Three times in this volume, a speaker stops what he is saying and quotes an old story. The sentence he uses to do it has an unusual shape: it names the story's two characters and does not name anyone who told it.
In this connection, an ancient history is recounted. This is about a conversation between X and Y.
Recounted by whom? The sentence declines to say. And each time it appears, the story that follows takes someone with real standing — the god of death's own officers, a room of senior mourners, a king — and shows them being corrected by someone with less.
That pairing is the finding. The volume has an attributed way of citing old stories and uses it elsewhere. Where it wants to demote an authority, it uses the version with no author.
The device appears at 6(1), 7(2) and 7(13).
At 6(1), Śuka: "In this connection, an ancient account is cited, about a conversation between the messengers of Vishnu and Yama. Listen to me."1
At 7(2), Hiraṇyakaśipu: "In this connection, an ancient history is recounted. This is about a conversation between Yama and the relatives of a dead person. Listen to it."2
At 7(13), Nārada: "In this connection, an ancient history is recounted. This is about a conversation between Prahlada and the sage Ajagara."3
Two components define the form. The citation verb is an agentless passive — is cited, is recounted — with no subject anywhere in the sentence. And the content is specified as a conversation between two named parties. The story is fully identified; its transmission is not identified at all.
The phrase "In this connection" occurs six times across the six Skandhas of this volume.4 Three carry both components. The other three are examined below, and one of them is the important one.
⚑ Correction to this build's PRD. The PRD described this device as "used only for that," implying the formula itself is unique to these three passages. It is not — the introductory phrase appears six times. What is exactly three is the conjunction of agentless passive and named dialogue pair. The claim survives in the sharper form and the loose form is withdrawn.
The fourth instance sits in the same neighbourhood as two of the three, and it does the opposite of all of them.
At 7(1), Śuka says: "In this connection, the divine rishi happily recited an ancient account, when he was asked by Ajatashatru at the great sacrifice."5
Same opening phrase. Same object — an ancient account. Everything else is supplied. There is a named reciter (the divine ṛṣi, Nārada), a named interrogator (Ajātaśatru, Yudhiṣṭhira), a named occasion (the great royal sacrifice), and even a named witnessed event that prompted the question — the lord of Chedi merging into Vāsudeva's body in front of the assembly. The text then adds that Nārada "was seated there, at the sacrifice," placing an eyewitness in the room.
This is maximum attribution. Every credential the other three withhold, this one volunteers.
And the material it introduces does not demote anybody. Yudhiṣṭhira asks how Śiśupāla, who hated Kṛṣṇa, could merge into him; Nārada answers with the doorkeepers' curse. The answer is startling in content but it supports the frame — it explains God's apparent partiality rather than overturning anyone who asserted it.
So within a single Skandha the volume runs both versions. Attributed citation, supporting material. Unattributed citation, material that takes someone down. The device is not a stylistic tic. It is selected.
At 6(1), the losers are the officers of the god of dharma. Parīkṣit has just told Śuka that atonement looks pointless — "like the washing of an elephant." Śuka answers before the story: ritual atonement works only for the already-knowing, devotion outperforms it, and a man who has withdrawn from Nārāyaṇa cannot be purified by prāyaścitta any more than rivers can clean a pitcher that held liquor.
Then he cites the ancient account. Inside it, Yama's messengers deliver a long and technically flawless exposition — dharma is what the Vedas state, the witnesses to action are enumerated, the liṅga-śarīra's sixteen parts are laid out, the guṇas compel action in the unliberated. It is correct, and it is the position of the officials whose entire remit is dharma. The chapter ends on their conclusion: "Since he has not atoned, he has to be purified through punishment."1
The reader has already been given the refutation. The rebuttal is delivered before the argument it rebuts, so the impeccable speech arrives pre-demoted. And the correcting party, Viṣṇu's messengers, "laughed" before replying.
At 7(2), the losers are the adults. Hiraṇyakaśipu is consoling Diti and the widowed daughters-in-law. He cites the ancient history, and in it Yama arrives "assuming the form of a boy" and opens with an assessment of the room: "These people are advanced in age. Despite having seen the ways of the world, they are confused." He goes on to address them as "O foolish people!" and "O ignorant ones!"2
The god of death could have appeared as himself. The text makes him a child specifically so that the correction runs uphill against age and status. And the story nests one level deeper still — inside Yama's speech, the male kuliṅga bird's lament is quoted directly, giving three levels of embedded speech before the moral lands.
