conjecture that these stones might originally have been kept on the parapets, to be thrown down by the defenders in case of attacks. This guess found early confirmation.; for on my subsequent visit to the Kan-su marches I saw such big stone ` ammunition' not merely often lying at the foot of the ancient watch-towers of the Tun-huang Limes, but still regularly stored for defensive use on the parapets of modern town walls and village forts built since the last Tungan rebellion.
SECTION III.—EXCAVATIONS IN THE MIRAN FORT
It was under the protecting cover of the slope of fine gravel and sand which had accumulated on the lee side of the east wall of the fort that my trial excavation of December 8 had first disclosed the remains of a series of small, roughly built quarters, which that wall had once helped to shelter from the bitter north-east winds. In my account of this first excavation,' I have already described how the thick layers of refuse filling these rooms had yielded Tibetan documents in plenty, besides other remains, making it abundantly clear that the ruined fort had been occupied for a long time during the period of Tibetan predominance in the Tarim Basin, roughly during the later half of the eighth and the whole of the ninth century A. D.
The excavations resumed on January 24, and continued without intermission during five days in the face of very trying atmospheric conditions, soon showed that my hope of finding more relics of that occupation was well founded. The previously cleared quarters, M.1. i—vii, had extended in an irregular line south of the north-east angle and, as the plan, Plate 3o, and the photograph, Fig. 114, then taken show, at a little distance from the east wall of the fort. It was in the space intervening, nowhere more than about seventeen feet wide, that the resumed digging soon disclosed another row of small rooms, M. 1. viii xi, which subsequently proved to extend right along the east wall to the neck of the south-east bastion. Whereas in the previous row only the first two rooms, i and ii, had been sufficiently covered by the slope of gravel and sand to retain their roofing, it had been preserved in the rooms viii—xi through the protection afforded by the rampart behind. But it was not until the consolidated masses of refuse which filled these rooms right up to the roof, constructed of rough Toghrak beams and layers of reed fascines, had been dug down to the floor, that I realized how much these accumulations of rubbish must have helped to strengthen the wall itself against wind-erosion.
The rooms themselves were built in a very rough and irregular fashion of coarse brickwork and clay, with walls of varying thickness. As the photograph, Fig. 115, of rooms viii and ix shows, the walls still retained in places their smoke-begrimed facing of plaster. The wall dividing these two rooms was only nine inches thick, and the floor of the second nearly three feet higher than that of the first. Similar differences in level between adjoining quarters were frequent, and indicated haphazard construction on various occasions. Only in a few, like xi and xiv, could low fireplaces of mud plaster be traced. Owing to the way in which the walls of friable clay and bricks were apt to crumble away when the refuse filling the rooms was being removed, it was often difficult to ascertain where approach to them had lain. Elsewhere, the entrances appeared to have been walled up with mud when the quarters abandoned to refuse had been completely filled up and ceased to be useful even as dust-bins.
The most remarkable feature about these humble quarters adjoining the east wall was the fact that a number of them, especially viii, ix and xiii, xiv, were found crammed with refuse, reaching almost to the roofing wherever this survived. In this there mingled thick litter of reed stràw, chippings of tamarisk wood, and sweepings of the hearth with what, for brevity's sake, we may call
See above, pp. S48 sq.