ETCHED IN STONE: SIXTEENTH-CENTURY VISUAL AND MATERIAL EVIDENCE OF ŚAIVA ASCETICS AND YOGIS IN COMPLEX NON-SEATED ĀSANAS AT VIJAYANAGARA

his article reassesses the history of postural yoga in precolonial India by drawing attention to recently discovered visual material evidence of non-seated postures carved onto the pillars of Vijayanagara temples at Hampi in Karnataka. Based on inscriptional evidence dating to the early 1500s CE, these sculptures represent important and overlooked early visual evidence for the practice of standing postures, inversions, and complex “pretzel-shaped” balancing postures in latemedieval South India. A number of sculptures bear a marked similarity to certain non-seated āsanas featured in more modern postural yoga systems, and might represent some of the earliest evidence of their existence. To contextualize these images and understand their significance within the larger history of yoga, the article begins with a preliminary genealogy of āsana and postural yoga traditions, highlighting a particular shift from seated to non-seated āsanas that is evinced in both the textual and visual-sculptural record. The author suggests that this shift in psychophysical functionality and praxis of yogic āsana may have opened up new anatomical potentials for engaging the body within a yogic context, and that this shift, alongside intermingling with much older traditions of asceticism (tapas), may partially explain the surge in complex non-seated āsanas featured in many yoga texts following the sixteenth century. Drawing upon other archeological sites, textual, epigraphical, and visual materials, the article makes the case that some of the ascetic figures in complex yogic postures sculpted at Hampi are depictions of Nātha yogis performing the techniques of Haṭhayoga.  T


JOURNAL OF
In this study, I will introduce one sample of recently discovered visual material that I hope can shine new light on the historical development of physical yoga postures (āsana) during the early sixteenth century in Vijayanagara, South India. In January 2016, during a period of brief field work at Hampi, I photographed and surveyed a 4 number of sculpted images of ascetics performing yogic āsanas, displayed on the many pillared halls of the great Vijayanagara temple complexes. While a few of these images have been documented by art historians Anna Dallapiccola, Anila Verghese (1998) and Richard Shaw (2011), most of the postures have not been identified, and many of the figures remain entirely unaccounted for in existing scholarship. Nor has there has been any attempt to understand these unique images within the larger history of physical yoga traditions in premodern India, which this article will seek to provide. 5 The observational and descriptive style of sixteenth-century Vijayanagara sculpted human figures (Dallapicolla and Verghese 1998, 10) make these material depictions of yogis in practice an important historical window onto yoga traditions in or around the 6 capital city of the empire. Sectarian markers featured on some of the sculpted figures allow us to identify certain figures as Nātha yogis, and when read alongside the textual record, suggest that the reliefs are artistic renderings of the techniques of medieval 7 The word "Hampi," anglicized from the Kannada haṃpe, derives its name from the eponymous goddess of 4 the region along the Tuṅgabhadrā river, Pampā (or Pampē), whose presence predates Vijayanagara rule. The letter 'p' in old Kannada often changes to 'h' in more modern registers of the language (Verghese 1995, 16). Technically, Hampi (or haṃpe) is the name of the village surrounding the main Virūpākṣa temple, while the remainder of the capital city is referred to as "Vijayanagara." However, today, the entire approximately 25 square kilometer site is commonly referred to as Hampi, or the Hampi ruins, recognized as a UNESCO World Heritage Site. In this article, I use the words "Hampi" and "Vijayanagara" interchangeably to refer to the capital city, although I favor the former when referring to the contemporary archeological sites.
In addition to Hampi, sculptures of non-seated āsanas are found in other south Indian temples, including 5 Śrīśailam, Śṛṅgeri, and the Raṅganātha temple at Śrīraṅgam, Tamil Nadu. Depictions of non-seated postures are also featured on the entrance gates at the famous Naṭarāja temple of Cidambaram, Tamil Nadu, though here the context is Śiva's dance (I thank Jason Birch for bringing this to my attention). Perhaps the earliest sculptures of non-seated āsanas have recently been observed by Mallinson (forthcoming b) on the northern gate at Dabhoi in Gujarat (1220Gujarat ( -1230. In this article, I use the word "yogi" when referring to the practitioners of yoga broadly speaking; and the 6 Sanskrit yogin when referring to a particular textual passage. Such yogis in the historical record were by and large ascetics-those who had renounced the conventional norms of society in pursuit of religious aims -though certainly not all ascetics practiced the methods that have constituted yoga (see Mallinson 2013a, 81: n. 1).
In this article, I use the term "medieval" to reference the period between 600-1600 CE, and "late-7 medieval" to denote its upper terminus between 1200-1600 CE. However, this is largely a heuristic device. For a historiographical analysis of the limits of the medieval period in India, see Wedemeyer (2013, 58-66).

JOURNAL OF
Haṭhayoga. In cases where more generic ascetics are depicted in non-seated postures, however, I suggest that this could be representing older traditions of ascetic postural practice (i.e., tapas) that influenced Haṭhayoga, as evinced by the incorporation of increasingly more non-seated āsanas in Haṭhayoga texts after the sixteenth century (Birch 2013;forthcoming 2018a).
The depictions of non-seated āsanas carved onto the pillared reliefs at Hampi are striking for their complexity and variation. The reliefs include standing postures, inversions, and unique "pretzel-shaped" balancing postures. Based on inscriptional evidence dating to the early 1500s CE, these sculptures represent important and 8 overlooked early visual evidence for the practice of advanced non-seated postures in late-medieval South India. Moreover, a number of images bear a marked similarity to certain non-seated āsanas featured in more modern postural yoga systems, and might represent some of the earliest evidence of their existence-visual, textual, or otherwise.
This article is organized into three parts. In the first, I begin with a preliminary genealogy of physical and postural yoga up to the medieval and early modern period, with particular attention to South India, in order better to assess the significance of the yogic imagery at Hampi in the early sixteenth century. Specifically, I highlight how an important shift from seated to non-seated āsanas is both evidenced in the texts and affirmed in the sculptural record, and argue, moreover, that this shift in function of 9 āsana opened up new anatomical potential and emancipatory ways of engaging the body in yogic praxis. Shifting our gaze to Vijayanagara, in part two, I provide a brief introduction to the empire in order to foreground the sociopolitical and religious context of the temples on the pillars of which these sculpted images are found, and to contextualize these yogic sculptures within the larger visual programs of the Vijayanagara temples. Turning to an analysis of the images, I attempt to identify Strictly speaking, the inscriptions provide dates for the erection of the temple complexes themselves, and 8 not explicitly for each of the detailed carvings therein. While there is, of course, always the chance of a pillared-sculpture being a later addition, replacement, or repair, without any evidence indicating such a renovation, I assume the dates of the temples to accord with their sculptures. All dates in this paper have been converted to the Common Era; thus, the "CE" will be dropped henceforth.
It should be cautioned that our understanding of the relationship between yoga praxis and textual 9 production is still emerging. A moment of textual codification does not necessarily correspond to a moment of innovation in yogic praxis or the invention of an āsana. It is, in fact, more historically likely that just as with the temple sculptures, a moment of codification indicates, not the newness of a particular āsana, but rather the historical presence of such a practice prior to the produced text or sculpture, which is only being recorded for the first time. texts. In part three, I turn to religious and yogic identity. Drawing upon other coeval archeological sites, textual, epigraphical, and visual materials, I make the case that some of the ascetic figures in complex yogic postures sculpted at Hampi are depictions of Nātha yogis performing the techniques of Haṭhayoga. I conclude by turning to an important literary source for the local region, the Pampāmāhātmya, to assess the context of these postures within the broader Śaiva milieu at Hampi. The importance of the visual material evidence at Hampi, when read against the textual record, suggests that various orders of yogis and ascetics were intermingling at Vijayanagara, and that this social milieu-alongside the shift in psychophysical function of āsana praxis-may partially explain why more non-seated āsanas found their way into yoga texts following the sixteenth century.

Towards a Genealogy of Āsana: From Seated to Non-seated Postures
Today, in colloquial language, the term "yoga" is virtually synonymous with the practice of āsana, or bodily postures. Yet despite the primacy of āsana in contemporary expressions of transnational, anglophone yoga (Singleton 2010), the early history of yoga in India is surprisingly sparse regarding āsana praxis. Mark Singleton's landmark monograph Yoga Body: The Origins of Modern Posture Practice (2010) has convincingly demonstrated that today's dominant forms of postural yoga are "not the outcome of a direct and unbroken lineage of haṭha yoga" but were, rather, birthed through "adaptation to new discourses of the body that resulted from India's encounter with modernity" (2010,33). Drawing upon the pioneering work of Norman Sjoman (1999), Gudrun Bühnemann (2007), and others, and surveying the textual record available to him at the time, Singleton concluded: In sum, the Indian tradition shows no evidence for the kind of posturebased practices that dominate transnational anglophone yoga today (32).
