Bhagavata Purana

The Bhagavata Purana is scripture to four traditions that agree on little else. The Gaudiyas of Bengal, the Pushtimarg of Vallabha, the Madhvas and the Nimbarkis all treat it as supreme, and all read it differently. Nobody can say when it was written.

What it is

Twelve skandhas, books, and about three hundred and thirty-five chapters. Its subject is Vishnu in his descents, and above all Krishna, whose life fills the tenth book: ninety chapters and 3,938 verses, close to twenty-eight percent of all the verses in the Purana.

The chapter total is itself unsettled, and the disagreement runs through the commentators. Jiva Gosvamin counted 335, which is the figure in K. S. Shastri’s edition and the one a modern reader will meet. Jonathan Edelmann reports that Shridhara Svamin, Madhva and Vallabha accepted 332, rejecting chapters 13 to 15 of book ten. Amarnath Roy, whom Ludo Rocher cites in a footnote as an example of how far some scholars will lean on what Puranas say about Puranas, reports it differently again: Shridhara commented on 335, the Padma Purana says 332, and Roy therefore went looking for three spurious chapters and found 1.3, 6.8 and 12.1. Edelmann and Roy contradict each other about which chapters are in question and about what Shridhara himself counted. C. V. Vaidya had set out the same comparison seven years earlier, in a passage Sheridan quotes, reporting that the three extra chapters had been pointed out without naming them. Rocher gives 335 himself in his own survey of the text, and prints the chapter-count for each of the twelve books in a footnote. The Ahmedabad critical edition, which excises later material, reports 329.

The tenth book is also the part most readers meet, and usually the only part. A widely available scholarly translation is Edwin Bryant’s Penguin volume, and it is book ten with chapters 1, 6 and 29 to 31 of book eleven: five chapters of a thirty-one-chapter book. The Oxford Centre for Hindu Studies runs a research project on this text which states its intention “to correct this exclusive focus on the tenth book,” and notes that the flood, the churning of the ocean and Narasimha are all elsewhere. So is the long Samkhya instruction Kapila gives his mother Devahuti in book three. So is the Uddhava Gita, Krishna’s last discourse, which occupies most of book eleven and is far less read.

The number in its own last chapter

In its final chapter the Bhagavata measures itself. It is reciting the standard table of the eighteen Puranas and their extents, and it closes the table with its own share:

evam puranasandohash chaturlaksha udahritah tatrashtadasha-sahasram shri-bhagavatam ishyate

The whole mass of the Puranas is declared to be four hundred thousand; of these the Shri-Bhagavata is held to be eighteen thousand. That is 12.13.9, and the figure stands four verses earlier inside the enumeration itself. Edelmann cites the claim as 12.13.3, which is the verse that announces the enumeration.

Count the stanzas and there are about fourteen thousand. Counting the distinct verse numbers in the GRETIL electronic text gives 14,067. Counting the verse numbers in the Bhaktivedanta text on vedabase.io, chapter by chapter across all 335 chapters, gives 14,090. Edelmann, in Brill’s Encyclopedia of Hinduism, states flatly that the text “is really 14,264.” Tagare, in his translation of the Narada Purana, records the Gita Press edition at 14,579. The catalogue description of the critical edition reports 13,129 stanzas and 452 prose passages, though that is a reconstruction which removes later matter and so is not measuring the same object as the others.

The resolution is not that the text is lying, and it is worth following, because the error it corrects is still in circulation. A traditional shloka count, a count of verses in the old reckoning, is not a count of stanzas. It counts granthas, the unit manuscript cataloguers use: thirty-two syllables, the length of a standard anushtubh stanza. Edelmann reports V. K. Chaturvedi’s calculation on that basis: about 17,998. The same answer had been reached long before by Gangasahaya, who wrote the Anvitarthaprakashika, and who is reported to have converted the whole book into anushtubh units three times, coming out a verse and a half short. That detail reaches us through the Gaudiya commentarial literature, Satyanarayana Dasa and Kundali Dasa’s edition of the Tattva-sandarbha, rather than from Gangasahaya himself, so what survives is a report of his count and not the count. The same page records that he counted the three disputed chapters in, and that without them, and without seven verses of 10.6, his total falls about two hundred and ten anushtubh units short.

