⚑ Title note: this build's PRD called it five flagged contradictions. A recount against the endnote layer gives four, plus one interpretive-difficulty note the PRD had miscounted as a fifth. The slug is kept for continuity; the count on this page is the corrected one.
A translator working on a composite text has three options when the text contradicts itself: smooth it, ignore it, or say so.
Debroy says so — four times, in the endnotes, in his own voice, about the Bhāgavata Purāṇa's internal consistency.
And once he does something else entirely. He denies a contradiction that nobody has raised.
One. "There is a contradiction with what has been said a little earlier."1 Flat, unelaborated, no attempt at resolution.
Two. "Therefore, he did not remember Krishna because he was scared. That is yet another apparent contradiction."2 The phrasing is the interesting part — yet another implies a running tally he is keeping and not publishing.
Three. "Dantavakra was Shishupala's cousin. He wasn't present at the royal sacrifice and was killed by Krishna later. Therefore, there is an inconsistency in the way this is stated."3 Here he supplies the evidence for the inconsistency rather than merely asserting it — an external fact about the narrative that the passage's own claim cannot accommodate.
Four. "Vishnu. There is a contradiction with what has been said about Kirtiman."4 The clearest cross-Skandha instance: a detail in the Tenth contradicting one established earlier.
Four flags across roughly 1,500 notes. Not a systematic apparatus, and not nothing.
The PRD for this build listed a fifth — a note on Hiraṇyakaśipu and which worlds survive a kalpa, quoted as "this contradicts what he has himself stated earlier."
The actual note does not say that. It says:
"This causes problems of interpretations that are not easily solvable. Hiranyakashipu seems to think that positions occupied by devotees of Vishnu (Vaishnavas) and Vishnu's world are destroyed at the end of a kalpa. He also seems to think that Brahma's position..."5
That is a translator flagging an interpretive difficulty in his own work — not easily solvable, seems to think — rather than flagging a self-contradiction in a character. Different category, different speaker, different implication.
The correction matters beyond arithmetic. A text that flags five internal contradictions is behaving one way; a translator who flags four contradictions and separately admits he cannot resolve a passage is doing two different things, and collapsing them loses the second.
The strangest note in the volume is this one:
"There is no contradiction. Brahma was about to speak, but allowed Vindhyavali to speak first."6
Nobody has raised a contradiction. The note appears at 8(22), where Vindhyāvalī speaks and then Brahmā speaks and Viṣṇu's reply is addressed to Brahmā — and Debroy has anticipated that a reader might notice the sequence looks odd, and has answered it before anyone could.
Three things follow.
He was reading for seams. The note is only produced by someone who has noticed the passage's structure and predicted a reader's objection to it.
The denial concedes that the seam is visible. A passage nobody would question does not attract a defence.
And this vault has now seen the same behaviour twice. The Śiva Purāṇa Vol. 1 build logged an identical move — a pre-emptive denial of a contradiction nobody raised. Two Debroy volumes, two unprompted defences. [n=2 SUGGESTIVE] across the translator's corpus, not across the Purāṇa.
The decisive test would be a third instance in another Debroy volume, or a case where the anticipated objection is one the vault can independently show a reader making.
Worth noting precisely, because the passage it protects is one this build has already flagged.
At 8(22) Vindhyāvalī makes the only argument in Skandha 8 that attacks the transaction's premise — nobody owns anything, so nothing was given, so there is no debt. Brahmā then speaks on entirely different grounds, and Viṣṇu's reply opens "O Brahma!"
So the strongest objection in the chapter is made, positioned immediately before the doctrine, and routed around. Debroy's note asserts that the sequencing is courtesy — Brahmā yielding the floor — rather than a structural gap.
That may well be right. It is also the reading that makes the seam disappear, produced by a translator whose declared method is to choose the simplest interpretation that fits, and offered where no one has asked.
The contradiction flags are one part of a wider disclosure layer, and the volume's other admissions are worth listing because they establish that the flagging is not exceptional behaviour.
Supplied material. "In the text, these sentences have the subject/subjects missing. We have added them, so that the meaning is clear."7 Subjects invented for clarity, disclosed.
Supplied punctuation-level attribution. "Though not explicitly stated, this seems to be the words of the gopas and the gopis. It has thus been placed within quotes."8 Speaker attribution assigned by quotation marks, disclosed.
Admitted difficulty. Four notes stating that shlokas have many complicated interpretations and that the simplest has been taken.
