The Maharthamanjari: seventy verses, and the seventy-first

The night in the pavilion

Some time around the year 1300, in the Cola country of the Tamil south, a man sat at night in a sacrifice pavilion. He had spent some time at the ordinary work before that night: worshipping the deity daily, reciting, meditating without moving, walking to the ends of the directions. On this night he satisfied the supreme Goddess with perfumes, flowers, unhusked rice and wine. He tasted what his own account calls three fierce cups of bliss. He took up the great state that lies between waking and dream. Then he sat, in his own words, overcome with wonder, his eyes wide as a blossoming white lotus.

We have this because he wrote it down himself, in Sanskrit verse, near the close of a long commentary on his own work. His given name was Goraksha. His teacher, Mahaprakasha, had given him a second name at initiation, and it is the second name that survived: Maheshvarananda.

What came out of that night was a short book with a long shadow. It is called the Maharthamanjari, the blossom-cluster of the mahartha, the Great Meaning. It is Maheshvarananda’s only surviving work and one of the fullest the Krama has left, and the Krama is a Kashmiri school that worshipped the phases of perception as goddesses. Between roughly three and five centuries separate the school’s ninth-century emergence from this book, depending on which of the two live datings one takes, and about sixteen hundred miles separate Kashmir from the town where it was written.

Three things about it are strange, and all three are deliberate.

He wrote the verses in Prakrit, which is the wrong language for scripture. He then wrote a vast Sanskrit commentary explaining his own verses, in the language he had just declined to use. And he states, in his own prose, that the book consists of seventy verses, before setting out a list of seventy-one.

None of these is a slip. Each one is the answer to a problem, and the problems are worth more than the answers.

What lived only in mouths

Before the account of that night, the commentary sets out where the teaching came from, and it is a lineage of voices rather than of books. Then comes the clause that explains the rest of the book:

yoginam vadaneshv eva tishthaty atyantadurlabha

It abides only in the mouths of the yogins, exceedingly hard to come by.

Read that as a description of an asset and it sounds like a boast. Read it as a description of a risk and it is the problem the Maharthamanjari exists to solve. A tradition carried in mouths is a tradition one generation from silence. A good deal of what is known about Krama literature survives as quotations inside his commentary, which is a fact about what reached us rather than a statement about his intentions.

He does not say he wrote to prevent a loss. The book reads as though he did, and that is a reading.

The language she was owed

Having decided to write, he had to decide what to write in, and the obvious answer was Sanskrit. Sanskrit was the language of scripture, of commentary, of every learned reader he could hope for. He used it for the commentary. He did not use it for the verses.

He gives his reason in an opening verse, and the reason is not literary. He asks that the goddess he met in a dream, named there as the fair-faced Siddhayogini, be pleased by these seventy gathas, single Prakrit stanzas. Then comes the word everything turns on, and the two streams of the text do not agree on it. The Kashmir edition of 1918 prints svochita, in languages suited to her. The southern line reads svapita, which would make the languages the dream’s rather than hers.

Either way the choice is answered to the goddess and not to the audience. The verses are in Maharashtri Prakrit, a Middle Indic literary language with its own long tradition of erotic and suggestive short verse, and not the language a tantra would be expected to arrive in, though Cox notes that works on poetics drew their proof-texts from exactly this kind of Prakrit lyric. The whole of it rests on one adjective, and the editions read that adjective two ways.

Whitney Cox puts the case differently, and more sharply. He notes that the Krama’s earlier non-Sanskrit works, with their artfully artless language, suited a soteriology of sudden awakening, and then says that this book’s aesthetics differ markedly from them. What Maheshvarananda gets from Prakrit, on Cox’s reading, is indeterminacy: the verses do not obscure their meaning so much as leave it open, and the Sanskrit commentary then unpacks and regiments the proliferation of meaning the Prakrit allows. The two halves of the book are a machine for opening and closing a text at once. Jackson Stephenson has made a comparable case about a different text, the Mahanayaprakasha of Shitikantha, a Kashmiri Krama work of ninety-four verses, arguing that such literary Middle Indic registers served as privileged vehicles for direct esoteric teaching while Sanskrit carried the discursive material. Its language is itself disputed: the usual description has been Old Kashmiri, and Stephenson argues it is literary Apabhramsha. Stephenson is careful on one point and the carelessness is easy to fall into: he declines to call that register a vernacular at all, holding its Apabhramsha to be artificial and literary, and objects to the word being used of it.

