mostly covered by débris, rises a second base which • presents a circular appearance with numerous shallow projections. It is possible that the plan which served as the original model for this base was the elaborate cruciform one of the Rawak Stûpa, as reproduced in the miniature votive Stupas to be presently mentioned, but' here applied without the stairs at the ends of the cross and with the projections of the numerous angles reduced. Above this base, over 15 feet high, follows a third apparently circular base, from which rises, over a kind of cushion moulding or plinth, the cylindrical, almost flat-topped dome. A very striking feature is the elaborate and disproportionately large member above, bearing a succession of big chatiras in brickwork. The total height of the Stûpa appears to be over 4o feet, while the diameter of the main base may be estimated at about 27 feet. The impression I received of the whole Stûpa was that it belonged to later Sung times, perhaps to the period (eleventh—thirteenth centuries) when westernmost Kan-su was subject to Tangut or Hsi-hsia rule. But without fixed chronological data, either here or in the case of certain Stupas at Turfân and the Shikchin ` Ming-oi ' of Kara-shahr which have features partially resembling, any attempt at dating must be distinctly hazardous.
There was no trace left of the deposit, if any, which the interior of the Stûpa may once have contained, and the hope of which, no doubt, had excited the treasure-seekers' efforts marked by the various big cuttings and tunnels. Their burrowings had not spared the row of small Stupas, about io—i 2 feet in diameter, and half a dozen in number, which line the northern edge of the terrace. Some had completely collapsed in consequence. Those still upright, showed a small interior chamber, nowhere more than two feet square, and in the case of two Stupas this was found to have been filled with hundreds of miniature clay Stupas after the fashion of those I had discovered deposited at the Khâdalik shrine Kha. vii." Most of these little votive offerings had been thrown outside and more or less injured by effacement or weathering. But there remained enough specimens fairly intact to show that they had all been reproduced from a few moulds. The most interesting of the specimens is So. a. oo6 (Plate cxxxix). It precisely reproduces the ground-plan of the Rawak Stûpa, with the four cruciform arms of the main base carrying stairs,72 but gives details of the superstructure which no longer survive there. The essential features are preserved also in So. a. 009 (Plate CXXxix), 0010, made from less carefully executed moulds, while So. a. oo7, oo8 (ibid.) are conventionalized representations corresponding in 'general type to the Khâdalik models.
The terrace portion to the south of the main Stûpa appears to have been once occupied by a temple. But the remains of this had been completely destroyed to the very foundations, perhaps for the sake of utilizing the building materials elsewhere. The existence of a shrine was, however, attested by_ numerous fragments of hard green-glazed pottery which evidently belonged to the tiled roof. The specimens So. a. COI-2 (Plate Iv) are moulded in relief with parts of winged dragon figures. In So. a. 003, 005 (Plate IV) we have fragments of glazed appliqué reliefs in stucco which probably decorated the temple walls. To the same may have belonged also the arm of a small moulded stoneware figure similarly glazed, So. a. oo4. I may note here that green-glazed pottery of similar type used for the decoration of the tiled roof and walls was found by me in plenty at a ruined shrine of Kara-khoto, belonging probably to the Hsi-hsia period, which I excavated in 1914. Some quarters adjoining the temple on the south-east, of which the walls were just traceable, were cleared without any finds.
The account of other structural remains which I noticed at the site may be brief ; for they were too scanty or indistinct to permit of a determination of their age. When returning northward from
" See above, pp. 16x, 194. chap. xxxx. sec. i, with Pl. 58, and another at Sahri-bahlbl,
12 Cf. Ancient Kholan, i. pp. 485 sq.; ii. Pl. XL; for Stein, Indian Archaeol. Survey, Annual Report, x911-12, p.