Kum. or. Bronze arrow-head, of type Lal S. ors, but slightly larger. Ferrule retains iron tang. Long triangular depressions in sides of ferrule between blades, not pierced through. See C. xcvi. oi6 (Pl. XXIII) ; T. xxii. f. 02-3 (PI. XLVII) ; also Ser., iv. Pl. XXIX, N. xis'. oo8. Excel-
lent condition. Length rA", gr. width ti". PI. Xxili. Kum. 02. Fr. of pottery, reddish-black, corroded. Gr. M. 2I".
Kum. og. Fr. of tamarisk wood, bleached and split, but hard. Length 5".
SECTION III.—AN ANCIENT TERMINAL BASIN last few marches still separating us from the westernmost portion of the Tun-huang Limes. Having
regard to the information obtained on my passage seven years earlier and on my subsequent
After a single day's halt at Bésh-toghrak, we set out on the morning of March 14th for the Ancient Chinese route east of Béshtoghrak.
explorations along the old border line, I could feel no doubt that the ancient Chinese route to Lou-lan must here have followed the same line as the present caravan track surveyed in 1907. The reasons for this belief have already been explained by me in Serindia.1 I have, I think,
proved in the same work that ` the San-lung = (" Three Ridges ") Sands ', which the Wei
lio's itinerary mentions as traversed at their northern extremity by the ` route of the centre ', are represented by the successive ridges of dunes that the present route crosses on the first march east of Bésh-toghrak ; also that the Chii-lu granary , f r, mentioned next to the west of the ` Three Ridges Sands ', may with great probability be located at or near Bésh-toghrak. That the ancient route to Lou-lan from the point where it passed out of the area protected by the ` Great Wall ' of Han times lay actually where the track to Lop now leads, was definitely proved in 1907 by the remains of the ancient watch-towers T. I and T. II which I then traced as an advanced line intended to guard it.2
Thus there remained no question of ancient topography to be solved on my renewed passage Problem
along this easternmost section of the Lou-lan route. But the observations which I had made in old lacus-
1907, when passing through what was manifestly an old lacustrine basin east of Bésh-toghrak, trine basin.
had raised a problem of distinct interest in connexion with the physical geography of this ground. As briefly set forth in the Personal Narrative of my former journey,3 they indicated an earlier direct connexion between this basin and the terminal course of the Su-to-ho, which at present comes to an end farther south. The same observations suggested that moisture, whether above or below the surface, might even now reach the basin from that side. The levelling operations detailed in the preceding section had definitely established that the basin was included in the drainage area of the ancient Lop Sea. Its connexion with the Su-lo-ho on the other side would, if proved, imply a vast extension eastwards of that drainage area. These considerations made it obviously important to carry out a closer survey of this ground, and at the same time also to ascertain the actual termination of the Su-lo-ho to the south.
In order to attain this object within the very narrow limits of time imposed by the exhausted Plan of
condition of camels and ponies and the probable difficulties about water, I again divided my party. surveys.
Muhammad Yâqûb was to carry his levelling by one more day's work to the nearest portion of the dry basin east of Bésh-toghrak, and then to follow the caravan track with the heavy baggage to a rendezvous near its eastern end. Lab Singh was directed to leave the caravan track near the