The appellations of particular āsanas often vary from text to text, as do their descriptions. Sometimes 10 within a single text, variant names of a single āsana are given (e.g., Haṭhapradīpikā 1.37). This is as true for premodern yoga texts, as it is among various contemporary yoga practice schools. Therefore, when I suggest Sanskrit names for sculpted āsanas at Hampi that have no textual or inscriptional appellation, this naming is, at best, provisional. In fact, it is clear that more than eighty-four āsana-s were practised in some traditions of Haṭha Yoga before the British arrived in India. The majority of these āsana-s were not seated poses, but complex and physically-demanding postures, some of which involved repetitive movement, breath control and the use of ropes.
[…] When the above latemediaeval yoga texts are taken into account within the broader history of Haṭha Yoga, it becomes apparent that there was a substantial increase in the number of āsana-s after the sixteenth century and, from the seventeenth century onwards, various lists of eighty-four or more āsana-s have been recorded. In contrast to this, very few āsana-s were mentioned or described in the early Haṭha texts, which can be dated from the twelfth to fifteenth centuries.
As Birch has demonstrated, the substantial increase in the number of āsanas taught in Sanskrit texts largely occurred in the seventeenth-century and thereafter. As I will demonstrate herein, the Vijayanagara temple sculptures of complex non-seated āsanas from the early sixteenth century are thus highly significant. The Hampi sculptures appear to anticipate the proliferation of āsana as evinced in the texts, while at the same time, some of the āsanas depicted at Hampi have no premodern textual referent (of Singleton himself has acknowledged the shortcomings of these now-outdated historical claims regarding 11 premodern āsana in the preface to the Serbian-language edition of the book (Singleton 2015, 4-5). He has been working to nuance this history through the recent publication of Roots of Yoga (Mallinson and Singleton 2017), and as one of the key researchers on the SOAS Haṭha Yoga Project. In what follows, and building upon Birch's study, I will gesture towards a genealogy of premodern āsana in order to historically locate the sculpted representations of yogic praxis found on the pillars at Hampi. This will by no means be an exhaustive account or history of premodern āsana, but rather will serve two heuristic and historiographical functions for the purpose of this study: 1) to provide a template for a working chronology of āsana; and 2) to highlight a notable shift in the employment and functionality of āsana-in theory and praxis-from seated to "non-seated" āsanas. 13 Āsana is commonly listed as one of several auxiliaries (aṅga) of yogic praxis featured throughout premodern yoga literature. Early usages of the word āsana in Sanskrit 14 texts refer to the physical act of sitting, or to the material seat one sits upon, often for 15 the purpose of meditation, breath-control, or visualization practices (in later tantric 16 traditions). Such meditative "seats" were utilized by ascetics (śramaṇa) across sectarian traditions in early India, including Buddhist, Jaina, Ājīvika, and Brāhmaṇical traditions, and are well-attested throughout Indian literature and visual art (Fig. 1). A case in point Admittedly, the phrase "non-seated āsana" is something of an oxymoron, if āsana is taken to mean "seat," 13 its primary meaning in Sanskrit; that is to say, a "non-seated seat." However, by the late-medieval period, the notion of āsana came to refer to virtually any physical posture, not only seats-and not only yogic ones. As Mallinson writes, "the use of the word āsana to describe any sort of physical posture appears to have become widespread by the early 14th century, when the Maithili Rasaratnākara used it (along with bandha) as a term to describe positions for sexual intercourse" (Mallinson 2011a, 776). By the turn of the first millennium, Sanskrit texts on yoga began including non-seated postures, such as mayūrāsana and kukkuṭāsana, in teachings on āsana as an auxiliary of yoga (e.g., Yogavasiṣṭha, Yogayājñavalkya, Haṭhapradīpikā, etc.).
For example, Bhagavadgītā 6.11-13. 15 One of the earliest literary references to an āsana as meditative "seat" is found in Aśvaghoṣa's 16 Buddhacarita (1st century), where the Buddha-to-be takes his cross-legged position under the bodhi tree, with firm resolve to remain seated until he has reached the final goal of nirvāṇa (Buddhacarita 12.120 The title Pātañjalayogaśāstra is adopted to reflect both sūtra and bhāṣya layers of the text. While 17 authorship of the bhāṣya has been most often ascribed (in both premodern sources and secondary scholarship) to the legendary sage Vyāsa or Vedavyāsa, some scholars, most recently Maas (2013), have challenged this view by advocating the theory of single authorship of both sūtra and bhāṣya by one Patañjali, thus rendering the bhāṣya an "auto-commentary" by the author himself. Not all scholars are in agreement with this position, however, and a discussion of divergent positions can be found in White (2014, 226-234   from later commentaries we learn that these āsanas are likely all seated postures for meditation such as padmāsana ("lotus posture"), bhadrāsana ("blessed posture"), 20 vīrāsana ("hero posture"), and svastikāsana ("auspicious posture"). They are to be employed by the yogi in order to still the body for a prolonged period of time, to assist in stilling the fluctuations of the mind (yogaś cittavṛttinirodhaḥ)-indeed, the cessative goal of Pātañjalayoga. In this way, establishing a "firm and comfortable" (sthirasukha) posture serves a functional role for cultivating single-pointed awareness, and operates within early Indian yoga systems as a foundational practice for stabilizing the body in order to control the breath (prāṇāyāma) and reign in the senses (pratyāhāra). Āsana thus enables the aspiring yogi to progress through the more subtle and refined inner auxiliaries (antaraṅga) of fixation (dhāraṇā), meditation (dhyāna), and ultimately, meditative absorption (samādhi).
This sedentary, ascetic nature of āsana was highlighted by Mircea Eliade, who applied 21 this view not only to Pātañjalayoga, but categorically to all later yoga traditions. 22 However, such an ahistorical and homogenous account of the continuity of āsana praxis in later Haṭhayogic sources is misleading and, as I will demonstrate, fails to account for the transition from seated to non-seated postures, and the re-envisioned functionality of āsana that ensued especially within medieval Haṭhayoga traditions. svastikāsanaṃ daṇḍāsanaṃ sopāśrayaṃ paryaṅkāsanaṃ krauñcaniṣadanaṃ hastiniṣadanam uṣṭraniṣadanaṃ samasaṃsthānaṃ sthiraprasrabdhir yathāsukhaṃ cety evamādi |). Interestingly, the author of the bhāṣya concludes this list of āsanas with the word ādi, the Sanskrit equivalent to "et cetera", implying that even by the fourth or fifth-century, other postures would have been known. This is a poignant reminder that lists of āsana in Sanskrit texts are not intended as exhaustive accounts of āsana "on the ground," but more likely conceived to serve as reference points for yogic praxis.
Maas has shown that the likely oldest account and description of these postures is found in the 20 Pātañjalayogasūtrabhāṣyavivaraṇa attributed to a Śaṅkara (Maas, forthcoming 2018).
Eliade viewed āsana as a "sign of transcending the human condition" through which the yogi becomes 21 like a "plant or a sacred statue" (Eliade 1958, 54).
Eliade states: "Lists and descriptions of āsanas are to be found in most of the tantric and Haṭha-yogic 22 treatises. The purpose of these meditational positions is always the same: 'absolute cessation of trouble from the pairs of opposites' (dvaṇḍvānabhighātaḥ)" (Eliade 1958, 54 While there are early textual references in the Mahābhārata and Rāmāyaṇa epics to ascetic sages and gods performing variations of non-seated postures, such as standing on one foot (ekapādasthita), raising the arms above the head (ūrdhvabāhu), and hanging upside down from a tree, these are not described in the classical texts as yogic āsanas per se, but rather as practices of tapas. However, as James Mallinson (2011a) has 23 demonstrated, these early physical acts of asceticism pre-figure many later bodily yogic technologies. Here, the ascetic asserts exacting control over his mental and bodily faculties in order to generate an internal spiritual heat (tapas), obtain boons from gods, and ultimately eliminate the accumulation of karmafreeing himself from the trappings and suffering of perpetual rebirth in saṃsāra.
Such ascetical feats are also well-known in Indian literature and art, one of the most famous examples being the relief of King Bhagīratha's penance carved at the circa seventh-century Pallava site, Mamallapuram, in coastal Tamil Nadu (Fig. 2).
By the turn of the first millennium, a new yogic orientation was underway which centered around the cultivation of the body and bodily techniques-including āsana. Around the eleventh and twelfth centuries, a corpus of Sanskrit texts teaching the techniques of Haṭhayoga (i.e., "the yoga of force") emerged, 24 systematizing and codifying this bodily-See, for example, Mahābhārata 1.81. 16,1.114.20,3.185.4,12.323.20,12.327.41,12.331.47. On 23 the relationship between descriptions of tapas and physical yoga in the Mahābhārata, see Hopkins (1901).
