Eighty-five titles and no text: the Saura tantras and how a tradition disappears

In 2018, in Varanasi, the Sampurnanand Sanskrit University published a book titled Saura Agama Tantra. Its author was Shitala Prasad Pandey, Professor and Head of the Department of Dharmagama at Banaras Hindu University. If anyone in India was placed to produce a Saura scripture, it was this man: the chair of a department devoted to the study of agama, revealed scripture, at the country’s most famous Sanskrit faculty, publishing in a tantra series, in the city where such things are still read.

The book is not an edition of a Saura agama. Its main text runs to a hundred and thirty-six pages of quotations gathered from the Veda, the Upanishads, the Puranas, the Mahabharata, the Upavedas and various shastras, assembled into an essay, with fifty-odd pages of appendices after it. The foreword, by the Vice-Chancellor of the publishing university, says so in Hindi and in plain words: the Saura, Ganapatya and Skanda agamas are called secondary for one reason only, the extreme scarcity of literature directly connected with them, because in the course of time their many sources were lost. What the author did, the foreword says, was sankalan. Compilation.

And on page eighty-six, in a chapter on temples and image-consecration, Professor Pandey records the moment that gives this article its subject. He had been told in conversation by Acharya Gayacharan Tripathi that a manuscript of a Saura-samuccaya was held in a library in Bhubaneswar. He writes: kintu mujhe ukta pandulipi prapt nahin hui hai. But I did not obtain that manuscript.

The only manuscript reproduced anywhere in the volume is a photocopy in an appendix. It is two leaves. It is a kavaca, a protective hymn, from the Raghunath Temple manuscript library in Jammu, catalogued there under the heading stotra, praise-poem. Its title is Rudrayamalokta-Suryakavaca: even the two-leaf hymn claims descent from a Śaiva tantra.

That is the state of the field, and it is not a failure of the scholar. There was nothing to find, and he says so.

A word about the word

“Saura” in what follows is a label of convenience, and it is worth saying so before it starts doing work. It covers a sixth-century priesthood serving Sun images; a canon-list of eighty-five titles whose contents have not been read for this article; a tenth-century Sun-tantra that is Śaiva on its first page; a Kashmirian Śākta cult of the Sun under another name; a caste that still exists; and a sect on a fourteenth-century Javanese legal board. Whether those were ever one tradition is exactly the question the missing books would have answered. They are gone, so the label has to stand in for the answer.

Eighty-five titles

We know how much of it there was.

Somewhere before the eleventh century, a Śaiva scripture called the Shrikanthiyasamhita set out a survey of the scriptural authorities of its world. It listed the eighteen Puranas. It listed a hundred and sixteen Pancharatra samhitas, the scriptures of the Vaishnava tantric tradition. It listed twenty-eight Siddhanta tantras, twenty-eight Garuda tantras, twenty Bhuta tantras, twenty-four Dakshina and twenty-four Vama tantras, and sixty-four of what it called the circle of Sadashiva. It is one of the fullest such surveys to reach us, and it was drawn from inside the tradition rather than by a later cataloguer. Readers of the two seats will recognise the milieu: this is the same Kashmirian world of canon-lists and scriptural bookkeeping in which the Jayadrathayamala classifies itself.

Between the Pancharatra list and the section on the five streams, the Shrikanthiyasamhita set out one more canon. Jürgen Hanneder, who published the transcript, summarises the section as the division of the canon of tantras taught by Surya. Alexis Sanderson counted the titles. There are eighty-five.

The Shrikanthiyasamhita itself is lost. Its canon section survives only because a Kashmirian compiler named Rajanaka Takshakavarta copied it into a digest called the Nityadisamgraha, and that digest survives in a single manuscript, now at the Bhandarkar Oriental Research Institute in Pune, known to scholarship through an apograph among Aurel Stein’s papers in the Bodleian. So the list of the Sun’s scriptures reaches us at two removes, preserved by a Śaiva who was cataloguing his rivals, in a codex that exists once.

The eighty-five titles themselves sit in Hanneder’s printed transcript. They have not been read for this article. What is reported here is the count.

Sanderson set out what the count means in The Saiva Age, his 2009 study of the rise of Śaivism, and the sentence is worth having exactly:

It appears that the Sauras, the initiated devotees of the Sun-god, possessed their own canon of scriptures, known, like those of the Śaivas and the Vaiṣṇava followers of the Pañcarātra, as Saṃhitās.

Read the words initiated devotees, and then read the words it appears. If the first is right, the tradition had teachers to confer initiation, a mantra to confer, and a line of people who had received it. Not one of those three is attested for a distinctly Saura lineage in any surviving source. The word is Sanderson’s inference from the shape of a canon-list, and this article rests on it.

