Every account of an extraordinary person is produced by people who already know who they are dealing with.
That is not a criticism; it is a structural fact. The people close enough to see anything are the people who came close because they believed something, and by the time they are in a position to observe, the observation has a frame around it. Everybody in the room is an insider, because being in the room is what insiders do.
Which makes one kind of witness worth more than all the others: somebody who stumbled in, saw the thing, and had no idea what they were looking at.
Antya-līlā 18 contains one, and he is the most valuable person in the book.1
A fisherman is working the sea at night. His net snags something heavy, and he pulls up what he takes for a corpse.
He touches it.
And catches ecstatic love off it — which he interprets as demonic possession.2
"As soon as I touched it, a ghost entered my heart."
He then describes the body he found, and this is where the passage becomes evidence rather than anecdote. Five to seven cubits long. Limbs three cubits. Its joints all separated beneath the skin, which is completely slack. Eyes open. Uttering the sound goṅ-goṅ.3
That is a clinical match for the Antya 14 description, produced by a man who was not present at Antya 14, does not belong to the community, and is at that moment trying to warn the devotees away from a corpse he thinks is haunted.
And then the detail that settles what kind of witness he is:
*"I have seen the Lord many times, but this is not He. This body is very deformed."*4
He knows Caitanya by sight. He is looking directly at him. And he does not recognise him.
That sentence is worth holding still for, because of the direction it runs in.
The obvious devotional use of a non-recognition is the opposite of this one. The corpus has a doctrine ready-made for people who fail to identify who they are dealing with — the concealed descent, the owls that cannot see the sun, non-recognition as exactly what a hidden divinity predicts.16 Every other instance in the book is absorbed that way.
Here the non-recognition is not about the fisherman's spiritual capacity at all. He gives his reason and it is physical: this body is very deformed. He is not failing to perceive a concealed glory. He is correctly reporting that the object in front of him does not look like the man he has seen many times, which is a statement about the body and not about him.
And that is precisely why it functions as evidence. A witness whose failure has a physical explanation is describing the physical thing — and the corpus, which had a doctrine available that would have converted the moment into further proof of concealment, does not reach for it here.
Take the evidential question seriously in both directions, because this is the strongest passage in the corpus for anybody trying to assess it.
What it gets you. An independent description, from outside the devotional community, of a physical presentation reported three chapters earlier — matching on the specific and unusual features rather than on the impressive ones. Separated joints under slack skin. Eyes open. The cubit measurements. And the goṅ-goṅ, which independently matches the Gambhīrā account in another chapter entirely.5
Nobody constructing corroboration would build it out of a man who fails to recognise the subject and calls the state demonic.
What it does not get you. The fisherman's testimony reaches us through the same channel as everything else — Svarūpa Dāmodara's notebook, Raghunātha's memory, Kṛṣṇadāsa's composition eighty years on.6 An independent witness reported by a dependent chain is not an independent source, and the corpus's own account of its transmission makes that unavoidable.
So the honest position is narrow and worth having. If the chain is reporting faithfully, this is genuine outside corroboration of a recurring physical presentation. If the chain is not, then the corpus has invented a witness who does not recognise its protagonist and reads its central evidence as demonic — which is a strange thing to invent, and the strangeness is the whole of the argument in its favour.
Then the part that is not about evidence at all, and is the best piece of practical intelligence in the chapter.
The fisherman is terrified. He is possessed, on his own account, and he has a specific and desperate worry: "if I die, who will take care of my wife and children?"7
And he reports a detail that a person making this up would not think of. The Nṛsiṁha mantra normally protects him from ghosts — he has the standard remedy, he knows how to use it, and it is failing:
*"This ghost overcomes me with redoubled strength when I chant the Nṛsiṁha mantra."*8
Now watch what Svarūpa Dāmodara does.
He does not explain. He does not correct the man's cosmology, does not tell him that what he touched was ecstatic love, does not identify the body, and does not offer the true account — all of which were available and all of which would have been useless.
He says: "I am a famous exorcist."9
Then he chants, slaps the fisherman three times, and pronounces the ghost gone. And the man is fine.
The move deserves a name because it generalises and almost nobody does it.
The fisherman's problem is not a false belief. It is terror inside a frame, and the frame is doing real work for him — it explains what happened, predicts what comes next, and specifies a remedy. It is a functioning model that has produced one alarming anomaly.
A true explanation from outside that frame is not an improvement on it. It is a demolition of it, offered to a frightened man at night, with nothing to replace it that he has any reason to accept. He would have been left with no account at all, which is worse than a wrong one.
Svarūpa Dāmodara supplies the exit the frame already contains. Possession has a remedy, the remedy is an exorcist, and an exorcist is a role that can be occupied by anybody the frame accepts. The three slaps and the chanting are not deception in any sense the fisherman would object to — they are the correct treatment for what he believes he has, administered by somebody who does not believe he has it.
And the corpus reports it without a flicker of discomfort, in a text that elsewhere adjudicates truthfulness at enormous length.10 Its own criterion — direction of benefit — clears this instantly, which is exactly the property that lets the criterion clear every other untruth in the book.
This is the chapter where the corpus's central evidential problem becomes fully visible, because all three readings are now on the record.
Divine ecstasy — the devotees, throughout, using the technical vocabulary: sāttvika-vikāra, the eight transformations, ardha-bāhya.11
Demonic possession — the fisherman, here, who touched it and caught it.
