They are not supported on
pilasters, as is the practice elsewhere, but rest on corbels, and, in
order to accommodate these corbels, the lateral niches, originally of
the same height as the central window, have been reduced in height. A
fragment of the original arch still remains, cut into by the wall arch
of the dome. The flat secondary arches crossing the chapel at each end
are similarly supported on corbels.

This view is confirmed by the examination of the plaster left upon the
walls. That plaster has four distinct coats or layers, upon all of
which eikons in tempera are painted.[499] The innermost coat is laid
between the transverse dome arches and the walls against which they are
built. Those arches, therefore, could not have formed parts of the
building when the first coat of plaster was laid, but must be later
additions.

In keeping with this fact, the second coat of painted plaster is found
laid both on the arches and on those portions of the old work which the
arches did not cover.

The secondary arches under the transverse arches at each end belong to a
yet later period, for where they have separated from the arches above
them, decorated plaster, which at one time formed part of the general
ornamentation of the building, is exposed to view. At this stage in the
history of the chapel the third coat of plaster was spread over the
walls, thus giving three coats on the oldest parts where unaltered--two
coats on the first alterations, and one coat on the second alterations.
The fourth coat of plaster is still later, marking some less serious
repair of the chapel.

The _voussoirs_ of the lateral dome arches should be noticed. They do
not radiate to the centre, but are laid flatter and radiate to a point
above the centre. This form of construction, occurring frequently in
Byzantine arches, is regarded by some authorities as a method of forming
an arch without centering. But in the case of the lateral wall arches
before us it occurs where centering could never have been required;
while the apse arch, where centering would have had structural value, is
formed with true radiating _voussoirs_. The failure of the _voussoirs_
to radiate to the centre therefore seems to be simply the result of
using untapered _voussoirs_ in which the arch form must be obtained by
wedge-shaped joints. For if these joints are carelessly formed, the
crown may very well be reached before the requisite amount of radiation
has been obtained.