The Saora, who are still there

The sentence that travelled

In 1916 the Government of India published a volume of a caste gazetteer compiled for the Central Provinces by R. V. Russell of the Indian Civil Service and Rai Bahadur Hira Lal. It runs through the tribes and castes of a territory that is now spread across Madhya Pradesh, Chhattisgarh and Maharashtra, article by article, in the manner of the genre. The article on the Savar carries, in its sixth section, one sentence about charms.

“Their charms, called Sabari mantras, are especially intended to appease the spirits of persons who have died a violent death.”

That sentence is the definition of a genre. It is the definition repeated in the popular literature, in the catalogue descriptions, and in more or less every paragraph anyone has written about what a śābara mantra is.

It is a district officer’s gazetteer line about a people in the Central Provinces, and it was never a report of anything the Sora do.

Four sets of hands held that sentence over a hundred and ten years, and every one of them made a decision about it.

The first pair wrote it. That is what a caste gazetteer is for: it states, in the flat voice of an administration, what the people of a district are held to do, so that the administration can go on knowing them. Their subject was the Savars of the Central Provinces. The Sora of the Odisha hills are not in the sentence.

Then a man who had the language got hold of it. G. V. Sitapati was a Telugu scholar, son of the man who compiled the dictionaries of Sora for the Government of Madras, and he published what he found on the Sora serially, in a regional journal in Andhra, over five years. In 1943, coming to the question of Sora adeptness in magic, he set the gazetteer’s article down entire, marked as a quotation, and called it a tradition come down to the present generation. Then, a few lines further on, with the chance to say in his own voice what the term meant, he said something else altogether, and much flatter.

The third had the field. Verrier Elwin worked in the very hills the sentence would come to be applied to, and in 1955 he published the fullest account of Saora religion that exists. Somewhere in the middle of it he wrote a footnote whose explicit business was to tell his readers that Russell and Hira Lal were describing somebody else. Into the last line of that same footnote, with no citation on it, he copied Russell and Hira Lal’s sentence.

The fourth is not a person at all. It is everything written since: the paperbacks, the catalogue entries, the introductions, the popular articles, all of them quoting Elwin, none of them aware that they are quoting a gazetteer.

The rest of this chapter is the route the sentence took, and the decision made at each pair of hands.

The footnote, and where it came from

The sentence reaches almost everybody through Verrier Elwin.

In 1955 the Oxford University Press published Elwin’s The Religion of an Indian Tribe, some two hundred and sixty-five thousand words by the machine count used here, on the gods, the dead, the shamans, the divination, the sacrifices and the funerals of the hill-dwelling Saora of Ganjam and Koraput. It is still the fullest account of Saora religion in existence. Nothing in this chapter displaces it, and a good deal of what follows is drawn from it.

In a section headed Disease caused by Sorcery, Elwin’s assessment is that dread of black magic counts for relatively little in how these people live, and that what a Saora household actually has to fear, where health and prosperity are concerned, is the dead and the gods. Then he concedes that Sitapati disagrees. Sitapati, he reports, holds a Saora to be more afraid of the sorcerer than of the spirit, and speaks of a covert body of sorcerers who take payment for injuring whoever their employer wants injured.

The footnote hangs off that concession. It opens with a citation to Sitapati’s 1943 instalment in the Journal of the Andhra Historical Research Society, and to one page of it. Then a parenthesis about a parallel in the Kathāsaritsāgara at VI.36, from which Elwin says no conclusions should be drawn. Then this.

”… When Russell and Hira Lal (vol. IV, p. 507) say that ‘the Savars are considered to be great sorcerers’ and quote the sayings — ‘The man bewitched by a Saora and the bullock tied up by a Rawat (grazier) cannot escape’ and ‘Verily the Saora is a cup of poison’ they are not speaking of our Hill Saoras. Sabara-mantras have long been famous: they are, as distinct from jadu-mantras, addressed only to the deified ghosts of those who have met with violent deaths.”

Verrier Elwin, The Religion of an Indian Tribe (Oxford University Press, 1955), p. 231, n. 2.

Look at where the citation is, and where it is not.

The reference to Sitapati stands at the head of the note, and it attaches backwards, to the sorcerer material in Elwin’s main text that the note is annotating. The definitional sentence stands at the end of the note. It carries no citation of its own. Elwin does not say where it comes from, because in the structure of the footnote he has already said: he has just named Russell and Hira Lal, quoted two of their proverbs, and denied that their people are his people. The definition follows the denial, and it is theirs.

