The Golagi matha: the monastery that survives only in its own praise

At Bheraghat on the Narmada, below Jabalpur, there is a temple with no roof. It is circular, open to the sky, and ringed on the inside with images of yoginis. In 1950, in the Indian Historical Quarterly, V. V. Mirashi used that building to find a monastery.

His paragraph is five sentences long and it is worth having whole.

“The hypethral temple became known as Golaki or the Round Temple from its shape. The matha or monastery established by its side became well known as Golaki matha. The Malkapuram pillar inscription says that the Golaki matha was situated in the Dahala mandala between the Bhagirathi and the Narmada. Dahala was the name of the home province of the Kalacuris with Tripuri, modern Tewar 6 miles from Jubbulpur, as its capital. Golaki matha was thus plainly identical with the matha at Bhera Ghat.”

The third and fourth sentences are evidence, and good evidence. The Malkapuram pillar does place the monastery in Dahala, between those two rivers. Dahala was the Kalachuri home province, and its capital was near Jabalpur. Nobody has answered that.

The first two sentences are the trouble, and they are the ones doing the precision work. Dahala is a province several hundred kilometres across. Bheraghat is a spot on a river bank. What carries the reader from the one to the other is a name: the temple is called round because it is round, the monastery beside it is therefore the Golaki matha, and the Golaki matha is therefore here. Nothing outside the etymology connects that particular building to that particular institution. And that is the foundation of the most widely repeated fact about this monastery.

The Golagi matha is in the school books. Three hundred thousand villages. A hospital. A lying-in house for women in labour. A kitchen where nobody was turned away, whatever their caste.

There is no archive. No accounts, no registers, no letters, no chronicle, nothing an auditor would recognise. There is one class of source and one only: prashastis, praise-inscriptions, commissioned by the institution and by the kings who endowed it, composed in Sanskrit verse by named poets and cut into rock by named engravers.

An institution whose entire surviving archive is its own publicity.

What follows is an attempt to say what can be known from that. It is not a debunking. The best of the famous claims survives the examination, and it comes out stranger than the version in circulation.

An institution with no address

The Bheraghat identification did not begin with Mirashi. It begins with a sentence in the 1930 edition of the Malkapuram pillar inscription, where J. Ramayya Pantulu, having explained that Dahala was another name for the Chedi country whose capital was Tripuri near Jabalpur, adds: “it is said that in this city there used to be a maṭha called the Gōlakimaṭha, from the name on which it was situated.” He gives no source for the report, and the city he means is Tripuri. The narrowing to Bheraghat, a few miles further down the Narmada, comes later. Hiralal had made the connection in 1927; L. N. Swamy, whose forty-three-page study of 1975 is the only work devoted wholly to the monastery, follows Hiralal rather than arguing it afresh. Mirashi supplied the argument, and the argument is half evidence and half circle. The half that is evidence gets you to a province. The half that is circle gets you to a river bank.

Against this stands a reading of the name itself. In a footnote in The Saiva Age, published in 2009, Alexis Sanderson points out that the earliest spellings are not Golaki at all. His witnesses are three. Nepalese palm-leaf manuscripts of Somasambhu’s ritual manual give Golagi. The Bangarh prashasti, cut in the time of the Pala king Nayapala, gives golagyas sa mahamathah. And the Chandrehe stone, he says, gives golaggikah, where Mirashi and Banerji both read golagnikah. So two of the three are inscriptions and one is a manuscript tradition, and only the third is a re-reading, contested at the level of a single consonant. The k forms, on his account, are a southern scribal substitution by scribes whose own languages do not distinguish the two sounds, which is why they dominate the Andhra records and the modern literature. The place, on his reading, is Gurgi, about twelve miles due east of Rewa, the site of once vast Shaiva ruins. He calls the Bheraghat view groundless, and says that his full reasons must be set out elsewhere.

It never was.

A full demonstration was promised and, so far as can be traced, never published. A full-text search of Sanderson’s subsequent work, including “The Saiva Literature” and the 2020 Festschrift volume, turns up no further discussion: no Golag-, no Golak-, no Gurgi, no Mattamay-, no Bhera-.

It has, however, been taken up by someone else who works on these texts. Dominic Goodall, writing in Cracow Indological Studies in 2014, places the monastery in the Rewa District of Madhya Pradesh and adds that “as Sanderson has demonstrated, the name of the place in Somasambhu’s time seems rather to have been Golagi”. He then uses the spelling in his own voice, twice, later in the same article. Rewa District is Gurgi’s district. It is not Jabalpur’s.

Goodall is not wholly consistent, and it is fairer to say so. On the page before, he calls the same man abbot of the Golakimatha near Tripuri, which is the Jabalpur side. What he endorses without hesitation is the name. On the place he wobbles, in one article, in the space of a page.

