Ask this volume where the hells are and it tells you twice. Once: below the seventh nether region, twenty-eight layers of five. Eight chapters later: "The hells are above the nether regions", and the layers are gone — a flat list of names.
Ask it what a headless shadow means. Death within a month. Three chapters later, in the same unbroken conversation with the same listener: loss of speech within six.
Ask it how many tattvas the Virajā oblation purifies, and here the pairing gets sharp — because both answers use the same rite and the same spoken sentence, and one enumerates twenty-six ending at Puruṣa while the other enumerates about fifty and keeps going.
None of these is a slip. A slip is a wrong number in a list. These are two complete, internally consistent technical systems occupying the same book and doing the same job, and the text never once puts them in the same room.
That distinction — coherent disagreement rather than scatter — is what makes the finding worth a page. It is not evidence that somebody miscounted. It is evidence about how the book was made.
Ch 322-6(8) locates them by digging down:
*"Under the earth, there is the seventh nether region of Tala, the region below it being enveloped in darkness. There are twenty-eight layers of hell there."*1
A layered grid follows — Ghora, Sughora, Atighora, Mahāghora, Ghorarūpa, Talatala and on — with a total asserted at four thousand.
Ch 330-6(16) opens by locating them the other way:
"O best among sages! The hells are above the nether regions. Sinners are cooked there."2
And what follows is not a grid. It is a flat roll of about twenty-three names — Raurava, Śūkara, Rodha, Tāla, Vivasana, Mahājvāla, Taptakumbha, Lavaṇa, Vilohita, Vaitaraṇī, Pūyavaha, Krimiṇa, Krimibhojana, Asipatravana, Lālābhakṣa, Sandaṃśa, Kālasūtra, Tamas, Avīcirodhana, Śvabhojana, Ruṣṭa, Mahāraurava, Śālmali — no layers, no multiplier, and a sin-to-hell mapping that is not ch 324's.2
⚑ Below versus above is not a detail. It is a flat contradiction about the geography of the cosmos, stated in the opening clause of each passage, eight chapters apart, in one Saṃhitā. Unflagged.
⛔ And a smaller one sits between apparatus and body in the same range: note 1195 lists the seven nether regions with Nitala and glosses "By Tala, one obviously means Patala", while ch 329's body lists them without Nitala and with Tala and Pātāla as two separate regions.3 [n=1 OBSERVED] — recorded because it is the translator's gloss being contradicted by the text seven chapters on, not a textual variant.
The second pairing is the tightest, because there is no frame change at all to explain it.
Ch 339-6(25), Śiva to Pārvatī:
*"If a person sees that his own shadow is without a head, or if he cannot see his shadow, he will not live for more than a month."*4
Ch 342-6(28), Śiva to Pārvatī, three chapters later, same continuous dialogue:
*"If he sees a shadow that is without a head, within a period of six months, the yogi will lose all power of speech."*5
⚑ Same omen, same speaker, same listener, no interruption — and the prognosis differs in both duration and outcome.
⚑ And the two are different in kind, not merely in detail. Ch 339's material is a general mortality calendar read off the body. Ch 342's is a colour-and-missing-limb code: white shadow = dharma, black = sin, red = bondage, yellow = enmity; no arms = death of relatives, no hips = death of the wife, no calves = loss of wealth, no feet = exile.5 A second divinatory apparatus, grafted on exactly as ch 330's list was grafted onto ch 322's grid.
⛔ Unflagged. Notes 1300–1310 cover ch 339's omens and none addresses it; nothing attaches at ch 342.
The third pairing is the decisive one, and it is decisive because the two passages share a verbatim ritual formula. Whatever else is going on, they are describing the same rite.
