The next tower, T. xxiii. o, standing at the end of a low terrace, was, however, fairly well preserved, and with its brickwork still rising to a height of about 15 feet afforded guidance. On clearing the refuse near it, two inscribed Chinese tablets were recovered and some miscellaneous objects, including part of a lacquered food bowl, T. xxIII. o. 01 (Pl. XLVII). From here onwards the line of the wall could be followed again over gravelly soil eastwards as far as T. xxiii. s. Of the towers intervening at distances of three-quarters of a mile to one mile, T. xxiii. p and r could be traced only in the form of low mounds. T. xxIII. q still rose to a height of about 12 feet and T. xx111. s of about 17 feet ; their masonry of ` regulation '-sized bricks included layers of reeds, in the former tower between every two courses, in the latter between every five. At none of these posts was refuse traced on the soft soil.
We now steered in the direction of a conspicuous tower, T. xxiii. t, visible to the east-southeast, for about a mile and a half, but found progress more and more difficult for the camels owing to spongy soil covered with soft slzôr. When within half a mile of our objective, a sheet of water flowing from the south finally stopped us. We were obliged to turn to the south and felt glad when, after a couple of miles' very trying progress, with the camels foundering in the bog, we gained practicable ground on a low clay terrace stretching to the south. Looking eastwards from it no towers could be sighted beyond T. xxiii. t ; but far away in the distance trees and farm buildings marked the northernmost outskirts of the Tun-huang oasis.
It was obviously impossible, at this season of inundation, to make our way to the east across ground reached by the spill ends of the Tun-huang canals. I therefore directed our march upon two ` Pao-t`ais ' visible to the south, until nightfall obliged us to halt. When the march, resumed on the morning of March 24th, had brought us along a line of low clay ridges flanked by marshes to the larger of the two towers sighted in the evening, it proved manifestly old but quite distinct in character from the watch-posts of the Limes. This tower, T. xxiii. u (see plan in Pl. 14), which, as I subsequently learned, bears the local designation of Yen-chi-tun, measured 29 feet square at its base and was constructed of salt-impregnated layers of clay and gravel reinforced by thin strata of reeds at intervals of 8 inches. An oblong enclosure, much decayed, showed that the tower was meant as a place of refuge, such as are often to be found near outlying homesteads of these western marches of Kan-su, exposed as they have been to raids and disturbances during recurrent periods of history right down to the last great Tungan rebellion.29 Large gaps in the sides of the tower proved that it could not be of modern origin. But whatever its age, it was clear that it could not be connected with the Limes line. This most probably passed from T. xxiii. t eastwards to the vicinity of the deserted town of Shih-pan-tung,30 and thence joined on to the section of the wall which in 1907 I had traced in the north-east of the Tun-huang oasis as far as the tower T. xxx.3'
The second and much smaller tower to the east proved obviously recent. So we moved on to the south-east, where a narrow gravel plateau now offered easy going for a time. Marshy depressions extended along it on either side, with open sheets of water fed by springs which obviously discharge subterranean drainage from the irrigated area to the west of the Tang-ho. Beyond these depressions a few scattered homesteads came into sight, apparently all deserted, sad mementoes of the destruction wrought here as elsewhere along the Kan-su border by the Tungan rebellion.
Finally, after having been brought up again and again by inundated ground and forced to make detours, we reached the wide gravel Sai and skirting it arrived at the edge of Tun-huang cultivation. After two months' continuous hard work in the desert, the familiar sight of the oasis, with
29 Regarding the little village forts known as p`u-tzû or S0 See Map No. 38. B. 4 ; Serindia, ii. p. 588.
pao-tzg, cf. Serindia, ii. p. 587. S1 See Serindia, ii. p. 603.