Sec. i] THE LIMES LINE NORTH-WEST OF TUN-HUANG 347
nearer than T. xxlii. b to be described presently. As the line of the wall coming from the east was traceable only up to the latter point and there abutted on the marshy foreshore, I was led to assume that for the intervening distance of fully five miles the ` wet border ' presented by the lake, here at its widest, was thought to provide sufficient defence. The ground here for over a mile from the shore is a dead flat and offers no natural position for a watch-station.
But the gap between T. xxii. f and T. xxiii. b was not left altogether unguarded ; for where the eastern edge of the belt of eroded terraces stretching from T. xxii. f approaches the northern extremity of a narrow plateau jutting out from the great Sai on the south we find the two towers T. xxiii and T. xxi11. a, close together, occupying a conspicuous position at the very end of the plateau just mentioned." I had already visited them in 1907, as the caravan route to Tun-huang passes immediately below them, and the results of my renewed examination will be found embodied in the account I have given of them in Serindia.16 Considering that the most convenient line of communication from Tun-huang towards Yü-mên and the western end of the Limes must always have passed here, it appears to me very probable that the occupation of this point by the station T. xxiii. a and its higher look-out post T. xxiii was intended to serve the double purpose of keeping a watch on the route, and of linking up the chain of the Limes guard-towers.
The group of watch-stations T. xxIII. b—g and the line of wall connecting them stretches across ground which on March loth to 22nd was found very boggy and in places almost impassable. This condition appeared to be due in the main to percolation of moisture from a depression to the south of T. xxiil. c and T. xxiil. d. The springs gathering in this are probably fed by subterranean drainage passing below the gravel Sai from the Tang-ho or river of Tun-huang. It was interesting to note that this depression is bordered to the south and east by stretches of ground heavily salt-encrusted, where the surface reproduces, in miniature as it were, the various conditions of shôr encountered in, and around, the bottom of the ancient Lop sea.
Thus, when on March 21st I transferred my camp to the spring-fed pool of Chien-ch`üan-tzû (known to the people from Lop as Shôr-bulak),17 we crossed, for two or three miles before reaching it, extensive patches of very hard corrugated salt crust recalling those encountered in crossing the
arm of the dried-up sea-bed north of Kum-kuduk. It was significant to observe that the road leading across them was worn quite smooth by traffic, and that its bottom lay three to four feet below the adjoining surface of hard salt. There could be no doubt that this sunk road was the result of traffic extending over long periods. The occasional passage of caravans and wood collectors' carts, such as use the road to-day, could not possibly have produced this condition. The experience gained here appeared to me a striking confirmation of what I had observed on the line of the ancient
Lou-Ian route where it crossed the bay of the dried-up sea-bed to the west of Camp cvi.18
We may, I believe, trace a notice both of the spring-fed pool of Chien-ch`üan-tzû and of the dried-up salt marsh to the north-west of it in the interesting topographical text of which a manuscript, recovered by me in 1907 from Chien-fo-tung and marked Ch. 917, has preserved a fragment. According to the translation which Dr. L. Giles has prepared of this manuscript and very kindly
15 Owing to a draughtsman's error which escaped my attention at the time of compiling the Map, the tower T. xxiii has been wrongly shown in No. 38. A. 4 some distance to the east of T. xxiii. a, whereas it lies quite close to the latter ; cf. Serindia, ii, p. 721. Similarly the narrow plateau tongue ought to have been shown with its end extending northward to the position of T. xxiii. a, instead of trending to the north-north-east.
16 Cf. Serindia, ii. pp. 720 sq. Some additional small
finds resulting from a supplementary clearing of the refuse heaps of T. xxiii. a are described in the List below.
17 As the water of the spring is by no means brackish, the name Shôr-bulak is evidently derived from the area of hard shôr crossed before it is reached from the west. This
accounts also for the local Chinese designation 0,,y
Chien ch`üan-tzû (` soda spring ') which it reproduces.