by D.D. Kosambi
Poona, India
Journal of the American Oriental Society
Vol. 66, No. 2, pp. 110-117 (8 pages)
(Apr-Jun, 1946)
NOTICE: THIS WORK MAY BE PROTECTED BY COPYRIGHT
YOU ARE REQUIRED TO READ THE COPYRIGHT NOTICE AT THIS LINK BEFORE YOU READ THE FOLLOWING WORK, THAT IS AVAILABLE SOLELY FOR PRIVATE STUDY, SCHOLARSHIP OR RESEARCH PURSUANT TO 17 U.S.C. SECTION 107 AND 108. IN THE EVENT THAT THE LIBRARY DETERMINES THAT UNLAWFUL COPYING OF THIS WORK HAS OCCURRED, THE LIBRARY HAS THE RIGHT TO BLOCK THE I.P. ADDRESS AT WHICH THE UNLAWFUL COPYING APPEARED TO HAVE OCCURRED. THANK YOU FOR RESPECTING THE RIGHTS OF COPYRIGHT OWNERS.
[Summary. This note, based upon a numerical and statistical analysis of the first five books (parvans) of the Mahabharata (Mbh) now available in the critical edition1 [The Mahabharata ... critically edited by V.S. Sukthankar (Poona, Bhandarkar Oriental Research Institute, 1933--).] presents evidence for the view that the Parvasamgraha (par.) figures give very accurate counts of some recension [a revised edition of a text]. The critical edition approximates to this recension, but in general reaches a little behind it. On the present manuscript evidence, there can be no hope of restoring the Par. recension itself. During the course of the study, the concept of textual fluidity can be given a somewhat more precise definition than hitherto, as regards the Mbh.]
1. The first three chapters (adhyayas) are palpably among the final additions made to the work during the long course of its original expansion. This is generally admitted because of the duplicated opening: lomaharsanaputra ugrasravah sutah pauraniko naimisaranye saunakasya kulapater dvadasavarsike satre (1.1.1. and 1.4.1) which occurs at the beginning of the first as well as of the fourth chapter. A reference to Mr. E.D. Kulkarni's counts2 [E.D. Kulkarni, The Parvasamgraha Figures, following this paper.] following this paper will show that these chapters are very much longer than any of the rest in the Adi [the beginning, and so forth]. Modern statistical analysis of the type utilized towards the end of this paper shows that there is very little chance of these three chapters belonging to the same distribution of adhyaya[chapter]-lengths as the remaining 222 in the same Book. To make the position clearer, however, it is worth while demonstrating that these three chapters (each of which is also a sub-parvan) are remarkably self-contradictory even for an epic that pays little attention to logical development of its narrative.
The first chapter, the Anukramani, corresponds to what would be, in a modern work, a combination of prefaces to several editions, reviews of the work itself, and the publisher's advertisements. The status of the work is very curiously described. The bard (suta) Ugrasravas, son of Lomaharsana, is versed in mythology (pauranika), so that no surprise need be felt when the narrative (akhyana) is called a purana (1.1.15); it was originally composed by the dark (krsna) Dvaipayana (1.1.8, 15). This puranic contention, incidentally, is borne out by the studies of Ruben, 3 [W. Ruben, JRAS 1941. 247-56 and 337-58; also A Volume of Eastern and Indian Studies presented to F. W. Thomas 188-203. See also M.S. Vaidya, ABORI Silver Jubilee volume 609-20.] which prove the claim of the Mbh. to be regarded as a purana, or even the purana. But the work is also a history (Itihasa) in 1.1.16; this too may pass muster because of the historical or pseudo-historical character of the main event, the great war described in the epic. The poem then rises a step higher in the scale of sanctity to parade as a samhita [the most ancient layer of text in the Vedas, consisting of mantras, hymns, prayers, litanies and benedictions.] (1.1.61), then as an upanisad [late Vedic Sanskrit texts of religious teachings which form the foundations of Hinduism. They are the most recent part of the oldest scriptures of Hinduism, the Vedas, that deal with meditation, philosophy, and ontological knowledge] (1.1.191), and finally reaches the peak of sanctity as a veda (1.1.204). There is nothing to show how or why this progressively higher status was achieved; the work is at one and the same time all of these. Yet the accepted divisions (of scripture) as given by Patanjali (Mahabhasya 1.1.1.5, Kielhorn, p. 9, 1. 20-25), or even by the Buddhists (Milindapanho, Trenckner, pp. 3, 178), show that purana, itihasa, upanisad, and veda are all distinct; it would be impossible for one work to appear simultaneously as two or more of these if the older tradition were followed. Moreover, the function of this variously described composition is shockingly different from that of any of the authentic Brahminical scriptures. It is in fact to delineate [describe] the three sastras of dharma, artha, kama; as it were, ethics, economics, erotics (a poor equivalent for kama). To anyone who took seriously the injunctions laid down for the observance of physical and mental purity during the study of sacred texts, such a claim would be blasphemous when made for any document on the level of a veda or even of an upanisad.
