Scholars in Pune Consulted 1,259 Manuscripts Over 47 Years to Produce the Mahabharata Critical Edition, 19 Volumes With More Than 89,000 Verses

On 22 September 1966, at the Bhandarkar Oriental Research Institute (BORI) in Pune, Sarvapalli Radhakrishnan, then President of India, released the finished critical edition of the Mahabharata. By the institute’s count, scholars had spent about five decades on it, consulting 1,259 manuscripts. The result is 19 volumes of more than 15,000 pages, and the constituted text, the single reading the editors settled on at each point, runs to over 89,000 verses across the epic’s 18 parvans, or books.

Counted from the formal start on 1 April 1919, the work took a little over 47 years. Variant readings from the manuscripts are printed at the foot of the page, so a reader can see what the editors chose and what they set aside. Simon Brodbeck, reviewing a 2018 study of the edition in the International Journal of Hindu Studies, writes that it “now forms the basis of all scholarship on the text.” The Mahabharata is itihasa, the tradition’s name for a telling of what happened, and this edition set out the wording that scholars now quote, with the alternatives beside it.

A call from Paris in 1897, and a lakh of rupees from Aundh

The idea came from Moriz Winternitz, who told the International Congress of Orientalists in Paris in 1897 that a critical edition was wanted as the only sound basis for studying the epic. V. S. Sukthankar, in his Prolegomena of 1933, records Winternitz’s own memory of the reception: most scholars thought a restored text was impossible. The International Association of Academies took the project up in 1904. Heinrich Lüders printed a specimen in 1908, 18 pages covering the first 67 stanzas of the first book with 29 manuscripts from European libraries. Winternitz reported in 1928 that nothing else had been printed, and Sukthankar adds that the First World War ended the plan.

BORI, founded on 6 July 1917 according to its own timeline, took it on soon after. At a meeting of its General Body on 6 July 1918, Balasaheb Pant Pratinidhi, the Chief of Aundh, urged a critical and illustrated edition and offered a lakh of rupees (100,000) in annual grants. On 1 April 1919 Sir Ramkrishna Gopal Bhandarkar, the doyen of Western India’s Sanskritists, inaugurated the work by beginning the collation of the invocation that opens some manuscripts. Collation means comparing each manuscript against a base text, word by word, and recording every difference.

The institute then appealed to governments, princes and donors. The title page of Volume 1 lists its patrons: the Ruler of Aundh, the Government of India, the provincial governments of Bombay, Madras and Burma, the states of Hyderabad, Baroda and Mysore, and the University of Bombay.

Ten scripts, a ruled sheet and a second pair of eyes

Sukthankar’s Prolegomena describes the working method in detail, for the first book, the Adiparvan. By his count about 235 manuscripts of it were known from library catalogues and from owners who sent them in. They were written in ten scripts: Devanagari (107), Bengali (32), Grantha (31), Telugu (28), Malayalam (26), Nepali (5), Sharada (3), and one each in Maithili, Kannada and Nandinagari. About 70 were collated and about 60 were used to build the text.

Collation ran at three centres. BORI could collate every script the work needed. Rabindranath Tagore’s Visvabharati in Bengal, under Pandit Vidhushekhara Bhattacharya, and the Saraswathi Mahal Library at Tanjore, under T. Sambamurthi Rao, were set up at first chiefly for the Bengali and the Telugu and Grantha manuscripts. A permanent staff of trained shastris (traditional scholars) and university graduates wrote each stanza of the Bombay printed edition on the top line of a ruled foolscap sheet. Variants were entered beneath it, one line per manuscript, syllable by syllable, with a four-inch column on the right for remarks and extra stanzas. A second batch of collators checked every sheet before it reached the editor.

The numbers per parvan were set by what survived. In 1933 Sukthankar reported that the whole epic stood collated from at least ten manuscripts, many parvans from twenty, some from thirty and a few from forty, while the first two chapters of the Adiparvan, which bear on the text of the entire epic, were collated from no fewer than sixty. BORI’s figure of 1,259 is the total for the finished edition.

Selection followed stated rules. The editors wanted as many scripts, and so as many provinces, as possible. They preferred old, fragmentary manuscripts to neat modern ones. Of the Nilakantha version, the most numerous, they took only three, because it is among the latest and had been printed several times. The oldest dated manuscript in the Adiparvan apparatus is Nepali, copied in 1511.

