Again, Salzenberg hardly makes it sufficiently clear that the large arches in the walls which fill the great vertical semicircles over the arcades on north and south sides, are in fact the inner surfaces of the arches which pass between the pairs of piers on north and south sides (seventy-two feet apart in this direction), and being the whole width of those piers (fifteen feet eight inches) on soffite they form the immense arches so well known on the outside. The semicircles of wall, each of which contains twelve windows, are now filled in beneath these arches, flush with their _inner faces_, and the arches therefore do not show to the interior through the decoration (Figs. 4, 36, 38). [Illustration: FIG. 35.--Plan of Upper Gallery as first designed.] [Illustration: FIG. 36.--Section of Aisles and Gallery.] Now Agathias (see page 30) says that at the restoration after the earthquake in 558, at the north and south arches they brought towards the inside “the portion of the building which was on the curve.” This, we think, must refer to the filling wall, in the arches of seventy-two feet span, which we suppose was formerly on the exterior, and thus left an upper gallery twelve feet wide and seventy-two feet long open to the interior. “And they made the arches wider to be in harmony with the others, thus making the equilateral symmetry more perfect. They thus reduced the vast space and formed an oblong design.” That is the arches of seventy-two feet, when filled up on the inside, were no longer visible, and the dome appeared to stand over arches of 100 feet span on north and south, as already on east and west, the transverse dimension of the church being lessened between these points by some twenty-four feet. Salzenberg understanding Agathias to refer to the apparent arches of 100 feet span on north and south is unable to offer any explanation. [Illustration: FIG. 37.--Plan of Basis of Dome as originally designed, with Additions A A containing stairs.] [Illustration: FIG. 38.--Section between Great and Secondary Orders.] The actual evidence in the church, we believe, fully bears out the interpretation here suggested. What we have called the secondary order of columns would pass exactly beneath the position given to this wall. These columns on the gallery floor are very strong, and a very strong row of arches runs along over them (see Fig. 38). Moreover the curtain walls in every other instance throughout the church are flush with the exterior.