You can't take a census of a city that's been dead for two thousand years. Pataliputra — Ashoka's capital, buried six to eight meters under modern Patna, impossible to excavate under a living city's foundations — left no tax rolls, no household registry, nothing that would tell you directly how many people lived there.
And yet historians confidently cite a number: somewhere between 900,000 and a million people. How do you get a population estimate for a city you can't dig up and can't count?
The answer turns out to be a chain of reasoning built from two other things entirely — a Greek diplomat's account of walls and towers, and a smaller city that archaeologists actually could dig up. Neither piece, on its own, tells you how many people lived in Pataliputra. Chained together, they do.
What survives is the Indica — a monograph by Megasthenes, the Greek envoy Seleucus sent to Chandragupta's court around 303 BCE. The original is lost, but extensive fragments survive quoted in later Greek and Roman writers.
Megasthenes wasn't a neutral observer — the historian Klaus Karttunen's sober verdict is that "he was not consciously distorting his view of India, but he placed his Indian experience within a Greek ideological framework and projected a Greek utopia on India... He was, however, selecting precisely those points which correspond well to the Greek conception of an ideal state."1 His account of Indian society and customs has to be read with real caution.
But his account of physical structures checks out differently. He gives precise dimensions for Pataliputra's fortifications — a defensive rampart, sixty-four gates, 570 towers.
The art historian Dieter Schlingloff cross-checked the numbers against known ancient engineering standards and found they line up with almost mathematical precision: "This distance corresponds almost exactly to the required 54 m [metre] interval between the towers... That the relation of the number of towers and wall intervals of Megasthenes corresponds exactly to the law confirms the reliability of his city plan."1
Wherever limited archaeological excavation has been possible — wooden defensive works predating anything found at other ancient Indian sites — it corroborates what Megasthenes described. He was, on matters of physical measurement, a careful and reliable observer, even where his ethnographic commentary wasn't.
That split matters more than it might seem. It means "is Megasthenes reliable" is the wrong question to ask about any given claim from the Indica. The right question is "is Megasthenes reliable about this specific kind of claim" — and the engineering data passes a test the ethnography fails.
So here's the move Schlingloff makes, and it's a clean piece of comparative reasoning worth naming as its own technique. Megasthenes gives you Pataliputra's physical footprint with confidence — a rampart 33.8 kilometers in circumference, enclosing 25.5 square kilometers.
That's enormous by ancient standards: three times the size of Alexandria, more than double Rome within the Aurelian walls, over eleven times the size of Athens.1
But Megasthenes never gives a population figure — he's silent on headcount. The physical measurements are solid; the number that actually matters (how many people lived inside them) is missing entirely. Size of the container tells you nothing about how full it was, unless you have some independent way to estimate density.
Schlingloff's solution: find a smaller ancient Indian city that has actually been excavated well enough to estimate its population density, then scale up by the ratio of physical size.
Kausambi, the next-largest city of ancient India, has been excavated enough to support a population estimate of 90,000 to 100,000 people. Schlingloff calculates Pataliputra was roughly ten times larger than Kausambi.
Multiply the ratio through: 900,000 to 1 million inhabitants.1 The entire estimate rests on one assumption doing all the work — that population density in Kausambi and Pataliputra was roughly comparable, so that a tenfold difference in area implies a tenfold difference in headcount.
Even discounting that figure substantially for uncertainty, you're still left with a city of well over half a million — one of the largest urban populations anywhere on earth at that moment in history. The demographer Tim Dyson, working independently with different density assumptions (230 persons per hectare across an estimated 1,350 to 2,200 hectares "within the ramparts"), arrives at a lower but structurally similar figure — 310,000 to 500,000 — and concludes Pataliputra was "likely... one of the world's most populous cities at the time."1 Two independent methods, different inputs, converging within the same order of magnitude.
The technique here isn't really about Pataliputra specifically — it's a general answer to a recurring problem in ancient demography: how do you estimate the population of a site you can measure but can't excavate, using a comparator site you can excavate but that isn't the site you actually care about?
Pataliputra is simply the case where the stakes happen to be highest — the largest city in the ancient Indian subcontinent, sitting directly under a modern city that will never permit the excavation that would settle the question directly.
The chain of inference has two separate reliability checks built in. First, does the unexcavated site's physical description hold up against independent corroboration (Megasthenes's tower-spacing math checking out against known engineering ratios)? Second, does the comparator site's own density estimate rest on real excavated evidence rather than another layer of guesswork (Kausambi's figure comes from actual archaeology, not a further extrapolation)?
Both checks have to pass independently. A physical description that checks out but a comparator built on guesswork gives you a precise-looking number resting on a soft foundation; a solid comparator paired with an unverified physical description gives you the reverse problem.
Only when both links hold does the final number deserve any confidence — and even then, Dyson's independent, lower estimate is the honest reminder that "extrapolation from a ratio" produces a plausible range, not a fact.
Break either link and the whole estimate collapses back into guesswork. If Megasthenes's tower-spacing math hadn't checked out against known engineering ratios, there'd be no reason to trust his rampart dimensions at all, and the entire extrapolation would have nothing solid to scale from.
If you're evaluating any ancient-demography claim built this way — not just Pataliputra's — here's the check to run.
Verify the witness's reliability domain-by-domain, not globally. Megasthenes is unreliable on ethnography and reliable on engineering. Don't accept or reject a source wholesale; ask which specific category of claim you're looking at and whether that category has independent corroboration.
This is the single most transferable step — most ancient witnesses, like most modern ones, are reliable about some things and unreliable about others, and the boundary rarely falls where a casual reader would guess.
Trace the comparator's own evidentiary base. Kausambi's density figure comes from actual excavation, not from a further chain of extrapolation.