One more thing, because it shows how a bad citation travels. English Wikipedia calls the monastery famed for its round temple shape and puts it, with a “probably”, near modern Jabalpur. Two references are attached. The first is Hartmut Scharfe, Education in Ancient India, page 183. Go to page 183. There is no “round” on it. No “shape”, no “circular”, no “yogini”. What is on it is a bare parenthesis, “(perhaps near modern Jabalpur in Madhya Pradesh)”, with no argument and no source. The second reference is Cynthia Talbot’s Precolonial India in Practice, pages 89 to 90 and 131 to 134, which treats the Golaki matha and its preceptors at the Kakatiya court. A full-text search of that book, run for this article, returns no occurrence of “Jabalpur” anywhere in it, and nothing about a round temple.

The round-temple etymology is Mirashi’s, from 1950. So far as either cited book can be searched, neither of them carries it.

And Mirashi’s real premise does not go away because half his paragraph is circular. D. C. Sircar, editing the same record in 1971, gives it independently: the pillar accounts for a Shaiva monastery of the Dahala-mandala, which he glosses as about the Jabalpur District, situated between the Bhagirathi and the Narmada. A province and two rivers a long way apart. Gurgi lies between them too.

So a circular argument for a location does not move a monastery. It means only that the location has never been narrowed properly, which is a different and emptier thing. Bheraghat has one real reason behind it and one bad one, and the bad one is the famous one.

And the g spelling is older than 2009. K. K. Handiqui was already writing Golagi in 1949, with a rival etymology attached: Golaki or Golagi, he reports, is said to be a contraction of Golagiri. If that is right the name is a hill, not a shape, and the round temple never came into it.

We do not know where the Golagi matha stood. That is the honest position, and it is the position this article keeps.

Four stones and a pillar

The reconstruction rests on five records, and they are very unequal.

The Ranod inscription, in northern Madhya Pradesh, was edited by Franz Kielhorn in the first volume of Epigraphia Indica in 1892. Sixty-four lines, sixty-nine verses, written across a surface six feet eleven inches high and two feet ten and a half inches broad, in letters between three quarters and seven eighths of an inch. Alexander Cunningham, who had seen it before Kielhorn, measured the whole slab, on the following page of his report, at seven feet four inches by three feet two inches; the two figures are not in conflict, because Cunningham measured the panel and Kielhorn the written surface. It carries no date. Kielhorn’s judgement, on the shape of the letters alone, has been the anchor for everything since: “The inscription is not dated, but it may on palæographical grounds be assigned to the end of the 10th or the beginning of the 11th century A.D.”

Chandrehe is the only one of the five whose date is cut in numerals. Twenty-seven lines across two slabs, twenty verses, recording the work of the ascetic Prabodhasiva. The year is given as Samvat 724, Phalguna sudi 5, the fifth day of the bright fortnight, and Mirashi observes that the numerals stand alone, with no era named. He supplies the Kalachuri era on palaeographic grounds, which is why the title of his own edition carries “(Kalachuri)” in a parenthesis. His conversion is unusually candid: “It corresponds, for the expired Kalachuri year 724, to Monday, the 10th February, 973 A.C. It does not admit of verification.”

Bilhari, a long dynastic prashasti of the Kalachuris of Dahala, is the operating budget. Thirty-three lines across an inscribed surface six feet three inches broad and three feet high. Kielhorn, who edited it in the same volume of Epigraphia Indica, reported that the inscription “may possibly have contained a date at the end of the last line; but, if such was the case, it has become entirely effaced”. Mirashi dates it twice in one volume, at about 975 on page 208 and at about 980 to 990 on page clii, the second figure matching his own dynastic table. The discrepancy is his, a decade wide, and it is unresolved.

Gurgi, undated, forty-nine lines and forty-six verses on a slab about four feet one inch by three feet four inches, was found among the ruins of the Gurgaj temple at Gurgi. It records a large land grant by the Kalachuri king Keyuravarsha, that is Yuvarajadeva I, to a great ascetic described as having completely mastered the Shaiva Siddhantas. The ascetic is not named in that portion of the record.

And Malkapuram, in the Guntur district of Andhra Pradesh, is a pillar rather than a slab: fourteen feet seven inches tall, inscribed on three sides, a hundred and eighty-two lines of Sanskrit followed by eighteen of Telugu prose. It is dated by a chronogram to Saka 1183, the cyclic year Durmati, which is 1261 of the Common Era. It is the constitution of a daughter house, and it is the richest of the five.

Three of the five carry no date at all. A fourth carries one its own editor will not vouch for. Only the pillar is securely dated, and the pillar was cut three hundred years later and a thousand kilometres away.

The stone its editor found worthless

Kielhorn published no translation of the Ranod inscription. He published an abstract, and he explained why in a sentence that has aged into an irony: “Considering that the language is generally plain and easy to understand, and that the inscription offers very little that can be of value to the historian, it will be sufficient to give the following short abstract of the contents :—”

A hundred and thirty years later, that record is the spine of the entire reconstruction of the Mattamayura lineage. It is where the line of teachers is set out. It is where a king’s initiation is described. The founding of the first two monasteries is recorded there too. Every survey of Shaiva monasticism in central India is downstream of it.

