Sec. i] OLD REMAINS WITHIN THE BAI DISTRICT 831
where the river debouches, near which the ruins are situated. These occupy, as the sketch-plan in Pl. 45 shows, the southernmost offshoot of a spur descending along the right bank of the wide river-bed where it emerges from the Tien-shan foot-hills. At the point where this end of the spur falls off, with a precipitous scarp, towards the river winding at a distance of about 6o yards along its foot, a number of small caves have been cut into the conglomerate rock. Some eight of these face the river ; about as many more are found on either side of a small Nullah which cuts into a little plateau bearing remains of a ruined tower and an adjoining structure marked 1.
In most of the little caves the rock walls expose the coarse rubble of which they are composed,
the original plastering having probably fallen off owing to the roughness and irregularity of the surface. It is therefore impossible to make out whether these small cellas, square or rectangular in shape, served as shrines or monastic quarters. Two somewhat larger caves, situated immediately below the structure mentioned above, and shown in the plan, Pl. 44, have retained portions of the plaster covering their walls ; they show traces of badly injured paintings, which, with the circumambulatory passage in the cave marked iii, conclusively prove that they served for Buddhist worship. The walls closing the front of these cellas, and dividing them from a narrow apartment between, were constructed of rough brickwork. These caves, like the rest, showed signs of having served as shelters at one time or another after worship had ceased there.
The ruined structure, 1, had been built on the artificially enlarged top of a small ridge, about
120 feet above the river-bed. The area enclosed by the badly decayed walls of stamped clay and gravel, measures roughly 4o yards by 26. Near the south-western corner, a fragment of the wall still rose to about 9 feet ; elsewhere the enclosure was marked only by low mounds of gravel. I was told that the whole interior had been dug up in the days of ` Bedaulat's ' rule, to obtain saltpetre from the soil. This suggests that the ruin had been occupied by quarters in ancient times.
The fact that the plateau-like top in which the spur terminates to the south and south-west is
covered with low scattered grave-mounds, obviously Muhammadan, probably indicates continued local worship at this spot. The very end of the spur to the south-east is occupied by the remains of a small township, enclosed by badly decayed walls, and defended on the north and west, the only sides easily accessible, by a well-marked fosse. This measures about 4o feet across at the top, and has a depth of to feet at the north-eastern corner, where it has been cut through the rock. The wall is built of large stones from the river-bed, and shows a thickness of about 3 feet on the north side, where its remains are best traceable. Elsewhere, owing to the steepness of the scarp, along the foot of which the river washes on the south and east, this roughly built wall has almost completely disappeared. The whole of the interior of the protected area, measuring about
140 yards by loo, is covered with stone heaps, the remains of the rubble-built walls of houses. The appearance of the whole strikingly recalled that of the Kafir-kôts on the Indus, and other ruined town sites near the hills of the Indian North-West Frontier,2 though the latter are on a much larger scale.
After halting for a day at Bai, the district head-quarters, out of regard for the attentive Chinese
magistrate, I once more left the line of the high road which we had struck there, in order to visit the ` Ming-oi ' which I had been informed at Kucha was situated in the barren hills to the south of the Muz-art river. After passing for some 10 miles through well-cultivated ground irrigated from the Kapsalang river, we crossed to the right bank of the Muz-art-darya, near the little marketplace of Un-bash (Map No. 12. D. 1). It was interesting to observe that though the bed of the river was fully a mile across, yet the water actually flowing in three small channels showed a total volume of only about 58o cubic feet per second. This volume seemed very modest compared with that of
2 See e. g. Stein, Archaeological Survey Report, NW. Frontier Province and Baluchistan, 1912, pp. 15 sq.
5 O 2
Cave- shrines at Tezak- kaghe.