There was, however, I believe, another form of Vihara even less capable of being repeated in the rock. It was pyramidal, and is the original of all the temples of southern India. Take, for instance, a description of one mentioned both by Fa Hian and Hiouen Thsang,[160] though neither of them, it must be confessed, ever saw it, which accounts in part for some absurdities in the description:--“The building,” says Fa Hian, “has altogether five storeys. The lowest is shaped into the form of an elephant, and has 500 stone cells in it; the second is in the form of a lion, and has 400 chambers; the third is shaped like a horse, and has 300 chambers; the fourth is in the form of an ox, and has 200 chambers; and the fifth is in the shape of a dove, and has 100 chambers in it”--and the account given of it by Hiouen Thsang is practically the same.[161] At first sight this looks wild enough; but if we substitute the assertion that the several storeys were adorned with elephants, lions, horses, &c., we get a mode of decoration which began at Karli, where a great range of elephants adorn the lower storey, and was continued with variations to Hullabîd, where, as we shall see further on, all these five animals are, in the 13th century, superimposed upon one another exactly as here recounted. [Illustration: 68. 69. Square and oblong Cells from a Bas-relief at Bharhut.] The opposite woodcut (No. 66), taken from one of the raths at Mahavellipore, probably correctly represents such a structure, and I believe also the form of a great many ancient viharas in India. The diagram (No. 67) is intended to explain what probably were the internal arrangements of such a structure. As far as it can be understood from the rock-cut examples we have, the centre was occupied by halls of varying dimensions according to height, supported by wooden posts above the ground-floor, and used as the common day-rooms of the monks. The sleeping-cells (Woodcuts Nos. 68, 69) were apparently on the terraces, and may have been such as are frequently represented in the bas-reliefs at Bharhut and elsewhere. Alternately they seem to have been square and oblong, and with smaller apartments between. Of course we must not take too literally a representation of a monastery, carried out solidly in the rock for a different purpose, as an absolutely correct representation of its original. The importance, however, of this form, as explaining the peculiarities of subsequent Buddhist and Dravidian architecture, is so great that it is well worth quoting here, though this will be more evident in the sequel than it can be at present. In construction the breadth, in a structural building, would probably have been greater in proportion to the height than in this example, but that is of little consequence for our present purposes.