Twice in 1,575 endnotes, the apparatus stops and says it does not know:
There are many difficulties in translating and interpreting this shloka. The meaning is not immediately obvious.1
What is intended is unclear. It could be the three Vedas, the three gunas, or past, present and future.2
The first is a warning without a solution. The second is a warning with three solutions, offered as alternatives, none preferred.
n. 890 is an ordinary scholarly hedge. Difficult shloka, meaning not obvious, reader be advised.
n. 1323 does something else. The problem is a word meaning three — a set of three that the text does not identify — and the note lists what it might be:
The three Vedas. The three guṇas. Past, present and future.
Those are not close. They are a scripture, a metaphysics and a temporality, and the passage reads differently under each. The note does not rank them, does not indicate a preference, and does not say which the translation assumes.
Which means the English above the note has already chosen and the note does not say what it chose.
That is the sharp edge of this species. A reader who reads only the body gets one reading. A reader who reads the note learns that two others were available and is not told which one they just consumed.
Set these two against the other seven species of intervention in the volume and they are the cheapest to disclose and the most useful to have.
We have taken a few liberties required admitting a departure.3 Translations usually say… this seems forced and unnecessary required taking a position against the field.4 Strictly speaking, Vidura was not a shudra required correcting the text.5
Admitting non-comprehension costs a translator nothing except the appearance of mastery, and it is the one species that improves the reader's position without asking them to trust anything.
And it is used twice.
Which is the finding, and it should be stated carefully. Two admissions in 1,575 notes is not evidence that the translator understood everything else. It is evidence about how often he said he did not — and the base rate of genuinely obscure passages in a Purāṇa is certainly higher than two.
The census's own accounting flags something beneath these two notes: a wider layer of interpreted in many different ways formulations scattered through the apparatus.6
Those are not admissions of non-comprehension. They are reports of disagreement — this is read variously — and they do a different thing. They tell the reader that the tradition is divided without telling them that the translator is stuck.
So there are three registers available and the volume uses all of them:
The tradition is divided — interpreted in many different ways. Frequent. I do not know — the meaning is not immediately obvious. Twice. The others are wrong — forced and unnecessary. Three times.4
The first shifts responsibility to the field. The third takes it. The second is the only one that leaves the reader without a position to adopt, and it is the rarest.
Which is not a criticism. A translation whose apparatus said I do not know on every genuinely hard shloka would be unusable. It is a note about what the ratio can and cannot tell you: the frequency of stated uncertainty is a fact about the apparatus, not about the text.
The vault has one comparison available and it is instructive.
The caterpillar simile at 4(29) is genuinely ambiguous — it supports a reading about continuity and a reading about metamorphosis, and the volume's entire transmigration mechanism turns on which one is taken.7
Debroy does not flag it as unclear. He flags it as far-fetched in the other direction — the metamorphosis reading is rejected, and the note reads as a verdict rather than as a difficulty.
So the volume has a passage where two readings produce incompatible doctrines, and the note treats it as settled; and a passage where three readings are possible, and the note treats it as open.
The difference between the two is not the degree of ambiguity. In one case the translator had a view and in the other he did not, and the apparatus reports the view rather than the ambiguity.
The practical consequence is narrow and worth stating.
Two flagged passages in a volume of this size means the flags are not a map of the text's difficulty. They are a map of two occasions on which the translator chose to say so.
Which means the correct inference from an unflagged passage is nothing at all. The absence of a note is not a claim of clarity, and treating it as one is the single easiest mistake to make with an annotated translation.
The flags are informative about the two passages they mark and uninformative about the rest. That is a small thing to know and it prevents a large error.
There is a second thing the apparatus does not say, and it is bigger than these two notes.
The volume answers the same question twice, differently, seven times — sheaths, ignorances, cosmic-body maps, avatāra catalogues, the Jaya–Vijaya curse, Parīkṣit's name, tattva schemes. Every occurrence is annotated. Not one note says the other occurrence exists.9
Set that beside n. 890 and n. 1323 and the two silences turn out to be different in kind.
The unclear-passage silence is local and reportable. The translator is looking at one shloka, cannot resolve it, and says so. The knowledge required to notice the problem is available at the desk.
