The Types of Tantra

Bhaskararaya reached the verse that closes the list of the sixty-four Tantras, and turned aside to deal with somebody. He was at work on a commentary of his own, on the Nityāṣoḍaśikārṇava, a Tantra of the Goddess, and had just finished establishing, against anyone who might doubt it, that these sixty-four books are śāstra, authoritative scripture, of the nature of the Veda itself. Then he stopped.

Somebody, he wrote, has babbled in a commentary on the Saundaryalaharī, the hymn to the Goddess that tradition assigns to Shankara, that these Tantras are avaidika, outside the Vedic path. Disregard it. It is the prattle of either a deceiver or a deluded man.

The somebody had been dead for two centuries. The modern edition’s index of names identifies him: Lakshmidhara, who commented on the Saundaryalaharī in the early sixteenth century at the Odishan court of Prataparudra Gajapati Deva, and who had written of these same sixty-four Tantras that they “are causes of the deluding of the worlds and are productive of ruin, because they stand far from the Vedic path.”

Two commentators, two centuries apart, in print, on whether the sixty-four Tantras are scripture at all.

And a few lines before the insult, Bhaskararaya had already conceded something stranger. Sarveṣām eva mate catuḥṣaṣṭisaṃkhyāsampattir durupapādā. On everyone’s view, he says, making the number come out to sixty-four is hard to establish. He then surveys how the commentators divide the titles, and the disagreement begins at the first entry. Shivananda’s Ṛjuvimarśinī reads it as two Tantras, the Mahāmāyā and the Śambara; Lakshmidhara reads it as one book called the Mahāmāyāśambara; a third commentary names no title there at all.

The ablest defender of the sixty-four Tantras cannot get the count to sixty-four, and cannot get agreement on item one.

That is what the question “what are the types of Tantra” opens onto. There is no single taxonomy. There are many, each datable, each with a verse address, each drawn by a school, and most of them a ranking wearing the clothes of a list.

What the question is asking

The general question, what Tantra is and what it is for, is answered at What is Tantra?. This page takes up the narrower one. Begin with an answer that is true enough to mislead. Āgama, Tantra and Saṃhitā are often presented as the three types of tantric scripture. They are closer to labels of provenance than to classes of practice. Scriptures of the Śaivas, the devotees of Shiva, tend to be called Āgama; those of the Śāktas, who worship the Goddess as ultimate, Tantra; those of the Vaiṣṇavas, the devotees of Vishnu, Saṃhitā. The tendency is real, and it leaks. Major Śaiva scriptures carry “Tantra” in their titles, the Svacchanda, the Netra, the Mālinīvijayottara, and Pāñcarātra texts are also called Āgamas. André Padoux lists tantra, āgama, saṃhitā and sūtra as near-interchangeable names for one kind of literature. The terminology is treated in full at What are the Āgamas and Tantras?, and the relation of this revelation to the Vedic one at Āgama and Veda.

The real classifications lie elsewhere, and they have Sanskrit names and verse numbers. The srotas, streams, sort scriptures by the mouth of God that spoke them, while the pīṭhas, seats, sort them by whose cult they teach; the numbered canons count them, in tens and eighteens and sixty-fours and twenty-eights and hundreds-and-eights; the āmnāyas are transmissions reckoned by compass direction, and the ācāras are modes of conduct.

No two of them agree.

What follows walks each in turn, and asks of each the question that explains its shape: who drew this, and where did he put himself?

The oldest division

The earliest surviving Śaiva Tantra is the Niśvāsatattvasaṃhitā, whose oldest layer Alexis Sanderson assigns to the period 450 to 550 of the Common Era. A later hand attached an introductory book to the front of the collection, and there, at Niśvāsamukha 1.26 to 1.27a, the sage Nandikeśvara sets out the whole of valid Indian religion in five levels.

Laukika, the worldly. Vaidika, the Vedic. Ādhyātmika, which here means the highest reach of Brahmanical religion as a Śaiva saw it, the Sāṅkhya and Yoga systems whose tattva scheme, their ladder of the constituents of reality, the Śaivas absorbed wholesale. Then atimārga, the outer path. Then the path called mantra.

This is the oldest datable classification in the corpus, and its ranking is not hidden. Adherents of the first three, on Christopher Wallis’s reading, attain the goals their own religions promise and stop short of definitive liberation, which the text reserves to Śaivism in its two divisions. A Śaiva book sorts the religious world and sets Śaiva revelation at the summit of it. Most of what follows does the same, and the exceptions are instructive enough to have a section of their own at the end.

From the fourth and fifth of those levels comes the division that has organised the modern study of the subject.

The outer path, the path of mantras, and a third branch

Sanderson’s formulation of 1988 is the one most later accounts work from: the teaching of Shiva divides between two great branches, the Outer Path, Atimārga, and the Path of Mantras, Mantramārga. The Atimārga, he wrote, is “accessible only to ascetics.” The Mantramārga “is open both to ascetics and to married home-dwellers.”

That is a scholar’s vocabulary laid over a distinction the tradition was already drawing. Abhinavagupta states the twofold himself, in the thirty-seventh chapter of the Tantrāloka, around the year 1000: there are two authoritative revealers, Śrīkaṇṭha and Lakuleśvara, and this teaching is of two currents. Śrīkaṇṭha speaks the Mantramārga. Lakuleśvara, who is Lakulīśa the Pāśupata founder, speaks the older path. The structure is the tradition’s own; Sanderson supplied the terminology from the Niśvāsa.

