The word they were given

There are people living today in the hills of southern Odisha, in Rayagada and Gajapati districts, who carry the name this whole book is built on. They did not choose it. It is written several ways in the record, Sabara and Saura and Sahar among them, and every one of those spellings was set down by somebody else.

That is the ordinary shape of a name like this. It is applied across a line, from the settled side, to the people on the far side of it: the ones up in the hills, the ones who live by the hunt, the ones out there. A word of that kind begins as a rough description of where a people lives. It rarely stays one.

By the time it is fixed in a Sanskrit lexicon it means a wild mountaineer tribe, and by the time the same entry has finished it means any savage at all, standing in a row with two other peoples’ names used the same way. A people has turned into an adjective, and the adjective is an insult.

Then the adjective does something stranger, and this is the turn the chapter exists for. It stops describing a body and starts describing a mouth. In the same dictionaries, just under the entry for the people, the same word names a grade of speech: a low form of Prākṛt, the sort spoken by mountaineers. The classical Indian manual of theatre has the convention from the other side, sorting the rough dialects of the stage by the peoples they are supposed to belong to, and handing the Śabara dialect to herdsmen and to men who range the forest. A playwright reaching for it never had to meet a Śabara. He had to know what a Śabara sounds like on a stage.

Hold those two things together, because everything in this treatise runs through the join. A word that names a people also names speech that is unfit for the high register. Any writer who wants to say this is not proper speech has the word ready to hand, and does not need the people at all.

Centuries later, and by people who had in all likelihood never met a Śabara, that word was attached to a whole class of mantra: blunt, vernacular, unmetrical, effective, and disreputable.

Somewhere between the dictionary and the street it picked up a third life as well. In Marathi it means sorcery.

A people, a slur, a grade of speech, a kind of magic. Four loads in one word, and the people at the beginning of the road are still there at the end of it.

The oldest shelf this chapter pulls the word off is an epic, and in the epic the Śabaras come out of a cow.

Born from the dung

There is a scene in the Mahābhārata in which a sage’s cow gives birth to armies.

Vasiṣṭha owns a cow called Nandinī. A king tries to take her. She will not be taken, and out of her body she throws up fighters to defend herself. The Critical Edition, at Ādiparvan 1.165.35 to 36, sorts them by the part of her they came out of. From her tail, the Pahlavas. From her dung, the Śabaras and the Śakas together. From her urine, the Yavanas. From her foam, a long line of others: Puṇḍras, Kirātas, Dramiḍas, Siṃhalas, Barbaras, Daradas, Mlecchas.

That is the origin the epic gives them. Not a migration, not a homeland, not a founding ancestor. A cow’s body, in the same breath as the Śakas, in a list that also produces the Pahlavas and the Yavanas.

Nothing else is said of them in these verses. No conduct, no way of life, no crime. Only the part of the cow, and the company they are listed in.

And the list is a list. It carries no graded scale. Nothing in the verses says dung outranks urine or is outranked by foam, and a scale read into them would run against the tradition’s own valuation, in which gomaya and gomūtra, cow-dung and cow-urine, are two of the five substances of the pañcagavya and are purifying. The scene is a cow producing armies to defend a sage, which is not straightforwardly a scene of contempt either. What the passage supports is one statement and this chapter holds to it: the Mahābhārata places the Śabaras among peoples produced from a cow’s body alongside the Śakas, the Pahlavas, the Yavanas, the Kirātas and the Mlecchas. That is a list of outsiders. It is not a ladder.

Everything that follows in this treatise depends on a single word, and by the time the word reaches a mantra it has been in company like that for a very long time. Śābara is the adjective. It is the word attached, centuries later and by people who had likely never met a Śabara, to a whole class of mantra: blunt, vernacular, unmetrical, effective, and disreputable. To ask what a śābara mantra is, you have to ask first what the Śabaras were taken to be.

The answer changes depending on which shelf you pull it from.