And because Kielhorn gave an abstract instead of a translation, the abstract is what everyone reads, and what the abstract leaves out has stayed out ever since. One example is the name of the place where the lineage says it began.

The stone also tells you who made it. Line 64 records that it was written on the stone by Rudra and engraved by Jejjâka. Verses sixty-eight and sixty-nine, which fall outside the abstract, name Devadatta as the composer of the eulogy and his son Haradatta as the one who delineated or copied it out. Four men, named, for one wall.

At the other end of the record, in verses sixty-five to sixty-seven, the register changes completely. Kielhorn summarises: “Verse 65 expresses the wish that the tank may last for ever ; and v. 66 forbids the planting of trees near it. And, finally, v. 67 forbids the ascetics to sleep on beds and to allow women to pass the night within the Maṭha.”

That is not eulogy. That is house rules, cut into stone, in the left end wall of the lower verandah of a building that is still standing. Cunningham described that building in the 1860s: “The main building is 48 feet long, 34 feet broad, and 21 feet high. The two storeys, which are exactly similar in their arrangement, consist of a long front verandah supported on four stout pillars, with a suite of three narrow rooms in one line at the back, which are dimly lighted by small stone-barred windows.”

Chandrehe says what they ate. Prasantasiva lived on fruit, on lotus-stalks and on roots. Prabodhasiva took priyala and amalaka fruit, greens and saluka, tended the fires, and, on Mirashi’s reading of verses ten and eleven, kept clear of the company of women all his life. Ranod’s rule and Chandrehe’s praise are two different stones, cut for two different patrons, describing one life.

Three narrow rooms, stone-barred windows, no beds. You can still walk into it.

The god in the deodar forest

Here is what the abstract dropped.

The sixth verse of the Ranod inscription is the charter myth of the whole lineage, and it names the place where the lineage begins. Kielhorn’s English summary passes over the place-name in silence, and the abstract is what most readers of this inscription have ever met. Mirashi found it, because he read the Sanskrit: his introduction to the corpus edition says that Shiva first initiated Brahma into this faith after a sacrifice in Daruvana. But finding it required going behind the abstract, which is precisely what an abstract exists to spare you. It is on the stone, printed on page 355 of Epigraphia Indica I, in a metrically sound Upajati stanza.

भक्त्योपसन्नं स्थिरभक्तिगम्यः पद्मासनं दारुवनेतिरम्ये । विधाय यज्ञं विधिना विधिज्ञः किलानुजग्राह पुरा पुरारिः ॥ ६

bhaktyopasannaṃ sthirabhaktigamyaḥ padmāsanaṃ dāruvane ‘tiramye | vidhāya yajñaṃ vidhinā vidhijñaḥ kilānujagrāha purā purāriḥ

Purari, knower of the rite, attainable through steadfast devotion, having performed a sacrifice according to rule in the exceedingly delightful Daruvana, formerly showed favour to the Lotus-seated one, who had approached him with devotion.

Ranod inscription, verse 6 · Epigraphia Indica I p. 355 #

Daruvana. The forest of deodars. Purari, the enemy of the three cities, is Shiva; the Lotus-seated one is Brahma; anujagrāha is “showed favour”.

Anyone who knows the Puranic Daruvana will feel the floor move. In the Puranic tellings Shiva comes into the pine forest as a naked beggar, the sages curse him, and the episode ends with the promulgation of the Pasupata vow. The Brahmanda Purana carries that ending. It is a Pasupata charter, and it is one of the most uncomfortable episodes in the corpus.

In the Ranod verse the scandal is simply gone. There is no curse and no nakedness. There is a correctly performed yajna, a fire sacrifice made according to rule, a devotee approaching in devotion, and a god who responds with favour. Out of that favour comes a line of teachers.

For an order whose business was the initiation of kings, the edit is the point. A myth that ends in a curse cannot be handed to a royal patron. A myth that ends in a sacrifice and a parampara, an unbroken line of teachers, can.

Two things get attached to this verse that are not in it. The first is initiation. Mirashi writes that Shiva “initiated” Brahma into this faith, which reads anujagrāha through the categories of the Shaiva Siddhanta, the dualist, ritual-centred school these houses belonged to. The stone says only that he showed favour. The second is Durvasas, the sage through whom several of these lineages later claim descent. He is not at Ranod. He surfaces three hundred years later and a thousand kilometres south, on the Malkapuram pillar, and again in the Dhureti plates, number 72 in Mirashi’s corpus, where an abbot is called an incarnation of Durvasas in the Kali age.

A lineage acquires its ancestors late.

