Sec. i] TO THE EASTERNMOST OUTPOST OF LOU-LAN 287
sides. The paring down of the sides of what must have been the foundation of a square tower or watch-post was due to exactly the same cause as had brought about the complete breaching and final disappearance of the walls of the L.A. enclosure where they faced the prevailing wind direction, while those lying parallel to it could still be traced over considerable stretches.12
The dimension of about twenty feet and a half that the extant fascine layer still preserved in the longitudinal direction corresponds exactly to the measurement which the square watch-towers of the Tun-huang Limes most frequently show at their base.13 The use of tamarisk brushwood for the fascines similarly agrees with the evidence afforded by the wall and towers of the Limes that the old Chinese engineers readily adopted for their constructive work whatever material the adjoining ground then offered and usually still provides.14 There were accordingly strong grounds for inferring that this tower at L.J. was of Chinese origin and of a period approximately contemporary with the Limes. That it had been tenanted as a look-out post, just like the towers of the Limes, was proved by the small heap of miscellaneous refuse I found half-hidden under fallen clay about four or five feet below the foot of the south-east corner of the surviving fascine layer. In this refuse were burnt pieces of tamarisk wood ; a mass of fragments of rush matting, L.J. 02 (Pl. XXVI) ; a small heap of reed-straw ; remains of a grass, L.J. 09, which Tokhta Akhûn declared to be yiken, such as grows along the shores of rivers and lagoons,'5 and plentiful chips of Toghrak wood. Fragments of rough pottery, L.J. 04-6, were picked up lower down on the slope.
However insignificant these finds were, and however modest the remains of the tower itself, their discovery on the Mesa L.J. afforded important guidance, and justified my feeling elation. The position of the ruined watch-post on the plane-table fell exactly into the same line to the northeast as all the sites we had traced from L.A. to the castrum L.E. and beyond. This appeared a sufficient indication that the ancient Chinese route had passed here, and that I should have to look for its immediate continuation in the same direction. It is true, this bearing was leading us away from the great valley-like depression between Besh-toghrak and Kum-kuduk, with its wells and grazing, through which passes the present caravan track from Lop to Tun-huang, and where alone water could be hoped for. These natural advantages must have imposed its use likewise for the ancient Lou-lan route. At first therefore it might seem somewhat disconcerting deliberately to turn away from the direction which would bring us to that track.
But from what the march of February, 1907, had shown me of the great salt-encrusted Lop sea-bed, and from the graphic description of Professor Elsworth Huntington, who a little more than a year before had pluckily crossed it from Kôshe-langza to Altmish-bulak,16 I knew that the vast expanse of hard salt crust offered most serious physical obstacles to a route leading straight across to the south-east from ancient Lou-lan. I also remembered what previous explorations had taught me of the remarkable skill with which Chinese enterprise always adapted itself to essential topographical features and thus economized effort and avoided needless risks.'' By continuing farther to the north-east and thus drawing nearer to the foot of the Kuruk-tâgh the Chinese pioneers of the ancient ` route of the centre ' were able to reduce the extent of the most difficult portion of the ground, that presented by the dry sea-bed with its corrugated surface of hard salt. At the cost of a detour they could thus avoid much hardship and escape the necessity of confronting physical obstacles which, added to all the other difficulties of the desert, must then as now
12 Cf. Serindia, i. p. 386 sqq., and for exactly corresponding observations as regards the badly breached walls of ruined towns near An-hsi and Ch`iao-tzû, ibid., iii. pp.1095 sq., 1102 sq.
13 See ibid., ii. p. 737 ; for watch-towers, twenty to twenty-one feet square at the base, cf. ibid., ii. pp. 591; 597, 600, 603 sq., 635, 644, &c.
14 See ibid., ii. pp. 736 sq.
15 Mr. Scully's list of ` Turki names of plants ' in Shaw, A Sketch of the Turki Language, Pt. n, p. 226, gives yakan. Reed-mace, Typha angustifolia.'
16 Cf. Huntington, Pulse of Asia, pp. 251 sqq.
17 Cf. e. g. Serindia, ii. pp. 583 sq., 632, 663, &c.
Evidence of occupation of watch-. tower.
Guidance from position of L.J.
Physical obstacles ahead.
Reasons for continuing to NE.