Two Śābaratantras

Two objects wear the name Śābaratantra. Nobody has read either of them.

The first is a title on a list. Some time around the turn of the ninth century, in eastern India, a man named Vairocana compiled a manual on how to install an image in a temple. The work is the Pratiṣṭhālakṣaṇasārasamuccaya, which unpacks to something like “a compendium of the essentials of the marks of installation.” Pratiṣṭhā is the rite of establishing a sacred image, the moment a carved stone becomes a presence that can be worshipped. A manual of that kind is a working document. It tells a priest what to do with a body of scripture, which means that at some point it has to say what that body of scripture consists of.

So Vairocana gives a list, laid out by stream, one column to each channel of the Śaiva revelation. Vairocana, on the identification Mark Dyczkowski carries, was a son of the Pāla king Dharmapāla, whose reign is placed at roughly 794 to 814 of the Common Era. A real man, in a real dynasty, at a datable court.

In the Gāruḍa column, the column of the scriptures that deal with snakebite and poison, items seven and eight read Śābaram and Kālaśābaram.

That is all anybody has. Two titles, in one column, in one list. Whether either text survives, in manuscript or in citation or in any form at all, is not established in this chapter, and nothing in this chapter is able to say what was in them.

The second object you can buy this afternoon. Ask at the shop nearest you in a North Indian bazaar for a Śābaratantra and one will be produced: a cheap ritual pamphlet, incantations set out in lists, the language moving between Sanskrit and Hindī and never settling in either. Presses at Calcutta, Kanpur, Moradabad, Haridwar, Kalyan and Varanasi have been turning out versions of it since 1884 and have never stopped. That one nobody has read either, in the sense that matters here. Behind the pamphlet stand a hundred-odd manuscripts scattered across about forty libraries from Kolkata to Kathmandu, and no scholar this research has reached has collated them, sorted them, or asked how many separate books are hiding under the titles.

So: a name in a Pāla canonical enumeration around 800 of the Common Era, and a bazaar paperback whose earliest dated manuscript was copied in 1855 or 1856 and whose earliest known imprint is Calcutta, 1884. Those are the firm dates in this chapter, and they are a little over a thousand years apart.

Between them, nothing. No chain of quotation. No intermediate copy. Not one writer in the intervening centuries pausing to say which Śābara he meant.

Chapter three met this collision in passing and stepped around it. This chapter stops and weighs it, because the collision is worse than it first looks, and because the silence in the middle is the most important thing in the file.

A list is a claim, not a shelf

A canonical list is a claim, made from inside a tradition, about what the tradition contains.

It is not a library shelf. It is not a catalogue of manuscripts anybody had in front of them. Dyczkowski’s whole book is a study of exactly this problem: the Śaiva scriptural canon exists in several competing enumerations, the lists disagree with each other, they name texts that have never been found, and they omit texts that survive in quantity. A title’s presence in a list of the ninth century proves that somebody in the ninth century thought a scripture of that name belonged in that place. It proves nothing else on its own.

So the honest form of the finding is narrower than “there was a ninth-century Śābara Tantra.” The honest form is this: by around 800 of the Common Era, a compiler working in a Pāla milieu believed that the poison-cure division of Śaiva scripture contained a text called Śābaram, and another called Kālaśābaram.

That second title deserves a moment. Kālaśābaram. Now set it beside a title from Goudriaan’s late group: the Kālīśābaratantra. In roman letters they are one vowel apart. In Sanskrit they are not close at all. Kāla is black, or time, or death. Kālī is the goddess. A reader moving quickly between a nineteenth-century print catalogue and a ninth-century canonical table will merge those two titles without noticing that anything has happened. No scholar reached by this research does that, and none warns against it either. The observation is this treatise’s, offered as a route by which confusion could arise, with no claim that any published writer has taken it.

The stream is a medical stream

The Gāruḍa Tantras take their name from Garuḍa, the great eagle, Viṣṇu’s mount and, in the older mythology, the standing enemy of serpents. The stream named after him is the stream of snakebite and poison.

That is worth saying plainly, because it changes what kind of book we are talking about. This is not a division of the canon concerned with liberation, or with the metaphysics of the syllable, or with the goddess. It is concerned with venom and its cure. It sits at the point where scripture and medicine meet, and the people who used it were dealing with an emergency in a village at night.

The standard work on this material is Michael Slouber, Early Tantric Medicine, Oxford University Press, 2017.

