Evidence page · voice
Debate in these texts is staged, named and answerable. Instruction is not. The rules that shaped a society for two thousand years are, with striking consistency, spoken by nobody, by a dying man, or by a figure who cannot be cross-examined.
Counted, and the result is absolute
| Text | āśrama | gṛhastha | vānaprastha | saṃnyāsa |
|---|---|---|---|---|
| Rigveda | 0 | 0 | 0 | 0 |
| Paippalāda Atharvaveda | 0 | 0 | 0 | 0 |
| Bṛhadāraṇyaka Upaniṣad | 21 | 5 | 3 | 21 |
| Chāndogya Upaniṣad | 17 | 24 | 2 | 0 |
| Āpastamba Dharmasūtra | 2 | 0 | 2 | 0 |
| Gautama Dharmasūtra | 6 | 4 | 0 | 0 |
| Baudhāyana Dharmasūtra | 9 | 4 | 4 | 4 |
| Manusmṛti | 40 | 30 | 2 | 8 |
Not one occurrence of āśrama, gṛhastha, vānaprastha or saṃnyāsa in the Rigveda or the Atharvaveda. The four-stage life does not exist in Vedic literature. It appears in the Dharmasūtras — c. 500–300 BCE, the same window as the Buddha and Mahāvīra.
And Olivelle showed the shape changed. In the earliest Dharmasūtra formulation the āśramas are four ALTERNATIVES a young man chooses between after studentship — not a sequence. The classical model, in which they are consecutive stages and renunciation comes last, is a later development.
Renunciation began as an option and was converted into a stage you reach at the end of a life of obligation.
That conversion neutralises it without banning it. A young man who could once choose the wandering life is now told he may have it — after studentship, after marriage, after sons, after a lifetime of ritual debt. The Buddha left at twenty-nine. Under the classical āśrama scheme he would have been out of order.
So I was wrong to describe the āśramas as always sequential. The sequence is the edit. And the Rigvedic figures who do resemble ascetics — the muni of RV 10.136, long-haired, wind-girt, and the vrātya bands of Atharvaveda 15 — are outside any stage system, because no stage system exists yet.
Named speech against rule language, measured across seven texts
The claim that the tradition argues in dialogue and legislates in anonymity is testable. Count the markers of named speech — uvāca, hovāca, abravīt, papraccha, brūhi — against the markers of rule language — the optatives kuryāt, kurvīta, bhavet, varjayet, vidyāt, plus daṇḍa and prāyaścitta, penalty and expiation.
Per 10,000 words. The multiplier at the right is rule language divided by named speech. Source texts: GRETIL.
In the Upaniṣads, named speech EXCEEDS rule language. And on a clean text the gap is far wider than the first measurement showed.
The standard digital Bṛhadāraṇyaka file bundles Śaṅkara's eighth-century commentary with the Upaniṣad — 85,467 words against roughly 17,000 of actual text. Measured on the commentary-free version (Śatapatha kāṇḍa 14): dialogue 76.0, rule 8.2 — a ratio of 9 to 1 the other way. The contaminated file gave 37.3 against 27.5, because Śaṅkara writes in rule language and the Upaniṣad does not.
Chāndogya, still commentary-bundled and therefore an underestimate: 42.6 against 26.1.
In the law codes the ratio inverts. Gautama: 41× more rule language than named speech. Manu: 19×. Vasiṣṭha: 13×. Āpastamba: 12×.
So the true spread is roughly 9:1 dialogue in the Bṛhadāraṇyaka against 41:1 rule in Gautama — a factor of about 370 between the two genres.
Where people are named, the texts argue. Where nobody is named, they bind.
Gautama's Dharmasūtra has 2.2 markers of named speech per 10,000 words and 88.4 of rule. It is almost entirely a voice with no owner issuing instructions with penalties attached.
The Khila Kāṇḍa — the Bṛhadāraṇyaka's supplementary sixth chapter, where the instruction to strike a woman sits — does not show higher rule density than the dialogue chapters. Chapters 1–5 measured 30.1; chapter 6 measured 14.2.
The hypothesis was badly formed. The Khila Kāṇḍa is a miscellaneous ritual appendix, not a law code; expecting law-code density from it tested the wrong thing. And the figures were taken from the commentary-bundled file, so they measure Śaṅkara as much as the text.
The observation that survives needs no density test: it is appended material with no named speaker, in a text otherwise built entirely from named debates.
What survives: the genre-level separation between dialogue literature and law literature is large and consistent. The claim that any particular anonymous passage is anonymous in order to bind is an interpretation, and this measurement does not establish it.
The śramaṇas did not have "āśramas." They had renunciation — the wandering mendicant, the bhikṣu, outside household and ritual entirely.
The Vedas have no framework for this at all. Zero occurrences of every term. What the Rigveda has is the muni of 10.136 — long-haired, wind-girt, drinking with Rudra — a wild figure with no institutional place.
Then the Dharmasūtras produce the āśrama system, in the same window as the Buddha, and in its earliest form it is a menu: four legitimate ways to live, chosen after studentship. Renunciation is one permitted option among four.
Then it becomes a sequence, and renunciation moves to the end — available only after a full life of study, marriage, sons and ritual obligation.
Olivelle's argument is that the system arose to accommodate renunciation, not that renunciation grew out of it. The framework is the answer to a challenge, not the source of it.
First it was banned by having no name. Then it was licensed as one option. Then it was permitted only at the end of a life.
Each step concedes more and controls more. The final form allows a brahmin to become exactly what the Buddha became — at sixty, with permission, after discharging every debt the system could name.