That second text matters for a reason beyond its argument. It means Maheshvarananda was not a lone eccentric writing scripture outside Sanskrit. A Kashmiri author had done the same thing in a different Middle Indic register.

The tidy version of that parallel does not survive contact with the evidence. The title Mahanayaprakasha attaches to more than one work: Dvivedi records two published editions, one in Prakrit gathas and one in Sanskrit, and reports that many of the passages the Parimala quotes under that name are found in the Sanskrit one, with some that are not, which he supposes belong to the same work published incomplete. So Maheshvarananda demonstrably read a book called the Mahanayaprakasha. Whether it was Shitikantha’s is another question, and Stephenson says plainly that it is unclear whether Maheshvarananda knew Shitikantha’s.

Seventy verses, and the seventy-first

The book ends by counting itself, and the number it gives is not the number on the page.

Its last verse salutes the Yogini, and it opens by naming what has gone before: paaḍasutta-sattai, which the author’s own Sanskrit gloss renders prakrita-sutra-saptati, the seventy Prakrit sutras. He is standing outside them when he says it. The verse that counts seventy is not one of the seventy.

That is not an editor’s inference. It is the text, on its last page, and it agrees with something the author had already said on his first. In the prose opening the commentary he explains that the whole work comes about through seventy gathas that function as the sutras of the tantra. The operative word is sutrayamana, acting as sutras. Seventy verses are load-bearing. Seventy carry the doctrine the commentary will unpack.

Between those two statements he sets out a synopsis of the contents, numbered, and it runs to seventy-one. The seventy-first entry is where the tantra-maker gives the purpose that is the reason for the tantra’s teaching, and what marks it out is one word: where the pattern would give an ordinal, he writes antyayam, in the last one.

So the seam is the author’s, and it is audible as well as countable. Cox describes the work as seventy arya verses and then a final verse in shardulavikridita, an art-metre, and an unusual one for Prakrit. The seventy move to one measure. The seventy-first changes step to say why the book exists and to count the others.

The printed editions then part company, and the reason is not settled here. Both southern editions, Trivandrum in 1919 and Varanasi in 1992, number the final gatha seventy-one. The Kashmir edition of 1918 ends its gathas at seventy, running one behind the southern numbering by the close. Dvivedi, printing seventy-one, nonetheless argues in his introduction for seventy, from parallel works of the region that run to the same number.

It would be neat to say the editors simply chose between the author’s two figures. This account cannot say that, because it has not collated the two texts, and a verse missing or divided differently on one side would produce exactly the same result. What is certain is the seam itself, and the seam is the author’s, not an editor’s.

The book that explains itself

The commentary is called the Parimala, the fragrance, which is what a blossom-cluster gives off. It is svopajna: the author’s own commentary on his own verses.

There is something faintly absurd about writing seventy allusive verses in a literary Prakrit and then explaining them at enormous length in the language you declined to use. It is also, on reflection, the only way the project works. The verses answer the goddess. The commentary answers the reader. He needed both, and he was not willing to let one job spoil the other.

The Parimala is why the book matters to anyone reconstructing the Krama today. Maheshvarananda quotes, and much of what he quotes is gone. Abhinavagupta’s Kramakeli survives in pieces, some of them on his pages. A work called the Kramasiddhi survives in a handful of extracts and nothing else, and Navjivan Rastogi reads it as a late work partly because no Kashmiri authority so far identified quotes it. His own teacher Mahaprakasha was described as a composer of tantras, and of all of that what survives is what the Parimala quotes. Dvivedi lists four of his hymns named there: the Manonushasanastotra, the Matangistotra, a Samvitstotra whose attribution he flags as unsteady, and the Anandatandavavilasastotra, on the play of the blissful tandava. Wenta prints a verse of that last one and reads it as a hymn to the dance of bliss, the anandatandava, which is the technical name of the dance at Chidambaram.

