Method page · the sources themselves

Everything that survives — and how little of it has been opened.

The archive of ancient South Asia was assembled by accident — by climate, by scribal effort, by which languages had patrons, by which sites got dug. It is larger and stranger than a list of text corpora suggests: it includes charred seeds, tooth enamel, cave sediment, living oral traditions, and at least four scripts nobody can read. Before any argument can be weighed, the archive needs describing.

HOW EACH BODY OF EVIDENCE WAS ACTUALLY USED no percentages — there is no denominator for "how much of the archaeological record exists" Sanskrit textual corpus COUNTED 1.8M words through one pipeline Archaeology — site records CATALOGUED 150 sites, 315 dated windows Ancient DNA literature READ published results, critically Epigraphy CITED never counted as a corpus Dravidian lexicon QUERIED DEDR, by keyword Near Eastern texts CITED in translation, for Meluhha Pali & Prakrit SECONDARY argued from description, not text Tamil corpus SECONDARY quoted from memory, never counted Chinese accounts SUMMARISED from general knowledge, unverified Numismatics UNTOUCHED Archaeobotany UNTOUCHED Isotope & palaeopathology UNTOUCHED Living oral tradition UNTOUCHED Indus script UNREADABLE structurally, not temporarily

Categories, not quantities.

The bodies of evidence — the full list

Not eight. Closer to twenty, and several are not textual at all
BodyWhat it can settleUsed here
WRITTEN
Sanskrit corpusWhat one literate tradition said, and how its vocabulary shiftedHeavily
Pali · Prakrit · Jain ĀgamaThe śramaṇa side, contemporary and independentThinly
Tamil corpusA complete literate tradition with no varna frameworkCited, not counted
Epigraphy — Aśokan, Tamil-Brāhmī, copper plates, temple wallsDated, located, administrative. Land, endowment, official languageCited, not counted
Outside observers — Greek, Roman, Syriac, Chinese, Arabic, PersianDescription from people with no stake in the self-accountBarely
NumismaticsDated, quantitative, tied to polity and metal supplyNo
Manuscript colophons and scribal recordsWhere and when a text was copied, and by whomNo
MATERIAL
ExcavationSettlement, technology, trade goods150 sites
ArchaeobotanyCharred seeds and phytoliths date crop domestication and spread. Millet, rice, wheat, cotton — this is how the agricultural argument gets settled, and it does not depend on any textNo
ZooarchaeologyWhich animals, when, and whether domesticated. The horse question is an archaeozoological question before it is a textual oneNo
Isotope analysisStrontium and oxygen in tooth enamel show whether an individual grew up where they were buried. Direct evidence of individual mobility — migration at the scale of one personNo
PalaeoclimateSpeleothems, lake cores, pollen. The Ghaggar-Hakra drying and the monsoon shift that ends Harappan urbanism are climate data, not archaeologyMentioned
Metallurgy and residue analysisOre sourcing, alloy composition, crucible technique. Wootz has no manual and a full material signatureNo
Iconography — Bharhut, Sanchi, Amaravati, AjantaWhat things looked like, who was depicted, and what instruments were playedOnce
BIOLOGICAL
Ancient DNAAncestry proportions, mixing datesYes
Modern population geneticsFounder effects, ROH, consanguinity structureYes
PalaeopathologyDisease, nutrition, violence and labour, read off skeletons. Whether a population was fed and whether it was beatenNo
LIVING
Historical linguistics beyond the two big familiesMunda, Tibeto-Burman, Nihali, Burushaski, Kusunda, the Vedda substrate, Andamanese. Each carries vocabulary not accounted for by Indo-Aryan or Dravidian. That is not the same as older — Munda is Austroasiatic and may have arrived later than Dravidian, and an isolate's vocabulary cannot be dated at all. Unaffiliated, not ancientBarely
Living oral traditionĀdivāsi law, medicine, ecology, genealogy and song — transmitted continuously, recorded rarelyNo
Ethnography and craft practiceTechniques still performed that no text describesNo
ToponymyPlace names outlive the languages that made them — the densest surviving substrate evidence anywherePartly
UNREAD
Indus scriptNothing linguistic, on present evidenceAs absence
OthersProto-Elamite, Linear Elamite, undeciphered Vindhyan and megalithic graffiti marksNo

The strongest evidence for when farming arrived is a charred seed, and this platform has never looked at one.

Outsider accounts credit the wrong place too

An independent observer is not a neutral one

Outside accounts are valuable because their authors had no stake in South Asian self-description. They are unreliable for a different reason: they credit whoever handed them the thing.

ObserverCreditsProblem
Greek writersEgypt, repeatedly, for astronomy, geometry and medicineAlexandria was where they encountered the material. Babylonian planetary theory and Indian material both reached them through Egypt, and both were credited to Egypt
Arabic writersGreece, for the philosophical corpusAccurate about the immediate source and silent about what the Greeks had absorbed. Al-Khwārizmī is the honourable exception — he titled his book after the Indian method
Syriac translatorsGreek authoritiesThey were the conduit and credited the language they translated from, not whatever lay behind it
Latin EuropeArabs — "Arabic numerals", "Damascus steel"Names the last merchant in the chain. The steel ingots came from South India and Sri Lanka
Chinese pilgrimsIndia, correctly and directlyThey travelled to the source. Which is why they are the most reliable outsiders on this list, and why their absence from this platform matters

The general rule: an observer credits the place they got it from, and the number of intermediaries determines how wrong that is. The Chinese pilgrims had none. Latin Europe had four. The reliability of an outside account is inversely proportional to the length of the chain behind it — and that is measurable, which makes it a usable test rather than a complaint.

