One named figure. One volume. Two accounts of how he came to change sex, and they agree about nothing except that it happened.
In the first, he is born male and Śiva curses him into womanhood, for reasons the chapter gives in five words.
In the second, he is born female — "no sons were born from that sacrifice. Instead, the famous daughter, Ila, was born" — and two gods grant him manhood, as a reward for "humility, self-control and devotion to the truth."
Opposite starting point. Opposite agent. Opposite valence. Forty-two chapters apart, in two Saṃhitās, unflagged.
And the endnote attached to the second one restates the first as "the standard story" without mentioning that the page it is sitting on says otherwise.
Line against line:1
| ch 308-5(4) (Koṭi Rudra) | ch 350-6(36) (Umā) | |
|---|---|---|
| Starting sex | male | female — born as Ilā from a sacrifice that produced no son |
| Cause | a curse — "For the sake of his beloved, Shiva cursed him, and his followers, that they would become women" | a boon — Mitra and Varuṇa: "You will become the son who will extend Manu's lineage… known as Sudyumna" |
| Agent | Śiva | Mitra and Varuṇa |
| Valence | punishment | reward, for "humility, self-control and devotion to the truth" |
| Resolution | "alternated between being a man for one month and a woman for one month. Eventually, he ceased to be a woman" | reverts to bear Purūravas by Budha; then "as a result of Shiva's favours, she again became the son, Sudyumna", and permanently "had the signs of being both a man and a woman" |
⚑ Note the endings especially. Ch 308 resolves into a settled male state. Ch 350 does not resolve at all — it ends in a permanent both, after a cycle that includes bearing a child.
⛔ Unflagged as a contradiction anywhere in the volume. This is the build's thirteenth unaided find, and it is mechanically checkable: two named accounts of one named figure, quoted in full, with no note connecting them.
The apparatus does attach to ch 350, and what it does there is the sharper half of the finding.
Endnote 1379, in full:
*"Sudyumna had wandered into a grove where Shiva and Parvati were amusing themselves in private… According to the standard story, that is how Sudyumna became Ila."*2
That is the ch 308 version — a curse, from Śiva, running male-to-female — supplied as the standard story, in a note attached to the chapter whose body says the opposite.
⚑ Three things are true at once here and they should not be collapsed.
The note is not wrong. The curse version is standard in the wider tradition, and the note is a service to the reader.
The note does not say the two conflict. It offers the received account beside an incompatible one and leaves the reader to notice, which most readers will not, because the note reads as background rather than as correction.
And the note supplies a motive the body of ch 308 does not give. Ch 308's whole explanation is "for the sake of his beloved" — five words. Note 1379 supplies the grove, the privacy and the intrusion. ⛔ That content is an import from the tradition, unmarked as external, in the same numbered sequence as notes that gloss a word or record an emendation.3
⚑ So the endnote is doing three jobs — background, silent alternative, and unsourced narrative supply — and the reader has no way to tell them apart.
The consequence the body only implies is stated by the apparatus, and it is the hardest single line of its kind in the volume.
Ch 350's body has Sudyumna decline a kingdom in favour of Purūravas, on Vasiṣṭha's advice, and gives no reason: "As a result of the words spoken by the great-souled Vasishtha, Sudyumna instated Pururava."
Endnote 1381:
"Having been a woman earlier, Sudyumna was not entitled to a kingdom and Vasishtha advised that it should be given to Pururava instead."4
⚑ Set that against the fact that Sudyumna appears in this same volume on a list of the saved — one of a hundred and thirty-three devotees whose worship is recorded as having worked.5
Qualified for liberation. Disqualified from a throne. The devotional record admits him without comment; the institutional order does not.
⛔ Provenance, stated precisely. The disqualification is the translator's reconstruction of a rule from an outcome. The body supplies the outcome and the advisor and stops. That is a weaker citation than a body quotation, and it must be reported as one wherever this line is used — which is exactly the discipline the whole endnote layer of this edition demands, because 48 notes hedge and several supply content the text does not contain.3
⚑ And what makes it worth the care is that the rule the note reconstructs is not otherwise stated anywhere. If it is right, the volume has an eligibility criterion for kingship that appears in no chapter and only in a footnote.
