what sort of relationship exists between these cognitions. In ĪPK 1.7.7 Utpaladeva considers the case in which one sees a surface and realizes that it is devoid of a certain object, a jar for instance. This is a case of anupalabdhi or ‘non-perception’. According to the Buddhists the mere perception of the absence of a jar is enough to prove that the surface is empty: a later correct perception would in fact negate the validity of the former one, just like the cognition of the mother-of-pearl wipes out that of silver.36 However, for the Pratyabhijñā this argument is unsound: the Buddhists are actually confusing two different ways of conceiving non-existence, an absolute one (tādātmyābhāva), in which the contents of the two cognitions (the invalidated-invalidating ones) are identical (the case of silver and mother-of-pearl), and a relative one (vyatirekābhāva) in which the contents differ. The Buddhist argument according to which the mere perception of mother-of-pearl invalidates the perception of silver is valid as far the tādātmyābhāva is concerned, but it does not hold for the vyatirekābhāva. According to Utapaladeva, the non-existence of an object on a certain surface is precisely an instance of vyatirekābhāva, and therefore it cannot be accounted for on the mere basis of a subsequent correct cognition.37 It is evident that the Śaiva is once again resorting to the idea that cognitions are self-revealing and self-confined. A surface and a hypothetical pot that may or may not be placed on it are the contents of two unrelated cognitions; these can be connected only by admitting the unifying activity of a knowing subject.38

The second case is more metaphysics oriented and regards the possibility of imposing unity over multiplicity. This topic—discussed in the second āhnika of the Kriyādhikāra—is of a broader significance, since it comes down to the question whether ‘metaphysical’ ideas—such as those of action, relation, time etc.—have epistemological legitimacy. The Buddhists neatly refute this possibility, by arguing that reality is made up of absolutely discrete and ever changing components. In their view, any attempt to reify this state of affairs produces conceptualized notions, vikalpas, which although of some practical use, are also inevitably inaccurate. As expected, the Śaivas claim the opposite. According to Utpaladeva, metaphysical ideas are legitimate, and they are so for two main reasons: they are “permanent and useful” (sthairyopayoga, see ĪPK 2.2.1). Let us focus on the first aspect. Here ‘permanence’ has an epistemological connotation, somehow connected to the bādhyabhādaka relationship we have discussed above. It means that the notion of, say, ‘action’ is never set aside by a later cognition and thus it remains valid over time. In the ĪPV Abhinavagupta makes clear that there is a substantial difference between the cognition of a certain action, such as ‘Caitra is going’, and that of ‘seeing two moons’, which takes place in erroneous judgements. The latter is in fact invalidated by a subsequent perception, while that is not the case for the former.39 The decisive argument is again the idea that a cognition is self-contained and impossible to be objectified: the Śaivas could even go as far as accepting that the notion of action is a vikalpa (that is to say a reification of a series of instantaneous movements in space and time that are actually independent from one another) but this would not change the fact that this vikalpa is in the end a cognition and, as such, self-restricted and unable to go beyond itself like all cognitions. How then to explain the persistence over time of metaphysical ideas, whose phenomenological appearance is so hard to deny? Again, this is by accepting the existence of a conscious subject who unifies the cognitions.

I think this quick survey is enough to show how tenacious the Śaivas were in applying the axiom of self-awareness of cognitions in order to defend the legitimacy of a knowing subject. Yet it still does not explicitly clarify the reason why they argue that the Buddhists’ understanding of this notion is in the end wrong. To appreciate it, one must pay attention to the definition of self-awareness provided by Abhinavagupta in the ĪPV on ĪPK 1.3.2. The passage is meant as an explanation of Utpaladeva’s expression dṛk svābhāsaiva, “a cognition is self-revealing”, contained in the corresponding stanza:

[In the verse] the word drś means “knowledge”. This knowledge differs from what is inert, insofar as its nature consists only of illuminating itself. For what is inert must be regarded as different from light. Hence, the expression ‘a cognition is self-illuminating’ (dṛk svābhāsa) means that: (1) the unfailing nature of a cognition is the capacity to illuminate (prakāśamānatā); or that (2) the proper nature of a cognition consists of illuminating itself.40

The crucial point is Abhinavagupta’s interpretation of the compound svābhāsa.41 In the first gloss, he takes ‘self-revealing’ as meaning that the own (sva) nature of a cognition consists in illuminating (ābhāsa) something else, a definition that highlights the intentional nature of knowledge, which is innately capable of illuminating, i.e., revealing, a certain content. The point is further corroborated by Bhāskara Kaṇṭha’s commentary on the ĪPV where Abhinavagupta’s prakāśamānatā is glossed as bāhyasaṃbandhiprakāśakartṛtva, loosely meaning “the capacity of being a knowing, illuminating agent in relation to external objects” (Abhinavagupta 1986, p. 126). Differently, with the second interpretation—according to which the own nature of a cognition consists in illuminating (ābhāsa) only itself (sva)—Abhinavagupta is highlighting the reflexive nature of knowledge. Thus, in few words the ĪPV seems to reproduce the dichotomy between the two types of svasaṃvedana we have seen displayed in the Buddhist exegesis. The most decisive remark comes nevertheless immediately after:

