The caves of this group are not in themselves of any great interest; but the locality in which they are situated was one of great importance in early Buddhist times. It was, in fact, so far as we at present know, the only place in Southern India where the Buddhists had any important establishments, or, at all events, no Buddhist remains have been found south of Kalinga, except those in this neighbourhood. This was probably owing to the fact, that it was from some port in the vicinity of the mouth of the Krishnâ and Godaveri that Java and Cambodia were colonised by Buddhists, and we know from the classical authorities that it was hence that communication was kept up between India and the Golden Chersonese at Thatun and Martaban. If no other evidence were available the existence of the Amrâvatî Tope within a few miles of Bêjwârâ is quite sufficient to prove how numerous and wealthy the Buddhist community must have, been in the fourth and fifth century. While the account given of it by Hiuen Thsang in the seventh shows how much of its previous importance, in Buddhist eyes, it retained even then.

Under these circumstances we might well expect that besides the Amrâvatî Tope, other remains might still be found there, and they probably will be when looked for. This however, has not hitherto been the case. The knowledge we do possess may be said to have been acquired almost accidentally, no thorough or scientific survey of the country having yet been attempted.

Bêjwârâ was the capital of the country of Dhanakacheka when Hiuen Thsang visited the place in 637 A.D., and he describes two great Buddhist establishments as existing in its immediate neighbourhood. One, the Purvaśila Sangarâma, as situated on a hill to the east of the city, where its remains can still be traced. To the westward of the city he describes the Avaraśila monastery, in his eyes a far more important establishment, and by which there seems little doubt he intended to designate the Amrâvatî Tope, situated on the opposite bank of the river, about 17 miles higher up. This was first explored by General Mackenzie in 1817-21, afterwards by Sir Walter Elliot, and recently by Mr. Sewell of the Madras Civil Service, and the results of their labours, except of the last named, are described in the second part, and last 56 of the plates of my Tree and Serpent Worship. Though it may not have been the most sacred, it certainly is, in an artistic point of view, the most important Buddhist monument that has yet been discovered in India, and is quite unique in the part of the country where it is situated. Its magnificence and the length of time it must have taken to execute its sculptures, prove that for a long period the Buddhists must not only have been all powerful in this part of India, but also the possessors of immense wealth, and it is consequently probable that other remains of the same class may still be found, and more especially that contemporary caves may still exist in the sides of the hills in its neighbourhood. Those that have hitherto been discovered, hardly answer to the expectations thus raised, while such as have been described belong to a much more, modern age, and to another religion. It will, consequently, only be when some contemporary series of caves is discovered that we can expect to find anything that is worthy to be classed with the sculptures of the Amrâvatî Tope.1

The front of the lower storey extends about 90 feet in length, and the excavation has been carried inward to various depths, leaving portions of three rows of massive square stone pillars partially hewn out. On this façade was carved an inscription in one line in the Vengî character “of about the seventh or eighth century”.

 The second floor is of much greater area, and has originally consisted of four separate apartments; a door has been broken through the dividing wall between the third and fourth, thus throwing them practically into one apartment. The façades of these four apartments represent—if my theory of the design is correct—the four fronts that would have been found in the second storey of a structural Vihara, though in that case they would have surrounded only one hall instead of four, as is the case here. The south or left side hall is about 19 feet square, the roof being supported by two plain pillars in front and two inside, all with heavy bracket capitals. At the back is a shrine cell, 10 feet square, with a vêdi or altar in the centre, and a runnel for water round it, for the conveyance of which to the outside a small channel was cut under the middle of the threshold. The front of this hall is ascended to by eight steps from a platform 10 feet broad. in front of it.

Outside is a cell in the left end of the platform, 6 feet by 4, and behind it a still smaller one, measuring only 3 feet by 2. On the rock above is a frieze of elephants and lions.

The façade above the hall has a frieze of geese; above this is a heavy projecting member, having the Chaitya-window ornament; and above this a row of five protuberances too weatherworn to be recognisable; and over this, again, is a carefully carved diaper pattern on a flat band. On the rock on the north side of the platform is a long inscription, in Telugu, of the thirteenth century of the Śaka era, recording large donations to the temple. Thus showing that it was still considered sacred in the fourteenth century after Christ. Long after Buddhism had entirely disappeared from India.

