The Svacchanda Tantra: the fight over what was in the cup

Some time around the year 1000, in Kashmir, a scholar sat down to explain a single word in a ritual instruction. The instruction told the worshipper how to make the guest-offering to a god, and it said the offering should be poured with fragrant liquor.

People, he said, were reading it as water.

The scholar’s name was Kshemaraja. He was the pupil of Abhinavagupta. He gave a reason for the word, and then went a long way past it. He wrote that anyone who reads surā, liquor, as water in a Bhairava scripture, in a rite whose whole purpose is to lift a person out of caste, is jātigrahagrasta. Seized by the demon of caste.

That commentary survives. So far as anyone knows it is the only Sanskrit commentary on this scripture that survives entire, so the argument reaches us almost entirely in the winner’s version.

A cup, and what was in it

The instruction sits in the second chapter, among the ordinary business of daily worship. It is a line of procedure, and it reads like one.

पश्चादर्घः प्रदातव्यः सुरया सुसुगन्धया । मुद्रां प्रदर्शयेत्पश्चात्त्रिधा त्रैकाल्यकर्मणि ॥१३६॥

paścād arghaḥ pradātavyaḥ surayā susugandhayā | mudrāṃ pradarśayet paścāt tridhā traikālyakarmaṇi

“Afterwards the guest-offering is to be given, with sweetly fragrant liquor; then one should show the hand-gesture three times, in the rite of the three daily junctures.”

Svacchandatantra 2.136 #

Svacchanda Tantra 2.136, printed at page 75 of the first volume of the Kashmir Series of Texts and Studies. The word is surā, and in its ordinary sense it means liquor.

Kshemaraja’s gloss runs directly beneath it on the same page. Here it is whole, cross-reference and all.

सुराया आनन्दहेतुत्वादेवमुक्तम् । ये तु जात्युद्धारपरभैरवरूपत्वोन्मीलकेऽप्यस्मिन् भैरवनये सुराशब्दं जलवाचिनमपि व्याचक्षते ते जातिग्रहग्रस्ताः

‘मद्यं मांसं तथा मत्स्यानन्यानि च वरानने । साचारांश्च निराचारांल्लिङ्गिनो न जुगुप्सयेत् ॥’ (५।४५)

इति भाविसमयोल्लङ्घिनः पशव एव ॥

surāyā ānandahetutvād evam uktam | ye tu jātyuddhāraparabhairavarūpatvonmīlake ‘py asmin bhairavanaye surāśabdaṃ jalavācinam api vyācakṣate te jātigrahagrastāḥ | “madyaṃ māṃsaṃ tathā matsyān anyāni ca varānane | sācārāṃśca nirācārāṃl liṅgino na jugupsayet ||” iti bhāvisamayollaṅghinaḥ paśava eva ||

“It is put that way because liquor is a cause of bliss. But as for those who, even in this Bhairava system, which reveals a man’s identity with the supreme Bhairava by the extraction of caste, explain the word surā as denoting water as well: they are seized by the demon of caste. ‘Wine, meat, and likewise fish, and other things too, O fair-faced one; and one should not feel disgust at those who wear the emblem, whether they keep the rules of conduct or keep none’ (5.45). Transgressors of that rule, which is taught further on, are simply beasts.”

Svacchandoddyota, on 2.136 #

The rule he throws at them is the scripture’s own, quoted inside the gloss. Anyone checking it should know that the Kashmir Series prints his citation as 5.45 while the same edition numbers the verse 5.48 when it comes to print it, at page 41 of the third volume; the Pondicherry electronic text agrees with 5.48.

And he names nobody. He writes ye tu … vyācakṣate, “but as for those who explain”. Who he means is contested, and the two scholars who have looked hardest at this gloss read it differently. Sanderson describes it as an attack on “the then current practice of substituting water for alcoholic liquor”, which makes it a quarrel with living people about what was going into the cup on the day he wrote. William Arraj reads the target as the earlier commentators, who “had evidently rejected the prescription by Svacchandatantram that Bhairava should be worshipped with an offering of alcohol”. The two readings are compatible if the practice in force simply was the dualist one.

There is a sentence people reach for to close that gap, and it needs reading carefully. In 2007 Sanderson writes that in both commentaries Kshemaraja stresses that his motive is to re-establish what he sees as the true, non-dualistic perspective of these texts, “which, he says, had been overshadowed by the prevailing view, that of the Saiddhāntika dualists, an unjustifiable inroad into non-Saiddhāntika territory that had both imposed a less enlightened reading of their teachings and led to the suppression of elements of non-dualistic practice (advaitācāraḥ), such as the offering of fermented liquor to Svacchandabhairava, that were vital to the proper performance of their rites”. The suppression, and the liquor, arrive inside a relative clause hung on which, he says, and the he is Kshemaraja. How much of the characterisation Sanderson is adopting and how much he is reporting can be argued about. What can be said is that he marks his distance thirty pages later, writing of what Kshemaraja saw as an unwarranted diluting of non-dualist practice. Taken at its most cautious, the sentence is one man’s account of a suppression, and he is the man who needed it to have happened.

Sanderson, in the 1988 survey of the Shaiva traditions that most later work starts from, gave Kshemaraja the point on this and took it away from him on everything else. His reason for giving it is a different reason from the one Kshemaraja gives. Writing on page 700, he says that in purely doctrinal terms Kshemaraja’s commentaries do violence to both of these texts, meaning the Svacchanda and the Netra, at least as much as the dualist commentaries must have been guilty of, and that neither Tantra fits either exegetical straitjacket. Then this:

“In the area of ritual, however, Kṣemarāja had a clear advantage. When, for example, he attacked the then current practice of substituting water for alcoholic liquor in Svacchandabhairava’s guest-offering, he was simply reinstating the text within the tradition of the Bhairava Tantras to which it properly belonged.”