A chain of names, and what is not in it

Verses eight to ten of the Ranod stone give the descent, and Kielhorn’s own spellings are worth keeping, because they carry information that a modernised transcription throws away.

”…Kadambaguhâdhivâsin, ‘the inhabitant of Kadambaguhâ’; from him sprang the sage Śaṅkhamaṭhikâdhipati, ‘the superintendent of Śaṅkhamaṭhikâ’; next came Terambipâla, ‘the protector of Terambi’; then Âmardakatîrthanâtha, ‘the lord of Âmardakatîrtha’; and after him Purandara (vv. 8-10).”

Look at what those four are. The dweller in Kadamba cave. The superintendent of Shankhamathika. The protector of Terambi. The lord of Amardaka-tirtha. They are not personal names. They are offices, each tied to a place. The first individual with a name of his own is the fifth, Purandara, and he is the one the record actually cares about.

The transaction follows in verses eleven to fifteen. A king called Avantivarman wanted initiation. He travelled to Upendrapura, brought Purandara back with him, received diksha, formal initiation, and paid for it with what the Sanskrit calls his nijarajyasara, which Sanderson renders as “the best part of the wealth of his kingdom”. Purandara then founded a matha at Mattamayura and a second at Aranipadra.

That is the founding charter of the network. A king buys initiation; the initiator builds houses with the proceeds.

Now the correction that catches most secondary accounts. The Ranod inscription grants no villages. It grants no wells, no gardens, no rampart. The wells, the gardens and the rampart belong to a different record altogether, the Gwalior Museum inscription of Patangasambhu, which Mirashi edited separately in 1962. Ranod records a tank, a rule about trees, and a rule about beds and women. If you read a summary that has Avantivarman endowing the Ranod house with wells and gardens, the summary has merged two stones.

After Purandara the record runs through Kavachasiva, Sadasiva and Hridayesa in verses sixteen to twenty-one. Three names, not four; the count is often given wrongly because the fourth figure, Vyomasiva, gets his own long eulogy across verses twenty-two to forty-two and is easily read back into the list.

Where Mattamayura itself was, nobody knows, and the ignorance is a hundred and thirty years old. Kielhorn in 1892: “Mattamayûra, the town of Avantivarman, and Upendrapura (or Vishṇupura ?), the original dwelling-place of the sage Purandara, I am unable to identify.” Mirashi is much more hedged on it than he is usually cited as being. His own words, in 1950: “Mattamayūra, the chief seat of this clan has not yet been identified. It must have been situated not far from Terahi, Ranod and Mahua… It may be identical with Kadwaha about 15 miles south of Ranod…” That is a possibility argued from the density of surviving temples, offered as a possibility, and it is routinely quoted as a settled identification. Sanderson leaves the site unlocated.

The monastery had an income

The Bilhari stone is where the network stops looking like a spiritual genealogy and starts looking like an organisation.

It gives a working abbatial succession, eight deep: Rudrasambhu, Mattamayuranatha, Dharmasambhu, Sadasiva, Madhumateya, Chudasiva, Hridayasiva, Aghorasiva. It gives a second, separate chain earlier in the record: Pavanasiva, Sabdasiva, Isvarasiva. And in verses fifty-seven and fifty-eight it records an administrative act rather than a miracle. The king Lakshmanaraja II brings Hridayasiva in and offers him the Vaidyanatha matha; Hridayasiva makes over the Nauhalesvara matha to his disciple Aghorasiva. A transfer of one house to make room for another. Houses, in the plural, held by a single man and reassigned.

Then, in verses seventy-nine to eighty-one, the record does something almost no prashasti does. It itemises revenue.

The names are worth keeping as the stone has them, because the values behind them are gone. A shodashika is levied on every khandika of salt, and one on every oil-mill each month. A paura is due for a couple of yugas per day. A kapardi is charged on every vithi, a shop. Four pauras are payable on an elephant offered for sale, and two on a horse.

These units are hard, and every popular retelling gets them wrong. A yuga in this passage is not a pair of oxen. Mirashi’s own footnote says the meaning is uncertain and suggests it is probably a voucher authorising the owner to display his goods in the market-place for a day. Kielhorn declined to translate the passage at all, calling it illegible or doubtful in part.

Even with the units unsettled, the shape of the thing is unmistakable. Salt by the measure, oil-mills by the month, shops, and a scale for livestock at the point of sale. This is not a lump endowment of land producing rent. This is a running income from a market. It is the best evidence in these five records that these houses were operating institutions with cash flow, and it survives because one Kalachuri king wanted his generosity recorded in verse.

Three hundred thousand villages

Now the number.

It appears once, on the Malkapuram pillar of 1261, in the phrase gramanam bhiksam trilaksim, a gift of three lakhs of villages. Everything else is transmission.

