for the fact that just where the two routes unite to pass between the fort and the little stream separating the site from the- lands of Shih-êrh-tun, there stands, as seen in the map, a small shrine, such as is always to be found at points marking an ancient ` Gate station '.
In the continued use of the route which leads past Shih-êrh-tun eastwards to Hua-hai-tzû or Ying-p`an (Map No. 40. D. 5) and thence to Su-chou we may with some probability recognize the reason to which the towers at Shih-êrh-tun itself and those found east of it as far as T. XLII. j owe their later repairs. They would naturally be used in times of trouble in order to safeguard an exposed route between two outlying settlements. But the very maintenance of these watchtowers down to recent times also helps to explain the designations borne by Shih-êrh-tun
and by the small hamlet of Shih-tun -f -- to about five miles farther east.21
Before carrying farther our survey of the line of the Limes, it will be convenient to record here certain hydrographical observations, made first in the vicinity of Shih-êrh-tun and then confirmed by others eastwards. They have an important bearing on the question of the Su-lo-ho drainage and present a distinct geographical interest. I had already observed, on the occasion of my reconnaissance to this ground from Yü-mên-hsien on September 21, 1907, that the lively little stream, passing Shih-êrh-tun on the north side and farther on expanding in a marshy bed towards Shih-tun, was flowing in an easterly direction, as duly shown in the map of Serindia (No. 85. A, B. 2). It was obvious that this streamlet, locally known as Hsi-wan-ho, could, like the canals which irrigate the lands of Shih-êrh-tun and Fang-pa-ying-tzû, a hamlet adjacent on the east, receive its water only from the Su-to-ho.
But the details of this connexion could not be ascertained on that rapid visit. It therefore remained an open question what becomes of this drainage, which takes an easterly course and thus flows in a direction exactly opposite to that which the Su-lo-ho follows from near its great bend. The problem deserved all the more attention because Professor Futterer's text and map indicated a westerly course for the stream passing Shih-êrh-tun and Shih-tun ; 22 and also because of the extensive lake or marsh bed which is shown by the Chinese ` Wu-ch`ang map ' to the north of the Su-lo-ho bend as well as to the east of it, and which on the strength of it has found its way into recent European cartography.23
A survey reconnaissance made by Lab Singh to the south of Shih-êrh-tun (Map No. 40. c. 5) ascertained that its irrigation and that of Fa-pa-ying-tzû is brought by canals from the Su-to-ho, which pass through the northernmost extension of the Yü-mên-hsien oasis and at their ends unite with the Hsi-wan-ho. This streamlet itself is fed by springs rising in a bed which comes from the south-west and undoubtedly represents an old flood channel of the Su-to-ho. To the south of T. XLII. a this bed is joined from the west by a marshy Nullah, which in 1907 was followed for over two miles to a point where it turns south in the direction of the Su-to-ho.24 The whole of this
21 Shih-êrh-tun means ` Tower xII ', Shih-tun ` Tower
x '. The popular explanation given to me in 5907 was that Shih-êrh-tun was ` twelve towers ', i. e. 120 li, distant from Hua-hai-tzti, and Shih-tun ten towers ', i.e. loo li, from the same, a watch-tower or usually a much-reduced representation of one being found on all modern Chinese highways of Kan-su and Hsin-chiang to mark a distance supposed to be to li.
The actual marching distance between Iiva-hai-tz i and Shih-êrh-tun along the regular route, however, is fully 40 miles. This, taking the 2 miles as the usual equivalent of 10 li in these provinces, would bring Shih-êrh-tun more correctly to the position of Tower x.
It appears to me, therefore, more likely that those designations were taken from such watch-towers, whether ancient or modern, as are actually seen close to the route. Counting from T. xLII. d, seven of these were met by us on the route up to Camp 126, and three more may well stand along the track from there to Ying-p`an which was not followed on our march.
22 See Futterer, i Vüste Gobi, p. 26.
23 Cf. ibid., p. 24 ; also Sheet A. I of the maps illustrating Count Széchenyi's expedition (1877-80) ; the Tibet map of the R. Geographical Society.
24 Details of the ground are shown more clearly in Serindia, v. Map No. 85. A. 2.