Sec. i] EXPEDITION TO THE KARA-DONG RUINS 447
heavy timber roofing, which once probably supported a second story, and was still partially preserved, added an element of risk for the labourers, as the rotten posts beneath could not be trusted to support it, and materials for propping it up could not be improvised. In spite of these difficulties my labourers managed, after two days' arduous work, to clear the three rooms on the north down to the original floor. No finds of any kind rewarded this effort. A small square structure close by, to the north-east, was roofless, but had retained parts of its plastered walls ; the construction resembled that observed in the Danddn-Uiliq ruins, reed matting and round sticks at regular intervals being used as a backing for the plaster.
To the north-east of K. i. the dunes had left a portion of the original ground-level bare,
except for the timber débris of some fallen structure (see Fig. 52, on the left). By careful levelling from this point I ascertained that the floor of the rooms occupying the top of the rampart lay 17 ft. above the ground-level in the interior of the quadrangle. The exposed position thus occupied fully explains the complete decay of these rooms. The rampart below them was proved, by a cutting which I had made in an eroded portion of it near the northeast corner, to have had a thickness of about 3o ft. at its foot. On its top a continuous layer of Toghrak and tamarisk branches about i ft. thick had served as a general foundation for the floor beams.
Amidst the sand and timber débris covering the top of the rampart and the outside
slopes fragments of coarse pottery, small broken pieces of metal objects, such as rivets and buckles, and of glass ; also shreds of felt and coarse cotton fabrics, could be picked up. Among these scanty finds, described under K. oo1, 002, two arrow-heads (K. ooi. b, K. 002. n), a bronze buckle (K. 002. a), a portion of the lip or foot of a glass vessel (K. 002. j), all shown in Plate LI, may be specially mentioned. An excellently-preserved small comb of sandal-wood, parabolic in shape, and of a pattern exactly the same as still commonly used throughout Northern India (K. 002. o), may have been an importation from that country. It is important to note that of the five Chinese copper coins found among or near the débris of the rampart, all bearing marks of long circulation, two are Wu-clan pieces, while the rest show no legend, I may conveniently mention here that, of the other nine coins picked up on eroded ground at some distance to the east and south of the ruin, one belongs to a `Vu-chu issue, the rest bearing no legend (see PI. LXXXIX, 15) or being too defaced for identification (see ibid., 6). In view of the chronological conclusion suggested by these coins, the few tiny pieces of flimsy paper without writing which I found in the débris of the quadrangle enclosure also possess, as we shall see, some antiquarian interest.
The only relatively well-preserved portion of this ruin was the large gateway (K. ii.),
21 ft. square, which was found to lead through the rampart on the eastern face of the enclosure. Fig. 53 shows it as it appeared, after clearing, seen from the east. The sand was here less high, and had left, even before excavation, some space free below the perfectly intact roof. The latter reached to within a foot of the top level of the rampart, and it was, no doubt, owing to the protection which the latter offered on either side that the timber of the walls and the massive posts supporting the roof had been preserved so well. Besides a central passage, io ft. broad, closed on the outside by a massive wooden door of two leaves, there were two side passages, 5 ft. broad. I noticed a somewhat similar arrangement in the gates of all modern Ya-mêns I visited. The folds of the door were 3 in. thick, and strengthened by massive cross-bars.
The walls dividing these passages consisted of a timber framework, massively but roughly constructed of Toghrak. The rows of sticks set vertically in it showed . that a covering of
Rampart and structures on it.
Finds of small antiques.
Coins found near ruins.
Gateway of quadrangle.
Construction of walls.