K ¯alavañcana in the Konkan: How a Vajray¯ana Hat.hayoga Tradition Cheated Buddhism’s Death in India

: In recent decades the relationship between tantric traditions of Buddhism and ´Saivism has been the subject of sustained scholarly enquiry. This article looks at a speciﬁc aspect of this relationship, that between Buddhist and ´Saiva traditions of practitioners of physical yoga, which came to be categorised in Sanskrit texts as hat.hayoga . Taking as its starting point the recent identiﬁcation as Buddhist of the c.11th-century Amr.tasiddhi , which is the earliest text to teach any of the methods of hat.hayoga and whose teachings are found in many subsequent non-Buddhist works, the article draws on a range of textual and material sources to identify the Konkan site of Kadri as a key location for the transition from Buddhist to N¯ath ´Saiva hat.hayoga traditions, and proposes that this transition may provide a model for how Buddhist teachings survived elsewhere in India after Buddhism’s demise there as a formal religion.


Introduction
It has long been recognised by indologists that Vajrayāna Buddhist and Nāth 1 Śaiva traditions have much in common, in particular adepts, sacred sites and metaphysical terminology.In recent years scholars have explained these commonalities either by pointing to the Nāths as their originators or by claiming that the two traditions share a common substratum.Early 20th-century Indian scholars, on the other hand, viewed Vajrayāna Buddhism as their source. 21 I use the designation "Nāth" here even though it was not current during the period under consideration (and I do the same for "Hindu").See (Mallinson 2011, p. 409) for a discussion of the usage of the term Nāth as the name of a grouping of yogi lineages.I use the vernacular form "Nāth" rather than the Sanskrit "Nātha" because it is in vernacular usage that the designation "Nāth" is most usually found. 2 (Dasgupta 1946, pp.194-95) denies the possibility of Buddhist origins for the Nāths, partly on the spurious grounds that they were the first alchemists so must have existed before the Pātañjalayogaśāstra because of its mention (4.1) of aus .adhi, medicinal herbs, and thus long predated the esoteric forms of Buddhism with which they have much in common.In east India, Nepal and Tibet, continues Dasgupta, the Nāths' traditions "got mixed up with those of the Buddhist Siddhācāryas", a process which was facilitated by their common heritage of tantra and yoga.White (1996, pp.106-9) suggests that Goraks .anātha, the second of the human Nāth gurus, was originally Śaiva before being made out in later myths to be Buddhist, concluding that "since no extant tantric or Siddha alchemical works, either Hindu or Buddhist, emerged out of Bengal prior to the thirteenth century, we need not concern ourselves any further with the imagined east Indian Buddhist origins of Gorakhnāth or the Nāth Siddhas".(White does not address the possibility of elements of Nāth tradition deriving from Buddhist traditions from other parts of India.)Briggs (1938, p. 151 n. 1) names the stages of development of Buddhism in Bengal as Mantrayāna, Vajrayāna and then Kālacakrayāna, and that " [t]hese Buddhist elements were absorbed into the Nāthamārga".Sen (1956, pp.281-86) writes of the Nāths' "Buddhist affiliation" and states that " [b]oth the Nātha cult and Vajrayāna had fundamental unity in their esoteric or yogic aspects", but makes no suggestion as to which of the two traditions came first.Templeman (1997, p. 957) talks of a "shared praxis" which persisted up to the seventeenth century.In contrast, the polymath writer, poet and scholar M.Govinda Pai, who was from the Tul .u region (whose Kadri monastery is the focus of this article), says that "the Nātha cult is known to have developed itself out of the Vajrayāna system of the The relationship between Vajrayāna and Nāth Śaiva traditions is but one part of the complex relationship between Vajrayāna and Śaivism as a whole, which has been the subject of detailed analysis since an article by Alexis Sanderson published in 1994 in which he demonstrated the dependence of certain Buddhist Yoginītantras on texts of the Śaiva Vidyāpīt .ha. 3 Shaman Hatley has shown that since the time of the earliest tantric texts there are likely to have been borrowings between the two traditions in both directions, albeit on a smaller scale than in the examples provided by Sanderson. 4 David Seyfort Ruegg and Francesco Sferra, while accepting the borrowings demonstrated by Sanderson, have argued the case for a shared substrate, with Sferra proposing that "Buddhist and Hindu Tantric traditions only appear to be distant from one another at the theoretical level when the common practices and 'substratum' are imbued with a doctrinal content". 5Sanderson has rejected the concept of a shared substratum because it is an entity that is only inferred, whereas everything we perceive in this context is either Śaiva or Buddhist. 6Buddhist traditions, in which he first appears in perhaps the 12th century. 10Tibetan hagiographic treatments of Vir ūpāks .a and textual cycles associated with him are particularly rich and diverse, 11 as are his depictions in Tibetan paintings 12 and statuary.Here I shall draw upon these Tibetan materials to note only (1) that they indicate that after spending his early life in east India, Vir ūpāks .a was active in the Deccan and the south; 13 (2) that the Sanskrit Amr .tasiddhi, whose teachings were attributed to Vir ūpāks .a, was translated into Tibetan (probably in the late 11th century) 14 and an associated cycle of Tibetan texts (usually referred to in Sanskrit back-translation as Amarasiddhi), 15 whose teachings were also attributed to Vir ūpāks .a, developed soon after; and (3) that Vir ūpāks .a is said to be the human conduit of the Tibetan Lamdre (lam 'bras) tradition, whose teachings include the practices of Trulkhor ('krul 'khor), some of which correspond to the three central techniques of the Amr .tasiddhi and which are predicated on an understanding of the yogic body first taught in the Amr .tasiddhi. 16he siddha Vir ūpāks .a has left few traces in Indian material and textual sources.Buddhist hagiographies mention a shrine to him at Somnath, 17 but there is no evidence of one there now.The temple of Bhīmeśvara at Draksharama in Telangana, which predates 1130 CE 18 and is associated with the Vir ūpāks .a legend in early Marathi texts and current Tibetan hagiographies, has a shrine to him; there are reliefs depicting him on the c. 1230 CE Mahudi Gate at Dabhoi in Gujarat (Figure 1), on the exterior wall of a c. 13th-century cave at Panhale Kaji in the north Konkan (Figure 2) and perhaps on the c. 14th-century Someśvara temple at Pimpri Dumala near Pune; 19 there is a Vir ūpāks .a cave on 10 There have been many Vir ūpāks .as other than Vir ūpāks .a the siddha: a Buddhist king known in early Pali sources (Malalasekera 1937, s.v.Vir ūpakkha; I thank Hartmut Buescher for this reference); one of four great kings of early Mahāyāna in the pre-5th century Suvarn .aprabhāsottama (6.1.1,6.3.1, 6.6.25;I thank Gergely Hidas for this reference); a form of Rudra mentioned in the Skandapurān .a (72.64 in the edition in preparation by Peter Bisschop et al.; I thank Professor Bisschop for this reference); one of eight yaks .as in the Śivadharma (pp.193-208); a form of Śiva whose teachings are given in the c. 12th-century Vir ūpapañcāśikā; and a form of Śiva which is the central deity of Vijayanagara.Monier-Williams (s.v.vir ūpacaks .us) gives many more references. 11See (Dowman 1985, pp.43-52) and (Davidson 2005, pp.49-54) for overviews of Vir ūpāks .a's legends (which are first found in the c. 12th-century Grub thob brgyad bcu rtsa bzhi'i lo rgyus of Smon grub shes rab, which is translated in Grunwedel 1916;  Robinson 1979 and Dowman 1985); and (Chattopadhyaya 1970, p. 404) for a list of works in Tibetan attributed to him. 12A painting from the Drigung tradition which predates 1217 CE is perhaps the earliest Tibetan depiction of Vir ūpāks .a (Luczanits 2006, p. 82). 13In the Grub thob brgyad bcu rtsa bzhi'i lo rgyus, Vir ūpa is said to have been born at Tripurā in East India and studied at the Somapurī vihāra, which is near Paharpur in Bangladesh (Robinson 1979, pp.27-28).Tāranātha says that Vir ūpa lived in Mahrata, i.e., Maharashtra, that he visited Srisailam and that his disciple Kāla Vir ūpa practised in the Konkan (Templeman 1983, p. 18 and Chattopadhyaya 1970, p. 215).Both Smon grub shes rab and Tāranātha also tell a story of Vir ūpa destroying an icon of Śiva.The former names the destroyed Śiva as Maheśvara and locates his temple in the unidentified land of Indra (Robinson 1979, pp.29-30); the latter names the Śiva Viśvanātha and locates his temple in Trili ṅga, i.e., the present-day Telangana region (Templeman 1983, p. 15).As a result of a transmission whose details are unknown to me, current Tibetan legend (see e.g., http://www.ludingfoundation.~org/Archive2016.htmlaccessed 7 June 2018) accords with the c. 1280 CE Marathi Līl .ācaritra in locating this episode at Bhīmeśvara, which is one of the three li ṅgas referred to in the name of Trili ṅga and whose temple complex at Draksharama houses a shrine to Vir ūpa; on the Līl .ācaritra's story, see footnote 13. 14 Schaeffer 2002. 15One of the texts attributed to Vir ūpāks .a in the Vanaratna codex (on which see Isaacson 2008) is entitled Amarasiddhi.Cowell and Eggeling (1876, p. 28) report the name of the text as Amarasiddhiyantrakam but a transcription of the text kindly shared with me by Péter-Dániel Szántó shows that its name is Amarasiddhi (f.47 recto and verso).Chattopadhyaya (1970,  p. 404) mentions an Amarasiddhivr .tti among Tibetan works attributed to Vir ūpāks .a. 16 Baker 2018, pp. 421-22. Schaeffer (2002, p. 527, n. 12) finds no connection between the Vir ūpāks .a of the Amr .tasiddhi and the Vir ūpāks .a of the lam 'bras tradition, but in the Vanaratna codex described by Isaacson (2008), one of whose texts is, as noted above, an Amarasiddhi of Vir ūpāks .a, after the text of the Marmopadeśa there is a lineage of teachers which starts from Vir ūpāks .a and which Isaacson (2008, pp.3-4) identifies as being very close to some of the lam 'bras lineages. 17See e.g., Tucci 1931, p. 690. 18 Vīra Rājendra Cod .ā, who flourished c. 1130 CE, is recorded as having made a donation to Bhīmeśvara at Drāks .ārāma in an undated inscription (Epigraphia Indica Vol.IV, p. 51). 19I am grateful to Amol Bankar for sharing with me these identifications of Vir ūpāks .a (personal communication 12 June 2018).