The distance between the two

Set the two objects side by side and the incompatibility is not subtle.

On one side, a title in a canonical enumeration compiled under a Pāla prince around 800 CE, sitting in the division of scripture that handles poison, inside a Sanskrit canon with its own transmission and its own commentarial world. On the other, a group of texts that Goudriaan and Gupta describe, in their survey of the whole Tantric and Śākta literature, as a cluster of their own, whose contents run, on their account, to little more than enumerations of incantations moving between Sanskrit and Hindī. That description is paraphrased here rather than quoted, because the 1981 volume is in copyright. The datable footprint of the second group is print, and it is the cheap ritual pamphlet of the North Indian bazaar.

The imprints exist, and they are named in print. Goudriaan and Gupta carry three of them: an edition of the Bṛhat-śābaratantra by Hariśaṃkar Śāstrī with a Hindī translation, Cawnpore 1900; a new edition at Moradabad in 1906; and an edition from the Lakṣmīveṅkaṭeśvara Press at Kalyan, Śaka 1841, which is 1918 or 1919 of the Common Era. All three sit in a footnote printed directly under the paragraph the survey gives this corpus.

Cawnpore, 1900, is the earliest of the three, and it is not the earliest imprint of this material. The union catalogue of Sanskrit works carries two that stand before it: a Śābaratantra from the Jyotiṣaprakāśa Press at Calcutta in 1884, and a printing in R. Chatterji’s periodical Aruṇodaya at Calcutta in 1890. Sixteen years, recovered by opening a catalogue rather than a survey. The rest of this chapter is written around what else was in it.

Set the whole run out in order.

Two of those rows carry a note, and both notes come from the same man. The Asiatic Society cataloguer, writing in 1939, records that Chatterji also published the Divyaśābaratantra under the name Śābara in eighteen pīṭhas, and again inside a compendium of indrajāla. And it is that cataloguer who reaches for the Kalyan edition as a comparandum while describing a manuscript.

Thirty-five years, five presses, four cities, and a periodical, from the first row of that table down to the Kalyan edition. Then a century of reprinting. Whatever else this corpus is, it is one of the more successful publishing properties of the North Indian bazaar press.

And it never stops. Behind the last row of the table stands whatever is on the shelf of the shop nearest the reader today.

Different century. Different division of the canon. Different purpose. And a completely different language situation, because the defining feature of Goudriaan’s corpus is that it does not stay in Sanskrit, and a ninth-century Gāruḍa scripture in a Pāla canonical list has no reason at all to be bilingual with Hindī, a language that did not yet exist in the form the later texts use.

Conflating them is an error of period and an error of genre at the same time. A reader who does it will place a bazaar pamphlet in the ninth century, or will imagine that a Pāla-era antivenom scripture was written half in Hindī. Both are worse than simply not knowing.

Counting is the trap, and the cataloguer says so

The instinct at this point is to count titles. Resist it, because the cataloguer whose page supplies them says explicitly that they cannot be counted.

Goudriaan, having named the Divya-, Kālī- and Siddha-śābaratantras, adds in the same sentence that the titles are confused and may occur after one another in the same manuscript. That caution governs everything this chapter does with the titles, and chapter three carries it too.

So: Goudriaan’s group carries a Bṛhat-, a Siddha-, a Divya- and a Kālī-śābaratantra, and the Pañcadrāviḍaśābaracintāmaṇi, and the man who lists them warns that the titles are confused and may run one after another inside a single manuscript. No inference about how many distinct works there are can be drawn from that list, and this chapter draws none. The ninth-century canonical table adds two more titles, from a different tradition and a different century.

The claim this chapter defends is the minimal one, and it does not need a count: the name covers at least two things that cannot be the same thing.

There is a third possibility standing behind both, and chapter three sets it out. Goudriaan opens the same paragraph by suggesting that there might be a connection between the term Sāvara and the generic name of the Saṃvara or Śambara Tantras known from Buddhist Tantrism. If that holds, then some part of what wears this name in Sanskrit sources is a Buddhist text-group under a Śaiva spelling, and the number of distinct things called Śābara goes up rather than down.

But titles are not the only countable thing.

What the corpus actually is

There is a question underneath all of this that is easier to ask sideways than to put directly. Not what does the name mean, not who claims it, not how old is it. Simply: what is there? How many texts, in how many copies, in whose custody.

