779 CE, and the Śabaras are already a costume

In a story finished in the year 779 of the Common Era, a lord in bark and berries leans toward a young man and speaks a formula into his ear.

That is the entire transaction. No book. No leaf, no tablet, no copy of anything. A mouth, an ear, and a distance of a few inches.

Look at what the lord is wearing. Garments of bark. Guñjā, the small hard scarlet seeds with a black eye that Indian forests supply by the handful. A bow. It is the full visual signature of the Śabara as Sanskrit and Prakrit literature had been drawing him for centuries: forest-dweller, hunter, outside the settled order, outside the language.

He is not a Śabara.

Nobody in the scene is. The men in the bark are Vidyādharas, knowledge-holders, the sky-going semi-divine aristocrats of Indian story, and the forest dress is something they have put on. The Prakrit says so twice over, in a word for the clothing and a word for the act of assuming it, and it says of a young woman in the same episode that she assumed Savara-hood, as though the identity came with sleeves. What is being spoken into the ear is a sāvara-vijjā, in Sanskrit a śābara-vidyā, a Śabara knowledge.

And the author stops to explain none of it.

That is the finding, and everything else in this chapter is the work of holding it steady. A novelist who pauses to gloss a costume has already lost the effect he was reaching for. This one does not pause. He dresses his sorcerer-princes as forest people, hangs a forest people’s name on the knowledge they are carrying, and moves on, apparently expecting the scene to read as ordinary to whoever picks the book up.

Which points, on the plainest inference available from that silence, at an audience that already had the picture. A reader in 779 who could take the scene as it stands knew what a Śabara looked like in a story, knew what kind of knowledge came dressed that way, and needed no more prompting than a handful of berries and a bark garment. The costume was legible long before anybody thought to write down a definition of it.

Twelve hundred years ago, at an early and securely dated place where this word can be caught, Śabara-ness is already something a person puts on.

The man who wrote the scene was a Jain monk. His name was Uddyotanasūri, his book is the Kuvalayamālā, and it is a campū, a composition that moves between prose and verse, written in Prakrit, and a romance besides. Kings, curses, rebirths, forests, merchants, a story that keeps opening into other stories. Somewhere in the middle of it, across two sections numbered 218 and 219 in the modern edition, a śābara-vidyā is named, and named more than once.

That is what this chapter is about. One episode in a Jain novel, and what it tells us that nobody quoting it seems to want to hear.

A date you can put on a calendar

Start with why the year can be stated at all, because almost nothing else in this file can be.

A Jain monk finished a book on a day the book itself counts out.

Not a decade. Not a reign. A day. The closing verses of the Kuvalayamālā say that seven hundred years of the Śaka era have run, less one day, and that the work was finished then. Those verses have been read here in the Prakrit of the edition prepared by A. N. Upadhye for the Siṅghī Jaina Granthamālā. The volume’s own Hindi front matter, by the series director Jinavijaya Muni, carries the conversion and uses it: Śaka 700 answers to Vikrama 835, which is 779 of the Common Era, and Jinavijaya had already used that colophon in 1919 to reset the date of Uddyotanasūri’s teacher Haribhadrasūri, against the traditional Vikrama 585.

Very little in the early history of this subject stands on ground that firm.

Christine Chojnacki and Hampa Nagarajaiah’s 2018 translation is titled A Jain Novel from 779 AD. That title reproduces the standard conversion. It does not derive it independently, and this chapter will not present it as a second witness; it is the same computation, quoted.

A bibliographic note, because the two volumes get cited as one. Siṅghī Jaina Granthamālā 45 is Part I, the Prakrit mūla with the critical apparatus and the Sanskrit and Hindi front matter. Upadhye’s long English Introduction, where the author, the date, the sources, the language and the cultural material are argued out, is in Part II, Siṅghī Jaina Granthamālā 46. This chapter’s page references are to Part I and are correct for the text. Part II is free in full at jainqq.org, its section on Uddyotana as author running from page 96 to page 112. It has not been read for this treatise, and no claim is made here about what Upadhye argues in it. Wherever Upadhye’s reasoning is wanted rather than his text, that is the volume to open, and nothing stands in the way of opening it.

Now what the date is worth, because the temptation is to make it worth the wrong thing.

It is not the earliest anything. No source in front of us claims that the Kuvalayamālā holds the first attestation of the term, and this treatise will not claim it either. Sanskrit and Prakrit literature is vast, unevenly edited, and full of unread manuscripts. Somebody may find the word earlier next year. The honest description is: an early and securely dated attestation.