Still more interesting is the nomenclature and extent of the work. To be sure, it is called the Mahabharata near the beginning (1.1.10), but the name throughout is merely the Bharata [capacity; loading/load; quantity]. The adjective is explained at the end (1.1. 209-9). The greatness is not derived from the extent but from its importance and weight, for it outweighed all the four vedas taken together when the gods and sages put it in the other pan of the balance. Thus, the maha is not an official part of the original title, just as the adjective divina was prefixed to Dante's Commedia by a later enthusiastic publisher. Moreover, we are told that the learned recognize as the Bharata a work of 24,000 (stanzas or sloka), without any of the minor episodes (upakhyana); to this the sage added an Anukramani chapter of 150 stanzas as an epitome (1.1. 61-2). Yet this composition of the sage, presumably Krsna-Dvaipayana-Vyasa, is the same Bharata as the present much more extensive work. The difference is trifling, the work is after all one which the twice-born possess in both the expanded and contracted forms: vistarais ca samasais ca dharyate yad dvijatibhih (1.1.25). The sage uttered it in both the expanded and contracted forms: vistiryaitan mahaj njanam rsih samksepam abravit (1.1.49). In fact, some are skilled in explaining the work, others only in sticking to the text: vyakhyatum kusalah kecid grantham dharayitum pare (1.1.51).
This can mean only thing thing: at the time the text of the critically edited Adi was composed there still existed a much shorter version of the epic which had full authority among the orthodox, and this new version could gain its own authority only by claiming to be the same as the shorter work in a more understandable form, not by ignoring the shorter work or by claiming to be a new one. This would be impossible for a veda or an upanisad, whose texts, at least in theory, are unalterable. So claims to special merit are made again and again, both for the whole of the present version (1.1. 191-8) and for the new beginning, the Anukramani itself (1.1. 999, 200, 206). The rather mixed claims to religious merit and scriptural equivalence surely indicate efforts made to gain sanction for what was, after all, a popular work -- a compendium of ethics, economics, erotics -- meant for persons of little learning whereas the real veda shied at such: bibhety alpasrutad vedo (1.1.204). The position at the time of composition must have been precarious, both as to authenticity and religious importance of the text we now possess.
2. This same performance, rather shifty as well as self-contradictory, continues in the second chapter, which is a list of contents of the entire work. Here we learn that the work as originally composed by Vyasa in a hundred parvans [books of the Mahabharata] was recast by the bard into eighteen and recited in that form in the Naimisa forest (1.2.70-1):
etat parvasatam purnam vyasenoktam mahatmana yathavat sutaputrena lomaharsanina punah kathitam naimisaranye parvany astadasaiva tu samaso bharatasyayam tatroktah parvasamgrahah
The list of original parvans does actually add up to a hundred, dispelling the controversies raised by those who based their conjectures on editions other than the critical. But there is a fundamental contradiction nevertheless in the entire passage, which is supposed to have been recited by Lomaharsani himself, beginning with 1.2.15. The beginning is innocuous enough, as the sages ask the bard to define the extent of an army corps (aksauhini) and he gives the size in quite fantastic numbers; then he says that eighteen such divisions were destroyed in the great war, gives the extent of the original narrative, and then says himself that he recited the work in its revised form before the sages -- when he has hardly begun the recital! However one may attempt to smooth the difficulty over, it is impossible to explain the contradiction entirely away. The use of the past participle kathitam spoils the game, which could have been saved by using the future tense.