One thing in ch 350 runs against the grain of everything else this volume does with a woman's stated position, and it should be recorded as an exception.
Ilā is told by Manu to "Follow me." Her whole reply:
*"I have been born as a portion of Mitra and Varuna. I will go their presence. There is no other dharma that appeals to me."*1
And she goes.
⚑ The volume's dominant pattern is that a woman's position is neither refuted nor accepted but routed around — by renaming, by non-sequitur, by deleting the ground she stood on, by purchase, by retroactive reattribution.6 Here the position is simply honoured. No argument, no transformation, no payment.
⛔ Recorded as one of three stated exceptions, and the pattern must be reported with all three: the tiger's advocate at ch 413, whose position is conceded in full against sustained argument; Devī at ch 362, who refuses Niśumbha's framing in one line and it holds; and Ilā here. ⚑ Two of the three are goddesses in a position of power. Ilā is not, which is what makes her the most interesting of the three.
Sudyumna is not the volume's only sex-transformation.
Ch 316-6(2): "as a result of his greed, a brahmana's son deceitfully pretended to be a woman. As a result of the powers of the vrata, he became a woman." Endnote 1138 supplies the names — Sumedha and Somavat — from the Skanda Purāṇa, again unmarked in the body.7
⚑ Both are Śaiva-vrata or Śaiva-curse etiologies, and neither is a simple reversal. The brahmana's son becomes a woman permanently. Sudyumna alternates monthly in one account and ends as "both" in the other. Nothing in this volume changes sex once and stops — except ch 308's closing clause, which is the only settled ending in the set.
⚑ And both figures end up in the devotee litanies, which is the pattern the whole cluster belongs to: the exemplary lives this volume tells are overwhelmingly of people its own access rules exclude, and note 1381 is the one place a text or its apparatus says so out loud.5
Worth stating, because the omissions are as consistent as the contradictions.
It does not reconcile the two accounts. It owns at least seven mechanisms for reconciling variant versions — kalpa-variance stated as a general hermeneutic, back-reference-and-decline, a doxography — and applies none of them here, or anywhere else in its dozen unnoticed doublets.8
It does not evaluate. Neither account moralises the transformation; ch 350's is explicitly a reward and ch 308's explicitly a punishment, and no third passage weighs them.
And it does not treat the state as an obstacle to devotion at any point. The disqualification, when it finally arrives, is from a kingdom — a civic office — and it arrives in a footnote.
⚑ ⛔ Reported throughout as this text's own material, in its own terms, and never as description of any historical practice, community or period. The vocabulary here is the volume's; nothing on this page is a claim about anything outside it.
You have two accounts of the same event in one body of material, and they contradict.
Collate before you decide which is right — and collate the frame, not just the facts. Agent, direction, valence, ending. ⚑ Here the two accounts differ on all four, which is a much stronger result than differing on details, and it is invisible if you only note "two versions of Sudyumna's story exist."
Check what the annotation layer does at the contradiction. This is the move that pays. An apparatus at a conflict site will do one of four things: flag it, resolve it, ignore it, or — the case here — quietly restate the other version as standard without saying it is other. The fourth is the most common and the least visible, because it reads as helpful background.
Separate what the body says from what a note supplies. The single hardest line about this figure is not in the text. It is a reconstruction, in a footnote, of a rule that appears in no chapter. Use it, cite it, and mark it — because the difference between the text says and an editor inferred is the whole difference between evidence and a reading.
And record the exception when you find one. Ilā's refusal is honoured in a corpus that almost never honours a refusal. The temptation is to leave it out because it complicates a clean pattern. A pattern with its three exceptions named is worth more than a pattern without them, and the exceptions are usually where the mechanism becomes visible.