Even admitting the existence of external objects, a light which is reflected on the material aspects of external things cannot be the proper nature of a cognition, because we maintain that the self-illuminating nature of a cognition is no more than the illumination of its own nature, in the form of the illumination of a different [external] thing.42

Here Pratyabhijñā’s fundamental contention is finally made explicit: a cognition is able to illuminate something else only by illuminating itself. The illumination of a content is actually nothing more than the manifestation of consciousness in an external form. This means that the intentional type of svasaṃvedana, whereby a cognition assumes itself as a content, is hardly justifiable if one keeps only cognitions in the picture, for that would imply the objectification of cognitions themselves:

One may object that if there is self-illumination of cognitions the cognition of an earlier perception will shine in a [later] recollection. [Utpaladeva] negates this idea, since “a cognition cannot be the object of another cognition”. If a cognition shines in another then it is no longer self-illuminated. For this is precisely the hallmark of self-illumination.43

It should be clear now why the Śaivas accuse the Buddhists of having construed a contradictory picture of svasaṃvedana: if the Buddhists concede that knowledge is able to assume itself as a content (i.e., the intentional type of svasaṃvedana) then its auto-luminous and sentient nature (i.e., the reflexivity type) is automatically contradicted. For their part, the Śaivas must nonetheless explain why self-awareness has an intentional connotation as well: in the end, our common experience is replete with myriads of cognitions that apparently have other cognitions as their content. Regarding this, the Śaivas believe that the presence of an intentional element in self-awareness is absolutely legitimate, provided one accepts the presence of a subject who guarantees the continuity of cognitions, something that a Buddhist would hardly agree with. In other words, the Pratyabhijñā grants to the Buddhists the possibility of building an epistemology based on cognitions as self-luminous. However, self-awareness of cognitions must always go together with their self-contained nature and the only way to comply with both requirements is to acknowledge the role, hence the existence, of a subject.44

This also makes it clear that there is an intimate relationship between svasaṃvedana and the self, as Utpaladeva himself says in a reconstructed fragment of the Vivṛti:

It has therefore been proved that being conscious of itself [on the part of cognition is pervaded by the light of the I, which is opposed to insentience—and insentience for its part, has the nature of “this”, which pervades the property of being knowable by others. Thus it is possible to deny that a cognition is knowable by other [cognition], because this property is pervaded by another that is in opposition to the pervading one.45

(Torella 1992, p. 337)

Here the author is giving a logical flavour to the idea we have been discussing at length. He affirms that between self-awareness of cognition and the “I” (i.e., the self) there is an invariable concomitance (vyāpti), meaning that whenever there is self-awareness one must surmise the existence of a self. The same holds true with regard to insentient beings and the notion of objectivity (idantā). If all this is posited, by claiming that a cognition can be the object of something else one would incur the fallacy of attributing to self-awareness a property (i.e., insentience) that is in contradiction with the pervading one (vyāpaka), i.e., the sentiency of the self.

We are now left with a last question to address, one that concerns the possible influence of earlier philosophies on the Pratyabhijñā’s discussion of svasaṃvedana. Were Utpaladeva and Abhinavagupta aware of earlier interpretations of this concept? Are there plausible antecedents to their extensive application of the axiom of non-objectification of cognitions? As is well known, historical reconstructions of premodern South Asia are problematic. However, since the Śaivas’ familiarity with the Buddhist thought is well documented, their discussion of svasaṃvedana can be explained on the basis of the debate that took place in Buddhist philosophy in the preceding centuries. Still, in the works of Utpaladeva and Abhinavagupta there are aspects of the question that are either hardly discussed in the Buddhist sources or have less relevance; in particular, Pratyabhijñā’s insistence on the purely reflexive nature of cognitions and the self-confined nature of knowledge seems to come from somewhere else. In this regard, Bhartṛhari, who predates all the Buddhist thinkers we have been discussing so far, is an author worth closer examination.

