the enclosing passage eastwards. These had evidently fallen down from the upper part of the wall now destroyed and had been subsequently protected by the sand. Here the four-petalled clematis appeared painted in terra-cotta colour over white stucco, with a scroll ornament in black resembling a cinctured garland in loops,
This style of decoration distinctly favoured the assumption that the ruined shrine belonged, like the Rawak Stûpa, to a period which may be approximately placed between the fourth and seventh century A. D. The total absence of Tang or later coins among the ` cash pieces picked up under my eyes on the neighbouring 'Tati', or brought to me from this neighbourhood, confirmed that assumption. With it also agreed the considerable depth, from ten to twelve feet, to which wind erosion had lowered the unprotected ground immediately adjoining the ruin. That the latter had suffered by fire, either before or after its abandonment, was suggested by plentiful charred débris within and near the cella and the blackened appearance of the wall stucco in places.
But it would be manifestly unsafe to draw from this isolated fact any conclusion as to why this whole tract was abandoned. On the other hand the day's observations were quite sufficient to prove that continuous cultivation must once have extended between the present north edge of the oasis near Suya and the Rawak Stûpa, an area some eight miles from south to north and now completely overrun by high dunes. The early date of abandonment which the extant data suggest for this tract has a special interest. In the first place the close vicinity of the Yurung-lash River, which still supplies this ground with abundant subsoil water, did not avail to protect it from lapsing into absolute desert. In the second place it is important to observe that an area so near to the main oasis was abandoned centuries before the desert sands were allowed to overrun the settlement of Dandûn-oilik, some sixty miles further out in the desert north-eastwards. It is a clear indication that the progress of general desiccation cannot by itself supply an adequate explanation of all such changes in the extent of cultivated areas.
On the morning of September 17, after a final survey of the dunes in relation to the walls still exposed, I said farewell to Rawak and proceeded to the ruined site of Kine-tokmak whence Raze, my old guide Turdi's stepson, and his ` treasure-seeking ' companions had brought to Khotan fragments of small stucco relievos which had once served for wall decoration, of exceptional hardness and yet withered and cracked by long exposure to the summer heat and fierce winds of the desert. By moving first for about three-quarters of a mile a little south of east my guides were able to show me a ' Tim ' they had previously mentioned ; this proved a badly decayed Stûpa mound, about fifteen feet high and almost completely enveloped by the foot of a big dune. Striking thence to the north-east we emerged after about half a mile from among the high dunes on to a belt of ground where the ridges of drift sand were low and small patches covered with pottery débris frequent. This was said to extend northward to the site known as 7umbe-kum which I had visited in April, 19oI.9
After having covered from Rawak a total distance of close on four miles I arrived at the `old house ' from which Raze and his men had brought their relievo fragments. It stood in the middle of a small plain, covered with dunes only six to ten feet high. The condition of the ruin corresponded exactly to the conditions I had deduced from those scanty relics. It consisted of the remains of ancient brick walls, forming the corner of a rectangular structure and traceable for about thirty-four feet on the south-east side and about forty feet on the south-west. If a much-broken wall section to the north-west belonged to the missing part of the quadrangle, the latter, when intact, would have measured about sixty-two by eighty-two feet, the longer sides facing to north-east and south-west. But the ground was far too much eroded to permit of safe measurements. The extant