The Mahabharata: Its Genesis and Growth (A Statistical Study) - An Old and Rare Book

The transmission of the Mahabharata is characterized by a truly bewildering mass of versions and sub-versions. Indeed, no other work illustrates the dictum, “as many manuscripts so many texts” me aptly than the Great Epic. This phenomenon of luxuriant growth and indiscriminate fusion of versions can be explained only on the assumption that, after its composition, the Mahabharata had, in the course of all these centuries been handed down in different forms and sizes from bard to bard through its text against partial corruption and elaboration or against arbitrary emendation and normalization. Even in its early phases, the Mahabharata text-tradition seems to have been not uniform and singular but multiple and polygenous. And this is, indeed, what is to be expected of a work which has proved to be a vital force in the life of a dynamic people.

It was natural that, with the advance of scientific Indology, which had become evident in India and outside in the course of the past hundred years, the need should have felt a critical edition of the Mahabharata. The credit for first voicing this need, in more or less clear terms, goes to Winternitz, who, in his paper presented the 11th International Congress of Orientalists held at Paris I 1897, laid special stress on this desideratum of a critical edition of the Mahabharata “as the only sound basis for all Mahabharata studies, nay, for all tidies connected with the epic literature of India.” Winternitz, who was convinced that a “critical edition of the Mahabharata was a sine qua non for all historical and critical research regarding the great epic of India,” persisted in his efforts, as the result of which the International Union of Academies resolved in 1904 to undertake the preparation of such an edition. Funds were raised for the purpose, and a specimen of the edition prepared b Luders was actually published for private circulation among scholars. But then came the first world war which interrupted all scholarly work in the West. However, soon after the Bhandarkar Oriental Research Institute was founded in 1917, its workers, in their youthful enthusiasm, decided to venture upon the critical edition, making a fresh start.

Event when the Critical Edition of the Mahabharata had been first mooted, it was visualized as a multifaceted research project. It was to have completed in 1966: 19 volumes, 13000 pages), (2) the critical edition of the Harivamsa which was traditionally regarded as the khilaparvan of the epic (completed in 1971: 2 volumes, 1782 pages), (3) the Pratika index to the two critical editions ((completed in 1972: 6 volumes, 4768 pages), (4) the publication of only the critically constituted text (completed in 1976: 5 volumes, 3150 pages), and (5) the Epilogue (the work on which is currently in progress). As was but to be expected, the Critical Edition has inspired quite a number of research monographs based on it. Indeed. Some such monographs were published even while the Critical Edition had still been in the process of completion. But most of these monographs have dealt with either the linguistic-literary or the religio-philosophical or the socio-historical aspect of the Mahabharata. I have now pleasure in presenting to the world of scholars a study of the Critical Edition from an entirely different point of view. In The Mahabharata: its Genesis and Growth, Shri M.r. Yardi seeks to trace the whole history of the emergence of the Great Epic by means of a statistical analysis of its Anustubh style. I do not wish to preempt any judgment regarding his methodology and results, but I do certainly want to say this much, namely, that, with his widely acknowledged mastery of both Sanskrit and Mathematics- Statistics and with a similar highly appreciated work about Shakespeare already to his credit, Shri Yardi is best qualified to attempt this kind of approach to the Critical Edition. But what struck me most is the seep devotion with which Shri Yardi occupied himself all these years with labour of love, notwithstanding the many tempting offers of remunerative assignments which he had been receiving since his retirement from the Indian Civil Service.

Some years ago, the Bharndarkar Oriental Research Institute published Shri Yardi’s The Yoga of Patanjali, which, I am happy to state, has received enthusiastic approbation from many discerning scholars. I have no doubt that the present work will easily emulate its predecessor in this respect.