What the lexicons hold

Start with the plainest sense. Monier-Williams gives śabara as “a wild mountaineer tribe in the Deccan,” and then records that in the later language the word is applied to any savage or barbarian, equating it there with kirāta, pulinda and bhilla. The specific becomes the generic. A named people becomes a slur.

Only the opening phrase there is verbatim from the lexicon, and only it carries quotation marks. The rest of the entry is paraphrased.

The Sanskrit lexicons put them in company. Kirāta, pulinda, bhilla: the hunter peoples, the forest peoples, the ones outside the settled agrarian order. A Kannada dictionary keeps the definition almost verbatim, “a wild mountaineer tribe, the members of which live by hunting.”

Hunting. Mountains. The forest edge.

A note is owed on how those entries reach us. Almost the whole lexical apparatus of this chapter comes from a single first-pass sweep across an aggregating reference site, which is the lowest evidential tier this project admits at all. The aggregation routes in each case to a named primary: Monier-Williams for the Sanskrit, the Kannada dictionary for the Dravidian entry, Hemacandra for the Jain classification, the Pali sources for the Buddhist. The routing is real. The aggregation is convenience. Not one of those entries has been opened in its own edition, and where a claim in this chapter stands only on the aggregation, the sentence carrying it says so.

The oldest name on the list

The earliest explicit reference this research has reached stands in the Aitareya Brāhmaṇa, at VII.18, numbered 33.6 in the other of the two systems in use. That is a statement about how far the reading went and not a claim about the record. The locus has been checked in Martin Haug’s translation, which carries the curse and the list of peoples in the order given below. What is still wanted is the Sanskrit at the same place, read beside a translation from an Indian editorial tradition.

The context is a curse. Viśvāmitra’s sons refuse him, and he curses them, and their descendants become the Dasyus. The text names them in a row: Andhras, Puṇḍras, Śabaras, Pulindas, Mūtibas. These are the peoples on the far side of the line, and the Brāhmaṇa explains their existence as a punishment that hardened into a bloodline.

Two things are happening in that sentence at once, and they are worth separating.

One is a genuine ethnographic memory. Somebody knew there were Andhras and Śabaras and Pulindas out there, and knew them well enough to list them in order.

The other is a theory of why they are out there, which is that a sage cursed them.

The list is data. The curse is ideology. The tradition hands you both in the same breath, and the reader’s whole job is to keep hold of one without swallowing the other.

Hold on to the date, too, because a later section of this chapter has to answer for it. A people called Śabaras are being named as a people, in a list of peoples, in Vedic prose. Whatever the nineteenth century did to this word, it did not invent the practice of reading it as the name of a population.

The Greek and Roman names, and who matched them

Outsiders are said to have seen them too. Ptolemy is reported to record a people he calls the Sabarai, and Pliny a people called Suari: two Mediterranean geographers writing about a subcontinent they knew mostly by report, both catching a name that sounds like this one, in roughly the right part of the world.

That pair is not independent corroboration. It is one identification, made once and then repeated, and whose identification it was is unsettled.

Rousseleau’s 2022 article gives it to Alexander Cunningham. Russell and Hira Lal, writing in 1916 in The Tribes and Castes of the Central Provinces of India, give it to George Grierson, and quote him from the India Census Report of 1901: “The most southerly forms of Munda speech are those spoken by the Savars and Gadabas of the north-east of Madras. The former have been identified with the Suari of Pliny and the Sabarae of Ptolemy.” Grierson’s own sentence uses the passive and names nobody. What Russell attributes to Cunningham directly, citing Cunningham’s own survey reports, is a different and stranger claim: that savar derives from the Scythian word for an axe, on the ground that the Savars are rarely seen without one. Russell prints that opinion, calls it highly speculative, and prints Crooke’s rebuttal beside it.

