National Institute of Informatics - Digital Silk Road Project
Digital Archive of Toyo Bunko Rare Books

712   THE JADE GATE BARRIER   [Chap. XIX

same direction a small refuse-heap had survived on the bare gravel, and this yielded over a dozen records on wood and bamboo, most of them much effaced. Among those deciphered by M. Chavannes, Doc. Nos. 392-97, is a complete slip, No. 392, dated in the year 58 B.C. It specifies the names of thirty-two garrison soldiers ' with their places of origin, among them the Fu-kuei canton being mentioned again. Among the other slips there are three containing fragments of texts, apparently medical. For miscellaneous finds in the same refuse deposit, including a fabric probably made of the fibre of the paper mulberry-tree, see the Descriptive List in Chapter XX below.

On descending from the edge of the plateau to the depression eastwards, the low mound marking the wall was very soon lost sight of amidst scrub and luxuriant Toghraks. Its direction was to the north-east, and there, only about a mile off, I discovered the remains of a small, completely broken tower, T. xvii. a, on the top of an isolated and steep terrace of clay. This rose itself like a tower to a height of some 5o feet by the very edge of a wide marshy area, which extended eastwards with several open lagoons as far as T. xix, and on this stretch of about three miles had obviously made defence by a wall quite unnecessary. To the north, too, there was water-logged, impassable bog more than a mile wide. The whole looked deceptively like a terminal basin of the Su-lo Ho, and it was only on a subsequent reconnaissance that I could locate to the north-west the outflow of the river, well screened behind a projecting plateau.3 Apart from fragments of the usual grey mat-marked pottery of the Han period, the only find made at T. xvii. a was the iron spear-point shown in Plate LIII.

It is behind the great natural barrier presented by that marsh belt, about four miles long from east to west and nearly two miles across in the middle, that there rises the imposing ruin (Fig. 186), overlooked by the tower T. xvIII, which has been briefly referred to already in connexion with my first journey to Tun-huang. When I returned for its closer exploration, towards the end of April, this huge structure still looked as puzzling as before. It was impossible to doubt its connexion with the Limes, the line of which, marked here by the towers T. xvii. a and T. xix, passed just in front of it. But neither what I had by that time learned of the general plan and arrangement of the Limes, nor the exact survey now made of the palace-like ruin, seemed at first to furnish any clue as to its true character and purpose.

The building, which the photograph in Fig. 186 shows as seen from the south, together with its enclosing walls, presented the imposing length of about 56o feet from east to west (see plan in Plate 41). If the idea of a Ya-mên or barrack had first suggested itself, this was quickly dispelled by an examination of the structural features. These comprised mainly three big halls, each 139 feet long and 481- feet wide within, which adjoined lengthwise and thus formed a continuous block facing due south. Their walls, 5} feet thick and constructed of solid layers of stamped clay from 3 to 32 inches in thickness, Occupied a terrace of hard clay. This terrace had been cut away to within to feet or so of the outer faces of the walls to provide a kind of natural base. It originally formed the northern end of a clay ridge, seen on the extreme left of Fig. 186, and had been separated from it by a deep cutting about 65 feet wide, which probably furnished a great portion of the material used for the construction of the walls. This natural base, even after the levelling that its top had obviously undergone, still stood fully 15 feet above the low-lying ground occupied by the enclosure, and thus helped to

Ruins of

T. xvii. a by marsh edge.

Big ruin overlooked by T. xvIII.

Structural features of ruin.

s On April 29, 1907, I measured here a volume of approximately 1,120 cubic feet per second. On May 13 the Su-lo Ho, where it flows out from the Khara-nor north of T. xx. c, had a volume of about 1,44o cubic feet per second. Without simultaneous measurements at these points and also

at T. xix it is impossible to estimate with any approach to exactness the amount of water which the Su-lo Ho carries into, and leaves behind in, riverine marshes between the outflow from the Khara-nbr and T. XVII.