A genre that was always bilingual

The Śābara corpus says of itself that it is old. Its charter puts the origin with Śiva, by whose grace a teacher received the mantras and gave them out to five pupils in five languages, for an age that needed rescuing. That is the tradition’s own account of where it comes from, and this chapter takes it seriously enough to test it.

Ask the archive the same question and the archive answers in a different register entirely.

Every copy of this material that anybody has catalogued is paper. Not one palm leaf appears in any description this chapter has read, in any collection, under any of the titles. What the cataloguers wrote down instead was country-made paper, a modern Nāgara hand, and, of the one complete copy of the charter text, that its appearance is fresh.

Fresh. Of the foundational book of the corpus.

There is exactly one explicit date in the whole Śābara block of the Asiatic Society’s catalogue, and it is a date of copying rather than of composition. A scribe finished his copy in Saṃvat 1912, which is 1855 or 1856 of the Common Era. The earliest hard dates anywhere near this corpus are that copying date and a Calcutta imprint of 1884, and the older of the two is younger than the electric telegraph.

Print runs the same way. The earliest imprint on record for any of this material is that Calcutta printing of 1884. After it come Cawnpore, Moradabad, Bombay, Delhi: the bazaar presses, cheap ritual manuals for a mass readership, sold in a channel the universities were not looking at.

And the copies are everywhere. Something with a Śābara title on it sits in about forty repositories, in something like a hundred and ten to a hundred and twenty manuscript items. Those items are catalogued under about twenty-three distinct headwords, which probably stand on five or six actual works, and chapter five sets that census out work by work. The charter text alone has been catalogued in Pondicherry, Lonavla, Chennai twice over, Varanasi, Trivandrum twice, Ujjain, Wai twice, Kolkata and Thanjavur.

Nobody has edited any of it. There is no critical edition of a single one of these texts, and so far as any source reached here records, no modern writer on the Śābara corpus has seen the one complete manuscript of its charter.

So here is the object at the centre of this chapter. A tradition that claims descent from Śiva, surviving on paper, in modern hands, scattered across forty collections, its oldest dated leaf written in the 1850s and its oldest printed edition sold out of a Calcutta press in 1884. None of that makes the texts young. A copy is not a composition, and an archive measures only itself. What it does mean is that everything anyone knows about this corpus is known from objects made recently, and every argument in this chapter has to be read in that light.

One paragraph in one European survey carries the formal description this whole corpus is usually read through. That paragraph is where the chapter has to start.

The paragraph everything rests on

In 1981, at Wiesbaden, the house of Harrassowitz published a volume by Teun Goudriaan and Sanjukta Gupta called Hindu Tantric and Śākta Literature, written for Jan Gonda’s History of Indian Literature. It is generally treated as the standard catalogue of the Tantric corpus in a European language. That is a conventional judgement and this chapter has not sourced it.

After the sections on the magical Tantras that precede it, a short paragraph opens a new group and runs over onto the following page.

Three things have to be said about that paragraph before the chapter uses it, and all three belong at the top rather than buried further down.

The largest is stated first. Between that 1981 paragraph and this chapter’s reader stand two catalogues: the Asiatic Society’s Tantra volume of 1939 and 1940, which describes the Śābara manuscripts leaf by leaf, and the University of Madras union catalogue, which lists every recorded copy of every one of them. Both have been read at page level for this chapter. What they show is set out in two sections below, and the short form is this: the text this book builds its charter on is extant, complete, in a public collection in Kolkata.

The paragraph is also not the whole definition of the category, because Elwin defines it by function and chapter five is built on that. And it is not the beginning of the scholarship. Indian scholars had this material in print decades before that paragraph was written. S. B. Dasgupta printed the Śābaratantra’s twelve-and-twelve Kāpālika list in English, from the Gorakṣasiddhāntasaṃgraha, in Obscure Religious Cults of 1946. Hazārīprasād Dvivedī printed it in Hindī, with both lists in a footnote, in Nāthasampradāya of 1950. Earlier than either of them, P. C. Bagchi named a Śāvara Tantra and its Kāpālika lists in the introduction to his edition of the Kaulajñānanirṇaya in 1934. The Indian scholarship that reached this material first is set out below.

What the catalogue actually says

The paragraph is short, and it is worth taking whole rather than in the half that gets quoted. It opens a group of its own: few texts known to belong to it, contents seldom running beyond enumerations of incantations, and those incantations often in Sanskrit alternating with Hindī. Then Goudriaan floats a possible connection between the term Sāvara and the generic name of the Saṃvara or Śambara Tantras of Buddhist Tantrism. He warns that the titles are confused and may occur one after another in the same manuscript. He reports Kāśīnātha Bhaṭṭa, an eighteenth-century author, trying in his Kāpālikamatavyavasthā to show a historical relation between the Śābara texts and the Kāpālikas, and judges there is more reason to connect them with the Nātha cult. And he names the group’s two main subjects: mastery of the mantra of Śābara, a minor deity of magical potency, and mantras of a variety of other gods such as Yoginīs, Kṣetrapāla, Gaṇeśa and Kālī, for magical purposes.

Every one of those points is worked below. The volume is in copyright and nothing in it is reproduced verbatim in this book.

One paragraph, inside a survey covering the whole of the Tantric and Śākta literature, is the entire description of a body of writing that has circulated in Indian print for well over a century and in Indian manuscript for longer. The genre is catalogued. It has a name, a shape, a handful of titles and a charter myth of its own. It is also, by the standards of Tantric studies, almost unexamined.

Three criteria, and one reading of them

The definition gives three criteria. Language. Deity. Purpose.

The corpus is not defined by register alone, by the language the words are in and by nothing else. The catalogue names the deity and it names the purpose, and it names them as the two main subjects of the texts.