So the honest form is this. Some nineteenth-century scholar matched two Greco-Latin names to a Sanskrit one, the match was in general circulation by 1901, and this chapter cannot say from the pages it has opened whether the first to make it was Cunningham, Grierson or someone earlier. Whether the Greek and Roman geographers were describing the people the Sanskrit texts call Śabaras is undetermined, and this chapter does not need it to be determined.

Neither the Ptolemy nor the Pliny passage has been opened in a named edition, and neither has Cunningham’s own page. What would close it is the Geography and the Naturalis Historia at the passage, and Cunningham beside them. No part of this chapter’s argument stands on any of the three.

Outside dharma

The Jain and Buddhist material is harder than the epic, and this section is the weakest-sourced in the chapter. It is written to say so in every sentence.

The aggregation reports Hemacandra, in the Triṣaṭṭiśalākāpuruṣacarita at 2.3, classing the Śabaras as a subdivision of the mlecchas, and it reports his definition of that category as people who do not know even the word dharma. If that is his sentence, it is one of the coldest in the corpus, because it describes people who lack the vocabulary rather than people who practise badly. The passage has not been opened in a printed edition and the wording above is the aggregation’s, not Hemacandra’s.

The same sweep reports Śabaras in the Pali canon as an aboriginal tribe, reports the Mahāprajñāpāramitāśāstra listing Śavara and Andhradeśa among the low places of birth, and reports the Milindapañha using Savara as a place whose inhabitants cannot appreciate red sandalwood. Each of those would be worth a section if it were sourced. None of them is, and no statement about how far these traditions agree can rest on four unopened texts.

What can be said is a direction and not a finding. In the material reached here, the Jain and Buddhist sources place the Śabaras where the brahmanical sources do, outside the order, and none of it has been read at page level. Chapter fourteen has to go back to this ground properly.

A grade of speech

Now the entry that changes the shape of the argument, and it comes from the same dictionaries, sitting quietly a line or two below the ethnonym.

Śābara, the adjective, is defined as a low form of the Prākṛta dialect, spoken by mountaineers.

Read that again. It is not a race. It is a register. It is the name of a way of speaking, graded and placed on a scale, at the bottom.

The dramaturgy has the same convention from the other side. The Nāṭyaśāstra, the classical manual of theatre, assigns speech to characters by type. Kings get Sanskrit. Others get Prākṛt. And at 17.49 the manual lists the vibhāṣās, the rough dialects of the stage, by the peoples they belong to: Śakāra, Ābhīra, Cāṇḍāla, Śavara, Dramila, Odra, and the speech of the forest-dwellers. Six verses later, at 17.55, it puts Śāvarī in the mouths of herdsmen and those who range the forest.

The verse number is not stable either. Hazārīprasād Dvivedī, in Hindī Sāhitya kī Bhūmikā, third revised edition of 1948, discusses the same passage and cites the śabara, ābhīra and cāṇḍāla listing at 17.48. This is a recension and numbering difference between printed editions rather than an error by anyone, and this treatise follows the numbering that gives 17.49 and 17.55. A reader checking Dvivedī will find an off-by-one and should not take it for a mistake.

Two cautions. The manual spells the word śavara and śāvarī, with the palatal sibilant and a short first vowel, which is one more entry in the spelling ledger this treatise keeps. And no verse in chapter 17 says that a Śabara character on stage speaks vibhāṣā. What 17.49 does is make the Śabara dialect one of the seven named vibhāṣās; what 17.55 does is assign that dialect by occupation.

A third caution, and it governs how the evidence in this chapter is counted. Monier-Williams and the Nāṭyaśāstra are not two witnesses. A nineteenth-century Sanskrit lexicon glossing śābara as a grade of Prākṛt is a digest of exactly the dramaturgical and Prakrit-grammarian tradition that Nāṭyaśāstra 17 heads. The dictionary entry is downstream of the manual. They are one leg seen twice.

