The name on everybody's lips

This corpus has a charter. It sits inside one of the corpus’s own texts, the Śābaracintāmaṇi, and it is the story the material tells about where it came from. The mantras were revealed to Ādinātha by the grace of Īśvara. Ādinātha proclaimed them to five pupils, in the languages of Gauḍa, Kerala, Karṇāṭa, Andhra and Gurjara.

Read the sentence again for the name that is missing from it.

Gorakhnāth is not in it. He is not the recipient, he is not one of the five, he is not the proclaimer. Ādinātha is known in this tradition as Matsyendranātha’s teacher, and Matsyendranātha is Gorakhnāth’s teacher, so the charter reaches back past Gorakhnāth, past his own guru, and comes to rest two full generations behind him at the origin point of the lineage. That is the tradition, in its own document, speaking about its own material. It does not say Gorakhnāth made these mantras. It says the mantras came before Gorakhnāth.

Now type the words “shabar mantra” into any search field in any Indian language and count how long it takes for a name to appear. It will be seconds. The name is Gorakhnāth, and the sentence attached to it is always some version of the same sentence: Gorakhnāth composed these mantras, or Gorakhnāth brought them down, or the nine Nāths made them in the Kali age for people who could not manage Sanskrit.

The sentence is everywhere. It is on the video thumbnails, in the printed prefaces, in the blog posts, in the app descriptions.

It is also, as far as this research has been able to establish, unsourced. Nobody who says it cites anything.

What has happened here has a shape, and the shape is not forgery and not fraud. Attribution drifts toward the most famous man in a lineage. Gorakhnāth is the name a reader in 2026 knows. The teacher’s teacher, who has no video channel, drops out of the sentence, and the material collects around whoever is left standing.

So the popular claim is a simplification of what the tradition said, running in the opposite direction from the tradition’s own evidence. That would be a small enough finding on its own. What makes this chapter worth writing is that when you go looking for what could be cited in place of the popular sentence, what you find is stranger and better than the claim itself: a Nāth text that quotes the Śābaratantra by name, argues at length with a rival about who is entitled to it, and never once says that a Nāth wrote it.

This chapter is about the difference between those two things.

The claim, and what it would need

Take the popular sentence at face value for a moment and ask what evidence would have to exist for it to be true.

At minimum: a text attributing the corpus to Gorakhnāth, or a manuscript colophon naming him as author, or a lineage record placing the composition in his hands, or a scholarly reconstruction dating the corpus to whatever period Gorakhnāth is placed in. Any of those would do. One would be enough to build on.

What the record actually contains is manuscript evidence, and it is thinner than the popular sentence needs. It is also messier than the popular sentence allows.

Goudriaan and Gupta report a manuscript in the Royal Asiatic Society of Bengal collection, number 6099, whose colophon names Gorakṣa as an author, and they record Gorakṣasiddhisopāna as one of the alternative titles carried by RASB number 8355. Mark Dyczkowski, working from the same collection and citing it independently, reports the Gorakṣasiddhisopāna colophon at Asiatic Society of Bengal number 8355, and adds what reads as a third witness: manuscript number 10542, which he says ascribes the work directly to Gorakhanātha. Hold that number. The section below turns on it.

Now the line that sits one line above the line usually quoted. Goudriaan records that the very same manuscript, RASB 8355, also carries the title Dattātreyasiddhisopāna.

Read that slowly. The single manuscript that supplies the Gorakṣa title also supplies a title ascribing the same work to Dattātreya. That is not a second witness for Gorakṣa. It is direct evidence that the ascription on that manuscript floats, and it belongs in the chapter for exactly that reason.

It is also not a stray. Shashibhusan Dasgupta, in Obscure Religious Cults, gives the list of nine Nāthas from the Tantramahārṇava as quoted in the Gorakṣasiddhāntasaṃgraha, and Dattātreya holds the west in it, alongside Gorakṣa in the east and Matsyendra in the south-east. So a Dattātreya title on a Nāth-adjacent manuscript is not an intruder from outside. It is a competing ascription from inside the same nine-name frame, and a scribe who reached for it was choosing between two authorities the tradition already had.

