occupied the cella wall facing the entrance. This passage at the back was vaulted, and might be decorated with frescoes or relievo statuary. The same disposition of the ground-plan prevails also in the cave-temples of the site.°
A third type, represented mainly among the larger shrines, corresponds closely to the one which is most common among the Buddhist sanctuaries of the Khotan sites (Dandan-oilik, Khadalik, etc.). Here we find a cella, usually square, approached through an antechapel and enclosed on the remaining three sides by a passage which communicates with the latter and served for the ' Pradaksina'. In the larger shrines of this type, e. g. x—xii, xxv, xxvi (Plates 52, 53), the passage widens somewhat oat the back of the cella and forms a chamber adorned with relievo statuary. Elsewhere the passages appear to have been decorated only with wall-paintings. The fourth and fifth types are confined to monuments for which a funerary character may be assumed with the greatest probability.
The fourth type in its external appearance presents itself as a Stûpa, with a cylindrical dome rising either direct from a low platform or from a base of varying shape which may be circular, polygonal, or square. Usually there is a square walled enclosure. The sketch-plans in Plates 52, 53 and Figs. 287, 288 will help to illustrate this type. The peculiarity of these ` Stûpas' is that their interior is invariably hollow, and that those in fair preservation still have an entrance by which the vaulted interior could be reached. In none of the vaulted circular chambers examined or cleared did I find either cult objects or human remains. And yet it is certain, both from their association with the monuments of the fifth type and from the exact analogy of the ` Stupas' found at the cemetery of ` Kosh-gumbaz ', outside the ruined town of Kara-khôja, which Professor Grünwedel has examined and described in some detail,' that they served a sepulchral purpose. This was definitely proved by finds of cinerary urns and boxes at the foot of the pillar-like structures (Figs. 28o, 288) which belong to the fifth type. Whether square or polygonal in shape, they have each a small walled enclosure, and taper in their superstructure. Their appearance is curiously reminiscent of Roman funeral monuments.8
Before I proceed to describe the results of the excavations carried on at different ruins I may conveniently record certain observations concerning the conditions in which I found them and which affect all remains of the site. At the first glance it was easy to notice that all the exposed portions of the ruins had suffered much from the destructive effects of rain and snow. I have already had occasion to refer to the climatic conditions of the Kara-shahr valley as being distinctly less arid than those prevailing in the central portions of the Tarim Basin or in its much reduced counterpart, the Turfan depression. Of the heavy rain-storms reported to visit the valley on not infrequent occasions during the summer 9 there was clear archaeological evidence in the almost complete effacement of the relievo decoration in plaster which the outside walls of some of the funeral structures appear to have borne at one time. The plain. plaster coating of temple walls, etc., has similarly been washed down almost everywhere.
The position of the ruins on sandstone terraces,' which, though low, are well above the level affected by the subsoil moisture of the surrounding plain, had fortunately saved them from the disintegrating effects of shör, so strikingly illustrated by the remains of other Kara-shahr sites. Yet, situated as they are within six miles of Baghdad-shahri and even less of the nearest shore of Lake Baghrash, the ruins could not escape the slow but constant deletory action of the atmospheric
8 See Grünwedel, Allbuddh. Kultslälten, pp. 895 sqq.;
Figs. 449, 454, 458.
' Cf. Grünwedel, Idikulschari, pp. x io sqq. ;' Allbuddh. Kullslälten, pp. 336 sqq.
8 For similar structures at the ' Kosh-gumbaz' of Kara
khbja, cf. Grünwedel, Idikutschari, Figs. 805, 8o8. In these Professor Grünwedel is prepared to recognize distant imitations of Roman funeral monuments ; cf. A llbuddh. Kullslälten, P. 336.
9 Cf. Grünwedel, Allbuddh. Kullslälten, pp. 192, 206.