Plate vIII), the Rawak walls again showing clearly the manner of their disposition. Everything in the subject and style of these aureoles and their decoration points to the closest relation between the work at both sites and an approximately contemporary origin.
Whether the larger appliqué figures of seated Buddhas, of which A.T. 017 (Plate VIII) is a specimen, served also to decorate aureoles, or had an independent position, cannot be decided. Nor
is it clear what place was occupied by the occasional female figures met with (iii. oo61 ; iv. oo66) or
by the demons of which the heads have survived (iv. 009 ; v. 0025). Even grotesque animals seem to have found here and there a place in the wall decoration (iii. oo82 ; iv. oo161). Architectural frag-
ments are rare (i. 0077 ; v. oo67, 0071). Where the mass of small appliqué pieces is so great, even
the negative evidence furnished by the absence of certain motifs may deserve some attention. On this occasion I may note that the Ak-terek collection does not include a single specimen of those
flying Gandharvis and small standing Buddhas which, from the evidence of the shrines of Dandân-
oilik, Khadalik, Kara-yantak, Farhad-Beg, would seem to have been favourite subjects of decorative sculpture in shrines of this region during Tang times. Marked differences can also be traced
between these Tang shrines on one side and the Rawak and Ak-terek group on the other in
the style of ornamental designs. In the former certain motifs present a distinctly more florid development while others, like the fleur-de-lis shaped Vajra, have disappeared altogether. But
detailed study of these changes must be left to workers of the future. Whether there was a corresponding development in the style of fresco decoration it is impossible to judge, since, as at Rawak, only the scantiest traces have survived of the painting of the Ak-terek temple walls (A.T. 00122).
That the architectural and decorative features traceable in the Ak-terek remains were typical of the Buddhist shrines of this region and period was proved by the visit I paid on September 19
to what my guides called the ' little Bat-khâna'. For about two miles we moved to the west-
south-west over ground where low dunes alternated with extensive stretches of ' Tad ' until the trees of Ak-terek village, which has given its name to the whole site, came clearly within view
southwards. There, on a flat débris-strewn area clear of dunes, which my guides called Siyelik,
I found the outlines of a small quadrangular Buddhist temple fully exposed and clearly recognizable by clay walls almost flush with the flat eroded loess soil around. The cella within measured about
twenty-five feet by twenty-three feet. Along its sides an outer passage, about six feet nine inches wide, could be recognized by its plastered flooring. The centre of the cella appears to have been occupied by a base, about nine feet square, the stuccoed faces of which, now decayed into friable clay, still survived to a height of about six inches. There can be no doubt that this, like the base found in the temple cella of Endere,11 had once served for the support of large stucco images.
Fragments of appliqué relievos in stucco exactly similar in style and material to those which once adorned the walls of the larger shrine could be picked up on the surface, particularly near the south
wall of the cella (Si. i). Some more turned up near the north-east corner (Si. ii), where I had the
interior of the cella cleared down to the original floor, only about one and a half feet below the level of the extant débris. Certain decorative fragments look almost as if produced from the same moulds
as corresponding appliqué relievos of small seated Buddhas, etc., found at the ' Ak-terek' ruin.12 Most probably an identical original had served for the production of the moulds. There were found also fragments of larger Buddha heads (Si. i. oo16; ii. 005) closely corresponding in type to those of Ak-terek.
The hardness of these fragments was unmistakably due to accidental burning, as at the larger shrine, and there were plentiful indications to show that the ruin after its abandonment and the