The Amaraughaprabodha: New Evidence on the Manuscript Transmission of an Early Work on Haṭha- and Rājayoga

The Amaraughaprabodha is a Sanskrit Śaiva yoga text attributed by its colophons to Gorakṣanātha. It was first published by Kalyani Devi Mallik in 1954 and has been discussed in various secondary sources. Most notably, Christian Bouy (1994, pp. 18–19) identified this work as a source text for the Haṭhapradīpikā of Svātmārāma (mid-fifteenth century). This article presents new manuscript evidence for a shorter recension of the Amaraughaprabodha than the one published by Mallik. Comparing the differences between the short and long recensions reveals that the structure of the shorter one is more cohesive and closer to the original design of the work. The close relationship of the Amaraughaprabodha's short recension with an eleventh-century Vajrayāna work on yoga called the Amṛtasiddhi provides unique insights into how early teachings on Haṭhayoga were formulated. Although the practice of the physical techniques is largely the same in both texts, the author of the Amaraughaprabodha removed or obscured Vajrayāna terminology, added Śaiva metaphysics and framed Haṭhayoga as subordinate to a Śaiva yoga known as Rājayoga. This article proposes that the Amaraughaprabodha's short recension is probably the earliest known work to combine Haṭha- with Rājayoga, on the basis of this recension’s close relationship with the Amṛtasiddhi, its rudimentary nature and the likelihood that Svātmārāma used it, and not the long recension, for composing the Haṭhapradīpikā.


Introduction
The Amaraughaprabodha, which literally means 'awakening a flood of nectar', is a Sanskrit yoga text that attributes its teachings to Gorakṡanātha, the alleged founder of the Nātha order and a physical type of yoga called Hat˙hayoga. This text was first published in 1954 by Kalyani Devi Mallik, whose edition is a transcription of one manuscript. The text has seventy-five verses and has been dated by Bouy (1994, pp. 18-19) and others as being prior to the mid-fifteenth century, on the basis that Svātmārāma, the author of the Haṭhapradīpikā, borrowed verses from it (Bouy 1994, p. 19). 1 This article aims to reassess these conclusions in light of newly discovered manuscript evidence which indicates that two recensions of the Amaraughaprabodha exist; a longer one, as published by Mallik, and a shorter one that is preserved by two unpublished manuscripts. An analysis of the manuscript transmission and the differences between the recensions reveals that the shorter recension is the older of the two and was probably the one known to Svātmārāma. Its rudimentary nature and close relationship with an eleventh-century Vajrayāna work called the Amṛtasiddhi make it probable that the short recension of the Amaraughaprabodha was one of the earliest works to teach a fourfold system of yoga that combined Hat˙ha-with Rājayoga. The article concludes by discussing the significance of these findings within the broader history of yoga.
Previous Attempts to Date the Text Bouy (1994, p. 19) examined the Amaraughaprabodha in Mallik's edition and identified twenty-two and half of its verses in the Haṭhapradīpikā. 2 In spite of the fact that Svātmārāma does not reveal the names of his sources, Bouy proposed that Svātmārāma borrowed the Amaraughaprabodha's verses by demonstrating that the Haṭhapradīpikā is an anthology (1994, pp. 80-86). If one accepts this logic behind the direction of borrowing, the Amaraughaprabodha was composed before the midfifteenth century. Bouy (1994, p. 19) also notes that the Upāsanāsārasaṅgraha, which he dates from the sixteenth to seventeenth century (1994, p. 91), cites the Amaraughaprabodha by name. This provides a certain, albeit more recent, terminus ad quem. 3  has identified at least five verses of the Amaraughaprabodha in the eleventh-century Amṛtasiddhi. 4 Furthermore, the version of the Amaraughaprabodha in Mallik's edition has borrowed a verse from the second chapter of the Amanaska (Birch 2011, p. 528), which can be dated to the eleventh or early twelfth century (Birch 2014, p. 406 n. 21), and another from the Dattātreyayogaśāstra, circa thirteenth century. 5 There is also a verse cited and attributed to the Śrīsampuṭa 6 and a short passage attributed to the Amaraughasaṃsiddhi. 7 These borrowings indicate that the Amaraughaprabodha in Mallik's edition is a compilation, the terminus a quo of which was the Dattātreyayogaśāstra, bearing in mind that the Śrīsampuṭa and the Amaraughasaṃsiddhi are unknown works. These observations led me to propose in an earlier publication (Birch 2011, p. 528) that the Amaraughaprabodha was probably composed in the fourteenth century, because it must have appeared after the earliest Hat˙ha-and Rājayoga texts and before the Haṭhapradīpikā. The discovery of new manuscript evidence requires that these conclusions be revised.

Authorship
Among the earliest modern publications that mention the Amaraughaprabodha in any detail are the first volume of Madras University's New Catalogus Catalogorum (1949) and Mallik's edition (1954). Both attribute it to Gorakṡanātha. Before these publications, the Amaraughaprabodha is absent in lists of Gorakṡanātha's works by modern scholars (e.g., Briggs 1938, pp. 251-257 andDvivedī 1950, pp. 98-100) and in studies on the Nāths (e.g., Dasgupta 1946, pp. 219-294). However, it has been included in more recent lists (e.g., Banerjea 1962, pp. 26-28, Gonda 1977 and studies (e.g., Bouy 1994, pp. 18-19, White 1996. The attribution of authorship to Gorakṡanātha is supported by the final colophon of the manuscript used by Mallik, that states: "the Amaraughaprabodha, which was composed by the honourable Gorakṡanātha, is complete." 8 The same attribution is made in colophons of all the available manuscripts. 9 The colophons were probably inspired by the mention of Gorakṡanātha in three of the text's verses (2, 65 and 74). Two of these verses (2 and 74), at the beginning and end of the text, assert that Gorakṡanātha taught the four yogas, which are the main topic of the work: The awakening, which is proof [of itself], was taught by Gorakṡanātha for those who have undertaken Laya and the other [yogas] and whose minds are quarrelsome.