So the most quoted footnote in this literature is a colonial gazetteer sentence of 1916, reproduced without attribution, inside a note whose explicit business is to state that the gazetteer is describing somebody else.

That footnote is the one passage of the book this treatise sets out as a quotation. It was read from a page image, and no emphasis has been added to it. Everything else of Elwin’s in these chapters is reported in this treatise’s own words.

What is printed on the next page

The volume of the Journal of the Andhra Historical Research Society that Elwin sends his reader to is the hardest document in this whole subject to lay hands on. It is missing from the Digital Library of India run, which has the volumes before it and the volumes after it and not that one. The South Asia Archive holds it behind a login. HathiTrust has it search-only and blocks the fetcher.

It has been obtained in full for this chapter: the July 1943 instalment of “The Soras”, as public-domain page scans on Wikimedia Commons, read by driving an optical-character service over the permitted page renderings one page at a time.

The page Elwin cites says what Elwin reports. Sitapati weighs the magical against the religious element and concludes that the Sora dreads the sorcerer more than the spirit, because a spirit’s wrath can be answered by propitiation while a sorcerer’s spell has to be met by a stronger spell and the sorcerer driven out. That much is a fair report.

The next page is the one nobody has cited.

Under a heading about Sora adeptness in magic, Sitapati sets out what he calls a tradition come down to the present generation, and to set it out he quotes Russell and Hira Lal’s Central Provinces article on the Savar, section 6, in full. The two proverbs are there. The sentence about the charms and the violently dead is there. He is quoting a gazetteer, and he marks it as a quotation.

Three lines further down the same page, under a heading of his own, Sitapati uses the term in his own voice for the first time. He observes that the classical literature of India contains references to Sābara mantrams, and he glosses the phrase. Sora magical spells.

That is the whole gloss. No addressee. No restriction to the violently dead. No contrast with jadu-mantras. The famous clauses are not his, and on the evidence of his own sentence he did not accept them, because when he had the chance to state what the term meant he stated something much flatter and much wider.

So the chain has three links and only one origin. Russell and Hira Lal write the sentence in 1916 about the Savars of the Central Provinces. Sitapati quotes it in 1943 as received tradition, immediately before setting it aside. Elwin relays it in 1955 without a citation, while stating that the population it describes is not the population he studied.

A sentence that has been repeated for a hundred and ten years was written once, by an administrator, about somebody else.

Three formulae, and whom they address

Two pages after the quotation, Sitapati prints the thing everyone assumed did not exist.

He gives two chants of the Sora medicine-man, in Sora transcription with English glosses. The first is used on approaching a medicinal plant before its roots are dug. The second is used sitting before the patient, at the moment of administering the compounded medicine. A page further on he records a third: words spoken to the offending spirit at the moment of sacrifice, after an amulet case of five metals holding a fourteen or fifteen drug compound, which he names as sorcery medicine, has been tied to the wrist or the waist of a person under sorcery.

None of the three is reproduced here, and nothing is said here about how any of them is delivered. That is the standing rule of this book and it does not bend for a decisive page.

What can be reported is the addressee, because the addressee is the argument.

The first chant is addressed to the fathers, the forebears, together with the gods and demigods of the past who planted and dug the plants, and it thanks them for the knowledge of what the plants do. The second is addressed to the fathers, the deities of old and the medicine-men of old; it presents the patient and the medicine, and pledges animal and bird sacrifice on recovery. The third is addressed to the offending spirit itself.

Ancestors. Old gods. The professional lineage of previous healers. The spirit currently doing the harm.

Not one of them is addressed to the deified ghosts of those who died violently.

These formulae are printed inside the very instalment Elwin cites, two and three pages after the page number he gives. They were available to anyone who opened the journal at the reference in the footnote’s first line.

The book that denies its own footnote

The second refutation is in Elwin.

Describing how Saora prayer actually works, Elwin states that there are no standard Saora prayers, no formulas and no fixed invocations. Shamans open with very long lists of names that vary considerably from place to place. Then come prophylactic prayers against sorcerers and their familiars. Then prayers that follow a general pattern while depending largely on the inspiration of the moment, with recurring clichés.

That is a flat denial, in the same monograph, sixteen pages after the footnote, that the people the footnote is used to characterise possess a fixed body of formulae at all.