The instrument for answering it has been in print for decades. The New Catalogus Catalogorum, compiled at the University of Madras, is the union catalogue of Sanskrit works and authors: it takes every manuscript catalogue published anywhere in the world and collates them under headwords. The volume carrying the Śābara block was published at Madras in 2014 under the editorship of Siniruddha Dash, and the block runs to four pages. Four pages. Everything recorded about this corpus, in every library on earth that has catalogued it, on four pages.

They carry about twenty-three distinct tantric Śābara headwords. Śābara, Śābarakalpa, Śābarakalpahanumantamantra, Śābaragāruḍa, Śābaracintāmaṇi, Śābaratantra, Śābaratantramantrasaṅgraha, Śābaraprayoga, Śābaramantra, Śābaramantraviṃśati, Śābaramantrasaṅgraha, Śābararahasyamantra, Śābaravidyā, Śābarī, Śābarīmantra, Śābarīyantra, Śābarottaratantra, Śābaropaniṣad, and the Divya-, Kālī-, Siddha- and Bṛhac- compounds entered in other volumes. Two more headwords in the same run, Śābarabhāṣya and Śābarakaustubha, belong to Śabarasvāmin the Mīmāṃsaka and have nothing to do with this corpus; the Buddhist Śābaripāda and Śābarīśvara are a separate matter again, and chapter sixteen has them.

Behind those twenty-three headwords stand perhaps five or six actual works, and one of them carries most of the mass.

Everything else is a scatter of short incantation booklets of two to twelve leaves, most of them flagged incomplete.

And there is proof inside the catalogue that the scatter is partly an illusion. One Asiatic Society entry describes a Śābaramantra of eight leaves, incomplete, and the cataloguer notes that everything after its sixth leaf agrees with the beginning of the Divyaśābaratantra. A booklet with its own title and its own place on a shelf, which is a piece of a larger text under a different name. That is one cataloguer catching one instance. Nobody has done the exercise systematically, and until somebody does, every count of Śābara “works” is an upper bound.

Roughly 110 to 120 manuscript items sit in about forty repositories, running from the Asiatic Society at Kolkata to Sampurnanand Sanskrit University at Varanasi, which after Kolkata holds more of this material than anyone, on through the Government Sanskrit Library and the Ganganatha Jha Institute at Allahabad, the Government Oriental Manuscripts Library at Chennai, Tanjore, Adyar, Wai, Ujjain, Jodhpur, Baroda, Lonavla, Mysore, Nagpur, Darbhanga, Cuttack, the Orissa State Museum at Bhubaneswar, the French Institute at Pondicherry, Panjab, Lucknow, Jaipur and Bombay University, and out to Nepal.

Nepal, which nobody had looked at

That last one closes a hole this treatise had left open, and had no way of closing while it was reading survey volumes instead of catalogues.

The union catalogue records three Nepalese witnesses to a Śābaratantra. One stands in Haraprasāda Śāstrī’s catalogue of the Durbar Library at Kathmandu. Two more stand in the Bṛhatsūcīpatram, the great inventory of the Nepalese national collection: one of them incomplete, and one of them attributed to Śrīkaṇṭha.

That name should stop a reader of this treatise. Śrīkaṇṭha is the twelfth and last name in the list of Siddhas that the Gorakṣasiddhāntasaṃgraha quotes from the Śābaratantra, the list chapter six works through. Aufrecht, in 1891, had already recorded a Śābaramahātantra by one Śrīkaṇṭhaśiva Paṇḍita. Whether the Kathmandu manuscript’s Śrīkaṇṭha is the Siddha of the list, the Paṇḍita of Aufrecht’s entry, or a third person entirely is not established, and nothing is built on it here.

The Nepalese-German Manuscript Cataloguing Project database at Hamburg, which is the modern instrument for Nepalese manuscripts, could not be reached. Its server presents a certificate that cannot be verified, and this project does not disable certificate verification to get at a source. So Nepal is covered here at second hand, through the union catalogue’s citation of two older Nepalese catalogues. What can be said is narrower than a database search and it is not nothing. Nepal has at least three Śābaratantra manuscripts on record, one of them with an author’s name attached, and no statement that the Nepalese collections are unexamined for this corpus can stand.

The verdict the manuscripts deliver

Now put the descriptions together, because they say something the titles do not.

