over 2,000 cubic feet which I had measured, less than a fortnight before, where the river debouches into the plain.3 It suggested that a very considerable portion of the supply of water then measured was derived from the several tributaries which the Muz-art river receives below Bai, and that the glaciers about Khan-tengri, which are the main sources of the river, had not yet begun to send down their summer flood. Moreover a good deal of the water actually available in the river, where it passes through the basin of Bai, was probably absorbed by the canals irrigating the belts of cultivation on either bank above and below Yaka-arik.
Pitching camp at the village of Jigdalik, I visited, on May 13th, the collection of cave-shrines locally known as ` Ming-oi'. They proved to be situated about 5 miles to the SSW., at the head of a narrow winding valley, which descends from the barren hill range dividing the basin of Bai from the desert plain north of the Tarim. Where the little valley widens to about 200 yards at the bottom, with steep cliffs of friable sandstone rising high on either side, we found three small springs close together. They issue amongst reeds, at the foot of a low terrace projecting from the rock wall on the west. Their water tasted perfectly fresh, though the little brook that they feed becomes brackish before it dies away a short distance lower down. The presence of drinkable water in this Thebais of absolutely bare ridges of much-eroded sandstone or shale accounts for the existence here of the numerous small caves, which once served as Buddhist shrines and monastic quarters. The very steep cliffs to the west contain at least six of these caves, along a front not exceeding a quarter of a mile. The somewhat lower spur to the east (Fig. 354) shows four close together, with several others hidden away in narrow side ravines or buried under detritus. The little ridge or terrace already mentioned, running across the bottom of the valley, appears to have once borne a series of Stûpas ; but of these even the foundations were difficult to trace, owing to the burrowings of treasure-seekers.
Of the caves on the western side, a small cella situated about 3o feet above the flat ground is the most accessible. As it contained a circumambulatory passage, it certainly had served as a shrine. The front had fallen in, and the interior was for the most part filled with hard stratified mud, which the rain had washed in. Five small niches in the central block of rock, and above the entrances to the side passages, retained traces of painting. About 200 yards farther south lies the main group of caves. The highest of these, seen on the extreme left of the photograph, and marked Jig. 1 in the sketch-plan (Pl. 44), was pointed out to me as that where Sahib `Ali, the Indian Ak-sakal of Kucha, guided by one of his local factotums, had dug up, about seven years (?) before, a large packet of manuscript leaves, which he had subsequently transmitted to Sir George Macartney. It is situated at a height of about 120 feet, and comprises, as the plan shows, a living apartment measuring about 12 feet by 14, provided with a window. It is entered through a passage 4 feet wide which communicates with it at the end by a doorway 3 feet wide. Grooves sunk into the floor and into the rock above prove that this was once closed by a wooden door. The northern wall of the living apartment holds a shallow recess for a fire-place, with five small receptacles above it intended to serve as cupboards.
The floor of the room was covered partly with a thick layer of dust and partly with refuse of straw and matting. Minute fragments from various paper manuscripts showing Brahmi characters of the Central-Asian Gupta type abounded among this refuse. I also recovered some twenty fragments of a similar character but larger, showing how roughly Sahib `All's clearing was done. It is probable that the cave had been searched before, and perhaps more than once, by people burrowing for treasure, with consequent damage to its manuscript remains. Half a dozen paper fragments of larger size were recovered from the passage, together with a very small piece each of
3 See above, ii. p. 8o8.