[…] The honourable Gorakṡanātha, who always abides in samādhi, taught Laya-, Mantra-and Hat˙hayoga solely for [the attainment of] Rājayoga. 10 It is probable that the scribe who added the colophon interpreted these statements as Gorakṡanātha referring to himself in the third person. However, it also possible to interpret these verses as statements made by an author within Gorakṡanātha's lineage, who believed that the teachings of the Amaraughaprabodha were first revealed by Gorakṡanātha. Therefore, these verses do not necessarily confirm that Gorakṡanātha was the author. Nonetheless, the sectarian milieu in which the text was composed is a Ś aiva siddha tradition, as evinced by the first verse, which pays homage to Ā dinātha, Mīnanātha (i.e., Matsyendranātha), Cauraṅgīnātha and Siddhabuddha, as well as by references to Ś iva elsewhere in the text. 11

Region
The manuscript used by Mallik (1954, p. 34) and the six surviving manuscripts of the Amaraughaprabodha are in south-Indian scripts. It is possible that the version published by Mallik was redacted in south India, because it has nine verses in 8 Mallik (1954, p. 55) (iti śrīmadgorakṣanāthaviracitaṃ amaraughaprabodhaṃ sampūrṇam). 9 G 1 (ity amaraugha[ṃ] gorakṣaviracitaṃ saṃpūrṇam); A 2 (ity amaraugho śrīgorakṣaviracitaṃ śatakaṃ samāpyate); B (śrīgorakṣanāthaviracitaṃ amaraughaprabodhaḥ saṃpūrṇaṃ); A 1 (iti śrīmadgorakṣanāthaviracitaṃ amaraughaprabodhaḥ sampūrṇaṃ); G 2 (iti śrīmadgorakṣanāthaviracitam amaraughaprabodhaḥ saṃpūrṇaṃ); T (iti śrīmadgorakṣanāthaviracitam amaraughaprabodhaḥ saṃpūrṇaḥ). 10 Amaraughaprabodha (Edition) 2: (layādipratipannānāṃ kalahotsukacetasām | gorakṣakeṇa kathitaḥ prabodhaḥ pratyayātmakaḥ || prabodhaḥ ] conj. : prabodha-Ed.). Amaraughaprabodha (Edition) 74 (śrīmadgorakṣanāthena sadāmaraughavartinā | layamantrahaṭhāḥ proktā rājayogāya kevalam). On the meaning of amaraugha as samādhi, see below. The compound sadāmaraughavartinā could also be understood as 'who always resides in the lineage (ogha) of the siddhas (amara).' 11 The beta recension of the Amaraughaprabodha has only a hemistich that pays homage to Cauraṅgīnātha and Buddhasiddha. Whether the salutations to Ā dinātha and Matysendranātha were lost in the transmission of this recension is difficult to say. Nonetheless, Ś iva is the object of meditation in both the Amaraughaprabodha's Mantrayoga (25) and Layayoga (27). Also, the Rājayogin's final accomplishment is to become similar to Ś iva (64). Other Ś aiva elements are discussed below. common with the fifth chapter of the Varāhopaniṣat. 12 This Upaniṡad is a compilation that was created in the mid-eighteenth century as part of the south-Indian corpus of one hundred and eight Upaniṡads (Bouy 1994, p. 106). If the long version of the Amaraughaprabodha was a source for this Upaniṡad, 13 then it would have been known in south India in the eighteenth century, which may account for why its surviving manuscripts are in south-Indian scripts. Nonetheless, there is new evidence, which I will discuss below, that suggests a shorter version of the Amaraughaprabodha was composed in south India.

Manuscript Transmission
There are six manuscripts of the Amaraughaprabodha reported in various catalogues by the Kaivalyadhama Yoga Institute's Descriptive Catalogue of Yoga Manuscripts (2005, pp. 22-25), 14 and five of them have been consulted for this article. 15 Also, another manuscript has been found and consulted at the Venkat˙eśvara Oriental Institute in Tirupati. 16 The six manuscripts consulted for this article are on palm-leaf and written in Grantha script. None of them have a scribal date. Four of the six preserve the version of the text in Mallik's edition. 17 Unfortunately, the single manuscript upon which Mallik's edition was based has been lost by the library that used to hold it. 18 This manuscript has not been available to researchers since at least 2004. 19 Although Mallik places several of her conjectures in round brackets, there is evidence to suggest that her transcription has tacit emendations and inaccuracies. 20 Two of the six manuscripts of the Amaraughaprabodha preserve a recension that is significantly shorter than Mallik's edition. 21 This recension has forty-six verses. Both manuscripts of the shorter recension are complete and do not contain any indication of lacunae.
The stemma of the manuscript transmission bifurcates into the four manuscripts of the long recension, which I shall call the theta hyparchetype, and the two of the short recension, the beta hyparchetype, as shown in Fig. 1. The manuscripts of each group are fairly close to one another, 22 although none can be dismissed as an apograph of another.
In this article, I shall argue that the beta hyparchetype predates the fifteenth century, whereas theta was possibly created sometime between the sixteenth and eighteenth centuries. If one includes Mallik's edition, there are seven available witnesses that can be used to create critical editions of both theta and beta. The I first requested the manuscript in 2004 and was told that the bundle to which it belonged could not be found. I have since requested it in 2010 and 2016, but to no avail. 20 The descriptive catalogue of the Government Oriental Manuscripts Library (Raṅgācārya and Bahadur 1910, pp. 3220-3221) transcribed the first four and last five verses of manuscript D-4339. When one compares this transcription to Mallik's edition, there are two instances where the catalogue has suggested a correction in brackets, which has been adopted by Mallik (3b catalogue tṛtīya(kaḥ), Mallik tṛtīyakaḥ and 4a catalogue laya(ḥ), Mallik layaḥ). Therefore, one wonders how many tacit emendations Mallik may have made. Without the manuscript at hand, it is not possible to determine whether the catalogue's transcription is more accurate than Mallik's. Nonetheless, the following discrepancies can be noted: several poor readings in the catalogue's transcription may have been tacitly emended by Mallik (e.g.,71c catalogue rājayogaṃ padaṃ, Mallik rājayogapadaṃ; 72d catalogue kleśāpaho, Mallik kleśāpahā). Also, Mallik may have introduced the following errors: 2d catalogue prabodhaḥ pratyayātmakaḥ, Mallik prabodhapratyayātmakaḥ; 4a catalogue pradiṣṭo, Mallik pradiṣṭaḥ; 4c catalogue mantrayogo, Mallik mantrayogaḥ; 72c catalogue bhogāspadaṃ, Mallik bhogāspadam; 73c catalogue proktā, Mallik proktāḥ. 21 Mss. Nos. R2831 and 70528, represented as G 1 and A 2 respectively, in this article. reconstruction of the relatively large section on Hat˙hayoga can be further improved by using the Amṛtasiddhi, the Śivasaṃhitā, and the Haṭhapradīpikā. Also, the fifth chapter of the Varāhopaniṣat is helpful for editing some of the additional verses of theta.

Differences between the Recensions
Both recensions have a similar structure, except for two significant differences. The structure (with these two differences in bold) has been summarised in Table 1. Notwithstanding significant variant readings, the sections on Mantra-, Laya-, Hat˙ha-and Rājayoga are largely the same. The most obvious differences between the two archetypes is, firstly, theta's ten additional verses on the four types of student and, secondly, its eighteen verses following Rājayoga, which I have called 'miscellaneous topics' for the sake of this discussion. Comparing these and other differences in theta and beta reveals that the structure of beta is more cohesive and closer to the original design of the work.
The first four verses of both theta and beta introduce the four yogas, which are the main topic of the text. In beta, this opening passage is followed by rhetorical verses on the efficacy of Rājayoga (5-9), the importance of the guru, semen (bindu) and resonance (nāda) (10-12) and the union of Ś iva and Ś akti (13). The last verse of this section introduces the teachings on the four yogas by asking how they are taught (14). None of verses 1-14 has yet been traced to an earlier source. 23 In contrast to this, two or three additional verses in the introductory section of theta can be traced or identified as coming from an earlier work. One verse derives from the second The possible exception is verse 2 ( Śivasaṃhitā 5.12). However, this verse is almost generic among texts that teach the fourfold system of yoga and may have found its way into the Śivasaṃhitā via another text. In its current form, it is unlikely that the Śivasaṃhitā predates the beta recension of the Amaraughaprabodha (see footnote 4).
chapter of the Amanaska and another is quoted with attribution to an unknown work called the Śrīsampuṭa. 24 Therefore, the redactors of theta increased the size of the Amaraughaprabodha's introduction by adding at least two verses from other texts. Furthermore, theta's seven verses on the 'four types of student' appear to have been inspired by the Amṛtasiddhi and possibly the Śivasaṃhitā, 25 and its additional section on 'miscellaneous topics' contains a verse which can be traced to the Dattātreyayogaśāstra. 26 The tracing of two verses to the Amanaska and the Dattatreyāyogaśāstra and the references in theta to the Śrīsampuṭa and the Amaraughasaṃsiddhi elicit the hypothesis that the redactors of theta simply added verses to beta. The strongest evidence in support of this hypothesis is that some of the additional verses distort the structure of the text. There are two significant instances of this. Firstly, the introduction of beta ends with the question of how the four yogas are taught. This is immediately followed by the teachings on Mantra-, Laya-, Hat˙ha-and Rājayoga, which has a logical structure, as can be seen in Table 2.