Furthermore, when the number of chapters in the revised first book, the Adiparvan (which does not figure in the original hundred), comes to be given, the division is again passed off as Vyasa's composition (1.2. 95-6). It remains to note that the Par. also ends with praise of the Bharata, the merit to be acquired by its recitation or by hearing it recited. The penultimate sloka (1.2. 242) is in the vasantatilaka metre, the final one being a praharsini (1.2. 243). Though both have the proper ending [x], they are unquestionably later 'fancy' metres, being almost unique examples in the critical text. Therefore, the boast that the narrative is the sustenance of all the leading poets (1.2. 241) also becomes suspiciously indicative of late composition.
In the third chapter, the Pausyaparvan, one need only note the horrible mess that has been made of the hymn to the Asvins. According to Renou,4 [L. Renou, A Volume of Eastern and Indian Studies presented to F.W. Thomas 177-87.]
'On sait que ce poeme imite deliberement la maniere des hymnes vediques ... Bref, si le style est pseudo-vedique, le rythme est upanisado-pali.' [Google translate: We know that this poem deliberately imitates the manner of vedic hymns ... In short, if the style is pseudo-vedic, the rhythm is upanisado-pali.]
This, with the extraordinarily corrupt text, goes to show that those who transmitted the narrative did claim to be Brahmins, did attempt to substantiate their further claim that this work was a veda, or at least had vedic embellishments: vedarthair bhusitasya ca (1.1. 16). But their scholarship was of no high order. We are not dealing here with a revival of Hinduism of the sort that must have occurred in the days of Pusyamitra and Patanjali, but a resurgence in a new form, oriented towards the people rather than based upon the possession of obscure and archaic ritual.
Yet, with these considerable doubts on the fitness and consistence of these prefatory chapters, we note that the manuscript evidence (apart from correspondences as between 1.1 and 1.56) completely proves their authenticity [??]; in fact even their popularly accepted text survives with comparatively trifling changes. This bring us to our real problem: What weight should we attach to the detailed figures given in the Par.? What do the numbers given for each book of the whole Mbh. indicate?
3. The late Dr. V.S. Sukthankar considered as negligible the editorial uses of the Par [book list].5 [Prolegomena to his Adiparvan, esp. xcvii-c; Introduction to his Aranyakaparvan; ABORI Silver Jubilee volume 549-58).] His objections to taking the Par. seriously may briefly be summarized as follows: (1) What text of the Par. [book list] should the editor take as authentic from so many divergent manuscript readings? (2) What would be the meaning of the unit sloka [a couplet of Sanskrit verse, especially one in which each line contains sixteen syllables], generally taken to indicate a stanza [verse], when the text contains all sorts of metres and long prose passages? What is the proper method of counting? (3) In any case the text contains so many variants that even if the Par. [book list] figures were unique and there were no difficulty about the method of counting, a sound critical method would be needed to determine the proper reading. The same method, however, would fix the entire text without recourse to the Par. [book list] (4) The text is fluid, so that its absolutely precise determination should not be considered as a serious possibility. Finally, he gave it as a guess that the Par. [book list] figures were probably obtained by counting a few pages at the beginning and end of the manuscripts carefully, and estimating the remainder by averaging the number per folio.
