1.3 — How Old Are These Stories, and How Do We Know

Ask when the Mahabharata was written and you will get answers ranging from 400 BCE to 3102 BCE. That spread is not a scandal, and it is not evidence that nobody knows anything. It is the result of a real ambiguity: a text and the events it describes are two different things with two different dates, and most arguments collapse them.

There are actually three dates to keep separate.

When the events happened, if they happened. When the story was first composed as a recognisable work. And when the text reached the form we can read, which for these epics is centuries after the second date.

How anything gets dated

Before the answers, the methods, because the methods are what make an answer worth anything.

Language change. Sanskrit changed over time in ways that are traceable, like English between Chaucer and now. The Vedic Sanskrit of the oldest hymns is noticeably different from the Classical Sanskrit of the epics, which is different again from later Puranic Sanskrit. A text can be placed in a rough band by its grammar and vocabulary alone, and a passage inserted later usually shows up because its language is younger than its neighbours.

What the text knows about. This is the most powerful tool available and it works in both directions. If a text mentions iron, it is after iron. If it mentions a city, it is after that city was founded. If it mentions coins, writing, particular kingdoms, particular kinds of warfare, that is a floor. Equally telling is what a text does not mention. The core of the Mahabharata is strikingly unaware of things that dominate later Indian life — it barely knows Buddhism, barely knows large-scale coinage, and its geography stops well short of what later texts take for granted.

Physical manuscripts. These give a hard latest-possible date but they are misleading on their own, because Indian manuscripts are written on palm leaf and birch bark, which rot. Copies were made and recopied continuously. The oldest surviving Mahabharata manuscripts are medieval, which tells you almost nothing about composition.

Cross-references from outside. When a Greek visitor, a Buddhist canonical text or an inscription mentions a work or a character, that is an independent floor. Greek writers in India after Alexander mention Herakles worshipped by a people around Mathura, which is very plausibly Krishna, and that places a Krishna cult before roughly 300 BCE.

Archaeology of the places named. The sites exist. Hastinapura, Indraprastha, Kurukshetra, Mathura, Ayodhya, Dwarka have all been excavated. Occupation layers can be dated. This does not confirm the story, but it tells you when there was actually a settlement there for a story to happen in.

And astronomy, which needs a warning. The epics contain descriptions of planetary positions and eclipses, and it is possible to run planetary motion backwards and find dates that fit. This has produced dozens of confident and mutually contradictory answers, from 5561 BCE to 1478 BCE. The method is unreliable here for a simple reason: the astronomical passages are among the most likely to have been inserted or embellished by later hands who knew astronomy, the descriptions are poetic rather than precise, and with enough free parameters a fit can be found for almost any target date. When someone gives you a precise year for the war based on planetary positions, that precision is manufactured.

The Ramayana

Composition of the core: roughly the fifth to fourth centuries BCE, with the story existing in oral form earlier.

The evidence is internal and consistent. Books two through six — the exile, the forest, the search, Lanka, the war — form a tight, coherent narrative with older language. Books one and seven, which contain most of the material presenting Rama as a divine incarnation and which include the ending discussed in 2.15, are linguistically later and structurally detachable. The epic's original hero is a man, and the divinity is a later layer. This is one of the most robust findings in the field and it changes how you read the whole thing.

The geography also helps. The Ramayana's world is northern India plus a long journey south, and its knowledge of the far south is vague in exactly the way an author who had not been there would be vague. It knows about Greeks and Scythians only in passages that are otherwise identifiable as late.

The text kept growing until roughly the second century CE, reaching about 24,000 verses.

And the events, if any? Unfixable. A Kosala kingdom certainly existed and was a real power in the Gangetic plain. Ayodhya is a real, continuously occupied city. Beyond that, the honest answer is that there is no way to date a possible historical Rama, and anyone who gives you a year is not working from evidence.

The Mahabharata

Composition: roughly the fourth century BCE to the fourth century CE, which is a span of about eight hundred years — and that span is the single most important fact about the book.

It did not grow evenly. The tradition itself remembers three stages, and remarkably the text preserves the memory of its own expansion: it says it was once called Jaya with 8,800 verses, then Bharata with 24,000, and finally the Mahabharata with about 100,000. A text that records its own tripling is telling you something honest about itself.

The layers are visible. The war narrative is tight and old. The enormous ethical and philosophical sections — the whole of Book 12, in which a dying Bhishma delivers many thousands of verses of political and moral instruction — are later insertions into a gap in the action. The Bhagavad Gita itself is very likely an insertion, though an old and superb one, and 5.1 goes into why that matters and why it does not diminish it.

The war, if there was one, is usually placed around 1000 to 900 BCE by scholars who accept a historical kernel, on the reasoning that this fits the Painted Grey Ware archaeological culture found at Hastinapura and Kurukshetra, fits the late Vedic political situation in which the Kuru people were the dominant power, and fits the epic's own picture of a world of chariots and clans rather than empires and cities.

Hastinapura's excavation produced one detail worth knowing: a layer showing the settlement was destroyed by a flood of the Ganga, and the epic tradition independently records that Hastinapura was abandoned after a flood and the capital moved to Kaushambi. Two independent lines agreeing on a mundane detail is exactly the kind of thing that suggests real memory underneath.

What the dating actually tells you

Three things, and each one changes how you read every later page.

First, these texts are compilations, not compositions. Nobody sat down and wrote the Mahabharata. Generations of reciters, editors and priests added to it, and every century that touched it left its concerns in it. When you find a passage that contradicts another passage — and there are many — you are usually looking at two different centuries arguing.

Second, the moral sophistication is genuinely ancient and the theology is often later. The hard ethical dilemmas, the awkward silences, the fact that the heroes do terrible things and the text does not flinch — these belong to the older layers. The passages that smooth everything over with divine explanation tend to be younger. Which means the difficult version is the original version, and this volume's willingness to sit with the difficulty is not modern squeamishness. It is closer to the source.

And third, growth over eight centuries is exactly the process 1.4 describes. A story that passes through forty generations of retelling does not stay the same size. Knowing the mechanism lets you read backwards through it.

What to carry from this page

Separate the date of an event from the date of the account of it. This one habit will save you from most bad arguments about ancient texts, and from a fair number about modern ones.

Be suspicious of precision that the evidence cannot support. A range of centuries offered honestly is worth more than a specific year offered confidently.

And notice when independent lines of evidence agree on something boring. A flood layer at Hastinapura and a tradition about a flood at Hastinapura is worth more than any number of dramatic confirmations, because nobody invents a boring detail.

Next: 1.4 shows the exact mechanics by which a plain story becomes a miraculous one over twenty-five centuries — with the same episode traced through four versions, so you can watch it happen.