4. Bhartṛhari on svasaṃvedana

4.1. The Jātisamuddeśa Section

In the monograph mentioned above, P. Williams suggests that the notion of ‘reflexive self-awareness’ may have been an innovation of the Madhyamaka-Yogācāra school, perhaps under the influence of Dharmakīrti’s thought: “The idea of portraying self-awareness as the quality of consciousness understood as the reverse of insentience (bems po) may well have originated with Śāntarakṣita” (Williams 1998, p. 25). Actually, this can hardly be the case. Though the fact has admittedly been overlooked so far, Bhartṛhari was really one of the earliest South Asian thinkers to argue explicitly for the self-luminous nature of knowledge and, consequently, to claim that a cognition cannot be objectified. Bhartṛhari concentrated his most decisive statements on the nature of svasaṃvedana in a specific part of the VP, that is at the very end of the Jātisamuddeśa, a subsection of the third kāṇḍa loosely dedicated to analysing competing views on the existence of universals. By drawing this chapter to a close, Bhartṛhari is faced with the need to offer an alternative theory to the Buddhist ontological standpoint whereby both things and cognitions are radically different from one another. He renders such a position in VP 3.3.101, where it is claimed that if things and cognitions appear to be somewhat similar, it is not because they really are so, but is due to the force of a conceptualized notion that is devoid of real existence.46 The 10th c. commentator Helārāja47 elaborates on this by saying that according to the Buddhists a unitary cognition is impossible due to cognitions’ intrinsic singularity.48 The Buddhist position is further spelled out in the two following kārikās, where it is claimed that if a conceptualized notion arises, this basically depends on the impossibility of expressing through language the absolute, radical difference existing between things and events.49 Immediately after this, Helārāja introduces the view of those who disagree with the aforementioned position but argue for the existence of real universals. In VP 3.1.104 Bhartṛhari calls them the supporters of the “connection theory” (saṃsargadarśana). This view is then expanded and thoroughly analysed, as if it were Bhartṛhari’s siddhānta, in four crucial verses: VP 3.1.105–106 and 3.1.109–110.50 Now, since the entire samuddeśa abounds with alternative theories on the nature of universal, and the formulation of these kārikās being quite succinct, one may doubt whether this saṃsarga stance is really the one Bhartṛhari is adopting. Actually, in this case there is no much room for doubt: first, this is the last view advocated in the samuddeśa. In addition, most importantly, we have a decisive remark by Helārāja who affirms in his commentary that this position is ‘our [Vaiyakaraṇa’s] stance’. Given the importance of the whole passage for establishing Bhartṛhari’s view on svasaṃvedana, it is worth reproducing it in its entirety, together with Helārāja’s explanation:

[VP 3.1.105] “The universal resides only in the content of a cognition and is subsidiary to cognitions. A cognition is never represented by another form, differently from what happens in the case of its content”.

[Helārāja] While the content of a cognition is represented, i.e., appears, in the form of a distinct universal, this is not the case for a cognition, which is never represented in the form of a distinct universal, because it is always accompanied by self-consciousness. In fact, the distinguishing mark of the content of a cognition is that its representation takes place through another form, whereas a cognition is dependent on the content but never becomes it. Still, since a cognition is restricted by its content, we must also accept that there is an ascertainment of identity between them, for the universal which resides in the content of a cognition is not different from cognitions, because the latter are grounded on contents. Now, being unacceptable that cognitions possess differentiation by nature—for they are dependent, insofar as they are determined by something external—such a differentiation must take place on the basis of an external, unitary cause. For example, in the representation of a pot such as “this is a pot” there is an external cause; so even if there is difference [between the cognition and the content] it is right to say that there is a universal contained in that [cognition].

Why then does a cognition not become a content and is it not represented through another form? Bhartṛhari explains:

[VP 3.1.106] “Just as a light is never illuminated by another light, similarly what has the form of a cognition cannot be apprehended by another cognition”.

Just as a lamp that illuminates a pot etc., does not require another lamp to illuminate itself, similarly a cognition illuminating a content does not require a different cognition to illuminate itself, for its self-luminosity is settled. If a cognition were illuminated, it would acquire the nature of an inert thing, for its nature of being an illuminator is what distinguishes it from an inert thing. Furthermore, if at the moment of the cognition of a content the illuminating cognition were lightless, there would be no apprehension of the object at all, because if a cognition is not sentient the content connected with such a cognition cannot be established. In addition, if the cognition is not perceived at the same time of its content what would be the use of a cognition occurring later in time? In fact, when a manifestation takes place, one could not have a knowledge directed at oneself such as “this object has been manifest to me”, since at that very moment the cognition directed at oneself would be insentient. Therefore, a cognition is self-luminous and it is not grasped by another cognition, for the property of being a cognisor is absent in a thing that is cognised by something else. Now since there is no representation of a cognition through another form, the cause of the unitary nature of this cognition is the undivided universal that resides in the content only. In addition, it is precisely because of this universal that one grasps an identity between an apprehended representation and another cognition. On the contrary, if the appearance of an identity between cognitions were to be explained on the basis of a universal inhering in them, then the notion of an external universal would be truly groundless; but in fact, the appearance of such an identity is to be explained differently. Being the contents of cognitions different from one another, in their case a unitary universal is justified. While, in the case of cognitions, it is correct to say that they do not have a common feature other than themselves. There is no parallelism with the objects because it has been proven that also this feature common to cognitions derives from the fact the cognitions are differentiated, insofar as they are determined by their content. So there is no place for universals in the realm of cognitions, because it has been established that the ascertainment of an identity between cognitions depends only on the common form present in the mental representations: so it is finally proven that cognitions have a unitary form, are self-illuminating and do not possess universals. Although the Vaiśeṣikās too claim that a universal resides in objects only and is subsidiary to the cognitions, they do not hold that cognitions have a unitary form, nor that they are self-illuminating. Nevertheless, we grammarians have another opinion. In any case, the view that is put forth here has been acquired through a reliable means of knowledge.

