The Search for the Original

A young German sat in Oxford with a Sanskrit hymn open before him and the money of a trading empire behind him. He was twenty-six. He would never set foot in India.

The hymn was old beyond confident dating, carried for centuries by voice before it was carried by hand. Friedrich Max Müller meant to fix it on the page.

The canon goes to press

The recovery of the Sanskrit canon in print was an act of nineteenth-century European scholarship, and it was funded by empire. Both facts are plain in the record. Müller’s edition of the Rigveda with Sayana’s medieval commentary appeared in six volumes between 1849 and 1874. Its printing cost was borne by the East India Company, whose Board of Directors agreed to bear it and commissioned the work; the title pages declared it published under the patronage of the East India Company. The scholar Arie L. Molendijk has documented the arrangement.

Müller did not stop at the Rigveda. From 1879 he edited the Sacred Books of the East for Oxford University Press, a series of fifty volumes completed in 1910, English translations of the scriptures of many traditions, produced by an international team. It was the largest single attempt yet made to put the sacred literatures of Asia between covers.

The series gave Europe its first broad library of translated scripture, the Vedas and Upanishads beside the texts of Buddhism, Jainism, Zoroastrianism, Confucianism, and Islam, fifty volumes rendered into English by an international team over thirty-one years. For most Western readers it was the only door into these literatures that existed. The choices made inside it, which texts to include, which to leave out, how to render a verse, shaped how a generation outside India understood Indian religion.

Note the funding fact and leave the motives alone. The work was real. It was also entangled in the apparatus of the Company that paid for it. A man who never crossed the ocean became, for a generation of European readers, the gatekeeper of a literature that had crossed no ocean and lived in no single book.

Printing solved one problem and exposed another. Once you set a text in type you must choose which words to print. And the manuscripts did not agree.

What a critical edition is

To choose, scholars borrowed a method built in Europe for the classics, often called Lachmann’s method after the philologist associated with it. The principle is simple to state and brutal to execute.

You gather every surviving manuscript of a work. Each one is a witness. You read them against one another, line by line, a process called collation. Scribes copying by hand make mistakes, and a copy inherits the mistakes of its parent. So manuscripts that share the same errors are likely descended from a common source. By grouping the witnesses according to their shared errors you can build a family tree, the stemma codicum, the lineage of the copies, descending from a lost common ancestor.

That lost ancestor is the archetype: the earliest reconstructable source from which all surviving copies descend. The work of building the tree is called recension (recensio); the work of correcting obvious errors in the reconstructed text is emendation (emendatio). What you cannot resolve, you do not hide.

Here is the discipline of the thing. The editor prints the readings judged oldest in the body of the page. Every rejected variant, every disagreement among the witnesses, is recorded below in the critical apparatus, usually a dense band of notes at the foot of each page. A reader can see not only what the editor chose but everything the editor refused. The apparatus is a confession built into the book.

This is the instrument. The question was whether it would work on the Mahabharata.

Pune, 1919

The longest poem in the world had never been so edited. There was a printed vulgate, the commonly received text in wide circulation, but no edition that had weighed the manuscripts against one another. In 1919 the Bhandarkar Oriental Research Institute in Pune resolved to make one.

The scale was without precedent in Indian textual scholarship. The whole-epic project would eventually draw on about 1,259 manuscripts, gathered from across the subcontinent in many scripts. In 1925 V. S. Sukthankar was appointed general editor, the post he held until his death in 1943. S. K. Belvalkar followed him, then P. L. Vaidya, with R. N. Dandekar among them. Completion was announced on 22 September 1966, forty-seven years after the work began.

What they produced is staggering as an object. The constituted text of all eighteen parvans, the books of the epic, runs to roughly 89,000 verses across some nineteen volumes and more than 13,000 pages. And under nearly every verse sits the apparatus, the recorded dissent of the manuscript tradition.

Read the dates carefully. The oldest dated manuscript Sukthankar could place in his apparatus for the first book is a Nepali copy of 1511 CE. The poem is reckoned, by tradition and by scholarship alike, to be far older than that. The witnesses were young. The text was old. Between them lay a silence of more than a thousand years that no manuscript could cross.

Sukthankar knew exactly what he was holding, and he wrote it down.

Sukthankar’s confession

In 1933 he published his Prolegomena, the long introduction to the first volume, the Adiparvan. It is one of the most honest documents in the history of textual scholarship, and because it is old scholarship it can be quoted plainly.

He saw the difficulty at once. “Where, one may well ask,” he wrote, “is the original of a whole literature?” The Mahabharata was not a book. It was a tradition wearing the shape of a book.

