One of the great goddess-names of the tradition comes from a mother trying to stop her daughter doing something.
Umā. The volume's translator breaks it down in a note: U + ma, meaning "Oh! Don't!"1
That's the etymology. A two-syllable interruption, caught mid-air and turned into a name. The girl is about to go and hurt herself with austerities and her mother says oh, don't — and the tradition keeps the sound.
It's a lovely piece of naming and the volume tells it twice.
Not twice as in retold. Twice as in two different occasions, fifteen chapters apart, with the mother saying it at two different points in her daughter's life, and no acknowledgement anywhere that the story has already been told.
Chapter 102. Pārvatī is a small child. The chapter is a childhood montage — she grows like the moon, plays with balls on the sandy bed of the Ganges, gets a tutor, recovers the learning of her previous life. Ordinary, affectionate, domestic.
Dropped into the middle of it:
"Her mother tried to restrain Kalika from performing austerities, exclaiming, 'Uma!' Hence, subsequently, the one with the beautiful face came to be known on earth as Uma."2
So the naming happens in childhood, in a chapter about childhood, as one detail among several. The child is already trying to do austerities and her mother already wants her to stop.
The word "hence" is doing real work. This is presented as the origin. Full stop, name acquired, story moves on.
Chapter 117. Pārvatī is grown. Nārada has given her the mantra. She's sending her friends to ask her parents for permission to go to the forest.
Her father agrees. Her mother doesn't. Menā argues at length — the body is delicate, austerities are hard, do it at home, we've never heard of women going to hermitages. Pārvatī persists. Menā is miserable and eventually relents.
And then:
"Since Mena restrained her from going to the forest and performing austerities, as a result of that reason, Shivaa obtained the name Uma then."3
As a result of that reason. Then. The same construction, the same causal claim, the same hence — attached to a different scene, a different age, and a different argument.
Put them together.
Chapter 102: a child, an unspecified moment, a single exclamation, name acquired. Chapter 117: an adult, a specific negotiation about the forest, a sustained argument, name acquired.
Both are stated as origins. Neither refers to the other. Nobody says and this is why she had already been called Umā, or the name she had borne since childhood now proved apt, which is the obvious repair and would take one clause.
The two accounts aren't compatible in any straightforward way. In chapter 102 she has the name from childhood, so by chapter 117 she has borne it for the entire intervening narrative — through the palm-reading, the service on the mountain, the burning of Kāma. Chapter 117 then says she obtained it then.
This is a doublet: the same story told twice from different sources or at different stages of composition, both preserved, neither reconciled.
The interesting part is what the apparatus does with it, which is: nothing.
Endnote 926 gives the etymology — U + ma, "Oh! Don't!" — and it attaches to chapter 102, the childhood version.1 It explains the wordplay competently.
It says nothing about chapter 117. It does not cross-reference. It does not flag an inconsistency.
That's worth sitting with, because this translator flags inconsistencies constantly and is not shy about it. Across this Khaṇḍa's endnote range he supplies a missing negative, declares that sentences don't belong where they are, announces "we have deviated from the text," and pre-emptively defends the text against a seam-reading nobody made.4 The apparatus is active and interventionist.
And this one goes past unmarked.
This is why the doublet matters beyond itself, and it needs stating carefully.
The build has been running a seam-detection lens for many clusters — tracking frame violations, vocative slips, numerical discrepancies, repeated blocks. The lens has a known and repeatedly-stated limitation: it has never found a violation the translator did not flag first. Five frame violations, all five caught by following his notes. That makes the lens a good reading of his attention and an unproven instrument in its own right.
Two unflagged numerical failures were found unaided in the Satī material — a gaṇa death-count that halves, and Vīrabhadra's arms halving between chapters.5 Those established something narrow and useful: his flagged arithmetic failures are not the complete set.
The Umā doublet is a third find of that kind, and it's a different species. Not arithmetic. A repeated aetiology — the text explaining the origin of the same name twice, incompatibly, in the same Khaṇḍa.
What this establishes is the same shape as the arithmetic finds and no more: the unflagged set includes narrative doublets, not only numerical slips. It does not establish that the lens works. It does not establish how many more there are. [n=1 OBSERVED] — one doublet. Reported, not named.
You want to find seams in a translated text without simply following the translator's footnotes.
Track aetiologies specifically. Origin stories for names, places, rites and features are the highest-yield category, because they're self-contained, they use flagging language (hence, therefore, as a result), and a text that acquires two of them for one item can't easily hide it. Keep a list as you read: item, occasion, chapter.
Watch for the causal connective and treat it as a marker. Hence she came to be known as is a claim of origin, not a mention of a name. Both Umā passages carry it. A second occurrence of that construction for the same item is the whole find — you don't need to interpret anything.
Then check whether the apparatus caught it, and record the answer either way. This is the step that converts a textual observation into a finding about method. If the translator flagged it, you've learned about his attention. If he didn't, you've learned something about the limits of using him as your instrument — and that's the more valuable result, because it's the one that can't be got by reading his notes.
Then be strict about what one instance can support. A doublet is a doublet. It is not evidence of a compilation history, not evidence about which version is older, not evidence about the number of hands involved. Those all require a sample this doesn't have.
Say the limits out loud, because this is exactly the kind of find that gets overstated.