What can be argued is narrower, and this chapter states it as an interpretation and not as a finding, because the source does not state it.

This treatise’s reading is that the language criterion is the one doing the distinguishing work. The reasoning is internal to the catalogue’s own arrangement. The magical Tantras immediately preceding this group in the same survey share the deity criterion and share the magical purpose; the Uḍḍīśa and its neighbours are texts of magical practice addressed to a spread of deities. What the survey gives the Śābara group that it does not give the group before it is the bilingualism. On that reading, register is what marks the boundary in this particular catalogue, at this particular page.

Goudriaan says none of this. He gives three features and nowhere identifies one as distinguishing. Nor is the reading propped up here with a generalisation about how the other divisions of the Tantric corpus are drawn, along doctrinal or sectarian lines: a claim of that size about the whole Tantric literature would need a source, and this project has none.

The popular literature meanwhile says something different again. In the retellings that circulate now, śābara mantras are the blunt ones, the commanding ones, the ones that work without initiation and without purity rules, the folk instrument set against the Vedic instrument set. On the evidence of the paragraph itself the functional half of that is close to the catalogue: the catalogue does say these are texts for magical purposes. What the catalogue does not supply is the opposition, the folk-against-Vedic frame, which is a modern addition.

The etymology, laid out

This is the chapter that owns the question, and it is going to be laid out and not smoothed. There are three proposals in the record and one of them is this treatise’s. Each gets its evidence, then the commitment, then what would overturn it.

Proposal one: the name is a grade of speech. Śābara is a lexical item meaning a low form of Prākṛt, spoken by mountaineers; it stands in the Nāṭyaśāstra’s list of stage vibhāṣās; and on Thapar’s account the classical outsider-vocabulary is first of all a vocabulary about language. A corpus of incantations that alternate Sanskrit with Hindī would then be named for what it sounds like. Chapter two builds this at length. The proposal has one thing the other two do not: it explains the catalogue’s own bilingualism criterion without needing any further step. It also now has a traditional witness. Mānas-pīyūṣa derives the name from सँवर, the word for Bhīl, and says Śiva composed the mantras in the Bhīl language. That is an Indian commentator making the ethnonym-to-speech derivation outright, though he makes it as a traditional etymology and not as historical philology.

Proposal two: the name is Buddhist. Goudriaan suggests, in the same paragraph, that the term Sāvara may be connected with the generic name of the Saṃvara or Śambara Tantras of Buddhist Tantrism. Saṃvara is a Buddhist deity, the Cakrasaṃvara of the Yoginī Tantras, and the saṃvara corpus is a large and well-studied body of Buddhist Tantric literature. If the connection holds, the name of this Hindu group is a borrowing from Buddhist Tantrism and says nothing about speech, forests, or the Śabara people. It is the only etymological proposal in the source, and a book that declines to follow it is obliged to carry it in full.

Proposal three: the name is a deity’s. This one comes from the same paragraph and is the easiest of the three to walk past. The catalogue says the first main subject of these texts is mastery of the mantra of Śābara, a minor deity of magical potency. If a corpus is named for the deity whose mantra it teaches, which is ordinary practice across the Tantric catalogue, then the Śābara Tantras are the Tantras of Śābara in exactly the way the Kālīśābaratantra is a Tantra of Kālī. That would make the register a consequence of the deity’s character rather than the source of the name. It is a live proposal and it has not been argued in print by anyone this project has read.

The three are not fully exclusive. A minor deity called Śābara could himself be named from the forest people or from their speech, in which case one and three are the same story with a step in the middle.

What would decide it. Three tests, and none of them is in hand.

The first is the direction and date of the borrowing. If Śābara-titled Hindu texts can be shown to postdate the circulation of Saṃvara literature in the same regions, proposal two gains; if the Śābara title is attested earlier, it loses. The last section of this chapter comes to a ninth-century Śābaram in the Gāruḍa canon which would be relevant if it could be shown to be the same word doing the same job, and chapter five says it cannot. That cuts both ways and this treatise will not use it selectively: a title this book calls uninterpretable when it would help the Nāth story cannot become interpretable when it would hurt the Buddhist one.

The second is phonology. Saṃvara to Sāvara is a plausible slide in Middle Indic, and so is śabara to sāvara. Somebody with the historical phonology at their fingertips could say which is the easier road, and this chapter cannot.

The third is content. If any Śābara text shows Saṃvara material, a Buddhist deity, a Buddhist maṇḍala, a Cakrasaṃvara formula surviving under a Hindu title, that would settle it in one stroke. Nothing in the sources reached here reports such a check having been made, and the texts are unedited.

What this treatise follows, and why. It follows proposal one. The reason is the fit between the name and the catalogue’s own primary criterion: the group is described as one whose incantations move between Sanskrit and Hindī, and there exists a Sanskrit word meaning precisely a non-Sanskritic grade of speech, in use as a technical term in the dramaturgical tradition for centuries before any of these texts. That is one step. Proposal two requires a borrowing whose route and date nobody has traced. Proposal three requires a deity nobody has otherwise documented.

What would overturn it. Any one of four things. A Saṃvara or Cakrasaṃvara element found inside a Śābara-titled text. A historical phonologist showing that Saṃvara to Sāvara is regular and śabara to sāvara is not. An independent attestation of Śābara as a deity in Tantric sources unconnected with this corpus, which would make proposal three the economical one. Or a Śābara-titled text securely dated before the Saṃvara literature reaches North India, which would not prove proposal one but would kill proposal two.

Goudriaan offers his connection as a possibility and does not argue it. The commitment above is this treatise’s, made on stated grounds, and it is a commitment and not a proof. Every later chapter that runs the register thesis carries a sentence naming what it loses if proposal two holds.

