there were five openings like big niches, and to these corresponded a series of vaulted passages on the south-east which, however, were almost completely filled with débris. Huge masses of fallen brick masonry made the examination of the interior altogether very difficult. It was only after experimental clearing in certain parts and continued study that it became possible to recover the plan of the structure in its basement story, as presented in Plate 50. The outer walls, which on the south-east face, where their foot is less encumbered by fallen masses of débris, still rise to a height of 24 feet, have everywhere a distinct slant inwards. They are built of bricks, 18" x 8" x 4" in size and fairly hard, some being partially burned. The wall facing south-east has a thickness of 6 feet 4 inches, while the longer one to the south-west measures in its unbroken portion fully 7 feet across. The whole structure appears to have formed at its base a rectangle of 62 feet by 53 outside. The basement story, which alone is still partially standing, contains within a central vaulted hall, about I I feet wide and close on 4o feet long, running from south-west to north-east and evidently once provided with an entrance from the former side. On each of the long sides of this hall lie five narrow chambers, measuring about 15 feet in length and 4 feet 2 inches across. They are vaulted at a height which could not be determined exactly, as the floor could nowhere be reached through the heavy débris within the available time. Vaulted openings, with their top some 3 feet below the line from which the vaults of the side-chambers spring, give access to them through the walls, 4 feet thick, of the central hall. On the opposite narrow side these chambers received light and ventilation from loopholes cut through the outer wall near the top of the vaulting.
Obviously a basement so massively constructed was capable of bearing a high and equally solid superstructure. But of this only a small portion has survived on the north-west, rising to a height of some 10 feet above the masonry that covers the vaults of the side-chambers. The latter are likely to have been useful for stores, etc., while the central hall evidently provided a cool place of the present keiner type for those who garrisoned this big tower or keep ; for as such it was obviously meant to serve. The fact that the south-east and north-east walls of the fort are only continuations of the corresponding walls of the keep iii suggests that this was constructed earlier and by itself. The pile rises to a considerable height over the flat ground stretching away, unbroken by any surface features, to the terminal lake-bed and the gravel glacis of the Kuruk-tagh ; even in its present ruined state it offers a very distant outlook. It was thus well adapted for guarding routes which give access from that side to the oases of Lukchun or Kara-khôja, the ancient Kao-ch`ang. Not far off to the east of the site passes a route leading from Lukchun to Singer, a tiny but strategically important oasis in the western Kuruk-tagh (see Map No. 55. D. 2), from which a number of tracks radiate towards the lower Tarim, the ancient Lou-lan area, and Kara-shahr.8
This fact adequately explains why the small stronghold may have originally been placed here. The fort is likely to have been added to it later to afford protection to those who cultivated the neighbouring area. The outer walled enclosure of irregular shape adjoining the fort to the northeast manifestly represents a still later enlargement, and its walls are of distinctly inferior strength. That there must at one time have existed around the fort and watch-station a settlement of such size as only cultivation of the surrounding ground could account for appears to me clearly indicated by the badly decayed remains of a much larger rectangular circumvallation built of stamped clay,
" The tower known as Béjân-tura (Map No. 59. A. 2) serves now an exactly corresponding purpose on the direct route from Singer to Turfân town., but did not appear to me very old when I had a chance of examining it in February, 1915. Bjân-tura lies close to the present western extremity of the terminal salt lake and nearly 400 feet lower than Chong-hassâr. The ground to the north of it is very marshy,
and in earlier times, when the extent of the salt lake was greater, may have been quite impassable. In that case all traffic from the side of Singer must have gone via Chonghassar. How distant a view the latter place commands was brought home to me on my visit to Man-tura. Though some twenty-five miles away, the ruined fort with its keep could clearly be sighted from it.