Three shelfmarks, two manuscripts

Read the two citations side by side and the manuscript plank looks like three catalogued witnesses, all three ascribing the work to Gorakṣa or Gorakhanātha, one of the three also carrying a title ascribing it to Dattātreya. Read the entries in the cataloguer’s own volume and the count comes down.

The catalogue is the Descriptive Catalogue of the Sanskrit Manuscripts in the Collections of the Royal Asiatic Society of Bengal, volume VIII, the Tantra volume, revised and edited by Chintaharan Chakravarti and printed at Calcutta in 1939 and 1940. It gives every manuscript two numbers: a catalogue serial, which is what “RASB 6099” means, and the Society’s accession number, which is what “MS 10542” means. They are different numbering systems for the same shelf.

RASB 6099 is MS 10542. One manuscript, cited twice, once by each of its numbers, by two different scholars, and easily taken for two witnesses. Goudriaan cites the serial; Dyczkowski cites the accession number; the object is the same object, and its colophon names Gorakṣa once and not twice.

What entry 6099 describes, under the accession number 10542, is a Śābaratantram on country-made paper: twenty-one leaves, Nāgara, complete in three chapters, its incantations principally in Hindī. It carries a date as well, Saṃvat 1912, which is 1855 or 1856 of the Common Era, and that is the only explicit date anywhere in the ten-entry Śābara block of the catalogue.

The same confusion of the two numbering systems accounts for the remaining witness. “RASB 8355” is not a catalogue serial at all; it is an accession number, and its catalogue entry is 6091, describing a Divyaśābaratantram on country-made paper, forty-one leaves, modern Nāgara, complete in fourteen chapters, with two stray leaves at the end carrying a portion of the beginning again. That is the codex whose alternative titles Goudriaan reports, and it holds a second work besides: part of it, accessioned as 8355B, is catalogued separately as entry 6095, the Kālīśābaram, complete in three chapters, with a section in Hindī after the third.

So the plank, stated at the shelf: two manuscripts in one collection. One of twenty-one leaves, dated 1855 or 1856, complete in three chapters, whose colophon names Gorakṣa. One of forty-one leaves, undated, complete in fourteen chapters, which carries the title Gorakṣasiddhisopāna and also the title Dattātreyasiddhisopāna, and which is bound together with a different work.

Every entry named here has been read at volume, entry and page, from page images of the printed catalogue. What remains unread is the manuscripts themselves, by this project and, so far as anything reached here records, by anybody writing on this corpus.

Say the uncomfortable half plainly, since this chapter set the test itself at the head of the last section. A manuscript colophon naming him as author was one of the four things that would do, and one colophon does it, with a second manuscript carrying his name in a title. The test is met. What the colophons cannot supply is the rest of the sentence: not a corpus, not composition, only an ascription written on a copy, and on the other copy an alternative ascription to somebody else.

They do supply one thing the popular sentence has no answer to, and it cuts against that sentence rather than for it. A date. The manuscript that names Gorakṣa was copied in 1855 or 1856, twenty-nine years before the first printing of a Śābaratantra at Calcutta. The scribe who wrote Gorakṣa’s name on that leaf was working in the reign of Bahādur Shāh Zafar, in a Bengal that already had steam presses. That is the age of the physical evidence for the Nāth ascription: not medieval, not Vijayanagara, not even Mughal. Mid-Victorian.

A date on a copy is a date on the copy. It says nothing about when the text was composed, and this chapter draws no compositional inference from it. What it does establish is the floor of the documentary record, and the floor is nineteenth-century.

A text that argues with an objection

The primary that carries the most weight in this area is the Gorakṣasiddhāntasaṃgraha, the Collection of Gorakṣa’s Established Doctrine, and its relevant passage was set out in chapter three: the citation formula तदुक्तं शाबरतन्त्रे, “thus it is said in the Śābaratantra,” followed by a list of twelve Kāpālikas and twelve disciples, with Gorakṣa, Nāgārjuna and Jālandhara among the names. Pages 14 to 15 of the Pāṇḍeya edition, Benares 1973, as cited by Mark Dyczkowski. The Gorakhnāth Mandir edition puts the same text within reach of anyone who wants to check it, and it has been opened at the citation formula for this treatise. The formula reads as given, in Devanāgarī, and the twenty-four names follow it in the order chapter three prints.

