=Discontinuity of terraces.=—When a stream sinks its channel into its flood-plain, it does not follow that a terrace remains on each side. Where the stream’s deepened channel is in the middle of its flood-plain, there is, temporarily, a terrace on either side; but wherever the deepened channel is at one margin of its flood-plain, a terrace remains on the other side only. Even where continuous at the outset, terraces soon become discontinuous, for all processes of subaërial erosion conspire to destroy them. A stream is likely to meander on its second and later flood-plains, as on its first and highest one. Wherever the meanders on its second flood-plain reach the borders of the first flood-plain, the terrace at that point disappears, and since the meanders are continually migrating, terraces are continually disappearing. The same would be true of the second terrace, if a second were developed. The removal of portions of a terrace by the sweep of meanders is likely to leave the remnants cuspate toward the stream.[91] Again, tributary streams, in bringing their channels into topographic adjustment with their mains, cut through the terraces of the latter. New gullies develop on the faces of the terraces and their heads work back across them, dissecting them still further. At the same time, sheet erosion and other phases of slope wash tend to drive the scarps of the terraces back toward the bluff beyond. By the time a second series of terraces is well developed, no more than meagre remnants of the first may remain. From the foregoing considerations it is clear that the extent to which river terraces once developed, now remain, is dependent in part on the length of time which has elapsed since the river sank its channel below them. Other things being equal, the greater their age the more meagre their remnants. Terraces developed from river plains formed chiefly by alluviation stand a better chance of long life than most other alluvial terraces. This is because of the configuration of the original valley, the aggradation of which gave origin to the plain. The principle involved is illustrated by Fig. 197. In developing the second flood-plain the river encounters the rock wall of the valley. This greatly retards lateral erosion, and the terrace above, _defended_[92] by the rock, is likely to be long-lived. [Illustration: +Fig.+ 197.—Diagram to show why certain terraces are longer lived than others.] Alluvial terraces, like rock shelves, are popularly thought to mark “old levels of the river.” In one sense this is true, but not in the sense in which the expression is commonly used. Every level, from the crest of the bounding bluffs to the bottom of a valley, is a level at which water ran for a longer or shorter time; but the terrace does not mean that the river was once so much larger than now as to fill the valley from its present channel to the level of the terraces.