616 AT RUINED SITES OF TURFAN [Chap. XVIII
The main features of the central portion of this group, which consists of a large cella flanked on either side by a row of vaulted rooms with small rock-cut recesses behind them, have been described by Professor Grünwedel.14 The mural paintings, already much injured, which he was able to examine in the exposed structures marked by him as A and B, b, had since suffered even worse damage, and the detached temples, C, D, once standing in front of them, had been reduced to almost shapeless masses of debris.'s On the other hand, a number of smaller ruins occupying the north-eastern end of the main terrace appeared to have remained in much the same state of advanced decay in which I remembered to have seen them on my cursory visit in November, 1907, and these I decided to have cleared.
The first structure opened, i in the sketch-plan Pl. 27, was a small cella, 12 feet square, containing the plinth of a Stûpa, 82 feet in diameter, which had proved to have been burrowed into. On the floor below were discovered a convolute of Chinese manuscript leaves and fragments of Chinese and Uigur texts. The approach to this cella lay through a small shrine, ii, once vaulted, measuring 16 feet by 7 ; this showed at its south-western end a recess which had once been occupied by a large image, as indicated by a stepped base in front. Of this image nothing survived but a portion of the lotus pedestal and remains of the feet. But in the layer of sand covering the floor of the room large numbers of small relievo fragments in stucco were recovered, mostly painted or gilded, which must have fallen from a decorative frieze high up on the walls before destruction overtook these.
In various ways these small fragments recalled the remains of the relievo friezes that had once adorned the walls of the temple cellas and passages Mi. x—xii at the Ming-oi site of Kara-shahr 1s Here, however, the scale of the figures filling the frieze must have been considerably smaller, in proportion to the reduced size of the shrine. It is impossible from the mass of stucco fragments to form any definite notion as to the scenes represented in the decorative scheme. So much, however, is clear, that human figures, as well as representations of the dead, must have been very numerous among them. Evidence of the former is afforded by the many small heads, Toy. iv. 072-122 ; ii. o8, 9, 58-60, &c. (Pl. LXXII), of different types and sizes, most of them apparently appliqué, and the hands, ii. 034-40, arms, 09, 17 ; ii. 080-94, and feet, ii. 017-33 (Pl. LXXII), also differing in size. The numerous skulls, 026-39, 52-8, &c. (Pl. LXXII), and the cadavers, 024-5 (Pl. LXXII), suggest work in the Tantric taste. Remains of monsters are represented by fragments like ii. oI I-15, 58 (Pl. LXXII). Among the few fragments of animal figures, the bird 07 (Pl. LXXII) may be mentioned. Ornamental details in the shape of jewels, rosettes, bands, &c., from drapery, 010-16, 70, 123-40, &c. (Pl. LXXII), are abundant and often richly coloured. The curling flames o8, 18-23, ii. 041-53 (Pl. LXXII), probably belonged to haloes. Among arms and implements, which dropped to safety at an early stage owing to their detached modelling, we find the Vajra, 0141 (Pl. LXXII), halberd points, 0142-3 ; ii. 0105, and mallet, 0144 (Pl. LXXII). Architectural fragments, ii. 073, and three votive relievos in clay, ii. 076-8 (Pl. LXXII), showing Buddha or Manjusri, may also be mentioned. Of the frescoes that once adorned the walls, only small pieces, ii. oi-6, executed in rich colours with details in raised gold, have escaped destruction ; in one of them, ii. oi, part of Vaisravana's banner is recognizable.
The passage, iii, leading into a small rock-cut shrine northward from which wall paintings appeared to have been removed, was next cleared. It yielded only some Chinese manuscript