What Dvivedī adds is worth more than a second copy of the same thing, and it is the point that carries forward. He reads Bharata’s language chapter, and Daṇḍin’s Kāvyādarśa 1.36 beside it, as describing literary languages and not dialects in the field. Bharata is allotting registers on a stage. Daṇḍin is naming what a poem in each is called. Neither is doing ethnography, and both are working with a fixed inventory of named speech-grades that a writer selects from.

A playwright reaching for it did not need to know a Śabara. He needed to know what a Śabara sounds like on a stage. The convention was portable in a way an ethnography never is. It travelled to anyone who had read the manual, in any city, in any century.

The word had become a stage direction.

The bridge is speech

Here is the turn this chapter exists for.

Romila Thapar, in “The Image of the Barbarian in Early India”, published in Comparative Studies in Society and History in October 1971, argued that mleccha operates primarily as a linguistic category. Her formulation, in the article’s own summary of itself, is that the earliest distinction made by the Aryan speakers was a linguistic distinction.

The mleccha is the one whose speech is unintelligible; the root behind the word, √mlech, means to speak indistinctly. On Thapar’s account the boundary the classical texts are policing is drawn in the first instance around language, and only afterward around blood, diet, ritual and territory.

The comparison with the Greek barbaros, whose formation works the same way, is a commonplace of the handbooks and it is not Thapar’s sentence. It is offered here as an analogy, unattributed because this project has not found the place where it is argued rather than assumed.

Follow that through, and the adjective śābara means, quite literally, in barbarian tongue. In the rough speech. In the dialect that is not Sanskrit and does not care to be.

Which is precisely what a śābara mantra is.

A mantra in a register, and not a mantra belonging to a tribe.

The whole class is named for how it sounds. It is vernacular where the Vedic mantra is Sanskrit. It is unmetrical where the Vedic mantra is bound to metre. It is plain-spoken, sometimes crude, and it works by direct address rather than by ritual grammar. Somebody with a classical ear heard that and reached for the word already sitting in the dictionary for speech of that grade.

Take the qualification before the sentence rather than after it, because the sentence is the one a reader will carry away. This is the most consequential claim in the treatise, and it is an argument rather than a citation. Its named legs are two: the lexical-and-dramaturgical tradition, which gives śābara as a grade of speech and stands behind both Monier-Williams and Nāṭyaśāstra 17.49 and 17.55; and Thapar 1971, held at abstract level, for mleccha as a category of speech. The inference that joins them to the mantra corpus is a third step.

One source consulted takes that step, and it is Indian, and traditional, and it sits on the page this whole treatise is built around.

The commentator who says it outright

Mānas-pīyūṣa, the standard Hindī compendium of Rāmcaritmānas commentary, edited by Añjanīnandan Śaraṇ and published by the Gita Press in 1968, glosses the śābara verse across two pages. In the note that closes that discussion it records the tradition that Śiva and Pārvatī appeared in Bhīl form in the Kali age to remove the sufferings of creatures; that Śiva composed the mass of śābara mantras in the Bhīl language; and that Gaṇeśa wrote them down at Pārvatī’s command. Then it gives the etymology in four words. सँवर, it says, is the word for Bhīl, and the name arose because the thing appeared in that speech and that form.

That is a derivation of the category’s name from the speech of a named forest people, stated flatly, inside the tradition, in Hindī. The step is not this treatise’s alone.

Two qualifications, both real.

It is a traditional etymology and not a historical one. Commentators derive words to explain them, and the derivation carries the authority of the tradition’s own understanding of the name rather than the authority of comparative philology. And the people it names are Bhīls, not Śabaras, which is its own small problem and its own small corroboration: Russell and Hira Lal, writing in 1916 out of the Central Provinces, record that in the classical stories of the Savars’ origin the first ancestor is sometimes described as a Bhīl. Two independent traditions, one Hindī and commentarial, one colonial and ethnographic, putting the same two names in the same slot.