The first editor did not believe it. In 1930, on page 152, Pantulu wrote: “The founder of this monastery was one Sadbhāva Śambhu who obtained a gift of three lacks of villages (or was it a villages fetching an income of Nishkas (coins) ? ) from the Kalachuri king Yuvarājadēva and gave away those villages to the Maṭha as an endowment.”

The doubt is inside the sentence, in the editor’s own parenthesis, in the first printing. What happened to it afterwards is the whole story.

Mirashi read the phrase as a count of villages, in the introduction to the standard corpus edition, and reached for an explanation rather than a check: if correct, he wrote, it would indicate that the king assigned one third of the total revenue of his home province of Dahala, which by tradition comprised nine lakhs of villages. That is a hedge, and it is a hedge that keeps the number. K. K. Handiqui in 1949 and Swamy in 1975 both keep it too.

The doubt did not die, though. It went into a footnote. In 1971, in the standard Indian account of this pillar, D. C. Sircar wrote: “The word grāma, ‘village’, may have been used in the present context in a modified sense or, probably, the number is inflated.” One sentence, at the bottom of page 156, with a cross-reference to his own work on historical geography. It is rarely quoted.

Sanderson took the question up again in 2009, in a footnote of his own, and what he actually does there is more interesting than a rejection. He prints the verses, translates gramanam bhiksam trilaksim as “a 300,000 endowment of villages”, and then says of Mirashi’s reading of that phrase: “This would indeed be a vast endowment, so vast indeed that I find it hard to accept his interpretation.”

His control is the Gurgi grant, and the control has to be stated carefully, because it is easy to overstate on both sides. “The Gurgi inscription mentions only about twenty villages and a town” is Sanderson’s own summary, not Mirashi’s count. Mirashi’s list on page 226 of the Gurgi place-names he could make out with certainty runs to sixteen, one of which he says may be a river. Either way the largest grant this lineage’s own stones trouble to itemise is of that order, while the eulogy claims three hundred thousand. The gap is four orders of magnitude.

But the control is not the argument. The argument is a reading. The pillar, Sanderson proposes, “need mean only that the endowment … was valued at 300,000 of some unspecified monetary unit”. And he does not claim the idea. “This alternative was already considered by PANTULU, the first editor of the Malkāpuram inscription,” he writes, and quotes the 1930 parenthesis back in full.

Then he does the thing that turns a guess into an argument. He finds a parallel. A sixth-century inscription from a site near Mrohaung in Arakan records the gift to a Buddhist monastery of a trisahasriko gramah, which, as that inscription’s editor pointed out, can only mean a village with a revenue yield of three thousand. The editor was D. C. Sircar. A number attached to a village in a grant can be what the village is worth rather than how many villages there are.

So the shape of this is not doubt, then certainty, then doubt again. A first editor raises a query in 1930 and answers it the conventional way in the same sentence. A corpus editor removes the query. A third scholar keeps it alive in a footnote in 1971. A fourth supplies, in 2009, the epigraphic parallel that makes it an argument. Eighty years to turn a parenthesis into evidence.

What has not happened is the next step. So far as this article has been able to establish, nobody has taken up Sanderson’s monetary reading in print, or tested it against the rest of the Malkapuram text, in the years since he proposed it.

That claim is only as good as what has been read, so here is what has not. Cynthia Talbot’s “Golaki Matha Inscriptions from Andhra Pradesh: A Study of a Saiva Monastic Lineage”, published in a felicitation volume in Delhi in 1987, is a dedicated treatment of exactly this body of records and is not available online. Tamara Sears published a monograph on these buildings, Worldly Gurus and Spiritual Kings, in 2014, and she is the scholar most likely to have tested the Gurgi identification directly; the only copy this article could reach is a restricted lending item. Until somebody reads those two, the question stays where it is.

What the pillar actually establishes

Strip the three lakhs out and the Malkapuram pillar is still remarkable, because it is the one record here that describes a monastery as a going concern. Bilhari gives the income. Malkapuram gives the payroll.

Two villages were given, one by the Kakatiya king Ganapatideva and one by his daughter Rudramba, and amalgamated into a single agrahara, a village settled on brahmins, named Visvesvara Golaki. Sixty brahmin families were brought up from the Tamil country to settle it. The land was measured in puttis by the standard rod of Penumbaka, and here Pantulu is more useful than he is usually given credit for. The word, he notes, is not used in that sense in the Guntur district at all. In the Godavari district it means eight acres. The acreage is therefore uncertain rather than unrecoverable, which is a different and better position than the one usually stated.

Then comes the establishment, and this is where a reader has to hold two editors at once, because Pantulu’s 1930 summary and Sircar’s 1971 account of the same lines do not itemise it identically.

