In his commentary on Haṭhapradīpikā 1.1 Brahmānanda first states that the author Svātmārāma starts appropriately with a maṅgala, a verse of adoration addressing Ādinātha, i.e. Śiva, but in accordance with his non-sectarian approach leaves room for a second interpretation of the word as Viṣṇu. A sentence later this is contradicted by his explanantion that Ādinātha, who first taught Yoga, taught it to Pārvatī, which limits the scope to Śiva.

The reading rājasaudham in the third verse quarter is well attested by the manuscripts, including α₁. However, the most common reading in manuscripts on the lower branches of the stemma is rājayogam, which was accepted by Brahmānanda in Jyotsnā 1.1 (see below). It appears that some scribes have made a concerted effort to replace words like saudha, vidyā and mārga with yoga in the opening verses of the text (see also 1.2d and 1.3b), even at the expense of the poetic imagery.

In light of the variants, which have led to our critical edition, Brahmānanda’s choices and interpretation of the second half cannot be upheld. This is Brahmānanda’s version and the relevant portion of his commentary:

1. Particle expressing a comparison (upamāvācaka): iva
2. Standard of comparison (upamāna): ladder leading to the terrace (saudhaprāpikā adhirohiṇī)
3. Thing that is compared (upameya): Haṭhapradīpikā that leads to Rājayoga (rājayogaprāpikā haṭhapradīpikā)
4. Common quality (samānadharma): Effortlessness (anāyāsena)

Most witnesses (including α₂) have rājayogam ajānatām (‘for those ignorant of Rājayoga’) in 1.3b. The reading rājamārgam ajānatām (α₁ and α₃), “for those unable to find the royal highway”, is more appropriate to the metaphor of being lost in darkness.

As Brahmānanda notes, the compound kṛpākaraḥ can be understood as one who is compassionate (kṛpā + kara) or one who is a mine (i.e. a rich source) of compassion (kṛpā + ākara). In the Devanāgarī transmission, the kṣamā of kṣamākaraḥ probably arose as a mistake for kṛpā.

The word athavā (‘or’) is well attested but difficult to construe here. Brahmānanda understands it as a conjunction (athavāśabdaḥ samuccaye), and this is how we have interpreted it. The variant mahāyogī in ζ₁ and other manuscripts (, J₄, J₁₁ and K_o) is probably an attempt to remove the difficulty of understanding athavā. One could adopt tathā, which is attested by α₃, but this would be at odds with the weight of the evidence supporting ’thavā.

In Śaiva texts which predate the Haṭha corpus, Mīnanātha and Matsyendra are one and the same, but they are differentiated in later Tibetan and Indian lists of siddhas (Mallinson 2019: 273 n. 35).

Two manuscripts of the α and δ groups have the variant reading virūpākṣaḥ savālikaḥ ( and ) for virūpākṣabileśayāḥ. In another α manuscript, , the original reading was savālmikaḥ, which may reflect an attempt at some stage to change savālikaḥ to a name similar to Vālmīki, the celebrated author of the Rāmāyaṇa.

The α manuscripts have goraṇṭaka, and several other manuscript groups have pauraṇṭaka. We are yet to find the name goraṇṭaka in other Sanskrit texts but it may be a Sanskrit rendering of Goraṇṭakuḍu, which is the name of a disciple of Gorakṣanātha in the Navanāthacaritramu (Jones 2017: 194 n. 3). The spelling koraṇṭaka is attested in the Haṭhābhyāsapaddhati, and it is reasonably well attested by manuscripts of the Haṭhapradīpikā, as well as those of the Haṭharatnāvalī (which also has gonandaka).

The compound siddhapāda could be a respectful affix to Surānanda, but it seems unlikely here because then the name would cross the pāda break.

It is possible that pūjyapāda is a respectful affix to the name Kaṇerī. The variant dhvaninātha may have resulted from a transposition of the first two syllables of nityanātha.

The α group supports kākacaṇḍīśvaro gayaḥ but we have not been able to find evidence for a Siddha called Gaya.