The scripture does have a way of not saying surā out loud. Its fifteenth chapter is a register of chummaka, code-words by which initiates were to refer to ritual things in company, and at 15.3 it prescribes that wine is to be known as harṣaṇa, the gladdener, and surā is called muditā, the joyful one. Kshemaraja glosses it without embarrassment: muditā means joy, and surā is so called because it causes joy. But the Svacchanda uses surā in its plain sense elsewhere, including in the very offering under dispute. Concealing a word is a different operation from redefining it.

The scripture belonged to a family of texts whose ascetic background was the skull and the cremation ground, and somebody had been quietly turning its liquor into water.

One caste alone

To see why the drink mattered you have to go two chapters further, into the great initiation liturgy, which is the longest ritual sequence in the book. The rite of initiation includes a procedure known as jātyuddhāra, the extraction of caste. The initiate’s birth is taken out of him.

Then the text says what that means, and it says it without any softening at all.

ब्राह्मणाः क्षत्रिया वैश्याः शूद्राश्चान्येऽथवा प्रिये । सर्वे ते समधर्माणः शिवधर्मे नियोजिताः ॥५४०॥

brāhmaṇāḥ kṣatriyā vaiśyāḥ śūdrāś cānye ‘thavā priye | sarve te samadharmāṇaḥ śivadharme niyojitāḥ

“Brahmins, kshatriyas, vaishyas, shudras, or others besides, my dear: all of them have the same duties, being yoked to the observance of Shiva.”

Svacchandatantra 4.540 #

The next verse says they all wear matted locks and ash, and that they eat in a single row, ekapaṅktibhujaḥ, and it names the row in the same breath: it is the row of the samayins, the lowest grade of initiate. The verse after that gives three further grades a row each. The putrakas, initiated for liberation. The sādhakas, adepts who are after powers instead. And the cumbakas, who are the officiants, the men who perform the rite. That word is not the chummaka of the fifteenth chapter, which is a code-word; the two are easy to confuse and the scripture uses both. Then the half-line: na prāgjātivibhedataḥ, not by the divisions of former caste.

Read one way, rank has survived initiation and is simply drawn somewhere else, which in a society organised around who may not eat beside whom is not a decorative change. Sanderson reads the same lines the same way, translating the further grades as rows of their own.

Kshemaraja’s note sits right there, wedged between the two halves of that verse on page 329 of the second Kashmir volume. Here, he writes, unlike in another tantra, initiates have no former-caste rite whatever, and the only distinction among them is by grade of initiation.

Then the text states the principle flat.

एकैव सा स्मृता जातिर्भैरवीया शिवाव्यया । तन्त्रमेतत्समाश्रित्य प्राग्जातिं न ह्युदीरयेत् ॥५४३॥

ekaiva sā smṛtā jātir bhairavīyā śivāvyayā | tantram etat samāśritya prāgjātiṃ na hy udīrayet

“That caste is taught to be one alone: Bhairava’s, auspicious, imperishable. Having taken one’s stand on this tantra, one must not utter anyone’s former caste.”

Svacchandatantra 4.543 #

And then it puts teeth in it. A man who names an initiate’s former birth becomes liable to expiation, and 4.545 tells him how long he will be cooked in hell for it: three of Rudra’s days, five of Keshava’s, a fortnight of Pitamaha’s.

One step in the same rite cuts against all of it. At 4.67 the caste is taken out, and in the very next words of the same verse it is put back: dvijatvāpādana, the conferral of twice-born status. The guru consecrates a sacred thread with the Bhairava mantra and gives it to him. The whole series of household sacraments is performed over him, from the rite of conception to the funeral rite. Twice, at 4.75 and again at 4.77, the text says he has been consecrated in the status of the twice-born. A share in Rudra follows. Nothing later in the chapter takes the thread back.

Arraj saw the difficulty and did not smooth it. The rite as this chapter has it, he writes, apparently represents a halfway stage back to the reacceptance of caste, and Kshemaraja has to work at pages 28 and 29 of the second Kashmir volume to show that it does not conflict with the rest. Elsewhere in the scripture, in the tenth chapter, where every grade of existence from a plant upward is purified along with all forty-eight brahmanical sacraments, Arraj reads the twice-born status as conferred only so that it can be negated. Two rites pointing two ways, inside one book. He adds one more thing, from Helene Brunner’s edition of the Somashambhupaddhati: the extraction of caste quietly disappears from the later handbooks of the Shaiva Siddhanta, the dualist school whose followers are called Saiddhantikas, altogether.

How far does the abolition reach? Further than the ritual, on the evidence of the verses themselves, which forbid the naming of a former caste at any time and put a hell behind the prohibition. It is tempting to set Sanderson and Arraj against each other on the point, and the temptation should be resisted, because the sentence usually quoted from Sanderson is a third of a sentence. Comparing the Bhairava and Shakta scriptures with the Saiddhantika ones, he writes that in them “we find accordingly a stronger rejection of caste in ceremonial contexts, a conviction that pride of caste is one of the factors that hold souls in bondage, and prohibitions against ever mentioning the birth-caste of a fellow initiate”. Only the first of those three is ceremonial. The second is a doctrine about what keeps a soul bound, and the third is a rule about how initiates are to speak of each other for the rest of their lives.

Arraj says something stronger, and says it against a different target. The Saiddhantika commentators, he writes, insisted on maintaining caste distinctions outside the ritual arena, while this text “unequivocally portrays the Saiva initiate, even outside of a ritual setting, as truly liberated and beyond distinctions of caste and impurity”. His is a claim about what the text means; Sanderson’s is a description of what a body of scripture contains. Read whole, they are not far apart, and the verses do not force a choice between them.