Five languages, named

The next page carries the charter.

On the second of those two pages Goudriaan describes a text called the Pañcadrāviḍaśābaracintāmaṇi, ascribed to Ādinātha, which contains vernacular versions of mantras to Kālī and other deities, those vernacular versions being said to have been revealed by Ādinātha’s five pupils. Ādinātha, the catalogue adds, is the figure known as Matsyendranātha’s teacher. Ādinātha obtained the mantras by Īśvara’s grace and proclaimed them to his pupils in the languages of Gauḍa, Kerala, Karṇāṭa, Andhra and Gurjara.

That is the sentence. Five vernaculars, named one by one: Bengal, Kerala, Karnataka, Andhra, Gujarat.

Three things this chapter must say before it says anything else about it.

The first is that this is the one place in the whole chapter where the foreign source is kept because there is no Indian equivalent for the entry itself. Dvivedī and Dasgupta both give the Ādinātha and Śābaratantra material earlier and, on the Kāpālika list, better. Neither of them gives the (Pañcadrāviḍa)śābaracintāmaṇi or its five named vernaculars. Goudriaan’s own footnote for it is a bibliographical pointer and nothing more: a catalogue entry number, and the remark that the catalogue prints the text’s opening.

The second is that pañcadrāviḍa is not a coinage of this text. It is the received brahmanical territorial classification, the pañcagauḍa and pañcadrāviḍa division of the brahmin communities of India, and it is a scheme of regions rather than a scheme of language families. Read the list Goudriaan actually prints. Gauḍa is Bengal. Gurjara is Gujarat. Neither is southern, and neither is Dravidian. So the charter’s five are neither five southern communities nor five Dravidian vernaculars. What the charter claims is a spread across the subcontinent, from Bengal to Gujarat to Kerala, and a pan-regional claim is a larger claim than a southern one.

The third is the one that changes the chapter, and it is set out in the next section.

Now the reading, and it is marked as a reading.

The usual origin story attached to this material, in the form it reaches a reader today, is a single sentence: Śiva made these mantras for the Kali age, for people who could not manage Sanskrit. It is a story about condescension, and it flattens the corpus into an inferior grade of scripture.

This treatise reads the charter as a story about distribution instead. One revelation, received once, given out through five pupils in five named language regions. On that reading vernacularity is a design feature of the transmission rather than a fall from Sanskrit.

Goudriaan’s sentence does not say that. It says Ādinātha got the mantras by Īśvara’s grace and proclaimed them to his pupils in five named languages. What is on the page is a revelation and five languages. The deliberateness, the fracture into regions, the refusal of a fall from Sanskrit: all of that is this treatise’s reading, it stays out of the declarative voice, and it is labelled above as this book’s.

Whether the design was ever realised is a separate question, and it is half answered. Of the three southern languages named, Kerala, Karṇāṭa and Andhra, two have manuscript evidence behind them. The Madras catalogue of Sanskrit manuscripts made from the old Fort St George collection describes a Śābaracintāmaṇi with some Telugu prose in it; the Tanjore catalogue describes a Śābararahasyamantra interspersed with Kannada. Andhra and Karṇāṭa, in the manuscripts, under Śābara titles. Kerala and Gujarat are still open, and the bulk of the datable corpus still sits in the Hindī belt, so the charter’s claim of a five-language spread is half evidenced and no more.

And the charter has a lineage attached. Ādinātha, Matsyendranātha’s teacher, is the origin point of the Nāth sampradāya. The charter of the Śābara corpus is a Nāth charter.

That clause about Ādinātha is a one-line catalogue remark in Goudriaan, footnoted to S. B. Dasgupta’s Obscure Religious Cults. It can be put on much stronger ground, and the stronger ground is Indian and earlier. P. C. Bagchi, in the introduction to his 1934 edition of Matsyendranātha’s Kaulajñānanirṇaya, documents the same identification from four independent Indian traditions. In the Haṭhayogapradīpikā’s list of Mahāsiddhas, Matsyendra stands second only to Śrī Ādinātha, whom Bagchi glosses as Śiva. In the Jñāneśvarī, the Marathi commentary on the Gītā that he dates to about 1290 of the Common Era, the succession runs Ādinātha, Matsyendra, Gorakṣa, Gahinī, Nivṛttinātha, the last being Jñānadeva’s elder brother and teacher. Bahiṇā Bāī, the Marathi saint who lived from 1628 to 1700 and was a disciple of Tukārām, gives a comparable guruparamparā. And Kabīr, at the turn of the fifteenth century, names the eighty-four siddhas in his dohās and names Gorakh and Bharthari among them, in the Nāgarī Pracāriṇī edition. Dvivedī, in Nāthasampradāya, states the same thing in one line: Ādinātha is simply another name for Śiva, and Matsyendranātha is the first human teacher of the lineage.

Sanskrit haṭhayoga, Marathi bhakti, Nāth hagiography and Hindī sant poetry, with dates, forty-seven years before the catalogue. That is what the clause is worth, and it is not worth it on Goudriaan’s authority.

One divergence about the Śābaracintāmaṇi can be located exactly. David Gordon White’s The Alchemical Body of 1996 mentions the title. That the title appears there is all this chapter claims from White, because it is all that has been verified from the page images available to it. Reports of what he says there give the work to Nityanātha Siddha and trace its teaching to Matsyendra himself, which sat oddly beside the catalogue’s ascription of the Pañcadrāviḍaśābaracintāmaṇi to Ādinātha, who is Matsyendra’s teacher.