When the critical Edition of the Adiparvan was completed by Dr. V. S. Sukthankar in 1933, it was hailed as ‘the most Max Muller’s edition of the Ryveda with Sayaba’s commentary’. The Critical Edition contained a complete critical apparatus, giving not only the constituted text, but also the variant readings of the Maa. Utilized and spurious passages which were excluded from the constituted text, the latter being in the critical notes and the appendices. Dr. Sukthankar further appended a prolegomena to this edition, in which he gave full account of the Mss. Available, their classification and the principles followed in the constitution of the text. This Aranyakaparvan. The work relating to the other parvans was completed by 1966, with matching Zeal and scholarship, by his successors in accordance with the principles laid down by him. A student of this great epic has now available to him a really critical edition of the Mahabharata, based on an extensive and carefully selected Ms. Material collected from all over India.

It became obvious that the critically constituted text was not the original Bharata, which was recited by Vaisampayana at the sanke-sacrifice of king Janamejaya. Dr. Sukthankar himself did not claim that it was a reconstruction of Ur-Mahabharata. It is stated in the Epic itself that this original text did not contain any episodes or legends and consisted of only 24,000 stanzas. On the other hand, the size of the original Bharata as given in the Epic. Furthermore, passages which are suspected to be interpolations on intrinsic grounds, such as the childish stories of the old maid and the five Indras, ostensibly inserted to justify the polyandrous marriage of Draupadi or the Bhargava legends, remotely connected, if at all, with Bharata war, had to be retained in the constituted text, as they are to be found in all versions. As observed by Winternitz, these interpolations must have come to be added at some earlier period to which the manuscript tradition did not reach back. A different approach was, therefore, needed to identity such earlier interpolations and recover the original Bharata. It was thought that a statistical approach would enable us to go a step further than the Critical Edition, and identity not only the original Bharata but also the interpolation which came be made over time.

A statistical analysis of the Anustubh style in the Critical Edition has disclosed, by the application of the tests of the homogeneity, five styles which are significantly different from one another. According to the internal evidence too in the Mbh. The number of compilers studies. Among these only adhyayas belonging to one style give a self-contained but bare account of the Bharata war and the number of verses which they contain comes close enough to 24,000. This is the Ur-Mahabharata, which was later expanded by four person, Suta, Sauti, the Harivamsakara and the author of the Pervasangraha. It had also been possible to determine the successive additions made by the redactors by the same method.

I take this opportunity to thank those who have helped me in the preparation and publication of the this work as a humble token of my gratitude to my teacher, the later Professor D. D. Kosambi, who encouraged me to taken up the study of statistic even after I had got into the Indian Civil Service and this study possible. Bothe the late Acharya V. P. Limaye and my teacher Dr. R. N. Dandekar, both of Bhandarkar Oriental Research Institute, showed keen interest in the progress of my work from the very beginning. Acharya Limaya pressed me to taken up this work to the exclusion of all other interests and finish it as early as possible, but as ill-luck would have it, he did not remain alive to see its completion. I a, greatly indebted to Dr. Dandekar for reacting to the results of this study and making valuable suggestion from time to time, and accepting this work for publication. I have greatly benefitted by the discussions which I had with Prof. D. G. Dhavale, an astronomer, who has made a deep study of the astronomical passages in the Mbh. I am thankful to him, as also to Dr. M. K. Dhavalikar and Dr. P. P. Apte, both of the Deccan College, the former for suggesting to me the latest publication on archaeological discoveries in the Kuruksetra region the latter for giving me facilities to refer to the work done by the Dictionary Department. I must also place on record the willing and prompt help I received from the librarians of the Deccan College, the Fergusson College and the Bhandarkar Oriental Research Institute, of which the librarian Shri V. L. Manjul helped me by preparing the author index. Finally I thank Shri V.V. Jayade and Shri A. N. Gokhale of the Institute for verifying some arithmetical work and for correcting the proofs respectively and Shri D. L. Mhaske and Shri E. R. QWalekar of the Institute’s Press for doing neat printing job.