And in the chapter Prayer and Trance, under the heading “The Invocation”, Elwin describes what the violently dead are actually for.

Every Saora prayer, he says, opens with an invocation, and it may run fifteen or twenty minutes. Its purpose is to drive away what is not wanted and to summon what is. It often begins by naming foreign ancestors, the familiars of sorcerers, the ghosts of the hanged and of those who hanged themselves, the murdered, those killed by tigers, those who fell from trees. They are named. They are offered a little wine and rice.

Then they are told to go away.

Elwin makes the same point earlier in the book: most ceremonies begin with invocations designed to keep sorcery away, and the shaman sometimes invites the familiars, along with the dangerous ghosts of the murdered and of suicides, to come, take wine and leave.

The violently dead do have a distinctive place in Saora ritual speech, and it is at the door. They are the class cleared out of the way so the rite can proceed. A gazetteer sentence turned that clearing into an address, and a generation of writers has taken the clearing-out for the point of the whole genre.

The rest of that invocation passage is a register analysis, and it is worth having whole. The invocation goes on to enumerate a great number of ancestors, tutelaries and gods, and Elwin notes that among them are names that have lost all meaning, whose referents nobody can now state. He lists some. Manninji, Sunninji, Jalbanji, Bulkanji, Radobanji, Raganji. Those are the same names Sitapati had recorded fourteen years earlier, in the previous instalment, where he says the fifteenth and sixteenth orders are always chanted together and so are the seventeenth and eighteenth, and that he could obtain no particulars about any of the four. Two men working separately, in different languages, found the same fossils in the same slot.

Elwin adds that the register shifts with the addressee: respectful to the sky deities, cautious to the dangerous gods, intimate and witty from a shaman to a tutelary spouse, plaintive from a widow to her husband’s ghost.

Address is exactly what the footnote claimed to be fixing. Elwin’s own field material shows it varying by the sentence.

Who recites, and how the plains got in

The page Elwin cites does one more thing, and what it does is social rather than definitional.

There are two classes of person credited with knowledge of sorcery among the Sora, and they are not the same class. The first is a secret body of sorcerers whose business is to injure enemies, or the enemies of whoever pays them. They are dreaded, and they are regarded with suspicion and contempt. That is the group Elwin quotes.

The second class disclaim practising sorcery, though they know it, and their business is to counteract it. They are not called sorcerers. They are reckoned with the medicine-men and hold an honourable position in Sora society. Sitapati records that these men chant spells to cure disease or to remedy an evil, and that whatever they do is accompanied by religious rite. The office is lexicalised in the language: Ramamurti’s Sora-English Dictionary of 1938 carries kudan-an, a priest, kudan-boj-an, a priestess, and kudan-mar-an, a male priest.

So the reciters are named, they are graded, and the grading is moral. The despised man and the honoured man use the same competence.

The language is Sora. The occasions are specific: approaching a plant before digging its roots, administering a compounded medicine, closing a sorcery treatment with the sacrifice of a fowl, a goat or a pig, and the pur-pur-an, the worship of a named spirit. No Brahmin appears anywhere in the account. No written text appears anywhere in the account.

Then, at the end of the instalment, Sitapati writes the sentence that matters most to the whole question this treatise has been asking.

Hindus of the plains, he says, have faith in the efficacy of Sora magical spells and Sora medicinal drugs. Several people of his own place consult Sora medicine-men. They have little regard for the Sora religious rites. They consent to the rites anyway, because the Sora medicine-man will not treat without them.

Read that again, because it is the only documented transmission mechanism in this entire literature.

Not a manuscript. Not a lineage of initiation. Not a Sanskrit compiler working from a tribal source. A sick person from the plains walking up into the hills, wanting the drug, sitting through a rite in a language he does not speak because that is the price of the drug, and walking back down having been present at something.

That is a route. It is small, it is undignified, it is entirely plausible, and it is the only one anybody has evidence for. Elwin cites this same page later in his own book and doubts it, at least for the Hill Saora. The doubt is on the record with the claim. What is not on the record anywhere is a better candidate.

The direction of travel

There is a lexical fact that settles one question and it has been sitting in a government dictionary since 1938.