The union catalogue answers it, and the answer is that the divergence is in the manuscripts. The New Catalogus Catalogorum, the union catalogue of Sanskrit works and authors, enters the work as by Ādinātha alias Siddhanātha, and records in the same breath that some colophons ascribe it instead to Nityanātha alias Nityanāthasiddha. Two attributions, both on the copies, neither invented by a modern scholar. White read one; Goudriaan reported the other; the scribes had already disagreed. That attribution matters beyond bookkeeping, because Nityanātha is the author of the Rasaratnākara, which drags this text into the rasa-śāstra and siddha orbit rather than leaving it purely in the mantra literature. The divergence is no longer a divergence between scholars. It is a divergence inside the tradition, which is a different and more interesting thing.

The text exists, and it is in Kolkata

Here is the third thing, and it is the largest.

The flat question to ask first about any text is whether the book survives, and if so, where. A title in a footnote, reached through what a European survey says a manuscript catalogue says, is not an answer to it.

It survives. It survives in roughly fifteen manuscript witnesses across about twelve repositories, and one of them is complete, described, and sitting at 1 Park Street in Kolkata.

The Descriptive Catalogue of the Sanskrit Manuscripts in the Collections of the Royal Asiatic Society of Bengal, in its Tantra volume, was begun by Mahāmahopādhyāya Haraprasāda Śāstrī and revised and edited by Chintaharan Chakravarti, printed at Calcutta in two parts in 1939 and 1940. Its Śābara manuscripts fill a run of consecutive entries, and the last of them is the one this treatise wanted.

The Śābaramantracintāmaṇiḥ. Country-made paper. Modern Nāgara. Appearance fresh. Complete in ten chapters.

Read that again, because the phrase in it that matters is the last one. Complete. Not a fragment, not two paṭalas of twenty-two, not the “inc.” that stands beside most of this corpus wherever it is catalogued. A whole text, in ten chapters, on country-made paper, in a public collection, catalogued in print in 1939, and open to any reader who writes to the Society and asks.

The catalogue prints the opening of the text. This book records that the opening exists and where it is printed, and does not reproduce it. That is the operative line this treatise keeps throughout, and it does not bend for a discovery.

No critical edition of the Śābaracintāmaṇi exists. None was found in any scholarly series. A commercial Sanskrit-and-Hindī edition of seventy-six pages is currently in print from a Varanasi house under a compound title; at that extent it cannot be a full presentation of a ten-chapter text of this length with any apparatus, and until a copy is opened nothing is claimed from it. What would close this: the Chowkhamba volume in hand, and its contents compared against the Asiatic Society’s description of the manuscript.

Where the other fourteen are

The New Catalogus Catalogorum, the University of Madras union catalogue of Sanskrit works and authors, gives the roll, in the volume published in 2014. Fifteen references stand under the headword. Some carry the bare title Śābaracintāmaṇi; some carry the fuller Pañcadrāviḍaśābaracintāmaṇi or Siddhaśābaracintāmaṇi with the attribution; several are marked incomplete; one of them is the Asiatic Society entry read above. The addresses are the thing worth reading.

Pondicherry, Lonavla, Chennai twice over, Varanasi, Trivandrum twice, Ujjain, Wai twice, Kolkata, Thanjavur. That is where the charter of this corpus has been sitting all along.

What the record looked like in 1891

One pre-1929 catalogue can be quoted rather than paraphrased, and it is worth quoting because it dates the ignorance.

Theodor Aufrecht’s Catalogus Catalogorum, part one, Leipzig 1891, gives the whole published Śābara tantric record in eight lines. Among them:

The entry for our text reads, in full: शाबरचिन्तामणि, mtr., and a single pointer to Gustav Oppert’s list of manuscripts in private libraries of southern India.

One line. One reference. In 1891 the entire recorded existence of this text was a number in Oppert’s list, and the whole Śābara tantric block in the world’s standard union catalogue ran to six or seven titles and about eight manuscript references altogether. In the union catalogue volume of 2014 the same headword carries fifteen manuscript witnesses. The corpus did not grow. The cataloguing did.

What the charter claim rests on

So the sentence this chapter can print is short.

The charter text is extant, complete, in ten chapters, in the Asiatic Society of Bengal, described in print since 1939. The manuscript was never hard to reach. It needed the union catalogue opened before the survey volume, which is the order any account of a corpus owes it.

What that does to the argument is small in doctrine and large in method. Nothing in this chapter’s reading of the charter changes: five languages, one revelation, five pupils, and the reading of it as distribution rather than condescension stands exactly where it stood. What it gains is an address, a physical description and a route to verification, and the next step for this treatise is a reproduction request rather than another argument.

The Asiatic Society entry has been read from page images of the printed catalogue. The manuscript itself has not been seen, by this project or, so far as any source reached here records, by anybody who has written about the Śābara corpus in the modern literature.

Two manuscripts wearing three numbers

Three shelfmarks circulate in connection with this charter text, and not one of the three belongs to it. The catalogue settles what each of them actually points at.

Two of the three sit in footnotes about something else. One of those notes attaches to Goudriaan’s caution about title-confusion; the other attaches to an attribution to Gorakṣa. The note that actually carries the Śābaracintāmaṇi is a third one, and it points at the Asiatic Society entry read above, with the remark that the catalogue prints the text’s beginning. The other two belong to a different problem in different manuscripts.

Two of the numbers are one object, cited twice under two systems: one is the serial the catalogue gives the entry, the other is the Society’s accession number for the manuscript described in it. The manuscript itself is a Śābaratantram, in Nāgara, complete in three chapters, and it carries the only explicit date anywhere in the Śābara block: Saṃvat 1912, which is 1855 or 1856 of the Common Era. The catalogue notes that its incantations are principally in Hindī.