Pantulu gives a physician and an accountant, six brahmin servants with two of them cooking, ten Virabhadras as village guards, twenty Viramushtis, ten artisans under a master craftsman, ten dancing girls, eight players of musical instruments, fourteen singers, six drummers, and one man from Kashmir. Sircar reads the same lines more fully and in places differently. He has two cooks who were brahmins, four servants, and six brahmins belonging to the monastery and the rest-house, which is twelve where Pantulu has six. He has twenty bhatas and Viramushtis together as constables, which is what a Viramushti seems to have been, ten artisans working in gold, copper, stone, bamboo and iron, and six karata players where Pantulu has drummers. And he supplies a detail Pantulu’s English drops: the ten Virabhadras are matted-haired men from the Coda country, brought up from the south.

Then Sircar does something no other reader of this pillar that this article has found has done. He adds them up. “Seventy-three other persons were given each a Puṭṭikā,” he writes, and a fraction of a nivartana besides, and then he lists them. Count his list and it comes to seventy-three exactly.

The text also contains its own audit. The ten dancing girls and the eight maddala players are paid at their own rate, one and a half puttikas apiece, above the seventy-three and below the teachers. After naming them the pillar refers back to them as aṣṭādaśānām eteṣām, “of these eighteen”. Ten plus eight. The arithmetic checks.

The teachers are counted too, and this is where a reader working only from Pantulu’s English will go wrong. His summary reads: “In addition to these there were brahmins well versed in the three Vedas, in grammar, in logic, and in general literature and five others skilled in philosophy.” Five is the only figure in that sentence, and it is easy to conclude that the record does not number its faculty. Sircar, working from the Sanskrit, numbers it: three professors for the Rig, Yajus and Sama Vedas; five teachers for pada, vakya, pramana, sahitya and agama; and a physician and a scribe. “Each of these ten persons received from Viśveśvara two Puṭṭikās of land.” Ten.

The superintendent’s allowance is given as one hundred nishkas, and it too is double-attested. The rules for his office are the closest thing in these records to a constitution. He must be an ascetic guru consecrated by a guru of the Golagi line, a brahmin, tranquil, honest, and a master of the Shaiva scriptures. And, Mirashi notes, the Shaiva community of the village could appoint a new acharya, a new presiding teacher, if the sitting one proved negligent or was guilty of misbehaviour.

A removable abbot. In a thirteenth-century Andhra village, written into the endowment.

One more thing about that establishment, since it sits oddly beside the Ranod wall. Ten dancers and fourteen singers are on this house’s payroll, and they are women. The rule cut at Ranod forbids any woman to pass the night inside the matha. Three centuries and a thousand kilometres separate the two stones, so they are not in contradiction with each other. They are two houses of one tradition, and the tradition evidently held both.

The lying-in house, and the kitchen

The charitable claims are the reason this monastery is famous outside specialist literature, and they sort into four different categories once you go to the stones.

The prasutisala, the lying-in house for women in labour, is real. It is on the Malkapuram pillar, endowed out of the same grant as the arogyasala, the health house, in the same clause. How rare it was is a separate question, and that is where the writing on this gets loose. Ranabir Chakravarti and Krishnendu Ray call it a unique case, not encountered previously. They assert that. They do not survey for it. Dominik Wujastyk surveyed. His 2022 article on the evidence for hospitals in early India is the fullest survey available, it records the birthing house with no uniqueness claim attached, it turns up no rival in forty-three pages, and it warns in its own voice that the corpus is by no means exhaustive. So the formulation this article uses is the small one. No other epigraphically attested maternity house is known to the current literature. That is less than the version in circulation. It is what the evidence will carry.

The feeding house is the one that looked for years like a contradiction. The pillar calls it a vipra-satra, a brahmin feeding house. At line forty-eight of the second side the endowment reads prasūtyārogyaśālābhyāṃ viprasatrāya cetaraṃ, assigning the remaining share to the lying-in house, the health house and the vipra-satra. That looks straightforwardly caste-restricted, and that is how it has usually been reported. The dual in prasutyarogyasalabhyam makes the lying-in house and the health house two institutions rather than one, which is also how Sircar reads them, though some later summaries treat them as a single compound establishment.

Twenty-three lines later the same pillar says who eats there.

ārabhya viprān anivāritānāṃ caṇḍāla-paryantam upāgatānām … anna-pradānāya ca sarvakālam akalpayad rājagurur mahātmā

…and for the giving of food, at all times, to those who have come and are not turned away, beginning with brahmins and extending as far as chandalas, the great-souled royal preceptor made provision.

Chandala names a hereditary group placed outside the four varnas altogether and treated as untouchable. It is the far end of the scale the pillar’s drafters recognised. The clause names both ends of that scale and then says nobody at either end is turned away.

One institution, two descriptions. Sircar, reading the whole record, has Visvesvara establishing one monastery and one perpetual rest-house, and there is no second kitchen anywhere in the text. The later lines refer back to the feeding house in the singular.