A man wrote a commentary on himself, and the commentary turned into an ark.

The hinge at verse thirty-four

A book of seventy verses that runs from metaphysics into practice has to turn somewhere, and this one turns in a place we can name.

Dvivedi describes the contents in a single sentence: the manner in which the thirty-six tattvas, the levels of reality, shine forth, and then the Krama procedure of worship, which expounds four things called the Shripitha, the Panchavaha, the Netratraya and the Vrindachakra. Aleksandra Wenta puts the division more precisely, taking the first half from Cox: the earlier part of the work expounds the thirty-six tattvas along the lines of the Pratyabhijna, the recognition philosophy of Kashmir, and from the thirty-fourth verse the second part is given over to Krama esoteric practice.

So the synthesis everyone describes in the abstract is visible as architecture. Pratyabhijna metaphysics for roughly half the book, Krama practice for the rest. He does not argue that the two belong together. He builds a book in which the second half stands on the first, and leaves the arrangement to make the case.

Five powers, and the gap between two thoughts

The Krama’s governing idea is that a single act of perception runs in stages, and that the stages have names and are worshipped. Maheshvarananda works with five of them, which he calls the Panchavaha, the Five Flows.

Wenta sets them out. Srishti, creation, is the object’s emergence into objectivity. Sthiti, maintenance, is its persisting for some duration. Samhara, withdrawal, is the shift from one object to another, and her example is the shift from a jar to a pillar. Anakhya, the nameless, is the gap between two cognitions, when the impressions withdraw into a potential that cannot be put into words. Bhasa, luminosity, is the state in which all objects are cognised at once as one with the primordial Self.

None of that is a modern reconstruction. One gatha instructs the practitioner to remember the Shripitha, the Panchavaha, the Netratraya and the Vrindachakra, and then the line of teachers and the five powers beginning with srishti; and the commentary on it immediately names them, all five, in a single line: srishti, sthiti, samhara, anakhya, bhasa. Elsewhere it quotes the Kramasadbhava running the same five in order. The commentary also defines the fourth directly:

When every discursive construction, pillar-ness and the rest, has subsided: that is anakhya.

The pillar is not an illustration Wenta brought to the text. It is the text’s own worked example, running the whole five-phase analysis on a pot and a pillar, and the clause above is the end of it. Jar and pillar are among the stock dummy objects of Sanskrit epistemology, which is the point: he is doing theory of perception with the ordinary furniture of the subject, and calling the stages goddesses while he does it.

There is a small technical note in the commentary that turns out to carry weight. For anakhya and bhasa, it says, no separate mantra is taught here.

That sounds like housekeeping, and it is a real asymmetry in the scheme. He names five phases throughout. The last two are then not given mantras of their own, and the commentary carries that as a quoted line rather than a remark of its own. Whether that is a trace of an older four-phase reckoning or simply how the ritual was taught, the text does not say, and this account will not put a motive behind it.

Sixty-four in the wheel of masses

The Vrindachakra is worth describing exactly, because it has been described loosely and because the text comes back to it a dozen times.

The commentary gives the etymology first. A vrinda is a mass, an aggregate, a heap. The Vrindachakra is the wheel of those masses. The gatha it expounds says that in the manifested fivefold aggregate there are sixty-four in the Vrindachakra, and the commentary confirms the arithmetic: in all, the Vrindachakra is sixty-four.

Then it adds a line that should be familiar by now. Shuddharaudreshvari, otherwise called Kalasamkarshani, is the sixty-fifth, and because she runs through all of them she is not counted separately. Sixty-four, and one that pervades the sixty-four and stands outside the count. The book that numbers its verses that way numbers its wheel that way too.

Its setting is the body, and the move the text makes is a familiar tantric one done carefully. Having established that the gross body is to be worshipped as the seat, the pitha, Maheshvarananda turns to the Vrindachakra in order to show that the subtle body is to be contemplated in the same way. The outer seat and the inner wheel are two applications of one idea.

