Another example may help to explain the evolution of a complex river system with many tributaries. A glance at the map of England (see diagram) shows that while the rivers of Northumberland and Durham flow independently into the sea, those of Yorkshire are united into a characteristic bunch, and all reach the ocean by means of the Humber. This estuary breaks through the high ground formed by the Wolds of Yorkshire and Lincolnshire, which consists of hard rock. At one time it is probable that the rivers of Yorkshire entered the sea separately, while the other great factor of the Humber, the Trent, mingled its waters with the present Witham. At this time the weathering of the land surface had not reached its present stage so the land would lie higher. In what is now the vale of York the rocks are much softer than where the Wolds now stand, and the present Ouse, which was at first a longitudinal tributary of a transverse stream, eating its way back through these soft rocks, tapped successively the streams flowing eastwards from the Pennines, and with the help of the abundant water so obtained was enabled to cut out the wide estuary of the Humber. [Illustration: Fig. 5.--Sketch-map of northern England, to show the position of the Tyne and Aire Gaps, and the peculiar character of the rivers of Yorkshire. The black areas are heights above 600 feet.] One other important point in connection with river-capture has been already suggested in the account given of the Feshie. In the little sketch-map we see clearly the shift of the watershed to the east. The ultimate cause of this shift is doubtless the fact that in Great Britain the rainfall diminishes to the east, so that, generally speaking, the westerly streams have more erosive power than the easterly. But the special interest of the case is simply that it may serve to suggest a fact not at first sight obvious, which is that water-partings are excessively unstable features. One set of streams is continually striving to encroach upon the others, and by capturing their headwaters to reduce their erosive power. A very striking example of capture on the large scale is seen in southern Patagonia, where the water-parting does not lie near the summit of the chain of the Andes, as might be expected, but considerably to the east, the western streams (or glaciers) having captured all the headwaters of the eastern streams, which lie in a region of much lower rainfall. The net result is that running water not only scours valleys in the sides of mountain chains, but also, sooner or later, wears away the crest itself, and with the assistance of the other agents of denudation tends to reduce the mountains to plains--or at least "peneplains." The deduction is, of course, old enough, but the recent emphasis placed upon river-capture helps us to realise it, showing us the actual "shift of the divide," or, in other words, the wearing down of the summit levels.