The gods have won a war and they are pleased with themselves. A figure appears that none of them recognises. It puts a single blade of grass on the ground and invites them to move it.
The wind god cannot move it. The fire god cannot burn it. The king of the gods raises his thunderbolt and the thunderbolt does nothing. Then the figure vanishes, and a goddess appears in the sky and explains who that was.
This volume tells that story twice, sixty-two chapters apart, in two different Saṃhitās — and a scripted sweep confirms the blade of grass appears in a test-of-the-gods context in exactly those two places and nowhere else.1
The setup matches. The test matches. The failures match. The vanishing matches. The goddess in the sky matches.
The identity of the figure does not. In one telling it is Śiva. In the other it is Devī herself.
Nothing anywhere in the volume remarks on this.
Directly, line against line:2
| ch 362-6(49) (Umā Saṃhitā) | ch 424-8.2(2) (Vāyavīya Uttara) | |
|---|---|---|
| Setup | the devas beat the dānavas and boast | the devas beat the asuras and boast |
| The visitant | "a mysterious mass of energy descended" | "Maheshvara assumed an excellent form, that of a yaksha" |
| The test | "move this blade of grass" | "He placed a blade of grass on the ground and asked each of the gods to move it" |
| Who fails | Vāyu first, then the Maruts, then all the gods | Indra's vajra, then Agni, Vāyu, Varuṇa |
| Indra | approaches — the energy vanishes instantly | asks "Who are you?" — it vanishes |
| The revelation | Umā manifests out of that same energy | Devī Haimavatī appears and explains |
| Who it was | Devī herself — "the supreme Shyama wished to destroy their pride" | Śiva — named by the frame narrator |
The episode is the Kena Upaniṣad's, where the answer is brahman. The Umā Saṃhitā gives it to the goddess and the Vāyavīya gives it to Śiva.
⛔ Unflagged. Endnotes 1488–1494 in the ch 362 range gloss mudrās, akṣauhiṇī and proper names; nothing addresses the repetition or the swap. No note attaches to ch 424 at all beyond term-glosses.
What makes this more than a variant is how little else moves.
Two tellings of one episode ordinarily differ in a dozen incidental ways — the order of the gods, the number of failures, the wording of the challenge, the setting. Here the differences are confined almost entirely to who the figure turns out to be, and to one structural consequence of that: in the Umā version the goddess who appears and the figure who vanished are the same being, emerging "out of that same energy", whereas in the Vāyavīya version they are two — Śiva vanishes and Devī arrives to explain him.
⚑ Which is exactly what a sectarian reassignment looks like from the inside. The plot does not need editing to move the identification; it only needs the last line changed and one seam re-stitched. The blade of grass, the failures and the vanishing are all indifferent to who is behind them, which is presumably why the episode has travelled so far and been claimed so often.
[n=1 OBSERVED] as a doublet; the collation is exact and both passages are quoted at length, which is what a single instance can carry.
Here is the part that is not just a doublet, and it is the sharpest thing on the page.
Ch 424's speaker names his own source, in the body, unprompted:
*"Shambhu's immeasurable deeds… are extremely wonderful. I have heard about them from the Upanishads."*3
Ch 362 contains no such statement. The same borrowed episode is narrated sixty-two chapters earlier with no attribution of any kind.
⚑⚑ So the volume tells one imported story twice — undisclosed with Devī as the answer, disclosed with Śiva as the answer.
That is not a claim about intent, and nothing here supports one. What it is, is a fact about where the acknowledgement sits: the acknowledgement travels with the Śaiva telling. A reader of the Umā Saṃhitā meets a Devī theophany with no indication it came from anywhere; a reader of the Vāyavīya meets a Śiva theophany that announces its own provenance.
⚑ And it fits a pattern the volume runs at three levels, which is the next section.
This volume carries at least three blocks of material from outside itself, and they are disclosed at three entirely different levels — one of which is nobody:4
| Import | Where | Disclosed by |
|---|---|---|
| Upaniṣadic — the Kena yakṣa (ch 424) | Vāyavīya | the speaker, in the body — "I have heard about them from the Upanishads" |
| Mahābhārata — the Pañcacūḍā dialogue (ch 338) | Umā | the translator, in an endnote — "the next chapter reproduces those sections from the Mahabharata, almost verbatim" |
| Devī Māhātmya — Suratha, Madhu-Kaiṭabha, Mahiṣa (ch 359–364) | Umā | nobody |
⚑ The one import the text discloses is the one the translator passes over. The one neither discloses is the largest — six chapters carrying a whole other work's frame, its three episodes and their order, with no note of any kind, from a translator who is himself the published translator of the Mārkaṇḍeya Purāṇa. [PLAUSIBLE — needs corroboration] on that identification; the collation has not been run.4
⚑ And the Kena episode now sits on both sides of the table at once — the same import, undisclosed at ch 362 and disclosed at ch 424, inside one volume. Which is the cleanest available demonstration that disclosure here is a property of the telling rather than of the material.