The name Allamaprabhudeva (sometimes Allama Prabhu Deva or Allama Prabhudeva in secondary literature) is frequently transmitted as allamaḥ prabhudevaḥ (cf. α₂), as though it were two names. However, manuscripts α₁, α₃ and π₁ do not have the visarga and write it as one name (i.e. allamaprabhudevaś ca). This is also the case in some manuscripts of the Haṭharatnāvalī (P,T,t1 in Gharote et al. 2009: 35 n. 8).

The names Nāgabodha, Nāgabodhi, and Nāgadeva are all possible in 1.8c. The adopted reading nāgabodhaś ca is attested across several primary groups of manuscripts. We consider the readings of α₁ (namioḍḍīśa) and α₂ (nāma tuṃḍīśa) to be corrupt.

The α₁ and α₂ reading of siddhaḥ kāpālikas is an exception among the manuscripts and seems too vague to be referring to someone within a lineage. Khaṇḍakāpālika is well attested by the remaining manuscripts (including α₃) and this name appears in other texts, e.g. Vajrapāṇi’s Laghutantraṭīkā, p. 45, where Khaṇḍakāpālika is the first of the 24 Vīras (vīrāḥ khaṇḍakāpālikādayaś caturviṃśatiḥ). It is likely to refer to an ascetic who carries a broken skull. Matsyendrasaṃhitā 33.2 mentions a practice for which one needs a khaṇḍakapāla and in the Saṃvaramaṇḍala of the Niṣpannayogāvalī, p. 26, Vajravārāhī is kapālakhaṇḍakṛtakaṭibhūṣaṇā. The compound khaṇḍakāpālika is found at Kathāsaritsāgara 18.2.6, but there khaṇḍa is being used in a derogatory sense (18.2.15 refers to the same character as a duṣṭakāpālika).

The reference to brahmāṇḍa (‘the world’) implies that these great Siddhas have attained liberation-in-life (jīvanmukti) and physical immortality.

The α group omits the second line of this verse, but this was probably the result of eyeskip caused by the repetition of maṭho haṭhaḥ. The reading jagatāṃ makes good sense with ādhārakamaṭhah in light of the cosmological notion that the worlds are supported by a tortoise. However, it appears not to have been understood by some and was changed in other witnesses to yuktānām (e.g., π in the collated group), which was later adopted by Brahmānanda in Jyotsnā 1.10.

Either the singular or plural of yogin could be read here. The singular is well attested among the testimonia, but the manuscript transmission favours the plural.

The term maṭhikā occurs in narrative literature and yoga texts in the sense of a small hut. For example, in the Kathāsaritsāgara (12.9.14, 29–30), maṭhikā refers to the small hut built in a cremation ground by a young Brahmin who makes as his bed the ashes of the dead girl he had hoped to marry. In several other stories (Kathāsaritsāgara 6.6.132, 10.5.89, 12.25.35), maṭhikā is the term used for the hut of an ascetic. In an elaborate description of the huts (maṭhikā) used for Haṭhayoga, the author of the Haṭhābhyāsapaddhati states that the dimensions of the hut are four “hastas” high and wide (there are various definitions of the term: 18 inches according to the Larger Petrograd Dictionary and Monier-Williams (1899), 48 inches according to the Smaller Petrograd Dictionary; the NWS lists even more variations, but the stipulation that the hut is the length of a bow found in the Gorakṣaśataka (on which see below) suggests that 18 inches was meant). The hut can be made of various materials, such as red earth, ashes, plaster and so on (Birch and Singleton 2019: 17–18).

In the Jyotsnā and printed editions of the Haṭhapradīpikā, including one by Digambara and Kokaje (1970: 6), this verse has the additional line, dhanuḥpramāṇaparyantaṃ śilāgnijalavarjite. This line derives from the Gorakṣaśataka (32cd), which has paryante instead of paryantaṃ. It stipulates that the hut should be built in a place measuring up to a bow length and free from rocks, fire and water. None of the early manuscripts has this line, which suggests that it was added at a later time. Nonetheless, it appears in over a dozen manuscripts that were consulted for this edition. These manuscripts are not close to an early hyparchetype of the text.