Bankar identifies as Vir ūpāks .a the Dabhoi image, which is one of a group of twelve siddhas of whom some are clearly Nāths (Shah 1957), because, despite considerable damage to the sculpture, it is evident that he is accompanied by a woman and that there are images of the sun and moon above him.Both these motifs are suggestive of the legend of Vir ūpāks .a in which he stops the sun's path through the sky so that a lady innkeeper will keep serving him and he will not have to pay his bill (see e.g., Robinson 1979, p. 29).A siddha depicted on the exterior of cave 14 at Panhale Kaji is sitting with a yogapat .t .a in a posture common in Tibetan images of Vir ūpāks .a and is accompanied by a woman who may be pouring him a drink.Only one Indian Buddhist text other than the Amr .tasiddhi mentions Vir ūpāks .a: the Caryāgīti, a c. 11th-century collection of fifty middle Indic dohā verses attributed to various siddhas.One of its dohās is by Biruā (i.e., Vir ūpa/Vir ūpāks .a), who, in highly esoteric language, summarises a yoga method which is similar to that of the Amr .tasiddhi but is couched in a metaphor of alcohol production Bankar's identification as Vir ūpāks .a of an image of a siddha at Pimpri Dumal (reproduced in Sarde 2014, p. 6, fig.10) is more tentative, being dependent upon the siddha, who is standing, being accompanied by an anthropomorphic image of S ūrya, the sun god, and pointing at the sky. 20I thank Amol Bankar for informing me of the Kāmākhyā Vir ūpa image, which depicts the tavern episode summarised in footnote 19.
rather than the Amr .tasiddhi's alchemy. 21The Caryāgīti's place of composition is uncertain.22I know of eight non-Buddhist texts which mention Vir ūpāks .a.One is the Varn .aratnākara, an early 14th-century Maithili compendium on a variety of subjects which gives a list of 84 siddhas and includes the name Vir ūpa. 23The remaining seven texts are connected in some way with the Śaiva Nāth tradition and are from the Deccan or south India.The two oldest are Mahānubhava works in Marathi (which corroborates statements elsewhere that Vir ūpāks .a was from the Maratha region): 24 the c. 1280 CE Līl .ācaritra, a hagiography of the Mahānubhava guru Cakradhara composed by his devotees, 25 and the Tattvasāra of Cā ṅgadeva, a compendium of Mahānubhava teachings which was completed in 1312 CE.After these Marathi works, the next texts to mention Vir ūpāks .a are the c. 1400 CE Telugu Navanāthacaritramu, an account of the lives of the nine Nāths composed by Gauran .a at Srisailam, the contemporaneous Vikramārkacaritramu of Jakkana, which includes Vir ūpāks .a's name among those of the nine Nāths, 26 and the Sanskrit Hat .hapradīpikā, a compilation by Svātmārāma of teachings on hat .hayoga whose parallels with the Navanāthacaritramu (and other Telugu texts) suggest it is from the same period and region.The Tārārahasya, a 16th-century Sanskrit treatise on the worship of the goddess Tārā by the Bengali author Brahmānandagiri, includes Vir ūpāks .a in a list of eight human gurus to be worshipped. 27The last non-Buddhist text to mention Vir ūpāks .a is the c. 17th-century Sanskrit Kadalīmañjunāthamāhātmya, a celebration of the temple of Mañjunātha at Kadalī (now known as Kadri, a part of the coastal town of Mangalore), in which Vir ūpāks .a is again one of nine Nāths, seven of whose stories, including that of Vir ūpāks .a, are taught in extenso.
In the Līl .ācaritra, Cakradhara tells his disciples how Vir ūpāks .a broke in two the Bhīmeśvara śivali ṅga at Drāks .ārāma in modern-day Telangana.Goraks .a converted Vir ūpāks .a from vajraolī, i.e., the [Buddhist] Vajra lineage, to amaraolī, the [ Śaiva] Amara lineage, and gave him the name Adan .d .īnātha. 28The Marathi Tattvasāra includes Vir ūpāks .a in a list of 84 siddhas. 29n the Navanāthacaritramu, Vir ūpāks .a is one of the nine Nāths of the text's title and his life story is told in detail.He is the second son of king Gan .yāvanta and Añjani, who were from Maharashtra. 30atsyendranātha, the first of the nine Nāths and the guru of the other eight, meets him in a forest.After Vir ūpāks .a recounts how he has eaten the heart of a bird which a hunter had informed him would make him a siddha, 31 a voice from the sky confirms his story and tells that of his previous birth, at which Matsyendra initiates him and instructs Goraks .a to teach him yoga.After receiving instruction, Vir ūpāks .a travels to the Karn .āt .a, Lalita, Kanauj, and Māl .ava regions, and initiates several disciples. 32In his Vikramārkacaritramu, Jakkana gives a list of nine Nāths almost identical to that of Gauran .a.He does not tell their stories but for each of them highlights one characteristic, which for Vir ūpāks .a is devotion. 33n the Hat .hapradīpikā nothing is said about Vir ūpāks .a other than that he, along with 28 other mahāsiddhas, broke the rod of time by means of the power of hat .hayoga. 34Vir ūpāks .a is at number 11 in the list, immediately after Caura ṅgi, Mīna (who is here differentiated from Matsyendra) 35 and Goraks .a.As will be explored in more detail below, these three siddhas are closely associated with Kadri and the Amaraughaprabodha, 36 approximately twenty verses from which are found in the Hat .hapradīpikā.
In six chapters and more than 300 verses the Kadalīmañjunāthamāhātmya gives a detailed account of several episodes in Vir ūpāks .a's life. 37He is born to a pious Brahmin couple from the northern Konkan as an incarnation of Mañjunātha of Kadalīvana, i.e., the Śiva of the Kadri temple and monastery complex.He wanders across India and has various adventures, including the conquest of a demon threatening all the gods; the assumption of the appearance of an ascetic and subsequent humbling of the Veda-obsessed brahmins of Drāks .ārāma (an episode which brings to mind the Līl .ācaritra's story of his breaking of the Bhīmeśvara li ṅga at Drāks .ārāma); 38 and a sojourn in Kāñcī, where he teaches the citizens by day and sports with women by night.This last episode includes an echo of the popular Tibetan story in which Vir ūpāks .a stops the sun in its course until the king pays the bill for his drinks. 39Here he falls for one of the women of Kāñcī and, in order to impress her, grabs the moon and makes it into a goblet with which to ply her with drink.The gods become concerned and, at Brahmā's instruction, go to Vir ūpāks .a and sing his praises, at which he puts the moon back in its rightful place.
With this, the historical trail left in India by Vir ūpāks .a goes cold, 40 but he has left enough clues for a tentative identification of the region in which his teachings were transmitted from Vajrayāna Buddhism to Nāth Śaivism.As noted above, Tibetan hagiographies point to south India and the Deccan as being central to his activities and almost all Indian material and textual sources associated with him are from the Konkan and Deccan. 41The Navanāthacaritramu identifies him as the author of the Amr .tasiddhi 42 and anchors his story in the Deccan: the Navanāthacaritramu was composed at the request of the pontiff of the Bhiks .āvr .tti monastery in Srisailam 43 and Vir ūpāks .a is said therein to have been born in present-day Maharashtra and to have travelled to Karnataka. 44The Hat .hapradīpikā adds nothing to his lore, but strengthens the connection with Srisailam since it has many parallels with the Navanāthacaritramu and was probably composed in the same region and in the same period. 45he Kadalīmañjunāthamāhātmya, one of whose manuscripts is likely to have been copied in 1730 CE and may not, in its present form, be much older than that, but which preserves some old legends from the Kadri site, 46 identifies Vir ūpāks .a with Śiva Mañjunātha of Kadri.All of the Nāths whose stories are told in the Kadalīmañjunāthamāhātmya are similarly identified with Mañjunātha, but Vir ūpāks .a is further tied to the region by virtue of being the only one to be given the epithets ko ṅkan .ādhīśa, ko ṅkan .eśvara and ko ṅkan .a. 47

Nāth Śaivism and Vajrayāna in the Konkan
Vir ūpāks .a's legend thus points to the south of the Indian subcontinent and in particular the Konkan as the likely location of the transition of his teachings from Vajrayāna to Nāth Śaivism, but gives little detail of how it might have happened, with only the Marathi Līl .ācaritra actually indicating a transition from Buddhism to Śaivism.I shall now widen the enquiry from Vir ūpāks .a to evidence for the presence of Vajrayāna and Nāth Śaivism in the Konkan, which will further support the supposition that such a transition occurred there and provide more detail on how it happened.