And a rival that would sink all of it

Goudriaan, in the one survey that catalogues these texts at all, suggests in the same paragraph that describes them that the name may connect instead to the Saṃvara or Śambara Tantras of Buddhist Tantrism. That is a derivation from a Buddhist text-group rather than from a grade of speech, and if it holds, the argument of the last three sections is not the explanation of the name. It would make the whole lexical apparatus of this chapter a description of a homonym.

Chapter three sets the two proposals out at length, gives the evidence for each, and says which this treatise follows and what would overturn it. A reader should hold both from here on, and should notice, each time a later chapter runs the speech reading, that the rival has not gone away.

Offered as this treatise’s reading of the evidence, then, and not as a settled finding:

The bridge is speech. It was never descent.

Then, sorcery

Somewhere between the dictionary and the street, the word acquires a third life.

In Marathi, śābara means magic, enchantment, witchcraft, sorcery. The compiler can be named: Molesworth, at the headword शाबर, gives exactly those four English words. Whether they are the primary sense of the entry is a claim about the structure of a page nobody here has seen, and it is not made here.

That would still be a large semantic distance for one word to travel. From a hunting people, to a grade of Prākṛt, to sorcery itself.

Half of that entry is closed and half is not. The compiler is Molesworth, the headword is शाबर, and the four English words are his. What is missing is a printed page: an edition, a year, a column.

There is a fourth sense, and it should be stated because it is a warning about method. In Āyurveda, the same sweep records śabara as a name for Lodhra, a medicinal tree. Nothing in the Lodhra entry has anything to do with tribes, dialects or spells. The word is simply polysemous. Anyone tracing a semantic chain across four fields is obliged to notice how easily an unrelated homonym can be pressed into a story it does not belong to. We are not building on the tree.

One further datum sits in the same sweep, and it needs handling with care. In Śākta material, Śavara is named as the god of the Jālandhara pīṭha, one of the seats of the goddess. If that held, and if the list carrying it could be dated, the tribal name would already be inside Tantric sacred geography long before any vernacular Nāth or Siddha corpus.

Neither condition is met, so no conclusion about the age of the association follows. No text is named for the attribution, pīṭha lists are unstable across recensions, and the presiding names vary. Nothing in this chapter rests on it.

What colonial ethnography actually did

In 1871, Alexander Cunningham published The Ancient Geography of India.

Raphaël Rousseleau, Professor of Anthropology at the University of Lausanne, writing in Politika on 1 June 2022 on the historiography of these peoples, states a mechanism in one clean line. His sentence, paraphrased, is that Cunningham turned a category into an ethnic name, and that this is a process which has gone on being repeated down to the present day. Rousseleau published in 2022 and is in copyright under this project’s pre-1929 rule, so his sentence is reported here in paraphrase and carries no quotation marks. The wording above is this treatise’s; the claim is his.

The exact words matter here more than usual. Rousseleau does not call the category a linguistic one, and the difference is the whole of his usefulness. He glosses śabara as a generic Sanskrit term meaning forest hunter, which is a descriptive category rather than a category of register. He is not a witness for the speech reading. He is a witness that a fluid classical term was pinned to a bounded people, and that is the only claim this chapter takes from him.

And the modern habit of reading the name as an ethnic label does not begin with Cunningham in 1871. This chapter’s own evidence stands against that. The Aitareya Brāhmaṇa names Śabaras among peoples. The Mahābhārata gives them an origin among peoples. Hemacandra, if the classification is his, sorts them among peoples. Reading the name as the name of a population is two millennia older than any surveyor.

Rousseleau’s claim is narrower: what changed in colonial ethnography was that a term the classical authors used loosely was fixed to a bounded, countable, locatable modern population. The census follows the surveyor. The ethnography follows the census. That is a real operation and it has real consequences, and it is not the invention of the ethnic reading.

And nothing in Rousseleau reaches the mantra literature at all. His article is about Jagannāth and the historiography of “tribal” origins in Odisha. It says nothing about śābara mantra, and it is not asked to.

The people who are still there

There is a last discomfort, and it should not be smoothed over.