What is attested is a priesthood. Around the year 550, Varahamihira, the astronomer of Ujjain, wrote a chapter on the installation of images, and in it he set out which priest to employ for which god: the Bhagavatas for Vishnu, the ash-wearing twice-born for Śiva, brahmins who know the mandala sequence for the Mothers, brahmins generally for Brahma, the Shakyas for the Buddha, the naked ones for the Jinas, and for Savitr the Sun, the Magas. He closes the verse with a principle: ye yam devam upashritah svavidhina tais tasya karya kriya. Whichever god men have taken refuge in, by those men should that god’s rite be done, according to their own prescribed method.

That is what the verse gives you. In the sixth century the Sun had a priesthood of his own and a rite performed by its own rule. It does not give you initiation, and it never uses the word Saura.

They had kings, too, for a while, though it is worth being careful about what a king’s title proves. Sanderson’s list of rulers who styled themselves paramasaura or paramadityabhakta, wholly devoted to the Sun, is short and clusters in the sixth and seventh centuries: a king of Padmakholi in Ganjam, a Maitraka of Valabhi, the three predecessors of Harsha at Kanauj, the Gurjaras of Bharukachchha. There are stragglers afterwards, Ramabhadra and Vinayakapala among the Gurjara-Pratiharas in the ninth century, the Senas of Bengal at the end of the twelfth. And then something happens that reads like a weather report. After Jayabhata the Second, the next three kings of that same Gurjara line, Dadda the Third, Jayabhata the Third and Ahirola, describe themselves as paramamaheshvara, wholly devoted to Śiva. Three successive reigns, sun to Śiva, recorded in a dynasty’s own copper plates.

As for the Senas: Lakshmanasena is paramasaura in one charter, paramanarasimha in another and paramavaishnava in a third. That is a man announcing a preference, not a membership.

Then Sanderson’s other sentence, the one this whole article stands on:

Unfortunately, no manuscript of any one of these Saura scriptures has surfaced; and the decline of Saurism as a distinct tradition, of which this is the consequence and evidence, is probably to be attributed, at least in part, to a failure to continue to attract patronage and so maintain its separate identity as Śaivism became more influential and encroached upon its territory.

Note the hedges. Probably. At least in part. This is a scholar inferring a cause from an absence and telling you that is what he is doing. Keep the hedge rather than launder it. Eighty-five titles and no text is a fact. Why is an inference.

The list everyone repeats is the wrong list

Before going further it is worth clearing away a sentence that appears in almost every popular account of tantra, because it is the reason people expect a Saura agama to be sitting on a shelf somewhere.

The sentence is: the tantras are divided into five, Śaiva, Vaishnava, Shakta, Saura and Ganapatya.

That is not a classification of tantric scripture. It is the panchayatana, the five-shrine arrangement of the Smarta household altar, where five images are set out with the worshipper’s chosen deity in the centre. Gudrun Bühnemann, in her study of Smarta ritual, gives its purpose: the idea, she writes, is to unite the principal deities of five sects, Śaivas, Vaishnavas, sun worshippers, Shaktas and worshippers of Ganapati. The textbook sentence is a domestic pūjā plate misread as a library catalogue, and we will come back to the plate.

The five srotas, the five streams issuing from the five faces of Sadashiva, are a genuine classification and a genuinely tantric one. They are Siddhanta, Garuda, Vama, Bhuta and Bhairava. They are entirely internal to Śaiva scripture. There is no Saura among them, no Vaishnava and no Ganapatya. And the six sects traditionally credited to Shankaracharya, which do include the Sun, belong to the hagiographies. No pre-modern Sanskrit text stating that scheme could be found for this article, and Bühnemann herself reports it only as something that is said to have been popularised by Shankara, footnoting P. V. Kane, who is in turn reporting tradition. (A different six appears later in this article, the six sub-sects of the Sauras that Anandagiri’s Shankaravijaya describes. The two schemes are unrelated and should not be run together.)

What the sources actually give is better than the textbook version. The Shrikanthiyasamhita puts the Saura canon next to the Pancharatra canon, in a survey that comes before and outside the section on the five streams. The Sun is not a branch of the Śaiva tree. He is a separate revelation with a separate library, which is a larger claim, not a smaller one.

And a second Kashmirian scripture names him too. The Netratantra, which Sanderson places between about 700 and about 850, and which Acharya would put in the first half of the ninth century at the latest, opens its ninth chapter by listing the systems of revelation, and its list has six: Vama, Dakshina, Siddhanta, Saura, Vaishnava, Vaidika. This is not the five faces of Sadashiva under another name. It is a different scheme doing a different job, and Kshemaraja, glossing it, calls its members srotas as well and says each has its own deities and its own observance, “not mixed.” The word does more than one job.

