unconnected commentary is seen right here in both manuscripts we have obtained.’] Arvind Sharma’s English translation (1983), which is unfortunately based on Pansîkar’s edition, is problematic in several places, although he occasionally consults Kashmirian variants given in the critical edition of the Gītā. For instance in 1.1 Sharma does not seem to notice that Abhinavagupta’s mūla is different from Pansîkar’s text, the former reading sarvakṣatra(with all the Kashmirian sources), the latter sarvakṣetra(with the Vulgate). Therefore, Sharma fails to translate the explanation of kṣatra derived from the root kṣadaccording to Abhinavagupta’s commentary. Similarly, it is unclear whether Gnoli’s Italian translation (1976) is based on Pansîkar’s or Lakshman Joo’s text of the Gītā, although his translation of the GAS is certainly based on Lakshman Joo’s edition. Nevertheless, it is possible that Gnoli follows Pansîkar, at least occasionally, since he seems to elude the difficulty already noticed about Sharma’s translation of GAS ad 1.1 for instance; Gnoli also appears to translate sarvakṣetrainstead of sarvakṣatra-, which is the variant supported by our MSS and by the semantic analysis given by Abhinavagupta. 10 This MS was provided by Pandit Kedāranāth, Durgāprasad’s son. It is possible that this MS is the same as our S5 in Sarada script, found in the BORI and also used in the Tirupati edition by Sankaranarayanan (discussed below), but the editor does not provide any information about the script. 11 It is said to come from Deccan College, Pune. Since variants are rarely reported in the Pansîkar edition, we cannot be certain that our MS B in Devanagarī (from the BORI) is identical with this. See also Schrader (1930, p. 4). 12 He states to have used three Śārada MSS, but also to have made a copy of the first one ‘with the help of many other manuscript copies’, which suggests that his so-called first MS (= Ka?) is the result of conflating several ones. 13 See for instance his conjectural reading samasirṣakatayā for samatayā in the avataraṇikā to 1.1, which is not supported by any evidence in the MSS at our disposal; or the reading anabhāvāt (anabhāva as a word being a hapax) for abhāvāt in 2.16, again without any manuscript support. The latter conjecture is probably due to the omission of a previous negation in his edition. 14 The Sanskrit text given in this translation agrees with Lakshman Joo’s edition, with the addition of typographic errors, which, surprisingly, often agree with those of the e-text (dated 2011) of the Muktabodha Indological Research Institute, also based on Lakshman Joo’s edition. See e.g., in mangala 2, mādhānyataḥ for pradhānyataḥ; in mangala 3, ākānkṣya for ākānkṣayā and prathiḥ for prathitaḥ; GAS ad 1.1: sarvakṣetrāṇām for sarvakṣatrāṇām; BhG 1.9, where he adds (agreeing with the Muktabodha e-text) a ’nāsti’ in the second hemistich, which produces faulty metre; BhG 1.26, where he reads pitṛn atha instead of pitṛn atha; BhG 18.62, where he reads tatprasādāt, which is the text of the Vulgate, Springer Abhinavagupta on the Kashmirian Gītā 16 17 Another great merit of Lakshman Joo’s edition is that he includes the text of the Gītā as probably read by Abhinavagupta." He does not say so, but his readings often (though not always) seem to agree with the Kashmirian recension of the Gītā, and it certainly corresponds on the whole to what Abhinavagupta probably read. Indeed, so much so that occasionally Lakshman Joo even changes the text of the Gītā in accordance with what he thinks Abhinavagupta had in front of him." Unfortunately, Lakshman Joo is silent on how he constitutes the text of the Gītā, which manuscripts he used for it and what editorial principles he adopted. This, however, did not deter the editors of the Mahabharata from using this edition of the Gītā as if it were a manuscript, saying that ’the printed text can reasonably be taken to represent Raina’s best MS’ (Belvalkar 1947, p. lvii). Therefore, the siglum S3 in the Gītā as printed in the critical edition of the Mahābhārata does not represent a manuscript, but Lakshman Joo’s text, which is a methodological problem. Moreover, the siglum Cg represents exactly the same text, for it is meant to stand for what Abhinavagupta read according to Lakshman Joo’s edition.19 18 Although it is evident from the small number of variants in the footnotes that Lakshman Joo did not report all the variants, he did so on a number of occasions, especially when the reading was somewhat problematic.20 Some of these variants are helpful for us in establishing the reading, therefore we have decided to report all Footnote 14 continued whereas the GAS comments on matprasādāt, with all the philosophical implications of that variant; GAS ad 18.62: hiranapotakāḥ (for harinapotakāḥ) and so on. See also kṣetrajñahaḥ (obviously a barbarism) in his translation of GAS ad 13.3, which is but what he must have heard or grasped of his master’s pronunciation of the visarga in kṣetrajñaḥ. This last error shows that Marjanovic must have based his translation on his notes reproducing his master’s (Prof. Shri Narayana Mishra’s) oral teaching. Another proof of this is his translation of pratiṣṭhā in GAS ad 14.27, which he translates as ‘seed’ instead of ‘seat’ (in aham eva hi brahmaṇaḥ pratiṣṭhā), a result of having misheard his master. 15 He reports variants of Ka, Ga and Gha for the text of the Gītā (e.g. p. 32 Ka, p. 33 Ga and once p. 111 Gha), which means that he had at least three manuscripts to constitute the mula. 16 The question of what can be called the Kashmirian recension of the Gītā is discussed below ‘The Kashmirian Gītā and Its Relation to the Vulgate.’ 17 This results in a completely unsupported reading of muḍhācāraḥ for mithyācāraḥ in 3.6d. Here, Abhinavagupta probably summarizes the verse by contracting two adjectives of the Gītā, vimūḍhātmā and mithyācāraḥ, into one mūḍhācāraḥ (the reading of all our MSS). Lakshman Joo, however, seems to assume that Abhinavagupta must have read mūḍhācāraḥ in the text of the Gītā itself, and replaces the received reading of mithyācāraḥ (read by all the Kashmirian sources as well as the Vulgate) by mūḍhācāraḥ in the text of the mula. Later on, the commentary on verse 12 cites this compound once more, and all MSS read mithyācāraḥ in the commentary. This is changed again into mūḍhācāraḥ by Lakshman Joo, this time to be consistent with his previous conjecture. We are grateful to Yuko Yokochi who contributed to solving this problem. 18 Another error of method’ in Belvalkar’s treatment of the Gītā recensions is pointed out by van Buitenen (1965, p. 103); for Belvalkar simply uses T.R. Chintamani’s select extracts to report Bhaskara’s readings. For these problems, see below ‘The Kashmirian Gītā and Its Relation to the Vulgate.’ 19 This results in such readings in the apparatus of the critical edition as mūḍhācāraḥ for mithyācāraḥ in 3.6d (see note 17), which is in fact Lakshman Joo’s conjecture, but appears as supported by S3 and Cg in the critical edition. 20 At least this is our impression, see e.g. in the commentary on 3.15, p. 36, where he rejects the reading of all his available MSS (probably considering a citation from Manu an interpolation); or in the mula at 3.39b, in which he rejects his MS Ga (indriyeșu ha), although this is what Abhinavagupta seems to read (utpattisamaye ’lakṣya indriyeṣu), and adopts indriyaiḥ saha of the Vulgate (agreeing with our S1 and S2). Springer L. Bansat-Boudon, J. Törzsök the variants found in Lakshman Joo’s edition, in addition to report of course the readings Lakshman Joo adopted, both for the commentary and for the Gītā. The third edition was published in 1985, in Tirupati, by Sankaranarayanan at the Sri Venkateswara University. In addition to Lakshman Joo’s edition, the editor also uses a Sārada manuscript (the same as our S5) and a rather corrupt Devanagarī one (our B), both from Pune. As we shall see, the additional manuscript evidence used here consists of two sources that are certainly not among the most useful ones. But more importantly, this edition does not question most of Lakshman Joo’s unmarked conjectures and reproduces them often without any sign to show that they are actually conjectures. In fact, the text of the commentary as given by Sankaranarayanan deviates very rarely from Lakshman Joo’s edition, except for the fact that pratīkas are used particularly generously. This overuse of the pratīkas is not based on manuscript evidence, but is a personal decision of the editor, ‘for the sake of convenience’ (p. liii). Similarly to Lakshman Joo, Sankaranarayanan also seems to interfere with the text of the Gītā when he deems it necessary, without any support from the manuscripts.22 Because of the heavy reliance of this edition on Lakshman Joo’s, and because we have consulted the MSS used in it ourselves, we have decided not to include the variants of this edition, unless they deviate from Lakshman Joo’s.2 From the description of the available editions, the need for a critical edition seems to be rather obvious. We have thus collected all manuscript sources we were aware of, out of which the most important ones seem to be the Saradā manuscripts. Manuscript Sources of the Edition 23 We have identified altogether ten manuscripts containing the Gītārthasamgraha, out of which five (S1, S2, S6, S7 and J) also contain the text of the Bhagavadgītā. All the ten manuscripts consulted are on paper, seven of them are written in Śāradā characters (under the siglum S) and three in Devanagarī. Out of these ten manuscripts, only eight have been fully collated and used to constitute the text. This was done because after collating the introductory verse and the first chapter, it turned out that two of the Devanagarī manuscripts are not particularly useful: one of them (from Jammu, see J2 below) is very corrupt and transmits a large number of 21 See e.g. at avataraṇikā to 1.1 cited above (accepting Lakshman Joo’s samaśīrṣakatayā for samatayā, or rather, correcting it to [or misprinting it as?] samaśīrṣataya); see also in the commentary on 3.12 printing mudhācāraḥ following Lakshman Joo rather than the MSS’ mithyācāraḥ. Neither of these is identified as a conjecture. 22 Sankaranarayanan had no additional evidence for establishing the text of the Gītā, therefore wherever his text reads differently, it is because of his (often silently made) conjectural emendations. They are sometimes very little supported (e.g. in 3.3a conjecturing puraikoktā for purā proktā without any MS evidence), but sometimes appear to be justified on the basis of Abhinavagupta’s text (such as indriyeṣu ha in 3.39b cited in note 20). 23 It does happen occasionally that the editor provides a very good conjecture or insight. At 3.39b cited above (note 20), indriyeșu ha is what Abhinavagupta seems to read, but Lakshman Joo adopts the Vulgate’s indriyaiḥ saha for some unaccountable reason. Sankaranarayanan’s edition adopts, rightly, indriyeșu ha (which is also the reading of Lakshman Joo’s MS Ga), without, however, noting whether it is a conjecture or MS reading. Springer Abhinavagupta on the Kashmirian Gītā obvious errors; the other one (from the BORI, see B below), although it contains fewer minor mistakes, often transmits corrupt or secondary readings (e.g. akṣāder for kṣader in 1.1, pāṇḍavaḥ for pāṇḍavāḥ in 1.1, seṣabuddya for viseṣabuddhyā in 1.35). If the readings of these two manuscripts had been included in the edition, it would have resulted in an inflated apparatus, which is already rather voluminous as it stands, and would have slowed down the collation procedure considerably. The variants in these two MSS for the introductory verse and the first chapter shall be made available in an appendix, which will show that sometimes both manuscripts transmit the same secondary reading. Furthermore, their readings also often agree with those of Pansîkar’s edition, which suggests that the manuscripts used in Pansîkar’s edition may also belong to this relatively corrupt subgroup. In fact our B and Pansîkar’s edition agree so often that initially we thought that Pansîkar used B, but this hypothesis is contradicted by some disagreements of variants. It must also be pointed out that this B is identical with manuscript B of the Tirupati edition published by Sankaranarayanan; one could therefore also consult that edition to see what B reads. However, comparing our collation of the first chapter with that of Sankaranarayanan’s edition, it seems that the latter omits to report a large number of small variants found in B, thus, it cannot be taken to represent B faithfully. Our list of MSS with the abbreviations used is as follows:2

.24 S1 Śārada on paper, Banaras Hindu University acc. no. C 42 (S. no. 3G/3425). 88 folios. Very clear and neat handwriting. Marginal additions by a second hand. Small vertical lines to separate words and avagraha signs to mark various elisions of the a are added in a less thick ink and probably by a second hand (these marks are not reported). Also contains the text of the Gītā, usually according to the Vulgate. A lot of insertions and glosses are seen in the margins, both in the original hand and in a second hand, whose akṣaras are thinner and whose writing is less regular. No date, but the writing was commissioned by a śaiva devotee called Nārāyaṇa. Catalogued in Tripathi (1971, pp. 334-5). S2

Śārada on paper, Banaras Hindu University acc. no. C 1016 (S. no. 3G/3446). 79 folios. Also contains the text of the Gītā. The text of the Gītā, when it runs through several verses, is usually indented. This Gītā follows mainly the Vulgate, but occasionally it is the only MS that has the good Kashmirian Gītā reading (in e.g. 1.10). Dated 1906-7 AD. Catalogued in Tripathi (1971, pp. 336–7). S3