The Atimārga belongs in a page about the types of Tantra because it is the thing that is not Tantra. Older, ascetic-only, pre-tantric, it has three divisions in Sanderson’s later work where it had two in 1988: the Pāñcārthika Pāśupata, derived from Lakulīśa, the Lākula, and the Kāpālika or Somasiddhānta, added as a third only after 1988, in which paper the word Kāpālika had been reserved for a Mantramārgic group instead. The skull-topped staff and the cranial alms-bowl, the emblems by which these ascetics are known in the literature, belong to the second and third divisions in common.

That revision matters more than a footnote. By 2014 Sanderson describes initiatory Śaivism as comprising three systems rather than two: the Atimārga, the Mantramārga, and those of “the predominantly Śākta Kulamārga,” the Kaula path, which he now derives from the third Atimārga division; in 1988 the Kaula had been a reform movement inside the Mantramārga canon. Three branches, where the 1988 map had two.

He keeps the twofold as well, and his reason is worth following closely. Atimārga and Mantramārga are the primary divisions of the Śaiva scriptures, because that is the division the insiders drew, in the fivefold of the Niśvāsamukha. And that fivefold has no slot for the Kulamārga at all.

So the tradition’s map has two branches and the scholar’s map has three, and the difference is no disagreement about facts. It is the gap between a classification drawn from inside a position and one drawn from outside all of them. That gap is the subject of this article.

One correction belongs here, because it concerns the word itself. Peter Bisschop has shown that atimārga is a Mantramārgic coinage: no Pāśupata, Lākula or Kālamukha scripture uses it of itself. What those ascetics did call their path was the atyāśrama, the state beyond the four life-stages, and the new term was built on that, so well built that Bisschop concludes whoever coined it “was closely familiar with the tradition.” It was not a slur thrown from a distance. He finds one place where a text uses Atimārga of itself, an unpublished twelfth-century Vārāṇasīmāhātmya in the Kaiser Library at Kathmandu, in a purely Pāśupata and Kāpālika setting with no mention of the Mantramārga. His essay appears in a volume published in Sanderson’s honour. This is refinement inside a school.

The twofold settles the outer boundary. The next scheme divides the inside.

Five streams from five faces

The most comprehensive classification of the Mantramārga’s own scriptures, and the one that became standard, divides them into five streams of revelation issuing from the mouths of the five faces of Sadāśiva, Shiva in the five-faced form in which he is held to utter scripture.

From the upper, upward-gazing Īśāna face come the scriptures of the Siddhānta, divided, as Sanderson puts it, artificially into two sub-canons, the ten Śivabhedas or Shiva-division Āgamas and the eighteen Rudrabhedas or Rudra-division. That is a Saiddhāntika scheme speaking, and it seats the Siddhānta highest. From the four faces that look to the cardinal directions comes the rest: the Vāmatantras from the north-facing Vāmadeva, teaching the cult of the four Sisters and their brother Tumburu; the Dakṣiṇatantras from the south-facing Aghora, teaching cults of Bhairavas and goddesses; the Gāruḍatantras from the east-facing Tatpuruṣa; the Bhūtatantras from the west-facing Sadyojāta.

The taxonomy is generated from the god’s own head. The shape of the canon is the shape of the deity, and the deity in question is the one the Siddhānta worships.

A long list of texts carries the scheme, among them the Śrīkaṇṭhī, the canon-list prefixed to one recension of the Kālottara, Vairocana’s tenth-century manual, the Mṛgendra and the Kāmika. One late Saiddhāntika source sorts the five by what they are for instead of by who spoke them. An anonymous commentary on Umāpatiśivācārya’s Śataratnasaṃgraha assigns liberation to the Siddhānta, the removal of poisons to the Gāruḍa stream, exorcism to the Bhūta, the destruction of enemies to the Bhairava, and to the Vāma stream sarvavaśīkaraṇa, the subjection of everything to one’s will.

Take that witness for what it is, a partisan list from a tradition arguing for its own supremacy. It still disposes of a common idea, which is that the left-hand stream is the gentle end of the spectrum. The mildness in the sources belongs to the Vāmadeva face, not to the function of the books that issue from it. A classification that files universal domination under the left hand is describing something other than a soft option.

Two of the five streams barely survive as canons.

The Bhūtatantras dealt with possession by bhūtas, pretas and piśācas, three classes of hungry and malevolent dead. The Gāruḍatantras dealt with snakebite and poisoning. Of this extensive literature, Sanderson writes, nearly all seems not to have reached us; the only survivors are two texts in Nepalese palm-leaf manuscripts claiming to be parts of a work called the Trottala. Michael Slouber has recovered what can be recovered, chiefly from a pre-eleventh-century digest called the Kriyākālaguṇottara and from material quoted into other tantras and the Purāṇas. His argument cuts against the easy description of this stream as the medical one: the Gāruḍatantras teach a complete tantric system with a soteriological dimension, in which the healer’s identification with mantra-deities is what makes a treatment work. The division between medicine and magic is ours.

The Vāma stream fared no better in India. Its cult of the four sister-goddesses with Tumburu faded at home, though it travelled: a Cambodian inscription of 1052 names four Vāma scriptures as the four mouths of Tumburu, and the ninth-century Khmer state cult knew the tradition well.

Other streams were named in the classical lists and left even less. The eighty-five titles of the Saura tantras are a category with no text behind it at all.

So a canon reported as fivefold is one crowded stream, one full one, and three in various states of ruin. The classification outlived most of what it classified.