A precision worth keeping, because it changes what kind of appropriation this is.
Inside the ch 424 story, Devī never names the yakṣa. Asked "Who was this?", her whole answer is:
*"He cannot be perceived by you. He is the one who makes the wheel of samsara… move around."*3
That is very close to the Upaniṣad's own answer, and it names no sect and no deity. The Śaiva identification is made by the frame narrator, outside the episode — "Maheshvara assumed an excellent form, that of a yaksha" — before the story starts.
⚑ So the borrowed material is left almost exactly as it was, and the claim is bolted on at the join. That is a much lighter operation than rewriting, and a much more durable one: the episode continues to work as an Upaniṣadic parable for anyone reading it on its own terms, and works as a Śaiva theophany for anyone who heard the framing sentence.
⚑ And it explains why the Umā version can make the opposite claim without touching the plot either. Both Saṃhitās annex the same story by editing its wrapper.
The two tellings are not perfectly symmetrical, and the asymmetry is in the theology rather than the plot.
The Vāyavīya's framing is a straightforward identification: Śiva was the yakṣa.
Ch 362's Devī goes considerably further, and takes the entire trimūrti in the first person:
*"I am Virinchi, the creator. I am Achyuta, the preserver. I am Rudra, the destroyer.… It is evident that I am the brahman."*2
⚑ Read against the Vāyavīya's version this is the stronger claim by some distance — not the yakṣa was our god but the three gods are me, and I am what the Upaniṣad was pointing at. The Saṃhitā that does not disclose its source makes the larger claim on it.
⚑ And the same chapter has her state her own gender as fluid — "Sometimes, I assume the form of a woman. Sometimes, I assume the form of a man. Sometimes, I assume the form of both" — and offer the volume's cleanest statement of its own driven-agent doctrine as reassurance: "A magician makes a wooden puppet dance. Like that, I, Ishvari, make all beings dance."2 ⛔ Logged; the puppet line belongs to the māyā material and is not developed here.
You have found the same story, dataset, argument or diagram in two places in a body of work, attributed differently. You want to know what happened.
*Collate before you interpret, and collate on the unchanged parts.* Put the two versions in a table and fill in every row, including the boring ones — setup, sequence, who acts, what fails. ⚑ The finding is usually in how much stays fixed. Two independent retellings of a shared source drift everywhere; a deliberate reassignment is startlingly conservative, because whoever did it only had to change the parts that carry the claim.
Then locate the attribution and ask which version carries it. Not whether the work cites its sources — most bodies of work cite inconsistently — but whether the citation travels with a particular reading. If one telling names its source and the other does not, that is a fact about the two tellings, and it is available without any theory about motive.
Look for the claim at the seam rather than in the content. The cheapest possible appropriation edits the wrapper: a framing sentence, a title, a byline, an introductory clause. The body stays usable in its original sense, which is what makes the operation both easy and hard to spot — there is nothing wrong with any sentence in the story. So read the joins: the sentence before the block begins and the sentence after it ends.
And when you find two claims on one object, resist adjudicating. Both tellings here are internally coherent, neither acknowledges the other, and the volume containing both never notices. The useful output is the collation and the disclosure-asymmetry, not a verdict about which one is prior — which would need evidence the text does not contain.
Evidence. Both chapters read in full and collated row by row;2 the scripted sweep confirming the blade-of-grass motif appears in a test-of-the-gods context exactly twice in the body, at lines 733 and 1319;1 the ch 424 attribution quoted from the body;3 the ch 362 endnote range (1488–1494) read in full and confirmed to gloss content only;2 the three-import table assembled from the complete endnote census.4
Tension 1 — the two versions may descend independently from the Upaniṣad rather than one from the other. ⚑ That reading is available and this page does not exclude it. What it does not explain is the conservatism — independent descent produces drift in the incidentals, and here the incidentals hold while the identification flips. [n=1 OBSERVED], and no rigour-register language attaches.
Tension 2 — the Devī Māhātmya identification is unconfirmed and the page leans on it once. Strike it from the three-import table entirely and the table still runs at two levels — body-disclosed and endnote-disclosed — which is enough for the asymmetry.