The value is in the second half of that phrase. Almost everything else in the Śābara file floats. The Śābara Tantras catalogued by Goudriaan and Gupta have no critical editions and no firm dates. The Śābaracintāmaṇi has a charter myth and a disputed author. The Nāth claim on the material runs through a text whose own date is argued. The popular literature, which is where most readers meet the subject, has no dates at all: it has the Kali age, which is a theological time rather than a historical one.

And here is a Prakrit romance that counts out the day it was finished.

That is the interest, and it is worth being exact about its limits. A dated floor under an eighth-century Jain use of a word. Whatever a sāvara-vijjā was understood to be by Uddyotanasūri and his audience, it was understood to be that, in a literary work with a colophon, before the end of the eighth century. Nothing in this chapter licenses carrying that understanding forward to a corpus a thousand years later. Chapter five will show why.

The place of composition has not been verified here. Upadhye’s praśasti discussion in Part II would fix it. It is not load-bearing for anything argued here.

The word in the primary is vidyā

Now a distinction worth slowing down for, because it is a small lesson in how the modern literature on this subject was built.

The compound in the primary text is sāvara-vijjā, the Prakrit of Sanskrit śābara-vidyā. Not śabaramantra. The text speaks of the सावरीए महाविजाए, the Sāvarī Mahāvijjā, and of the महासावरी विजा, the Mahāsāvarī vijjā, and in the next section of a treasury of सव्व-सिद्ध-सावर-विजा, all the perfected Sāvara vijjās. The word running through the passage is विजा, vijjā, Sanskrit vidyā, knowledge, and in its technical sense a formula.

Every Devanāgarī string in this chapter carries व in the middle and not ब: सावरीए, महासावरी, and so on down the chapter. That is the reading this project takes to be Upadhye’s, and it is only as good as the scan behind it. A second machine-read witness obtained independently reads साबरीए, महासाबरी and साबर-विजा-कोसो with ब, and mixes सवर with सबर for the ethnonym inside three lines, which is the signature of exactly the confusion named here rather than a second vote for either reading. The ethnonym in these lines is spelled सवर, savara, with a short first vowel and the same व. What would close it beyond doubt is Upadhye’s pages 132 and 133 photographed rather than machine-read, because the scan this project holds is exactly the kind of scan that turns व into ब.

Where did śabaramantra come from, then? From English, at least on the route that leads to the reference sites. V. S. Agrawala’s “A Cultural Note on the Kuvalayamālā of Uddyotanasūri”, published in the 1970 volume, paraphrases the passage into English and renders the term with the word mantra, which for an English-reading audience is the obvious and reasonable choice: mantra is the word that has travelled, vidyā is the word that has not. That is what a translator does.

The date on that volume is not the date of writing. Vasudeva Sharana Agrawala died in 1966, so the note appears in the 1970 volume posthumously and was written at some earlier date this chapter cannot fix. And Agrawala 1970 has not been opened for this treatise; that mantra enters through his English paraphrase is reported at second hand. What would close it is the volume itself, at the page where the passage is rendered.

The online reference site wisdomlib, which is where a very large share of casual research on Indian religious vocabulary now begins, carries śabaramantra as a Sanskrit headword with the Kuvalayamālā cited behind it. Two things can be said about that with confidence. The headword exists. The passage it cites does not contain the word. How the lemma arose is not established here. The route from English paraphrase to Devanāgarī headword to downstream citation is plausible, and it is not demonstrated.

One conclusion does not follow from any of it: that śabaramantra is a term no primary text uses. That is false, and this treatise’s own chapters refute it. Tulsīdās writes साबर मंत्र at Rāmcaritmānas, Bālkāṇḍ 1.15.3, four centuries before Agrawala, and chapter one quotes the line at length. Kāśīnātha Bhaṭṭa, in the eighteenth century, in Sanskrit, calls the twelve-and-twelve list the proclaimers of the Śābaramantras, and Goudriaan reports him. The claim that survives is narrow and it is enough: this passage, in 779, does not say mantra. It says vijjā.

This matters beyond tidiness, because vidyā and mantra are not simply two words for the same object. The convention in Tantric usage, broadly stated, is that a vidyā is the formula of a goddess, feminine in grammar and in reference, while mantra is the wider and default term. The passage here is consistent with that: Sāvarī and Mahāsāvarī are feminine forms, and what is being cultivated in the story is addressed as a Mahāvijjā, a great vidyā. Reading mantra over the top of that quietly erases the gender of the thing being named.