Evidence. Both accounts read in full in place and collated row by row;1 endnotes 1379 and 1381 read at their own lines and quoted in full;24 the ch 316 second case read in place with endnote 1138;7 the devotee-litany census confirming Sudyumna's presence among the saved.5
Tension 1 — the two accounts might be reconcilable as a sequence. A reader determined to harmonise could propose that ch 350's Ilā-to-Sudyumna precedes ch 308's Sudyumna-to-Ilā. ⚑ The text does not do this, the agents differ, the valences are opposite, and ch 350's ending is "both" while ch 308's is "ceased to be a woman." The harmonisation is available and is not the text's.
Tension 2 — note 1381's rule is unattested in the body. Marked twice above. ⚑ It is the most-quoted-looking line in the material and the least securely sourced, and any vault page using it must carry the qualification.
Tension 3 — "unflagged" is a claim about the notes in range, not about all 2,124. The endnote layer was censused completely and no note connects the two accounts; that is a strong negative but it is a negative on a scripted sweep for intervention markers, and a note that merely mentioned both without flagging would not necessarily have been caught.
Open question. Ch 308's curse falls on Sudyumna "and his followers." Nothing in the volume develops the followers. Do they recur anywhere in the corpus — and if the transformation is collective rather than individual, is ch 308 an etiology for something the text no longer explains?
Against The Corpus of the Disqualified: the two pages share note 1381 and use it for different work. There it is one data point in a system that admits demons and courtesans to liberation while restricting worship rights by birth. Here it is the specific case, and the specificity matters: the disqualification is from a throne, not from the tradition, and it is supplied by an editor rather than by the text.
Against The Devotees Who Exist Only in the Endnotes: note 1379 is a fourth species for that page's catalogue. Not a circular cross-reference, not an unmarked import, not a hedge — a restatement of a contradicted version as the standard one, on the page that contradicts it. ⚑ Same apparatus, same typeface, and the reader distinguishes it from a gloss only by knowing the body.
Against Only What a Character Asks: forty-two chapters and a Saṃhitā boundary. No listener is present at both accounts, so nothing in the book can compare them — and this is one of the cases where an interlocutor would plainly have caught it, since both accounts name the same figure.
Against Closure Without Refutation: Ilā's honoured refusal is a third stated exception to that page's pattern, and the only one whose speaker is not a goddess in a position of power. ⚑ Which weakens the hypothesis that rank is the variable and strengthens the alternative: the routed-around cases are all ones in which a woman asserts something about herself, and Ilā is asserting where she is going rather than what she is.
Zero handshakes, and the reason is stated rather than filled. The obvious reaches — into psychology on identity, or into behavioural mechanics on category-assignment — would require this page to treat the volume's material as evidence about anything outside itself, and it is not. The page's whole discipline is that this is one text's two incompatible accounts of one of its own figures, plus an apparatus doing three jobs at once; a cross-domain connection here would have to smuggle in a claim the page has been careful not to make. One genuinely load-bearing handshake would be better than this note; none is available that does not breach the framing rule.
Sharpest implication. One named figure, one volume, two accounts of the same transformation that agree on nothing but that it happened. Ch 308 has him born male and cursed into womanhood by Śiva; ch 350 has him born female as Ilā and granted manhood by Mitra and Varuṇa as a reward for "humility, self-control and devotion to the truth." Opposite starting point, opposite agent, opposite valence — and opposite endings, since ch 308 closes on "eventually, he ceased to be a woman" while ch 350 ends permanently as "both." Forty-two chapters apart, in two Saṃhitās, unflagged. And the apparatus at the site does the most invisible thing available to it: *endnote 1379 supplies the ch 308 version as "the standard story", in a note attached to the chapter whose body contradicts it, without saying they conflict — and supplies a motive the body of ch 308 never gives.* Then endnote 1381 states the consequence the text only implies: "Having been a woman earlier, Sudyumna was not entitled to a kingdom." ⚑ He is on this volume's list of the saved. Qualified for liberation, disqualified from a throne — and the rule that disqualifies him appears in no chapter, only in a footnote.
Generative questions.