conditions produced by the vicinity of this great sheet of water and marsh-land. From these we suffered a good deal, though .December belongs to the driest portion of the Turkestan year. During the greatest part of our stay at the site an icy mist, rising from the lake on the south, enveloped ruins and camp ; together with minimum temperatures down to 42 degrees Fahrenheit below freezing-point it made work very trying. The nightly hoar-frost practically amounted to a light snowfall. It continued to cover the ground even when the sun fitfully struggled through. There was constant difficulty about getting adequate light for photographs, and it was only during the last day or two that the atmosphere cleared sufficiently to allow most of them to be taken. In them, too, the hoar-frost is conspicuous:
The excavations had not proceeded far before it became quite clear that in most of the shrines the damage caused by fire was even greater than that due to moisture. That the whole site had been subjected to a big conflagration is certain. The effects of this were most striking in the larger temples, where evidently the amount of timber inserted in the walls, etc., and of other inflammable materials had been great. Here the burning had hardened all the masses of fallen brickwork, making excavation slow and difficult. But fortunately it had at the same time reduced much of the smaller relievos in friable stucco to the condition of hard, if somewhat brittle, terra-cottas. Thus incendiary fury had helped to preserve them where the bigger sculptures were either completely smashed or else had decayed through damp beyond all hope of removal. But even in the small cellas, where there was little to feed the ;fire and where consequently the sun-dried brickwork had remained unaffected by the heat, evidence of wilful destruction was obvious.
None of the numerous Chinese coins, found mostly in places where they must have been deposited as votive gifts at image bases, etc., are of issues later than the end of the eighth century." On the other hand, finds of Uigur manuscript remains and sgraffiti indicate that the shrines still continued to be visited as places of worship at least during the earlier period of Uigur dominion (ninth—tenth century). Thus the idea suggests itself that this wholesale burning may have been caused by iconoclastic zeal during one of the early Muhammadan invasions following the conversion of the Karluk Turks in the second half of the tenth century. The fact that the temple ruins of the Khi5ra site, higher up in the valley of the Khaidu-gol, which will be described further on, have also been destroyed by fire seems to support this conclusion.
In describing the different ruins and the finds made among those which were excavated I propose to follow their topographical order, beginning at each group from the south-west. The shrine (xxv in Plate 52) at the southern end of group I, which extends along the westernmost of the previously mentioned three ridges, proved to have suffered badly from moisture. Its cella measures about 20 feet square, and has in front a large antechapel occupying the top of a terrace that is walled up against the slope of the ridge and approached over a flight of stairs. This and part of the antechapel showed signs of some previous clearing by Professor Grünwedel's party. Débris lying to a height of 7 feet and more filled the cella and the enclosing passage. Excavations were carried down to the floor in the northern passage and in part of the cella. In the latter they brought to light remains of over a dozen small painted panels, all unfortunately badly perished through damp. There were pieces, too, of a small wooden arch, Mi. xxv. 001-2 (Plate CXxvIii), decorated with relievo figures of Buddhas. Other miscellaneous minor objects are described in the List below.
10 As seen from Appendix B, out of thirty-three coins from the site, thirty-one were discovered, as it were in situ, within shrines. Only five of them belong to pre-Tang issues (with the legends JVu-chu or Huo-ch'iian); eight show the legend K'ai yiian, current throughout the Tang coinage ; six are of
the Ta-li period (A.D. 766-79), and not less than fourteen bear the nien-hao Chien-chung (A. D. 780-3). It is worth noting that the last nien-hao is the latest found on Tang coins from Khotan sites as described in the list of Ancient Khotan, i. pp. 575 sqq.