Śārada on paper, Banaras Hindu University acc. no. C 1099 (S. no. 3G/3297). 34 folios. Contains only the commentary. Dated samvat 82. Catalogued in Tripāṭhī (1971, pp. 332-3). S4 = Sārada on paper, Banaras Hindu University acc. no. C 3981. (S. no. 3G/3444) 117 folios, bound as a book. Very clear and neat hand, but several small errors. Contains only the commentary. The MS starts with other texts, such as a certain Bhuvanamālinīkalpavivaraṇa. Lots of marginalia and sometimes very idiosyncratic readings (in e.g. the commentary on 1.1 aparihartavyāni for apahartavyāni). Ends 24 Note that we use digital photos for the collation of these MSS. Springer L. Bansat-Boudon, J. Törzsök abruptly in the middle of the Stavacintamani with commentary, page 126 of KSTS ed. Not dated. Catalogued in Tripathi (1971, pp. 336-7). S5 = Śārada, Bhandarkar Oriental Research Institute, 163/1883-84. (New no. 1517.) The same as MS S in Sankaranarayanan’s Tirupati edition. Not catalogued. For a detailed description, see Sankaranarayanan (1985 part 1, pp. xlix ff.). Contains only the commentary. Several marginal glosses. No date. Many small errors. S6=Śārada, Shrinagar, Oriental Research Library, n. 2308. Bound as a book. Large, thickly written pages. Contains only the commentary with pratīkas till verse 2.11. Verse 2.11, which is missing in the Vulgate, is found in S6, which has its first citation of the Gītā at this point including the preceding speaker indication. It then omits the subsequent verses till 14 and gives the full text of the Gītā from verse 15 onwards. Dated 1896-7. Written by a certain Kṛṣṇadāsa. S7 = Śārada, Shrinagar, Oriental Research Library, no. 1612. This manuscript also contains extracts from Rāmakaṇṭha’s commentary on the Gītā. Includes the text of the Gītā. The Gītā is written in larger characters on each page with a wide margin, and extracts from the commentaries are copied in the margins at the appropriate places. There are numerous omissions in Abhinavagupta’s commentary, the scribe (or a scribe somewhere in the transmission) perhaps made a deliberate choice of copying only certain passages. Given the format, it seems that first the text of the Gītā was written in full, and then the commentaries were added to it, i.e. this Gītā may have come from outside the transmission of the commentary. Whatever the case may be, this Gītā seems to represent the Kashmirian recension and, in addition to being conform to Abhinavagupta’s commentary, it also agrees mostly with Ramakantha’s and Bhaskara’s readings and Schrader’s edition of the Kashmirian Gītā. Incomplete, stops at 16.11c, no date. J = Jammu, Shri Ranbir Sanskrit Research Library Jammu Tawi, N. 209. 64 folios. Devanagari. Contains the text of the Gītā. A lot of obvious mistakes, but often seems to retain original readings going with the Śārada MSS. Seems to transmit the Kashmirian recension of the Gītā on the whole. No date. Catalogued in Stein (1894, p. 195) as Bhagavadgītārthasamgraha under no. 930, said to be written in modern Kashmirian writing (navīnā kāśmīrikī lipiḥ). Manuscripts consulted but not collated for the edition: J2 = Jammu, Shri Ranbir Sanskrit Research Library Jammu Tawi, N. 250. J2 Devanagari. Contains only the commentary. Often agrees with B, both having secondary readings. Incomplete. Stops at commentary on 18.67. No date. B = Bhandarkar Oriental Research Institute MS no. 422/1875-76 (new no. 28), in Devanagari. Same as MS B in Sankaranarayanan’s 1985 edition described in part 1, pp. xlix ff. Contains only the commentary. With many secondary readings, often agreeing also with J2. Dated 1895. Springer Abhinavagupta on the Kashmirian Gītā The Kashmirian Gītā and Its Relation to the Vulgate While preparing the edition of the commentary, it became increasingly apparent that a new edition of the mūla, i.e. the Kashmirian Gītā, was inevitable. Five of our manuscripts also include the text of the Gītā, but in spite of this evidence, it is sometimes very difficult to decide what Abhinavagupta could have read, given the concise nature of the commentary and because the Kashmirian manuscripts also seem to have been influenced by the Vulgate.25 We therefore realized that external sources also needed to be consulted to reproduce Abhinavagupta’s Kashmirian Gītā, which led us to consider various questions concerning the Kashmirian recension and its status as compared to the Vulgate. The textual reconstruction of the Gītā should certainly start with that of the Mahābhārata, since, no matter how we try to establish the different layers of composition of the epic, the Gītā is part of it. For a start, a definition of what the Vulgate consists of should be found. Now, the Vulgate of the epic itself is considered, by scholarly consensus, to be identical with the Poona edition by Kinjawadekar (1929-1936), which includes Nilakantha’s commentary. One could therefore identify the Vulgate of the Bhagavadgītā with what is printed in the Poona edition, which is what Belvalkar (1941, p. 18, note 3) and Biardeau (2002 I, pp. 1820) do. 26 Next to this edition, however, we also have the text of the Gītā that Śankara (and his school, as emphasized by Schrader 1935, p. 148) comments upon. Because of Sankara’s importance in Indian thought, and because he is perhaps the earliest known commentator of the text,27 his mula (Apațe 1936; Gokhale 1950) has come to be seen as the Gītā by several scholars, such as Schrader (1930, p. 18, 1935, p. 148, 1934, p. 352), T.R. Chintamani (1941, p. xx), van Buitenen (1965, p. 104) and Kato (2014, p. 1146). Let us note that Śankara does not comment on the first 57 stanzas, which includes the entire first chapter. It is interesting to observe in this context that Tadpatrikar in his edition of Ramakantha’s commentary on the Kashmirian Gītā gives Sankara’s variants for comparison (śānkarapāṭhaḥ) at each 28 25 The strong influence of the Vulgate even on Kashmirian sources was already pointed out by Schrader (1930, p. 2). Belvalkar (1945) expressed doubts about the existence of a Kashmirian recension and the rather overpowering influence of the Vulgate, maintaining that in fact no real Kashmirian recension had ever existed. We can prove at least part of Schrader’s argument, namely that variants from the Vulgate influenced the copyists: for when we consulted and checked the e-text of Lakshman Joo’s edition prepared at the Muktabodha Institute, it turned out that the copyist of the e-text replaced several times the Kashmirian readings of Lakshman Joo’s edition with those of the Vulgate. Thus, the overpowering influence of the Vulgate can be felt even to this day. E.g. in 2.1c the e-text reads viṣīdantam (Vulgate) for sīdamānam (Kashmirian/Lakshman Joo); 2.12 [= 2.11 Vulgate] reads asocyan anvaśocas tvam (Vulgate) for Lakshman Joo’s Kashmirian asocyan anusocams tvam; 2.55 [= 2.53 Vulgate] reads yadā sthāsyati niścalā (Vulgate) for yadā sthāsyati niścitā (Kashmirian/Lakshman Joo). 26 We take Sankara to be the author of the Gītā commentary attributed to him, although the authenticity of this text has been questioned several times (see e.g. Sarma 1933a, Chintamani 1941, p. xxiv, Mayeda 1965). 27 This depends on the identification and date of Bhāskara, who may have been Sankara’s immediate successor or even his contemporary. For a discussion, see the Appendix. 28 This implies that one cannot speak of the Vulgate for this chapter before Rāmānuja (11th-12th c.), as van Buitenen (1965, p. 103) observes. For more on this question, see Bansat-Boudon (2015, p. 93). Springer L. Bansat-Boudon, J. Törzsök verse, below the Kashmirian reading. He thus also appears to consider Śankara’s text the Vulgate. What may be called the third Vulgate was edited as part of the critical edition of the Mahabharata and reprinted in 1945 separately by Belvalkar. This text corresponds mostly to Sankara’s text except for 14 minor variants and has been considered the Vulgate by some scholars such as Gnoli (1976, pp. 39-40) and even, somewhat surprisingly, by van Buitenen in his translation of the Gītā, published in 1981.29 Since this text has become regarded as the received text of the Gītā, we refer to this edition in our apparatus. Let us remark here, nevertheless, that in this part of the Mahābhārata, the editors did not maintain their general principle of following the Kashmirian recension on the whole. 30 As has been suggested,31 this is perhaps due to the importance of Sankara and his version of the Gītā, which was too wellknown to replace, even if Belvalkar seems to argue otherwise.32 Whatever the reasons were for this choice, here we have, by and large, Śankara’s Gītā again. Next to these three, which are slightly different versions of what we may call the Vulgate of the Gītā, the existence of a distinct Kashmirian recension was noticed early on.33 Otto Schrader’s study (Schrader 1930) of the Kashmirian recension is a very valuable source for us, for he notes all the variants compared to the Vulgate, on the basis of a birch-bark manuscript of Sarada characters (dated 1750) and on the basis of what he manages to reconstruct as the Kashmirian readings from Rāmakantha’s and Abhinavagupta’s commentaries.34 Schrader did remark that the Kashmirian readings appeared to be primary at several places.35 His opinion started a heated debate36 that has flared up even recently, but contrarily to the original debate, without any actual philological arguments and purely on a moral-political basis.3 37 29 See van Buitenen (1981, p. xii): ‘The text reproduced is that presented by S.K. Belvalkar as part of the critical edition of the Mahabharata […]. For the Gītā itself this is virtually the vulgate.’ It is surprising that van Buitenen bases his translation on the Vulgate after showing the numerous problems it contains in his paper of 1965. See below. 30 See van Buitenen (1965, p. 101): ‘In the critical edition of the Bhagavadgītā the remarkable fact emerges that, while the MSS. of the Saradā Kashmir tradition are generally the most authoritative for the earliest text of the epic, including the Bhismaparvan where the Gītā is found, nevertheless they are found to be late and secondary as far as the text of the Gītā is concerned.’ 31 This is not explicitly stated by van Buitenen, who writes (1965, p. 102): ‘The earliest testimony concerning the Gītā is that of Sankara, and the Gītā text adopted by the Editor is really Sankara’s text with but 14 highly insignificant variants.’ 32 As pointed out above, his argument is that no real Kashmirian recension ever existed. 33 Note that Belvalkar (1941, pp. 25–6) does not consider the Kashmirian recension a regional one, but a Śaiva sectarian one. For more on this question, see Bansat-Boudon (2015, pp. 94-5). 34 On the history of this discovery, see Bansat-Boudon (2015, p. 92). 35 For numerous examples, see Schrader (1930, pp. 12ff.). 36 See Schrader (1930) and (1935), Edgerton (1932) and Belvalkar (1939, 1941, 1945). For more on this debate, see Bansat-Boudon (2015, pp. 93–5). 37 See Adluri-Bagchee (2016). The authors do admit that ‘a complete evaluation of Schrader’s claims is only possible by re-examining all his sources and comparing these with the editor’s decisions in the Critical Edition.’ Perhaps because they do not intend to take up such a time-consuming task, they do not discuss any textual passages thoroughly from a philological point of view, and when they mention SpringerAbhinavagupta on the Kashmirian Gītā We do not consider it our task to take sides in this controversy, but, as we shall point out, we do think that at certain places the Kashmirian recension offers what appears to be an earlier version of the text than the Vulgate, without this being the case everywhere. What is certainly important for us is the existence of Schrader’s work, which, in addition to Lakshman Joo’s Gītā, gives us support when considering certain variants to be Kashmirian and helps us to reconstruct Abhinavagupta’s mūla. In addition to Schrader’s invaluable work on the Kashmirian recension, several other editions have proved to be helpful for the reconstruction of Abhinavagupta’s Gītā. Most importantly, the texts of two other early commentators who were Kashmirian or at least used some form of the Kashmirian Gītā must be taken into account. One is Bhaskara’s Bhagavadāśayānusaraṇa (available only for the first nine chapters), which may tentatively be dated to the 9th century. This Bhaskara is certainly identical with the one who also wrote a commentary on the Brahmasutra (Kato 2011), and it seems that his Gītā, although perhaps not fully agreeing with what may be called the Kashmirian recension, was certainly close to it in many important respects (for more discussion, see our Appendix).38 Because of Bhaskara’s relatively early date, and since he was possibly Kashmirian, his testimony is very important when reconstructing the Kashmirian recension. Unlike Abhinavagupta’s, his commentary is quite detailed and fills in the gaps wherever Abhinavagupta’s commentary is lacking or too concise to reconstruct his reading. We have consulted the only edition of the text published by Subhadra Jha (Benares, Varanaseya Sanskrit Vishvavidyalaya, 1965), of which a preliminary study had been provided by van Buitenen in the form of an article dated 1965. Curiously, although van Buitenen examined Bhaskara’s commentary on the Kashmirian Gītā and showed that its readings were better than those of the Vulgate, he adopts an ambiguous position when he translates the BhG (1981). He claims (p. xii) to follow Belvalkar’s critical edition, but to ‘have added occasional variant readings, for which there is better authority’ and to ‘have appended in [his] textual notes a further list of readings based on the early commentary of Bhaskara.