So the figure is not simply wrong. It is in a unit that was forgotten. The reconciliation is also not exact. Count the syllables of the GRETIL electronic text and divide by thirty-two and the answer is a little over seventeen thousand; deduplicate the half-lines GRETIL reprints in the longer metres and it falls by about four hundred and fifty; while the chapter-end colophons GRETIL omits, which the method that source describes expressly includes, would add a few hundred back. So the grantha reading closes most of the gap and not the whole of it. Say the Bhagavata has eighteen thousand verses and you are wrong; say the text overstates itself and you have missed what the number counts.

Rocher adds a detail in a footnote. Across the Puranic lists he collected, “only the Bhagavata appears consistently with 18,000, and the Linga with 11,000 (but it is omitted at Vayu ch. 104).” Most of the genre’s extent-claims wander between lists. Two barely do. The footnote is attached to Rocher’s own argument that the compilers of these lists were working towards a round total of four hundred thousand and allowed themselves “a considerable degree of latitude” in the details, and it is there to mark the Bhagavata as one of the two exceptions to that latitude. The unique consistency is a fact about the lists. What it does not by itself establish is what the figure was counting.

When it was written, and why nobody can prove it

The spread of scholarly opinion is not narrow. Rocher’s table of proposed dates runs from the twelfth century before the Common Era to the thirteenth after it. The two proposals at the early end are outliers, and Tagare calls the earliest of them untenable, on the ground that the Purana’s language is far more modern than Vedic. The proposal at the late end is an outlier too. It belongs to those who thought the book was written by the grammarian Vopadeva; Sheridan calls that theory untenable, Farquhar had already argued it was impossible, and Rocher reports that it has generally been rejected. Name both ends as outliers and the live proposals in Rocher’s table still run from about 200 to about 1000, eight centuries, with Sheridan putting his own limits at 500 to 1000. The arguments are worth more than the verdict.

The one external reference the live proposals all accept is al-Biruni’s survey of India, around 1030. It carries less than it is usually made to carry. He gives two lists of the Puranas. The first, taken down from dictation, does not contain the Bhagavata at all, and he closes it by saying he has himself seen portions of only three Puranas: the Matsya, the Aditya and the Vayu. The Bhagavata appears only in his second list, which he says was read to him out of the Vishnu Purana, standing fifth and glossed “i.e. Vasudeva.”

What it proves is read at very different strengths by serious people, and in the same direction. Friedhelm Hardy takes the list entry to mean the Purana “had in some form already gained prestige as a maha-purana” by 1030, and stops there, warning that it “does not necessarily mean that certain passages could not have been added later.” R. C. Hazra pushes the same inference much further back: the gloss shows that rival claims to the name had already begun, so the text’s position must by then have been long established, and it “cannot possibly be later than 800 A.D.” One datum, two inferences at different strengths, and a terminus rather than a date.

The early anchors, and the manuscripts that broke them. Abhinavagupta, in Kashmir around the turn of the eleventh century, appears to quote the Bhagavata in his commentary on the Gita. F. O. Schrader reported it in 1930 and hedged it in the same breath, “if authentical”; Hardy noted in 1983 that neither Schrader nor van Buitenen had given a reason for the doubt. A reason existed. Hazra had printed one in 1940, reporting an observation made in ABORI XV, 1933 to 1934: in an old manuscript in Sharada script the verses stand on the margin rather than in the running text of the commentary. Hazra’s verdict is flat: the verses “are not reliable.” He is careful only about the mechanism. It “may be,” he writes, that a marginal note was absorbed into the text by later copyists. A verse of the Bhagavata quoted in the Uttaragita-bhashya, a work ascribed to Gaudapada, fails on two grounds, and Hazra gives them in that order. The verse is absent from a manuscript that appears to preserve an older version of the bhashya. And the bhashya is demonstrably open to later addition, since another manuscript of it carries a Tantra quotation that stands in no printed edition. Hardy and Tagare press a separate objection, that the work is probably not by Shankara’s teacher’s teacher at all. Two parallels in the Mathara-vritti fail differently again: they are missing from Paramartha’s sixth-century Chinese version of that text, so the Sanskrit cannot be used to date the Bhagavata earlier.

And here is the part that changes how the demolitions read. The scholar who assembled those three demolitions was not clearing the ground for a late date. Hazra dated the Bhagavata to the first half of the sixth century, and held that Ramanuja was not unacquainted with the text. For that he pointed to an article of B. N. Krishnamurti Sharma’s in ABORI XIV, not to any quotation of his own. Tagare later put the claim concretely, by naming a Bhagavata verse quoted in Ramanuja’s Vedanta-tattva-sara. That ground does not hold. Hardy showed in 1983, answering a different writer who had leaned on it, that the Vedanta-tattva-sara opens by paying obeisance to Venkatanatha, whose dates are 1268 to 1368, and is the work of a later Shrivaishnava called Ramanujacharya. Sheridan’s flat statement further down, that there are no references to the Bhagavata in Ramanuja, stands. Hazra built his positive case from the Kurma Purana, from a contents-list in the Narada Purana that matches our text, and from the observation that the Bhagavata seems familiar with ajatavada, the doctrine that nothing is ever really originated, while showing no awareness of the anirvacaniyata that Shankara went on to posit. He discarded all three of those early anchors while holding to an early date, and counted the last of them among the evidence that confirmed it.

