The Bhāgavata gives its roster of Viṣṇu's descents in the First Skandha, numbered, in order — fourteenth Narasiṃha, fifteenth the dwarf, sixteenth the axe-wielder, seventeenth Satyavatī's son, nineteenth and twentieth Rāma and Kṛṣṇa, then Buddha, then Kalki.1
Then it gives the roster again in the Second Skandha, in a different chapter, to a different purpose, and the second telling is not the first with different wording. It is a different list.
Neither passage mentions the other. Neither is marked as a summary, a variant, or a correction. A reader who meets one and not the other has a complete account; a reader who meets both has two.
The two are easy to conflate because they share most of their contents. The differences are structural rather than lexical.
The first list is numbered. The second is not. Chapter 1(3) counts — in his fourteenth, in the sixteenth descent, in the nineteenth and the twentieth — and the count is what makes it a list. Chapter 2(7) narrates episodes in sequence without ordinals, which means it has no total.
The first list names Buddha. The second does not. At 1(3): he will be born as Jina's son, in the land of Kikata and will be named Buddha.1 At 2(7) the same figure is described only by function — He will confound their intelligence, with a desire to tempt them. He will don many kinds of attire and speak to them about what is not dharma — and it is Debroy's endnote that supplies the identification, with an unusually frank hedge: "Though this is by no means obvious, this is interpreted as a reference to Buddha."2
The second list is longer where the first is compressed. Chapter 2(7) spends paragraphs on the destruction of named enemies — Pauṇḍraka, Śālva, Balvala, Dantavakra, Śambara, Rukmī, the Kāmbojas, Matsyas, Kurus, Sṛñjayas — where 1(3) covers the same descent in a clause.
And the second is in the future tense throughout. That is a large enough matter to have its own page.3
A numbered list makes a claim a narrative sequence does not: that the set is closed and countable.
Once you say the sixteenth, you have committed to there being a fifteenth and a seventeenth and to somebody being able to say how many there are in total. The count is a promise about completeness.
Chapter 2(7) makes no such promise. It tells you what he did, or rather will do, one thing after another, and the sequence could be extended at either end without contradiction.
So the two passages are not two versions of one doctrine about how many avatāras there are. Only one of them has a doctrine about how many there are. The other has a story.
Here is what makes the numbered list stranger than it first appears: it never gives its own total.
The ordinals run up to the twentieth, and then two more arrive without numbers — Buddha and Kalki, both introduced by when kali yuga ensues and when it is the conjunction of two yugas rather than by an ordinal. So the passage counts to twenty and then stops counting, while continuing to enumerate.
Twenty-two is the number the tradition supplies. It is not the number the passage supplies. The passage supplies twenty ordinals and two more entries.
And the two that fall outside the numbering are exactly the two that have not happened. The ordinals cover what is past; the unnumbered pair covers what is to come. Which suggests the numbering is not a census of the complete set but a count of the completed ones — a running total rather than a final one.
If that reading holds, the list is open by construction, and the tradition's fixed number is a later closure of something the text left running.
Register: this is a reading of where the ordinals stop, in a translation. The text offers no comment on why the last two lack numbers. What can be shown is that they do, and that the boundary falls exactly at the tense change.
A question worth putting to a volume with two avatāra lists is whether it has a third.
It does not. Varāha and Kapila both arrive in Skandhas 3 and 4 as fully narrated episodes — the boar lifting the earth, the sage instructing his mother — with all the weight a descent could carry, and neither is presented as an item in a roster. They are events in the story, not entries in a catalogue.
So the count stays at two lists. That matters because it constrains what kind of disagreement this is. The volume is not accumulating rosters loosely; it has exactly two, in the two Skandhas that are doing framing work, and it never reconciles them.
The obvious question is which list the volume thinks is right, and the answer is that nothing in it addresses the question.
There is no as was said earlier, no in brief, no some say. Each list is delivered in its own chapter as though it were the account, to a listener who has in one case already heard the other one.
That is the same behaviour this volume shows with the sheath count, with the cosmic body's anatomy, and with the elemental derivation: the same content stated twice, differently, each version locally coherent, neither flagged. What is distinctive here is only the subject matter. A count of divine descents is closer to the doctrinal centre than a count of the egg's wrappers.