On the "force" of haṭhayoga, see Birch (2011).  soteriological treatises authorizing instructions for the aspiring yogi in matters of lifestyle and diet, the proper locale for praxis, a rudimentary metaphysics and theory of the body and its subtle energy channels, and overall, mapping a progressive curriculum of yogic practice, outlining the techniques said to culminate in the liberative state of samādhi (i.e., Rājayoga).
As Mallinson's work has convincingly shown, the Haṭhayoga texts represent a unique synthesis of (at least) two historical ascetico-yogic streams. The first is the older ascetic tapas traditions of munis or sages, illustrated by descriptions of yogic asceticism in the epics, and which can be traced at least as far back as the hagiographical accounts of the Buddha's forays as an ascetic in the Pali canon. The second stream is found in medieval tantric yoga traditions, in particular, Vajrayāna Buddhism (Mallinson forthcoming a) and Śaiva tantra, especially the Kaula Paścimāmnāya, or "Western Transmission," yoginī cult associated with the great siddha, Matsyendranātha (Kiss 2009;Mallinson 2011a). Many of the early Haṭhayoga treatises are attributed to Gorakṣanātha (Hindi: Gorakhnāth), the famed disciple of Matsyendranātha and alleged founder of the fledgling "Nātha sampradāya" (about which we shall say more later). Other texts are 26 attributed to a variety of authors, most of whom we know little about, and which reveal Birch has identified the earliest occurrences of the term haṭhayoga in eighth-to eleventh-century 25 exegetical tantric Buddhist literature, wherein it is spoken of as a last-resort option for the Mantramārgin whose mantras are ineffective (Birch 2011, 535). However, it is not until the second millennium that these techniques were championed in Sanskrit literature. According to Mallinson (2011a, forthcoming a), the tantric Buddhist Amṛtasiddhi (c. 11th century) is the earliest text to prescribe some of the mudrās and bandhas which would be subsequently taught in almost all Haṭhayoga texts, while the first Sanskrit text to explicitly teach a system of Haṭhayoga and call it as such is the Vaiṣṇava Dattātreyayogaśāstra (c. 12-13th century).
While the majority of secondary scholarship on Nāthas tends to identify a distinct unified "Nāthism," 26 sampradāya, or panth, organized around the central teachings and personality of Gorakṣanātha (see, for example , Briggs 1938;Das Gupta 1946;Banerjea 1962;Lorenzen and Muñoz 2011), Mallinson has challenged the historical claims of such an institutionalized sampradāya. Although its historical gurus, Matsyendranātha (c. 9th century) and his disciple Gorakṣanātha (c. 12th century), are known to have lived much earlier, according to Mallinson, "The earliest references to the Nāth ascetic order as an organized entity date to the beginning of the 17th century" (Mallinson 2011b). During the interim historical period, "there are numerous references to both ascetic and householder Nāths in texts, inscriptions, iconography, and historical reports" (Mallinson 2011b), however, according to Mallinson these refer to a loose body of charismatic individuals, teachings, and practices, not to any systematized Nātha doctrine or school. This argument, however, has been called into question by Monika Horstmann, who, reexamining the question of a Nātha order in light of unpublished vernacular Hindi-related sources, seeks to push back the emergence of an organized Nātha sampradāya to a period prior to the sixteenth century (Horstmann 2014 Pāramārthaprakāśike of the Vīraśaiva scholar, Nijaguṇa Śivayogin (c. 15-17th century). It is possible that the Haṭhapradīpikā of Svātmārāma (15th century) was composed in Andhra Pradesh (Reddy 1982, 15), possibly at or around the famous Śrīśailam temple. 29 Śrīnivāsa, the author of the Haṭharatnāvalī (17th century), resided in the Tirabhukta region of Andhra, and according to Reddy, was likely a Telugu Brahmin (Reddy 1982, 14). Furthermore, many of the late-medieval Yoga Upaniṣads were compiled in the south by followers of Śaṅkara's Advaitavedānta (Bouy 1994 Gahininātha, Nivṛttinātha, Jñāndeva (Kripananda 1989). Thus, Jñāndeva's guru's guru was a disciple of Gorakṣanātha, or to put it otherwise, Gorakṣanātha was Jñāndeva's great-grandfather-guru.
I have located over a dozen manuscripts of the Śivayogapradīpikā from various archives in South India, all 28 written in southern scripts, and I believe we can be quite confident of the text's southern origins. Despite the title of the Ānandāśrama edition (1978 [1907]), the majority of these manuscripts are titled and listed as Śivayogapradīpikā, rather than -dīpikā, and indeed, the text refers to itself as the Śivayogapradīpikā (1.2, 5.58); I am thus inclined to adopt the latter title. The author of this text, Cennasadāśivayogin, was likely a Vīraśaiva (e.g., Śivayogapradīpikā 3.61, 3.63), and it became an important yoga treatise within Vīraśaiva communities, as evidenced by a later Kannada commentary attributed to Basavārādhya, and a prose Kannada rendition and commentary, the Pāramārthaprakāśike of Nijaguṇa Śivayogin.
Personal communication with James Mallinson. February 20, 2018. 29 A corpus of 108 Upaniṣads was compiled and commented on by Upaniṣadbrahmayogin around 1750. 30 Among these are found the so-called "Yoga Upaniṣads," nine of which Bouy observed, "bear a striking resemblance with works of the Nātha traditions." These are the Nādabindu, Dhyānabindu, Yogacūḍāmaṇi, Nirvāṇa, Maṇḍalabrāhmaṇa, Śāṇḍilya, Yogaśikhā, Yogakuṇḍalī, and the Saubhāgyalakṣmī. According to Bouy, "all these nine Upaniṣads, with the possible exception of the Saubhāgyalakṣmī, were either enlarged or wholly composed […] in South India by an Advaitin" (Bouy 1994, 6 during the fifteen-seventeenth centuries-providing precedent for the yogic sculptures we will encounter at Hampi in the early sixteenth century.
In general, authors of medieval yoga texts do not appear to have been concerned with providing exhaustive accounts of yogic theory and practice, but rather were keen to 31 integrate disparate traditions and techniques of yoga and attempt to synthesize them in a coherent and systematic manner -perhaps for a broader, more public audience 32 (Birch 2013(Birch , 2015(Birch , forthcoming 2018a(Birch , 2018bMallinson 2014Mallinson , 2016. Although we cannot be certain of the material life and function of medieval yoga manuscripts, it is unlikely that they would have been used by yogis as how-to-guides, or to replace the oral instruction of a personal guru. Instead, the texts offer us historical moments of codification, synthesis, and reform, visions of what were perhaps oral renunciate traditions being recorded for the first time, or as was often the case, reinterpretations of techniques from earlier Sanskrit texts, detectable from the authors' highly intertextual borrowing of shared verses. Yet, for as much as the texts say, there is perhaps as much left unsaid. Relying solely on the textual record thus leaves us with an incomplete and partisan history of yoga's past. As I aim to show in this study, the visual record of sculpted yogis performing non-seated āsanas at Hampi may help to fill some of these lacunae, offering another type of codification rendered by Vijayanagara artists in stone. This is not to elevate the visual material over the texts, but rather is an exercise in reading them together. As we shall see, the temple sculptures do not always simply affirm the prescriptions of yogic texts, but as artistic renditions of yogis in practice, also complicate and add variance to our understanding of premodern yoga traditions. In looking to the corpus of medieval yoga texts, Birch has noted the gradual shift from seated āsanas used primarily for meditation, as in the Pātañjalayogaśāstra, to more complex non-seated āsanas, including balancing postures and inversions utilized for bodily-purification, harnessing subtle energies in the body, as well as for therapeutic aims (Birch 2013(Birch , forthcoming 2018a Yogabīja, etc., teach relatively few āsanas; and they are entirely seated postures such as padmāsana or siddhāsana. 33 Mallinson has noted that the earliest textual references to non-seated āsanas have been found in large Vaiṣṇava and Śaiva tantric compendiums written around the turn of the first millennium (Mallinson 2014). While they are not devoted exclusively to the aims of yoga, they contain important yogic teachings, which to date remain sparsely studied. The Vaiṣṇava Pāñcarātrika Vimānārcanākalpa (c. 10th century) teaches the two-armed balancing posture, mayūrāsana ("peacock posture"), while the Kubjikāmatatantra (c. 34 10th century), the "principal scripture" of the Paścimāmnāya branch of Kaula Śaivism associated with Matsyendranātha (Sanderson 2002, 1), is perhaps the first text to teach another important two-armed balancing posture, kukkuṭāsana ("cock posture"), where it is described as a tantric karaṇa. During the twelfth through fifteenth centuries, for the 35 first time in Yogaśāstras, we begin to find these same non-seated āsanas. of the techniques of medieval yoga en vogue during the middle of the second millennium (Bouy 1994;Mallinson 2011aMallinson , 2014. This important treatise came to represent a locus classicus for the Haṭhayoga tradition, evident from the number of later works and commentaries that hold it in such esteem. In his consolidation of medieval yoga 38 systems, Svātmārāma describes fifteen āsanas including the non-seated postures mayūrāsana, kukkuṭāsana, uttānakūrmaka ("upward facing tortoise"), dhanurāsana ("bow posture"), and the supine śavāsana ("corpse posture"). The Haṭhapradīpikā becomes a touchstone from which we can witness a marked shift in the sixteenth century and onwards, as new āsanas were gradually introduced in texts. For example, we find twenty-one āsanas in the Persian Baḥr al-ḥayāt, thirty-two in the Gheraṇḍasaṃhitā, and then a tradition of eighty-four āsanas displayed in the Haṭharatnāvalī and the Hindi Brajbhāṣā Jogapradīpyakā. A unique illustrated manuscript of the latter has been brought to light and studied by Bühnemann (2007). The proliferation of āsanas in texts following the fifteenth-century Haṭhapradīpikā has been argued for by Birch, who in bringing to light unpublished yoga manuscripts from the seventeenth and eighteenth centuries, reveals the rapid expansion, innovation, and creativity of āsana in the early modern period (Birch 2013; forthcoming 2018a). Birch has found in texts such as the Yogacintāmaṇi, Haṭhābhyāsapaddhati, and the Siddhāntamuktāvalī, descriptions of over one hundred āsanas, including seated, standing, balancing, and dynamic moving āsanas, and even āsanas involving the use of ropes and a wall-all before the colonial period.
