John Brockington and the Sanskrit Epics

John Brockington and the Sanskrit Epics: Limits of Statistical Approaches John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Vishwa Adluri and Joydeep Bagchee Thus, the unfacts, did we possess them, are too imprecisely few to warrant our certitude, the evidencegivers by legpoll too untrustworthily irreperible where his adjugers are semmingly freak threes but his judicandees plainly minus twos. —James Joyce, Finnegans Wake Introduction This paper is a revised and expanded version of a paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8) in September, 2017.1 It incorporates several suggestions and criticisms the participants made.2 We were fortunate John Brockington could participate, as he provided valuable clarification of his notion of “Venn diagrams” as an alternative to the stemma codicum. At the outset we thank him for offering to share a draft version of the conference paper he is currently working on. The occasion for this paper was the discrepant reception of the Mahābhārata critical edition, ranging from complete rejection to the suggestion that, despite “the impossibility of constructing a stemma codicum on the classical model” for the Mahābhārata, the critical edition is “not only on the whole the best we have […] but also permit[s] further advances in our understanding of [the Mahābhārata].”3 These different views necessitate a reassessment of the edition, especially in light of recent calls for a “revised” critical edition.4 As one of the most prolific writers on the critical editions of the 1 Vishwa Adluri and Joydeep Bagchee, “Contra Lachmann: Criticisms (Brockington, et al.) of the Mahābhārata Critical Edition,” paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 15, 2017, IUC, Dubrovnik, Croatia. 2 All quotations from the session are from Joydeep Bagchee’s electronic recording. A compete transcript is currently under preparation. Besides John and Mary Brockington, Mislav Ježić, Przemyslaw Szczurek, Sven Sellmer, Vidyut and Ashok Aklujkar, and Simon Brodbeck also attended. We thank Ashok Aklujkar and Simon Brodbeck for detailed comments and suggestions for revision. 3 All references are to V. S. Sukthankar, et al., eds., The Mahābhārata for the First Time Critically Edited, 19 vols. (Pune: Bhandarkar Oriental Research Institute, 1933–66). 4 See “International Conference on Mahabharata Manthan: A Critical Revisit to Tangible and Intangible Heritage, 19th–21st July 2017, Resolution [sic] & Recommendations,” computer printout, 2017. The resolution notes that “the Critical Edition of the Mahabharata published by BORI Pune 1  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Mahābhārata and the Rāmāyaṇa, Brockington’s views are necessarily at the center of this reassessment.5 But we also evaluate the work of other scholars who either provide support for or affirm his conclusions. We are aware that this paper presumes significant earlier work. We could not always provide a full preview of our reasons for rejecting specific positions or interpretations of the Mahābhārata critical edition (for example, Brockington’s suggestion that “the constituted text is […] an approximation to a written redaction of the text that became normative,” which is a variation on Andreas Bigger’s hypothesis that “the critical edition represents an attempt to approximate [a] normative redaction”).6 All such discussions must await publication of our forthcoming book Philology and Criticism.7 The book also outlines our reasons for thinking that the case for oral precursors of the Rāmāyaṇa and the Mahābhārata is not made.8 We are aware, also, that more can and needs to be said about the two critical editions (that is, of the Mahābhārata and the Rāmāyaṇa). But we hope that the present paper, minimally, provides an impetus for further discussion(s) at a future venue. should be revised by a committee which will consult the manuscripts left out, the Arabic, Persian and Indonesian and other South-East Asian versions. The committee will study inconsistencies and contradictions and ancient commentaries on these with a view to suggesting resolutions. It will take into account findings of lexical and stylistic research.” 5 This paper focuses primarily on the Mahābhārata. However, Brockington frequently cites his experience with the Rāmāyaṇa for the Mahābhārata. This paper accordingly discusses both. 6 Andreas Bigger, Balarāma im Mahābhārata: Seine Darstellung im Rahmen des Textes und seiner Entwicklung (Wiesbaden: Harrassowitz, 1998), 14. 7 Vishwa Adluri and Joydeep Bagchee, Philology and Criticism: A Guide to Mahābhārata Textual Criticism (London: Anthem, 2018). The hypothesis that the critical edition reconstructs a “normative redaction” is discussed in chapter 1. 8 See ibid., especially chapters 1–2. To clarify a pervasive misunderstanding, it is not that we think oral transmission is completely excluded: we think scholars have not made the case for it. Their attempts to reinstate the oral hypothesis after the critical edition are circular and question-begging. The orality hypothesis frequently conceals real deficits in their knowledge of textual criticism, to say nothing of the fact that they can more easily make claims about a non-existent text since these claims are unverifiable. Sullivan penetratingly observes: “As [Kevin] McGrath writes in his most recent volume (McGrath 2013: 10), ‘In this book I examine what I envision to be an earlier and thoroughly heroic status of the poem when it still existed in a preliterate form, examining the narrative as it concerns Kṛṣṇa as he exists in that hypothetically ‘earlier’ telling’. In fact, we do not have such a text, so McGrath is examining the text we have while omitting portions he chooses to ignore, namely, parts he regards as didactic, concerning ‘edification’ (McGrath 2004: 5).” Bruce M. Sullivan, “An Overview of Mahābhārata Scholarship: A Perspective on the State of the Field,” Religion Compass 10, no. 7 (2016): 170. We are also not unaware that the search for “oral” versions of the Rāmāyaṇa and the Mahābhārata is rooted in what J. A. B. van Buitenen calls “the antibrahminism of Western scholarship of the last century.” J. A. B. van Buitenen, “The Mahābhārata: Introduction,” in The Mahābhārata: I. The Book of the Beginning (Chicago: University of Chicago Press, 1973), xxi. 2  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Redaction Criticism Critics of the Mahābhārata critical edition frequently cite the occurrence of contradictions, duplications, and other inconsistencies in the constituted text as an argument against the critical edition.9 Setting out from M. A. Mehendale’s statement that “the task of removing the spurious matter that still remains in the critically constituted text and bringing it as close as possible to the oldest version of the Mbh. has been left by them [the editors] to future students of the text,”10 these scholars (among them John Brockington, Pradip Bhattacharya, and Mislav Ježić) argue for the use of “higher criticism” to provide a more authoritative and more consistent text purged of these errors.11 In their view, a combination of criteria such as internal consistency, likelihood of occurrence, common sense, and style enable a closer approximation of the hypothetical Kṣatriya “Ur-epic” they consider the Mahābhārata’s source. 12 M. A. 9 For examples, see M. A. Mehendale, “Interpolations in the Mahābhārata,” Annals of the Bhandarkar Oriental Research Institute 82, no. 1/4 (2001): 193–212; M. A. Mehendale, “The Critical Edition of the Mahābhārata: Its Achievement and Limitations,” Annals of the Bhandarkar Oriental Research Institute 88 (2007): 1–16; and M. A. Mehendale, “The Critical Edition of the Mahābhārata: Its Constitution, Achievements, and Limitations,” in Text and Variations of the Mahābhārata: Contextual, Regional and Performative Traditions, ed. Kalyan Kumar Chakravarty (New Delhi: National Mission for Manuscripts, 2009), 3–23. 10 Mehendale, “Interpolations in the Mahābhārata,” 196. 11 Mehendale’s work has inspired not only Brockington but also Bhattacharya. A complete bibliography of his work is available in Vishwa Adluri and Alf Hiltebeitel, “Redressing the Undisrobing of Draupadī,” in Indology’s Pulse: Arts in Context. Essays Presented to Doris Meth Srinivasan in Admiration of Her Scholarly Research, ed. Corinna Wessels-Mevissen and Gerd J. R. Mevissen, with the assistance of Arundhati Banerji and Vinay Kumar Gupta (New Delhi: Aryan Books International, forthcoming). For Brockington’s reliance on Mehendale, see his Paris conference paper (cited later) and note 62. In contrast, Ježić primarily cites the German Mahābhārata and Bhagavadgītā critics Adolf Holtzmann Jr., Hermann Oldenberg, Richard Garbe, and Rudolf Otto. For his sources and his ideas of “textual criticism” see Vishwa Adluri and Joydeep Bagchee, “Paradigm Lost: The Application of the Historical- Critical Method to the Bhagavadgītā,” International Journal of Hindu Studies 20, no. 2 (2016): 199– 301. To our knowledge, Ježić does not cite Mehendale. 12 For a recent example, see Przemyslaw Szczurek, “In Search of the Beginning of the Strīparvan,” paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 13, 2017, IUC, Dubrovnik, Croatia. In lieu of the paper, we quote from Szczurek’s abstract. “While reading the initial part of the Strīparvan (i.e. MBh. 11,1–8 + appendix I of the Critical Edition, CE, or 11,1–9 of the Bombay Edition, BE)—up to the beginning of the descriptions of the stricte funeral operations at Kurukṣetra (starting with 11,9 of CE or 11,10 of BE)— the reader favouring a diachronic attitude towards the text of the epic may have doubts regarding the original unity of all these chapters. Twice there, in a very similar way, Janamejaya asks Vaiśampāyana to describe to him the events that happened after the military operations and the night slaughter (11,1.1– 3  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Mehendale, for example, argues that, despite the editors’ efforts at “removing much spurious matter, their text remains still burdened with much which could equally he spurious.” He notes that the editors “are in no sense presenting to the readers the ‘original’ Mbh. attributed to Vyāsa and recited by Vaiśaṁpāyana, not even the version that was narrated by Sūta Ugraśravas in the Naimiṣa forest to the sages assembled there for the long sacrificial session of Śaunaka.”13 Mehendale identifies three types of interpolations: 1. Those revealed by ms. evidence and set aside by the editors, 2. those that are not revealed by ms. evidence and part of the constituted text but are quite obvious due to contradictions in consecutive stanzas, and 3. those revealed either by contradictions that are not so obvious because they are not found in consecutive passages but are found in the text as a whole, or by the criterion of intrinsic probability.14 By implication, future editors or readers should remove or reconcile these contradictions to approximate better the original, even if “arriving at the truly ‘original’ text of Vyāsa is well nigh impossible.”15 The problem with this suggestion, however, is that, lacking objective criteria, we cannot determine, between two conflicting accounts, which is the more original. Neither is it certain that the original text was free of contradiction. The source of the “error” may lie in the author’s copy itself. The critical edition revealed that the contradictions were a feature of the 3; App. I, lines 1–4). Three times Dhṛtarāṣṭra falls to the ground unconscious as a result of the loss of his sons (11,1.9; 11,8.1 [in spite of his statement in 11,3.1 that his pain has been overcome]; App. I, l. 13– 14 = 11,9.8 BE), and three different interlocutors (Saṁjaya, Vidura and Vyāsa) address their consolatory speeches to Dhṛtarāṣṭra in successive chapters. Even with a superficial reading of the text, the whole series of consolations, advice, ethical and philosophical remarks directed to Dhṛtarāṣṭra (along with the images, parables, comparisons and metaphors placed in them) can give the impression of a multi-layered construction. The author of this paper [Szczurek] makes an attempt to look through the content of the initial parts of the Strīparvan, to comment upon the information contained in the Critical Apparatus of CE, to reflect upon the role of the part deleted from CE and marked as appendix I, to point out and classify the repetitions contained in the text under examination. He also ponders the question of whether it is possible to talk here about the history of this part or its relative chronology. Namely, whether preserved manuscripts and previous epic studies allow the presentation of a hypothesis of an older (original (?) and most probably shorter) version of the beginning of this parvan as well as its gradual development and enrichment of both the argumentation and the content of the great epic.” Despite repeated requests, Szczurek declined to share his paper. 13 Mehendale, “Interpolations in the Mahābhārata,” 195. 14 Ibid. 15 Ibid., 196. 4  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches common ancestor of all surviving manuscripts examined for the edition (by definition, the archetype). It would be irresponsible to emend the constituted text to fit our preconceptions of what the original looked like. Mehendale misleadingly presents his “application of ‘higher criticism’ to the text” as a continuation of the editors’ project, even though they explicitly distinguished their work from so-called higher criticism. 16 Sukthankar is scathing in his rejection of the latter.17 The two approaches have different genealogies and pursue opposed 16 Mehendale footnotes the statement “the task of removing the spurious matter that still remains in the constituted text and bringing it as close as possible to the older version of the Mbh. has been left by them [the Mahābhārata editors] to future students of the text” (ibid., 196) with a reference to Edgerton. “This task has been characterized as the application of ‘higher criticism’ to the text. See F. Edgerton, Introduction to Sabhā-parvan, p. XXXIII.” Ibid., 196, n. 12. In M. A. Mehendale, “The Critical Edition of the Mahābhārata,” he notes: “The Editors of the critical edition have already pointed out in their respective introductions to the parvans edited by them instances of such contradictions, especially when they occur in passages close to each other. They have, however, not said which of the two contradictory passages could be original and which spurious. That, according to them, would mean going beyond the scope of the work they had undertaken. In all such cases, taking decisions would be the task of later researchers by adopting ‘higher criticism’” (ibid., 7–8). The attached footnote reads: “To try to make the text consistent on such points would be to enter the realm of ‘higher criticism.’ F. Edgerton, Introduction to Sabhāparvan, p. xxxiii.” Ibid., 7–8, n. 16. Mehendale misrepresents Edgerton both times. What Edgerton actually says is: “To try to make the text consistent on such points would be to enter the realm of ‘higher criticism.’ I agree with Sukthankar that our job is different.” Franklin Edgerton, “Introduction,” in The Sabhāparvan for the First Time Critically Edited (Poona: Bhandarkar Oriental Research Institute, 1944), xxxiii (italics added). The reference is to V. S. Sukthankar, “Prolegomena,” in The Ādiparvan for the First Time Critically Edited, vol. 1 (Poona: Bhandarkar Oriental Research Institute, 1933), lxxxvii. Sukthankar notes: “The above examples will show that the diaskeuasts did not always employ any great art—I may add, fortunately—in conflating two discrepant accounts of an incident, which is by no means an easy task. To resolve such anomalies, however, is beyond the scope of this edition, since the entire manuscript evidence unanimously supports the conflation, which is too old and deep-rooted to be treated by the ordinary principles of textual criticism. If we went about, at this stage of our work, athetizing such passages as were self-contradictory or as contradicted the data of some other part of the epic, there would not be much left of the Mahābhārata to edit in the end.” 17 See V. S. Sukthankar, On the Meaning of the Mahābhārata (Bombay: Asiatic Society, 1957), 10–11: “The Mahābhārata is in short a veritable chaos, containing some good and much useless matter. It is a great pity that a fine heroic poem, which, may even be found to contain precious germs of ancient Indian history, should have been thus ruined by its careless custodians. But it is not quite beyond redemption. A skillful surgical operation—technically called ‘Higher Criticism’—could still disentangle the submerged ‘epic core’ from the adventitious matter—known to textual critics as ‘Interpolation’—in which it lies embedded. The Mahābhārata Problem thus reduces itself to the discovery of criteria which will enable us to analyse the poem and to dissect out the ‘epic nucleus’ from the spurious additions with which it is deeply incrusted. This is the ‘Analytical Theory’ of the origin and the character of the Mahābhārata, which was espoused by the majority of the Western critics of the Great Epic of India, chief among them being Lassen, Weber, Ludwig, Sörensen, Hopkins and Winternitz.” And see also ibid., 30–31 for Sukthankar’s rejection of the oral epic: “Higher Criticism would have us search for the lost ‘epic nucleus,’ which is apparently something immensely worth possessing. With that end in view it proceeds 5  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches aims.18 Even assuming that, in a handful of readings, we can restore not just the text of the archetype, but of one of its ancestors, we cannot reach the mythic oral epic Mahābhārata critics seek.19 The Bhandarkar editors aimed to “to reconstruct the oldest form of the text which it is possible to reach, on the basis of the manuscript material available,”20 whereas Mehendale and Brockington think they can reconstruct the “oral” epic they consider its source.21 Neither is it by the method of athetizing certain lines, passages, chapters, or even whole books. These are spurious, and all the rest is the work of one great poet. This method has been applied to comparatively more recent and also much simpler works, about whose historical context we happen to be better informed and where it would be much more legitimate. Yet even in these cases it has, as is well known, completely broken down. Very little reflection is needed to convince one that a mere process of stripping off what we regard as spurious will not automatically leave us with the pure and unalloyed ‘original.’ As we analyse the poem back towards its source, it proves to have not one source but many. What shall we do then? We know nothing about the hypothetical ‘nucleus.’ Moreover the nucleus we may discover in our analytical adventures is likely to prove to be not the ‘original’ we are looking for, but merely a projection of our own feelings. On the other hand we have got the poem, about which there is no doubt, and we may be able to puzzle out a good deal about its meaning, its inner meaning, if we tried. Let us then focus our thoughts upon that and try to understand it as best as we can. I believe we shall find in the poem itself something far greater and nobler than the lost paradise of the primitive Kṣatriya tale of love and war, for which the Western savants have been vainly searching and which the Indian people had long outgrown and discarded.” 18 Soulen and Soulen note: “Lower Criticism is an unhappy term, now of infrequent parlance, characterizing TEXTUAL CRITICISM in contrast to so-called higher criticism, i.e., all other forms of BIBLICAL CRITICISM. The term has fallen into disuse because of its pejorative sound coupled with the increasing acknowledgment that textual criticism is both important and complex.” Richard N. Soulen and R. Kendall Soulen, Handbook of Biblical Criticism, 4th ed. (Louisville, KY: Westminster John Knox Press, 2011), 121 (capitalization and block text in original). We make the stronger point: textual criticism is not a preparatory or auxiliary science to biblical criticism. 19 Trovato speaks of “staunch ‘reconstructionists’, who believe that the task of a scientific edition is not merely to transcribe a manuscript, or to reconstruct the archetype of surviving manuscripts, but to use the archetype as a point of departure, using all available means—linguistic, stylistic or metrical information, historical data, etc.—to try to come as close as possible to the lost original, detecting and correcting, as far as possible, but always as rationally and transparently as possible, the errors shared by surviving copies.” Paolo Trovato, Everything You Always Wanted to Know about Lachmann’s Method: A Non-standard Handbook of Genealogical Textual Criticism in the Age of Post-structuralism, Cladistics, and Copy-Text (Padua: Libreriauniversitaria, 2014), 15. These conditions are simply not given in the case of the Mahābhārata, a text whose earliest version is dated to the second century AD. Brockington, moreover, does not think a single author existed, speaking, rather, of “bards […] who recited in short songs […] the glorious deeds of their lords.” John Brockington, The Sanskrit Epics (Leiden: Brill, 1998), 19. 20 Sukthankar, “Prolegomena,” lxxxvi (Sukthankar’s italics). 21 For Brockington’s comments on the Sanskrit epics’ oral beginnings, see John Brockington, “The Textualization of the Sanskrit Epics,” in Textualization of Oral Epics, ed. Lauri Honko (Berlin and New York: Mouton de Gruyter, 2000), 193–216. He declares: “It hardly needs affirming that the Mahābhārata and the Rāmāyaṇa together represent the culmination of a lengthy tradition of oral poetry, transmitted through recitation by sūtas or bards” (ibid., 193), but the evidence never comes. The 6  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches clear what Mehendale’s criteria for “contradiction” are. 22 Standards for consistency are idiosyncratic. Where one reader sees a contradiction, another may find no difficulty.23 This is why the place for resolving contradictions is commentary, as, indeed, the work of glossators shows.24 More important, once we begin emending the text, where do we stop? Mehendale’s attached footnote merely states: “All references to the epic are to their Critical Editions, Sukthankar et al. 1933–66; Bhatt and Shah 1960–75. I use the term ‘* passages’ generically to designate all material excluded from the text of the Critical Edition in either the critical apparatus or Appendix I (unless context clearly indicates a more limited usage).” Elsewhere Brockington suggests: “In both the Sanskrit epics, the Rāmāyaṇa and the Mahābhārata, their framework stories present them as oral compositions; this suggests the importance of the roles not only of the bard or reciter but also of the audience whose presence is integral to this introductory frame.” Ibid., 193. Insofar as he refers to the epic’s narration at Śaunaka’s sattra in the Naimiṣa Forest, the statement is false. Ugraśravas neither composes the Mahābhārata before his audience nor refers to an anterior “oral composition.” In this first reference, the Mahābhārata is spoken of as a collection Vyāsa authored and taught his students (Mahābhārata 1.1.15– 19 and 63–64). The transmission is first Brahmanic and only then bardic if we understand this term to mean a narrator of “ancient Lore” (sūta paurāṇika) rather than the representative “of a lengthy tradition of oral poetry,” Brockington imagines him as. Contrary to Brockington’s claim that “the Mahābhārata in its frame indicates very clearly its oral nature. […] Here we have the epic presented as if being narrated then and there, with reciters and audience actually within the text” (ibid., 202), the sages assembled in the Naimiṣa Forest ask to hear Vyāsa’s composition (dvaipāyanena yat proktaṁ purāṇaṁ paramarṣiṇā, Mahābhārata 1.1.15). The bard responds that he will narrate Vyāsa’s “entire thought” (mataṁ kṛtsnaṁ, 1.1.23). The Mahābhārata’s account of its transmission is more complex than most scholars imagine. It is keenly attuned to issues of memory, philosophical insight, and narrative time. See Vishwa Adluri, “The Perils of Textual Transmission: Decapitation and Recapitulation,” Seminar 608, The Enduring Epic: A Symposium on Some Concerns Raised in the Mahābhārata (2010): 48–54 and Vishwa Adluri, “Literary Violence and Literal Salvation: Śaunaka Interprets the Mahābhārata,” Exemplar: The Journal of South Asian Studies 1, no. 2 (2012): 45–68. 22 Neither Mehendale nor Brockington (nor Bhattacharya, who cites Mehendale and Brockington in support of his criticisms of the critical edition) provide objective criteria for contradictions. They rely on everyday experience and common sense expectations of the text, but consider neither the manuscript evidence nor the commentarial tradition, even though the latter often preserve evidence of objective difficulties in the text as well as providing the surest guide to the text’s history of reception. For a discussion see Joydeep Bagchee, “The Epic’s Singularization to Come: The Śakuntalā and Yayāti Upākhyānas in the Southern Recension of the Mahābhārata,” in Argument and Design: The Unity of the Mahābhārata, ed. Vishwa Adluri and Joydeep Bagchee (Leiden and Boston: Brill, 2016), 83–126, especially 111, n. 83. Their approach thus suffers from overhastiness understood in the full sense it acquired in the Enlightenment. 23 For criticisms of Mehendale and Bhattacharya see Adluri and Hiltebeitel, “Redressing the Undisrobing of Draupadī.” 