Now one may further object: just as in a cognition such as “this is a pot” one ascertains the pot, in the same way in the cognition “this is the cognition of the pot” a pot is ascertained as well. What is then the difference here? The difference is that if a knowledge consisting of the cognition of an object were not perceived, it would not be denoted by language either, because linguistic denotation is preceded by perception. Moreover, a cognition is not able to perceive itself through itself, because this would contradict its own function. However, let us suppose that a cognition, because of its conscious nature, illuminate spontaneously the object even if it is unknown. Let us admit that there is a difference to such an extent between this cognition and external objects. In this case, the point is that in order to know that cognition, another cognition is required. However, a series of cognitions is inappropriate as well, because ordinary activity is absolutely based on a firm discrimination between cognitions and contents of cognitions, and because another second-order knowledge is irrelevant. With regard to this Bhartṛhari says:

[VP 3.1.109] “The cognition ‘this is the cognition of a pot’ is different from the cognition of the pot. The cognition of the pot comes down to the external object”.

[Helārāja] “The cognition ‘this is the cognition of the pot” is a cognition that has the form of a pot and so forth. When it is grasped by another cognition, as when one says: “I had this cognition”, the cognition that is grasping another—namely, that which has a cognition as its object—has features that differ from those of the cognition of a given pot, because it does not arise directly from the object. For only cognitions that are the content of a further cognition derive from the object, not the others, which are [in turn] founded on cognitions. The reason is that there is an intervention [of a cognition] between the two. A relationship of ‘apprehender’ and ‘apprehended’ is established due to the force of a similarity. But the cognition of a pot appears in the other [i.e., second-order] cognition as having different properties. Hence it is not grasped [by the second-order cognition]. Now it might be objected that, due to the continuity (anusaṃdhāna) of the element ‘pot’ the [apprehending] cognition is similar to the [apprehended] one. But then the author says that also “the cognition of the cognition of the pot comes down to the external object”, meaning that it does not come down to cognitions. Therefore, a cognition is never the support of another, for it always shines by itself. This is the meaning of the verse.

Others explain: the cognition that consists in ‘knowing the pot’ is different from the cognition of the pot whose object is the external thing: a cognition which is not engaged with an object and is confined to itself is not conscious. Since the cognition consisting in an object like a pot—which is distinct from the former one and arises from the apprehended thing—is totally subordinated to the apprehended thing and refers to the external object. For, if there is no illumination of an object the existence of a cognition will be groundless. Therefore, even if one thinks in terms of ‘cognition of cognitions’ because there is no difference between the two cognitions in terms of awareness, it is nonetheless correct to say that the content of ‘the cognition of a cognition’ is always the external object.

Others explain ‘the cognition of the cognition of a pot’ by saying that its object is not the ‘cognition of a pot’, but that it has a different nature and is devoid of form. For the consciousness-nature of a cognition is not reflected in what is imagined to be its cognition. The relation between apprehender and apprehended is based on similarity. If that were not the case, since there is no distinction in terms of awareness, there would be no rule to determine what is the perception of what. However, a cognition, whose content is an external object like a pot, conforms itself to that very object and assumes its form.

Or again Bhartṛhari says: “The cognition ‘this is the cognition of the pot’ is different from the cognition of the pot. What is the difference between the two? The author says: “The cognition of the pot comes down to the external object”. Even if this cognition is formless, it nonetheless conforms itself to, determines and ascertains the object in terms of “this has such a form”. However, the cognition “this is the cognition of the pot” does not conform itself to the nature of the cognition (‘this is a pot’). In fact, by ascertaining the ‘cognition of the pot’ as a content, the cognition of the cognition of the pot can at most ascertain that the cognition of the pot is some form of knowledge; yet the conscious nature of the cognition of the pot is not reflected in the cognition of the cognition of the pot, as it happens in the case of the cognition of another person. A cognition is determined by another on the basis of a certain linguistic expression, such as “he had the cognition of a pot”, but this knowledge of another cognition is not manifested for others as it is for oneself, hence the cognition of the cognition of a pot is different from the cognition of a pot and does not determine it. Having an apprehending nature, a cognition is manifested by itself and is therefore said to be self-cognizing. It does not illuminate itself as if it were an object, and it is never the object of its own activity.

Now, one may object that an apprehended object is not different from the cognition and it is apprehended on the basis of a cognition such as “this is the knowledge of a pot”. In this way it is precisely the cognition which is cognised. To remove this doubt Bhartṛhari says:

[VP 3.1.110] For what has the nature of a cognition is not apprehended as having that of an object. Cognition’s own nature is not grasped separately from the object.

Bhartṛhari’s refutation is valid also if one supports the theory that a cognition is devoid of objects (nirākāravāda), for also in that case there is a cognition that represents another cognition in the form of a content, such as “this is the cognition of a pot”. This second-order cognition does not determine the proper nature of the first one in separation from it. Moreover, the manifestation in the form of an apprehended content is not the proper nature of a cognition; it is rather a temporary qualification, just like the colours blue etc. are for a crystal. In addition to this, the subjective aspect of a cognition is not cognised by another cognition as distinct from the objective aspect: due to the pureness of cognition, another cognition is not admissible. Hence, due to the impossibility of having a cognition if the content is not cognised, even what is supposed to be the ‘cognition of a cognition’ depends on the content. The content exists as the representation, and is established as different from that representation, just as when one says: “this object is known”; this is also the case of a cognition like “this is the cognition of blue” since that cognition is rooted in a blue thing: a cognition is apprehended as arising from a content.

