As the excavations were extended further towards the centre of the circumvallated area and along the north face (see Figs. 117, iz8), they disclosed an agglomeration of hovels, many of them very small and all built with even greater irregularity in ground-plan and floor-level than those found on the east side. As the walls were mostly quite thin here, built of mere clay or the coarsest brickwork, and in addition had suffered much from erosion, the disposition of the various hovels and small cellar-like apartments was as difficult to trace as that of a rabbit warren. As work had to be carried on in different places simultaneously and finds continued in rapid succession, my attention was necessarily distracted, and the record of structural remains, kept and entered in the plan, could not aim at full precision, and had in part to be left schematic. The striking irregularity in the levels observed among the broken walls in the centre of the area of the fort and towards the foot of the great south bastion suggested that these were remains of structures erected at different periods, one above the débris of another. The rooms xl, xlii,, xliv yielded a considerable number of documents.
The rooms unearthed near the north wall were somewhat less confusing in arrangement, and as they were often in better preservation, it was possible to make quite sure that most of the small ones possessed no entrances, and had evidently been approached from above. Some with bottoms four or five feet deeper than the rest had probably served as places of storage for the supplies and chattels of the motley collection of soldiers and petty officials whom the stronghold sheltered at different periods. Such use would account for the curious platforms built across the corners in some of these underground rooms, e. g. xxvi. Sheaves of wheat, barley, and millet mixed plentifully with the sand and refuse in some of them,2 while written remains were wholly absent. Other rooms, like xxiv, xxvii, xxviii, seemed to have served as modest living quarters, and in these a fair number of Tibetan records was found near the floor. From a row of small compartments also, cut into the north rampart, near xxxi, a few Tibetan fragments were recovered. But a far more interesting discovery rewarded the clearing of the modest apartment xxxii, where, on the floor and under a cover of only two feet of refuse and sand, there were brought to light three fairly large fragments of paper in Runic Turkish, the only finds in that script from any of my sites in the Tarim Basin. I shall have to recur to them further on. To the west of xxxii the protecting cover of stable refuse thinned out, and erosion had left but the scantiest traces of other small structures which had once extended towards the fort gate. The cluster of mud hovels near the centre of the area of the fort would probably have shared the same fate if a consolidated crust of sheep-dung, deposited over their remains at some period, had not provided protection.
Among the relics of archaeological interest which had found a safe refuge in the rich layers of refuse preserved within the ruined fort, Tibetan records on wood and paper were by far the most abundant. Not counting pieces which from various causes have become illegible, the total number of such documents amounts to over a thousand. Complete effacement of the writing in many instances was not to be wondered at, seeing the close association of these remains with the amount of decayed animal and vegetable matter which had also found a resting-place in the walled-in dustbins. It often encrusted the written records so thickly that it required quite as much care to discover and extract them as subsequently to clean them.
In the case of both classes of documents outward indications had made it probable from the first that the majority of these Tibetan records belonged to official or private correspondence. A considerable proportion of the narrow slip-like tablets, to which form practically the whole of the wooden stationery at this site was confined, showed at the left end of their obverse raised oblong seal sockets, as seen in M. i. vii. 35; viii. 22 (Plate CLxxI). These often still retained grey clay from the seal impressions once filling them. The notches generally observed in the longer side rims of
I For specimens see M.1. xxvii—xxviii. ooi ; for other specimens of grain see M. ow, I. oo6o in List below.