One detail is worth carrying out of that reading: the Sanskrit prints शाबरतन्त्रे and the facing Hindī gloss prints साबरतन्त्र, which is the śa and sa alternation chapter five tracks, running here in the opposite direction from the Rāmcaritmānas.

The book itself, described

The edition itself is worth describing, since so much rests on it.

Gorakṣasiddhāntasaṃgrahaḥ, edited by Śrī Janārdana Śāstrī Pāṇḍeya, under the general editorship of Dr Bhāgīratha Prasāda Tripāṭhī, “Vāgīśa Śāstrī,” Director of the Research Institute of the Vārāṇaseya Sanskrit Viśvavidyālaya. It stands in the Sarasvatī Bhavana Granthamālā, the university’s own series, printed at Vārāṇasī in Vikrama 2030, Śaka 1895, 1973 of the Common Era. Its title page calls it parivardhitaṃ saṃśodhitaṃ ca saṃskaraṇam, a revised and enlarged edition, which means it revises an earlier one. First printing of that edition: five hundred copies, priced at ten rupees fifty.

Five hundred copies. That is the print run behind the text this treatise has been calling its strongest primary.

Now the two things that matter more than the imprint.

The Gorakhnāth Mandir edition says outright that the book does not survive complete. Its Hindī introduction states that the Gorakṣasiddhāntasaṃgraha is not available in its full form, and locates the value of what does survive in its quotations from rare works. It then lists those works, and it names the Śābaratantra first, ahead of the Kapilagītā of the Padmapurāṇa, the Yoga Upaniṣads, Gorakṣanātha’s own Amanaska-yoga and Yogabīja, the Siddhasiddhāntapaddhati, the Gorakṣasaṃhitā, the Vivekamārtaṇḍa, the Avadhūtagītā, the Ādināthasaṃhitā, the Tantramahārṇava, the Haṭhayogapradīpikā, the Śivasaṃhitā and a dozen more. The edition’s own editors, in other words, describe the book exactly as Dvivedī did in 1950 and as this chapter uses it: a damaged anthology whose worth is the shelf it quotes.

And neither edition states what manuscripts it was made from. The Pāṇḍeya edition’s Sanskrit prāstāvikam, its whole introductory essay, has been read. It is doctrinal: the Nātha lineage, the nine Nāthas, the avadhūta, the date of Gorakṣanātha. It names no manuscript, gives no sigla, and carries no variant apparatus beyond an occasional single-letter footnote. The Gorakhnāth Mandir introduction names none either. So the text of the twelve-Siddha list that three modern scholars and this treatise have all worked from rests on an editorial base that neither of its two editions discloses. What would close it: the identity, date and manuscript basis of the first edition, which the 1973 title page implies and does not name. That is one visit to the Sarasvatī Bhavana Library at Varanasi.

What is actually being quoted

One more thing about the passage, and it changes what the quotation can be used to prove.

Look at what stands after tad uktaṃ śābaratantre. It is a list of names: Ādinātha, Anādi, Kāla, Atikālaka, Karāla, Vikarāla, Mahākāla, Kālabhairavanātha, Baṭuka, Bhūtanātha, Vīranātha, and Śrīkaṇṭha reckoned the twelfth. Then the twelve pupils. The editor summarises it in his own introduction in one clause, that according to the Śābaratantra there are twelve Siddhas, and the printed index of the edition carries the entry शाबरतन्त्रम् against one page and nowhere else.

So the sole indexed Śābara locus in the whole book is a genealogy. No incantation is quoted. No rite is described. Nothing operative crosses from the Śābaratantra into the Gorakṣasiddhāntasaṃgraha at all. What crosses is a list of who the teachers were.

That fact does two jobs at once, and they pull in opposite directions.