Nāth Śaivism in the Konkan
Nāth Śaivism has a long history in the Konkan.Western India and the Konkan region were important centres of early Paścimāmnāya Kaula Śaivism, elements of which were preserved by subsequent Nāth lineages. 48Matsyendranātha, the revealer of the Kaula doctrine in the Kali age and the first human guru of the later Nāth sampradāya, is closely associated with the region.His consort was called Ku ṅkun .āmbā, "the mother of the Konkan"; 49 a c. 13th-century anthology of his teachings, the Matsyendrasam .hitā, is associated with the Konkan and was composed either there or in the Tamil region; 50 and among the earliest material evidence of the Nāth sampradāya is the c. 13th-century ch. 5 on the sun and moon) and which, like the Amr .tasiddhi, gives its chapters the unusual designations of viveka and laks .an .a. On the date of the Vivekadarpan .a, see Reinelt 2000, pp. 93-95. 43 Navanāthacaritramu canto 1 (Jones 2018, pp.185-86). 44Navanāthacaritramu cantos 4-5 (Jones 2018, pp.189-94). 45Mallinson forthcoming b. 46 The Kadalīmañjunāthamāhātmya is a palimpsest of Buddhist, Śaiva and Vais .n .ava teachings, reflecting the passing of the control of the Mañjunātha temple from Vajrayāna Buddhists to Nāth Śaivas and then Mādhva Vais .n .avas, with the latter responsible for its final redaction.For an overview of its contexts and context, see Nagaraju 1969. 47 Kadalīmañjunāthamāhātmya 14.9, 48.7, 50.11, 50.17, 53.27  (Schoterman 1975, pp.934-35); (Sanderson 2011, pp.44-45 and 2014, pp.62-64), and (Mallinson 2011, pp.412-14).The Śaiva tradition of the southern Nāths is Śāmbhava, a variant of the Paścimāmnāya (Kiss forthcoming).The oldest known statues of Nāths date from the 12th century CE onwards and are found in western India and the Deccan (Sarde 2017, pp.96,  108-10). 49Tantrāloka 29.32. 50 Kiss (forthcoming, p. 32) says that the Matsyendrasam .hitā was "composed in South India, probably in the Tamil region, or alternatively, around Goa, in the 13th-century".Kiss chooses the Tamil region over the Konkan because of the mention in the Matsyendrasam .hitā of the god Śāstr ., whom he identifies as a specifically Tamil deity, but the Kadri Mañjunāth temple has an image of Śāstr .dated to the twelfth century and other, earlier images of Śāstr .are found in the region (Bhatt 1975,  pp.354-55) and plates 290 and 291.Of the two toponyms mentioned in the Matsyendrasam .hitā, the first, Gomanta, is likely to be in present-day Goa (Kiss forthcoming, pp.30-31) while the location of the second, Al ūra, whose king's dead body was taken over by Matsyendra, is uncertain.Kiss gives various possible identifications of Al ūra, including Ellora, Eluru in Andhra Pradesh and Vellore (Kiss forthcoming, p. 31).The five manuscripts of the Matsyendrasam .hitā all have statuary at Panhale Kaji on the Konkan coast between Mumbai and Goa, which includes at least two images of Matsyendra. 51he monastery at Kadri is today the most important Nāth site in south India.Statues of three Nāths at the Mañjunātha temple below the monastery have been dated on stylistic grounds to the 14th or 15th centuries. 52The first firmly datable evidence of the presence of the Nāths at Kadri is in a Kannada inscription of 1475 which records a grant of land to Ma ṅgalanātha Od .eya "at the pure place of Śiva, Kadire", the income from which will support the worship of G ōraks .anātha and Candranātha, a deity at the temple. 53Bhatt notes an inscription from Bārak ūru dated 1490 CE which refers to the position of arasu or king at Kadri in the context of one Subuddhinātha Od .eya, disciple of Anupamanātha Od .eya, and to the worship of G ōrakhanātha at the mat .ha of the latter. 54Subsequent descriptions of the head of the Kadri monastery also call him the "King of the Yogīs" and until the demise of the Vijayanagara kingdom and resultant depredations by the Nāyaka kings he lived in great style.Writing at the beginning of the 16th century, the Italian traveller Ludovico di Varthema said that the king ruled over 30,000 people and travelled about India with an impressive entourage including a troop of warrior yogis. 55By the 17th century the king lived in much reduced circumstances 56 and it seems that at this point the Kadri monastery was taken over by a northern Nāth lineage intent on creating a pan-Indian yogi order.By 1820 CE Kadri was recognised as the southern seat of the four seats of the Nāth sampradāya. 57To this day the head of the Kadri monastery, who retains the title of "King Yogi" (rājā yogī), is drawn from northern Nāth lineages.The changes at Kadri in the 17th century resulted in the migration southwards from Mangalore to northern Kerala of the yogi caste associated with the previous royal lineage. 58Some members of the caste remained at Kadri, but they have a fraught relationship with the King Yogi, whom they look to for leadership but who cannot speak their language.The King Yogi is elected once every 12 years at a council of the Nāth sampradāya at the Nasik Simhastha festival (which nowadays is recognised as one of the four Kumbh Melas), after which he and several hundred Nāths walk barefoot for six months to Kadri, where he is installed in a lengthy consecration ritual called the rājyābhis .eka, which has parallels with traditional royal initiations. 59the reading al ūrādhipatim .bh ūpam .dāks . in .ātyam .purā kila (55.3ab).The text is likely to have been redacted in the early 19th century at the request of the Mahārāja of Jodhpur (Kiss forthcoming, p. 32); perhaps the original reading, obscure to north Indian scribes, was ālupādhipatim ., referring to a king of the Ālupa dynasty, which ruled the Tul .u region from the early centuries CE until the end of the 14th century (Bhatt 1975, p. 18), and whose capital was at Mangalore from the 11th to 13th centuries, the likely period of composition of the Matsyendrasam .hitā.In the Navanāthacaritramu, the king who plays the same rôle in the story of Matsyendra rules over Kadri, which at the time would have been under Ālupa rule.The Matsyendrasam .hitā does not, however, mention Kadri in its version of the story of Goraks .a rescuing Matsyendra.Supporting a Tamil origin for the Matsyendrasam .hitā is its identification of Goraks .a as a Col .ā king (pat .alas 1 and 55) and its naming at 55.20 of Matsyendra's son and grandson as Kharparīśa and Vyālīndra, both of whom are associated with alchemy in various Tamil traditions but are not mentioned in texts associated with Kadri (or Srisailam). 51Deshpande 1986.Panhale Kaji has various Nāth statues, including a group of 9 or perhaps 10 siddhas.A group of 12 Nāth siddhas, dating to 1230 CE, is found at Dabhoi near Ahmedabad in Gujarat.Bhatt 1975, p. 299 and plates 303 and 304(a). 53I thank Manu Devadevan for sharing with me his transcription and translation of this inscription.Bouillier (2008, p. 96)  reports that the inscription refers to Candranātha as a king (arasu) but Devadevan tells me that this is not the case.Bhatt (1975, p. 295) notes that an inscription from Mangalore dated 1434 CE mentions "the gift of land to one Jugādikun .d .ala J ōgi-Purus .a by J ōgi-Od .eya alias Chaut .a". 54 Bhatt 1975, p. 294. 55 Badger 1863, pp. 111-13. 56 Grey 1892, pp.345-52. 57Tashrīh al-Aqvām .chp. 104 (British Library Board, Add.27255 f.399).I thank Bruce Wannell for translating this passage for me. 58This migration is recorded in the Teyyam performances regularly put on by the C ōyī (the vernacular for yogī) caste in northern Kerala, in which it is precipitated by the death of the king's son and the resultant ending of his lineage (Freeman 2006, pp.167-69). 59On the rājyābhis .eka, see Bouillier 2008, chp.6.

