The word, reclaimed or not

In Rayagada, and in Gajapati, the people whose name is on this book are still there.

They are called Sora, or Saora, or Savara, depending on who is doing the calling. They speak a language that is neither Sanskrit nor the regional Oriya of their neighbours. They have priests and they have priestesses, and their lexicon is native and dense for sorcery, for exorcism, for the breaking of a spell by the uttering of a counter-spell. In 2008 one of them wrote their history: Giridhar Gamang, a Sora, a former Chief Minister of Odisha, grandson of the Sora religious leader Mangey Gomang, arguing his people’s case from literary sources and from inscriptions.

That book is the one book on this subject that this treatise does not open.

Hold those two facts together, because the whole epilogue hangs off them. A national industry sells mantras under a version of these people’s name. Twenty-one chapters have now been written about it, and the single volume written from inside the name is not read here.

So begin with what that costs. A book can establish what a word meant, when it was used, by whom and to what end. This one has spent twenty-one chapters doing exactly that, and it can stand behind it. It cannot establish who is entitled to say whether a tradition carried under that word is being honoured or merely endured. That is the question the epilogue was called for, and the treatise does not have the standing to answer it.

Standing is where the last chapter ended. Standing is this one’s whole problem.

What it can do is set down what it found, refuse two comfortable readings, name what is still unreached, and say what would change its mind. The largest of the unreached things has a name and an address in an Indian city. And the one sentence everybody quotes turns out never to have been traced back to whoever first said it.

What the book found

Two sentences have to be separated here, because they are easily printed as one. One of them is a finding. The other is an inference across a bridge.

The finding. The Sanskrit lexeme śābara carries the sense “in barbarian speech.” Chapter 2 assembles the lexical record for it, and nothing in this treatise or outside it touches that.

The inference. This corpus is called Śābara because its mantras are in that grade of speech, which would make the name a linguistic category rather than a lineage. That is a bridge from the dictionary word to the title of a body of texts, and there is a named rival standing on the same bridge. Goudriaan suggests, in a single sentence in the paragraph where he identifies the corpus, that the term Sāvara may be connected with the generic name of the Saṃvara or Śambara Tantras of Buddhist Tantrism. If he is right, the name of this Hindu group is a borrowing from Buddhist Tantrism, and, in chapter 3’s own words, it says nothing whatever about speech, forests, or the Śabara people. Chapter 3 neither adopts his suggestion nor dismisses it, and tells the reader to hold two candidate etymologies at once.

The epilogue’s job is to say what the treatise thinks, so here is what it thinks and why, and here is what would change its mind.

It commits to the speech reading, for four reasons drawn from its own chapters. The lexical sense is attested and Goudriaan’s derivation does not touch it. The corpus’s own charter text, the Śābaracintāmaṇi, ascribed to Ādinātha on an ascription chapter 3 leaves contested, declares the mantras proclaimed in five named regional languages, which is the corpus explaining its own name in terms of speech. In 779 CE, in Uddyotanasūri’s Kuvalayamālā, the practitioners of the Śābarī vidyā are Vidyādharas dressed as Śabaras, so that Śabara-ness is already a costume in the earliest securely dated attestation, which is what a category behaves like and not what a lineage behaves like. And around 1574 Tulsīdās reaches for these mantras to justify writing the life of Rāma in Avadhi and seals the passage with bhāṣā bhaniti prabhāu, the power of a vernacular composition, which is a reader inside the tradition treating the name as a claim about language four hundred and fifty years ago.

Against that stands one undeveloped sentence from the field’s only cataloguer, offered as a possibility, with no evidence given in the place where it is given.

The exit condition is this. If a manuscript colophon, a text, or a piece of scholarship connects the Hindu Śābara titles to the Buddhist Saṃvara or Cakrasaṃvara literature by transmission rather than by resemblance of sound, the commitment goes, and the arguments in chapter 16 and in the two sections below go with it. Chapter 15 does not, because its costume argument concerns the ethnonym inside the Jagannātha legend and never uses the corpus’s name as evidence. A reader who wants to break this book should start there, with Goudriaan’s sentence, and not with the dictionary.