It strengthens the passage as evidence, because the same list reaches us by two further routes that do not pass through this book. Kāśīnātha Bhaṭṭa quotes it in his Kāpālikamatavyavasthā in the eighteenth century. P. C. Bagchi reports it from Nepalese material in his 1934 introduction. Three independent witnesses to one passage make it, by a distance, the most securely attested piece of Śābaratantra content anywhere in the record. And a copy of a Śābaratantra in Kathmandu is catalogued under the name of Śrīkaṇṭha, who is the twelfth name in that very list, which is the sort of small convergence that is either meaningful or a coincidence and cannot yet be told apart.

And it weakens what the quotation proves about the corpus, because a lineage list is the part of a tantra least dependent on the tantra. Genealogies travel. They get copied into introductions, recited in initiations, and attached to whatever authority needs them. When a Nāth compiler writes “thus it is said in the Śābaratantra” and then gives twelve names, he establishes that a text of that title was circulating and was worth citing. He does not establish that he had read its mantras, or that the Śābaratantra he had is the Śābaratantra on any surviving shelf.

That citation is the best primary this treatise has in the Nāth question, and it has to be used for what it is: doctrinal genealogy rather than ritual content, from a book its own editors say is incomplete, in an edition that does not name its manuscripts.

The passage is not this treatise’s discovery, and three scholars got to it first, in three languages, across four decades.

David Lorenzen translates the whole passage in The Kāpālikas and Kālāmukhas and gives the twelve and the twelve immediately after it. Shashibhusan Dasgupta prints both lists in English in Obscure Religious Cults, published by the University of Calcutta in 1946, with a footnote naming the Śabara-tantra as quoted in the Gorakṣa-siddhānta-saṃgraha. Hazārīprasād Dvivedī has it in Hindi in Nāthasampradāya, published at Allahabad in 1950, with both lists of twelve printed in his own footnote and Śābaratantra entered in his index. Dasgupta’s English list matches the Sanskrit mūla held for this project name for name, including Bhūtanātha and Vairāgya, where the Dvivedī scan drops one and varies the other. Where the English lists are wanted, Dasgupta is the accurate report; where the text is wanted, it is available in Sanskrit and should be read there.

Which Dasgupta, and which pagination. Every Dasgupta page here is the University of Calcutta first edition of 1946, read at the running head. The later Firma K. L. Mukhopadhyay edition runs about thirty pages lower through this material and is the copy the field’s own citations use, Lorenzen’s footnote 110 and Goudriaan’s note 44 among them. Chapter 3 sets the problem out in full.

The Bagchi and Dvivedī pages have a status worth stating too. Open scans of both cited editions are on the Internet Archive, and the text layer of the 1934 Kaulajñānanirṇaya reads its Roman introduction correctly, so that introduction can be searched by anybody who wants to. The Bagchi pages used below, 19, 21 to 22, 28 and 34, and the Dvivedī pages 4 and 5, were read for this project and have not been re-checked against a page image by anybody outside it. They are consulted and checkable. Nobody outside this project has checked them.

Now the sentence wrapped around the list, which is what this chapter came for.

The text is arguing with an objection it can hear coming. In Lorenzen’s rendering it heads off the reader who would take the Kāpālika connection for Kāpālika doctrine, claims for itself instead the Avadhūta doctrine, the doctrine of the one who has shaken off every bond, and then concedes that the Kāpālika doctrine too was revealed by the Nātha, who alone propagates that path.

Those lines can be read as a lineage annexing a corpus in writing. That reading needs one word to mean one thing, and the field does not agree that it does.

Lorenzen, translating the same lines, glosses Nātha in a parenthesis as Śiva. On his reading the sentence says only that Śiva revealed the Kāpālika path too, which is an ordinary Śaiva claim about the origin of a revelation and not an act of annexation by anybody.

P. C. Bagchi, editing Matsyendranātha’s Kaulajñānanirṇaya for the Calcutta Sanskrit Series in 1934, refers to the same twenty-four-name passage, under the title Sāvara Tantra, and describes the twelve gurus in it as twelve forms of Śiva, with Mīnanātha, Gorakṣa, Carpaṭa and others among the twelve disciples. That is an Indian scholar reading the list from the Sanskrit, forty-seven years before Goudriaan, and reading its upper half exactly as Lorenzen does.