The syntax of this verse is problematic. One would expect the features of the hut, which are listed in the first three quarters of the verse, to be in the nominative case. Then, the words idaṃ lakṣaṇaṃ in the fourth quarter would refer back to them. However, the compounds in the first three verse-quarters appear to qualify lakṣaṇa as though they were adjectives, and this seems to have been the way the verse was composed.

The manuscripts preserve many different readings at the end of the compound beginning with arandhragarta. We have adopted piṭaka, which usually means “a basket” but can also mean “a boil or blister,” because it is well attested and might here refer to bumps on the floors or walls that would make them uneven. Another possibility is piṭharaṃ, which can have the sense of potsherds and would here mean that the hut should be free of rubbish on the floor. One would expect a word for a defect in a hut that is similar to, but not the same as, cracks (randhra) and holes (garta). For this reason, the reading vivaraṃ looks like a patch, as its meaning does not add anything to randhragarta. The reading viṭapaṃ (‘the young branch of a tree or creeper’) attested in some manuscripts of the Haṭhapradīpikā is difficult to construe in this context unless it was intended to refer to creepers or branches that might invade or encroach upon the hut.

Manuscripts of several groups, namely ε, η, and π, have bādhojjitaṃ, whereas γ, δ and the Yogacintāmaṇi have the more easily understood reading of jantūjjhitaṃ (‘free from creatures’). The α group is split on this, with α₃ (bodhojhitaṃ) closer to bādhojjhitaṃ and α₁ (jyaṃtyūpsitaṃ) and α₂ (jaṃtūṣṇitaṃ) closer to jantūjjhitaṃ. We have adopted the more unusual reading of bādhojjhitaṃ with the support of a similar description of a hut in Suśrutasaṃhitā 6.17.67 (gṛhe nirābādhe).

Since many scribes do not use an avagraha, we cannot be sure whether to understand niyamagrahaḥ in pāda b as having a negative prefix. Although yama and niyama are not included in the Haṭhapradīpikā as auxiliaries of Haṭhayoga, verse 2.14 implies that some niyamas are necessary at least in the early stages of establishing a practice. However, verse 3.78 suggests that a yogi who does not practice niyama might obtain success in yoga through the practice of vajrolī. Thirty of the manuscripts consulted for this critical edition contain additional verses on ten yama and ten niyama, which are usually inserted after 1.16 and derive from either the Śāradātilakatantra (25.7–8) or the Vasiṣṭhasaṃhitā (1.38, 1.53).

α and several other groups of manuscripts have tattvajñānārthadarśanāt (as found in Bhagavadgītā 13.11), tattvajñānāc ca darśanāt or something very similar in the second pāda of the verse, but darśanāt (α₁) by itself is problematic: a vision of what? The early sources of this verse, in particular the Śivadharmottara, indicate that the second verse quarter read as santoṣāt tattvadarśanāt, which makes much better sense of the word darśanāt so we have adopted that reading.

It should also be noted that the word tattva could have a more specific meaning in the Haṭhapradīpikā (4.32–33) as Svātmārāma states that it is a synonym of samādhi. In other yoga texts, it can sometimes refer to the practices of yoga (e.g. tritattva in Amṛtasiddhi 13.12, 14.2–3) or, more generally, to the highest reality or truth (e.g. Amanaska 1.2, 1.20–21, 2.17 etc.).

The reading aṅgapāṭavam is attested among many of the early manuscripts, including the α group. Although this compound rarely appears in other yoga texts, a similar term, śarīrapāṭava, occurs in the Śivasaṃhitā (2.35) as one of the benefits bestowed by digestive fire (vaiśvānarāgni), which indicates that the word pāṭava was used in relation to the body and the benefits of yoga. The compound aṅgapāṭava seems to imply the optimal functioning of the body. The variant reading, aṅgalāghava (‘lightness of the limbs’ or ‘dexterity’) is more common in yoga texts and similar formulations occur even in works known to Svātmārāma, such as the Dattātreyayogaśāstra (śarīralaghutā) and the Amanaska ([...] laghutvaṃ ca śarīrasyopajāyate). It appears that the less common term aṅgapāṭavam was changed to the more widely used notion of aṅgalāghava, perhaps early in the transmission, as the latter is attested by manuscripts in several early groups (i.e. γ, δ, and π).