What it supports. The volume gives two incompatible origins for one name, fifteen chapters apart, both marked as origins, neither acknowledging the other, and the translator does not flag it.
What it does not support. Any claim about composition history. Doublets arise from combined recensions, but they also arise from oral performance, from scribal duplication, from a compiler restating a detail for readers who joined late, and from an author simply not remembering. Nothing here distinguishes those.
What would sharpen it. A third aetiology for the same name, or a second doublet of the same species elsewhere in the volume. Either would move this off [n=1 OBSERVED]. Chapters 121–150 are unread and both are live there.
One thing worth noticing but not building on. Both versions place the naming in the same conflict — mother versus daughter, over austerity. Whatever else is going on, the volume is consistent about what the interruption was about, and inconsistent only about when. That could mean the doublet is two tellings of one remembered thing. It could mean the conflict is so central it attracts the name twice. The evidence doesn't choose.
⚑ ENRICHED 2026-08-11 — Volume 2 shows the text handling this exact problem correctly, twice, which makes the silence here worse.
This page's finding is a doublet with no acknowledgement: two etymologies for one name, fifteen chapters apart, neither referring to the other. Volume 2 supplies the control the page lacked — the corpus owns a device built for precisely this situation, and deploys it.
Chapter 163, on Gaṇeśa's birth: "Because of differences in kalpas, the story of Ganesha's birth is spoken about in different ways. I will now speak about Ganesha's origin during the kalpa known as Shveta."C1
Chapter 298, on Gautama's two curses — and this one is raised by the listeners themselves: "These differences occur because incidents have occurred in different kalpas… All of you have lovingly heard the earlier account too."C2
That is a working doctrinal solvent for incompatible versions, stated in plain terms, twice.
And it is not the only one. Volume 2 also runs back-reference-and-decline at ch 239 — "I have told you everything about his conduct when I spoke of Sati's conduct. Since you have heard it many times, I will not speak about in great detail" — and announced re-telling at ch 286 and ch 217.C3
⚑ Three mechanisms, five deployments, and none of them applied to the Umā doublet. Which materially strengthens this page's claim without adding a single new instance to it: the silence is not the silence of a tradition that had no way of speaking. It is the silence of one that had the sentence ready and did not use it.
The full four-doublet inventory and the disclosure toolkit are carried by The Four Doublets and the Four Disclosures, where the Umā case is named as the Volume 1 precedent for the species.
Zero handshakes, stated rather than invented. This is a philological find about one text's internal consistency. The candidates that surfaced — memory-reconstruction research, oral-formulaic transmission — would each have required importing a framework the source supplies no purchase for, and the connection would have been the vault's, not the material's.
Strongest support. Both passages are explicit aetiologies using causal connectives, in the same Khaṇḍa, fifteen chapters apart.23 Endnote 926 glosses the etymology and attaches to only one of them.1 The translator's demonstrated willingness to flag inconsistencies elsewhere in this same endnote range makes the silence here meaningful rather than merely absent.4
Tension, preserved. The volume asserts two incompatible occasions for one naming and reconciles neither. This page does not reconcile them either.
Open. Is there a third aetiology for Umā anywhere in ch 121–150? Unread; nothing here claims what they contain.
Open. Are there other doublets of this species in the volume? One instance is what the range read supports.
Open. Does the same name-conflict — mother against daughter over austerity — attach to any of Pārvatī's other names? Aparṇā comes from her giving up leaves,6 and Gaurī from the outcome of the austerities; neither involves Menā.
The page this most directly speaks to is The Vocative That Does Not Fit, and the relationship is corrective rather than confirming.
That page carries the build's most instructive failure: a "compound signature" written up on two instances in the language of rigour, cited forward by two other pages, and then collapsing to two out of five as more evidence arrived. The obituary is retained there deliberately. It's the reason every structural finding in this build now carries a sample-size tag.
This doublet is exactly the kind of find that could repeat that mistake. It's satisfying, it's unflagged, it feels like it proves something about the method, and it is one instance. So: [n=1 OBSERVED], no name, no -lens suffix, no claim that it validates anything.
What it does do — and this is the part worth keeping — is add a second species to the small set of things found without the translator's help. The two unflagged numerical discrepancies from the Satī material were arithmetic. This is narrative. Together they say the unflagged set isn't confined to one kind of error, which is a modest claim and a real one.
And read against Six Disclosed Repairs, the contrast does work: that page tracks what the translator tells you he changed. This one marks a place where he had every opportunity and said nothing. The two together describe the shape of his attention from both sides.
Sharpest implication. The volume explains the origin of one of its central names twice, incompatibly, fifteen chapters apart, both times with a causal connective marking it as the origin — and the translator, who flags inconsistencies constantly across this same endnote range, walks past it. The find is small and its value is what it says about the instrument: the unflagged set includes narrative doublets, not only arithmetic.
Generative questions.
Both versions site the naming in the same conflict — mother against daughter, over austerity. Is the doublet two tellings of one thing, or a conflict so central it attracts an aetiology twice?
The translator's attention is demonstrably sharp on numbers, frames and vocatives, and silent here. Is aetiology simply outside what he's watching for, or is a doublet less visible than a slip because both halves read correctly on their own?
If chapter 102 gives her the name in childhood, she carries it through the palm-reading, the mountain service and the burning of Kāma. Does any passage between 102 and 117 use it — and would that make chapter 117's claim harder or easier to explain?