‘39 Footnote 37 continued variants, they dismiss them as irrelevant for the reader ‘using the text for self-reflection and selftransformation.’ 38 On the question of the relation of Bhāskara’s Gītā and the Kashmirian Gītā, see van Buitenen (1965, p. 104), which, however, does not solve all the problems. He concludes that ‘for the Bhagavadgītā too the K[ashmirian] tradition carries on a text, however deteriorated here and there, that was authentic and of which we have the earliest record in Bhaskara’s bhāṣya.’ See also our Appendix. 39 In fact, in his textual notes (p. 161) van Buitenen does not give any list of Bhaskara’s readings, but only refers to his 1965 paper ‘A Contribution to the Critical Edition of the Bhagavadgītā’ and adds: ‘I have accepted into the text only those readings that make any difference in the translation.’ However, in reality, van Buitenen retains very few of Bhaskara’s readings even when they would make a difference in the translation. He accepts, for instance, the transposition between Bhisma and Bhima in 1.10, ta eva naḥ sthita yoddhum prāṇāms tyaktvā sudustyajān (for the Vulgate’s ta ime ‘vasthita yuddhe prāṇāms tyaktvā dhanāni ca, on this reading, see below, note 46) in 1.33cd, paradharmodayad api (for the Vulgate’s paradharmo bhayāvahaḥ) in 3.35d, and vā ‘in the sense of iva’ (for the Vulgate’s ca) in 3.39d. It is nevertheless rather puzzling that in his translation, van Buitenen does not retain the very first ‘Kashmirian’ variant (1.1b), which is also read by Bhaskara, namely sarvakṣatrasamāgame (for the Vulgate’s samaveta yuyutsavaḥ), although van Buitenen (1965, p. 103) starts his list of Bhaskara’s variants with this one, observing that it is stylistically superior to the Vulgate’s version. Springer L. Bansat-Boudon, J. Törzsök The other commentary is Ramakantha’s 40 Sarvatobhadra (950-1000), which is unambiguously of Kashmirian origin and is equally indispensable. Schrader was the first to have identified it in the form of a MS (MS 3271, Library of the India Office), dated 1750, written in Kashmirian nāgarī, and copied from an even older Saradā original (Schrader 1930, pp. 1-4, Chintamani 1941, p. xiii); and with the discovery of this commentary, he also launched the debate on the Kashmirian Gītā. We have consulted three editions of this text, published by S.N. Tadpatrikar (Poona 1939), T. R. Chintamani (Madras 1941), and Madhusudan Kaul Sastri (Srinagar 1943), of which Chintamani’s has proved to be the most faithful to Ramakaṇṭha’s commentary as well as to the Kashmirian Gītā (see at 2.5 discussed in note 59.) Ramakantha’s text is available for the whole Gītā, although his commentary is very succinct on the first chapter. His testimony is as valuable for the constitution of the Kashmirian recension as Bhaskara’s. There exists a third published commentary on the Kashmirian Gītā: the Anandavardhini, written by Anandavardhana in Kashmir, in the 17th century. He should not be confused with the famous author of the Dhvanyaloka (9th century). Belvalkar published his commentary in 1941, which made him change his mind concerning the Kashmirian recension of the Gītā: he admits in his introduction that another version of the Gītā did exist, which was different from the Vulgate. He nevertheless sees it as a sectarian, Śaiva version rather than a regional, i.e. Kashmirian one. Since this commentary is much later than Abhinavagupta’s and was certainly more exposed to the influence of the Vulgate, we have decided not to use it for our edition in general, but to consult it only occasionally, to be in a better position to evaluate the transmission. After establishing the external sources we may use in addition to Abhinavagupta’s commentary and the manuscript sources thereof, let us look at the different cases in which an editorial decision must be made as to which reading to accept. We have thus far encountered three different situations. 41

Firstly, when a reading differs from the Vulgate and is supported by Abhinavagupta’s commentary, it evidently needs to be adopted, for our primary aim is to reconstruct Abhinavagupta’s Gītā. In such cases, there are three different ways in which the readings of our MSS may be distributed. 40 (a) In several cases, most or all of our MSS follow Abhinavagupta’s reading, thus fully supporting our choice against the Vulgate.* 42 Rāmakaṇṭha, the author of the Sarvatobhadra, is probably identical with the author of the Spandavivṛti (1913, KSTS 6) and of one of the commentaries on the Stavacintāmaņi, known from citations. See Bansat-Boudon-Tripathi (2011, p. 257, n. 1168). Note that Kaul’s edition (1943) uses three MSS, among which ‘one transcript by Rājānaka Lakshmana Brahmacharin of Srinagar,’ that is Lakshman Joo, first editor of the Gītārthasamgraha, which shows Lakshman Joo’s interest for other commentaries of the Kashmirian Gītā in addition to the GAS. 41 We have decided not to provide a stemma of our manuscripts, neither for the commentary nor for the mūla, for the transmission is clearly cross-contaminated. We shall nevertheless provide some indication as to the relationship of the manuscripts, as far as such a relationship can be determined. 42 One such case is in 2.19b, where we have adopted vināsino ‘prameyasya against the Vulgate’s anāsino ‘prameyasya. Here, the change seems to have been made from vināsino to anasino, in other words, the Kashmirian version seems to be primary. Vināsino describes bodies (nom. plural dehāḥ in the verse), but Springer Abhinavagupta on the Kashmirian Gītā (b) In a number of cases, only some of our MSS agree with what Abhinavagupta reads, most commonly S2 and S7 (usually also followed by J), while S, and S6 reproduce the Vulgate." 43 (c) It also happens that none of our MSS reads what Abhinavagupta does. In these cases, we still have the evidence of Lakshman Joo’s edition and Schrader’s Kashmirian Gītā in most cases, which often confirm that Abhinavagupta’s reading did survive in the Kashmirian transmission at least somewhere. Nevertheless, it can also happen that we decide against the readings of these editions if Abhinavagupta’s commentary clearly reads differently, and opt for Abhinavagupta’s reading." 44 45 The cases (b) and (c) are important in that they show the overwhelming influence of the Vulgate, which must have been strong enough for the scribes to go even against the commentary’s reading they copied.
Secondly, there are variants at places that Abhinavagupta left uncommented, therefore we cannot be sure what he read. Our manuscripts may have various distributions of variants, some agreeing with the Vulgate, others not; and it may even happen that all or almost all agree with the Vulgate. In such cases, we primarily considered the testimony of other commentaries of Kashmirian origin, Footnote 42 continued because of the adjective qualifying the soul in the genitive right next to it (‘prameyasya… i.e. sarīriņaḥ), it must have been felt more appropriate to have another adjective of the soul here too, in the genitive. Only S1’s version before correction agrees with the Vulgate. Note that this variant was not listed by Schrader, although Ramakantha clearly reads vināsino in his commentary (vināsino vinaśvarasvabhāvatvāt) and this is what Chintamani prints in his edition (although Tadpatrikar’s edition retains anāśino without any variant or note). Similarly, in 2.63, we have adopted yattasyapi/yat tasyāpi against the Vulgate’s yatato ‘pi. Abhinavagupta’s gloss is clear, for he interprets both yattasya and yat tasya. Our MSS are unanimous on this reading, and Ramakantha’s commentary also supports it in the editions. The Vulgate may again be secondary, possibly to disambiguate yat(-)tasya. 43 In 2.71d, for instance, we have the very minor variant sā rātriḥ in Abhinavagupta’s commentary and our S2, S7, J as well as in Lakshman Joo’s edition and in Schrader’s text against the Vulgate’s să niśā, followed by our S1 and S6. Note that although both Chintamani and Tadpatrikar print niśā here in their editions, Ramakantha must have read rātriḥ, as he glosses it with nisa. In this case, one could argue both ways to explain the variant. Incidentally, the same distribution of the MSS readings can be observed in 2.12b, which reads prajñavan nābhibhāṣase in the Kashmirian version for the Vulgate’s prajñāvādāmś ca bhāṣase, but here, we have not got Abhinavagupta’s gloss (only Ramakantha’s). For a discussion of these two variants, of which the Kashmirian seems to us certainly primary, see Schrader (1930, pp. 13-4, under II.11). Schrader (1930, p. 13) remarks that Speijer 1902 already noticed the textual problem. For a different opinion, see Adluri-Bagchee (2016, p. 11). On the basis of this common distribution of readings, and because our S, seems to be the closest to the Kashmirian recension, we have opted for S2 and S7 when they went against S1 and S6 and the Vulgate, such as in 2.60d sthiraprajñas tadocyate (supported by Ramakaṇṭha again) for the Vulgate’s tasya prajñā pratisthita. 44 We also assume that Lakshman Joo had some manuscript evidence for what he printed. In 2.63a, for instance, all our five available MSS read tāni sarvāņi samyamya with the Vulgate, but Abhinavagupta clearly reads what is transmitted in Schrader’s and Lakshman Joo’s editions: tāni samyamya manasā; for he glosses it with ya evam manasa indriyāņi niyamayati […]. 45 See the example cited above of 3.39b in notes 20 and 23. Springer L. Bansat-Boudon, J. Törzsök most importantly Ramakaṇṭha’s and Bhaskara’s readings, whenever they were available." 46 47 It must be noted that we looked first of all at the commentaries themselves for support, and not the readings of the editions, which in some cases proved to go against the commentator(s). For a commented word, especially if there is more than a mere gloss, has more chance to survive intact in the transmission of the commentary; and the two commentators mentioned were certainly much closer to Abhinavagupta in time than any of our manuscripts. Sometimes we needed to take into account (and adopt) even simple glosses, although they are more likely to get corrupted, for they provided at least some commentarial support. Similarly, if a variant is supported by other Kashmirian citations or summaries of the Gītā (even if outside the commentarial tradition of the text), we have decided to accept it as belonging to the Kashmirian recension.4 49
Thirdly, there remain a number of instances in which Abhinavagupta is silent, and we have not got any unambiguous support from the other Kashmirian commentators, nor from Kashmirian citations. We have established two guiding principles for these cases. 46 (a) If the reading of the Vulgate appears slightly more correct grammatically or syntactically, we assume that it results from standardization, as is the case in In 1.33d, for instance, three of our four available MSS (S1, S2 and J) agree with the Vulgate in reading prāṇāms tyaktvā dhanāni ca, and this is also what Lakshman Joo adopts. Abhinavagupta does not comment this expression, and we only have one of our MSS (S7) and Schrader’s Gītā that read prāṇāms tyaktva sudustyajan. Unfortunately, Ramakantha has no commentary on this part of the text, but both Chintamani and Tadpatrikar print the latter reading, therefore this is what appears to be transmitted with Ramakaṇṭha’s commentary in MSS having both his commentary and the mūla. Luckily, Bhaskara’s commentary is available and it further confirms that the Kashmirian reading is prānāms tyaktvā sudustyajan, for it says sudustyajan prāṇāms tyaktvā. This example also shows that our S7 is often more faithful to the Kashmirian reading than our other MSS. Let us remark here that the expression prāṇāms tyaktvā sudustyajān or tyaktvā prāṇān sudustyajān is a typical expression of the Mahabharata (5 occurrences in the critical edition). 47 Let us consider the case of 2.6d. The Vulgate as well as our S1 and S2 have te ‘vasthitaḥ pramukhe dhārtarāṣṭrāḥ, against which our S7 and J, together with Lakshman Joo and Schrader, have te naḥ sthitaḥ pramukhe dhārtarāṣṭrāḥ. Neither Abhinavagupta, nor Ramakaṇtha glosses the word, although the editions of Ramakantha’s commentary read naḥ. Looking at Bhaskara’s text, the edition reads te ‘vasthitaḥ, going with the Vulgate. However, Bhaskara gives the following gloss: ta evāsmākam pramukhe sammukhe sthita iti-which shows with the word asmākam that he certainly read naḥ in the text. Therefore, we have adopted naḥ in our edition. 48 In 1.47c, we only have our S, and Schrader’s text that read utsṛjya, against the rest of our MSS (S1, S2 and J), Lakshman Joo and the Vulgate, which give visrjya. If we look at the other Kashmirian commentaries, Ramakantha gives no gloss (although the editions give utsṛjya), but Bhaskara has dhanur utsrjya in his commentary, supporting the variant utsṛjya, albeit without giving a synonym. 