To the right of this, and projecting about 10 feet further forward, is the principal or central hall of the whole, 29 ft. 9 in. wide by 31 ft. deep, and varying in height from 7 ft. 3 in. to 8 ft. The roof is supported on sixteen square pillars chamfered in the middle of the shafts, arranged in four parallel rows, with pilasters in line with each row, which are advanced from 2 to 3 feet into the cave. At the back is a shrine, about 13 feet square, with an empty vêdi or place for an image against the back wall, as in t.he Râvaṇa-kâ-Khâi at Elurâ. On each side the shrine door are two standing figures cut in niches, one of them being Nârasiñha or the man-lion avatâra of Vishṇu. Over the head of the door is a roll ornament or toraṇa held by a pair of makaras, or conventional Saurians, and carrying some object in the centre which rests on the back of an animal. On the left side of the hall, at the back, is a deep niche containing a figure of Gaṇêśa, “which, like the others, has been heavily covered with plaster.”

The four pillars in the back row are much weatherworn, and some of them are broken away. They have been sculptured with arabesques and lotuses, and on one a group of a man and his wife with a female attendant. The pillars in the next and front rows are almost entirely destroyed also. The bases and capitals of the second row are covered with lotuses, animal and human figures,  &c., one group containing a figure of Mâruti or Hanuman. Outside, on a portion of the rock face, is an elephant, with a man supporting its trunk.

The third apartment has originally consisted of two rooms, that on the left measuring 19 ft. 9 in. wide by 17 ft. 7 in. deep, and its roof supported by four pillars bearing arabesque and lotus ornaments. At the back is a cell, 11 feet square, with a pedestal for the image. The other room was 17 feet deep by 13 wide, and has also a shrine, 4 feet square, with dwârpâlas or doorkeepers at the entrance to it. On the west wall is a sculpture (perhaps of Vishṇu in Vaikuntha) in which the principal male figure is seated on a couch with his wives and attendants, and with musical performers represented in front. The four pillars of this room have also arabesque and lotus ornaments on their capitals.

A stair in the left side of the large hall leads up to the third storey, and lands in a great hall, 52 ft. 9 in. long by 30 ft. 3 in. deep, including the verandah, which is arranged on the same plan as the Bâdâmi caves. First there is a long verandah, with six pillars and two pilasters in front; then in the back of the verandah, separating it from the hall, are four pillars in the middle, and a wall at each end extending the length of the opening between a pair of pillars, and carved in front with a dwârpâla. The hall itself, about 8 feet high, has two rows of six pillars each from end to end. There is no shrine in the back wall, but a cell, 12 ft. 9 in. square, in the left end. The pillars that support the hall are square masses, the corners of the middle section of each having been chamfered off so as to make that portion of each octagonal. On the front sides of the upper portions of each have been sculptured the avatâras of Vishṇu, and other figures; the lower portions bearing elephants and siṅhas or lions. At the left end of the back wall, and partly on the return of the end wall, is a figure of Vishṇu, as represented in the left end of the great cave at Bâdâmi, seated on the body of the serpent Ânanta, while the hoods of the snake overshadow his head. He is four armed, holding the śankha and chakra emblems in his hands, and is attended by Lakshmî. At the sides were thirteen figures, each about 2 feet high, listening to his discourse or worshipping him, but two of them are broken away. The local Brahmans call it “Vishṇu and the Ṛishis.” In the right end wall of this hall has been cut a gigantic recumbent figure of Nârâyaṇa, 17 feet long, resting on Śesha, the great serpent, whose seven hoods canopy his head (woodcut 24). At his feet are two colossal figures, 8 feet high, and above and below the extended arm of Vishṇu are attendant figures, with Brahma seated on the lotus that springs from Vishṇu’s navel.

In front of the verandah is a platform, 48 feet long by 19 feet broad, forming part of the roof of the storey below. On the northern half sits a fat male figure similar to what is found on some of the roofs of Kailâsa, and on the hall in front of the Dâśa Avatâra at Elurâ; on each side of him is a lion. On the southern half have been similar figures, but only the bases remain.

The upper storey is reached by a series of steps in the rock at the left or south side. It represents the circular or domical termination which crowns every square pyramidal temple, in the Dravidian style of architecture, in the south of India, without a single exception, so far as I know. Here it is of course flattened out to meet the exigencies of rock construction, but all its features are easily recognisable, and are identical with those found elsewhere. It stands on a plain platform over the roof of the verandah of the third storey with three circular cells or shrines in the back wall with a bench round each. They are apparently unfinished, but their existence here is interesting, as showing that the upper storey or domical part of these Viharas was intended to be inhabited. As it happens that at Mahâvallipur they are solid we have no other absolute proof that this was the case.