Acknowledged incompleteness in the source. "As will be apparent, this is not a complete listing of all of Vasudeva's wives."9
And one substantive reversal against consensus — the Devāpi note, which flips who is at fault in a twelve-year drought.
A translator who supplies subjects, assigns speakers, admits four hard passages, notes a defective list and reverses one consensus reading is doing considerably more than rendering. The value of the volume is that all of it is on the record.
Read the apparatus as evidence about the text and about the translator, and keep the two separate. Four contradiction flags are evidence about the Purāṇa's composite construction. The denial and the difficulty notes are evidence about Debroy. Conflating them — as this build's own PRD did — inflates the first at the expense of the second.
Treat an unprompted defence as a marker. The most useful note in the volume is the one answering a question nobody asked, because it identifies a passage the translator himself found structurally uncomfortable. Anywhere in a scholarly apparatus, a defence with no attacker is a map to the seam.
Count, do not accept counts. The PRD's figure was wrong in a specific and instructive way: a note that felt like a contradiction flag was recorded as one. Any enumeration inherited from a prior pass should be recounted against the source before a page is built on it.
And note what disclosure buys. None of these admissions makes the translation less reliable. All of them make it more usable, because a reader can now bound their claims. A rendering with no such notes is not more accurate; it is only less legible.
The flag is unelaborated. There is a contradiction with what has been said a little earlier. No resolution offered. That is a translator declining to smooth, and it is the honest form.
The phrasing implies a tally. Yet another apparent contradiction. Where an annotator counts without publishing the count, more exist than are marked.
A defence appears with no attack. The single strongest signal in an apparatus.
Supplied material is disclosed at the punctuation level. Quotation marks assigned to an unattributed speech is an editorial act, and a translator who flags it is flagging the small ones too.
Strongest support. All four flags, the denial, and the five other disclosed interventions are quoted verbatim from the endnote layer, and the miscounted fifth is quoted alongside the PRD's version of it so the correction is checkable.
Tension the page preserves. The denial may simply be correct — Brahmā may indeed have yielded the floor — and it is also the reading that removes a seam this build independently flagged from the body text. Both stand.
Absence-claim discipline. Four flags is what a search of the endnote layers for Skandhas 5–10 returns. The endnote layer is ~1,500 notes and was searched by pattern rather than read entire; the claim is four found, not four exist.
Open question. Yet another apparent contradiction implies a running count. If Debroy is tracking them and publishing only some, the selection criterion is unknown — and the four published are all of the kind a careful reader would hit anyway, which may be the criterion.
The vault's Śiva Purāṇa build already logged one instance of the pre-emptive-denial behaviour in Debroy Vol. 1, and this is the second across his corpus.
That makes it a finding about a translator, not about the Purāṇas, and the vault should file it accordingly. Debroy has now twice anticipated a seam-reading and answered it unasked, in two different texts, in two different decades of composite material.
The convergence is worth holding lightly at [n=2 SUGGESTIVE]. Two instances in a corpus of thousands of notes is a thin sample, and the alternative explanation — that both passages genuinely have an innocent reading and he supplied it — is not excluded by anything here.
What is not thin is the disclosure practice as a whole. Across two volumes this vault now has a translator who flags contradictions, admits hard passages, discloses supplied subjects, discloses assigned quotation marks, notes defective source lists, and records deviations from consensus. Whatever else is true of the rendering, the vault can bound its claims against it — which is more than it can do for most of the translated primary texts it holds.
With Barnum and Planned Partial Disclosure (business). The reversal that page documents is a showman who published his own tricks and profited by it: where every other case says conceal the mechanism, disclosure increased the audience's willingness to be worked on, because being told the mechanism feels like being on the inside of it.
A scholarly apparatus is the same structure with the incentives inverted, and the comparison sharpens both. Debroy discloses supplied subjects, assigned quotation marks, four hard passages and one consensus reversal — and the disclosure has the identical effect on the reader's disposition. Trust goes up.
What neither page states alone: disclosure produces the same trust movement whether or not the disclosing party benefits from it, which means trust-gained-by-disclosure is not evidence that the disclosure was against interest. Barnum's audience and Debroy's reader both feel let in. The difference is not in the mechanism or in the reader's response — it is entirely in whether the undisclosed remainder matters, and the reader is in no position to assess that either way. [n=1 OBSERVED]
Sharpest implication. Four flagged contradictions across roughly 1,500 notes, one phrased to imply an unpublished tally, one interpretive-difficulty note this build's own PRD had miscounted as a fifth — and one defence of a contradiction nobody raised, protecting the exact passage where the chapter's strongest objection is routed around.
Generative questions.