What survives on any reading is the pattern. The Jagannātha story acquired its Śabara in the retelling (chapter 15). The Buddhist Śavaripa is a hunter as literate monks imagined him (chapter 16). In print the corpus is a commercial genre with a price list (chapter 17), and in the present its footprint is a marketing footprint (chapter 18). Where documented harm exists in the world around it, it attaches to a ritual-specialist role and to an accusation, and the specialist’s part in that accusation is to confirm a conflict already under way (chapters 19 and 20).

Two more findings belong on that list, and both are of the kind that only arrives when somebody opens a catalogue.

The corpus has a shape, and the shape is late, thin and bilingual. Chapter 5 works the census out of the catalogues, and a long list of titles turns out to stand on a small number of actual works. Not one Śābara manuscript described in the Asiatic Society’s own catalogue is on palm-leaf, which in an Indian tantric collection is the strongest single date signal there is. Every one of them is paper, the script is Nāgara throughout and repeatedly modern Nāgara, most witnesses are flagged incomplete, and the extents are tiny, a few leaves at a time. Exactly one manuscript in the whole block carries a date, and it is a nineteenth-century date. The contents are bilingual wherever the cataloguers looked closely, with Hindi sections, Telugu prose, Kannada, and one description that says in as many words that its manuscript is in Hindi and corrupt Sanskrit. Then the presses arrive, in 1884 in Calcutta and after that everywhere, and chapter 17 is the rest of that story. What the catalogues show is a genuine manuscript tradition that is also a shallow one, and a print afterlife out of all proportion to it.

And the count is not one book. Chapters 10 to 12 start from a single Delhi paperback, and one paperback is a case study wearing the clothes of a claim. Eight independently compiled printed books have since been counted by the same method, with the method published, and the appeal vocabulary survives the widening. It intensifies. On duhāī, the cry of the wronged, the Delhi paperback turns out to sit at the floor of the eight and not at the centre of them, so widening the base raised the finding rather than diluting it. Two controls from the same bazaar shelf, an Indrajāla compendium and an Uḍḍīśatantra, score at or near zero on the same four terms, which means the vocabulary belongs to the Śābara compilations specifically and not to Hindi occult printing in general. Chapter 10 carries the counts, book by book. The claim moves from one book to the printed popular corpus, and it moves upward.

Where the readings landed

Several of the readings a book of this kind arrives with do not survive its own counting. These are the ones that did not, and the settled versions of them.

The oath vocabulary is not a ghost in the surviving corpus. It looks like one on a search run in a single regional orthography, and chapter 10, which counts both spellings of ān, shows that it is not. The corpus itself names duhāī and ān together in a single phrase, duhāī yā ān, as though they were interchangeable instruments.

Nor is there an opposition between appeal and oath to build on. The distinction looks natural, the appeal made in another’s name set against the oath that binds, and an Indian dictionary closes it: the Hindī Śabdasāgara glosses the two words by each other. Generalising from one regional spelling is the same operation, in the same direction, as generalising from one regional usage.

The command word belongs to one book. Chapter 11 reads ājñā, the word of command, as characteristic of the corpus, and on the eight-book comparison it is characteristic of the standing witness, which carries the highest ājñā rate in the whole independent set while the others fall away behind it. That is a partial disconfirmation of this treatise’s own reading, and it is printed as one.

The lock has a text. The exemption in the kīlana doctrine, the claim that certain mantras are exempt from the lock laid on the Kali-yuga, has a textual basis in the Mānas-mayaṅka commentary, reachable through the Mānas-pīyūṣa, and the popular compilation this treatise counts from states the doctrine in its own hits. A negative here is a negative about what has been read.