24 Krešimir Krnic provides valuable examples in “Squaring the Circle: Commentarial and Intratextual Explanation of Controversial Places in Vālmīki’s Rāmāyaṇa,” paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 14, 2017, IUC, Dubrovnik, Croatia. As he notes, the commentators were not unaware of contradictions in the text. But they made strenuous efforts to preserve the text’s integrity in keeping with the first principle of hermeneutics. 7  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches examples of the third type of interpolation highlight the difficulty. Despite his invocation of “intrinsic probability” to justify their removal, the manuscript evidence is unambiguous.25 His arguments for removing a segment of adhyāya 41 of the Bhīṣmaparvan reveal the subjective and ad hoc nature of his criteria.26 He argues, 25 Mehendale, “Interpolations in the Mahābhārata,” 201: “We shall now consider the third kind of interpolations, viz. such as are revealed to be of doubtful nature due to internal contradiction or due to the criterion of intrinsic probability. The contradictions, however, are not obvious since they do not occur in the consecutive passages but are revealed by the study of different Adhyāyas in different parvans.” The expression is Sukthankar’s, who introduces it in the context of a discussion about the passages found only in some recensions. The question arises whether they are insertions in the recensions that contain them or omissions in the rest. Sukthankar argues that “the intrinsic probability is wholly on the side of those manuscripts that lack these accretions [the reference is to additions in γ or the vulgate].” Sukthankar, “Prolegomena,” liv. Compare also ibid., lviii: “Here the evidence of documentary and intrinsic probability is almost equally balanced; and documentary probability points in one direction, while intrinsic probability points in the other” and lxxxii: “There are indeed yet more difficult cases, where the evidence pro et contra of documentary and intrinsic probability is equally balanced, as far as we can at present judge. In such cases we are forced to look for small things which look suspicious and lead us to probabilities, not facts.” Sukthankar does not use “intrinsic probability” in relation to presumed contradictions and Mehendale misconstrues his meaning. 26 “We now come to an instance of an interpolation, which could be an innovation of some narrator of the poem. It need not be case of borrowing a from another version. Dhṛtarāṣṭra asked Saṁjaya: who struck first, my warriors or those of the Pāṇḍavas? (6.22.18). The account of the war then should have followed. But the Gītā intervenes, and at the end of it Arjuna became ready to fight (6.23–6.40). Now at least with the next adhyāya (41) the war description should have begun. But again it does not. It begins actually with adhyāya 42. The intervening adhyāya 41 reports a very strange and most ridiculous incident. Just when the first arrow was about to be shot Yudhiṣṭhira, we are told, got down from his chariot, dropped his armour and weapons and, with folded hands, started on foot towards the chariot of Bhīṣma. The rest of the Pāṇḍava, confounded, and Kṛṣṇa followed suit. Many other kings did the same, the Pāṇḍavas asked Yudhiṣṭhira the meaning of his curious behaviour, but he would not reply. Kṛṣṇa, then, told them that Yudhiṣṭhira was proceeding to Bhīṣma, Droṇa and other elders on the opposite side to seek their permission to fight. Yudhiṣṭhira did go to Bhīṣma and sought his permission. He did not stop there. When he was granted a boon by Bhīṣma, Yudhiṣṭhira asked him how the Pāṇḍavas could defeat him. This he asked within the hearing of all those who could listen to the conversation. Saṁjaya too could listen to it to report it later to Dhṛtarāṣṭra. Bhīṣma hesitated to reveal immediately the secret of his defeat but promised to do that later. Yudhiṣṭhira next approached Droṇa. Everything happened exactly as before, the only difference was that in the case of Droṇa, Yudhiṣṭhira did not just ask how to defeat him, but went a step further and asked how to kill him (vadhopāyaṁ vadātmanaḥ 6.41.58). Droṇa did not hesitate like Bhīṣma and straightaway told him that any one could kill him once he had laid down the arms after learning some extremely bad news from a person whom he could trust (śāstraṁ cāhaṁ raṇe jahyāṁ śrutvā sumahad apriyam / śraddheyavākyāt puruṣād etat satyaṁ bravīmi te 6.4.61). When Yudhiṣṭhira next met Kṛpa, after completing the initial formalities, he did not ask him either how to defeat him or how to kill him. He just kept quiet. Kṛpa, who in all probability had heard what had happened in Yudhiṣṭhira’s earlier meetings with Bhīṣma and Droṇa, told him without his asking that he could never be killed (avadhyo’haṁ mahīpāla 6.41.69). The meeting ended there, and Yudhiṣṭhira approached Śalya. He requested Śalya that later in the war when he would become the charioteer of 8  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches This long episode of seventy-eight stanzas (6.41.6–83) is clearly of a secondary nature. Its spurious nature is revealed by the fact that it shows advance knowledge on the part of the inventor of events which took place later during the war. Everything happens later exactly as foreseen here. The incident is also intrinsically improbable. It is impossible to believe that Yudhiṣṭhira put such questions and make request to the leaders of the opponents on the battlefield. The replies given and the request agreed to by the leaders amount to treason. It is surprising that Duryodhana did not object immediately to what had happened. Later during the war, Duryodhana is occasionally seen finding fault with Bhīṣma and Droṇa for not fighting seriously with the Pāṇḍavas, but he never utters a word of reproach about their objectionable behaviour as reported in the above episode.27 Mehendale’s claim rests on several implicit assumptions about the nature of the Mahābhārata. It presumes that the text refers to a real, historical war and that its characters are real individuals, who always behaved consistently. But what proves that Yudhiṣṭhira could not have asked these questions—his character as a historical individual or his character as the epic envisions it? And why should either be immutable? Neither need we presume an “inventor” with “advance knowledge […] of events which took place later during the war,” because an author has the identical knowledge. As their creator, he can anticipate future events in his narrative or mold his characters’ actions to suit the plot’s requirements. Mehendale confuses narrated events with actual events that unfold along a historical timeline. The argument not only denies the poet the freedom to shape his narrative; it also makes several assumptions about the hypothetical revisionists’ motives. Except that Mehendale does not attribute the insertion to “Brahmanic revisionists,” who inserted their “Brahmanic apologetic” into the text, the argument resembles Georg von Simson’s idea of a Brahmanic revision of a heroic Kṣatriya epic.28 For comparison, we cite the earlier account: Adhyāya 6.41, the last of our insertion, manifests just as late and unrealistic an account [as the Bhagavadgītā]. Yudhiṣṭhira, followed by the other sons of Pāṇḍu, suddenly lays down his arms and approaches the opposing army to ask Bhīṣma, Droṇa, Kṛpa and Śalya for their blessings. In Karṇa he should lower Karṇa’s spirits (sūtaputrasya saṁgrāme kāryo tajovadhas tvayā 6.41.81). After getting this assurance from Śalya, Yudhiṣṭhira returned to his side.” Mehendale, “Interpolations in the Mahābhārata,” 210–11. 27 Mehendale, “Interpolations in the Mahābhārata,” 211. 28 Georg von Simson, “Die Einschaltung der Bhagavadgītā im Bhīṣmaparvan des Mahābhārata,” Indo- Iranian Journal 11, no.3 (1969): 159–74. 9  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches response to Yudhiṣṭhira’s question, how he may be defeated in battle, Bhīṣma replies (6.41.43) that he does not know any foe who can overcome him. The time for his death has not come; Yudhiṣṭhira should come again. This is a clear reference to 6.103 where Yudhiṣṭhira’s second visit to Bhīṣma is described. In the night between the ninth and the tenth day, the day of the decisive battle against Bhīṣma, Yudhiṣṭhira expresses his concern about Bhīṣma’s invincibility and solicits his advice. Kṛṣṇa agrees to the suggestion and the Pāṇḍavas approach Bhīṣma. He reveals to them that he will not fight against Śikhaṇḍin who was born a woman. Arjuna can therefore bring him down if he were to use Śikhaṇḍin as his shield. . . . It is thus likely that the visit of the Pāṇḍavas to Bhīṣma in 6.103 is just as late an insertion as its counterpart in 6.41.29 Von Simson continues: It is thus likely that the Pāṇḍavas’ visit to Bhīṣma in 6.103 is a late insertion like its counterpart in 6.41. The motivation for both interpolations is clear and corresponds to that for the Bhagavadgītā insertion: the later Brahmanic revisionists of the Mahābhārata thought they had to justify the conduct of the Pāṇḍavas during the battle; first, the battle against the grandfather and teacher required an ideological justification (Bhagavadgītā); thereafter, the base trick by means of which the Pāṇḍavas defeated Bhīṣma had to be excused: the victim himself makes the suggestion and the killers cannot be criticized when they accept his suggestion. This becomes quite clear in Yudhiṣṭhira’s conversation with Droṇa, whom he visits next after Bhīṣma. 6.41.47 (cf. v. 64) yotsye vigatakalmaṣaḥ (I wish to fight without stain) Yudhiṣṭhira says, and in response to the question of how he may be defeated, Droṇa indicates the sole means that would motivate him to lay down his arms. 6.41.61: śastraṁ cāhaṁ raṇe jahyāṁ śrutvā sumahad apriyam | śraddheyavākyāt puruṣād etat satyaṁ bravīmi te ||. I would lay down arms in battle if I learned of bad news from a man whose word one can trust, this I promise you. Granted, this is not as direct a demand to use an unfair trick as in 6.103, [but] the author’s intent is manifestly the same: a justification of the Pāṇḍavas’ ignoble tactics, which must have been embarrassing to the devout Brahmans of a later age.” Ibid., 173–74. In the history of Mahābhārata scholarship, attempts to prove the Bhagavadgītā a forgery (either all or parts of it) led to peculiar results.30 But none is as tendentious and racist as Georg von Simson’s, which, citing specious “text-critical grounds” (textkritische Anhaltspunkte), excised 29 Ibid., 171–72. 30 See Adluri and Bagchee, “Paradigm Lost,” especially pages 259–61 for an overview. 10  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches the entire Bhagavadgītā and291 additional verses from the Mahābhārata.31 Von Simson’s reason for excising the Bhagavadgītā was its alleged Brahmanic character, which conflicted with his anti-Brahmanism. Racial and anti-semitic prejudices, however, are not tenable grounds for a reconstruction.32 The attempts to reconstruct a Bhagavadgītā free of Brahmanism did not lead to objective, scientifically verifiable results.33 The following table summarizes these attempts: Year Indologist “Critical” criterion Verses retained (out of 700) 1893 Adolf Only pantheistic elements in the Gītā are original 164 Holtzmann Jr. 1905 Richard Garbe Only theistic elements in the Gītā are original 528 1910 F. Otto Schrader Original Gītā, as part of the pre-Viṣṇuite Mahābhārata, ends with 85 2.38 1914 Richard Garbe Garbe’s second attempt incorporating Winternitz’s suggestions 353 1918 Hermann Jacobi Only epic elements are original 71 1919 Hermann Key to the poem in 2.39; everything thereafter belongs to the 85 or 82 Oldenberg “didactic poem” 1930 Jarl Charpentier Amalgam of other scholars’ ideas about the “original” Gītā 66 1934 Rudolf Otto Only elements relating to Arjuna’s “situation” are original; 139 Gītā an instance of the numinous experience of the mysterium 31 We evaluate von Simson’s reasons in Vishwa Adluri and Joydeep Bagchee, The Nay Science: A History of German Indology (New York: Oxford University Press, 2014), 277–96: “The Method Becomes Autonomous.” 32 On anti-Brahmanism as a form of anti-semitism see Vishwa Adluri and Joydeep Bagchee, “Jews and Hindus in Indology,” paper published on Academia.edu; https://www.academia.edu/30937643/Jews_and_Hindus_in_Indology (accessed October 10, 2017). As discussed in Adluri and Bagchee, The Nay Science, 331–39, German Indologists, who claimed that “Brahmans” corrupted the primitive versions of these texts (identified with a “heroic” Aryan or Indo- German race), were recurring to long-standing Lutheran prejudices against the rabbinic tradition, 33 Brockington cites von Simson’s view of the Bhagavadgītā as “a late piece of Brahmin apologetic” verbatim (on the problems with this unattributed citation, see later), but provides neither an analysis nor a justification. See John Brockington, “The Bhagavadgītā: Text and Context,” in The Fruits of Our Desiring: An Inquiry into the Ethics of the Bhagavadgītā for Our Times, ed. Julius Lipner (Calgary: Bayeux, 1997), 28–47, reprinted in The Bhagavad Gita: A New Translation, Contexts, Criticism, trans. Gavin Flood and Charles Martin, ed. Gavin Flood, Norton Critical Editions (New York: W. W. Norton & Company, 2014), 135–53. 11  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches tremendum found preeminently in Luther’s “On the Bondage of the Will” 1937 Jakob Wilhelm Gītā is a “metaphysics of battle and action” that combines “the two 140 Hauer life poles of the Indo-Germanic nature” 1969 Georg von The Gītā is a “secondary interpolation” following a first 700 Simson interpolation from 6.16.21 to 6.20.22 and 6.42.1 1986 Mislav Ježić Attempt to validate German scholars’ ideas of the Bhagavadgītā: 60 “The poetic parts of the Gītā are relatively more ancient than the didactic parts attached to them” Extrapolated to the Mahābhārata, these attempts at identifying an “original” Gītā based on a priori conceptions of its contents and aim would lead to vastly different texts and, in some cases, to no text at all. This is the opposite of an objective textual criticism aware of its responsibility to the reader and its duty to preserve texts.34 A critical edition in the sense of an edition free of contradictions is neither achievable nor is it necessarily an improvement over the present text.35 34 Leonardi expresses this responsibility well: “The concept of a critical edition [also] implies a responsibility to propose a text that, while meeting the requirements of science and elucidating the manuscript tradition, is not merely accessible to specialists, but recovers a work of the past for a contemporary public to read; one that does not reproduce a document, but interprets the tradition as a whole as a means to transfer its textual reality into something that is readable today.” Lino Leonardi, “Il testo come ipotesi (critica del manoscritto-base),” Medioevo romanzo 35 (2011): 6 (Trovato’s translation). Indology’s greatest flaw is that it lacks a similar sense of responsibility to say nothing of a readership to which it could feel responsibility. Compare Adluri and Bagchee, The Nay Science, 372: “On one hand, the Orientalists claimed justification for their work in the fact that they were offering a nonconfessional take on Indian texts. On the other, the texts studied by them had lost all reference to their originating communities. They were no longer the texts they had originally been. The ‘knowledge’ the Orientalists attained in this process was thus pertinent only for them, premised on objects that existed only for them.” 35 Contrary to the expectation that the critical edition provide a completely coherent text, it attempts, rather, to undo the emendations of scribes and editors. A critical edition prefers difficult readings over easier ones (the difficiliores over faciliores) and generally works against the tendency to banalization. F. A. Wolf famously wrote: “when the witnesses require it, a true recension replaces attractive readings with less attractive ones. It takes off bandages and lays bare the sores. Finally, it cures not only manifest ills, as bad doctors do, but hidden ones too.” F. A. Wolf, Prolegomena to Homer, trans. with an introduction and notes by A. Grafton, G. W. Most, and J. E. G. Zetzel (Princeton, NJ: Princeton University Press, 1985), 44. Sukthankar explicitly noted the advantages of the vulgate over the constituted text: “The place for resolving contradictions is interpretation, as, indeed, the work of glossators shows. The Vulgate text of the Mahābhārata is fairly readable and will appear in places, at first sight, to be even ‘better’ than the critical text, because the former has been purged by the continuous emendations of scholars for centuries. A whole army of anonymous scholars and poets must have worked at the text to make it smooth and easy of comprehension and to increase its popularity and usefulness by adding to it 12  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Until its advocates establish objective, universally valid, and scientifically verifiable criteria for its reconstruction, we cannot agree on the text. Indeed, as we demonstrated, the only unifying feature of their various reconstructions was their racial and anti-Brahmanic prejudices, which led them to propose an “Ur-epic” that supposedly embodied the heroic values of the Aryan race.36 Source Criticism Few have dedicated themselves to studying the Sanskrit epics and their critical editions like John Brockington. 37 Recent proposals for a revision of the Mahābhārata critical edition unsurprisingly invoke his views. In this section, we consider his recent statement on the critical editions of the Rāmāyaṇa and the Mahābhārata, presented on the occasion of the international colloquium “Enjeux de la philologie indienne: Traditions, éditions, traductions/transferts,” December 5–7, 2016, Collège de France, Paris.38 We first quote the relevant portion of his lecture (the complete recording is available online):39 interesting anecdotes, incorporating into it current and popular versions and explanations, bringing it in a line with the ethical, moral, religious and political ideas of essentially different ages.” Sukthankar, “Prolegomena,” ciii. Mehendale and Bhattacharya do not seek a critical text, but a thoroughly modernized text revised to meet their expectations of consistency, sense, and social propriety. They stand precisely in the tradition of that “whole army of anonymous scholars and poets” that continually revised the Mahābhārata to make it conform to contemporary norms and ideals. 36 See Adluri and Bagchee, The Nay Science, chapters 1–3. In Adluri and Bagchee, “Paradigm Lost,” we showed that although P. L. Bhargava and Mislav Ježić, by the lights of their own criteria, should have attained different results from their German peers, they finessed their results to validate their colleagues’ view of a “Brahmanic” takeover of an original heroic Kṣatriya epic. 37 In Eli Franco’s assessment Brockington is “arguably the greatest living scholar on Indian epic literature.” Eli Franco, review of The Nay Science: A History of German Indology, by Vishwa Adluri and Joydeep Bagchee, South Asia: Journal of South Asian Studies 39, no. 3 (2016): 696. We discuss Franco’s reasons for praising Brockington in Vishwa Adluri and Joydeep Bagchee, “Theses on Indology,” paper published on Academia.edu, https://www.academia.edu/30584042/Theses_on_Indology (accessed October 13, 2017). See, especially, ibid., 29: “his [Brockington’s] work is a pastiche of German views from the past two centuries.” 38 For the conference theme, program and abstracts see: http://www.iran-inde.cnrs.fr/evenements- scientifiques/colloques-et-conferences-2016/enjeux-de-la-philologie-indienne-traditions-editions- traductions-transferts.html?date_debut=2017-03-28&lang=fr (accessed August 30, 2017). The conference was organized by Jean-Noël Robert (Collège de France, CRCAO) and a scientific committee comprising Lyne Bansat-Boudon (EPHE), Silvia D’Intino (CNRS), Philippe Hoffmann (EPHE), Charles Malamoud (EPHE), and Jean-Noël Robert (Collège de France). 39 John Brockington, “Regions and Recensions, Scripts and Manuscripts: The Textual History of the Rāmāyaṇa and Mahābhārata,” https://www.college-de-france.fr/site/jean-noel-robert/symposium- 13  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Fifty year after completion of the Mahābhārata critical edition is an appropriate time to review its success, its limitations, and its impact on epic scholarship along with that of the Rāmāyaṇa critical edition, completed in 1975. I have myself studied the Rāmāyaṇa more fully and shall rely on that for much of what I present, but I hope to make my remarks relevant to both epics, and more generally to the themes of this conference. Although they are classified by different names in the Indian tradition, for convenience I shall use the term epics to refer to both texts together. The basic, much debated issue is whether and how far oral performance gave way to manuscript transmission. External evidence is extremely meager and so internal evidence— formulaic diction, repetition, duplication of episodes and the like—has to be the prime source of information about this process of transition. I start with external evidence, which is only relevant for the Mahābhārata. Pāṇini’s Aṣṭadhyāyi lists the addition of the suffix ka to the names Vāsudeva or Arjuna, with the meaning of “one who has bhakti towards them” and mentions both Mahābhārata and some other names, while Patañjali mentions Yudhiṣṭhira and Arjuna as older and younger brothers. None of this though gives us much clue about what sort of text the grammarians knew, but in [unclear word] date it must have been oral. […] The internal evidence has mainly been seen in terms of formulaic diction under the influence of the work of Parry and Lord on Homer and on the south Slavic epics. Indeed, the name oral formulaic theory usually given to this approach privileges formulae, perhaps unduly so. It is important to remember that while oral poetry is characteristically formulaic, this does not automatically mean that all formulaic poetry is oral. But what it means for us is that further markers of orality are needed before any text can reasonably be regarded as oral. Berkeley Peabody, working on Hesiod, produced a set of criteria at five levels: phoneme, formula, enjambment, theme, and song, and suggested that positive indications were needed for each before a text could firmly be considered a direct product of an oral tradition. While we must be aware of the limitations of extrapolating from one language and culture to another, without allowing for the inevitable differences, these tests offered me a convenient framework to assemble the evidence for the Sanskrit epics since they cover a broader range of evidence than do other schemes and confirmed with varying degrees of certainty the oral origins of both epics. Another feature which seems more characteristic of oral poetry than written literature is ring composition. Its earliest significant application to either epic was Renate Söhnen’s monograph on speeches and dialogues from the Rāmāyaṇa. She argues persuasively for frequent deliberate structuring of speeches by anaphora, refrains, parallelism and the like but especially by means of ring composition. An oral origin for both epics has major implications for the duration of their growth period, which then has implications for issues about design and deliberate literary construction across their whole text 2016-12-06-14h00.htm (accessed June 30, 2017). We thank Silvia D’Intino for the link. The website incorrectly lists the title as “Religions and Recensions, [etc.]” We thank Brockington for the hint. 14  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches as well as for our assessment of the socio-cultural and religious contexts in which they evolved. The older view of this process of growth as essentially random has long been discarded by scholars who take a diachronic approach, through retained by those who take a narrowly synchronic approach as a stick with which to beat their opponents—discarded in favor of a more nuanced view of it as a design modified and adapted over successive periods. […] The publication of the critical editions has generated debate, focused mainly on the Mahābhārata about the validity of the criteria used for the selection of readings and even the value of these editions. Some scholars notably Madeleine Biardeau have rejected the critical edition, regarding it as in essence a fiction and prefer to use the vulgate text, in effect the text established by Nīlakaṇṭha Caturdhara late in the seventeenth century when he wrote his Bhāratabhavadīpa commentary. Often overlooked, though not by Biardeau herself, is that Nīlakaṇṭha explicitly compared several manuscripts and was well aware of regional divergences. His comments to this effect come prominently at the start of his commentary. His text is avowedly eclectic while aiming at being comprehensive. Appeal to or preference for the vulgate as more authentic is therefore misguided in my view. The critical editions are not only on the whole the best we have, despite their limitations that I discuss next, but also permit further advances in our understanding of them. Both critical editions are based on the readings of just a selection of the vast number of manuscripts known. A larger number was initially examined, but then reduced to a