Although the above question appears to have been emended in the edition of the Amaraughaprabodha (see Table 3), the manuscript readings suggest that the same question was posed in theta, which is then followed by a passage on the four types of student. Had the redactors of theta been more careful, they might have rewritten the initial question to ask about the types of student to whom each yoga should be taught, but this does not appear to have been done. As shown in Table 3, it is apparent that the section on the four types of student (in bold) was inserted between the question on the four yogas (17cd) and the explanation of each of them .
Secondly, much of the content of theta's additional passage on 'miscellaneous topics' (56)(57)(58)(59)(60)(61)(62)(63)(64)(65)(66)(67)(68)(69)(70)(71)(72)(73) is extraneous to the main topic of the Amaraughaprabodha, which is the four yogas. This passage can be seen as consisting of three topics. The first (56)(57)(58)(59)(60)(61) concerns the five elements (pañcabhūta). The final statement is that their loss leads to death and their retention (dhāraṇa) to life. 27 The motivation for adding these verses might have been to elaborate on the notion of immortality, which is introduced early in the text and mentioned as a result of Mantra-and Hat˙hayoga. 28 However, this section does not indicate how the practice of any one of the four yogas might retain the five elements and the main section on the four yogas does not mention the five elements.
The second miscellaneous topic is on how the yogin can manipulate the sun and moon in the body. After completing an unspecified practice eight times for three hours, the yogin swallows his breaths and thereby replenishes his moon and controls both his sun and semen (bindu). Bliss and the attainment of Ś ivahood follow. 29 The In the edition of the Amaraughaprabodha, the question has been tacitly emended to reflect the use of the word dīyate in verse 24 on the type of students to whom each yoga is taught (mṛdave dīyate mantro madhyāya laya ucyate | adhimātre haṭhaṃ dadyād amaraughaṃ mahattare) 27 Amaraughaprabodha (theta) 61ab (itthaṃ bhūtakṣayān mṛtyur jīvitaṃ bhūtadhāraṇāt).
28 The introduction of theta has a verse that asks why one who knows the essence of semen (bindu) and the internal resonance (nāda) would fear death, among other things. Amaraughaprabodha (theta) 15: "He who knows the essence of both because of the teachings of a true guru, where is the fear of bodily affliction, old age, disease, sin or death [for him]?" (yo jānāty anayoḥ sāraṃ sadguror upadeśataḥ | kāyakleśajarāvyādhipāpamṛtyubhayaṃ kutaḥ). By repeating oṃ and meditating on Ś iva in the heart, death is destroyed: Amaraughaprabodha ( . When the mind is motionless and steady, the breath goes into the central channel. The sun reaches its zenith and semen becomes controlled. The yogin is filled with bliss and becomes like a Ś iva. Then, all the supreme powers are seen [in him] after ten months" (yāmāṣṭakakṛtābhyāsāt sarvāñ śvāsān grasaty asau | sa ṣoḍaśakalopetaḥ śaśī tiṣṭhati pūrakāt ||62|| nistaraṅge sthire citte vāyur bhavati madhyagaḥ | ravir ūrdhvapadaṃ yāti bindur āyāti vaśyatām ||63|| ānandapūrito yogī jāyate śivasannibhaḥ | tadaiśvaryaguṇāḥ sarve dṛśyante daśamāsataḥ ||64|| motivation for adding these verses might have been to support the following comment in the section on Hat˙hayoga, "One should know the connection between the moon, sun and fire in order to [attain] immortality." 30 However, the additional passage introduces new terminology for the sun and moon (i.e., śaśī, ravi and ṣoḍaśakalā). It may be an attempt to summarise a yoga similar to that of the Amṛtasiddhi, in which case the practice done eight times for three hours would be the three mudrās of that text and the Amaraughaprabodha 31 . Furthermore, it is followed by a verse that suggests that the preceding verses were borrowed from a text called the Amaraughasaṃsiddhi, which is also attributed to Gorakṣanātha.
Thus, in the Amaraughasaṃsiddhi, the awakening that is brought about by adeptness in the methods of Layayoga and so forth and is proof [of its own efficacy], has been revealed by Gorakṡanātha. 32 Therefore, it is quite likely that this passage was borrowed from another text, which was probably on the same four yogas (i.e., Mantra, Laya, Hat˙ha and Rājayoga/ amaraugha) taught by Gorakṡanātha, but contained different terminology to the short version of the Amaraughaprabodha.
The third miscellaneous topic is mainly on Rājayoga and liberation-in-life (jīvanmukti). The content of this passage is relevant to the central concern of the text, which is the four yogas that result in liberation. Nonetheless, it begins with a verse which was most probably taken from a Buddhist text. This verse mentions the signs of success that arise from practice, and it closely parallels a verse from the Sekoddeśa. 33 This indicates that the redactors of theta had access to Buddhist material or were using texts on the four yogas that had content borrowed from Buddhist works.
A subsequent verse in this section introduces a view of liberation-in-life which goes beyond that implied by beta. Whereas beta ends with the yogin's attainment of samādhi and his becoming a second Ś iva, additional verses in theta, including one from the Dattātreyayogaśāstra, state that the yogin emerges from samādhi to live in the world however he so wishes: 34 After [liberation-in-life has been achieved], no one at all lives and no one will die. Having obtained the state of Rājayoga, which subjugates all beings, [the yogin] can do anything or nothing, behaving as he pleases. The king of yogis There is no apparent reason as to why the passage on five elements is followed by another on the sun and moon and then further verses on Rājayoga and liberation. The infelicitous position of these additional verses suggests that this section of theta was hastily redacted and inserted into the Amaraughaprabodha. The redactor appears to have used at least one unknown work on the topic of amaraugha (i.e., the Amaraughasaṃsiddhi), as well as Buddhist material and an early yoga text on the same four yogas (i.e., the Dattātreyayogaśāstra).