It also outlived the agreement that there were five.

Two maps that each put the other at the bottom

The five-stream model became standard, which has hidden the fact that it was fought over while it formed. Two texts draw the streams in different shapes, and between them they frame the whole argument of this article.

The first is the Brahmayāmala, also called the Picumata, an early Śaiva scripture of the Bhairava cults and of the yoginīs, the fierce female powers at their centre, its old core placed by Shaman Hatley at perhaps 650 to 700. Its thirty-ninth chapter is a determination of the streams of revelation. It gives three streams, not five, attaches each to one of the three Śaktis of the great God, and grades them by the guṇas, the three constituent qualities of everything manifest.

Dakṣiṇa, the right, goes with sattva, lucidity, and is śuddha, pure. Vāma, the left, goes with rajas, energy, and is miśra, mixed. Madhyama, the middle, goes with tamas, inertia, and is described by a compound meaning stained with the whole mass of impurity.

Then the chapter says which books belong where. The right stream holds the bhairavatantras, subdivided by four seats, those of vidyā, a female mantra-deity, of mantra, of mudrā, ritual gesture, and of maṇḍala, the drawn diagram. The left holds the vāmatantras. And the middle, the tamas one, holds the siddhāntatantras: the Vijaya, the Niśvāsa, the Svāyambhuva, the Vātula, the Vīrabhadra, the Raurava, the Makuṭa, the Vīreśa.

The Śaiva Siddhānta, which the five-stream scheme crowns, is here the stream of inertia. And the Niśvāsa, the oldest surviving tantra, sits in the lowest of the three.

The chapter is gentler about it than that sounds, and the qualification is worth having. It likens the ascent from the middle stream through the left to the right to husking rice, then removing the bran, then cooking the clean grain, with the soul passing through the three over successive births. A ranking figured as stages, not as a scrapheap. Sanderson, who edits the simile, still reads the arrangement as a hierarchy in ascending order designed to assert non-Saiddhāntika superiority, and both things are true of it at once. The same chapter adds a fourth and lower stream, the adhaḥsrotas, holding alchemy, herbs, exorcism, snakebite cure and, placed last, the Vaiṣṇava Pāñcarātra scriptures, and then elaborates its three into ten.

P. C. Bagchi printed the Sanskrit of the guṇa passage from a Nepalese manuscript in 1939, Hatley has edited the chapter, and Sanderson records the same scheme independently, reading it as a modification of the fivefold rather than a precursor to it. Hatley’s judgment runs the other way: the text’s model looks archaic, because it shows no awareness of the five-stream scheme that became normative. That disagreement has not been settled, and it cannot be settled by a date, because no attestation of the fivefold is dated earlier than the Brahmayāmala. One further caution for anyone following this up: English summaries circulating online, Wikipedia’s among them, reverse the vāma and madhyama assignments. Bagchi’s printed text gives the order above.

Now the mirror. Sarvātmaśambhu, in the Siddhāntaprakāśikā, returns the compliment with interest. Gāruḍadakṣiṇavāmabhūtāni tv adhaḥsrotāṃsi. The Gāruḍa, the Dakṣiṇa, the Vāma and the Bhūta tantras are the lower streams, and the name Mantramārga belongs to the Siddhānta alone. A verse of the Kāmika states the Saiddhāntika order from the top down in the same spirit: the Siddhānta is the Mantratantra, the Atimārga below it, the ādhyātmika lower, the Vedic lower still.

Set the two side by side and the shape of the whole subject shows: a Bhairava text that consigns the Siddhānta to the stream of inertia, and a Saiddhāntika text that answers by flushing the Bhairava tantras into the drains, both of them working in the shared vocabulary of the five streams and both presenting the result as a determination of what revelation actually looks like. Neither is describing a shelf.

The next scheme is the one most readers have met, and it does its ranking inside a single verse.

Ten, eighteen, sixty-four

Abhinavagupta, Tantrāloka 1.18: daśāṣṭādaśavasvaṣṭabhinnaṃ yacchāsanaṃ vibhoḥ, tatsāraṃ trikaśāstraṃ hi tatsāraṃ mālinīmatam. The teaching of the Lord, divided into ten, into eighteen, and into eight times eight; the essence of that is the Trika teaching, and the essence of that is the doctrine of the Mālinī, which is the Mālinīvijayottara, the one scripture he is about to expound.

Read the halves in sequence. In the first he classifies the entire revelation. In the second he puts his own school above all of it, and his own text above his school. The classification and the self-placement are one sentence.

Note what that does to the sixty-four. They are not the summit of this scheme. Two things stand above them, and Abhinavagupta says elsewhere exactly how they function: initiation into the sixty-four Bhairava tantras is what you give to someone already initiated into the ten and the eighteen. A rung, explicitly.

The source behind the verse is a text Sanderson calls the Śrīkaṇṭhī-Srotobheda and Jayaratha the Śrīśrīkaṇṭhī. No copy of it survives as a book. What survives is a long excerpt copied into a Kashmiri ritual compendium, the Nityādisaṃgrahapaddhati, from which Jürgen Hanneder edited it in 1998; Jayaratha quotes a stretch of fifteen verses from it. Commenting on 1.18 in the middle of the thirteenth century, Jayaratha first glosses the arithmetic. Vasu in the verse means the Vasus, a group of eight deities, so eight multiplied by eight gives the sixty-four Bhairava divisions. Then he quotes the Śrīkaṇṭhī, and the quotation gives the sixty-four as eight aṣṭakas, eight groups of eight:

Bhairava, Yāmala, Mata, Maṅgala, Cakrāṣṭaka, Śikhāṣṭaka, Bahurūpa, Vāgīśa. The verse numbers the last two itself, “the seventh” and “the eighth,” so the order is fixed in this recension and cannot be shuffled. It is frequently printed wrongly.

