Author — Anonymous (traditionally ascribed to Krishna Dvaipayana Vedavyasa); translated by Bibek Debroy
Year — Text dated by the translator to c. 500–1000 CE. This translation: Penguin Books, the third and final volume.
Original file — EPUB, ingested 2026-08-13 from Downloads/Telegram Desktop/. ⚠ Not yet placed in RAW/books/. The file should be moved there for permanence; every citation on downstream pages currently traces to a download path that will not survive a disk clean. The same warning stands on the Volume 2 stub and has not yet been acted on.
Source type — Primary text in unabridged translation
Publisher — Penguin Books
Sanskrit source text — Nag Publishers, Delhi. A transliterated text is available through the University of Gottingen.
Coverage — 123 printed chapters, ~185,000 words. Tenth Skandha chapters 12–90, Eleventh Skandha 1–31, Twelfth Skandha 1–13. The volume resumes exactly where Volume 2 broke off, at 10(11). Two chapters listed in the front matter are not printed — see Provenance below.
Endnote layer — Three note files, roughly 2,900 notes: Tenth Skandha (continuing Volume 2's numbering), Eleventh, and Twelfth, plus separate notes to the Introduction. Mined in full during ingest across two disjoint diagnostic sweeps.
Volume 3 carries the Purana's most famous material and its least defended. The Tenth Skandha runs the whole Krishna biography from the Agha episode to the close of the Dvaraka years; the Eleventh is the Uddhava-gita and the destruction of the Yadus; the Twelfth is the kali-yuga prophecy, the dynastic lists, and the book's account of its own transmission.
The through-line is that the text repeatedly argues against itself and lets the arguments stand. Balarama tells God to his face that what he has done is "wicked" and "contemptible" and is never answered (10.54). A gopi delivers a three-count charge-sheet against Krishna drawn from three separate avataras — he shot Valin from cover like a hunter, he disfigured a woman who desired him, he ate Bali's offerings and then bound him — and no reply is printed (10.47). Shishupala's attack on Krishna's legitimacy is procedural rather than theological (no varna, no ashrama, a cursed lineage, a sea-fort held "like bandits") and is met with a beheading rather than a rebuttal (10.74). And at 10.77 the narrator interrupts his own narration to attack the verse he has just delivered, calling the rishis who say God can grieve "inconsistent in their reasoning."
The second running argument is that the qualification the tradition tests for is not the qualification the text rewards. Ritualists refuse God food and their unlettered wives walk out of the house and are right (10.23). Cowherds who ask for nothing aloud are shown the world beyond tamas (10.28). At 10.87 the Vedas themselves state, in the volume's most technical chapter, that the state reached by yogis "was also obtained by your enemies, who constantly remembered you," and by "the women, whose minds were attached to your sturdy arms."
The third is an unresolved disagreement about who may approach, and it is confined to the Eleventh Skandha. Five positions: Chamasa restricts the account to the twice-born and places "women, shudras and others like that" outside it as objects of pity (11.5); Krishna names "vaishyas, shudras, women, outcastes" among those who reached him by association alone (11.12); he says devotion purifies even one "born as a svapacha" and six lines later says keep women far away (11.14); kriya-yoga is said to bring "the greatest benefit for women and shudras" (11.27); and the teaching is released to both "as long as they possess devotion" (11.29). The caste bar and the sex bar do not move together. Birth-status is overruled and never re-imposed. Sex is overruled twice and re-imposed twice. That asymmetry, not a flat self-contradiction, is what the text does.
Classification rationale. Filed as primary-text on the same basis as Volumes 1–2 and the three Shiva Purana volumes: the object of study is the Purana. Debroy's contribution is translation and annotation, not argument.
The translator's stated method constrains every citation, and he re-states it here at the worst possible places. From the Introduction: "I have consciously done what is akin to a contemporary Manmatha Nath Dutt translation. Without distorting, I have chosen the simplest possible interpretation or translation that fits." In this volume he re-declares it twice, at the two most doctrinally loaded set-pieces in the book — note 192 on the rasa: "There are various interpretations of Krishna's use of yoga maya. There are also extensive commentaries on these five chapters. We have deliberately kept the translation simple." And note 861 on the Shruti-gita: "The shlokas are subject to numerous and complicated interpretations and we have chosen the simplest interpretation possible." Any page drawing on the rasa chapters or the Shruti-gita must carry this. A doctrinal claim sourced here is sourced to Debroy's simplest-reading English.