Not one Śābara manuscript described in the Asiatic Society’s Tantra catalogue is on palm-leaf. Every one is paper: “country-made” or “foolscap.” In an Indian tantric collection that is the single loudest date signal there is. Palm-leaf is what pre-modern transmission looks like in eastern and southern India, and there is none of it here.

The script is Nāgara throughout, and the cataloguer repeatedly writes “modern Nāgara.” The extents are small, from a couple of leaves to forty-one at the top of the range. The great majority of witnesses across the whole union catalogue are flagged incomplete. And there is exactly one explicit date in the entire block: Saṃvat 1912, which is 1855 or 1856 of the Common Era. One more manuscript ends a section with a date the catalogue prints and nobody has read.

One date. In a corpus of a hundred-odd manuscripts. And it falls twenty-nine years before the first printing.

Then the contents, which have been described here already and now do a second job. Hindī incantations. A Hindī section after the third chapter. Hindī and corrupt Sanskrit. Explanatory notes in an unnamed vernacular under the yantras. Telugu prose. Kannada interspersed. Vernacular meaning. Those are cataloguers’ field notes from four cities, written to record contents and not to make an argument, and together they confirm from the manuscripts what Goudriaan defines from the armchair. The bilingualism is real, it is everywhere, and it was documented before anybody wrote a definition of the genre.

So the honest description of this corpus, which this treatise can now give and could not give before:

A genuine manuscript tradition, and a shallow one. A few dozen late paper copies of five or six short bilingual works, none of them demonstrably older than the nineteenth century, concentrated in Varanasi, Calcutta, Allahabad and the Madras libraries, with outliers as far as Kathmandu and Pondicherry, which the Calcutta and Bombay presses then industrialised between 1884 and 1919 and which has been reprinted commercially ever since.

Read what that does to the argument rather than to the mood. It confirms Goudriaan’s judgement that not many texts belong to this group. It removes any claim about the corpus’s antiquity that rests on witnesses, because the witnesses are young. And it leaves the corpus’s authority standing on three things that are not antiquity: an attribution to Ādinātha and the Nāth line, made inside the texts; one quoted list of twelve Siddhas that turns up in three independent places; and the fact that the texts are openly, structurally vernacular, which is a claim about how mantra works rather than a claim about how old the book is.

A treatise that says these are ancient texts is making a claim the manuscripts will not support.

The catalogue contradicts itself, and we are not going to fix it

One thing found in the union catalogue cannot be resolved and will be reported rather than tidied.

The Kālīśābara contains an internal list of Śābaratantra titles: a canon that the corpus draws up for itself. The union catalogue mines that list, and cites the same Asiatic Society manuscript for it every time.

In the volume that carries the Śābara block, the New Catalogus Catalogorum describes the Śābaratantra as one of the twelve Śābaratantras named in the Kālīśābara.

In an earlier volume of the same catalogue, in three separate entries, it describes titles as one of the nineteen Śābaratantras named in the Kālīśābara.

Twelve and nineteen. Same catalogue, same source manuscript, two numbers, and no note anywhere acknowledging the difference. A third volume adds that the Divyaśābaratantra is the sixth in the series, which is consistent with either.

This chapter reports the contradiction and does not resolve it. Resolving it would require reading the manuscript, and the manuscript is five leaves long: entry 6095, folios 36 to 41 of the codex accessioned as 8355, complete in three chapters, with its opening printed in the catalogue at pages 279 to 281. It is the cheapest piece of verification anywhere in this project. A reader with a morning at 1 Park Street could settle a discrepancy in the standard union catalogue of Sanskrit literature, which is the sort of sentence that tells you how little attention this corpus has had.

An exact parallel, from an Indian scholar, on a person

The two-texts problem has a twin, and it was documented in Hindi thirty-one years before Goudriaan’s catalogue.

Hazārīprasād Dvivedī, in Nāthasampradāya, published by the Hindustani Academy at Allahabad with a publisher’s note dated 31 March 1950, works through the Siddhas contemporary with Gorakhnāth. Under the entry सबर, Sabara, he writes that there have been two Siddhas of this name. One is the guru of Lūhipā, the Kāyastha of King Dharmapāla, whose reign Dvivedī gives as 769 to 809. The other is a Siddha of the tenth century. And then the sentence this chapter needs: the two have been mixed into one another.

