Sec. ii] EXCAVATIONS AROUND AND WITHIN TANG FORT OF ENDERE 277
In discussing the Endere Site in my former Report, I called attention to the interesting fact that Hsüan-tsang, who passed along the desert route from Ni jana, or Niya, to CM-ma-ito-na, i.e. the present Charchan, on his return journey to China about A.D. 645, found no inhabited place on the ten days' march.6 But after describing the great desert of drifting sands ', which he entered east of Niya, in terms closely recalling Marco Polo's account of the great desert between Lop-neir and Sha-chou, the pilgrim arrived, about four marches after leaving Niya and six from Charchan, at ruins of abandoned settlements which the tradition of his time ascribed to the ` ancient Tu-huo-lo kingdom '. This country, he tells us, had long been deserted, and all its towns were unoccupied wastes.6
The topographical evidence derived from the distances which Hsüan-tsang records from Niya and to Charchan respectively, and which I have discussed already in my former Report, made it quite certain that the remains of the abandoned settlements which Hsüan-tsang saw were to be looked for in the close vicinity, if not in the very locality of the ruined fort which I had explored in 19oI.7 Yet seeing that about A.D. 645 this tract was already a waste abandoned to the desert, it had then seemed rather puzzling how to account for the existence in it of the ruins excavated by me, which were undoubtedly occupied during the early part of the eighth century. But now as soon as I had verified Sadak's find by the discovery of more Kharosthi tablets in the same ruined structure, and under my own eyes, I felt sure of the right explanation. Clearly here we had definite archaeological evidence of an old site in the desert having been reoccupied after the lapse of centuries—and a fresh instance of the often proved accuracy of Hsüan-tsang's topographical statements. The Kharosthi records on wood now brought to light are shown by their palaeographic character to date from the same period as those of the Niya Site which, as we have seen, was abandoned towards the end of the third century A.D. Hence the conclusion is obvious that the small house, E. vi, yielding these tablets must have belonged to the earlier settlement which Hsüan-tsang found completely deserted and in ruins.
That the area must have subsequently come under occupation again, probably in consequence of the improved conditions and increased traffic eastwards which followed the establishment of Chinese control over the Tarim Basin within a little over ten years after Hsüan-tsang's passage,8 is proved by the ruined fort ; for my excavations of 1901 showed that this served for a Chinese garrison at the commencement of the eighth century A.D. But even the condition of the earlier structure, E. vi itself, testified to this reoccupation ; only thus did it seem possible to account for the layers of straw, plentifully mixed with wheat stalks and grains (for specimens, see E. vi. i. 001), and of stable refuse which extended uniformly both over the top of the broken walls, and the débris of bricks filling the rooms between. Evidently when the site was resettled in the second half of the seventh century some one had found it convenient to erect his homestead or stable over the mound formed by the tumble-down ruin, just as near Domoko, on ground reconquered from the desert, I had seen the tops of tamarisk-cones utilized as sites for the new settlers' dwellings.s
The search, which was continued on November 9 in the vicinity of the Tang fort, revealed
8 Cf. Ancient Kholan, i. p. 435.
8 See Julien, Mémoires, ii. pp. 246 sq. ; \\ratters, Yuan Chwang, ii. p. 305 ; Rémusat, Ville de Kholan, p. 65.
7 With reference to the distances indicated in Anciens Kholan, i. p. 435, note 11, I may add now that the road distance from the Endere fort to Charchan, as determined by actual measurement with the cyclometer, amounted to 115 miles. The length of route followed by me in 1901 between the Endere Site and Niya was about 98 miles. In comparing these distances with the six and four marches respectively,
which Hsüan-tsang reckons, it must be borne in mind that the caravan track from the Endere River to Charchan lies along an almost straight line, while the route I followed in 1901 from Endere to Niya winds considerably as the map in Ancient Kholan shows. In Hsüan-tsang's time a more direct route might have been in use. The direct distance from the Endere fort to Niya as measured on my map is 86 miles.
8 Cf. Ancient Kholan, i. p. 59.
° See ibid., i. p. 454 ; Desert Cathay, i. p. 252.