So the house is named for its brahmin clientele and endowed to feed everybody who comes, with the word anivaritanam, “not turned away”, carrying the whole weight. Both facts are true at once, and each one alone is a distortion. That is a real and specific piece of information about how caste worked in a thirteenth-century endowment, and it is more useful than either the sanitised version or the cynical one.

Pantulu had it whole in 1930, rendering the clause as a choultry for feeding all classes of persons, from the brahman down to the chandala, with the Sanskrit in a parenthesis after it. Sircar had it whole again in 1971, in one sentence, with the tension left in: “The word Vipra-satra means a rest-house established for the use of Brāhamaṇas, though it is stated that there was arrangement for feeding, at all times, without any obstruction, all people from Brāhmaṇa to Caṇḍāla, who came and asked for food.” The though is the whole finding. Later summaries keep the first half and drop the clause after it, and the drop is what produces the two rival versions that are still in circulation.

The other two claims do not belong here. The fifteen-bed hospital regularly attributed to the Golagi matha is from the Tirumukkudal inscription of the Chola king Virarajendra. The famous charity house open to every kind of person is at Belagave, and it is Kalamukha rather than Saiddhantika, described in an inscription of 1162 that Fleet edited:

”— a place where food is always given to the poor, the helpless, the lame, the blind, and the deaf, and to professional story-tellers, singers, musicians, bards, players, and minstrels whose duty it is to awaken their masters with music and songs, and to the naked and the crippled, and to (Jain and Buddhist) mendicants, to (Brâhmaṇ) mendicants who carry a single staff and also those who carry a triple staff, to haṁsa and paramahaṁsa ascetics, and to all other beggars from many countries,— a place where many helpless sick people are harboured and treated,— a place of assurance of safety for all living creatures,”

That is a magnificent record and it is somebody else’s.

And women’s education, which appears in a good deal of popular writing about the Golagi matha, is not in the pillar. What is in the pillar is smaller and better. A body of Samavedin brahmins from Gauda kept the estate’s written accounts and held land for doing it, and, on Sircar’s reading of the clause, if one of them died without issue his wife would enjoy the land provided she took up the work herself. Not a widow’s pension. A literate accounting tenure, inherited on condition that the woman does the accounting.

Sircar prints a question mark against his own rendering of that condition, so it is a reading rather than a certainty. It is still the only place in this corpus where a woman is contemplated doing skilled clerical work, and it has been passed over for a claim about women’s education that the record does not make.

Where the stones touch the books

For most of this article the monastery is an economic object: villages, tolls, staff lists, a removable abbot. There is one point where it becomes a place where somebody sat down and wrote.

Probably the most influential Saiddhantika ritual manual of the period, on Sanderson’s estimate, the Kriyakandakramavali of Somasambhu, was composed at this monastery while its author was abbot there. It is the handbook behind an enormous amount of later Shaiva practice. Its rites were absorbed almost entire into the Agnipurana, which means they entered the Puranic mainstream and travelled far beyond the Agamic tradition that produced them. Manuscripts survive in Kashmir, in Nepal and in the South, and one Nepalese palm-leaf fragment transmits part of its reparation rites with the Shaiva mantra-deities swapped out for Buddhist ones.

There is a detail in its colophons worth stating carefully, because it is usually stated flatly. The Kashmirian manuscripts, printed in the Kashmir Series, close with a verse dating the work to Vikrama 1130, which is 1073 or 1074 of the Common Era. The Devakottai edition, prepared from Grantha manuscripts, carries the same verse with the year given as Vikrama 1153, which is 1096 or 1097. Two different dates in an otherwise identical verse indicate conscious revision rather than corruption, Sanderson argues, and the text probably circulated in two editions.

There may be a third. An early Nepalese palm-leaf manuscript ends instead with a regnal year written sasama, which means nothing. If that is an error for dasama, the tenth, and if the king it names is Yashahkarna of Tripuri, the year is 1082. If it is an error for prathama, the first, the year is 1073 again and Kashmir is right.

So the single most-copied Shaiva handbook of its century is a book whose date of composition is itself a small textual problem, and the problem is one consonant deep.

This is the moment where the epigraphic record and the manuscript record touch. A house that shows up in eulogies as a recipient of villages is also the address on the title page of a book that shaped ritual practice across the subcontinent for centuries.

What happened to them

The lineage moved south, and the scholarship on that is thinner than it should be. Noboru Karashima, Y. Subbarayalu and P. Shanmugam published an epigraphical study of Tamil mathas in two parts in the Indian Historical Review, in 2010 and 2011. The full text is behind a paywall and this article has read only the two abstracts, which is worth saying plainly, so here is what those abstracts say. The first speaks of an influx of Shaiva ascetics from north India, of mathas localising and Tamilising the Brahmanical orthodoxy those ascetics brought, and it names the Golaki matha, originating in the north, as the case that shows the process, ending in the establishment of Saiva Siddhanta philosophy in the Tamil country in the thirteenth century. The second speaks of an influx of north Indian brahmin ascetics into the Chola country in the eleventh century, and of participation in matha activities from the twelfth by various people including low-ranked jatis, the birth-groups of the caste order.