No diacritical marks, by deliberate choice. Downstream pages use vault diacritics in prose and quote Debroy verbatim without them inside blockquotes.
No Critical Edition of any Purana exists. The colophon at 12(13) states the Bhagavata "is said to have eighteen thousand" verses; the Introduction to the same volume states this text carries just over 14,000 across 335 chapters. This is not a translator's error — the colophon reports the traditional ascription and Debroy discusses the gap — but the volume does carry its own self-description failing to match the edition printing it.
Volumes 1–2 are separately stubbed and their ranges are separately bounded. Volume 3 makes no claim about what Skandhas 1–9 or Tenth Skandha chapters 1–11 contain. In particular, two exempla cited at 10(72) — "the hunter" and "the pigeon" — are invoked as common knowledge and never narrated anywhere in this volume; whether Volumes 1–2 supply them has not been checked, and no downstream page may claim they are absent from the corpus. [n=2 OBSERVED], bounded to this volume.
The Tenth Skandha chapter-and-shloka table in this volume's front matter lists Chapter 10(85): 59 shlokas and Chapter 10(86): 59 shlokas. Neither is printed. The body runs 10(84) → 10(87) with no break, no ellipsis, and no editorial note. Verified at the raw XHTML level, not merely in a converted text file: the strings do not occur as headings in any body file. The content is genuinely absent rather than merged under a neighbour — standard 10.85 (Krishna's instruction to Vasudeva, the restoration of Devaki's six dead sons, the descent to Sutala) and standard 10.86 (Arjuna's abduction of Subhadra; Shrutadeva and Bahulashva) leave no trace anywhere in the Tenth Skandha files. And the endnote layer cross-references one of them as though present — note 673: "She was married to Arjuna, but that incident is described in Chapter 10(86)." By the book's own count, 118 shlokas are unaccounted for, and the omission is nowhere disclosed.
A second absence of the same species, sixty chapters later. At 11(7) the avadhuta enumerates his twenty-four preceptors — "the earth, the air, space, water, fire, the moon, the sun, the pigeon, the python, the ocean, the moth…" — and promises "I will tell you exactly what I have learnt from each of them." The teaching runs earth → air → space → water → fire and then jumps straight to the python at 11(8). The moon, the sun and the pigeon are never taught. Checked across the whole Eleventh Skandha, not just the chapter: "moon" occurs four times and not one is a teaching passage; "sun" occurs in the 11(7)–11(9) range only inside the enumeration itself. Debroy does not flag it.
⚠ Bounding, and it is not negotiable on any downstream page. Both are claims about this EPUB as ingested. Three consecutive items dropped from a list, and two consecutive chapters dropped from a Skandha, are precisely the signature a defect at any production stage would leave, and it cannot be distinguished from a recension difference using this file. No claim is made about Debroy's Sanskrit or about the print edition. [n=1 OBSERVED] each; [n=2 SUGGESTIVE] on this volume's apparatus disagreeing with its own body. Whether Volumes 1–2 do anything comparable is unchecked and is not claimed either way.
Beyond the two above: 12(13)'s eighteen thousand against the Introduction's 14,000; the Jambavan duel at twenty-eight days in narration (10.56) and twenty-seven in Jambavati's own first-person account (10.83); "ten things" followed by nine at 11(13); the twenty-four preceptors announced, then exceeded by the mountain, the tree and the practitioner's own body; and ten Shunga kings claimed with nine named at 12(1). Only the last is flagged — note 1313: "The name of Agnimitra's father, Pushyamitra, is missing in the text."
⚑ The boundary matters more than the count. A deliberate pass over every announced enumeration in the Eleventh and Twelfth Skandhas found the technical lists correct: the twelve yama and twelve niyama at 11(19), the eight major and ten minor siddhis at 11(15) (with five further powers explicitly placed outside the eighteen by the text itself), and the ten characteristics of a Purana at 12(7). Every one of the seven failures is bibliographic or narrative — chapters, verses, dynastic kings, days of a duel, teachers in a story, inputs to a guna — and not one is a technical doctrinal enumeration. The apparatus transmitting the system is intact; the apparatus transmitting the story and the book is not. ⚠ Stated at the register the evidence carries: this is a distributional observation over one volume in one translation, and an obvious alternative explanation is available and untested — doctrinal lists are short, closed and mnemonically drilled, while narrative counts are long, open and easy to corrupt. The observation is recorded; the explanation is not chosen.
