wall, and possibly a ditch as well.

For the

camp 20 acres would be small, 60 acres not impossible; it might accommodate two whole legions with its auxiliaries, less commonly three or four. For the fort 3 to 4 acres and 6 to 7 acres are common sizes; 8 acres is large and 10 acres almost unknown. A cohort would have been a large garrison. Notwithstanding these fundamental distinctions, the analogies between camp and fort are striking. The principles that elaborated the one were equally applicable to the laying out of the other. There is the same quadrangular shape, the same symmetrically situated gates, and systematic arrangement of the interior.

Two classes of Roman forts confront the observer: the one of 3 to 6 acres in size, with its walls supported by internal buttress-towers; the other of 8 acres, with external towers and a thicker wall. We shall at once designate these the early and late types respectively-by early" meaning the end of the first and beginning of the second century, by "late" more generally the third and fourth centuries. By employing these terms we somewhat forestall some of the conclusions of the examination we shall make; yet the distinction is characteristic, and the terms fairly general and convenient. It will be our object to see how far they are justified, and how far they may be modified or narrowed.

Of the early class of fort, distinguishable by its internal towers, there are also two kinds :-(1) The fort of 3 to 4 acres; (2) the fort of 6 to 7 acres.

Internally these two types present some striking contrasts, though in principle the same. Some examples are figured in the upper row of the second plate. Both large and small are symmetrical,

The

with internal towers; a prætorium is in the centre, its courtyard facing the main street, which in the smaller forts bisects, and in the larger trisects the interior. But, whereas in the smaller the prætorium (the official quarters), and possibly some baths or granary, are usually the only stone buildings within the enclosure, the larger type is filled, by contrast, with systems of long rectangular parallel buildings, which divide up the interior into a series of symmetrical rectangles. These details, not being connected with the defences, are not shewn in our illustration, but they serve to draw the distinction. The smaller types pictured are the outlines of Melandra Castle, in Derbyshire, and of Hardknott Castle, in Cumberland. former is protected by a thin wall and supporting mound (as described), and the latter by a stouter wall only. In neither case is there definite appearance of a ditch around. The larger type is chiefly familiar along The Wall, as at Chesters and Housesteads, but is found also in other sites. forming part of the same system of defences, as at Ribchester, on the Ribble, in Lancashire. It is surrounded by a stout wall, and sometimes by a ditch also. The wall of each type is supported by buttress towers, which surmount the corners and flank the gates. Commonly, these have no chamber, or a vault merely, on the ground floor; but the upper storey was probably a roofed chamber, giving access either way to the rampart walk. In the larger class, the longer interval between the side gates and further corners was generally strengthened by a tower also. Sometimes the number of towers was more numerous, the shorter intervals intervening between gates and corners having one and the larger two such supports. In a well-known case, figured on the third plate, of