Kailāsa Saṃhitā, ch 378-7(13):
*"After this, the Viraja oblations will be offered. This must be done so that the twenty-six tattvas in the body merge and become purified. He will state, 'Let the tattvas in my body become purified.'… He will mention the tattvas in the due order, starting with earth and ending with Purusha."*6
Vāyavīya Saṃhitā, ch 420-8.1(33):
*"He will chant, 'Let the tattvas in my body be purified.' These tattvas are the five elements, the five tanmatras, the five senses of perception, the five organs of action, the seven dhatus, the five breaths of life, prana and the others, the mind, the intellect, ahamkara, fame, the gunas, Prakriti, Purusha, attachment, Vidya, Kala, Niyati, Kaala, Maya, pure Vidya, Maheshvara, Sadashiva, Shakti and Shiva's tattva."*7
Counted: thirty-two in the numbered groups plus eighteen singles — about fifty. And the list does not stop at Puruṣa. It runs on through rāga, Vidyā, Kalā, Niyati, Kāla, Māyā, śuddhavidyā, Maheśvara, Sadāśiva, Śakti and Śiva-tattva.7
⚑⚑ That is not two counts of one thing. It is two metaphysical systems. The Kailāsa's twenty-six ending at Puruṣa is the Sāṃkhya scheme. The Vāyavīya's fifty continuing into the pure tattvas is the Śaiva Siddhānta hierarchy — whose vocabulary the same Saṃhitā has already established elsewhere, paśu/pāśa/pati, the five kañcukas, mala.8
Same rite. Same spoken sentence. Two ontologies. Unflagged in both.
⛔ This is what takes the species from [n=2 SUGGESTIVE] to [n≥3 PATTERN], and it is the cleanest of the three precisely because the shared formula removes the escape route. Two hell-systems could be two traditions about a place. Two shadow-codes could be two divinatory manuals. Two enumerations under one spoken formula in one rite cannot be either.
The Praṇava's component count divides the same way, and a sweep shows each Saṃhitā says its number twice.
| Saṃhitā | Where | What it says |
|---|---|---|
| Kailāsa | ch 368-7(3), line 803 | "Pranava is said to have five components—the three matras, bindu and nada" |
| Kailāsa | ch 371-7(6), line 821 | "With bindu and sound as the fifth, Pranava has five components" |
| Vāyavīya | ch 428-8.2(6), line 1339 | "The Upanishads speak of four matras—'A', 'U', 'M' and nada" |
| Vāyavīya | ch 457-8.2(35), line 1551 | "The akshara then divided itself into four parts—the three matras… and the subsequent half-matra" |
9
⚑⚑ This is the finding, stated as sharply as the evidence allows: each Saṃhitā is internally consistent and they disagree with each other. Corruption scatters. Two coherent positions, each repeated, is the signature of independent composition — and it converts a structural inference about this book into something demonstrable.
⚑ And note the asymmetry in how each grounds itself. Ch 428 attributes its four to "the Upanishads"; the Kailāsa — the Saṃhitā that cites named external authorities constantly — simply asserts five, twice.
⭐ The strongest single item here was found by chasing a line reference, not by an instrument, and it is the one place where the divergence and the doublet are visible in the same sentence.
Ch 374-7(9), line 849 (Kailāsa) and ch 419-8.1(32), line 1261 (Vāyavīya) are near-verbatim. A character-level comparison finds about 2,900 of roughly 3,200 characters shared in runs of sixty or more — a little over ninety per cent.15 Whole stretches match exactly, including the eight-name derivation of Śiva, "Depending on the condition, Sadashiva assumes these five names", and the closing "There is no atman who is superior to him. Therefore, Paramatman is Shiva himself."
And inside that shared passage, one number differs:
| Kailāsa, ch 374-7(9) | Vāyavīya, ch 419-8.1(32) | |
|---|---|---|
| The clause | "Even when they possess senses to accomplish the five kinds of jnana" | "Even when they possess senses to accomplish the ten kinds of jnana" |
| The endnote | 1620 — "Through the five senses." | 1893 — "Through the five senses of perception and the five organs of action." |
⚑⚑ The apparatus annotates both, consistently with each local number, and never notices they are the same passage. The proof that it is the same passage is in the notes themselves: 1619 and 1892 are word-identical — "Vaidya (physician) of samsara" — 273 notes apart.15
⛔ This complicates the independence reading rather than confirming it, and the complication is the honest part. Two Saṃhitās that never agree are one thing; two Saṃhitās that share a paragraph and disagree inside it are another. Whatever the relationship is, it is not pure independence — there is contact, and the divergence sits in the middle of the contact.
⚑ What the case does establish beyond argument is the signature. Five against ten, in identical surrounding prose, glossed twice by an editor who annotated the same words twice without recognising them — that is not scatter, and it is not corruption of a single exemplar. It is two versions of one passage, both preserved.
The volume is not short of machinery for handling exactly this.
