Text in context

Judging by the large amount of both direct and indirect references to the followers of Haṭhayoga, we can safely state that this form of yoga became a widespread practice on the Indian subcontinent, especially between the 13^th and the 15^th centuries, and probably well onto the 17^th. No doubt, it became a significant ascetic school with a strong influence in South Asian religious and cultural spheres. Gorakhnāth, the proverbial representative of this form of yoga, continues to play a crucial role in these spheres and his rather shadowy historicity is tantamount to the Nāth tradition’s own fuzzy historicity. A good number of works have been attributed to Gorakh and allusions to him are available in a wide range of sources. Regardless of whether others cited Gorakh as an authorial figure, or instead attempted to ignore him (see e.g. ^{Lorenzen and Thukral, 2005}), what remains clear is that Gorakhnāth -or a Gorakh-related sect- was a key player in the game of religious groups and identities in pre-modern India.

As concerns the Yogabīja, one of the particularities of the text is the fact that its contents do not correspond completely to the topics dealt with by other reputed and later texts on Haṭha (e.g. the Haṭhapradīpikā and the Gheraṇḍasaṃhitā). This is despite the fact that the Yogabīja is purportedly an important Haṭha treatise. The Sanskrit bīja means “seed, semen, grain, or germ” and can imply an “origin” or “kernel”. Thus “yoga-bīja” can be translated as “the seed of yoga”. Accordingly the text claims to delve into the essence and core of yoga, and so it explains some practices for breath control, such as mudrās (“seals”) and bandhas (“bonds, ligatures”) such as jālandhara, mūlabandha, or uḍḍiyāna. It also describes different types of knowledge (passim), of ascetics (§34 ff.), of siddhis or powers (passim), or of yogas (§145-152), among other yogic items. It is also possible that the text belongs to a relatively early phase of Haṭha literature (circa 12^th-15^th cent.), a matter that is gaining increasing attention from specialists.

Both the Gorakṣaśataka and the Yogabīja (YB) are attributed to Gorakhnāth, the legendary founder of the Nāth Panth who, according to a colophon of the YB in one of the printed editions is also known as Maheśvara (vide infra). The exact location of the text’s composition is uncertain, but ^{Dasgupta (1995: 373)} suggested Bengal as the probable place of origin for the YB and for kindred, Gorakh-related texts. The yogic corpus including these texts, as well as the Amanaskayoga, the Siddhasiddhāntapaddhati, and others, probably belongs to the period 12^th to 15^th century in pre-modern India, more or less coinciding with the Delhi Sultanate. During this period there was a documented exchange between Sufis, fakirs, and yogis, a topic beyond the scope of the present article, yet of noteworthy attention. Some Haṭha-related texts have been focused upon by scholars, notably the Ṣaṭcakranirūpaṇa (Avalon 1959), the Amanaskayoga (^{Birch, 2005}), and the Khecarīvidyā (^{Mallinson, 2007}). There is also a group of some twenty so-called Yoga Upaniṣads which were likely composed and collected well into the 18^th century which have been edited, translated, and discussed (see bibliography and ^{Bouy, 1994}). Although some scholarly attention has been drawn toward the relationship between these Upaniṣads and other texts on Haṭha, there still remains a great deal of work to be done in order to fully comprehend the relationship between them.

The YB and the Yogaśikhā Upaniṣad (sometimes only Śikhā Upaniṣad) were discovered to share some verses by scholars such as ^{Bouy (1994, p. 112)} and ^{Feuerstein (2001, p. 421)} but no researcher has as yet attempted to establish the history and actual relationship between the two texts. The early Amanaskayoga was probably composed around the 12^th-13^th century but most certainly not after the 16^th (^{White, 2009, p. 46}; ^{Birch, 2005, p. 4}). On the other hand, the Haṭhapradīpikā, the Gherāṇḍasaṃhitā, and the Śivasaṃhitā date collectively to the period between the 15^th and 17^th centuries. In my translation of the YB, I stated that the probable dates of its composition were 1450-1500 CE (^{Muñoz, 2014, p. 476}). However, and bearing in mind that the Haṭhapradīpikā belongs to the 15^th century and is very likely posterior to the YB (cf. ^{Bouy, 1994, pp. 81-86}), we may place the YB slightly earlier chronologically, but still after the circa 1400 CE Gorakṣaśataka (^{Mallinson, 2012, p. 263}). If it is accepted that the Gorakṣaśataka does indeed belong to the very early 15^th century and that the YB was composed soon thereafter, it appears possible that the YB was written between 1400 and 1450, some decades before the Haṭhapradīpikā, which seems to be a much more systematic and developed account of Haṭha practices. This latter text, therefore, seems to follow a (long?) line of previous texts upon which it draws.

