There is a thing almost nobody does in public, and the reason is obvious once you have watched somebody try.
They state, accurately and in order, the reasons their claim is weak — and then make the claim.
It is not the same as hedging. Hedging softens the claim. It is not the same as caveating, which fences the claim in. This is a full accounting of the evidential position, followed by an assertion that the accounting does not support, and it looks so much like self-defeat that people avoid it instinctively.
At Antya-līlā 14, at the point where his material is least believable, Kṛṣṇadāsa Kavirāja stops the narrative and does exactly that.1
*"No one has witnessed such bodily changes elsewhere, nor has anyone read of them in the revealed scriptures… These ecstasies are not described in the śāstras, and they are inconceivable to common men. Therefore people in general do not believe in them. Raghunātha dāsa Gosvāmī lived continuously with Śrī Caitanya Mahāprabhu. I am simply recording whatever I have heard from him. Although common men do not believe in these pastimes, I believe in them totally."*2
Four statements and a conclusion, and it is worth separating them because each is a concession.
One: no corroborating witness elsewhere. Nothing like this is reported outside his material.
Two: no scriptural support. In a tradition whose entire method of authorisation is scriptural precedent, he concedes that the precedent does not exist.3
Three: the claim is generally disbelieved. He states the reception accurately rather than pleading it away.
Four: a single named eyewitness. Not many devotees saw — one man, named, whose continuous presence is the whole of the chain.
And then: I believe in them totally.
Take those four as an evidential position and the concessions are more complete than most critics would extract.
He has surrendered independent attestation, scriptural warrant, general plausibility and multiple witnesses — which is essentially the entire list of grounds available in his intellectual environment for asserting anything.
And he names the eyewitness. That is the most exposing thing on the list. A vague appeal to the devotees is unfalsifiable and comfortable; naming Raghunātha dāsa places the whole weight on one identifiable person and invites exactly the question that follows.
Nobody does this accidentally. A man who wanted the material believed and had not thought about the problem writes it down and moves on. The passage exists because he stopped, and stopping is the choice.
And there is one further concession buried in the wording that is easy to pass over.
I am simply recording whatever I have heard from him.
He is not claiming to have heard it from Caitanya. The sentence names his source as Raghunātha, and Raghunātha's qualification is stated as continuous presence — not as having witnessed these particular episodes, and not as having been told them by anybody in particular.
Which leaves the actual chain one link longer than the sentence appears to concede. Somebody saw the body at the Siṁha-dvāra; that account reached Svārūpa Dāmodara's notebook; the notebook reached Raghunātha; Raghunātha reached Kṛṣṇadāsa; and the book is eighty years after the events.13
He does not spell that out and he does not obscure it either. The corpus supplies every link elsewhere, in the ordinary course of describing itself, and this passage names the one closest to him.
Which is the pattern worth extracting. A source that concedes the link nearest to itself has done the hardest part — the near link is the one they could most easily have left vague, and the distant links are the ones a reader can reconstruct without help.
The natural reading is that he is overriding his own evidence, and that is not quite what the sentence does.
I believe in them totally is a report about himself. It is not an argument that the reader should believe, and it does not claim that the four concessions have been answered.
Which relocates the whole passage. He is not saying the evidence is sufficient. He is saying that his position does not rest on the evidence, and — crucially — he is telling the reader that, so that the reader is not left thinking it does.
That is a specific and unusual honesty and it should be credited at its actual size. It does not make the claims more likely. What it does is remove the reader's risk of being misled about the grounds — and being misled about grounds is the thing partisan documents normally do.
A text that overstates its evidence is dangerous. A text that states its evidence accurately and then asserts on faith is merely one you can decide about. The second is a much better object to hold, and it is rarer.
The placement is not incidental and it repays attention.
It appears at Antya 14 — the elongated body, nine feet with the joints six inches apart under the skin, no breath at the nostrils.4 The most physically impossible material in the corpus.
Not at the identity claim. The assertion that this man is Kṛṣṇa in Rādhā's sentiment is the book's central and least testable proposition, and it receives no such passage. It is argued for at length, from scripture, with the full apparatus.5
The self-criticism appears where the scriptural apparatus fails. He can argue the identity claim because precedent exists; he cannot argue the bodily transformations because he has just conceded that none does — and the concession arrives precisely at the point where his ordinary method has run out.
Which is what an honest author looks like from outside: the epistemics get discussed exactly where they become a problem, and nowhere else.
And it is also what a well-managed text looks like. Disclosing the weakest link buys credibility for everything around it, and the page cannot distinguish the two readings — though it can note that the passage costs him the corpus's most spectacular material and gains him nothing a reader was demanding.
The passage changes what a reader can do with the whole book, in both directions.
It licenses scepticism about the transformations specifically. The author has said the material has no independent attestation and one witness. A reader declining to believe it is agreeing with him about the evidence and differing only about faith.
And it raises the value of everything else. A text that concedes its weakest point has demonstrated that its reports are not being uniformly managed — which is the same property the fisherman's demonic reading and the epilepsy sentence supply, from the other direction.6
Those three together are the corpus's real epistemic asset. An outside witness whose reading damages the claim, a naturalistic explanation in the protagonist's own mouth, and an author stating that his central marvel has no support. None of them had to be there.
