Sec. i] EXPEDITION TO THE KARA-DONG RUINS 451
small bits of paper also points to a period not very distant from that ascertained for the ruined fort of the latter site. As to the specific cause of the abandonment it would be hazardous to express any definite opinion. A variety of causes, which we have no means of tracing, might have led to the transfer of the post to another locality, or to its complete desertion.
It is easy, of course, to suggest a change in the river-course which deprived the station of convenient access to water. That these lateral shifts of the Keriya river must have been in old times quite as frequent and marked as they are now is evident. Yet it is noteworthy that the direct distance between Kara-dong and the Toldâ.ma river-bed, which still receives water during the summer floods, is only about five miles, i. e. considerably less than the distance separating recent beds of the river, e.g. below Kochkar-Öghil or Tonguz-baste. When returning from Kara-dong on March i8 I passed, within only one mile of the main ruin, a depression occupied by a large number of small Toghrak trees, all dead, yet clearly showing by their appearance that the time when they died cannot have been very remote. It would appear that they all sprang up together during a temporary return of a river-branch in this direction, but that the period of revived subsoil moisture was not long enough to permit these hardy pioneers of vegetation to grow up fully. Further on I reached the first groups of living Toghraks within about two and a half miles from the ruin, growing comfortably between the westernmost of the three Dawâns, 50 to 6o feet high, which my map marks. Taking into account all conditions, I am inclined to look upon the position of the Kara-dong ruins as striking evidence how relatively small a change the main direction of the Keriya river course has undergone during the last twelve hundred years or so. A reference to the questions discussed above with reference to the Dandan-Uiliq site will show the antiquarian interest attaching to this observation 17.
Lateral shifts of river-course.
K. ooi. Small objects in metal, glass, &c., found amidst débris of Kara-dong quadrangle.
Small fragment of carved bronze plate. 1" x A".
Iron point, long square pyramid-shaped, with tang. Probably an arrow-head, but possibly a drill. Length 2", width of thickest part A". See Pl. LI.
Wire-like strip of iron. Length 2".
Small irregularly-shaped piece of iron. Length Am.
String of 3 seed pearls, on each side of which a pink (coral) bead, the remainder being greenish glass beads, all strung on a (modern) thread. See PI. LII.
Small fragment of thin yellowish glass ; curved.
Small copper rivet.
[K. 002. Objects found on eroded ground east of Kara-dong quadrangle.]
K. 002, a. Bronze buckle. The portion which was attached to the garment or belt is double, one plate being on each side of material, to which it was secured by four rivets. Loop of buckle rectangular, and lying across it is the tongue, which is broad at base, and rapidly narrows to ordinary thickness, broadening again slightly before reaching angular point. The buckle is flat on inner side, and rounded on outer. In this example there appears to be
a raised decorative strengthening piece on the outer plate, very similar in form to the tongue, but broader. Outer plate 1" x rg". Loop â" x rr. See Pl. LI.
K. 002. b. Bronze buckle similar to K. 002. a. Outer plate r x i $" ; loop â" x
K. 002. C. Double plate of buckle, similar to K. 002. a. Hinge broken and loop, &c. absent. Rivets remain.
K. 002. d. Double plate of buckle ; inner plate has become detached and is sticking to front of outer plate. Rivets remain. :4" x . ".
K. 002. e. Outer plate of bronze buckle with 3 rivets remaining. $" x
K. 002. f. Thin bronze plate, broken at one corner. One hole in one corner only. â" x 11"
K. 002. g. Thin bronze plate ; 2 rivets attached. â" x ".
K. 002. h. Thin bronze plate ; inner plate of buckle. 4 rivet holes. z." x â".
K. 002. 1. Fragment of thin bronze plate. Portion of rivet hole visible. A" x
'7 See above, pp. 285 sqq.