Sanderson translates the passage in The Saiva Age at pages 293 and 294, and anyone who goes to check him against the printed Kashmir text should be ready for two small divergences. His footnote gives the passage as 4.539c to 545 and then prints a second “545” over the verse the Kashmir edition numbers 4.546. And in that verse he prints siddhim, the highest attainment, where the Kashmir edition reads gatim, the highest goal. The second half of that same verse puts siddhi and mukti side by side, so the text is not fastidious about keeping them apart, and the difference here is small. The Kashmir edition’s footnotes, which record where its manuscripts disagree, carry no note at all on that page. The three manuscripts its preface describes are all Kashmiri Sharada, and the two further witnesses its footnotes cite are described nowhere at all, as a later section of this article shows. Sanderson’s own footnote names no witness either. That silence settles nothing about what he had in front of him. Neither point disturbs his argument. This is the ordinary friction of two scholars reading two families of manuscripts.

So: a scripture that removes birth at the door, and an argument about whether the cup in the same rite held wine or water. Kshemaraja joins those two himself, and not loosely. In the gloss on the cup he calls this the Bhairava system that lifts a man out of caste, and the word he uses for it is jātyuddhāra, the initiation chapter’s own word, sitting at one end of the sentence whose other end is the accusation. Arraj put his finger on it: Kshemaraja, he writes, argues that commentators who gloss the alcohol as water are denying the rite that extracted the initiate’s caste. Sanderson, who quotes the same polemic, explains it differently, by canon rather than by caste. On his reading Kshemaraja was reinstating the text within the family of scriptures it belonged to. Kshemaraja’s reason is caste. Sanderson’s is genre.

Whether the men he was attacking had in fact been putting water in the cup, or only water in the word, is the one thing his commentary cannot be asked, since no commentary from that side survives complete.

What the book is

Strip away the argument and look at the object.

Fifteen chapters, called paṭalas, almost all of it in the ordinary thirty-two-syllable metre. The verse total is not settled. The scripture itself is no help: at 1.5 it says it is a thousand million verses long, and the book in your hands is the summary. Arraj, working from the printed edition, counts 3,648 and a half, a half-verse being left over. Adding up the highest verse number in each chapter of the Pondicherry electronic text gives 3,679. The printed edition states no total anywhere, though the Jammu and Kashmir Research and Publications Department’s own catalogue of the series calls it “an extensive work consisting of 3780 verses”. That one is hard to credit. The same sentence gives the tenth chapter 1,280 verses and the fourteenth 28, and both are exactly right, which leaves about a hundred verses to be found somewhere in the other thirteen. The catalogue states no method, misnumbers one volume of the edition it is describing, and lists six volume numbers for a work published in seven parts. There is no critical edition of this text, so there is no critically established total.

The shape of the book is lopsided in a way that tells you something. The first chapter gives the mantras. The second gives daily worship. The third and fourth give the preparatory rites and then the initiation, and the fourth is enormous, 546 verses of liturgy. The fifth gives a second mode of initiation, the sixth and ninth and thirteenth belong to the sādhaka, and the eighth matches adept to deity. The seventh is on time. The eleventh and twelfth work through the planes of reality. The fourteenth gives eighteen hand-gestures, which Arraj reads as imitating the weapons and poses of the god’s eighteen arms. The fifteenth gives the chummaka, the coded speech by which adepts referred to things in public.

And the tenth chapter is 1,280 verses long, which is more than a third of the entire scripture and more than twice the size of the initiation chapter.

That chapter is the bhuvanādhvan, the path of the worlds: the Shaiva cosmology, absorbing and extending the whole Puranic map. At the bottom sits Kalagnirudra, the Rudra of the fire of time, from whom pours the fire that eats the worlds at the end of an age. Above him the hells, and there are more of them than anyone is going to name. At 10.31 the text puts the number at five hundred million. At 10.32, before a single name, it says it will give the principal ones, a hundred and forty of them, and tells the Goddess to listen. The names begin at 10.33 and run on for pages, and Kshemaraja goes through them explaining what each one means. He says where the explanations come from. Most are quoted out of another scripture, the Śrīparā: the first hell is Avichi, so called because it is covered with waves of pus and blood and mud, iti śrīparāyāṃ niruktāḥ, thus etymologised in the Śrīparā. Where that source runs out he glosses a name himself. Arraj’s comment is that the list surpasses the enumerations found in the Puranas, and he notices something else about it. The list runs from page fifteen of that volume to page thirty, and then at page thirty-nine the dialogue announces, once again, that the most important hells will now be enumerated. Two lists have been sewn together, and the seam is still showing.

Seams like that are what Arraj’s reconstruction is built out of. He proposed that the book grew out of six sources. At the bottom a Bhairava source, the base document, carrying the rites of the god, and combined with it in the same first stage a stratum of short early Shaiva practices with no tie to Bhairava at all. Then the scriptural Shaiva layer, the agamic ritual that turned a manual into a scripture. The cosmology is not a seventh source in that scheme but a complication inside it: he reads the tenth to twelfth chapters as preserving, in several half-standardised versions, a process that other texts collapse into two stages, the early Shaiva and the scriptural Shaiva; and elsewhere he treats the tenth book’s swelling as late. Then a tantric source supplying material piecemeal wherever the text was receptive to it. Then later Bhairava material arriving in finished, self-contained blocks. And at the top a Kaula source: the fifteenth chapter, the code-words, bolted on with the barest of links. Running through all six he posits a final redactorial hand, smoothing. This is the one full account anyone has published of how this particular book came to be shaped the way it is, and it comes from an unpublished 1988 dissertation whose dating framework has since been overtaken. Nobody has taken it up since, and nobody has replaced it either.

A cosmology like that reads as decoration until you see what it is for. The officiant has to know the worlds in order to burn a candidate free of them, world by world, in the initiation. As Kshemaraja explains it, when a world is purified it can no longer obstruct the initiate, because whatever he had accumulated that would have had to be discharged there is now cancelled. On his reading, the map is an itinerary, and the point of drawing it is to be able to burn it.