Things being so—since one cannot grasp the conscious nature of a cognition through another cognition—an analysis of cognitions in terms of universals that differ from the cognitions does not hold. On the contrary, since there is no distinction between the objective aspect of a cognition and a cognition, the cause of the comprehension of the cognition must be the unitary universal present in the object. Therefore, when there is an objective aspect of a cognition that is identical to the cognition, positing another universal is useless, because this objective aspect is grasped by another cognition precisely so i.e., on the basis of the universal present in the object. Nor is there a fallacy of mutual dependency: there is in fact a difference of activity, since the universal contained in the object produces a unitary cognition, and the unitary cognition expresses the universal in the object. This is unquestionable. If cognitions are determined as identical without the postulation of a universal, the same does not apply for the objects, since cognitions are indeed determined as identical on the basis of a universal, but in the way we have just described. One may object: if the cognition aspect is not regarded as different from the objective aspect, the cognition aspect cannot be established by itself. This is true: the objective aspect is always experienced in concomitance with the cognition aspect. Nevertheless, there are also objects of pleasures to be enjoyed, which follow (standard cognitions?), which are perceived as having the nature of the seat of the sense of the I, and which have qualities different from those (standard) objects of knowledge that are, on the contrary, detached from the seat of the sense of the I. On this basis, cognitions are regarded as having a nature distinguished from that of the object: cognitions such as pleasure etc., even when grasped through another cognition, they are never cognised.

This experience is established and we have dealt with it more than diffusely.

So the universal is confirmed on the basis of a reliable means of knowledge, and when denoted by language it is capable of accomplishing a linguistic usage that deals with both visible and invisible things. Hence the existence of the category of universal is established".51

The starting point of Bhartṛhari’s discussion is the commonsensical experience in which everyone perceives an identity between cognitions. Now, is such an identity based on a common feature, a universal that is instantiated within the various cognitions? The answer is a definite no: a common element actually exists, but it does not pertain to cognitions but rather to their contents. As Helārāja explains, the main difference between the content of a cognition (a pot, the colour blue, the action “running” etc.) and the cognition itself lies in the fact that the former is explainable in terms of universal features, whereas for the latter such a representation is not possible. In the case of contents, a common element is in fact identified by the cognitions themselves. However, in the case of cognitions, what is going to do the job? The only common feature cognitions share is their conscious nature (saṃvidākāra), but such a conscious nature cannot be cognised by another cognition, since then it would become itself a content and thus would lose its most distinguishing characteristic. That we normally perceive a similarity between instances of knowledge depends on the fact that cognitions reflect or assume the universals of their contents. Such a state of affairs is always valid, since in Bhartṛhari’s view any cognition is always related to a content. VP 3.1.106 expands on the reason why a cognition is never objectified by another. Bhartṛhari brings in the example of light: as a source of light is never illuminated by another one, likewise a cognition is never cognised by another one. Helārāja adds further information by specifying that a cognition is always self-revealing and this is the distinguishing mark of the living. Moreover, a cognition is always perceived at the time of the cognition of its content, since otherwise one would need one of a second-order to explain the first and so forth. Then VP 3.1.109 affirms that ‘the cognition of x’ is quite a different thing from ‘the cognition of the cognition of x’. Bhartṛhari states that this difference is caused by the fact that ‘the cognition of x’ comes down to the content, thus meaning that the cognition of x is produced directly by the content, whereas this is not the case for ‘the cognition of the cognition of x’. Helārāja gives four reasons to explain the difference between these two kinds of cognitions. For the sake of clarity, let us call C¹ ‘the cognition of x’ and C² ‘the cognition of the cognition of x’, where x is a pot. In the first interpretation, the pot in C¹ has different characteristics from the pot in C². The pot in C¹ is produced directly from the content, the one in C² only indirectly. This means that the two are not similar and a relationship of ‘apprehender and apprehended’ cannot be established, since such a relationship requires similarity. The conclusion is that C² does not cognize C¹. In the second interpretation, C² must necessarily have a content, for a cognition without a content is unthinkable. The content is clearly the same of C¹, i.e., the pot. The conclusion is again that C² does not cognize C¹; it is directly cognizing the pot. In the third interpretation, C¹ cannot cognize a pot and be cognised by C² at the same time, for in that case the distinction between an apprehender subject and an apprehended content would vanish, and with it, any possibility of dependable knowledge. Again, the conclusion is that C² does not cognize C¹. Finally, in the fourth interpretation, C² can at the most recognise C¹ as some form of cognition, but it cannot cognise the conscious nature of C¹. It is similar to the case of someone cognising the cognition of someone else: one can certainly cognize the content of the cognition of another person, but cannot cognize it in the same way as that person. Therefore, C² does not cognize C¹ and the objectification of a cognition is inadmissible. Finally, VP 3.1.110 is meant to meet the last criticism of the pūrvapakṣin who seeks to identify the cognition with its content. If a cognition and its content are identified, then one is allowed to claim that cognitions are able to assume themselves as their own content. In the stanza Bhartṛhari refutes this view, by conceding that a cognition is certainly never devoid of a content but also emphasizing that between cognitions and contents there is a substantial difference. According to him, whenever we suppose that a cognition x is having another cognition y as its content, we are just cognizing the universal represented in x, which is generated by the content of y. Helārāja expands on this idea by stressing that ‘cognitions of cognitions’ are always rooted in the original object: if a cognition appears to be the content of another it is just a temporary occurrence, as it happens in the case of a crystal assuming a certain colour. At the end, Helārāja returns to the problem of the universal by restating the grammarians’ position on the issue, according to which there is no need to postulate the existence of universals of cognitions because the similarity we recognise in knowledge is adequately explained on the basis of the universals present in external objects. In Helārāja’s words, “the universal contained in the object produces a unitary cognition, and the unitary cognition expresses the universal in the object”.

