long to the genus
_Rozites_, and the species was named _R. gongylophora_. A microscopic
examination of the particles of which the garden is composed shows
that they contain remains of leaves; bits of epidermis, stomata,
spiral vessels, etc., occur in them.

If a nest is broken into and the fungus garden scattered, the ants
collect it as quickly as possible, especially the younger parts,
taking as much trouble over it as over the larvae. They also cover it
up again as soon as possible to protect it from the light. A nest, 1
metre x 50 cm., was opened, and in twenty-four hours the ants had put
on a new roof 10 cm. deep.

Some ants' nests were placed under a bell jar and supplied with
leaves; they made no use of them and presently died. If they were
supplied with a piece of "garden," they rebuilt it and covered it so
far as they could. It was seen to shrink from day to day, the ants
bringing out the old pieces and adding them to the wall; finally it
was exhausted and the ants died. Others were starved for five days,
and then supplied with a bit of garden; they at once began to eat the
Kohl-rabi clumps. Finally, by supplying the ants with bits of garden,
a damp sandy floor, and fresh leaves, they were induced to build in
captivity. The dish in which they worked was covered by a glass lid,
and when this was covered with a dark cloth or otherwise kept dark,
the ants built under it without covering the garden. In this way the
whole process was observed. An ant bringing in a piece of leaf
proceeds to cut it into halves, repeating the process till it has got
a very small piece left, which it holds between its fore feet and
turns round, crushing it in its jaws until the whole is reduced to a
round ball of pulp about .25 mm. thick. This it then takes and adds to
the garden. So well is the kneading performed that no single cell
remains uninjured, and it was observed that the hyphae of the fungus
grew through and round one of these particles within a few hours. Belt
supposed that this process was performed by the small workers
above-mentioned, but it is not so, as we have just seen. The small
workers perform the function of weeding the garden, and this is so
well done that a portion of it removed and grown in a nutrient
solution gives a perfectly pure culture, not even containing bacteria!

In the course of these investigations it was found that somewhat
similar fungus gardens occur in the nests of the hairy ant,
_Apterostigma_,