The making of the text

Where the text comes from

The Mahabharata tells the story of its own making, and scholars tell a different one. Both are here, kept apart. The first part is what the Critical Edition says, chapter by chapter. The second is what the manuscripts, the inscriptions and the editors say, and it is the one place on this site with dates. The third is how tradition says it was written down, Ganesha and all.

What the text says about itself

The epic does not begin with Shantanu. It begins with a bard, a sacrifice and a question, and it says who composed it, how long it took, what it is called and how big it is. Every line below is in the Critical Edition, cited as parva.chapter or parva.chapter.verse.

A bard at a sacrifice

The text opens in the Naimisha forest, where the bard Ugrashravas, son of Lomaharshana, arrives at the twelve-year sacrifice of the sage Shaunaka and is asked what he has heard. He has come from the snake sacrifice of King Janamejaya, where he heard the Bharata recited by Vaishampayana, and he tells it again. The whole epic is a retelling, at one remove, of a recital at a ritual. CE 1.1

Composed by a man inside it

The composer is Krishna Dvaipayana, called Vyasa, the arranger, because he divided the one Veda into four. He is Satyavati's son by the sage Parashara, born on an island in the Yamuna, and so the grandfather of both the Pandavas and the Kauravas, and a character in the story at every turn. The text says he rose early every day and worked at the poem for three years. CE 1.54.5; 1.56.32; 1.57

Taught to a pupil, recited to a king

Vyasa teaches the poem to his pupil Vaishampayana. When Janamejaya, Parikshit's son and Arjuna's great-grandson, holds a sacrifice to destroy the snakes that killed his father, Vyasa comes to it, and the king asks to hear how his ancestors quarrelled. Vyasa tells Vaishampayana, seated beside him, to recite it. The recital that follows is the epic. CE 1.54–1.56

Its own name and size

The text calls itself Jaya, victory, a history that one who wants to conquer should hear. It says Vyasa made a Bharata collection of twenty-four thousand verses without the side-stories, which the learned call the Bharata proper, and that the whole runs to a hundred thousand verses. Its second chapter is a table of contents that counts the chapters and verses of every parva; it gives the Drona Parva, for instance, one hundred and seventy chapters and 8,909 verses. CE 1.1.61; 1.2.167; 1.56.13; 1.56.19

Three beginnings

The bard says that some brahmanas learn the Bharata starting from Manu, others from the story of Astika, others from King Uparichara. Even inside the text there is more than one place to begin. CE 1.1.50

What is here is elsewhere

The line the epic is best known by is its own claim about itself: on dharma, on profit, on pleasure and on release, what is here may be found elsewhere, but what is not here is nowhere. CE 1.56.33

The fifth Veda

Vyasa, the text says, taught the four Vedas with the Mahabharata as the fifth. That is the claim the tradition accepted: not a poem but a scripture that tells its truths as a story. CE 1.57.74

How it was written down, as tradition tells it

TraditionalThe vulgate, Adi Parva chapter 1 (Ganguli, Book 1, section 1); the caves at Mana are pilgrim tradition; the tusk is later still

When Vyasa had finished composing the poem in his mind, he sat wondering how to teach it to his pupils, and Brahma, the creator, came to his hermitage. Vyasa bowed and said that he had made a poem that held everything, the Vedas and their meaning, the histories, the duties of the four orders and the four goals of life, and that there was no one on earth who could write it down. Brahma told him to think of Ganesha.

Vyasa thought of Ganesha, and Ganesha came at once. He agreed to be the scribe on one condition: his pen must never stop, not for a moment. Vyasa agreed, and set a condition of his own: write nothing you have not first understood. Then he began to dictate, and whenever he needed time to think ahead he tied a knot in the verse, a line so tight that even Ganesha, who knows everything, had to pause over it. The bard says there are eight thousand eight hundred such verses, and that no one to this day has untied them all.

Where it happened, the poem does not say, not even in the vulgate. Tradition gives a place: the village of Mana above Badrinath, at the head of the Alakananda valley, where pilgrims are shown a Vyasa cave, whose layered roof is said to be the pages of the poem, and a Ganesha cave a little below it. The atlas marks Badari as the mountain retreat the Pandavas reach on their pilgrimage and pass on the great departure; the caves are on the same road. Badari on the map →

The story that Ganesha's pen broke and he snapped off his own tusk to keep writing, so that his promise would hold, is not in the vulgate passage either. It is a later tradition, told to explain why Ganesha is shown with a single tusk, and it belongs with the story rather than with the text.

The Pune editors left the whole episode out of the Critical Edition, because it is missing from the Southern manuscripts and from the oldest Northern ones, and recorded it in the notes as a later insertion. It is told here in full because the atlas would be incomplete without it: it is how most people picture the making of the poem, and Vyasa's condition, write nothing you have not understood, is still the best one-line description of what the Mahabharata asks of a reader.

What the manuscripts and the scholars say

The story’s own timeline has no dates, and the atlas gives it none. The history of the book is another matter. This is the outline most scholars would accept, with the disagreements marked and the works to read named; none of it is drawn from the text itself.