The earliest anchor anyone proposes is Jain, and it turns on how a list is read. The Nandi-sutra, whose Valabhi redaction belongs to the fifth or early sixth century, runs through the learning of outsiders. Tagare quotes the roll in abridged form, ending Shashtitantra, Mathara, Purana Bhagavata, Patanjali and an “etc.”, and he finds it significant that the Bhagavata is mentioned there at all, at a stage when the Mahabharata was still simply the Bharata, and that it stands after the Purana. The list he prints is silently abridged in the middle too. He gives thirteen of the first twenty items and drops seven, and one of the seven is the one that matters. In the Prakrit, grammar, vagarana, stands between the Purana and the Bhagavata, so the adjacency is not there to argue from. Ghasilal’s edition prints the item as bhagavayam at sutra 41, Umaravkunwar’s as bhagavam at sutra 42. And a bare Bhagavata sitting in a run of system names, the Shashtitantra a few items earlier and Patanjali immediately after, could as easily mean Bhagavata or Pancharatra sectarian literature as our text. That Mathara may well be the same Mathara whose vritti, its running commentary, appears two paragraphs above. Ghasilal’s own commentary cuts against it, glossing the item as a Nyaya work establishing the sixteen categories. Everywhere else the Mathara-vritti is a Samkhya text, so the gloss is probably his slip, but it is his edition this paragraph has been leaning on.

The late case rests on the Tamil south. Hardy argued in detail that passages of the Bhagavata are translation-paraphrases of the Tamil poems of the Alvars, and that the text is preoccupied with the Pandya country. His own conclusion was “the ninth or earlier tenth century,” and he was candid that his dating of three of the four Alvars concerned “rests on less firm grounds than one would wish.” The familiar figure of 850 to 900 belongs to Thomas Hopkins. The whole argument also turns on a premise the other side disputes, which is the direction of borrowing.

The hardest fact for that case is a silence. Daniel Sheridan states it plainly: there are no references to the Bhagavata in Ramanuja, and none in Yamuna, whose dates he gives as 918 to 1038. Two major Shrivaishnava theologians of the next three centuries, working in the same country, never cite it. Van Buitenen said the reticence “needs to be explained,” and in the same passage that arguments from silence “are never conclusive.” Sheridan, having stated the difficulty, still settles on around 900 and a southern origin.

Two more positions belong on the table. Van Buitenen is better known in this debate for an argument about the language: the Bhagavata’s Sanskrit deliberately archaises towards Vedic, which he called a unique phenomenon, a text consciously attempting to archaise at a moment when Sanskrit letters were in decline. Van Buitenen’s own reading is that the archaising is conscious. Tagare would regard the same archaic passages as the text’s oldest layer. And Edwin Bryant questions both the southern provenance and the late time frame, and wonders whether the text might not have reached its final form by the Gupta period.

The firmest later ground belongs to Madhva, and it is epigraphic rather than hagiographical. Dasgupta reports Bhandarkar’s argument from the inscriptions of Madhva’s disciple Narahari Tirtha. One at the Shrikurma temple is dated shaka 1203, which is 1281, and others put Narahari’s activity between shaka 1186 and 1215. On that reading Madhva cannot have died as early as the monastery lists imply, and their date must instead be his birth: 1197 to 1276. Others, B. N. K. Sharma among them, read the same material for 1238 to 1317. The inscriptions are not in dispute. What the two camps read differently is the monastery list. His Bhagavata-tatparya-nirnaya, a digest rather than a running commentary, is the earliest surviving exposition of the text by an author whose dates can be argued from anything but tradition.

What the tradition holds

That the book is Vyasa’s own, received in samadhi, absorbed contemplation, and that it is the akritrima-bhashya, the uncontrived commentary, on his own Vedanta Sutra, which is the Brahma Sutra under its other name. Jiva Gosvamin set it out in the sixteenth century in the Tattva-sandarbha, and the argument is about authority before it is about chronology: if the Vedanta Sutra already carries a self-established commentary by its own author, then every other commentary on the Brahma Sutra is arvachina, recent, each made up out of its author’s own head, and worth regarding only so far as it follows the Bhagavata. Jiva names none of them.