Look at what sits inside group number six. The Śikhāṣṭaka holds the Bhairavī, the Vīṇā, the Vīṇāmaṇi, the Saṃmoha, the Ḍāmara, the Atharvaka, the Kabandha and the Śiraścheda. Sanderson identifies four of the titles in this canon as Vāma works taken into it. The stream that had a face of Shiva to itself has been folded into one group of eight inside a Bhairava canon. His description of the move is exact: the later classifications “submerge the Vāma within the Dakṣiṇa,” reflecting, he proposes, a time when the Vāma had gone marginal, and the sixty-four are listed, in his phrase, with evident artificiality.

How artificial shows in the witnesses. The eight-ogdoad scheme survives in a second source, the forty-fourth chapter of the first book of the Jayadrathayāmala, whose title announces the determination of Sadāśiva’s eight eights. The two witnesses agree on the structure. On the question of which Vāma works were absorbed into it, their lists share not one member.

The site treats the lower two divisions of this scheme at The 28 Śaiva Āgamas, and the Śākta set of sixty-four, which no scholarship has yet set beside this one, at The 64 Śākta Tantras. Abhinavagupta’s life and the shape of his synthesis are at Who was Abhinavagupta?, and the philosophy he built on it at What is Kashmir Shaivism?.

A second classification was running alongside this one, drawn by the other side.

The two seats

The pīṭha scheme divides the non-Saiddhāntika scriptures, leaving the Bhūta and Gāruḍa streams out of account, into two collections called seats. The Mantrapīṭha holds the Bhairava-centred texts, mantra being grammatically masculine, and is headed by the Svacchanda, the only one of its eight Bhairava texts that survives entire. The Vidyāpīṭha holds the Goddess-centred ones, vidyā being feminine and meaning, in these sources, a female mantra-deity. Texts of the Vidyāpīṭha generally assign themselves, in their own colophons, to the Vidyāpīṭha within the Bhairava stream. The site sets the scheme out in full, contents and all, at The two seats.

What belongs here is Sanderson’s observation about who drew it. The division into Bhairava-centred and Goddess-centred texts is made in works that themselves belong to the second category, so it articulates a Śākta perspective on the canon. And Abhinavagupta, characteristically, works with four seats where the Jayadrathayāmala has two, the same four the Brahmayāmala had already named, and puts them in ascending order: Maṇḍalapīṭha, Mudrāpīṭha, Mantrapīṭha, Vidyāpīṭha, with the Siddhayogeśvarīmata highest inside the Vidyāpīṭha and the Mālinīvijayottara its essence. The same book wins again.

Which is a convenient place to total up.

Numbers that will not add up

Here is the numbered apparatus, set against what can be counted.

Twenty-eight Śaiva Āgamas. Ten Śivabhedas and eighteen Rudrabhedas, with subsidiary scriptures, upāgamas, bringing the traditional South Indian total to two hundred and seven or two hundred and eight. Dominic Goodall counts six of the twenty-eight as demonstrably early: the Kiraṇa, the Niśvāsa, the Rauravasūtrasaṅgraha, the Svāyambhuvasūtrasaṅgraha, the Pauṣkarapārameśvara and the Parākhya. His criterion is citation rather than manuscript date. A surviving listed Siddhānta counts as early when the Kashmirian exegetes of Rāmakaṇṭha’s lineage, who worked in the tenth and eleventh centuries, quote it. Three major Saiddhāntika scriptures with early witnesses and commentaries are absent from the list altogether: the Mṛgendra, the Mataṅgapārameśvara and the Sarvajñānottara. And of the South Indian scriptures that now carry the canonical titles, Sanderson writes that while their titles do stand in the old lists, they “are unlikely to be, or are certainly not, works of that period”; where an early text under one of those titles is quoted by the early commentators, the passages preserved do not establish identity with the southern books of the same name.

One hundred and eight Pāñcarātra Saṃhitās. The named titles across all sources exceed two hundred, and a fraction survive. The cleanest proof that these numbers are theological is Surendranath Dasgupta’s observation that the Mārkaṇḍeyasaṃhitā speaks of one hundred and eight Saṃhitās and then gives a list of ninety-one. The corpus is at The Pāñcarātra Saṃhitās.

Sixty-four Śākta Tantras. Several mutually divergent enumerations survive, and the site lays them beside one another at The 64 Śākta Tantras. Akṣaya Kumāra Maitra, introducing the Kulacūḍāmaṇi in 1915, observed that several of the sixty-four could not be identified with certainty and that its list diverged from the Vāmakeśvara’s by about nine titles. Two of the lists carry the point further. The Sarvollāsa cites the Ṭoḍalatantra’s sixty-four, a list noted even there to take in some modern Tantras. And the Mahāsiddhasāratantra drops doctrine as the principle of sorting and goes by region instead; a related report of that scheme gives sixty-four in each of three territories, a total of one hundred and ninety-two.