49 We have not yet met particular cases; but works that we intend to consider include for instance the Mokṣopāya, Kṣemendra’s Bhāratamañjarī (see Schrader 1935, p. 147), Yogaraja’s (11th c.) commentary on the Paramarthasara of Abhinavagupta, and Jayaratha’s (13th c.) commentary on the Tantraloka. It is worth noting that Yogaraja appears to quote the Vulgate text; yet, out of this, no definitive conclusion can be drawn, for Yogaraja’s Gītā citations may well have been assimilated to the Vulgate in the transmission. More than a century later, Jayaratha still cites only the Kashmirian recension. On this point, see Schrader (1930, p. 2, n. 1 and 1935, pp. 147-8), Bansat-Boudon (2015, p. 95). We also intend to take into account the Parimala ad Mahārthamañjarī, which quotes the Kashmirian text (see ad 70). Springer Abhinavagupta on the Kashmirian Gītā a number of unambiguous examples given below. Therefore, we accept the reading that is potentially perceived as less correct, for the direction of change from less to more correct forms is the general rule in the transmission of epic and purāņic sources. (b) If it is not possible to establish a clear difference of grammar or syntax, then we assume that the reading that does not agree with the Vulgate is more likely to be the Kashmirian reading, therefore we accept it as such. This principle of going for the ‘deviant’ reading is based on the following two observations: First, it is more likely that such ‘deviant’ readings are primary in the Kashmirian recension, simply because it is much more likely that the influential Vulgate became dominant against such readings (and contaminated the Kashmirian transmission) than that ‘deviant’ readings arose against the widely accepted Vulgate. This principle was also used in the reconstruction of the Mokṣopāya, whose editors chose to prefer readings that were not those of the widely circulating Yogavāsiṣṭha." 50 Second, readings not agreeing with the Vulgate have been confirmed as Kashmirian by the other Kashmirian commentators on several occasions.51 We have concluded from this, too, that it is much more likely in general that the ‘deviant’ version is the Kashmirian one.52 Therefore, in such ambiguous cases we consistently adopt all deviations from the Vulgate as Kashmirian and thus apply an ’extremist’ point of view whereby we reconstruct, in the end, the most deviant Kashmirian version as compared to the Vulgate.53 Some Interesting Examples of Abhinavagupta’s Interpretation = Abhinavagupta provides us with a rather unusual interpretation of the text from the very beginning. He tells us that some people understand the place name ‘Kurukṣetra’ to mean the field of the sense organs (kuru karaṇa) and this is where the battle takes place, i.e. in the body. Although he first presents this interpretation as if it belonged to others (kecit), he then continues his own 50 See Krause-Stinner in Mokṣopāya vol.1, pp. xxviii ff. and Hanneder in Mokṣopāya vol. 2. p. ix. See, for instance, the examples cited in parts 1 and 2 in this section. 51 52 In 2.9c, all our manuscript sources and the Vulgate read na yotsya iti, against Lakshman Joo’s na yotsyāmīti. Schrader does not mention a variant here, from which we must conclude that his text agreed with the Vulgate. Looking at the other commentaries, Chintamani prints yotsye, while Tadpatrikar gives yotsyāmi for Ramakantha’s mula, but Ramakantha does not say anything at this point. Bhaskara’s mūla has yotsyāmi printed and here we seem to have a confirmation from the commentary, which also has yotsyāmi. However, at this point, the commentary has a variant, which is yotsye-thus, no conclusion can be drawn from it. In this case, we have more support for the variant of the Vulgate, at least in numbers. Our policy is nevertheless to choose what differs from the Vulgate, i.e. we adopt na yotsyāmīti. 53 In this, our conclusion and method agree in principle with Schrader’s way of reconstructing the Kashmirian recension, for he also opts for an ’extremist’ position by going for the ‘deviant’ readings as Kashmirian. Our main aim is, however, to reconstruct Abhinavagupta’s mūla, and for this reason, our choices need not agree with his, nor with those of editors who edited other Kashmirian commentaries. Springer L. Bansat-Boudon, J. Törzsök understanding in the same vein, explaining that the battle is fought between the forces of knowledge, represented by the Pandavas, and ignorance or mistaken thought constructs (samkalpāḥ), embodied by the Kauravas. Similarly, dharmakṣetre is understood to mean ‘in field of the [supreme] dharma’, which, according to Abhinavagupta, is seeing one’s self truly; and the locative sarvakṣatrasamāgame (which is the Kashmirian reading for the pāda in which the Vulgate reads samaveta yuyutsavaḥ) denotes for him the conflict of all internal forces such as passion and lack of passion, anger and patience etc. After this interpretation, which places the whole battle inside the self, it does not come as a surprise in 3.11 that he glosses the word ‘gods’ (devāḥ) by ‘goddesses of the senses’ (karaṇeśvaryaḥ), remarking that they are well-known in secret teachings (rahasyaśāstraprasiddhāḥ, referring to Krama teachings). Propitiating deities (devān bhāvaya-) thus denotes, according to this understanding, making the goddesses of the senses devour the objects of the senses (viṣayan bhakṣaya-). If one performs their propitiation, those deities shall assimilate one to the Self (tadātmasādbhāvana) and one shall obtain final release. This idea of liberation is further elaborated in the commentary on verses 3.14-15. These verses of the Gītā describe the way in which creatures are nourished by food (anna), which is produced thanks to rain (parjanya), rain being secured by sacrifice (vajña), which in turn is offered in ritual (karman). Ritual then has its origin in Brahman/Vedic injunction, which comes from the Imperishable (akṣara). This passage is again given an esoteric interpretation in Abhinavagupta’s commentary. For him, food stands for the objects of enjoyment or of experience (bhogya), rain is the enjoyer or subject of experience (bhoktr) and sacrifice is the act of enjoyment (bhogakriya). Ritual act (karman) in turn means the autonomy of the power of action (kriyāśaktisvātantrya), and this autonomy derives from the supreme Self (Brahman) that manifests itself out of its free will. The supreme Self is an active entity, which comes from the Eternal One (akṣara), i.e. from pure, undisturbed (praśānta) consciousness (samvit). In Abhinavagupta’s understanding, these six elements enumerated form a six-spoked wheel, which revolves and may lead, depending on its construction, to the maintenance of mundane existence or to the attainment of final release. At the end of this chapter, Abhinavagupta proposes another secret teaching about the last verses. These verses (3.47-8) mention the Atman as the ultimate entity beyond intellect (buddhi); but Abhinavagupta suggests that for those who know the secret teachings (rahasyavid) it is not Atman but the ultimate (para) ‘I’ (ahamkāra) that is referred to here, which is an affirmation of the identity of the self and the universe in the form of ‘I am everything’ (sarvam aham). The term ahamkara here does not refer to the limited and limiting ahamkāra of the Samkhya, the ego, which belongs to the bound soul (pasu). Rather this paro ‘hamkāraḥ stands for the pure, absolute aham, ‘I’, from which the central Saiva notion of ahantā is constructed. With such awareness of non-duality, it is not possible to be split up and experience feelings such as anger, which is based on duality. Thus, it is this ‘I’ of nondual 54 54 buddher yaḥ paratra vartate paro ‘hamkāraḥ sarvam aham ity abhedātmā sa khalu paramo ‘bhedaḥ. Springer Abhinavagupta on the Kashmirian Gītā consciousness that one must fully grasp, in order to annihilate anger that comes from ignorance with it." 55 Some Interesting Cases of Variants in the Kashmirian Recension of the Gītā Compared to the Vulgate In what follows, two kinds of variants shall be discussed. First, those that affect the meaning of a verse, second those that do not change the meaning as compared to the Vulgate, but which show some form of language variation. Belvalkar (1945) argued that the variations found in the so-called Kashmirian recension are too trifle to consider it an actual recension and that no major variations can be found in the different recensions or rather, versions, of the Gītā. Although we agree that these versions do not constitute fundamentally different texts, the differences are important enough to speak of recensions. In some places the readings are different enough so that readers, whether specialists or not, may want to know about their existence. As our example of verse 2.5 shows below, it is not the same whether it is Arjuna who is said to desire wealth or his elders. Variants Affecting the Meaning In the following three examples, the Kashmirian recension appears to yield better sense in our opinion. Moreover, it seems that in each case, the change from the Kashmirian version to the Vulgate is easier to explain than the other way round. Verse 1.10 The first one was already discussed in detail by van Buitenen (1965, pp. 99ff.) in the context of Bhaskara’s commentary:" Kashmirian recension 56 aparyāptam tad asmākam balam bhīmābhiraksitam | paryāptam tv idam eteṣām balam bhiṣmābhirakṣitam || Vulgate aparyāptam tad asmākam balam bhīṣmābhirakṣitam | paryāptam tv idam eteṣām balam bhīmābhirakṣitam || 57 As van Buitenen also argues, the Kashmirian version yields good meaning (NB it is Duryodhana who speaks): ‘That army protected by Bhima is not large for us; but this army [of ours] protected by Bhīṣma is large for them.’ 55 For some other examples of a Śaiva interpretation, see Bansat-Boudon (2015, pp. 98–101). 56 Let us remark that Abhinavagupta’s commentary also supports this reading. 57 Van Buitenen (1965) argues for Bhaskara’s version, which he considers different from and earlier than the Kashmirian recension (see Appendix); in any case these versions or recensions agree on the reading of this verse. Springer L. Bansat-Boudon, J. Törzsök The Vulgate, which swaps Bhīma and Bhīṣma, requires the readers to interpret aparyāpta and paryāpta in an unnatural way: ‘Our army protected by Bhīṣma is not limited in number; but this army of theirs, protected by Bhima, is limited in number.’ For paryāpta normally means ‘abundant, copious, full’; but here, if we take the Vulgate, it must be interpreted to mean ’limited in number’ in order to yield the required meaning. Indeed, it seems to us that the meaning of paryāpta recorded in dictionaries in the sense of ’limited in number’ is only to be found at this place, in other words, this meaning was invented only to explain this passage. In addition to the fact that the Kashmirian reading yields better meaning, the direction of change is easy to explain from the Kashmirian recension to the Vulgate. For asmākam and eteṣām were probably wrongly understood in the possessive sense (‘our army/their army’ rather than ‘for us/for them’), which triggered the exchange of the two names. Verse 2.5 Kashmirian recension: gurun ahatvā hi mahānubhāvāñ chreyas cartum bhaikṣyam apīha loke | na tv arthakāmas gurun nihatya bhuñjiya bhogan rudhirapradigdhān || For it is better to live on alms in this world without killing one’s noble elders than to kill one’s elders due to one’s desire for wealth in such a way that I would enjoy pleasures tainted with blood. Vulgate: gurun na hatvā hi mahānubhāvāñ chreyaś bhoktum bhaikṣyam apiha loke | hatvārthakāmāms tu gurun ihaiva bhunjiya bhogan rudhirapradigdhān || For it is better to live on alms in this world without killing one’s noble elders. Having killed my elders who desire wealth, I would enjoy pleasures tainted with blood. The first line does not create any difference in meaning, although it may be remarked that the Kashmirian recension has a more idiomatic version (bhaikṣyam car-). The second line is also more idiomatic in that śreyas or varam commonly stands with na (tu), but there is an asymmetrical construction there in that instead of a second infinitive, the text has an optative of the first person singular. Indeed, it may be this asymmetrical syntax that triggered the change in the Vulgate. The real difference in meaning, however, lies in that the Kashmirian variant in the nominative singular, arthakāmas, attributes the desire for wealth to the subject (Arjuna/ the general subject), while the Vulgate makes the elders desire wealth (arthakāmāms tu). The latter version seems less satisfactory, for the question is not so much whether one is to kill greedy elders, but rather whether one is to kill one’s elders for wealth, a question that is also referred to in 1.35. The Kashmirian reading is indirectly supported by Abhinavagupta’s commentary, for he identifies this part of the verse as pointing out a plan for a particular 58 Van Buitenen (1965) shows that commentators also struggled to find an appropriate meaning here. Oddly, Schrader (1930, p. 23) does not point out this variant. Springer Abhinavagupta on the Kashmirian Gītā result (phalaviseṣānusamdhāna), which would be wealth here. The nominative singular is clearly supported by Ramakantha’s commentary, who glosses the word in question with dhanalipsuḥ ‘wishing to obtain wealth’ in the nominative singular." Bhaskara’s gloss also agrees with the Kashmirian reading, paraphrasing arthakāmas with dhanābhilāṣī san. 59 Verse 3.35cd Kashmirian recension: svadharme nidhanam śreyaḥ paradharmodayad api || Vulgate: svadharme nidhanam śreyaḥ paradharmo bhayāvahaḥ || The Kashmirian recension has a comparative with an appropriate complement in the ablative: ‘death according to one’s own dharma is better even than prosperity according to another person’s dharma.’ The Vulgate lacks this ablative and stresses more the disastrous consequences if one follows someone else’s dharma: ‘death according to one’s own dharma is better, another person’s dharma brings disaster." The Kashmirian recension has, once again, a variant that reads smoother with a comparative complete with the ablative. It also brings out the contrast deathprosperity nidhana-udaya. The Vulgate gives the impression of trying to insist more on how bad it is to follow someone else’s dharma, for it excludes the possibility of prosperity in that case and states more categorically that following another’s dharma can only be wrong. This normative and ethical intention, namely to reject more categorically the moral fault of observing another’s dharma, seems to be the reason for which the Kashmirian version could have been changed into the Vulgate’s reading. Slightly Irregular Forms and/or Usage As pointed out above, the Kashmirian recension seems to use slightly incorrect forms or syntax at several places. These all seem to be primary compared to the readings of the Vulgate, the direction of change being more likely to go towards the (hyper-)correction of such irregularities. In particular, the Kashmirian recension uses the middle form of verbs that commonly exist only in the active, most notably, the middle participle sīdamānain 1.28ab, 2.1c, 2.10d, which are always hypercorrected in the Vulgate to vişīdant-. Similarly, in 3.36c the Kashmirian recension uses the middle participle anicchamāna (anicchamāno ‘pi balād ākramyeva niyojitaḥ), which is duly corrected in the Vulgate to anicchan (anicchann api vārṣṇeya balad iva niyojitaḥ). In the famous sentence (2.3), which in the Vulgate reads with the injunctive klaibyam mā sma gamaḥ pārtha (‘do not be unmanly, O Pārtha’), the Kashmirian recension uses the 59 Note that this reading is given by Chintamani, who rejects the variants dharmalipsuḥ and arthalipsuḥ, while Tadpatrikar retains the Vulgate’s arthakāmāms in the Gītā and dharmalipsuḥ in the commentary, rejecting dhanalipsuḥ. In our opinion, this is among the examples that show that Chintamani’s edition is more reliable. Springer L. Bansat-Boudon, J. Törzsök plain imperative (mā klaibyam gaccha kaunteya), which is of common usage in more popular genres such as the epic (for the epic usage, see Oberlies 2003, p. 185 and references), but is frowned upon by grammarians. Although these slight irregularities do not affect the meaning, they are interesting to note, for they may give us a somewhat different picture of the composition of the text. In all these instances, both concerning change of meaning and grammatical differences, our conclusion is that the Kashmirian recension represents an earlier version of the text compared to the Vulgate. This, however, does not mean that we can affirm the priority of the Kashmirian recension. At several places, the Kashmirian recension includes extra verses not found in the Vulgate.60 Additional verses usually attest chronologically later accretion, but the evaluation of the additional verses in the Kashmirian recension is not straightforward. For this reason, and because we have not yet examined all the variants thoroughly, we cannot say anything definitive at this point about the relationship of the two recensions. Conclusion We hope that the descriptions and examples given above have shown convincingly that Abhinavagupta’s commentary as well as his Kashmirian Gītā merit a critical edition. We believe that the mula and the commentary are crucial for our understanding of the history of Indian religious thought and that they cannot be fully grasped if they are not reconstructed on a sound philological basis. Although collating all the sources and weighing the evidence carefully in every case may not always appear labour efficient enough, we hope that the result of our work will be useful both for scholars working on Saivism and for those who study the textual history of the Bhagavadgītā and the ideas it conveys. 61 Appendix: Bhaskara on the Kashmirian Gītā: A New Hypothesis Bhaskara’s Testimony Bhaskara is a very important witness for our edition, no matter how his exact identity and date are determined, for, as shown by his readings, he comments on the Kashmirian Gītā or on a version of what can be identified as the Kashmirian Gītā, and gives a detailed commentary which allows us to reconstruct his mūla. It is generally taken for granted that he was a Vedantin who also composed a Brahmasutrabhāṣya [BSBh]62 and that therefore he must have preceded 60 For instance in chapter 2 between verses 10 and 11 and between verses 48 and 49 of the Vulgate. (These are verses 11 and 50 in our edition of the Kashmirian recension.) 61 For some notable studies, see Malinar (2007) and Ježić (1979, 1986, 2009). 62 That the Vedantin Bhaskara, who belongs to the jñānakarmasamuccayavāda school comments on the Kashmirian Gītā is taken for granted by all scholars having worked in one way or the other on the Bhagavadāśayānusaraṇa, although van Buitenen has a rather unique opinion on the status of Bhaskara’s Gītā, as we shall see. See also Ingalls (1967). SpringerAbhinavagupta on the Kashmirian Gītā Abhinavagupta, who may even refer to him in his GAS as a previous commentator of the Gītā (GAS on 18.2).63 64 Indeed, it is obvious from the edition of Bhaskara’s Bhagavadgītābhāṣya [GBh] also entitled Bhagavadāśayānusarana (Subhadra Jha 1965, see also van Buitenen 1965, which is a preliminary text published as an article, and Raghavan 1968) that it has Kashmirian variants from the very first verse, in which Bhaskara reads, with Abhinavagupta and Ramakantha, sarvakṣatrasamāgame for the Vulgate’s samavetā yuyutsavaḥ (1.1b). Note also Bhaskara’s reading of 6.7 (after Jayatīrtha’s testimony, since the verse is missing in the MSS), which again agrees with that of Ramakantha, Abhinavagupta and Anandavardhana. Moreover, he also includes the ‘additional’ verse starting with tvam mānuṣenopahatāntarātmā after Vulgate 2.10, characteristic of the Kashmirian recension. Van Buitenen (1965), Subhadra Jha (1965), Raghavan (1968) and Kato (2011, 2014) have dealt with Bhaskara’s so-called Kashmirian version, which appears (see, notably, van Buitenen’s study, 1965) to be an ’extremist’ version of the Kashmirian Gītā,65 and, most probably, the most ancient evidence of its existence. Which Bhaskara? Yet, the very identity of this Bhaskara has been much debated, especially because the question of Bhaskara’s identity is tightly linked with the question of the Gītā’s two recensions, the Kashmirian Gītā and the Vulgate, as we shall see. The rediscovery of Bhaskara’s GBh and its tentative reconstruction from two fragmented MSS were a milestone in this debate. As pointed out by previous scholars,66 the name Bhaskara appears three times in the tradition as a commentator of the Gītā, mainly in a Vedantin context, with the notable exception of Abhinavagupta’s GAS. In the GAS, Bhaskara is merely mentioned as a previous eminent (tatrabhavant) exegete," whereas in Vedantin literature his name occurs in discussions of arguments proper to the Vedanta. 67 One of the two Vedantin references is found in Vedantadeśika’s (1268-1370) commentary on the Gītā, the Tatparyacandrika, which discusses several interpretations of Bhaskara’s GBh from the perspective of the Visiṣṭādvaita. These include 63 GAS ad 18.2: atra cadhyaye yad avasiṣṭam vaktavyam asti tat praktanair eva tatrabhavadbhaṭṭabhaskarādibhir vitatya vimṛṣṭam iti, ‘Whatever remains to be said on this chapter has been discussed at length by previous [exegetes] such as the revered Bhaṭṭabhaskara, etc.’ We shall discuss the question below. 64 Van Buitenen (1965, pp. 106-109) discusses the variants in the first chapter and lists them from the second chapter to the fifth. Although he gives the variant sarvakṣatrasamāgame (1.1b), he does not retain it in his 1981 translation, see above, n. 39. 65 ‘Extremist’ for it differs on most points from the Vulgate whenever one of the Kashmirian versions differs. However, Bhaskara’s mula does not record all the ‘additional’ verses read in Ramakantha’s and Abhinavagupta’s mūlas and also ‘shows differences from accepted Kashmirian readings’ (Raghavan 1968, p. 282, n. 10). 66 See Schrader (1934, p. 350, n. 8), Sarma (1933), Raghavan (1968, p. 282), and Kato (2011, p. vii). See also Sarma 1933, p. 670. 67 Springer L. Bansat-Boudon, J. Törzsök his understanding of BhG 18.64 (66 in editions of Vedāntadeśika’s text), from which we may infer that Bhāskara’s GBh was complete." 68 The second one occurs in Jayatīrtha’s work (14th century), whose Prameyadipikā, dvaita subcommentary on Madhva’s Gītābhāṣya, criticizes Bhaskara’s interpretations and even refutes Bhaskara’s reading of 6.7b (parātmasu samā matiḥ) as a ‘conjecture’ (kṛtrimaḥ pāṭhaḥ). To do so, he uses two main arguments: 1. that Bhaskara rejects the traditional reading (sampradāyagatam pāṭham visṛjya), which has paramātmā samāhitaḥ (that of the Vulgate commented upon by Madhva), and 2. that Bhāskara’s reading creates redundancy (punarukti) with verses 9 and 32 of the chapter.69 A kṛtrimaḥ pathaḥ for which, unfortunately, there is no direct evidence from Bhāskara’s GBh lacunary MSS, but which has proved, nevertheless, to be a Kashmirian reading: that of Ramakantha (all editions) and that of Abhinavagupta (all editions). For here, Abhinavagupta does not only quote the pratīka as he usually does, but the first three pādas, ‘as though to ensure the correct reading of pada 2,’ as Schrader (1930, p. 16) observes, before he rightly concludes: ‘There seems to have been an early dispute on this sloka." ,70 The discovery of Bhaskara’s GBh confirmed the existence of a Bhaskara, author of a commentary on the Gītā, who is attested by the Vedantin tradition (and perhaps by Abhinavagupta, a question which we examine below), and who is also the author of a BSBh. The discovery was made by V. Raghavan, who first found a fragmentary MS in the Banaras Sarasvati Bhavan in the forties, which preserved chapter 1 to chapter 7, although with a few lacunae. In 1953-4, the same scholar found another fragment in Sārada in the Wellcome Library, London, which ran from ch. 4 to ch. 9, ‘with a gap in chapter 6 coinciding more or less with the one in this chapter in the Banares ms.’ (Raghavan 1968, pp. 281-2). This rediscovered but incomplete text was then published in Subhadra Jha (1965), by collating both fragmentary MSS.71 71 It should be noted, however, that Bhaskara’s commentary on the Gītā had already been known to Chintamani (1941, p. xxviii),72 the editor of Ramakantha’s commentary on the Kashmirian Gītā. In his edition, he speaks of a ‘fragmentary copy,’ breaking off after chapter 7 verse 16, which is undoubtedly the same incomplete manuscript that was found by Raghavan (1968, p. 281) in the Sarasvati 68 See Sarma (1933, pp. 669-70). 69 Six passages of the Prameyadīpikā refer polemically to Bhaskara’s GBh (prastāva, 2.54, 6.7, 3.4, 3.42, 2.47; see Schrader 1934, p. 350); on Bhaskara’s reading of 6.7b, see Schrader (1930, p. 16 and 1934, p. 350, n. 8), Sarma (1933, pp. 672-73), Raghavan (1968, p. 282) and Kato (2011, p. vii). 70 Even Anandavardhana’s commentary confirms this Kashmirian reading, despite the somewhat arbitrary decision of Belvalkar’s edition to retain paramātmasamā matiḥ. For Belvalkar himself notes that parātmasu samā matiḥ is the reading of his two MSS for the Kashmirian Gītā and of all the Kashmirian commentaries, including that of Anandavardhana. 71 On this rediscovery, see notably Raghavan 1968, pp. 281-2. For details on this edition, see Kato 2011, pp. vi-vii. 72 Despite Raghavan’s assertion (1968, p. 281): ‘Bhaskara’s Bhasya on the third Vedanta Prasthāna, the Bhagavadgītā, although found in the Banares fragment, had not become known among scholars.’ Springer Abhinavagupta on the Kashmirian Gītā 73 Bhavan, Banares, and was later used by Subhadra Jha along with the newly discovered MS of the Wellcome library, London. Chintamani (xliii-lxi) even establishes a comparative chart of the Kashmirian readings, which includes Bhaskara’s readings up to the place where the MS breaks off." He also compares parallel passages of Bhaskara’s and Sankara’s commentaries on the Gītā (xxviiixxx). This comparison is later completed by Raghavan’s examination of the edited text (1968, pp. 283-292), which shows the way in which Bhaskara’s and Sankara’s commentaries interact. From the references Vedāntadeśika and Jayatīrtha make to Bhaskara’s GBh, it was natural to infer that Bhaskara was most probably a jñānakarmasamuccayavādin (see Sarma 1933, p. 669; Chintamani 1941, pp. xxvii, xxx). The discovery of the actual text, albeit lacunary, confirmed the hypothesis. Nevertheless, there remains a riddle: Abhinavagupta also mentions in the GAS (ad 18.2, see above) a previous eminent exegete, named Bhaṭṭabhaskara. Who is he? Is he Bhaskara, the Vedantin, or another Bhaskara, the Kashmirian Śaiva, son of Divakara (hence his other name of Divakaravatsa), well-known as the author of the Śivasūtravārttika (published in the KSTS 4), of the lost Kakṣyāstotra (a hymn to the Goddess only known from citations), and of the lost Vivekāñjana (also known by a citation in the Isvarapratyabhijñāvimarśinī, vol. I, p. 10, saying: yad āha bhaṭṭadivākaravatso vivekāñjane). 74 At a time when nothing was known of the fragmentary MSS of Bhaskara’s GBh, Schrader (1935, pp. 351-2) asserted that the Bhaskara Abhinavagupta speaks of must undoubtedly be a Kashmirian. Schrader’s demonstration is based on four main arguments: 1. Abhinavagupta cannot have referred, especially with such reverence, to a scholar who was not Kashmirian, to an ‘outsider;‘75 2. the title ‘Bhatta,’ which Abhinavagupta adds to Bhaskara’s name,76 is hereditary in Kallata’s line of pupils (and nowhere is the Vedantin Bhaskara thus designated); 3. Abhinavagupta cannot have explicitly referred to a Vedantin, since he clearly makes the point in the avataraṇikā of 1.1 that the Gītā is not meant to teach jñānakarmasamuccaya;77 4. moreover, it is impossible that Abhinavagupta ‘recommended a commentator whose 73 Editing the Anandavardhini, Belvalkar (1941) also gives Bhaskara’s Kashmirian readings, probably from Chintamani’s table. 