Dvivedī reads it the other way. Working through the same section of the Gorakṣasiddhāntasaṃgraha, he takes the passage, together with the Ṣoḍaśanityātantra on the nine Nāthas and the section on the Śākta ācāras, as the Nāth texts’ own testimony that the Kaula path of the Tāntrikas and the Kāpālika doctrine are both Nāth-following. That is the annexation reading, stated in Hindi in 1950 by the scholar who wrote the book on the lineage.

The Sanskrit नाथ एव permits both, and nothing available here decides it. So the honest statement is this. Lorenzen and Bagchi take the Nātha of these lines, and the twelve at the head of the list, as Śiva. Dvivedī reads the same section as the Nāth tradition claiming the Kaula and Kāpālika streams for itself. This treatise cannot adjudicate between them and does not.

What survives on either reading is smaller than the annexation reading would make it, and it is still worth having. Whoever the Nātha is, the author of the Gorakṣasiddhāntasaṃgraha is doing three things in one breath. He concedes the Kāpālika material rather than hiding it. He refuses the Kāpālika label for his own school, which is a defensive move, because Kāpālika was a slur long before the eighteenth century. And he routes the whole thing back through a single source of revelation that his own school stands closest to.

That is a claim about authority. It is not a claim about authorship, and it does not become one on any reading.

What Lorenzen actually concludes

Lorenzen’s monograph is not silent on this file. He translates the passage, and later in the book he makes a concession that reads, on its own, like a willingness to credit the tradition’s own account. Both of those are on the page. A reader who stops there stops one page short of the verdict.

One page after the passage both chapters lean on, Lorenzen states that as a historical document the late Gorakṣasiddhāntasaṃgraha is virtually useless; that the similarities between the Kāpālika and Kānphaṭa schools are common to the whole of the Tantric tradition; and that historical speculations built on them are of little value.

That is the field’s principal authority on the Kāpālikas, in the paragraph immediately following the Śābaratantra list, declaring the citing text historically worthless.

And that later concession, which reads at a glance like Lorenzen crediting the Śābara material, is about something else. What he allows there is that the twelvefold division of the Kāpālika sect may have some factual justification behind it, because the same tradition of twelve founders turns up in unrelated sources, the Ḍiṇḍima commentary among them. That is a narrow concession about one motif in one list. It is not an endorsement of the Śābaratantra, of its age, or of anything the Gorakṣasiddhāntasaṃgraha says about it.

So the sentence this chapter is left with is uncomfortable and it stays. This treatise’s best primary in this area is a text that the standard monograph on the sect it describes calls virtually useless as history. Two things follow, and both are honest.

The first is that nothing dated can be built on it. The chapter does not try. Every use made of the passage here is a use of its grammar: a citation formula points at a text the writer did not write and expects the reader to accept it as already standing. That is a fact about the sentence, and it survives even if every historical claim in the text around it is worthless.

The second is that Lorenzen’s verdict is itself a judgment about historical value, made in a book about two Śaiva sects, and it does not settle what the passage is evidence of. Dasgupta and Dvivedī both use the same passage, knowing it is late, as evidence about how the Nāth tradition arranged its own genealogy. That is a different question, and Lorenzen was not answering it.

David Gordon White is reported to place the Gorakṣasiddhāntasaṃgraha at around the seventeenth century, and to read the Śābaratantra it quotes as relatively recent. That dating is not in this project’s research file, no page has been checked for it, and no argument in this chapter rests on it. What would close it: The Alchemical Body (1996) at page level, starting at page 98. Until somebody does that, the seventeenth century is hearsay and is treated as hearsay here.

The rival claimant, and how thin he is

There is exactly one competing bid on the record, and chapter three has already weighed it: Kāśīnātha Bhaṭṭa, an eighteenth-century author, whose Kāpālikamatavyavasthā pulls the same twelve-and-twelve list into a Kāpālika frame and names its members the proclaimers of the Śābara mantras. Goudriaan reports him and immediately doubts him. There is, he writes, more reason to connect them with the Nātha cult.

