Sec. i] FROM AN-HSI TO HAMI ; HSÜAN-TSANG'S DESERT CROSSING 1145
he feels refreshed by a cool breeze and finds rest in short sleep. A divine vision seen in his dream urges him to move ahead. After about io li his horse, which also had managed to get on its legs afresh, suddenly turns into another direction and after a few more li carries him to a plot of green pasture. Having allowed his horse to graze he is about to move on when he discovers a pool of clear water and feels saved. After a day's halt at this spot he continues his journey with a fresh supply of water and fodder, and emerging from the desert arrives at I-wu.
If we compare the account of Hsiian-tsang's desert crossing here briefly summarized with the actual topography of the route from An-hsi to Hâmi, as set forth above in its main features, we cannot fail to recognize their close accord in essential points, as well as an apparent lacuna in the text of the Life as at present available. This makes the pilgrim proceed in a single march from the first signal-tower to the fourth. But this is clearly in contradiction to the plain statement contained in a previously quoted passage of the Life, which reproduces the information given to Hsiian-tsang at Kua-chou concerning the route : 12 ` To the north-west, beyond this barrier, there are five signal-towers where the guards entrusted with keeping the look-out reside. They are a hundred li apart, one from the other.' We are thus led to assume that Hsiian-tsang in reality had to cover four marches from the river before reaching the fourth tower, and that in the narrative
presented by the extant text two of these marches have been left unrecorded.
Once allowing for this lacuna, of which the exact explanation cannot now be traced, but which unfortunately has its only too frequent counterparts elsewhere in the Life, we find it easy to reconcile the information recorded about the stages and incidents of the desert journey with the topographical facts. That the position indicated for the first signal-tower clearly points to the present Pei-tan-tzû, the first stage from An-hsi, has been shown above.13 The 480 li reckoned from the Su-lo Ho to the fifth signal-tower are in remarkably exact agreement with the 96 miles marching distance recorded by cyclometer on our journey between the river and Hsing-hsing-hsia, the fifth halting-place on the present road. The statement about the dreaded Mo-ho-yen desert extending beyond the fifth signal-tower is in perfect accord with the change which the character of the ground undergoes after leaving Hsing-hsing-hsia. Nor is there any difficulty about showing that all the matter-of-fact indications which can be gathered from the narrative about Hsüang-tsang's crossing of this desert are consistent with what the map indicates.
We are told that the traveller, having been advised to avoid the fifth signal-tower, i. e. H singhsing-hsia, turned off from the main route at the fourth tower in order to reach the ` Spring of the Wild Horses ', at a distance of a hundred li. When he failed to find this and thought of regaining the fourth tower, he is said to have turned back to the east.'* This makes it quite clear that the Yeh-ma-ch`üan spring to which he had been directed must have lain in a westerly direction. Now a reference to the Russian Trans-frontier map shows that the route from Tun-huang to Hâmi, as surveyed by Captain Roborovsky's expedition, passes at a distance of about 3o miles west of Ma-lien-ching-tzû before joining the An-hsi—Hami road at K`u-shui,'b and that one of its halting-places with water is to be found at about that distance to the west-north-west of Ma-lien-ching-tzû. Thus the existence, in the past or present, of a spring approximately in the position assumed for the Yeh-ma-ch`üan which Hsiian-tsang vainly sought for becomes plausible enough. That the pilgrim unguided failed to find it is an experience with which I became only too often and painfully
12 See above, p. 1097 ; Julien, Vie, p. 17.
'g Cf. above, p. 7098. This form of the name, as communicated to me by Mr. Li at An-hsi, seems more correct than the Pi-ling-tell of Map No. 81. c. i.
" See Julien, Vie, p. 29.