Read what has just happened. The same name, the same two-centuries-apart problem, the same undetected merger, on a person instead of a book, in the same Pāla reign that dates the Pratiṣṭhālakṣaṇasārasamuccaya. An Indian scholar had already found and named the pattern, in the Nāth material, in Hindi, in 1950, and nobody carried it across to the titles.

Dvivedī adds, on the authority of Dr Benoytosh Bhattacharyya, that a Śabarapāda is the first propagator of the Kurukullā cult in the middle of the seventh century and a disciple of Nāgārjuna, and also the original propagator of the Chinnamastā cult. Chapter sixteen owns that figure. What it establishes here is that the name Śabara, attached to knowledge and to power, is already doing multiple unrelated jobs across traditions and centuries before any of the texts in this chapter were printed.

The spelling will not help you

The natural instinct, once you suspect two texts hiding under one name, is to look at the orthography. Surely the letters would separate them.

They do not, and the reason is the opposite of the obvious one. The letters are not the least stable thing in this dossier. They are among the most regular things in it.

On the pages that have been read, the forms behave. शबर to शाबर is vṛddhi, the standard Sanskrit lengthening that turns a noun into an adjective of belonging, and chapter seven sets the derivation out. Prakrit turns intervocalic b into v; Avadhi turns palatal ś into dental s. Those are sound changes every Sanskrit word undergoes in those languages. Measured that way, śābara is exactly as stable as any other word crossing the same boundaries, and no argument from instability can be built on it.

What is left is a different and more useful problem. Each occurrence takes the regular form of the language it happens to be sitting in, and editors, poets and cataloguers reach for different regular forms without comment.

Upadhye’s Kuvalayamālā prints Prakrit सवर for the people and सावर for the knowledge, and his apparatus records copyists writing the short form in both places. The mūla of the Rāmcaritmānas at Bālkāṇḍ 1.15.3 reads Avadhi साबर. The Gorakṣasiddhāntasaṃgraha prints its Sanskrit citation formula with शाबरतन्त्रे and the Hindī gloss facing it on the same opening reads साबरतन्त्र. Goudriaan’s survey enters the group as Śābara and offers Sāvara in a parenthesis in the same sentence. P. C. Bagchi, editing Matsyendranātha’s Kaulajñānanirṇaya for the Calcutta Sanskrit Series in 1934, refers to the twenty-four-Kāpālika passage under the title Sāvara Tantra.

Five witnesses, five defensible spellings, and no misspelling among them.

That is why there is no orthographic test. Nothing in the spelling of Śābaram in the Gāruḍa column of the ninth-century list marks it off from the Śābara Tantras of the bazaar press, and nothing in the spelling links them either. The evidence that would normally do this work is not corrupt. It is uninformative, which is worse, because corrupt evidence announces itself.

A name in the poison ward

Now the part that is a suggestion and will be marked as one.

The addendum to this treatise’s research file establishes one negative here: no documented harm was found attaching to Śābara mantras by name in any source this project reached. That is a statement about the sources reached and it is not a statement about the world. This treatise has no evidence about the purposes for which Śābara material is used on the ground, and any sentence claiming that its use is predominantly protective or curative would be invented. What would close it: an ethnography of practitioners who name the corpus.

What the record does carry is adjacent and thinner. Jhāḍ-phūṅk, sweeping and blowing, the healer’s physical gesture over a patient, is named in the Chhattisgarh statute of 2005. Read the section rather than the word. At section 2, sub-section 3, the Act defines an ojha as a person who claims to possess the power to control, cure or treat tonahi, or any person or animal alleged to be affected by tonahi, and to render that power harmless, by jharphook, totka, tantra-mantra or by any means. So jharphook is named there among the means by which an ojha claims to cure, inside a definition whose purpose is to make the activity punishable. That is a legislature describing a claimed healing practice in order to prohibit it. It is a fact about what a statute thought it was regulating. It is not a fact about this corpus.

It would be easy to read the Act as judging that jharphook functions as a treatment, and easy in the direction that flatters this treatise, toward the protective and curative register and away from the coercive one. The Act’s judgement runs the other way. It names the practice in order to prohibit it.

Chapter ten counts the corpus’s own vocabulary from the inside and finds its characteristic move to be duhāī, the appeal in somebody’s name, which the standard Hindi lexicographers class as a swearing-by. The coercive register is present in the corpus, and no reader should carry away from this chapter that it is absent.

And here is a text called Śābara sitting, around 800 CE, in the division of the Śaiva canon devoted to healing poison.