The word for the fivefold aggregate is pancha-skandha, which is also the Buddhist term for the five heaps a person is analysed into, and it is tempting to call that a borrowing. The commentary will not let you. It names the five immediately, and they are grades of siddhas, sixteen and twenty-four and twelve and eight and four, aligned with the five flows and set in a frame it calls the city of eight. Skandha is ordinary Sanskrit for a trunk or a division. The same test that disposed of the pillar a few paragraphs ago disposes of this.

The Buddhist question is real elsewhere in the book. Dvivedi flags the fifty-fourth gatha and its commentary taking on momentariness and emptiness directly. Rastogi goes further, and his claim is doctrinal rather than verbal: he holds that philosophical Buddhism marked the vital Krama dicta, and reads anakhya itself, the nameless fourth phase, as possibly a Krama reworking of Madhyamaka emptiness. That is a real argument about a real term. It is simply not this diagram, on this word.

One caution about the Vrindachakra, and it is a divergence rather than a dispute. Navjivan Rastogi reports that on Maheshvarananda’s own valuation it is the most important single aspect of the Krama system, thinks the Srichakra may have inspired it, and holds that the idea travelled from the Krama into the Tripura system afterwards. Wenta, writing a long study of this author’s ritual thought, does not mention the Vrindachakra at all. Two serious readers, one text, and one of them does not see what the other reports the author calling central. We do not adjudicate that.

What the school called itself

The 1919 preface records that the work designates its school by three names: Pratyabhijna Shastra, Samvit-Tantra, and Auttaradvaita. That third name is harder to render than it looks. Auttara is the strengthened form of uttara, which means higher, and also northern. So Auttaradvaita is the non-duality of the higher, or of the north, and the choice is not settled. Wenta reports that Maheshvarananda called his own system the anuttaramnaya, the transmission of the unsurpassed, which is built on the negation of that same word uttara. Whether the two names are variants of one idea or two different ones is open, and it would be convenient rather than honest to collapse them.

What it was, in practice, is a question with a careful answer. Count the words and the answer comes back one-sided: krama and its compounds run to something over three hundred across the printed volume, the editor’s own introduction and indexes included, against under sixty for pratyabhijna. That is a crude measure. Of the book itself it sees only the Sanskrit half, and it is blind to the Prakrit verses, where krama loses its r and comes out as kama. It points the same way the scholarship does. Maheshvarananda, Wenta writes, was well acquainted with the Krama system and its scriptural basis, and he regarded its framework as the highest and most esoteric of all the Shaiva traditions.

And yet she also holds that his distinctive originality lies in his attempt at a fresh synthesis of the various Shaiva and Shakta traditions, producing what she calls a unique amalgam of Pratyabhijna, Trika, Krama, Kaula and Shrividya. Rastogi and Cox both class the book as a Krama work.

The honest description is therefore double, and the two halves do not cancel. He is a Krama author who ranks the Krama above everything else. His originality is what he builds around it. He is not a neutral synthesist and he is not a purist, and any account that makes him simply one or the other has tidied something that was not tidy.

The god who is doing this

If worship is a sequence of cognitive phases, the obvious question is what is left of worship. Maheshvarananda’s answer is the one his Kashmiri predecessors had already reached, and he takes it further than they did.

Following Abhinavagupta and Kshemaraja, he moves the rite inward. Worship, in Wenta’s summary of his position, is redefined as the thought-reflection, the paramarsha, on one’s own true nature. At the forty-second verse the external rites lapse once the supreme reality is known. At the forty-seventh, or the forty-third depending on which of Wenta’s own citations one follows, the claim is put flatly: the great luminosity is the deity and none other. Alongside it the text sets an inward linga, a linga of consciousness.

From the Pratyabhijna he takes the piece that makes this work, which is a theory of a free agent underlying the whole cognitive process. Without it the five phases are a description of how perception decays. With it they are something a self is doing. At the seventeenth verse, maya, usually the problem, becomes the summit of Shiva’s freedom. At the eighth, quoting Somananda’s Shivadrishti, Shiva plays out this and that the way a king plays the part of a soldier.