Arrangement and printed versions

As far as I have been able to surmise, there are three versions of the YB which have been edited and published to date, all of which have been incorporated in my transcription. Gorakhpur’s YB was edited by Rāmlāl Śrīvāstav and is composed of 190 verses, which amount to around 360 lines due to verse variations. I shall refer to it as either the Gorakhpur recension or Śrīvāstav’s edition. As a whole, the YB presents recurrent grammatical irregularities and ambiguous lexicography, as in the other two recensions. Besides a Hindi introduction, this edition provides a commentary (tīkā) also in Hindi. The printing is not of a good quality and hence there are many instances in which vocalic matras have apparently faded out: notably, what many times looks like a long ā was presumably a prior, original o. Some confusion arises with the long ī too. Anusvāras as well as some avagrahas also tend to be either erased or not printed correctly and are therefore provided in the critical transcription according to the rules of Standard Sanskrit.

There are at least two publications of this version, one which was included in a periodical issued by the Gorakhnāth Temple, comprising also the Gorakṣaśataka, the Siddhasiddhāntapaddhati, the Vivekamārtaṇḍa and the Amaraughaprabodha, and an independent imprint, also published by the Gorakhnāth Temple (see bibliography). There are some orthographic and grammatical disagreements among them, which is intriguing since both editions were published under the supervision of Śrīvāstav. No remark on this has been made by the editors. Also, the Hindi tīkā differs between editions.

Unfortunately, Śrīvāstav does not mention which, and how many, manuscripts were utilized for his version. Judging from the obvious grammatical imperfections, it seems likely the he relied on hand-written manuscripts rather than on a printed edition, but this remains as yet to be confirmed. As noted above, though not extremely common, this recension does not commit wholly to regular ślokas (two-line verses), but has one-line verses (around eighteen) and three-line verses (two), whose distribution and/or numbering in most cases do not correspond with those found in the other two recensions. The title is indicated by a non-declined compound: “yogabīja”.

Although not more than a mere conjecture, I am inclined to believe that this edition may coincide with a potential “earlier” version than the other two printed recensions at my disposal. In these two, there is an ostensible attempt to correct apparent mistakes in a conceivable “Gorakhpur recension” (yet, sometimes the amendment does not seem entirely correct).

A second edition was published in Delhi under the supervision of Brahmamitra Awasthi, heretofore labelled as either the Delhi recension or Awasthi’s edition. This version is correctly labelled as “yogabījam”. It apparently has two verses less than the Gorakhpur edition. However, apart from order changes and lexical discrepancies, it should be noted that Awasthi’s edition presents a serious issue with verse numbering: after verse §16 (pp. 55-57), the numbering re-begins at §111, so that to what seems to be a whole of 182 verses we have to add six more; thus giving us a total of 188 verses approximately. A second issue then comes to light after we resume the numbering, at §120, where there is a long sequence of one-line verses (adding up to almost thirty). Moreover, after §132 the numbering inexplicably changes to §154 (later on, at least twice there seems to be some error when giving numbers for §170) and the again after §179 the numbering goes back to §140-41. All of this confusion gives one the feeling that this Delhi recension, with all its shifting and jumping of verse-numbers, tries to catch up with the Gorakhpur recension after having failed to go on a par since the early stages of the text.

