The power in the phoneme

Take the man this literature was written for, and grant him everything.

He did not find his formula in a book. He had it from a living teacher, face to face, with that teacher’s warrant behind it, and he keeps it to himself afterwards, because the tradition treats silence about it as part of the maintenance. He has done what he was told to do with it, for as long as he was told to. Nothing has happened.

The tradition has an answer for him, and the answer is not that he was insincere.

The answer is that the object in his mouth may itself be faulty.

Somewhere in the fifteenth chapter of the Kulārṇavatantra, one of the Kaula scriptures of the Tantric tradition, there is a list. It is sixty items long, and it is a list of things that can be wrong with a mantra.

Not sixty errors a reciter can make. The flaws sit in the formula itself, in the arrangement of its syllables, present before anyone opens their mouth. Item sixteen on that list is kīlita: tied, pinned, nailed down. A mantra in that condition is understood to be inert. You can have received it correctly, from someone entitled to give it, and hold it faithfully, and it will do nothing, because the thing itself is faulty.

So a mantra can be locked.

And the tradition does not leave the man holding a locked thing with no recourse. It has a name for the fault, which means somebody had to be able to recognise it. It has a sequence of named refinements that repair a defective formula, which means somebody had to be able to carry them out. And the person who can do either is the same person the formula had to come from in the first place: a teacher who holds it, is qualified to pass it on, and stands between the man and the object at every point where the object could fail.

The lock has a key-holder. That is the whole architecture of this chapter in one sentence, and the rest of the treatise is a refusal of it.

Which tells you something about what the tradition thought it was handling.

You do not write a sixty-item fault taxonomy for a prayer. Prayers do not have faults. They have sincerity, or they lack it. A sixty-item fault taxonomy is what you write for a machine.

This chapter is about that machine, built out at some length, because the corpus this treatise is following claims to work without any of it.

What a mantra is made of

Begin with the object.

In the Tantric literature a mantra is rarely a sentence. It is an assembly, and the tradition can name its parts and does so routinely, in a standard preamble attached to the formula when it is written down or handed on. There is a ṛṣi, the seer to whom it was first disclosed. There is a chandas, the metre it belongs to. There is a devatā, the deity who is its presiding power. There is a viniyoga, the application, the purpose this particular formula is designated for. There is a kīlaka, the wedge, which Bühnemann lists among the items the Tantric texts specify for a longer mantra, and which is named here and not described. The scope matters, and it is narrower than it looks. In Bühnemann’s account the phrase “for every mantra” governs the first four items, the seer, the metre, the deity and the application. The seed, the power and the wedge are introduced separately, for the case of longer mantras. The wedge is the item chapter 9 turns on, and it does not belong to every mantra. And there are, in most of them, one or more bīja syllables: seeds, single sounds carrying no dictionary meaning at all, which the tradition treats as the load-bearing elements of the whole structure.

Take the shape of that seriously for a moment. The meaningless part is the operative part. The syllables that could not be translated into any sentence are the ones the system regards as doing the work. The rest of the assembly, the seer and the metre and the deity and the application, functions as specification around them: where this came from, what it is for, whose power stands behind it, how it is to be fitted.

None of those components will be reproduced in this treatise. Not a bīja, not a formula, not a famous one offered as illustration. The operative line of this whole work is that it describes a technology and never supplies it, and this is the chapter where that line is under the most pressure. You are being told that a mechanism has parts. You are not being handed the parts.

The related discipline of nyāsa, the placement of sounds onto the body, sits alongside this and works on the same principle: sound treated as something that can be located, installed and made to bear. It is named here for completeness and left alone.

It arrives from a person

The second thing about the Tantric mantra is that you cannot pick one up.

It comes through dīkṣā, initiation, from a living teacher who holds it and is qualified to pass it on. That transmission is sealed. The formula is not published to be taken; where the compendia are printed, the working forms are frequently encoded. The tradition’s own position on this is blunt and it is on the record. Goudriaan and Gupta, surveying the literature, report the Kaulāvalinirṇaya at 1.20 and following warning the aspirant against working from a book: the fool who acts on what he has looked up in writing, without having had it from his teacher’s mouth, is destroyed like everyone else who tries it.