The narrow negative around that bid still holds, and it has been checked in the text and in the printed index. Lorenzen never mentions Kāśīnātha Bhaṭṭa. He never mentions the Kāpālikamatavyavasthā. So the eighteenth-century bid in particular has not been taken up by the one scholar best placed to take it up, while the passage that quotes the Śābaratantra has been, and has been dismissed as history.

So the contest for this corpus runs between a Nāth text making a considered claim about authority, an eighteenth-century author whose particular treatise nobody in the field has since engaged, and Goudriaan’s one-line preference for the Nāths. On the current evidence that preference is the defensible reading of where the material sits. It is not evidence that anybody in the Nāth lineage wrote it.

A sourcing warning belongs here, because it governs how any of these pages can be quoted. The reading “Katha cult,” which is loose in circulation, is machine garbage. Goudriaan’s printed page reads Nātha. The scan in circulation renders capital N as K and b as h through that passage, and the same file turns Śabara into Sahara, Gorakṣa into Gobaksa and Nāgārjuna into Kagarjuna. Anybody quoting Goudriaan on the Nāths from a full-text search rather than from the printed page is at risk of quoting the scanner.

The string Sahara is the awkward one, because two separate mechanisms produce it. One is that scanner failure. The other is ordinary sound change. Russell and Hira Lal, at IV.502 of The Tribes and Castes of the Central Provinces of India in 1916, record that in the Uriya country v changes to b and an aspirate is interpolated, so that Savara became Sabra or Sahara, as Gaur has become Gahra. Sahara is therefore an attested vernacular form of the name, recorded by a named source already in this treatise’s bibliography and already quoted twice in chapter two. A b-to-h in this particular word is a regular Odia-country change as well as a known scanner artefact, and the string by itself is evidence of neither.

That is the narrow ground the whole question stands on. A string that two separate mechanisms can produce cannot carry an argument until somebody establishes which mechanism produced it, and that includes the “Sahara” that appears in a 1938 footnote read from another scan of the same vintage. What is wanted there is a page image of the footnote. None has been obtained.

What the scholarship on the Nāths does and does not contain

Now the part of this chapter that has no precedent in the popular literature.

If the Nāth attribution were substantive, you would expect to find this corpus in the modern scholarship on the Nāths. Not necessarily prominently. Somewhere.

Two of the best recent books on the order were searched in full for this treatise, both born-digital and therefore genuinely searchable.

Véronique Bouillier, Monastic Wanderers, Routledge, 2017, the major ethnographic and historical study of the Nāth monastic order in the field. The whole book was searched for Śabara and for Śābara, 122,607 words of it on the extraction used here. Occurrences: zero. A second reader repeated the search on a separate extraction of the same file, against śabara, sahara and savara with and without diacritics, and got the same nothing. The two extractions differ a little in length, which is ordinary variance between extractors and changes nothing about a zero. This is the one negative in the chapter that two people have obtained independently, and it is the one to lean on.

The Power of the Nāth Yogīs, edited by Daniela Bevilacqua and Eloisa Stuparich, Amsterdam University Press, 2022. A search of an extracted text of that volume, 129,907 words of it, returned no occurrence of the word. That second result is reported as a result and not as a fact about the book. A later verification pass could obtain no copy for independent re-testing, so nobody has yet been able to re-run the search and get the same answer. One unrepeatable extraction is a weaker thing than a repeated one.

One phantom needs killing while we are here. A title in circulation, “Bouillier, The Power of the Nath Yogis,” does not exist. Bouillier contributed to the 2022 Amsterdam volume; her own monograph is Monastic Wanderers. The two have been welded into a book nobody wrote, and that book now gets cited.

Now the fact that cuts the other way, and it has to be met before the word silence is used at all.

There is a Śābara inside the Nāth lineage lists.

Dasgupta prints in English the list of Mahāsiddhas from the first chapter of the Haṭhayogapradīpikā, verses five to nine, and its order runs Ādi-nātha, Matsyendra, Śābara, Ānanda-bhairava, Cauraṅgī, Mīna, Gorakṣa, Virūpākṣa, Bileśaya, and onward. Śābara is the third name in the canonical haṭhayoga text of the Nāth tradition, standing between Matsyendra and Gorakṣa’s own generation. Dasgupta then gives Albert Gruenwedel’s list of the eighty-four Siddhas from Tibetan sources, in which Śābara, or Śabarī, stands fifth, with Mīna eighth, Gorakṣa ninth and Cauraṅgī tenth.