All in all, the verses follow almost faithfully the order in Śrīvāstav’s edition, but sometimes the numbering and organization of individual lines do not match. For example: verses §5 and §6 are inverted in Awasthi’s edition and Awasthi’s verse §144 is not in Śrīvāstav’s. Finally, in spite of being a bit shorter, Awasthi’s edition manages to “include” almost all of the Gorakhpur recension. Another difference between them is that Awasthi’s edition tends to bear recurrent sandhi blunders: e.g. yogāt candra instead of yogāc candra. Besides a free English translation, this edition is also accompanied by a Hindi paraphrase. Like Śrīvāstav, Awasthi gives no hint of his sources, but they seem to be altogether different from those in the Gorakhpur edition. The only thing Awasthi does in the footnotes is indicate similarities between the Yogaśikhā Upaniṣad and the YB (“tulanīya yogaṣikhopaniṣad…”).

Lastly, there is the Yogaśikhā Upaniṣad (YŚU), included in the Yoga Upaniṣads. The edition I have used was prepared by Mahadeva Sastri in Madras. Consequently, it will be designated either the Madras recension or Sastri’s edition. The text is divided in two parts, or adhyāyas; around 80% of the first adhyāya (178 verses in total) is basically comprised of the Yogabījā.

Unlike the other two recensions, the YŚU is very consistent in terms of its verse-structure, i.e. all of the verses are composed of two lines regardless of the metre. It is very probable that this upaniṣad is a late work, even later than the Haṭhapradīpikā and the Gherāṇḍasaṃhitā. It is also likely that the people behind the composition of the YŚU were not members or adherents of the Nāth order. This can be surmised from the fact that in some instances the text eliminates references to the nāthas (cf. YB §7, §136, §145, §189). In general terms, Sastri’s edition of the Yoga Upaniṣads appears more carefully realized than the other two YB editions; this may well be in consonance with the nature of most recensions from south India of various texts. These tend to offer rather “purified” versions (^{Törzsök, 1999, p. xxvii}). Yet the apparent lack of mistakes in the YŚU does not necessarily imply a more “original” or “pure” reading of a hypothetical YB original text; these readings, on the contrary, can rather convey “corrections” stemming from ideological considerations, as is evident from the change of speakers in the dialogue.

Sastri’s edition was based on five manuscripts located in the Adyar Library and a Bombay printed edition of the 108 upaniṣads, which was itself based on a South Indian ms. Sastri’s edition has a Sanskrit commentary by Śrī Upaniṣad-brahmayogin, but no paraphrase or translation (an English translation of the Yoga Upaniṣads was later published independently). The recension’s rendering of the text tends to coincide with some readings of the Delhi recension, though this is not always the case. At times Sastri gives variant readings from his sources, which may coincide with one reading in either D or G, but which for some reason were discarded. It needs to be stressed that nearly without exception the verses from the YB not repeated in the YŚU are, significantly, either Devī’s speeches or Ῑśvara’s replies to her. It also happens that a discourse by Devī becomes merged in the main body of the text (e.g. §24-25), and is, therefore, not singled out from the principal voice.

We know for sure that the sources for the YŚU were located in South India, but cannot be so certain about the sources for the Gorakhpur and the Delhi printed versions. It may be the case that both the Gorakhpur and the Delhi recensions were based on mss stored in locations such as Jodhpur, in Rajasthan, and Kathmandu, Nepal, where we know that there is a vast array of yoga-related texts to be found. Allegedly, the Madras recension relied at least on one northern recension (arguably close to the Delhi recension). One could of course argue that it may very well be that the line of derivation moves in reverse; that is to say, one of the northern recensions was based on a southern one, but in this case I feel confident that it is not so because a) the Madras recension does seem to betray a clear intention of purging non-orthodox elements from the text while retaining most of the content and original order, and b) as Otto Shrader has proved the current, southern versions of the Yoga Upaniṣads are greatly indebted -and have expanded on- earlier northern, and sometimes shorter, recensions (^{Bouy, 1994, p. 30}). Moreover, I feel that what looks like grammatical correctness in the Madras version is an indication of a later, more polished recension.

There are at least two other editions: an 1899 edition from Kaunpur and a 1956 one from Mangalore, the second of which seems particularly close to the YŚU’s first chapter (^{Bouy, 1994, p. 105}). Unfortunately I have not been able to consult either edition. Unlike the Gorakhpur edition, these appear to have 180 verses, which more or less coincide with Awasthi’s edition.

