Sec. i] EXPEDITION TO THE KARA-DONG RUINS 449
character of the remains indicated a relatively large but roughly-built dwelling. The very fact of its survival suggested that what structures may have once existed near by must have been of a still less substantial type.
It seems to me very significant that here, too, just as in all the structures of the ruined quadrangle, only Toghrak was used for the timber-work. This species of wild poplar grows plentifully, as we have seen, in the jungles of all the rivers which lose themselves in the sand ; it manages to survive even in desert tracts which surface water never reaches now. But its twisted knotty trunks and branches by no means furnish as good a building material as the Terek or white poplar, the Jigda, and other trees planted in cultivated areas. At DandanUiliq, at the Niya and Endere Sites, only timber of these latter trees appears to have been used for the framework of houses, and this accordingly bore there a far more finished look than at Kara-dong. At those other old sites the dead trunks of Terek and other trees depending on systematic irrigation formed a conspicuous feature. But around Kara-dong I looked for them in vain. Dead trees rising from between the sand-dunes were plentiful on the ground close to the east and south of the main ruin, but they were all old Toghraks or tamarisks, such as are still found growing luxuriantly in a broad belt west of the relatively recent riverbed of Toldama. I think it justifiable to conclude from this observation that cultivation could not have existed to any considerable extent in the vicinity of the Kara-dong site at the period from which its buildings date.
What then can have been the purpose of the great ruined quadrangle, situated, as it evidently was, in the forest land between the desert and the river, and contrasting so strikingly by its size with the modest character and limited extent of the other remains ? Keeping in view the position and the peculiar plan of the structure, I think the suggestion may be hazarded that we have here the ruin of an ancient ` Langar ' or roadside Sarai, built primarily to afford accommodation for a large number of people, yet capable of defence, too, if the need should arise. Such a suggestion of course presupposes traffic along the Keriya Darya at an early period, and for the existence of such, I believe, adequate historical and topographical evidence can be adduced.
Mirza Haidar, when speaking of the rivers which empty their waters into the great lake of the Turkestan desert, i. e. the Lop-Nor, distinctly mentions the Keriya Darya along with ` the river of Yarkand, the Ak-Kash (i. e. the Yurung-kash), and the Kira-Kish ' and ` the Chârchân' s. Seeing how thoroughly well acquainted from personal experience the Moghul historian was with the Khotan region, and how exact his description of it has proved to be, it seems very unlikely that he would have made this statement unless the Keriya river was still in his days believed to reach the Tarim. Also in the Turki legendary of Mahmûd Karam Kabuli, which purports to relate events of the twelfth century A. D., a force of Muhammadan warriors, coming from the conquest of Ak-su, is described as attacking the infidels in the vicinity of Keriya by the route of the Keriya river 10.
Whatever the historical value of this tradition may be, M. Grenard was fully justified in quoting it in connexion with the plan which an energetic Amban of Keriya formed in 1893 for
See Tärikh-i-Rashidi, p. 406. The text calls the lake Xuk Naur, i. e. Koko-Nor, but the description given of it shows beyond all doubt that Mirza Haidar means the Lop-Nor. Subsequently he refers to ` the Kara Murân of Khit£i,' i. e. the Huang-ho issuing from one end of the lake,—a reflex, as Mr. Ney Elias duly recognized, of the ancient
Chinese legend which supposes a subterraneous connexion between the Lop-Nor and the Huang-ho (see, e. g., Rémusat, Ville de Kholan, p. 2). This traditional belief accounts for the confusion of the two names.
10 See Grenard, Mission D. de Rhin:, iii, p. 44.