The nineteenth verse is where it lands, and Wenta treats it as the key verse for the whole dancing-god reading. Shiva performs the Five Acts as the purusha, the person, the self that acts. The classic Shaiva five are emission, maintenance, withdrawal, concealment and grace, and this book does not quite keep them: in the commentary on that very verse the fourth act is the experience of the fourth state, the turiya, which is anakhya under another name. The two sets are being folded into one another as he writes, and the commentary says so, naming one pair of acts concealment and grace in one aspect and nameless and luminous in another. The purusha’s agency turns out to be identical with Shiva’s. A line she quotes from the commentary on that verse puts the upshot in the first person: in the end I am Shiva, the great Dancer. It belongs to the Parimala rather than to the gatha.

That last word is doing real work. Wenta reads the whole project through performance: pure consciousness is the dancer and actor of a world-drama, dance is the enactment of the Five Acts in the inner sphere of ritual, and what is original in Maheshvarananda is a link between ritual, cognition and theatrical performance. Chidambaram, where the scholarship places him, is the temple of the dancing Shiva, and his own teacher Mahaprakasha had written a hymn to the dance of bliss there.

One caution, and it cuts both ways. The word Nataraja does not appear anywhere in the printed text. But the text does name the dancing god, as Natesha, the standard Sanskrit equivalent, and it does so in its own closing self-description. So the dancing-god reading is not imported from the town; it is in the text, under a different name. What remains Wenta’s own is the weight she puts on it, the reading of the whole work through performance, and the identification with Chidambaram’s deity in particular. That is a strong reading and it is a reading.

Two men named Shivananda

Anyone reading further into this literature will meet the name Shivananda in two roles that matter here, and conflating them wrecks the whole picture. Rastogi, who numbers more than two of them, warns that there are quite a few Shivanandas and that irresponsible identification must be avoided.

The first Shivananda stands at the head of the Krama itself, the recipient of the visionary transmission from which the lineage descends, placed by one account at Oddiyana and by Rastogi in Kashmir itself. The second is Maheshvarananda’s own grand-teacher, his teacher’s teacher, and Rastogi heads a whole section of his study with exactly that description. He was a Shrividya author who wrote the Rjuvimarshini, a commentary on the principal Sri Vidya scripture, along with the Subhagodaya and its companions, and the fourth of the introductory verses of this book salutes him by name.

The name was common and the confusion is old. The point of separating them is not pedantry. It is that the second Shivananda is where this book touches a different tradition.

Here the connection is worth stating carefully, because it is easy to overstate. The word Shrividya does not appear in the Prakrit verses at all. What does exist is a shared form. Dvivedi argues the seventy-verse count partly from parallels: Shivananda’s Subhagodaya and its companion works, and Shrikanthananda’s Nishkalakramacharcha, also run to seventy. A Krama text and a group of Shrividya texts from the same region and the same few generations are built to the same measure.

That is a structural link to set beside the personal one, and together they are stronger than either alone. Vocabulary can be borrowed by anyone. A teacher’s teacher who wrote the standard Shrividya commentary, and a shared measure running through both bodies of work, are not a coincidence of the same century. Wenta counts Shrividya as one of five streams in the amalgam she describes, so the traffic is not only formal.

When he lived, and where

Everything above sits at a date, and the date has moved twice.

Ganapati Sastri, editing the work in 1919 with three manuscripts in front of him, wrote the plainest version available to him: the date of the author is not known with any certainty. That was an honest report of the evidence he had.

Navjivan Rastogi, in 1979, placed him at the close of the twelfth century or the beginning of the thirteenth, between about 1175 and 1225. Dvivedi in 1992 put him in the fourteenth century. Whitney Cox writes that he flourished around the turn of the fourteenth century, and for the date itself he does not argue from the text: he refers the reader to Alexis Sanderson, who concludes that Maheshvarananda will have been active around 1275 to 1325.