Hereafter recensions are identified by place of publication as follows: Śrīvāstav’s YB edition (G); Awasthi’s YB edition (D); M. Sastri’s YŚU edition (M).

Meter and style

The YB’s prosodic structure is that of vṛttapadyas, i.e. stanzas organized by the number of syllables in each pāda or hemistich. For the most part the text is composed in anuṣṭubh stanzas (eight syllables per pāda) shifting between forms. This is usually restricted, however, to the pathyā form (with a tendency for ya-gaṇa, or bacchius trochaic: ̮--; and ra-gaṇa: -̮-); also, the syllabic pattern at the end of the second and forth pāda of each stanza tends to be a regular iambic (̮-̮-). There are four verses in a different meter in the YB: §66, which is triṣṭubh (11 syllables in each pāda) in indravajrā meter; §101, a śakvarī (14 syllables) in vasantatilaka; §137 a jagatī (12 syllables) in a more or less vaṃśastha meter; and §138, again a triṣṭubh oscillating between indravajrā and upajāti.

The stylistic quality of the Sanskrit is allegedly poor in aesthetic terms, yet it is quite consistent with other kindred texts. In general terms, the grammatical and lexical irregularities are of the usual sort in this type of literature, which has been given various names. Some scholars have tended to call it Tantric Sanskrit, for its primary manifestation is found in Tantric and/or āgamic literature; others speak of Aiśa language because it is said to derive from God Śiva (i.e. Ῑśa), an idea first expounded by the 11^th century philosopher Kṣemarāja. There are a number of specialists dealing with this linguistic phenomenon as well as with the prosody of said literature (see for example ^{Goudriaan & Schoterman 1988, pp. 44-109}; ^{Törzsök, 1999, pp. xxvi-lxix}; ^{Mallinson, 2007, pp. 15-16}). They have attempted to schematize a functional grammar of Aiśa language and have emphatically suggested that grammatical incongruities derive basically from prosodic concerns.

In terms of stanza division, we can also identify inconsistencies between the available versions. As noted before, for the most part the YB is in anuṣṭubh versification, but there are some departures. Besides the other verse forms, two recensions (G and D) have both one-line and three-line verses, which accounts in large part for the incongruence in numbering the verses. It is doubtful that three-line verses are meant to function as a form of the gāyatrī meter, but not impossible. Some examples of one-liner units in the transcription are §43, §59, §72, §109, §133; three-line verses are §152 and §163. M is definitely more consistent: it presents us with a constant series of couplets throughout the text; M also provides headings to indicate changes of topic, usually at the expense of the sequence of couplets. For example, more often than not a śloka is split in order to allow for a heading and change of topic.

Transcription and collation

My transcription is presented in the Roman alphabet, supported by diacritical marks as specified by the international transliteration system for Sanskrit (IAST). When two or more words in the original become blended due to graphic rules in the devanāgarῑ script, I have decided to write these words separately (e.g. devairapi → devair api). When a union is due to nominal compounds (samāsa) or phonetic rules (sandhi), the coalition of words has been preserved (e.g. yogamārge, sarvadoṣavivarjitaḥ, tatrāsau).

The work undertaken for this partial critical transcription was based on the three “recensions” mentioned above, i.e. four printed versions. The significant variant readings between recensions have been indicated in the notes to the text. At some points, I have only indicated the alternative readings; at others, I have provided readings from all recensions. Sandhi deviations have not been indicated in the notes; neither have visarga or anusvāra omissions. Needless to say the “corrected” reading provided by this transcription may imply a rethinking of both the Spanish and English translations. Given the fact that habitually apparent inaccuracies in all existent transcriptions are due to metrical concerns, I have opted here to provide a semantically and grammatically accurate reading when possible. A particularly difficult passage is §79cd: the two key words there (rajjau>rajjuḥ>rajjvā / rajvī), in whatever form, seem to contradict the grammar or semantics of the rest of the pāda.¹ Other passages in haṭha literature that deal with a rope (rajju) are Śivasaṃhitā 1.38 and 1.43 but are completely different from YB´s passage; Śivasaṃhitā deals with the standard analogy between rope and snake to explain the lack of true knowledge instead of the interweaving of mind and breath that the YB aims to.