At 7(13), the loser is a king. Nārada cites the conversation between Prahlāda and the sage Ajagara. Prahlāda arrives with advisers; he is the volume's supreme devotee and lord of the asuras. The man he questions is lying on bare ground, "covered with dust all over his body," and — the text is emphatic — "through his deeds, form, words or signs, people were unable to understand his varna or ashrama, who he was and who he was not."3
The sage answers, and closes by conceding that his teaching "differs from the sacred texts of the worlds."
This is the extreme case. An unverifiable teaching, contrary to scripture, delivered by a man with no determinable caste or life-stage, cited from a source that is not named. Every layer at which a claim could be checked has been removed, and the party being corrected is the highest-status listener in the volume.
And the corrector is always the lower-ranked party. Set the three side by side and the regularity is exact.
| Passage | Demoted | Corrector | Corrector's standing |
|---|---|---|---|
| 6(1) | Yama's messengers | Viṣṇu's messengers | servants; they laugh first |
| 7(2) | senior grieving relatives | Yama as a boy | a child |
| 7(13) | Prahlāda, king of asuras | the sage Ajagara | no determinable varṇa or āśrama |
In none of the three does the correction come from above. A senior sage does not descend to settle the matter. The instrument is consistently someone the demoted party would outrank in any ordinary encounter — and the volume arranges that deliberately, most visibly at 7(2), where a god adopts a child's body in order to do it.
The unattributed citation and the low-ranked corrector are the same move made twice. Both strip away the grounds on which the correction could be resisted. You cannot dismiss the source, because there is no source. You cannot appeal to rank, because the speaker has already conceded he has none.
An attributed citation is attackable through its attribution. If Nārada said it at Yudhiṣṭhira's sacrifice, the claim inherits Nārada's standing — which is high, but it is also a surface. A listener who wanted to resist could ask what Nārada knew, or why he was believed.
The agentless passive removes that surface. An ancient history is recounted offers nothing to push against. It is old, it is transmitted, and there is no one in the chain to interrogate.
This is worth stating precisely, because it inverts the intuitive account of how citation works. Naming a source usually strengthens a claim. Here the volume names a source when the material is uncontroversial and withholds one when the material has to defeat someone — which means, in this text, attribution is a liability whenever the claim's job is to overturn a credentialed party. Give the opponent a name and you have given them somewhere to aim.
The one loose end is 7(13)'s annotation, and it cuts against the device. Debroy's endnote identifies the unidentifiable sage as Dattātreya, deduced from the First Skandha.6 The passage's whole discipline is that the man cannot be placed; the note places him. An apparatus that resolves an ambiguity the text was maintaining does not merely add information — it removes the mechanism.
The remaining two occurrences of "In this connection" have one component each, and neither introduces a dialogue.
At 5(6): "In this connection, it has been said, 'The mind is fickle and one should never contract a friendship with it…'"4 Agentless, but what follows is a maxim, not a conversation. Nobody is being corrected; a proposition is being reinforced.
At 8(19): "In this connection, there is a chant from the sacred texts. Hear it from me."4 The source is named as a category — the sacred texts — and again the content is not a dialogue.
Both matter to the finding as controls. They show that the agentless construction alone is not the device, and that citing an old thing alone is not the device. Only the conjunction is, and only the conjunction demotes.
Read the frame before the quoted material, and note which came first. At 6(1) the refutation precedes the argument. A quoted position that arrives after its rebuttal is not being presented for evaluation; it is being exhibited. Reading the speech on its own merits misses that it has already been answered.
Check whether a text has more than one citation formula. The finding here was unavailable from any single passage. It required collecting every instance of one phrase across the corpus and noticing that they were not all the same shape. Where two formulae coexist, the choice between them carries information.
Treat agentless passives in citation as a marked choice, not as style. It is said, is recounted, has been reported — where a text elsewhere shows it can name a transmitter, the places it declines to are worth listing.
Watch what the annotation does to an unplaceable figure. Where the text withholds an identification on purpose, a note supplying one is a substantive intervention. Record the text's version and the note's version separately rather than merging them.
The citation names the characters and not the teller. Full specificity about content, zero about transmission, in the same sentence.
The corrector outranks nobody. Servants, children, men without caste. Where the instrument of correction is consistently low-status, the demotion is designed rather than incidental.
