and that the Limes extended along a line farther south. So I proceeded for about two miles to the south-south-east in the direction of a slight eminence which looked as if it might mark a ruined post. But when we reached it after crossing ground badly cut up by wind-erosion, it proved to be a dead tamarisk-cone, close to the dry river-bed. Of the Limes line no trace could be seen anywhere in this neighbourhood.13
Convinced by this reconnaissance that the border line must have had its continuation farther north, I returned in that direction and at a point about a mile and a half to the east-south-east of T. xxxvii. f came upon the Limes agger, constructed of layers of brushwood and earth exactly as between T. xxxvii. a—e. But strangely enough there were two lines of it, separated by about 90 yards of wind-eroded ground and here nearly parallel to each other. Half a mile to the southeast the two lines united at T. xxxvir. h, a tower completely decayed into a shapeless mound, but clearly marked as a watch-post by the semicircle which the agger formed round it, and by abundant pottery debris. Further examination showed that the more southerly of the two lines of agger could be clearly traced to a point about half a mile east of T. xxxvii. f, where it struck the big gravel dike at a blunt angle. The northern one joined on to the dike where it ended about a mile to the east-south-east of this point and thus linked the dike to the agger at T. xxxvii. h.
In the absence of definite archaeological evidence, we must resort to conjecture to account for the strange duplication of this short section of the Limes line. After close consideration of the ground the following explanation commended itself to me. It seems likely that the dike, whatever its origin was, already existed before Wu-ti's Limes was pushed forward to Tun-huang and beyond. At first the new line, as marked by the southern agger, was made to join this dike near T. xxxvii. f, which, by reason of its natural clay terrace, offered a convenient position for a watch-station. Some time later it was noticed by those responsible for guarding this portion of the Limes that the eastern segment of the dike, not having been utilized in the alignment of the border wall, effectively masked the ground in front of the newly built agger and thus made the guarding of the latter more difficult. In order to rectify the mistake made in the first alignment, this eastern segment of the dike, which before had been left, as it were, hanging in the air, was accordingly joined up with the Limes line at T. xxxvii. h. The line between this post and T. xxxvri. f was thereby pushed a little farther north than before.
Whether the explanation here offered or another is the right one, it seems difficult not to recognize in this duplication of the agger one more sign of the hasty and careless construction which appears to have prevailed along the Limes on either side of An-hsi. This inferior construction manifests itself very strikingly in the substitution of a mere agger of earth and loosely laid brushwood for the solid wall carefully built with regular fascines which we found all along the Tun-huang Limes as far east as T. xxxv. To the same cause, even more perhaps than to less arid conditions of climate, may be attributed the complete decay of almost all the watch-towers between An-hsi and the Limes section explored in 1907 to the north-east of Tun-huang. In the absence of documentary evidence, it would serve no useful purpose, after the lapse of two thousand years, to conjecture the reason of this inferiority of construction. Circumstances of a purely accidental character may have had as much to do with it as considerations of a topographical or quasi-strategic nature.
13 The small enclosure, seen to the south and marked with g on the map, had the appearance, when seen through the binocular, of a ruined village fort or p`ao-ta, such as I had noticed at several points beyond the present cultivated area