Sec. iii] BY THE HAN LIMES TO AN-HSI 367
before reaching the next traceable remains of a watch-tower, T. xxxvii. d. This had decayed into a small shapeless clay mound ; but the semicircle which the wall made round it, with a diameter of 5o yards, rendered its position quite certain.
From here onwards we found rows of low Yârdangs approaching close to the line followed
by the Limes. Their axis lay almost due east and west, whereas the Limes agger followed an alignment of S. 97° E. It was here reduced once more to the condition of a low gravel-covered mound. In this form we managed to trace it to another completely decayed tower, T. xxxvrt. e, and thence for about three-quarters of a mile farther, until we lost it in a belt of closely packed Yârdangs flanking the old river-bed previously mentioned. At this point the late hour and our distance from the river where we should have to look for the night's camp obliged us to suspend our search. But the sight of a ruined tower to the north-east showed us where to resume it next morning.
From Camp cxxv, which we were lucky to be able to locate and reach after nightfall, as it
was pitched low down by the river, I started back on April 12th, Easter Sunday, to the Limes line. While moving with the baggage towards the previously noted cart-track which it was to follow to the western edge of the An-hsi oasis, we came unexpectedly, after proceeding a mile or so amidst luxuriant scrub, upon a small area of abandoned cultivation. The.canal that once carried water to it from the east was still clearly traceable, and as no wind-erosion had as yet set in on the fields, the date of abandonment could not be remote.
The tower, T. xxxvir. f, sighted on the preceding evening, was at once shown to be old by the
abundance of pottery debris of Han type found lying about its foot. I t still rose to a height of some i8 feet and measured about as much on each side of its square base. Its construction was unusual ; for fissures in the mass of stamped clay showed that this formed only a kind of outer casing for a core of natural clay. An isolated erosion terrace, about io feet in diameter and of irregular shape, had apparently been utilized for building this watch-post. In order to secure cohesion, wooden beams had been fixed into square holes cut into the natural clay, and then the stamped clay had been built up around the beams. The latter, as well as smaller pieces used for this wooden reinforcement, still survived. But their very rotten condition was plain indication of climatic
conditions less arid than those to which the Limes remains west of Tun-huang bore witness. On clearing the top of the tower the natural clay emerged and showed by its red colour that signal-fires had been kept burning there. The only refuse traced near T. xxxvtr. f consisted of a heap of brushwood, mainly small twigs, such as occurs in the agger of this section of the Limes, and some animal bones.
At first the position of this tower appeared distinctly puzzling. To the south-west of it no clear alignment could be made out connecting it with the agger traced to T. xxxvtr. e. The ground was broken up by closely packed Yardangs, though gravel mounds of short length seemed to crop up here and there. To the north of the tower, however, I noticed that a broad dike-like bank of
gravel, about 70 feet across at its foot and for the most part 15 to i6 feet high, started at a distance of about 5o yards and extended in a slightly curving line eastwards for about a mile. It looked very different from any part of the Limes wall I had traced so far, but somewhat resembled the broad gravel embankment which in 1907 I found running across the bare Sai from Nan-hu to the Tun-huang river."
Feeling uncertain about the relation between this strange-looking gravel dike and the tower T. xxxvii. f, I at first considered it possible that the latter might have been intended merely as an advanced post for observing the ground otherwise screened by the dike, perhaps of earlier origin, 12 See Serindia, ii. pp. 6ro, 6t7 sqq.