There are eighteen major Puranas. This is standard, stated in several of the texts themselves, and treated as settled.
Here they are, with their shloka counts, alphabetically as Debroy gives them: Agni (15,400), Bhagavata (18,000), Brahma (10,000), Brahmanda (12,000), Brahmavaivarta (18,000), Garuda (19,000), Kurma (17,000), Linga (11,000), Markandeya (9,000), Matsya (14,000), Narada (25,000), Padma (55,000), Shiva (24,000), Skanda (81,100), Vamana (10,000), Varaha (24,000), Vayu (24,000), Vishnu (23,000).1
Total: 409,500 shlokas, roughly eight million words — four times the Mahabharata.2
Now the disagreements, and there are four.
Sometimes Bhavishya (14,500) is in and Vayu is out. Sometimes Vayu is in and Agni is out. Some lists include Narasimha and Vayu while dropping Brahmanda and Garuda. And "Bhagavata" is ambiguous — it may mean the Bhagavata Purana or the Devi Bhagavata Purana.3
So the number eighteen is fixed and the membership isn't.
That's the finding, and it's worth stating plainly: eighteen is a constant with variable contents.
Nobody counted the available texts and arrived at eighteen. The slot count was settled, and then different communities filled the slots differently — and when a new text needed a place, something else came out.
Debroy is careful to note that it "is not obvious that this number of eighteen existed right from the beginning."4 But by the time the lists are being written down, eighteen is the thing that doesn't move.
That is how canon usually behaves. The shape hardens before the contents do, and the contests happen inside a fixed frame.
There's a detail in how the list is presented that carries more than it looks.
Fourth, when a list is given, the order also indicates some notion of priority or importance. Since that varies from text to text, our listing is simply alphabetical, according to the English alphabet.5
Every Purana that gives the list is also, by the sequence it chooses, making a claim about rank. And the sequences differ, because each text is composed by people with a view about which god comes first.
Debroy sidesteps it by using the English alphabet — a scheme with no stake in the fight, imported from outside the tradition entirely.
Which is a small methodological decision with a large effect. His neutrality is only available because he is using a foreign ordering principle. There is no neutral order available from inside.
Put the shloka counts side by side and the corpus stops looking like a set of eighteen comparable books.
Skanda has 81,100. Padma has 55,000. Markandeya has 9,000 — Skanda is nine times its size. Shiva sits at 24,000, below the median-ish middle and less than a third of Skanda.
Three implications follow.
The Mahapurana category is not about magnitude. A 9,000-shloka text and an 81,100-shloka text hold equal formal status. Whatever "major" means, it isn't length.
The Shiva Purana is a mid-sized text arguing it outranks all of them. Its supremacy claims — that all other Puranas "roar in this world" only until it arrives — are made from the middle of the distribution.6
And the biggest one is the one it borrowed from. Chapters 1.1–1.7, the seven arguing hardest for this text's supremacy, are from the Skanda Purana.7 The 24,000-shloka text ranks itself above the 81,100-shloka text using material the larger one supplied. See The Praise That Came From Another Purana.
The corpus gets sorted two ways, and Debroy is unimpressed by both.
By deity. Brahma-Puranas (creation themes), Vaishnava Puranas, Shaiva Puranas — with Shiva grouped alongside Agni, Linga, Skanda and Vayu. His verdict: there is "a grain of truth" but all three gods feature in every Purana, so "beyond the relative superiority of Vishnu vis-à-vis Shiva, the taxonomy probably doesn't serve much purpose."8
By guna. Sattva, rajas, tamas, from a passage in the Padma Purana's Uttara Khanda, recited by Shiva to Parvati. Debroy calls this "even more tenuous."9
And here is what that second scheme says. The sattva Puranas — pure — are the six Vaishnava ones. The tamas Puranas — darkness, ignorance — are Agni, Kurma, Linga, Matsya, Skanda and Shiva.
So a classification found in a Vaishnava text places the Vaishnava Puranas in the highest category and the Shaiva ones in the lowest, and puts the judgement in Shiva's own mouth speaking to his wife.
That is not a taxonomy. It is a sectarian verdict wearing one, and the mechanism — attributing your ranking to your rival's god — is worth noting as a technique.
Two boundary cases sharpen what a Purana is.
The Harivamsha is "sometimes loosely described as a Purana, but strictly speaking, it is not." It's an addendum to the Mahabharata.10 So proximity to the epic doesn't qualify a text.
Upa-Puranas are minor Puranas, often said to number eighteen as well, with no country-wide consensus on the list. The Śaiva corpus has several — Shivadharma, Shivadharmapurva, Shivarahasya, Maheshvara.11
And that second category is where the knife sits. Hazra's position is that the Shiva Purana is one of these — "the comparatively late sectarian Upapurana called 'Siva-purana'" — raised to Mahapurana status by "late sectarian zeal."12
The list this page describes is exactly the thing that text is arguing its way onto. See The Disputed Status of This Text.
You are looking at the list and trying to work out what you are holding.
It's on the list. Thirteenth alphabetically, 24,000 shlokas, sitting between Markandeya and Skanda.