So the Saura stream is attested by name, twice, by the people best placed to leave it out. What it never got was a stream of its own in the Śaiva map of the five faces. It got something else instead, and we will come to it.

The one book that survives, and whose side it is on

In Acta Orientalia 80, Diwakar Acharya published a study of a text he described as the earliest surviving tantra text on Sun worship available by far. It is called the Saurasamhita. It survives in one palm-leaf manuscript in the National Archives in Kathmandu, dated Nepal Samvat 69, which is 949 of the Common Era, together with four twentieth-century paper transcripts of it, two still in Kathmandu, one at the Wellcome Institute in London and one at M. S. University in Baroda. It is among the oldest tantric manuscripts anywhere in the world.

It was copied by a Buddhist householder, with a Mahayanist motive, wishing peace and prosperity for all.

Thirteen chapters, a little more than five hundred verses, a dialogue between Maheshvara Śiva and his son Skanda Kartikeya. It nicknames itself prayojanatilaka, a text marked with purpose. It argues that Aditya is called aditya because he belongs to the beginning, and that even Prajapati and Kala were made after him. It says he is so called because he carries off the darkness standing in all the quarters, by releasing his rays. It says there is neither chariot nor horses. It says that without the Sun there would be only darkness, and no distinction of dharma and adharma, and therefore the Sun witnesses and maintains everything.

It also says this, at chapter one, verse fifteen:

adityam tu shivam vindyach chivam adityam eva cha | nanatvam yas tu gachcheta yatnenapi na sidhyati ||

One should know Aditya as Śiva, and Śiva precisely as Aditya. He who goes over to the view of their difference will not attain success even with effort.

(Saurasamhita 1.15, as quoted by Sanderson; the translation is mine.)

The book’s colophon places it in the Vathula, a line of Śaiva scriptural transmission, and Acharya notes that it matches the Kalottara class of Śaiva texts in style and structure. It was obviously produced by some Śaivas, he writes, to promote identity of the Sun god with Śiva. The first half-line of its verse 1.10 is shared word for word with the Paushkaraparameshvara and with the Shrikanthiyasamhita.

There is a smaller detail that carries further than the doctrine does. In the Agamic Śaiva ritual systems, the leftovers of the worship, the nirmalya, are received by a figure called Chanda or Chandesha, who also forgives the worshipper his mistakes. The Saurasamhita has one too. It calls him Tejaschanda, which is, in Acharya’s words, a solar version of the Śaiva deity Chanda. And Tejaschanda does not stay in the Nepalese manuscript: when King Bhojadeva of Dhara writes his Sun-worship sequence into a Śaiva ritual manual in the eleventh century, the rite ends by consigning the leftovers to Tejaschanda by name. The solar Chanda travelled into the royal Śaiva paddhati along with the rite. One could read that as a Saura functionary being matched by the Śaivas rather than invented by them, and nothing in the text rules that out.

Acharya’s edition is still forthcoming. As of now, the earliest surviving tantra on Sun-worship is a book that two scholars have described and nobody can buy.

And there is one more thing inside it. The Sun’s heart-mantra in the Saurasamhita, which Acharya calls the khasholka mantra after the name of the Sun it invokes, turns on a word that already appears in the Taittiriya Aranyaka, a late Vedic text. Neither of that Aranyaka’s two great commentators recognised it as a Sun-mantra. Sayana read the word one way, Bhatta Bhaskara another, and each forced an abstract sense onto it, claiming it was dedicated to the supreme lord. Bhatta Bhaskara is the earlier of the two, so by his time the reading was already gone. Acharya’s suggestion is that the Saurasamhita had superimposed a separate archaic and popular mantra of the Sun onto a pro-Śaiva ritual system it created or inherited.

Something older is inside the Śaiva wrapper, and by the late medieval period the two men best placed to read it could not.

Three displacements, on three different clocks

What follows is the case for saying the tradition was displaced rather than defeated. It comes in three parts, and the first thing to say about them is that they are not three stages. They ran on different clocks, one of them starts earlier than anything called Saura, and only the first is a sequence in which something is taken away.

One: the liturgy is annexed

From the early eleventh century, the Sun’s worship becomes a compulsory limb of somebody else’s.