There is one other addition in theta that provides some insight into how it was redacted. In the section on Hat˙hayoga, a verse has been added to the beginning of the description of mahāvedha. 36 The first hemistich of this verse occurs in the Amṛtasiddhi (11.3cd), but the context is different. In the Amṛtasiddhi, this hemistich describes the yogin's posture in mahāmudrā, whereas in theta it is appended to a passage on mahāvedha. 37 The second hemistich of the additional verse was taken from a different chapter of the Amṛtasiddhi (14.5cd). The inclusion of both hemistiches reveals that the redactor of theta used the Amṛtasiddhi to augment and emend the original Amaraughaprabodha. Nonetheless, despite having access to the source text, it appears that this redactor did not understand the practice of mahāvedha in the Amṛtasiddhi, nor any Hat˙hayoga text for that matter, because this mudrā is not supposed to be done with both hands holding one extended leg. 38 One might ask whether the opposite hypothesis to the one I have just discussed is at all likely. In other words, could beta be a more recent, truncated version of theta or, more to the point, can beta be seen as a deliberate attempt to rectify the problems of theta? Perhaps, a redactor omitted the passages in theta that are unrelated to the four yogas and removed theta's worst imperfections. It is true that the absence of 35 Amaraughaprabodha (theta) 71-73 (na jīvati tataḥ ko 'pi na ca ko 'pi mariṣyati | rājayogapadaṃ prāpya sarvasattvavaśaṅkaram ||71|| sarvaṃ kuryān na kuryād vā yathāruci viceṣṭitam ||72|| nagnaḥ ko 'pi guhāsu divyavasanaḥ kaupīnavāsāḥ kva cid divyastrīsuratānvito 'pi kuha cit sa brahmacārī kva cit | bhikṣāhārarataḥ kva cit kva cid api prāpnoti bhogāspadaṃ sarvatrāpratibaddhavṛttir akhilakleśāpahā yogirāṭ || 71b mariṣyati ] B, T, Ed : mariṣyatmi A1 : Dattātreyayogaśāstra 161 (rājayogavaraṃ prāpya sarvasattvavaśaṃkaram | sarvaṃ kuryān na vā kuryād yathāruci viceṣṭitam). 36 Amaraughaprabodha (theta) 37: (savyaṃ prasāritaṃ pādaṃ karābhyāṃ dhārayed dṛḍham | āndolanaṃ tataḥ kuryāc charīrasya trimārgataḥ || 37cd kuryāc charīrasya ] B : kūryāch śarīrasya A 1 , G 2 : kūryāt śarīrasya T). Amaraughaprabodha (theta) 37ab = Amṛtasiddhi 11.3cd = Haṭhapradīpikā 3.10cd Śivasaṃhitā 4.27ab (descriptions of mahāmudrā). Amaraughaprabodha (theta) 37cd = Amṛtasiddhi 14.5cd. 37 Amṛtasiddhi 11.3cd: "The [yogin] should hold firmly with both hands the left leg which has been extended" (savyaṃ prasāritaṃ pādaṃ karābhyāṃ dhārayed dṛḍham). Cf. Amaraughaprabodha (theta) 37ab (see footnote 35). 38 As far as I'm aware, the theta recension of the Amaraughaprabodha (37ab) is unique in instructing that one leg should be extended for the practice of mahāvedha. some of theta's passages, such as the four types of student and the five elements, makes the structure of beta more coherent. However, if creating a better structure were the purpose behind beta, one would have to explain why theta's final verses on Rājayoga and liberation-in-life were deliberately removed, because they are relevant to the central topic of the four yogas. Furthermore, had the redactor of beta simply removed problematic verses of theta, it is unclear why this redactor also removed verses that are coherent but were borrowed from other texts, such as the Amanaska and the Dattātreyayogaśāstra. As we shall see below, the hypothesis that beta followed theta becomes even less likely when one attempts to date and understand the content of both recensions within the broader history of works that teach the same four yogas.

Dating the Recensions
The terminus a quo of the beta recension of the Amaraughaprabodha is the Amṛtasiddhi and not the Dattātreyayogaśāstra. The Amṛtasiddhi was composed before 1160 CE, which is the date of a colophon in the oldest available manuscript. 39 This Vajrayāna work does not teach a system of yoga called Hat˙ha-or Rājayoga, but it was nonetheless a source text for three important Haṫhayogic mudrās, namely mahāmudrā, mahābandha and mahāvedha, and four distinct stages of yoga (ārambha, ghaṭa, paricaya and niṣpatti). The close relationship between the Amṛtasiddhi and the beta recension of the Amaraughaprabodha and the fact that this recension does not borrow from another work on Haṫha-and Rājayoga make it probable that the Amaraughaprabodha was one of the earliest works to formulate a fourfold system which included Haṫha-and Rājayoga.
The terminus ad quem of the beta recension remains the Haṭhapradīpikā, as Bouy determined. However, it is probable that Svātmārāma knew only the beta recension, because none of the additional twenty-nine verses of theta can be found in the Haṭhapradīpikā. These include theta's additional verses on Rājayoga, which would have been relevant to the fourth chapter of the Haṭhapradīpikā. Although the following evidence is not conclusive, there are two differences between theta and beta that are significant enough to suggest that Svātmārāma borrowed from beta. Firstly, in one place, the order of the hemistiches diverges in beta and theta, and the order in beta is the same as that in the Haṭhapradīpikā. 40 The second difference concerns the additional verse on mahāvedha in theta. The Haṭhapradīpikā's description of the yogin's posture for mahāvedha is more, but not entirely, consistent with beta than theta. 41 A comparison of the variant readings of theta and beta with the critical edition of the Haṭhapradīpikā indicates that beta has twice the number of significant variant readings in common the Haṭhapradīpikā than theta (see "Appendix"). Although this result is worth noting, it is not in itself conclusive evidence for assessing whether theta or beta was the source of the Haṭhapradīpikā, because the comparison is based on a small sample of verses and on a critical edition of the Haṭhapradīpikā that is not dependable nor comprehensive in the manuscripts it reports. 42 Nonetheless, this comparison raises an interesting question: if theta postdates the Haṭhapradīpikā, why would it have readings in common with the Haṭhapradīpikā that are not found in beta? If it is indeed true that theta was redacted sometime after the Haṭhapradīpikā, then its transmission must have been contaminated by manuscripts of the Haṭhapradīpikā. This contamination seems plausible because its redactor was, after all, inserting material from other yoga texts.
The terminus ad quem of the Amaraughaprabodha's theta recension may be the eighteenth-century Varāhopaniṣat. It remains somewhat uncertain as to whether theta was a source for this Upaniṡad, because the available manuscripts of the Upāsanāsārasaṅgraha are incomplete and, as Bouy (1994, p. 19) noted, the fifth chapter of the Varāhopaniṣat contains a large passage that was borrowed from the Upāsanāsārasaṅgraha. 43 The only other evidence, as far as I am aware, for the theta's terminus ad quem is the four palm leaf manuscripts that preserve it, none of which is dated nor likely to be older than the eighteenth century. 41 Haṭhapradīpikā 3.26-27ab (mahābandhasthito yogī kṛtvā pūrakam ekadhīḥ | vāyūnāṃ gatim āvṛtya nibhṛtaṃ kaṇṭhamudrayā || samahastayugo bhūmau sphicau sanāḍayec chanaiḥ). Cf. Amaraughaprabodha (beta) 27-28ab (punar āsphālayed kaṭyāṃ susthiraṃ kaṇṭhamudrayā | vāyūnāṃ gatim ārudhya kṛtvā pūrakakumbhakau || samahastayugo bhūmau samapādayugas tathā) and Amaraughaprabodha (theta) 37 (see footnote 35). The posture for mahāvedha in the Amaraughaprabodha is similar to that described in the Amṛtasiddhi (i.e., a squatting type position in which the legs and arms are symmetrical), whereas in the Haṭhapradīpikā the posture is a cross-legged position, as stipulated for mahābandha (see Haṭhapradīpikā 3.19). 42 The only critical edition of the Haṭhapradīpikā is that by Swami Digambaraji and Pt. Raghunatha Shastri Kokaje in 1970 at the Kaivalyadhama, S.M.Y.M. Samiti. Most of the manuscripts which they used for this edition are from libraries in the state of Mahārāṡṫra, so it is unlikely that regional differences in the transmission of this pan-Indic work are represented by their edition. More importantly, the editors appear to have been unaware of the significant number of parallel verses in early yoga texts such as the Dattātreyayogaśāstra, the Amṛtasiddhi, the Amanaska, etc. Also, the critical apparatus is negative and it appears to include only the most important variants; the oldest manuscript has not been consulted (Bouy 1994, p. 84 n. 357) and the section on vajrolī has been distorted by the editors' belief that Haṫhayoga did not teach transgressive practices; e.g., the verse on amarolī which describes it as drinking urine and emanating from a Kāpālika tradition has been relegated to a footnote (Haṭhapradīpikā 1998: 112 n. 170), in spite of the fact that most of the manuscripts have it. 43 In footnote 3, I mentioned that the Upāsanāsārasaṅgraha cites with attribution Amaraughaprabodha 38-41. These verse numbers are based on Mallik's edition of the Amaraughaprabodha (i.e., the theta recension), which Bouy used. However, it is very likely that the Upāsanāsārasaṅgraha borrowed from the beta recension of the Amaraughaprabodha (27-30) and not theta, because an additional verse on mahāmudrā in theta (for details, see footnote 35) is omitted from the Upāsanāsārasaṅgraha's passage on this mudrā, which it attributes to the Amaraughaprabodha.