Thus I somehow decided to artificially reconstruct the passage. That is to say that I have shown more concern for a “possible” correction rather than complying with phonetic constraints. In this sense, my transcription is closer to Birch’s critical edition of the Amanaskayoga than, for example, Törzsök’s edition of the Siddhayogeśvarīmata. I have also retained G’s numbering of stanzas, even to the point of accepting the one- and three-line verse forms.

Although I have devoted a great deal of effort to comparing the different editions and, as far as possible, have attempted to amend incorrect readings and errata, the task is far from completed. It has to be stressed that this transcription is based solely on printed materials: a full critical edition will have to resort to extant manuscripts, as well as other possible printed editions as they become available. Access to these archives from Latin America is rather difficult because of time and financial issues, as well as for bureaucratic reasons (cf. ^{Birch, 2005, pp. 23-24}). Even though a closer look at manuscripts will have to be made, we can accept that the printed versions consulted here were prepared from surviving manuscripts. Any later critical work will have to determine what actual sources were drawn upon in each of the printed recensions and suggest a more definite “standard” archetype.

Nevertheless, despite the possibility of misreadings from the editors, I deemed it viable to provide a more or less homogenous and reliable unified transcription from the four printed editions due to the high frequency of verses with which the same verses appear in the different recensions. As for lexical discrepancies, some of these stem from both obvious errata from scribes as well as compliance with metrical requirements, while some others demonstrate ideological dissent.² Notably all references to a dialogue between Īśvara and Devī (YB) have been modified so as to present a conversation between Śaṅkara and Padmasaṃbhava (YŚU), apparently a more orthodox scheme for a religious colloquium. We musk ask ourselves, however, is this a teaching from Śiva (or the Advaitin philosopher) expounded to the reputed Buddhist teacher? If so, there may still be some underlying tantric undertones.

Inconsistencies among G and D suggest that they were based on different manuscripts, but these differences also reinforce the thesis that various early versions of the text did in fact offer a dialogue between Īśvara and Devī, not Śaṅkara and Padmasaṃbhava, which points toward a later “sanitization” of allegedly śākta elements in the text. It is also sensible to conclude that some YB recensions were earlier than a YŚU recension.

G presents a total of 364 pādas, whereas D gives 379 or 373, depending on how we count verses in dispute (vide supra). Finally, M (included in the YŚU) contains 356 pādas, almost equal to G, although in principle M is shorter than either G or D due to its rejection of Devī´s speeches as I have indicated above. Among the recensions, there are 303 pādas shared by all three-which is also the number of half-verses shared by G and M. The number increases if we compare only G and D: 357 shared half-verses. This quantity decreases if we compare D with M: 312 half-verses, three of which are unique to these texts and absent in G. (These I have included in the appendix to the transcription.) The correspondence of M with G is roughly YŚU 1.1ab-1.69ab, and 1.79cd-1.164cd. The rest of the first adhyāya goes from 1.165ab through 1.178cd. The opening of M coincides with G 4ab. I have only rarely made mention of omitted passages from the recensions. In order to provide a more detailed account, a correspondence chart is under preparation.

I have adopted Śrīvāstav’s two editions from Gorakhpur (G) as my starting point, on the assumption that they may very well represent an early stage of the YB. When needed, I have referred to these publications as either G¹ (1982 edition) or G² (1991 edition); if no distinction is made, it is because the two editions agree on the point under scrutiny. I have then compared this with both Awasthi’s edition from Delhi (D) and the YŚU recension (M), which seems to be a more carefully executed version in editorial terms. When two sources agree, I have sometimes opted for the reading presented in these two works, unless the single independent reading comes from G and can be logically read and/or construed. When discrepancies prove difficult to reconcile, I have tried to look for recurrence in other texts, particularly the Haṭhapradīpikā. This is despite the fact that there aren’t many shared verses among them; perhaps eighteen or less. Amendments not supported by either source are marked with the symbol º after the word in question. An asterisk (*) indicates the faulty numbered stanzas in D, which I have not attempted to correct. I hope to carry out a more thorough and comparative survey as soon as time permits.