The Mahabharata -- The Text Behind the World's Longest Epic

Most people who say they have read the Mahabharata mean they have read a retelling: the Pandavas and Kauravas, the dice game, the war at Kurukshetra, the Bhagavad Gita. Very few have opened the text as a text and asked the more basic questions a book invites. How is this book actually organised? Who is supposed to have composed it? And through how many mouths does its story pass before it reaches a reader at all? This article stays with those three questions. It does not retell the dice game, which Eternal Raga covers scene by scene elsewhere on this site, and it does not reopen the historicity debate, which has its own dedicated article here. What follows instead is the architecture: the eighteen parvas that divide the text, the scale that makes it the longest epic poem to survive from the ancient world, the authorship tradition set against what textual scholarship actually finds, and the layered frame narrative through which the story reaches every reader, then and now.

The tradition has a name for the epic's size: Shatasahasri Samhita, the hundred-thousand-verse collection, built from shlokas, the standard couplet of classical Sanskrit narrative verse. A hundred thousand shlokas is the figure the text itself claims and the figure repeated across the Puranas that describe it. Rendered in a unit modern readers can picture, that comes to close to 1.8 million words, a length commonly compared to roughly ten times the combined length of Homer's Iliad and Odyssey, though the precise multiple shifts depending on which edition of the Greek epics is doing the counting. However the comparison is run, the result does not change: the Mahabharata is the longest single epic poem to survive from the ancient world, by a margin wide enough that the claim rarely needs softening.

What does need softening is the exact verse count in what can actually be checked against surviving manuscripts. The Pune Critical Edition, compiled at the Bhandarkar Oriental Research Institute (BORI) between 1919 and 1966 by comparing more than 1,250 manuscripts gathered from across the subcontinent, arrived at a constituted text of a little under 75,000 verses, commonly cited as 73,784; adding the appended Harivamsha brings the total closer to 80,000. The gap between that figure and the traditional 100,000 is not a scandal, and it is not evidence that the tradition exaggerated. It is what a text copied by hand across two millennia and dozens of regional lineages should look like: a stable narrative core, surrounded by a wide penumbra of passages that some manuscript families kept and others dropped, which the Critical Edition's editors recorded in a textual apparatus rather than silently including or silently erasing.

Hindu tradition names a single author: Krishna Dvaipayana Vyasa, also called Veda Vyasa, born to the sage Parashara and the fisherwoman Satyavati on a river island, dvaipa, which gives him his name. Tradition holds that Vyasa composed the epic in its entirety and then needed a scribe fast enough to keep pace with his recitation. Ganesha volunteered, on one condition: Vyasa could not pause. Vyasa agreed, on a counter-condition of his own: Ganesha had to understand every verse before writing it down. Whenever Vyasa needed time to compose the next difficult passage, he inserted a verse dense enough to make Ganesha stop and think, buying himself room to work. In the most famous embellishment of the story, Ganesha's pen breaks partway through, and he snaps off his own tusk to keep writing rather than interrupt the recitation.

That specific embellishment is later than most readers assume. The Ganesha-as-scribe episode does not appear in the oldest manuscript layers behind the Mahabharata; it entered the tradition around the 9th or 10th century CE, present in only some manuscript families (37 of the 59 consulted for the Pune Critical Edition carry it at all), and the Critical Edition's editors placed it in a footnote within an appendix rather than in the constituted text. Even that added episode, in its earliest form, says nothing about a broken tusk; the tusk detail is a further layer added on top of an already late addition. None of this makes the story less loved. It makes it a useful small case study in how a beloved detail of who wrote the Mahabharata can itself be a later chapter in the text's own transmission history.

Set against the devotional tradition, textual scholarship treats single authorship, by Vyasa or anyone else, as the wrong kind of question to ask of a text like this. John Brockington's influential account of the Sanskrit epics argues for gradual accretion: an oral epic core, focused on conflict between the Kuru and Panchala dynasties, that grew through recognisable stages, absorbing genealogical, didactic, and devotional material as it moved from oral performance toward a fixed written form. Alf Hiltebeitel takes a more architected view, arguing that a formative version of the epic came together in a comparatively compressed window, roughly the mid-2nd century BCE to the turn of the era, shaped with enough deliberate design that he resists calling it pure accretion, even while accepting that further material was layered on afterward. The two scholars disagree sharply on how planned the process was; neither of them argues for a single sitting or a single hand across the whole text. The working range most often cited across the field for the epic reaching something close to its current form is roughly the 4th century BCE to the 4th century CE, a span of some eight hundred years, built on oral material about a Kuru-Panchala conflict that may itself reach back further still.

The Pune Critical Edition project did not attempt to settle a date of composition at all. Its editors, working under general editor V.S. Sukthankar and his successors across nearly five decades, set out to reconstruct the earliest recoverable common ancestor of the many surviving manuscript recensions, not to identify an original text by an original author. That is a meaningfully different task, and the distinction matters. Asking who wrote the Mahabharata and when turns out to be the wrong shape of question for a text transmitted this way, in the same honest sense this site uses for the Puranas elsewhere: a scripture built by many hands across many centuries is not diminished by that fact. It is simply describing what a living, orally rooted, continuously copied tradition actually looks like once scholars go looking for its edges.