The eight best-documented bodies, in detail

BodyExtent and conditionWhat it can settle
Sanskrit textual corpus Vast, continuously copied, now largely digitised. But: manuscripts are late — palm leaf rots in a monsoon climate, so a text from 1200 BCE may survive in copies from 1400 CE What one literate tradition said, and how its vocabulary changed. Not what anyone did
Epigraphy The Aśokan edicts are the earliest securely dated South Asian texts — and they are in Prakrit, not Sanskrit. Below them, hundreds of thousands of inscriptions: Tamil-Brāhmī cave labels, temple walls, copper plates recording land grants Dated, located, and largely non-literary. Land, endowments, administration, and the languages actually in official use
Archaeology 150 sites in this platform's atlas; the real total is far larger and unevenly distributed by excavation funding Material culture, settlement, technology, trade goods. Never intention
Ancient DNA A few hundred South and Central Asian genomes, sparse and heavily skewed by preservation — hot wet climates destroy DNA, so South Indian and eastern samples are scarce Ancestry proportions and mixing dates. Not language, not culture, not direction of travel
Pali and Prakrit corpora Large, well-edited, digitised — and preserved outside India, in Sri Lanka, Burma, Thailand and Tibet The śramaṇa side of every argument on this platform. An independent literate tradition contemporary with the Sanskrit one
Tamil corpus Sangam, Tolkāppiyam, Tirukkuṟaḷ, Tēvāram. Physically rescued from decaying palm leaf in the 1880s by U. V. Swaminatha Iyer A complete literate tradition with no varna framework. The control case for almost every claim here
Outside observers Megasthenes, the Periplus, the Chinese pilgrims Faxian, Xuanzang and Yijing, later Arab geographers Independent, dated, non-brahmanical description of Indian society from people with no stake in its self-account
The Indus corpus Around 4,000 inscribed objects. Mean sign sequence roughly five characters. Longest known text about seventeen. Undeciphered Nothing linguistic, and this is a hard constraint rather than a temporary one — see below

The three gaps that matter most

1 · The Chinese pilgrims are absent from this platform entirely

Faxian (c. 400 CE), Xuanzang (c. 630–645) and Yijing (c. 671–695) travelled in India, studied there, and wrote detailed accounts in Chinese. Xuanzang spent over a decade in the subcontinent, studied at Nālandā, and described kingdoms, monasteries, social divisions and religious practice systematically.

They are the best outsider evidence that exists — dated, extensive, written by literate observers with no investment in brahmanical self-description, and preserved in a completely independent manuscript tradition.

A platform that argues the Sanskrit record is self-interested has not consulted the most substantial independent description of the society that record describes.

This is the single largest omission here, and it is not a minor one. Their accounts bear directly on the Vedism-versus-Brahmanism argument, on the state of Buddhist institutions, and on how caste appeared to someone encountering it from outside.

2 · Epigraphy is cited but not counted

The Aśokan edicts are the earliest dated texts in South Asia, and they repeatedly pair bamhana and samana — brahmins and śramaṇas — as two categories deserving equal respect. That pairing, from a third-century BCE royal inscription, is harder evidence for the two-traditions argument than anything in the Sanskrit corpus.

They are in Prakrit, not Sanskrit — the state's language of public communication was not the priestly language.

And below the edicts sit hundreds of thousands of inscriptions recording land grants, endowments, temple administration and community rights. The land page here argues from Chola tenure categories that come from exactly this material, without treating it as a corpus that could be counted.

3 · The Indus constraint, stated properly

Roughly 4,000 inscribed objects. A mean sequence of about five signs. The longest known text is around seventeen.

That is not merely a small corpus. It is a corpus with a structural problem. Decipherment historically requires either a bilingual text, a known related language, or enough running text for statistical attack. The Indus material offers none of the three: no bilingual has been found, no descendant language is agreed, and five-sign sequences cannot support the distributional analysis that cracked Linear B.

So every claim about Indus language, religion or social structure is inference from objects — and this platform's Meluhha page is built from Mesopotamian sources precisely because the only people who wrote about the Indus in a readable script were its trading partners.

The honest position is that this may not be solvable, and that "not yet deciphered" quietly implies a solvability the corpus size does not support.

What the archive systematically over- and under-represents

OverUnder
Dry climates — northwest India, Central Asia, EgyptWet climates — Kerala, Bengal, the Deccan coasts. Palm leaf and DNA both fail there
Stone and fired clayWood, textile, leather, thatch — most of what most people used
Endowed institutions — temples, monasteries, courtsHouseholds, workshops, fields
Languages with patrons and copyistsLanguages without either
Sites near modern infrastructure and fundingSites that are remote, submerged, or under a living settlement
Whatever a colonial administration needed — law, revenue, landWhatever it did not
The compounding problem

These biases are not independent. A dry region with stone architecture, endowed institutions and a patronised written language will be represented many times over. A wet region with wooden architecture, household transmission and an unwritten language will be represented essentially not at all — and the two are not evenly distributed across South Asia.

The archive is not a thin version of the past. It is a heavily structured selection, and the structure runs in one direction.

Priorities, stated as a work order

1The Chinese pilgrims. Independent, dated, and directly on this platform's central question. The largest gap
2A Tamil corpus in the same pipeline as the Sanskrit. Aṟam against dharma, tiṇai vocabulary, caste terms across Sangam / Tēvāram / later bhakti. Nobody has put the two side by side with one instrument
3The Aśokan edicts as a counted corpusbamhana against samana, dhamma against dharma, and what a third-century BCE state actually legislated
4Pali and Prakrit. The śramaṇa side is argued here from secondary description rather than from text
5Numismatics and inscriptional land records — the only large dated datasets on this list that are quantitative by nature