In the Haṭhayoga literature, āsana continues to serve a foundational role for engaging in prāṇāyāma and other auxiliaries of yoga, however, is also associated with bodily  (1852). (Bouy 1996;Bühnemann 2007, 8;Birch 2011, 548 , stretching, pressing, twisting, bending, balancing, etc.) to cure the disorders (doṣa) and diseases (roga) of the body, develop health (ārogya), and manipulate subtle energy (kuṇḍalinī, prāṇa) within the yogi's body. I suggest that this significant shift in the psychophysical functionality and purpose of āsana opened new avenues for the anatomical potential of the body in yogic praxis. As previously noted, although the growing development of āsana traditions can be detected in Sanskrit texts following the fifteenth century, much of the proliferation of non-seated āsanas in yoga literature does not occur until the seventeenth and eighteenth centuries (Birch 2013, forthcoming 2018a). As we will see, this makes the sculpted images of yogis performing complex non-seated āsanas on the temple pillars at Vijayanagara in the early 1500s especially noteworthy.
In order better to locate yoga within the sociopolitical and historical context of late-medieval Vijayanagara, let us now turn to a brief introduction to the empire, and the importance of religion and state-sponsored temple building for the rāyas, or kings, upon whose temples the yogis are etched in stone.

Vijayanagara Temples, Religion, and Sculptural Style
The Vijayanagara Empire is named after its capital, "the city" (nagara) of "victory" (vijaya), popularly known today as Hampi, which lies along the Tuṅgabhadrā river in northern Karnataka. Perhaps the greatest empire South India has known, the Vijayanagara kings reigned from approximately 1336-1565 CE, rivaling the Deccan Sultanates and Mughal Empire of the north by creating a vibrant cosmopolitan civilization and polity. While an early historiography of Vijayanagara often framed the rise of the empire in religious ideological terms-as a Hindu kingdom that sought to assert its sovereignty vis-à-vis the oppression of an impending Muslim north-more recent scholarship has complicated and enriched the simplicity of this dichotomous religious narrative. Vijayanagara rulers not only allowed Islamic worship and the building of mosques within the capital city (Verghese 1995, 125-28) rājagurus, and their epigraphical sign-off was "Śrīvirūpākṣa," in deference to their patron deity, the "uneven-eyed" Śiva. Yet, there was a marked transition in the kingdom from Śaivism to Vaiṣṇavism under later Saḷuva and Tuḷuva rule, when the Śrīvaiṣṇavism of Rāmānuja received state-favor beginning with Sāḷuva Narasiṃha (r. 1485-1491). Furthermore, Jainism, which had for centuries been the dominant religion in the Karnataka region (although its influence had waned considerably by the Vijayanagara period), continued to receive patronage by Vijayanagara kings (Verghese 1995, 7). Indeed, to claim Vijayanagara as a uniformly Hindu polity fails to pick up these subtleties and varieties of religious difference, only briefly mentioned here. As Valerie 50 Stoker has demonstrated, the religious diversity at Vijayanagara amounted to, not simply an ethos of religious pluralism or ecumenicalism, but the cultural and economic conditions of realpolitik that also gave rise to hostile polemics and sectarian rivalry between competing groups of Śaivas and Vaiṣṇavas, as well as Mādhva Brahmins and Śrīvaiṣṇavas, all of whom were contesting for royal patronage within the capital (Stoker 2011(Stoker , 2016. The royalties for the victors of such patronage resulted in one of the enduring legacies and greatest contributions of the empire, what historians refer to as the "Vijayanagara temple style," which drew upon earlier Hoysaḷa, Kākaṭīya, and Cōḷa styles, and matured into its own by the mid-fifteenth and early-sixteenth centuries (Stein 1989, 111). The Vijayanagara rulers commissioned the building of many great temple complexes within the capital city, particularly during the reigns of the great Kṛṣṇadevarāya (r. 1509-1529) See, for example, the royal Saṅgama inscription of 1352, issued by Devarāya II, "Sultan among Hindu 47 kings" (hindurāyasuratrāṇa), in Wagoner (1996, 853).
There is some debate regarding the sectarian affiliation of the early Saṅgama kings. Many contend that 48 the early Saṅgamas were disciples of the influential Smārta Brahmin, Vidyāraṇya of the Śṛṅgeri Maṭha, while according to Verghese, "a careful study of the epigraphical and literary sources reveal that the rājagurus of the early Saṅgamas were Kāḷāmukhas" (Verghese 1995, 7). For a critical assessment of the religious orientation of the Vijayanagara rulers, see Clark (2006, 193-202).
Sanderson observes that the form "Kālāmukha" is simply the south-Indian version of the term 49 "Kālamukha" (Sanderson 2006, 151-52). Given the provenance of this study, then, unless directly quoting, I use the appellation Kālāmukha.
According to historians Catherine Asher and Cynthia Talbot, "This interpretation, which sees the 50 Vijayanagara kingdom as inspired by and imbued with a deep sense of Hindu nationalism, is clearly anachronistic-a case of projecting a present-day situation back into the past" (Asher and Talbot 2006, 64 and his half-brother and successor, Acyutarāya (r. 1529-1542). These temples functioned not only as vital centers of trade, calendrical festivals, and religious worship within the city, but also served to boast the cultural power of the empire over other neighboring chiefdoms in nearby Tamil and Andhra lands (Stein 1989, 112). Vijayanagara sculptural tradition continued to mature towards the end of the fifteenth and early sixteenth centuries, as the attention of sculptors began to focus "on the embellishment of newly built pillared halls" within the great temple complexes (Dallapicolla and Verghese 1998, 8). Artisans (śilpin) began to experiment with techniques for carving figures out of the large granite boulders native to the Hampi region. They created composite pillars "constituted by central shafts and clusters of subsidiary colonettes and piers adorned with three-dimensional" human, animal, and divine figures "emerging," as it were, out of the standing pillars, "thereby giving the impression of free-standing statues" (Dallapicolla and Verghese 1998, 8). Thousands of these three-dimensional sculpted figures can be found today within the remains of the many halls (maṇḍapa) decorating the major temple complexes at Hampi. The iconographic programs of these maṇḍapas feature a wide range of both idealized and According to Dallapicolla and Verghese, the oldest "narrative friezes" are located on the Mahānavamī 51 platform in the "royal centre" of the capital, "depicting life at court, with the king in all his might and splendour" (Dallapicolla and Verghese 1998, 26). There we see descriptive renditions of wrestling and boxing matches, hunting scenes, parading horses, music, dance (Dallapicolla and Verghese 1998, pls. 94-98), and narrative reliefs of the principle festival of Mahānavamī Vijayadaśamī, a sumptuous celebration commemorating the victory of the goddess Durgā over the evil demon Mahiṣāsura. This annual celebration became a kṣatriya ritual utilized by Vijayanagara rulers to gather and inspect the military before embarking for battle (Dallapicolla and Verghese 1998, 26-27), and left a lasting impression on foreign visitors such as the Portuguese, Domingo Paes (Sewell 1972(Sewell [1900, 262-64). descriptive elements. While for sculptures of deities and religious icons, Vijayanagara sculptors generally followed the artistic guidelines of prescriptive texts (śilpaśāstra), according to Dallapicolla and Verghese, for "the portrayal of human beings, however, they relied more on observation than on prescriptions of the texts" (Dallapicolla and Verghese 1998, 10). This is borne out by the detailed attention given to the figures' varied styles of dress, jewelry, hairstyles, accoutrements, and in the case of yogic postures, the sculptors' close attention to specific positions and physical anatomy. The composite temple pillars display granite reliefs of many types of sculpted figures: animals and birds, courtiers, soldiers, foreigners, musicians, female and male dancers, religious virtuosos, and numerous ascetics-including, as Verghese has noted, some "complicated yōgic ones" (Verghese 1995, 111). However, according to Verghese, "since they are represented in a stereotyped manner, it is almost impossible to group them according to sects and sub-sects" (Verghese 1995, 111).
