556 TO GUCHEN AND ACROSS THE T`IEN-SHAN [Chap. XVI
mediaeval times by Pei-t`ing, is but the reproduction of a far more ancient designation.5 Hsü Sung had also duly recognized the identity of Pei-t`ing with the town which, when in possession of the Western Turks, was known as the town of Kagan-stùpa ` the Khan's Stûpa ' (Ko-han fou-eu ) and under this name is mentioned also in Hsüan-tsang's Life.'
It is not within the scope of this work to show in detail that this location of Pei-t`ing is fully borne out by the very numerous passages of Tang texts in which this important centre figures ; M. Chavannes's work, the Documents sur les Tou-kiue occidentaux, has rendered them accessible to students who are not Sinologues. Still less can I attempt to discuss the question as to which particular five towns were supposed at different periods to be comprised in the popular territorial term of which the familiar Bésh-balik is the mediaeval Turki rendering. It must suffice here to point out that the Tang itinerary relating to the route via Pa-no-pea to Turfân brings us by its bearings and distances exactly to the ruined site identified by Hsü Sung with Pei-t`ing.
On the morning of October loth I started from Jimasa northward for Hup`u-tztic,'' the village beyond which the ruined town was said to be situated. The way to it led for about five miles through cultivated ground, crossed by a number of canals and deep-cut nullahs in which springs gather into small streams. It is the water from the latter—kara-su, to use the Turki term current in the oases along the foot of the K`un-lun—which irrigates the fields lower down, while higher up, near the town of Jimasa, irrigation is supplied by streams descending from the mountains. An abundance of elms and other trees along the boundaries of the fields suggested a fertile soil. But there were signs that in this tract cultivation had not yet recovered from the devastation attending the Tungan rebellion. Within the village of Hu-p`u-tzii, wholly Chinese and enclosed by a tumble-down clay wall, most of the houses were in ruin. Passing fields and groves for about another two miles, we arrived at a wide open area. There, to the west of a broad marshy nullah drained by a stream flowing northward rise the remains of the massive clay walls that once protected the ruined town.
The outer walls, as seen in the sketch-plan Pl. 23, appear to have once enclosed a roughly rectangular area, measuring approximately 2,16o yards from north to south and 1,26o yards from east to west. But the north-eastern portion of this circumvallation has completely disappeared, evidently through the eroding action of the stream which drains the marshy depression above mentioned and which here takes a turn to the north-west. Other portions also of the walls, both of the outer circumvallation and of an irregular-shaped inner enclosure, have suffered great decay. In places it was only in the course of plane-tabling that the connexion between the succession of detached clay mounds could be clearly traced. Judging from the best-preserved segments near the north-western corner the outer walls appear to have had a thickness of about 3o feet at their base and a height of over 20 feet, with bastions still more massive at the corners. The inner enclosure built against the east wall would seem to have had about the same strength and has suffered equally. Along portions of the outer enclosure cave-like rooms have been cut into the clay ; judging from their smoke-begrimed appearance they once served as dwellings. Those seen on the inside of the north-western corner bastion (Fig. 294) had apparently been recently searched. The floors of some others were cleared in our presence without any finds resulting.
The condition of far-advanced decay displayed by the walls, in places approaching to complete effacement, seemed to me to prove both that the site had long been abandoned and that there was
s Regarding the identity of Bésh-balik with Pei-t`ing, see Bretschneider, loc. cit., i. p. 66 ; ii. pp. 27 sqq. See also Ritter, Asien, i, pp. 382 sqq., for mediaeval notices, taken from A. Rémusat and Klaproth. The location of Bésh-balik
at Urumchi, accepted by Klaproth, Mémoires relatifs à l'Asie, ii. 355 sqq., has proved erroneous.
6 Cf. Chavannes, Turcs occid., p. 193.
6a The spelling Huo-p`u-tzü in Map No. 28. c.1 is erroneous.