1. The large size and rolled shape of the boulders in the conglomerates. To move blocks several tons in weight, and not only to move them but to wear them into more or less rounded forms, must have required the operation of strong currents of water. The coarse detritus intercalated among the basalts is quite comparable to the shingle of a modern river, which descends with rapidity and in ample volume from a range of hills. 2. The evidence that the materials of the conglomerates are not entirely local, but include a marked proportion of foreign stones. The proofs of transport are admirably exhibited by pieces of Torridon Sandstone, epidotic grit, quartzite, and other hard rocks none of which occur _in situ_ except at some distance from Canna. These stones are often not merely rounded, but so well smoothed and polished as to show that they must have been rolled along for some considerable time in water. 3. The lenticular character and rapid lithological variations of the strata, both laterally and vertically. The coarse conglomerates die out as they are followed along their outcrop and pass into finer sediment. They seem to occur in irregular banks, which may not be more than 200 feet broad, like the shingle-banks of a river. The coarser sediment generally lies in the lower part of the sedimentary group. But cases may be observed, such as that shown in Fig. 269, where fine sediment, laid down upon the bottom conglomerate, has subsequently been overspread by another inroad of coarse shingle. Such alternations are not difficult to understand if they are looked upon as indicating the successive floods and quieter intervals of a river. For these reasons I regard the platforms of sedimentary materials intercalated among the basalts of Canna and Sanday as the successive flood-plains of a river which, like the rivers that traverse the lava-deserts of Iceland, flowed perhaps in many separate channels across the basalt-fields of the Inner Hebrides, and was liable to have its course shifted from time to time by fresh volcanic eruptions. That this river came from the east or north-east and had its source among the Western Highlands of Inverness-shire, may be inferred from the nature of the stones which it has carried for 30 miles or more along its bed. And that it crossed in its course the tract of Torridon Sandstone, of which a portion still remains in Rum, is manifest from the abundance of the fragments of that formation in the conglomerates.