Now this last explanation of his contradicts (2) above in that it implies the existence of some method and unit of counting. In fact, Sukthankar himself believed in Haraprasad Sastri's traditional interpretation (Adiparvan xcix) that the sloka here was to be taken in its restricted meaning of 32 syllables, and everything in the manuscript down to the numerals counted accordingly. As to (1), the critic might very well ask why Sukthankar did not check his own critically edited Par. [book list] against the actual contents of the books of the critical edition. As a matter of fact, the editors (among them S.K. De who noticed that the numbered slokas in his Udyoga amount to 6063, far shorter than any Par. [book list] figure) pay very little attention to the Par. [book list] Sukthankar's final paper shows his state of doubt clearly, when he attempts to settle the Par. [book list] figures by comparison of his own critical text with the Javanese and the Southern recensions; this contradicts his own usually rigorous and brilliantly applied method, in that his text of the Par. [book list] is based on an extensive manuscript apparatus whereas the other two are single versions, edited by rather doubtful methods.
The one real, valid objection is, of course, (3) above: that a sound canon of text-criticism would be needed to select the proper readings from a vast and conflicting array of manuscripts, and this by itself determines the entire text without any attention to the Par. The question that still remains would be that of the original function of the Par. itself, not of its use to the editor. No criticism made here is to be taken as directed in any way against Sukthankar's editorial method, for which any intelligent person could have nothing but the highest admiration. If we can see a considerable distance through the epic wilderness today, it is only because of the path he cleared for us with giant stride.
The theory of the fluid text is more difficult to deal with. The idea seems intuitively clear to any Indian without further explanation, though a western scholar like Edgerton naturally dismisses it as meaningless. In any case, it is beyond doubt impossible to restore the absolute archetype. For example, suta uvaca may be replaced in the recitative directions by sautir uvaca, or lomaharsanir uvaca, and to the uncritical Indian eye (which alone roamed over the manuscripts till a century ago) [??] there is no real difference. To the question 'which is the real reading?' the reply would be 'any or all of them!' But was there any fluidity about the slokas themselves? Should not the wavy line that runs through so much of the critical text be taken as evidence of fluidity? In justice to Sukthankar, it should be said that in spite of his undefined concept of fluidity, he himself defended his own critical text, even the wavy line readings, with consummate skill, as may be seen in so many of his epic studies. I believe I owe it to his memory to interpolate here a personal record which will make it clear that the theory of the 'fluid text' never interfered with his faith in the superb critical method. When the text of the Adi [the beginning, and so forth] was virtually complete, I questioned him as to the possibility of invalidation by some future discovery of a manuscript much older than any used in his critical apparatus. He considered the idea in his usual calm, judicial manner and said that the possibility could certainly not be denied. Then with a sudden flash of the inner spirit, so rare with him, he burst out, 'No! This text will stand no matter what discoveries are made later on. At the very most, an occasional wavy line reading may be exchanged for another in the critical apparatus.' I had the good fortune of reminding him of this remark, which he had himself forgotten, when the famous Rajaguru manuscript was discovered in Nepal, to confirm in spite of its antiquity not only his accepted readings but many even of his judicious emendations which were not to be found in any manuscript of the original critical apparatus.6 [V.S. Sukthankar, Epic Studies VII, in ABORI 19. 201-62 (1938).] So, his undefined concept of the fluid text cannot be of any use to us here.
4. The most striking feature of the criticism levelled against Sukthankar's neglect of the Par. [Book list], and of his own defence, was the total absence on both sides of any numerical data. The critical text had never been properly counted. Having admitted the justice of this criticism of mine (made in private conversation), Sukthankar directed his staff of learned scribes, pandits, and sastris in the collation department to count the edited text by any method they considered traditional. The work was interrupted by his sudden and untimely death. Whatever had been done was handed over to me, but a brief check sufficed to convince me that the data so gathered was useless. Besides actual errors of counting and arithmetic, the omission of adhyaya [chapter] colophons and the uvaca ["said"] insertions made serious differences inevitable. My own supplementary counts, however, made it clear that the Par. [book list] was a document to be taken seriously, and for that reason the entire published text had to be counted again by some one who would take the utmost care, but who had no theory to prove or disprove as I did.
