which the removal of all this sand would have cost, since the clearing of the north-west and northeast corners of the cella showed that the remains of terra-cotta relievos were here relatively few and in no way different in character from those which the clearing of the enclosing passage yielded in
As this clearing proceeded, first along the central part of the north passage (A. T. i) and subsequently towards its east and west corners (A. T. ii, iii), the rapidly increasing finds of small terracotta-like relievo pieces enabled me to make certain general observations. In the first place I was able to convince myself that these finds were not confined wholly to decorative fragments from large haloes, etc., but also included, though in much more restricted numbers, detached fragments such as fingers, ears, curls of hair, etc., which had once belonged to relievo images up to life-size. That the walls of the enclosing passage must have been once adorned by larger sculptures was thus conclusively proved. But no other indication whatever of such sculptures had survived in the débris layer which covered the original floor to a height of about two and a half feet. It appeared the obvious explanation that they had been made of friable clay which crumbled away, just as in the strata of Yôtkan, without leaving a trace.
But the question thus raised as to why the smaller fragments had escaped the same fate soon led me to realize that this complete decay had been preceded and aided by a destructive factor of another kind. The general reddish colour of the clay in débris and wall remnants alike, the discovery of minute fragments of completely charred wood, the manifestly vitrified appearance which the surface of some of the relievo fragments showed, and other clear signs of accidental burning made it certain that the shrine had first suffered from a great conflagration. Then it became easy to see that it was the heat produced by this accidental burning which had given the terra-cotta-like hardness and colouring to the smaller appliqué relievos of the walls, as well as to such detached fragments as fingers, ears, head ornaments, etc., from the larger sculptures. Originally the material of all must have been the same friable stucco, i. e. sun-dried clay, which I had found in the Rawak relievos and statues alike. The heat of the conflagration had not been intense and lasting enough to penetrate the clay masses of the larger sculptures ; but in the case of the small appliqué relievos and of those free-standing parts of sculptures which the fire could envelop, it had sufficed to produce the effect of a regular ` firing ' and thus to protect them. The chemical analysis which Sir Arthur Church has been kind enough to furnish of a specimen from these stuccos, in Appendix D, shows that the result of this burning has been a substance closely resembling a sound red terra-cotta. Excepting such pieces, all the sculptures had completely decayed under atmospheric influences while exposed, and subsequently through moisture when the ground was levelled and probably kept under cultivation.
That some time had elapsed between the first destruction of the shrine and the complete levelling of its walls was indicated by a stratum of loess-like soil about one and a half feet thick intervening between the débris-layer over the original ground and a second layer of débris, about six inches in thickness, which was close to the present surface and contained many small relievos. The intermediate layer of loess was manifestly due to the same accumulation of fine alluvial dust which still proceeds all over the oases of the Tarim Basin wherever there is enough moisture to retain it, and which steadily raises the ground level .5 It was on the top of this layer that the hardened relievos fell when the walls were finally pulled down to make room for cultivation or some later structure. Fresh layers of loess then came to cover up and protect them until occupation finally ceased and the ground was abandoned to the desert dunes.
Wind erosion, ceaselessly proceeding between the shifting dunes, had now begun in turn
a See Ancient Kholan, i. 198.