Sec. i] SCULPTURED REMAINS OF RUIN M. II 491
period through which the once inhabited area appears to have passed before becoming the desolate arid waste that it is now.1e
This leads us to the important question, still remaining to be examined, of the approximate periods to which the construction of this shrine and its abandonment are likely to belong. That the construction must have preceded the Tibetan occupation of the site will, I think, have been made sufficiently clear by the observations already recorded about the actual remains. That the abandonment also preceded it is made very probable by the total absence of any finds suggesting Tibetan character or origin. But it may help us to more definite chronological conclusions if we briefly review the available indications as a whole.
In the first place, some safe guidance can be derived from the architectural decoration of the Vihara base. We have already seen that this is closely linked up by its Indo-Persian pilasters with Graeco-Buddhist models of Gandhara. Now it should be specially emphasized that this architectural element can also be traced at other ruined sites of the Tarim Basin, and in specimens which seem to indicate definite chronological limits for the Miran pilasters both upwards and downwards. On the upper side we have, as already shown above, the important testimony of the wooden double-brackets, ending in down-turned volutes, from three different ruined sites of Lou-lan. They prove that the very shape of this member, as displayed by the pilasters of M. II, must have been prevalent in actual decorative use throughout the Lop region as early as the third century A. D. Considering the very conservative fashion in which the Buddhist art of Eastern Turkestan has treated the forms derived from Graeco-Buddhist models, we might well be in doubt as to the downward limit of the period during which this particular shape of the Indo-Persian double-bracket remained current, were it not for certain finds, made among the ruins of Endere and the Domoko tract, which prove that by the
i Tang period both this double-bracket and the Indo-Persian column bearing it had already undergone
an unmistakable change.
Taking the double-bracket first, we have the very interesting specimen in wood, F. it. ii. or,
i from the site of Farhad Beg-yailaki, some ten miles north-north-west of Khadalik, which is shown in
f Plate XVII and well repays attention. Here we have before us an unquestionably later develop-
ment. In it a modified, but yet clearly recognizable, form of the volute-ended Indo-Persian double-bracket is surmounted by a second identical in all essential features with the double-brackets characteristic of the Niya Site, of which specimens are reproduced in Plate XVIII and also in Ancient Kholan, Plate LXIX.17 We also find the same combination in the pair of wooden double-brackets, Kha. v. 003. a, b (Plate XVII), from the Khadalik temple ruins. Here, too, the lower portion represents a modified and later form of the Indo-Persian double-bracket, the voluted ends appearing as a particularly striking feature both at top and bottom. The upper portion is a double-bracket of the Niya Site pattern, treated very plainly and lacking the floral carving of the under-surface at the ends. Comparing the Farhad Beg-yailaki and Khadalik specimens, it seems to me that the former stands distinctly nearer to the models from which both the constituent portions are derived. It is certain that the site of Khadalik was abandoned in Tang times, towards the close of the eighth century.'$ As to the ruins of Farhad Beg-yailaki, we shall have occasion to show below that they probably belong to the centuries immediately preceding the Tang period.l$a
The columns which carried those Farhad Beg-yailaki and Khadalik double-brackets were not found in either case. But the latter site at least furnished a sample of the probable appearance of these columns in the lathe-turned wooden pillar found in Kha. ix and seen in Fig. 42.19 That its