G. V. Ramamurti spent over thirty years on Sora and compiled the Manual of the So:ra (or Savara) Language in 1931, an English-Sora Dictionary in 1933 and a Sora-English Dictionary in 1938, all published under the authority of the Government of Madras. He was G. V. Sitapati’s father. The 1938 dictionary carries the headword montoron, sometimes mantoron, and glosses it: a charm, a spell. The etymology is marked. Sanskrit, through Oriya.

The Sora word for a charm is a loan of Sanskrit mantra, arriving by way of the neighbouring regional language.

Set that against what the son published in the same year. In October 1938, Sitapati infers from Sora music, religious practice, magico-religious charms and metaphysical speculation that the Soras once lived near the Aryans of the Vedic period, took up some of their culture, and influenced Aryan thought in the field of magico-religious charms. It is inside that sentence, in a parenthesis, that he glosses the charms as popularly known by the name Sābara mantrams. A footnote on the same page promises a detailed discussion in the instalments to come.

The instalment that promise pointed to has been read.

The elaboration is in the 1943 instalment, and it is a comparison. Sitapati sets Sora healing practice beside Atharvavedic healing material, printing Griffith’s public-domain renderings of Atharvaveda VI.127, VI.129 and VIII.7 as comparanda for two ideas: disease as a physical thing to be forced out of a body, and plants as consecrated divine agents. W. H. R. Rivers’s Medicine, Magic and Religion supplies the theoretical frame.

That is the whole argument. It is a typological parallel. There is no linguistic evidence in it, no textual evidence, no historical evidence, and nothing that would establish transmission in either direction.

So the strongest statement in the literature that the corpus owes something to the people, followed to the page where its author said he would prove it, turns out to rest on a resemblance. And in the same family, in the same decade, in a dictionary compiled by the man who taught the author the language, the one word in question is recorded moving the other way.

The lexicon is not the whole story and it does not settle what the ritual owes to whom. Ramamurti’s dictionary is native and rich for sorcery, exorcism, priesthood and counter-spelling: a verb for breaking a spell by uttering a counter-spell, a dense field of terms for witch, wizard, sorceress and the evil eye, an interjection glossed as the word with which the priest opens his address, and a headword marked simply as used in incantations with no ordinary sense given at all. That last one is the register showing through the dictionary’s own conventions. The forms are withheld here.

The asymmetry is the finding. Native vocabulary, thick, for everything the ritual specialist does. A borrowed word, from Sanskrit through Oriya, for the thing he says.

The sound this treatise never heard

The Śābara corpus is handled everywhere in this treatise as text. The Sora sources have been saying, since 1931, that the thing in front of them is sound.

Ramamurti devotes a section of the 1931 Manual to what he calls tag-words. A principal word is reinforced by a partner word of the same grammatical form, rhyming with it, of roughly the same syllable count, taking the same prefixes, suffixes, infixes and particles, so that the pair comes out symmetrical and balanced. The partner may be a current synonym. It may equally be an archaic word, an obsolete word, a dialect form, a coinage or a pro-form. He enumerates ten types with examples, the ninth and tenth being the obsolete and the archaic, and he states that this is the structure of the sentence in the impassioned speech of the Soras and in their songs. He notes the resemblance of some tag-words to Santali jingles and compares the formation with couplet formation in the Tai languages.

That is a formal description of a ritual register, published by an Indian scholar working inside the language, twenty-four years before Elwin’s monograph. It is the description Piers Vitebsky picks up again in Dialogues with the Dead, where he analyses paired and parallel diction in Sora ritual speech, cites Ramamurti for the Sora material, and sets it beside K. Mahapatra on echo-formation in Gta’, N. H. Zide on the same, and R. D. Munda on parallelism in traditional Mundari poetry. Vitebsky leaves open whether the feature is Munda-specific or pan-Indian.

In a footnote to that same section of the Manual, Ramamurti records that a small number of sounds occur in hymns and chants, and that some occur in ordinary songs while not occurring in ordinary speech. That is a phonological marking of register and not merely a lexical one.

Sitapati adds the lexical marking. Parsing a divine name in the earlier instalment, he records jangadan, an old Sora word used in incantations, meaning life. He notes that the root is widely attested across Kol and Munda languages, in Santali, Mundari and Birhor, with a spread of ordinary senses: seed, kernel, stone of fruit, bone, wood, fibre, stalk. The sense “life” belongs to the chanted register alone. The same root does one job in speech and a different job inside a chant.

A word that lives only inside the chant is the signature of a real one.