The remaining number is not a catalogue serial at all. It is an accession number, and the entry describing it is a Divyaśābaratantram, in modern Nāgara, complete in fourteen chapters. That codex carries a second work as well: a run of its last folios, accessioned under a letter suffix of the same number, is catalogued separately as the Kālīśābaram, complete in three chapters, with a section in Hindī after its third chapter.

So the three shelfmarks are two manuscripts, and neither of them is the Śābaracintāmaṇi. Chapter six works out the count that rests on them.

That second codex is worth keeping in view for a second reason. It is the strongest single piece of evidence for the point Goudriaan makes and this chapter has already accepted: a Divyaśābaratantra and a Kālīśābara bound in one book, with chapter colophons inside the first that call it Divyaśābara in two places and Siddhaśābara in a third. The titles do not merely resemble each other in a catalogue. They alternate inside a single physical object, in the hand of one scribe.

The charter has an older cousin, in another religion

Here is the strongest new thing in this chapter, and it comes from an Indian scholar reading a Buddhist manuscript.

S. B. Dasgupta, in Obscure Religious Cults, published by the University of Calcutta in 1946, reports from the Laghukālacakratantrarājaṭīkā, entitled Vimalaprabhā, that the doctrines of the Pāramitā-naya are written wholly in Sanskrit, while those of the Mantra-naya are explained in Sanskrit, in Prakrit, in Apabhraṃśa, and beyond those in tongues outside the Sanskritic family, of which the Savaras’ is his example. His footnote records that he read it in manuscript, at the Royal Asiatic Society of Bengal.

Sit with what that is.

The Vimalaprabhā is the great commentary on the Laghukālacakratantra, a Buddhist Tantric work of the first rank, and it is centuries older than anything else in this chapter. It is not Śaiva. It is not Nāth. It has no interest in the Śābara Tantras and no knowledge of them. And it states, as a principle of its own tradition, that mantra-teaching is delivered in non-Sanskritic speech, and it names the Savaras as the exemplar of that speech.

That is the vernacular charter of chapter three, arriving from an entirely different religion, on a manuscript read by an Indian scholar in Calcutta in the 1940s.

What it does to this treatise is precise, and it is worth being exact rather than triumphant. It does not prove that the Pañcadrāviḍaśābaracintāmaṇi’s five languages are historical. It does not connect the Buddhist text to the Hindu corpus. What it establishes is that the idea the Śābara charter expresses, that mantra is properly given in the languages that are not Sanskrit, and that the Savaras’ speech is the type case of such a language, is a pan-Indic Tantric commonplace attested in a major Buddhist commentary, and not a late Nāth apology for a vernacular corpus. It also lands squarely on the etymology laid out above: here is a Buddhist Tantric source using Savara for a grade of non-Sanskritic speech, which is proposal one, inside the very tradition proposal two would derive the name from.

Which Dasgupta, and which pagination. Obscure Religious Cults has been printed more than once and the printings do not agree on where anything sits. Every Dasgupta page given in this treatise is the University of Calcutta first edition of 1946, read at the running head printed on the leaf. A later edition from Firma K. L. Mukhopadhyay runs about thirty pages lower through this material, and it is the copy the field’s own citations use: Lorenzen and Goudriaan both cite Dasgupta by the Mukhopadhyay numbers, which is why the Dasgupta pages given in their footnotes will not answer in the 1946 volume. A reader checking a reference in this treatise should open the first edition.

The named texts

The titles are few, and the catalogue lists them with a warning.

The warning first, because it governs everything after it. In principle, Goudriaan writes, there are a Divya-, a Kālī- and a Siddha-śābaratantra, but the titles are confused and may occur one after another in the same manuscript, and his note points to the Asiatic Society codex read above, where the title Dattātreyasiddhisopāna also occurs. Four flat existence statements, one per title, would be a misreading of that sentence.

They cannot be counted from the titles alone. A codex in which two of these works are bound together, with a third title surfacing in the chapter colophons of the first, and which also carries a title ascribing the work to Dattātreya, is evidence that the titles float. What can be said is that the names Divya-, Kālī- and Siddha-śābaratantra are on the record, that a Bṛhat-śābaratantra is on the record separately with printed editions, and that how many distinct works stand behind those names has to be settled by counting manuscripts rather than headwords. Chapter five does that count. No inference in this book rests on there being four or five separate Śābara Tantras arrived at by reading a list of titles.

The Siddha- or Siddhi-śābaratantra is referred to in the Kakṣapuṭatantra, and Goudriaan records two manuscripts of it in the Asiatic Society, both incomplete. The catalogue describes them: one discoloured, with explanatory notes in a vernacular written under the yantra diagrams on its last leaves; the other in two portions in two different hands, with a section near its end that carries a date the catalogue prints and this project has not read. The Kakṣapuṭatantra citation gives the text no age, because the Kakṣapuṭatantra’s own date is itself argued and is not established anywhere in this project. What the citation gives is a relative order and nothing else: one text was quoting the other before both reached their surviving manuscripts.

The Bṛhat-śābaratantra has been printed, and this chapter can say where. Goudriaan records an edition by Hariśaṃkar Śāstrī with a Hindī translation, Cawnpore 1900; a new edition, Moradabad 1906; and an edition from the Lakṣmīveṅkaṭeśvara Press, Kalyan, Śaka 1841, which is 1918 or 1919 of the Common Era. He points from there to the India Office Library Sanskrit catalogue.

Two imprints stand earlier than any of those. The New Catalogus Catalogorum records a Śābaratantra printed by the Jyotiṣaprakāśa Press at Calcutta in 1884, and a printing in R. Chatterji’s periodical Aruṇodaya, at Calcutta in 1890. The Asiatic Society catalogue mentions the same Chatterji: the Divyaśābaratantra, it notes, was published under the name Śābara in eighteen pīṭhas by Chatterji in the Aruṇodaya and in the Indrajālādisaṃgraha. The earliest imprint on record for this corpus is therefore Calcutta, 1884. Cawnpore, 1900, stands sixteen years down the run, and this treatise dates the corpus’s printed footprint from the earlier year throughout.

