e point of view of economy. In the
previous constructions, the sides of the reed supplied the greater part
of the walls and the work was limited to one partition for each cell.
Here, except at the actual periphery, where the tube itself supplies a
foundation, everything has to be obtained by sheer building: the floor,
the ceiling, the walls of the many-sided compartment are one and all
made of mortar. The structure is almost as costly in materials as that
of the Chalicodoma or the Pelopaeus.

It must be pretty difficult, too, when one thinks of its irregularity.
Fitting as best she can the projecting angles of the new cell into the
recessed corners of the cell already built, the Osmia runs up walls
more or less curved, upright or slanting, which intersect one another at
various points, so that each compartment requires a new and
complicated plan of construction, which is very different from the
circular-partition style of architecture, with its row of parallel
dividing-disks. Moreover, in this composite arrangement, the size of the
recesses left available by the earlier work to some extent decides
the assessment of the sexes, for, according to the dimensions of those
recesses, the walls erected take in now a larger space, the home of
a female, and now a smaller space, the home of a male. Roomy quarters
therefore have a double drawback for the Osmia: they greatly increase
the outlay in materials; and also they establish in the lower layers,
among the females, males who, because of their earlier hatching, would
be much better placed near the mouth of the nest. I am convinced of it:
if the Osmia refuses big reeds and accepts them only in the last resort,
when there are no others, it is because she objects to additional labour
and to the mixture of the sexes.

The Snail-shell, then, is but an indifferent home for her, which she
is quite ready to abandon should a better offer. Its expanding cavity
represents an average between the favourite small cylinder and the
unpopular large cylinder, which is accepted only when there is no other
obtainable. The first whorls of the spiral are too narrow to be of use
to the Osmia, but the middle ones have the right diameter for cocoons
arranged in single file. Here things happen as in a first-class reed,
for the helical curve in no way affects the method of structure employed
for a rectilinear series of cells. Circular partitions are erected at
the required distances, with or without a s