The internal-disagreement silence is global and structurally unreportable. To flag the seven-sheath passage as disagreeing with the eight-sheath passage, the annotator would need to be holding chapter 2(10) in mind while writing a note on 2(1) — four hundred pages away and hundreds of notes earlier.
A footnote cannot see another footnote. That is not a limitation of this apparatus; it is what an apparatus is.
Which means the two admissions on this page mark the outer edge of what the format can report. The meaning is not immediately obvious is the most a per-passage note can say about a difficulty it can see. The seven disagreements are difficulties no per-passage note is positioned to see at all.
You are working from an annotated source — a commented codebase, a translated document, an edited dataset — and the annotations are your only view of the editor's judgement.
Count the registers, not the notes. The field disagrees, I do not know, and the others are wrong are three different admissions with three different costs, and the ratio between them tells you more about the editor than any individual note does. An apparatus that never uses the middle register is not more certain; it is less legible.
Never read an absence as a claim. The unflagged passages are not the clear ones. They are the ones nobody chose to mark, and the reasons for not marking include clarity, confidence, oversight and haste — which are indistinguishable from outside.
And check the flagged ambiguity against the load-bearing one. The most consequential ambiguity in this volume is unflagged and the least consequential is flagged with three options. Editors mark what puzzled them, not what matters, and those two sets overlap by accident.
The strongest evidence that the flags are about the annotator rather than the text is n. 1323's structure — three incompatible options offered, none preferred, above a translation that has already chosen one silently.
Tension — the volume's most doctrinally consequential ambiguity is not flagged as unclear but resolved as far-fetched.7 Preserved.
Tension — two admissions in 1,575 notes cannot be a map of the text's difficulty, and the volume offers no way to tell which unflagged passages are hard. Preserved unresolved.
Open question. n. 1323's three options — Vedas, guṇas, or past-present-future. Which does the English above it assume, and can it be recovered from the wording?
The commentarial tradition on the Bhāgavata is enormous and its standard operation is the opposite of this note: where a passage admits three readings, a commentator argues for one and gives reasons.
Debroy's method is declared and different — the simplest sense, minimal doctrinal commitment — and n. 1323 is that method producing its characteristic output: the options are listed and the choice is not defended.8
Which is honest and it has a cost the tradition's method does not. A commentator's choice is arguable; an undefended choice is invisible. The reader of a commentary knows what was decided and why. The reader of this translation gets the decision embedded in the English and a footnote saying the question was open.
The vault holds both as legitimate. What it records is that they fail differently: commentary can be wrong in ways you can see, and this can be wrong in ways you cannot.
With Confidence, Anxiety and the Four Levels of Competence. That page holds the competence ladder — unconscious incompetence, conscious incompetence, conscious competence, unconscious competence — and the finding that the second rung is where anxiety peaks, because you can see the gap and cannot yet close it.
The meaning is not immediately obvious is a statement from the second rung, written down and published. And the interesting thing about the volume is how rarely it appears: twice, in an apparatus of 1,575 notes.
What the pairing produces: the ladder is normally used developmentally, tracking one person's progress through a skill. This case uses it as a disclosure ratio, and that turns out to be a different measurement.
An expert operating at the fourth rung — unconscious competence — produces confident output and does not experience most decisions as decisions. That is what makes the level valuable and it is also what makes it unreportable: you cannot flag an ambiguity you did not notice resolving. n. 1323 exists because that one stayed conscious; the caterpillar note exists because that one felt like a judgement rather than a difficulty.
Which suggests a diagnostic that applies to any expert-produced artefact. The rate of I do not know in a body of expert work is not a measure of the material's difficulty or of the expert's honesty. It is a measure of how much of their processing remained conscious — and the more skilled the practitioner, the lower that rate goes, for reasons entirely unrelated to how much they actually resolved correctly.
Sharpest implication. n. 1323 lists three incompatible readings — the Vedas, the guṇas, past-present-future — offers none as preferred, and sits above an English translation that has already picked one silently. Two such admissions in 1,575 notes cannot be a map of the text's difficulty; they are a record of two occasions when a decision stayed conscious enough to report. And the volume's most consequential ambiguity, the one its transmigration doctrine turns on, is not flagged at all.
Generative questions.