Finds at Within the area enclosed by the wall on the top of the Mesa, pottery debris was abundant,
T. XXIII. c. clear evidence that the station had been tenanted by a considerable number of men and for a long time. From a layer of refuse between the western knoll and that occupied by the tower I extracted, lying almost on the surface, the well-preserved Chinese record on wood, T. xxiii. c. 4. Eleven more wooden slips were subsequently discovered here and among the rubbish lying close to the tower. Among the miscellaneous small objects recovered may be mentioned the fragment of a wooden bowl decorated in lacquer, T. xxiii. c. oz ; the piece of an iron hoe-blade, 04 (Pl. XLVII) ; and two bronze arrow-heads of the ` regulation ' type of Han times, 06-7 (Pl. XLVII). [Among the records recovered T. xxiii. c. 4 has proved to contain the draft of a private letter. T. xxiii. c. 016 indicates the exact place of origin in distant Ho-nan of a soldier who has died. In T. xxiii. c. 022
the local name of Po-hu 4A occurs again.24 T. xxiii. c. 023 contains the fragment of a calendar from the year 4 B. c.]
Position of At a distance of about a mile and a half to the north could be seen the watch-tower T. xxru. e,
advanced situated in advance of the Limes line. The lake shore which it guarded was masked from direct tower
T. xxiii. e. observation from T. xxiii. c by a gravel-covered ridge. Owing to the boggy condition of the
ground it proved impossible at the time to reach this tower. It evidently served a purpose similar to that which accounted for the placing of the advanced post T. ix. a, thrown out beyond the line of the westernmost Limes.25 It was needed for the better protection of an important point where this section of the Limes rested its flank on the Khara-nôr and moreover, as the map shows, formed a projecting angle. The presence of this angle itself was probably due to the desire to take advantage of the very wide outlook commanded by the height of T. XXIII. c. This extended far away to the north-east, to the point where the Su-to-ho, joined by the deltaic branches of the Tang-ho, enters the Khara-nôr. It ranged also over the bare plain to the east where the salt-encrusted depression coming from Chien-ch`üan-tztl is fringed by a wind-eroded area with rows of low Ydrdangs.
Remains From T. XXIII. c the line of wall, now decayed into a low mound, ran to the south-east towards
of watch- a far-stretching Mesa about a hundred feet high and less than a mile off. The ground, salt-tower
T. xxIII. d. encrusted throughout and boggy, could scarcely have supported us but for the firmer track offered
by the mound marking the Limes line. On the top of the Mesa a completely decayed mound marked the remains of the tower T. xxiit. d. Around it pottery debris was plentiful. The specimens collected here and described in the List below comprise pieces of glazed stoneware such as those which had been found by me at T. xxix.26 Mr. R. L. Hobson, of the British Museum, attributes these partly to 'rang and partly to Sung times. The evidence of later occupation of the place which these pieces afford is borne out by four fragments of porcelain, T. xxiii. d. 010-13, which were also found here, three belonging to the same vessel and all roughly painted in blue.27 What special reason accounts for this later occupation I am unable to indicate, as no route likely to have remained in continued use leads past this point.
Tower At T. xxiii. d the line of the wall took a turn to the ENE. and at a distance of nearly a mile
anT. d XXIII. fell. brought us to a narrow isolated clay ridge (see plan in Pl. 14) running east and west and about its
a hundred yards long. Its summit, rising about 35 feet above the adjoining ground and only 25 feet across at its widest, bore the tower T. xxiii. f (Fig. 202). This was built of lumps of clay, with layers of thin Toghrak branches inserted to reinforce the masonry. It measured 14 feet at the base and still rose to a height of about i6 feet. No other structural remains were traceable
24 See above, pp. 345 sq.
25 See Serindia, ii. p. 662 ; Map No. 35. D. 4.
26 Cf. ibid., ii. pp. boo sq., 788 sq.
27 For finds of porcelain apparently similar in type at T. xi, see ibid., ii, pp. 667, 773.