The Lo-los hold thousands of Chinese in slavery; and more are added yearly to the number.”—­H.C.] Two routes run from Ch’eng-tu fu to Yun-nan; these fork at Ya-chau and thenceforward are entirely separated by this barrier.  To the east of it is the route which descends the Min River to Siu-chau, and then passes by Chao-tong and Tong-chuan to Yun-nan fu:  to the west of the barrier is a route leading through Kien-ch’ang to Ta-li fu, but throwing off a branch from Ning-yuan southward in the direction of Yun-nan fu.

This road from Ch’eng-tu fu to Ta-li by Ya-chau and Ning-yuan appears to be that by which the greater part of the goods for Bhamo and Ava used to travel before the recent Mahomedan rebellion; it is almost certainly the road by which Kublai, in 1253, during the reign of his brother Mangku Kaan, advanced to the conquest of Ta-li, then the head of an independent kingdom in Western Yun-nan.  As far as Ts’ing-k’i hien, 3 marches beyond Ya-chau, this route coincides with the great Tibet road by Ta-t’sien lu and Bathang to L’hasa, and then it diverges to the left.

We may now say without hesitation that by this road Marco travelled.  His Tibet commences with the mountain region near Ya-chau; his 20 days’ journey through a devastated and dispeopled tract is the journey to Ning-yuan fu.  Even now, from Ts’ing-k’i onwards for several days, not a single inhabited place is seen.  The official route from Ya-chau to Ning-yuan lays down 13 stages, but it generally takes from 15 to 18 days.  Polo, whose journeys seem often to have been shorter than the modern average,[2] took 20.  On descending from the highlands he comes once more into a populated region, and enters the charming Valley of Kien-ch’ang.  This valley, with its capital near the upper extremity, its numerous towns and villages, its cassia, its spiced wine, and its termination southward on the River of the Golden Sands, is CAINDU.  The traveller’s road from Ningyuan to Yunnanfu probably lay through Hwei-li, and the Kin-sha Kiang would be crossed as already indicated, near its most southerly bend, and almost due north of Yun-nan fu. (See Richthofen as quoted at pp. 45-46.)

As regards the name of CAINDU or GHEINDU (as in G.T.), I think we may safely recognise in the last syllable the do which is so frequent a termination of Tibetan names (Amdo, Tsiamdo, etc.); whilst the Cain, as Baron Richthofen has pointed out, probably survives in the first part of the name Kienchang.

[Baber writes (pp. 80-81):  “Colonel Yule sees in the word Caindu a variation of ‘Chien-ch’ang,’ and supposes the syllable ‘du’ to be the same as the termination ‘du,’ ‘do,’ or ‘tu,’ so frequent in Tibetan names.  In such names, however, ‘do’ never means a district, but always a confluence, or a town near a confluence, as might almost be guessed from a map of Tibet....  Unsatisfied with Colonel Yule’s identification, I cast about for another, and thought