One other point is worth notice, because it illustrates another way in which the analogy of youth and maturity holds good. The youthful river, with its interrupted slope, its lakes and falls, does not permit the water to flow off with the same regularity as the mature river with its smoothed outlines. The mature river is thus a more perfect instrument of drainage. It is not necessary for our purpose to consider in detail the characteristic forms of river erosion. It may be sufficient to notice that rapids and waterfalls are due to the varying hardness of the rocks forming the bed of the river, and that the normal course of events is the transition from waterfall to rapid, and from rapid to stream flowing quietly at the bottom of a rocky gorge. Long gorges or canyons tend to occur in regions where river erosion is not greatly assisted by the other eroding agents. As a general rule, as the river cuts its way down, the other agents cut back the walls so exposed, so that a wide valley is formed. But a river does not only eat out its bed in its valley track. A necessary consequence of this erosion is that it is also able to eat back the slope on which it is rising, as a result of the smoothing out of the curves of its bed, so that its source retreats further and further into the mountain. In regions of abundant rainfall every slope is abundantly supplied with streams, and therefore those streams which cut back their region of origin most rapidly will necessarily encroach upon their neighbours' territory. They therefore tend to tap some of the tributaries of the other streams, a phenomenon which has sometimes considerable human importance, and has been extensively studied of late years under the name of river-capture. Some examples may serve to make the phenomenon clear. Every one who has travelled up the Rhone valley in Switzerland has noted the enormous number of lateral streams, of all sizes, which tumble down the mountain sides into the Rhone. These streams on, _e. g._, the south side, are, roughly speaking, parallel to each other, and to a large extent enter the main stream independently. That is, for the most part they are very youthful streams. In some cases, however, _e. g._ in the case of the Dranse and the Visp, the drainage is of a more advanced character, and we find a large stream with tributaries of considerable size as distinct from mere torrents. A glance at any great river system on the map, _e. g._ the Mississippi, the Amazon, etc., will show that the condition of a great stream with many tributaries is normal in a district where the drainage is of the developed type. How are the two conditions, that of numerous parallel mountain torrents and that of a great river system, related to one another? There is no doubt that capture, the encroachment of one stream upon the territory of another, has played an important part in the process. [Illustration: Fig. 4.--Sketch-map to illustrate river-capture.]