According to traditional belief, the epic Mahabharat has undergone a metamorphosis from 8,800 verses of the Jaya, through 24,000 verses of the Bharata, to the present text of one lakh verses over a long period of time. The internal evidence in the Mahabharata itself suggests that there were at least two redactions of it, Vaisamoayana’s Bharata consisting of 24,000 slokas and Mahabharata of about 82,00 slokas according to the count given in adhyaya 2 of the period of ten years, is sift, by the application of methods available in the theory of statistics, the original Bharata and the additions made from time to time by subsequent redactors.

Such a study has been made possible by the happy circumstance that the pic has been written for the most part in the Anustubh metre, which has a flexible pattern. This metre consists of 32 syllables, eight in each quarter (pada), and the only requirement it has to satisfy is that the fifth syllable of each quarter should be short, the sixth long and the seventh alternately long and short, with a few permissible variations. A syllable is long if it contains a long vowel or has a short vowel followed by a conjunct consonant. For the purpose of this study I have taken up the lines (slokardhas), which will then each have the fifth, the thirteenth and the fifteenth syllable short and the sixth, the seventh and the fourteenth long. The remaining ten syllables namely the first to the fourth, the ninth to the twenlfth and the eight and the sixteenth can be either short or long. The stylistic variations in the Anustubh slokas could arise from the natural propensity of the author to make use, unconsciously of course, of more or less long syllables, where they are free to do so.

For the purposes of this study, the Critical Edition of the Mbh. Brought out by the Bandarkar Oriental Research Institute, Pune, has been used. The Institute has also brought out four volumes of the Mbh. Text as constituted in the Critical Editions and given each parvan a table of contents (anukramanika) giving a number of adhyyas relating to definite topics. Then the statistical study proceeds in two stages: first we take a group of adhyayas (with a few changes where necessary) relating to a topic and count the number of lines as well as the number of long syllables in each of the ten positions where they can be either long or short. It is obvious that all the adhyayas which consist wholly or mainly of long metre stanzas or prose passages will have to be excluded. I have also excluded those adhyayas which contained very few i.e. less than ten slokas, but now find that their inclusion would have made no difference to the final result. The bouve exercise gives a two-way distribution of the count of long syllables, the rows giving the counts in the ten positions adhyaya-wise, and the columns the position-wise count for the different adhyayas. We then ask the question, does this group of adhyayas exhibit a single style or in the language of statistics does this distribution satisfy the test of homogeneity? For this we carry out what is known as the Analysis of variance and after making allowance for the different sizes of adhyayas, compare the variance between the adhyayas with the variation within the adhyayas with their appropriate degrees of freedom. If these two estimates of variances are found to be not significantly different by the application of statistical tests of significance, then the group of adhyayas is taken to be homogeneous in style.

In the Analysis of Variance, the total of squares is separated into parts corresponding to the sources of variation. Here the total sum of squares 90786.9 is first divided into parts, 78279.3 due to variation between the adhyayas and 12507.6 due to variation within the adhyayas. A large part of the variation between the adhyayas is, however, due to the differences in the sizes of the different adhyayas. If we work out the regression of the adhyayas means on the number of lines, the sum of squares due to this regression comes to 77930.6. The sum of squares due to variation between the adhyayas is, further divided into two parts, namely 77930.6 due to regression and 348.7 due to deviation from this regression. An estimate of variance is obtained by dividing the sum o squares by its degress of freedom. We have now two estimates of variance, one accounting for the variation between the adhyayas, namely 218.8 (d.f.16) after making allowance for their differing sizes and the other accounting for the variation within the adhyayas , namely 77.2 (d. f. 162). Now thequotient of two estimates of a variance for a normal population has a distribution discovered by Fisher, who tabulated it in the form z=log?vF. Snedecor named the variance ration F after Fisher and has given a Table for the distribution of F itself. The quotient F in this case is 28, which is less than its 5per cent tabulated value 2.05 for n1 = 162, n2 = 16. This shows that the variation between the adhyayas is not significant. All the eighteen adhyayas have, therefore, the same style, and so the statistical analysis does not support the theory of the multiple authorship of the Bhagavadgita.

