the fort at Weisbaden, the number is considerably increased, but the towers are smaller. Besides being architectural supports, these may be looked upon as sentry boxes, in the sense that the sentries and guards upon the ramparts might seek ready shelter within their upper chambers. Those who look for an extended analogy, may see in them a survival of sentry boxes along the ramparts of a temporary camp. An overhead platform, crossing the entrances from tower to tower, enabled the gates also to be protected.

Let us now look at the change found to have taken place in the type of the later class, of which four well-known examples, from the Saxon shore, are represented on the second plate. At Richborough, in Kent, we find the internal towers preserved, indeed, in idea, by a slight inward projection of the towers flanking the western gateway. Otherwise, all the towers are external; those at the corners of the enclosure are round (being three-quarters of the arc), while those intervening are rectangular, with the same projection. The gate towers, likewise, project to a similar distance. There is no ditch. The walls are thicker and higher; the area enclosed is rectangular, and of eight acres. There is no symmetrical arrangements of the gates; a postern, strongly defended (with overhead gallery), alone remains on the northern side; while the main gateway, on the west, was possibly narrow, with a single arch only, and a long flanking defence. At Porchester there is a similar effect, only the bastions are all rounded, and are in some cases hollow. At Burgh Castle and at Pevensey there is less attempt at symmetry of form; in the latter case, particularly, there is an obvious deviation of shape, to suit the contour of the rise on which it

stands. The points of difference, so far as the defences are concerned, are apparent from these illustrations. The walls are thicker and higher, the area is larger, the gates are smaller, the ditch has disappeared, and the towers are become external. The change is no incidental one; the later class of fort is a new type, modelled on new principles of defence. Let us look more closely into the details of this new class, to see whether it may be possible to discern any motive that induced the change.

The various forms of the new type of external buttress are illustrated by a few details on the third plate. The Romans had arrived at a system of defences which in principle survived, with little modification, until utterly new and modern methods of attack were introduced. Witness, for instance, the analogy which is afforded between the bastions. of Richborough (Kent) of the third and fourth century and those of Bodiam Castle (Sussex) of the fourteenth. The new principle, which separates the late types from the old, and which dominates the whole of the later work, may be illustrated by a few simple examples. In the diagram of two towers defending an intervening wall, the one (a), rectangular, the other (B) circular, it is clear that each tower equally protects by direct aim the wall itself. Yet the form в holds over the form a a distinct advantage in being able to receive support at the extreme outer point b from the adjoining towers, whereas the corresponding point a of the rectangular tower is left undefended except from above. [The principle of machicolation designed to meet such case is well illustrated at Bodiam Castle.] The further development of the bastion in Roman works is illustrated rather by efforts at modification or extension of the same principle.