Ch 400-8.1(13) states a general hermeneutic for why accounts disagree — "in different kalpas, Brahma, Vishnu and Maheshvara are born from each other… Based on what happened in one particular kalpa, maharshis speak about their relative power" — which would licence every one of these pairings in a sentence.10 The same Saṃhitā uses back-reference-and-decline correctly several times. Ch 363-6(50) explicitly harmonises three names of one goddess: *"Shatakshi, Shakambhari and Durga are the same. All three are identical."*11
⚑ The apparatus for saying these are two views of one thing exists, is understood, and is in competent use. It is applied to a goddess's three names and never to the two hells.
⛔ Which is the reading this page rests on, and it is worth stating as a reading rather than a fact: a text this comfortable with announced multiplicity is probably not tolerating these pairings — it is not seeing them. They are blind spots, not licensed variation. That distinction is inferential and it is marked as such.
You have found two versions of something in one body of material and want to know what you are looking at.
First, check whether each version is internally consistent. ⚑ This is the whole diagnostic and it is cheap to run. Scatter — a number that wanders, a name spelled four ways — is corruption or carelessness. Two positions, each stated more than once, each coherent, is two sources. The Praṇava case gives four data points and resolves in one pass.
Then look for a shared invariant that neither side would have invented separately. ⚑ The Virajā passages share a spoken sentence verbatim. That is what rules out two different rites and forces the conclusion. Without an invariant you have two accounts; with one you have two treatments of a single thing.
Then ask what reconciliation machinery the source has, and whether it was used here. ⛔ If a document has a standard way of saying these two accounts are both true under different conditions, and it uses that machinery elsewhere and not here, the silence is data. It tells you the discrepancy was not noticed rather than not minded.
And check the apparatus separately from the text. ⚑ Two of this corpus's tattva counts come from the translator, not the book — one in an endnote, one in the Introduction, and the Introduction's number appears nowhere in the body.12 An editor's count sits in the same typeface as the author's and is not marked as different, which means an inventory of "what this text says" will silently include things it does not say.
Evidence. Every passage read in full in place on the source file: ch 322-6(8) line 331 and ch 330-6(16) line 379 for the hells;12 ch 339-6(25) line 467 and ch 342-6(28) line 499 for the shadows;45 ch 378-7(13) line 891 and ch 420-8.1(33) line 1271 for the Virajā oblation;67 ch 368-7(3) line 803, ch 371-7(6) line 821, ch 428-8.2(6) line 1339 and ch 457-8.2(35) line 1551 for the Praṇava.9
⛔ Cited in the Saṃhitā-internal form throughout — the running chapter numbers in this edition are not unique, and the fact is stated nowhere in it.13
Tension 1 — the four pairings are not equally strong and are not presented as though they were. ⚑ The Virajā and Praṇava pairings are strong (a shared formula; four data points, two on each side). The hells and the shadows are single-instance oppositions and would each be [n=1 OBSERVED] alone. The [n≥3 PATTERN] is on the species — incompatible technical systems coexisting — not on any one of them.
Tension 2 — one nearby pairing may be principled rather than contradictory. Ch 374-7(9) gives "beyond the twenty-three tattvas… Purusha is spoken of as the twenty-fifth" — 23 + Prakṛti + Puruṣa = 25.14 ⛔ 25 is classical Sāṃkhya and 26 is Seśvara-Sāṃkhya, so the 25/26 pair may be a real distinction rather than a discrepancy. It is deliberately not folded into the Virajā finding, and is recorded here so nobody folds it in later.
Tension 3 — the ch 374 / ch 419 doublet cuts against the independence reading. ⚑ Four pairings say the two Saṃhitās never agreed; the fifth says they share a paragraph. ⛔ Both are on this page and neither is subordinated to the other. The most this build will say is that the relationship involves contact and that the divergence sits inside the contact — which is a different and harder claim than "two independent compositions", and the evidence to settle it is not here. [n=1 OBSERVED] on the doublet.
Tension 4 — "not noticed rather than not minded" is an inference. ⚑ It rests on the reconciliation machinery being demonstrably available and demonstrably unused. That is good evidence and it is not proof, and an alternative reading — that a compiler saw the divergences and preserved both deliberately, without comment, as received material — fits everything on this page equally well.
Open question. The Praṇava count divides cleanly on the Saṃhitā boundary, two against two. Does the tattva count divide the same way when every statement in the volume is collated — or is the Kailāsa/Vāyavīya line only one of several seams?
Against The Kailāsa Is a Different Book: this page supplies that one's hardest evidence. ⚑ The Praṇava pairing is not an argument that the Kailāsa differs in flavour; it is two coherent positions each stated twice, on a technical question, split exactly on the Saṃhitā boundary. That is the difference between a stylistic impression and a demonstration.
