You ask somebody a question and they do not reply.
And you learn something from that. Not nothing — something specific, and usually something you are quite confident about. They are protecting someone. They are embarrassed. They have been told not to say. The silence gets read, and the reading is almost always about the relationship between the person and the thing you asked about.
That inference is unavoidable and it is also unfalsifiable, because a silence is compatible with every explanation including the boring ones. You cannot be argued out of a reading of a silence — there is nothing in it to contradict you.
Antya-līlā 15 contains the purest instance of that operation this vault holds, and it is performed on plants.1
Caitanya, in the gopīs' mood, recites their questioning of the flora — and he is doing it in an actual garden by the sea at Purī, which he has taken for Vṛndāvana.2
The trees are asked where Kṛṣṇa has gone. They give no answer.
And the conclusion drawn is not that trees do not speak:
"Since all of these trees belong to the male class, all of them must be friends of Kṛṣṇa" — and therefore they will not betray him.3
Then the creepers are asked, on the reasoning that being feminine, "they are like friends to us."
They stay silent too. And the conclusion adapts:
*"These plants are all Kṛṣṇa's maidservants, and out of fear they will not speak to us."*4
Set the two inferences side by side, because the structure is the entire point.
Silence from the masculine trees → they are his friends and are covering for him. Silence from the feminine creepers → they are his maidservants and are afraid to speak.
Two opposite social premises. Identical evidence. And both conclusions arrive at the same place: Kṛṣṇa is being protected.
That is not sloppy reasoning; it is reasoning that cannot fail. The hypothesis — everyone is on his side — is confirmed by silence from allies and by silence from subordinates, which exhausts the available cases. There is no silence that would have counted against it.
And the adaptation happens in real time. The second inference is built after the creepers fail to answer, using the material the first inference did not need — which is the signature of a conclusion held fixed while the reasoning is refitted around it.
The mechanism producing the premises is the strangest thing here and it is entirely explicit.
The plants are sorted by grammatical gender. Vṛkṣa is masculine, so trees are male, so they are Kṛṣṇa's friends. Creepers are feminine, so they are female, so they are the gopīs' natural allies.
A fact about noun class has been converted directly into a fact about loyalty.
That is worth sitting with, because it is not obviously more absurd than a great deal of ordinary social inference. Categories that are arbitrary at the point of assignment routinely become the basis for expectations about allegiance — the department somebody works in, the school they went to, the side of a family they came from. None of those causes loyalty either. All of them get treated as evidence of it, and the treatment is usually as automatic and as unexamined as this.
What makes the passage useful is that the arbitrariness is visible. Nobody reading it will be tempted to think that Sanskrit noun class actually predicts a plant's allegiance. The identical operation applied to a human category is invisible from inside, and that is the only difference.
There is a structural fact about the scene that the corpus does not remark on, and it is worth stating plainly.
Caitanya, in the gopīs' mood, is reciting the gopīs' interrogation of the trees of Vṛndāvana — to actual trees, in a real garden, by the sea, at Purī.5
That is at least three layers deep. A man in another person's emotional condition, performing a remembered speech from a story, addressed to physical objects that he has identified as the objects in the story.
And the layers do not stay separate. The garden is not a stand-in for Vṛndāvana; on his own account it is Vṛndāvana, and the trees being addressed are therefore the trees being asked about in the recitation.6
Which means the recitation is not a performance of a scene. It is the scene, occurring, with the same questions put to the same trees — and there is no vantage available inside the passage from which to say which of the layers is the one actually happening.
Strip the setting and the operation is one anybody would recognise.
A person who has lost access to somebody is going around asking, and treating everything as evidence. The route the answers get processed through is a fixed conclusion — he is being kept from me — and every input, including no input, confirms it.
That is not a feature of the gopīs' condition. It is a feature of the conclusion's shape. A hypothesis that predicts both cooperation and refusal has no observation left that could count against it, and anybody holding one will find the world unanimous.
And the corpus supplies the honest detail: they keep asking. The creepers are questioned after the trees have failed, on a fresh premise. The confirmation is not satisfying — a hypothesis confirmed by everything is confirmed by nothing, and something in the asking knows it, because the asking does not stop.
