Sec. vii] CONCLUSION OF WORK AT TURFAN 717
for manuring earth have been at work here so long and so assiduously that the natural soil has been reached in most ruins of this main portion of the town and only light sand deposited by the winds covers it now. How high the debris accumulations within the basement walls may have been, and how much of their archaeologically interesting ` refuse ' contents has been lost by transport to the fields of Yar-mahalla, it is now impossible to determine. We owe it, however, to these industrious, if humble and unscientific excavators, that the question as to the town's water-supply can be answered with certainty. It was doubly important at a site which owed its occupation mainly, if not solely, to its character as a natural stronghold. Owing to this removal of the debris it is possible to trace in most of the ruined houses the wells, circular and usually 2 to 22 feet in diameter, by which the inhabitants drew their water from the natural level of subterranean drainage about a hundred feet lower down. It is clear that the skill with which the modern Karéz diggers carry their shafts down to depths quite as great, and moreover connect them with perfectly levelled tunnels, is an old inheritance at Turfan.
Apart from the main central thoroughfare above referred to there are two or three narrower streets running parallel to it on the east. They and the transverse streets are again connected by a network of small alleys and lanes cut from the rock and curiously resembling the ` galis ' of an Indian town or those to be found in the heart of many an Italian city little changed since the Middle Ages. I noticed that in these alleys and also in the main roads the walls facing outwards show but comparatively few openings, a feature common to most Eastern towns from the Mediterranean to the Yellow Sea. I found no structural features indicating the location of a Bazar. But the great axial thoroughfare is broad enough to have permitted the erection of booths built of mud bricks or with wattle and plaster walls, such as are to be found in most Turkestan towns ready for temporary occupation on the customary weekly market days. Or else local trade may well have been conducted in the suburbs. Considering the small area available on the plateau and the necessity of suburbs as an adjunct to the official capital, they may safely be assumed to have stood on the opposite side of the eastern ghol or Yar where now stretch the lands of Yar-khoto village. The town on its isolated plateau could never have served the purposes of trade, as do the present towns of Turfan or Lukchun, since it was inaccessible both to camels and carts.
Near the point where the above-mentioned straight reach of the main thoroughfare ends, and the road takes a somewhat winding course, much of the plateau surface retains its original level, as seen in the panoramic view, Fig. 326, having been occupied only here and there by houses. Into a big clay terrace thus left unoccupied by the side of the road a spacious cave has been dug which looks as if it had served some public purpose, perhaps as a market or guard station. Some little distance to the east of this and by the side of a transverse road is a curious group of large chambers formed by vaults of natural clay and opening on two sides of an open court sunk into the ground. The place is known by the name of Zinc/an, ` the prison '. Whether it really served this purpose seems doubtful. As the view taken from this point (Fig. 329) shows, the southern area of the town site, which may be said to commence here, is only partially covered with buildings, mostly standing separate and none of them of any height.
There is a stretch of open ground but no clearly marked road leading towards the gate by which the track winding down the cliff face is gained. Close to this stands a small brick-built dome, which by its ` Mihrab ' or prayer niche turned to the west can clearly be recognized as a mosque. This suggests that the site of Chiao-ho or Yâr-khoto was not altogether abandoned by the time that Islam was established in the territory. But since the Ming Annals mention the city of T`u-lu-fan, i. e. Turfan, as the chief place of the territory,14 it is not likely that Yar-khoto, situated only about
14 See Bretschneider, Med. Researches, ii. p. 189.