With the remarkable exception of the section on Dùn Beag, to be immediately referred to, no trace of any eroded channel of this river through the lavas of the great volcanic plain has been preserved. Possibly frequent invasions of its bed by streams of basalt from different vents hindered it from remaining long enough in one course to erode anything like a gorge or canon. But, in any case, the main channel of the river probably lay rather to the east of the present islands of Canna and Sanday, on ground which is now covered by the sea. The banks or sheets of boulder-conglomerate undoubtedly show where its current swept with great force over the lava-plain, but the manner in which these coarser materials are so often covered with fine silt suggests that the sedimentary materials now visible were rather deposited on the low grounds over which the steam rushed in times of flood. Pools of water would often be left after such inundations, and in these depressions silt would gradually accumulate, partly carried in suspension by the river, partly washed in by rain, while drift-wood that found its way into these eddies, and leaves blown into them from the trees and shrubs of the surrounding country, would remain for some time afloat and would be the last of the detritus to sink to the bottom. Hence, no doubt, the carbonaceous character of the hardened silt in the upper part of each intercalation of sediment. If we were to look upon the volcanic materials in the conglomerates as derived from the subærial disintegration of the fields of basalt, we should be compelled to admit a very large amount of erosion of the surface of the volcanic plain during the period when the river flowed over that tract. It would be necessary to suppose not only that there was a considerable rainfall, but that the differences of temperature, either from day to night, or from summer to winter, were so great as to split up the lavas at the surface, in order to provide the river with the blocks which it has rolled into rounded boulders. I do not think, however, that such a deduction would be sound. If we compare the materials that have filled up the large eruptive vent at the east end of Canna (to be afterwards described) with the great majority of the blocks in the coarse conglomerates, we cannot fail to note their strong resemblance. The abundance of lumps of slaggy lava in the river-shingle corresponds with their predominance in the agglomerate of the vent. The boulders of basalt, dolerite, and andesite which crowd the conglomerates need not have been derived from the action of atmospheric waste on the lava-fields, but might quite well have been mainly supplied by the demolition of volcanic cones of fragmental materials. [Illustration: Fig. 272.--View of the Dùn Beag, Sanday, seen from the south. (From a Photograph by Miss Thom.)]