Named speech versus anonymous instruction
| Passage | Attributed to | Can it be answered? |
|---|---|---|
| BU 3.6, 3.8 — Gārgī's challenge | Gārgī and Yājñavalkya, before a named king, in a named hall | YES. Two people, on the record, with an audience |
| BU 2.4 — Maitreyī refuses the property | Maitreyī and Yājñavalkya | YES |
| ChU 5.3 — Pravāhaṇa teaches Uddālaka | A named Kṣatriya king, who says the knowledge never passed to a Brahmin | YES — and the concession is the reason the passage matters |
| BU 6.4.7 — strike the woman who refuses | NOBODY. Khila Kāṇḍa, the supplement. Impersonal instructional voice | NO. There is no speaker to hold to it |
| RV 10.90 — the four varnas | Nobody in the hymn. The Anukramaṇī assigns it to "Nārāyaṇa" — a later index naming a deity | NO |
| Manu, entire | Manu, the mythological first man, relaying Brahmā | NO. The author is a cosmological figure |
| Anuśāsana Parva — 6,500 verses on gifts to brahmins | Bhīṣma, dying on a bed of arrows | NO. A deathbed speech is structurally unanswerable |
| Gītā 18.41–48 — duties by varna | Krishna, as God | NO |
| The Dharmasūtras | Āpastamba, Gautama, Baudhāyana, Vasiṣṭha — eponymous sages, not documented persons | NO |
| Śambūka's execution | Uttarakāṇḍa — the book most widely held to be a later addition | NO |
Every binding rule above is spoken by something that cannot be questioned — an impersonal voice, a mythological ancestor, a god, a dying man, or a later index.
Every passage where the hierarchy is challenged is spoken by a named person in a room with other people in it. Gārgī has an audience. Pravāhaṇa has a pupil. Maitreyī has a husband who has to answer her.
Dialogue is where the tradition argues. Anonymity is where it legislates.
This is not a coincidence of genre. It is what genre is for. A debate invites reply by construction. An unattributed instruction forecloses it by construction. And the material that shaped the social order is almost entirely in the second form.
| Who | Identified as | Where | What they do | Impact |
|---|---|---|---|---|
| Gārgī Vācaknavī | Daughter of Vacaknu, patronymic only. No husband, no household mentioned — she appears purely as a disputant | Janaka's hall, Videha — the eastern frontier, beyond the Sadānīrā | Asks on what the world is woven and climbs the chain to Brahman; is threatened; returns in a second round and asks about the imperishable | The only woman in the Upaniṣads present as a professional debater rather than a wife. Later tradition names her a brahmavādinī |
| Maitreyī | Yājñavalkya's wife; the text contrasts her with Kātyāyanī, who has "only women's knowledge" | The household, at the moment of his departure | Refuses the property settlement and asks for the teaching instead | The exchange that follows is among the most quoted in the Upaniṣads — and it begins with a woman declining money |
| Pravāhaṇa Jaivali | Kṣatriya king of Pañcāla | Pañcāla — not the east, correcting a common claim | Teaches the five fires and two paths to the Brahmin Uddālaka, saying the knowledge never before passed to a Brahmin | The doctrine survives in two independent recensions. The caste-womb clause is in only one |
| Aśvapati Kaikeya | King of Kekaya — the northwest | ChU 5.11 | Teaches six Brahmins who come to him. Refuses their gifts | Erased. Janaka, who paid, became proverbial. Aśvapati, who refused payment, did not |
| Satyakāma Jābāla | Son of Jabālā, a servant woman. Does not know his father | ChU 4.4 | Tells the truth about his parentage; is accepted because he told it | Preserved, and repeatedly explained away by later commentators |
| Kavaṣa Ailūṣa | Son of a dāsī, a slave woman | Aitareya Brāhmaṇa | Driven from the sacrifice by other priests; the river follows him and he is recalled | The tradition preserved its own counter-example and did not delete it |
| Śākalya | Brahmin, Yājñavalkya's rival | BU 3.9 | Questions past the limit | His head shatters. He dies; his bones are lost on the road. The threat made to Gārgī is carried out on a man |
Three of the seven are not Brahmins. Two are kings; one is the son of a slave. Two are women. One does not know his father.
The Upaniṣads are a record of Brahmins being taught by people the later law code would bar from the texts. And the tradition kept the record.
That is the strongest single argument on this platform, and it is made entirely out of material the tradition itself transmitted.
| Passage | Structural position | What it establishes |
|---|---|---|
| BU chapter 6 — Khila Kāṇḍa | Explicitly a supplement, appended after the main body | Conception procedure; the instruction at 6.4.7 to strike a woman who refuses |
| RV 10.90 | Latest book, three of its key words occurring nowhere else in the corpus | The four varnas |
| Uttarakāṇḍa | Widely held to be a later addition to the Rāmāyaṇa | Śambūka's execution; Sītā's second exile |
| Anuśāsana Parva | Didactic book attached to a war narrative it does not affect | Gifts to brahmins; the degradation lists |
| ChU 5.10.7's womb list | Inside a named king's speech, but absent from the parallel recension | Birth-rank as earned by conduct |
In a tradition that preserves attribution obsessively, anonymity is a choice. This is a literature that records which poet's family owns which book, which teacher heard what from whom, which school disgorged its learning to which rival. The gotra system exists to track descent from named composers.
Against that background, an unattributed instruction is conspicuous. It is not that nobody remembered who said it. It is that the passage functions better with no one behind it.
And the Anukramaṇī shows the discomfort: where a hymn has no author, the index assigns one — 10.90 to "Nārāyaṇa." The tradition would rather name a deity than leave a blank.