The Tantric vidyā / mantra convention is stated here from general usage. No source consulted for this treatise fixes it specifically for eighth-century Jain Prakrit narrative, where the terminology may work differently.

Write sāvara-vijjā, or śābara-vidyā when the Sanskrit form is wanted. It is what the page says.

Into the ear

The line of the episode most often quoted is one clause standing under the §219 head: महासावरी विजा कण्णे. The Mahāsāvarī vijjā, into the ear.

In the story, the overlord in Śabara dress hands it to the youths at the ear, one person to one person. That is a narrative detail in a Jain romance, and it is reported here as one. Nobody in the scene is a Śabara; the giver is the savarāhivai, the Savara-overlord of the story, and the story has already told us that these are Vidyādharas who have put on Śabara dress.

Sit with that for a moment before moving on, because it is doing more work than it looks. The whole modern life of this material is printed. Chapter three ended on the finding that the genre was bilingual before anybody called it folk, and bilingualism is a fact about texts; the corpus as a reader meets it today is a cheap edition, a paperback, a PDF, a video. Every popular account of Śābara material is a book about a book.

The one early, securely dated scene of this material in use that this research has reached has no book in it. It has a mouth and an ear and a distance of a few inches.

This is not a claim that the Kuvalayamālā is describing a real transmission practice among real people. It is a Jain romance and it is describing a scene. What can be said is narrower: an eighth-century author put that scene in front of an audience without explaining it, which suggests he expected it to read as ordinary. A formula of this kind moves by being spoken into a listener’s ear, from one person to one person, in a forest, by a man with the authority to give it.

Nothing in that scene is reproducible and nothing here is meant to be. This treatise does not print working formulas, does not describe rites, does not give counts or materials or times, and this chapter breaks none of that. What it records is that the text names a vidyā, and names the manner of its handing over as speech into the ear. The content is not on the page for us. It never was.

The book keeps the secret by not being the medium.

The men in the bark and the berries

Here is the finding, and it is the reason this chapter exists.

The practitioners in the episode are Vidyādharas. The word means knowledge-holder, and in the narrative literature of India the Vidyādharas are a class of semi-divine beings who travel through the air and hold vidyās: powerful, glamorous, aristocratic, entirely at home in Jain and Buddhist and Brahmanical story collections alike. They are not a forest people. They are not a tribe. They are, in the machinery of Indian romance, closer to what a later literature would call sorcerer-princes.

And in this episode they are dressed as Śabaras.

The Prakrit word has now been read. It is वेस, veṣa, dress or guise, governed by पडिवण्ण, pratipanna, assumed or taken on: पडिवण्णो पत्त-सवर-वेसं of one figure, पडिवण्ण-सवर-वेसाण of the group, and of the young woman पडिवण्णा सवरित्तणं, she assumed Savara-hood. “Dressed as” is right. “Disguised as” would also be defensible. The treatise keeps the weaker word, now by choice rather than by caution, and every later chapter inherits it.

Now read the wardrobe this chapter opened on the way a philologist has to. Berries, bark, bow: it is a wardrobe, not a working kit. Nothing in it is an ingredient, nothing in it is attached to the vidyā as a requirement, and the text supplies it for the same reason a novelist supplies a costume. This treatise owns that inventory here and does not reprint it later.

Section 219 then gives us a Vidyādhara overlord named सवरसीलो, Savarasīla, described as सव्व-सिद्ध-सावर-विजा-कोसो, a treasury of all perfected Sāvara vijjās. He is a former overlord, one who has renounced; it is his son, called Savara-senāvai, the Savara commander, who installs the crystal image in the cave. A Vidyādhara. Carrying a Savara name. Holding the Sāvara knowledge. And passing the name down to a son who carries it too.

The renunciation and the son’s installation of the image are both in §219 as printed. Which of the two the treasury epithet attaches to has not been pinned down here beyond the word order, and Upadhye’s own construing of the sentence would close it.

The episode is therefore not an ethnographic report, and it should never have been read as one. Nobody in it is a Śabara. What the text gives us is a literary convention already fully formed by 779: to signal that a body of potent, outside-the-fence knowledge is in play, dress your characters as Śabaras and put the word śābara on the knowledge.

Twelve hundred years ago, Śabara-ness is a costume.

The same costume, centuries later, in another religion

The costume is where this chapter can go furthest, and it can go there on sourced ground rather than on assertion.