This chapter opened on the people in Rayagada and Gajapati, and it has spent everything since on a word that outsiders put on them. They are not a lexical problem. They are a population with an address, and the two have to be set beside each other.

They are seldom mentioned in the accounts of śābara mantra read for this treatise. What cannot be printed is the negative in its widest form, that no scholar anywhere links this corpus to this people by descent. That would be a universal claim across a literature nobody has surveyed, and it is false on the evidence already in this project’s files.

Russell and Hira Lal, in volume four of The Tribes and Castes of the Central Provinces of India, published by Macmillan in 1916, write of the Savars:

“The Savars are considered to be great sorcerers. … Their charms, called Sabari mantras, are especially intended to appease the spirits of persons who have died a violent death.”

Named scholarship. In print. Attaching the mantras to the people, by name, as their charms, with a stated function. It is a hundred and ten years old and it has been sitting inside the one footnote this treatise quotes more often than any other sentence in the literature: Verrier Elwin, in The Religion of an Indian Tribe of 1955, cites exactly that passage of Russell and Hira Lal, reproduces the two proverbs about Savar sorcery that follow it, and then pushes the whole reputation away from his own Hill Saoras. Elwin’s own next sentence, the one this treatise has used for years as its definition-by-function, says that sabara-mantras have long been famous and that their addressees are the divinised dead of the violently killed.

Read Russell’s sentence and Elwin’s side by side. They are the same claim. Elwin cites Russell on that page, and what he adds is the distinction: those are not our Hill Saoras. That leaves an unattractive possibility open, and this chapter names it rather than resolving it: Elwin’s famous functional definition of sabara-mantras may be Russell’s, restated.

So the honest position. There is at least one body of named scholarship, in print since 1916, that links the mantras to the people by name and treats them as that people’s charms. Whether that linkage is a historical finding or a repetition of a regional reputation is a different question, and Elwin, who worked among the Saora for years, thought it was the second. Beyond Russell and Elwin, nobody in the literature this project has read has seriously tried to test the connection.

That last sentence is a claim about a literature, made from a reading list that is not complete, and it is scoped to that list. Chapter fourteen makes the proper pass through the modern Saora ethnography.

That is a gap. Naming it correctly is the least this treatise can do.

What the word carries

So the word arrives at the mantra corpus carrying several loads at once.

A hunting people at the forest edge, named among peoples in the Aitareya Brāhmaṇa and given an origin among peoples in the Mahābhārata. A slur for any savage. A category of people the Jain and Buddhist material reached here places outside dharma. A grade of rough speech, fixed in the lexical and dramaturgical tradition and named among the stage dialects. And, in the living vernacular, sorcery.

Two of those still rest on entries nobody in this project has opened in a printed edition, and the sentences carrying them say so. Set against them, the chapter has five claims of a different order, each with a locus: Mahābhārata 1.165.35 to 36 in the Critical Edition, read as a list of peoples and not as a scale; Nāṭyaśāstra 17.49 and 17.55, with Dvivedī’s 17.48 recorded beside them; Thapar in CSSH, 1971, at abstract level; Rousseleau in Politika, 1 June 2022, paraphrased and scoped to what he claims; and Russell and Hira Lal, 1916, quoted exactly and cutting against this chapter. Those five carry the chapter. Everything else is texture, and a reader is entitled to know which is which.

The tradition that used the word for these mantras appears to have been doing something precise with it: marking a register, unpolished, direct, outside the Sanskritic order, and effective anyway. On that reading the name was a description of how the words sound in the mouth. A Hindī commentator on Tulsīdās says the same thing in his own vocabulary, and a rival etymology in the standard catalogue says the name came from somewhere else entirely.

The next chapter goes to the mantras themselves, and to the two questions the name has been standing in front of all along: what, structurally, makes a mantra śābara, and where the word in that phrase actually came from. Chapter three owns the etymology and settles which road this book takes.

The word was given to them. They did not choose it.

We can at least read it with the evidence on both sides of the page.