He had a figure for how it had grown. The text spread “not only upwards and downwards, but also laterally, like the Nyagrodha tree, growing on all sides,” the banyan that drops new trunks from its branches until you cannot find the first stem. And he named the cause. The variation made sense, he wrote, “as soon as we assume that the epic was handed down from bard to bard originally by word of mouth.”

That assumption broke his borrowed instrument. The European method had been built for manuscripts copied from manuscripts, a single line running back to one archetype. An oral, many-branched tradition does not behave that way. Sukthankar said so without flinching: “the genetic method, operating with an archetype and a stemma codicum, cannot strictly be applied to fluid texts and conflated manuscripts.” The tool fit the task only with great limitation.

So he refused the grand claim. He did not pretend to recover the shatasahasri samhita, the traditional hundred-thousand-verse original ascribed to Vyasa. “A wholly certain and satisfactory restoration of the text to its pristine form,” he wrote, “may be a task now beyond the powers of criticism.” He chose a humbler and more exact word for what he had made. He called it the constituted text: the oldest form the surviving evidence allows us to recover, and no more than that.

And he located the real worth of the edition below its body, not in it. The work, he said, was “a veritable thesaurus of the Mahabharata tradition,” because in its apparatus it preserved the variants, the regional voices, the whole living spread of the banyan. He had not found Vyasa. He had built the most complete map ever made of how the poem actually existed.

This is worth holding still over. Sukthankar had the prestige of European philology behind him and the patience of decades in front of him, and he used both to argue against the very promise that gave the method its glamour. The promise was the lost original, the single fountainhead. He set it down and walked away from it. What he kept was the discipline of recording everything, the apparatus as an honest ledger of doubt.

Baroda, and the same problem

The method crossed to the other epic. At the Oriental Institute in Baroda, collation of the Ramayana of Valmiki began in 1951. Seven volumes followed, one for each kanda or book, published between 1960 and 1975, under the first general editor G. H. Bhatt and later U. P. Shah.

Tradition calls the Ramayana the adikavya, the first poem, composed whole by a single seer. The manuscripts told the now-familiar story instead. Many recensions, regional families, a text that had moved and grown across centuries. The same instrument, the same care, the same result: a constituted text resting on an apparatus of difference.

Two of the foundational poems of a civilization had been edited by the most rigorous method available. And the method, applied honestly, kept returning the same verdict. There was a recoverable text. Whether there was an original was another question entirely.

The question that will not close

Here the scholars divide, and an honest account names the positions without ruling between them.

Sukthankar holds the cautious ground. The constituted text is the oldest recoverable form, reached by comparison of the witnesses, and he said with care that it was not Vyasa’s lost composition. The stemmatic method, he warned, applies only with severe limitation to a fluid and, in his word, polygenous text, one born from many sources rather than one. He recovered what could be recovered and drew the line there.

The French scholar Madeleine Biardeau pressed harder. In a 1968 essay in the journal Purana, her position later relayed and discussed by Alf Hiltebeitel, she argued that the critical edition had reconstructed a text that “never existed.” A fundamentally oral epic, multiple by region from the start, has no single archetype waiting to be found. The constituted text, on her reading, carried little heuristic value as a thing that ever stood whole in the world. The edition’s true value, she agreed, lay in the apparatus, which preserved the versions that had actually lived.

A third position stands between them. Andreas Bigger and James Fitzgerald have argued that the edition recovers neither Vyasa’s original nor a phantom, but a later normative redaction: a real historical moment when the long-fluid text was edited and fixed into an authoritative shape. By this reading the edition recovers a genuine text-form. Just not the first one.

Set the verdict down plainly. Sukthankar reconstructed the oldest recoverable text and said exactly that. Biardeau asked whether “oldest recoverable” and “original” are even the same question for a poem like this. Tradition holds that one author, Vyasa, dictated a hundred thousand verses to Ganesha, and that Valmiki sang the first poem entire. The record shows two epics that grew over centuries in many hands, whose oldest surviving manuscripts are only about five centuries old, and whose single authorial original is, on the evidence, unrecoverable and may never have existed as one object at all.

You can read the cross-currents elsewhere on this site, in what the Mahabharata is and in what the Ramayana is, and you can trace how these poems reached the press at all in the print chapter of this volume.

The scholars built a magnificent book to find the original. What they found, recorded line by line in the apparatus beneath the text, was that the poem had always been a library.

The next chapter follows that library off the page entirely, into the era of the screen, where the text multiplies again. See the library without walls.

The original was never there to be found.

It had become, instead, everything the apparatus could hold.