He also gives the sequence. Across the earlier instalment he sets out twenty-one named orders of deities and spirits, and the ordering is liturgical rather than taxonomic. The hill spirits Manninji and Sunninji are thought of first at every ceremony and generally open the invocation with which any religious function begins. The fourteenth order has no worship of its own and is enumerated only in the invocations and chantings. The fifteenth and sixteenth are always chanted together, as are the seventeenth and eighteenth. The grandfathers and grandmothers, the ancestral spirits, come generally after all the others. The twenty-first order is the souls of the dead before the guar, not yet ancestral, hovering near the house, hungry and thirsty, seen and heard by nobody.

And then Sitapati says the thing that ought to stop anybody who writes about this corpus.

The priests will not chant the hymns except when they are actually officiating. When they do chant, he writes, the words are very difficult to commit to writing, because they are either murmured in low tones or chanted too rapidly to be followed. He therefore states plainly that his own list of deities and spirits cannot claim to be complete or accurate.

The man best placed in the world to transcribe this material told his readers, in print, in 1941, that he could not transcribe it.

Everything downstream of that sentence is partial by the author’s own declaration. And the reason is not a failure of diligence. It is a description of the object: a restricted register, not performed outside office, delivered either below the threshold of hearing or above the speed of the ear.

Elwin’s material agrees, from a different direction. Elsewhere in the same chapter he describes prayer offered by one shaman or by two, three or seven together, officiants chanting in a row, three praying in unison night and day, seven praying in chorus, and he notes that the congregation frequently pays no attention at all. That is liturgy as ambient labour. Later he records that the great official dreams and visions are recited as if they were poems, which he says they very often are.

The modern literature has caught up in two places and both are recent. Rolf Killius, writing in the Jahrbuch der Staatlichen Ethnographischen Sammlungen Sachsen in 2013, gives the only dedicated ethnomusicology of the Sora located in this research. He documents the kading-pane ensemble, reserved for ancestor ritual festivals and not used at weddings, and names its instruments, among them a C-shaped trumpet used to call the ancestors to the dance. He notes that music-making is open to anyone who takes an interest and that vocal music is predominantly performed by women, and he records that mantras precede trance rituals and are recited by the kuran. He gives no treatment of the violently dead as a musical category at all. Killius has been summarised here from a repository preview, not read in full.

Stefano Beggiora, in the Irish Journal of Anthropology in 2015, opens with a shamanic invocation of the Lanjia Saora in translation, an opening formula that summons the sonums, the shamans of previous generations, the ancestral fathers and mothers, and gods and goddesses from distant lands, to the place of worship. The text is withheld here. His source is his own Sonum. Spiriti della giungla of 2010, which therefore prints Sora invocations at length. His subject is the anital wall painting as a ritual instrument of the kuran, and what Christian conversion and NGO-driven revival have done to it.

Four independent witnesses across ninety years, two of them Indian and working in the language, agree on the formal profile. The corpus is paired. It is sequenced. It is register-marked lexically, phonologically and musically. And it is improvised within pattern, which is Elwin’s denial of fixed formulae stated the other way round.

Whatever “Śābara mantra” names in Sanskrit and Tantric usage, it does not name a fixed corpus recited by the Hill Saora, because on the testimony of the two men who watched most closely there is no fixed corpus to recite.

Vitebsky, and what a failed search does not prove

The other major ethnography of this people is Piers Vitebsky’s Dialogues with the Dead, Cambridge University Press, 1993. It is partially opened here, and the result has to be stated with its method attached.

The term śābara mantra, in any spelling, was not found in Vitebsky’s book. That is a much weaker statement than it sounds.

The Internet Archive holds the book as a lending item with the full-text endpoints dead or returning errors throughout this work. HathiTrust is blocked. What did function was Google Books in-volume search, and on that tool the search terms sabara, shabara, mantra, charm, spell, incantation, formula, magic and sorcery all returned nothing.

So did Elwin. So did Sitapati, Saora, trance and menhir.

Those are control terms known to be in the book. One of the snippets the same tool did return names Elwin and Sitapati in its own text. The index for this title is demonstrably incomplete, and a zero on it is evidence about the index. The honest position is that no engagement with the Sanskritic Śābara-mantra category has been found in Vitebsky, by the means available, and that the search is inconclusive. Nothing in his bibliographic essay of 2017 or his Encyclopedia of World Cultures entry suggests he engages with it. His interest is Sora-internal throughout. That is a reasonable expectation and it is not a demonstrated absence.