Turn to Shashibhusan Dasgupta, Obscure Religious Cults as Background of Bengali Literature, the Calcutta doctoral study published in 1946, which remains one of the fullest Indian treatments of the Siddha and Nāth material. Dasgupta gives, in English, the caryā songs attributed to Śabara-pāda, a Buddhist Siddha of the Apabhraṃśa song tradition. Discussing the lists of eighty-four Siddhas, he places the Buddhist Siddhācāryas of that corpus in the tenth to the twelfth century of the Common Era. The Śabarī girl stands on the high mountain, decked with peacock feathers, wearing a garland of guñjā. The singer calls out to the exhilarated Savara, the mad Savara. The Savara spreads a bed.

Guñjā berries. A mountain. A Savara. Two to four centuries after Uddyotanasūri, on Dasgupta’s own dating of the singers, in a different religion, in a different language, in a genre with nothing else in common with a Jain Prakrit romance.

That gap is a gap between two texts, and it is stated loosely on purpose, because the man may not be one man. No source in front of us supports a round figure for it, and chapter five cuts against one: Dvivedī, at page 140 of Nāthasampradāya, reports Benoytosh Bhattacharyya placing a Śabarapāda in the middle of the seventh century, which is before the Kuvalayamālā rather than after it, while the singers Dasgupta dates belong to the tenth century and later. The Śabarapāda of the caryā songs and the seventh-century Śabarapāda of chapter five may well be two men wearing one name. That is chapter five’s own thesis about this word, and it applies here.

A page status. Open scans of both cited editions are on the Internet Archive, and the text layer of the Bagchi volume reads its Roman introduction correctly. The Dvivedī page cited just above, page 140 of Nāthasampradāya, and the Bagchi page 19 cited at the end of this chapter, were read for this project and have not been re-checked against a page image by anybody outside it. Any reader can open both books and check them.

That is the point, and it is the strongest thing the 779 passage supports. The Śabara costume is a stable literary convention of the Indic story-world, shared across Jain, Buddhist and Brahmanical writing, and it is not ethnography in any of them. Neither text is reporting on a people. Both are reaching for the same set of signals.

Dasgupta supplies a second datum, and it is larger. He reports, from a manuscript of the Laghukālacakratantrarāja-ṭīkā known as the 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.

Read what that sentence is doing. A Buddhist Tantric commentary of the first rank states, as a principle, that mantra-teaching is delivered in non-Sanskritic speech, and names the Savaras as the example. Not as a concession. As a description of how the Mantra-naya works.

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.

Say plainly what this does and does not do. It does not connect the Kuvalayamālā to any later Śābara corpus. It shows that the association between mantra in non-Sanskritic speech and the name of the Savaras is attested in a second tradition, independently, and is treated there as a doctrinal principle rather than as a curiosity.

A cave, and a crystal Ṛṣabhanātha

One more feature of the episode, and it is the one most often left out of the summaries.

The scenery of the episode, and it is scenery, puts the Vidyādharas in their Śabara dress in a mountain cave, before a crystal image of Ṛṣabhanātha. The story dedicates the third of its kāyotsargas, a Jain observance named here and not described, to the Sāvarī Mahāvijjā. These are the furnishings of a romance. They are not instructions, and this chapter reports them the way one reports the set of a play.

Ṛṣabhanātha is the first tīrthaṅkara, the first of the twenty-four ford-makers of Jainism. Uddyotanasūri is a Jain monk. The Kuvalayamālā is a Jain work, written to Jain ends, with a Jain moral architecture running under its romance.

Say plainly what that means. An early and securely dated attestation of a śābara-vidyā sits inside a tradition that is not Hindu at all.

Not a Śaiva tantra. Not a Nāth text. Not a Purāṇa. A Jain Prakrit romance, performing Jain devotion in front of a Jain image, using Śābara colouring for its atmosphere and śābara-vidyā for its power.

There is a dichotomy to refuse here, and it is tempting: that if the category were a lineage it could not travel like this, and that because it travels like this it must be a register. That imposes two modern categories on a literature that used neither. Vidyā material and Vidyādhara lore move constantly between Jain, Buddhist and Śaiva narrative without being either registers or initiatory lineages in the Nāth sense, and Dasgupta’s Buddhist Śabara-pāda songs are a demonstration of exactly that movement. The chapter’s own strongest scene, moreover, is transmission-shaped: a formula given at the ear, one person to one person, by a man with authority to give it.