manageable number. The principles followed were first enunciated by Sukthankar in his Prolegomena to the Ādiparvan of the Mahābhārata and were broadly followed by all editors for both epics. These principles comprise grouping the manuscripts mainly by script into versions, relying on agreement between these versions for accepting any passage into the text in default of attestation in all manuscripts used, and accepting the two axioms of the textus simplicior and the lectio difficilior. Sukthankar recognized the impossibility of constructing a stemma codicum on the classical model and relied on grouping the script versions into primarily the northern and southern recensions. Since these are too far apart to permit the establishment of an archetype solely on the basis of manuscript readings, his—and the other editors’—constituted text is necessarily eclectic though based on careful judgment on the basis of the available evidence and is an approximation to a written redaction of the text that became normative. The same considerations apply mutatis mutandis to the Rāmāyaṇa. The first limitation that must be recognized is the long period of transmission of the texts—around two millennia on Hiltebeitel’s view that the Mahābhārata was composed between the mid-second century BC and the year zero by a committee of “out-of-sorts Brahmins”; more like two and a half millennia according to various other scholars. That is to be set against the recent date of most manuscripts. The oldest dated manuscript used for the Mahābhārata critical edition is dated 1261 AD while for the Rāmāyaṇa critical edition it is one dated 1020. However, for both epics, 15  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches the majority of the manuscripts used come from the sixteenth to nineteenth centuries. Most are essentially complete, whereas many older manuscripts lack portions of their text. The editors’ understandable tendency despite their good intentions to choose complete texts may well have resulted in the use of younger manuscripts. Moreover, several scripts are not represented: Oriya, Kannada, and Nandināgari. My own investigations into an Oriya script Rāmāyaṇa manuscript and of another one in Nevārī script shows that such further evidence leads to a reassessment of the concept of script-based versions in favor of one based more on regional affinities. My investigation of another one in Malayālam script in the Trivandrum collection necessitates a reassessment of the manuscript M4, previously dismissed as contaminated. These two manuscripts share features which indicate that they represent an alternative and probably older because shorter Malayālam recension. The view that the various scripts generated relatively isolated manuscript traditions with little contamination between them goes back as far as Lüders. It was reaffirmed by Sukthankar in his Prolegomena. I quote: “The superficial difference of scripts corresponds, as a matter of fact, to deep underlying textual differences.” But he goes on to state that in practice matters are never as clear cut. Moreover, most literate Indians nowadays know more than one script and there is no reason to think that such was not always the case. Besides a scribe might have listened to an oral presentation of either epic or perhaps the reading aloud of a text in an area or by an individual who was using a script he did not know and he could have incorporated some episodes that remained in his memory into a subsequent manuscript copying. Certainly there is ample evidence of scribes deliberately comparing manuscripts or at least checking one from another, in particular through variant readings found in the margins or interlinearly, sometimes even in a different script from the manuscript itself. The supposed barriers of scripts are in fact highly permeable. We should I submit be thinking more in terms of regional groupings that may cut across script differences. The number of marginal or interlinear additions also shows that the texts were commonly enlarged in successive copyings. Accidental loss is always a possibility but much less likely than accretion. This is of course what underlies the principle of the textus simplicior which governs Sukthankar’s preference, followed by the other Mahābhārata editors, for the Śāradā script manuscripts and other closely related ones, despite their relatively recent date. The Malayālam manuscripts also tend to contain a shorter text. Perhaps a feature well known in linguistics that in language change the periphery tends to be more conservative than the center also applies here. On the other hand, attestation in all manuscripts is not infallible proof of a passage being part of the original text. It could have entered the tradition early on and then spread to all extant manuscripts, assisted by this habit of comparing manuscripts. The tendency to include everything found in the exemplars available may well account for the duplication of episodes that we find in both texts. The scribe found varying versions in two manuscripts before him and 16  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches rather than choose one and reject the other, he incorporated both in an attempt at comprehensiveness at the expense of consistency. So the text even of the critical editions cannot be regarded as a completely coherent whole. A well known example is the dice game in the Sabhāparvan of the Mahābhārata and Mehendale has collected multiple examples of interpolations and internal contradictions. Such lateral transfer of passages or episodes, verses between manuscripts is not the only complicating factor. Scribes did not or could not always copy a single exemplar throughout quite apart from comparing with others. […] Whereas the premise that manuscripts written in the same script belong together has significant limitations then, the broader differentiation into northern and southern recensions has a greater measure of validity. But even here we must recognize that a significant number of manuscripts combine features of both. This is probably more often by conflation of the two at a relatively recent stage in the chain of transmission but perhaps sometimes by retention of older features. Again I cite as my example my research on the Trivandrum Malayālam script manuscript. Although it is close to M4 of the Ayodhyākāṇḍa of the Rāmāyaṇa, neither could have copied from the other and their closeness is marked by common absence of material rather than additions, suggesting a fairly early date. Neither unfortunately is dated but may be of sixteenth century date. However, they both combine features of the present northern and southern recensions. Within the northern recension, they align particularly with two manuscripts D1 and [D]2 which are usually assigned to the western subrecension within the northern recension and to a lesser extent with a third V1. The evidence of these two manuscripts, especially when linked with that from another fragmentary manuscript from Trivandrum, shows that they represent a definite alternative tradition current within Kerala differing markedly from that of M1–3 and that it has clear links with the so-called western recension. A related point is that the supposed uniformity of the southern recension of the Rāmāyaṇa has been overstated. All in all, the evidence of a large number of Rāmāyaṇa manuscripts shows that the simple opposition between the northern and southern recensions, however useful it may once have been as a heuristic device does not adequately reflect the complexities of the chain of transmission involved and consequently the question of their relative value is a more complex issue than was recognized in the constitution of the critical edition text. The same is no doubt the case for the Mahābhārata too. […] So far I have mainly being demonstrating limitations in the way that the critical editions of both epics classified the manuscript evidence into recensions and script-based versions, relying on what is only a modification of the stemma concept. Can we go beyond this in our understanding of manuscript relationships? I will put forward two possible ways of doing so. Thirty years ago, I suggested that the model of the Venn diagram could be borrowed from mathematics and logic as it had already been borrowed into linguistics. This uses circles or ellipses overlapping as necessary to represent a relationship between separate groups or sets of items. In the case of the 17  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches manuscripts I was studying this means that a circle or ellipse representing M4 and the Trivandrum manuscript will partly overlap with others for the other Malayālam manuscripts with D1 and [D]2 and V1, in this way showing to some extent at least the multiple allegiances involved. The drawbacks, as you will realize, are that this is a decidedly visual model rapidly becoming too complex to describe verbally and that even visually the patterns soon become highly intricate and less easy to interpret. The other is not just a model but even more a new way of establishing relationships between manuscripts developed for the Mahābhārata by Wendy Philipps-Rodriguez. This uses phylogenetic algorithms based on the principles of cladistics taken from biology to plot the degree of closeness between manuscripts, producing what she has called unrooted trees. The length of the branch of each manuscript showing the degree of divergence from the rest individually and collectively. This is a more objective way of identifying the actual relationships involved and was, for example, the means of identifying the change of alignment of the Sabhā manuscript D6 that I noted earlier. It is reassuring to find that the Mahābhārata critical edition is located near the center closer to the northern manuscripts reflecting both the editorial preference for them and the general connectedness [correctness?] of the readings chosen. Variant readings and new patterns of omission or addition are found in virtually every manuscript examined. Many of course are trivial but others are significant for building up a truer picture of the complex relationship between the various recensions and versions. There is therefore a continuing need for further examination of the manuscript evidence. However, both the approaches just noted rely heavily on the information found in the apparatus of the critical editions and this will inevitably be true of any other initiative. We can only go beyond the critical editions of the Mahābhārata and the Rāmāyaṇa because of the firm basis that they have provided for further research.40 40 Ibid. (all italics ours). Brockington refers to Renate Söhnen, Untersuchungen zur Komposition von Reden und Gesprächen im Rāmāyaṇa, 2 vols., Studien zur Indologie und Iranistik 6 (Reinbek: Dr. Inge Wezler, Verlag für orientalistische Fachpublikationen, 1979). For his review, see John Brockington, review of Untersuchungen zur Komposition von Reden und Gesprächen im Rāmāyaṇa, by Renate Söhnen, Journal of the Royal Asiatic Society of Great Britain and Ireland 114, no. 1 (1982): 65: “The first volume begins with various general observations, such as the utility in an oral narrative of clear markers at the beginnings and endings of speeches. The author then distinguishes three types of dialogue, dramatic (infrequent, despite providing all four passages chosen for detailed study), narrative, and emotive. She takes as her basis the Bombay edition on the grounds that, unlike the Critical edition, it represents a relatively uniform text. However, she lacks the courage of her convictions, regularly discussing the Critical edition’s readings, frequently—especially when it fits her analysis—accepting its rejection of stanzas found in the Bombay text, and occasionally suggesting that certain passages found in both texts are not original. Her comments on such occasions are sensible but inconsistent with her own initial position. Nevertheless, Söhnen argues persuasively for frequent deliberate structuring of speeches, especially for ‘Ringkomposition’, the organization of a speech symmetrically around a central passage (less commonly there is no central point as such but just an axis of symmetry). There is room for 18  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Before we examine Brockington’s suggestion that the Venn diagram is an alternative to the stemma codicum for representing manuscript relationships, let us first consider the sources of his other theories. Brockington does not provide citations, but his work is clearly indebted to German scholars. The assertion that the editors grouped “the manuscripts mainly by script into versions” recalls Grünendahl’s claim in his 1993 article “Zur Klassifizierung von Mahābhārata- Handschriften.”41 The statement “the premise that manuscripts written in the same script belong together” paraphrases Grünendahl’s “Schriftartprämisse.” Indeed, this is how the term has been translated in English-language summaries of Grünendahl. 42 The suggestion that discussion about details of her analysis, and in particular the varied nature of what constitutes a ring, for the similarity may rest on theme, mode of expression, wording, or merely the attitude of the speaker, while the balancing elements may differ considerably in length. As the author recognizes, the interpretation therefore involves a substantial subjective element. More objective evidence of attention to structure in speeches is provided by the undoubted greater frequency of refrains than in the narrative parts, and also of chiasmus (to which she gives a more extended meaning than I would).” 41 Reinhold Grünendahl, “Zur Klassifizierung von Mahābhārata-Handschriften,” in Studien zur Indologie und Buddhismuskunde, ed. Jens Uwe-Hartmann and Petra Kiefer-Pülz, Indica et Tibetica 22 (Bonn: Indica et Tibetica Verlag, 1993), 101–30. Alternately, Grünendahl may himself owe the error to Brockington, who asserts, “Apart from the Devanāgarī manuscripts, the [Rāmāyaṇa] Critical Edition groups its manuscripts by the script employed but, as I stress in the previous paper, the numbers [sic] of manuscripts in each version varies considerably. The point is equally valid of the Critical Edition of the Mahābhārata […].” John Brockington, “The Text of the Rāmāyaṇa,” Indologica Taurinensia 15–16, Proceedings of the Seventh World Sanskrit Conference, Leiden, August 23–29, 1987 (1989–1990): 80. The error repeats in John Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” Journal of the Asiatic Society of Calcutta 28, no. 3 (1986): 14–24, reprinted in Epic Threads: John Brockington on the Sanskrit Epics, ed. Greg Bailey and Mary Brockington (New Delhi: Oxford University Press, 2000), 195–206 (verbatim on page 197). (All references hereafter to revised the 2000 edition.) We cannot determine who borrowed the thesis from whom, since neither scholar cites the other. Oskar von Hinüber, review of The Sanskrit Epics, by John Brockington, Wiener Zeitschrift für die Kunde Südasiens 46 (2002): 268 notes the pointed omission: “Only very rarely has a title that deserved mention escaped the author’s notice, whose overview of scholarship is current up until the book’s very publication date. This applies, for example, to the important contribution by R. Grünendahl, Zur Klassifizierung von Mahābhārata- Handschriften (in: Studien zur Indologie und Buddhismuskunde. Festgabe ... für Heinz Bechert. [Indica et Tibetica 22]. Bonn 1993, p. 101–130), who with good reasons disputes the connection between the script and recension, which V. S. Sukthankar had already questioned.” The reasons for this silence may have to do with Brockington’s negative review of Grünendahl’s dissertation. See John Brockington, review of Viṣṇudharmāḥ: Precepts for the Worship of Viṣṇu Part 1: Adhyāyas 1–43, by Reinhold Grünendahl, Journal of the Royal Asiatic Society of Great Britain and Ireland 2 (1984): 291– 92. 42 Bigger’s “English Summary” states: “To sort their mss. for the edition, SUKTHANKAR and his followers used what GRÜNENDAHL (1993b) called the premise of scripts (Schriftartenprämisse). GRÜNENDAHL showed that this idea, namely that all mss. written in a certain script belong together, is not tenable, but it may still be useful as a first criterion.” Bigger, Balarāma im Mahābhārata, 170. Mahadevan circumscribes Grünendahl’s notion thus: “the texts of a given script hang together” and “A 19  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches “Sukthankar recognized the impossibility of constructing a stemma codicum on the classical model and relied on grouping the script versions into primarily the northern and southern recensions” reproduces Grünendahl’s claim “on the basis of this theoretical fundament, which I in the following would like to characterize by the term Schriftartprämisse, Sukthankar then groups the manuscripts included for his edition of the Ādiparvan into ‘recensions’ and ‘versions’ in the Prolegomena published in the 7th fascicule (of 1933).”43 In contrast, the latter half of Brockington’s claim that the Mahābhārata editors “classified the manuscript evidence into recensions and script-based versions, relying on what is only a modification of the stemma concept” owes more to Bigger’s notion that “for the editors of the critical edition, the Schriftartenprämisse [sic] replaces the stemma.”44 The notion that the “closeness” between two manuscripts can be “marked by common absence of material rather than additions” recalls Grünendahl’s argument that “the proximity of D2.5 to the K group manifests precisely not in terms of shared textual additions that are present but in terms of textual additions that are not present, […] lends especial weight to the passages mentioned.” 45 Except that Brockington translates “gemeinsam” with “common” rather than “shared” and simplifies Grünendahl’s “nicht vorhandenen Textzusätzen” to “absence of material” the sentences are equivalent. Indeed, like Grünendahl, who argues that the similarities between K2 and K1 are “a case of not present textual additions, […] permits us to once again infer a relationship between these two manuscripts that reaches far back,” 46 Brockington thinks that the circumstance that the Trivandrum manuscript and the Āyodhyakāṇḍa manuscript M4 exhibit “common absence of material rather than additions,” suggests “a fairly early date.” The circumstance that two critical edition (CE) of a work, by definition, is an assemblage of its available texts, anchored on the twin maxims of textual scholarship: brevior lectio praeferenda est (the shortest text is to be preferred) and Schriftartenprämisse (manuscripts [MSS] in the same script belong together.).” T. P. Mahadevan, “The Śakuntalā-Yayāti ‘Transposition’, the Southern Recension of the Mahābhārata, and V. S. Sukthankar,” in The Churning of the Epics and Purāṇas at the 15th World Sanskrit Conference, ed. Simon Brodbeck, Alf Hiltebeitel, and Adam Bowles (New Delhi: Rashtriya Sanskrit Sansthan and D. K. Printworld, forthcoming), 60 and T. P. Mahadevan, “Three Rails of the Mahābhārata Text Tradition,” Journal of Vaishnava Studies 19, no. 2 (2011): 23. 43 Grünendahl, “Zur Klassifizierung von Mahābhārata-Handschriften,” 104. 44 Bigger, Balarāma im Mahābhārata, 118. 45 Grünendahl, “Zur Klassifizierung von Mahābhārata-Handschriften,” 117 (Grünendahl’s italics). See Adluri and Bagchee, Philology and Criticism, chapter 3 for criticisms. 46 Grünendahl, “Zur Klassifizierung von Mahābhārata-Handschriften,” 121 (italics in original). 20  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches manuscripts lack the same passages, however, does not prove them closely related, since they need not owe this absence to an ancestor more recent than the archetype or, indeed, the first source of the tradition. Neither do we need to assume the “influence” of one manuscript on the other to explain this absence. 47 Brockington erroneously thinks that if the Trivandrum manuscript and M4 lack the same passages this means that they derive from a more recent common ancestor than the one they share with the manuscripts containing the passages. But this is false. If the passages are insertions, the Trivandrum manuscript and M4 could be more closely related with the manuscripts that contain them than with each other. The circumstance that they both lack the passages could be coincidental. Only an evaluation of their readings and, more specifically, of their shared errors of transcription will permit us to conclude their filiation. Brockington does not address this issue. Like Grünendahl, he thinks that the circumstance that the two manuscripts lack the same passage justifies him in assuming “a definite alternative tradition current within Kerala,” even though, as yet, he only has two groups of manuscripts— the interpolated and the non-interpolated—and has not shown that the latter constitute a family in the genealogical sense. Whereas the idea that the Bhandarkar editors classified manuscripts by script is owed to Grünendahl, the idea that the critical edition reconstructs “a written redaction of the text that became normative” is borrowed from Bigger. Bigger calls the Mahābhārata critical edition “a redaction that became normative” or “had a normative effect” (normativ gewirkt hat).48 He also declares: “What makes it [the archetype reconstructed in the critical edition] so important is the fact that this redaction came to dominate the whole written tradition of the Sanskrit MBh; it became normative for this part of the transmission.”49 Brockington’s source could be Oskar von Hinüber, who argues that “What has arisen [in the critical edition] is not the Ur-text […] 47 See Adluri and Bagchee, Philology and Criticism, chapter 3. 48 Bigger, Balarāma im Mahābhārata, 15. 49 Andreas Bigger, “The Normative Redaction of the Mahābhārata: Possibilities and Limitations of a Working Hypothesis,” in Stages and Transitions: Temporal and Historical Frameworks in Epic and Purāṇic Literature, ed. Mary Brockington (Zagreb: Croatian Academy of Arts and Sciences, 2002), 20 (italics added). 21  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches but a completely new normalized recension.”50 But since von Hinüber himself draws on Bigger, it is likelier Brockington is quoting Bigger. Other passages suggest a close and careful reading of Bigger’s work. For convenience’s sake, we list Brockington’s comments alongside their probable source: 50 Oskar von Hinüber, “Der vernachlässigte Wortlaut. Die Problematik der Herausgabe buddhistischer Sanskrit-Texte, Zur Überlieferung, Kritik und Edition alter und neuerer Text,” in Kleine Schriften, Teil I, ed. Harry Falk and Walter Slaje (Wiesbaden: Harrassowitz, 2008), 436. 22  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches John Brockington Andreas Bigger “Besides a scribe might have listened to an “Different copyists inserted passages from oral presentation of either epic or perhaps other versions—partly from direct the reading aloud of a text in an area or by an recollection, partly from other (younger?) individual who was using a script he did not written versions—into the transcript.”51 know and he could have incorporated some “These additions were either composed by episodes that remained in his memory into a the scribes themselves, written down from subsequent manuscript copying.” memory (that is, an oral text was written down) or copied from another source.”52 “Attestation in all manuscripts is not “Such passages can be interpreted in two infallible proof of a passage being part of the ways: on one hand, we can assume these original text. It could have entered the passages were very popular and hence tradition early on and then spread to all transmitted across different manuscript extant manuscripts, assisted by this habit of traditions.”53 comparing manuscripts.” “In terms of their content, all these passages are of secondary significance for the Mahābhārata’s further course. We cannot identify any plausible reason why a later redactor should have considered these passages worthy of addition. It appears even less plausible that these passages attained such popularity that they could spread all across India.”54 51 Bigger, Balarāma im Mahābhārata, 111. 52 Ibid., 118. 53 Ibid., 120. 54 Ibid., 121. 23  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches The oral nature of Brockington’s lecture makes it difficult to provide exact references. Most likely, his source is Bigger. At any rate, we find no evidence he developed the idea independently. 55 Regardless of its source, the idea that the critical edition reconstructs a “normative redaction” is false.56 Like Bigger, Brockington finds it convenient to suggest that the “constituted text […] is an approximation to a written redaction of the text that became normative” because this redefinition lets him rehabilitate the hypothesis of a heroic Kṣatriya epic corrupted by Brahmans.57 Like the German critics he draws on, Brockington identifies the 55 Brockington cites Andreas Bigger, “Balarāma im Mahābhārata,” Asiatische Studien/ Études Asiatiques xlviii, no. 1 (1994): 1297–99 with the words: “These episodes are briefly studied by Andreas Bigger (1994), as a foretaste of his thesis on the same subject,” but does not indicate whether he has read this latter work. See Brockington, The Sanskrit Epics, 288, n. 129. Most likely, it had not appeared at the time (both works appeared concurrently in 1998). There is no reference to the Mahābhārata critical edition as a “normative redaction” and the term “redaction” itself appears only sparingly (for example, on pages 21 and 156; on page 71 it is linked to the work of Georg von Simson). Whenever Brockington discovered the idea that the critical edition reconstructs a normative redaction, it must be after 1998. 