What Vitebsky does document bears on the footnote directly, and it cuts both ways.

He sorts the Sora dead by manner of dying, and a category name does three jobs at once: it names the manner of death, it names a site in the landscape where those so killed reside, and it names the collectivity of everyone who died that way. Leopard-Sonum takes those killed by leopards, with a residence at a rock outside each village and a collectivity called the Leopard-People. Sun-Sonum covers suicide and accidental death, and its space runs upward into the sky rather than horizontally across the ground. Ra’tud-Sonum belongs to paths and journeys, sudden onset, blood coughed up, with a house at a main path junction outside each village and outposts along the paths. He groups Leopard, Convulsion, Sun, Smallpox and Sorcery together as the acutely contagious deaths, the ones that propagate.

So the violently dead are a real Sora category with their own sites and their own names, and that much of the gazetteer’s intuition survives.

But Vitebsky’s whole book argues against treating any of these as a fixed class of ghost. His governing claim is that sonum is best rendered as one person’s memory of another they have known, a relational and biographical category rather than a demonology, and that a person moves between states over time. The violently dead are addressed because they are somebody’s dead. A definition that pins an entire genre to the deified ghosts of those who met violent deaths is the exact kind of static demonology his book was written to dismantle.

He also documents named speech-forms. His book prints named Sora chants with translations, and none of them is reproduced here.

The one that matters most is where he reports that specific melodic signatures belong to particular sub-categories of spirit and are chanted over ritual objects.

The corpus is musically indexed. It is not merely verbal. That sentence went into print in 1993, and it changes what kind of object the material is.

And Vitebsky supplies one more thing, in his open bibliographic essay of 2017, which is the hardest document found in this whole exercise. Elwin knew Hindi and Gondi and no Munda language, and did not know Sora. His Sora material reached him through Sora speakers, then Pano go-betweens, then a local Oriya dialect, then Hindi, then Elwin. Vitebsky calls the book the product of several filters, observes that it carries individual Sora words and long narratives in fluent English while carrying no joined-up Sora phrases, and states that Elwin’s depiction of Saora prayer could never have been realistic on any theoretical framework.

An uncited definition, restating a gazetteer about a different population, written by a man working through a four-stage interpretive chain in a language he did not have, and contradicted by his own fieldwork two hundred and thirty pages later in the same volume.

That is the foundation the whole literature stands on.

The scholarly definition, given whole

There is a second archive in which the same operation is easy to perform, and the entry is worth setting out whole.

In Hindu Tantric and Śākta Literature, Teun Goudriaan and Sanjukta Gupta describe the Śābara or Sāvara Tantras as a small group whose contents are mostly lists of incantations, with Sanskrit and Hindī alternating. That is the register sentence. It is one sentence of a paragraph that goes on to propose an etymology, weigh a rival from Buddhist Saṃvara, and say what the texts are for, which is command of the mantra belonging to Śabara and of assorted other gods, to magical ends.

So there is a subject-matter definition in the same paragraph as the register definition, an etymology from a deity of that name, and a second rival etymology from Buddhist Saṃvara. Chapter 3 owns the etymological argument and holds those rivals open. Lifting one sentence out of that paragraph and calling it the definition is a smaller version of exactly what Elwin did with Russell.

The people with an address

Now be concrete about who is being fenced off, because the difficulty of this subject comes from handling a people as a term. The sources for this section are Odisha’s own.

The Tribal and Harijan Research-cum-Training Institute in Bhubaneswar published The Saora, edited by N. Patnaik, on Republic Day 1989. It prints the district-wise Saora population from the 1981 Census across all thirteen districts of the state: 182,344 men, 187,716 women, 370,060 in all, with the largest counts in Sambalpur, Ganjam, Bolangir, Koraput and Puri. The same page gives Saora literacy at 14.47 per cent against 34.23 per cent for the state. It names the areas of greatest compactness as the Parlakhemundi sub-division of Ganjam and the Gunupur sub-division of Koraput, says that the section of the tribe known as Lanjia Saora is found in those contiguous areas, and notes that Saoras elsewhere are heavily Hinduised.

The successor institute, the SCSTRTI, in Particularly Vulnerable Tribal Groups of Odisha, volume I, edited by A. B. Ota and S. C. Mohanty, 2015, places the Lanjia Saora in the highlands of Rayagada, Gajapati and Ganjam. The two statements agree. Districts were reorganised in between, and Gunupur, which was in Koraput in 1989, is in Rayagada now.