The Cawnpore, Moradabad and Kalyan imprints are real and they are named in Goudriaan and Gupta’s own note, printed directly below the paragraph this chapter has been reading. What is still wanted is any one of those editions in hand, with the title page read.

One shelf across sits a neighbouring text with a longer print record. In a footnote a page earlier, Goudriaan records editions of the Uḍḍīśa, a text of magical practice: R. M. Chattopadhyaya’s Indrajālādisaṃgraha, Calcutta 1879; Moradabad 1898; the Lakṣmīveṅkaṭeśvar Press at Bombay, 1954; Hind Pustak Bhandar at Delhi; and an edition by M. P. Vyas at Varanasi with no date. Many cheap editions of that kind, he remarks, are in circulation.

Read the addresses. Calcutta, Moradabad, Delhi, the Bombay commercial presses: this is the publishing world that put out cheap ritual manuals and pothī editions for a mass readership. Material of this kind did not enter the modern world through a university seminar. It entered through the bazaar press, and it has stayed in that channel ever since. Part of the reason the scholarship is thin may simply be that these texts never looked like scholarship’s kind of object.

Two negatives, and both are scoped. No critical edition of any of these texts was found, and that negative rests on two searches rather than one, because the union catalogue of Sanskrit works has been searched at the Śābara headwords and records none either. No date of composition for any of them is given in any source reached here. A date of copying now is: Saṃvat 1912, 1855 or 1856 of the Common Era, on the Śābaratantram described above. That is the only explicit date in the whole Asiatic Society block, and a date on a copy is not a date on a text.

And one positive the chapter’s own title has been waiting for. This treatise has argued from a European survey’s single sentence that the genre alternates Sanskrit with Hindī. The manuscript descriptions confirm it from the other side, independently, in the cataloguers’ own field notes: incantations principally in Hindī in the dated Śābaratantram, a Hindī section after the third chapter of the Kālīśābaram, “Hindī and corrupt Sanskrit” in another entry of the same volume, explanatory notes in an unnamed vernacular under the yantras of a Siddhaśābaratantra, Telugu prose in a Madras copy of the Śābaracintāmaṇi, Kannada running through a Tanjore Śābararahasyamantra, and a Panjab copy of the Śābaratantra catalogued as carrying vernacular meaning. Four collections, in Calcutta, Madras, Tanjore and the Panjab, described by cataloguers who were recording contents rather than making an argument about register. The bilingualism is not a European scholar’s interpretation of the genre. It is what the manuscripts look like when somebody opens them and writes down what is inside.

What the Nāths say about it

The Śābara corpus does not only get claimed by the Nāths from outside. A Nāth text claims it from within.

The Gorakṣasiddhāntasaṃgraha, the Collection of Gorakṣa’s Established Doctrine, quotes a Śābaratantra by name. The Sanskrit, in the Gorakhnāth Mandir edition held here, runs:

ननु कापालिकप्रसंगकथनादेतेषां कापालिकमतमिति कैश्चिज्ज्ञायत इति, वस्तुतु नास्त्येवास्माकं मतं त्ववधूतमेव । परन्तु कापालिकमपि मतं नाथेनैव प्रकटीकृतम् एतन्मार्गस्य प्रकटकर्ता नाथ एव । तदुक्तं शाबरतन्त्रे

The sense, in indirect speech: some, on account of the discussion of the Kāpālika connection, take this to be the Kāpālika doctrine; it is not so, ours being the Avadhūta doctrine, the doctrine of the one who has shaken off all bonds; but the Kāpālika doctrine too was revealed by the Nātha, and the propagator of that path is the Nātha alone. Then the citation formula, tad uktaṃ śābaratantre, thus it is said in the Śābaratantra, and a list follows: twelve Kāpālikas from Ādinātha to Śrīkaṇṭha, and their twelve pupils from Nāgārjuna to Malayārjuna, among them Gorakṣa, Carpaṭa and Jālandhara.

Two small things in that Sanskrit before the large one. The citation formula spells the title with the palatal sibilant, शाबरतन्त्रे, while the Hindī rendering printed below it on the same page of the same edition spells it साबरतन्त्र. And the name at the head of the list of twelve is Ādinātha, who is Śiva.

The sense above is reported in indirect speech rather than given as a translation, and the reason is worth stating. Any close English rendering of these lines comes out near-identical to David Lorenzen’s published translation in The Kāpālikas and Kālāmukhas, which is where this project first met the passage, and a treatise cannot present that as its own. So the Sanskrit stands, and the sense is reported. The passage sits in Janārdana Pāṇḍeya’s Benares edition of 1973, as cited by Mark Dyczkowski; Lorenzen cites it from a differently paginated edition, which is a difference of printing rather than of text.

The framing is a claim of ownership, made carefully. The Nāth author concedes the Kāpālika material, refuses the Kāpālika label for his own school, and absorbs the Kāpālikas into a Nāth revelation anyway.

Dvivedī reads it exactly that way, and he read it in 1950. In Nāthasampradāya he sets out the same passage and draws the inference in Hindī: the author concedes that the Kāpālika doctrine too was revealed by the Nātha, and the reason he can concede it is that in the Śābaratantra the first of the twelve Kāpālika teachers is Ādinātha himself and several of the twelve pupils are chief teachers of the Nāth path. His footnote prints both lists of twelve, and his book’s printed index carries the entry शाबर तंत्र. On the same page he quotes the Ṣoḍaśanityātantra for Śiva’s statement that the nine Nāthas propagated in the world the Tantras he had spoken. Dasgupta had given both, in English, in Obscure Religious Cults in 1946, footnoting the first to the Śābaratantra as quoted in the Gorakṣasiddhāntasaṃgraha and the second to the Ṣoḍaśanityātantra in the same source, and adding from the Tantramahārṇava the scheme that seats eight Nāthas in the eight directions with Ādinātha at the centre in Kurukṣetra.