Epilepsy — named by Caitanya himself, to Pāṭhān soldiers, when a naturalistic account was what would save four men's lives.12
Three vocabularies. One body. No adjudication anywhere in the book.
And the corpus is not concealing this; it is the corpus that supplies all three. A text arranging its evidence would have dropped the fisherman's reading or corrected it on the page, and this one records it, lets it stand, and moves on.
Which is the strongest reason to take the corpus seriously as a source and the strongest reason not to take its verdict. A document that preserves the readings that damage it has demonstrated that its reports are not being filtered for effect — and has thereby also demonstrated that the reports do not settle what they are reports of.
Layer A supplies the netting, the touch, the possession, the physical description, the failure of the Nṛsiṁha mantra, the non-recognition, and the exorcism.13
Layer B expands the episode into the potency of contact with a pure devotee and the transmission of ecstatic love by touch.14
Worth marking the difference, because Layer B's reading is coherent and it costs the passage its best feature. On the commentary's account the fisherman caught real ecstatic love and misdescribed it, which converts him from an independent observer into a beneficiary — a man who received something and lacked the vocabulary to name it.
Layer A leaves him an observer. He describes a body accurately, names the state wrongly by the community's lights, and goes home. The corroboration is only worth something on the second reading, and only the second reading is what the verse actually gives.
You want to know whether an account is being filtered. Look for the witness who was not told. Every insider report carries its frame with it; the value of an outsider is not that they are more honest but that their errors are uncorrelated with the community's.
Somebody is frightened inside a frame you think is wrong. Supply the exit their frame already contains, and argue later or never. A true explanation offered to a frightened person is a demolition with nothing to move into, and the remedy their own model specifies will work because their model is what is generating the fear.
You are assessing corroboration. Ask whether the independent witness reached you independently. An outside observer reported by an inside chain is worth much less than it feels like, and it is one of the easiest errors to make with historical testimony because the outsider's voice is so distinctive on the page.
You have a source that preserves damaging material. Rate its reports up and its verdicts down. The two move in opposite directions and it is worth keeping them apart: the fisherman's presence means the descriptions were not curated; it does not mean the community's reading of them was right.
The evidence is a continuous scene with an outside observer's physical description, an explicit alternative diagnosis, a failed standard remedy, a non-recognition, and a staged exorcism.15
The tension the source doesn't close: the same body is divine ecstasy, demonic possession and epilepsy, on the corpus's own pages, and nothing adjudicates. All three framings are supplied by the text and it does not choose between them or acknowledge that it has offered three.
A second one. The corroboration is genuinely independent in content and entirely dependent in transmission. The corpus cannot supply a witness who reaches the reader by any route except its own, and this is the closest it gets.
Open: whether the fisherman's account was recorded because it corroborated or because it happened. The corpus gives no sign of understanding the passage as evidence, which is mildly in favour of the second — but a text that did understand it would look the same.
This page and The Body That Changes Shape are one presentation described from inside and outside. That page has the household's measurements — cubits, six inches, temperature; this one has a stranger's, taken at night, from a man trying to warn people off a corpse — and the two match on the features neither had reason to invent.
Against The Epilepsy Cover Story it is the third framing of the same phenomenon and the mirror of the first. There the subject supplies a naturalistic account to outsiders in order to save lives; here an outsider supplies a supernatural one and the insiders play along in order to calm him — and the corpus approves of both without noticing it has described one technique.
And it belongs beside The Servants of the Household, which holds the corpus's other untruths. A servant lying to donors, two staff running a cover story about lunch, and now a secretary claiming to be a famous exorcist — all cleared by the same criterion, and the criterion has no content clause at all.
Aram: Masquerade-Possession and Mental Illness — that page's mechanism is a diagnostic one. In Igbo practice, masquerade-possession is normally contained: the ritual opens, the spirit inhabits the performer, the performance closes, the spirit releases. Aram is the name for the failure of that release — the same possession, outside the contained architecture, not letting go. It is classified as a subtype of Ara, mental illness, and the community recognises the signs.
Set that beside the fisherman and something neither page reaches becomes available.
The Igbo framework has two categories where this corpus has one. Contained possession is the practice working; uncontained possession is illness, with a name and a recognised presentation. The distinguishing variable is not the phenomenology — it is whether there is a ritual boundary the state is inside of.
The Caitanya corpus has no such boundary and therefore no such distinction. There is no ritual that opens and closes the state. It arrives when it arrives, in a bedroom or on a beach at night, and it is divya-unmāda on every occasion because the category is unconditioned by containment.
Which gives the finding. A tradition that ritually contains a possession state acquires, for free, the vocabulary to say that a particular instance has gone wrong — the failure of release is diagnosable precisely because release was supposed to happen. A tradition with an uncontained state has no such term available, and the corpus's six chapters confirm it: nobody ever says this has gone on too long, because there is nothing it is going on too long relative to.
And the fisherman is the accident that makes the comparison possible. He arrives holding a possession framework and applies it correctly by his own lights — a spirit has entered, the standard mantra has failed, an exorcist is needed. Two traditions with possession vocabulary, one body, and the only one with a release condition is the one the community rejected.
Sharpest implication. The most valuable witness to any extraordinary person is the one who had not been told, and their value is that their errors are uncorrelated with the community's — not that they are more honest. And a tradition that ritually contains an altered state gets a term for this one has gone wrong for free; a tradition whose version is uncontained cannot say it at all, which is why nobody in six chapters ever does.
Generative questions.