The evidence is in the ritual manuals. King Bhojadeva of Dhara, who ruled around 1018 to 1060 and whose Siddhantasarapaddhati is one of the great codifications of Saiddhantika practice, includes a Sun-worship sequence as a required preliminary to the regular worship of Śiva. The Somashambhupaddhati, composed towards the end of the same century in dependence on Bhoja’s text, carries it too. Acharya reaches the same date from the other side: from the early eleventh century onwards, he writes, all Siddhanta Śaivas gradually adopted sun-worship as found in the Saurasamhita as a compulsory element of the regular worship of Śiva.

In the Saiddhantika manuals, that is what the rite becomes. Kept entire, and no longer its own end.

There is a Śaiva verse often quoted alongside this, and it is worth quoting carefully because it says less than it appears to. The Brihatkalottara, at folio 108v of the manuscript Sanderson cites:

bauddhavaishnavapancharthe saurakalamukhadishu | shaivah sarvadhikari syan na shaive ‘mi kathamchana ||

The Śaiva Guru has competence that extends into all religious systems, the Buddhist, Vaiṣṇava, Pañcārtha-Pāśupata, Saura, Kālamukha, and others; but Gurus of those have absolutely no competence to act in the Śaiva system.

(Sanderson’s translation, with his square brackets removed for readability.)

The verse is not about the Sun. It is the standard Śaiva claim of universal ritual competence, made about five named systems and an “and others,” and Saura is a list item in it. What it establishes is that the tradition doing the annexing had a doctrine of jurisdiction ready to hand, and wrote it one way only. It does not establish that Saura was annexed. The Śaiva could work anywhere. Nobody could work in his house. If you want the shape of the Śaiva canon that made such a claim thinkable, it is on the site already.

Two: the Sun was always also a planet

The second displacement is the one usually told as a sequence, and the sources will not allow it.

The simplest proof is a single book. The verse listing the priesthoods of India, the Magas among them, is Varahamihira’s. So is the Brihajjataka, one of the founding textbooks of Indian genethlialogy, and the Brihajjataka opens by invoking the Sun as the lamp of the three worlds, the Self to those who know the Self, the sacrifice of those who sacrifice, the lord of the dissolution and arising and maintenance of the worlds. Then it turns the page and begins its chapter on the planets: the Sun is the soul of Time, and first of the nine. Around the year 550 the two Suns are not a before and an after. They are consecutive chapters of one manual by one author.

The Yajnavalkya Smriti shows what the second of them looks like as a rite. At verses 1.294 to 1.308, in the numbering of the Mitakshara vulgate, it sets out the grahayajna, the rite of the planets, for anyone who desires prosperity, peace, rain, long life or increase, and, it adds in the same breath, for anyone working sorcery. Then it names them, in order: suryah somo mahiputrah somaputro brihaspatih, shukrah shanaishcharo rahuh ketush cheti grahah smritah.

Surya is item one of nine. Not the god above the nine. The first of them. The text assigns each of the nine a substance, and the Sun’s is copper. And it lays down the operative principle at 1.307: whichever planet is ill-placed for a man, that one he should worship with effort. Worship indexed to affliction.

How old that passage is has been disputed. Bühnemann says flatly that the chapter carrying the Vinayaka and graha worship is not an original part of the work. Kane had already answered the older form of that position sixty years earlier, arguing from the discovery of the Manavagrihya that the Vinayaka section is no proof of lateness, though his argument bears on the Vinayaka verses rather than on the planetary ones. The Smriti as a whole is usually placed in the third to fifth centuries. The argument above does not depend on which way this falls. Varahamihira alone establishes the coexistence.

What the Smriti adds is the shape of the thing. The planetary Sun does not arrive to replace the denominational Sun. It runs alongside it, and it is the frame that outlived the other one.

Which makes the choice of mantra the interesting thing. Look at which Vedic verse the Yajnavalkya Smriti assigns to the Sun as his planetary formula. It gives the opening word only, in the usual manner: akrishnena. That is Rigveda 1.35.2, from Hiranyastupa Angirasa’s hymn to Savitr, the god moving through the dark space in his golden chariot, beholding the worlds.

The Vedic hymn to the sovereign of the sky, doing duty as the propitiation formula for planet number one.

The visual record fills in the rest. Stephen Markel’s study of the planetary deities dates the first known representations of them to Madhya Pradesh around 500, and the addition of Ketu to the eight, completing the group of nine in art, to the early seventh century. The textual roster is older: Varahamihira’s own list of the planets already runs to nine, Rahu and Ketu included. What Markel’s dating gives is the moment the navagraha becomes a thing you can carve, and Surya is the figure the set assembles around: he is, Markel writes, by far the most important and oldest of the deities, with a much greater body of legend attached to him than to any of the others and the most abundant representations.