The entire passage shows how Bhartṛhari’s ideas on “cognitions of cognitions” were close to those of the Pratyabhijñā. One should also consider that, although Helārāja’s commentary provides welcome additional material, the basic information is all contained in the stanzas: first, a cognition cannot be objectified by another, because the knowledge that derives from an object is radically different from the one produced by another cognition; second, there is no need to postulate a universal of cognitions because ‘pure’ cognitions have only consciousness as their common feature, and consciousness cannot be cognised by anything but itself; third, a cognition is always dependent on a content. The only crucial notion that Bhartṛhari does not mention in these stanzas—even though Helārāja does—is that whereby only the living are self-aware of their own cognitions. Nevertheless in other kārikās of the VP the idea is clearly expressed.52

Before seeing how the Pratyabhijñā used Bhartṛhari’s concepts concretely, let us consider few further passages of the VP dealing with the same problems.

4.2. Bhartṛhari on svasaṃvedana: Further Remarks

Apart from the discussion found at the end of the Jātisamuddeśa, other considerations on the nature of cognitions and self-awareness are found throughout the VP. The most substantial is possibly the one offered by three kārikās of the Saṃbandhasamuddeśa: 3.3.23–24 and 3.3.26. The first two restate the concept we discussed above as follows:

For on a cognition that has the nature of a doubt, which is established as instrumental to determine an object and which does not abandon its proper nature, on top of that one cannot apply a further doubt.

When a subsequent determinative cognition is applied to an original one, then the original determinative one does not retain its distinguishing feature.53

Helārāja’s remarks helps to unpack the statement of Bhartṛhari:

When there is a ‘determinative cognition’—i.e., a cognition which has the form of a specific ascertainment, that is to say, whose content is an object being ascertained—then another self-restricted, determinative cognition (nirṇaya) cannot be applied to the first, which is directed at establishing the object. For the own nature of a cognition lies in its dependence on an object. Hence, if one in conceiving a cognition eliminates that, one will be led away from the core nature of the cognition, precisely because that cognition would freely get the status of an object. We have in fact previously proven that ‘the cognition of a pot’ cannot be the object of another cognition, since that would imply the loss of the cognition’s defining nature, because the cognition would abandon its nature of subject (viṣayitvatyāgāt) by becoming the object of another cognition (jñānāntaraviṣayatve). Just as in a unitary cognitive event a cognition that is engaged with some other object cannot be itself the object of knowledge—since its activity is not directed at itself—likewise, the expressive aspect of a linguistic unit cannot be the expressed one: the expressive aspect of the word “inexpressible” does not exclude that its meaning is expressible, because that activity of the word is not in contradiction with itself.54

The novelty here is the connection, which Helārāja makes explicit, between the nature of a cognition and that of a linguistic expression, a similarity that Bhartṛhari also stresses in VP 3.2.26:

A linguistic unit which is employed as expressive cannot be the expressed one.

That by which something is cognized cannot be cognized by something else in the same context.55

As for this Helārāja comments:

‘Being denoted’ (abhideyatā) is a notion that excludes that of ‘being engaged in the action of denoting something else’ if applied at the same time; for a thing that is employed to express something else cannot turn back to itself. It is indeed the very nature of things that what is endowed with the capacity of the subject cannot be at the same time the locus of the capacity of the object; this means that it is contradictory, etc., to attribute to the same locus, at the same time, both independence and dependence.56

Apart from this section, there are also further passages to take into account. Consider, for instance, the following early kārikā from the first kāṇḍa (Bhartṛhari 1966, p. 108) where Bhartṛhari already introduces the idea that knowledge has the unique capacity of illuminating both itself and its content, as well as the notion that there is a special affinity between language and cognition:

“Just as in a cognition both the content and the cognition itself are perceived, in the same way in a linguistic expression both the meaning and the form of that very linguistic expression are manifested”.57

Later in the same chapter Bhartṛhari seems to hint at the idea that it is useless to talk about cognition independently from a content;58 in the second chapter (Bhartṛhari 1983, p. 169) he explicitly says that a cognition never appears in a ‘pure’ state and that is always coloured by the object.59

Further, in a passage of the Dravyasamuddeśa, in a context meant to show how non-existing entities manifest themselves as existent, Bhartṛhari specifies that although a cognition cannot be posited without a content, there is nonetheless a sharp difference between the nature of an object and that of the cognition:

Just as the qualities of an object are utterly non-existent in the cognition, and that which is utterly non-identical is established as identical.