Then, having received it, you keep it. The privacy is not decorative. It is treated as a condition of function, and one of the ten refinements named later in this chapter, gupti, is concealment itself, listed as a procedure with a name, as though not saying it aloud were part of the maintenance.

So the object has a chain of custody. It has a named origin, a named holder, a named recipient, and a rule about disclosure.

That is four requirements before the mantra has done anything at all.

The literature of what goes wrong

Now the diagnostic material, which is the substance of this chapter and is unusually well documented.

The Kulārṇavatantra, at 15.65–69, gives sixty defects, and kīlita is the sixteenth of them. That is the anchor citation for the whole discussion, and its status has to be stated exactly. The Kulārṇavatantra has not been opened for this treatise. What is in hand is Gudrun Bühnemann’s report of it, and her table. Every Kulārṇavatantra citation in this chapter is therefore held at exactly one remove, in the same way the Śāradātilaka is. Bühnemann renders kīlita plainly as “tied”; the fuller gloss above is this chapter’s own, and a reader who goes to her page will find the shorter word.

It is not the only list. Five further compendia enumerate the defects of a mantra, and the whole set is small enough to print.

Two notes on that table before anything is read off it. The Bṛhattantrasāra is the one witness in the set that cannot be cited by verse at all. And the Nāradapurāṇa is given here in the numbering the standard Sanskrit electronic corpus uses; Bühnemann cites it bare, as 64.14–19. Her “Selecting and Perfecting Mantras in Hindu Tantrism,” published in 1991, is the standard modern study of the whole procedure and the route into all of it.

Three counts, then, across six compendia. Sixty once, forty-nine three times over, twenty-five twice.

All six belong in the count, and the two easiest to overlook, the Bṛhattantrasāra and the Kaulāvalinirṇaya, are the two that carry the most weight, because both are repetitions and repetition is what the argument turns on.

The forty-nine in the Nāradapurāṇa is not taken on anybody’s authority. The defects were counted for this treatise, verse by verse, in the Sanskrit, across the six verses that carry them, and they come to forty-nine. The passage then announces at 1.64.20 that the defects have been declared and their marks will follow.

One precision to carry away from that count and no further. In the Nāradapurāṇa’s list kīlita stands seventh. The number sixteen belongs to the Kulārṇavatantra’s ordering and travels nowhere with the defect. Anyone citing “kīlita, defect sixteen” from a Purāṇic page is citing the wrong book’s arithmetic.

They say two things at once.

They disagree, and the disagreement is worth more than any one list on its own. A fixed revealed checklist would be identical everywhere. Sixty flaws in a Kaula scripture and twenty-five in a later handbook is the signature of a live technical field with schools in it, practitioners arguing about categories, some lumping and some splitting, some traditions recognising faults that others did not think were faults.

Then the forty-nine that repeats, and it repeats three times. The same count stands in the Śāradātilaka and in the Bṛhattantrasāra, which are Tantric compendia, and in the Nāradapurāṇa, which is a Purāṇa: a different genre, a different audience, a different transmission history. Two literatures that were not doing the same job are carrying the same technical list. Which way that borrowing ran, this treatise does not know. That it ran is hard to avoid.

The material was argued over, and it travelled. This was contested expertise, and it circulated. People worked on it.

What the marks actually are

The Nāradapurāṇa does not stop at naming the defects. It gives their marks, at 1.64.20 to 57, and they are the most probative primary datum in this chapter and the next.

Here is the one that matters:

Nāradapurāṇa, Pūrvabhāga, 1.64.28a: māprā namāmi ca padaṃ nāsti yasminsa kīlitaḥ

That mantra in which the specified element is absent is kīlita. Which element, this treatise will not unpack, and the surrounding verses that specify the others stay where they are; the shape of the sentence is the whole of the finding here.

