an. These arches were made out by an extra quantity of concrete
on the under side for decoration, and on the upper in the case of the
great arch, so as to form a roof, the well-known roll and flat Italian
tiles being embedded in the mortar. Many and large fragments of
this roof were found lying on the deposit that had partially filled
the ruins previous to the fall of the roof, and are still carefully
preserved. A large fragment, 18ft. long by about 3ft. wide, and 1ft.
9in. thick, that has slipped down, as it were, from the western end,
in the position in which it was discovered, was formed of solid tiles,
with an arch of tiles 1ft. 8in. long,[21] the roof having sufficient
abutment on this side for a solid construction.[22] This arch gives
the form of the window that lighted the bath on the western end.

[Footnote 21: The arches in the adjoining apartment west of this were
built of a sort of a tufa.]

[Footnote 22: On the falling of the roof one of the piers was thrust
out of the perpendicular, the upper half toppling over, and the lower
would have again returned to its original position had a stone not
fallen into the vertical joint, catching the pilaster as a wedge.
The pier is still fixed out of the perpendicular by the stone in the
joint.]

The vaulting of the side aisles, or rather that over the _schola_,
was arched from pier to pier longitudinally and transversely, the
quadrangular spaces being in all probability simply groined; but
a fragment of box tiles found almost leads one to think that these
spaces were vaulted by a domical vault, springing either from
pendentives in the angles of the vaults, more common in later work,
or from a slight cornice on a level with the apex of the arches. The
vault, if there was one, over the semi-circular _exedrae_ must have
been hemispherical. From the number of roofing tiles of local stone,
shaped into hexagons, found, I think these arcades were roofed in
with them, placed overlapping each other, giving a very good effect.
Similar tiles were dug up at Wroxeter, and I have found slates of the
same shape in the Roman villa I have been excavating for Mr. Chas. I.
Elton, F.S.A., M.P., at Whitestaunton Manor. The form of these slates
deserves copying; a roof covered by them is far lighter than that of
rectangular slabs and more picturesque. The walls on the sides towards
the hall, and externally, so far as I have been able to ascertain, are
covered with the usual red plaster, s