How many worlds are there? At 11.196 the text says only śataśo ‘tha sahasraśaḥ, by hundreds and by thousands, and it never gives a total. Where it does give a number, at 10.1185, fifty-nine worlds in the plane of Ishvara are to be purified at initiation, and Kshemaraja has to go back over the list he has just given and count it again to make it come out at fifty-nine. Anyone who gives you a confident total for the worlds of the Svacchanda is giving you a commentator’s arithmetic.

A caution belongs here about the shelf this book is filed on. Calling it a scripture of the cremation grounds leans on its place in the Bhairava canon, and Arraj thinks the filing was done later than the book.

His evidence is a small internal contradiction, and the text does classify itself, so the contradiction is easy to follow. At 1.5 the Goddess calls it the catuṣpīṭha mahātantra, the great tantra of four seats. Kshemaraja glosses the four as mantra, vidyā, maṇḍala and mudrā, and quotes another scripture, the Sarvavīra, for the scheme. But that quotation, printed on page 10 of the first Kashmir volume, files the Svacchanda under the seat of mantras alone, alongside the Caṇḍa, the Krodha and the Unmatta. The authority the commentator reaches for puts the book on one shelf while the commentator is insisting it covers all four. Arraj reads the discrepancy as a sign that the fourfold scheme is a later way of arranging texts, and no part of how anyone composed one. About the dualist and non-dualist labels he is blunter, calling that schema an ex post facto superimposition, and about the canonical number sixty-four blunter still: a factitious and retrospective construction.

He also grants the filing clerk a point. Classing the Svacchanda as a mantra text, he says, is an understandable assertion given the actual predominance of mantra practice in its content. And none of this makes him a sceptic about where the book came from. His own reconstruction of it puts wandering ascetics and cremation-ground rites at the very bottom. His doubt attaches to the cataloguing system. The material inside the book he takes as it comes.

One more thing the name misleads about. Svacchandabhairava is one deity among the Bhairavas, and the name is not a synonym for the family. At 2.88 to 2.94 the text describes a specific deity: five faces, three eyes on each of them, eighteen arms, skull-garland and sword and shield and severed head and the skull-topped staff called a khaṭvāṅga, a lion skin worn with an elephant hide over it, and the upper face the colour of crystal. At 1.75 Aghoreshvari is the one who goes upon Svacchanda’s lap. Sanderson, on page 670, calls him transitional in type between the calm Sadashiva of the Shaiva Siddhanta, the dualist school of the twenty-eight Agamas, and the gods of the Kapalika mainstream, the skull-bearing ascetics of the cremation grounds. Milder elements, is his phrase. He says in the same breath that the scripture also teaches secondary forms, Kotaraksha and Vyadhibhaksha, which stand closer to the standard Bhairavas of that Kapalika tradition. The mild one is the one who ended up the household god of the Kashmiri brahmins.

The clock in the breath

Take one chapter as a specimen. The seventh is about time, and it opens by dividing time in two: saura, the time of the sun, and ādhyātmika, the time inside a person.

The proportion is in Arraj’s page ranges. External time runs over three pages of the printed edition. Internal time runs over the next hundred.

The premise is the old correspondence between the small world and the large one, and the text pushes it much further than a metaphor. At 7.90 it states that six finger-breadths of the breath’s course make one sign of the zodiac, and the rest is the arithmetic that follows: six signs rising is the sun’s northern course, and the full circuit is a day and a night. The circuit runs thirty-six finger-breadths from the heart-lotus, out past the tip of the nose and twelve breadths beyond it, to the place the commentary calls the dvādaśānta, the end of the twelve, and back. An eighth of that is one watch of three hours. Coughing and sighing shorten or lengthen the course, and where the bright and dark fortnights overlap you get, inside the body, the equivalent of a solar or lunar eclipse. Kshemaraja notes that these happen by the procedure stated in the astronomical treatises.

Follow it far enough and a single breath contains twelve years, and then sixty. What the chapter claims, in Arraj’s summary: these practices are extolled for yielding, in a condensed cycle of a single breath, all the success that would normally require a long sequence of external rituals.

Then the chapter turns, and it turns towards death.

An adept who has learned to read his own breath can read his own end in it. At 7.180 to 7.182, six watches of flow in the right channel means three months of life. Four watches, one month. Two watches, eight days. The text is so compressed here that Kshemaraja records conflicting interpretations of it.

And then it says something unexpected. It offers the sign to a person who has done none of this.

कर्णरन्ध्रकृताङ्गुष्ठो घोषं न शृणुते यदा । मरणं तस्य देवेशि षण्मासेन विनिर्दिशेत् ॥१८७॥

karṇarandhrakṛtāṅguṣṭho ghoṣaṃ na śṛṇute yadā | maraṇaṃ tasya deveśi ṣaṇmāsena vinirdiśet

“When a man has put his thumb into the opening of his ear and does not hear the roar, one should declare his death within six months, O mistress of the gods.”

Svacchandatantra 7.187 #

The verse before it introduces this as knowledge of death ayuktasyāpi ca prāṇe, for one who is not disciplined in the breath at all. The verse after it says that when he can still hear the roar but no longer the finer sound within it, he has a month. The text calls that finer sound a cīravāk-ciñcinī, and Kshemaraja glosses it as a sound like the ciñcinī, the chirp made by the creature called cīravāc. The dictionaries know it under near neighbours of that name, cīrī and cīrīvāka. A cricket, named for the noise it makes.

It requires no initiation and nobody to teach it to you. You stop your ear and listen, and the scripture tells you what the silence means.

The commentator and his instruments

All of it reaches us through one man’s hands. Kshemaraja worked roughly between 1000 and 1050. Navjivan Rastogi, in the first volume of his study of the Krama, places him at the close of the tenth century or the beginning of the eleventh, and reports that K. C. Pandey counted eighteen works ascribed to him, while another scholar’s list runs to twenty. He also calls him Abhinavagupta’s cousin as well as his pupil, though he rests that on Pandey’s identification of Kshemaraja’s father with Abhinavagupta’s uncle Vamadeva, and Sanderson calls him only the pupil.

