I to 3 feet in height crop out everywhere, and on their tops potsherds and fragments of brick lie thickest.
The abundance of small pieces of porcelain left no doubt from the first that the date of the abandonment of this site was late, and closer examination of the surface remains that I could trace confirmed this. Besides a small fort with stamped clay walls about 6o yards square, not far from the eastern belt of high dunes and half-smothered by sand, I found about a mile to the south-west of the high road a walled enclosure some 300 yards square (Fig. 252), resembling in type those ch`êng within which most of the villages and small towns on this border seek shelter at the present day. The walls, about 8 feet thick at the top, still rise to a height of circ. zo feet and bear at the northeast corner a massively built square tower. The drift-sand lying to a depth of about I I feet under the shelter of the inner side of the east wall, just as observed at An-hsi, showed that the prevailing winds come from the east. Evidently the drift-sand is carried to this area from the bed of the Kan-chou River, here fully two miles wide. At the same time the little extent of breaching observed on the face of the east wall proved that the erosive power of the winds was far from being as great on this ground as it is in the An-hsi region. Within the circumvallation numerous fragments of brick marked the foundations of houses, and the lines of roads crossing at right angles were still traceable. Evidently all structural remains had been destroyed by people from the neighbouring inhabited areas in search of building materials. Porcelain fragments were plentiful within the walled enclosure, most of the specimens taken here and elsewhere belonging to the Ming dynasty.8 A third circumvallation visited near the northern edge of the ` Tati ' area was of approximately the same size as the last, and had walls equally well preserved. Close to its west face a Tang coin with the legend
ai-yüan was picked up on wind-eroded ground. Taking into account Mr. Li's statement that metal objects of i'`ang times have been found at Hei-shui-kuo, and the fact that among the decorated fragments of stoneware and porcelain I picked up there are some which Mr. Hobson ascribes to the Sung period,° I am led to conclude that the site was already occupied at that time, and probably earlier also, though its abandonment does not date back further than the close of the Ming dynasty. I may add in conclusion that many of the fragments of hard-burnt bricks found widely scattered over this extensive area looked as if they might have belonged not to buildings but to graves.
A reconnaissance made across the richly cultivated ground to the north of Hei-shui-kuo enabled me to ascertain that across the river, along the narrow strip of arable ground left between its right bank and the foot of the barren hill range northward, there stretched a line of watch-towers evidently connected with the border line of Ming times. In 1914 I verified this assumption by tracing actual remains of the Ming wall on my descent along the right bank of the river. I may conveniently record here that the line of this later ` Great Wall ', closely hugging the narrow strips of cultivation on that bank, was then traced to near the village of Hsiang-p`u (Map No. 91. D. 2), whence it evidently was continued across the river westwards.
At Kao-t`ai, a place of some antiquity and local importance,10 the high road passes the point where the belt of cultivable ground, between the here utterly barren glacis of the Nan-shan and the desert hill range across the Kan-chou River, is reduced to its minimum width, less than five miles in all.11 One march further, at Hua-chuan-tzû (Map No. 91. D. 2), the high road leaves cultivation behind
8 Cf. the Descriptive List at the close of this section, with Mr. Hobson's analysis of the ceramic fragments.
See Kan-chou. 001-5, 007, 009 in Descriptive List below.
10 See Map No. 93. A. 3. Kao-t`ai is duly mentioned as Gaula in IIâji Muhammad's remarkably accurate list of caravan
stages from Kan-chou right through to Kashgar, Samarkand, and beyond ; cf. Yule-Cordier, Cathay, i. p. 293.
" The map (No. 93. A. 3) needs a slight correction here. The river approaches Kao-t`ai within a mile or so, and the village land of Li-po, on the opposite bank, is less than a mile wide.