In the Baths of Diocletian, the Basilica of Maxentius, and other great Roman halls, this was met by cross walls pierced only by small archways, and placed at intervals, dividing the aisles into chambers, each of which was covered by a short barrel vault at right angles to that over the central space (Fig. 1). This, however, would be inconsistent with the uses of a church, and, indeed, applies to a _groined_ rather than a barrel vault, though a very similar expedient was sometimes used by the Romanesque builders, by covering the aisles with cross barrel vaults, as those above described, supported by arches across the aisles, instead of by cross walls (Fig. 2). Another system was to cover the aisles by a half or little more than a half longitudinal barrel roof, forming a continuous arched buttress to the continuous central vault (Fig. 3). This gave them a perfectly vaulted building of trustworthy construction, provided only that the aisle walls were of sufficient strength. The barrel vaults were often both strengthened and their monotony relieved by arched ribs added to their thickness over each pillar of the nave, and _repeated_ over the aisles, while these planes of extra strength were carried through to the exterior in the form of buttresses of small projection against the aisle walls (Fig. 4). [Illustration: Fig. 2.] [Illustration: Fig. 3.] [Illustration: Fig. 4.] The builders of such churches were not, however, ignorant of the principles of the _groined_ or _intersecting_ vault formed by the inter-penetration of two demi-cylinders, and so largely used by the Romans. They did not use them in such buildings, because their main vault rising into the roof, they could not, under the same roof-plane, introduce the intersecting vaults,--though this had been effected in Roman structures by a series of cross gables over the _cross_ vaults. In churches of the same kind, however, we find the groined vault used to carry a gallery in the aisles, all the rest remaining as before (Fig. 5). [Illustration: Fig. 5.] It would appear that the obvious mechanical advantages it offered led at an early period, in the south of France, to the substitution of the _pointed_ for the round arch in the great vault of churches of this construction; but I will suppose for the present the semicircle to be strictly adhered to. The great defect in such a church as I am supposing would be want of light in the nave from the absence of clerestory windows; and as such windows had been in use from the days of the earliest basilicas, this loss would be fully appreciated. [Illustration: Fig. 6.]