Vajrayāna in the Konkan
Buddhism was well established in the Konkan when Hsüan-Tsang visited the region in the seventh century, 60 and Vajrayāna had a small but significant presence there and in surrounding areas from the sixth to thirteenth centuries, and perhaps later.Sixth-century statues of Tārā and Avalokiteśvara are found in the western Deccan. 61The colophon of the ninth-century Cakrasam .varapañjikā of Jayabhadra (who was also known as Ko ṅkan .apāda) says that its author visited a temple of Tārā at the Konkan site of Mahābimba. 62A statue of Mañjughos .a from the Kadri monastery but now in the Mangalore government museum dates to the ninth century or earlier. 63One of the 29 caves at Panhale Kaji contains a tenth-century image of the Vajrayāna deity Acala (Figure 3). 60(Sastri 1939, pp.104-5).On early traces of Buddhism in Karnataka see (Nagaraju 1983, pp.6-10). 61Acri 2016, p. 8. 62 (Szántó 2016b, p. 2).The exact location of Mahābimba is uncertain.I know of two further possible references to it.Tāranātha says that in Ko ṅkan .a his guru Buddhaguptanātha saw "the self-created image of Mañjuśrī in the middle of a pond.It is called Jñānakāya... Then he saw also the bimbakāya which looks like a rainbow raising the st ūpa of the accumulated vapour beyond touch" (Tucci 1931, p. 696).A manuscript of the As .t .asāhasrikā Prajñāpāramitā dated 1015 CE mentions a pilgrimage site in Ko ṅkan .a called Mahāviśva, which Szántó (2012, Vol.1/2 p. 40, fn.61) suggests may be a corruption of Mahābimba.A puzzling verse in the long recension of the Amaraughaprabodha (67 in Jason Birch's edition, from which the variant readings below are taken), which does not appear to fit its context and is also found, with significant variants, at Sekoddeśa 26, indicates that these two unusual compounds may refer to a single object, which lights up various heavenly bodies (including smoke (dh ūma • ) and specks of dust ( • marīci • ), suggestive of the vapour reported by Tāranātha): The early ninth-century Vajrayāna adepts Dharmākara and his fellow initiate Pālitapāda lived in the Konkan. 64Pālitapāda was twice visited there, probably at Kadri, by Jñānapāda, the founder of an important eponymous tradition of exegesis of the Guhyasamājatantra. 65The Vajrayāna teacher Śākyamitra visited the Konkan region in the tenth century 66 and the same century saw the composition of the Hevajratantra, which includes Ko ṅkan .a in a list of 24 sacred sites. 67An illustrated manuscript of the As .t . asāhasrikā Prajñāpāramitā dated 1015 CE 68 includes six sites in the Konkan among 72 Vajrayāna places of worship. 69An inscription at Bal .l .igāve in Karnataka (80 kilometres inland from Gokarn .a) dated 1065 CE records the construction of a vihāra containing images of Buddha, Lokeśvara and Tārā Bhagavatī 70 and another inscription from the same location dated 1067 CE records the establishment of a temple of Tārā Bhagavatī. 71An 11th-century stone statue of Tārā from Bal .l .igāve is still visible at the site and may be one of those mentioned in the contemporaneous inscriptions. 72As will be explored in more detail below, the Kadri Mañjunātha temple contains bronze sculptures of the Vajrayāna deities Lokeśvara and Mañjuvajra (as well as a bronze of the Buddha), with an 11th-century inscription recording the establishment of the Lokeśvara image in the kadirikā vihāra, i.e., a Buddhist monastery at Kadri.An image dated to the 11th century in the Lokanātha temple at Hat .t .iya ṅgad .i, Coondapur, might be of Jāmbāla. 73A 12th-century inscription at Dharmavolal (today known as D .ambal ., 60 km east of Hubli) records the worship of Buddha and Tārā. 74An image of Aks .obhya from Puttige, M ūd .abidure, dates to the 12th or 13th century. 75A 13th-century inscription from the village of Kol .ivad .(20 km east of Hubli) records the worship of Tārā. 76According to the Jain exegete Vīrānandī, Buddhist ascetics called Ājīvakas were active in the Kanara region in the 12th century. 77Finally, two Tul .u inscriptions, one dating to 1187 CE, the other to 1545 CE, mention Buddhism. 78he accounts of tantric Buddhism in India by the 17th-century Tibetan scholar Tāranātha include several references to the Konkan.Tāranātha's histories are notoriously unreliable, but some of his reports of teachers visiting the Konkan are corroborated by the older evidence noted above.Thus he says that Jayabhadra lived for some time at Mahābimba, 79 describes Jñānapāda's visit to Pālitapāda and his initiation into the Guhyasamājatantra, 80 and confirms that Pālitapāda's co-initiate Dharmākara was from the Konkan. 81Tāranātha makes several more mentions of the Konkan in his accounts of early Vajrayāna adepts and teachers. 82riting of more recent events, Tāranātha says that Śāntigupta, the guru of his own guru Buddhaguptanātha, who lived during the 16th century and had close links with the Nāth sam .pradāya, went to Suvarn .adhvaja in the Konkan, which was ". . .a noble well-proportioned place.Its monastic colleges flourished.There were about fifty fully ordained monks there and at most about one thousand upasakas". 83Subsequently "[t]he monks of Maharata and Kongkuna invited him and he went to all their temples giving empowerments, upadesas, alms, sermons on the tantras, etc., and he clearly explained the Vajrayana teachings". 84hese references to a flourishing Vajrayāna tradition in 16th-century Konkan are intriguing, but are likely to be garbled reports from earlier times.In Tāranātha's detailed account of the travels of his guru Buddhaguptanātha, he writes the following: Then in Ko ṅkan .a he embarked and went to the west up to an island called h .gro ling [,] in Sanskrit Dramiladvīpa.In the language of the Muhammadans, the barbarians and [the inhabitants] of the small island, it is called la sam lo ra na so (in Śambh: sam lo ra na so).In that island the teachings of the guhyamantras are largely diffused.He heard these from a pan .d .it called Sumati who had acquired the mystic revelations (abhijñā), the mystic power of the Sam .vara (tantra) and of the Hevajra (tantra) and then he learnt the detailed explanation of the Hevajratantra.This Hevajratantra belongs to the system of the Ācārya Padmasambhava.Generally speaking, the tradition of the fourfold tantras is still uninterrupted in that island, and if we except the sublime and largely diffused Kālacakratantra, whatever is in India is also there such as the (Vajra)kīlatantra and the Tantra of the daśakrodhas, many Heruka-tantras, Vajrapān .i, mkhah .ldi ṅ (Garud .a), Māmakī, Mahākāla, etc. Then the sublime order of Hayagrīva which is largely spread in India is to be found there.Moreover there are many sacred teachings (chos) belonging to the Tantras expounded by Padmasambhava.Though the community is numerous, the rules of the discipline are not so pure.The monks wear black garments and usually drink intoxicating liquors . . ." (Tucci 1931, p. 690).
The likely location for this small island is among those near Karwar on the coast of northern Karnataka, with Anjediva the best candidate, 85 but it and all the other islands off the Konkan coast were taken over by the Portuguese at the beginning of the 16th century, well before Buddhaguptanātha would have visited the region. 86Tāranātha appears to have been mistaking Jesuit priests for bibulous black-clad Buddhists. 8783 Templeman 1983, pp. 82-83. 84 Templeman 1983, p. 94. 85 (Yule and Burnell 1903, p. 28) s.v.Anchediva: "c.1345.-IbnBatuta gives no name, but Anjediva is certainly the island of which he thus speaks: "We left behind the island (of Sindāb ūr or Goa) passing close to it, and cast anchor by a small island near the mainland, where there was a temple, with a grove and a reservoir of water.When we had landed on this little island we found there a Jogi leaning against the wall of a Budhkhānah or house of idols."-IbnBatuta, iv.63." Anjediva was taken over by the Portuguese in 1505 (Mathew 1988, p. 163).The name of the island given by Tāranātha appears to be San Lorenzo, but the Portuguese church on Anjediva has always been known as Nossa Senhora das Brotas (Our Lady of the Springs) in homage to the island's good supply of fresh water.San Lorenzo was a name for Madagascar, but that is by no means a "small island", so it seems that either Tāranātha was again conflating his sources or he was, as suggested by Templeman (1997, p. 962, n. 38), referring to a Portuguese settlement by that name elsewhere in the Konkan.The only such reference I have found is to a church of San Lorenzo in Goa mentioned by della Valle in the early 17th century, but which was no longer standing in the 19th century (Grey 1892, p. 495). 86Anjediva is now under the control of the Indian Navy and closed to visitors, including local Christians wanting to visit its two churches.I visited the neighbouring Kurumgad, another possible candidate for Tāranātha's island, in March 2016, only to discover that it was covered in Portuguese fortifications dating to the beginning of the 16th century and to be told that all the other habitable islands in the vicinity were similarly fortified. 87Tucci 1931, p. 692, n. 2.

Vajrayāna-Śaiva Interaction
Of the Konkan sites mentioned above, there are two where both Vajrayāna and Śaiva traditions flourished and which are thus possible locations for a transition from the former to the latter. 88

Panhale Kaji
At Panhale Kaji, which is eight kilometres inland from the Konkan coast about halfway between Mumbai and Goa, there are 29 rock-cut caves dating from approximately the 6th century CE onwards. 89s noted above, one of them houses a statue of the Vajrayāna deity Acala.Two others contain multiple images of Nāths, including depictions of Matsyendra overhearing Śiva teach Pārvatī the Kaula doctrine, Vir ūpāks .a, 90 groups of 9 (or perhaps 10) and 84 siddhas, and a relief of Tripurasundarī.An impressive but damaged statue of Goraks .a was also found at the site when it was restored in the 1970s. 91here are no textual or epigraphic sources to suggest that either the Vajrayāna or Nāth presence at Panhale Kaji was of wider significance, and the site fell into disuse from the 14th century onwards.

Kadri
It is the Kadri Mañjunātha temple and monastery complex that gives us our best evidence of links between Vajrayāna and Nāth traditions in the Konkan, with a wealth of material showing that both traditions flourished there. 92There is a gap of more than four centuries between the last firmly datable evidence of Vajrayāna Buddhism at the site and the first of a Nāth presence, but there is much to indicate continuity between the two.
The Kadri statuary is a key indicator of such continuity.The three bronzes housed in passages to the north and south of the central shrine of Mañjunātha (a svayambh ū or "self-born" li ṅga representing Śiva), are perhaps the finest Buddhist images still worshipped in India (albeit now as Hindu deities, on which see below).The inscription on the plinth of the single bronze in the corridor on the south side of the Mañjunātha shrine records how the Āl .upa king Kundavarman established an image of the god Lokeśvara in the Kadirikā vihāra on the 16th of January 1068 CE. 93 The iconography and workmanship of the image support this date and the two other bronzes appear to be of a similar age.
