The alignment of training walls or groynes should be such as will give the best channel consistent with economy in cost. The best channel is generally that which is most free from sharp bends. It is assumed for the present that no cuts or diversions of such lengths as to materially alter the gradient are to be made, but that a certain amount of choice of alignment is afforded by the reduced width of the trained channel and by small diversions or easings of bends. It is sometimes said that straight reaches are objectionable because the stream will tend to wander from side to side and cause shoals, whereas in a bend there will be no such tendency. The difficulty as to shoaling will be greatest at low water, but it is likely to be serious only when the width between the training walls is too great. If the width cannot be reduced to such an extent as to do away with the trouble, it may be better to adopt a curved course. The width between the training walls should generally be the same throughout, whether the reaches are straight or curved, but in view of the preceding remarks it may be desirable, where a reach cannot be otherwise than straight and where shoaling is feared, to give the straight portion a reduced width with of course a greater depth, and similarly to reduce the width at reverse changes of curvature. In curves which are at all sharp the curvature should be rather sharper in the middle of the curve than at the ends (CHAP. IV., _Art. 8_). 4. =Alteration of Depth or Water-Level.=--When the width of a stream is altered, the depth of water--the gradient being supposed to be unchanged--must alter in the opposite manner. A narrowing of the channel by training necessitates an increase in the depth of water, and the same remark applies if an arm of the stream is closed. The increase in depth may be effected either by raising the water-level or by lowering the bed--as may be convenient--or both. If the bed is to be lowered and is of hard clay, it may be necessary to dredge it and, when this has been done, training may be unnecessary. If the bed is of soft mud, a dredged channel is likely to fill up again, and training alone will be the method to adopt. If the bed is moderately hard, say compact sand, it may be suitable to train the channel first and then to dredge if necessary. In any case, shoals of hard material may have to be dredged or rocks, whether these form shoals or lateral obstructions, to be blasted or otherwise broken up (_Art. 2_). In cases where it is desired to raise the water-level without any lowering of the bed, training is of course necessary. In any case in which the bed is likely to scour to a lower level than is desired, or if the bed is to be raised, the measures described in CHAP. V., _Art. 6_, may be adopted, but they are hardly likely to be suitable and satisfactory in all cases.