“Along the base and sides of this hill,” according to Mr. Boswell2, “there are remains of a considerable number of rock-caves and temples, evidently of Buddhist origin.” “There is a rock-temple in two storeys close to the village, which has recently been utilised as a granary.” “In various places the figures of elephants and other animals in the Buddhist style of representation3 are to be seen depicted. At one place there is a Mantapam or porch cut out of the rock and supported by stone pillars, more solitary cells, and lastly a rock temple (that of Undavilli) in four storeys of considerable proportions.”

Among these it may hereafter be possible for someone thoroughly familiar with the details of Buddhist architecture to identify the grande caverne in which, according to the traditions reported by the Chinese pilgrim Hiuen Thsang, Bhâvaviveka resided awaiting the coming of Maitreya Buddha to dissipate his doubts.4

There is, however, nothing about this Undavilli cave that could have been considered as old in Hiuen Thsang's time, and there is no form or feature about it that could at any time be ascribed to the Buddhists, while from the nature of its plan, and its being constructed in the rock, it is impossible that all the Buddhist details—if they ever existed—could have been so altered and obliterated as to be no longer recognisable. We may say we now know exactly what the Rock-cut Architecture of the Buddhists was during the seventh and eighth centuries to which this cave certainly belongs, and it was not like this. At the same time, if the date assigned to the Raths at Mahâvallipur, to be described in the next chapter, is correct, we may feel equal confidence in asserting that we know what the style was, which the Hindus adopted in the south of India, at the time when these caves were excavated, and it is as nearly as may be identical with what we find here. Everything about this cave is Hindu, and belongs to that religion, and is comparatively modern—almost certainly after Hiuen Thsang's time. It is, in fact, like the Kailâsa at Elurâ, only another instance of the manner in which the Hindus about the eighth century appropriated Buddhist sites, and superseded their rock-cut temples by others belonging to their own form of faith. They, however, differ so essentially in many important particulars, that with a little familiarity, it seems impossible to mistake the one for the other. If this is so, it is clear that this Undavilli cave never could have belonged to the Buddhists. It is as essentially Brahmanical as any of the caves belonging to that sect at Badâmi or Elurâ, of about the same age, though by a  curious inversion of the usual routine, its forms are as certainly copied from those of Buddhist viharas, like the raths at Mahâvallipur, to be described in the next chapter. Proving as clearly as can well be done, that at the age when they were excavated, the Brahmins in the south of India had no original style of their own, and were consequently forced to borrow one from their rivals.

  • 1. In a paper read to the Royal Asiatic Society on the 17th of November last, Mr. Sewell adheres to the opinion he expressed in his original report to the Madras Government, that the Avaraśila Sangarâma of Hiuen Thsang was not identical with the Amrâvatî Tope, but was a “rock cut” vihara situated on the side of a hill immediately overhanging the city of Bêjwârâ. He admits that there are no remains of any structural buildings on that hill, which could have belonged to ancient times, and no trace of the “caverns” mentioned by the pilgrim. All he contends for is that there are platforms cut here and there in “the rock” on which he thinks the buildings of the monastery may have been erected.

    Although it may fairly be admitted that the language of Hiuen Thsang may bear the interpretation Mr. Sewell puts upon it, it is so deficient in precision that it may with equal fairness be argued that the expression which he considers descriptive of the monastery in reality applies to the road. The “Via Sacra”, with its statues and rest places, which its founder constructed to lead from the city to the sacred spot. As the case now stands, we have before us the substantial fact of the existence of the Amrâvatî Tope, which from our knowledge of the sculptures found in the Gandhara monasteries we know was “adorned with all the art of the palaces of Bactria,” and very similar in style to them. On the other hand, we have only a hill side which has in some places been cut down to afford platforms for buildings, but of what form and of what age we have no suggestion. Under these circumstances, and with the knowledge we now possess of Buddhist cave architecture, it is probably safe to assert, that no such combination as Mr. Sewell suggests, of rock cut with structural buildings exists in India, and till some such are discovered I must be excused if I decline to register these “platforms” among the “Cave Temples of India,” or to believe that Hiuen Thsang did not mention Amrâvatî Tope under the designation of the Avaraśila Sangharama.[/fn.]

    The principal cave that has yet been discovered in this neighbourhood is situated in a small isolated hill, about a mile from the town of Bejwâḍâ (the Bejwara or Bezwara of the maps), and is a four or rather five storeyed Vaishṇava temple, dedicated to Anantasena or Nârayaṇa. It has been suspected of having been originally excavated as a Buddhist Vihara; but there is certainly no sufficient evidence ~to justify such a supposition. It is entirely Brahmanical in all its arrangements, and very similar to the contemporary caves belonging to that religion at Bâdâmi or Elurâ, and can from the character of its sculptures hardly date further back that the 7th or 8th century of our era. It probably should be attributed to some of the Châlukya kings of Vengi, who like the elder branch of that family ruling at Bâdâmi, and later at Kalyaṇa, were worshippers of Vishṇu.