In part three, I will expand on and revise Verghese's preliminary assessment to reconsider the question of sectarian identity of the ascetics performing complex yogic āsanas at Hampi. I argue that the observational and descriptive orientation of Vijayanagara artisans suggests that these figures performing yogic postures were not simply idealized or generic renderings of semi-divine beings, but rather point to contact with living yoga and ascetic traditions under the empire. As we proceed, I will highlight evidence that suggests the influence of tantric Śaiva and siddha traditions on the iconographic programs at Vijayanagara, and argue that at least some of the ascetics in question are Nātha yogis performing the techniques of Haṭhayoga. Let us now turn to the sculpted reliefs.

Locating Yoga at Vijayanagara: Yogis Etched in Stone
The first set of images is located in the Tiruvēṅgaḷanātha temple complex, one of the last major temples built before the fall of the Vijayanagara empire in 1565. Incorrectly referred to today as the Acyutarāya, or Acyutadevarāya temple, the foundational inscription reveals that the temple was installed by the emperor's brother-in-law, the "Mahāmaṇḍalēśvara" Hiriya Tirumalarāja Voḍeya, on Sunday April 26, 1534. The 52 temple is located in the "sacred centre" of the capital city, south of the Tuṅgabhadrā river, and situated below the iconic Mataṅga hill (Fig. 3). The complex runs along the north-south axis, with the garbhagṛha at its center, housing the Śrīvaiṣṇava deity, Tiruvēṅgaḷanātha (also known as Veṅkaṭeśvara)-although today no image of the deity    In the far north-west corner of the outer wall (prākāra) of the main temple is a hundredcolumned maṇḍapa featuring some of the most elaborately carved images in all of 54 Hampi. Given its multi-tiered structure and acoustical capacity, it was likely used as a performance hall for religious recitations and ceremonies (Fig. 4) During the first half of the sixteenth century, as royal allegiance shifted from Śaivism to Vaiṣṇavism, the 53 Tiruvēṅgaḷanātha cult was immensely popular and received generous patronage under Tuḷuva rule. According to Verghese, "the existence of eight temples dedicated to this deity in the city and its suburbs is revealed by epigraphical and sculptural evidence," and by the rule of Sadāśiva (r. 1542-65), Tiruvēṅgaḷanātha "had become the most important Vaishnava deity in the empire" (Verghese 1996, 189-90).
The hundred-pillared maṇḍapa was a common feature of sixteenth-century Vijayanagara temple building. 54 In addition to the one found at Tiruvēṅgaḷanātha, there are also hundred-pillared maṇḍapas at the Virūpākṣa, Viṭṭhala, and Mālyavanta Raghunātha temple complexes. For more architectural details of these and other major Vijayanagara temples and structures, see Michell and Wagoner (2001). Dallapicolla at the Tiruvēṅgaḷanātha temple   55 complex, Hampi, January, 2016. Vasundhara Filliozat similarly observes the acoustical capacity of this hundred-pillared maṇḍapa: "Good acoustics is another extraordinary quality of these maṇḍapas. If a person sings from the platform, which is at the other end of the maṇḍapa, one can hear the artist even at the entrance" (Filliozat 1985, 309).

Personal communication with Anila Verghese and Anna
This type of sectarian cross-pollination in sculpted imagery at Hampi is not unique to Vijayanagara, but 56 is common in many regions and imperial temples of southern India.
The experimental nature of Vijayanagara sculpture traditions in relation to other south Indian dynasties 57 should not be overstated, however, and requires further comparative study of southern dynastic temple sites including Cōḷa and Hoysaḷa, which is beyond the scope of this study. For an example of Vijayanagara appropriation of earlier Cāḷukya temple styles, see Wagoner (2007 does not appear to be any particular spatial sequence or measured schema to the orientation of pillar sculptures. While the majority of Hampi ascetics are featured in 58 traditional seated yogic āsanas, it is the unique reliefs of more complex non-seated āsanas that warrant attention herein. Walking up the eastern steps of the maṇḍapa, along the first row of pillars second to the north, on the middle panel facing east, is a sculpted figure performing a one-handed balancing posture (Fig. 5). The ascetic's legs are crossed and lifted in a figure-four formation. The left arm extends downwards and through the aperture of the legs, resting the weight of the body on the left hand. The right arm is extended outwards, the hand fixed in a downward-facing mudrā with the thumb and index-finger touching. The figure has large hooped earrings, and his matted hair (jaṭā) is fixed in an ornate bun at the top of his head. In between his legs, is the faint protrusion of the ascetic's cloth dhotī, who otherwise appears naked.
The first pillar of the following row depicts a more challenging one-handed balancing āsana carved on the middle panel, facing east (Fig. 6). With the left leg fastened behind the head, the figure's left arm extends down, balancing his weight on the left hand. The right leg is bent and lifted, while the right arm hooks underneath the back of the knee. In his right hand, although difficult to make out, the ascetic appears to be holding an 59 akṣamālā (likely rudrākṣa seeds, or possibly bone ornaments ) for the practice of mantra 60 recitation (japa). Again he wears a large hooped earring on the right side, another variation of the medieval matted hair bun atop the head, and an armlet or piece of jewelry wrapped around the upper right arm-all key iconographic features encountered throughout many of the yogic sculptures at Hampi. Though the same postures are found repeated at Hampi like a visual trope, sometimes within a single maṇḍapa (Fig. 7), the specificity of the bodily configurations displayed in this and other sculptures of āsana suggests that they are not simply generic or idealized yogic postures, but rather particular ones.
It is worth pausing here to note that these postures do not look particularly sthirasukha, or suitable for prolonged meditation à la Patañjali. Thus, they seem to suggest ulterior "It is unfortunate that the granite of the sculptures is no longer covered with the original plaster work and, therefore, the fine and sensitive modeling cannot now be appreciated" (Dallapicolla and Verghese 1998, 7).
A necklace of bone ornaments is one of the standard accoutrements of the antinomian Śaiva ascetic 60 order, the Kāpālika "skull-bearers," about which I shall say more below (see Törzsök 2012   here I suggest that this ascetico-yogic recycling is also displayed visually in the Vijayanagara sculptural record. Further, while they may have simply served as visual embellishments for the Vijayanagara sculptors, depictions of yogis with the mudrā or akṣamālā in-hand might also reflect a difference in practice than what can be gleaned simply from the texts. Nowhere, to my knowledge, in a medieval text of Haṭhayoga, does an author prescribe non-seated āsanas to be practiced simultaneously with hand mudrās, or while engaging in japa-which is, of course, not to say that yogis did not perform them. Mantrayoga, although featured alongside Haṭhayoga as one of the four standard yogas in medieval yoga texts, is often conspicuously absent from the Haṭha synthesis. In its place is 63 64 often the ajapā mantra, the "unpronounced" mantra of the exhalation and inhalation of For example, on the relationship between tapas and yoga in the Mahābhārata, see Hopkins (1901). The
Mallinson has argued that the earlier tapas practices of ascetics paved the way for many of the āsanas and 62 mudrās of medieval Haṭhayoga, as evidenced in the reuse and adaptation of such techniques in later yoga texts (Mallinson 2011a). A salient example of this ascetico-yogic recycling is in the eighteenth-century Jogapradīpyakā, which although the text does not call its yoga Haṭhayoga, includes many haṭha techniques within the framework of aṣṭāṅgayoga. Among its eighty-four āsanas, the Jogapradīpyakā describes the inversion, tapakāra āsana ("the ascetic's posture"), which appears to be a reformulation of the ancient "bat penance" (Pali: vaggulivata) austerity, in which the practitioner hangs by the legs upside down from a tree. Underneath Figure 6, on the south-facing lower panel of the same pillar, is a bearded ascetic standing on one leg, balancing on the toes of his extended left foot, hands pressed together in añjalimudrā (Fig. 8). His long flowing dreadlocks (jaṭā), cleverly depicted by the artisan, extend outwards like the branches of a tree. His other features are stark: the prominent mustache and beard, large nose and arched eyebrows, large pendant earrings, the loin cloth fixed around his waist, and a jeweled plate or neck ornament draped around his chest. His fixed standing posture is suggestive of the 67 ancient tapas practice of standing-on-one-leg (ekapādasthita) (Fig. 2). However, by at least the eighteenth century, yoga tradition would come to know this type of standing posture as a distinct āsana, namely vṛkṣāsana, the "tree posture." As the Gheraṇḍasaṃhitā states: And having placed the right foot on the area at the base of the left thigh, then [the yogin] stands on the ground like a tree. This they know as the Tree Posture (vṛkṣāsana). 68 The figure in the Hampi sculpture goes one further in his tapas, balancing on the tip of the toes, increasing the difficulty of this standing posture. The tree-like rendering of On the ajapā mantra, see e.g. Gorakṣaśataka 42-44 (Nowotny ed.) and Śivayogapradīpikā 2.28. While the 65 Śivayogapradīpikā describes the Mantrayogin as one who performs the one-, two-, six-, or eight-syllabled mantras (1.5), the author later equates Mantrayoga with Ajapāyoga (2.26a), and proceeds to provide a description of its technique (2.28-34). Thus, while the Śivayogapradīpikā acknowledges that some yogis practice various mantras, it advises for the higher course of Śivayoga only the repetition of the "unpronounced" ajapā mantra.