74 The title Kakṣyāstotra is referred to in Iśvarapratyabhijñāvivṛtivimarśinī, vol. 2, pp. 301, 328, and vol. 3, p. 388 and in Yogaraja’s commentary ad Paramarthasara 51. 75 A point which could be corroborated by a traditional legendary narrative, recorded by the advaitin Anandanubhava, which shows Kashmirian pandits to be rather intolerant to outsiders from the South (Raghavan 1968, pp. 282-3). 76 This honorific is confirmed in various sources such as in Bhaskara’s Śivasūtravārttika, in which his name is repeatedly prefixed by bhaṭṭa. See the colophon to the first section: śrīmadbhaṭṭadivākarātmajaśrīmadācārya-bhaṭṭabhaskaraviracite sivasūtravārtike [sic] […] (KSTS 4: 28). Abhinavagupta also refers to him as Bhaṭṭadivākaravatsa (see Īsvarapratyabhijñāvivṛtivimarśinī vol. 2, pp. 13, 14, 145) as well as Yogaraja ad Paramarthasara 35. 77 Against Lakshman Joo’s na tu jñānakarmaṇī samaśīrṣakatayā [unmarked conj.] samucciyete ity atra tātparyam, Schrader reads here, as we do in our edition in progress: na tu jñānakarmaṇī samatayā samucciyete ity atra tātparyam (see above, n. 13). We have nevertheless a different understanding of the sentence in the translation of ch. 1, on which Lyne Bansat-Boudon is working. Springer L. Bansat-Boudon, J. Törzsök authorities (Brahmasūtra and Vedas) ignore, if not reject, his own sources of inspiration (Śivasūtra and Āgamas).’ Indeed, as suggested by some, Abhinavagupta could well have vehemently criticized Bhaskara’s views at some places such as ad 9.33-5 (Sankaranarayanan 1985 part 2, p. 203, n. 40 and 41, Gnoli 1976, p. 29, n. 23, and pp. 180-1, n. 14), although without naming him,78 yet it does not follow that the Bhaskara to whom Abhinavagupta refers as tatrabhavant in GAS ad 18.2 (and whom he recommends to his readers for a more detailed exegesis than his own) is the same as the Bhāskara whose views he denounces. Therefore, we believe that Schrader’s arguments are valid, and remain so even after the rediscovery of the GBh by Bhaskara, the Vedantin. Later on, Chintamani (1941, p. xxvii) was apparently of the same opinion, although less explicitly, on the mere basis of Bhaskara’s being referred to by Abhinavagupta as a commentator of the Gītā." 79 Thus, there may have been two exegetes of the BhG with the name Bhaskara: the Vedantin, whose text has been rediscovered, and who is also referred to in some Vedantin witnesses and the Kashmirian Śaiva, of whom the only evidence would be Abhinavagupta’s assertion, the difficulty being that not only is his Gītābhāṣya lost, but we do not even have any mention of its title. 80 The one with whom we are concerned in the process of editing and translating the Kashmirian Gītā is the Vedāntin Bhaskara, who, we believe, is not the Bhaskara referred to by Abhinavagupta in the context of the Gītā.81 Nevertheless, it is not impossible that Abhinavagupta knew (see above, n. 78) of the Vedantin Bhaskara and of his commentary on the Gītā. If this were the case, it could speak in favour of the Kashmirian origin of Bhaskara, as we will propose below. Indeed, it could explain that, as a man from Kashmir, Abhinavagupta, although disagreeing with Bhaskara, could have referred to his commentary on the Gītā. At the same time, in Abhinavagupta’s GAS no trace is found of a criticism of Sankara’s Gītā interpretation, even covertly, thus Abhinavagupta does not appear to be aware of Sankara’s commentary. In any case, it hardly needs to be demonstrated that Bhaskara the Vedantin is also the author of the well-known BSBh. First, it is expected of a Vedantin to comment 78 According to this view, Abhinavagupta’s detailed refutation is often directed against what seems to echo the very words of Bhaskara’s GBh as it has come down to us (see Sankaranarayanan 1985 part 2, p. 203, n. 40 and 41). 79 Chintamani (1941, p. xxvii): ‘We know of the Kashmirian author Bhāskara […]. To this Bhaskara and his commentary on the Gītā, reference is made by Abhinavagupta.’ Raghavan (1968, p. 283, and n. 9) is therefore wrong when attributing the identification of the two Bhaskaras to Chintamani: “The Bhaskara cited by Abhinavagupta here [ad BhG 18.2] had been taken as our Bhāskara [the Vedāntin] by some other scholars too.’ Note that Kato (2011, p. xxvi) eludes the difficulty, since he appears to consider silently the mention of Bhaskara by Abhinavagupta to be among the proofs of a Vedantin Bhaskara. 80 After Schrader, Chintamani and Raghavan (1968, p. 282). 81 As opposed to van Buitenen’s assertion (1965, p. 105): ‘Of the Bhaṭṭabhaskara mentioned by Abhinavagupta we know only that he had commented on the Gītā, but unless we assume that there were two Bhaskaras who commented on the Gītā we may safely conclude that Jayatirtha’s Bhaskara and Abhinavagupta’s were the same person.’ Springer Abhinavagupta on the Kashmirian Gītā 82 upon the three prasthānas of the Vedanta, namely the BS, the BhG and the Upanisads, just as Sankara does. Now in addition to Bhaskara’s BSBh discovered in 1915 (even though the text is poorly edited, see Kato 2011, p. x),” an internal reference by Bhaskara points to the existence of a commentary on the Upanisads (or at least on the Chandogyopaniṣad, see van Buitenen (1961, pp. 269ff.) and Kato 2011, pp. v-vi) written by him (BSBh ad 3.1.8). It is therefore not unreasonable to assume that Bhaskara also commented upon the Gītā, following Sankara’s example. Furthermore, it is beyond doubt that Bhaskara’s two commentaries support the same doctrine belonging to the jñānakarmasamuccaya school, and that both give their author an opportunity to criticize Śankara’s interpretations. Finally, as pointed out by Kato, the BSBh itself offers an additional clue: BhG 18. 61ab, as quoted in the BSBh 1.2.6, has the Kashmirian reading which is attested in Ramakantha’s and Abhinavagupta’s commentaries on the Gītā, and which can be considered to be also the reading of Bhaskara, in spite of the absence of the actual passage in the MSS of his GBh (Kato 2011, p. xxviii [with a typo: xiii for xvii] and 2014, p. 1148).83 Date Since Bhaskara criticizes Sankara’s interpretations of both the Gītā and the BS, it is reasonable to consider Bhaskara posterior to Sankara but anterior to Vacaspatimiśra (second half of the 9th century), whose Bhāmatī on 1.1.4 and elsewhere criticizes the views of Bhaskara’s BSBh (see Schrader 1935, p. 348, van Buitenen 1961, pp. 268-273 and 1965, p. 105 and Raghavan 1968, p. 292).84 Alternatively, one may assume that the debate is reciprocal between the two Vedantin thinkers, and that Bhaskara could have been more or less contemporary with Sankara.85 In that case, Bhaskara’s mūla would be the earliest evidence (c. 9th century) of the Kashmirian Gītā, Rāmakaṇṭha (or Rājānaka Rāma or Rāmakavi) being tentatively dated in the second half of the tenth century (ca. 950-1000 CE, see Sanderson 2007, p. 411), Abhinavagupta at the turn of the tenth and eleventh centuries (c. 975-1025 CE, see Sanderson 2007, p. 303) and Anandavardhana (not to be confused with the author of the Dhvanyaloka) in the 17th century. 86 82 Since then van Buitenen worked on a preliminary text of a revised edition, which Kato 2011 completed. 83 The BSBh quoting BhG 18.61ab reads (against the Vulgate: isvaraḥ sarvabhūtānām hṛddese ‘rjuna tisthati): isvaraḥ sarvabhūtānām hrdy eșa vasate ‘rjuna, which is the Kashmirian variant given by both Abhinavagupta and Ramakaṇṭha (see Kato 2014, p. 1148). 84 He would also be anterior to Abhinavagupta, if we accept that Abhinavagupta silently refutes Bhaskara’s interpretations at some places. 85 See Sarma (1933, pp. 666, 668) refuted by Schrader (1935, pp. 349-350), van Buitenen (1961, pp. 268ff.) refuted by Ingalls (1967, p. 61, n. 2). See also van Buitenen (1965, p. 105, n. 26). For a summary of the question, see Kato 2011, pp. xxiv-xxv. 86 He claims himself to be the direct pupil of Utpaladeva (Sanderson 2007, p. 352). Springer L. Bansat-Boudon, J. Törzsök Provenance: Is Bhaskara from the South? Bhaskara’s provenance seems to be at least as controversial as his date. One may wonder why he should hail from the South, as assumed by Raghavan and others such as van Buitenen (1965, p. 105)87 and Gnoli (1976, p. 29 and n. 23). One of the main arguments for his Southern origin comes from Anandānubhava’s assertion, himself from the South (Kanchipuram), according to which Bhaskara is from Karnataka (Raghavan 1968, pp. 282-3). Now, although Bhaskara is a Vedāntin and knows Sankara’s works (the two commentaries on the BS and the BhG), this is not enough to conclude that he is of Southern origin, as is Sankara; for, in that case, it remains unexplained why Bhaskara retains Kashmirian readings for his GBh.88 Moreover, Abhinavagupta seems to criticize exegetic views that resemble Bhaskara’s (see above). If those views can be shown to be Bhaskara’s own (rather than more general views of samuccayavādins), then this could be an additional argument for Bhaskara’s Kashmirian provenance. Is Bhaskara’s Gītā Kashmirian? Before the discovery of Bhaskara’s GBh, Schrader (1930, p. 16, 1934, p. 350, n. 5) deduced from the testimony of Bhaskara’s reading of BhG 6.7 (as quoted and discussed by Jayatīrtha at the same verse) that Bhaskara may have had before him the Vulgate of the Gītā with only one or two Kashmirian readings. Jayatīrtha himself, unaware of the existence of any Kashmirian Gītā, takes a Kashmirian reading for a conjecture (kṛtrimaḥ pāṭhaḥ), as mentioned above. Nevertheless Schrader’s conclusion and Jayatīrtha’s hypothesis have been refuted by the discovery of Bhaskara’s GBh MSS and their examination: the Gītā which Bhāskara comments upon is indeed the Kashmirian Gītā, or in any case a Gītā with readings that are mainly Kashmirian or agree with those of the other commentators of the Kashmirian recension. In his edition of Ramakantha’s commentary, Chintamani (1941, pp. xxxi ff.), who had an incomplete MS of Bhaskara’s commentary before him, already showed that the GBh by Bhaskara mostly conforms to the Kashmirian version, at least as far as the first seven chapters are concerned. This was further shown in detail by van Buitenen (1965), although in a rather twisted way, as we shall see. 87 He does so on the basis of a somewhat surprising argument: ‘Elsewhere I have raised the question whether Sankara and Bhaskara were not contemporaries-there is a bit of evidence to suggest that, in which case his home would be likely to be South-India.’ Here van Buitenen refers (105, n. 26) to his 1961 paper (note the typo: he dates his paper to 1962). 88 As Chintamani (1941, p. xxxi) lucidly puts it about Bhaskara, author of the GBh: ‘The most interesting feature of this commentary is that it follows the Kashmirian Recension in most of the places. […] Is this Bhaskara identical with the Kashmirian author or is he the same as the Vedantin? If he is the Kashmirian, he shows himself to be aware of the works of Śankara. If he is not, how did Kashmirian readings find a place in his commentary? [our emphasis]’ Springer Abhinavagupta on the Kashmirian Gītā For, in order to reconciliate, we suspect, Bhaskara’s alleged Southern origin and the Kashmirian character of his Gītā, van Buitenen proposes an ingenious yet contrived thesis. He takes it for granted that the author of the GBh is a Vedāntin from the South, who lived in the 9th century, and that he commented on a Gītā different from Sankara’s Vulgate but of equal authority, while also having Sankara’s Vulgate before him. This hypothesis was already made by Schrader, although in a very limited context, that is on the sole basis of Jayatīrtha’s remark, since Bhaskara’s GBh was still unknown (1930, p. 16, n. 1 and 1934, p. 350, n. 5; see above). It was then also confirmed by Raghavan (1968, pp. 283ff.), through a meticulous examination of parallel passages in Sankara’s and Bhaskara’s commentaries. In the context, van Buitenen’s conclusion (1965, pp. 104-5) is quite audacious and surprising: Bhaskara’s Gītā would have thus been the ‘prototype’ of the Kashmirian Gītā, in other words, van Buitenen considers it to predate the so-called Kashmirian Gītā: […] Bh.[āskara] had, besides Sankara’s Vulgate, another text which must be considered a Vorlage of what now survives as the Kashmir version. It is, however, not identical with it, an important fact not recognized by Belvalkar. […]. We do not know whether Bhaskara was a Kashmirian. We do know he had a text different from, sometimes superior to, the Vulgate of Sankara and prior to, and consistently superior to, a version of the Gītā now known from Kashmir sources. […] I believe that the conclusion is unavoidable that in the ninth century there existed a text of the Bhagavadgītā which had equal authority with that used by Sankara; that it existed outside Kashmir; and that it is the prototype of the so-called Kashmir version. The consequence of this conclusion is that the Kashmir version is late and secondary not to the Vulgate, but to Bhaskara’s text.’ One can see that van Buitenen’s implicit conclusion is that there were three versions of the Gītā: the mula commented upon by Sankara and the mula commented upon by Bhaskara (both authors, therefore both texts, from the South), as well as a third one, the so-called Kashmirian recension, whose prototype had been Bhaskara’s Gītā. In that case, a number of queries may arise:
How come the version of the Gītā van Buitenen identifies as a ‘Vorlage’ or a ‘prototype’ of the later Kashmirian recension was known only from that unique example of Bhaskara’s Gītā? Why did it disappear from the South and elsewhere?
