villagers of Ch`iao-tzti who came out to watch my proceedings, that within memory of man herdsmen grazing their ponies on the steppe northward used occasionally to seek shelter against the icy blasts of the winter in a few ruined structures evidently of later date and then still upright. There was evidence that the ruined town even at the present day saw temporary residents, people who came to collect saltpetre from the soil once occupied by buildings. Little smoke-begrimed caves at the foot of the west wall were pointed out to me as having been dug out by such humble folk for quarters.
From a tiny mud-built shrine of quite modern date, which rose in front of one of these troglodyte dwellings and on what manifestly was a worked-out rubbish-heap, an interesting small antique was brought to me by the intelligent petty officer detailed to escort my party. It is the well-carved upper arm in wood, So. oo9 (Plate xLVII), belonging to some statue evidently life-size, decorated in low relief with ornaments that represent the patterns of two garments. Their design, partly floral, distinctly recalls Gandhara motifs, including the four-petalled clematis-like flower familiar from the wood-carvings of the Niya and Lou-lan Sites. At the time when it was brought to me remains of bright colouring still adhered. That the statue to which this relic once belonged must have been of early, perhaps pre-Tang, date seems clear. Unfortunately there was nothing to indicate its original place of discovery. Even thus it suffices to show that the mounds and drift-sand within Soyang-ch'êng may yet hide objects of archaeological value, besides mere small débris of the ` Tati ' type. But systematic clearing will cost time and labour.
Outside the walled town to the north-west, and at a distance of about a thousand yards from its north-west corner (see Plate 46), two massive structures of stamped clay attracted my attention. From a distance they looked like small forts. But on reaching them, beyond the much-decayed remains of an outer enclosing wall on the north, I was puzzled at first by their peculiar nature. The well-built clay walls, fully 20 feet thick and rising to a considerable height, formed solid squares measuring about 70 feet on each side ; they showed neither a proper entrance nor any stairs or other arrangement for reaching the interior. Having made my way into the further of these strange structures by a gap which had formed at one of its corners through the parting of the walls, I vainly searched for remains to explain its purpose until I noticed some fissured planks lying loose on the sand which half filled the enclosure. Their length was just that needed for a coffin. Subsequent scraping of the sand revealed in one corner similar planks still in situ, with badly decayed remains of a skeleton between them. It was clear that these remarkably massive walls had been intended to protect an abode of the dead. They might have long defied wind-erosion, but evidently failed to keep out the sand—and intruders in search of treasure. I was unable to ascertain how the enclosure was originally entered. Inspection of the surrounding ground, where it had been left bare by drift-sand and consequently eroded, showed that the whole of it had served as a cemetery. But of the graves and their contents erosion had spared nothing but small fragments of human bones.
The most striking individual structure of the site is, however, the large and once elaborately decorated Stûpa, marked a on the map. It rises on a terrace, partly artificial, partly due to erosion around, less than a mile to the east of the outer east wall of So-yang-ch`êng. Fig. 238. shows it from the south, while in Fig. 237 it appears as seen from the south-west along with some of the smaller Stupas which adjoin it in a row to the north and north-east. The damage caused by the large cuttings which treasure-seekers had made into what represents the upper base, and the destruction of much of the thick coating of hard yellow stucco once covering the whole, made accurate measurements difficult within the available time. But the photographs will suffice to show that in shape and proportions this Stûpa differs considerably from those I have had occasion to examine in the Tarim Basin, and manifestly represents a later type. On a lower base, apparently square and