    The great interest of this cave for our present purposes, lies in its enabling us to carry one step further back our researches into the external appearance of the structural Buddhist Viharas, which have disappeared from the land. In describiug the Râni kâ Nûr, at Udayagiri (ante, p. 78) it was pointed out that the upper storey there, and in the Vaikuntha cave, were set back, not so much from constructional motives, as in imitation of the forms of the structural Viharas of the period. Here we have the same system carried out through four—possibly five—different storeys. It is true the exact section of the cave may, to some extent, have been adapted to the natural slope of the hill, but it hardly seems doubtful that the successive terraces are adaptations to rock forms of the platforms which formed essential features of pyramidal Viharas of the Buddhists, and which became afterwards the fundamental idea of the Dravidian style of architecture, in the hands of the Brahmans of the south.

    As already mentioned, the Undavilli cave is four storeys in height one above the other, but there is a fifth storey in front, shown in the view, woodcut No. 22, to the right, a little detached, but which may have been intended to be connected with and made part of the original design. The lowest of the four connected storeys is so entirely unfinished, and we cannot even guess what form it was ultimately intended to take, and how far it might be extended towards a lower one still, which certainly was commenced to the right, and may have been intended to extend across the whole front.

    When describing the Râni kâ Nûr at Udayagiri, it was suggested that the three sides of the court were really intended to represent the three sides of a pyramidal Vihara turned inside out. If this cave at Undavilli is carefully examined, it seems almost certain that it equally represents three sides of a similar building, its centre being three intercolumniations in width. The sides on the second storey having, or being intended to have, five, which was a greater number than it was possible to give to the centre from its situation, flattened out on the rock. In the third storey they were all reduced to three intercolumniations, and the uppermost storey of all was only the dome which all the Viharas had, flattened out. These storeys in a structural Vihara would be in wood. The lowest only, if I am correct, in stone, and consequently more solid, and not admitting of the same minute sub-divisions. To all these points we shall have occasion to revert presently when describing the Mahâvallipur Raths, but this cave is almost equally interesting, as a copy of a pre-existing form of building, but not being carved out of an isolated block, it is flattened out into a façade, which is not at first sight so obviously a copy of a Vihara as they are. Notwithstanding this, however, it seems hardly to admit of any doubt, that though so essentially Brahmanical in its dedication, this cave is intended for as literal a copy as could well be made, in the rock, of one of the Buddhist Viharas that must have abounded in the neighbourhood at the time it was executed. Even if we did not know from Hiuen Thsang's account how essentially Bejwara was a Buddhist colony in the seventh century, the ruins at Amrâvatî would be quite sufficient to show that every form of Buddhist architecture, in all probability, existed on the spot at the time it was excavated, and, as we gather from the result, were the only models the Hindus, at that time, had to copy, when designing structures for their own intruding faith.

    To these points we shall revert presently, but meanwhile to finish our description of this cave the following particulars based upon Mr. Sewell's plans and report.Report by Mr. R. Sewell, M.C.S., iissued by Governmeut of Madras, 1st Nov.1878. No. 1620, Pub. Dep., on which and the plans prepared by Mr. Peters, together with the notes of Sir Walter Elliot (Ind. Ant., vol. v. p. 80), this account is biased.

  • 2. Report to the Madras Government, 1870.
  • 3. It is difficult to say what the “Buddhist style of representation” of an elephant really is. There is a large bas-relief of an elephant at Ajaṇṭâ. and two others at Kuḍâ. In Buddhist caves, and many smaller ones on friezes; in the Hindu Kailâsa at Elurâ, there are many in alto-rilievo, and two free standing; there are four or five free standing ones at Ambâ, a bas-relief at Karusâ, and there was a colossal free standing one at Elephanta, all Brahmanical; one free standing one andd several in bas-relief at Mahâvallipur; and there is a free standing one and many heads, &c. in the Jaina temples at Elurâ, but no antiquary can show that each sect had its “style” of representing elephants. The carving of all figures varies more or less with the age in which they were executed, but “elephants” less than almost any other figure, and usually they are better carved than any other animal.
  • 4. Mémoires sur les Cont. Occid., tom. ii. p. 110. It is to be remarked that Hiuen Thsang says he “rested in the palace of the Asuras,” not in a Buddhist temple.