278 TO THE RUINS OF DANDAN-UILIQ [CHAP. XVII.
running as usual from N.W. to S.E., we crossed in the evening a belt of ground where dead trees were seen emerging from heavy sand. Shrivelled and bleached as they appeared, Turdi and the men could recognise among them trunks of the ` Terek ' or poplar, the willow and other planted trees, unmistakable proofs that we had reached the area of ancient cultivation. About one and a half miles further to the S.E. we came upon stretches of bare loess with an extensive line of hollows, curiously resembling a dry river course, yet undoubtedly only a result of wind erosion. In one of these steep-banked hollows we succeeded in digging a well, and thus saved ourselves a search in the dark for the spot which alone, according to Turdi's knowledge, offered water in the immediate vicinity of the ruins. On the following morning, the 18th of December, after turning a great Dawan, Turdi guided us to this spot, and a couple of miles further south I found myself amidst the ruined houses which mark the site of Dandan-Uiliq.
Scattered in small isolated groups over an area which my subsequent survey showed to extend for about one and a half miles from north to south with a width of three-quarters of a mile, there rose from among the low dunes the remains of buildings modest in size, but of manifest antiquity. Where the sand had blown away, the walls constructed throughout of a timber framework with thick layers of plaster were exposed to view, broken down to within a few feet from the ground. Elsewhere in places covered by low dunes the walls could still be made out by the rows of wooden posts rising from the sand. All structural remains left exposed showed signs of having been " explored " by treasure-seekers, and the marks of the damage done by their operations were only too evident. Yet even thus the ruins, on a first hurried inspection, furnished unmistakable proofs of their character and approximate date. In the remains of frescoes which had once adorned the much-injured walls in some of the larger rooms, I could easily recognise representations of Buddhas and Bodhisattvas. These plainly indicated that I was standing in the ruins of Buddhist places of worship. Peculiarities in the style of the frescoes seemed to mark the last