Sec. iii] THE RUINS OF ARA-TAM AND LAPCHUK 1153
winter occasionally to a depth of 6-8 inches, though it never lies for more than a few weeks. But at Bardash, only about 8 miles higher up the valley, snow was declared to cover the ground for several months each year and rain to be fairly frequent.
The clearing of shrine A. u yielded no better result This consisted of an outer hall about 23 feet square, half of it on a level about 3 feet higher than the rest, and of a cella, 13 feet by 14. Painted stucco fragments showed that the base running round the walls of the latter had once borne relievo images. Two square bases (Plate 47), cut from the natural clay soil which the clearing disclosed on either side of A. 11, may once have borne Stûpas of small size.
The second group of ruins, shown as A. 111 in the plan, Plate 48, consists mainly of a series of cave-shrines cut into a ridge of gravelly clay that extends for a total length of about 40o feet from
east to west, as seen in Figs. 192, 2J5. This ridge is completely isolated from the foot of the rocky
hill chain to the north. It rises to a maximum height of 6o feet at its eastern end, where two flights of stairs once led up to its top. Into its southern face there have been cut five cellas which had
their walls partly formed by the natural clay, and partly by masonry of sun-dried bricks, as seen in Figs. 257, 258, 260. Their roof, originally vaulted, is likely to: have been everywhere constructed of brickwork. Each of them appears to have contained a colossal Buddha image, which was either carved mainly from the live rock or else built up in plaster which timber pieces, set into the brickwork facing of the wall behind, helped to secure.
Of these cellas, which appear to have measured from about 20 to 25 feet square, the first, A. 111. i, at the western foot of the ridge had its walls for the most part built of bricks and is now almost
wholly destroyed. The second, as seen in Figs. 258, 26o, had its walls faced with bricks, and the
stucco plastering still survived on part of the north wall and the extant portion of the vaulting above it. The colossal image set up against the north wall had completely disappeared. The stuccoed
surface of the wall had been painted in tempera, the surviving portions showing mainly a diaper of
small seated Buddha figures, best preserved in the north-east corner (Fig. 260). The figures, about 8 inches high, were painted alternately with red and brown robes over a background of light green,
the whole closely recalling the similar diaper decoration so common on the walls of the ` Thousand Buddhas' caves ' of Tun-huang. The very shallow squinches set above the corners of the square walls retained traces of bold floral decoration, also in a style resembling that of Chien-fo-tung.
In cella A. 111. iii (Fig. 257) it was still possible to make out the base of a colossal image and the screen at its back carved from the rock. Behind this led a narrow passage intended for circum-
ambulation. The upper portion of the statue was probably built up of bricks thickly coated with
plaster ; the heavy beam which once supported the head is seen in the photograph. This also shows what little remained of the wall-painting, which in the north-east corner, on the right, was bold scroll-
work in dark red and green, probably forming part of the flame border of a large vesica. In the
squinches I could make out traces of divine figures in flowing drapery, mainly of bright green, probably representing Lokapâlas as found in corresponding positions below the vaulting of Chtien-
fo-tung cellas. Bands of floral decoration formed medallions around them just as there. Courses
of projecting mouldings below the foot of the vaulting were modelled in clay over bundles of reed-straw, just as found in shrines at Danclân-oilik and elsewhere. Cella iv showed a similar arrange-
ment of screen and passage behind the colossal image, here mainly carved from the rock and, as far
as I could judge from the shape of the surviving mass, seated. The plaster surface of the walls had completely disappeared. The same was the case also in cella v, evidently owing to the effect of rain-
water flowing down the walls from the top of the ridge. The front wall to the south had entirely fallen. From the position of the lowest moulding, still surviving below what was once the octagonal drum of the dome, it appeared that this cella was higher than the rest, as might also be inferred
1374 7 H