That is an alignment worth noticing, and it is offered as a suggestion and nothing else. A modern corpus whose adjacent practices a legislature described in the language of curing, even while moving to prohibit them, shares a name with a scripture filed under cure. It would be very satisfying to draw a line between those two points.

The line cannot be drawn. Not on anything in front of us. There is no chain of intermediate attestations running from a ninth-century Gāruḍa title to a nineteenth-century Hindī pamphlet. We do not know the contents of the earlier text. We do not know whether its name refers to the Śabara people, to the barbarous-speech adjective, to the minor deity Goudriaan glosses, or to something else again. Continuity across a thousand years cannot be inferred from a shared word, and chapter four’s eighth-century episode is bound by the same rule: it testifies about 779 and about nothing later.

So it stays a suggestion, marked as one, and it is explicitly not demonstrable across a millennium of silence. What would close it: Slouber, first; then any citation of the Gāruḍa Śābaram in the intervening centuries. Absent those, this is an observation about a coincidence of names, and the reader is entitled to weigh it at exactly that.

Nobody has had occasion to notice

There is a reason nobody in the sources reached has sorted this out, and the reason is itself a finding.

Goudriaan and Gupta give the late corpus one paragraph in a survey volume covering the whole of the Tantric and Śākta literature. Dyczkowski gives the Gāruḍa titles two cells in a table. David Lorenzen, whose The Kāpālikas and Kālāmukhas is the standard monograph on the sect that claims this material, indexes the Śabara-tantra and discusses the passage that quotes it, and chapter six sets out what he actually concludes about it. Verrier Elwin mentions sabara-mantras once, in a footnote, and does two separate things in it. He pushes them away from the Hill Saora he studied. He also says what they are: formulae whose only addressees are the divinised dead of the violently killed.

That second half is the one worth weighing. A definition by addressee is a definition by what the incantations are for, and it sits against chapter three’s formulation that the one scholarly definition of this corpus fixes it by register. Two chapters of this treatise pull against each other on the point, and the pull is recorded here rather than smoothed over. Chapter fourteen owns the Elwin footnote and works the conflict out there.

So the field’s attention to these titles is small, and it is not nothing. The confusion is not the result of anyone’s carelessness. It is the result of nobody putting the objects on the same page. Two texts a thousand years apart can wear the same name indefinitely if no monograph ever has cause to compare them. Dvivedī, in 1950, put two people of the same name on the same page and saw the merger at once. Nobody has done it for the books.

And the cataloguers, who are the one profession that did look, were not in a position to notice either. A cataloguer’s job is the object in front of him. Chakravarti at Calcutta in 1939 described ten Śābara manuscripts with exemplary care, noticed that one booklet was a piece of another text, noticed which chapter colophons disagreed with which, and had no occasion to ask what a ninth-century list in a Pāla installation manual had to do with any of it. The Madras compilers collated his entries with everyone else’s and produced four pages that contain, without comment, both the census and the contradiction inside it. The information has been sitting in public print since 1939 and 2014 respectively. What has been missing is somebody reading the two catalogues against each other and against a survey volume, which is an afternoon’s work.

What a search returns

Type “Śābaratantra” into a search field today and what comes back is a blur.

Astrological services. Pamphlet reprints. Video uploads promising results. A handful of encyclopedic paragraphs that fold every occurrence of the word into a single undifferentiated tradition attributed to Śiva in the Kali age. Somewhere in that blur is a ninth-century poison-cure scripture from a Pāla canonical list, and somewhere else in it is a genre of bilingual incantation pamphlets from the North Indian bazaar presses, and nothing in the results tells the reader that these are different objects.

Distinguishing them settles no mystery and produces no new doctrine. It removes a mistake before the mistake is made. That is unglamorous work, and it is exactly the work a treatise exists to do, because nobody else is going to do it for a title this obscure.

What the catalogues add to that is an answer rather than a warning. The question underneath the blur was never which of the results is the real Śābaratantra. It was what is there at all, and there is now a number for it: twenty-three names, five or six books, a hundred-odd paper copies, forty libraries, one date, and a printing history that starts in a Calcutta press yard in 1884 and has outlived every one of its editors.

At least two things wear the name. We have now measured one of them, glanced at the other, and can prove no relation between them.

The next chapter turns to the lineage that claims the later corpus as its own, and asks how late that claim actually is.