254 THE RUINS OF DANDAN-UILIQ [Chap. IX
which first suggested to me, on closer study of the photograph shown enlarged in Plate II, that the strange subject of the fresco might possibly be explained by the legend which Hsüan-tsang relates of the Naga's widow residing in a stream east of the Khotan capital and her wooing by the self-sacrificing minister. We have already had occasion to consider the legend in detail in connexion with the ` Drum-lake Convent' 12. Seeing that without exception the figures or scenes appearing in the decoration of the shrines excavated by me at this and other sites, whether sculptured or painted, bore a religious or mythological character, the presumption seems a ftriori justified that some legendary scene was intended here also. As far as my knowledge of Buddhist iconography goes, there is no legend affording as suitable an interpretation for our fresco as the one of the Naga's widow. An observation of local archaeology comes to support the conjecture. The popularity of two other Khotan legends connected with particular sites, those of the sacred rats and of the silk-bringing princess, is attested beyond all doubt by pictorial representations found at Dandan-Uiliq. Hence it is natural to look in the first place among the stories of sacred local lore as recorded by Hsüan-tsang for a clue to this otherwise unexplained scene.
Figures from Starting then from the figure of the riderless horse, it is clear that it would be most
legend of the appropriately accounted for by the legend which represented the minister's horse as the bringer Naga's
widow. of his message and miraculous gift after his own disappearance. The minister himself might
be recognized both in the swimming figure with head and shoulders just visible above the water, and in the small nude male which tries to rise by holding to the side of the woman. Such repetition for the purpose of indicating successive stages of a legend is a device as well known to old Indian as to mediaeval art of the West. The disproportion in size between the female and the male figures is another feature easily explained on the basis of the suggested interpretation. The divine Nagini would, in accordance with a convention which Gandhara art borrowed from the declining antique, be necessarily shown far larger than the mortal wooing her 13. Finally, we may reasonably expect to see the watery home of the Nagini indicated in the painting by a regularly enclosed tank, seeing that the same manner of representation is ordinarily resorted to by the Gandhara sculptors when their reliefs have to show the dwelling places of Nagas 14.
In the face of these features supporting the identification, the difficulties presented by two negative ones should not be ignored. In view of Hsüan-tsang's story, it is strange to miss in a painted representation of it the miraculous drum with which the minister's horse was supposed to have returned, and neither in the original fresco nor in its photograph could I trace any object resembling this. It is further noteworthy that the figure of the woman, however richly adorned, bears no indication of her character as a Nagini or semi-divine being. In Gandhara sculpture a snake-hood above the human head would have been ordinarily resorted to to mark the Naga's consort", while even if this characteristic emblem was unknown to the Buddhist art of Khotan, there was at least the aureole conveniently at the artist's hand to mark superhuman origin. It must, however, be borne in mind that Kashmir tradition knew Nagas in a purely human shape 16. It may also be doubted whether the legend, as heard by Hsüan-tsang, reckoned the Nagini among divine beings proper, seeing that it related her first husband's death.
Suggestions However this may be, it is clear that the suggested identification of the scene must be
models. 12 See above, p. 227. 15 Comp. Foucher, L'Arl du Gandhdra, i. p. 384; Grün-
is See Grünwedel-Burgess, Buddhist art, p. 138. wedel-Burgess, pp. 1o6 sq.
14 Comp. Foucher, L' Art du Gandhdra, i. p. 388 ; Figs. 18 See my note, Rajal. i. 3o ; also i. 220.
194-196; Grünwedel-Burgess, loc. cit., Figs. 57, 59.