He was not the first to write about this scripture, and he is careful to tell us so, because he is arguing. He names names.

Sadyojyotih he attacks under one of the several names he uses for him, Khetapala, laying out his account of how grace descends and then dismantling it across pages 84 to 86 of the third Kashmir volume. Narayanakantha he handles more coldly. Having reported his reading at 4.208, on page 130 of the second volume, he writes that whether it is right or wrong the discerning may judge, and enough. And in the tenth chapter, at page 211 of the fifth, he names Rajanaka Bhullaka, bṛhaṭṭīkākāra, the author of the big commentary, says he expounded the verse svarucyā, according to his own fancy, and rules that it is to be disregarded as being without substance.

That last one is the loss. Of the three, only Bhullaka is identifiable as having written on the Svacchanda itself, and Sanderson records that commentary as lost. Narayanakantha’s surviving commentary is on the Mṛgendra; Sadyojyotih’s are on the Svāyambhuvasūtrasaṅgraha and the Raurava. With those two, Kshemaraja is often rejecting readings of comparable passages elsewhere and carrying the fight over to this text. As for the word bṛhaṭṭīkā, it will not settle whether Bhullaka’s commentary bore it as a title. Sanderson uses the same word to describe a lengthy commentary Narayanakantha wrote on Sadyojyotih’s Tattvasaṅgraha, which has itself not come to light and is known only because a South Indian Saiddhantika, Aghorashiva, mentions it, saying he wrote a short commentary on the same work for readers the long one had defeated. Long commentaries on this material had a way of not surviving.

One man in that circle he does not attack. Brihaspati, author of the lost Śivatanuśāstra, was a Saiddhantika as well, and on the fourth page of the entire commentary Kshemaraja cites him for support, with the honorific, as one of his warrants for etymologising the name Bhairava the way he does. Abhinavagupta does the same thing in the Tantraloka. That is one citation, and a single citation should not be made to carry a thesis about the dualists at large. But see it once: on the fourth page of a commentary written to demolish the dualist reading, the warrant for the method is a dualist.

What is lost is not quite everything. Bhullaka survives in the sentences his opponent reports in order to demolish them. And one stray opinion from some other commentary got out. A Kashmiri initiation manual, the Kalādīkṣāpaddhati, settles a question about whether the candidate should be shaved by citing the Svacchanda and adding that “in the commentary on that text and related sources we read that his purification is by bathing alone”. Sanderson, printing the passage, notices something and says it in a footnote: the commentary mentioned here is not Kshemaraja’s, because on that quarter-verse, 4.219c, Kshemaraja explains only why the man bathes. The manual’s own phrase is ṭīkādāv api, in the commentary and so forth, so the inference that a second commentary on the Svacchanda is meant is an inference, not a statement. It is a reasonable one: Bhullaka’s commentary is attested and lost. If it holds, then a Kashmiri ritualist was still consulting a commentator other than Kshemaraja, and what survives of that man is a single ruling: that a bath is enough.

The older literature gets these men’s dates wrong, and this article nearly did too. Arraj, in 1988, placed Sadyojyotih and Brihaspati around the year 900, and Narayanakantha around 1075, which would put him a generation after the man who quotes and dismisses him. Sanderson has since fixed the Kantha family, in the 2007 study of Kashmiri exegesis: Narayanakantha and his son Bhatta Ramakantha were both active in the second half of the tenth century, a generation before Kshemaraja. And in a long article of 2006 he moved Sadyojyotih back to about 675 to 725, with Brihaspati more tentatively at about 650 to 750. Dominic Goodall and others have followed him.

That pushes the earliest Saiddhantika reading of Shaiva scripture back by two centuries, and opens a long silence in the transmitted exegesis between those founders and the first Kashmiri commentators. Where the Svacchanda itself sits in that stretch has never been fixed, and it is not a blank either. On the same page as the footnote about its provenance, Sanderson says that the Netra Tantra was composed in Kashmir between about 700 and 850, that the Svacchanda was used by the redactor of that text, and that its practice was therefore probably already current in Kashmir at that time. So on his reckoning the book is probably older than 850, and probably in use in the valley at least a century and a half before Kshemaraja sat down to explain it.

There is one datable Kashmiri report from inside that stretch, though it reaches us from three centuries later, in Kalhana’s chronicle of about 1148. At Rajatarangini 5.55, in the reign of Avantivarman, who ruled from 855 or 856 to 883, a minister goes to have a man beheaded at a temple of Bhairava that stands, with a circle of the Mothers, near the shrine of Bhutesha. Kalhana mentions no book. What his sentence establishes is that a Bhairava temple with a Mother-circle stood in the valley in exactly the decades Sanderson is talking about.

Arraj declined to fix a date, giving only outer brackets: not before the seventh and eighth-century flowering of Shaiva royal patronage, not later than Abhinavagupta in the tenth, probably earlier than Vasugupta in the early ninth. In an anonymous composite scripture, he argued, a datable detail dates only the passage it sits in. His chronology has in any case been overtaken. The Niśvāsatattvasaṃhitā, which Goodall and Isaacson have since made the case for as the earliest surviving Shaiva tantra, goes unmentioned in his dissertation; it was unedited when he wrote.

Kshemaraja’s instruments are not what a modern reader expects a commentator to use. Arraj documents each of the four that follow, though he sets them out in an order of his own.

He re-splits words. The Goddess’s imperative kathayasva, “tell!”, he divides as kathaya sva, so that sva becomes a vocative addressed to the Self-formed supreme lord, and a non-dual doctrine falls out of a verb ending. Arraj lists four places where he does it and one where the same word is left alone.

He etymologises rather than translates, preferring the derived sense of a word to its conventional one, which as Arraj observes immeasurably widened the scope for sectarian commentary. Kailasa becomes head, cardamom seed and seat. Even a throwaway vocative used to fill out a line, varānane, “O fair-faced one”, is unfolded into she whose face is the pulsation of Shakti, the entry into the supreme state.

