Home › What the text actually says

The Mahābhārata critical edition — forty-seven years, and it printed what it left out

The popular claimNobody has ever seriously established what the Mahābhārata's text actually is.

The two pages before this one describe a problem: texts copied by hand acquire lines, and a reader wants to know which lines. This page is about what happens when somebody decides to actually find out, at scale, for the largest epic on earth.

It is a better story than almost anyone outside the field knows, and it is not a story about doubt. It is a story about an enormous, patient, public act of scholarship that the tradition carried out on itself.

The undertaking

The Bhandarkar Oriental Research Institute in Poona began the work in 1919. The first fascicule appeared in 1927. The constituted text of the eighteen parvans was completed in 1966, and the Harivaṁśa a few years after.

V. S. Sukthankar led it until his death in 1943; S. K. Belvalkar and P. L. Vaidya carried it to the end. Most of the people who started it did not see it finished.

Manuscripts were gathered from across the subcontinent and beyond — in Śāradā script from Kashmir, in Bengali, Devanāgarī, Telugu, Grantha, Malayāḷam, on birch bark and palm leaf. Over a thousand were assembled; a selected set, chosen to represent every independent line of descent, was collated word by word.

The method, in one paragraph

Manuscripts that share distinctive errors share an ancestor. Sort them into families on that basis, and the families fall into two great recensions, Northern and Southern, which diverged early. Where both agree, the reading is old. Where they differ, the editor weighs which reading better explains the existence of the other — a shorter, harder reading usually being older than a longer, smoother one, because scribes clarify and expand far more often than they compress.

The Kashmiri Śāradā manuscript of the Ādiparvan mattered disproportionately, because its line of descent had been separate for so long that its agreement with a Southern reading is powerful evidence of age.

The result is shorter

The version most widely printed and quoted — the vulgate, carrying Nīlakaṇṭha's seventeenth-century commentary — runs to roughly a hundred thousand verses. The critical text constitutes something closer to seventy-four thousand.

That gap is the point of the whole exercise, and it is worth being exact about what it means. It does not mean twenty-six thousand verses were forged. It means they entered the tradition at various points after the recoverable archetype, in various regions, and the edition can now say where and roughly when.

One finding deserves its own sentence, because people assume the opposite. The Bhagavadgītā came through essentially intact. Across recensions that disagree about a great deal, that section is remarkably stable — which is itself a result, arrived at by a method that had no interest in producing it.

And it printed everything it rejected

This is the part that ought to be famous and is not.

The edition does not present a clean text and quietly discard the rest. Every rejected line is published: the critical apparatus records the variants at the foot of each page, the star passages mark lines found in some manuscripts but not accepted, and the appendices carry longer excluded passages in full.

Nothing was destroyed and nothing was hidden. A reader who disagrees with an editorial decision can see exactly what was decided, on what evidence, and read the excluded line for themselves on the same page.

We will say plainly why that matters to us. It is the same discipline this whole site runs on — the calculators showing their arithmetic, the booklet naming the readings it withholds and counting them, this section printing its debated field even when the finding is strong. The Poona editors did it at the scale of a hundred thousand verses across forty-seven years, and they did it first.

What the editors said it was

Sukthankar was careful, and later readers have often been less careful than he was. The constituted text is presented as the earliest recoverable form of the written transmission — not as the poem Vyāsa composed, not as an original in any romantic sense.

Behind the earliest manuscripts lies an oral tradition that manuscripts cannot reach. Saying so is not a weakness of the edition; it is the edition being honest about the limit of its own instrument, which is the thing that makes the rest of it trustworthy.

What this page does not say

It does not say the vulgate is fake or that anyone reciting it is reciting a forgery. The vulgate is the text of the living tradition in most of India, it carries a great commentary, and the verses the critical edition sets aside have been loved and quoted for centuries. Later is not false.

It does not say the critical text settles every question. The debated note above is not a formality: whether a method built for copied manuscripts suits a work with a long oral life is a genuine argument, and individual readings are contested one at a time, which is how a scholarly edition is supposed to be used.

And it does not offer this as a defence of anything. The edition is not evidence that the tradition is good; it is evidence that the tradition contains, and paid for, forty-seven years of people checking.

Source — The Bhandarkar Oriental Research Institute critical edition of the Mahabharata (Poona, 1919-1966), general editors V. S. Sukthankar, S. K. Belvalkar and P. L. Vaidya; Sukthankar's Prolegomena to the Adiparvan; the edition's own critical apparatus, star passages and appendices

Every claim on this site names its source, and every calculator shows its arithmetic.

The free calculators →