At 12(4) the transmission is given as "In ancient times, the undecaying rishi Narayana told Narada about this Purana samhita and Narada told Krishna Dvaipayana about it" — and note 1354 objects: "However, in Chapter 2(9), we were told Brahma recounted this to Narada." Nine chapters later, 12(13) gives a third chain: "he illuminated Brahma with this lamp of jnana. Brahma passed it on to the sage Narada." 12(4) and 12(13) do not agree with each other, and the note catches only the Skandha-2 divergence, not the local one. A text whose most-repeated claim is the reliability of its own lineage gives that lineage two ways inside one Skandha and three across the corpus.
Volume 2's stub records the long-range cross-reference system as closing — seven forward promises, six paid off inside the volume. That characterisation is correct for Volume 2 read alone and should be enriched rather than corrected. Volume 3's own strongest instance of the system working is Narada's forward index at 10(37), which lists the remaining plot in order fifty chapters early and holds. But the endnotes also record three substantive disagreements with Skandha 9, which is in Volume 2 and already ingested: note 503 (Rajadhidevi is Anakadundubhi's sister at 9(24), his father's sister here at 10(58)), note 512 (Shrutakirti, same species, same chapter), and note 1307 (Puranjaya "has been referred to as Ripunjaya" at 9(22)). ⚠ These are the translator's own observations of divergence between his own volumes, all three in genealogy. No claim about the Sanskrit.
Flagged internal contradictions (6). Note 533 (the parijata: "There is a contradiction with what has been stated in Chapter 10(50)"); 557 (Rukmi's death referenced before it happens — "This is an inconsistency"); 1373 (the Veda-classification cycle: "This should read at the end of every dvapara yuga, not at the beginning"); 500 (Arjuna's weapons — "There is an obvious consistency, since we have been told Arjuna already possessed these"; reads as a typo for inconsistency and is recorded as printed, not silently repaired); 1313 (the missing Shunga king); 1396 (the eighteen-Purana list: "Since Bhavishya Purana has been included, Vayu Purana is missing from the list").
One pre-emptive defence of a contradiction nobody raised — and it is the third in the corpus. Note 741, on the Shalva passage: "There is no such contradiction, because Krishna was behaving like an ordinary human being." The identical behaviour is already logged at Bhagavata Volume 2 note 1058 and Shiva Purana Volume 1 note 1015. [n≥3 PATTERN] across three volumes by the same translator. ⚑ And its placement is the finding: it sits on 10(77), the one passage where the text itself interrupts to deny that God can grieve. The translator adds a defence on top of the text's own defence, of the same claim, in the same place.
One reading taken against the standard one, and it is a caste reading. Note 138, on the pulinda women at 10(21): "There is no reason to take pulinda women to mean women from backward and aboriginal classes, as most interpretations render it." The most consequential disclosed intervention in the volume, and it runs against the majority reading in the direction of removing a caste ascription. Any page touching 10(21) must say the translation declines the standard reading here. Volume 2's equivalent (note 1322, Devapi) reversed who was at fault in a drought.
One flag that a claim in the text does not work. Note 315, on the Yayati curse Krishna cites at 10(45) to refuse the throne: "This doesn't solve the problem, since Ugrasena was also descended from Yadu." The volume's largest political act is justified by a rule the translator points out does not cover the man it installs.
Disclosed additions and liberties (5). Notes 1239 ("We have added it"), 1267 ("we have added the mantra for clarity"), 47 ("we have taken some liberties"), 388 ("We have taken liberties with the text in this sentence, since the meaning isn't obvious"), 597 ("The text says the garments were not white. This is interpreted as blue, because of what follows").
Deliberate non-translations, disclosed (2). Note 433 — shyama left untranslated, glossed as a woman under sixteen who has not borne children, "we have deliberately not translated." Note 784 — dvija retained "since it doesn't necessarily mean a brahmana."
Interpretations offered without textual warrant, and said so (2). Note 631 ("This is interpreted as Satyabhama, but there is nothing in the text to suggest this"); 1038 ("Though not mentioned by name, this seems to be addressed to Rama").
Admissions of unresolvable difficulty (6+). Notes 116, 755 and 756 ("It is impossible to be more specific than that", printed twice identically for adjacent tirthas), 845, 995, 1072 ("extremely difficult to translate… the answer is no, there is no free will" — the translator supplying a philosophical verdict the text leaves implicit), and 128, on the Gopi-gita: the speaker cannot be determined, "whether a single gopi is singing this, whether they are singing this collectively, or whether they take turns. Probably the last."