It is possible that an akṣamālā (along with a mantra) might have been used to count the length of time 66 that the practitioner would hold an āsana, although we are not aware of any textual evidence supporting this. I thank Jason Birch for this suggestion.
The jeweled neck ornament may indeed be the rucaka, one of the six insignia (mudrikāṣaṭka) of the 67 Kāpālika ascetics. The other five Kāpālika insignia are the necklace (kaṇṭhikā), earring (kuṇḍala), the crestjewel (śikhāmaṇi), ashes (bhasma), and sacred thread (yajñopavītam), see Lorenzen (1972 Perched upon the Nallamala hills, above the banks of the Kṛṣṇā river in modern-day Andhra Pradesh, about 175 miles (385km) north-east of Hampi, Śrīśailam is an important temple site that has been long-renowned as a great center and pilgrimage destination for Hindu and Buddhist siddhas and yogis. Numerous sculpted reliefs bearing their narratives and iconography (Shaw 1997;Linrothe 2006). During the Vijayanagara period, Śrīśailam was one of the most important Śaiva pilgrimage centers of the region, and was visited and patronized by Vijayanagara kings, including the great Kṛṣṇadevārāya (Shaw 2011, 237). Richard Shaw has highlighted important iconographic associations and shared features of ascetic imagery across Hampi and Śrīśailam, as well as at the Smārta Brahmin advaita maṭhas in Śṛṅgeri, Karnataka, often linked to the early rulers of the Vijayanagara empire. His comparative work yields the presence of a shared "siddha iconography"-ascetics adorned with large hooped earrings, long matted, sometimes flaming hair, waistbands, necklaces, armlets, hands in mudrās, and other yogic accessories and props such as the "yoga staff " (yogadaṇḍa) and "yoga strap" (yogapaṭṭa)-found across all three of these important Vijayanagara temple sites (Shaw 2011, 242). One particularly rich site for this type of imagery is the prākāra wall surrounding the Mallikārjuna temple at Śrīśailam, which, like the many temples at Hampi, features a dazzling array of ascetic and siddha figures, some in yogic postures.
Based on shared architectural and iconographic themes, it is highly likely that the sculpted reliefs of ascetics at Śrīśailam and Hampi were fashioned during a close period of production, and possible that its artisans would have been aware of each other, if not from a shared occupational network. 69 Along the north side of the Mallikārjuna prākāra, one encounters a seated ascetic figure with long matted hair (Fig. 9), whose features (despite the absence of the neck ornament) are so similar to the one at Hampi (Fig. 8) that one wonders whether the artisans were depicting a unique personality or ascetic figure. Indeed, many of the Śrīśailam reliefs depict narrative cycles of legendary siddhas, including widely known stories of prominent Nātha yogis like Matsyendranātha, Gorakṣanātha, and Cauraṅgī (Linrothe 2006a, 128-32). Could this dreadlocked ascetic also be a celebrated siddha or An inscription in the Mallikārjuna temple states that it was renovated in 1510-11 at the hands of the 69 "master mason," Kondoju (Shaw 1997, 162;Linrothe 2006a, 127). However, according to Shaw, "[i]t seems likely that the carvings of the thousands of panels on the prakara was [sic] not carried out at a single period but, perhaps, over several creative periods during the fifteenth and early sixteenth centuries" (Shaw 2011, 239). Photograph courtesy of Rob Linrothe. The sheer variety of siddha images on the prakāra walls tempts one to believe the artists knew them firsthand. The conviction with which poses, gestures, hair styles and behaviors are rendered would seem to suggest familiarity with the many siddha types that we may assume were present at the time, just as they continue to congregate there today […] familiarity with actual models on the part of the artists at Śrīśailam, which textual evidence also encourages us to assume, is not a far-fetched notion (Linrothe 2006a, 128).

SETH POWELL | Etched in
Although the historicity of many siddha figures is often beyond reach, in medieval India, they were undoubtedly celebrated as real people, often as teachers (Linrothe 2006b, 82). It is likely that artisans drew upon larger regional and even transregional siddha visual tropes, and then adapted those conventions in local ways, perhaps 71 based on first-hand contact with siddha and yoga traditions. So too, at Hampi, the variety, attention to bodily position, and specificity of the yogic āsanas-especially when paralleled with the textual evidence-suggest the familiarity of the sculptors with yogis present in or around the capital city.
Returning to Hampi, moving into the It is possible that the standing ascetic in Figure 9 is a depiction of Arjuna performing tapas, as narrated in 70 a well-known episode from the Mahābhārata. Indeed, other sculpted reliefs from south Indian temples depict Arjuna's tapas in a very similar manner of standing on one leg (e.g., at Cidambaram and other temples in Tamil Nadu). However, such sculptures always feature the iconic boar below or surrounding Arjuna, who, as the story goes, interrupted Arjuna's tapas. The boar is notably absent from the Hampi figure which thus raises doubts as to whether this is Arjuna. I am grateful to Jason Birch for bringing this possibility to my attention. One should encircle the neck with the two feet, face turned upwards, supported by the hands. May the "Lord of Snakes Posture" (phaṇīndra), which destroys all disease, always bestow happiness upon you. 74 Continuing the tour of Tiruvēṅgaḷanātha, in an open maṇḍapa on the outer northern side of the main temple shrine, third row from the west, second pillar from the north, on the bottom panel, facing east, we see another more difficult variation of the same posture (Fig. 11). Here only, the figure is balancing entirely on the left hand, while the right arm extends upward, palm facing the sky. A simple loin cloth covers his groin, while a pointed cap crowns the top of his head. While at first glance, this may appear to be yet another of the many interesting hair styles adorning sculpted ascetics at Hampi, a closer examination reveals that it is a hat; and moreover, possibly an 75 indicator of the sectarian affiliation of this particular yogi. Mughal and later Jodhpur paintings often depict Nātha yogis wearing pointed black caps (Fig. 12) similar in shape In Jois' Ashtanga Vinyasa, pressing down and balancing on both hands is the dynamic vinyāsa movement, 72 rather than a static āsana, used to transition from dvipādaśīrṣāsana to the next posture in the sequence, yoganidrāsana. See: https://www.ashtangayoga.info/practice/intermediate-series-nadi-shodhana/item/ way-out-303/. Accessed on: December 29, 2016. I am grateful to Naomi Worth for clarifying this for me.