Mr. Kulkarni took up the problem at my suggestion, and I have every reason to believe that his work is satisfactory, on both the above scores. My criterion for the counts was derived from the recitative purpose of this (or any other ancient Indian) text: every syllable that is pronounced must be counted. Thus, only the avagraha, the numerals at the end of each sloka, the adhyaya-heading numbers, and the vertical bars that terminate each half-sloka had all to be taken as silent punctuation marks, therefore not counted. There does, therefore, remain a slight ambiguity; but the result, it seems to me, justifies my reasonable criterion. Taking the counts of just five books out of a total of eighteen would seem rather premature, but there remains no alternative to one who is not certain of surviving to the completion of a critically edited text. The results may be summarized in
Table I
Books / Adi / Sabha / Aranyaka / Virata / Udyoga
Par. chapters / 218 / 72 / 269 / 67 / 186
Actual chapters / 225 / 72 / 299 / 67 / 197
Par. slokas / 7984 / 2511 / 11664 / 2050 / 6698
'Official' stanzas /7196 / 2390 / 10316 / 1824 / 6063
syllable/32 / 7964.125 / 2599.53125 / 11421.000 / 2036.000 / 6754.4375
What do these figures tell us?
In the first place, the chapters are never less than the number indicated in the Par., actually greater in three out of the five cases; moreover, the adhyaya[chapter]-number automatically counts itself by means of the colophons. If this indicates changes in the text as such, they should naturally be in the direction of inflation, not contraction. But if we count the official number of stanzas, remembering that even the prose passages are divided into convenient though not exactly regular lengths and numbered as slokas, we find that in every case without exception the observed number comes out to be far below that expected from the Par. The different is very highly significant, using the statistician's chi-square test. So, the method of counting did not take sloka as a stanza. Taking the total number of syllables divided by 32, we see at once that we come extraordinarily close to the Par. sloka-number, and that the text we now possess does actually represent something of very nearly the Par. [book list] lengths. The adhyaya[chapter] discrepancy may be explained by the fact that a chapter is a recitative unit, not a narrative one; many of them break off in the middle of some incident and the next begins at the right place. Hence, the most plausible interpretation would be that the text length has been restored to an excellent approximation, but that the adhyayas[chapters] were broken up into more convenient lengths since the Par. days, hence could not be restored as neatly as the text.
The method of counting, therefore, may be taken as reasonably well determined. The 32 syllable units is traditional even in prose counts of Jaina and Mahayana Buddhist works; according to my colleague Dr. V.V. Gokhale, the latter take asvasa to mean the unit of 32 syllables, explaining it as the number that can conveniently be uttered in one breath. The method of counting may ultimately be traced back to the dual system to be seen in the weights of Mohenjo Daro. The work of Pingala on the Vedic metre gives a highly artificial series progressing by differences of four syllables; the quadragesimal system survives to the present day in Indian currency, and is used in accounting with its own special notation. The statistical argument of the previous paragraph is that while 'official' stanzas as sloka would give, except for Sabha [Book 2], probabilities of less than one in a million (by chi-square), the 32 syllable unit gives a significant discrepancy only for the Aranyaka [Book 3], where the probability is about 2.75 percent of so great a difference in either direction from the expected total of 11,664. Finally, if we count the very shortest of all parvans, the Mahaprasthanika (no. 17), from the Vulgate, we find its stanzas to be 109, whereas the 32 syllable unit gives 120 and a fraction over, very close indeed to the Par. figure of 120 slokas. It cannot be argued that so short a parvan had to be counted by some method of approximation.