The three images are currently worshipped as Brahmā, Vis .n .u and Vyāsa, and have been since at least 1730 CE because they are identified as such in the Kadalīmañjunāthamāhātmya. 94Their 88 The location of a thousand-armed Avalokiteśvara at a site called Śivapura in a 1025 CE manuscript of the As .t .asāhasrikā Prajñāpāramitā (see p. 10) indicates another possible site of Vajrayāna-Śaiva interaction (I thank Andrea Acri for suggesting this in an email dated 24th January 2019).Bankar (2013) and Sarde (2016) describe early Buddhist sites in the Deccan which were later occupied by Nāths, but there is no evidence of Vajrayāna Buddhism at the sites, nor of direct links between the Buddhist and Nāth traditions. 89 Deshpande (1986, p. 121) dates the oldest caves at Panhale Kaji to before the 5th century, but Rees (forthcoming) has shown that a later date, probably in the latter half of the first millennium, is more likely. 90See figure 91 Deshpande 1986, pp. 146-48. 92 On Kadri, see Bouillier 2008, chp. 4 and Bouillier 2009; also Saletore 1937; Pai 1946; (Bhatt 1975, pp.287-97) and (Freeman 2006, pp.164-67). 93 As it stands this is an unlikely formulation which must correspond to the year 4069 (4000 + 68 + 1) of Kaliyuga, i.e., 967-68 CE, and this date has been repeated in most secondary literature on Kadri.Pai, however, without viewing the inscription itself, demonstrated that • gate must be a misreading of • śate (ga and śa are similar in the Grantha script in which the inscription is written), not only because • śate is much better Sanskrit, but also because the tithi given only makes sense if the year is 1068 rather than 968 (Pai 1946, pp.60-62).Dominic Goodall visited Kadri in 2017 and confirmed that the correct reading is • śate (personal communication 6th February 2017). 94Kadalīmañjunāthamāhātmya 13.10-11: uttarārāmamadhye tu sthito vis .n .ur manoharah .| iconography, however, is clearly Buddhist.The statue on the inscribed plinth is worshipped as Brahmā, but is in fact a form of the Bodhisattva Mañjuśrī, most likely Mañjuvajra (Figure 4). 95Of the two bronzes in the northern corridor, that which is worshipped as Vis .n .u is a form of the Bodhisattva Lokeśvara (better known as Avalokiteśvara), who can be identified by the representation of the cosmic Buddha Amitābha on his crown of matted hair, his antelope skin, his hand gestures and his sattvaparya ṅka seated position (Figure 5). 96ālaye mañjunāthathasya p ūrvottaragato harih .|| p ūrvadaks . in .ayor bhāge bhagavām .ś caturānanah .| tayor uttarabhāge tu vyāsah .satyavatīsutah .|| "In the middle of the northern grove is situated a delightful Vis .n .u.In the abode of Mañjunātha, to the northeast, is Hari.In the area to the southeast is the four-faced Lord [i.e., Brahmā].To the north of both of them is Vyāsa, the son of Satyavatī."I do not know what image is being referred to as Vis .n .u here; it may no longer be at the site. 95Alexis Sanderson first suggested this to me, in a meeting in December 2010.His identification was subsequently confirmed to me by Christian Luczanits (email communication 30 January 2018). 96I thank Christian Luczanits for this identification (email communication 26 April 2016).The bronze worshipped as Vyāsa is the Buddha, in a form which closely matches Sri Lankan, Tamil and some Southeast Asian Buddha images; his seated position with the legs half-crossed (sattvaparya ṅka) rarely occurs in Buddha images from India, apart from in the Tamil region (Figure 6).establishment of an image of Lokeśvara.It seems that at some point the image of Lokeśvara was removed from its original plinth and replaced with the image of Mañjuvajra. 98 In the Mangalore Museum is a stone statue of Matsyendranātha, the first human guru of the Nāth sam .pradāya (Figure 7). 99The statue is said in its label to be from the Kadri monastery and is cracked across its neck, which may account for its having been removed from its original location.The m ūrti shares many features with the Lokeśvara bronze: it is three and a half feet high; it is seated in sattvaparya ṅka, an easy cross-legged position; its right hand rests on its right thigh, palm open; 100 and it has a crown of jat .ā, matted locks, on which is depicted an indistinct icon.This latter is of particular significance, because no other non-Buddhist sculpture in India includes such an element.
Today the Matsyendra shrine in the Kadri monastery houses an apparently newer and whole m ūrti of Matsyendra, which is very similar to that found in the museum and was presumably made to replace it (Figure 8). 98Comparison of photographs of the image show that its composition has been altered in recent years, and confirm that the m ūrti is not of one piece with the plinth.The photographs in the Government of Madras's Annual Report for Epigraphy for the Year Ending 31 March 1921 (Plate 1), Bhatt (1975, plates 300 and 301) and Shetti (1988, fig.10) show the prabhāvalī (halo) behind Mañjuvajra higher than it is today, and the image now has a metal sheet between it and the plinth.The sheet has small posts at its rear which support the prabhāvalī in its new position.In the oldest photograph Mañjuvajra holds a lotus flower in each of his middle left and right hands as he does today, while in the photographs found in Bhatt 1975 and  Shetti 1988 the lotuses are not in Mañjuvajra's hands but in those of his two attendants. 99The image is identifiable as Matsyendra (whose name means "lord of fish") because the subject is seated on a fish. 100The Lokeśvara image has four arms and his other right hand may have held something, but is now empty.Matsyendra, being human-born, has only two arms.There are two small but significant differences between the two Matsyendra images.In both statues Matsyendra wears on a thread around his neck a si ṅgī, the small horn signifying membership of a Nāth lineage.That on the older statue is shaped like an antelope's horn, while that on the later statue is a whistle like that worn by yogis of today's Nāth sam .pradāya.Mughal miniatures depict the earlier style of horn from the mid 16th-century onwards; the whistles are not seen until the 18th century. 101n its label in the museum, the older image is dated to the 10th century.I do not know the grounds for this dating, but Bhatt (1975, plate 302(b)) gives the same date so may be its source.He accepts the incorrect reading • gate found in the published transcription of the inscription on the plinth of "Lokeśvara" (i.e., Mañjuvajra) so takes the image to be dated 968 CE and proposes the 10th-century date for Matsyendra because he believes the Nāth tradition to have predated Buddhism at the site.I see no need to propose such an early date for the Matsyendra image in the Mangalore Museum.The other difference between the two images of Matsyendra is that the icon on the crown of jat .ā on the older image is unclear (Figure 9), unlike that of the newer one (Figure 10).Images of Bodhisattvas wear small representations of one of the five Tathāgatas, indicating their lineage.The cosmic Buddha on Lokeśvara is Amitābha (Figure 11) and that on Mañjuvajra (and on his prabhāvalī) is Aks .obhya (Figure 12).The icon on the newer Matsyendra is clearly a four-armed deity with the jat .āmukut .a, club and fish seat of the Matsyendra image itself, together with the d .amaru drum usually carried by Śiva, especially in his Bhairava form.Pai (1946, p. 64) identifies this icon as Ādinātha, i.e., the form of Śiva seen as the founder of the Nāth sam .pradāya.The fish seat, however, which is also faintly identifiable on the older Matsyendra image in the Mangalore Museum, indicates that this is an image of Matsyendra himself, while his four arms suggest that he has been deified.The 1068 CE inscription on the plinth supporting the image of Mañjughos .a at the Mañjunātha temple says that an image of Lokeśvara was established in the Kadirikā vihāra, i.e., a Buddhist monastery at Kadri.No mention of a separate temple is made in the inscription, but today there are two sites at Kadri, the Mañjunātha temple and, a few minutes walk up a neighbouring hill, the Kadri mat .ha or monastery.It is unclear whether the Kadri temple and monastery have always been separate sites.The first reference to the existence of both is from the report of a visit to Mangalore of Pietro della Valle in 1624 CE.102There are meditation caves at the monastery which date to approximately the 12th to 13th centuries CE. 103 The oldest inscription in the temple courtyard is dated śaka 1308, i.e., 1385 or 1386 CE.104 Today the temple is in the control of Mādhva priests and the Nāth presence is confined to the monastery,105 but three stone statues in the middle of the south, west and north exterior walls of the inner shrine of the temple which sport yogi iconography (earrings, jat .ā and, for two of them, a cross-legged seated position) are said in accompanying labels to be Matsyendranātha (Figure 13), Goraks .anātha (Figure 14) and Caura ṅginātha (Figure 15).106Bhatt dates these images to the 14th or 15th century, 107 which seems plausible and renders the ascription of a date of the 10th-century to the image of Matsyendra in the Mangalore Museum even more unlikely, since that statue has the si ṅgī or horn otherwise found only in north Indian depictions of Nāth yogis from the 15th century onwards, and which is not found on the three images in the 107 Bhatt 1975, p. 299 and plates 303 and 304(a).
Mañjunātha temple.The statues indicate that there was previously a stronger Nāth presence at the temple than there is now.