  1. 4th century BCE

    Panini's grammar knows the word Mahabharata and forms names for devotees of Vasudeva and Arjuna; a household ritual text of about the same age lists 'the Bharata and the Mahabharata' among the teachers to be honoured with water. Some form of the poem, and perhaps two sizes of it, existed by then.

    Panini, Ashtadhyayi 4.3.98 and 6.2.38; Ashvalayana Grihyasutra 3.4.4
  2. c. 400 BCE to 400 CE

    The span most scholars give for the poem's growth from a bardic tale of the Kuru war to the eighteen-book work we have, with the teaching books and the Gita gathered in along the way. John Brockington's survey sets out that case. Alf Hiltebeitel argues instead for a deliberate composition by a small group within a much shorter span, around the first century BCE. James Fitzgerald reads the Shanti Parva as a response to Mauryan and post-Mauryan kingship. The atlas takes no side, and its timeline carries no dates.

    Brockington, The Sanskrit Epics (1998); Hiltebeitel, Rethinking the Mahabharata (2001); Fitzgerald, The Mahabharata, vol. 7 (2004)
  3. 2nd to 3rd century CE

    The oldest physical trace. A fragmentary manuscript from a Buddhist site near Kucha on the Silk Road lists the parvas of a Mahabharata, but not the Virata Parva or the Anushasana Parva, so the list of eighteen was still settling.

    The Spitzer manuscript; Schlingloff, 'The oldest extant parvan-list of the Mahabharata', JAOS 89 (1969)
  4. 6th century CE

    A copper-plate land grant quotes the epic as scripture and calls it the collection of a hundred thousand verses, the size the text claims for itself.

    Khoh copper plates of Sarvanatha, 533 CE
  5. 1st millennium onward

    Two manuscript families grow apart: a Northern recension in Sharada, Nepali, Devanagari, Maithili and Bengali scripts, and a Southern recension in Grantha, Telugu and Malayalam, longer and more ornate. Translations and retellings begin: the Old Javanese parwas around the year 1000, the Persian Razmnama made for Akbar in the 1580s, and the Mahabharatas in Telugu, Kannada, Tamil, Bengali, Odia and other languages, each with additions of its own.

    Sukthankar, Prolegomena to the Adi Parva (1933)
  6. 17th century

    Nilakantha Chaturdhara of Maharashtra writes his commentary on a full text he assembled from many manuscripts. Printed at Bombay in 1863, it becomes the vulgate, the version most readers used for the next century and the one Ganguli translated into English between 1883 and 1896. Most of the stories on the atlas's Not in the text page live here.

    Nilakantha, Bharatabhavadipa; the Bombay edition (1863); Ganguli's translation (1883–1896)
  7. 1919 to 1966

    The Bhandarkar Oriental Research Institute at Pune sets out to recover the oldest text that the manuscripts allow, collating copies from every region and keeping only the readings the families agree on. V. S. Sukthankar leads the work from 1925 and publishes the Adi Parva in 1933 with a Prolegomena that explains the method; the last of the nineteen volumes appears in 1966. This is the Critical Edition the atlas follows.

    Sukthankar, Belvalkar, Vaidya et al., The Mahabharata, critically edited (1933–1966)
  8. 1973 onward

    J. A. B. van Buitenen begins the first English translation of the Critical Edition at Chicago and dies in 1979 with the first five books done; James Fitzgerald and others carry it on. Bibek Debroy completes a full translation in ten volumes between 2010 and 2014. Peter Brook stages the epic in 1985 and B. R. Chopra's television serial runs from 1988 to 1990; on this site both count as Adaptation.

    van Buitenen (1973–1978); Fitzgerald (2004); Debroy (2010–2014)

The Critical Edition by the numbers

18parvas
1,995chapters
73,816verses

Counted from the Critical Edition e-text, each numbered verse once. The tradition's hundred thousand includes the Harivamsha and a great deal that the editors set aside.

What tradition adds

These are the best-known stories about the making of the poem. They are real as tradition, and they are not in the Critical Edition. They are also listed on Not in the text.

  1. The story that Vyasa dictated the poem to Ganesha, who agreed to write it down only if Vyasa never paused, and that Vyasa slowed him with eight thousand eight hundred knotted verses, stands at the very start of the vulgate's first chapter and in almost every retelling. The Critical Edition has no Ganesha and no scribe: Vyasa composes, teaches, and the poem is recited aloud. The passage is missing from the Southern manuscripts and from the oldest Northern ones, so the editors set it aside as a later insertion.

    Traditional

  2. The neat story that the poem grew in three steps, Jaya of 8,800 verses, then Bharata of 24,000, then Mahabharata of 100,000, is a modern reading of the text's own numbers. The Critical Edition gives the name Jaya (1.56.19), the 24,000 (1.1.61) and the 100,000 (1.56.13); the 8,800 comes only from the Ganesha passage, which it does not have.

    Traditional

How the atlas uses this

Everything on the map and in the event panels follows the Critical Edition; where the vulgate or a retelling differs, the difference is labelled and collected. The bard’s frame is why the story rail begins with Shantanu rather than with Manu, and it is the atlas’s closing event, How the story was told, where the Ganesha story also appears as a labelled note. The translations and references behind each page are on the Sources page.