It we now take up the group of the first 22 adhyayas of the Bhismaparvan, we find that the sum of squares due to variation within the adhyayas is 16060.8 with variance 81.1152 (d. f. 198). We have now to examine whether the adhyayas I-22 have the same style as the Bhagavadgita. For this we have to test the homogeneity of the two variances 81.1152 (d. f. 198) and 77.2074 (d. f. 162). We calculate F as the quotient of the larger variance by the smaller one and the double the tabular probability. The latter follows from the first that F so calculated is always greater than 1, so that only the upper part of the distribution id used (Snedecor, 10.13). In order to test the hypothesis of homogeneity or otherwise between the two variances, it would be necessary to utilise a Table giving 2.5 percentage points of F distribution. Such a Table id given by Bowkef and Lieberman in their book Engineering Statistics. They have also given a formula for calculating the values of F different degrees of freedom. In this instance, the value of F is the quotient 81.1152 / 77.2074=1.05 which is less than its formula value 1.35 for n1 = 198, n2 = 162. This shows that the adhyayas 1-22 of the Bhismaparvan have the same style as the Bhagavadgita.

If we go to the next group of adhyayas 41-60 (excluding adhyayas 56 consisting of long metre stanzas), we get the sum of squares due to variation within the adhyayas 45196.8 with variance 264.31 (d. f. 171). Applying the same method, we find 64 adhyayas 41-60, 71-94 and 95-117 (excluding adhyayas which contain wholly or mainly long metre stanzas) belong to the same style represented by the variance 258.78 with 576 degrees of freedom. We call this style A style. The remaining 50 adhyayas of the Bhisma,parvan, namely 1-40 and 61-70 designated as the B style, are represented by a variance and take the quotient F of the higher variance by the lower, then F> 1 = 3.12, which exceeds the 2.5 per cent (formula) value 1.19 for d. f. n1 = 576, n2 =450. These two styles are, therefore, significantly different.

By the successive e application of this method, we have discovered five styles, which are significantly different from one another. The whole of Karnaparvan, for instance, exhibits the A style and has the variance 278.56 with 522 degrees of freedom. The Sriparvan has a style named the alpha with a variance 55.72 with 234 degrees of freedom. As mentioned earlier, the 50 adhyayas 1-40 and 61-70 of the Bhismaparvan exibit the B style, represented by a variance of 82.82 with 450 degrees of freedom. Adhyayas 13-90 of the Aranyakaparvan (excluding adhyayas 24, 26 and 35 containing wholly or mainly long metre stanzas) disclose a variance 177.33 (d. f. 675). This style is designated as the C style. There is still a fifth style belonging to the six adhyayas 106-110 and 135 of the Anusasanaparvan. This style is named the Beta style and has a variance 738.81 with 54 degrees of freedom. It may be noted that these variances differ from one another not marginally but significantly. The Mbh, itself mentions that the Harivamsa forms a supplement (khila) to it (1.2.233). The Hrivamsa exhibits the C style except for twelve adhyayas including the Usasvapna which belong to the Beta style.

Of the parvans which exhibit a single a single style, Karnaparvan is written in the A style, while the Striparvan as also the Asramavasika, Mausala, Mahaprsthanika and the Svargarohana tken together exhibit the alpha style. The Sabha, virata, Bhisma, Drona, Salya and Sauptika parvans disclose two styles. Of these Sabha and Virata have alpha and B styles. The Bhisma and Drona A and B styles, Sulya A and C styles and Sauptika A and alpha styles. The Aranyaka and Anusasana exhibit three styles, the Aranyaka containing alpha, B and C style adhyayas and the Anusasana A, B, and Beta style adhyayas. The Adiparvan discloses four styles, the exception being the C style, while the Santiparvan has all the five styles.