ld
support the flat roof with ease, and the whole upper structure could be
made of sufficient thickness to exclude both the heat and the rain, while
the present appearance of the ruins is naturally accounted for.

Those who have lived in the East, those, even, who have extended a visit to
Athens as far as Eleusis or Megara, must have stretched themselves, more
than once, under the stars, and, on the flat roofs of their temporary
resting-places, sought that rest that was not to be found in the hot and
narrow chambers within. They must then have noticed, as I have more than
once, a large stone cylinder in one corner. In Greece and Asia Minor, it
will be in most cases a "drum" from some antique column, or a funerary
_cippus_, abstracted by the peasantry from some neighbouring ruin. This
morsel of Paros or Pentelic has to perform the office of a roller. When
some heavy fall of rain by wetting and softening the upper surface of the
terrace, gives an opportunity for repairing the ravages of a long drought,
the stone is taken backwards and forwards over the yielding pise. It closes
the cracks, kills the weeds that if left to themselves would soon transform
the roof into a field, and makes the surface as firm as a threshing-floor.

The roofs of Assyrian buildings must have required the same kind of
treatment, and we know that in the present day it is actually practised. M.
Place mentions rollers of limestone, weighing from two to three
hundredweight, pierced at each end with a square hole into which wooden
spindles were inserted to facilitate their management.[196] A certain
number of these rollers were found within the chambers, into which they
must have fallen with the roofs. As soon as the terraces ceased to receive
the care necessary for keeping down the weeds and shrubs and keeping out
the water, the process of disintegration must have been rapid. The rains
would soon convert cracks into gaping breaches, and at the end of a few
years, every storm would bring down a part of the roof. A century would be
enough to destroy the vaults, and with them the upper parts of the walls to
which they were closely allied by the skill of the constructor. The
disappearance of the archivolts and the great heaps of _debris_ are thus
accounted for. The roof materials were too soft, however, to damage in
their fall the figures in high relief or in the round that decorated the
chambers beneath, or the carved slabs with which their walls were lin