One thing worth crediting, because this page is otherwise a catalogue of unmarked discrepancies.
Debroy's note on the Buddha passage at 2(7) does not assert. Though this is by no means obvious, this is interpreted as a reference to Buddha. He names the identification as an interpretation, flags that the text does not support it plainly, and leaves the reader able to disagree.
And at 1(3), where the text does name Buddha, his note on Jina's son offers two readings rather than one — Jina as a name of the Buddha himself, or jina as an adjective applied to Śuddhodana, Siddhārtha's father, meaning victorious.4
That is careful work at exactly the point where a translator's own commitments would be hardest to keep out. It makes the absence of any note connecting the two lists more conspicuous rather than less: the same hand that hedges a single identification does not remark that the passage it appears in is the volume's second roster.
You have two documents that describe the same thing — a product's feature set as marketing has it and as the docs have it, a process as written and as practised, a history told by two departments.
The instinct is to diff the contents and reconcile the differences. Do that second.
First, check whether the two are the same kind of artefact. One of these lists is numbered and one is narrative. That is not a difference in content; it is a difference in what the document is claiming. A numbered version asserts completeness and invites the question how many. A narrative version asserts sequence and does not. You can reconcile two counts. You cannot reconcile a count with a story, and trying to is how a narrative gets retroactively treated as a census.
Then find where each stops committing. The ordinals here run out precisely where the tense changes. That boundary is usually the most informative thing in the document — it marks what the author was willing to be pinned to.
Only then compare contents, and expect the differences to cluster around whatever each version was for. This volume's second list dwells on the destruction of named enemies because it is addressing a king from that lineage. The first does not, because it is establishing a framework.
The strongest evidence that the two passages are different in kind rather than in wording is the numbering. One counts and the other does not, and the one that counts stops counting exactly where the events stop being past.
Tension — the volume gives two avatāra rosters, in two Skandhas, differing in numbering, in whether Buddha is named, and in tense, with no cross-reference and no primacy assigned. Preserved unresolved.
Open question. The ordinals stop at twenty and two unnumbered entries follow. Is the numbering a count of completed descents rather than of the whole set — and if so, is the tradition's figure of twenty-two a closure the text does not authorise?
This vault's Vaiṣṇava material generally treats the twenty-two-avatāra list as the Bhāgavata's signature contribution, distinguished from the shorter ten-item scheme and cited as evidence of the text's expansive view of descent.
The volume itself is less settled than that. It gives the expansive list once, gives a second and differently constructed account in the next Skandha, never states a total, and does not indicate which is authoritative. The number twenty-two is arrived at by counting entries in one passage and continuing past the point where the passage stops counting.
Seven Sheaths, or Eight documents the same volume stating a count in one chapter and a different one nine chapters later, both glossed locally. The avatāra case is the same behaviour with a higher-stakes object and a subtler mechanism: here the two versions do not even disagree about a number, because only one of them offers one.
With Brand Universe Flexibility: Marvel vs Star Wars. That page works two franchise models for how much a new instalment may vary from what came before — one demanding continuity, the other treating variance as a feature, with the connecting thread in the second case being shared characters and ongoing plot rather than consistency of register.
The Bhāgavata's two rosters behave like the second model. The descents are held together by the figure descending, not by agreement between the accounts, and a reader encountering both is not expected to experience a contradiction.
What the pairing produces: the franchise analysis assumes an audience that could audit continuity if it chose — the material is finite, indexed, and endlessly compared by people who care. Under those conditions tolerance for variance is a genuine editorial choice, and it has a cost the page can measure. The Purāṇic case removes the auditing audience entirely: the two lists sit in different chapters of a recited text, encountered days apart in a seven-day telling, and nothing in the practice of hearing brings them into contact. Which separates two things the franchise framing runs together — a canon can be inconsistent by permission or inconsistent by inaccessibility, and only the first is a decision. The second produces the same surface freedom with none of the deliberation, and it collapses the moment the text becomes searchable. Which is what happened to this one.
Sharpest implication. The tradition's famous count of twenty-two is assembled by continuing a numbering the text stops. The passage counts what has happened and declines to number what has not — so the fixed roster is a later act of closure applied to a list that was, in its own construction, still open at the end.
Generative questions.