He denies that scripture contains any explanatory filler at all, following Abhinavagupta, so that nothing in the text can be set aside as mere praise. And where the Sanskrit is simply not standard, he does not emend it. He labels it aiśa, the usage of the Lord, and moves on.

Arraj’s verdict on all this is more careful than a soundbite. Both Kshemaraja and the Saiddhantika commentators before him, he says, had systematic prejudgements that coloured their work, and a cultural and temporal gap separated all of them from the men who compiled the scripture. So the quality of any given passage of commentary may range, in his words, from biased falsification to literal recording of a living tradition. And then, in the next sentence, he gives him back his ordinary motive. Alongside the polemic, he writes, Kshemaraja probably wanted to offer a clear and coherent guide to the ritual and practices surrounding the scriptural cult of Bhairava.

You can even see where the tradition runs out. Arraj noticed that Kshemaraja’s remarks go thin and perfunctory over some of the Bhairava sections, the ones the other commentators had felt obliged to explain away, and reads that thinning as a measure of the distance between him and the circles the material came from. The commentary is at its most confident where the practice was still alive, and it thins as that runs out.

He also failed at the thing he most wanted. Arraj’s conclusion is that in finding a direct and unequivocal assertion of non-dualism in the scripture, Kshemaraja appeared to have little success, because explicit dualism and non-dualism lay outside the intellectual world of the men who assembled it. He was arguing a case the text does not contain, against opponents arguing a case it does not contain either.

At the foot of page 146 of the last volume, after the closing colophon, he says why he wrote it.

गतानुगतिकप्रोक्तभेदव्याख्यातमोऽपनुत् । तेनाद्वैतामृतस्फीतः स्वच्छन्दोद्द्योत उम्भितः ॥

gatānugatika-prokta-bheda-vyākhyā-tamo ‘panut | tenādvaitāmṛta-sphītaḥ svacchandoddyota umbhitaḥ

“Therefore this Svacchandoddyota, dispeller of the darkness of the dualist commentary uttered by mere followers of the beaten track, swollen with the nectar of non-duality, has been put together.”

Svacchandoddyota, at the end of paṭala 15 #

Uddyota means illumination. He was quite clear about what he thought he was lighting up, and about whose lamp he was putting out.

Who tamed it

In the literature this article draws on, nobody argues that the Kashmirian cult ended up anything but domesticated. The god of the cremation ground became the household deity of Kashmiri brahmins. Almost nobody claims to know who did it, and the question turns out to be harder to ask than it looks.

Start with what Sanderson does not say. His 1988 account is usually read as blaming the Shaiva Siddhanta, and it does not. The Shaiva cult of the majority in Kashmir, he wrote, was that of Svacchandabhairava, which derived its authority from the Svacchandatantra of the Mantrapitha section of the Tantras of Bhairava, and was therefore, strictly speaking, a tradition of the Kapalika-based left. The Mantrapitha is one of the two seats into which the canon-lists sort these scriptures. Then the sentence everyone quotes half of: none the less, the Kashmiri practised a thoroughly domesticated form of the cult, and in the tenth century the Shaiva Siddhanta, though not its source, had taken advantage of this to bring it under the sway of its doctrines. The domestication is already done inside that sentence. It gets no date and no agent. What the tenth century dates is the taking advantage.

Nineteen years later he offered a mechanism, and hedged it twice. In the 2007 study of Kashmiri exegesis, at page 429, he says it is tempting to conclude that this middle-ground Shaivism was the mainstream all along, and that the Siddhanta had little impact on the core practice of the majority. What it did do, on that reading, is supply a view of how Shaiva practice ought to stand towards brahmanical norms. That view encouraged and justified a drift away from the harder observances. And the drift, he says, would probably have occurred anyway, as the routinization you would expect in any tradition that won wide acceptance inside brahmanical society. The Siddhanta encouraged and justified it without being necessary to it.

Two things about that sentence usually get lost. The first is a clause in the middle of it. The Siddhanta had little impact on the core practice of the majority like the Shakta Shaivism of the Trika and Krama. Kshemaraja’s own school is inside the same demotion.

The second is that Sanderson does not let the tempting conclusion stand. On the same page he turns on it. The very manuals whose existence shows how dominant the two Bhairava cults became, he says, also contain evidence that must make us hesitate to conclude that this state of affairs goes back to the golden age of Shaiva exegesis. His reason is that those manuals are eclectic where the Tantras and their learned commentators are not, and that they record what the local officiants actually did. Then the sentence that turns the picture round: now, the Siddhanta figures very strongly in this respect. His example is a Shiva-vase set up in the preliminaries of a Shaiva cremation, whose principal god carries a Bhairava name and turns out, on inspection, to be an iconographic variant of the Siddhanta’s own ten-armed, five-faced Sadashiva. So the caution runs the other way from a second hedge on routinization. The practice that survived carries a good deal of Siddhanta in it.

Two pages on he draws the moral, and it is the sentence this whole section turns on. The dominance of the Bhairava and Shakta cults in the later evidence, he writes, may be more an effect of the Siddhanta’s demise than its cause, and the weight of Saiddhantika elements inside the surviving manuals should be enough to inhibit the facile conclusion that the Siddhanta had always been a marginal phenomenon in Kashmir. And he thinks the cause of that demise more probably political than doctrinal: the coming of Muslim rule in the fourteenth century, and with it the withdrawal of royal patronage from the public sphere that had been the Saiddhantikas’ own territory.

The one full alternative comes from Arraj, writing in the same year as Sanderson’s first account and apparently without knowledge of that particular essay. He put the taming inside the text’s own history. On his reconstruction the base document belonged to wandering ascetics, the worship of Bhairava was then remodelled along the lines of the cult of Shiva, patronage arrived, institutions hardened, and eventually, in his words, “the ritual became mechanical, caste distinctions were introduced and the category of the adept virtually dropped.” That is close to what Sanderson would later call routinization, relocated several centuries earlier and out of Kashmir altogether. It is also where his reading of the caste verses does its work: if the scripture really did put the initiate beyond caste outside the ritual as well as inside it, then a later reintroduction of caste distinctions is a change to the religion itself, and something a reconstruction can date by.