What survives is a statement about eighth-century usage, and it is worth having. By 779 the Śabara costume and the word attached to it were portable enough for a Jain monk to pick both up and use them without explaining either.

That he was confident his audience would recognise the signal is an inference from the absence of explanation, not a statement any source makes about his intentions.

Whether the word in 779 and the word in a nineteenth-century Hindī pamphlet name one category or two is not settled here, and chapter five will argue that it cannot be settled at all. Chapter three sets out the standing rival to this treatise’s reading of the name, Goudriaan’s suggestion of a connection with the Buddhist Saṃvara or Śambara Tantras. If that connection holds, the eighth-century sāvara-vijjā and the late Śābaratantras may be a homonym doing two unconnected jobs, and this chapter’s finding stays a finding about the Kuvalayamālā alone.

सवर, सावर, साबर

Finally, the spelling, which is not a footnote. And this chapter has just walked into its own subject.

Upadhye’s main text alternates two forms across the episode. The ethnonym is सवर, savara, short first vowel. The name of the knowledge is सावर, sāvara, long first vowel. Both carry व in the middle.

The apparatus has been read, at the two pages, and what it records is narrower and more useful than any general claim about wavering spellings.

Where the printed text gives the epithet सव्व-सिद्ध-सावर-विजा-कोसो, the apparatus records a manuscript reading सिद्धसवर, with the short vowel. Where it gives सावर-विजाहिवइणो, a manuscript reads सवरविज्जा, again short. Two copyists, in the two places where the compound names the knowledge, wrote the ethnonym form instead.

Now be a philologist about it.

शबर to शाबर is not instability. It is vṛddhi, the regular lengthening Sanskrit uses to build an adjective of belonging from a noun, and chapter seven sets the derivation out. Prakrit intervocalic b to v, and Avadhi palatal ś to dental s, are not instability either. They are what every Sanskrit word does when it crosses into those languages. By that test śābara is exactly as stable as rāja or bhāṣā, which is to say it is behaving normally. A conclusion that words with stable homes do not do this would be a philological error, and nothing in this treatise rests on one.

What is left is small, and it is real.

First, the apparatus variants above: in the eighth-century transmission itself, scribes did not always hold the derivational distinction between the people and the knowledge, which is a fact about copyists rather than about the word’s home. Second, printed editions correcting their own base texts in opposite directions. The Gorakṣasiddhāntasaṃgraha, cited in the last chapter, prints the Sanskrit citation formula with शाबरतन्त्रे, palatal, while the Hindī gloss facing it on the same opening reads साबरतन्त्र, dental. The mūla of the Rāmcaritmānas at Bālkāṇḍ 1.15.3 reads साबर. Goudriaan’s survey enters the group as Śābara and immediately offers Sāvara in a parenthesis. P. C. Bagchi, editing Matsyendranātha’s Kaulajñānanirṇaya in 1934, refers to the same twenty-four-Kāpālika passage under the title Sāvara Tantra.

That is the honest residue. The word takes the regular form of whatever language it is standing in, and editors and cataloguers reach for different regular forms of it without comment. It is a reason why spelling cannot be used as a test to separate one Śābara text from another, which is chapter five’s problem. It is not evidence about how the word travelled, and this chapter does not offer it as any.

What 779 leaves us with

Strip the episode back to what it can carry.

A Jain monk finished a Prakrit romance whose closing verses count out seven hundred Śaka years less a day, which is March of 779. In it he names a sāvara-vijjā, calls it the Sāvarī Mahāvijjā, has it spoken into the ear of young men by an overlord in Śabara dress, and gives a renounced Vidyādhara lord named Savarasīla the epithet of a treasury of all perfected Sāvara vijjās. The men doing this are Vidyādharas in bark and berries with bows, worshipping a crystal Ṛṣabhanātha in a cave. Two to four centuries later a Buddhist Siddha called Śabara-pāda, or some other man of that name, sings of a Śabarī girl in peacock feathers and guñjā on a mountain, and a Buddhist commentary of the first rank, whose century is nowhere dated in this treatise, states that mantra-teaching goes into non-Sanskritic speech and names the Savaras as its instance. And no primary in the 779 passage says mantra: the word is विजा, vijjā.

The next chapter goes to the other early trap in this material: two entirely different things wearing the name Śābara, one of them a title in a canonical list compiled around 800 of the Common Era, in the division of scripture that treats poison, and one of them a corpus whose earliest datable footprint in this research is print, at Calcutta in 1884.

The costume is older than the corpus.

Nobody has yet found the person underneath it.