Twenty-three texts in the two seats of the non-Saiddhāntika canon, by the tradition’s own reckoning: eight in the Mantrapīṭha and fifteen in the Vidyāpīṭha. Six of the twenty-three are titles that may never have denoted real books. Of two of them, the Viśvādya and the Vidyābheda, Wallis says they are scriptures still to be revealed at the end of the present age; Sanderson, citing the exegetical record, says something adjacent and colder, that the evidence draws their reality into question.

The verse that made the number sixty-four famous deserves reading with its own instability showing. Saundaryalaharī 31:

चतुष्षष्ट्या तन्त्रैः सकलमतिसन्धाय भुवनं स्थितस्तत्तत्सिद्धिप्रसवपरतन्त्रैः पशुपतिः । पुनस्त्वन्निर्बन्धादखिलपुरुषार्थैकघटना- स्वतन्त्रं ते तन्त्रं क्षितितलमवातीतरदिदम् ॥ ३१ ॥

catuṣṣaṣṭyā tantraiḥ sakalam atisandhāya bhuvanaṃ | sthitas tattatsiddhiprasavaparatantraiḥ paśupatiḥ | punas tvannirbandhād akhilapuruṣārthaikaghaṭanā-svatantraṃ te tantraṃ kṣititalam avātītarad idam

“Having worked upon the whole world with the sixty-four Tantras, each of them dependent on the production of its own particular siddhi, its own attainment, Paśupati, the Lord of Creatures, rested. Then, at your insistence, he brought down onto the earth this Tantra of yours, self-sufficient, alone able to accomplish every human aim.”

Saundarya Laharī 31 #

The word rendered “worked upon” is atisandhāya, whose sense runs to having outwitted, or deluded. A great many manuscripts read instead abhisaṃdhāya, having controlled, and by the count in W. Norman Brown’s apparatus they are the majority, by something like two to one. Brown printed abhi-, and said in print why the choice is not a quiet one. It “generates a certain amount of heat among adherents to the Śrīvidyā,” and users of his reading had seemed to him to regard it as close to sacrilege to say that Shiva had been practising deceit.

The reading printed above is the other one. It is the text of the South Indian commentarial line and of every printed edition that carries Lakshmidhara, which makes it the vulgate of the living tradition and the minority of the manuscript base at the same time.

So the verse on which the number rests divides its own witnesses at exactly the point where it says what the sixty-four were for. And on either reading it ranks them: the sixty-four are the partial instruments, superseded by one complete Tantra of the Goddess.

As for the charges Lakshmidhara reads into that verse, they have a provenance, and it is not Śākta.

The scriptures made to delude

There is a classification of the tantric canon drawn from entirely outside it.

In the fifteenth chapter of the Kūrmapurāṇa, Vishnu proposes to Shiva that between them they make scriptures to delude those who stand outside the Veda. Vimohanāya śāstrāṇi kariṣyāmo. Rudra then makes them, cakāra mohaśāstrāṇi, and the text names what he made: kāpālaṃ nākulaṃ vāmaṃ bhairavaṃ pūrvapaścimam, pañcarātraṃ pāśupataṃ tathānyāni sahasraśaḥ. The Kāpāla, the Nākula, the Vāma, the Bhairava of eastern and western kinds, the Pāñcarātra, the Pāśupata, and others by the thousand.

Every one of those is a real division of the tantric literature. The Vāma and the Bhairava are two of the five streams. The Pāśupata and the Nākula are two of the three Atimārga divisions. The Pāñcarātra is the Vaiṣṇava canon. The list is accurate, it is in the canon’s own vocabulary, and its organising principle is that the books exist in order to mislead.

The same Purāṇa does it again elsewhere, grading the Kāpāla, the Pāñcarātra, the Yāmala, the Vāma and the Jain scriptures as tāmasa, of the quality of inertia, with an end to match. Grading scripture by the three qualities was a standard Purāṇic instrument; the Matsyapurāṇa turns it on the Purāṇas themselves. So the Brahmayāmala’s graded streams are an instance of a wider habit and not a peculiarity, with one difference that matters: in the Purāṇas tāmasa is a terminus, and in the Brahmayāmala it is a rung you climb off.

R. C. Hazra, working through the Purāṇic records in 1940, supplied the Varāhapurāṇa’s version of the device with a motive: a Pāñcarātra tradition losing ground to the Āgamic Śaivas had Rudra himself denounce the ascetics who stood outside the Veda. And the device did not stay in Purāṇic hands. Sanderson notes a Saiddhāntika commentator reaching for one of these very passages to use against the non-Saiddhāntika Śaivas: a blanket condemnation of tantric scripture, repurposed by a tantric school against its rivals.

Now go back to the quarrel this article opened with.

Lakshmidhara does not merely prefer the harsher reading of the verse. He writes that the sixty-four Tantras deceive even those who know them thoroughly; that they aim at nothing beyond worldly attainment; that they stand far from the Vedic path. He reserves Vedic standing to five Āgamas alone, the Śubhāgamapañcaka, and identifies the practice taught there as samayācāra, conduct according to the covenant. Those are the Purāṇic charges, in Purāṇic vocabulary, inside a commentary on a hymn to the Goddess.

Which is why Bhaskararaya does not answer calmly. A philological dispute about one word would not need that much heat. He is refusing a frame, and the frame is older and larger than his opponent: the sixty-four are śāstra of the nature of the Veda, he insists, because the Tantras are a complement of the Upaniṣads. The insult is proportionate to what was being claimed.

Two positions, then, on one canon, taken two centuries apart by two commentators on the same hymn. It is a device for deluding those outside the Veda. It is scripture of the nature of the Veda. Both are classifications, and the second had to be argued because the first was available.