One qualification, and Arraj supplies it himself in his very next sentence. That trajectory is the institutionalised Shaiva sect’s at large. The Bhairava side, he writes, which had royal patronage and temple building of its own, retained stronger connections to the ascetic groups and stayed open to innovation coming in from the periphery, and he reads the Svacchanda’s successive interpolations of esoteric material as the evidence. Patronage arrived on that side too, in other words, and did not produce the same result. So his routinization describes the current his text partly swam against, more than it accounts for how this cult got tame.

Two further voices belong here, and neither is answering quite this question. Mark Dyczkowski, in 1988, describes the Svacchanda cult as the mild thing that filled a space left by something wilder: the controversial Kaula rites, he writes in one passing sentence, gradually ceased to be performed, giving way to tamer and milder cults, such as that of Svacchandabhairava. His subject there is why the Agamas fell out of use in Kashmir, and the period is later than the one at issue here. And Gavin Flood, in 2025, objects to the shape of explanations of this kind. Writing about the Netra Tantra, he grants Sanderson’s account of it as royal protective apparatus and then says it is not the whole story, because in the chapters he examines the impulse is theological and not political, and the culture a scripture creates is inherent in the text and does not come only from the forces around it. He says nothing about the Svacchanda, and his target is a political explanation where Sanderson’s account of this cult is a sociological one, so the objection does not land here as squarely as it looks. What it does do is name a habit to watch for: reaching first for a cause outside the book.

So the tally is thinner than a list of names suggests. Sanderson floats a mechanism and then, on the same page, produces the evidence that cuts against the picture it belongs to. Arraj offers a full reconstruction and then exempts the Bhairava side from it himself. Dyczkowski is describing a later collapse and Flood a different text.

What the five accounts do establish between them is worth having. The taming is undated: nobody puts a year on it and nobody names an agent. The Siddhanta was not the source of the cult and is not required to explain its softening. And whatever softened it, the Siddhanta’s fingerprints are all over the manuals that survived. That is not an answer to who tamed it. It is a fairly exact description of the shape of the hole where the answer should be.

There is a further wrinkle, and it needs stating carefully, because it is easy to overstate. In a footnote running across pages 428 and 429 of the same study, Sanderson adds one flat sentence: he sees no reason to think the Svacchanda too was Kashmirian. That sentence records where the evidence runs out, and Arraj concludes only for a north Indian origin. Neither of them says where the book was made. What is demonstrably Kashmiri is the exegesis and the household worship built on the book.

A Buddhist objection, answered in an appendix

The dualists were the opponents inside the house. The Shaiva claim about initiation is a strong one: a rite performed by another person, on your body, can end your rebirth. Somebody outside was always going to object.

Dharmakirti did, in the Pramanasiddhi chapter of his Pramanavarttika, and he made two arguments. The first is that a ritual which works on seeds is not enough to stop the birth of persons, because if sterilising a seed is what liberates, then rubbing it with oil or scorching it in fire ought to liberate too. The second attacks a specific rite in which the candidate is weighed before and after his initiation and found lighter, offered as proof that his sin had been removed. Sin, says Dharmakirti, has no weight, because sin has no form. Let the man be weightless. It proves nothing.

Somadeva Vasudeva, writing in 2020, follows Diwakar Acharya in tracing the target of that attack to the early dualist Siddhanta, which matters, because it means Dharmakirti was not arguing with Kshemaraja’s tradition at all. Kshemaraja answered anyway, centuries downstream, in a short tract appended to his commentary. He tells us so himself, in his commentary on Bhatta Narayana’s Stavacintāmaṇi: he analysed the question in detail, he says, in his Svacchanda commentary, at the end of the fifth chapter, in the context of justifying initiation. The tract is there, about eleven pages of it, running from page 73 to page 84 of the third Kashmir volume.

His answer runs like this. Oil applied to a seed can only obstruct a capacity the seed already has. Mantras do something further, because initiation both eradicates and confers: it removes the latent impressions of bondage, and it gives the true state of knowledge. Dharmakirti’s analogy models only half of what is being claimed. And the reason mantras can do it, Kshemaraja says, is acintyaprabhāvatvāt, because their power is unthinkable.

Whether that is an argument or the end of one is left to the reader. What is odd is how his own teacher introduces the same material. The twentieth chapter of the Tantraloka is sixteen verses on these probative signs, and Abhinavagupta opens it by calling this the initiation that gives comfort to the simple-minded. Then he explains it anyway, having told the reader first who it is for.

What Dharmakirti was attacking, and what Kshemaraja spent those eleven pages defending, is the same proposition the caste verses rest on: that a rite performed on a man by another man changes what he is. Take that away and jātyuddhāra is theatre. The Buddhist saw the load-bearing wall and went for it.

What survives

The Svacchanda survived in three scripts and two transmissions. Sharada and Devanagari manuscripts carry the text together with Kshemaraja’s commentary. The Newari manuscripts of Nepal carry the text alone, with no commentary at all, and Arraj holds that they may preserve a version Kshemaraja never saw, which is why he wanted two critical editions of this scripture and not one. The oldest witness anyone cites is a palm-leaf in the National Archives in Kathmandu, MS 1-224, filmed as NGMPP reel B 28/18, Newari script, dated 1067 or 1068; Sanderson catalogues it, and a transcript at Pondicherry, IFP T. 507, under a longer title, Svacchandalalitabhairavatantra. Even the name of the book is not quite stable across the two transmissions.