A fully attributed version of the same formula exists nearby. This is what converts a stylistic observation into a finding. Without 7(1), the three unattributed citations could be the text's only way of doing this.
The rebuttal precedes the argument. Where a frame answers a position before quoting it, the quotation is exhibition, not argument.
The demoted party's speech is technically correct. At 6(1) the Yamadūtas are not wrong about the Vedas. Demotion here operates on standing, not on accuracy — which is a different thing from refutation and should be recorded as such.
Strongest support. All six occurrences of the introductory phrase were collected by pattern search across the body text of Skandhas 5–10 and are individually quoted or characterised above. The three-instance claim is about a conjunction of two features, both of which are visible in the quoted sentences. The near-miss at 7(1) is quoted in full.
Tension the page preserves. Whether the volume intends the demotion or has inherited a formulaic convention is not decidable from inside the text. Purāṇic Sanskrit has a stock citation verse for exactly this move, and its recurrence may be transmission rather than design. The page's answer is that the 7(1) contrast makes design more likely, since a stock formula would not be swapped for an attributed one precisely where the material stops demoting anyone. That is an argument, not a demonstration.
Boundedness to the English. The agentless passive is a feature of Debroy's rendering. The underlying construction is impersonal in the source, but how impersonal, and whether 7(1)'s attribution is equally explicit in the Sanskrit, cannot be settled from this translation. The distribution is real; the grammatical characterisation is bounded to this English text.
Absence-claim discipline. The device appears exactly three times is a claim about a search on the string In this connection. A citation introduced by different words — there is an old story, the ancients tell — would not have been caught, and no exhaustive sweep for paraphrases was run.
Structural finding at n=3, with a control. Three instances plus one contrasting near-miss plus two partial cousins. That is enough to state the pattern and not enough to call it a rule of the genre. The register here is held to what four data points support.
Open question. Does Volume 1 contain the same split — agentless citation for demotions, attributed citation for supporting material? The concurrent Vol. 1 ingest could settle it with one string search, and the answer would move this from a volume-level observation to a corpus-level one.
This page sits against two things the build has already established, and complicates one of them.
The volume gives impeccable doctrine to characters whose conduct it contradicts — Hiraṇyakaśipu at 7(2), Kaṁsa at 10(4). The 7(2) instance is now doubly strange, because Hiraṇyakaśipu is not merely delivering correct doctrine; he is deploying a rhetorical device correctly, citing an old story in the precise form the volume reserves for humbling the self-important, in the same chapter in which he has ordered the earth burned. The device works. He uses it well. Diti is consoled and gives up sorrowing. Nothing about his own conduct changes.
The build has also found that this text's narrators are stacked and that the stacking is load-bearing. What this page adds is that the stack is not only depth, it is a gradient of accountability. Each embedding step removes a party who could be asked to answer for the claim, and the volume drops to the least accountable level exactly when the claim is least defensible from authority.
The convergence with the rest of the build is the usual one, and it is getting hard to call coincidence: the volume executes a distinction with total consistency and never articulates it. Three demotions, one shared grammatical shape, one contrasting control in the same Skandha, and no character or annotation anywhere that remarks on it.
Authority Construction and the Architecture of Belief — attribution as a liability, not only an asset.
Greene's charlatan builds a source. The procedure is investment in visible credentials — mystique, apparatus, a cause to rally to — on the premise that belief attaches to whoever most convincingly occupies the authority position. Attribution is the product.
The Bhāgavata runs the inverse and gets the same result. Where its claim has to defeat a credentialed party, it removes every credential from its own side: no named transmitter, and a speaker with no determinable caste, or a child, or a laughing servant. The claim wins by being unattackable rather than by being impressive.
The insight neither produces alone: source-naming is a trade, not a good. Greene's account works from a zero baseline, where the audience has no committed position and credentials are pure gain. The Purāṇa supplies the other case — an audience with a standing authority already in the room — and there, naming your source hands the opponent something to aim at, so the optimal attribution is none. Read together, the two say that the value of citing a source is a function of what the audience is already committed to, which neither the influence literature's zero-baseline model nor the text's silent practice states on its own.
Sharpest implication. This volume has two ways to cite an old story. The attributed one names a reciter, an interrogator and an occasion, and introduces material that overturns nobody. The unattributed one names only the two characters, and every time it appears, someone with standing is about to be corrected by someone without any.
Generative questions.