The ladder, and the word that turns around

The schemes so far have sorted books. This one sorts conduct, and it is reported wrongly more often than any of them.

It is a ladder of seven, in the second chapter of the Kulārṇava Tantra, at verses seven and eight, which can be read in Tārānātha Vidyāratna’s edition with the numerals printed beside the lines:

sarvebhyaś cottamā vedā vedebhyo vaiṣṇavaṃ param vaiṣṇavād uttamaṃ śaivaṃ śaivād dakṣiṇam uttamam dakṣiṇād uttamaṃ vāmaṃ vāmāt siddhāntam uttamam siddhāntād uttamaṃ kaulaṃ kaulāt parataraṃ na hi

Higher than all the rest are the Vedas; higher than the Vedas, the vaiṣṇava; higher than the vaiṣṇava, the śaiva; higher than the śaiva, the dakṣiṇa; higher than the dakṣiṇa, the vāma; higher than the vāma, the siddhānta; higher than the siddhānta, the kaula, the way of the Kula; and higher than the kaula there is nothing at all. — Kulārṇava Tantra 2.7–8

Three things about that passage, each of which the popular account loses.

First, dakṣiṇa is the fourth rung and vāma the fifth. The right stands below the left. Whatever the two words are doing here, they are two steps on a staircase, and the left is the higher step.

Second, the verses never use the word ācāra, conduct. The terms are plain substantives and the first is a plural, “the Vedas.” The naming of the seven as vedācāra, vaiṣṇavācāra and the rest, and the grouping of the lower rungs as paśvācāra, the conduct of the paśu or bound creature, come from the commentators and from Woodroffe, who takes the grouping to cover the first four and locates the characteristics of each in the twenty-fourth chapter of the Viśvasāra Tantra.

Third, the count is unstable inside its own edition. Tārānātha’s footnote to verse eight records a compendium that quotes the passage in a variant form, running dakṣiṇa, siddhānta, vāma, aghora, yoga, kaula, which with the first three rungs makes nine, and he closes the note by observing that on that view there are nine ācāras. A classification disagreeing with itself in the apparatus of its own editor.

As for what vāma means here, M. P. Pandit’s rendering glosses it as the reverse path, in which the natural outward turn, pravṛtti, is turned inward into nivṛtti, the turn toward cessation. Nothing in the gloss concerns sex. And this is the very text the popular literature cites as the wine-and-sex Tantra.

The deeper confusion is about the two words themselves, and it is worth taking apart carefully, because the dramatic version of it is wrong.

In the srotas scheme, dakṣiṇa and vāma are compass bearings on a god’s head. Aghora faces south, the right as an icon is reckoned, and his stream is the Bhairava material; Vāmadeva faces north, the left, and his stream is the Caturbhaginī material, the books of the four Sisters. In the ācāra scheme the words are about ritual handedness and carry a purity judgment: dakṣiṇācāra is Veda-congruent, caste-observant, publicly avowable conduct, and vāmācāra is transgressive conduct.

Two different motivations for one pair of words, each explicit in its own sources. Specialists move between them without effort; English popular writing collapses them constantly, which is why the right-and-left literature is such a swamp.

What cannot be said is that the scriptural scheme simply inverts the moral sense of the conduct scheme. It does not assign one consistently. In the Brahmayāmala, dakṣiṇa is sattva and pure; in the Siddhāntaprakāśikā it is thrown in with the Vāma among the drains. A compass bearing on an icon carries no verdict until some school supplies one, and the schools supplied different ones.

A further scheme, internal to Sri Vidya and later, gives four ācāras: Samaya, Dakṣiṇa, Kaula and Vāma. It is not the Kulārṇava’s seven, and the site sets it out at Kameshvara and the four acharas. Kaula and Samaya, which popular lists present as two of the types of Tantra, are two modes of worship within one tradition, disputed by the two commentators who opened this article, at The Kaula-Samaya dispute. What the pañcamakāra, the five substances whose names begin with the letter M, actually prescribes is at The panchamakara question.

The third of Sanderson’s branches classifies by neither books nor conduct.

The transmissions, reckoned by direction

The Kulamārga leaves the mouths of God alone and sorts teachings by the quarter they came from. These are the āmnāyas, the transmissions.

An early account gives four, in the Ciñciṇīmatasārasamuccaya: the eastern around the goddess Kuleśvarī, standing close to the Trika; the western around the hunchbacked Kubjikā; the southern around Kāmeśvarī; and the northern around Kālī Kālasaṅkarṣaṇī, which is the Krama. The site follows each of those, and the five and the sixth of the later reckoning, at The six amnayas.

What belongs here is what the scheme did next, because the furniture is unusually visible. The Tripurā or Sri Vidya tradition grew out of the Kaula traditions and declined to take a seat among the four. It did not assign itself to the southern transmission; it claimed to transcend all four. The southern slot was in due course filled by a goddess called Bhoginī, who appears to have been invented for the purpose. And in the late South Indian systematisation a sixth transmission is stacked on a fifth, the Kulārṇava’s own late Kaulism taking the Ūrdhvāmnāya, the upper transmission, and Tripurā’s mantras the Anuttarāmnāya above it, the transmission than which there is none higher.

A vacancy filled by a goddess made for the vacancy, and a tradition that refused to be classified until a slot could be built above the whole scheme.