Where does the Kashmirian Gītā come from exactly? For it follows from van Buitenen’s argument that Bhaskara’s Gītā travelled to Kashmir in one way or another, and that it was adopted there and transmitted as a Kashmirian version up to Anandavardhana. This is particularly improbable in the Kashmirian context, which is known to be rather conservative and resistant to external intrusion. Springer L. Bansat-Boudon, J. Törzsök 89 Furthermore, such a sequence of events would imply that the Kashmirian Gītā is a version without much regional peculiarity, which would be rather difficult to maintain, since it came to be the current version among Kashmirian commentators of the Gītā, and since (with the notable exception of Jayatīrtha in 14th century) it has not been referred to by non-Kashmirian authors. In other words, why should one take the Kashmirian version of the Gītā away from Kashmir, unless one tries to take Bhaskara away from Kashmir on the mere assumption that he was from the South?90 We propose the following hypothesis: since Bhaskara has the Kashmirian readings on the whole and since one of the two MSS of his GBh is in Śāradā, he may well have come from Kashmir, whether he was Kashmirian by origin or not. At the same time, if we assume he comes from Kashmir, how could he have known Sankara’s Vulgate, of which no other Kashmirian commentator of the Gītā appears to be aware (Schrader 1930, pp. 7-8)? There may be at least two possibilities:
either Bhaskara remained in Kashmir and Sankara’s Vulgate somehow reached him there,
or Bhaskara came from Kashmir (hence his attachment to the Kashmirian readings), but may have travelled to the South or at least somewhere considerably south of Kashmir. This may be attested by the geographical origin of some of the BSBh MSS (see Kato 2011, and below). He could have become acquainted there with both Sankara’s GBh and BSBh, therefore he had both the Kashmirian Gītā as his main text before him and Sankara’s GBh with its own mula (hence his quotations or anuvada from Sankara, see Raghavan 1968, pp. 283ff.). Then Bhaskara himself or his text may have gone back to Kashmir. Our hypothesis is almost identical with van Buitenen’s scenario, according to which Bhaskara had two texts before him, the so-called prototype of the Kashmirian Gītā and Sankara’s Vulgate. Nevertheless, in our hypothesis there is no need to postulate Bhāskara’s Gītā as the prototype of a third, Kashmirian one. Bhāskara had the Kashmirian Gītā in one particular form and Sankara’s Vulgate before him. We thus consider the Kashmirian character of Bhaskara’s Gītā to be original. Note that the hypothesis that he travelled to the South, or at least somewhere considerably south of Kashmir, could be corroborated by the fact that he is wellknown in the South, as witnessed by Jayatirtha and Anandanubhava, and that for his BSBh, there are MSS from every part of India. While van Buitenen mainly used BSBH MSS in Southern scripts (perhaps because he assumed that it was a text from the South), Kato (2011, pp. xiii-xviii) discovered one MS from Bengal and one from the North (nāgarī). The existence of Bhaskara’s BSBh MSS from the North may throw a different light on the question of Bhaskara’s origin, even if no firm conclusion can be drawn from the mere distribution of manuscripts. 89 This is, notably, Belvalkar’s thesis (1941), see below. 90 This is indeed a long-lasting supposition that most scholars who studied the question share and maintain as a fact, namely that Bhaskara was not a Kashmirian (for a summary, see Kato 2011, pp. xxvixxvii). Springer Abhinavagupta on the Kashmirian Gītā In any case, it is easier to imagine such a scenario rather than that the whole of Kashmir came to adopt a Gītā from the South (as implied by van Buitenen’s thesis), only to replace it later by the Vulgate. It is obvious that Bhaskara knew Sankara’s Vulgate of the Gītā. Nevertheless, due to much uncertainty about the author and his text, the question raised by Schrader (1930, p. 8 and 1934, p. 352), namely whether the Vulgate was known or not in Kashmir by the time of Abhinavagupta, remains unanswered.“1 Here again, several assumptions are possible. Either Bhaskara’s Gītā commentary criticizing Sankara and his Vulgate never came back to Kashmir, or it came back to Kashmir but for some reason Abhinavagupta did not refer to Sankara’s Vulgate thus criticized. It may have been because he did not accept this line of transmission or because it was not yet wide-spread or well-known (and let us not forget that Abhinavagupta wrote his GAS when he was still something of a fresher, see Sanderson 2007, p. 359) or perhaps because he did not intend to argue with it at length in a short commentary (samgraha). Yet, it can also be assumed that, although Bhaskara’s Gītā and his commentary must have come back to Kashmir (after all, we have a MS of his GBh from Kashmir), his criticism of Śankara and Sankara’s variants were simply not known or wide-spread in Abhinavagupta’s time and even thereafter. It must have taken some time for the Vulgate to become the dominant version of the Gītā in India in general, and perhaps even more time to usurp the place of the Kashmirian Gītā in Kashmir. It follows that, if Bhaskara, author of the GBh entitled Bhagavadāśayānusaraṇa in its colophon, is identified as a Kashmirian Vedāntin, he is not Bhāskara, the Kashmirian Śaiva, author of the Sivasūtravārttika, the Vivekāñjana and the Kakṣyāstotra. Thus, provided our hypothesis is admitted, Bhaskara would represent a unique example of a pandit being acquainted with both traditions of transmission of the Gītā. Although the Kashmirian tradition seems to have had a fairly independent life, Bhaskara would have provided a rare example of a point of contact. Acknowledgements We are extremely grateful to a number of people who all helped us in one way or another to have access to or copies of manuscripts, in particular to Prof. Bettina Bäumer, Prof. K.D. Tripathi and his very helpful student Dr Narendra Dutt Tiwari, to his Highness the Mahārāja of Jammu 91 According to Schrader (1930, p. 2 and 1935, p. 352), there was no sign of acquaintance with the Vulgate in Kashmir before the decline of the Pratyabhijñā school. Schrader’s position is contested by Chintamani 1941, pp. xxii-xxiii, saying that Abhinavagupta usually adopts Rāmakantha’s readings but sometimes also agrees with the Vulgate. Chintamani then states: ‘So the vulgate must have been current in Kashmir in the days of Abhinavagupta.’ He subsequently adds two arguments: 1. the traditional accounts of Sankara’s travels to the North: ‘for the tradition is persistent that Sankara travelled up to Nepal. And one of the traditions refers to the visit of Sankara to Kashmir itself. However much we may discountenance these traditions, there is nothing inherently against the spread of the works and fame of Sankara in Kashmir’ (xxii-xxiv); 2. the fact that some passages of Ramakantha’s commentary seem to refer to Sankara’s interpretations of the Gītā (xxv-xxvii). Finally, Chintamani concludes (xxviii): ‘It will be therefore difficult to regard Ramakantha as unacquainted with Sankara and authors of his school of thought.’ The first argument is based on a legend, which does not need to be taken for a fact. Now the question of Ramakantha’s alleged references to Sankara is difficult to confirm or refute without examining them in great detail, which would be beyond the scope of this study. Let us note, however, that Tadpatrikar (1939, p. 5), who edited Rāmakaṇṭha’s commentary, did not find any explicit mention of the two Gītā recensions, which suggests that Sankara and his commentary were not known in Kashmir. Springer L. Bansat-Boudon, J. Törzsök and Kashmir Dr Karan Singh and his secretary Mr S.K. Vohra, to Prof. Françoise Nalini Delvoye and to Prof. Ganesh U. Thite. We would also like to thank the various institutions whose manuscripts we have been able to use, namely the Banaras Hindu University, the Bhandarkar Oriental Research Institute in Pune, the Oriental Research Library of Shrinagar and the Shri Ranbir Sanskrit Research Library, Jammu, as well as the Indira Gandhi National Centre for the Arts. Special thanks are due to all the librarians in these institutions for their invaluable help. We are also indebted to Prof. Jürgen Hanneder, Marburg, and to Ms Nicole Merkel-Hilf, South Asia Institute, Heidelberg, for their help with bibliographical data. Abbreviations and Bibliography Authors and works BhG: Bhagavadgītā GBh: Bhagavadgītābhāṣya BS: Brahmasutra BSBh: Brahmasutrabhāṣya GAS: Gītārthasamgraha Institutions, journals and series ABORI: Annals of the BORI ALS: Adyar Library Series ASS: Anandasrama Sanskrit Series BORI: Bhandarkar Oriental Research Institute CSS: Chowkhamba Sanskrit Series EFEO: École française d’Extrême-Orient HOS: Harvard Oriental Series IFP: Institut français de Pondichery IHQ: Indian Historical Quarterly JAOS: Journal of the American Oriental Society JRAS: Journal of the Royal Asiatic Society} [of Great Britain and Ireland, Londres KSTS: Kashmir Series of Texts and Studies TSS: Trivandrum Sanskrit Series WZKS: Wiener Zeitschrift für die Kunde Südasiens A. Sources (Texts and translations) Bhagavadgītā [BhG] Text Vulgate editions
Poona edition (Kinjawadekar, with Nilakantha’s commentary): Kinjawadekar, Ramachandrashastri (1929-1936): Shriman Mahâbhâratam, with Bhârata Bhawadeepa par Nilakantha, 8 vols. Poona, Citrashala Press.
Sankara’s mula: • · Apate, V.G. (1936): Śrīmadbhagavadgītā, ānandagirikṛtaṭīkāsamvali-taśamkarabhāṣyasametā. Benares, Anandāśrama Press (ASS 34); Gokhale, D.V. (1950): Śrīmadbhagavagītā Śrīśankarabhagavatpādā-cāryaviracitena bhāṣyeṇa sahitā. Poona, Oriental Book Agency (Poona Oriental Series 1).
Critical edition: Belvalkar, S.K. (1945): The Bhagavadgītā, being reprint of relevant parts of Bhismaparvan from BOR Institute’s edition of the Mahābhārata. For the first time critically edited by SpringerAbhinavagupta on the Kashmirian Gītā Shripad Krishna Belvalkar. Poona, BORI [Extract from the MBh C. Ed, Poona, BORI, 1933-1972 by Sukthankar, Belvalkar & Vaidya].
Belvalkar, S.K. (1947): [Critical edition of the] Bhismaparvan [of the Mahabharata]. Mahābhārata vol. VII. Poona, BORI. Kashmirian Gītā editions
Belvalkar, S.K. (1941): Śrīmad-Bhagavad-Gita with the Jñānakarmasamuccaya’ Commentary of Ananda[vardhana]. Edited from an unique Śāradā Ms. Poona, Bilvakuñja Publishing House.
Chintamani, T.R. (1941): Śrīmadbhagavadgītā with Sarvatobhadra of Rājānaka Rāmakaṇṭha. Madras, University of Madras (Madras University Sanskrit Series 14).
Jha, S. [Subhadropādhyāya] (1965): Bhagavadgītā Bhāṣya by Bhāskarācārya [Śrīmadbhagavadgītāyāḥ : bhagavadāśayānusaraṇābhidhānabhāṣyam bhagavadbhaskaraviracitam]. Benares, Varanaseya Sanskrit Vishvavidyalaya (Sarasvati Bhavana Granthamālā 94).
Kaul Shastri, M. (1943): The Bhagavadgītā with the Commentary called Sarvatobhadra by Rājānaka Ramakantha. Srinagar, Research Department, Jammu & Kashmir Government (KSTS 64).
See Lakshman Raina [Lakshman Joo] (1933) at GAS below.
Schrader, F.O. (1930): The Kashmir Recension of the Bhagavadgītā. Stuttgart, Kohlhammer (Contributions to Indian Philology and History of Religion 3).