It is important to note that I am not making a genealogical argument here that contemporary yogis were 73 practicing some sort of medieval Vijayanagara yoga, nor am I arguing that the ascetic depicted at Tiruvēṅgaḷanātha was practicing dvipādaśīrśasana per se. My point, rather, in referencing Jois' Ashtanga Vinyasa system is to show that some of the sculpted images at Hampi appear to at least anticipate some of the physical forms common to expressions of modern postural yoga. On the detailed variety of hairstyles as reflected in Vijayanagara sculpture and painting, see Kumari 75 (1995, 37-44   Nearby, in the south-western corner of the maṇḍapa, we find a pillar featuring a unique pair of balancing postures displayed side-by-side on the upper panels (Fig. 13). The image on the right (facing north) depicts a yogi performing a difficult posture from the front; while the image on the left (facing east), displays another posture from the unusual posterior view. Although difficult to confirm, given their proximity on the pillar and their similarity postural type, it is possible the artisans were depicting the same yogi performing two distinct postures, with the sense of movement from one posture to another. The image on the right (Fig. 14) is a familiar Hampi posture: the right leg is bent in front of the body, with the right arm tucked underneath the right knee. The left leg is wrapped behind the head. The yogi balances on his left hand, while holding an akṣamālā in his right hand. The image on the left (Fig. 15), however, is quite different. Displayed from behind, the yogi's left leg is bent horizontally, resting the left foot on the inner right thigh. His right leg extends upwards, wrapping around the back of the neck. The yogi leans over his left leg, pressing down and balancing his weight on both hands. This posture is strikingly similar to the modern yogic āsana, virañcyāsana ("the posture of the [sage] Virañci"). A black and white photograph (Fig. 16) depicts a young T.R.S. Sharma, a student of the influential twentieth-century yoga revivalist Tirumalai Kṛṣṇamācārya, performing virañcyāsana in front of the Mysore Palace in 1941. We know of no premodern yoga text that describes a posture named 76 virañcyāsana, or of a balancing āsana described in this fashion-making the visual depiction at Hampi particularly noteworthy. 77 Hampi ascetics in non-seated yoga postures are depicted not only at the Tiruvēṅgaḷanātha temple, however, but also at several other major temple complexes within the capital. Two inverted postures are of particular interest. The first is found in the free-standing kalyāṇamaṇḍapa in the north-eastern corner of the Viṭṭhala temple complex. Here, on the upper panel of a pillar in the far north-eastern corner of the 78 maṇḍapa is a depiction of a yogi balancing upside down, with the soles of the feet pressed together (Fig. 17). He presses his weight off the left hand, while resting his chin This photograph is also published in Singleton (2010, 187). 76 For a seated variation of virañcyāsana, see Iyengar (1994Iyengar ( [1966, pls. 386-89). 77 Michell and Wagoner note that this particular "structure is without any historical records," though 78 construction of the Viṭṭhala temple may have begun in 1505 by the first powerful Tuḷuva king, Vīra Narasiṃha. Additional structures within the Viṭṭhala complex bear inscriptions dating from 1513-1545. The kalyāṇamaṇḍapa likely falls somewhere during this period of the first half of the sixteenth century. For this, and more details of the Viṭṭhala temple's architecture, see Michell and Wagoner (2001, Vol. 1: 217-29    The second inversion is located in an outer maṇḍapa of the so-called "underground" Virūpākṣa or Prasanna temple in the "royal centre" of the city. Figure 18 depicts a wild 80 looking, long-haired yogi balancing upside down in what appears to be a type of reversed padmāsana-a strong backbend in which the legs are crossed behind and resting on the buttocks. The lines of his loin cloth are visible, indicating his ascetic stature. His hands and chin balances on a yogadaṇḍa, that rests on some sort of moundpossibly the fires of a pañcāgni-like tapas practice, or more likely, the skeletal remains 81 or pyres of a cremation ground (śmaśāna) utilized in tantric sādhana. The use of yogic props like the yogadaṇḍa as depicted in these sculptures also appears unique when read against the textual record. To my knowledge, no premodern yoga text prescribes yogadaṇḍas to aid the practice of non-seated āsanas. Yet, when read across 82 Vijayanagara temple sites, it is not particularly surprising. Numerous images of yogis and ascetics are found to be utilizing yogadaṇḍas across the pillared-carvings of temples at Hampi, Śrīśailam, and Śṛṅgeri-although they are typically supporting seated postures, rather than inverted ones .
Four more non-seated āsanas are found at the iconic Virūpākṣa temple, the history of which long predates Vijayanagara rule. However, important architectural renovations were commissioned during the reign of Kṛṣṇadevarāya (1509-1529). In the south-east 83 corner of the complex, within the south colonnade, on the second row of columns, third from the wall, on the middle south-facing panel is another type of advanced twohanded balancing posture (Fig. 19). The yogi presses his hands onto a raised platform. The left leg wraps around the extended left arm and tucks under the right, while the right leg tucks back, pressing into the back of the head. Again we see the large hooped According to Fritz and Michell, the site, which features a "15th-century gopura," was once partly buried 80 "underground" but has been fully exposed through excavation. "The core sanctuary is now empty," but an inscription reveals "that the temple was originally consecrated to Virupaksha. Considering its proximity to many of the residential structures of the royal centre, the 'underground' temple may have been used by members of the king's household" (Fritz and Michell 2003, 75).
However, that is not to say that the texts prohibit the use of daṇḍas. Omission does not always equal 82 absence. Yogaśāstras often omit details that may have been provided by a guru, and that may have varied from one tradition to another, such as sequences, breath count, time in posture, etc. I am grateful to Jason Birch for suggesting this.

SETH POWELL | Etched in
Located in the south-western corner of the Virūpākṣa enclosure is another hundredpillared maṇḍapa, attributed to Kṛṣṇadevarāya (Fritz and Michell 2003, 61). Here, are two more non-seated āsanas of particular significance. As one walks up the steps to enter the maṇḍapa, on the first row of the northern side, along the second pillar, on the upper panel facing-south, is a yogi depicted in what appears to be a variation of kukkuṭāsana ("cock posture"). His arms are pressed down through the legs, the soles of the feet 85 touching (Fig. 21). Again, we see the flame-like jaṭā hairstyle. This sculpture and the one at Śrīśailam (Linrothe 2006, 139) are perhaps two of the earliest known visual depictions of kukkuṭāsana.
Within the same maṇḍapa, towards the south-east corner, fourth row in from the southern wall, third pillar from the east, we find another non-seated āsana sculpted on the middle panel, facing east (Fig. 22). Here we find what is commonly known in  Haṭhābhyāsapaddhati appears to describe a posture in which the yogi's own body presses up off the ground to become the shape of a sofa. Unlike the modern ūrdhvadhanurāsana in which the shoulders, knees, feet and hands are all held in one parallel line (see Iyengar 1994Iyengar [1966, pls. 479-82), in Figure 22 the posture is shown with the yogi's hips and torso twisted toward the side (frontal view), the knees and feet pointing outwards to his left.
While this posture might appear at first-glance similar to other sculptural depictions of dance, acrobatics, or other physical traditions of premodern India (see, e.g., Dallapicolla and Verghese 1998, pl. 102), the familiar siddha accoutrements of the figure's garb clearly indicate that this is indeed an ascetic or yogi: the loin cloth, right-arm band, flaming jaṭā hairstyle, and necklace. As we shall see, a closer inspection of the necklace in particular reveals important sectarian information that allows us to confidently identify this figure as a Nātha yogi. This data will also permit us to reflect more broadly on the religious identity and milieu of the sculpted figures at Hampi.
The earliest textual reference and visual depiction of ūrdhvadhanurāsana I am aware of is from a circa 86 nineteenth-century illustrated Jaina manuscript entitled Yogāsanam (34).

The Presence of Nātha Yogis at Vijayanagara
While much ink has been spilt by scholars concerning the Nāthas, or Kānphaṭās (lit. "split-eared"), in the north-western, northern, and eastern parts of South Asia 89 (Dasgupta 1976;Lorenzen and Muñoz 2012), the textual, epigraphical, and sculptural records reveal that the Nāthas had a strong presence in South India in the periods prior to, during, and post-Vijayanagara, particularly in the larger Deccan and Andhra regions (Saletore 1937). Mallinson has argued that the Nāthas likely originated in the Deccan, where the "majority of early textual and epigraphic references to Matsyendra and Gorakṣa" are found (Mallinson 2011b). For example, in a late sixteenth-century illustrated manuscript of the Bāburnāma, an encampment of Nāthas features yogis sporting the siṅgī around their necks (Fig. 24). More local to Vijayanagara, in the early 1520s, the Portuguese traveler and visitor to the They are "split-eared" on account of the large thick hooped earrings modern adherents wear through the 89 cartilage of their ears, which are split open by the guru upon initiation. However, as Mallinson has observed, the piercing of the cartilage rather than the lobe of the ear appears to be a later development, and does not appear visually on Nāthas until Mughal paintings in the second half of the eighteenth century (Mallinson 2013b). Indeed, the sculpted reliefs of Nāthas at Hampi feature large earrings pierced through the lobes only. Although Figure 29 reveals Matsyendranātha with large split or plugged earrings, they are clearly pierced through the lobe, and not the cartilage as we find amongst Nātha initiates today. Thus, the characterization is a bit misleading, when Verghese states that Matsyendranātha is "occasionally shown with the slit ear-lobes and large earrings typical of the kānphaṭas" (Verghese 1995, 113).