The Āl .upa king Kundavarman, who the 1068 CE inscription says established the image of Lokeśvara, is described in the sixth verse of the inscription as pādāravindabhramaro bālacandraśikhāman .eh ., "a bee at the lotus feet of he whose crest-jewel is a young moon", i.e., a devotee of Śiva.There are no indications that the region's primary religion was ever Buddhism; when Kadri was Buddhist it would have been an enclave within a Śaiva kingdom.Today the central shrine of the temple at Kadri contains a self-born li ṅga worshipped as a form of Śiva called Mañjunātha.This name of Śiva is attested nowhere other than the Tul .u region (and possibly its environs), 108 where it is understood to mean "the gentle Lord", mañju meaning "gentle" in Sanskrit.It is of course also redolent of Mañjuśrī (or Mañjuvajra, the form in which Mañjuśrī is found at Kadri), for which it is an alternative name, 109 and it is likely that the primary focus of worship at Kadri when it was Buddhist was the Bodhisattva Mañjuvajra, a sādhana of whom was written there by Jñānapāda in the ninth century 110 and an icon of whom was installed there in the eleventh century.Within two centuries of the establishment of the Lokeśvara image at the Kadirikā vihāra, the temple may already have become a Śaiva shrine: in an inscription of the Āl .upa queen Ballamahādevī who ruled 1277-92, she calls herself a worshipper of Machinātha. 111Bhatt sees this as possibly the first reference to Śiva as Mañjunātha. 112achinātha could alternatively refer to Matsyendra, whose name is a compound of matsya, "fish", and indra, "lord": ma[c]chi is a middle Indic form of matsya, and nātha a synonym of indra.Two late 14th-century inscriptions at Kadri mention the worship of Mañjinātha 113 and della Valle, who visited the site in 1624, says that the temple deity (which he was not allowed to see) was called "Moginato". 114n contrast, two copper plate Tul .u inscriptions dated 1329 CE kept inside the Kadri monastery mention Kadire Mañjunātheśvara and Kadre Śrī Mañjunāth. 115he specific referents of the name Mañjunātha and its variants as found in historical sources pertaining to Kadri are thus often uncertain, but they are likely to have referred, in sequence, to the Buddhist deity Mañjuvajra, the Nāth Matsyendra, and a localised form of Śiva.
Textual, material and epigraphic sources point to Kadri's importance as a centre of Vajrayāna Buddhism from the early 9th to late 11th centuries, and indicate that it had become Śaiva by the early 14th century at the latest.The similarities between the statues of Lokeśvara and Matsyendra from the site indicate a direct continuity between Vajrayāna and specifically Nāth Śaiva traditions, but the Nāth presence is not confirmed by epigraphic evidence until the late 15th century, and the details of the transition from Buddhism to Śaivism are not inferrable from material and epigraphic sources.If we return to our textual sources, however, some further information may be obtained.
The c. 1400 Telugu Navanāthacaritramu of Gauran .a tells the stories of seven of the nine Nāths of its title, including Matsyendra, who is the guru of the other eight, and Mañjunātha, the only one of the eight to be identified as the son of Matsyendra.Mañjunātha's story takes place in Mangalore116 and is one of the earliest tellings of a famous Nāth legend found in many different sources and versions. 117atsyendra travels to Mangalore with a group of his disciples.On their arrival the king of the city dies.His minister Prabuddha conceals his death, but Matsyendra, through his yogic sight, knows what has happened.He decides to leave his body and enter that of the king in order to experience worldly pleasures and thereby confirm that the way of the yogi is superior to worldly life.He uses a yogic technique to enter and reanimate the body of the king.Prabuddha realises what has happened but decides not to tell the people and advises Matsyendra on how to rule.Matsyendra enjoys sexual relations with his queens, and fathers a son by the chief queen.His disciples, who have been guarding his body in a mountain cave in the meantime, realise that he has forgotten himself and Goraks .a is sent to rescue him.Goraks .a manages to convince Matsyendra to shed all his worldly attachments with the exception of his love for his son.Goraks .a kills the boy and the shock brings Matsyendra to his senses.He returns to his body, Goraks .a revives the prince, and Prabuddha joins Matsyendra after arranging the king's succession.
Matsyendra's son is initiated as a siddha, placed on a k ūrmāsana and given the name Mañjunātha.Leaving him there, Matsyendra and the other siddhas go to a cave in the Narendra mountains, 118 where Prabuddha is initiated, thus becoming one of the nine Nāths of the text's title, and given the name Buddhasiddha.Matsyendra instructs his disciples in yoga and they go on a tour of the the holy sites of north India to disseminate his teachings.
The story is in part an origin myth for the shrine of Mañjunātha at Kadri.Mañjunātha is placed on a k ūrmāsana, which Śaivasiddhānta Pratis .t .hātantras identify as the altar on which an image of a deity is to be installed. 119He is left in Kadri, while Matsyendra and his disciples continue to have adventures across India.Goraks .a's rescue of his guru Matsyendra from a life of debauchery is an allegory of his reformation of Nāth religious practice, i.e., the triumph of his celibate hat .hayoga over Matsyendra's Kaula sexual rites. 120Sexual ritual was central to the practices of the Vajrayāna traditions which flourished at Kadri; the Kadri Matsyendra legend as told in the Navanāthacaritramu may also reflect the specific takeover by celibate Nāth yogis of the Kadri Vajrayāna tradition.There are faint traces of Buddhism in the Navanāthacaritramu.Gauran .a mentions two types of yoga, a royal (rāculī) method as practised by Matsyendra, and a "peerless" (anuturya) method which Matsyendra teaches to Caura ṅgi. 121The name anuturya may be a contraction, metri causa, 122 of anuttarīya, which is unattested as a name for a type of yoga in Sanskrit but which, as anuttara, has parallels in Tibetan works. 123The element buddha in the pre-and post-initiation names of the minister of the Kadri king, Prabuddha and Buddhasiddha, is also suggestive of a connection with Buddhism.
If we turn now to doctrinal texts, we find further details of the transition from Vajrayāna Buddhism to Nāth Śaivism at Kadri.This is demonstrated most clearly by the parallels between the Buddhist Amr .tasiddhi and the Śaiva Amaraughaprabodha.The Amaraughaprabodha borrows five verses from the Amr .tasiddhi and paraphrases it extensively elsewhere. 124As shown above, the Amr .tasiddhi's place of composition is uncertain but is likely to have been in the Deccan.There is similarly no firm evidence for the place of composition of the Amaraughaprabodha, but its dependence upon the Amr .tasiddhi suggests a southern origin, which is supported by all its extant manuscripts being written in southern scripts and further corroborated by parallels with other southern texts and traditions, which will be explored below.
In addition, the new formulation of yoga taught in the Amaraughaprabodha may be traced to both the Amr .tasiddhi and tantric Buddhist traditions of hat .hayoga which are known to have been practised at Kadri.The Buddhist scholar Jñānapāda, who, as we have seen, is likely to have resided at Kadri during the first half of the ninth century, wrote a sādhana of Mañjuvajra, who is the chief deity of the Guhyasamāja in his teaching125 and is depicted in the most spectacular of the three Kadri bronzes.The Guhyasamājatantra advocates hat .hayoga, albeit as a fallback method of achieving awakening. 126he Amaraughaprabodha draws on this understanding of the term hat .hayoga when it uses it as the name of the yoga method of the Amr .tasiddhi, which, it states, is subordinate to rājayoga.As suggested by the primacy of celibate practice implied by the Navanāthacaritramu's legend of Matsyendra's rescue by Goraks .a (who, like all his co-disciples, is avowedly celibate), the Amaraughaprabodha teaches that the control of bindu, i.e., semen, is central to hat .hayoga.It refers to the practice of vajrolimudrā, a method of controlling semen which is likely to have been part of certain traditions of Vajrayāna sexual ritual, but makes no mention of its use in such rites, saying that in the Amara tradition it is equanimity and the flowing of the breath in the central channel. 127This transition from Vajrayāna to Śaiva hat .ha practice evident in the Amaraughaprabodha is echoed in the story from the Līl .ācaritra referred to above,128 in which Vir ūpa, the first teacher of the yoga of the Amr .tasiddhi, is said to switch allegiance from the Vajra to Amara lineage after his conversion by Goraks .a.The polyvalence of the term ogha in the name of the Amaraughaprabodha (whose teachings are attributed to Goraks .a), allows it to mean both "Awakening (prabodha) by the Stream (ogha) of the Nectar of Immortality (amara)" and "Awakening in the Amara lineage (ogha)".
The strongest indication of a link between Kadri and the Amaraughaprabodha is found in the siddhas it invokes.It opens with a ma ṅgala verse to Ādinātha, Mīnanātha,129 Caura ṅginātha and Siddhabuddha, and in the next verse ascribes its teachings to Goraks .a. 130 Ādinātha is the name of the form of Śiva to which many Nāth lineages trace their origin.The triad of Mīna, Goraks .a and Caura ṅgi has primacy only in Nāth lineages from the Deccan and Konkan.Caura ṅgi rarely features in Nāth traditions from further north 131 and no Sanskrit hat .ha texts apart from the Amaraughaprabodha, Hat .hapradīpikā and derivative works mention him.Matsyendra, Goraks .a and Caura ṅgi are the main protagonists in both the Navanāthacaritramu and Kadalīmañjunāthamāhātmya. 132 The invocation of these three Nāths thus supports a southern origin for the Amaraughaprabodha, with Kadri a strong candidate not only because they are central to its māhātmya, but also because, as we have seen, the same triad adorns the main shrine of Mañjunātha.