The claim has a foundation in the text, and it is easy to check. Brahma Sutra 1.1.2 reads janmady asya yatah, “from which the origin and so forth of this.” The Bhagavata opens:

janmadyasya yato ‘nvayad itaratash cartheshv abhijnah svarat

Those same three words open the Purana, compounded and carried into sandhi as janmadyasya yato. Whatever one concludes about the date, the book announces on its opening line what it intends to be.

The commentators, and the awkwardness at the centre

Madhva wrote before the commentators the schools came to use, and Tagare still calls Shridhara the oldest of the Vedanta annotators he surveys, while taking Shridhara’s own wording to imply a Bhagavata commentary by Citsukha, whom he dates 1220 to 1284. What Madhva wrote Tagare describes, in passing and while discussing someone else, as “not a Commentary but a digest of the Bhagavata Purana.” The commentary that mattered most to what came after is the Bhavarthadipika of Shridhara Svamin, whom P. K. Gode placed between about 1350 and 1450. Shridhara leans Advaitic. Jonathan Edelmann places the writing of the commentary in the Govardhana Pitha, a monastery at Puri he traces to Shankara, and Ravi Gupta reports that Edelmann and Ananta Sukla, following earlier writers, take Shridhara to have been abbot of an Advaita monastery there, while adding that his institutional affiliation is still a question requiring further historical research. Farquhar had been flatter a century earlier, calling him mahant of that monastery outright in 1920. Tagare records a separate legend that has the commentary approved at Kashi.

So Jiva Gosvamin and the Gaudiya commentators after him built their exegesis on a commentator from the school they spent their lives arguing against. Jiva presents his own commentary as no more than an elucidation of Shridhara’s, and Ravi Gupta finds that he follows Shridhara’s reading of almost every verse he quotes, a dependence Stuart Elkman reads less strongly. Tagare adds, without naming them, that commentators of other Vedanta schools “bodily incorporate his Commentary into their own.” And the friction shows. The Chaitanya-charitamrita has Vallabha Bhatta tell Chaitanya that he has refuted Shridhara’s explanations and cannot accept them. Chaitanya’s answer is blunt even in translation. Later in the same episode he tells Vallabha to produce his commentary on the Bhagavata “following in the footsteps of Shridhara Svami.”

They each found room. Vallabha wrote the Subodhini, Jiva Gosvamin the Kramasandarbha, Vijayadhvaja the Padaratnavali, Vishvanatha Chakravartin the Sararthadarshini, and for the Nimbarkis Shukadeva the Siddhantapradipa. The text allows it. One of its most contested verses, 1.2.11, says that those who know call the one reality brahman, paramatman and bhagavan, three names for a knowledge without a second, and the schools read that sequence differently; Sheridan takes the final position of bhagavan to raise it above the other two.

What is not in it

Radha.

The text that is the scriptural ground of the traditions most devoted to her never names her. The gopis, the cowherd women of Vraja, are there. So is the rasa dance, the circle dance of book ten. One unnamed favourite is singled out and then lost in the forest.

That episode is where the argument lives. At 10.30.28 the others say of her anayaradhito nunam bhagavan, by her the Lord was surely propitiated, and whether the name Radha is concealed in that aradhita has been asked for a century. Winternitz answered it yes in his Geschichte der indischen Litteratur and no in the English History of Indian Literature that grew out of it. What can be said without taking a side is narrower, and still worth saying: the name as a name is absent. Across all twelve books of the GRETIL electronic text the string radha, with both vowels long, occurs eleven times, every one of them inside a longer word and every one of them something else. Seven are forms of pradhanika or pradhanya, principal or predominant, two of them carrying the strict sense of pradhana, primal matter: one at 3.26.11 in Kapila’s Samkhya instruction, and one at 12.12.7 in the closing table of contents, where it names the creation account of book two. One is anuradha, a lunar mansion. Two are places where a word ending in r runs straight into one beginning adha, again with both vowels long. The last is paradhavan, running away.

Radha reaches the tradition through other texts, and the devotion that reaches its height in Bengal reads her back into a book that, at the very least, never says her name.

For the god at the centre of it, see who is Krishna. For the devotion this book did more than any other text to shape, see what is bhakti. For where it sits among the eighteen, see what are the Puranas. For how a corpus this size survived without a library, see the unwritten library.