Bagchi has the same thing from the other side. He sets out the Haṭhayoga tradition’s spiritual descent as Matsyendra, Savara, Cauraṅgī, Mīna, Gorakṣa, Virūpākṣa, Carpaṭi, and, in a footnote abstracting the Kaulajñānanirṇaya, he records Sabarī-pāda among the Siddhas worshipped at Kāmākhyā, Pūrṇagiri, Oḍiāna and Arbuda, and Sabara-pāda in the line of gurus.

So the association between the name Śābara and the Nāth-Siddha world is not thin, and it was not invisible to anybody. It is printed in the tradition’s own foundational haṭhayoga verses, in the Tibetan Siddha lists, in Matsyendranātha’s own text as Bagchi edits it, and in the standard Indian study of the Nāth cult. Chapter sixteen takes the figure up.

What that does to this chapter is precise. The zeros in Bouillier and in the Amsterdam volume are still zeros, and they still mean something: a vernacular incantation corpus circulating through the bazaar press does not appear in two modern scholarly books about a monastic order. They do not mean that the name Śābara is absent from Nāth material, because it is not. Those are two different claims and they are easy to run together.

Dasgupta also supplies the sceptical control this chapter needs, and it comes from inside Indian scholarship rather than from outside it. He argues that eighty-four is a mystical number in this tradition, attached to the postures, to the cycles of birth, to the liṅgas of the Skanda Purāṇa, and concludes that this justifies doubt about the historical nature of the eighty-four Siddhas tradition altogether. A Śābara standing fifth in a list of eighty-four is therefore evidence about a list, not about a man.

The Nāth claim on the Śābaratantras is made by Nāth texts, by two nineteenth-century manuscripts in one Calcutta collection, and by an eighteenth-century controversialist, and it is barely touched by the modern scholarship on the order. It is carried by the tradition, about itself, in its own documents. That is not a refutation. It is a precise description of what kind of claim it is: an internal one.

How late is the frame

There is an obvious chronological objection to the popular sentence, and it does not work. It is worth setting out why, because the reasoning is easy to run backwards.

The nine Nāths, the Navnāth, are a systematised list, and the systematising is later than the figures in it. That much is generally accepted, and Dasgupta says so plainly: even within the nine-Nātha tradition the lists do not agree with one another, and he prints two that do not. What this research has not established is when the roster fixed. The date is not in hand.

The objection runs like this: if the order took institutional shape in the fifteenth or sixteenth or seventeenth century, the popular sentence would be spoken from inside a frame that postdates the corpus it describes.

That is backwards on this treatise’s own evidence. Chapters three and five establish that the late corpus has no date of composition at all, and that its earliest datable footprints are a manuscript copied in 1855 or 1856 and a Calcutta printing of 1884. A frame that fixed in the sixteenth century would then predate every datable thing about the corpus by three hundred years, which reverses the conclusion.

The manuscript census tightens that rather than loosening it. Not one Śābara manuscript in the Asiatic Society’s Tantra catalogue is on palm-leaf; all are paper, most are modern Nāgara, and the one date in the block is 1855 or 1856. Whatever century the Nāth frame fixed in, the paper this corpus survives on is younger than the railway.

So the chronological objection is simply not available to this treatise. The corpus is undated. The frame is undated. The relative order of the two cannot be established from anything in hand, and a reader who wants that argument will have to get it somewhere else.

What can be said is narrower and it is not nothing. The claim runs one way in the documents. The Gorakṣasiddhāntasaṃgraha quotes the Śābaratantra; the Śābaratantra does not quote it. A citation formula is a downstream instrument. Whatever century the citing text belongs to, it belongs to a century in which the Śābaratantra was old enough to be quoted at by an author who felt no need to introduce it.

The tradition reaches behind him

Now back to the charter this chapter opened on, because it has an address and a body of documentation behind it, and both are worth having.