And the shape is a general classificatory rule about an arrangement of syllables. Something is missing from the string, and a formula with that gap is called pinned. The neighbours on the list are built the same way: at 1.64.22 a formula with a certain doubling is ruddha, obstructed, and yields its result only with great difficulty; at 1.64.25 one with a certain letter in three positions is badhira, deaf, and gives little for the effort. Cut, blind, burnt, immobilised, ashamed, deluded. The whole vocabulary is medical and mechanical at once, and every entry in it is a statement about how the syllables sit.

So kīlita is a diagnosis. In the tradition’s own most technical statement of the matter, a mantra is pinned the way a machined part is out of tolerance. That is what chapter 9 is built on.

The repair, named and not described

The diagnosis has a corresponding repair. A defective mantra is not simply discarded. The literature names the yonimudrā as one remedy, or else the sequence known as the ten saṃskāras, the ten refinements, which run: janana, jīvana, tāḍana, bodhana, abhiṣeka, vimalīkaraṇa, āpyāyana, tarpaṇa, dīpana, gupti. The names are fixed at Kulārṇavatantra 15.71–72. The Śāradātilaka explains them at 2.114–22.

Bühnemann’s own placement is worth following exactly, because a looser one is easy to reach. She gives the ten names from the Kulārṇavatantra and the explanations from the Śāradātilaka, and her single reference to Mantramahodadhi 24.107 governs one detail of gupti alone. So the ten refinements stand six verses below the sixty defects, in the same chapter of the same book. Diagnosis and repair are neighbours on the page.

There is a larger gap in the same place, and it is a gap in this treatise’s source balance rather than a judgement about where the best work sits. The whole Tantric apparatus in this chapter rests on one Western journal article. No Indian scholarship on it has been read directly: not Gopīnāth Kavirāj, not V. V. Dvivedī, not the Sanskrit commentaries, not any Indian edition of either text. The Nāradapurāṇa, which is read directly and carries the chapter’s best evidence, is the one thing here that does not run through a single foreign study, and it is not enough on its own. What would close it: an Indian edition of the Śāradātilaka with its commentary, read at 2.64–70 and 2.114–22, and a direct reading of the Indian scholarship on mantra defects.

Those ten names are the taxonomy, and the taxonomy is all this chapter will give. What each refinement consists of, how any one of them is carried out, in what order, by whom, with what, is in the literature, and as far as this treatise is concerned it stays there. The categories are the finding. The procedures are not on offer here at any word count.

But hold the shape of the sequence, because the shape is the argument. It opens with a word for birth and closes with a word for concealment. Something is brought into being, worked on through a series of named stages, and then hidden.

That is a manufacturing process with a final security step.

Dates, and what they do to the argument

The compendia are datable, and the dates stand in the opening lines of Bühnemann’s article.

Bühnemann places the Kulārṇavatantra between the eleventh and the fifteenth centuries of the Common Era. She fixes the Śāradātilaka through its commentary, Rāghavabhaṭṭa’s Padārthādarśa, of 1494. The Mantramahodadhi is Mahīdhara’s, of 1588 or 1589. The Tantrarāja, with its Manoramā commentary, is 1603 or 1604. To that can be added Goudriaan and Gupta, who assign the Kaulāvalinirṇaya, by Jñānānandagiri Paramahaṃsa, to the sixteenth century.

Set that against the argument above and the argument gets sharper. A Kaula scripture from somewhere in a four-century band carries sixty defects. A handbook of 1588 is still adding detail to a single one of the ten refinements. A compendium of 1603 carries twenty-five defects and not sixty.

Half a millennium of work, and no convergence at the end of it.

Two of the six remain undated here. Bühnemann does not date the Bṛhattantrasāra or the Nāradapurāṇa, and no entry for the Bṛhattantrasāra was found in Goudriaan and Gupta on the searches run. Those two dates are outstanding and the chapter does not lean on them.

A technology with a maintenance schedule

Assemble what has now been established, and read what it implies.

A mantra has to be selected, because there are many and they are not interchangeable, and the compendia contain elaborate procedures for determining which formula suits which practitioner. It has to be examined, because it may carry one of sixty defects and the defect will not announce itself. It may have to be purified, through a named sequence of refinements. It must be transmitted by someone qualified, and concealed by whoever receives it.

