In chapter 2(1), Śuka describes the cosmic body he is about to hand Parīkṣit as an object of concentration:
This body is like an egg and is covered in seven sheaths.1
Nine chapters later, closing the same description for the same listener, he says:
I have described to you the gross body of the illustrious one. On the outside, it is covered by eight sheaths, the earth being the first.2
Same speaker. Same listener. Same object, named the same way. Two counts.
Neither passage refers to the other. Neither is hedged. Between them sits the entire second Skandha's cosmology, delivered as one continuous teaching to a man with seven days left.
What turns this from a curiosity into something usable is that Debroy annotates both, and the two notes can be laid side by side.
On the seven:
Interpreted as earth, water, fire, air, space, ahamkara and Mahat.3
On the eight, glossing "the earth being the first":
The other seven are made out of water, fire, air, space, Mahat, ahamkara and Prakriti.4
Count them out. The first list: earth, water, fire, air, space, ahaṃkāra, Mahat. Seven. The second: earth, then water, fire, air, space, Mahat, ahaṃkāra, Prakṛti. Eight.
The lists are identical except that the second one adds Prakṛti. Everything else is the same set of five elements plus the two psychic principles. The entire disagreement between chapter 2(1) and chapter 2(10) reduces to a single question: is Prakṛti a sheath, or is it the thing the sheaths are made of?
That is not a rounding error. It is a real position in Sāṅkhya, and it is the position on which the whole scheme's boundary depends.
A smaller detail, easy to miss and worth having.
The first note ends "…space, ahamkara and Mahat." The second ends "…space, Mahat, ahamkara and Prakriti."
Ahaṃkāra and Mahat have swapped. In the standard Sāṅkhya derivation the order matters — Mahat arises from Prakṛti and ahaṃkāra from Mahat, so a nested set of sheaths running outward-in should put Mahat inside ahaṃkāra or outside it consistently, not both ways.
So the two notes disagree twice: on how many sheaths there are, and on which of the two innermost comes first. Neither note mentions the other exists.
Take the question seriously rather than treating it as bookkeeping, because the answer changes the shape of the cosmos.
Under the seven, the sheaths run from earth inward to Mahat, and Prakṛti sits outside the whole enclosure as the unmanifest ground that the enclosure is precipitated out of. The egg is a thing in Prakṛti. Prakṛti is not part of the packaging.
Under the eight, Prakṛti is the outermost sheath — a layer of the egg, the last wrapper before whatever is beyond. Which means the egg has an outside that is not Prakṛti, and the unmanifest ground has been demoted into an item on a list.
Those are different cosmologies. In one, the manifest world is a bubble inside an unmanifest medium; in the other, the unmanifest is itself the bubble's skin. A practitioner meditating outward through the sheaths reaches, at the last layer, either the end of the packaging or the end of manifestation itself — and the two are not the same destination.
Chapter 3(11) adds a scale to the structure without settling the count:
This cosmic egg is united with transformations and their various manifestations. On the outside, it extends for fifty crore. Each sheath is ten times larger than the preceding one. When they enter together, they are like a paramanu.5
Ten times per layer. Which means the count is not decorative — it is an exponent. Seven sheaths and eight sheaths differ by a factor of ten in the size of the whole, and the text supplies the multiplier without ever fixing the number of times to apply it.
Register: this is one internal disagreement, and it belongs to a pattern already logged across this volume rather than standing alone. The volume says several things twice and differently and glosses each version locally without flagging any of them. This is the cleanest instance, because here the two glosses are by the same hand, are eight words apart in content, and can be diffed.
The volume's other internal disagreements are harder to pin, because their two versions differ in more than one way at once — different chapters, speakers, purposes, vocabulary.
This one is controlled. Same speaker, same listener, same continuous teaching, same object under the same description, same annotator, and a difference reducing to one named item. There is nowhere for it to hide in context, which makes it the right instance to point at when the pattern needs demonstrating rather than asserting. Only the count moves.
There is a reason this discrepancy bites harder than a disagreement about, say, which god sits at the mouth.
The sheaths are not scenery. They are the thing the practitioner passes outward through — nested layers, each ten times the last, traversed on the way out of the manifest. A count of sheaths is therefore a count of stages in a procedure, and a procedure whose number of stages is unfixed is a procedure with an unfixed endpoint.
Compare it to the anatomy collisions elsewhere in the same chapters. There, the practitioner holds one limb at a time and releases it, so a slot occupied twice never surfaces. Here, the layers are meant to be gone through in sequence to the last one, and knowing you have reached the last one requires knowing how many there were.
