Made visible · counted, not asserted

Even in the later Vedic corpus, the four castes are thin, ritual, and rare.

The caste order claims to be the structuring reality of Vedic society. So we counted. In the Taittirīya Saṃhitā — a large later-Vedic text of some 163,000 words — the four class-names are individually scarce, wildly unequal, and cluster together as a ranked set in exactly two places, both ritual. The "untouchable" is absent entirely. A society organised by four fixed castes does not leave a record this thin.

The count.

Taittirīya Saṃhitā · 163,264 pada-tokens · whole-token matched

The Taittirīya is stored as a pada text — word-by-word, cleanly separated — which means the terms can be counted honestly, without the substring noise that inflates unspaced texts. Every occurrence of each class-word, matched as a whole token:

Whole-token counts in the Taittirīya Saṃhitā (163,264 tokens), this platform's corpus scan. "kṣatriya" is rare (6) because the older text prefers rājanya (27) for the warrior class — itself a sign the fixed four-term scheme is not yet standardised. Reproducible: the text and method are published.

What the numbers say.

three findings

1 · The priest-word dominates; the lower classes are scarce. Brāhmaṇa (84) outnumbers vaiśya (8) and śūdra (9) roughly tenfold. The text is overwhelmingly concerned with the priestly class it was produced by — not with a balanced four-part society.

2 · The warrior-term isn't even fixed yet. "Kṣatriya" — the standard later word — appears just 6 times; the older rājanya (27) is preferred. A society running on four canonical varṇas would not still be wavering over what to call one of them.

3 · The untouchable does not exist here. Caṇḍāla: zero. As on the platform's jātization page, the cruellest category of the caste system is simply absent from the later Vedic layer — it arrives with the law codes.

The four-varṇa order clusters in exactly two places.

and both are ritual, not social

The real test is not how often each word appears, but whether the four-fold ranked set appears as a description of society. Scanning the whole text, the four class-names co-occur as a group in only two loci — and neither is a social census. Both are ritual schemas, assigning verses or qualities to classes within a rite:

In both, the varṇas appear because a ritual needs to distribute something — stanzas, offerings, symbolic qualities — across the social categories. This is the fourfold order as a liturgical device, invoked in the ritual manual, not the fourfold order as the lived, enforced, birth-bound structure of society that the later law codes would build. The scheme exists; it is not yet the cage.

TS 2.5.10.1, TS 6.2.5.3 — the only two loci in the Taittirīya Saṃhitā where 3+ of the four class-names co-occur; both ritual-assignment contexts. Machine-verified on this platform's corpus. machine-verified · corpus

What this establishes — and its limits.

the discipline

• Verified, reproducible: the whole-token counts and the two-loci co-occurrence result, from the Taittirīya pada text on this platform. Anyone can re-run it.
• The claim: the four-varṇa order is thin, unequal, name-unstable, and ritually (not socially) deployed in this later-Vedic Saṃhitā; the outcaste is absent. This is consistent with — and adds corpus weight to — the platform's thesis that rigid birth-caste is a later Dharmaśāstra construction.
• NOT claimed: that varṇa was unknown or that no ranking existed. It existed as a scheme — the two loci prove it. The claim is about its thinness and ritual character, not its absence.
• NOT claimed: raw-count comparison across texts as a precise diachronic curve — the Śatapatha is unspaced and cannot be disambiguated the same way, so this page reports only what the cleanly tokenizable Taittirīya supports. (The Śatapatha caṇḍāla-absence holds for books 1–13, the Brāhmaṇa proper; the one occurrence is in book 14, the Upaniṣad — its other counts are not published here as figures.)

The ledger.