The three lists are worth comparing, because they disagree in instructive ways. Checked against the Sanskrit of the Gorakhnāth Mandir edition, Dasgupta’s English is the most accurate of the modern reports: it carries Bhūtanātha and Vairāgya where the scan of Dvivedī’s footnote drops the one and reads Vairāgī for the other. Lorenzen notes that Dasgupta reads Vaikalika where the printed text has Atikālaka, and says he does not know where Dasgupta got it. The Sanskrit in hand reads अतिकालकः. Four scholars, one list, and the names will not sit still either.

Two page statuses. Open scans of both cited editions are on the Internet Archive: Dvivedī’s Nāthasampradāya, Hindustani Academy, Allahabad, 1950, complete; and Bagchi’s Kaulajñānanirṇaya, Calcutta Sanskrit Series III, 1934, whose text layer reads its Roman introduction correctly and is searchable. The Bagchi and Dvivedī pages used in this chapter stand as consulted here and not re-checked against a page image by anybody outside this project. Any reader can open both books and check them.

What Lorenzen actually says about the Gorakṣasiddhāntasaṃgraha

David Lorenzen’s The Kāpālikas and Kālāmukhas is the monograph the field goes to for this sect. It carries an index entry for the Śabara-tantra, and the pages behind that entry have to be read to the end, because the field’s principal authority engages the Śābara material in order to reject what is usually built on it.

The page immediately after the Śābaratantra list carries Lorenzen’s verdict. Reporting the argument some modern authorities have built, that the Kānphaṭa school was a later transformation of the older Kāpālika order, he answers it in substance as follows, and this is a paraphrase and not his words: as a historical document the late Gorakṣasiddhāntasaṃgraha is virtually useless; the similarities between the two schools, meat-eating, wine, magical powers obtained through yoga, dwelling in cremation grounds, are common to the whole of the Tantric tradition; and such historical speculations are therefore of little value.

Virtually useless as a historical document. That is the field’s standard authority on the Kāpālikas, one page after the passage this treatise leans on, on the text this treatise has called its strongest primary.

And the concession he does make has to be measured exactly. Later in the same book Lorenzen writes that Baṭuka appears among the twelve original Kāpālika sages in the Śabara-tantra list quoted in the Gorakṣasiddhāntasaṃgraha, that the twelve were evidently taken to be the founders of twelve divisions of the sect, and that the presence of this tradition in unrelated sources suggests there may have been some factual justification for it. The unrelated source he has just been discussing is Ānandagiri and the Ḍiṇḍima commentary on Mādhava, which mention twelve sorts of Kāpālikas beginning with the Baṭukas without naming them. So the thing he allows may have a factual basis is the twelvefold division of the Kāpālika sect, corroborated between two independent traditions. It is not the Śābaratantra. It is not the Gorakṣasiddhāntasaṃgraha. It is one motif inside them, and a narrow concession about a number is a long way from a scholar crediting the corpus.

So what is left, and can this chapter still use the text?

It can, for one job and not for another, and the difference is the whole of the matter.

Lorenzen’s verdict is aimed at a specific inference: that the Gorakṣasiddhāntasaṃgraha can be used to reconstruct the historical relation between the Kāpālika and Kānphaṭa orders. On that question he is right and this treatise accepts him without reservation. No claim that the Nāths historically absorbed the Kāpālikas, or that the Kāpālika-Nāth relation can be dated from this text, can stand on it, and chapter six has to be read against this section.

What the Gorakṣasiddhāntasaṃgraha is good for is something the historical verdict does not touch. It is a compilation of citations. Dvivedī, examining it in 1950, describes it as an incomplete and hastily made book whose Sanskrit is loose and in places incorrect, by an author who was nevertheless widely read, and he counts the works it quotes: forty-eight of them, many no longer available. That is the Indian scholar who knows the text best, saying both halves at once. The book is bad history and it is a good witness to what was on a Nāth compiler’s shelf.

And what this chapter takes from it is exactly that kind of fact. Not that the Kāpālikas became the Nāths. Only that a Nāth compilation quotes a text called Śābaratantra by name, with an ordinary citation formula, and reproduces a list from it. That is evidence about the circulation of a title, and it survives Lorenzen’s verdict intact, because Lorenzen himself uses the Gorakṣasiddhāntasaṃgraha for precisely that while judging it worthless for anything more.

There is a second brick, and it does not pass through the Gorakṣasiddhāntasaṃgraha at all. P. C. Bagchi, in his 1934 introduction, cites a Śāvara Tantra directly, on the authority of Sylvain Lévi’s Nepalese materials and a 1930 article by Giuseppe Tucci in the Journal of the Asiatic Society of Bengal, and reports it as containing twenty-four Kāpālikas: twelve gurus, who are the twelve forms of Śiva, and twelve disciples, among whom he finds Mīnanātha, Gorakṣa and Carpaṭa. An independent route to the same list, through Nepalese manuscript material, and in print before either Dasgupta or Dvivedī reached it.

Bagchi’s clause about the twelve forms of Śiva is the most useful sentence in this section, and it points the same way Lorenzen does. If the twelve gurus of the list are twelve forms of Śiva, the list was never a roll of historical persons. Ādinātha at its head is Śiva, as Dvivedī says. Reading it as a chronicle of a sect was a category error before anyone got to the question of whether the Gorakṣasiddhāntasaṃgraha is reliable.