According to Briggs, the siṅgī is often made from the horn of a deer or rhinoceros. He relates the Nātha 90 use of the horn to a legend involving king Bhartṛhari, a key disciple of Gorakṣanātha. "Once his seventy queens urged him to go hunting. While he was away he ran across a herd of seventy hinds and one stag; but was unable to overtake the stag. Finally, a hind asked the stag to allow himself to be shot, and he agreed, on certain conditions, one of which was that his horn should be used for the Yogi's whistle." For this, and descriptions of the modern use of such a horn in Nātha daily praxis, see Briggs (1989Briggs ( [1938, 11-12).
According to Mallinson (  Hundred-pillared maṇḍapa, Tiruvēṅgaḷanātha temple complex (author's photograph). By the middle of the second millennium, there is literary and visual evidence to suggest that a Nātha order was taking shape in South India. The fifteenth-century Telugu Navanāthacaritramu ("Deeds of the Nine Nāthas") by Gauraṇa, a Vīraśaiva scholar based at Śrīśailam in Andhra, is the first text to provide a list of nine Nāthas, including important hagiographies of seminal Nātha figures like Matysendranātha and Gorakṣanātha, and regionalized tales of their yogic feats and attainments (Reddy 2016). By the early sixteenth century, at Hampi, we begin to see the development of a unique south-Indian navanātha iconographic program, sculpted visually across the Vijayanagara temple pillars: distinct ascetic figures seated on animal vehicles (vāhana).
In the hundred-pillared maṇḍapa at Tiruvēṅgaḷanātha, for example, a Śaiva ascetic is seated atop a large scorpion blowing a long horn (Fig. 30), and another on a tortoise. At the Virūpākṣa temple, an ascetic rides atop a strange "cross between a boar and a mouse" (Dallapicolla and Verghese 1998, 80). Most prominent, however, are more than one hundred images of Matysendranātha seated upon his iconic fish vāhana, found 95 within the maṇḍapas of almost every major temple complex at . 96 Several of these images also include the Nātha siṅgī draped around his neck (Fig. 27). While these ascetics sculpted on animals are not grouped collectively in any particular fashion, and are scattered individually across the many temples at Hampi, by the late sixteenth and seventeenth centuries, a more fully-fledged and consolidated navanātha iconographic program can be detected in the surrounding region. Verghese has drawn 97 attention to a visual panel of Nāthas at the Someśvara temple at Ulsoor, a suburb of modern Bangalore (Verghese 2000). Here, in this post-Vijayanagara, Nayaka representation, are nine Nāthas sitting prominently on their animal vāhanas. In addition to the fish, scorpion, boar, and tortoise witnessed at Hampi, at Someśvara, Nāthas are also found seated on a bear, antelope, makara (sea creature), snake, and lion ( Fig. 31). A very similar navanātha visual program is found nearby at the Jalakateśvara On the myth cycles relating Matsyendranātha and fish, see Briggs (1989Briggs ( [1938, 231-32); Dasgupta (1976, 95 382-84); White (1996, 222-229      Although the bodily configuration prescribed in the text differs from the visual depiction of Figure 28 at Hampi, it is plausible that this eponymous seated twist was 101 inspired by such sculptural traditions of Matsyendranātha. Before his section on āsana, Svātmārāma informs readers that some of the postures he will disclose are known by I am grateful to Anila Verghese for directing me to both of these temples, and drawing my attention to 98 the navanātha visual programs. This is not to say, however, that there was a singular unified or monolithic Nātha sampradāya. Different 99 lists of nine Nāthas continued to appear in the texts, and different visual configurations of Nāthas on animal vāhanas appeared in sculpture, demonstrating the fluidity and changing nature of Nātha identity and representation in the premodern period. In the opening chapter of the Haṭhapradīpikā (1.9), Svātmārāma places himself in a genealogy of mahāsiddhas, tracing the practice of Haṭhayoga back to Ādinātha (i.e. Śiva), Matsyendranātha, Gorakṣanātha, Cauraṅgī and Allama Prabhu, among others. Mallinson has argued that the Haṭhapradīpikā should thus be conceived as a siddha yoga text, rather than an explicitly Nātha one, for the cast of mahāsiddhas in Svātmārāma's list are certainly not all Nātha yogis, and represent more of a disparate group of siddhas than a singular tradition or sampradāya (Mallinson 2014, 226: n. 2). In particular, and relevant to Vijayanagara, is the curious inclusion of Allama Prabhu, the celebrated Vīraśaiva yogi-saint of Karnataka. Hagiographical accounts in Kannada depict polemical tales between Allama and Gorakṣanātha, with the former making fun of the yogic feats and getting the best of the latter. However, despite Gorakṣanātha's denigration in the 102 local imaginaire, his purported guru appears to have been immensely popular in the region. The ubiquitous presence of Matsyendranātha imagery across nearly all Vijayanagara temples  attests to the strong influence of this tantric siddha in South India, and to the yogi orders who invoked him.

The Glory of Pampā: Śaiva Ascetics and Yogis at Hampi
Regarding the identity of sculpted yogi figures at Hampi, as previously noted, Vijayanagara art historian Verghese remarks, "since they are represented in a stereotyped manner, it is almost impossible to group them according to sects and subsects" (Verghese 1995, 111). However, in a later publication, she does accord them the category of Śaiva ascetics (Dallapicolla and Verghese 1998, 81). Here Dallapicolla and Verghese are drawing on the Pampāmāhātmya, an important literary source that attests to the presence of Śaiva ascetics at Vijayanagara. The Pampāmāhātmya, the "Glory of Pampā" (also known as the Hemakūṭakhaṇḍa of the Skandapurāṇa), is a local sthalapurāṇa Portion" (madhyamabhāga) of the Pampāmāhātmya, there is a revealing account of Śaiva religious praxis at Hampi. Having descended from the celestial path, a group of seven sages arrive at Pampākṣetra (i.e., Hampi) and bathe in the holy Tuṅgabhadrā river. After performing the appropriate religious rites, they enter the sacred area of Pampā through See the Śūnyasaṃpādane Upadeśa 21.

102
For a general summary of the extant manuscripts, contents, and the challenges of dating the 103 Pampāmāhātmya, see Das (2006) and Evensen (2007, 285-313 accord with the sculptures of ascetics and yogis on the temple pillars at Hampi. When read alongside the extant sculptural record, this regional account in the Pampāmāhātmya seems to suggest a rich milieu of Śaiva ascetics and yogis intermingling at Vijayanagara.
Of the several Śaiva orders mentioned in this passage, however, the Nāthas are notably absent-though the text does seem to distinguish a particular group of Śivayogins. While the Pampāmāhātmya, like most māhātmya and sthalapurāṇa literature, which is based on multiple layers of redaction, is challenging to date, the omission of Nāthas 107 as a particular Śaiva order might be suggestive of an earlier date of composition; or it may point to the Nāthas' early lack of independence from other Śaiva movements of the medieval period in the Hampi and larger Vijayanagara regions.
Verghese acknowledges that many of the ascetics sculpted at Hampi were, indeed, likely Nāthas. Noting the abundance of Matsyendranātha images, she suggests that "the nāthas were probably to be found in the city" (Verghese 1995, 113). However, she cautions that in terms of the yogic sculptures, based on a lack of epigraphical and literary evidence, it is not possible "to specifically identify any of the yōgis depicted in the reliefs as nāthas" (Verghese 1995, 113). Yet, the visual and material evidence of sectarian markers such as the siṅgī, pointed cap, and other contemporaneous visual, material, and textual evidence of Nāthas in the region suggest otherwise. Moreover, the appearance of such iconic sectarian Nātha-markers in the early 1500s, along with the cultivation of a regionalized navanātha iconographic program, suggest the growing independence and development of a specific Nātha tradition or sampradāya, even if, as Mallinson contends, it was not yet unified in the institutional sense we find today.
The sculpted images at Hampi further nuance our understanding of the development of a south Indian Nātha tradition, as well as the techniques and traditions of Haṭhayogain this case, the development of non-seated yogic āsanas. Assessing the available historiographical materials at Vijayanagara, I believe we can identify many of these yogic figures generally as tantric siddhas, and be confident that at least some of these were depictions of Nātha yogis practicing āsana. That is not to suggest, however, that all of the yogis carved in complex non-seated āsanas were Nāthas, nor should they be taken as representing a singular sectarian order or group of yogis. Further critical study of the Anthony Evensen (2007, 285) notes the considerable range of opinions regarding the date of the 107 Pampāmāhātmya, from as early as the ninth to tenth centuries to as late as the sixteenth century. Though Evensen provides compelling topographical evidence for an earlier date of composition, he cautiously concludes: "We can only say that the text was probably written before the construction of the Raghunātha temple on Mālyavanta hill which itself is undated but certainly in existence by 1559" (Evensen 2007, 310).