In fact, it is rather difficult to settle the conjecture of Sukthankar's (ABORI Silver Jubilee volume, 550) that the sloka counts were estimated by folio averages. Taking modern printed and lithographed pothi-styled works and my own random sample counts as a guide, I conclude that the Adi figures, for example, could have been estimated to about 15 slokas (taking 2s as the confidence interval). With Gupta period copper plate inscriptions, the estimate could have been made to about 25 slokas. But this is very misleading in that the line and not the page is taken as my fundamental unit; or what amounts to the same thing, a constant number of lines per page has been assumed. I have seen no long manuscript where there are no gaps, no abbreviations, and the number of lines is exactly the same from page to page; so, such accurate estimation would seem definitely out of the question. In any case, none of the Par. figure variants can possibly have been obtained by such estimation because these variants make a point of changing the Par. reading as little as possible. The last two digits are usually retained, some change being made in the hundreds or the thousands to bring the number nearer to the actual bulk of the parvan in the recension concerned. I feel justified in maintaining that the original Par. figures must have been obtained by the diaskeuast [editor] from accurate counts, syllable by syllable, of some text before him. It follows that he had a recension which he considered as the norm. The puranic taste for large numbers and grasp of the necessary arithmetic are both demonstrated in the aksauhini [described in the Mahabharata as a battle formation consisting of 21,870 chariots (Sanskrit ratha); 21,870 elephants (Sanskrit gaja); 65,610 horses (Sanskrit turaga) and 109,350 infantry (Sanskrit pada sainyam) as per the Mahabharata (Adi Parva 2.15-23).[1][2] Thus one akshauhini consisted of 218,700 warriors (not including the charioteers, who didn't fight).] definition which precedes the counts (1.2. 15-23)
5. What interpretation should be give to the accurate sloka counts coexisting with manifestly wrong adhyaya[chapter] figures?
As has been said before, the adhyaya[chapter] number has generally been increased for recitative purposes. The average number of slokas calculated from the Par. figures themselves would be, for the five parvans before us: 36.62; 34.86; 43.36; 30.6; 36.01. It is to be noticed that the highest average, the Aranyaka, is cut down to an actual 39.01 by the great addition of 30 adhyayas[chapters]. The Udyoga average was reduced by the increase of 11 in adhyaya[chapter] number, the Adi has only seven more chapters than the Par. text gives, but we must note that the average length is still further reduced if one leaves out the first three -- palpably later -- chapters which are inordinately long. The two shortest parvans, Sabha [Book 1] and Virata [Book 4], have also the lowest mean value for chapter length, and it is to be noted that they have exactly the same number of adhyayas[chapers] as originally given by the Par. It will be interesting to see whether this tendency towards reduction of adhyaya[chapter]-length to a more convenient average is present in later parvans. The entire work, according to the Par., has 82,136 slokas in 1948 chapters, which gives the average of 42.16, much higher than in any book of the critical text so far available.
Every additional adhyaya[chapter] means, even if no sloka is changed, increase in the text by at least one colophon. Yet our critical edition shows reduction in the sloka number except in the Udyoga [Book 5] parvan, where the difficulties of both editor and tabulator are increased by such passages as the Sanatsujatiya (5. 42-45), and in Sabha [Book 2], discussed later. The discrepancy is most serious for the Aranyaka [Book 3], where all Indian variants of the Par. agree as to 11,664 slokas, and only the Javanese version gives a different reading which, incidentally, is significantly closer to the number actually counted. This parvan is remarkable (Introduction to the Aranyakaparvan xvi) in that here the Northern recension is longer than the Southern, contrary to all other examples so far known.
Incomplete as the evidence is, one may venture upon a plausible conjecture to explain these observed facts. The Mbh. in its formative period was the property of a single Brahminic gens, the Bhargavas7 [V.S. Sukthankar, Epic Studies VI, in ABORI 18. 1-76 (1936).] The story was becoming very popular, having spread to South India by the end of the fourth century, as is seen by Buddhaghosa's reference and the injunction that it was not lawful for good Buddhists to attend such recitals: Dighanikaya Atthakatha, PTS 1.84: Akkhanam ti Bharata-Ramayanadi; tam yasmim thane kathiyati, tattha gantum na vattati. Before the completion of the second century of the Gupta era our epic has gained tremendous authority; Vyasa is quoted in making land grants, and the Mahabharata is cited as such by name. We have come far from the days when the ridiculous etymology of 1.1. 209 mahattvaa bharavattvac ca mahabharatam ucyate was necessary. But the body of the narrative must have been complete by the beginning of the fourth century if Puranas recast in their present form by 330 A.D. 8 [F.E. Pargiter, The Purana Text of the Dynasties of the Kali Age (Oxford, 1913).] could take the Mbh. text as base.9 [See fn. 3.]
