tempera painting which help us to form some idea of what the decoration of the higher parts of the wall may have been like. The largest and best-preserved of these fragments were found leaning in three closely packed layers against that part of the wall which retained the lunettes iv—vi of the dado. The photograph in Fig. 127 shows them cleared of smaller débris, but before removal. Their preservation was very probably due to their having slid down at a time when sufficient sand and soft débris had already accumulated within the passage to stop the fall of the pieces of fresco as they broke loose from the wall through one or another cause. It seemed reasonable to assume that the innermost panel (now seen in M. III. 003, Plate XLII) was the first to be stopped in its fall, and that it represented a portion of the frieze nearest to the dado. It is scarcely necessary to point out that the wall from which these painted plaster panels had slipped must have remained standing for some time after ; otherwise its fall would have been bound to crush these frail pieces of plaster into dust at its foot. It was probably débris from the vaulting which helped to bury them safely before the wall, too, fell.
That the decoration of the wall above the dado must have included more than one painted frieze can be safely concluded, not merely from the analogy of the fresco friezes subsequently discovered in the neighbouring rotunda M. v, but also from actually surviving fragments which clearly prove the existence of at least two friezes in M. III. Along the top of the large panel M. III. 003 (Plate XLII), which we shall have to discuss presently, there runs a black band with the remains of a grey one above it. Now in the fragment M. III. oo18 (Plate xLIV) we find the foot of a human figure resting on a band of white which itself is succeeded below by a grey and then a black one. M. III. 0036 (Plate xLV), too, shows a white and then a grey band below the feet of two figures, the edge further down being broken. It is clear that in all ,three pieces we have remnants of the same triplicate band which divided two upper friezes.
In my Personal Narrative I have fully described the difficult operations by which I succeeded, at the expense of much care and personal exertion, in rescuing the broken pieces of friable painted plaster once belonging to these fresco friezes and in packing them so safely that they subsequently survived, without further appreciable damage, all the risks involved in transporting them over thousands of miles across deserts and high mountain ranges.6 I had entertained little hope at the time that these brittle panes of mud plaster could be brought to safety, over such a distance and such ground, in a condition still permitting of careful arrangement and study. I had all the more reason to feel gratified at the result of my efforts when, three years later, Mr. F. H. Andrews, with the help of my second assistant, Mr. Droop, was able to put together from these disjecta membra panels so large and well preserved as those shown in Plates XLII and XLIII, besides a considerable number of smaller ones, some of which are illustrated in Plates XLIV and XLV.
The successful rescue of these fragments has enabled expert eyes, such as those of Mr. F. H. Andrews and M. A. Foucher, to observe many points which are of interest for the history of Buddhist art in Central Asia, and of which my notes and photographs, even if taken under less hampering conditions, could not possibly have preserved an adequate record. All the same, I should not attempt to deal at once with these scanty remnants of the painted friezes, were it not for the definite evidence which the exactly analogous fresco frieze discovered in shrine M. v furnishes as to the general scheme of the composition, and did not the subjects treated in them point so plainly to the connexion with Graeco-Buddhist art, as known to us from the sculptures of Gandhara. If we previously examine these frieze fragments, it will be easier for us afterwards to make sure of the true descent and significance of the fascinating ` angels ' in the dado which might otherwise puzzle us.
s See Desert Cathay, i. pp. 463 sqq.