The story reaches every reader through a frame nested at least three levels deep, and the Mahabharata is unusually deliberate about making that frame visible rather than hiding it. At the outermost level, a bard named Ugrashrava Sauti, son of Lomaharshana, arrives at Naimisharanya, a forest hermitage, during a pause in a twelve-year sacrifice being conducted by the sage Shaunaka. The assembled sages ask Sauti for a story, and Sauti obliges by narrating the Mahabharata, in full, exactly as he says he once heard it recited. This outer conversation, Sauti speaking to Shaunaka's sages, is the frame a reader actually enters when opening the text: everything that follows is presented as Sauti's retelling, not as an unmediated narration.

What Sauti heard, in turn, was Vaishampayana, a direct disciple of Vyasa, narrating the epic to King Janamejaya at Janamejaya's own royal sacrifice. That gives the text a second interior frame, one level further in: Vaishampayana speaking to Janamejaya, which Sauti is now repeating to Shaunaka's sages, who are in turn the reason the text exists on the page at all for anyone reading it later. Vyasa himself appears inside this structure too, not only as the epic's traditional author but as a character present at Janamejaya's sacrifice, who is said to have directed Vaishampayana to narrate the story his disciple had learned from him. The author, in other words, steps into his own frame narrative as a witness to its telling.

Janamejaya's sacrifice was not a routine ritual occasion. It was a sarpa satra, a snake sacrifice, and Janamejaya convened it out of grief and rage. His father, Parikshit, grandson of Arjuna and the king who inherited the Kuru throne after the Kurukshetra war, died of snakebite. The immediate cause, as the tradition tells it, was an insult: Parikshit, out hunting and unable to get an answer from a sage named Shamika who was absorbed in silent meditation, draped a dead snake around the sage's neck in irritation and walked away. Shamika's son, the young ascetic Shringin, learned of the insult and, in fury, cursed Parikshit to die of snakebite within seven days. The serpent king Takshaka carried out the curse.

Janamejaya, on succeeding his father, resolved to destroy the entire race of serpents in retaliation, and the sarpa satra was the mechanism: a Vedic fire ritual designed to draw every snake in existence into the sacrificial flames by the power of mantra. Thousands of snakes were destroyed before a young sage named Astika, born of a Naga mother and a Brahmin father and therefore standing between both sides of the conflict, intervened and persuaded Janamejaya to halt the sacrifice before Takshaka himself, the one who had actually killed Parikshit, could be consumed. It is in the intervals of this same sacrifice, called for revenge and interrupted by mercy, that Vaishampayana narrates the Mahabharata to Janamejaya.

The connection between the frame and the framed story is closer than it first appears, because the Adi Parva, the very first book of the Mahabharata, opens by narrating the sarpa satra and the Astika episode as part of its own content, before moving on to Vyasa's birth and the genealogy of the Kuru line. The occasion of the telling and the beginning of what is told are not two separate things happening at a tidy distance from each other. They overlap: the first book a reader encounters is partly an account of the very sacrifice at which the whole epic was first recited aloud to Janamejaya. A story about a king's grief and a race nearly exterminated becomes, within a few chapters, the doorway into an eighteen-book epic about a different war and a different set of grieving survivors.

This nested structure, story inside story inside story, each with its own narrator and its own occasion, is one of the Mahabharata's most distinctive formal features, and it is easy to miss entirely when the text is encountered only through retellings that begin directly with Bhishma or Shantanu. Read as it presents itself, the epic never lets a reader forget that a story is always being told by someone, to someone, for some reason, and that the reason is itself worth listening to.

None of this structural picture explains why the epic is still read, argued about, and adapted three thousand years after its oldest layers were first composed. Much of that answer sits inside episodes this article has deliberately left alone. The dice game in the Sabha Parva, where Yudhishthira stakes and loses his brothers, himself, and finally Draupadi while the assembled elders of Hastinapura watch and say nothing, is arguably the single hinge on which the entire plot turns. Eternal Raga covers that scene in full, verse by verse, elsewhere on this site, and there is no reason to retell it here. The same holds for the historicity question, whether the war, the dynasties, and the geography the epic names correspond to anything archaeology, astronomy, or geology can actually confirm. That question also has its own dedicated article on this site, weighing the evidence honestly in both directions rather than settling it either way.

What belongs in this article instead is the structural fact that makes the Mahabharata unlike almost anything else surviving from the ancient world: a text that narrates its own transmission as part of its own content. Every reader who opens the epic is, formally speaking, sitting where the sages at Shaunaka's sacrifice sat, listening to Sauti repeat what Vaishampayana once told Janamejaya, at a sacrifice convened because a king's father was killed by a snake and halted because a boy born of two warring worlds asked for mercy. The book does not just contain a story. It contains an account of how that story survived long enough to be told at all, layer folded inside layer, until it reaches whoever happens to be reading it next.