There is a last turn, and Varahamihira reports it himself. At Brihatsamhita 3.1 to 3.2 he observes that the old treatises put the Sun’s southern turning at the middle of Ashlesha and the northern at the beginning of Dhanishtha, and that this must certainly have been so at some time, since the earlier books say so. But at present, he writes, the turning is from the beginning of Karkataka. And then the phrase that makes him a scientist: pratyakshaparikshanair vyaktih. It is made evident by direct observation. He gives the method too, a distant marker sighted at sunrise and sunset, or the entry and exit of a shadow across a large circle.

What he had measured, without naming it, is the effect we now call precession. He offered no mechanism for it in this passage, and where he did theorise, in the Panchasiddhantika, David Pingree finds him explaining the change by a theory of trepidation, an oscillation which is precession’s rival and which is wrong. Pingree’s own conclusion is that Indian astronomy before the twelfth century ran three systems side by side, two of them versions of trepidation and one simple precession, and that the reasons for choosing among them can by and large no longer be reconstructed.

The sidereal zodiac was kept, and the drift was allowed to stand. Makara Sankranti is still reckoned by the Sun’s entry into the sidereal sign, which now falls around the fourteenth of January, while the solstice it commemorates falls around the twenty-first of December. The three weeks between them are the same phenomenon Varahamihira described, still running, and never taken out of the reckoning. The two roads of the Sun, the uttarayana and the dakshinayana, are set out on the site in the Sun’s two roads; what jyotisha made of them is this.

The Sun’s institutional home did not stop being a temple and become a computation. It was always both. What changed is which of the two had a future.

Three: the seat that was never a survival

The third item is the one usually described as survival, and the honest thing to say is that it is neither survival nor dispossession. It is a popular claim that does not stand up, and the reason to include it is that people will bring it.

On the Smarta household altar, five deities are set out, and Surya is one of the five. That much is true, and Bühnemann gives the arrangement’s purpose as the uniting of the five sects. But three things complicate the tidy story.

The five were never fixed. In earlier sources, Bühnemann notes, Brahma appears instead of Durga or Ganesha; the seventeenth-century Pujaprakasha of Mitramishra is organised around Vishnu, Śiva, Surya, Durga and Brahma, with no Ganesha at all. The canonical modern pentad is a late settlement, and across its versions Surya’s seat is steadier than Ganesha’s.

The Sun could hold the centre, and did. The Nityacharapradipa of Narasimha Vajapeyin, composed after 1400 and before 1565, arranges a surya-panchayatana with Surya named first, Śiva, Brahma, Vishnu and Durga around him, and again no Ganesha. Bühnemann notes that the text was popular in Orissa. A Sun-centred altar, in the Konark region, in the very century the cult is supposed to be finished.

And this is subordination rather than digestion. Bühnemann is explicit: the four deities worshipped along with the central deity are not considered to be manifestations of the latter, even though their position at the time of worship is clearly subordinate. The panchayatana demotes the Sun. It does not dissolve him.

What it does not do at all is explain how anything survived. There is no initiation on that altar, no canon, no lineage and no separate goal, and the arrangement is far too late to be a mechanism for anything. The five-shrine building plan is sixth-century architecture. The verse governing the pūjā is attributed, by a later digest, to Bopadeva at the Yadava court of Devagiri in the thirteenth century, and it circulates through the ritual compendia of Mitramishra, Kamalakarabhatta and Nilakantha in the first half of the seventeenth, three men who were near-contemporaries and, in the case of the last two, first cousins. Anyone who writes that Shankaracharya instituted the panchayatana and that is how the Sun survived has compressed an eighth-century philosopher, a sixth-century floor plan and a seventeenth-century rule book into a single sentence.

One thing should be said plainly here, because it is a join this article makes and its main source does not. The word panchayatana does not occur anywhere in The Saiva Age. Connecting the arrangement to the fate of the Saura tradition, even to dismiss it, is my inference and not Sanderson’s.

The tests

A story this tidy should be attacked, so here are three places it could have come apart, with what would have broken it stated before the answer.

Kashmir. Kashmir has by far the richest surviving tantric literature in India, and it has Martand, where, in Sanderson’s words, Lalitaditya built his majestic temple of the Sun in the mid-eighth century. A Kashmirian Saura text with a Kashmirian tantric commentary would complicate this article badly.

There is one, and it is worth being exact about what it is.

The Sambapanchashika is fifty verses in praise of Surya. It is a stotra, a hymn, and not a scripture: no revelation frame, no initiation, no ritual system. It is probably of the eighth century, composed in Kashmir in the century of Martand, and it comes down to us with a commentary by Kshemaraja, the same Kashmirian Śaiva who glossed the Netratantra. Martina Palladino, whose fieldwork appears later in this article, describes it as a text that perfectly represents the tantric Saura tradition which later disappeared. It has been in print with Kshemaraja’s commentary since 1889, and Bettina Baumer has written a study of it.