The institute’s journal Adivasi has been carrying Saora material out of Bhubaneswar since 1955. In its 2007 volume Harihar Das and Devi Prasad Das give the Saora magico-religious specialist as Kudan if male and Kudanboi if female, a person who goes into trance, communicates with the aggrieved spirit and conducts the sacrifice; the village priest is the Buyya. The same issue notes that the Buyya’s position is ascribed and the Kudan’s is achieved.

The terminology does not agree across the sources, and the disagreement belongs in a book about a word. Adivasi in 1997, in an SCSTRTI study of eighty-six tribal medicine-men across five districts, of whom nine are Sabar, tabulates the Saora magician as Disari. Adivasi in 2007, and the 1989 THRTI volume, give Kudan and Kudanboi. Ramamurti’s dictionary of 1938 gives kudan-an and kudan-boj-an. Killius in 2013 gives kuran, and so does Beggiora in 2015. Elwin gives other terms again.

Six sources across eighty years, and no settled orthography for the office at the centre of the practice.

The point of the paragraph is not the vocabulary. It is that the question has an address. Anyone who wants to know what a Saora ritual specialist does can go to Rayagada or Gajapati and ask, in this decade, as Elwin did in his, and can read what the state’s own research institute has been printing twice a year for seventy years.

Almost nobody writing about śābara mantras has done either.

Seven districts, and the district that survived

One correlation belongs here, planted and refused.

The ActionAid Association and Odisha State Commission for Women study Witch-Hunting in Odisha, released on 20 December 2021, states that 83 per cent of the state’s witch-hunting and witch-branding cases are concentrated in seven districts, and names them: Mayurbhanj, Sundargarh, Rayagada, Malkangiri, Keonjhar, Gajapati and Ganjam. The same list appears twice more in the same report. The report’s 102 case studies were collected from twelve districts.

Rayagada is on the list. The doubt arose because Down To Earth, reporting the launch on the day of release, gave six districts and dropped Rayagada. The report is the document and the newspaper is not.

Two limits on the figure, both taken from the report itself. It gives no source and no method for the 83 per cent anywhere. And elsewhere it states the figure and then says the study collected its data from those districts because of it, which makes the percentage the sampling criterion. The 102 case files therefore cannot corroborate it.

Rayagada, Gajapati and Ganjam are the districts the SCSTRTI gives for the Lanjia Saora, and the parent districts of Elwin’s fieldwork.

The inference sits there waiting and it is refused on the record. No source in this research disaggregates witch-hunting victims by tribe. The overlap is geographic and nothing more. Converting it into a claim about the Saora would repeat, in a modern register, the operation this treatise has spent the chapter documenting. What would close it is a dataset reporting victims and accusers by community rather than by district, and nothing reached here does that. The harm ledger belongs to chapter 20, with its scale and its caveats.

Why the sentence kept being copied

The chain everybody draws is wrong in one specific way.

The drawn chain is this: a gazetteer attaches the charms to the people in 1916, an Indian scholar attaches them again and harder in 1938, and an English ethnographer detaches them in 1955. That version has a disagreement in it. Two ethnographers with different readings of the same people, and only one of them easy to read.

There is no disagreement. There is a sentence, and it was copied.

Russell and Hira Lal wrote it about the Central Provinces Savars. Sitapati reprinted it as a tradition and then declined to use it. Elwin reprinted it without a citation while denying that its people were his people, and forty pages either side of it he wrote things that contradict it: that there are no fixed Saora formulae, and that the violently dead are named in order to be sent away. Everything necessary to unmake the definition was inside those two books all along.

The mechanism is dull and it is worth naming exactly, because it is the mechanism and not the error that generalises. The English monograph is indexed, digitised, searchable and in every library. The Telugu scholar’s serial article is in a regional journal whose volume XIV sits behind a login, out of the national digital run, and blocked at two aggregators. One is easy to reach and one is not, and the easy one carries a sentence that makes reading the hard one look unnecessary. A research process that follows the path of least resistance will reproduce the sentence every time.

And the volume is obtainable. It has been on a public repository throughout, in a category anyone could have opened, and the only thing standing between a reader and the page after the one Elwin cited is the assumption that the footnote was a finding.