What none of them does is make the marvels more likely. The corpus's candour is about its chain of custody and the chain is still what it is — one notebook, one memoriser, one author, eighty years.7
And candour and accuracy are independent properties, which is the thing this passage makes unusually easy to see.
A source can be scrupulously honest about how it knows what it knows and still be reporting something that did not happen — because honesty operates on the account of the evidence and not on the events. Kṛṣṇadāsa has told the truth about his position with more precision than most historians manage, and every word of that could hold while the elongated body did not.
Which is worth stating because the two get conflated constantly in both directions. Readers who find the candour impressive raise their estimate of the marvels; readers who find the marvels impossible lower their estimate of the candour. Both moves are errors, and the passage is the cleanest available demonstration that the properties come apart.
Layer A supplies the entire passage — the concessions and the assertion, in the author's own voice.8
Layer B expands the surrounding material into the doctrine of the sāttvika transformations and the inconceivable nature of the Lord's body.9
Worth marking what the commentary does with the self-criticism, which is not much. Layer B has no use for a passage conceding that the śāstras are silent, because its whole method is to establish continuity with the śāstras.
So the corpus's most sceptical moment belongs to the seventeenth century, and a reader working from the purports will not meet it — which is the same finding this build has recorded on the "plea of being attacked by Muslims" correction, and on the measurements at Antya 14.10 Three times now, the sharpest epistemic material in this corpus is Layer A's and the commentary passes it.
You are making a claim your evidence does not support. State the evidential position in full and then say you believe it anyway. It costs the claim nothing it had and it removes the reader's risk of being misled about grounds — which is the harm that matters and the one overstatement causes.
You are assessing a partisan source. Look for where it discusses its own epistemics. An honest text raises the question exactly where its method fails and nowhere else, and the location of the discussion is more informative than its content.
You are tempted to appeal to unnamed witnesses. Naming one is more exposing and worth more. Many devotees saw it cannot be checked or challenged; Raghunātha dāsa lived continuously with him invites the obvious question and answers it in advance.
You have found a source that concedes its weakest point. Raise your estimate of its other reports and not of the conceded one. The two move independently, and conflating them is the standard error in both directions.
The evidence is a single first-person passage in which the author suspends the narrative to state four evidential concessions and one assertion of belief.11
The tension the source doesn't close: the corpus disclaims its own authorship and this passage is unmistakably authorial. I am simply recording whatever I have heard and my body is like a wooden doll are the same man in the same book, and the first is a claim about his method that the second denies he has.12
A second one. The self-criticism appears at the physically impossible material and not at the identity claim, which is the less testable of the two. The epistemics are discussed where the scriptural method fails rather than where the claim is largest.
Open: whether the passage is candour or management. Disclosing the weakest link buys credibility for everything around it, and no observation available from the text distinguishes an honest author from a well-advised one.
This page and The Disclaimed Author are the same man contradicting himself about what he is doing. There he is a wooden doll and the Deity gave the order; here he is a recorder with a named source, stating his own evidentiary position — and the second is the sentence of somebody who knows he is an author.
Against Ardha-bāhya it is the chain of custody named from the top. That page reconstructs the constraint from the corpus's own three-states doctrine; this has the author simply stating the chain — one man, continuously present, and everything downstream from him.
And it belongs beside The Fisherman and the Ghost. Both are the corpus preserving material that damages it, and between them they establish the habit as a habit rather than an accident — an outside witness who reads the state as demonic, and an author who says his central marvel has no support.
Faxian: Witness vs Hearsay — that page's mechanism is a single sentence doing all the work. Faxian describes a five-storey rock monastery in extraordinary detail and then marks the boundary: this part he saw, that part he was told. The discipline is not scepticism about the marvellous; it is the maintenance of a visible line between the two categories, sustained across a long text.
Kṛṣṇadāsa is doing the same operation from the other side, and the pairing shows what each method can and cannot buy.
Faxian's line is drawn per-item, in passing, as he goes. It is cheap to maintain, it never interrupts, and it lets a reader grade every claim individually. The cost is that it says nothing about what he believes — the reader gets categories and no commitment.
Kṛṣṇadāsa draws the line once, expensively, with a stated position on both sides. He tells the reader the evidence is insufficient and that he accepts it, which Faxian never does about anything.
Which gives the finding neither page reaches. Marking the boundary and declaring your position across it are separable, and almost every source does at most one. The per-item marker is more useful and less honest — it declines to say what the writer thinks. The stated commitment is more honest and less useful — it tells you where the author stands and leaves you to grade the material yourself.
A source doing both would be the ideal and this vault has not found one. Faxian marks and does not commit; Kṛṣṇadāsa commits and marks once, at the single worst point, which is where the marking was cheapest to make and most necessary.
Sharpest implication. Marking what you witnessed and declaring what you believe are separable operations, and nearly every source performs at most one — the per-item marker is more useful and quietly withholds the writer's own position, while the stated commitment is more honest and leaves the reader to grade everything alone. And the location of an author's epistemic discussion is more informative than its content: an honest one raises the question exactly where their method fails.
Generative questions.