Chapter three set the Śābaracintāmaṇi charter out as Goudriaan reports it, in the words this chapter opened on: the mantras revealed to Ādinātha by the grace of Īśvara, proclaimed to five pupils in the languages of Gauḍa, Kerala, Karṇāṭa, Andhra and Gurjara. Two things have to travel with that sentence, and chapter three prints both in full. Those five languages are not five Dravidian languages; Gauḍa is Bengal and Gurjara is Gujarat, and pañcadrāviḍa in the title is a territorial classification of brahmin communities rather than a language family. And the charter is not a rumour about a text. It sits in a catalogued manuscript in the Asiatic Society, twenty-seven leaves, complete in ten chapters, which is neither of the two manuscripts this chapter has been arguing over. The charter text exists, it is complete, and it has an address.

What concerns us here is one clause inside it, the clause the popular retellings drop.

Ādinātha is known as Matsyendranātha’s teacher.

Goudriaan states that in a single catalogue line. Bagchi had it forty-seven years earlier and had it better, because what he works from is primaries rather than catalogue remarks. The Haṭhayogapradīpikā puts Matsyendra’s name second only to Ādinātha, that is to say Śiva. The Jñāneśvarī, the Marathi commentary on the Gītā generally dated to about 1290 CE, runs its lineage Ādinātha, Matsyendra, Gorakṣa, Gahinī, Nivṛttinātha, that last name being Jñānadeva’s own elder brother and guru. Bahiṇā Bāī, born 1628 and died 1700, a disciple of Tukārām, carries a comparable guruparamparā. Kabīr reaches for the eighty-four Siddhas in his dohās and names Gorakh and Bharthari, which Bagchi cites to the Nāgarī Pracāriṇī edition of the Kabīr-granthāvalī. Four independent Indian traditions, and Bagchi’s own caution travels with them: the Haṭhayogapradīpikā list is not very old, since like all late traditions it treats Mīnanātha as a person distinct from Matsyendra.

Follow the generations again, with all of that behind them. Gorakhnāth’s teacher is Matsyendranātha. Matsyendranātha’s teacher is Ādinātha, whom Bagchi and Lorenzen both take to be Śiva under another name. The charter reaches past Gorakhnāth, past his own guru, and lands two full steps behind him, at the origin point of the lineage. Not one of the traditions Bagchi assembles was written to settle an argument about the Śābara corpus, and every one of them holds that sequence anyway.

So where does Gorakhnāth’s name actually enter this record? As one of twenty-four in a quoted list, alongside Nāgārjuna and Jālandhara. And on the colophons of catalogued manuscripts, one of which also names Dattātreya. That is the whole of it.

The drift this chapter opened on is a general habit of traditions and it is not unique to this corpus. Attribution moves toward the most famous name in a lineage, and the further back you go, the fainter the names get, until the guru’s guru is only a word in a formula somebody else is quoting.

Notice what every claimant in this chapter is doing. The Nāth author routes the Kāpālika doctrine back through the Nātha. Kāśīnātha Bhaṭṭa routes the same list into the Kāpālikas. The charter has Ādinātha receive the mantras by grace and hand them on in five vernaculars. Not one of them says: we invented this, we made it work. Every one of them says: this is ours to authorise. Those are the moves of people securing a permission, and permissions are only worth securing where something stands in the way.

The colophons are the exception, and this chapter has to say so rather than let the pattern swallow them. A colophon naming Gorakṣa as the author of the copy in front of it is an authorship claim. It is also the smallest witness in the whole file: one unnamed scribe, on twenty-one leaves of country-made paper, in 1855 or 1856, writing about the manuscript under his own hand, with a second codex across the room carrying Gorakṣa’s name in one title and Dattātreya’s in another.

The next chapter goes to what stood in the way. It is not a rival lineage and not a rival sect. It is a theory of sacred speech that locates power in the exact phoneme of one particular language, and it had already ruled out almost every mouth that would ever say a śābara mantra from the rite it governs.

The Nāths did claim this corpus. They claimed it carefully, in writing, and in argument with a rival.

The texts that argue about it never say Gorakhnāth wrote it. Two manuscripts in one Calcutta collection do, and one of the two could not decide between him and Dattātreya.