On the historical implications of these two traditions of postural practice in early Haṭhayoga, see Mallinson 2016: 119–122 and Birch 2018a: 45–46.

One might wonder how the soles of the feet could be placed between the knees and thighs. Brahmānanda explains that the region of the shank near the knee should be understood by the word ‘knee’ in this verse (atra jānuśabdena jānusaṃnihito jaṅghāpradeśo grāhyaḥ jānusaṃnihito jaṅghāpradeśaḥ). This is consistent with the earliest known description of svastikāsana, found in the Pātañjalayogaśāstravivaraṇa (2.46), which states that the big toe of one foot is tucked in between the shank and thigh of the other so it is not seen (dakṣiṇaṃ pādāṅguṣṭhaṃ savyenorujaṅghena parigṛhyādṛśyaṃ kṛtvā tathā savyaṃ pādāṅguṣṭhaṃ dakṣiṇenorujaṅghenādṛśyaṃ parigṛhya yathā ca pārṣṇibhyāṃ vṛṣaṇayor apīḍaṇaṃ tathā yenāste tat svastikam āsanam). For a discussion of svastikāsana in the Pātañjalayoga tradition, see Maas 2018: 68–69. The descriptions of svastikāsana in early Śaiva Tantras do not mention the inserting of the toes between the knees and thighs (see Goodall 2004: 348–350, fn. 371).

This posture first appears in some Vaiṣṇava Saṃhitās that predate the Haṭhapradīpikā, including the Ahirbudhnyasaṃhitā and the Vasiṣṭhasaṃhitā, which is likely to have been the source of this verse. The position of the ankles is the same in all the source texts. The Ahirbudhnyasaṃhitā adds that the hands are crossed behind the back and hold the big toes. For illustrations of six possible positions of the arms and hands, see Gharote, Jha, Devnath, Sakhalkar 2006: 111–113.

Although most witnesses have tathā in 1.21a, the word atha has been accepted because it is attested by α₃ and η₁, the sources and the testimonia. It appears to be verse filler here rather than indicating a temporal sequence of actions. Svātmārāma borrowed the verse on vīrāsana from the Vasiṣṭhasaṃhitā, the redactor of which appears to have adapted its first line from a description of this posture in the Śāradātilakatantra. This would explain the rather strange syntax of the Vasiṣṭhasaṃhitā’s version, in which adhaḥ kṛtvā was changed to athaikasmin, and tathetaram became ca saṃsthitam. It seems that saṃsthitaṃ must be understood with ūruṃ in the third pāda in the sense of saṃsthāpya (i.e. ‘having placed’).

Different versions of vīrāsana are found in earlier tantric and yogic works, such as the Kiraṇatantra, Hemacandra’s Yogaśāstra and commentaries on the Pātañjalayogaśāstra. For a discussion of some of these, see Maas 2018: 66–68.

In the first quarter of the verse, the witnesses are split between nirudhya (‘having blocked’), nibadhya (‘having bound’), niyamya (‘having restrained’) and niṣpīḍya (‘having pressed’). The source, the Vasiṣṭhasaṃhitā, and two manuscripts of the γ and π groups support nirudhya whereas α₂ and the testimonia support niyamya and α₃ has niṣpīḍya. While nirudhya makes sense here (i.e. ‘having blocked or closed the anus...’), we have adopted niṣpīḍya because it is better attested among the α, ε and η groups.

The word vyutkrameṇa describes the position of the ankles. Its basic meaning is ‘against the normal direction.’ In āsana descriptions it usually means ‘crossed’ (see e.g. Vasiṣṭhasaṃhitā 1.71), which is how we have understood it here. It could also mean ‘turned out’: if the yogi is in a kneeling-type position, turning the feet out would bring the ankles together, blocking the perineal area. See Yoga Mīmāṃsā, vol. 8, no. 2 (1965: 29–30) for a discussion of vyutkrameṇa and the position of the ankles in kūrmāsana, and vol. 8, no. 2, figures 3–6 for photographs of a practitioner performing this āsana.