Every one of those is a place where the thing can fail. Wrong selection, undetected defect, incomplete refinement, unqualified transmitter, broken secrecy. The system is candid about all of them. It is a technology with a maintenance schedule, and it operates under something close to a licensing regime.

Now the part that matters for this treatise.

Every one of those requirements is also a gate.

Consider who stands on the far side of each. The mantras are in Sanskrit, and so is the entire diagnostic literature that governs them, which means the language is a gate. Transmission requires initiation, which is a gate held by another person, and the Kaulāvalinirṇaya passage above exists precisely to keep it shut against readers. You need physical access to a qualified teacher, which in practice has meant geography, money and time. And the diagnostic apparatus is a written literature, which means literacy, or a teacher who has it on your behalf.

That is four filters standing between an ordinary person and a working mantra, and they compound.

The tradition has its own account of why the gates exist, and it deserves to be stated rather than skipped. A powerful instrument handed to an unprepared person is held to harm the person; the qualification is protective; the secrecy protects the formula and the holder both; the teacher exists because a text cannot correct you and a person can. Serious practitioners have argued this for a very long time and it is not a foolish argument. How far the eligibility rules of these traditions also functioned as social gatekeeping is a live scholarly dispute, and this treatise has not read the eligibility literature directly, so it takes no position and makes no claim.

What can be said without adjudicating anything is narrower and harder to argue with. A gate built to protect is still a gate. Whatever its purpose, it has an effect, and the effect is that some people get through and some do not.

Then look at what Tulsīdās describes

Now set the Śābara corpus beside all of that.

Chapter 1 established what the famous verse in the Bālkāṇḍ says and what it is doing there. Return to the middle of it:

अनमिल आखर अरथ न जापू।

anamila ākhara aratha na jāpū Ill-matched letters; no settled meaning, and no japa.

Rāmcaritmānas, Bālkāṇḍ 1.15.3 · first half of the second line only, not a complete chaupāī #

Anmil ākhar: ill-matched letters, syllables that do not fit together properly. Arath na jāpū: this is the half-line the commentators fight over, and chapter 13 fights over it again.

The standard Hindī compendium of Mānas commentary, Mānas-pīyūṣa, edited by Añjanīnandan Śaraṇ and published by the Gita Press at Gorakhpur in 1968, sets out the construction of that half-line, in its own phrase, in several ways. Its base gloss takes the letters as ill-matched and says they have neither correct sense nor japa, then, inside the same sentence, offers an alternative in which they have no sense and japa alone is what counts. Three further construals follow, each attributed on the page by a siglum rather than a name. Two of them put the mismatch between the letters and the sense and leave the meaning standing, differing over whether what is missing is a prescribed japa, the count and rule that other mantras must satisfy before they yield anything, or whether repetition alone is what brings the fruit. The last reads the letters as simply unjoined, in breach of the grammatical rules the commentary lists.

Four readings, then, counting the base gloss with its embedded alternative as one and the three lettered construals as three. Chapter 13 counts them the same way, attributes them the same way, and takes the last of the three.

It is tempting to quote the blunt reading alone, the one that gives the letters no proper meaning and no japa either. That is one of the four and it is reported here as one of four. The commentarial tradition has not settled this line, and printing one construal as the gloss flattens a live argument in the very source being cited.

Two things about the citation itself. Mānas-pīyūṣa numbers this verse dohā 15, chaupāī 5–6, while this treatise cites Bāl 1.15.3 throughout; the editor explains in the edition itself that he revised the dohā-numbering after collating old manuscripts, so the two conventions are both live and a reader has to be told which one is in use. And the sigla behind the three lettered construals are expanded in the compendium’s own key, which was not legible in the scan worked from, so none of them is attributed here to anybody by name. They are attributed to Mānas-pīyūṣa reporting unexpanded sigla, and chapter 13 says the same of them.

Read the verse against the machine just described, item by item, and the correspondence is exact and inverted.

