A little bridge, apparently frail, spans the torrent just above the entrance to this gorge, and from it one perceives, being fashioned in the rocks below, concavities similar to those to which reference has just been made. The torrent is seen hurrying forwards. Not everywhere. In some places the water strikes projecting angles, and, thrown back by them, remains almost stationary, eddying round and round: in others, obstructions fling it up in fountains, which play perpetually on the _under_ surfaces of overhanging masses; and sometimes do so in such a way that the water not only works upon the under surfaces, but round the corner; that is to say, upon the surfaces which are _not_ opposed to the general direction of the current. In all cases _concavities_ are being produced. Projecting angles are rounded, it is true, and are more or less convex, but they are overlooked on account of the prevalence of concave forms. Cause and effect help each other here. The inequalities of the torrent bed and walls cause its eddyings, and the eddies fashion the concavities. The more profound the latter become, the more disturbance is caused in the water. The destruction of the rocks proceeds at an ever-increasing rate; for the larger the amount of surface that is exposed, the greater are the opportunities for the assaults of heat and cold. When water is in the form of glacier it has not the power of making concavities, such as these, in rocks, and of working upon surfaces which are not opposed to the direction of the current. Its nature is changed; it operates in a different way, and it leaves marks which are readily distinguished from those produced by torrent-action. The prevailing forms which result from glacier-action are more or less _convex_. Ultimately, all angles and almost all curves are obliterated, and large areas of flat surfaces are produced. This perfection of abrasion is rarely found, except in such localities as have sustained a grinding much more severe than that which has occurred in the Alps; and, generally speaking, the dictum of the veteran geologist Studer, quoted below, is undoubtedly true.(77) Not merely can the operations of extinct glaciers be traced in detail by means of the bosses of rock popularly termed _roches moutonnées_, but their effects in the aggregate, on a range of mountains or an entire country, can be recognised sometimes at a distance of fifteen or twenty miles from the incessant repetition of these convex forms.