So Kashmir does not give silence. What it gives is a hymn, which is a weaker survivor than a scripture, and that is itself the finding.

And then look at what the hymn contains. Śiva is not named in it. Bäumer calls it a purely Saura hymn, and only Vishnu is mentioned, twice, though Palladino notes on the same page that the hymn’s three forms of Surya include a manifestation of Śiva in his eight forms, so the separation is not quite absolute. What is not in doubt is where the identification becomes explicit. It is Kshemaraja’s commentary that performs it, reading the Sun’s light as the Light of Consciousness. That is the annexation happening on the page: a Saura hymn on one side and a Śaiva commentator absorbing it on the other, and the hymn reaches us because he did.

The rest of the Kashmirian solar material is Śaiva through and through. There is a Śākta cult of the Sun as Vira or Vireshvara with the goddess Bhargashikha, known only from citations in Abhinavagupta, Jayaratha and Kshemaraja, and, proving it was not merely local, in the Sekanirdeshapanjika of the east Indian Buddhist Ramapala, who calls the Bhargashikha a Shakta tantra. Sanderson, who states a conclusion plainly when the evidence allows one, declines to say whether any of it preserves Saura material: in the absence of properly Saura literature, he writes, it is impossible to be sure. The philosophical side of that world is on the site at Kashmir Śaivism.

The inscriptions. A Saura acharya named in an inscription, with a monastery or an initiation lineage attached, would end this article.

Sanderson’s method throughout The Saiva Age is to trace Śaiva Siddhanta teachers, their monasteries, their initiation lineages and their royal appointments through the epigraphic record, region by region and century by century. He has a machine built to find an organised initiatory tradition in stone. Turned on the Sun, his survey finds kings and nothing else: no Saura acharya, no Saura monastery, no Saura initiation lineage, and no Saura scripture quoted in an Indian inscription.

That is a statement about what one very good survey found, not about every stone in India, and it needs a qualification that comes from elsewhere. Saura books went on being named for centuries in the digest literature. R. C. Hazra traced the lost Suryapurana, of which no manuscript has come down, through a verse quoted from it in Chandeshvara’s Tithinirnaya and a Berlin manuscript of the Bhavishyottara naming it as a source; and he found that the Saura work declared by Narada has become extinct altogether, leaving only a line of the Bhavishya Purana mentioning it, itself quoted in the Kritya-kalpataru, the Kritya-ratnakara and the Nityacharapradipa. Aufrecht’s Catalogus Catalogorum records a Sauratantra cited by Kshirasvamin, commenting on the Amarakosha, probably in the eleventh century. The books were being named. They were not being copied.

Bhaskararaya. The last test is not an absence. It is a choice, and it is the sharpest thing in the article.

Bhaskararaya Makhin is the pre-eminent Shrividya commentator of the eighteenth century, the author of the Setubandha and of the standard commentary on the Lalita Sahasranama. In 1708, twenty years before that commentary, he wrote a small book called the Trichabhaskara, on a tricha, the group of three Rigvedic verses at 1.50.11 to 13 addressed to the Sun. He argues the Sun’s supremacy at length from the Bhagavata, the Brahma Purana, the Kurma Purana and the Bhavishyottara. He cites the Shankaravijaya, in the recension ascribed to Anandagiri, on six kinds of Sauras. He is, on this evidence, a genuine devotee of the Sun.

And then he sets the tantras aside on purpose, and he does it knowing exactly what he is setting aside. Saura rites are found in the tantras in many forms, he writes, and he names them, the Bhuvaneshi, the Chaturakshara, the Prayojana, the Tilaka. Nevertheless, since procedures using the tricha are supreme by reason of being Vedic, those alone are expounded here.

The Baroda edition of the Trichabhaskara has not been consulted directly for this article, and the quotations above come to me at second hand; the six Saura sub-sects, at least, are independently attested from Anandagiri by J. N. Banerjea.

By 1708, for the man in the best position in India to use a Saura tantra, the living solar tradition was Vedic and the tantric material was a list of names he could recite and chose not to use. That is the argument of Agama and Veda running in an unexpected direction.

What it looks like when the books do survive

There is a comparison worth making here, and it needs to be made carefully, because two cases prove nothing.