56 See Adluri and Bagchee, Philology and Criticism, chapter 1. 57 For Brockington’s views of the Mahābhārata’s “brāhman redactors,” see The Sanskrit Epics, 11, 19, 237, and 332. In John Brockington and Mary Brockington, Rāma the Steadfast: An Early Form of the Rāmāyaṇa (London: Penguin Books, 2006), the distinction between the Vedic-Brahmanic and the Epic-Heroic traditions is encapsulated thus: “Transmission of the Rāmāyaṇa and Mahābhārata in their oral stages had been a different process from transmission of the Vedic texts that preceded them, which in traditional circles still retain their original oral form. Knowledge of the Vedic texts was restricted to the brāhmanic class, and the words themselves took on a ritual, even a magic, significance largely unrelated to their semantic meaning. They were handed down verbatim over the centuries without variation, and the reciter’s or hearer’s understanding of them—or lack of it—was immaterial. The case of the Rāmāyaṇa and the Mahābhārata has been completely the opposite. They were heroic tales, narratives conceived as entertainment. Both began, and both continued for several centuries, as the preserve of the warrior kṣatriya class, before being taken over by brāhman redactors.” Ibid., xxiv. And see also ibid., 364: “Around the first to third centuries AD brāhman rather than kṣatriya values came to dominate the text. The details of the process are not entirely clear, but seem to have involved a change to a new set of redactors; probably at this time the text was first committed to writing, although it continued to be present orally.” Most recently, Mislav Ježić has revived the hypothesis. distinguishing “brahmanical, hieratic poetry” from “aristocratic, heroic poetry.” Here is how he expresses their distinction and transformation: “Vedic poetry is brahmanical, hieratic poetry. Formulations about celestials, divinities, world and the powers inherent in it are not narratives, no stories, and are not to be taken literally, but are expressions of a hidden, invisible order, which can be understood as righteous if seen in its truth (ṛta), in which the sages have insight (dhī), and which they express in an appropriately covert, enigmatic manner (brahman, brahmodya). We may term Vedic mantras or brahmans, or formulations in them myths, but we should be aware of the fact that they are not narrations, not even elliptical ones, and that celestials praised in them are not anthropomorphic, have no clear genealogies, and are essentially invisible to those who have no insights into their nature. Epic poetry is aristocratic, heroic poetry at the base, as evident in the case of the Mahābhārata, which can represent a form of archaic court poetry, which may attain a certain poetical refinement close to kāvya, as in the case of 24  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches task of textual criticism with recovering the hypothetical Kṣatriya epic.58 His work consistently seeks to legitimate the hypothesis of an oral, bardic epic.59 He objects that “both critical editions Rāmāyaṇa. It describes individual heroes who are human and mortal, or partly superhuman, but certainly visible, for whom their genealogy is an essential condition of their nobility, whose feats can and should be narrated to glorify their superior virtues, strength and skills. They serve as a model for other members of the aristocracy to follow. This epic poetry, if it reaches out for universal hieratic mythic models, either to describe the relationship of mortals and immortals, or to enhance the glory of heroes by comparing it with that of celestials, or even by elevating them to the level of immortal divinities, may assume a form of religious poetry. This tendency can be reinforced if epic compositions are recited at religious ceremonies, or if epic tradition starts being transmitted by brahmans or priests. The idea that some epic heroes are sons of celestials, or that some of them are even incarnations (avatāra) of God, gave the Mahābhārata and Rāmāyaṇa, and even more their khilas, their religious significance. If hieratic topics, like the origin or creation of the world, its dissolution and renewal, the cosmic periods, and origin of celestials and supernatural persons, become the material for epic poets and singers, they will elaborate these by means of their poetical technique, and give the immortals anthropomorphic features, fixed genealogy, and individual (instead of regular or cyclical) feats, which can be narrated and tend to be understood literally, as mythological facts. They need not be decoded, solved as riddles, but may be taken at their face value, remembered and transmitted (smṛti), in the same way as the genealogies of heroes and (legendary or historical) kings. That is the literary genre of transmitting ancient stories, ancient lore, namely the Purāṇas.” Mislav Ježić, “Vedic Myths and Epico-Purāṇic Mythology—Transformation of the Sense of Vision Into Narration,” paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 12, 2017, IUC, Dubrovnik, Croatia. Insofar as they oppose Vedic literature to the Mahābhārata, all such attempts go back to Adolf Holtzmann, Jr., specifically his claim that: “Alongside this epic literature ran the religious literature of the Brahmans, both completely independent of each other, only occasionally making use of each other. Epic and Veda are equally ancient; no bridge leads from the Veda to the Mahābhārata; “the world of the Veda is a world by itself” […] of course, the epic knows of the Veda, albeit, to conclude from the Mahābhārata, only most superficially; but, on the whole, the two literatures, the Brahmanic and the heroic, are completely independent of each other.” Adolf Holtzmann, Jr. Zur Geschichte und Kritik des Mahābhārata (Kiel: C. F. Haessler, 1892), 61–62. Despite Ježić’s attempt to separate the Mahābhārata from the Veda, itihāsa is a Vedic genre. The Mahābhārata is thus no less “brahmanical” and “hieratic” than the Vedas. The very term “epic poetry” is a misnomer. It has no equivalent in the Indian tradition as several conference participants noted. Its use for the Mahābhārata originated with Christian Lassen who used it in the sense of “the narration [Epos] of the battle of the Kurus and Pândavas.” Only later, once Holtzmann Jr. identified the Mahābhārata with the remains of a common “Indo-Germanic epic” that constituted the “epic genetic inheritance” (episches Erbgut) of the “Indo-Germanic people,” did it enter scholarly discourse as a term for the entire Mahābhārata. Ibid., 42–43. For Lassen’s use of the term, see Christian Lassen, “Beiträge zur Kunde des Indischen Alterthums aus dem Mahâbhârata I: Allgemeines über das Mahâbhârata,” Zeitschrift für die Kunde des Morgenlandes 1 (1837): 61–86. Ježić’s distinction is illegitimate and one of his terms rooted in nineteenth-century ethno-nationalism, as conference participants noted. 58 This level of dependence would usually raise suspicions of plagiarism. In Brockington’s case, it is likelier he lost sight of when he was speaking on the German scholars’ behalf and when in his own name, that is, if a distinction exists. We know of only one instance of demonstrable plagiarism, but here also it is likelier Brockington forgot he was repeating in translation something his German source already said. The passage occurs in Brockington, “The Bhagavadgītā: Text and Context,” 32: “Also, in the visit of the 25  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches are based on the readings of just a selection of the vast number of manuscripts known,” but eliminatio codicum descriptorum is well understood in textual criticism. There is no reason to Pāṇḍavas to the enemy, which follows the Bhagavadgītā, we see a late piece of Brahmin apologetic, which equally with the second visit to Bhīṣma (6.103.40–84) must have been inserted for the moral justification of the Pāṇḍavas.” It is a literal translation of von Simson, “Die Einschaltung der Bhagavadgītā im Bhīṣmaparvan des Mahābhārata,” 174: “Auch in dem auf die Bhagavadgītā folgenden Besuch der Pāṇḍavas im feindlichen Heer (6.41) sehen wir ein spätes Stück brahmanischer Apologetik, das ebenso wie der zweite Besuch bei Bhīṣma (6.103.40–84) zur moralischen Rechtfertigung der Pāṇḍavas eingefügt wurde.” Brockington does not enclose the passage in quotation marks, and no citation is provided. 59 Brockington begins The Sanskrit Epics with this distinction: “It is clear that the Mahābhārata and the Rāmāyaṇa represent the culmination of a lengthy tradition of oral poetry, transmitted through recitation by the sūtas or bards. The oral origins of the Sanskrit epics have, indeed, long been recognised.” Ibid., 3 (italics added). Thereafter, he cites several German sources. And see ibid., 18–19 for similar claims: “No doubt eulogies of heroes and cycles of stories underlie the extant epics; traditionally the Mahābhārata had an earlier and perhaps more clearly epic version in the Bhārata. The first stage in the establishment of an epic tradition would then have been the progressive clustering of ballads and other material around some central theme, whether the personal vicissitudes of a hero or events of particular importance. Eventually, these cycles may have been linked together into works with a more complex plot. The Mahābhārata and Rāmāyaṇa would on this view constitute the end-point of a similar evolutionary process which can be identified, according to some, through a number of intermediate stages. The reciters and preservers of this heroic poetry would be the sūtas or bards, who were attached to the courts of chieftains and who recited in short songs, and on major festivals the glorious deeds of their lords […].” Later he adds: “Although the two extant epics are based on such traditional bardic material, in the course of time more religious and didactic material was included in them, to such an extent that the Mahābhārata in particular has rather lost the character of an epic, since so much of its volume is didactic material. On the face of it, the Mahābhārata does give some indication of its own development. The very setting of the scene suggests the process of expansion in transmission by which a Bhārata lay of 24,000 verses grew to the Mahābhārata of 100,000. Elsewhere there occurs an instance not only of self-reference but also of its narration by brāhmans (brāhmaṇāḥ kathayiṣyanti mahābhāratam āhavam, 5.139.56ab). In the first stage of growth the basic story must have been subject to expansion from within, then in a second stage mythologised and in a third stage entirely taken over by the brāhmanical tradition; finally, after the epic was committed to writing, the number of manuscripts needed for the purpose seems to have become virtually a library of Indian tradition, to which new material of all sorts could be added.” Ibid., 20–21 (except for Sanskrit, all italics ours). The passage quotes van Buitenen near verbatim, who writes: “The original story was in the first phase of complication expanded from within, in the second phase mythologized, in the third phrase brahminized. One might even discern a fourth phase, after the epic was first written down, when this collection of manuscripts became, as it were, a library to which new books could be added.” Van Buitenen, “The Mahābhārata: Introduction,” xxiii. In contrast, the idea of a “Bhārata lay of 24,000 verses” is probably owed to Hopkins’s idea of “Bhārata (Kuru) lays,” allegedly “combined into one, but with no evidence of an epic before 400 B.C.” But van Buitenen also mentions a “Bhārata of 24,000 couplets” and an “old Bhārata lay.” Ibid., xxiii and xxiv. Except that he changes “The Bhārata” to “a Bhārata lay” and replaces “couplets” with “verses,” Brockington’s statement is identical with van Buitenen’s “Thus The Bhārata of 24,000 couplets grew to The Mahābhārata of 100,000.” For the Hopkins reference, see Edward Washburn Hopkins, The Great Epic of India: Character and Origin of the Mahābhārata (New York: Charles Scribner’s Sons, 1901), 397. 26  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches include all manuscripts known if the majority can be eliminated as codices descriptii. Manuscripts that the editor discarded as inferior would only clutter the apparatus. Brockington’s objection would carry weight if he could show that the editors overlooked a significant manuscript, occurring high up in the text’s transmission. But he explicitly negates this possibility.60 He claims that “attestation in all manuscripts is not infallible proof of a passage being part of the original text. It could have entered the tradition early on and then spread to all extant manuscripts, assisted by this habit of comparing manuscripts,” but if all manuscripts agree in their reading, there is no debate between them. Editorial judgment is not needed.61 To print a different reading in the face of their unanimous testimony is to enter the realm of a subjective ars critica scarcely worth the name. Brockington argues that “the text even of the critical editions [of the Rāmāyaṇa and the Mahābhārata] cannot be regarded as a completely coherent whole,” but this is to misunderstand what a critical edition can achieve. The circumstance that the constituted text contains “multiple examples of interpolations and internal contradictions” is not an argument against it.62 Contrary to Brockington’s belief, we 60 Brockington, “The Text of the Rāmāyaṇa,” 79–80: “These eleven manuscripts add well over a third to the number used for the Critical Edition of the Ayodhyākāṇḍa (and more than a quarter even to those that were initially collated). It is not surprising, therefore, that together they add substantially to the body of variant readings recorded in the critical apparatus. This material is significant, even if of no great value in itself, for building up a truer picture of the complex relationship between the different recensions and versions.” (Italics added.) 61 What principle should we follow if not the manuscripts’ univocal testimony? If they are obviously corrupt, the editor may propose a conjecture, though this situation will be rare, since a problematic reading is unlikely to be conserved in all manuscripts (Contini: diffraction). Sukthankar therefore rightly preferred documented readings (Sukthankar, “Prolegomena,” xcii: “Emendation has played a very inconspicuous role in the preparation of the constituted text. Interpretation has in general been given preference over emendation. Even in the case of corrupt passages, the reading of some manuscript or other gives sense, though it may not be the original sense, not even a wholly satisfactory sense. Precipitate emendation is, however, to be deprecated; for experience has shown that but a small proportion of scholars’ corrections are really amendments.”). 62 The argument recalls Pradip Bhattacharya’s: “However, no attempt appears to have been made to resolve issues of repetition and inconsistency. One would expect that in the 20th century scholars would seek to make good these lacunae. Instead, the CE contains contradictions that are the result of rigidly observing the principle that whatever appeared in the largest number of manuscripts had to be included, irrespective of intra-textual consistency which was argued to be ‘the realm of higher criticism’.” Pradip Bhattacharya, “Problems with the Critical Edition: the Attempted Disrobing of Draupadī,” paper presented at the 15th World Sanskrit Conference, New Delhi, January 5–10, 2012. In this case, it is not easy to establish an obvious dependency, and probably both arrived at it independently. But the shared error reflects a prevailing confusion about critical editions. Despite many excellent technical handbooks, many still think a critical edition is an edition that has been criticized. They are unaware that critical 27  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches cannot arrive at a more original text merely by purging the constituted text of perceived contradictions. Mehendale’s identification of “contradictions” does not offer an objective basis for emending the text.63 Like his predecessor, Brockington would only arrive at a modernized version of the text reflecting contemporary prejudices.64 edition refers to an edition based on an exhaustive survey of the manuscripts that uses clearly enunciated principles and simple rules to reconstruct a definite state of the text. The term is frequently applied to a Lachmannian or Neo-Lachmannian edition and hence as a synonym for the genealogical-reconstructive edition. Brockington’s and Bhattacharya’s criticisms are irrelevant because Sukthankar anticipated the charge, and warned against expecting more from the edition than it could provide: “The reader will find that the constituted text is by no means smooth. It contains fresh instances of loose and archaic linguistic forms and constructions, anacoluthons and lack of syntactical concord. There remain many contradictions and superfluities. There is evident lack of finish in the hidden parts. These blemishes—if they be blemishes in epic poetry, which is dynamic poetry, with no necessary pretensions to niceties of style, in the narrower sense of the term—must have been inherent in the old poem. Where they are met with in the critical text, they are not speculative fiction; they are documented by the manuscripts themselves or at least are inferable from them with a high degree of probability.” Sukthankar, “Prolegomena,” ciii–civ. Like Mehendale and Bhattacharya, what Brockington seeks is not a critical edition, but an emended edition meeting his expectations of simplicity and consistency. From a critic’s perspective, this is the well-known tendency of banalization. 63 The reference to Mehendale does little to bolster Brockington’s claims. Brockington likely cites him because Mehendale, like him, proceeds from the assumption that the Mahābhārata “was orally transmitted for many centuries.” As we saw, Mehendale divides interpolations in the Mahābhārata into three categories. Insofar as Brockington refers to the constituted text, he cannot mean the first category. Insofar as his claim concerns the second category, the manuscript evidence is unambiguous and no Lachmannian or neo-Lachmannian editor would emend the text against the manuscripts’ uniform testimony. This leaves the third possibility—that Brockington, like Mehendale, thinks the critical edition does not provide a facsimile of the “original” epic, because of its perceived inconsistencies. This suggests that Brockington’s work aims, rather, at a subjective Konjekturalkritik. Insofar as he thinks this continues the editors’ work he commits the same error as Mehendale earlier when he suggested his work contributes to Mahābhārata textual criticism. 64 See Adluri and Bagchee, The Nay Science, 153: “In spite of Fitzgerald’s and McGrath’s efforts, however, the Holtzmannian Ur-epic remained elusive. As far back as they pushed this ideal epic—and in his most recent iteration McGrath has advanced a date of the Bronze Age—they were still unable to find evidence of it and for the simple reason that its origins lay neither in the fourth century CE nor in the fourth century BCE nor even in the twelfth century BCE, but in the nineteenth century CE when the thesis was first proposed by Lassen. The idea of an original epic, as we have seen, was a specifically German notion answering to specifically German needs. German scholars had deployed the idea in pursuit of an ideal of a heroic Aryan race. It permitted them to make polemical points against Roman Catholicism as being essentially alien to the spirit of the German people. It permitted them to identify defining traits of the ‘Indo-Germanic’ or ‘Aryan’ peoples—nobility, free-spiritedness, suspicion of priestly authority, a warrior culture, and so on—in contradistinction to the ‘Semitic’ tribes. It permitted them to undertake a sustained critique of non-Germanic cultures as well as to issue stern warnings of what might befall the German nation if it permitted itself to be overrun by these cultures. Beyond these ends, the theory of a heroic epic had no relevance to Mahābhārata criticism.” 28  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Form Criticism Having considered Brockington’s sources, let us address his other contribution to Mahābhārata studies: the suggestion that the “Venn diagram” is an alternative to the stemma codicum. Before we look at his suggestion, let us briefly review the concept of a Venn diagram. A Venn diagram, named after its inventor John Venn, is used to express the logical relationship between classes or sets of objects.65 The diagram uses closed plane figures to denote sets, while the region they enclose stands for the sets’ members or elements. Sets can be given either an intensional definition (for example, A is the set of the letters of the alphabet) or an extensional definition (in which case, the convention is to list its members within braces; thus: A = {a, b, c, d…}). Once the sets under discussion are defined, the Venn diagram can be used to provide a graphical representation of all possible logical relations between them (for instance, their intersection, union, difference, and complement). Let us see how Brockington introduces the suggestion: Thirty years ago, I suggested that the model of the Venn diagram could be borrowed from mathematics and logic as it had already been borrowed into linguistics. This uses circles or ellipses overlapping as necessary to represent a relationship between separate groups or sets of 65 This reference is essential. It comprises Venn’s main innovation over the earlier Euler diagram. Edwards notes: “When such pictorial representation [the Euler diagram] is extended to more complex sets of logical propositions, however, difficulties set in quickly. For example, a logician faced with a series of propositions might wish to ascertain whether they were mutually consistent, and Euler diagrams do not provide a graphical ‘algorithm’ for settling such a question. Venn’s own description of the impasse, in 1880, is conclusive: ‘… we cannot readily break up a complicated problem into successive steps which can be taken independently. We have, in fact, to solve the problem first, by determining what are the actual mutual relations of the classes involved, and then to draw the circles representing this final result; we cannot work step-by-step towards the conclusion by aid of our figures.’ Venn had a better idea […].” A. W. F. Edwards, Cogwheels of the Mind: The Story of Venn Diagrams (Baltimore and London: Johns Hopkins University Press, 2004), 3–4. Venn himself explains his advance over Euler thus: “Whereas the Eulerian plan endeavoured at once and directly to represent propositions, or relations of class terms to one another, we shall find it best to begin by representing only classes, and then proceed to modify these in some way to make them indicate what our propositions have to say. How, then, shall we represent all the subclasses which two or more class terms can produce? Bear in mind that what we have to indicate is the successive duplication of the number of subdivisions produced by the introduction of each successive term, and we shall see our way to a very important departure from the Eulerian conception. All that we have to do is to draw our figures, say circles, so that each successive one which we introduce shall intersect once, and once only, all the subdivisions already existing, and then we have what may be called a general framework indicating every possible combination producible by the given class terms.” John Venn, “On the Diagrammatic and Mechanical Representation of Propositions and Reasonings,” cited in Edwards, Cogwheels of the Mind, 4. There is no Venn diagram not based on a prior establishment or analysis of sets. This is the greatest problem with Brockington’s work: he does not conceptualize much less define his sets. 29  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches items. In the case of the manuscripts I was studying this means that a circle or ellipse representing M4 and the Trivandrum manuscript will partly overlap with others for the other Malayālam manuscripts with D1 and [D]2 and V1, in this way showing to some extent at least the multiple allegiances involved.66 Since these comments are not very illuminative, let us look at his earlier reference: Perhaps I should at this point remark that the Critical Edition’s system of assigning manuscripts to recensions or to script versions is nearly as limiting as the construction of a stemma codicum (which does have the merit of corresponding on the whole to the way that manuscripts are propagated). The drawback is that it still involves an either/or situation, where a particular manuscript is assigned to a particular recension or version, or else is dismissed as ‘contaminated’ or a ‘Mischcodex’, descriptions which seem almost to have become terms of opprobrium. Tentatively, I would like to propose that a better model might be the Venn diagram (which has already been borrowed from mathematics into linguistics to considerable advantage; see the sketch overleaf). Thus V1, which I was last considering, might be represented by a circle or oval mainly overlapping with one representing the Ñ and B mss but also partly overlapping with one representing the NW and W recensions. To a certain extent this is a reversion to the stemma codicum in that it attempts to depict the different influences or originals which have formed a specific manuscript, while avoiding the impossible task of constructing the entire chain of transmission involved.67 66 Brockington, “Regions and Recensions, Scripts and Manuscripts.” Brockington does not provide a citation for the use of Venn diagrams in linguistics, but all the examples we examined had two features in common: (1) they defined the members of the sets under consideration; (2) they employed Venn diagrams to depict either the extent of shared features between two languages or to classify certain types of statements. Neither feature applies in Brockington’s case. 67 Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” 203–204. The proposal was probably announced earlier at the 23rd International Congress for Asian and North African Studies in Hamburg, as the following abstract records: “Of the varied manuscripts examined, the most significant is ms. 14052 (unfortunately containing only half the Ayodhyākāṇḍa) for it is closely related to M4 and probably of similar age. Their closeness is marked by omissions—more exactly absence of material— rather than additions, indicating relative antiquity. However, omissions unique to each show that, despite their regular shared readings, neither could have been copied from the other, nor probably from an immediate ancestor. […] Such details suggest that the CE scheme of recensions and subrecensions is too limiting. I propose as a possible model the Venn diagram, which can better represent overlapping influences and avoids a straight either/or classification.” John Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” Zeitschrift der Deutschen Morgenländischen Gesellschaft, Supplement IX, Proceedings of the XXXII International Congress for Asian and North African Studies, Hamburg, 25th– 30th August 1986 (1992): 509–10. Brockington also refers to the Venn diagram in his latest published work: “Elsewhere I have proposed that a better model than either the Critical Edition’s system of script versions or a stemma codicum would be the Venn diagram, which with its overlapping circles or ovals 30  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches As an example, he presents the following diagram: Figure 1. Brockington’s “Venn diagram.” Note: Reproduced from Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” 204. The same diagram occurs in the handout accompanying Brockington, “Regions and Recensions, Scripts and Manuscripts.” Brockington’s label is misleading since the diagram does not depict manuscript relationships, but only common members (and only partially so, since he bases it on prominent features like interpolations). Moreover, the graphic is not a Venn diagram. Brockington neither defines the sets nor populates them with members (a problem he now acknowledges). Although the diagram is not a Venn diagram, in the following we refer to it as one and only gradually introduce corrections to bring it closer to a real Venn diagram. can better indicate degrees of commonality between various manuscripts, as well as being capable of arrangement to a certain extent in a form reminiscent of geographical relationships.” John Brockington, “Is the Script Relevant? Further Evidence from a Nevārī-Script Rāmāyaṇa Manuscript,” in Epic and Argument in Sanskrit Literary History. Essays in Honor of Robert P. Goldman, ed. Sheldon Pollock (Delhi: Manohar, 2010), 24. 31  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches There are several problems with Brockington’s proposal. Let us first consider his reasons for thinking a stemma codicum is unachievable for the Mahābhārata and then evaluate his misconceptions about the Venn diagram. 1. Manuscript Filiation Brockington erroneously links manuscript filiation with “alignment” and, from the observation that the manuscripts’ “alignment” changes, concludes that a stemma codicum is unachievable.68 Both assertions are false. Agreement in truth is not evidence of filiation and so a change in the manuscripts’ agreements should not prevent us from identifying their true filiation, which will reveal itself in a small albeit informative number of significant errors. Brockington is misled because he focuses only on selected passages, above all, obvious features like additions and omissions, which just because of their itinerant nature will mislead us as to the manuscripts’ filiation.69 Moreover, the promiscuity of more recent and inferior manuscripts should not 68 Compare Brockington, “Regions and Recensions, Scripts and Manuscripts”: “Within the northern recension, they align particularly with two manuscripts D1 and [D]2 which are usually assigned to the western subrecension within the northern recension and to a lesser extent with a third V1.” Compare also Bigger, Balarāma im Mahābhārata, 113 and 133, n. 88. For Grünendahl’s use of the term (sich anschließen), see Grünendahl, “Zur Klassifizierung von Mahābhārata-Handschriften,” 126: “Additionally, there are nine passages where, according to the list all the Ñ manuscripts feature the reading of γ, but, according to the critical apparatus, individual manuscripts align themselves with the reading of ν and the southern recension.” 69 The following is typical: “Within this same span of the first thirty sargas, in addition to its unique omissions, the Trivandrum manuscript omits two other passages found in M4 but omitted by some other manuscripts; the second of these is interesting, since it shows the Trivandrum manuscript diverging from M4 on one of the few occasions when M4 goes with the S recension in these first thirty sargas. After sarga 31 (its own sarga 34) M4 shows a greater tendency to side with the S recension, and so in general does the Trivandrum manuscript, but it still omits some S material occurring in M4 (such as 42.25cd + 1047*, and 1107*1) and includes some N material lacking in M4 (such as 1097*3–4 + 1098*1–2). Equally, though, it includes some S material absent from M4 (for example, 50.11 and 1200*), so it is not possible to simply to say that it shows a greater tendency than M4 to retain the affinity with the N recension. Indeed, there are two instances where it seems to give even clearer evidence than M4 of combining both N and S traditions. These are at the start of sarga 50, where it has both 1217* (read by many N mss and M4) and 1218* (read by S mss, including some Devanāgarī mss), and at 54.16cd where it has both the text (representing the S version) and the N variant successively. Moreover, at 50.11, although its reading is basically that of the text (= S), it has a unique variant reading in the first pāda which is similar to the N substitute (1197*1 pr.).” Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” 202–203. 32  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches prevent us from fitting the older and better ones into a stemma.70 Brockington goes astray because, rather than study the entire tradition or individual manuscripts within the context of the larger tradition, he focuses on late and inferior specimens, and uses their recidivism to claim that, as their filiation cannot be satisfactorily determined, the very concept of a stemma codicum must be abandoned.71 He evidently confuses the stemma codicum with the “real tree,” that is, “the unknowable, but certainly much larger and more complex ensemble of all copies that ever existed.”72 From the circumstance that a genealogical tree in this sense is unattainable, he falsely concludes that we should use Venn diagrams, instead, to represent “the multiple 70 As West also notes: M. L. West, Textual Criticism and Editorial Technique (Stuttgart: B. G. Teubner, 1973), 14–15. 71 Brockington’s approach is peculiar. He writes: “In view of the large numbers of Rāmāyaṇa manuscripts preserved in various collections, and in particular of their relative lateness (and therefore the impracticality of establishing a stemma codicum), the number of manuscripts used for the Critical Edition is really quite small. The largest number—41—was used for the Uttarakāṇḍa and the smallest— 29—was for the Ayodhyākāṇḍa and also for the Āraṇya and Sundara kāṇḍas, although admittedly a larger number was initially collated (e.g. 43 for Ayodhyā). It is not surprising, therefore, that variant readings and new patterns of omission or addition can be found in virtually every new manuscript examined. I would argue that this material is significant if we are to build up a truer picture of the complex relationships between the different recensions and versions.” Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” 197. Compare also John Brockington, “A Malayāḷam-Script Rāmāyaṇa Manuscript,” Indologica Taurinensia 21–22 (1995–1996): 79: “However, the Critical Edition necessarily utilised a relatively small number of manuscripts (29 in the case of the Ayodhyākāṇḍa) and so variant readings and new patterns of omission or addition can be found in virtually every new manuscript examined.” The late manuscripts probably deserve classification as inutiles. Even if they contain good readings, their lateness makes it unlikely they will affect the stemma. Brockington has not grasped the principle of eliminatio codicum descriptorum, which Trovato elucidates as follows: “In operative terms, if all the sources of a contaminated ms. or of a subfamily of contaminated mss. are preserved, there is no need to use the contaminated ms. or subfamily, which can be eliminated as codices descripti, or rather inutiles.” The sole exception to this rule is: “If, instead, a ms. that descends to some extent from a known exemplar contains high-quality variants not found in any known ms. or subfamily, we shall need to assume that ‘extra-stemmatic’ (Timpanaro) or rather ‘extra-archetypal’ contamination has occurred.” Trovato, Everything You Always Wanted to Know about Lachmann’s Method, 134. But here also, rather than alter the stemmatic relationships of older family members, this is more likely to trigger a revision in our judgment about (the utility of) late and inferior specimens. Moreover, the circumstance that “new patterns of omission or addition can be found in virtually every new manuscript examined” does not negate the relationships already established, since they were not established using “patterns of omission or addition.” Possibly, what Brockington means is that some discarded manuscripts contained better readings due to their scribes’ access to good, old sources. But he does not argue from the lectio difficilior. He merely claims that several lower-value manuscripts exist. This is like saying we should not establish family trees for blue-bloods because we cannot establish the parentage of the bastard children of prostitutes. 72 Trovato, Everything You Always Wanted to Know about Lachmann’s Method, 80. 33  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches allegiances involved.”73 Not only does this misunderstand the stemma codicum’s function (to depict, as unambiguously as possible, a manuscript’s descent from others, either surviving or hypothesized as having existed).74 It also overlooks that a Venn diagram cannot replace the stemma as it contains no genealogically relevant information. From Brockington’s diagram, we could establish neither the ancestors of the manuscripts listed nor the part of their shared text owed to a common ancestor. Indeed, there is no place to locate an ancestor: unlike the stemma codicum, which, although two-dimensional, uses the illusion of a third plane to convey the impression of height and thus temporal succession, a Venn diagram remains trapped within the plane of the paper. Brockington depicts the Northwestern and Western tradition by one circle, the Northeastern by another, and the Southern by a third. But this is surely false, since the circles do not intersect and the three groups of manuscripts do not contain completely different texts.75 If the diagram is to have any sense, the intersecting regions must represent readings shared between manuscripts and hence their inherited text. But which regions? In Brockington’s diagram, some manuscripts are completely unrelated to the others, which must mean they are not Mahābhārata manuscripts at 73 Brockington, “Regions and Recensions, Scripts and Manuscripts.” The reference is to Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,”, specifically the claim that “to a certain extent this is a reversion to the stemma codicum in that it attempts to depict the different influences or originals which have formed a specific manuscript, while avoiding the impossible task of constructing the entire chain of transmission involved.” Ibid., 204. This is a novel interpretation. A stemma codicum would only “depict the different influences or originals which have formed a specific manuscript” if every manuscript had lines not only diverging from it but also converging on it, that is to say, if every manuscript was copied not from one exemplar, but several. Though not impossible, this is an unusual interpretation of the stemma, since genealogical reconstruction under these conditions is impossible. 74 If he wanted, an editor could draw dotted lines converging on manuscripts to depict the direction and extent of contamination (for example, by varying the lines’ thickness or affixing numbers). Belvalkar actually undertook this. But besides mapping the extent of contamination, his stemma did not provide new insights. It did not affect the relationships already established. From an editor’s perspective, showing “multiple allegiances” is less relevant than establishing the manuscripts’ descent as unambiguously as possible. The tradition may, of course, be so hopelessly contaminated that we cannot reconstruct an archetype. In this case another solution like editing the best manuscript must be pursued. But this is not Brockington’s contention. He does not recommend another editorial strategy. He defends the critical editions as “on the whole the best we have.” Brockington probably emphasizes the “multiple allegiances involved” because he has not collated the manuscripts himself and does not seek to provide an edition, critical or otherwise. 75 But see figure 5 later, for what we think Brockington is really trying to express. Here also the circles must intersect even though they have no common members. A Venn diagram must indicate all possible logical combinations of sets. 34  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches all. Even discounting these outliers, between two intersecting regions, which one indicates the manuscripts’ inherited text? Both cannot be original. If M4 ∩ V1 ∩ Ñ ∩B represents their common text and M4 ∩ D1.2 ∩ V1 represents their common text, one of the agreements must be secondary. More precisely, neither one can be completely original. Both intersecting regions will include verses or readings that are shared because they are inherited unchanged from the archetype and others that are shared either because they are owed to a more recent common ancestor than the archetype or because they were transmitted horizontally by contamination. Brockington must disambiguate between the part of M4 ∩ V1 ∩ Ñ ∩B owed to the archetype, that is, is a subset of θ (since not all of θ’s reading will be transmitted unchanged to its descendants) and the part owed to later sources; and likewise for M4 ∩ D1.2 ∩ V1. But since this is the issue at stake, the Venn diagram cannot help him. Manuscripts like M1–3 must also share in the “archetypal” inheritance. Brockington has not mapped the common text of all manuscripts, but only selective affinities based on prominent characteristics.76 2. The Venn Diagram Even if we grant Brockington that “variant readings and new patterns of omission or addition are found in virtually every manuscript examined” and they permit us to “build […] up a truer picture of the complex relationship between the various recensions and versions,” the problems 76 We did not find a single example of complete collations in the articles we examined. To our knowledge, Brockington has never edited a manuscript. The closest he comes to providing collations is in Brockington, “A Malayāḷam-Script Rāmāyaṇa Manuscript,” 77–91, but even this does not amount to a complete collation. Although he provides a detailed account of variants for the first twenty sargas, he then focuses on “major features,” especially additions and omissions. The following passage is typical: “The Trivandrum manuscript agrees with D1.2 M4 in, for example, omitting—as well as 9.14ef, 18–19, 26, 43, 10.6cd, 21–25, 13.25, 18.19cd (not D2) and 40 (not D2)—20.23c–34d, App. I 12.35–39, 58, 80–81 and 24.16, in inserting 168*, 203* and 362* (also in B2.4), in reading 620*4 after 632*, in the sequence of stanzas at 25.4–14, and in very many common readings elsewhere. It is still more interesting to note the frequency with which it agrees with D1 or D1.2 only against M4. It agrees with D1 only but not M4 at 17.24b, 451*3, 452*5, the insertion of 469* (also in D5), 489*19–20 (but not 37–40), 508*1 and 570*2; it agrees with D1.2 but not M4 at 281*2, 295*3 (also in D5), 320*3, 14.18c, App. I 12.80, 623*2 and 658*1. These are certainly not enough to indicate a link between this manuscript and these two of the supposed Western recension manuscripts to the exclusion of M4, but it does suggest that D1.2 are aligned in some way with this alternative Malayāḷam recension of M4 and ms. 14052, which is still more obvious when we note the common readings, insertions and omissions of these four manuscripts. There are also traces of links with V1 either of all four manuscripts or of the two Malayāḷam-script ones.” Ibid., 88–89. 35  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches with his “Venn diagram” remain. For starters, what is his set? If M4, V1, etc. denote sets whose members are the lines or words contained in the respective manuscripts, how does he account for repetition of lines or words given that every element of a set must be unique and that the order of elements is irrelevant? He notes that the Mahābhārata contains many repetitions (he calls them “formulaic expressions”).77 If we define M4 as the set of lines contained in (the manuscript) M4, Brockington must first eliminate all repetitions from the manuscript. To understand the implications, consider the following account from The Sanskrit Epics: Pāda-length phrases comprising a personal name and an epithet are the commonest formulaic expressions; however, they are not used just at random but fulfil a definite narrative function. The other main types are introductions and conclusions to speeches, various verbal formulae expressing emotion or emphasis, certain descriptive and hyperbolic phrases, stock expressions for battle scenes, phrases of time, place and number, proverbs and similar expressions, and stereotyped similes. Such formulae occur mainly in the second and fourth pādas, except those used after the end of speeches, which for obvious reasons usually occur in the first (or less commonly the third) pāda; the metrical pattern of the śloka means, however, that there are regularly different sets of formulae for the odd pādas from those for the even pādas. It is also noteworthy how often a phrase or passage is repeated within a short space of its first occurrence. In several instances this results from the exact wording of a message being repeated or some event already described being narrated to another character. By contrast, the use of refrains, which occurs mainly in speeches, has a definite emphatic purpose, as has parallelism within the verse, whereas another type of repetitiveness typifies certain didactic passages.78 Brockington cannot preserve any of the elements that, in his view, demonstrate the Mahābhārata’s “oral” origins. A line that repeats in the first adhyāya and the sixth may be listed only once, even if he considers the second occurrence significant for the manuscripts’ “alignment.” He could partially circumvent the problem by defining the verse as the smallest unit of analysis. He can now retain these lines despite their repetition (though identical verses must still be eliminated).79 But what he gains in material for comparison (that is, the number of 77 See Brockington, The Sanskrit Epics, 103–111. He estimates that “the proportion [of formulaic expressions] for the Mahābhārata is at least as great as in the Rāmāyaṇa where around one-third of the ślokas contain significant formulaic material.” Ibid., 103. 78 Ibid., 105–106. 79 Brockington’s calculations show that the likelihood of repetition decreases as we move away from the pāda. But this also applies across manuscripts, triggering the paradox we encountered earlier. “In my 36  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches elements in the set) he loses in accuracy, since he can no longer consider partial matches. For instance, if M4 and V1 contain a verse with an identical first pāda, he cannot consider this pāda common to both sets, since their members are the complete verses only. Inevitably, this eliminates the majority of genealogically informative features like shared significant errors. A smaller semantic unit than the complete verse thus appears preferable. But consider what happens when we define the members of the sets as the words contained in the manuscripts. Except for their first occurrence, we must now eliminate all frequently occurring words, including particles, proper nouns, and common verbs. After elimination, we could identify elements common to M4 and V1. But not every shared word is genealogically significant. Many shared words will be commonplace names or terms. Even if M4 and V1 deviate from the other manuscripts in that they share not only common names but also words not found in the other manuscripts, this may not matter. The common elements could occur in different parts. Unlike a sequence, the order of elements in a set is irrelevant. Thus, the set M4 = {a, b, c, d, e} is identical with the set V1 = {d, e, c, a, b}, which can be rewritten as {a, b, c, d, e}. From a genealogical perspective, however, it is the occurrence of c in the same position in V1 as in M4 that is significant. Set notation cannot preserve this distinction and so we would consider V1 identical with Ñ = {a, b, c, d, e}, even though its text is closer to M4’s than Ñ’s, which contains the same elements in a different order.80 article on stereotyped expressions (1970: 201–11) I calculated on the basis of the Ayodhyā to Kiṣkindhā kāṇḍas that: ‘The average proportion [of stereotyped pādas] over all three kāṇḍas is about 1 in 22 or 4.5%, which means that one in eleven lines or one in five or six stanzas in fact contains a full pāda found in identical wording elsewhere. But if all instances where the verbal similarity is less exact and where the resemblance extends over less than a pāda are included, then between 30% and 40% of all śloka stanzas contain some stereotyped material. Though usually printed as two lines, in origin—and still to a large extent in the epics—the śloka actually consists of four pādas of eight syllables. In my usage, a formulate consists of a full pāda that is completely fixed and a formulaic element is one where at least five syllables and two words are involved. Calculations on a different basis, such as those of Grintser (1974), give figures of around 80% of battle scenes consisting of formulae and formulaic expressions and 40–50% of narrative chapters.” Ibid., 103, n. 30. 80 Ignoring the difficulty that agreement in the truth is not evidence of filiation, since Brockington is not arguing in terms of filiation, but merely the extent of shared text. 37  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Brockington had a clarification for his diagram at the recent DICSEP 8.81 In a session on Venn diagrams as a tool of analysis, he explained that when he proposed the Venn diagram, it was “obvious” that the members of the sets were the manuscripts themselves. Thus, the set M4 contains the manuscript M4, V1 contains the manuscript V1, and so on. We must confess, this possibility did not strike us. If this is all Brockington means some of our criticisms are moot. But this explanation does not cohere with his diagram. Let us first address a basic problem. If the set M4 contains the manuscript M4, V1 contains the manuscript V1, and so on, no circle should overlap. M4 contains only M4 and hence no elements in common with V1 (which contains only V1). Their intersection is the empty set: M4 ∩ V1 = { } or ∅. This also holds for the intersection of the remaining sets. Our diagram would look like this: Figure 2. A “Venn diagram” depicting sets, whose members are individual manuscripts. Actually, the circles should all overlap, because the Venn diagram must indicate all possible logical combinations of the sets, even if the containing intersections are empty (in which case, they can be shaded to indicate that they have no members). Not only is this extremely difficult in the case of ten sets; we also wish to remain with Brockington’s understanding of “Venn 81 Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 11–16, 2017, IUC, Dubrovnik, Croatia. All references hereafter are to Joydeep Bagchee’s electronic recording. 