Layer A supplies the interrogation, the gender reasoning, the two conclusions and the garden setting.7
Layer B expands the passage into the gopīs' separation and the Vṛndāvana pastimes.8
Worth marking what neither does. Neither treats the reasoning as reasoning. The gender-to-allegiance inference is presented and passed over; no layer asks whether it is sound, and no layer notices that the two conclusions are incompatible premises producing one result.
Which is what you would expect — the passage is not offered as an argument and nobody was ever meant to audit it. It is only an argument to a reader looking for one, and it is worth saying that this page is that reader rather than pretending the corpus intended the analysis.
You have read a meaning into somebody's silence. Ask what silence would have meant the opposite. If no version of not-answering would have counted against your reading, you have not learned anything from it — and the confidence you feel is a property of the hypothesis rather than of the evidence.
You are inferring allegiance from category membership. Check when the category was assigned and by whom. Grammatical gender is the clean case because its arbitrariness is undeniable; the categories people actually use are arbitrary at the point of assignment too, and the arbitrariness is simply harder to see from inside.
Your explanation adapted when the first version failed. Notice which part moved. A premise that gets rebuilt while the conclusion holds still is a conclusion being defended, and the rebuilding usually feels like being thorough.
You keep asking after you have your answer. That is information. A hypothesis confirmed by every possible observation does not settle anything, and the continued asking is the part of you that has registered it.
The evidence is a short recited passage with two explicit inferences from silence, both grounded in grammatical gender, arriving at one conclusion, inside a nested setting the narrator describes plainly.9
The tension the source doesn't close: the two inferences rest on opposite social premises and are treated as one line of reasoning. Friends who protect and subordinates who fear are different explanations, and the passage moves between them without noticing it has.
A second one. The scene is a recitation of an interrogation, performed on the objects the interrogation was about, by somebody who holds the location to be the location. No layer of the corpus supplies a vantage from which to say which layer is occurring.
Open: whether the gender reasoning is the gopīs', Caitanya's, or a compositional flourish of the author's. It is a pun-adjacent move available only in Sanskrit, and the man who wrote it down eighty years later had a demonstrated taste for exactly that kind of construction.10
This page and Citra-jalpa and the Bumblebee Song are the same speech-form with different staffing. There a bee is enlisted as interlocutor because it arrived from the right direction; here a garden is sorted into allies and subordinates by noun class — and both solve the form's one requirement, which is an addressee that cannot answer.
Against Vṛndāvana Overwrites Purī it is the same misidentification followed one step further. That page has him acting on the overwrite — running at the dune, jumping in the sea; here he interrogates the overwritten object, which is the only instance in the set where the substituted landscape is asked a question rather than approached.
And it belongs beside The Yogi of My Mind, where the mendicant of the mind begs "from citizens, trees and creepers alike." The same flattening of persons and plants into one addressable class, in adjacent chapters, in the same voice, unremarked both times.
Confirmation Bias as Ancient Problem — that page's mechanism is the sickle-thief: a man suspects his neighbour, and the neighbour's every movement looks like guilt; the sickle turns up, and the identical conduct now looks innocent. The evidence never changed; the prior did, and the observer cannot see from inside which of the two he is running.
The interrogation is the same machinery with one component removed, and the removal is what makes the pairing worth having.
The sickle-thief has behaviour to misread. The neighbour does things — walks a certain way, looks a certain way — and the prior distorts the interpretation of real signals.
Here there is no signal at all. Silence from a tree is not ambiguous data; it is the absence of data. And the prior generates a full reading out of it anyway — twice, on incompatible premises.
Which gives the finding neither page reaches. A prior does not need input to produce output. The sickle-thief story implies that bias operates on evidence, which suggests that reducing ambiguity would help; this passage shows the same mechanism running at zero evidence and arriving at conclusions of full confidence. Absence is the most permissive input there is, because it constrains nothing — and the readings people hold most immovably are disproportionately readings of what somebody did not say.
Sharpest implication. A prior does not need input to produce output, and silence is the most permissive input there is — it constrains nothing, so it confirms whatever is already held, twice over on incompatible premises if required. Which is why readings of what somebody did not say are held more immovably than readings of what they did, and why what silence would have meant the opposite is the only question that gets any purchase on them.
Generative questions.