Marichi is a Buddhist goddess of light and concealment, and she is assimilated to the Sun’s iconography in a specific and deliberate way. Benoytosh Bhattacharyya sets out the contrast: like the Hindu Sun-god she has a chariot, but hers is drawn by seven pigs where the Sun’s is drawn by seven horses, and where the Sun’s charioteer is the legless Aruna, hers is a legless goddess or Rahu, a head without a body. The Tibetan lamas invoke her at the advent of the morning.

Her canon did not disappear. It is in the Tibetan Kangyur, in the Action Tantra section, catalogued at Toh 564, 565 and 566, all three now translated into English. Her Sanskrit sadhanas are in the Sadhanamala. In Japan she became Marishiten, and David Hall’s study of her cult is subtitled its impact on the Japanese warrior.

It would be neat to say that the difference is carriage rather than doctrine, and the neatness is exactly the problem. Marichi’s texts survived because a state that had adopted Buddhism translated its tantric corpus wholesale into Tibetan and Chinese, which is doctrine and carriage at once, inseparable. And the counter-case is already on this page: the Shrikanthiyasamhita’s hundred and sixteen Pancharatra samhitas never entered a Kangyur, and a great many of them survive.

So the pair does not prove a rule. What it isolates is the question worth asking: what kind of institution has to exist for a body of books to be recopied for a thousand years. Marichi had one. The Pancharatra, with a hundred and sixteen titles and no Kangyur, had another. The Sauras, so far as anything shows, had neither, and Sanderson’s answer to why is the one on the page above: they stopped attracting patronage.

What is left

A great deal, if you count honestly, and almost none of it is Saura.

The most widely prescribed daily verse in brahmanical Hinduism is a solar verse. Rigveda 3.62.10 is addressed to Savitr, and Manu 2.78 and 2.101 fix it in the twice-daily sandhya, to be murmured standing at the morning juncture until the sun appears. It survived the extinction of the tradition that made the Sun supreme for a simple reason: it was never Saura property. It was Vedic, and Vedic is what the brahmanical mainstream kept. Its own story is told in the Gayatri, and the wider family of solar deities it belongs to in the Vedic gods.

The Sun kept a chapter in other people’s mantra-manuals. The big printed Shakta compendia all have one: the Prapanchasara at chapter fifteen, the Sharadatilaka at chapter fourteen, Mahidhara’s Mantramahodadhi at chapter fifteen, Krishnananda’s Tantrasara in its second pariccheda. Gudrun Bühnemann has catalogued the first three of those pantheons in detail. The architecture in them is the one the Saurasamhita also uses, with the Sun in the middle of a lotus and nine powers around him ending in Sarvatomukhi.

It would be convenient to call that a trace of the lost Saura liturgy, and it will not bear the weight. The same nine powers, ending in the same Sarvatomukhi, are already in the Agni Purana, which is a far more widely copied book than any Saura tantra ever was, and so are the seat and the Sun’s khasholka heart-mantra. What the correspondence shows is not descent. It shows that Sun-worship had a standard shape that everybody used, which is a duller finding and a truer one, and which cuts against the idea that the Saura tradition had a distinctive liturgy to lose in the first place.

One detail may connect. The Prapanchasara’s Sun chapter turns on a four-syllable mantra, a caturakshara, and caturakshara is one of the four tantric Saura systems Bhaskararaya names in 1708 and declines to use. The word is a generic descriptor, so the identification stays unproved. The coincidence is still worth recording.

Chhath is thriving, in Bihar and eastern Uttar Pradesh and the Nepal Tarai, and its presiding figure is Chhathi Maiya, whom scholarship generally identifies with Shashthi, the sixth-day goddess of childbirth and child protection, though Rana P. B. Singh’s ethnogeography of the festival calls her simply the sun goddess. The festival is named for the sixth lunar day. The texts usually cited for its antiquity, the Rigvedic sun hymns and the Gayatri, say nothing of a four-day river vow with offerings at sunset and sunrise, and no passage prescribing or describing the festival has been identified in the scholarship. Its documented history runs back to nineteenth-century district reports. Alexander Cunningham recorded two sun festivals held on the sixth day of the bright fortnight of Chaitra and Karttika at Deo-Baranark in Bihar, a village whose Sun-temple endowment is the subject of an undated Later Gupta charter; the note reaches me through Fleet’s edition of that charter rather than from Cunningham’s report directly. Beyond the 1880s the trail does not run with any security.

What it is, on the evidence, is a women-centred household goddess-vow whose ritual address is the sun, which is precisely why it outlived temple Sauraism: it was never Saura religion in the first place. Vow traditions of that kind are exactly what survives when sectarian institutions collapse. The Vedic claim is unsupported and it is unnecessary. Nothing about the festival’s standing depends on it, and nothing in the paragraph above touches its authenticity or its meaning to the people who keep it.