Chronology
Newly discovered Puranas manuscripts from the medieval centuries has attracted scholarly attention and the conclusion that the Puranic literature has gone through slow redaction and text corruption over time, as well as sudden deletion of numerous chapters and its replacement with new content to an extent that the currently circulating Puranas are entirely different from those that existed before 11th century, or 16th century.
For example, a newly discovered palm-leaf manuscript of Skanda Purana in Nepal has been dated to be from 810 CE, but is entirely different from versions of Skanda Purana that have been circulating in South Asia since the colonial era. Further discoveries of four more manuscripts, each different, suggest that document has gone through major redactions twice, first likely before the 12th century, and the second very large change sometime in the 15th-16th century for unknown reasons.
The different versions of manuscripts of Skanda Purana suggest that "minor" redactions, interpolations and corruption of the ideas in the text over time.
Rocher states that the date of the composition of each Purana remains a contested issue. Dimmitt and van Buitenen state that each of the Puranas manuscripts is encyclopedic in style, and it is difficult to ascertain when, where, why and by whom these were written:As they exist today, the Puranas are a stratified literature. Each titled work consists of material that has grown by numerous accretions in successive historical eras. Thus no Purana has a single date of composition. (...) It is as if they were libraries to which new volumes have been continuously added, not necessarily at the end of the shelf, but randomly.
— Cornelia Dimmitt and J.A.B. van Buitenen, Classical Hindu Mythology: A Reader in the Sanskrit Puranas[117]
Forgeries
Many of the extant manuscripts were written on palm leaf or copied during the British India colonial era, some in the 19th century. The scholarship on various Puranas, has suffered from frequent forgeries, states Ludo Rocher, where liberties in the transmission of Puranas were normal and those who copied older manuscripts replaced words or added new content to fit the theory that the colonial scholars were keen on publishing.
-- Puranas, by Wikipedia
Granting the first three chapters of the first book as final additions, one can only regard the Par. as an attempt to fix the text when an enormously increased demand was already in sight.
Even at the beginning of such a period of enhanced popularity, it would be most difficult to retain the text as the sole property of a small or at least a compact group of bards. Some attempt at fixing the text would, therefore, be logical and necessary. Now it is clear that not even the prodigious memory of a good Indian pandita could retain the entire work; but a parvan or two, say 10,000 slokas in all, would not be considered too much to memorize, as there are old-fashioned Brahmins even today -- in rapidly decreasing number -- who demand this of their boys. The memory of any person old enough to be considered as an authority represents, in the main, the tradition of some thirty years earlier, apart form possible defects of transmission arising in his own mind. The very existence of the Par. shows that there could have been no local versions (leaving aside the older and shorter Bharata) of the Mbh. then being recited, comparable in divergence to our present Northern and Southern recensions; but the need for a Par. with accurate sloka counts would seem to indicate that minor variants had already begun to be committed to writing. This would occur at a period when the work was still a popular recital, not a super-veda. That is, when the subject matter was more important than the exact words in which the story was recited. Thus, the Parvasamgraha recension would not be the original source of our present manuscript evidence, but these manuscripts would derive (at several removes) from various slightly divergent versions written, or at least memorized, a generation or so earlier than the text of the Par. diaskeuast [editor] attempted to fix. For, there was no way of putting a 4th century manuscript before the public in a vast number of copies, no way of erasing variants from human memories. The veterans would still have to be consulted when more copies were to be made; the copies do not necessarily derive from a manuscript, but are very likely to be taken down by dictation from memory. For all these reasons, the critical edition indicates, in my own opinion, a pre-Par. state of fluidity. We are not restoring the actual Par. recension in general. The process of taking the highest common factor from best available manuscripts reaches to some point, or rather to several points, on the parent stem of tradition from which both the Par. recension and our present manuscript evidence would have been derived. The result is in many ways analogous to the starred forms of the philologist which explains numerous extant cognate derivatives, though all the starred forms themselves need not have existed simultaneously in any single living language. Our critical edition, therefore, may not represent any particular version or recension recited at one time, let alone the Par. recension. The Mbh. was as living as any language, as changeable; the wavy lines in our printed text may go back to a stage behind the Par. But even further, the text as reduced to print in the critical edition may be taken as the starred form of a great and living tradition. This is the meaning we must ascribe to Sukthankar's concept of fluidity, and to his idea that an absolute archetype would remain an ideal impossible of attainment (Adiparvan ciii).