38  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches diagrams” as we explore its problems, while gradually proceeding towards a real Venn diagram.82 But even as drawn, Brockington’s explanation does not quite fit the diagram. The set D1.2 contains two manuscripts (D1 and D2). The set M4+14052 likewise contain two manuscripts (M4 and 14052), whereas the set M1–3 contains three manuscripts (M1, M2, and M3). These three sets differ quantitatively from the remainder. Although this may not make a difference, since the intersection of D1.2 with any other set is also the empty set (and likewise for M4+14052 and M1–3), it does raise the question: why are these seven manuscripts grouped into sets? Do they have a common property? If not, and we interpret Brockington’s statement that each set contains only the manuscript it is named for strictly, our diagram should appear thus: Figure 3. The corrected “Venn diagram.” This diagram is more accurate, but it does not solve the problem. If the sets’ members are the individual manuscripts themselves, what about NW+W, NE, and S, given that no manuscripts correspond to them? Either NW+W, NE, and S are empty sets, in which case we can eliminate them, or we must define them differently. If they mean anything, NW+W must denote the set of all sets that contain manuscripts belonging to the Northwestern and Western group, NE 82 Higher order Venn diagrams exist and there is an entire field dedicated to studying (and drawing) these formations. But Brockington has plainly not studied them. We could move to a real, founded discussion of Venn diagrams, but this would not serve the purpose of exploring Brockington’s ideas of textual criticism. 39  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches must denote the set of all sets that contain manuscripts belonging to the Northeastern group, whereas S must denote the set of all sets that contain manuscripts belonging to the Southern group. Our diagram will look like this: Figure 4. A “Venn diagram” of sets of sets that contain manuscripts of the S and NE groups. NW+W remains an empty set. But we can now show that whereas NE contains six members (NE = {Ñ, D4, V1, B, D1, D2}), S contains five members (S = {M4, 14052, M1, M2, M3}). We can, moreover, make logical inferences, for example, that NE and S are disjoint sets (they have no elements in common). But what do we gain by defining NE as the set of all sets that contain manuscripts belonging to the Northeastern group and likewise S as the set of all sets that contain manuscripts belonging to the Southern group? It is senseless to place each manuscript in a set that contains no other manuscript and then group those sets into other, larger sets.83 A more intuitive definition is that NE is the set of all manuscripts belonging to the Northeastern group, whereas S is the set of all manuscripts belonging to the Southern group. We can update our diagram as follows: 83 Brockington has his reasons for framing the requirement thus. He needs M4, V1, etc. to be members of the sets containing them but, simultaneously, also sets capable of containing other members. This reflects a fundamental incoherence in his mind about what should be the class and what its members. 40  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Figure 5. A “Venn diagram” of sets containing manuscripts of the S and NE groups. This diagram corresponds more to a standard Venn diagram and its uses.84 But it no longer maps the resemblances or shared text between manuscripts. Rather, it represents the grouping of manuscripts after their analysis into groups, that is to say, after we have determined that Ñ, V1, B, D4, D1, and D2 belong to the NE group, and so on. Most important, it does not “show […] the multiple allegiances involved,”85 since every manuscript either belongs to the set NE or does not belong to it; and likewise for S. Brockington may object that, actually, his point was that certain manuscripts have features of the Northeastern and Southern traditions, and hence should be considered members of both. Similarly, D1 and D2 in his original diagram represent manuscripts occurring halfway between the Northwestern and Northeastern traditions, rather than, as we have it, to the Northeastern group alone. He may therefore propose returning to a modified version of his original diagram: 84 It is not, of course, a Venn diagram, which must contain all possible regions even if empty. Like the previous examples, this is actually a Euler diagram. 85 Brockington, “Regions and Recensions, Scripts and Manuscripts.” 41  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Figure 6. A modified version of Brockington’s original “Venn diagram.” This is a valid objection, but the correct way to represent this situation is as follows: Figure 7. The corrected “Venn diagram.” 42  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches This is a “true” Venn diagram in contrast to Brockington’s.86 It shows that the sets NW+W and NE and S and NE each have two common members, whereas NW+W and S have no common members. Expressed in set notation, we have: NW+W ∩ NE = {D1, D2}; S ∩ NE = {M4, 14052}; and NW+W ∩ S = { } or ∅. But contrary to Brockington’s claim, this diagram does not let us analyze the relations between manuscripts. It only lets us represent their relationships after analysis, that is, after we establish the universal set (the set of manuscripts to be distributed) and determine which manuscripts belong in which of our finite sets NW+W, NE, and S. Although we can now see at a glance which manuscripts are common to the sets, its advantages are not immediately apparent. We can neither express our reasons for assigning a manuscripts to two sets, for instance, that M4 has features of the S and NE groups, nor can we express the extent to which it does so, for instance, that M4 is a composite of 40% of S and 60% of NE. In set theory, an element is either a member of a set or not a member of a set. Venn diagrams neither obey cardinality (their size does not indicate the number of members in the set) nor can they be used to express affinities between members (for instance, by placing certain members nearer others), whereas Brockington’s original diagram tried to show precisely this, for instance, by displacing one manuscript towards another or by making some circles larger than others. But this is not a valid use of the Venn diagram. A Venn diagram is a schematic, that is, abstract diagram. The size, shape, and position of circles or the location of elements within those circles, beyond the mere fact of intersecting or containing common elements, is irrelevant to their relationships. The real problem, however, is neither the Venn diagram nor that it cannot show what Brockington wishes it to. It is that Brockington himself has not thought through what he wishes to show. Is it the manuscripts’ belonging to the regional groups? In that case, the groups are the sets and the manuscripts their members. Or is it the text shared between manuscripts? In that case, the manuscripts are the sets and their readings (words, lines, or verses) the members. Without clarity on this fundamental issue, Brockington cannot make headway. If we now return to his original diagram (figure 1), we shall see that all his problems stem from a 86 The regions NW+W ∩ S and NW+W ∩ NE ∩ S are missing, so this is not yet a true Venn diagram. But we wanted to make the minimum number of changes possible as we proceed stepwise from Brockington’s “Venn diagram” towards the real thing. 43  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches failure to conceptualize what is the class and what is the member. On one hand, the diagram shows manuscripts contained either in one of the three regional groups or spanning the region between two groups. On the other, it shows manuscripts overlapping with each other such that their intersection subtends the extent of shared text. These two alternatives, however, are mutually exclusive. We can use the Venn diagram either to demonstrate the manuscripts’ membership of groups or to demonstrate the readings’ membership of manuscripts, but not both. For instance, if we wanted to show that 40% of M4’s text is shared with V1, we could populate their intersection with 40% of their members, specifically the elements actually common to both. We could either list the verse numbers or use some other kind of notation to indicate their members. But where should we place this intersection? If we placed it inside S, M4 would be entirely an S manuscript, whereas Brockington seeks to show that it is a composite of the S and NE traditions. Alternatively, if we placed the intersection of M4 and V1 inside NE, M4 would be partly a member of S and partly not a member of S, which cannot be. Every element is either a member of a set or not a member of that set. Brockington must choose between the alternatives. The following diagram clarifies why we cannot simultaneously express the relation between the sets NW+W, S, and NE and their members, the sets M4, V1, etc., and the sets M4, V1, etc. and their members, the lines or words they contain: 44  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Figure 8. Expressing relations between two orders of sets. This diagram returns to treating NW+W, S, and NE’s members as sets that contain the manuscripts M4, V1, etc. rather than as the manuscripts M4, V1, etc. themselves. This is essential: unless M4, V1, etc. are themselves sets capable of containing members, we cannot express the relationship between their members in turn (for instance, whether they are common to the sets M4, V1, etc. or found in just one manuscript, etc.). Even though the problem should already be evident, let us proceed. The region 14052 ∩ V1 indicates the elements common to 14052 and V1. The diagram displaces the circle denoting S so that 14052 remains inside it (recall that 14052 can either be a member of S or not a member of S, but it cannot be partly inside and partly outside S). The diagram similarly displaces the circle denoting NE so that V1 remains inside it. This already contradicts Brockington’s aim, since he wanted to show that 14052 was only partially an S manuscript (it also features elements characteristic of the NE manuscripts). More important, V1 is now partly contained inside S and partly outside it, whereas 14052 is partly contained inside NE and partly outside it, despite our best effort not to contravene the requirement that every element be either an element of a set or not an element of the set. Neither outcome is permissible. V1 can be either a member of NE or a member of S or a member of both. If it is a member of both, it must be so to the same extent, that is to say, it 45  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches must satisfy the conditions of being NE and being S equally. This also holds for 14052. In this case, 14052 and V1 are members common to S and NE and we would locate them in the intersection of S and NE (S ∩ NE = {14052, V1}). Brockington may object that this is exactly what he wished to show. Indeed, all we need do is shrink the circles denoting S and NE so that 14052 and V1 fit inside their intersection. But consider the outcome: Figure 9. The source of the error. Initially, this diagram appears to meet all Brockington’s requirements. It expresses the intersection of 14052 and V1, while preserving their status as elements of S and NE. Indeed, it also apparently indicates their halfway position between S and NE. In truth, however, the diagram does not show that 14052 is partly an S manuscript and partly an NE manuscript. Rather, it shows that 14052 is both an S manuscript and an NE manuscript. Indeed, it is an S manuscript to the same extent that it is an NE manuscript. Every line or every member of 14052 is now a member of S and likewise a member of NE. Contrary to Brockington’s aim, which was to show that 14052 owed part of its text to S and part of it to NE, that is to say, that it was a composite of the two traditions, this diagram shows that it owes all of its text to S and 46  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches all of it to NE. There is simply no way that Brockington can show what he hopes using the Venn diagram.87 For the sake of argument, assume that Brockington discards the first option. He decides it is less important to show whether the manuscripts are members of one of the three groups or of two or three. He only seeks to show the common elements of certain manuscripts. At this point, all the problems we identified earlier with defining their members resurface. What are the sets’ members—the verses, lines, pādas, or words of the manuscripts? What do we do with repetitions in a manuscript? Do we preserve as much of its text as possible (in which case the elements must be as large as possible to eliminate repetition, albeit at the price of finer resemblances between manuscripts) or do we preserve as much nuance as possible? How useful is it to indicate crude percentages of shared text, if they neither represent genealogical relationships nor indicate the extent of inheritance from the archetype or, indeed, a more recent ancestor? To circumvent these problems Brockington must first define the universal set: the set of passages (or elements) to be distributed. He may specify that his examination is restricted to the starred and appendix passages. Moreover, to avoid the problem of repetition (whether of words or longer units like pādas) he regards these passages in their entirety. He seeks to show that, of x insertions, whereas M4 contains y and V1 contains z, w of their insertions are common. This is a valid use of the Venn diagram. But it is limited precisely to a consideration of genealogically irrelevant aspects. It is analogous to cataloguing the number of shared frocks and ribbons the children in various families have after they have been identified as members of their respective families and their genealogical descent established. Although useful for certain kinds of investigations (for instance, if we wanted to establish that the children in family A had the most toys, but the children in family B also owned 30% of those toys), it does not address the problem of filiation. Indeed, it can only be undertaken once we have assigned the children to families, so that, effectively, what Brockington is doing is using information contained in the Mahābhārata’s appendices to “regroup” the manuscripts by their interpolations. The only value 87 The source of the problem was already apparent in the last paragraph. On one hand, we defined the members of NW+W, S, and NE as the sets containing the manuscripts M4, V1, etc. On the other, we defined M4, V1, etc. as sets that contained other members besides the manuscripts themselves, namely, the lines or words found in the manuscripts. There is real contradiction here: do the sets M4, V1, etc. contain manuscripts or the contents of the manuscripts? Put another way, are their members individual manuscripts or the elements of those manuscripts? 47  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches of doing so is that it provides a visual representation of the appendix passages different manuscripts share, but it cannot tell us anything about their filiation. The following diagram clarifies what Brockington most probably envisions when he proposes that we use the “Venn diagram […] to represent a relationship between separate groups or sets of items […] in this way showing to some extent at least the multiple allegiances involved.” Figure 10. Mapping interpolations using a “Venn diagram.” This diagram comes closest to reconciling what Brockington envisions. But it does not accurately represent the distribution of interpolations in the Rāmāyaṇa (remember that although Brockington features the diagram in his assessment of the Mahābhārata critical edition, the sigla actually stand for Rāmāyaṇa manuscripts). The diagram is limited to eleven manuscripts, which it groups into seven sets, so that it only features seven circles. It cannot handle the full complexity of the tradition. Even if we limited its use to representing interpolations, it would quickly break down. Granted that Brockington recognizes that “this is a decidedly visual model rapidly becoming too complex to describe verbally and […] even visually the patterns soon become highly intricate and less easy to interpret.” But he does not realize that it cannot represent even the limited set of “patterns” he wishes. This is where the final failure of Brockington’s attempt at creating a “Venn diagram” manifests. A Venn diagram must represent all possible logical combinations between sets. Their number is given as 2n where 48  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches n represents the number of sets in the diagram. Thus, for seven sets, the required number of regions is 128, whereas for Brockington’s original proposal (figure 1), which featured ten sets, the number is 1024. Even our corrected diagram (figure 10) is far from meeting this condition. D1.2 and M1–3 do not intersect. D1.2 and Ñ likewise do not intersect. There is no region where only M4+14052 and Ñ meet. There is also no region where only V1 and B meet. Despite every attempt to save his hypothesis, this final diagram is also not a Venn diagram! Not only is it not a Venn diagram; it also does not depict all possible combinations of interpolations in manuscripts, but only those Brockington selectively identifies. Insofar as it represents anything, the diagram is symbolic of his approach to textual criticism: presenting partial information, selectively identifying affinities, and using those affinities to draw erroneous conclusions based on a flawed method. Let us draw some conclusions: 1. The Venn diagram is useless as a tool of genealogical analysis. 2. It is also useless as a tool of analysis of the commonalities between manuscripts. 3. It can be used to represent manuscript groups, but not to analyze them. We can use it only once we have undertaken an analysis. 4. The Venn diagram can be used to represent the distribution of interpolations, but this presumes a prior stemmatic analysis to identify the interpolations as interpolations. 5. The Venn diagram cannot represent all interpolations, but only a limited number in a small group of manuscripts at a time. We can thus use it to study a selection of the evidence, but only if we clarify, in advance, our aims. Any conclusions we draw must be contextualized vis-à-vis the complete manuscript tradition. Finally, there is a possibility that what Brockington contemplates is not a Venn diagram at all, but simply some kind of “graphical representation.” In our dialogue, he admitted as much, saying that he merely wished to represent, in a rough way, the manuscripts’ relationships, remembering, of course, that by “relationships” he means not genealogical relationships but common features like interpolations. The general consensus that emerged was that his diagram was not a Venn diagram and many confusions arose from his erroneous attribution of the term to it. Provided he no longer calls his figures Venn diagrams, understands that they contain no genealogically relevant information, and acknowledges that they are drawn on the basis of the 49  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches editors’ existing analysis of the manuscripts’ relations of filiation and presume their labor in identifying and cataloguing interpolations, we see no objection to their continued use. We merely suggest that in future they be referred to as Brockingtonian sketches to avoid confusion. 3. Summary Brockington recommends the Venn diagram as permitting an initial, albeit crude, sorting of manuscripts. But he overlooks the following: 1. Not all shared variants have the same value. Only some variants are genealogically significant, whereas the great majority reflect minor orthographic changes and the like. 2. Not all shared passages are evidence of a common inheritance. Some passages will have been transmitted horizontally by contamination and hence will mislead us as to the manuscripts’ true filiation. 3. The kind of resemblance—placement, repetition, extent, minor deviations, and changes to neighboring text—contains more genealogically relevant information than the Venn diagram can capture. Set theory requires that we treat verses or variants as identical members of their containing sets, which implies an unacceptable loss in precision. In his desire to “mathematize” the manuscripts’ relationships, Brockington needlessly discards valuable information. 88 His proposal is both excessive and insufficient: excessive 88 “Mathematize” in the sense that he seeks to reduce them to numerical figures and percentages and not in the sense of logical, comprehensible, and universally verifiable principles. As his confusions about the Venn diagram attest, Brockington is far from a mathematical approach in the latter sense. Examples of his “mathematical” approach are so numerous in The Sanskrit Epics that we cannot cite all of them. The following are indicative only. “Although the commonest formulae occur throughout the Mahābhārata there are a significant number which have a more limited distribution, sometimes but not always determined by the difference of subject matter between different parts, while there is also a marked variation in relative frequency between the text and the * passages for many (which some occurring preponderantly or exclusively in the text, and others mainly or exclusively in the * passages). For example, paurajānapadā janāḥ occurs 11 times in the text only, sa saṃprahāras tumulaḥ occurs 10 times in the text of books 3 and 6–8 only, and the simile visphūrjitam ivāśaneḥ occurs 14 times in the text only; by contrast, prayatenāntarātmanā occurs 5 times in the * passages only, naimiṣāraṇyavāsin occurs 5 times in the * passages and just twice in the constituted text (significantly at 1.1.3b and 13.6d), devadevaḥ sanātanaḥ occurs 10 times in * passages and three times in the text (at 1.61.90b, 3.86.24b and 15.35.16b), the simile haviṣā kṛṣṇavartmeva occurs 6 times in * passages only, and the compound sarvadevanamaskṛta occurs 7 times in * passages only. In a particularly interesting pair of formulae sarvapāpaiḥ pramucyate, occurring in the even pādas (the commoner position for most formulae), is found more often in the text, with 26 occurrences there (more than half of them in the 50  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches because it entails a drastic simplification of the manuscripts’ contents beyond even that he undertakes when he reduces their relationships to quantitative figures; insufficient because it does not adequately conceptualize the sets’ elements. He neither defines his sets nor explains how we should treat partial matches or repeated verses, and so on. His proposal is thus inadequate to the task. A Venn diagram based on grouping manuscripts by their most visible differences appears an easier alternative. It replaces careful study of the manuscripts’ readings and, above all, of their shared errors of transcription with a handful of prominent characteristics like shared interpolations. But the move from a simple quantitative relationship (for example, manuscript x shares 30% of the text of manuscript y) to its diagrammatic representation presents insurmountable obstacles. Do placement and semantic order count? What about accidental identities? If M4 and V1 share a word or sequence of words not found in the other manuscripts, albeit in different chapters or books, is this evidence of affinity? Or, rather, is it a coincidence, even though M4 ∩ V1 will now contain more members than their respective intersections with the other manuscripts? Brockington justifies his suggestion that editors use Venn diagrams “to represent a relationship between separate groups or sets of items” on the grounds that it evades “the impossible task of constructing the entire chain of transmission involved,” but he has considered neither the nature of those relationships (intersection, union, or complement, or is their nature, rather, one of filiation?) nor the nature of the elements involved (are they additions, omissions, words, or verses?).89 He emphasizes that the “Trivandrum manuscript” frequently omits the same passages as M4, even though the Venn diagram is singularly unsuited Āraṇyakaparvan) against 19 in * passages, whereas mucyate sarvapāpebhyaḥ occurring in odd pādas is found mainly in * passages, with 7 occurrences against 3 in the text (at 3.34.76a, 198.52e and 12.201.35c).” Brockington, The Sanskrit Epics, 109–110. And see ibid., 111: “The distribution pattern for all these types of repeats is indeed quite revealing. There are 7 in the Ādiparvan, 4 in the Sabhāparvan (with another 6 in * passages), 43 in the Āraṇyakaparvan (of which 30 occur in 3.80–83), 4 in the Virāṭaparvan (with another 3 in * passages), 11 in the Udyogaparvan, 8 in the Bhīṣmaparvan, 7 in the Droṇaparvan (and 9 in * passages), 6 in the Karṇaparvan (and 1 in a * passage), 5 in the Śalyaparvan, none in the Sauptika and Strī parvans, 25 in the Śāntiparvan (with 8 in * passages), 37 in the Anuśāsanaparvan (and 27 in * passages), 16 in the Āśvamedhikaparvan (and 19 in * passages), and none in the last four books.” Entire segments of The Sanskrit Epics consist of such statistical details, doubtless aided by Muneo Tokunaga’s “electronic texts of […] the Rāmāyaṇa and the Mahābhārata” (Tokunaga is cited and acknowledged on pages 519 and 521). 89 Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” 204. 