What is still shut

Four doors, and they are not all shut to the same degree.

The later parts of the 1943 volume, in which the second instalment was promised, are on Wikimedia Commons, which is where part 1 was got. The title page reads “VOL XIV October 1943 & January & April 1944 PARTS 2, 3 & 4”. It carries three articles and none of them is Sitapati’s: Lanka Sundaram at pp. 29 to 46, M. Upendra Sarma at pp. 47 to 48, A. S. Tyagaraju at pp. 49 to 66. Its closing note, headed EXPLANATION, reads: “This issue may be treated as the first instalment of Vol XIV Parts 2, 3 & 4. The second instalment will be issued shortly.” So the issue in which the continuation was due exists, is public, and does not contain it. That puts a floor under the point: the promised description of the ceremonies, rites and festivals, and the promised chapter on the Sora language, did not appear in the place they were promised for, and a reader can open the issue and see that they did not.

What is still missing is the further instalment that closing note announced, and the 1944 to 1945 volume, vol. XV, which is not on Commons and is not in the Digital Library of India run. Sitapati promised, in a footnote to the earlier instalment, a detailed description of the ceremonies, rites and festivals, and at the end of that instalment a chapter on the Sora language. Neither has been located in any instalment. Two independent bibliographies, Elwin’s own and Vitebsky’s of 2017, close the Sora series at 1943, which is good evidence and is not proof.

Sitapati’s “Sora Musical Instruments”, in the Bulletin du Musée d’Ethnographie du Trocadéro (Paris, 1933), has not been obtained. Elwin gives it as number 5 in both his Preface and his bibliography; one modern secondary source gives number 3. The discrepancy needs resolving against the Bulletin. His “Sora Songs and Poetry” has no recovered venue or date at all. Elwin’s remark on the latter is itself a finding: he says Sitapati published a charming account of Saora songs which he unfortunately translated into English rhymed verse. The one published Indian treatment of Sora song, by the one scholar with the language, reached print with its form destroyed.

Two institutional monographs are downloadable and unread: Savaras: A Monograph by M. V. Krishnarao, D. Hanumantharao and R. Pratap, 1972, and The Tribes of Andhra Pradesh, 1964, both from the Tribal Cultural Research and Training Institute. Sitakant Mahapatra, the senior Indian authority on the oral poetry of Odisha’s tribes, has not been read at all. That is the largest unworked Indian seam in this subject and it is unworked here.

Giridhar Gamang’s Sabara Tribe in Indian History, 2008, is an open download: the Kitab Mahal edition, 106 pages, on the Internet Archive, with a title page carrying the spelling Gamang and “First Published 2008”. It is a book on Sora history by a Sora author, and it is the obvious thing for anybody working on this word to read. It is unread here. Its optical character recognition is a Devanāgarī model run over English and is worthless, so the page images have to be read one at a time, and nothing whatever is asserted in this chapter about what Gamang says in it.

One thing this chapter’s own evidence settles, and it is worth saying flatly because the opposite is easy to assume. The Saoras’ own religion is not without mantras. Sitapati printed three formulae in 1943, Ramamurti’s lexicon carries a headword marked as used in incantations, and the register is marked lexically, phonologically and musically across four independent witnesses.

The name and the people

The next chapter goes to the one story in which a Śabara stands at the centre of a great Indian tradition rather than at its margin: Viśvāvasu in the forest with the blue god, and the temple at Puri that is said to have grown from his shrine. It is far more contested than the popular telling allows, and the affirmers are read there alongside the sceptics.

For now, the shape of this one is enough.

A district officer wrote a sentence about the Savars of the Central Provinces in 1916. A Telugu scholar with the Sora language quoted it in 1943 as a tradition, and then defined the term his own way, without it. An English ethnographer relayed it in 1955 with no citation and with the population disclaimed, in a book that elsewhere denies the Saora have fixed formulae at all. And everything written since has repeated it on Elwin’s authority, without once asking where he got it, while what the printed corpus actually says sat on the shelf, countable, in eight editions anybody could buy.

The bridge this book argues for is speech rather than blood, and that argument stands. What does not stand is the definition attached to it, and with the definition goes the tidiest thing the colonial archive ever handed this subject: a people, named as a caste, issued with a body of charms they were never asked about.

The mantras are still called Śābara.

The Saora are still in Rayagada and Gajapati, and nobody has been up to ask them.