Sanderson draws the comparison himself: Tripurā transcends the four just as the Siddhānta claimed its superiority within the Mantramārga, by making its scriptures come forth from the mouth of Īśāna. He notes too that the Paraśurāmakalpasūtra derives five āmnāyas from the five faces exactly as the Mantramārga derived its streams. The device travels, and so does the furniture.

That is the Indian apparatus. What reached English was something else.

How two rituals became two teams

The right-and-left division is old, and it is Indian. Hugh Urban, writing on Abhinavagupta’s gradation of the schools, renders the indigenous categories in exactly those English words. What the nineteenth century supplied was the flattening of a staircase into a scoreboard, and then the attachment of black magic to one side of it.

Each link in the chain can be dated, and the usual account is wrong in its most quoted one.

William Ward, a Baptist missionary writing from Serampore, had the literal glosses in print by 1815: “Vamacharéé, from vam, the left hand, and acharin, practice”; “Dakshinacharéé, from dukshinu, the right (hand).” What the vamacharees actually did he declined to say. In the London edition of 1817 he wrote that here are things “too abominable to enter the ears of man, and impossible to be revealed to a Christian public.” His own index entry is briefer: rites “too abominable to be recorded.”

The division becomes the standard English classification with Horace Hayman Wilson’s “Sketch of the Religious Sects of the Hindus,” which appeared in two instalments. In the first, in 1828, the worshippers of Śakti “were then, as now, divided into two classes, a right and left hand order.” In the second, in 1832, the year he was elected the first Boden Professor of Sanskrit at Oxford, the sect is “divided into two leading branches, the Dakshinácháris and Vámácháris, or followers of the right hand and left hand ritual,” and a few pages later the Vámís “mean the left hand worshippers, or those who adopt a ritual contrary to that which is usual.”

Wilson wrote ritual, order, form of worship, description. He did not write path. That substitution came later and did most of the damage.

On the same pages where he gave English its two-team binary, Wilson printed, in his own footnotes, both sides of a quarrel that had been running inside the tradition for centuries. From the Śyāmā Rahasya he quotes a flat condemnation: the Vāma ritual was intended for Śūdras only; a Brahman who takes spirituous liquor forfeits his Brahmanical character; the practice is brutality and should never be performed. And from the Kulārṇava he quotes, in English, in 1832, the ascending ladder set out earlier in this article, ending that the left hand division is better than the right, that the Siddhānta is better still, that the Kaula is better than the Siddhānta, and that nothing is better than the Kaula.

He had the ladder in front of him. He printed it. What passed into English was the scoreboard.

The moral grading was then fixed in the lexicon. Monier-Williams’s first edition of 1872 glosses dakṣiṇācāra as “a worshipper of Sakti according to the purer or right hand ritual,” and vāmācāra as “the left hand ritual or doctrine of the Tantras” practised “according to the grosser system,” and carries an entry for vāma-mārga, “the left hand way or ritual.” Purer and grosser: a partisan position from inside a Sanskrit argument, printed as a definition, in the most authoritative English reference work on the language.

The word path, with black magic attached, is Blavatsky’s, and it arrives later than it is usually said to. It is routinely credited to Isis Unveiled of 1877. The phrase does not occur in that book, which mentions neither Tantra, nor the Śāktas, nor vāmācāra. It appears in The Secret Doctrine of 1888, where humanity branches into “the Right- and the Left-hand Paths of Knowledge” and the germs of White and Black Magic are sown. The equation with vāmācāra is made only in The Theosophical Glossary of 1892, which splits the Śāktas into right-hand and left-hand, glosses them as white and black magicians, and gives Wilson’s two classes and the lexicon’s moral grading with nothing added but the colours.

Four stages, each on a printed page, none of them forced on English by the Sanskrit. One more map gets handed to the English-language reader, and it belongs to a different religion.

A different religion’s answer

A reader searching for the types of Tantra in English will mostly be shown four: Kriyā, Caryā, Yoga and Anuttarayoga. Be plain about what that is. It is a Buddhist classification of Buddhist scriptures, and on the best current reading it was fixed in Tibet. It is not a division of Hindu Tantra, and a reader who arrives expecting one and finds the other has been misdirected by a search engine, not by either tradition.

Stated fairly, as the Sarma or New Translation schools hold it: Kriyā, action tantra, emphasising external conduct and ritual purity, the deity regarded as superior to the practitioner; Caryā, conduct tantra, balancing external ritual with internal contemplation, the deity as companion; Yoga tantra, inner meditation on reality, the practitioner inseparable from the deity; and Anuttarayoga, highest yoga tantra, with its generation and completion phases, subdivided into Mother, Father and Non-dual classes.

Three qualifications, all from inside Buddhist studies.

The fourfold is not the whole of Tibetan Buddhism. The Nyingma schools work with a nine-vehicle system whose tantric portion has six classes, three outer and three inner, the inner three being Mahāyoga, Anuyoga and Atiyoga.

The fourfold does not appear to be Indian. Anthony Tribe’s account is that the Indian tradition’s own self-representation was fivefold, Kriyā, Caryā, Yoga, Mahāyoga and Yoginī, a scheme that tracks the actual chronology of the texts, and that the commonly encountered fourfold “appears to be Tibetan rather than Indian in origin.” Jacob Dalton, who studied how the Tibetan schemes were built and rebuilt between the eighth and twelfth centuries, puts it harder: the four classes are a late and uniquely Tibetan innovation, perhaps of the twelfth century. Indian authors had a twofold, a threefold, a fivefold, and fourfolds of other shapes. Of the Indian authorities later Tibetan scholasticism cited in support, Dalton concludes, none employed the system that became standard.

