592 AT RUINED SITES OF TURFAN [Chap. XVIII
windows ; for thus alone could the passage round the circular interior wall have received its lighting. This supposition is borne out by the presence in the circular wall of eight small arched windows about io inches wide, at a height of about a foot from the floor of the passage, giving on the interior, and each opposite to a corner formed by the octagonal facets of the outer wall. That the interior once had a double vaulting, as found in other Stûpas of Kosh-gumbaz,8 is made probable by another set of eight ` windows ' which penetrate through the whole thickness of the base, 14 feet, into the inner chamber, at a height of about 3i feet above the ground level and about as much below the floor level of the passage. These windows were 2 feet 3 inches wide outside, reaching a height of 3 feet with their arched points. At about a foot and a half from their inner end these window passages appeared to have been partially blocked by brickwork, perhaps used in pinjory' fashion.
In the course of clearing the passage a considerable number of fine fresco fragments were recovered which evidently had belonged to a painted frieze higher up. None of them were found less than 5 to 6 feet above the passage flooring, which suggests that the walls of the passage below this level had been left without decoration. Some of the fresco fragments were found lying one in front of the other, as if they had slid down successively, as had happened with the frescoes of the Mirän shrine M. III.8 The entrance to the passage, if it had one, must have been from the south ; but there both the octagonal outer and the inner circular walls showed a big breach, made, no doubt, by those who searched the sepulchral structure after it had become a ruin. Against the south-eastern facet of the base was built a small annexe, originally double-storied, containing on the ground floor two vaulted rooms, each 15 feet long. The one nearest to the sepulchral structure was only 5 feet wide and may have once contained stairs leading to the upper rooms ; but this could not be determined. The fact that the enclosing wall of the quadrangular courtyard, against which the annexe was built, stood on this side i8 feet away from the main structure and only 4 feet away on the opposite side suggests that the annexe was contemporary with the former. In this case it may possibly have served as a convenient place for performing funeral rites or keeping articles needed in connexion with these, &c.
That the sepulchral structure itself had at some later period been put to use as quarters or as a storing-place appeared probable from a large layer of straw and brushwood that came to light at the bottom of the western portion of the passage. We also found there a large but torn sheet of paper with Chinese writing, apparently a commercial document ; fragments of leaves with Brahmi and Uigur writing ; and some small wooden slips, one bearing Uigur script. A leaf with some fragments in Uigur was also recovered on the floor of the eastern side of the passage, besides textile remains, including the pieces of printed silk, Kao. III. E. 01. a, b (Pl. LXXXVI) and the fragments of silk tapestry, Kao. III. E. 02. a. Other objects found here were the wooden weighting board of a small banner, Kao. III. obi, and the black-lacquered inkpot Kao. III. E. 03 (PI. LXXI).
Quite distinct from these remains is the large hoard of metal objects which was unearthed on the north-eastern side of the passage, well above the floor. Before, however, turning to this cache, brief reference may be made here to the remains of the original fresco decoration of the passage, which are fully described in the List below but await illustration. The two largest pieces, Kao. III. 019-20, each close on two feet in length and about one in height, show the figures of male donors kneeling in front of altars or pedestals with sacred emblems (?) and accompanied by remains of Uigur inscriptions. Some of the bearded heads preserved show interesting details of features and head-dress which may help towards an approximate dating. Among the smaller fragments, Kao. III.
s See Serindia, p. 498•
8 Cf. Griinwedel, Idikutschari, p. III. i.