51  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches to capture such information.90 No precedent exists for defining a set in terms of members it does not contain. We could, at most, identify the set’s complement, but this would not contain all of the set’s members rather than comprising a mixture of elements and omissions. Brockington also overlooks that if the “entire chain of transmission” cannot be reconstructed it is especially unlikely that the relationship of the later and worse manuscripts can be stemmatized. Rather than focus on the part of the transmission that can be reconstructed, his solution is to map—imperfectly and with severe limitations—the extent of indiscriminate borrowing between the deteriores. He then reduces this borrowing to quantitative relationships and recommends using overlapping circles “showing […] the multiple allegiances involved.”91 The method’s supposed advantages, above all, the ability to overlap different circles, indicating a manuscript’s composite nature, are bought at the price of being unable to represent its (true) descent. Tradition Criticism Brockington’s approach is counter to everything editors have attempted for the past two centuries, when they sought to determine manuscript filiation based on a systematic recensio; counter also to the principle that an editor must first establish the “vertical” constants of the tradition before he can determine the extent of “horizontal” transmission.92 These errors are understandable insofar as Brockington has never edited a work and never produced full collations of a manuscript. All his examples are drawn from the * passages the editors identified.93 Besides the questionable nature of drawing conclusions about the Mahābhārata 90 See Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” 509–10 and Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” 202–203. Both passages are quoted earlier. 91 Brockington, “Regions and Recensions, Scripts and Manuscripts.” 92 Compare Jean Irigoin, “Quelques réflexions sur le concept d’archétype,” Revue d’histoire des textes 7 (1977): 242–43: “He [the editor] should ensure, by all the means provided by his science and his ingenuity, that he discovers therein the constants of the ‘vertical’ tradition. It is at this price, and at this price only, that he will be able to determine the reality, and eventually the amplitude, of a ‘horizontal’ transmission.” 93 Thus at Brockington, “Textual Studies in Vālmīki’s Rāmāyaṇa,” 203: “Vaidya states unequivocally about M4 that ‘when it goes with N, it is invariably associated with the NE version.’ Certainly, if we proceed by negative evidence, this is broadly true, since the N *passages not found in M4 are those of the manuscripts in the grouping Ś1 D1–7, which represents the NW and W recensions; the largest number are in fact of the grouping Ś1 D4–7, which does not however correspond at all exactly to the grouping 52  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches based on the Rāmāyaṇa, Brockington’s work exhibits, once again, the tendency among Indologists to make a great display of their “critical” acumen. His statement that “the critical editions are […] on the whole the best we have” is baffling because if the stemma does not hold neither does the reconstruction based on it, and Brockington does not think we can draw up stemmata for the Sanskrit epics. A critical edition differs from an edition based on the editor’s conjectures or the concord of a majority of witnesses precisely in that it relies on a stemma for most of the sifting between variants.94 Yet Brockington argues that, Essentially the point at issue is that the Critical Edition’s system of assigning manuscripts to recensions or to script versions is almost as limiting as constructing a stemma codicum (which does have the merit of corresponding on the whole to the way that manuscripts are propagated). The drawback is that this still involves an either/or situation, where a particular manuscript is assigned to a particular recension or version, or else is casually dismissed as ‘contaminated’ or a ‘mischcodex’. The reality is that script boundaries are by no means as rigid into recensions of the Critical Edition, in particular by the inclusion of D4, a supposedly NE ms, with the NW and W mss. On the positive side we may note the frequency of the complementary grouping Ñ V1 B D1.2 M4 which, besides M4, includes D1.2 (supposedly W) with the otherwise NE mss. Even before we consider the additional evidence from the Trivandrum manuscript, there are then grounds for questioning the assignment of the N mss to subrecensions. As we have just seen, D4 often sides with NW mss and not with NE mss, making it sensible to consider it rather a representative of the NW recension in the Critical Edition’s terms; within the NW recension D4 and D7 seem to form a sub- group. It is particularly noticeable that there are a considerable number of occasions when M4 and the Trivandrum manuscript share common readings, common omissions or common additions with D1.2 only, while nearly as common is the grouping V1 D1.2 M4 and the Trivandrum ms. On the other hand, V1 usually goes with the Newari and Bengali mss to constitute the NE recension.” 94 Contini is persuasive here. He calls stemmata “an objective and mechanical tool, invented to sort out, in the first instance, the quarrel between variants that are in themselves equally acceptable: […] variants that are pronounced erroneous serve to brand as erroneous those that have remained equally acceptable.” Gianfranco Contini, “La «Vita» francese «die Sant’Alessio» e l’arte di pubblicare i testi antichi (1970),” in Frammenti di filologia romanza: Scritti di ecdotica e linguistica (1932–1989), ed. Giancarlo Breschi, vol. 2 (Florence: Edizioni del Galluzo, 2007), 961. Brockington either overlooks the distinction between manifest and latent errors or he overlooks the problem that, having eliminated the stemma, he cannot identify, much less correct, latent errors. Eklund is good on the stemma’s role in the latter: “A very serious criticism of the stemmatic method is that it is a single, gigantic vicious circle […]. This objection may seem well founded but it may, nevertheless be rejected […]. As I have just pointed out, the stemma is built up by means of the manifest errors but then it is not these same manifest errors which are corrected by means of the stemma […]. No, the stemma is used to enable a choice to be made as soon as latent errors are encountered.” Sten Eklund, “The Traditional or the Stemmatic Editorial Technique?” Eranos 104 (2006–2007): 12. Both passages are cited (Contini also translated) in Trovato, Everything You Always Wanted to Know about Lachmann’s Method, 59. 53  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches as this scheme of classification would suggest and that the only effective way of tracing alignments is by plotting the shared features of specific manuscripts.95 Although intended as his definitive statement on the Rāmāyaṇa (and mutatis mutandis the Mahābhārata),96 it is less conclusive than he imagines. Consider the following problems: 1. The critical edition does not “assign manuscripts to recensions or script versions.” It undertakes a genealogical analysis of their relations of filiation. 2. The stemma does not correspond “on the whole to the way that manuscripts are propagated.” It is the best means yet found for representing manuscript filiation. Brockington’s proposed alternative—the Venn diagram—is unequal to the task. We await clarification of his statement that “a new way of establishing relationships between manuscripts [has been] developed for the Mahābhārata by Wendy Philipps-Rodriguez,” although at the recent DICSEP 8 he did not elaborate.97 3. The “either/or situation, where a particular manuscript is assigned to a particular recension or version, or else is casually dismissed as ‘contaminated’ or a ‘mischcodex’” does not arise from “the Critical Edition’s system of assigning manuscripts to recensions or to script versions” (which it does not do). It is a relic of the fact that we can either establish a manuscript’s descent with a degree of certainty or it resists stemmatization. Ultimately, it is a consequence of whether vertical or horizontal transmission dominated in the manuscript’s formation, and thus an unavoidable feature of manuscript transmission. Brockington’s proposed solution cannot eliminate the problem. The Venn diagram appears to cope better with this ambiguity, but only because it is less sensitive to filiation. To argue that the Venn diagram is preferable as its blurry nature corresponds to the blurring of manuscript boundaries 95 Brockington, “Is the Script Relevant?” 18. 96 The attached footnote makes this explicit: “the point that script boundaries are less significant than assumed by the editors of the Critical Edition is one that is valid for the Mahābhārata also, as research being done in Cambridge under John Smith by a former student of mine is demonstrating by the rather different technique of applying cladistics (see Wendy J. Phillips-Rodriguez, ‘The Evolution of a Sanskrit Epic: Some Genetic Considerations about Scripts’, in The Evolution of Texts: Confronting Stemmatological and Genetical Methods, ed. C. Macé, P. Baret, A. Bozzi, L. Cignoni, Linguistica Computazionale 24–25 (IEPI, Pisa-Roma).” Ibid., 27–28, n. 4 (italics added). 97 We referred to the diagrams contained in Wendy Phillips-Rodriguez, C. J. Howe, and H. F. Windram, “Chi-Squares and the Phenomenon of Change of Exemplar in the Dyūtaparvan,” in Sanskrit Computational Linguistics, ed. Gérard Huet, Amba Kulkarni, and Peter Scharf (Heidelberg: Springer, 2009), 383–86. 54  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches is like suggesting we should look at indistinct objects through a frosted lens because we will then not notice their indefinite boundaries. 4. The objection that “script boundaries are by no means as rigid as this scheme of classification […] suggest[s]” reflects a justified anxiety that horizontal transmission played a greater role in an exemplar’s formation than vertical transmission. But we must first establish this. The stemma is the best means for this.98 If all Brockington means by the statement is that manuscripts rejected as late and inferior nonetheless contain good readings, this principle is well known in textual criticism (recentiores, non deteriores). But does it repay to study the manuscripts for a few potentially superior readings? Brockington, at any rate, appears to think not, since he has neither collated a manuscript completely nor proposed a single conjecture that betters the editors’ readings. 5. Finally, although it is true that the “only effective way of tracing alignments is by plotting the shared features of specific manuscripts” this is to misunderstand the aim of textual criticism. Agreements between manuscripts are the easiest way to sort them, but offer only a preliminary criterion. They do not reveal their genealogical relationships. If two or more manuscripts agree against the rest, which of the two groups preserves the correct reading? A table of agreements like the one Brockington proposes cannot eliminate what Trovato calls “the ineluctability of critical judgment.” “Plotting the shared features of specific manuscripts” will most likely provide the same information as a simple observation.99 The proposal is also not 98 Reeve rightly notes that “scholars [... who] believe that contamination and interpolation were so common that stemmatic method seldom or never works [...] are biting off the hand that fed them, because unless one can actually see scribes contaminating and interpolating under one’s nose it is only by applying stemmatic method that one can detect contamination and interpolation.” Michael D. Reeve, “Stemmatic Method: ‘Qualcosa che non funziona?’” in The Role of the Book in Medieval Culture: Proceedings of the Oxford International Symposium, September 26–October 1, 1982, ed. P. Ganz, Bibliologia 3 (Turnhout: Brepols, 1986), 65. 99 This is also the opinion of no less an authority than West: “Of the value of other sorting methods, in particular statistical methods, I remain sceptical. A numerical table of significant agreements between every two manuscripts, as described on p. 38, will provide objective confirmation of groupings suggested by casual inspection, and will indicate how clear-cut they are (e.g. how much more often GHIKL agree with each other than with other manuscripts); but simply collecting the evidence, without reducing it to figures, will probably have given a clear enough picture already. Indeed, where groups of three or more manuscripts are concerned it will have given a clearer one, for the information that A agrees with Β 81 times, Β with C 92 times, and A with C 79 times does not enable us to deduce that ABC all agree together 73 times, or even once.” West, Textual Criticism and Editorial Technique, 46. 55  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches new. Before Brockington, John Griffith proposed a similar approach, albeit in greater detail.100 West provides a concise summary of Griffith’s procedure: A more elaborate way of using such a table has recently been advocated by J. G. Griffith. It involves comparing manuscripts in respect of the whole numerical series of each one’s agreement with the others; no distinction is made between primary and secondary readings, though agreements that may represent coincidental innovation are excluded. Suppose manuscript A shows the following numbers of agreements with the others: B F G H J K L N O P R T U V Z 5 4 6 4 5 4 6 6 4 4 4 5 6 4 4 0 4 1 9 2 8 2 2 8 2 4 3 1 7 3 For each manuscript a similar series of numbers is found. The one whose series matches that of A most closely is then grouped most closely with A. Finally the manuscripts are all arranged in a ‘spectrum’: those with the most dissimilar patterns of agreement appear at opposite ends, with a continuous gradation from one end to the other, while certain clusters or ‘taxa’ mark themselves off along the line.101 Griffith’s tabulation of agreements is more systematic, but in essential respects it parallels Brockington’s approach. His diagram on page 121 can be considered a precursor of Brockington’s “Venn diagram” as it likewise aims to provide a graphic representation of manuscript agreement. But whereas Brockington uses circles and ellipses to represent his manuscripts and the points within the enclosed plane figures to represent their elements (their intersections representing the common elements), Griffith adopts a more elegant solution: an array that lists manuscripts along the x- and y-axes and uses three kinds of hatching to indicate either shared readings or the extent of sharing.102 What his approach gains in being able to 100 John G. Griffith, “A Taxonomic Study of the Manuscript Tradition of Juvenal,” Museum Helveticum 25, no. 2 (1968): 101–38. 101 West, Textual Criticism and Editorial Technique, 46–47. 102 Griffith provides the following, not entirely perspicuous explanation: “A dot is then placed in each square that corresponds to a coincidence of reading in each pair of mss. The agreement of an isolated pair is recorded by a single dot in the appropriate square; three dots are needed for a trio of mss. in agreement, 6 for a quartet, and so on. Obviously an aberration or singularity on the part of an individual ms. has to be neglected, if only because there is nowhere to register it. (One might in such a case record the agreement of the other 15, but it is doubtful whether the extra labour would affect the result and isolated eccentricities are more likely to falsify the picture than to clarify it.) The only critical activity 56  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches represent different kinds of agreement, it loses in not being able to represent multiple coincidences, which must be read off from the array by considering pairs of manuscripts at a time. Vice versa, Brockington’s approach cannot capture verse position and order and remains completely abstract since he never populates his sets or their intersections with members. Differences in representation aside, the two approaches share several features. Brockington shares Griffith’s emphasis on the “open” nature of the tradition,103 the suggestion that the editors’ selection of manuscripts was random, that many exemplars were ignored or not known, that attention should focus rather on identifying alignments or affinities between manuscripts, 104 and the idea that, beneath the confused mass of contaminated exemplars resistant to genealogical classification, statistical compilation of similarities can reveal distinct demanded of the recorder is that he should be alert to exclude from the count any non-significant similarities, such as lines lost by homoeoteleuton (such as I 87–88) or homoearchon (as at II 24–25). Spelling trivialities are also inadmissible, as in these details scribes may have unconsciously conformed to ingrained habit and so not reproduced faithfully what was before them. The recording process is not however completed by merely filling in the Squares wherever there is agreement between pairs of mss. and leaving it at that. Thus at V 116, PRA agree in spumat, as is shown by the three entries in distinctive hachuring (white bands running from ‘North-West’ to ‘South-East’) in diagram I below[.] But the agreement of the other 13 in the alternative fumat must be plotted in too, as has been done in the diagram by a rather different hachuring, with the white bands running from ‘South-West’ to ‘North- East’. This second operation entails 78 recording acts, in addition to the three already made: in practise no distinction need be made between the agreement within a small and the agreement within a large group, which in diagram I has only been done for clarity of exposition. Obviously the least tedious distribution to record is a split of 8 mss. against 8, which requires only 28+28 or 56 recording acts in all. The maximum number occurs when 14 mss. agree against 2: this works out at 91+1 or 92 in all. Where mss. split into 3 groups, each group is treated separately, as at VI 322. There AFNPR agree in fluctum, LO in frictum and the rest (BGHJKTUVZ) in fructum. This calls for 10+1+36 recording acts, or 47 in all, as is shown in diagram I. Reference to the diagram and its attendant key should dispel obscurity.” Griffith, “A Taxonomic Study of the Manuscript Tradition of Juvenal,” 120–22. 103 “This enquiry sets out to see how far an unsophisticated but, I hope, judicious use of taxonomical methods (4.22 f., below) may lead towards a rational sorting of those mss. which, being fully collated, represent for this purpose the ‘open’ or ‘horizontally interpolated’ element in the tradition.” Ibid., 102. 104 “I now turn to those other mss. which have been collated in sufficient detail to make quantitative comparison possible. There is an undoubted element of randomness inherent in the data, in as much as the choice of mss. for collation was made by different individuals long ago, and was, so far as can be seen, largely a haphazard one. Indeed much of the work of collation was undertaken before many of the possible candidates for scrutiny were known or even listed. It may well be that the choice was on the whole not a bad one. Whatever its merits or failings, it is only practical to make the most of the information at present available. Any improvement to the text of the author will thus be consequential, neither is this an effort to establish an order of merit of the witnesses being examined. This would be tantamount to reviving the meaningless concept of a ‘best ms.’.” Ibid., 112–13. 57  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches patterns or groups of manuscripts.105 Griffith also echoes Brockington’s preference for arguing from agreement and his aversion to evaluating readings when he writes: “It is essential to remember that the operation is conducted in terms of similarities and differences between the several mss. Thus there is no question of forming provisional judgments about the rightness or wrongness of a reading which is open to doubt.”106 Like Brockington, he is skeptical of the stemmatic method and considers it especially inadequate for open traditions. 107 Like Brockington, Griffith thinks a statistical approach has an advantage over the latter in that that it reduces manuscript filiation to quantitative terms, and hence is “free” of its subjective and potentially circular nature.108 Griffith also expresses a similar distaste for complete collations.109 More important, he concurs with Brockington that a stochastic sampling of data can replace “a true, continuous, and systematic recension” (Wolf).110 Like Brockington, Griffith does not 105 “What I set out to do is to explore a promising method of achieving a meaningful calibration of the superficially disparate array of manuscript-characteristics in terms of ‘near-neighbour’ affinity, which is quite another matter. If an arrangement of clusters of mss. can be detected underneath the confusion of data before us, the path to the next stage of enquiry may be clearer than it would otherwise be. If the process is properly conducted, it should result in a scale of such near-neighbour relationships, with the most heavily interpolated documents coming together at one end of the scale and as far apart from the sincere witnesses as possible. Clusters with less pronounced characteristics should appear in the middle section of the line, which can be thought of as similar to a spectrum-line in the field of Physics or Chemistry.” Ibid., 113. 106 Ibid. 107 “From what has just been said, it will be clear that one must cut oneself loose from the preconceptions and limitations of the conventional family-tree presentation, whose inadequacy has become increasingly evident as the intricacies of such open traditions come to be better understood. A further advantage of the process is that the investigator is entirely ignorant of the way the data he is analysing will work out until his counting is completed and he comes to the final stage of the resolution of the figures he has arrived at for each section of the satires.” Ibid. 108 See ibid.: “In the conventional methods of evaluating mss. there is an inherent danger of circular argument, from which the taxonomic process is free.” 109 “I take the view that the return on the labour of further full-scale collation is unlikely to be justified. The use of reliable sampling techniques to associate promising witnesses, as yet imperfectly collated or unknown, with well-established taxa is obviously a matter that will have to be faced at the next stage. It is obviously impracticable to re-calculate similarity-matrices de novo to take account of new material as it arises; fortunately other techniques (such as ‘split matching-scores’ and the like) may enable the classification to be achieved with an acceptable economy of effort.” Ibid., 132. 110 “So long as one is dealing with compact arrays of data which show characteristics common to the whole group of specimens being studied, the accepted methods are generally adequate, in biological study as elsewhere. Such systems are now called ‘monothetic’ and their treatment and interpretation are mostly straightforward. They are obviously analogous to a ‘closed’ textual tradition, in which features such as lacunae of identical extent common to a whole group of mss. They tell their own story. ‘Polythetic’ systems are another matter. Here characteristics are not universal to the group under 58  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches differentiate typological from genealogical classification. Again, like Brockington, he argues that “the taxonomic process” can “disentangle the underlying relations of the interpolated mss. and penetrate to the pattern of their ‘genetic histories’ latent beneath the over-burden of interpolation.” 111 Both scholars propose graphical schemes that allegedly overcome the stemma’s weakness in representing contaminated traditions. Griffiths argues, “in really complicated cases an admirable ingenuity is often displayed in adorning the stemma with an elegant arabesque of dotted lines to show the presumed channels of contamination.”112 But they overlook that the data they wish to capture is less relevant for reconstructing the archetype than the data their schemes fail to capture. Even discounting the problems with Brockington’s “Venn diagram,” West’s reservations about Griffith’s statistical approach are a serious objection: The trouble with this kind of analysis is that it is not clear what useful conclusions can be drawn from it. Two manuscripts may be grouped together just because they show no particular tendency to agree with any manuscript more than any other, in other words because they are equally promiscuous, even if they have no special similarity with each other textually. In some investigation: thus their exist in nature birds without wings, mammals lacking red corpuscles and countless other ‘untidy’ phenomena to bedevil classification. Concentration on single characteristics leads nowhere in such situations, and attention has to be directed to an aggregate of properties. Certainly, acute observation, aided by luck, may detect some striking features of resemblance, but discovery of this kind is at best fortuitous and at worst misleading, in that what has been observed may have come in only at a late stage in the evolutionary process, and so mean less than it might seem. Obviously, for the taxonomical approach to be valid, a substantial number of points of similarity between specimens or groups of specimens is essential. There seems general agreement that a minimum of 40 mutually unrelated elements should be taken for any single operation, but a figure of the order of 60–70 is safer. If therefore in a literary text a significant divergence occurs about once in every 4 or 5 lines, reliable results may be expected where stretches of not less than 300 lines have been treated; in most of what follows considerably longer blocks have been used.” Ibid., 114–15. 