Similarly the forms of the transformations are utterly non-existent in reality, and yet that which is utterly non-identical appears as identical.60

Concerning this point, Helārāja’s commentary is even more straightforward, as it puts the question in the well-known terms of the dichotomy between what is inert (the object) and what is sentient (the cognition). Interestingly, he flavours Bhartṛhari’s affirmations with arguments that come straight from Dharmakīrti, perhaps with the mediation of a later commentator:

According to the Vijñānavāda view, since what is manifested as having the form of the content of a cognition actually does not exist, a quality like the colour ‘blue’ etc. is self-contained, inert and absolutely absent in a sentient cognition. Thus the author says that there is no similarity between what is sentient and what is inert on the basis of some part. With regard to this it has been affirmed: ‘If there were similarity èbetween a cognition and its object] on the basis of one aspect, then everything would apprehend everything. But, on the other hand, if there were similarity in all aspects then a cognition would cease to be a cognition’.61

To conclude, the passages above show how Bhartṛhari was well aware of and upheld the idea that cognitions are by nature self-reflexive; moreover, he was one of the earliest thinkers to regard svasaṃvedana as the hallmark of living entities and to claim that due to the force of its auto-luminosity a cognition can never be objectified. These positions were all-well known to the Śaivas.

4.3. Abhinavagupta’s Quotations of the VP Stanzas

The most straightforward proof that the Śaivas were aware of Bhartṛhari’s discussion of svasaṃvedana is the fact that in the ĪPVV Abhinavagupta cites two of the stanzas of the Jātisamuddeśa we examined above, VP 3.1.106 and 109. The first is quoted precisely where one is expecting to find it, that is, in the third āhnika of the Jñānādhikāra, where Utpaladeva concentrates most of his considerations on self-awareness. More precisely, Abhinavagupta quotes VP 3.1.106 while commenting Utpaladeva’s vivṛti on ĪPK 1.3.7. Here the author claims that ordinary reality would be inexplicable without the activity of unification of cognitions that is exclusive to the knowing subject. The self is then ultimately identified with Śiva and, by appealing to a quotation from the Bhagavadgītā, it is regarded as the source of “memory, knowledge and exclusion”. Specifically, Abhinavagupta is referring to the notion of non-objectification of knowledge to show that the way the Buddhists prove non-existence is wrong. He says:

Utpaladeva now examines the Buddhist way of establishing non-existence with the intention to show that the aforementioned idea whereby ordinary apprehension, directed at itself or at other things, is produced on the basis of ‘a distinction from what is different’ does not hold. With regard to this, it has repeatedly been proven that knowledge has a unitary, self-illuminating nature. As the venerable Bhartṛhari said: “Just as a light is never illuminated by another light, in the same way what has the nature of knowledge is never made visible in another knowledge”. Therefore, on the basis of the fact that a cognition is not liable to be merged with another cognition, it might be said that if there were two cognitions, a third one should be there to know them. However, here we have the knowledge of a single cognition, hence two cognitions are not present. With regard to this, a pot is the counterexample.62

The second stanza Abhinavagupta quotes, VP 3.1.109, comes in a different section of the work, whose content is nevertheless similar to the previous one. It is in the fourth āhnika of the Jñānādhikāra, precisely in ĪPK 1.4.6, where Utpaladeva maintains that memory never operates on the original perception: the common phrasing ‘I had this perception in the past’ is just a linguistic analysis of the more accurate sentence: ‘this thing was perceived by me in the past’. In this regard Abhinavagupta brings in Bhartṛhari right at the beginning of his discussion by saying:

One may object that when a pot is remembered, that is to say it is in the condition of an object of knowledge, then the original cognition should come together with the knowing subject, but if that original cognition is remembered as an object of cognition, what else can be said? We reply. Even if we say ‘I had a perception of a pot’ it is the pot the object of the cognition. As the venerable Bhartṛhari said: “The cognition ‘I have the cognition of a pot’ is different from the cognition of the pot. The cognition of the pot comes down to the object’.63

Abhinavagupta is thus quoting Bhartṛhari as an authoritative source to corroborate Pratyabhijñā’s conception of svasaṃvedana. It is worth noting that in the quotation of VP 3.1.109 he is interpreting the ‘cognition of the cognition of the pot’ as directly perceiving the pot, a position that is similar to Helārāja’s second explanation we discussed above.