Two recensions, sorted by script

The manuscripts fall into a Northern and a Southern recension, a recension being a distinct line of the text, and each divides into regional versions. Sukthankar found that script tracks version closely, since scribes mostly copied manuscripts in their own script, though Devanagari copies mix versions freely. The two recensions differ in readings, in passages present or absent, and in the order of passages. For the Adiparvan he gave the Bombay printed text as about 8,460 stanzas and the Southern text edited by P. P. S. Sastri as 9,984, a gap of about 1,524. The epic’s own table of contents, in the constituted text, lists 7,984.

His working rule was to prefer a reading on which independent versions agree, while allowing that mixing between versions made certainty impossible. He also used outside witnesses: the commentaries of Devabodha, Arjunamishra and Nilakantha, and three epitomes of the eleventh century, an Old Javanese adaptation of about 1000 CE, Nannaya’s Telugu version and Kshemendra’s Bharatamanjari. Nannaya’s own count for the first book is 9,984 stanzas, which Sukthankar read as showing that the text Nannaya used was about the size of the surviving Southern manuscripts.

What the editors said the text is

Sukthankar was plain about the limits. The constituted text, he wrote, is not a reconstruction of the Ur-Mahabharata, the “ideal but impossible desideratum” of an original form, and not a copy of what Vaishampayana recited before Janamejaya. It claims to be a version “as old as the extant manuscript material will permit us to reach.” He added that an average chapter may hold a stanza of the second century BCE beside one of the second century CE, in his estimate.

He grounded this in the epic’s own account. In the Adiparvan, Vaishampayana, a pupil of Vyasa, recites the work at Janamejaya’s sacrifice, and Sauti later recites it again from memory. Vyasa is also said to have taught five pupils, each of whom is reported to have published a separate version. Sukthankar read this as the tradition’s own record of a text that was recited freely before it was written down, and he printed a wavy line under the important words and phrases that were less than certain, “obviating all false sense of security.”

Brodbeck’s review describes the edition’s text as hypothesised to be the archetype, the common ancestor of the surviving manuscripts. He reports that Adluri and Bagchee call attempts to speak of versions earlier than that archetype misguided, and he adds that most earlier speculation about the epic’s text history gained no support from the edition. He also notes that the book’s introduction says the archetype depends on which manuscripts happened to survive and may have been one of several versions in circulation.

The editors who carried it from 1925 to 1966

A tentative edition of the Virata Parvan by N. B. Utgikar appeared in 1923 and was sent free to leading Sanskritists in India, Europe and America for comment. Sukthankar writes that the opinions were favourable and many suggestions were adopted. BORI’s timeline records his appointment as general editor on 1 August 1925, and the first fascicule of the edition on 6 July 1927. The Adiparvan was finished in 1933 in seven fascicules, with a 110-page Prolegomena and 881 pages of text and notes. Reviewers abroad noticed it as it came: L. D. Barnett in the Bulletin of the School of Oriental Studies in February 1930, and R. P. Dewhurst in the Journal of the Royal Asiatic Society in July 1932.

The title page of Volume 1 names a dozen scholars as cooperating, Indian and European. Winternitz is among them, as are P. V. Kane and the Jesuit R. Zimmermann. Sukthankar led until his death on 21 January 1943, according to News Bharati. S. K. Belvalkar became general editor on 1 April 1943, and P. L. Vaidya on 1 April 1961 when Belvalkar retired. BORI’s page also names S. K. De and R. N. Dandekar among the editors.

After 1966

BORI formally began a critical edition of the related Harivamsha on 19 November 1954, when President Rajendra Prasad inaugurated it. It was completed in November 1971. The constituted text of the Mahabharata and the Harivamsha together appeared in September 1976 and BORI lists it as five volumes. A Cultural Index of the epic, edited by M. A. Mehendale, has two volumes published so far, built on a collection of more than 1,50,000 (150,000) cards.

A ThePrint column names the Pune text as the base of Bibek Debroy’s English translation, which it says was completed in 2014. For the epic itself and its story, see what the Mahabharata is, and for the best-known section inside it, the Bhagavad Gita, which sits in the Bhishma Parva.

On BORI’s page, checked on 29 October 2026, the constituted text is available on CD-ROM with search functions. Digitising the critical apparatus and the Cultural Index is listed there as a future task.