111 “One question has been deliberately left undiscussed. Has the taxonomic process effectively disentangled the underlying relations of the interpolated mss. and penetrated to the pattern of their ‘genetic histories’ latent beneath the over-burden of interpolation they now present to our observation? The degree of regularity of their behaviour as disclosed in this study makes this at least a tenable supposition. It has been put to me that the position should be stated more strongly: that it is incumbent on those who are sceptical of the method or of its results in Juvenal’s case to indicate the kind of explanation which in their view would cover both the superficial disarray of the data and the underlying regularity revealed by the analysis, certified as this is by rigorous and accepted criteria of statistical probability.” Ibid., 134. 112 Ibid., 119. 59  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches cases it is evident that the taxa reflect real affinity-groups, in others it is not. In any case we are given no guidance as to the distribution of ancient readings.113 Griffith’s and Brockington’s work illustrates the limits of statistical approaches to manuscript classification.114 A mere tabulation of agreements (in Brockington’s case, also “omissions”) between manuscripts cannot replace a genealogical analysis. Above all, it does not let us reconstruct earlier stages of the text, despite Brockington’s belief that removing certain kinds of repetitions (those he considers “Brahmanic”) and collecting others (those he considers remnants of “oral epic poetry”) lets him reconstruct an older text than the archetype. Conclusion Criticisms of the critical editions of the Sanskrit epics reflect wider trends in textual criticism, where the neo-Lachmannian edition has fallen out of favor. In an age when few scholars have either the patience or the resources to create an edition based on an exhaustive survey of manuscript sources and a consideration of their readings, ad hoc criticisms based on partial investigations are appealing. Yet, as this overview revealed, these criticisms are often based on an insufficient familiarity with the material, or an insufficient knowledge of the principles involved, or both. The charge that the “critical editions are based on the readings of just a selection of the vast number of manuscripts known,” for instance, is only relevant if we can show that it incorrectly discarded some higher value manuscripts. Invoking “a new way of 113 West, Textual Criticism and Editorial Technique, 47. 114 Unlike Griffith, who applies a statistical approach only to the manuscripts, Brockington also applies it to the constituted text of both epics itself. On one hand, he lists interpolations and “omissions” in manuscripts, and uses this to undermine the critical editions’ claims to authority. On the other, he lists repeated words or verses in their constituted texts, and uses this to argue for the works’ oral origins. Both are, of course, ultimately consonant, because they facilitate his goal of undermining the transmitted text for bearing theological meanings he rejects. Pollock notes the connection: “In the most recent discussion Brockington (1984) stubbornly reaffirms Jacobi’s view: ‘Far from being a Vaiṣṇava epic, Vālmīki’s Rāmāyaṇa is no religious epic at all. It is lamentable that misunderstanding of this point … should still persist so longer after Jacobi’s explicit declaration’ (p. 13). Brockington employs an elaborate five-stage scheme of text evolution to demonstrate the developing conception of the hero (from one who is ‘thoroughly human’ to a god, pp. 218–25).” Sheldon I. Pollock, “Introduction,” in The Rāmāyaṇa of Vālmīki: An Epic of Ancient India, ed. Robert P. Goldman, vol. III: Āraṇyakāṇḍa, trans. Sheldon I. Pollock (Delhi: Motilal Banarsidass, 2007), 17, n. 21. The reference is to John Brockington, Righteous Rāma, The Evolution of an Epic (New York: Oxford University Press, 1984). For Pollock’s rejection of the thesis of a “‘heroic epic’ [transformed] according to a later theological program,” see Pollock, “Introduction,” 18–19. 60  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches establishing relationships between manuscripts,” likewise, requires that we first demonstrate we understand the principles involved.115 Before we raise unfounded and misleading objections to the critical editions, let us first learn to appreciate the editors’ achievements, remembering that they were men of great probity and dedication,116 who worked untiringly for an inconsequential 115 Trovato’s judgment is apt and devastating: “More generally, as many of the considerations discussed above suggest, one gains the clear impression that almost all the debatable elements discussed so far, and possibly even the search for a computer based alternative, arise from inadequate knowledge of traditional philological theory and methods.” Trovato, Everything You Always Wanted to Know about Lachmann’s Method, 199. 116 For an insight into the sense of humility and service that motivated the Mahābhārata editors, see S. K. Belvalkar, “Introduction,” in The Bhīṣmaparvan for the First Time Critically Edited (Poona: Bhandarkar Oriental Research Institute, 1947), cxxxii (“The edition of the Bhīṣmaparvan, completed by me after several years’ labour, is now offered, as a sort of vāṅmaya-śrāddha, to the revered memory of Ramakrishna Gopal Bhandarkar on the occasion of his twenty-second death-anniversary. May it be regarded as a tribute worth of that great savant, and may it also be, in accordance with BG. 18.70, acceptable to Lord Kṛṣṇa as a proper iṣṭi of the Jñānayayajña that is to redound to His Glory!”) and V. S. Sukthankar, “Introduction,” in The Āraṇyakaparvan for the First Time Critically Edited, vol. 1 (Poona: Bhandarkar Oriental Research Institute, 1942), xxxviii (“If Mahārṣi Kṛṣṇa Dvaipāyana Vyāsa tells us that he has cried himself hoarse urging people to follow the Path of Duty: ūrdhvabāhurviraumyeṣa na ca kaś cicchṛṇoti mām | dharmādarthaś ca kāmaśca sa kimarthe na sevyate ||, his shouting with uplifted arms has not been entirely in vain. He has not failed in his mission. Across the reverberating corridors of Time, we his descendants can still hear dimly his clarion call to Duty. It is in response to that call and in a spirit of reverent homage to that sage of unfathomable wisdom that embodied Voice of the Collective Unconscious of the Indian people we offer this work, pledged to broadcast to mankind, in this hour of its need and its peril, the luminous message of the Maharṣi: na jātu kāmānna bhayānna lobhād; dharme tyajejjīvitasyāpi hetoḥ | nityo dharmaḥ sukhaduḥkhe tvanitye; jīvo nityo heturasya tvanityaḥ ||.”) (All italics Sukthankar’s.) See also V. S. Sukthankar, statement reproduced in “Bhandarkar Oriental Research Institute, Poona Silver Jubilee Celebrations 4th and 5th of January 1943. Roger Lumley, N. C. Kelkar, J. R. Gharpure, R. N. Dandekar, C. Kunhan Raja, V. M. Apte, P. V. Bapat, A. P. Karmarkar, B. R. Kulkarni, H. D. Sankalia, C. R. Sankaran, S. R. Shende, Ludwik Sternbach,” Annals of the Bhandarkar Oriental Research Institute 24, no. 1/2 (1943): xlvii: “There is a danger that in our pseudo-scientific mood, we may be tempted to discard this great book, thinking that we have out-grown it. That would be a capital blunder! That would in fact mean nothing but an indication of our will to commit suicide, national suicide, the signal of our national extinction. For never was truer word spoken than when the late German Indologist Hermann Oldenberg said that ‘in the Mahābhārata breathes the united soul of India, and the individual souls of her people.’ And why is that? Because the Mahābhārata is the national saga of India. It is, in other words, the content of our collective unconscious. And just for that reason it refuses to be discarded. We must therefore grasp this great book with both hands and face it squarely. Then we shall recognize that it is our past which has prolonged itself into the present. We are it: I mean the real WE! Shall we be guilty of strangling our own soul? NEVER!” (Italics and capitalization Sukthankar’s.) Penned on January 5th January 1943, sixteen days before he died, they are the last words Sukthankar wrote. 61  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches remuneration. 117 We should also require scholars who criticize the available edition(s) to provide alternatives or, at least, explain which text we should use.118 Brockington criticizes “appeal to or preference for the vulgate” as “misguided,” but simultaneously argues that “the text even of the critical editions cannot be regarded as a completely coherent whole.” In fact, neither the receptus nor the constitutio textus should have binding value for us. No less than Grünendahl, who has yet to produce his promised critical edition of the Nepālī recension, Brockington must tell us what editorial approach he favors, what manuscripts he proposes editing, and how he plans to go about it. Until someone proposes a better text, the objections to the existing editions stemming from higher criticism are against argument and good sense. As Brockington’s work attests, speculative, a priori histories based on unprovable claims of “oral” transmission have only led to grievous misunderstandings of the manuscript evidence. 117 P. L. Vaidya, the last General Editor, recollects: “The late Dr. Sukthankar, the first General Editor, was paid a salary of Rs. 500 per month for a period of nearly 18 years. After his death in 1943, the Institute found that it could pay to the General Editor not more than Rs. 250 a month. When, therefore, a scholar of the eminence of Dr. Belvalkar, who had held a post in the Indian Education Service, was appointed General Editor, we could not offer him more than Rs. 250 per month as salary. In order to make this paltry sum look dignified, we started referring to his ‘salary’ as ‘honorarium of Rs. 3,000 per annum.’ Out of his 18 years of General Editorship, Dr. Belvalkar received this honorarium for10 years only, and, although the cost of living was steadily rising all those years, the amount of his honorarium of Rs. 3,000 per annum actually happened to dwindle down to a zero. The Parvan-editors were paid at the rate of Rs. 5/-per printed page of demi quarto size, no doubt, but this rate, in my opinion as well as in actual experience, worked out to be not more than one rupee a day. The departmental staff was paid a salary every month, but, according to Government Audit Party visiting the Institute annually, it was far below the standard.” General Editor P. L Vaidya’s statement read out on September 22, 1966, reprinted in “The Completion of The Critical Edition of the Mahābhārata,” Annals of the Bhandarkar Oriental Research Institute 47, no. 1/4 (1966): iii–iv (italics in original). 118 This also applies to Reinhold Grünendahl, who, after casting suspicion on Sukthankar in “Zur Klassifizierung von Mahābhārata-Handschriften,” states that “a more detailed evaluation of the Nepali manuscripts of the Mahābhārata in general and of the Ādiparvan manuscript Ñ4 in particular must be reserved for a work currently in preparation.” Ibid., 129–30. Twenty-five years later this work is yet to appear (despite being told “the conditions for this [task] are provided by the work of the Nepal-German Manuscript Preservation Project”; ibid., 130). In fact, the sole collations we possess for this manuscript are Sukthankar’s, provided in 1939 in V. S. Sukthankar, “Epic Studies VII: The Oldest Extant Ms. of the Ādiparvan,” Annals of the Bhandarkar Oriental Institute 19 (1938–39): 201–62. Grünendahl himself relies on them, though, like Brockington, he neglects readings for interpolations. Not one Mahābhārata critic has addressed himself to the task Sukthankar set when he wrote: “In the light of such variants, some readings of the Critical Edition will have to be reconsidered, and it may, in sporadic cases, be found necessary even to alter slightly the readings of the Ādiparvan in the Critical Edition. That work must however be left over now for further research by a future generation of scholars.” Ibid., 209. 62  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches References Note: Brockington has published under the names “John L. Brockington” and “J. L. Brockington” as well as “John Brockington.” As per Chicago, all bibliographic entries have been standardized and listed under “John Brockington.” Adluri, Vishwa. “Literary Violence and Literal Salvation: Śaunaka Interprets the Mahābhārata.” Exemplar: The Journal of South Asian Studies 1, no. 2 (2012): 45–68. ———. “The Perils of Textual Transmission: Decapitation and Recapitulation.” Seminar 608, The Enduring Epic: A Symposium on Some Concerns Raised in the Mahābhārata (2010): 48–54. Adluri, Vishwa, and Alf Hiltebeitel. “Redressing the Undisrobing of Draupadī.” In Indology’s Pulse: Arts in Context. Essays Presented to Doris Meth Srinivasan in Admiration of Her Scholarly Research, edited by Corinna Wessels-Mevissen and Gerd J. R. Mevissen, with the assistance of Arundhati Banerji and Vinay Kumar Gupta. New Delhi: Aryan Books International, forthcoming. Adluri, Vishwa, and Joydeep Bagchee. “Contra Lachmann: Criticisms (Brockington, et al.) of the Mahābhārata Critical Edition.” Paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 15, 2017, IUC, Dubrovnik, Croatia. ———. “Jews and Hindus in Indology.” Paper published on Academia.edu. https://www.academia.edu/30937643/Jews_and_Hindus_in_Indology (accessed October 10, 2017). ———. The Nay Science: A History of German Indology. New York: Oxford University Press, 2014. ———. “Paradigm Lost: The Application of the Historical-Critical Method to the Bhagavadgītā.” International Journal of Hindu Studies 20, no. 2 (2016): 199–301. 63  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches ———. Philology and Criticism: A Guide to Mahābhārata Textual Criticism. London: Anthem, 2018. ———. “Theses on Indology.” Paper published on Academia.edu. https://www.academia.edu/30584042/Theses_on_Indology (accessed October 13, 2017). Bagchee, Joydeep. “The Epic’s Singularization to Come: The Śakuntalā and Yayāti Upākhyānas in the Southern Recension of the Mahābhārata.” In Argument and Design: The Unity of the Mahābhārata, edited by Vishwa Adluri and Joydeep Bagchee, 83–126. Leiden and Boston: Brill, 2016. Belvalkar, S. K. “Introduction.” In The Bhīṣmaparvan for the First Time Critically Edited. ix– cxxxii. Poona: Bhandarkar Oriental Research Institute, 1947. “Bhandarkar Oriental Research Institute, Poona Silver Jubilee Celebrations 4th and 5th of January 1943. Roger Lumley, N. C. Kelkar, J. R. Gharpure, R. N. Dandekar, C. Kunhan Raja, V. M. Apte, P. V. Bapat, A. P. Karmarkar, B. R. Kulkarni, H. D. Sankalia, C. R. Sankaran, S. R. Shende, Ludwik Sternbach.” Annals of the Bhandarkar Oriental Research Institute 24, no. 1/2 (1943): iii–xlvii. Bhattacharya, Pradip. “Problems with the Critical Edition: the Attempted Disrobing of Draupadī.” Paper presented at the 15th World Sanskrit Conference, New Delhi, January 5–10, 2012. Bigger, Andreas. “Balarāma im Mahābhārata.” Asiatische Studien/ Études Asiatiques xlviii, no. 1 (1994): 1297–99. ———. Balarāma im Mahābhārata: Seine Darstellung im Rahmen des Textes und seiner Entwicklung. Wiesbaden: Harrassowitz, 1998. ———. “The Normative Redaction of the Mahābhārata: Possibilities and Limitations of a Working Hypothesis.” In Stages and Transitions: Temporal and Historical Frameworks in Epic and Purāṇic Literature. Proceedings of the Second Dubrovnik International Conference on the Sanskrit Epics and Purāṇas, edited by Mary Brockington, 21–33. Zagreb: Croatian Academy of Arts and Sciences, 2002. Van Buitenen, J. A. B. “The Mahābhārata: Introduction.” In The Mahābhārata: I. The Book of the Beginning. xiii–xliv. Chicago: University of Chicago Press, 1973. 64  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Brockington, John. “The Bhagavadgītā: Text and Context.” In The Fruits of Our Desiring: An Inquiry into the Ethics of the Bhagavadgītā for Our Times, ed. Julius Lipner, 28–47. Calgary: Bayeux, 1997. Reprinted in The Bhagavad Gita: A New Translation, Contexts, Criticism. Translated by Gavin Flood and Charles Martin. Edited by Gavin Flood, 135–53. Norton Critical Editions. New York: W. W. Norton & Company, 2014. ———. “Is the Script Relevant? Further Evidence from a Nevārī-Script Rāmāyaṇa Manuscript.” In Epic and Argument in Sanskrit Literary History. Essays in Honor of Robert P. Goldman, edited by Sheldon Pollock, 17–29. Delhi: Manohar, 2010. ———. “Regions and Recensions, Scripts and Manuscripts: The Textual History of the Rāmāyaṇa and Mahābhārata.” https://www.college-de-france.fr/site/jean-noel- robert/symposium-2016-12-06-14h00.htm (accessed June 30, 2017). ———. Review of Untersuchungen zur Komposition von Reden und Gesprächen im Rāmāyaṇa, by Renate Söhnen. Journal of the Royal Asiatic Society of Great Britain and Ireland 114, no. 1 (1982): 65. ———. Review of Viṣṇudharmāḥ: Precepts for the Worship of Viṣṇu Part 1: Adhyāyas 1–43, by Reinhold Grünendahl. Journal of the Royal Asiatic Society of Great Britain and Ireland 2 (1984): 291–92. ———. Righteous Rāma, The Evolution of an Epic. New York: Oxford University Press, 1984. ———. The Sanskrit Epics. Leiden: Brill, 1998. ———. “The Text of the Rāmāyaṇa.” Indologica Taurinensia 15–16, Proceedings of the Seventh World Sanskrit Conference, Leiden, August 23–29, 1987 (1989–1990): 79– 90. ———. “Textual Studies in Vālmīkī’s Rāmāyaṇa.” Journal of the Asiatic Society of Calcutta 28, no. 3 (1986): 14–24, reprinted in Epic Threads: John Brockington on the Sanskrit Epics, edited by Greg Bailey and Mary Brockington, 195–206. New Delhi: Oxford University Press, 2000. ———. “The Textualization of the Sanskrit Epics.” In Textualization of Oral Epics, edited by Lauri Honko, 193–216. Berlin and New York: Mouton de Gruyter, 2000. Brockington, John, and Mary Brockington. Rāma the Steadfast: An Early Form of the Rāmāyaṇa. London: Penguin Books, 2006. 65  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches “The Completion of The Critical Edition of the Mahābhārata.” Annals of the Bhandarkar Oriental Research Institute 47, no. 1/4 (1966): i–xiv. Contini, Gianfranco. “La «Vita» francese «die Sant’Alessio» e l’arte di pubblicare i testi antichi (1970).” In Frammenti di filologia romanza: Scritti di ecdotica e linguistica (1932–1989). Edited by Giancarlo Breschi. Vol. 2, 957–85. Florence: Edizioni del Galluzo, 2007. Edgerton, Franklin. “Introduction.” In The Sabhāparvan for the First Time Critically Edited. ix–l. Poona: Bhandarkar Oriental Research Institute, 1944. Edwards, A. W. F. Cogwheels of the Mind: The Story of Venn Diagrams. Baltimore and London: Johns Hopkins University Press, 2004. Eklund, Sten. “The Traditional or the Stemmatic Editorial Technique?” Eranos 104 (2006– 2007): 5–18. Franco, Eli. Review of The Nay Science: A History of German Indology, by Vishwa Adluri and Joydeep Bagchee, South Asia: Journal of South Asian Studies 39, no. 3 (2016): 695–713. Griffith, John G. “A Taxonomic Study of the Manuscript Tradition of Juvenal.” Museum Helveticum 25, no. 2 (1968): 101–38. Grünendahl, Reinhold. “Zur Klassifizierung von Mahābhārata-Handschriften.” In Studien zur Indologie und Buddhismuskunde, edited by Jens Uwe-Hartmann and Petra Kiefer-Pülz, 101–30. Indica et Tibetica 22. Bonn: Indica et Tibetica Verlag, 1993. von Hinüber, Oskar. Review of The Sanskrit Epics, by John Brockington. Wiener Zeitschrift für die Kunde Südasiens 46 (2002): 268–69. ———. “Der vernachlässigte Wortlaut. Die Problematik der Herausgabe buddhistischer Sanskrit-Texte, Zur Überlieferung, Kritik und Edition alter und neuerer Text.” In Kleine Schriften, Teil I, edited by Harry Falk and Walter Slaje, 431–50. Wiesbaden: Harrassowitz, 2008. Holtzmann, Adolf Jr. Zur Geschichte und Kritik des Mahābhārata. Kiel: C. F. Haessler, 1892. Hopkins, Edward Washburn. The Great Epic of India: Character and Origin of the Mahābhārata. New York: Charles Scribner’s Sons, 1901. 66  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches “International Conference on Mahabharata Manthan: A Critical Revisit to Tangible and Intangible Heritage, 19th–21st July 2017, Resolution [sic] & Recommendations.” Computer printout, 2017. Irigoin, Jean. “Quelques réflexions sur le concept d’archétype.” Revue d’histoire des textes 7 (1977): 235–45. Ježić, Mislav. “Vedic Myths and Epico-Purāṇic Mythology—Transformation of the Sense of Vision Into Narration.” Paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 12, 2017, IUC, Dubrovnik, Croatia. Krnic, Krešimir. “Squaring the Circle: Commentarial and Intratextual Explanation of Controversial Places in Vālmīki’s Rāmāyaṇa.” Paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 14, 2017, IUC, Dubrovnik, Croatia. Lassen, Christian. “Beiträge zur Kunde des Indischen Alterthums aus dem Mahâbhârata I: Allgemeines über das Mahâbhârata.” Zeitschrift für die Kunde des Morgenlandes 1 (1837): 61–86. Leonardi, Lino. “Il testo come ipotesi (critica del manoscritto-base).” Medioevo romanzo 35 (2011): 5–34. Mahadevan, T. P. “The Śakuntalā-Yayāti ‘Transposition’, the Southern Recension of the Mahābhārata, and V. S. Sukthankar.” In The Churning of the Epics and Purāṇas at the 15th World Sanskrit Conference, edited by Simon Brodbeck, Alf Hiltebeitel, and Adam Bowles, 46–70. New Delhi: Rashtriya Sanskrit Sansthan and D. K. Printworld. ———. “Three Rails of the Mahābhārata Text Tradition.” Journal of Vaishnava Studies 19, no. 2 (2011): 23–69. Mehendale, M. A. “The Critical Edition of the Mahābhārata: Its Achievement and Limitations.” Annals of the Bhandarkar Oriental Research Institute 88 (2007): 1–16. ———. “The Critical Edition of the Mahābhārata: Its Constitution, Achievements, and Limitations.” In Text and Variations of the Mahābhārata: Contextual, Regional and Performative Traditions, edited by Kalyan Kumar Chakravarty, 3–23. New Delhi: National Mission for Manuscripts, 2009. 67  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches ———. “Interpolations in the Mahābhārata.” Annals of the Bhandarkar Oriental Research Institute 82, no. 1/4 (2001): 193–212. Phillips-Rodriguez, Wendy J. “The Evolution of a Sanskrit Epic: Some Genetic Considerations about Scripts.” In The Evolution of Texts: Confronting Stemmatological and Genetical Methods, edited by C. Macé, P. Baret, A. Bozzi, L. Cignoni, 175–90. Linguistica Computazionale 24–25. Pisa: Istituti editoriali e poligrafici internazionali, 2006. Phillips-Rodriguez, Wendy, C. J. Howe, and H. F. Windram. “Chi-Squares and the Phenomenon of Change of Exemplar in the Dyūtaparvan.” In Sanskrit Computational Linguistics, edited by Gérard Huet, Amba Kulkarni, and Peter Scharf, 380–90. Heidelberg: Springer, 2009. Pollock, Sheldon I. “Introduction.” In The Rāmāyaṇa of Vālmīki: An Epic of Ancient India. Edited by Robert P. Goldman. Vol. 3: Āraṇyakāṇḍa. Translated by Sheldon I. Pollock. 3–84. Delhi: Motilal Banarsidass, 2007. Reeve, Michael D. “Stemmatic Method: ‘Qualcosa che non funziona?’” In The Role of the Book in Medieval Culture: Proceedings of the Oxford International Symposium, September 26–October 1, 1982, edited by P. Ganz, 57–70. Bibliologia 3. Turnhout: Brepols, 1986. Von Simson, Georg. “Die Einschaltung der Bhagavadgītā im Bhīṣmaparvan des Mahābhārata.” Indo-Iranian Journal 11, no.3 (1969): 159–74. Söhnen, Renate. Untersuchungen zur Komposition von Reden und Gesprächen im Rāmāyaṇa, 2 vols. Studien zur Indologie und Iranistik 6. Reinbek: Dr. Inge Wezler, Verlag für orientalistische Fachpublikationen, 1979. Soulen, Richard N., and R. Kendall Soulen. Handbook of Biblical Criticism, 4th ed. Louisville, KY: Westminster John Knox Press, 2011. Sukthankar, V. S. “Epic Studies VII: The Oldest Extant Ms. of the Ādiparvan.” Annals of the Bhandarkar Oriental Institute 19 (1938–39): 201–62. ———. “Introduction.” In The Āraṇyakaparvan for the First Time Critically Edited, vol. 1. i– xxxviii. Poona: Bhandarkar Oriental Research Institute, 1942. ———. On the Meaning of the Mahābhārata. Bombay: Asiatic Society, 1957. ———. “Prolegomena.” In The Ādiparvan for the First Time Critically Edited, vol. 1. i–cx. Poona: Bhandarkar Oriental Research Institute, 1933. 68  John Brockington and the Sanskrit Epics: Limits of Statistical Approaches Sukthankar, V. S. et al., eds. The Mahābhārata for the First Time Critically Edited, 19 vols. Pune: Bhandarkar Oriental Research Institute, 1933–66. Sullivan, Bruce M. “An Overview of Mahābhārata Scholarship: A Perspective on the State of the Field.” Religion Compass 10, no. 7 (2016): 165–75. Szczurek, Przemyslaw. “In Search of the Beginning of the Strīparvan.” Paper presented at the Eighth Dubrovnik International Conference on the Sanskrit Epics and Purāṇas (DICSEP 8), September 13, 2017, IUC, Dubrovnik, Croatia. Trovato, Paolo. Everything You Always Wanted to Know about Lachmann’s Method: A Non- standard Handbook of Genealogical Textual Criticism in the Age of Post-structuralism, Cladistics, and Copy-Text. Padua: Libreriauniversitaria, 2014. West, M. L. Textual Criticism and Editorial Technique. Stuttgart: B. G. Teubner, 1973. Wolf, F. A. Prolegomena to Homer. Translated with an introduction and notes by A. Grafton, G. W. Most, and J. E. G. Zetzel. Princeton, NJ: Princeton University Press, 1985. 69