The Islamic register is dense, and by an order of magnitude. The standing witness carries pīr, the Sufi master, beside paigambar, the prophet, and Mohammadā vīr, and most of the independent compilations run that vocabulary denser again; chapter 12 has the counts. A narrow search in a single spelling comes back with one occurrence, and one occurrence turns a register into a curiosity.

The field’s principal authority on the Kāpālikas does touch this material. David Lorenzen’s own index carries Śabara-tantra, twice. In one place he quotes the Gorakṣasiddhāntasaṃgraha citing the Śabara-tantra for the twelve sages to whom the Kāpālika doctrine was revealed, Ādinātha at their head. A few pages later he observes that the presence of the tradition in unrelated sources suggests there may have been some factual justification for it. The narrow negative that does hold is this: Lorenzen never mentions Kāśīnātha Bhaṭṭa and never mentions the Kāpālikamatavyavasthā.

The charter text’s five languages are not five Dravidian vernaculars, and two of them are not Dravidian at all. The five are Gauḍa, Kerala, Karṇāṭa, Andhra and Gurjara, which is to say Bengal, Kerala, the Kannada country, Andhra and Gujarat. Pañcadrāviḍa, the word in the title, is the classical brahmin territorial division of the subcontinent’s southern half, and it is a statement about where brahmins live rather than about what family a language belongs to.

The six acts have a chapter, and it is a negative one. Chapter 21, “The six acts”, reports that the classical six-act grid, śānti, vaśīkaraṇa, stambhana, vidveṣa, uccāṭana, māraṇa, does not organise this corpus at all, and that the tradition carries its own ethical critique of the coercive members, from the upapātaka ruling in the Manusmṛti down to a Jain compendium that keeps the sixth act on its list and renames it the forbidden one.

And the 1916 gazetteer entry of Russell and Hira Lal does attach the corpus to the people. What follows from that is larger than the entry, and it is the subject of the last section of this book.

Against the easy reclamation

There is a comfortable modern reading available here, and it should be refused.

The reading goes like this. A word for barbarian has become a word for a living popular tradition. A slur has been turned inside out, worn with pride, made into a badge. The vernacular has won over the brahminical, the forest over the settlement, the crude effective syllable over the metrically perfect one. Reclamation.

It is a good story. The evidence does not support it.

The word was never reclaimed by the Śabaras. There is no point in the record, from the earliest lexical strata chapter 2 assembles to the marketing footprint of 2026, at which the people carrying the name take the name up and turn it. What happened instead is that the word was repurposed, repeatedly, by people standing somewhere else. Jain narrators used it for a costume. A Nāth charter used it for a claim about language. A poet in Avadhi used it to defend his book. Bombay and Banaras presses used it to sell cheap grimoires beside their conjuring manuals and their fireworks catalogues. A channel uses it now for the reason a press did then.

Every one of those repurposings was profitable, in reputation or in cash, for somebody who was not a Śabara.

That is not reclamation. Reclamation is what a community does with a word aimed at it. This is a word that kept being borrowed over the heads of the people it named, by parties who had, in most cases, never met one.

Against the easy condemnation

The opposite reading is equally available, and the treatise cannot make it either.

The corpus is not a demonstrated theft. Nothing this research reached demonstrates a line of transmission from the ritual practice of the historical Śabaras or the modern Sora into this body of mantra. No text, no manuscript, no ethnography traces the passage. The manuscript survey makes that harder rather than easier: a corpus whose oldest dated witness is a paper copy of 1855 or 1856 is not well placed to carry a borrowing out of the hills.

A stronger sentence than that is available, and it will not hold: that no modern scholarship asserts a descent link, and that the assertion is found only in the colonial archive at Russell and Hira Lal 1916. It is false, and where it fails is the subject of the closing section here.

What can be said, and all that can be said, is this. The descent claim is on the record from at least 1916 and it has never been demonstrated in either direction. What sits opposite it is a chain of category. The label travelled as a category of language, exactly as the dictionaries said it did, until Alexander Cunningham published The Ancient Geography of India in 1871 and identified the Śabaras of the Sanskrit and Greco-Latin sources as one people scattered across several places. Raphaël Rousseleau’s conclusion, paraphrased here because his 2022 article is in copyright, is that Cunningham turned a category into an ethnic name, and that the same move went on being made right down to the present.