The mention of Siddhabuddha in the Amaraughaprabodha's ma ṅgala verse points more specifically to Kadri as the location of that text's composition.Unlike the other three siddhas named in the verse, Siddhabuddha is little known; within the Sanskrit hat .ha corpus his name is otherwise found only in the list of 29 siddhas in the Hat .hapradīpikā and texts deriving from it. 133The Hat .hapradīpikā borrows twenty verses directly from the Amaraughaprabodha and may have taken Siddhabuddha's name from its ma ṅgala verse.Arguing against this possibility, however, is that in the Hat .hapradīpikā's list of siddhas, Siddhabuddha is not grouped with the other three siddhas named in the Amaraughaprabodha's ma ṅgala verse.A more likely source, or milieu, for the Hat .hapradīpikā to have taken the name Siddhabuddha from is suggested by the only other texts known to me in which he is mentioned.These are two Telugu works, the c. 1400 CE Navanāthacaritramu and the contemporaneous Vikramārkacaritramu of Jakkana.The Hat .hapradīpikā has several parallels with the Navanāthacaritramu, in particular in the names of the siddhas found in both texts, 134 some of which are otherwise obscure.As we have seen, the Navanāthacaritramu includes Buddhasiddha among the nine Nāths of its title.That this Buddhasiddha is simply Siddhabuddha with the elements of his name transposed is indicated by his name being found as Siddhabuddha in the Vikramārkacaritramu, which gives a list of nine Nāths almost identical to that of the Navanāthacaritramu (Mañjunātha is replaced by Ādinātha).A further detail supports the identification of the Siddhabuddha of the Amaraughaprabodha with the Buddhasiddha/Siddhabuddha of the Navanāthacaritramu and Vikramārkacaritramu.In the Navanāthacaritramu, before his initiation Buddhasiddha is Prabuddha, the crafty minister of the king of Mangalore.The Vikramārkacaritramu highlights a quality of each of the nine Nāths.Siddhabuddha comes after Ādinātha, Matsyendra, Sāra ṅga (the pre-initiatory name of Caura ṅgi) and Goraks .a, and is praised for his intellect. 135In the Amaraughaprabodha's ma ṅgala verse Siddhabuddha is the only siddha to have an epithet, which is dhīmate, "the wise one".These parallels allow for a tentative identification of the Siddhabuddha of the Amaraughaprabodha's ma ṅgala verse with the Buddhasiddha of the Navanāthacaritramu, who resided at Kadri.

Final Remarks
Our available evidence indicates that the transition from Buddhism to Śaivism at Kadri was peaceful.No textual sources related to Kadri suggest a violent takeover and the Vajrayāna bronzes at the site remained more or less in situ without being damaged. 136Furthermore, the later Nāth statues of Matsyendra are modelled on that of the Buddhist Lokeśvara, indicating direct continuity between the Nāth tradition at Kadri and its Vajrayāna forerunner.Other than the icon on Matsyendra's crown of matted locks, however, all material traces of Buddhism were removed, or, as in the case of the three bronzes, reassigned as Hindu.This is true also for the Amaraughaprabodha, which, in adopting and His story was current in Bengal and Mithila (it is told in the Gorakh Vijay cycle) and is popular in Punjab, where he is known as P ūran .Bhagat (see e.g., White 1996, pp.298-99).Caura ṅgi's legend is also found in the Grub thob brgyad cu rtsa bzhi'i lo rgyus, a 12th-century Tibetan account of the lives of the 84 siddhas (the story is similar to that taught in the Navanāthacaritramu) and he is mentioned in the 13th-century Marathi Jñāneśvarī (pp.1730-40). 132The bulk of the Kadalīmañjunāthamāhātmya, chp.15-45, is devoted to their exploits. 133Hat .hapradīpikā 1.6. 134Mallinson forthcoming b. 135 Jones 2018, p. 199, n. 7: siddhabuddhuni buddhicittam . bunam . jerci. 136 Similarly, the Nāth presence at Panhale Kaji did not result in the destruction or removal of Buddhist statuary.
adapting the teachings of the Amr .tasiddhi, removes features that are obviously Buddhist but leaves traces of more obscure Buddhist doctrines which were not recognised by its redactors. 137he Śaiva tradition which took over at Kadri thus absorbed elements of the earlier Vajrayāna tradition but tried to efface anything overtly Buddhist.This is in contrast with the Newar tradition in the Kathmandu Valley, where Avalokiteśvara/Lokeśvara and Matsyendra have been identified since at least the 15th century, and Buddhism and Śaivism remain inextricably entangled.13819th-century chronicles claim an 11th-century Karnatakan origin for the Newar kings and mention connections with the Konkan,139 hinting at a possible origin in south India for the Newari identification of Avalokiteśvara and Matsyendra, and at links with the Kadri tradition, but the earliest record of their identification in Nepal whose date is certain is from the 15th century. 140Buddhism's survival in the Kathmandu Valley is in stark contrast with its almost complete demise in India.There is no clear reason for either of these outcomes, but the Kathmandu Valley's isolation was responsible for other divergences from the history of the subcontinent, such as its not being subjugated by the Mughals or British, and is likely also to be at least partly responsible for Buddhism's survival there.
The apparently peaceful transition at Kadri does not support arguments that Brahmanism or Islam violently removed Buddhism from India. 141Textual and epigraphical sources do however suggest that interactions between tantric Buddhists and Śaivas in the region of Kadri were not always friendly.Desai notes two late 12th-century south Indian inscriptions from Karnataka which indicate that Śaiva aggression may have been responsible for the destruction of Buddhist vihāras. 142And, as we have seen above, several texts tell the story of Vir ūpa himself destroying a śivali ṅga at Drāks .ārāma. 143he Nityāhnikatilaka, a text of the Paścimāmnāya Kaula tradition from which some Nāth lineages developed, reports that Unmanīśanātha destroyed 125,000 Buddhist vihāras in heaven, as a result of which nine Buddhists fell to earth and were subsequently initiated by him. 144ut these hints of a violent transfer are exceptional.Other sources point to debate or instruction as the means by which one doctrine superseded the other.Unsurprisingly, the Śaiva sources at our disposal almost always tell of Buddhists capitulating to Śaiva arguments.Of particular relevance to the topic in hand is a story in the Līl .ācaritra of Vir ūpa, after his conversion by Goraks .a from Buddhism to Śaivism, 145 defeating in debate ks .apan .akas, i.e., Jain or Buddhist ascetics, and re-establishing the Vedic religion at Ellora. 146Kularatnoddyota 11.31 includes the Buddha (and the Jina and Matsyendra as well as many others) in a list of siddhas who were taught the Kaulajñāna of the P ūrvāmnāya by Mitranātha. 147I know of one instance in which the transmission went the other way, suggesting a willingness by Śaiva traditions to accept at least some Buddhist teachings: the S .at .sāhasrasam .hitā, a Paścimāmnāya Kaula Śaiva text, tells the story of eight Nāths being given teachings by a teacher whose name, Vajrabodhi, indicates a Buddhist affiliation. 148he demise of Buddhism, at Kadri and elsewhere, thus seems to have usually been peaceful.It may in fact have been imperceptible to all but a few.Boundaries between Śaiva and Vajrayāna siddha traditions were porous.Some siddhas were clearly originally either Śaiva or Vajrayāna, 149 but others cannot be tied to one or other tradition. 150The Amr .tasiddhi has some features which definitively identify it as Buddhist, such as a metaphysics that has only four basic elements, not the five of Hindu traditions, but it is also peculiarly Śaiva for a Buddhist work, including such un-Buddhist notions as jīvanmukti, liberation while living. 151Patronage and devotion were similarly non-partisan.We have seen how the Kadri Lokeśvara inscription records the installation of a Buddhist icon by a Śaiva king.Further north in the subcontinent, some Gāhad .avāla and Pāla monarchs had both Buddhist and Kaula gurus. 152Thus there was a shared siddha world, especially of praxis, from which, in India, the Śaiva siddha tradition emerged, mostly free from Buddhist vestiges.
The above evidence for the existence of Vajrayāna Buddhism in the Konkan is compelling, but scant.Vajrayāna was never a state religion in the Tul .u region, whose rulers were Śaiva, but the 1068 CE Kadri inscription shows that it was patronised by at least one Tul .u monarch.It may also have had patrons from further afield.Gomin Avighnakāra, a Buddhist devotee from Gaud .a, made an endowment at Kanheri in the northern Konkan in 854 CE 153 and it is not impossible that Kadri was also supported by patrons from Gaud .a.But there is much more evidence for a Tamil connection.Unlike all other inscriptions at Kadri, which are written in the Kannada script and make no reference to Buddhism, the 1068 CE inscription recording the installation of Lokeśvara is in Grantha, the script used in the Tamil region for writing Sanskrit.In the c. 13th-century Matsyendrasam .hitā, which was composed in either the Tamil or Konkan regions, 154 Goraks .a is a Col .ā king before he becomes a yogi. 155The Tamil Tirumantiram, whose composition, like the installation of Lokeśvara at Kadri, 148 S .at .sāhasrasam .hitā ff.351v-353r, for an edition of which by Alexis Sanderson and me, see https:\discretionary{-}{}{}//www.\discretionary{-}{}{}academia.\discretionary{-}{}{}edu/35397806/Mallinson$_$Harvard$_$Talk$_$Nov$_$2017_Handout.The instruction takes place at Arbuda, i.e., Mount Abu.There are no material remains at Abu from the period under consideration which indicate either a Buddhist or Nāth presence, but there is material evidence for both traditions at Taranga, a site 70 kilometres due south of Abu best known for its magnificent 13th-century Jain Ajītnāth temple.On the opposite side of the Siddhaśilā hill at Taranga from the Ajītnāth temple is a small temple complex with two shrines containing images now worshipped as Hindu goddesses called Dhāran .Mātā and Tāran .Mātā.Dhāran .Mātā is flanked by six slightly smaller sculptures.The priests at the shrine dress her in the attire of a goddess, but she and the others are all male figures whose iconography indicates that they are Buddhist, with some dating to perhaps the 6th century (I thank Ken Ishikawa for this tentative dating).Tāran .Mātā is a beautiful female image who has been identified as Kurukullā (Sompura 1969, pp.29-30); see also (Sompura 1968, p. 72) and (Rawat 2009), but is in fact the Buddhist goddess Tārā, as can be inferred from her retinue (I thank Christian Luczanits for this identification (personal communication 21st June 2017)).