The Significance of the Amaraughaprabodha's Beta Recension in the History of Yoga
The close relationship of the beta recension of the Amaraughaprabodha with the Amṛtasiddhi, or perhaps a rudimentary version of the Amṛtasiddhi, provides unique insights into how early teachings on Hat˙hayoga were formulated. Unlike early Hat˙ha-and Rājayoga texts, the Amṛtasiddhi contains extensive, detailed passages on the theory behind the practice, which reveal that its teachings were intended for esoteric Buddhists who had rejected deity yoga. 44 As I shall discuss below, the redactor of the Amaraughaprabodha borrowed and modified only select portions of the Amṛtasiddhi, omitting much of the theory in the process, and introduced new ideas to formulate a system of Hat˙hayoga that would have appealed to Ś aivas. Also, the sparing application of doctrine and metaphysics in the Amaraughaprabodha probably made its yoga more accessible to people of other religions.
The section on Haṫhayoga in the Amaraughaprabodha is a terse account of the practice of the three mudrās and the four stages of yoga that are similar to those in the Amṛtasiddhi. As seen in Table 4, the content of the Amaraughaprabodha largely derives from nine of the Amṛtasiddhi's thirty-six chapters.
In spite of this shared content, a significant difference between the Amaraughaprabodha and the Amṛtasiddhi is that the latter has extensive chapters on the theory and metaphysics underlying the practice of the mudrās and the stages of yoga that follow it. These additional chapters are in bold in Table 5. Most of the content of these chapters is not mentioned at all in the Amaraughaprabodha. Metaphysics, such as the sun, moon and fire, are mentioned only in passing in the Amaraughaprabodha's section on Haṫhayoga. Theory regarding matter (prakṛti), mind and the five bodily winds is absent. In the chapters of the Amṛtasiddhi that follow the practice of the three mudrās (i.e., chapters 11-14), many supernatural effects are enunciated and nearly all of these have been omitted by the redactor of the Amaraughaprabodha. Therefore, it can be generally said that, if the redactor of the Amaraughaprabodha did indeed use the Amṛtasiddhi, the process of redacting the section on Hat˙hayoga was reductionist in the extreme and orientated almost entirely towards the practice of physical techniques and the results. This is also characteristic of other early works on Hat˙hayoga. As the physical techniques, which came to define Hat˙hayoga, moved from one tradition to another, the theory underlying them was minimal and subject to syncretization.
The redactor of the Amaraughaprabodha did not altogether omit the terminology and theory that was peculiar to Buddhism. Terms were retained that Ś aivas would have understood in contexts of yoga, such as mahāmudrā, ānanda and śūnya, on account of their use in earlier Śaiva works. Other terms that could be understood in a general sense, such as vimarda ('pounding') and vicitra ('various'), were also retained. Nonetheless, when this terminology occurs in the Amṛtasiddhi, its meaning is specific to a system of four blisses, moments and voids that was unique to esoteric Buddhism. Furthermore, some of Amṛtasiddhi's terminology that probably derived from Rasaśāstra, such as vedha, maraṇa and puṭa, was incorporated by the redactor of the Amaraughaprabodha, perhaps because the meaning of these terms was known outside of esoteric Buddhism.
The Śaiva orientation of the Amaraughaprabodha is established by the attribution of its teachings to Gorakṡanātha and verses that mention Śiva, Ś akti or the liṅga in the introduction and the sections on Mantra-, Laya-, Hat˙ha-and Rājayoga. Although the redactor integrates terminology and metaphysics of the Amṛtasiddhi, the explicit references to Ś iva, Ś akti and the liṅga affirm the Ś aiva a The four types are mṛdu, madhya, adhimātra and adhimātratara. These chapters probably inspired the section on the four types of students in the theta recension of the Amaraughaprabodha orientation of the work. A good example of this is seen in the penultimate verse of the introduction, in which notions of the sun and moon that are central to the teachings of the Amṛtasiddhi are subsumed under the goal of uniting Ś iva and Ś akti above the aperture of the skull (brahmarandhra): When the lord of thoughts is still and the sun enters completely the path of Meru (i.e., suṣumnā); when its fiery state has increased, the moon melts and the body is quickly made full; when an abundance of bliss rises up and the darkness of delusion, such as [the thoughts] 'yours' and 'mine', departs, and when the aperture of the skull bursts open, the extraordinary and unprecedented union of Ś iva and Ś akti prevails. 45 The above verse combines the metaphysics of the Amṛtasiddhi with those of earlier Ś aiva traditions. Moving the sun into the central channel (meru) is mentioned in a passage of the Amṛtasiddhi (4.5-10) that describes a process of uniting the sun and moon. The aim of the Amṛtasiddhi's yoga is to prevent the sun devouring the moon's nectar. However, rather than conserving the nectar in the moon, the above verse incorporates the different idea of the moon melting and filling the body, presumably, with nectar. The notion of flooding the body with nectar in order to purify and rejuvenate it can be found in early Saiddhāntika works, including the Kiraṇa and Mṛgendratantra (Tāntrikābhidhānakośa 2000 vol 1, p. 138). The verse concludes with uniting Ś iva and Ś akti, rather than the sun and moon. The association of śakti, or more specifically kuṇḍalinī, with nectar may not have been new to Ś aivas, because a nectarean kuṇḍalinī (kuṇḍalī) is mentioned in the pre-tenth century Ś ākta scripture, the Tantrasadbhāva (Tāntrikābhidhānakośa 2000 vol 1, p. 136.), 46 and kuṇḍalinī is associated with nectar (amṛta) in a pre-tenth century Saiddhāntika work, the Sārdhatriśatikālottara (12.1-2). 