That is the traffic running one way. It also ran the other. The Malkapuram agrahara, founded in 1261 by a lineage whose charter myth is set in a north Indian pine forest, was settled with sixty brahmin families brought up from the Tamil country. Ascetics went south. Householders came north. Both movements sit in the same body of records, two centuries apart, and neither is usually told alongside the other.

In central India the record continues, and it is more crowded than the story usually told about this period.

At Kadwaha there is a slab of Vikrama Samvat 1366, that is 1309 or 1310, in the Bhutesvara temple inside the garhi, the fort enclosure. Tamara Sears, who published a study of these buildings in 2009 and a monograph on them in 2014, reads it as recording that an ascetic named Bhutesvara performed special worship at a Shiva linga after mlecchas, which she renders “barbarians” and reads as Muslim forces, had overrun the earth. The record describes nineteen acts of evil, pataka, at least one of which damaged the jaladhara, the tank beside the temple. Bhutesvara gave a new tank and undertook severe austerities. It is a record of depredation and of repair, and it is not, as it is sometimes reported, an account of an assault on the monastery itself.

Sears draws a distinction that is more useful than either the rupture narrative or the continuity narrative. Some reuse at Kadwaha is deliberate and ideological: a mosque was built on a platform of 8.45 by 12.85 metres, its mihrab carrying a verse from the third sura of the Qur’an and a hadith, co-opting the westward axis of the site. Some is casual: a third storey was simply added to the monastery. And transformation, she argues, does not necessarily imply rupture. At least fourteen other temples at the site stayed in use. Pilgrimage inscriptions record Shaiva, Vaishnava and Jain visitors through the fourteenth and fifteenth centuries. A slab of Vikrama Samvat 1403, which Sears converts to 1347, names three malikas, which she renders “landholders”, whose names are composite Sanskrit and Persian.

At Surwaya, in the same group of sites, the fort gate is dated Vikrama Samvat 1350, that is 1293 or 1294, under a queen named Salahanadevi.

Fortified monasteries, a mosque on a temple platform, three landholders with half-Persian names, and pilgrims still writing on the walls a hundred and fifty years later. All of it at one site, all of it in the record.

What is still not settled

A good deal, and the honest thing is to list it.

We do not know what the Bilhari tolls were worth. The units are uncertain, and one of them defeated Kielhorn entirely.

We do not know whether the Vyomasiva eulogised at Ranod is the same man as the author of the Vyomavati, the Nyaya-Vaisheshika commentary. That identification was proposed by Dasharatha Sharma in the Indian Historical Quarterly, volume X, at pages 165 and following, and Mirashi endorsed it in 1950 with a “probably” in his text and a flat “It has turned out to be correct” in his footnote, adding a real piece of evidence: a verse in the Gwalior Museum inscription says that this Vyomasambhu “by his commentary, expounded the true nature of the real padartha”. But the specialists in that philosophical literature have simply not taken it up. Handiqui discusses the Ranod Vyomasiva at length and never makes the identification. Sanderson does not endorse it. The Gwalior verse is real and worth citing. The identification is not established.

We do not know how the Golagi houses stood in relation to the Kalamukhas. Swamy classifies the Mattamayura, Amardaka and Golagi schools as dualistic Shaivism, that is Shaiva Siddhanta, and treats the Kapalika, Bhairava, Pasupata and Kaula streams as rivals. The word Kalamukha occurs once in his study, inside a doxographic list. David Lorenzen, who wrote the standard study of the Kapalikas and Kalamukhas, does not discuss this network. The question has been bypassed rather than answered.

And we do not know how far the Malkapuram pillar can be trusted about the early abbots. Mirashi was inclined, in his own words, to look with suspicion on its statements about the early pontiffs of the Golaki matha. Sircar answered him at length and thought the reasons weak. That is the sharpest disagreement in the literature about this record, it is between two Indian epigraphists, and it is nowhere in the popular accounts, which quote the pillar as though nobody had ever questioned it. It also complicates the picture of Mirashi as the credulous keeper of the number: on this pillar he was the sceptic.

None of this is a reason to distrust the record. It is the record doing what stone records do. A prashasti is written to be read by a god, a king and posterity, in that order, and it tells you what its patron wanted fixed. It fixes the lineage, the initiation, the gift and the rule. It does not fix the acreage, the accounts, or the address.

What survives of the Golagi matha is a set of claims it made about itself, and a handful of details it let slip while making them. The details are the better part. A tank that was meant to last forever. Three narrow rooms with barred windows. Ten dancers, eight players, eighteen in all. A kitchen named for brahmins and endowed to turn nobody away. A rule against beds.

The three hundred thousand villages exist on one pillar and nowhere else.

The rule against beds is still on the wall.