If the comparator city's own number were itself an estimate built on guesswork, the whole chain would be extrapolation stacked on extrapolation, and the final confidence should collapse accordingly.
Check whether independent methods converge. Schlingloff's ratio-based method and Dyson's density-based method used different inputs and different assumptions, yet landed in the same broad order of magnitude. Convergence from independent methods is meaningfully stronger evidence than either method standing alone — divergence would have been a red flag.
Report the range, not a false point-estimate. The honest output of this method is "roughly half a million to a million," not "973,000." Any presentation that manufactures false precision from a ratio-based estimate has broken the discipline the method depends on.
Treat divergence between independent estimates as information, not noise. Schlingloff's and Dyson's numbers don't match exactly. That gap isn't a flaw to average away — it's a rough measure of how much uncertainty the underlying method actually carries, and it should travel with the estimate whenever it's cited elsewhere.
The corroboration is genuinely strong on the physical-dimensions side — Megasthenes's numbers check out against engineering standards and against what limited excavation has confirmed. That's independent, checkable corroboration, not just internal consistency within Megasthenes's own account.
The population figure itself is the weaker link in the chain: it depends on Kausambi being a valid comparator for Pataliputra's density (a different city, on different terrain, functioning as a different kind of urban center — imperial capital versus regional center), and on the size-ratio itself being accurate.
Neither of those two dependencies can be independently verified the way the tower-spacing math could be. They're reasonable assumptions, not confirmed facts, and the honest reading of the final population figure has to carry that distinction forward.
Dyson's independently-derived lower estimate (310,000–500,000) versus Schlingloff's higher one (900,000–1,000,000) is not a resolved disagreement — both remain live in the scholarship, and Olivelle presents both without adjudicating between them. The gap between the two isn't small: Schlingloff's low end and Dyson's high end nearly touch, but Schlingloff's high end is roughly double Dyson's high end.
That's a real, unresolved range, not a rounding difference — and any claim citing a single figure for Pataliputra's population should be read as picking one end of a genuine scholarly disagreement rather than reporting a settled fact.
Neither Schlingloff nor Dyson has the kind of decisive corroborating evidence that would let one figure retire the other. Barring new excavation directly under Patna — unlikely given the living city sitting on top of the site — this is probably a permanent range, not a temporary disagreement waiting for more data.
This material sits entirely outside Lahiri's existing vault coverage of the Mauryan capital and territory — her hub treats Pataliputra as backdrop rather than as a demographic puzzle worth its own methodological treatment.
Olivelle's chapter isolates the population-extrapolation problem as a case study in method, which gives the vault a worked example of comparative-archaeological reasoning it didn't previously have on the Ashokan corpus.
Within Olivelle's own chapter, Schlingloff and Dyson function as an internal author-tension of their own — two independent scholars, different eras, different assumptions, genuinely different numbers, both cited without either being declared correct. Olivelle's choice to present both rather than pick a winner is itself an application of the portrait-not-biography discipline documented elsewhere in this vault: state the possible and the plausible, and resist the pull toward false certainty.
Business: Business Index — Schlingloff's method is structurally identical to how a modern analyst estimates an unmeasurable quantity — total addressable market, a competitor's headcount, a private company's revenue — from a measurable proxy plus a known ratio from a comparable, better-documented case.
Walk the mechanism on both sides. An analyst estimating a private competitor's revenue might know its office square footage (measurable, like Megasthenes's rampart dimensions) but not its headcount or sales (unmeasurable, like Pataliputra's population). The move: find a public company in the same industry with known revenue-per-square-foot, then scale the private competitor's footage by that ratio. The reliability of the final estimate depends entirely on two checks, identical to the ancient case: is the proxy measurement itself independently verified, and is the ratio derived from real data (audited public filings) rather than another guess.
The insight neither domain produces alone: ancient historians and modern market analysts are running the identical inferential move — chained comparator-ratio extrapolation — and both disciplines share the same failure mode when either link in the chain goes unchecked. A market analyst who skips verifying the comparator company's own numbers is making Schlingloff's exact error, just with spreadsheets instead of Greek fragments; recognizing the shared structure means a due-diligence checklist built for one domain transfers directly to the other.
Creative Practice: Historiography Hub — MacCulloch's "may have" discipline shows up here in its most quantitative form: Dyson's own hedge, calling his estimate "just an educated guess drawing on the work of several other writers," is the numerical equivalent of a historian's verbal hedge.
Walk both sides of the parallel. A narrative historian hedges with language — "may have," "probably," "it is likely that." A quantitative historian hedges with a range instead of a point, and with an explicit statement of method ("educated guess drawing on the work of several other writers") rather than presenting the number as measured fact. Both moves do the same epistemic work: they keep the reader's confidence calibrated to the actual evidence rather than to the rhetorical or numerical form the claim happens to take.
The insight neither domain produces alone: a number stated with false confidence (900,000-1,000,000, presented as settled) is just as much an overclaim as an unhedged narrative sentence would be. The discipline of flagging uncertainty has to survive translation into figures, not just prose, or the reader inherits a precision the evidence never actually supported — readers are far more likely to distrust a hedged sentence than a hedged-looking number, which makes numerical overclaiming the more dangerous and more common failure mode of the two.
Sharpest Implication
If Pataliputra genuinely held somewhere in the range of half a million to a million people in the third century BCE, then Ashoka governed, from a single capital, an urban population comparable to or larger than any city in the Mediterranean world at the same moment — and this fact is recoverable at all only because two independent scholars, working a century apart and using different methods, happened to land in the same broad range. Ancient demography this significant survives on the thinness of convergent extrapolation, not direct evidence — a genuinely humbling ratio of confidence to certainty for a fact this consequential.
Generative Questions