The names kurkuṭa and kurkkuṭa in some manuscripts are variant spellings of kukkuṭa attested in the Pañcatantra (MW).

The oldest dated manuscript, η₁, has kukkuṭāsanavat kṛtvā, which is a simpler alternative to the widely attested reading kukkuṭāsanabandhasthaḥ (including α₂ and α₃), which we have accepted. Since there is no known source for this verse other than the Haṭhapradīpikā, it appears that the reading of η₁ was an isolated attempt to simplify the syntax.

We have adopted the reading dhanurākarṣaṇaṃ kṛtvā, which is in the δ group, as well as the principal testimonia (i.e. the Yogacintāmaṇi and Haṭharatnāvalī), because it fits the overall syntax of the verse, unlike the readings with kṛṣṭaṃ (for kṛtvā) that are found in the early manuscripts. It is curious that kṛṣṭaṃ is so well attested because it seems redundant with ākarṣaṇaṃ. The following reading in Godāvaramiśra’s Yogacintāmaṇi (f. 40r) appears to be an attempt to make sense of kṛṣṭaṃ: dhanurākarṣavat kṛṣṭaṃ dhanurāsanam ucyate.

A different version of dhanurāsana is described in the Haṭhayogasaṃhitā. On the two versions of dhanurāsana, see Hargreaves and Birch 2017.

One of the collated manuscripts of the Haṭhapradīpikā, ζ₃, has a scribal comment stating that dhanurāsana should be done continuously (anavarata) on the left and right sides (tatra ekam dhanurākarṣaṇāsanam āsanaṃ savyāpasavyapādahastābhyām [abhy]ased anavaratam). This would make dhanurāsana a dynamic practice as shown in this video.

In the second verse quarter, most of the manuscript groups have a compound with doṣṇā at the end, as seen also in the Yogacintāmaṇi, Haṭharatnāvalī and Yuktabhavadeva. The instrumental ending (‘with the hand’) works well with the gerund (pragṛhya) in the third verse quarter and the object (dakṣpādaṃ) in the first quarter. This reading indicates that the right foot is held by the hand of the arm that is wrapped around the outside of the left leg, which would be the right hand (dakṣadoṣṇā) rather than the left (vāmadoṣṇā), as shown in Figure 1.

α₃ and most manuscripts of the Haṭharatnāvalī have dakṣapādo in the first pāda. This reading yields the same meaning as the adopted one if read with vāmapādaṃ in the second. However, α₃ and manuscripts of the Haṭharatnāvalī read vāmadoṣṇā, which is not good because it leaves the gerund without an object.

The version of this verse in Jyotsnā (1.26), which is supported by some manuscripts in two important groups, η and π, has two objects of the gerund, namely the left and right feet, without an instrumental or conjunctive particle. In his commentarial remarks, Brahmānanda proposes that the left foot is grasped by the right hand and the right foot by the left foot, as seen in Figure 2.

Here we understand jaṭhara, which usually refers to the stomach, to mean “tough” or “persistent” (cf. Amarakośa 3.3.189: jaṭharaḥ kaṭhine’pi syād adhastād api cādharaḥ).

In 1.27d, the compound daṇḍasthiratvaṃ (‘steadiness of the spine’) is attested by all the important manuscript groups and testimonia, so it was likely original. However, the Jyotsnā (1.27d) has candrasthiratvaṃ (‘steadiness of the moon’), and this reading is well-attested in many manuscripts that are lower on the stemma. Brahmānanda understands steadiness here as ‘the absence of flow’ (sthiratvaṃ kṣaraṇābhāvaṃ), a reference to the moon retaining its nectar.

The reading dorbhyāṃ padāgradvitayaṃ is well attested but is somewhat strange because dos usually means ‘the arm’ rather than the hands. The variant dvābhyāṃ karābhyāṃ dvitayaṃ, “with both hands”, appears to be an attempt to remove dorbhyāṃ, but it introduces the problem of the toes not being mentioned.

