The lower story of the north-western pier is covered with a flat circular roof resting on four pendentives, while the upper story is open to the timbers, and rises higher than the roof of the church, as though it were the base of some kind of tower. It presents no indications of pendentives or of a start in vaulting. The original eastern wall of the church has been almost totally torn down and replaced by a straight wall of Turkish construction. Traces of three apses at that end of the building can, however, still be discerned; for the points at which the curve of the central apse started are visible on either side of the Turkish wall, and the northern apse shows on the exterior. The northern and southern walls are lighted by large triple windows, divided by shafts and descending to a marble parapet near the floor (Plate IV.). The dome, which is large in proportion to the church, is a polygon of sixteen sides. It rests directly on pendentives, but has a comparatively high external drum above the roof. It is pierced by sixteen windows which follow the curve of the dome. The flat, straight external cornice above them is Turkish, and there is good reason to suspect that the dome, taken as a whole, is Turkish work, for it strongly resembles the Turkish domes found in S. Theodosia, SS. Peter and Mark, and S. Andrew in Krisei. The vaults, moreover, below the dome are very much distorted; and the pointed eastern arch like the eastern wall appears to be Turkish. When portions of the building so closely connected with the dome have undergone Turkish repairs, it is not strange that the dome itself should also have received similar treatment. In the western faces of the piers that carry the eastern arch large marble frames of considerable beauty are inserted. The sills are carved and rest on two short columns; two slender pilasters of verd antique form the sides; and above them is a flat cornice enriched with overhanging leaves of acanthus and a small bust in the centre. Within the frames is a large marble slab. Dr. Freshfield thinks these frames formed part of the eikonostasis, but on that view the bema would have been unusually large. The more probable position of the eikonostasis was across the arch nearer the apse. In that case the frames just described formed part of the general decoration of the building, although, at the same time, they may have enclosed isolated eikons. Eikons in a similar position are found in S. Saviour in the Chora (Plate LXXXVI.).