The other characteristic peculiarity of these _roches moutonnées_ consists in the direction of the glacier-scratches, which ascend the slope to its summit in a direct line on one side, while they deviate to the right and left on the other sides of the knoll, more or less obliquely according to its steepness. Occasionally, large boulders may be found perched on the very summit of such prominences. Their position is inexplicable by the supposition of currents as the cause of their transportation. Any current strong enough to carry a boulder to such a height would of course sweep it on with it. This phenomenon finds, however, an easy explanation in the glacial theory. The thickness of such a sheet of ice is of course less above such a hill or mound than over the lower levels adjoining it. Not only will the ice melt, therefore, more readily at this spot, but, as ice is transparent to heat, the summit of the prominence will become warmed by the rays of the sun, and will itself facilitate the melting of the ice above it. On the breaking up of the ice, therefore, such a spot will be the first to yield, and allow the boulders carried on the back of the glacier to fall into the hollow thus formed, where they will rest upon the projecting rock left uncovered. This is no theoretical explanation; there are such cases in Switzerland, where holes in the ice are formed immediately above the summit of hills or prominences over which the glacier passes, and into which it drops its burdens. Of course, where the ice is constantly renewed over such a spot by the onward progress of the glacier, these materials may be carried off again; but if we suppose such a case to occur at the breaking up of the glacier-period, when the ice was disappearing forever from such a spot, it is easy to account for the poising of these large boulders on prominent peaks or ledges.
The appearances about the _roches moutonnées_, especially the straight scratches and grooves on the side up which the ice ascended, have led to a mistaken view of the mode in which large boulders are transported by ice. It has been supposed, by those who, while they accepted the glacial theory, were not wholly conversant with the mode of action of glaciers, that, in passing through the bottom of a valley, for instance, the glacier would take up large boulders, and, carrying them along with it, would push them up such a slope and deposit them on its summit. It is true that large boulders may sometimes be found in front of glaciers among the materials of their terminal moraines, and may, upon any advance of the glacier, be pushed forward by it. But I know of no example of erratic boulders being carried to considerable distances and raised from lower to higher levels by this means. All the angular boulders perched upon prominent rocks must have fallen upon the surface of the glacier in the upper part of its course, where rocky ledges rise above its surface and send down their broken fragments. The surface of any boulder carried under the ice, or pushed along for any distance at its terminus, would show the friction and pressure to which it had been subjected. In this connection it should be remembered that in the case of large glaciers low hills form no obstacle to their onward progress, especially when the glacier is thick enough to cover them completely, and even to rise far above them. The _roches moutonnées_ about the Grimsel show that hills many hundred feet high have been passed over by the great glacier of the Aar, when it descended as far as Meyringen, without having seemingly influenced its onward progress.