The use of the word paścima to mean the central channel is found at Yogabīja 95 (paścimamārgataḥ), 108 (paścime pathi), 117 (paścimadvāramārgeṇa) and 121 (paścimaṃ). Cf. the usages of paścimamārga in Dattātreyayogaśāstra 140 and Śivasaṃhitā 3.110 (from which this verse is likely to be derived). Brahmānanda understands paścima as referring to the Suṣumṇā (Jyotsnā 1.29): paścimavāhinaṃ paścimena paścimamārgeṇa suṣumṇāmārgeṇa vahatīti paścimavāhī.

There is no direct source of this verse, but it has the same elements as two verses in the Vasiṣṭhasaṃhitā (1.76–77), which are themselves derived from earlier Vaiṣṇava sources (see Mallinson 2014: 227 n. 9). The compound uccāsanaḥ in the third verse quarter seems to approximate in a somewhat vague way the Vasiṣṭhasaṃhitā’s reading samunnataśiraḥpādaḥ.

In the second verse quarter, the pronoun in tatkūrpara° refers to the two hands. This is stated more explicitly (i.e. hastayoḥ kūrparau) in Vasiṣṭhasaṃhitā 1.76c and Yogayājñavalkya 3.15c.

The word tu is often used to introduce a new posture, but in this case seems to be a verse filler.

In the first and third verse quarters, Svātmārāma appears to have rewritten Śivasaṃhitā 3.96 to include the information that it was Śiva (śivena) who taught the eighty-four āsanas, whereas in the source Śiva is himself speaking. Svātmārāma also changes the meaning of the second half of the verse, as the Śivasaṃhitā states that Śiva picked out the four best postures and taught them, whereas in the Haṭhapradīpikā it reads as though Svātmārāma himself is responsible for picking out the four best postures and teaching them. There are other instances in the Haṭhapradīpikā where Svātmārāma borrows a verse with a first person verb (e.g. 3.43). However, in this instance, he may have intended to indicate that he chose the four postures coming after this verse (i.e. siddha, padma, siṃha and bhadra) because the Śivasaṃhitā follows 3.96 with teachings on the postures called siddha, padma, paścimottāna and svastika. The vocative here (sakhe) is also found in Haṭhapradīpikā 4.56, which is also likely to be an authorial verse.

It is likely that the original version of this verse contained the vocative with the imperative form of the verb (sakhe tiṣṭha). There are other instances where Svātmārāma included a verse with the vocative (e.g. 4.7c, 4.56b, 4.72d, 4.74c) as though the text were a dialogue. Other versions of this verse are transmitted by some manuscripts of the Haṭhapradīpikā, in which the vocative and imperative verb have been removed. In these cases, sukhe and sukham are difficult to construe because the context suggests that the intended meaning was that one should always sit in siddhāsana (as opposed to the other three āsanas), rather than the prescription to always sit in a comfortable siddhāsana.

In the second verse quarter, the adopted reading ekahṛdayo is supported by two manuscripts of the Haṭhapradīpikā (J₂ and ) and is close to the α₁ reading ekahṛdaye. It is also attested by the six-chapter Vivekamārtaṇḍa and the Yogacintāmaṇi, which attributes this verse to an unkown work called the Pavanayogasaṅgraha. In this case, ekahṛdayaḥ appears to describe the yogi as having his mind focused on one thing. There are many variations of this verse quarter in the Haṭhapradīpikā manuscripts, as well as in the manuscripts of the sources and testimonia. Most of the collated witnesses have athaikam eva hṛdaye dhṛtvā, which is close to the adopted reading. Other readings allude here to the practice of the Jālandhara lock, in which the chin is placed on the chest. This is most clearly seen in the Jyotsnā’s version, hṛdaye kṛtvā hanuṃ susthiram (‘having put the jaw firmly on the chest’). The other well-attested reading, athaikaṃ eva niyataṃ, was an attempt to fix the problem of hṛdaye by replacing it with niyatam, which must be read with meḍhre pādam athaikaṃ (‘having fixed one foot on the penis’), but niyatam is redundant here because of vinyaset in the first verse quarter.

Svātmārāma’s introductory and following remarks to verse 1.36 indicate that he preferred the siddhāsana of the Vivekamārtaṇḍa over the version taught as muktāsana in the Vasiṣṭhasaṃhitā and Yogayājñavalkya.