When scour of the sides of a channel occurs it may occur by direct action of the stream on the sides near the water-level, or by action at or near the toe of the slope, which causes the upper part of the bank to fall in. Such falling in is generally more or less irregular, and the bank presents an uneven appearance. The fallen pieces of bank may remain, more or less intact, especially if they are held together by the roots of grasses, etc., where they fell, and prevent further scour occurring along the toe of the slope. Falling in of banks is most liable to occur in large streams and with light soils. It may be caused by the waves which are produced by steamers and boats or, especially in broad streams, by wind. The action on the banks at bends is discussed in _Art. 8_. Thus in designing a channel according to the principles laid down in _Art. 6_, the question of action on the sides of the channel has to be dealt with as follows. Whether or not the velocity is to be low, relatively to the depth, _i.e._ whether or not deposit on the bed is more likely to occur than scour, care can be taken not to make it actually too low, and not to make it actually too high, particularly if the soil is light and friable. With ordinary soils a mean velocity of 3·3 feet per second in the channel is generally safe as regards scour of the sides. Any velocity of more than 3·5 feet per second may give trouble. A velocity of less than 1 foot per second is likely to give rise to deposit on the sides. In channels in alluvial soils the falling in of banks is sometimes said to occur more when the stream is falling than at other times. This has been noticed on both the Mississippi and the Indus. The cause has been said to be the draining out of water which had percolated into the bank, the water in flowing out carrying some sand with it. The effect of this cannot however be great. 8. =Action at Bends.=--At a bend, owing to the action of centrifugal force and to cross-currents caused thereby, there is a deposit near the convex bank and a corresponding deepening--unless the bed is too hard to be scoured--near the concave bank. The water-level at the concave bank is slightly higher than at the convex bank. The greatest velocity instead of being in mid-stream is nearer the concave bank. As the transverse current and transverse surface slope cannot commence or end abruptly, there is a certain length in which they vary. In this length the radius of curvature of the bend and the form of the cross-section also tend to vary. This can often be seen in plans of river bends, the curvature being less sharp towards the ends.