47 Also, there is a precedent in the Ś aiva work called the Jayadrathayāmala for the practice of a mudrā (karaṇa) that releases nectar and unites Ś iva and Ś akti. 48 45 Amaraughaprabodha (beta) 13 (nirvāte cittarāje viśati khararucau merumārgaṃ samantād udrikte udrikte vahnibhāve dravati śaśadhare pūrayaty āśu kāye | udyaty ānandavṛnde tyajati tavamametyādimohāndhakāre prodbhinne brahmarandhre jayati śivaśivāsaṅgamaḥ ko 'py apūrvaḥ || 13a nirvāte ] : niryāte Ed. 13a cittarāje ] G 1 , A 2 : cittarāte B, A 2 , G 2 , T, Ed. 13a viśati ] G 1 , A 2 : vrajati B, A 2 , G 2 , T, Ed. 13a khararucau ] B, A 2 , G 2 , T, Ed : khararuce G 1 , A 2 . 13a merumārgaṃ ] B, A 2 , G 2 , T, Ed : merudurge G 1 , A 2 . 13b udrikte ] conj. : udrakte T : udrajñe B, A 2 , G 2 : durvṛtte G 1 , A 2 : dudrajñe Ed. 13b dravati ] G 1 , A 2 : sravati B, A 2 , G 2 , Ed : +++ T. 13b śaśadhare ] ++++ T. 13b pūrayaty āśu kāye ] B, A 2 , G 2 , Ed : pūrayitvā tu kāyaṃ G 1 , A 2 : ++yaty āśu kāye T. udyaty ānandavṛnde ] A 2 , G 2 , T, Ed : utp +utyānantakande G 1 : unnatyānandakande A 2 : udayatyānandavṛnde B (unmetr.). 13c tyajati ] jatyat A 2 . 13c tava-] śiva-A 2 . 13c -mametyādi ] A 2 , G 2 , T, Ed : -mamevandi B : -mukhetyādi-G 1 , A 2 . 13c unnatyānandakande ] ++hāndhakāre T. 13d śivaśivāsaṅgamaḥ ] A 2 , Ed : śivaśivāsaṅgamaṃ B, T, G 1 : śivaśi+saṃgamaḥ A 2 : śivāsaṅgamaṃ G 2 (unmetr.). 13d apūrvaḥ ] apūrvam B. 46 On the date of the Tantrasadbhāva, see Sanderson (2001, p. 4 (n.1), 20-35). 47 See Hatley (forthcoming,. On the date of this work, see Goodall (2004, p. lxxxvii). 48 For the reference and translation of this passage, see Mallinson (2007, p. 21, 177 n. 79). Also, two fourteenth-century works, the Khecarīvidyā (3.1-15) and a section on yoga in the Śārṅgadharapaddhati, contains passages on flooding of the body with nectar by uniting Ś iva and Ś akti (Mallinson 2007, pp. 28, 131).
Although the Amaraughaprabodha's section on Hat˙hayoga corresponds most closely to the yoga of the Amṛtasiddhi, Ś aiva terminology and metaphysics have been introduced at key places to give the impression that this is a Ś aiva form of yoga. In order to demonstrate this, as well as examine other issues around the relationship between the two texts and the Ś aiva conception of Hat˙hayoga, I shall compare the descriptions of mahāmudrā and the results of practising it in both texts. The Amaraughaprabodha describes it as follows: Having pressed the perineum with the left foot and having held the extended right foot with both hands, [ As depicted in Fig. 2, the yogin in mahāmudrā is in an asymmetrical position, with one leg straight and the other bent. The heel of the bent leg presses against the perineum, his hands hold the foot of the extended leg, over which his torso is folded forward.
Furthermore, the Amaraughaprabodha has the additional instruction that the yogin should breathe in through the mouth (highlighted in bold in Table 6), and several other directives are missing, such as focusing the mind on the crossroads (catuḥpatha). These significant differences raise the question of whether the redactor of the Amaraughaprabodha used the Amṛtasiddhi or some other (currently unknown) source to redact the section on Haṫhayoga. If it was the Amṛtasiddhi, then the redactor radically truncated and modified the Amṛtasiddhi's account of mahāmudrā. Some of the added details and omissions might be explained by variations or alternative views of the practice. However, it is more difficult to 50 Amṛtasiddhi 11.3-10 (yoniṃ saṃpīḍya vāmena pādamūlena yatnataḥ | savyaṃ prasāritaṃ pādaṃ karābhyāṃ dhārayed dṛḍham ||3|| āsane kaṭim āropya cibukaṃ hṛdayopari | nava dvārāṇi saṃyamya kukṣim āpūrya vāyunā ||4|| cittaṃ catuḥpathe kṛtvā ārabhet prāṇayantraṇam | candrārkayor gatiṃ bhaṅktvā kuryād vāyunivāraṇam ||5|| jāraṇeyaṃ kaṣāyasya cāraṇaṃ bindunādayoḥ | cālanaṃ sarvanāḍīṇāṃ analasya ca dīpanaṃ ||6|| kāyavākcittayogena kāyavākcittasādhanaṃ | bhaved abhyāsato 'vaśyaṃ mārgārūḍhasya yoginaḥ ||7|| anayā mudrayā sarvaṃ jāyate yogino dhruvaṃ | tasmād enāṃ prayatnena mahāmudrāṃ samabhyaset ||8|| sarvāsām eva mudrāṇāṃ mahatīyaṃ svayaṃbhuvā | mahāmudrāṃ ca tenaināṃ vadanti vibudhottamāḥ ||9|| anayā bādhyate mṛtyur hiteyaṃ tena sarvadā | cetasā vidhṛtā yena yantraṃ tenaiva nirjitam). propose reasons for why a redactor would truncate the Amṛtasiddhi's account to the extent seen in the Amaraughaprabodha. Indeed, it seems possible, if not probable, that the Amaraughaprabodha has preserved an older and more rudimentary account of this yoga than that of the Amṛtasiddhi. One might further propose that this older work was a source text for the Amṛtasiddhi, because the latter has verses in common with the Amaraughaprabodha. Even if the Amaraughaprabodha's teachings on Hat˙hayoga derive from a text that was older than the Amṛtasiddhi, the original source must have been composed in a Vajrayāna milieu because the remnants of this tradition's distinct system of four blisses, voids and moments are apparent in the Amaraughaprabodha. Therefore, the most significant change made by the redactor of the Amaraughaprabodha to the description of mahāmudrā was the introduction of the Ś aiva metaphysics of kuṇḍalinī (highlighted in bold in Table 6).
In the Amaraughaprabodha, mahāmudrā is likened, in effect, to a stick that is used to kill a snake. It strikes the coiled kuṇḍalinī and forces her to become straight. Such notions of force, as well as the metaphysics of kuṇḍalinī, are absent in the Amṛtasiddhi. So, it would seem that Ś aivas understood the notion of force (haṭha) in Hat˙hayoga as referring to the forceful effect of its practice on kuṇḍalinī. This understanding probably extended to the general aim of the three mudrās, which was to force kuṇḍalinī and prāṇa up through the central channel to pierce the three knots (granthi). Similar forceful effects, in particular on apānavāyu, semen and kuṇḍalinī, can be found in other early works on Hat˙hayoga and, in my opinion, provide the strongest indication for the intended meaning of haṭhayoga in the formative phase of this type of yoga (Birch 2011, pp. 544-545).