A category into an ethnic name. He does not say linguistic. He says category, and putting the word “linguistic” into that phrase would make a living scholar appear to endorse this treatise’s own thesis about the word. Chapter 2 follows what Cunningham’s error did downstream.

Which means that most people using the word today are not appropriating a people. They are using a word whose history has been invisible to them, in a chain of transmission where the invisibility was manufactured four generations back and maintained by everyone downstream, including the scholars. Including this one.

What the book is for

Underneath the borrowing there is a real claim, and it is the reason any of this was worth twenty-one chapters.

The claim is that power in language is not confined to the sanctioned phonemes of a sanctioned tongue.

Set it against chapter 7, where the relation between word and meaning is eternal and unmade and the efficacy of Vedic speech rests on its being exactly the speech it is. Set it against chapter 8, where the Tantric apparatus builds an extraordinarily careful machinery around the syllable, perfects it, lists what can spoil it, and gates its use behind an initiation. Both systems are serious. Both locate the power in the form, and both keep a door on the form.

The Śābara claim is that the appeal works from wherever you happen to be standing, in whatever you happen to speak. And the vocabulary chapters 10 to 12 counted, now across eight books instead of one, duhāī and ān together and glossing each other in the Hindī Śabdasāgara, is the vocabulary of somebody with no court to go to, no lineage certificate, no Sanskrit, who cries out in a name that is not his and binds by it anyway.

Whatever else the corpus is, and it is many things, some of them shabby, the claim at the centre of it is a claim on behalf of the unauthorised.

It deserved better custodians than it has had.

What is still open

There is no specimen in this book, and that is a position rather than an omission.

Twenty-one chapters have described a genre and not once exhibited it. A specialist will notice. The decision is settled and the reasoning is short. The genre’s form is exhibited throughout, and it is exhibited from documents: in the word-counts of chapters 10 to 12, in the advertising copy of the Veṅkaṭeśvar catalogue in chapter 17, in the lexicographers’ glosses in chapters 2 and 10, in the catalogue descriptions of the manuscripts themselves. A reader finishes this book knowing what the register sounds like, what its recurring vocabulary is, who it addresses and at what price it was sold. The genre’s function is withheld: no working mantra, no vidhi, no materials, no timings, no treatment described. Chapter 20 is the reason. In a field where a ritual specialist’s word is the instrument that converts a private quarrel into a public verdict, a book that supplied operative material would be adding to the stock of that authority, and it would be joining the trade chapter 18 describes while pretending to study it. The form is the scholarship. The function is somebody else’s business, and this book has read enough case files to leave it alone.

Four boundaries on the evidence, stated plainly.

There is still no critical edition of anything in this corpus. The counting has widened from one bazaar text to eight, which answers the objection that a single paperback carries a case study and not a claim, and it does not answer the deeper one. Eight uncritical popular compilations are eight uncritical popular compilations. The counts are good evidence about what the printed popular corpus looks like, which is exactly the claim they are now used for, and they are evidence about nothing else.

Nothing here has been verse-gated. No Devanāgarī in this treatise has been through the house’s verse verification. The four digital copies of the chapter 1 verse agree with each other, and four digital copies are not four sources: all of them descend from the Gita Press line of text, and under any honest test of independence they are one witness repeated. The reading has still not been checked against a printed page, and that verse is the most-quoted thing in the treatise.

The etymological question is open, on the terms set out at the head of this chapter.

The charter text is not lost. The Śābaracintāmaṇi is extant. About fifteen witnesses are catalogued across roughly a dozen repositories, and one of them is whole: described, ten chapters, nothing missing, sitting on a shelf in the Asiatic Society at Kolkata. No critical edition exists. Every argument that leans on the charter text leans, at present, on a description of a codex rather than on that one, while the thing itself lies in an accessible library in an Indian city.

