It is, I think, tolerably clear that the wall of basalt against which this conglomerate has been laid down is one of erosion. The beds of basalt have here been trenched by some agent which has likewise scooped out the soft underlying shales, and even cut them away from under their protecting cover of basalt. There can be little hesitation in regarding this agent as a water-course, which for some considerable interval of time continued to dig its channel through the hard basalts. There is not room enough between the basalt-wall of Dùn Beag and the opposite cliffs of the shore (where no trace of this conglomerate is to be seen) for any large stream to have found its way. I do not therefore seek to identify this relic of an ancient waterway with the channel of the main river which deposited the conglomerate bands of Canna and Sanday. More probably it was either a mere torrential chasm, or a tributary stream draining a certain part of the volcanic plateau and allowed to retain its channel long enough to be able to erode it to a depth of nearly 50 feet. Erosion had reached down through the underlying tuffs to the slaggy basalt below, but before it had made any progress in that sheet its operations were brought to an end at this locality by the floods that swept in the coarse shingle, and by the subsequent stream of basalt of which a mere outlying fragment now forms the upper third of the stack (_e_, Fig. 274). That the ravine or gully of Dùn Beag probably lay within the reach of the floods of the main river, may be inferred from the number and size of the far-transported rocks in its conglomerate. It was filled up gradually, but the conditions of deposition remained little changed during the process, except that the largest blocks of rock were swept into the chasm in the earlier part of its history, while much smaller and more water-worn shingle were introduced towards the close. Denudation, which has performed such marvels in the topography of the West of Scotland since older Tertiary time, has here obliterated every trace of this ancient gully, save the little fragment of one of the walls which survives in the stack of Dùn Beag. When in the course of centuries this picturesque obelisk shall have yielded to the action of the elements, the last leaflet of one of the most interesting chapters in the geological history of the Inner Hebrides will have been destroyed. The question naturally arises--What was the subsequent history of the river which has left so many records of its floods entombed among the basalts of Canna and Sanday? In particular, can any connection be traced or plausibly conjectured between it and the river-bed preserved under the Scuir of Eigg? To this question I shall return after the evidence for the existence and date of the latter stream has been laid before the reader.