Against The Saṃhitā That Grew: the two pages describe different compositional phenomena and should not be merged. ⚑ That page's Umā runs far over the shloka budget the text publishes for itself and is where every imported frame was found — expansion by absorption. The Kailāsa and Vāyavīya land close to budget and simply never agreed — compositional independence. Two mechanisms, two bodies of evidence.
Against Only What a Character Asks: the mechanism that explains the silence. ⚑ This text detects an inconsistency only when a named listener raises one, and no listener is present at a technical enumeration. Nobody asks Sanatkumāra why the hells moved, so nothing fires — which is why the reconciliation machinery can be fluent and unused in the same volume.
Against Named Circularity, Unnoticed Contradiction: the same asymmetry from the other side. That page holds that the text names and argues from circular relations it constructs deliberately, while walking past contradictions it did not construct. ⚑ These four pairings are the largest single deposit of the second kind.
Against The Apparatus Is the Translator's: two of the corpus's five tattva counts are his — 31 in an endnote and 36 in the Introduction, the latter unattested anywhere in the body.12 ⚑ So a reader counting this text's positions on tattvas from the printed page will get five and the text supplies three.
Wang Mang and the Interpolated Texts — that page holds an emperor inserting passages into a revered text because the author was five centuries dead and nobody could check.
Put the two side by side and what falls out is a detectability rule, which neither case yields alone.
Wang Mang's interpolations worked because there was nothing to contradict them. The inserted passage had one job — to sound like Confucius — and no rival system inside the same corpus was making a competing technical claim on the same point. There is no internal signature to find, which is why that case turns on external evidence and motive rather than on the text.
This volume's material is the opposite configuration and is therefore legible. Nobody was hiding anything, and the seam shows anyway — because two bodies of material each carried a complete, self-consistent technical position, and complete positions collide where isolated insertions do not. The Praṇava case is the cleanest demonstration available: four statements, two on each side, each side repeating itself, splitting on a boundary that was drawn for other reasons.
⚑ Which yields the rule in a form that travels. What makes composite material detectable is not that it was inserted badly — it is that the inserted material had to commit to a system. A forger writing one sentence leaves no signature. A tradition contributing a ritual manual brings its whole ontology with it, and the ontology is what shows. The bigger and more competent the contribution, the more detectable it is — which inverts the usual intuition that large interpolations are harder to hide than small ones only because of their length.
Running back the other way, the Purāṇic case supplies the control the historical one lacks. Wang Mang's insertions are undetectable partly because we cannot see the alternative; here the alternative is printed forty chapters away in the same book. ⛔ A corpus that contains its own rival version is a corpus where source criticism can be run at all — and this volume, by preserving both hell-systems instead of choosing, made itself readable in a way a tidier compiler would have prevented.
Sharpest implication. "Under the earth, there is the seventh nether region of Tala, the region below it being enveloped in darkness. There are twenty-eight layers of hell there." Eight chapters later: *"The hells are above the nether regions."* A headless shadow means death within a month at ch 339 and loss of speech within six at ch 342 — same speaker, same listener, same unbroken conversation. And the Virajā oblation is performed in two Saṃhitās with the same spoken sentence — "Let the tattvas in my body be purified" — purifying twenty-six tattvas "ending with Purusha" in one and about fifty that run straight past him into Māyā, śuddhavidyā, Sadāśiva and Śiva-tattva in the other. ⚑⚑ Same rite, same words, Sāṃkhya on one side and Śaiva Siddhānta on the other, unflagged in both. Then the Praṇava settles what kind of thing this is: five components in the Kailāsa, stated twice; four in the Vāyavīya, stated twice, one of them citing "the Upanishads" for its number. Each side is internally consistent and they disagree with each other — and corruption does not produce that, corruption produces scatter. ⭐ And the sharpest case turned up by accident: ch 374 and ch 419 are ninety per cent the same paragraph, word for word across the Saṃhitā boundary, and inside it one says "the five kinds of jnana" and the other "the ten" — each glossed by the translator to fit its own local number, in notes whose neighbours are word-identical, 273 apart. He annotated the same passage twice without recognising it. Meanwhile the volume owns a doctrine that would licence all of it — "based on what happened in one particular kalpa, maharshis speak about their relative power" — and spends it instead on harmonising three names of one goddess. It can say these are two views of one thing fluently, and it never says it about anything it did not already know was one thing.
Generative questions.