There is a verse where the two transmissions can be caught disagreeing. At 4.470 the text lists the insignia handed to a newly consecrated master, and the Kashmir edition reads chatraṃ pādukam āsanam, parasol, sandals, seat. The Kathmandu palm-leaf reads chatrapādukacāmarāḥ instead, parasol, sandals, fly-whisk, and so does the Pondicherry transcript of the Grantha line. That would be a small thing on its own. What makes it useful is the third witness. Bhojadeva of Dhara, who ruled in Malwa in the first half of the eleventh century and whose Siddhāntasārapaddhati draws on this verse, has the fly-whisk too. An author in central India, datable within decades, was reading the version Nepal and the south preserve and not the one Kashmir printed. Sanderson sets the three readings side by side in The Saiva Age.

The printed edition was made by Pandit Madhusudan Kaul Shastri and appeared between 1921 and 1935 in the Kashmir Series of Texts and Studies, six volumes in seven parts. His preface names three manuscripts, all in Sharada. One was copied by Rajanaka Ratnakantha in the Shaka year 1608, which is 1686 or 1687, worm-eaten and repaired here and there. The footnotes cite variants by Devanagari sigla, and the first three, क, ख and ग, are presumably those three manuscripts. But two more sigla appear. A single footnote on page 34 of the second volume cites four at once, ख, ग, घ and ङ. घ and ङ stand together again, alongside others, on page 248 of the first volume and page 129 of the second, and घ turns up without ङ on pages 98 and 106 of the third, alone on the first of those. क and घ are cited side by side on page 18 of the third. So the apparatus draws on five witnesses. No volume after the first carries a preface, and the fourth and fifth are never described anywhere. The edition collates five manuscripts and tells you about three.

The series itself was founded in 1911 by Jagadish Chandra Chatterji, whose title page gives him as Director of the Archaeological and Research Department, Kashmir State. Three years later, as the second volume of his own series, Chatterji published Kashmir Shaivaism, and the tradition has carried a version of that name ever since. He did not claim to have invented it. His first sentence says that “the first beginnings of what has been called ‘Kashmir Shaivaism’, to distinguish it from other forms of Shaivaism known and still practised in different parts of India, may have to be traced” to the Shiva Sutras, the volume he had printed first. What he did do is fix the scope of the name, in his own subtitle: a brief introduction to the history, literature and doctrines of the Advaita Shaiva philosophy of Kashmir, specifically called the Trika System.

The label names a philosophy and a school. This scripture is a ritual manual with no doctrinal credo, and what is philosophical about it was supplied by a commentator writing a very long time after it was fixed. The series he founded printed this scripture in seven parts, long after he had left it; by 1918 the volumes were coming out under another editor. The name he fixed describes what was written about it.

And that is roughly where it still sits. No complete English translation of the Svacchanda Tantra has been published, as far as we have been able to establish in September 2026. Arraj translated fragments as appendices to his thesis. Simone McCarter has translated passages from the first, second and fourth chapters, most of them from the second. Satya Prakash Singh and Swami Maheshvarananda included one chapter, the seventh, on the breath, in a 2016 volume of selected chapters from five Kashmiri Shaiva texts. The whole text is available in Sanskrit with a Hindi commentary, in Radheshyam Chaturvedi’s Chowkhamba edition of 2005 and in Paramhansa Mishra’s. For the rest, you read the Sanskrit or you read about it.

Sanderson recorded, twice, that the ritual manuals built on this scripture and on the Netra were the practical guides of Kashmir’s Shaiva officiants and other initiates down to recent times, and were in use until recently among the Tantric family priests. In 1990, during the insurgency, almost the whole Kashmiri Pandit community left the valley. What became of the officiant families who had used these manuals, and of the manuals themselves, we have found no published account of. Swami Lakshman Joo, who died in 1991 and is usually described as the last master of the tradition, left a published corpus of the other Kashmiri schools: the Trika of the three goddesses, the Pratyabhijna of recognition, the Krama of the sequence, which is the cycle of the goddesses of consciousness, and the Spanda of the divine vibration. Nothing on the Svacchanda appears in the list of his works kept by the Ishwar Ashram Trust, or in the Lakshmanjoo Academy’s series of the classic texts, though those are organisational listings and not a scholarly catalogue. Sanderson, writing about him in a centenary volume in 2007, puts it more sharply. What Lakshman Joo carried was a Trika of meditation and gnosis. The ritual tradition, the one the Svacchanda and Netra manuals had served, was the part that had already gone.

The manuscripts are being photographed. A British Library endangered-archives project, EAP886, has digitised about two thousand manuscripts from the Raghunath Temple collection in Jammu. Its own description calls the collections particularly vulnerable to pests and other environmental conditions, and to the political instability in the region, especially given that the temple has been targeted by two separate suicide attacks in the past. The Kashmir Series volumes are all online, free, scanned page by page, including the pages quoted in this article.

Who was copied

Kshemaraja lost the argument he thought he was having, the one about whether this book teaches non-duality, and won the one that turned out to matter. Sanderson says so plainly. On the next page of that 1988 survey he writes that the attempt to throw off the influence of the Shaiva Siddhanta was entirely successful. We can say only three things, he goes on: that the corpus of anonymous Kashmiri Shaiva ritual texts is completely non-dualistic in the manner Abhinavagupta and Kshemaraja defined, that this corpus records a tradition which must go back at least five hundred years, and that after the eleventh century there is no trace of any Kashmiri literature in the doctrine or the liturgy of the Siddhanta. How quickly it happened, he says first, cannot be seen from the evidence so far uncovered.

That sits oddly beside the Saiddhantika fingerprints all over the surviving ritual manuals, and the two statements are nineteen years apart. His own way of holding them together is the one quoted above: the Siddhanta went, and the dominance of what was left may be the effect of its going rather than the cause.

So the outcome is settled. What is lost is the argument itself, because the manuscripts that went on being copied were his, and when a press in Bombay finally set the text in type, his was the reading that came with it.

What his opponents said about this book survives in the sentences he reported in order to break them, and one ruling more, from a commentary nobody has identified, that a bath was enough.