gently rising Sai towards the outermost Pei-shan hills far away could be observed from here over nearly a day's march, and similarly the foot of the gravel glacis towards An-hsi.
Finds at It was truly an ideal look-out place, created by nature, and the tower T. XL. c which crowned
watch- the summit of the spur showed that its advantages had not been neglected by those who guarded
T. XL. c. the Limes in Han times. It was built of bricks of the regular size, with a layer of tamarisk brush-
wood after every three courses. It measured 23 feet square at the base and still stood to a height of about 13 feet, though on the south and south-west much of the masonry had fallen owing apparently to the subsidence of the slope. I noticed that on the east face one course of bricks standing vertically was inserted between two of the usual horizontal courses, a style of masonry very rarely found in the old structures of the Limes. Yet there could be no possible doubt that the tower had been built and occupied in Han times ; for among the refuse found in a small conning-place on the top, only 4 feet square, there was discovered, to my special satisfaction, a small but perfectly preserved ` shaving ' from a wooden document, bearing Chinese characters in the fine brush strokes characteristic of Han times ; 8 also two blank fragments of writing ` slips '. Small objects in wood and dressed leather found among straw of wheat, reeds, &c., are enumerated in the List below.
Mound On descending from the spur we came upon the agger again, about two furlongs to the east
marking of T. xL. b ; it was here built of layers of gravelly earth and tamarisk brushwood, and we were
Limes. able to follow it for a mile to Camp 122. It clung closely to the slope of the rocky hillocks that
here line the north side of the defile at a distance of only about 200 yards from the present river-bed. The line of the Limes was completely commanded by the crest of these hillocks rising from ioo to 150 feet above it. This clearly illustrates the fact that the line was intended, anyhow along this stretch, not for military defence but only to secure greater safety in policing the border. Where this agger descended from the foot of the hillocks to the alluvial ` thalweg ' of the river its brushwood layers had completely rotted. But about 400 yards farther to the west we found its place taken by the gravel mound already referred to, running straight towards T. xL. a and at this end still fully 20 feet high. It occurred to me that this substitution of a high and broad mound for the agger might have been prompted by the fact that the stretch of flat ground over which the mound runs is liable to inundation at the time of big floods.
Volume of After examining this ground afresh on the morning of April loth, I crossed the river to its
Su-to-ho. left bank. The water of the Su-lo-ho flowed here in a single channel about 45 yards wide and about three feet deep in the middle, with a volume of about 1,600 cubic feet per second. A comparison of this volume with that observed in the Tang-ho at Tun-huang two weeks earlier in 1907 suggests that the supply of water carried by the Su-lo-ho is less than that of its tributary until the glaciers and big snow fields at the former's head-waters begin to melt much later in the season.°
Temple on On the left bank of the river, at a point a little below our Camp 122, we found the mouth of
left river a small Nullah that descends from the Wan-shan-tzti spur occupied by a ruined temple, well bank,
built and of recent appearance (Fig. 213). Its name was subsequently given as Lao-chiin-miao 14. It had apparently been destroyed during the Tungan rebellion. Two steep rocky ridges flanking the mouth of the Nullah were each occupied by a miniature Stûpa and a small square cella, also in ruins.
Wan-shan- The end of the Wan-shan-tzti spur undoubtedly marks a point which for topographical and
suited for military reasons might have served very conveniently for the flanking defence of the line of the
`Gate' Limes, before this was extended towards Tun-huang. Probability thus supports in some degree
station. what was assumed in Serindia, viz. that the defile here crossed by the road coming from Yü-mên-
[8 M. Maspero finds in it mention of a hou lA or fire-signal tower.] ° See Serindia, ii. p. 582.