4.4. Pratyabhijñā on Memory and Bhartṛhari’s Liminal Cognitions

Before drawing the conclusions of this essay, we must turn our attention to a last question, perhaps more peripheral, but nevertheless suggestive of a proximity between the Pratyabhijñā’s epistemology and that of Bhartṛhari. The question concerns the way memory is accounted for in each. Although Bhartṛhari does not delve into the problem, some of his epistemological conceptions are useful for explaining how recollection works. I am referring in particular to what Vincenzo Vergiani has recently called ‘liminal perceptions’ (see (Vergiani 2012)). Although Bhartṛhari strongly argues for the presence of language in all cognitive acts—strictly speaking, in his view there is no room for pure perceptual cognitions devoid of conceptualization (nirvikalpakajñāna)—there are cases in which this basic tenet appears to be less compelling. As Vergiani has noted, the Vṛtti records an occurrence of the expression avikalpa jñāna, a phrase apparently at odds with the aforementioned principle whereby conceptualization, i.e., language, permeates all knowledge. Therefore, what is a liminal cognition? Vergiani defines it as “a primordial mental state which exists in every living creature and consists of the awareness of oneself as other than one’s surroundings, but at the same time inevitably reflects an acknowledgement of the surrounding world in its bare spatial and temporal existence” (see (Vergiani 2012, p. 525)). In other words, everyone at any one time, is struck by huge numbers of sensorial stimuli. Since these are not necessarily conceptualized at the moment in which they take place, they would seem fit for being categorized as nirvikalpaka cognitions. However, actually, Bhartṛhari believes that even for liminal knowledge conceptualization is always at work, albeit in a subtler form. What is of interest to our discussion is the reason Bhartṛhari provides for his claim. According to him the fact that liminal cognitions too are vikalpakajñāna is proven by their responsiveness to memory. The most obvious passage that discusses this question is the Vṛtti on the rather notorious VP 1.131, where Bhartṛhari states that all knowledge is imbued with language. In this connection the author comments:

As in the case of somebody’s verbal potentiality in its contracted condition, a non-conceptualized knowledge does not bring about any verbal usage whatsoever, even if produced in relation to known objects. To explain: even the cognition of somebody quickly walking, acquired by entering in contact with grass, lumps of clay and so forth, is a kind of cognitive state in which the seed of a verbal potentiality is present. In it, once manifested the expressive powers of the words—which are explicit or implicit, make grasp the object and are fixed for any object—one cognizes, that is, linguistically denotes, the manifestation of a well defined form (vyaktarūpapratyavabhāsa), which is consistent with knowledge, which has the nature of the object, and which is obtained and concretized by a cognition permeated by language and in accordance with the expressive power of the words.64

Then he adds an apparently hasty remark that is nonetheless crucial:

When the linguistic seeds are manifested due to other reasons, [the manifestation of a well defined form] is the cause of memory.65

Even if Bhartṛhari’s discussion is as usual broad-ranging, one can safely conclude that for him the underlying feature of all knowledge is a cognitive state which is language-permeated. As in the case of the Pratyabhijñā and the Buddhists, it is important to stress that also here memory works as a testing ground for the thesis. The responsiveness of liminal cognitions for recollection is a proof that all knowledge, even one that admittedly appears far from verbal conceptualization, is indeed conceptualized: the fact that at a later time one is capable of recollecting and verbalizing an object or an event that one did not consciously notice at the moment of perception is a proof that the original cognition was already potentially capable of being verbalized, since otherwise the subsequent, actual linguistic utterance would never take place. Referring again to Matilal’s scheme introduced above, Bhartṛhari’s ideas can thus be categorized as belonging to a particular form of T⁴: a cognition comes to be aware of itself only under certain circumstances, but its auto-luminous—as well as linguistic—nature is an innate, ever-present quality. However, what does this cognitive state consist in? If we are talking about a permanent feature that all cognitive states possess at any moment in time, then one is allowed to call this state ‘consciousness’, which is exactly what Bhartṛhari does in VP 1.134.66 In the end, the fundamental idea is that consciousness comes down to be a high-order cognitive state, one that is linguistically informed and that, in turn, informs all the others. This picture had a tremendous impact on Utpaladeva and Abhinavagupta’s philosophy, in particular on the development of the concept of pratyavamarśa, or reflective awareness.67

5. Conclusions

On the basis of the discussion above we can arrive at the following conclusions:

(1)

Generally speaking, as already anticipated by the work of Torella and others, Bhartṛhari’s presence in the Pratyabhijñā literature is not incidental but absolutely functional to the aims of the school, as well as to the actual unfolding of the apologetic debate with the Buddhists.

(2)

More specifically, the Śaivas’ strongest argument against the views of the Buddhists on ontology, epistemology and, even more crucially, on religious and soteriological questions is a strict interpretation of the notion of svasaṃvedana. In this regard, Bhartṛhari is a documented source of inspiration, especially in relation to the notion that a cognition is always restricted to itself and is never the content of another. Whether the Pratyabhijñā thinkers were acquainted with other sources (i.e., Buddhist) affirming the same principle is of course of historical importance and is a question that future research will hopefully address, but even in that case, the fact remains that when the Śaivas looked for an external authority to support their interpretation of svasaṃvedana they quoted Bhartṛhari.

(3)

Finally, it is also a fact that some ideas that Bhartṛhari discusses in relation to svasaṃvedana are clearly endorsed by later Buddhist thinkers. It is enough to mention the very notion that a cognition is necessarily self-revealing but also the characterization of knowledge as the hallmark of the living. This leads to the fascinating but also extremely complex question of the relationship between Bhartṛhari and the Buddhists. As far as svasaṃvedana is concerned, for instance, one cannot rule out a priori the possibility that Dignāga, who knew the VP, was aware of Bhartṛhari’s analysis. This probably would not change the fact that the Buddhist Pramāṇavādins were the first to formalize the question in accurate philosophical terms, but it would certainly give a somehow different perspective to a debate that was so central in the epistemological discourse of premodern South Asia.