Both the Amaraughaprabodha and the Amṛtasiddhi use terms from Rasaśāstra, such as jāraṇa ('digesting') and cāraṇa ('assimilating'), to describe the effects of mahāmudrā. In the Amaraughaprabodha's description, the pressing of the perineum and the application of the throat lock (bandha) seem to create something akin to an alchemical vessel with two halves or lids (dvipuṭa). 51 Presumably, this vessel is the central channel after it has been blocked above and below by the locks, which envelop and hold kuṇḍalinī in a death-like state (maraṇāvasthā). Although the meaning of maraṇāvasthā is not entirely clear to me in the context of raising kuṇḍalinī to meet Śiva in or above the brahmarandhra, it seems that the redactor may have been alluding to the process of 'killing' (māraṇa), which in Rasaśāstra refers to heating a substance in a burning pit (puṭa) until it turns to ash. 52 These alchemical metaphors, which are also found in the Amṛtasiddhi, imply that kuṇḍalinī is transformed irreversibly from her coiled to erect state in the central channel. 53 The Amaraughaprabodha's description of the second mudrā called mahābandha includes two metaphysical terms, namely triveṇī and kedāra, that do not occur in the Amṛtasiddhi. 54 The technical term triveṇī, which usually means the confluence of the three main channels in other Haṫha-and Rājayoga texts, 55 would probably have been known to a Ś aiva audience judging by its appearance in earlier Ś aiva works, such as the Kubjikāmatatantra, in which it is mentioned along with the channels (5.170-72) and located in the navel (25.93). The Kuñcitāṅghristava, which was composed in Cidambaram in the fourteenth century, locates it in the heart. This confluence, also known as the prayāga, was considered one of eight sacred places in early Ś aivism, as attested in works such as the Niśvāsaguhya (1.29-34) and the Tantrasadbhāva (15.62). 56 Kedāra, which is generally located in the head, 57 is included in lists of forty sacred sites (tīrtha) in "the Ś ivadharma and throughout the scriptural authorities of the Mantramārga" (Sanderson 2003, p. 405). 58 Both these terms appear to have been added by the redactor of the Amaraughaprabodha to make the practice of mahābandha appeal to a Ś aiva audience.
The account of mahāvedha in the Amaraughaprabodha does not introduce terminology that is peculiar to Ś aivism and absent in the Amṛtasiddhi. In fact, the challenge the redactor appeared to face in composing the remaining portion of the section on Hat˙hayoga was modifying or removing concepts unique to Buddhism, in particular the technical terms from the sexual yoga of some Vajrayāna works, which incorporate systems of four blisses, moments and voids (Sferra 2000, pp. 31-33), as seen in Table 7.
The Amṛtasiddhi incorporates this terminology, with the exception of viramānanda, vilakṣaṇakṣaṇa and sarvaśūnya. The Amaraughaprabodha does not mention the four moments at all, but includes three blisses and voids. Although this specific system of blisses and voids may be unique to Vajrayāna, terms such as ānanda, paramānanda, śūnya and mahāśūnya are scattered throughout Ś aiva works. Furthermore, some Ś aivas were probably familiar with progressive stages of sounds and void-like meditative states in their own tradition. 59 Therefore, it would seem that the redactor of the Amaraughaprabodha deliberately excluded terminology that was specific to Vajrayāna, but was willing to include terms that occur in Ś aiva works. This strategy is evident in a comparison of the following parallel passages of the Amṛtasiddhi and the Amaraughaprabodha. In the first example, the vicitra moment has been emended to a 'vicitra' sound: The Amṛtasiddhi: When the first stage is completed and Brahmā's knot pierced, then, a biting [sensation] and the perception of a little bliss in the central channel arise. One also enters into voidness and the vicitra moment arises. 60 The Amaraughaprabodha: Because of piercing Brahmā's knot, bliss arises in the void. The unstruck resonance, which has various [musical] sounds (vicitrakvaṇaka), is heard in the body. 61 Whether the redactor of the Amaraughaprabodha was responsible for this unobtrusive emendation is not certain because the south-Indian manuscripts of the Amṛtasiddhi also preserve vicitrakvaṇa, which is one of many instances where the Indian transmission has changed technical terms of Vajrayāna into something more intelligible to non-Buddhist yogins. 62 Therefore, it is possible that vicitrakṣaṇa had already become vicitrakvaṇa in a south-Indian version of the Amṛtasiddhi by the time the Amaraughaprabodha was written. 63 In another example, the redactor of the Amaraughaprabodha removed the reference to the vimarda moment by omitting the word for moment (kṣaṇa). In this case, vimarda would have been understand by a non-Buddhist in its more general sense of 'pounding': The Amṛtasiddhi: […] When the second stage is complete, the [state] beyond the void arises. Then, the sound of the bherī drum occurs in the middle channel and the vimarda moment. 64 The Amaraughaprabodha: […] Then, because of piercing Viṡṅu's knot, a pounding [sound] (vimarda), which is indicative of supreme bliss, arises beyond the void. Then, the sound of a bherī drum occurs. 65 It is worth asking why the redactor of the Amaraughaprabodha retained any of the Amṛtasiddhi's technical terminology peculiar to Vajrayāna, given that less ingenuity would have been required to remove it altogether. For example, one might question the need of a verse at the end of the description of mahāvedha in the Amaraughaprabodha, which states that the yogin should know the conjunction of the moon, sun and fire in order to achieve immortality. 66 These metaphysical notions are important in the Amṛtasiddhi, but none are mentioned elsewhere in the Amaraughaprabodha's section on Hat˙hayoga. The inclusion of this verse, as well as terms like vicitra and vimarda, suggests that the intended audience of the Amaraughaprabodha was familiar with the Buddhist origins of this yoga. Therefore, although the redactor of the Amaraughaprabodha may have composed the text to lay claim to Hat˙hayoga as a Ś aiva practice, it seems that the system of blisses, voids and sounds could not be dispensed with entirely, perhaps because this type of Vajrayāna yoga was still known at the time of writing. The likelihood of this and the possibility that the redactor was using a more rudimentary version of the Amṛtasiddhi points to an early date (i.e., the twelfth century) for the composition of the short recension of the Amaraughaprabodha. Be this as it may, it seems that the short recension of the Amaraughaprabodha was composed close to the time of the Amṛtasiddhi, when the yoga of the latter was still known, especially in communities transitioning from 62 For a discussion and more examples of such changes, see Mallinson (2016b). 63 I would like to thank Dominic Goodall for pointing this out to me. 64 Amṛtasiddhi 20.7 (saṃpūrṇāyāṃ dvitīyāyām atiśūnyaṃ prajāyate | bherīśabdas tadā madhye vimardakṣaṇasaṃbhavaḥ). 65 Amaraughaprabodha (beta) 38 (viṣṇugranthes tato bhedāt paramānandasūcakaḥ | atiśūnye vimardaś ca bherīśabdas tato bhavet || 38a tato ] G 1 , A 2 : tathā B, A 1 , G 2 , T, Ed. 38a bhedāt A 2 : vedhāt G 2 , A 1 , T, Ed : om. B (unmetr.). atiśūnye ] A 1 , G 2 , T, Ed : atiśūnyo G 1 , A 2 , B. 38d bherīśabdas ] bherīnādas A 2 . 38d tato ] G 1 , A 2 : tathā B, Ed : tadā A 1 , G 2 , T). 66 Amaraughaprabodha (beta) 30ab (somasūryāgnisaṃbandhaṃ jānīyād amṛtāya vai | 30a -saṃbandhaṃ ] A 1 , G 2 , T, Ed : -saṃbandhā B : -saṃbandhāj G 1 , A 2 . 30b jānīyād ] dāniryyād B).