So this is the case where the volume's habit of local glossing actually costs something operational, rather than only costing consistency.
It is worth being precise about the failure, because it is a small and specific one and overstating it would be its own error.
Debroy is not wrong at either site. The seven-list is a defensible reading of chapter 2(1) and the eight-list is a defensible reading of chapter 2(10) — indeed the second is not a reading at all, since the text says eight and says the earth being the first, so naming the other seven is close to arithmetic.
What is missing is the cross-reference. Four words — "contrast 2(1), which has seven" — would have converted an invisible discrepancy into a documented one. The apparatus demonstrably has that move available: elsewhere in this volume it reports a numerical check that fails and says so.6
So this is not a translator concealing a problem. It is a translator working locally, note by note, in exactly the way the practitioner is instructed to work limb by limb — and inheriting the same blind spot.
You have a definition that appears in more than one place — an onboarding doc and a README, a spec and its implementation, two decks by the same person six months apart.
Do not read them for agreement. Reading for agreement is how this survives; both versions are plausible, and plausibility is all a reading check tests.
Put them in two columns and diff the items, not the prose. The count first, then the membership, then the order. That sequence matters: the count is the cheapest check and catches the most, membership tells you what the disagreement is actually about, and order catches the disagreements that membership hides — the two lists here contain the same seven names in two different sequences, which no summary would have surfaced.
Then, when you find the delta, resist fixing it immediately. Ask what each version buys. Prakṛti-as-sheath and Prakṛti-as-ground are not a typo and a truth; they are two positions, and someone chose each of them for a reason that may still be live. The delta is a record of a decision, and you want the decision before you overwrite it.
The strongest evidence that this is a genuine doctrinal disagreement rather than a slip is that the delta is exactly one principled item. Not a miscount, not a dropped element — Prakṛti, whose membership is precisely the contested boundary in the scheme. A careless error would have lost a random layer.
Tension — the same continuous teaching states the sheath count as seven and as eight, and the same apparatus glosses both without cross-reference. The multiplier at 3(11) makes the difference a factor of ten in the size of the manifest cosmos. Preserved unresolved.
Open question. Is Prakṛti inside the egg or outside it in this text's own view? Neither passage argues the point, and the practice — meditating outward through the layers — would terminate in a different place depending on the answer.
This vault's Sāṅkhya material treats the Prakṛti boundary as load-bearing: Prakṛti is the unmanifest ground from which the tattvas evolve, and it is categorically not one of them. On that account the seven-list is correct and the eight-list has made a category error.
But the Bhāgavata is not doing textbook Sāṅkhya, and elsewhere in this volume it runs more than one tattva scheme without reconciling them either. The eight-list may be reproducing a Purāṇic convention in which the egg's wrappers are counted inclusively up to and including the unmanifest — a different tradition of counting rather than a mistake within one.
What can be said without choosing: the text does not adjudicate, the apparatus does not notice, and the reader is handed both. Two Incompatible Maps of the Same Body documents the same behaviour in the same two chapters with a different object — an anatomy assigned three times over, unremarked. Whatever is producing this is not local to the sheaths.
With The Six Methods of Controlling Options. That page opens on a count that does not match its contents: Greene's chapter is titled and referred to as a set of six, and the chapter carries seven bolded headings. The vault's own page records the mismatch in its first paragraph and then works the material anyway.
Set that beside the sheaths. Two sources, two counts, in both cases a named total that the enumerated members do not add up to, and in both cases the discrepancy sitting in plain sight in a text people have used for a long time without stumbling.
What the pairing produces: the two cases fail in opposite directions, and the pair makes the mechanism visible. Greene's list has more members than its stated number — the count was fixed first, as a title, and the content grew past it. The Bhāgavata has more counts than its list has variants — the members held nearly constant and the number moved. Which locates what neither shows alone: in an enumerated scheme the number and the membership are maintained by different processes and drift independently. The number is rhetoric — it gets quoted, titled, remembered. The membership is content. Audit them separately, because being wrong about how many you have and being wrong about which they are are different defects with different fixes.
Sharpest implication. A discrepancy survives not by being hidden but by never being adjacent to itself. Nine chapters is enough distance to make seven and eight both feel obvious, and the annotator standing outside the text, whose whole job is the view the practitioner does not have, inherited the same distance and the same blindness.
Generative questions.