And the name of the top class is doubly secondhand. Anuttarayoga, Tribe writes, has not been found in any surviving Sanskrit manuscript. The Indian sources use Mahāyoga, Yogottara, Yoganiruttara or Yoginītantra; Dalton traces the familiar form to a conflation in Tibet and then to a back-formation in Western scholarship.

None of which dismisses a living tradition’s framework, which is internally coherent and which its teachers are entitled to hold. It is a claim about textual history, and it is the claim this article has been making about the Hindu schemes throughout. The scheme was built after the texts, by people with a position in them.

The two traditions are also less separate than the search results suggest. Wallis reports a suggestion from Somadeva Vasudeva that the Buddhist term Mantranaya, the way of mantras, is calqued on the older Śaiva Mantramārga; the Buddhist sources had their own standing pair of pāramitānaya, the way of the perfections, and mantranaya to build it from, so the suggestion is worth recording without leaning on it. What is solider is the address. The Krama’s scriptures and the Buddhist Yoginī tantras both claim Oḍḍiyāna, the Swat valley, as the place of their revelation, though Sanderson reads the Krama’s claim as myth rather than fact.

The site compares the traditions more generally at Hinduism and Buddhism: the difference.

Where the pattern breaks

The corpus does not support this argument everywhere, and the places it refuses are worth knowing.

Not every list ranks. Vairocana’s tenth-century manual is the clearest case, because it does both things between two covers: it carries the ranked five-stream canon in one chapter, and seventy verses later it gives twenty-one Pratiṣṭhātantras, scriptures of temple and image installation, assembled by subject matter across the Siddhānta’s internal divisions with no summit and no claim to one. The Saiddhāntika ritual digests cite across the whole canon without grading what they cite, which is why Sanderson can argue from their silences: they describe without arranging. And a colophon placing a text in the Vidyāpīṭha within the Bhairava stream is self-location, not self-elevation.

One primary author objects to the habit in so many words. Bhaṭṭa Rāmakaṇṭha, commenting on the Mataṅga in the late tenth century, notes the convention of opening a commentary by ranking one’s own scripture above the Lākula and the Mausula and above the Vāma, Dakṣiṇa, Bhūta and Gāruḍa tantras, and calls it utsūtram, which Sanderson renders a matter not pertinent to the text at that point. His objection is to where the ranking is put, not to whether it is true. It is still a Śaiva exegete finding the reflex tiresome.

Nor does the division fall neatly between kinds of book, because the Śrīkaṇṭhī is an inventory and it ranks. What separates it from Vairocana’s installation canon is less the business each text is in than where in the text the ranking sits: at the head of a scheme, where a claim is being staked, or down in the body of a list, where the author is cataloguing.

So the honest form of the argument is narrower than the pattern first suggests. Where a text is staking a position among the religions or the streams, it ranks, and it ranks from where it stands. Where it is inventorying, installing images, or telling a priest what to do on a Tuesday, it frequently does not.

Sanderson draws the methodological consequence. Writing of the real division of function between the Saiddhāntika and the Śākta traditions, he warns that this duality of function “is not readily observed by reading the insiders’ texts, since these belong to one or other of the divisions and all claim to be offering the same goals.”

Which leaves the reader with a practical problem, and it has a practical answer.

How to tell which map you are holding

For a reader who has to work out what a given book or teacher means, a short set of tests.

Scriptures sorted by which mouth of God spoke them, in five streams? The srotas system, Saiddhāntika in origin, and the Siddhānta is at the summit. If the same device yields three streams graded by the guṇas, it is the Brahmayāmala’s, and the Siddhānta is at the bottom.

Mantrapīṭha and Vidyāpīṭha? The pīṭha system, Śākta in origin. The Goddess texts win.

Ten, eighteen and sixty-four? The Śrīkaṇṭhī’s canon, which grades the three tiers itself and puts the sixty-four above the ten and the eighteen. In Abhinavagupta’s hands it becomes a staircase whose top step is one book, and the book is his.

Transmissions by compass direction? The āmnāyas, and the Kulamārga. Count them: four is the classical scheme, and every number above four is somebody claiming to stand outside it.

Seven ascending modes of conduct? The Kulārṇava’s ladder, Kaula highest, vāma above dakṣiṇa. Four modes with Samaya among them is the later Sri Vidya scheme instead.

Two teams, right and left, the left meaning sex or black magic? The pair of words is Indian. The scoreboard is Wilson’s and the black magic is Blavatsky’s.

Four classes rising to Anuttarayoga? Tibetan Buddhist, answering a different question inside a different religion.

One test covers most of them. Find what the scheme puts at the top of the streams, and you have usually found the school that drew it.

That is the answer to “what are the types of Tantra.” A shelf of precise, dated, locally coherent classifications, no two of which reconcile, drawn up by people who mostly had a view about where they stood in them. The Siddhānta is the summit and the Siddhānta is the sediment. The sixty-four are scripture of the nature of the Veda and a device for deluding the unwary. The left hand is the rung above the right and the mark of the damned. Each of those was somebody’s considered position, held in Sanskrit, with reasons.

The ādhīnams, the monastic centres of the Śaiva Siddhānta, still order their liturgy from the twenty-eight. The Sri Vidya maṭhas, the seats of its teaching lineages, still hold the Samaya line Lakshmidhara defended.

What the schemes were arguing over never needed any of them to agree.