On the hill itself is a cave, known locally as jogan dā guphā, "the cave of the yogis", which has a small makeshift shrine centred on a freestanding panel depicting in relief four Buddhist images of seated meditating figures.On the top of the hill is a small temple from perhaps the 19th-century (I thank Crispin Branfoot for confirming my tentative dating), which now houses the pādukā of a Jain saint but has eight statues of Nāth yogis on its roof, suggesting that it was originally a Nāth shrine.Tāranātha mentions a yoginī disciple of Śāntigupta (his guru's guru) called Tāra ṅgā, who travelled widely, including to the Konkan (Templeman 1983, p. 93). 149E.g.Matsyendra, who is mentioned in the early 11th-century Tantrāloka as the propounder of the Śaiva Kaula doctrine in the kali age (1.7, 29.32 and 30.102), or Vir ūpa, who is Buddhist in early sources such as the Grub thob brgyad cu rtsa bzhi'i lo rgyus and the Līl .ācaritra. 150E.g.Goraks .a, whose earliest textual mentions, which date to the 13th century, are found in both Śaiva and Buddhist texts (Mallinson 2014, p. 233, n. 28). 151Thus for Schaeffer, the first scholar to study closely the Amr .tasiddhi, it "cannot be comfortably classified as either Buddhist or non-Buddhist" (2002, p. 515). 152See (Szántó 2017, pp.7-8) on the Gāhad .āvala Jayacandra, and the Nityakaumudī (NGMPP B35-26 and NAK 4/324), which is a commentary on the Śaiva Vīracandra's 1072 CE Nityaprakāśa whose colophon says that it was commissioned by the Pāla monarch Rāmapāla's guru, Śambhudatta.I thank Péter-Dániel Szántó for this reference. 153(Gokhale 1990, p. 70); see ibid., p. 10 for more on connections between Kanheri and Gaud .a. 154 See footnote 50. 155Matsyendrasam .hitā pat .alas 1 and 55.
dates to the second half of the 11th century CE, 156 has both Buddhist and Śaiva elements in its yoga, and several features of its cult are shared with the Kadri yogi tradition. 157The bronzes of Lokeśvara and Mañjuvajra at Kadri are similar to contemporaneous Col .ā pieces. 158The Kadri Buddha may also be from the Tamil region, or perhaps Sri Lanka or southeast Asia, bringing to mind Mangalore's importance as a trading port (whose cosmopolitanism in the 12th-century has been vividly reimagined by Amitav Ghosh). 159A connection between Kadri and Sri Lanka is supported by the inclusion of Matsyendranātha among a list of eight forms of Avalokiteśvara (who in Sri Lanka is also commonly known as Nātha) described in the Śariputra, a text of uncertain date used by the makers of images in Sri Lankan temples. 160In addition, the Guhyasamāja exegete Jayabhadra, who spent some time in the Konkan before travelling to Vikramaśīla, was originally from Sri Lanka. 161Kadri was at the western extremity of a Buddhist maritime world whose hub was Sri Lanka and which extended to southeast Asia.The possibility of Nāth influence on the mixed Śaiva-Buddhist traditions of Java and Bali is hinted at by the presence of an image of Matsyendra on a statue of Cāmun .d .ī in Java dated 1292. 162The links between Kadri and the Tamil sphere of influence point to Kadri being supported by the Col .ā empire, which would have been connected to Mangalore by sea and land: the most important centres of Buddhism in the Col .ā empire were the ports of Kanchi and Nagapattinam, and its territory reached near Mangalore until its demise in the 14th century.Like Kundavarman, who established the Lokeśvara image at Kadri, certain Col .ā rulers such as Rājarāja I (pp.985-1014 CE) were Śaivas who also patronised Buddhism. 163The Kadri vihāra may thus have benefited from Col .ā patronage as well as that of its local overlords.When these sources of support dried up, either as the result of increasingly exclusive patronage of Śaivism by both the Āl .upa 164 and Col .ā kings or the demise of the Col .ā empire, the Kadri Buddhist tradition was left isolated and unsupported. 156The Tirumantiram mentions the Kālacakra[tantra] (Tirumantiram "III chp.14"; see (Venkatraman 1990, p. 193) for this reference, which I have been unable to confirm).The Kālacakratantra may be dated to between 1025 and 1040 CE (Newman 1998).The Tirumantiram is cited in a commentary on the Yāpparu ṅgalakkārigai by Gun .asāgara, who was active c.1100 CE (Venkatraman 1990, p. 193).A text-critical study of the Tirumantiram remains a desideratum; if the text as it is currently constituted does not contain additions to its earliest layer, it may thus be dated to the second half of the 11th century CE. (On the implausibility of the very early datings of the Tirumantiram often found in secondary sources, see Goodall 1998, p. xxxvii, n. 85 and Goodall 2004, pp.xxix-xxx.) 157The Tirumantiram makes no mention of the Nāth tradition but its doctrinal parallels with the Matsyendrasam .hitā lead Kiss (forthcoming, pp.51-52) to conclude that the two texts are "not completely unrelated".The Tirumantiram's yoga method has much in common with that of the Amr .tasiddhi, but, like the hat .hayoga of other Vajrayāna texts and in contrast with the celibate yoga of the Amr .tasiddhi, includes sexual intercourse without ejaculation.(In the sexual ritual taught in pat .ala 40 of the Matsyendrasam .hitā the yogi is to ejaculate.)The Tirumantiram mentions the Buddhist Kālacakratantra (see note 156) and celebrates the Śaiva site of Chidambaram, which is also a cultic centre for the Keralan caste yogis who moved south from Kadri in perhaps the 17th century (Freeman 2006, pp.172-73). 158The Kadri Lokeśvara, for example, resembles a Col .ā Śiva sold by Christies in New York in 2015: https://www.christies.com/lotfinder/Lot/a-large-bronze-figure-of-shiva-south-5875978-details.aspx 159Ghosh 1994. 160 Kasthuri (2016, p. 160) notes that the Śāriputra has been variously dated to between the 5th to 15th centuries.Its mention of Matsyendra indicates that it was composed in the latter few centuries of this period.Paranavitana (1928, pp.60-62)  Avalokiteśvara/Nātha is identified with eight Nāthas in the Śāriputra: Śiva Nātha, Brahma Nātha, Vis .n .unātha, Gaurī Nātha, Matsyendra Nātha, Bhadra Nātha, Bauddha Nātha and Gan .a Nātha (ibid.).Additionally, Avalokiteśvara's role as the protector of mariners in Sri Lanka (Bopearachchi 2014, p. 182), is echoed by Matsyendra's identification as a fisherman in many of the various legends associated with him. 161Jayabhadra gives his place of birth in his Pañjikā (Szántó 2016a, p. 2). 162See (Scheurleer 2008, p. 292, n. 21).I thank Andrea Acri for drawing my attention to this statue (personal communication 24th January 2019). 163(Veluppillai 2013, pp.65-77); see also (Acri 2018, p. 13). 164 Saletore (1936, pp.384-85) argues that the strong Śaivism of Tul .u rulers was responsible for disappearance of Buddhism from the Tul .u region.
Padmanabh Jaini, analysing the very different historical outcomes for Buddhism and Jainism in India and citing the example of Kadri, argues convincingly that Hinduism was able to assimilate Buddhism -and not Jainism -because the Bodhisattvas that were central to Buddhist devotional cults could be reconfigured as Śaiva deities and siddhas. 165The Vajrayāna cult at Kadri had the Bodhisattva Mañjuvajra as its central deity, hence Jñānapāda's writing a sādhana of him there in the ninth century and the name of the temple being Mañjunātha, an alternative for Mañjuvajra.But then the Śaiva king Kundavarman established an image of the Bodhisattva Lokeśvara in the temple.Lokeśvara's iconography is closely related to that of Śiva -with whom he shares a dual role as ruler and ascetic yogi166 -and thus also to that of Matsyendra, who subsequently became the central object of worship at the Kadri monastery.Legend has it that Matsyendra used the yogic technique of parakāyapraveśa, entry into another's body, to revive the dead king of Kadri; historical sources indicate that Kundavarman's act of inclusive benevolence at Kadri paved the way for a moribund tantric Buddhism to use the same technique to achieve the yogic aim of kālavañcana, cheating death, by entering both the iconic body of Matsyendra and the corpus of teachings attributed to his putative disciple Goraks .a.
Funding: The research for this article was carried out as part of the Hatha Yoga Project (hyp.soas.ac.uk).This project has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement no.647963).

Figure 6 .
Figure 6.Buddha, Mañjunātha temple, Kadri.These identifications are well supported by comparison with other such images.Muddying the waters, however, is the inscription on the plinth of the Mañjuvajra image, which records the