Tension 3 — the frame-narrator point cuts both ways. If the takeover is a wrapper edit, then so is the Umā version's, and neither Saṃhitā has done more work than the other. The page says so rather than treating the Śaiva version as the intruder.
Open question. The Kena's own answer is brahman, and ch 362's Devī says "I am the brahman" while ch 424's Devī says only that he "cannot be perceived by you." Which of the two is closer to the Upaniṣad's wording? A collation against the Kena would settle whether one telling preserves the source more faithfully while the other claims it more strongly — and that is a source-map task, not adjudicated here.
Against The Saṃhitā That Grew: the ch 362 telling sits in the Umā Saṃhitā, which runs forty-six per cent over the shloka budget the text publishes for itself and which independently carries every other import in the volume. ⚑ A second Kena telling in the over-budget Saṃhitā is exactly what that page predicts — and it arrived after the prediction, which is the only reason it counts for anything.
Against The Kailāsa Is a Different Book: the contrast in kind of divergence is instructive. That page's two Saṃhitās disagree about technical doctrine — twenty-six tattvas against fifty — while running the same rite. Here two Saṃhitās disagree about who a story is about while running the same plot. Doctrinal independence and narrative annexation are different phenomena and the volume exhibits both, in the same pair of sections.
Against Only What a Character Asks: sixty-two chapters and a Saṃhitā boundary, so no listener is present at both tellings, so nothing in the book can notice. ⚑ This is the single cleanest instance of that page's mechanism, because the two passages are near-identical in everything except the one thing that would have to be compared.
Against The Vocative That Does Not Fit: notably, no vocative leaks here. The Kena episode is imported cleanly, with the claim made outside it — which is what a wrapper edit produces, as against the paste edits that leave the host's addresses inside the guest's speech. Two different import operations, two different signatures, and only one of them is detectable by that instrument.
The Founder Problem in Historical Perspective — that page's territory is how traditions acquire origins retroactively: a movement needing a beginning and finding one, with the attribution doing work the underlying material never asked to do.
The two Kena tellings supply an unusually clean specimen, because the same episode is claimed twice, by two parties, in one book, and the underlying story is untouched by either claim.
What that isolates is a distinction the founder-problem literature usually cannot make. Retroactive attribution is normally studied where the material and the claim are entangled — a saying, a teaching, a law that has been shaped by whoever now owns it, so that recovering the pre-attribution version is guesswork. Here the attribution is demonstrably separable. Delete the framing sentence from ch 424 and you have an Upaniṣadic parable; delete the closing identification from ch 362 and you have the same parable again. Both claims are external to the thing claimed, and you can see it because there are two of them.
⚑ The generalisable move: when a tradition annexes existing material, look for whether the claim is load-bearing inside the material or bolted to its edges. Edge-attachment is cheaper, more common, and far more durable — the material keeps working for everyone, so nobody has an incentive to challenge the wrapper, and over time the wrapper is what gets transmitted. Content-level rewriting is rarer and leaves scars that a later reader can find.
And back the other way: the founder-problem page supplies the reason this is worth more than a curiosity. A tradition does not need to have originated anything to have a founder-story about it — and a corpus that annexes the same parable twice, for two different deities, sixty-two chapters apart, is showing you the mechanism operating on a scale small enough to be watched.
Sharpest implication. The Kena Upaniṣad's yakṣa episode — the unrecognised figure, the blade of grass, the wind that cannot move it and the thunderbolt that does nothing — is told twice in this volume, sixty-two chapters apart, in two different Saṃhitās. The setup, the test, the failures, the vanishing and the goddess in the sky all match; a scripted sweep confirms the motif appears in exactly those two places. What differs is who the figure turns out to be: Devī herself in the Umā Saṃhitā, Śiva in the Vāyavīya — where the Upaniṣad's own answer is brahman. Unflagged; the notes gloss mudrās and proper names and say nothing. And the disclosure is the finding: ch 424's speaker names his source in the body — "I have heard about them from the Upanishads" — and ch 362 says nothing at all. So the same import appears at two disclosure levels inside one volume, undisclosed with the goddess and disclosed with the god, in a corpus whose three known imports are disclosed by the speaker, by the translator, and by nobody, respectively. ⚑ And the annexation is done at the seam, not in the story: inside the episode Devī never names the yakṣa — "He cannot be perceived by you" — while the sectarian identification is made by the frame narrator before the story begins. The borrowed material is left as it was and the claim is bolted to its edge, which is why both Saṃhitās could take it without either having to touch a line.
Generative questions.