So there are two live positions and they are a century apart. Rastogi’s early dating turns on where he places the grand-teacher Shivananda, whom he puts at 1125 to 1175 and whom current scholarship moves a hundred years later. The later bracket is where the field now sits, and this article’s figure of 1300 comes from it. That is a choice between scholars, not a neutral summary, and the dissenting voice belongs to the man who is otherwise the fullest guide to this author in English.

On place there is more agreement. Cox puts him at Chidambaram in the Cola country and Wenta does the same. Rastogi locates the Cola country in modern Karnataka, which Cox and Wenta’s Tamil placement rejects, and he also records that Maheshvarananda travelled widely before he began to write.

The text is not silent about where he came from, which is worth saying carefully because one verse in it is easy to misread. The verse that introduces the lineage gives the Cola epithet to the teacher: Mahaprakasha is the crest-jewel of the Cola country, and Goraksha enters in the next verse only as his disciple. Read alone, that verse makes the teacher a Cola man and says nothing about the pupil. But it does not stand alone. Some seventy verses later, at the very close of the commentary, the author lists his own distinctions, and the Cola lands head the list, beside his father Madhava and his teacher Prakasha. Ganapati Sastri’s own Sanskrit preface is blunter than his English one: the Cola country, he writes, is known from the book to be the birthplace of this man and of his teacher. Both. He was reporting the closing verses, not guessing from the lineage verse alone.

A hymn with a Tamil name

In the commentary Maheshvarananda names his own works, and the 1919 preface prints the list in two columns. Ten titles. The Maharthamanjari and the Parimala, which we have. Then the Padukodaya, the Samvidullasa, the Komalavallistava, the Parastrota, a Sukta, the Maharthodaya, the Kundalabharana and the Mukundakeli, which we do not.

Later scholarship counts eleven and possibly twelve, working through the Parimala rather than through the editor’s summary. Most of the discrepancy is not a conflict: Ganapati Sastri printed a list, and others going through the text found more than the list held. On one title there is a real disagreement, since Dvivedi reads the Sukta of that list as no work at all, taking the word in its ordinary sense of a fine saying. The 1919 editor’s own inventory of his author was incomplete, and we can watch the gap.

Two of the names look, at first glance, like they point away from everything else here. The Komalavallistava praises a goddess whose name is current in the Tamil country, and a Komalavalli well known in that country is a Vaishnava goddess, the consort at Kumbakonam. The Mukundakeli carries a name of Vishnu in its title.

The obvious inference is that a Shaiva author was writing to local and even Vaishnava deities, and it is not Rastogi’s. He classes the Komalavallistava itself as a Krama work, on the author’s own statement, and sorts the rest of the lost titles by kind: the Maharthodaya Krama, the Samvidullasa and the Sukta philosophical, and only three of them, the ones that survive as bare names, probably literary pieces. So the names are not a window onto some other devotional life. They are a reminder that a Krama author in the Tamil country wrote under local names and expected to be understood.

What the list does show, without any inference at all, is the shape of the loss. Of the eleven that later scholarship counts, two survive whole, six are extracts inside a commentary, and three are nothing but a name. On the shorter list the 1919 editor printed, two are bare names and the eleventh title is not there at all.

Three manuscripts in Malayalam characters

At the end of his preface, Ganapati Sastri records what he had to work with.

The edition of the work is based on the following three palm leaf manuscripts in Malayalam characters: [the sigla are printed in Devanagari] 1. (ka) A worn out manuscript belonging to Mr. Thekkunkoor Raja. 2. (kha) Belonging to Mr. C. Narayana Bhattathiri, Thiruvalla, beginning with the 10th Gatha. 3. (ga) Lent by Mr. Aryan Narayanan Moossathu, Kottayam.

Then: “Trivandrum. T. Ganapati Sastri.”

Thekkunkoor, as the preface spells it. Thiruvalla. Kottayam. Trivandrum. The three manuscripts he used were all Keralan and all in Malayalam script, and one of them does not begin until the tenth verse. He does not say whether he looked for others.