Sabha [Book 2] needs special comment as its editor does not subscribe to the fluid text theory and believes (this is my impression, as the volume has not yet been issued) that the archetype has been restored in all essentials. This may certainly be true, but the archetype cannot possibly be the version before the original composer of the Par. For, in the first place, the serious excess of 88 slokas and 17 letters is still less serious than the default found if sloka is interpreted as 'official' stanza. The discrepancy between our counts and the Par. figure remains much too large to be explained by Sukthankar's approximation theory, for any reasonable method of approximation would give much closer results. The only variant in Par. readings that might help us here is the 2611 of D1, and this is not likely to be received with any favour, seeing the uniformity of the most reliable manuscripts, and the poor show that the D group makes in contrast with the rest. There are no extra adhyayas[chapters], and there seem to be no specially difficult passages. All that can be said is that the tristubh [a Vedic meter of 44 syllables (four padas of eleven syllables each), or any hymn composed in this meter. It is the most prevalent meter of the Rigveda.] stanzas, particularly the 'hypermetric' ones, tend to bunch towards the end of the parvan, though it is difficult to see how this helps in the matter under consideration. It should not be taken as an unreasonable conclusion if one decides that Sabha derives from a later or at least more inflated version than the Par. recension [revised edition], just as Aranyaka seems to have developed from an older one. It may be noted in passing that the actual figure is very near to the 2590 in the colophon of S1, (perhaps repeated from the missing S1 Par.) the manuscript being the one most favoured by our editors.
6. Something may be done from more refined statistics applied to the study of adhyaya[chapter]-lengths. Mr. G.U. Yule1 [G.U. Yule, Biometrika 30. 363-90 (1938).] has reached the conclusion, by use of Pearsonian statistics, that the Imitatio Christi shows sentence-lengths compatible with the style of Thomas a Kempis, not with that of Gerson (Groote being left out of consideration). We cannot discuss sentence lengths here, as the work is mostly in verse without modern punctuation. But adhyaya[chapter]-lengths may be so discussed. The statistics I use derives from R.A. Fisher; and I leave out all details and technical terms, merely giving enough description of the metho to show anyone who knows the subject how my conclusions were drawn; at the same time, the pure philologist will not be bored.
The distribution of adhyaya[chapter]-lengths is not normal (using the word always in its technical sense). The curves would be asymmetrical, having a long tail to the right, and a flatter top than in the familiar bell-shaped curve that everyone has seen. To overcome this, I use a well-known device, of taking logarithms. The distributions are again tested for normality. As, even in the worst cases, anormality is sensibly reduced, one can apply the technique known as analysis of variance to see whether the variation between sub-parvans is of the same order as that within a sub-parvan (of the same book). Normality of the entire distribution in logarithmic units would not necessarily prove, but would be compatible with, the hypothesis of composition by one person at one time; similarly, uniform sub-division into sub-parvans would be compatible with the idea that the recension was unitary or done by several similarly constituted person. The results are as follows.
The shortest parvans, Virata and Sabha, are both normally distributed and show uniform division, so that they may have been derived from works of unitary authorship; of course, no one would maintain that Virata, statistically the most satisfactory of all the books before us, represents the Par. recension, if only for the wavy line that runs so prominently through the printed text. Adi and Udyoga are anormal, the sub-parvans also vary materially among themselves in both cases, even when we leave out the first three chapters of Adi on grounds made clear at the beginning of this paper. Aranyaka is decidedly skew-positive platykurtic, but surprisingly enough bears no sign of uneven division into sub-parvans. That is, it may have been composed by different people at different times, but the process of subdivision was more or less uniform.