Then you notice the footnotes. Vayu is on some lists instead of this one — and the substitution is not accidental, because the reason given elsewhere is that "Śiva" in most enumerations is really a title for the Vāyavīya.13 So on some accounts your text's slot belongs to a different book.
You notice the guna scheme puts it in tamas.
You notice a serious scholar declined to treat it as a separate Purana at all.
And you notice that none of this appears in the text itself. Chapter 9 tells you it is the crest of all the virtuous Puranas and that hundreds of other texts aren't worth a sixteenth of it.
The confidence is entirely on the inside and the doubt is entirely on the outside, and both are printed in the same volume, about two hundred pages apart.
One correction the list makes to a common assumption, and it changes what these texts are for.
Debroy is explicit that the five defining topics don't exhaust the contents. The Puranas carry substantial material on geography, on geological change — shifts in the courses of rivers — and on astronomy. "Therefore, those five attributes shouldn't suggest the Puranas have nothing more. They do, and they have, therefore, been described as encyclopedias."15
And they are regional. "Bharatavarsha is vast and heterogeneous and each Purana may very well have originated in one particular part of the country," with the depth of geographical detail varying accordingly — each text knows its own region best.16
That reframes the eighteen. Not eighteen competing statements of doctrine, but eighteen regional compendia, each holding the cosmology, ritual practice, dynastic history, geography and law of somewhere in particular.
Which explains a good deal about the disputes. Membership lists vary by region because the lists were made in regions. And the general practice of dharma, Debroy notes, rests "much more on the Puranas than on the Vedas" — culture, law, ritual, architecture and iconography all come from here.17 These were the working reference books, not the holy of holies.
Strongest support is that Debroy gives the list, the counts, the four disputes and both classification schemes in one place, with his own assessment of each.
Tension one. Eighteen is fixed and membership varies. Either the number is arbitrary and got fixed early, or texts were fitted to a pre-existing sacred count. Debroy notes the number may not be original and does not choose.
Tension two. The guna scheme is a Vaishnava ranking placed in Shiva's mouth. Debroy dismisses it as tenuous but reports it. The vault should record it as evidence of sectarian technique rather than as a taxonomy — but it is in the corpus, and some readers took it seriously.
Tension three. The shloka counts are stated and, Debroy says, generally exceed what any actual edition contains. "Either those many shlokas did not originally exist, or they have been lost."14 So the list's own numbers are aspirational.
Open question. Which Puranas' lists produce which membership disputes? Debroy reports the four variations without saying which texts hold which position. That mapping would show whether the disputes track sectarian lines — which the Vayu/Shiva substitution suggests they do.
The vault has no page on canon formation in this domain, and its Śaiva corpus has been operating without one.
A Tantra Is a Text (Reclaim the Source) argues for returning to the vetted corpus rather than to internet material. The argument holds. This page supplies what "vetted" means in practice: a list with a fixed length, contested membership, two incompatible ranking schemes, and at least one member whose inclusion a major scholar rejected.
That doesn't undermine the advice. A contested canon is still incomparably better than no canon. But the vault should stop treating "the corpus" as a settled object with a clean boundary, because the boundary is where the arguing happens — and the Shiva Purana's own supremacy rhetoric is best understood as a text arguing at that boundary rather than speaking from safely inside it.
The Founder Problem in Historical Perspective (history). The history corpus documents retrospective construction — how a tradition builds an account of its own origins that licenses its present shape.
Canon formation is that process applied to a set rather than a person, and this case shows the machinery unusually clearly: a fixed slot count, texts moving in and out, and each list's ordering asserting a rank that the next list contradicts.
What the pairing produces: a canon with a fixed size is a zero-sum arena, and that changes what its members' rhetoric is for. If eighteen is constant, a text arguing for its own greatness is not only praising itself — it is arguing that something else should come out. The Shiva Purana's supremacy stack reads differently once you know Vayu occupies its slot on competing lists. The history pages treat self-promotion as identity-building; here it is territorial, and the territory is finite.
Authority: Institutional Override (behavioral-mechanics). The behavioral corpus documents invoking a higher authority to settle what the speaker cannot settle himself.
The guna scheme is a refined version: a Vaishnava text ranks the Vaishnava Puranas as pure and the Shaiva ones as tamas, and attributes the ranking to Shiva, speaking to Parvati.
The override isn't to a neutral higher authority. It is to the opposing side's own god, conceding your verdict about you.
The insight neither yields alone: the strongest form of the override is a claimed concession from the party being ranked, because it removes the appearance of self-interest entirely. The behavioral pages document escalation upward to shared authorities. This escalates sideways, into the rival's mouth — which is unfalsifiable in the same way everything in a nested-attribution corpus is unfalsifiable, and considerably more damaging than a claim made in one's own voice.
Sharpest implication. Eighteen is a constant with variable contents — the slot count hardened before the membership did, which makes the corpus a zero-sum arena where a text arguing for its own greatness is arguing that something else comes out. The Shiva Purana is mid-sized at 24,000 shlokas, claims supremacy over all of them, borrowed that claim from the 81,100-shloka Skanda, and is placed in the tamas category by a scheme that puts the verdict in Shiva's own mouth.
Generative questions.