Ellora is a kaleidoscopic expression of religious architecture on a monumental scale its mile long basaltic scarp punctuated by thirty-four major rock-cut temples. Within these sanctuaries carved into the hills northwest of Aurangabad, we encounter hundreds of larger than life size images of gods of three major religions -Saivism, Buddhism, and Jainism. After one visit, or even after years of study, this is too much to absorb at once. Each visit reveals new caves are being discovered and paintings revealed our understanding of Ellora continues to grow and change.How can we create a structure for these experiences in order to reveal Ellora's manifold meanings? Various studies have described the context layout and iconography of its Brahmanical temples but less attention has been paid to explicating the place of its twelve Buddhist caves in the wider spheres of artistic and religious movements of the early medieval period. The present work will explore the historical artistic and doctrinal dimensions of its Buddhist period focusing on the mandalas that provided interative schemes for the temples.Last of the great rock-cut sites of the Western Deccan, Ellora's Buddhist cave temples were excavated during the seventh and early eighth centuries when there was a surge of activity at Buddhist centers throughout India and beyond to Asia. Thses twelve temples are like a museum preserving in situ a visual record of the early development of tantric Buddhist art from the relatively simple from seen in Cave 6 (ca. 600 C. E.) to the splendid and complex perfection of the latest Cave 12 (ca, 730). The artisans and monks who created Ellora's caves worked within the centuries old tradition of Buddhist rock cut architecture but at every stage new ideas in particular innovations in iconography illustrate the creative integration of old and new.Its importance resides in three related features. First its Buddhist sculpture includes many images new or unique in the seventh and early eighth centuries. Second because it is a rock-cut site in situ images are preserved in a programmatic context lost at it built sites which were more easily disturbed or even plundered. Third mandalas carved both in relief and expanded into the large-scale three dimension shrine programs of the earliest and lasted caves presents us with surprisingly early but still clear- cut evidence for organizational schemes of the images. The mandalawhose basic meaning can refer to anything round or circular has in religious contexts an extended meaning as an arrangement -condensed or quite complex -of patterns or pictures of deities used in rituals to guide worship or meditation. As chapter 1 will explain in greater detail, mandalas were a diagnostic and important component of the teachings encompassed in the general development of tantric Buddhism; they were carried for example by tantric teachers from India to Ladakh, and by early Shingon teachers from China to Japan and were displayed in temple sanctums. This use of mandalas has long been recognized in later tantric iconographical texts and at Buddhist sites of leter periods but is encountered at Ellora at a relatively time. The iconography of the sculptures is in some cases comparable to that of images found at contemporary or stylistic relationships between Ellora and sites in the traditional centers of Buddhism has deterred scholars from discussing Ellora in the broader context of Buddhist art in India.This, essentially an art historical study is complicated by a common methodological problem: many of its images have no clear relationship to texts that might explain their meaning or record iconographical details. While it may be assumed that to object can be made to speak as eloquently of itself as a dialogue between object text can it is also true that ideal pairings are seldom encountered in the histories of early art. This problem has been explored in greater methodological detail outside of Indian art history. For example in the fifth and sixth centuries texts may seldom appear to be the direct inspiration for a given image. Yet at the same time it can be shown that they both embody common "thought patterns." If example are chosen randomly this assumption may not appear to be legitimate. However when texts and image come from the same milieu, even on (like the early Byzantine period) in which text and image rarely come from precisely the same time and place the assumption has come validity. Moreover the more complicated the pattern of motifs repeated in image and text the greater the likelihood that a similar pattern of meanings underlies them.For Ellora less precision is possible for correspondences than might be made compared with Byzantine studies where rather precise correlations can be drawn between, for example rhetorical style and narrative art. Despite the steadily increasing number of iconographic texts being discovered edited and published there is so far no known text that confirms in a systematic way to the overall iconography of Ellora's Buddhist caves. Yet we know from the dissemination of teachings, texts and Buddhist art in the seventh through ninth centuries that new ideas were rapidly circulating as the Buddhist world expanded and it is from this more dispersed material that analogies must be drawn. Moreover given the death of precisely correlated written and visual material in a study like this one the monument itself must be examined with particular care. I would argue further that Ellora's visual "texts" can help push back the horizon for the development of a kind of Buddhism which we know from later written texts and monuments, was expounded throughout most of Asia.Art historians may be more comfortable than historians of religion with the proposition that the should suspend our expectation of finding a textual standard that would confirm and validate the art historical evidence. Yet even among south Asian art historians a tendency persists to seek a formal literary text to "prove" the meaining of what we observe in sculpture or architecture. Consider for example a recent study which offers a new interpretation of the concept of Buddhist relics based on a close reading of inscriptions from third-century Nagarjunakonda. The author notes that previous studies of this site have been hampered by art historians adherence to formal literary sources a "perfunctory preference Which is quite common in historical works on Indian Buddhism [that] can have only little relationship to what practicing Buddhists actually did. At the Very least it rather effectively impedes an adequate appraisal of other kinds of sources." Although this refers primarily to the use of various kinds of written sources the argument might be taken to have equal relevance for even more varied sources, including sculpture and architecture. This is of particular importance for places like Ellora where no literary sources- formal or overwise -are available to "explain" the structure and by inference Buddhist Practices of the site.Because Ellora is unique, previous studies have slighted its relationships to other Buddhist monuments. Yet as chapter 5 will show it is related in significant ways to other Buddhist sites on the "periphery" of the Buddhist world in the seventh and eighth centuries. These sites preserve iconographical information either specific to teachings disseminated beyond the Buddhist heartland or long since lost from those central places so much disturbed by the events of later history. Studies of certain iconographic problems and comparative studies of specific sites will explore these relationships which taken together create a picture of early tantric Buddhism as a dynamic international movement whose teachings spread rapidly throughout the Buddhist world affecting sites on the "periphery" perhaps without mediation through the heartland. In this way we can imagine- taking two traditional definitions of mandala- its unfolding to include both the diagrams which govern its sculptural programs and less literally in broader perspective the association of a loose circle of places like it, related by iconography and may infer, tantric teachings.To facilitate understanding of this sometimes-complicated material "iconographic plans" are included for each of the ten caves than contains important sculpture in addition to the larger scale illustrations provided. Images are illustrated in their approximate location in the cave but rotated so that all are legible from the same (reader's) direction. All photographs are by the author unless otherwise indicated. Sanskrit words are transliterated according to the standard orthographic system and are defined where necessary the first time they appear. Place names follow spellings in J. Schwartzberg, A Historical Atlas of South Asia (1978).