A teaching that the text itself says arose in Kashmir, composed in a Maharashtri literary register in the Tamil country, reached the twentieth century on Kerala palm leaf, in an alphabet neither its subject nor its author had used, with one of the three copies already worn out when it arrived.

It would be tidy to say the book survived by leaving. It did not only leave.

The two first lines

There is a second edition, and it is easy to miss alongside the first. In 1918, a year before Trivandrum, Mukunda Rama Shastri published the Maharthamanjari in the Kashmir Series of Texts and Studies. Its commentary is the shorter of the two. It was edited for the Research Department at Srinagar, where Ganapati Sastri’s rests on three Kerala palm-leaf manuscripts. The Trivandrum preface does not mention it.

Two editions, one year apart, from opposite ends of the subcontinent.

The Trivandrum and Varanasi editions, which stand in the southern line, open the work like this:

णमिऊण णिच्चसुद्धे गुळुणो चळणे महप्पआसस्स । गढइ महत्थमंजरिमिमिणं सुरहिं महेसराणंदो ॥ १ ॥

namiuna niccasuddhe guluno calane mahappaasassa | gadhai mahatthamanjarim imiṇam surahim mahesaranamdo ||

“Having bowed to the eternally pure feet of the teacher Mahaprakasha, Maheshvarananda strings together this fragrant Maharthamanjari.” (our translation, made from the Sanskrit chaya printed with the verse; “fragrant” follows the reading surabhim, on which the editions differ)

Mahārthamañjarī 1 #

The first word of the book is an act of obeisance, and the name it bows to is his teacher’s. The verb repays attention, though not quite in the way it first appears. Gadhai renders grathnati, he strings, and it is tempting to set that against composing, as if he were disclaiming authorship for something humbler. He is not. Grathnati is the ordinary Sanskrit verb for composing a book, and a grantha, the standard word for a book, is literally a thing strung together. The pun is not his invention. It is built into the language, and the title collects on it: a manjari is a blossom-cluster, separate flowers made one without any of them ceasing to be itself. Wenta reports that the Parimala states its aim in those terms, as the gathering of doctrinal flowers into a single spray. The book is named after its own method.

The Kashmir edition does not print those lines. It has the same verse in the same place, and several of the words differ: guruno where the southern text has guluno, canene for calane, a different form of the stringing verb, and the opening gerund printed as two words, the second of them corrupt. The Sanskrit paraphrase beneath agrees in substance in the two editions that print one here and still differs in detail, since Varanasi prints surabhi where the Kashmir edition prints surabhim, and the Kashmir paraphrase differs again at nityavishuddhau. That two paraphrases are independent enough to disagree is what makes the divergence in the Prakrit worth reporting at all.

Here the honest account has to stop short of the satisfying one. It is tempting to say that the Kerala line kept the Prakrit and the Kashmiri scribes lost it. That would be a verdict about manuscripts, and what has been compared here is two printed editions. Ganapati Sastri’s retroflex letters may be his own normalisation toward the Prakrit the grammarians describe. The Kashmiri forms may be an artefact of copying out of Sharada script, where letters of this shape are easily confused. Or guruno may simply be the older reading, or the Kashmirian one. Nothing in a printed page settles which, and no manuscript behind either edition has been collated for this account.

What can be said is that the two streams diverge at nearly every word of the first line, and that at this verse neither reading is obviously the degraded one. On the recensions as wholes the judgement is not ours to make and has been made: Cox, who has collated the manuscripts, calls the version that circulated in Kashmir truncated and inferior. Both carry the author’s prose synopsis, though only the Kashmir editor sets numerals against its entries, running them to seventy-one while his gathas stop at seventy. And the adjective the whole argument about Prakrit turns on, svochita against svapita, is a split between them; Cox reads it the Kashmir way too, emending against the Varanasi edition he works from. A book that travelled this far came down in two states, and the second state is why the first can be checked at all.

He wrote, in the clause quoted near the start, that the knowledge abides only in the mouths of the yogins, exceedingly hard to come by. He wrote it down.

Twice, as it turned out, and in two different shapes.