What remains genuinely unreached begins with that complete Śābaracintāmaṇi in Kolkata. Everything this treatise says about the charter text, including the five-language passage that is one of its four reasons for committing to the speech reading, currently rests on Goudriaan and Gupta’s paragraph and on somebody else’s description of the manuscript. One reading room closes it.

Beyond it: the second volume of the Śābaramantrasāgara, whose compiler announces in the first volume that a further part will draw on the sects current today, Hindu, Muslim and Sikh among them, with its Islamic material gathered from faqīrs and ʿāmils, the natural test of chapter 12’s finding about the prophet vocabulary, and open on the Internet Archive, where the eight-book search did not go. H.-G. Türstig’s study of abhicāra in the Wiener Zeitschrift für die Kunde Südasiens for 1985, the only monographic treatment of the category that the six acts systematise, which may well carry the dating argument for the crystallisation of the scheme that nobody else supplies; chapter 21 was written without it. The Nepalese-German manuscript cataloguing database, which failed to open on every attempt in this research: the catalogues already record three Nepalese witnesses of the Śābaratantra, so the claim that nobody has looked at Nepal is answerable, and a query there is the cheapest remaining test of whether the corpus is older than its surviving paper.

Two things about that second volume are easy to get wrong. It is not undigitised: it is openly downloadable, a full 600 ppi scan of Khaṇḍelvāl’s uttara bhāga from Chaukhambā Surabhāratī, the same publisher and the same series as the pūrvabhāga this treatise already counts from, and its archive metadata is bare, which is presumably why searches slide past it. And its announced division is not Muslim and Jain material. The volume’s own preface puts the scope at the sects current today, हिन्दू, मुस्लिम, सिख, Hindu, Muslim and Sikh. Jaina occurs twice in the whole volume against ten occurrences of muslim, and nothing in the preface makes Jain material half of the announced scope. So the test chapter 12 calls untestable can be run. It has not been run here.

And the book this chapter opened on. Giridhar Gamang, Sabara Tribe in Indian History, 2008, catalogued also under the surname Gomang. A Sora writing Sora history, a former Chief Minister of Odisha, the grandson of the Sora religious leader Mangey Gomang, reported by Rousseleau to have argued the positive case from literary sources and from inscriptions. The publisher’s own file returns a bot challenge and the web archive refuses the request, which is what a search that stops there finds. The book is an open download of the 2008 Kitab Mahal edition, 106 pages, sitting in the largest free repository there is, and it is unread here.

One negative has to be reported as unproven rather than as a result. Piers Vitebsky’s Sora ethnography was searched, term by term, for any engagement with the Sanskritic category, and every term came back empty. So did the controls. Words certainly in the book returned nothing, which means the tool failed and the search did not run. The honest statement is that no engagement was found and that the method could not have found one.

Each of those is a hole with a name on it. A treatise that names what it could not reach is worth more than one that quietly does not.

Who decides

Which leaves the question, and it does not have an answer.

Can a tradition be honoured under a slur? The treatise can establish what the word meant, when it was used, by whom, and to what end. It has done that. It cannot establish whether a Sora in Rayagada or Gajapati, watching a national mantra industry trade under a version of her people’s name, experiences that as tribute, as theft, as absurdity, or as nothing at all. That judgement belongs to her.

There is an ending available that says the question was never put to the people whose name it is. There is a second that says it was put once, in 1938, by an Indian scholar nobody went back to.

Both are too kind, and the way they fail is the last finding in this book.

Start with the sentence. Every account of this genre, this treatise’s included, has defined a śābara mantra by its addressee: these are the charms for the spirits of the violently dead. The sentence comes from Verrier Elwin’s footnote of 1955, and in that footnote it carries no citation at all. Follow it back and it is not Elwin’s sentence. It is R. V. Russell and Rai Bahadur Hira Lal’s, in the article “Savar” in The Tribes and Castes of the Central Provinces of India, published in 1916: the Savars’ charms, called Sabari mantras, are “especially intended to appease the spirits of persons who have died a violent death.”

