80 JOURNAL OF THE EOYAL H(3RTICULTUEAL SOCIETY. 
rocks in such a place is confined almost entirely to those few species of 
ferns and other plants (mostly of the rougher class) which wall exist 
under the named conditions, a site of this kind should be avoided. 
The better class of mountain plants require an abundance of light ; 
and though bamboos and trees or shrubs at a considerable distance may 
be a help in forming an admirable background, the site for our rock-garden 
should be neither obstructed by them, nor should it be within reach o£ 
their roots. If it cannot be avoided to have large trees, especially Elm 
trees, within such a distance as to cause apprehensions lest their roots 
might eventually reach the site selected, such roots can often be rendered 
harmless by an underground wall of concrete, which would make their 
encroachment impossible. 
Pkeparatory Woek will always be necessary to produce that 
unevenness of surface without which the imparting of a bold character is 
impossible. But even on a perfectly level site it can generally be so 
arranged that, where soil is excavated to produce depressions, the material 
thus gained may be utilised close by for filling up the portions for 
w^hich a higher level is desired. 
Where the surface is naturally undulating, it will often be found just 
as well to take advantage of what nature has done already, and, instead 
of filling up a depression in the ground or levelling down a hillock, it will 
often be found much more advantageous to make the hill still higher and 
the valley still deeper, and thus emphasise them both. 
When ground is excavated it should always be sunk to a level lower 
than is actually required to allow for a sufficient supply of good soil to 
go back. But where elevations are to be made it would be a mistake to- 
raise the ground to anything approaching the intended height during the 
preparatory stage. 
I have always found it a far better plan to commence the elevations- 
with the stones and not with the soil, which latter should be filled in behind 
the stones as the work proceeds. It will be a great saving of labour 
(especially if the rock-garden is to be on a large scale) to postpone at least 
some of the preparatory excavations till the work of rock-building 
actually commences, when the excavated soil can be filled in behind the 
stones as required without having to be moved twice. But where this 
would be inconvenient, good soil and inferior soil should at any rate be 
kept apart, and stored till required at some convenient spot close to the 
work. For ponds and streamlets special preparations are required, with 
w^hich I will deal later on. 
That the preparatory work must include proper drainage goes with- 
out saying. 
What Stone to Use is a question of no little importance, for on 
this much of the success of the work must naturally depend. If a stone 
quarry is near at hand, it will in most cases be found preferable to use 
the material which is easiest to be obtained, unless this should be of a 
nature quite unsuitable. 
But when a choice of several kinds of stone can be had I should 
certainly give preference to pieces showing an old w^eather-beaten surface^ 
toned down by long exposure, and perhaps even covered with moss and. 
lichens. 