One negative about Lorenzen holds. He nowhere mentions Kāśīnātha Bhaṭṭa, and nowhere mentions the Kāpālikamatavyavasthā. That has been checked in his index and text and holds at zero. His appendices and bibliography have not been checked line by line, and a negative printed this flatly should be.

So the honest statement of the Kāpālika question. A list of twelve and twelve, headed by Ādinātha and Nāgārjuna, is attributed to a Śābaratantra in a Nāth compilation and, by a separate route, in Nepalese material. Its twelve seniors are forms of Śiva. Lorenzen knows the list, judges the compilation that carries it useless as history, and allows only that the twelvefold division of the sect may have a factual basis because it turns up in unrelated sources. Kāśīnātha Bhaṭṭa, in the eighteenth century, took that list and called its proclaimers Kāpālikas; Lorenzen has never seen him; and Goudriaan, who has, judges there is more reason to connect them with the Nāth cult.

The Kāpālika association may be real. It may equally be an eighteenth-century scholar’s label for material whose actual home was the Nāth orders. On what is in front of us, we cannot tell, and anyone who tells you otherwise is going beyond the evidence.

Two texts, one name, and a gap in the record

There is a trap in this material, and it is easy to fall into.

Mark Dyczkowski’s The Canon of the Śaivāgama (SUNY, 1988) reproduces the canonical list of Tantras given in the Pratiṣṭhālakṣaṇasārasamuccaya. The list is organised by stream. In the Gāruḍa column, the stream concerned with snakebite, poison and venom-cure, items seven and eight are Śābaram and Kālaśābaram.

Dyczkowski adds that the author of the Pratiṣṭhālakṣaṇasārasamuccaya, Vairocana, was the son of the Bengali Pāla king Dharmapāla, whose reign he places from 794 to 814 of the Common Era, and that the list therefore furnishes an ante quem for the works in it. That puts a scripture named Śābara inside a ninth-century canonical list, in the poison-cure stream of the Śaiva canon.

This is not Goudriaan’s corpus. Goudriaan’s Śābara Tantras are a late collection of incantations moving between Sanskrit and Hindī, and the earliest hard dates anywhere near them are a Calcutta imprint of 1884 and a copying date of 1855 or 1856 on one manuscript. Set the ninth-century list beside those and the interval is a little over a thousand years of record. Same word, different period, different genre, different world.

That interval measures records rather than compositions. The late corpus is undated, so the distance between the two as texts is unknown and may be far smaller. Anyone printing a figure for it, including this chapter, is measuring the archive rather than the literature.

The scholar to read on the Gāruḍa stream is Michael Slouber, and his Early Tantric Medicine (Oxford, 2017) is the standard work on that corpus. It has not been opened for this chapter, and nothing is said here about what it argues.

Two texts carry the name. We have looked at one of them.

What this chapter does not print

A reader who has come this far is entitled to ask why a chapter about a corpus of incantations has not shown a single incantation.

The answer is a position and not an oversight. This treatise exhibits the form of the genre and withholds its function. The form is on every page of this chapter: the catalogue’s description of contents that seldom go beyond enumerations of incantations, the alternation of Sanskrit with Hindī, the deities addressed, the imprints of the presses that carried the material to a mass readership, the lexicographers’ glosses and, in chapters ten and eleven, the distributions of the formulae themselves counted across a corpus. That is enough to characterise a genre. It is not enough to operate one.

Chapter twenty is the reason, and it makes the case from the documented record of what this material has been used for and against whom. A reader who wants the argument rather than the rule will find it there. What is being withheld throughout is the operative line: no working formula, no ritual prescription, no count of repetition, no materials, no timing.

The shape of what we have

Strip it back to what can be defended.

A genre exists and is catalogued. The catalogue’s definition gives three criteria: incantations moving between Sanskrit and Hindī, a minor deity called Śābara along with Yoginīs and Kṣetrapāla and Gaṇeśa and Kālī, and magical purposes. This treatise reads the first as the one doing the distinguishing work, and says so as its own reading, and the manuscript descriptions now support that reading from a direction the survey did not supply. The same paragraph floats a Buddhist derivation of the name, which this chapter has laid out beside the two rivals and declined to adopt while stating what would overturn its choice. The catalogue also warns that the titles are confused and may run one after another in a single manuscript, and an Asiatic Society codex shows exactly that happening, so titles cannot be counted as texts. Its own charter says the revelation was given out in five named vernaculars, from Bengal to Gujarat to Kerala, by the teacher of Matsyendranātha; and that charter text is extant, complete in ten chapters, on paper, in Kolkata, with about fourteen further witnesses scattered from Pondicherry to Ujjain and no critical edition of any of them. A Buddhist commentary of far greater age says mantra-teaching is given in non-Sanskritic speech and names the Savaras as the case in point. A Nāth compilation quotes a Śābaratantra by name, and Nepalese material reached by Bagchi quotes one too; the compilation is worthless as history on the judgement of the field’s authority on the sect, and it remains a witness to a title in circulation. The eighteenth-century Sanskrit author who called that list’s proclaimers Kāpālikas is unknown to that authority. The corpus has been in print since Calcutta, 1884. And a completely different text, a thousand years earlier in the record, wears the same name in a canon of poison-cures.

That is the corpus. It is small, it is under-edited, it is bilingual in the manuscripts and not only in the survey, and it is more Nāth than anything else.

The next chapter leaves the catalogue and goes to the oldest securely dated place where this word can be caught in use, a Prakrit narrative finished in 779 of the Common Era.

The genre was bilingual before anyone thought to call it folk.

Whether bilingualism is what its name was ever about is a question this chapter has now answered as far as the evidence allows, and named the price of being wrong.