That book is a colonial gazetteer. It is describing the Savars of the Central Provinces. Elwin names it, quotes its neighbourhood, and says in the same footnote that Russell and Hira Lal are not speaking of the Hill Saoras he himself studied. He prints the definition anyway. And the route by which it reached him runs through the Indian scholar in the middle: G. V. Sitapati, in the Journal of the Andhra Historical Research Society for 1943, quotes the gazetteer paragraph in full, under a heading about the Sora reputation for magic, and quotes it explicitly as a tradition that has come down to the present generation. He is reporting hearsay, and he marks it as hearsay.

Three lines further down the same page, Sitapati gives his own gloss of the term, and it is a different thing altogether. He observes that classical Indian literature refers to Sabara mantrams, and he glosses them simply as Sora magical spells. No violent death. No addressee. No contrast with any other class of charm. The restriction that fixed the genre for a century is in the quotation and not in the scholar.

Then the formulae. Sitapati prints two chants of the Sora medicine-man, one for approaching a plant before its roots are dug and one for sitting before the patient. This treatise reports what they address and withholds what they say. They address the fathers, the gods and demigods of old who planted and dug the plants, and the medicine-men of old. They do not address the violently dead. And Elwin’s own ethnography, hundreds of pages after the footnote that started all this, says that where the violently dead are named at the opening of a Saora invocation they are being cleared out of the way, invited to take a little wine and rice and go, so that they will not spoil the rite. Elsewhere in the same book he says the thing that ends the matter: that there are no standard Saora prayers, no formulas and no fixed invocations at all.

So a gazetteer line about one population, relayed as hearsay by a scholar who did not endorse it, was reproduced without a citation by an ethnographer who disclaimed it in the same breath, and became the definition of a genre. Chapter 14 carries the chain link by link.

And the direction of travel reverses.

Sitapati’s own claim, in the instalment of 1938, is that the Soras lived for a time near the Aryans of the Vedic period, took on some of their culture, and, in his phrase, “influenced the Aryan thought” in the field of magico-religious charms. That is a descent claim made by a modern Indian academic in a peer-reviewed journal, running from the tribe into the Sanskritic corpus. Followed to the place where he actually argues it, in the instalment of 1943, it turns out to rest on typological parallel: Sora healing set beside Atharvavedic healing, disease conceived as a physical thing to be expelled, plants as consecrated agents, with a public-domain translation of the Atharvan hymns printed as comparanda and a 1920s anthropology of medicine as the theoretical prop. He offers no linguistic, textual or historical evidence of transmission in either direction.

His father’s dictionary offers some, and it runs the other way. In the Sora-English Dictionary of 1938, G. V. Ramamurti records the Sora word for a charm or a spell and marks its etymology: Sanskrit, through Oriya. The Sora lexicon is native and dense for sorcery, for exorcism, for the priest and the priestess, for the act of breaking a spell by uttering a counter-spell. For the single word meaning a charm it borrowed, and it borrowed through the neighbouring regional language. Whatever the direction of ritual influence, at the level of the lexicon the traffic runs into Sora.

What is left, then, is not a story about scholarship that failed to ask the people. It is worse and it is duller. Somebody asked. Sitapati went and asked, in the language, for five years, across three volumes of an Indian journal. The answer he brought back was flattened into a gazetteer sentence about somebody else, and the sentence was what got quoted, for seventy years.

That changes what can be claimed here. It does not change what is not known. Whether Sitapati was right about the direction of influence is undecided, and his father’s dictionary is evidence against him.

This book opened with a poet who needed a warrant to write in his own language, went looking for one, found it in a class of mantra named after a people, and won his argument so completely that four hundred and fifty years later his poem is still recited by people who could not read a line of Sanskrit.

He got to make his case in his own words.

So did Sitapati, in 1938, in the journal of the Andhra Historical Research Society.

What got quoted was Russell.