The Tantric mantra is an assembly of syllables whose arrangement can be defective; these are stated to be ill-matched from the start, and no repair sequence is prescribed for them. The Tantric mantra has a viniyoga and a defined application; these have no settled sense. The Tantric mantra has a counted, ruled recitation; these have none, on the reading this treatise follows. The Tantric mantra has a seer, a metre, a presiding deity and a chain of custody through initiation; the verse names no ṛṣi, no chandas, no dīkṣā, and no teacher.

Selection, examination, purification, qualified transmission. Not one of the four is present.

There is a commentator on this same page who saw the whole structure and put it as an accounting. Mānas-pīyūṣa, in its third note on the line, reports the frame: a mantra fails to bear fruit if any one of four conditions goes unmet, the letters garbled, the sense absent, the puraścaraṇa not performed by rule, the japa not done by rule. The śābara mantras satisfy none of the four and bear fruit anyway, by the pratāp of Maheśa alone. The commentator adds that the effect is instant.

Four defects and no repair, and the thing works. That is the tradition’s own scholar stating this treatise’s problem, in print at Gorakhpur, long before this treatise arrived at it.

And then the second half of the line:

प्रगट प्रभाउ महेस प्रतापू॥

pragaṭa prabhāu mahesa pratāpū Their power is manifest, by the majesty of Maheśa.

Rāmcaritmānas, Bālkāṇḍ 1.15.3 · second half of the second line only; the chaupāī number ॥३॥ stands after it in the source and is dropped here because only the fragment is quoted #

Their power is manifest, out in the open, by the majesty of Maheśa.

That is the entire warrant. No mechanism is supplied, and none of the missing components is explained away. The verse does not claim the meaning is hidden from the uninitiated. It does not claim the syllables are secretly well-formed. It does not claim there is a vidhi known only to a lineage. It concedes the defect in full and grounds the result in a person.

The claim is stranger than the slogan

It would be easy, and wrong, to summarise all of this as democratisation. The Nāths broke the Sanskrit monopoly; the vernacular came for the mantra; power to the people. That reading is available, it is popular, and it flattens the thing until it is no longer interesting.

Look at what is actually being asserted.

The Śābara claim does not say the apparatus is mistaken. Nothing in this material argues that the sixty defects are fiction, that the ten refinements are superstition, that initiation is a racket or that the Kulārṇavatantra got it wrong. The Tantric machine is left standing, entirely intact, unrefuted.

The claim is that in one specific case the machine is unnecessary.

And the reason offered has nothing to do with any hidden virtue in the syllables. The reason is authority, and the verse is precise about whose.

Hara and Girijā made them. Śiva and Pārvatī, the two of them, and the grammar settles it rather than the piety. The relative pronoun the chaupāī uses for the makers is jinha, which is plural and oblique; a single maker in that construction would have taken jehiṁ, and the poet did not write jehiṁ. The reading is reinforced twice over, by so umesa in the next chaupāī and by the favour of hara gauri in dohā 15. Chapter 1 sets this out. It is worth restating here because this is the paragraph where the temptation to drop the goddess is strongest, and the popular literature drops her constantly.

What the verse does put in the singular is the warrant. The power stands out in the open by the pratāpa of Maheśa. So the exemption is not earned by the mantras, and it is not earned by the practitioner. It is granted from outside, by someone else, and the corpus can only point away from itself to say so.

That is a much stranger position than “the Nāths democratised mantra.” It is a claim of licensed exception inside a system whose rules are conceded to be real. It leaves the licensing regime standing and asserts one permit that nobody has to qualify for, issued by the highest possible authority, verifiable by nothing except the fact that people say it works.

Which raises the obvious question, and it is the question chapter 9 exists to answer. If the Śābara corpus is exempt, some text somewhere must say so. The doctrine of pinning and unpinning is real, it is in the primary literature, and the popular account of it has a specific and testable shape. Chapter 9 goes looking for the passage that grants the exemption, and it finds one, in a language and a genre nobody was searching.

The apparatus described here is not being refused by the Śābara material. It is simply not present in it, and the corpus does not explain itself.

Tulsīdās grants the defect and asserts the power, and names the authority it runs on.

He names no procedure. There is not one to name.