Arithmetic supplied by the translator (2). Note 1324 (the 1,200-year kali figure "is interpreted as 1200 years of the gods, though the text doesn't explicitly state that"); 1325 (the missing 200 years of the yuga cycle).
What he catches and what he misses is diagnosable. The one repeated block he flags is note 183 on 10(26): "Except the last sentence, Garga's quote is a repetition of shlokas from Chapter 10(8)." Set against two he does not flag — the tirtha block spoken three times, once by Shiva at 12(10), and the wasp figure used once as atrocity and once as liberation — the difference is that the doublet he catches is a quotation of a named speaker that a reader could look up, and the ones he misses are figures recurring in different mouths. [n=1 caught / n=2 missed], stated at that register and no stronger. Every page relying on his notes to establish an absence should carry this characterisation.
One vault finding corrected in place during ingest. The poor brahmana of 10(80)–10(81) — the flattened-rice friend, universally known in the devotional repertoire as Sudama — is unnamed in the body throughout. Note 783 supplies the name and identifies him with the Mathura garland-maker of 10(41): "From Chapter 10(41), we can deduce this was Sudamna/Sudama." That identification is Debroy's deduction, disclosed as such, and is not the mainstream one. Downstream pages must not adopt it as the text's, and any page built on it is evidence bearing on the identification, not evidence of a contradiction in the Purana.
Three standing flags apply corpus-wide and are named on downstream pages rather than re-argued per page.
🚩 MOTIVATED REASONING — structural. This is a sectarian Vaishnava text, and Volume 3 is where the sectarian claim is least disguised: Shiva is defeated by a yawn at 10(63), Vishnu slanders him to an asura in order to rescue him at 10(88), and Bhrigu's three-subject insult experiment at 10(89) is a sectarian verdict reached by controlled experiment with the method printed. ⚑ And the same volume routes three of its best lines through Shiva — satsanga and the tirtha block at 12(10), and the kriya-yoga lineage grounding women's inclusion at 11(27), which rests on "The illustrious Bhava spoke about it to the goddess." [n≥3 PATTERN]. This complicates the flag rather than cancelling it. Pages on sectarian material must hold both.
🚩 TRANSLATION DISTORTION — present and only partly mitigated. Mitigated by Debroy's word-for-word aim and his habit of flagging his own repairs. Not mitigated by his stated method, which selects the simplest reading that fits and says so — twice, in this volume, at the rasa and the Shruti-gita.
🚩 SECONDARY WITHOUT PRIMARY. No Sanskrit is consulted anywhere in this vault's use of this source. Claims are about what Debroy's English says.
Gender and caste material must not be averaged or softened. This volume contains Chamasa's exclusion (11.5); the outcaste attribute list and the varna-from-the-cosmic-body passage (11.17) against "in krita yuga there was only one varna… known as hamsa" in the same chapter; the brahmana-property sermon at 10(64) ("Even if he strikes you a lot and curses you, always bow down before him") against "prostrate before a horse, a chandala, a cow or an ass" at 11(29); the kali-yuga misogyny at 12(3); one entry in the monsoon catalogue comparing lightning that will not stay in one place to "a woman driven by desire"; two named territories called unclean at 11(21); and the mass suttee at 11(31). Against these sit the feeding programme Krishna writes at 10(24), which specifies food "to dogs, shvapakas, chandalas and outcastes" inside the very rite he designs to replace Indra's; the ritualists' wives at 10(23); the eight queens' first-person marriage accounts at 10(83); and the Vedas' own statement at 10(87) that the women reached what the yogis reached. Pages hold the spread rather than resolving it.
Confidence tagging. Claims drawn from the body are claims about what this text says, cited by chapter reference, and carry no web-source confidence tag. Claims about dating, transmission, shloka counts or rival translations drawn from the Introduction are tagged [SCHOLARLY — translator's introduction].
Evidence tagging for structural findings. Assertions this vault makes about the text — recurring formulae, repeated blocks, pattern counts — carry [n=1 OBSERVED], [n=2 SUGGESTIVE] or [n≥3 PATTERN], and no rigour-register language below n=3.
Absence claims. No page may assert what this volume does not contain outside ranges actually read, and no page may extend an absence claim to Volumes 1–2, which were not re-read for this ingest.