Tāḍapatrīkara, S.N. [Tatpatrikar] (1939): Rājānakarāmakavikṛta-sarvatobhadrakhyaṭīkāyutā kāśmir apāṭhānusāriņī śrīmadbhagavadgītā. Poona, Anandāśrama Press (ASS 112). Text & Translation (Vulgate) van Buitenen, J.A.B. (1981): The Bhagavadgītā in the Mahabharata: a bilingual edition. Chicago, Chicago University Press. [Claims (p. xii) to follow Belvalkar’s critical edition and to have added ‘occasional variants readings’ and ‘appended a list of readings based on the early commentary of Bhaskara’ (although, vB only refers to his 1965 paper: ‘A Contribution to the Critical Edition of the Bhagavadgītā’, on p. 161)]. Gītārthasamgraha [GAS] Text [BhG+ GAS]
Lakshman Raina [Lakshman Joo] (1933): Śrīmad Bhagavad Gita with Commentary by Mahāmaheśvara Rājānaka Abhinavagupta. Srinagar, Kashmir Pratap Steam Press.
Pansikar, W.L. Shastri (1912): The Bhagavadgītā with the Sankarabhāṣya comm. and the subcomm. of Anandagiri, the Nilakanthi comm., the Bhāṣyotkarṣadīpikā comm. of Dhanapati, the Śrīdhari comm., the Gitarthasangraha comm. of Abhinavagupta, and the Guḍārthadīpikā comm. of Madhusudana with the Guḍārthatattvāloka subcomm. of Dharmadatta (Bacchāśarman). Bombay, Nirnaya Sagar Press. [The Sanskrit text of the Gītā proves to be that of the Vulgate.
Sankaranarayanan, S. (1985): Śrīmadbhagavadgītā with Gītārthasangraha of Abhinavagupta. Tirupati, Śrī Venkateswara University, Oriental Research Institute (S.V. University Oriental Series 11). Pt 1: Text. [Follows mainly Lakshman Joo’s edition]. Translation (BhG + GAS)
Gnoli, R. (1976): Il Canto del Beato (Bhagavadgītā). Turin, Unione Tipografico-Editrice Torinese. [From both the Introductory Note and the translation, it is not clear what Sanskrit text Gnoli follows for the GAS: that of Pansikar or that of Lakshman Joo? Probably that of Pansikar] Springer L. Bansat-Boudon, J. Törzsök
Marjanovic, B. (2002): Abhinavagupta’s Commentary on the Bhagavad Gita. Gītārtha-samgraha. Translated from Sanskrit with Introduction and Notes by Boris Marjanovic. Delhi, Indica.
Sankaranarayanan, S. (1985): Śrīmadbhagavadgītā with Gitarthasangraha of Abhinavagupta. Tirupati, Śrī Venkateswara University, Oriental Research Institute (S.V. University Oriental Series 11). Pt 2: Translation. Translation (GAS only) Sharma, A. (1983): Gītārthasamgraha, Translated with an Introductory Study. Leiden, Brill. [Seems to follow Pansikar’s edition]. Other Sources Bhāmati Shastri, J.L. (1980): Brahmasutra-Šankarabhāṣyam, with the Commentaries Bhāṣyaratnaprabha of Govindānanda, Bhamati of Vacaspatimiśra, Nyayanirnaya of Anandagiri. Delhi, Motilal Banarsidass. Bhāratamañjarī Śivadatta, P.-Parab, K.P. (1898): The Bhāratamañjarī of Kṣemendra. Bombay, Nirṇaya-Sâgara Press. Brahmasūtrabhāṣya (Śārīrakamīmāṇsābhāṣya) by Bhāskara [BSBh]. Dvivedin, V.P. (1915): Brahmasutra with a Commentary by Bhaskarāchārya. Varanasi, Chowkhamba Sanskrit Series Office (CSS 20). Kato, T. (2011): The First Two Chapters of Bhaskara’s Sārīrakamīmāmsābhāṣya: Critically Edited with an Introduction, Notes and an Appendix. Online Publication, ULB Sachsen-Anhalt, 2011, UNR: http:/ nbn-resolving.de/urn:nbn:de:gbv:3:4-9304 (last access February 2015). Isvarapratyabhijñāvimarsinī Subramanya Iyer, K.A.-Pandey, K.C. [Dwivedi, R.C. (gen. ed.)] (1986): Isvara-Pratyabhijñā-Vimarsinī of Abhinavagupta: Doctrine of Divine Recognition. Vol. 1 and 2: Sanskrit text with the Commentary Bhaskari (Subramanya Iyer, K.A.Pandey, K.C.); vol. 3: English Translation (Pandey, K.C.). Delhi, Motilal Banarsidass. [1st ed. Allahabad, Lucknow, 1930-1954 (The Princess of Wales Saraswati Bhavana Texts 70, 83, 84)]. Isvarapratyabhijñāvivṛtivimarsinī Shastri, M. Kaul (1938-1943): Isvarapratyabhijñā Vivṛtivimarsini by Abhinavagupta. 3 vols. Srinagar, Research Department, Jammu & Kashmir Government (KSTS 60, 62, 65). Mahārthamañjarī Ganapati Śāstrī, T. (1919): The Maharthamanjari with the Commentary Parimala of Mahesvarananda. Trivandrum, Government of His Highness the Maharajah of Travancore (TSS 66). Moksopāya Anonymus Casmiriensis Mokṣopāya. Textedition Teil 1 Das Erste und Zweite Buch Vairagyaprakaraṇa, Mumukṣuvyavahāraprakaraṇa. Kritische Edition von Suzanne Krause-Stinner. Wiesbaden: Harrassowitz Verlag. 2011. Anonymus Casmiriensis Mokṣopāya. Textedition Teil 2 Das Dritte Buch Utpattiprakaraṇa. Kritische Edition von Jürgen Hanneder, Peter Stephan und Stanislav Jager. Wiesbaden: Harrassowitz Verlag. 2011.n. Paramarthasara Bansat-Boudon, L.-Tripathi, K.D. (2011): An Introduction to Tantric Philosophy: The Paramarthasara of Abhinavagupta with the Commentary of Yogarāja. [translated by Lyne Bansat-Boudon and Kamaleshadatta Tripathi. Introduction, notes, critically revised Sanskrit text, appendix, indices by Lyne Bansat-Boudon]. Londres/New York, Routledge (Routledge Studies in Tantric Traditions 3). Prameyadīpikā Sadhale, S.J.S. (1935-38): The Bhagavad-Gita with eleven Commentaries. First Collection. 3 vols. (vol. 1: chap. 1-6; vol. 2: chap. 7-12; vol. 3: chap. 13-18). Bombay, Gujarati Printing Press [2nd ed.]. Śivasūtravārttika Chatterji, J.C. (1916): Shiva Sūtra Värttika by Bhaskara along with the Shiva-Sutra-Vṛtti by Kṣema Raja [an extract from Śivasūtravimarśinī (cf. Preface)], and the Spanda Kārikās with the vritti by Kallata. Srinagar, Research Department, Kashmir State (KSTS 4-5). Spandakārikävivṛti Springer Abhinavagupta on the Kashmirian Gītā Chatterji, J.C. (1913): The Spanda Kārikās with the Vivriti of Ramakantha. Srinagar, Research Department, Jammu & Kashmir State (KSTS 6). Stavacintamani Rāma Shāstrī, M. (1918): The Stava-Chintamani of Bhaṭṭa Nārāyaṇa, with Commentary by Kshemarāja. Srinagar, Research Department, Jammu & Kashmir State (KSTS 10). Tantraloka Kaul Shastrī, M. (1918-1938): The Tantraloka of Abhinava Gupta with Commentary by Rajānaka Jayaratha. 12 vols. Srinagar, Research Department, Jammu & Kashmir State (KSTS 23, 28, 29, 30, 35, 36, 41, 47, 52, 57, 58, 59). Tātparyacandrikā See Sadhale, S.J.S. (1935-38). B. Studies Adluri, V.-Bagchee, J. (2016): ‘Who’s Zoomin’ Who? Bhagavadgītā Recensions in India and Germany’, International Journal of Dharma Studies, 4,4: 4-41. Bansat-Boudon, L. (2015): ‘Le Gitärthasamgraha: une lecture sivaïte de la Bhagavadgītā (2010-2014)’, Annuaire de l’École pratique des hautes études, Section des sciences religieuses. Résumé des conférences et travaux (2013-2014). t. 122. Paris, EPHE Ve, 2015: 89–104. Belvalkar, S.K. (1939): ‘The so-called Kashmir Recension of the Bhagavad-Gītā’, New Indian Antiquary, 2, 4: 211-251. Biardeau, M. (2002): Le Mahabharata. Un récit fondateur du brahmanisme et son interprétation. Paris, Seuil, 2002, 2 vols. van Buitenen, J.A.B. (1961): ‘The Relative dates of Śamkara and Bhaskara’, Adyar Library Bulletin, 25: 268-273 [reprint in L. Rocher (ed.), Studies in Indian Literature and Philosophy. Collected Articles of J.A.B. van Buitenen. Delhi, Motilal Banarsidass, 1988: 187-190]. van Buitenen, J.A.B. (1965): ‘A Contribution to the Critical Edition of the Bhagavadgītā’, JAOS, 85,1: 99-109. Chatterji, J.C. (1914): Kashmir Shaivaism (being a brief Introduction to the History, Literature and Doctrines of the Advaita Shaiva Philosophy of Kashmir, Specifically called Trika System). Srinagar, Research Department, Jammu & Kashmir State (KSTS 2). Edgerton, F. (1932): ‘The Kashmir Recension of the Bhagavad Gītā by F. Otto Schrader’, JAOS, 52,1: 68-75. Ingalls, D.H.H. (1967): ‘Bhāskara the Vedantin’, Philosophy East and West, 17, 1-4: 61–67. Ježić, M. (1979): ‘The First Yoga Layer in the Bhagavadgita’. In Sinha, J.P. (ed.) Ludwik Sternbach Felicitation Volume, Part One. Lucknow, Akhila Bharatiya Sanskrit Parisad, 1979: 545-557. Ježić, M. (1986): ‘Textual Layers of the Bhagavadgītā as Traces of Indian Cultural History’. In Morgenroth, W. (ed.), Sanskrit and World Culture: Proceedings of the Fourth World Sanskrit Conference (Schriften zur Geschichte und Kultur des Alten Orients 18). Berlin, Akademie-Verlag, 1986: 628-638. Ježić, M. (2009): ‘The Triṣṭubh Hymn in the Bhagavadgītā’. In Koskikallio, P. (ed.) [gen. ed.: Ježić, M.], Parallels and Comparisons. Proceedings of the Fourth Dubrovnik International Conference on the Sanskrit Epics and Purāņas, September 2005. Zagreb, Croatian Academy of Sciences and Arts, 2009: 31-66. Kato, T. (2010): ‘Bhaskara’s Brahmasūtrabhāṣya–An Unpublished Edition by J.A.B van Buitenen’, WZKS, 52-3: 295-305. Kato, T. (2014): ‘A Note on the Kashmirian Recension of the Bhagavadgītā: Gītā Passages Quoted in Bhaskara’s Gītābhāṣya and Brahmasutrabhāṣya’, Journal of Indian and Buddhist Studies, 62,3: 1144-1150. Malinar, A. (2007): The Bhagavadgītā: Doctrines and Contexts. Cambridge, Cambridge University Press. Mayeda, S. (1965): ‘The Authenticity of the Bhagavadgītābhāṣya Ascribed to Sankara’, WZKS, 9: 155

Oberlies, Th. (2003): A Grammar of Epic Sanskrit. Berlin/New York: De Gruyter. Springer L. Bansat-Boudon, J. Törzsök Raghavan, V. (1967): ‘The Upadeśasahasrī of Śankarācārya and the Mutual Chronology of Sankarācārya and Bhaskara”, WZKS, 11: 137-139. Raghavan, V. (1968): ‘Bhaskara’s Gītābhāṣya’, WZKS, 12-13: 281-294. Sanderson, A. (2007): ‘The Śaiva exegesis of Kashmir’. In D. Goodall-A. Padoux (eds), Mélanges tantriques à la mémoire d’Hélène Brunner. Pondicherry, IFP-EFEO (Indologie 106): 231-442. Sarma, B.N.K. (1933): ‘Bhaskara: a forgotten commentator on the Gītā’, IHQ, 9: 663–677. Sarma, B.N.K. (1933a): Śamkara’s Authorship of the Gītā-Bhāṣya’, ABORI, 15: 39–60. Schrader, F.O. (1933): ‘Rezensionen der Bhagavadgītā’. In Stein, O.-Gampert, W. (eds): Festschrift Moriz Winternitz. Leipzig, Otto Harrassowitz: 40-51. Schrader, F.O. (1934): ‘Ancient Gītā commentaries’, IHQ, 10.2: 348-357. Schrader, F.O. (1935): ‘On the Form of the Bhagavadgītā Contained in the Kashmirian Mahābhārata’, JRAS, 1: 146-149. Stein, M.A. (1894): Catalogue of Sanskrit manuscripts in the Raghunatha temple Library of the Maharaja of Jammu and Kashmir. Bombay: Nirnaya-Sagara Press/ London: Luzac & Co./ Leipzig: O. Harrassowitz. Tripathi, R.S. (1971): Descriptive Catalogue of Sanskrit manuscripts in Gaekwada Library, Bharat Kalā Bhavana Library and Samskrit Mahā-Vidyalaya Library, Banaras Hindu University. Banaras: Banaras Hindu University. Springer