And the priests are still there. The Shakadvipiya or Maga brahmins persist in Bihar, Bengal, Rajasthan and Uttar Pradesh, and they still keep a solar practice: Martina Palladino’s fieldwork records daily surya-puja at dawn and sunset with the water-offering as its core act, matching the Puranic prescription, and Ratha Saptami as the great festival of the Rajasthan families.

The historical record shows something more complicated than continuity. In 1137 a Maga poet named Gangadhara had it cut into stone at Govindpur in Bihar that his people came from beyond the milk-ocean and that Samba himself brought them there. Read the rest of that inscription and its opening invocation is to Vishvambhara, Vishnu; its object is the building of a tank; and one of its verses praises the poet’s own father, Manoratha, as a man whose mind was purified by the most sacred texts of the Śiva doctrine. Another records that when Manoratha and his wife were childless, Śiva came to them in a dream and told them to worship him. The solar origin is carried as a charter of descent. The family’s religious affiliations, as it chooses to record them, are elsewhere.

Palladino, interviewing Shakadvipiya families at Bikaner in November 2016, asked about the avyanga, the girdle the Puranic texts prescribe for the Sun’s devotees and which scholars have long identified with the Zoroastrian kusti. It is conferred in a child’s eighth year, and privately: a hidden samskara, one of her informants said, only between the kid and the sun. It is also nearly gone. This was only in the ancient times, he told her. Now very few people are worried about that, even the Shakdvipiyas today are not worried about that. When she named the Avestan cognate, neither informant had heard of it; they asked her to repeat the word, and agreed it was possible the two were related.

The connection now reaches the community from outside it, and the community is not troubled by that.

As for Konark, the largest Sun temple ever attempted in India: when the Archaeological Survey cleared the sanctum of debris, there was no presiding deity in it. A wooden image that perished, a disc, an image removed to Puri, or a shrine never consecrated at all. In the Survey’s own words, in the absence of reliable data, speculation has been rife. There is no chariot festival at Konark. The Chandrabhaga Mela is held each year three kilometres away at the beach that gives it its name, which is pilgrimage to a place rather than the daily worship of an installed deity. The chariot is the building, and it is stopped.

Java, and then Bali

Sanderson’s own verdict on where the tradition ended is not in India at all.

Only in the Majapahit kingdom of East Java do we hear of the survival of adherents of a distinct Saura denomination. There a royal charter of c. 1350 tells us that a board of six learned men appointed to adjudicate law suits included two adherents of this tradition.

Their names are Marmanatha and Jayasmara, and the charter calls them men of the sorapaksha, the Saura sect. The last Sauras on record are two judges on a Javanese legal board around 1350.

And a little further east, the thing itself did not stop.

The daily liturgy of the Balinese padanda, the Śaiva priest, is called Surya-sevana. When Sylvain Levi published the Balinese Vedaparikrama in 1933, the evidently tantric character of the whole ritual was, in the phrase Acharya quotes from him, something he did not fail to perceive. Then he read through every tantra he could reach, looking for a Sanskrit Indian original for a twilight worship that combines Surya and Śiva in one deity, Parama Adityashiva or Parama Shivaditya. He could not find one.

Diwakar Acharya found it. A set of verses used in Bali as a Suryastava, a praise of the Sun, appears directly taken from the second chapter of the Saurasamhita, and he prints the two texts one under the other with the divergent syllables underlined. He calls it a set of four; the passage he sets out runs to about six.

His argument for the direction of travel rests on a single word, and it is worth knowing how thin and how good the evidence is. The Saurasamhita is Śiva speaking to his son, so its verses address Skanda by name, in the vocative. In a Balinese hymn of praise to the Sun there is no Skanda and no dialogue, and the vocative is simply redundant. It is still there, in one Balinese manuscript. The others read kandah, and Levi read kenda, and Acharya argues for skanda partly on the strength of the parallel it then explains. The reading is his, and he says so in a footnote, though his main text puts it more strongly than that.

But if he is right, a Balinese priest at dawn is calling Skanda by name, in a hymn that has no Skanda in it, because the vocative came across the sea inside the verse and nobody took it out.

The cult of Śiva-Surya, Acharya writes, has lost its independent existence in South Asia, whereas in Bali it has not only maintained its independence but also taken central stage.

Eighty-five books were written and eighty-five books were lost. What is left of them in India is a chapter at the front of somebody else’s liturgy, a Sun chapter in somebody else’s mantra-manual, a copper disc first among nine, a corner of a household altar, a verse that was Vedic before it was ever solar, and one wheel of a stone chariot on the ten-rupee note.

What is left of them in Bali is a priest, at dawn, saying the words.