Buddhism to Ś aivism. 67 This supports James Mallinson's proposal (2016b, p. 11 n. 25 and forthcoming) that the Amaraughaprabodha was composed in such a community at Kadri in Karnataka, owing to the text's opening invocation to the wise Siddhabuddha, a disciple of Matsyendranātha from that locality.
Apart from introducing Ś aiva terms and metaphysics and obscuring or omitting Vajrayāna ones, the Śaiva appropriation of the Amṛtasiddhi's yoga is achieved by making it subordinate to Rājayoga. This so-called 'king of all yogas' probably had an older association with Ś aivism, as evinced by a Ś aiva work known as the Amanaska, which teaches the attainment of the no-mind state by the practice of śāmbhavī mudrā. The Amanaska was known in Gujarat in the mid-twelfth century (Birch 2014, p. 406 n. 21), so it was probably composed in the eleventh or early twelfth century. The redactor of the Amaraughaprabodha alludes to the Ś aiva origins of Rājayoga by referring to the liṅga in the final verse on Rājayoga, 68 and by defining the term amaraugha (literally, 'a flood of nectar' or 'a multitude of immortals') as Rājayoga. 69 It is likely that the term amaraugha would have reminded Ś aivas of the divine stream of teachings known as the divyaugha in earlier Kaula scriptures, 70 while also conveying meanings close to that of amṛtasiddhi ('the attainment of nectar/immortality'). Moreover, any association Hat˙hayoga might have had with Buddhism when the Amaraughaprabodha was composed would surely not have deterred Ś aivas from adopting a system of yoga in which Hat˙hayoga was only one option of several auxiliary methods for the attainment of Rājayoga.
Furthermore, the redactor of the Amaraughaprabodha appears to have distanced Hat˙hayoga from the yoga of the Amṛtasiddhi by omitting any mention of semen (bindu) in the section on Hat˙hayoga. In keeping with the sexual yoga of the Kālacakra tradition (Sferra 2000, p. 32), the Amṛtasiddhi is very much concerned with retaining semen. An introductory verse of the Amaraughaprabodha states that there are two types of Hat˙hayoga: one practised with the internal resonance (nāda) and the other with semen (bindu). 71 Although the yogin is advised to avoid frequenting fire, women and roads in the initial stage of practice, 72 the emphasis on internal sounds and the absence of references to semen indicate that the first type of Hat˙hayoga, rather than the second, was taught. A further distinction can be seen in the way both works interpret nectar. In the Amaraughaprabodha, nectar (amara) is associated with Rājayoga, which is defined as meditative absorption, whereas in the Amṛtasiddhi nectar (amṛta) corresponds with semen in the male body. 73 Some of the metaphysics and terminology specific to the Amṛtasiddhi continued to be used by authors of works on Hat˙hayoga up until the nineteenth century, partly because of the significant amount of verses that Svātmārāma borrowed from the Amaraughaprabodha's section on Hat˙hayoga. The passage on the four stages of yoga, in which the knots are pierced and various blisses, sounds and voids arise, was incorporated into the fourth chapter of the Haṭhapradīpikā as the practice of nādānusandhāna ('fusing the mind with the internal resonance'). 74 In other cases, the original meaning of the technical terminology was lost and reinterpreted. For example, the alchemical terms dvipuṭa and puṭadvaya have been interpreted by commentators as dvināsāpụta ('the two nostrils') and the channels iḍā and piṅgalā. 75 In the Yogacintāmaṇi, Ś ivānanda rewrote the verse on kuṇḍalinī to say that she destroys the state of death, and Bhavadeva interpreted it to mean that, after the practice of mahāmudrā, kuṇḍalinī resides in the two nostrils. 76 Furthermore, it is likely that Svātmārāma and others understood a verse on the conjunction of the moon, sun and fire as referring to the intersection of the channels called iḍā, piṅgalā and suṣumnā, 77 because this meaning of moon, sun and fire can be found in early Ś aiva works and even two yoga texts, the Vivekamārtaṇḍa and the Śivasaṃhitā, known to him. 78

Conclusion
The relationship between the Amṛtasiddhi and the beta recension of the Amaraughaprabodha provides a window onto the early history of Hat˙hayoga, which was shaped significantly by the exchange of ideas and techniques between Buddhism and Śaivism in south India around the 11th and 12th centuries. It seems that some sects associated with these religions devised physical methods of yoga on the understanding that prāṇa could be deliberately manipulated within the body to attain samādhi and liberation with greater efficacy and certainty than other methods. Even though proponents claimed that Hat˙hayoga would cure diseases, old age, suffering, death and so on, 79 the notion that forcefully manipulating prāṇa was dangerous (Birch 2011, pp. 538-539), as well as broader trends toward gnosis and ritual in these religions, may have prevented this type of yoga from becoming a central teaching of Śaivism or Vajrayāna. Nonetheless, Haṫhayoga evolved as an auxiliary practice in systems of yoga associated with Gorakṡanātha.
Comparing the Amaraughaprabodha with the Amṛtasiddhi reveals that the doctrine and metaphysics used to teach a physical yoga practice could change significantly from one tradition to another, while the physical techniques remained essentially the same. Furthermore, although the physical practice of the Amṛtasiddhi's yoga is the same as that of the Amaraughaprabodha, only a fraction of the former's underlying doctrine was adopted by the latter. Physical yoga could be taught with minimal doctrine, and the transposition of its techniques from one sect to another was probably a contributing factor to the doctrinal simplicity of early Hat˙hayoga, which relied upon basic ideas that were extracted from Vajrayāna and Ś aivism. Its simplicity would have made it easy to disseminate and adapt for people of various religions and social circumstances, which might explain why its teachings endured for most of the second millennium. A core group of physical techniques, in particular the mudrās and bandhas, came to define Hat˙hayoga, across sectarian divides, throughout its history and made it a unique method for achieving samādhi and liberation.
The short recension of the Amaraughaprabodha is probably the earliest work to combine Hat˙ha-with Rājayoga and it likely predates the Dattātreyayogaśāstra (13th century), which is a more syncretic text, by at least a century. The number and sophistication of physical techniques associated with Hat˙hayoga gradually increased in the centuries following the Amaraughaprabodha, but the doctrine and theory remained simple until shortly after the Haṭhapradīpikā, when Haṫhayoga was absorbed by Brahmanical traditions. This resulted in the physical practices being merged with more complex systems of the subtle body and philosophical teachings from Pātañjalayoga, Vedānta and Tantra. Thus, by the eighteenth century, some erudite Brahmins were writing large compendiums on Haṫhayoga, such as the Haṭhasaṅketacandrikā, and others composed encyclopaedic digests on yoga that incorporated the distinct methods of Hat˙hayoga. Also, at this time, Upaniṡads based on Haṫhayoga texts were created and references to Hat˙hayoga appeared in commentaries on the Pātañjalayogaśāstra, such as Vijñānabhikṡu's Yogavārttika and Nārāyaṅatīrtha's Yogasiddhāntacandrikā. This process of assimilation marked the decline of Haṫhayoga as a distinct method, insofar as its physical techniques and its emphasis on bodily health had become essential components of more general notions of yoga in the modern period.