— 105 — 
In this connection attention is drawn to the Taklamakan- 
desert where S. HEDIN travelled for days through a desert 
perfectly devoid of vegetation, and far away from any human 
habitations. 
If it has been thus established that the development of 
the Sand-desert has in all probability been from Barchans 
to Hummock-desert and Desert-plains, that it has proceeded 
from the most shifting condition to the more stable, then we 
have at the same time traced the process of development of 
the vegetation. This process is expressed by the order in 
which the various sand-desert vegetations were described in 
the preceding pages, and it may be shortly summed up: 
1. Aristida pennata. 
2. Ammodendron, Calligonum and other desert trees. A 
few herbs, mostly annuals. 
3. Desert-trees, smaller, but growing more closely. Seve- 
ral herbs, among which perennials are dominant. Annual 
halophytes in the valleys. (Hummock-desert). 
4. Small desert-trees (or none). No halophytes (or few). 
(Desert of the Sand-plains). 
To this process of development the definition of forma- 
tion by Moss might be applied. What this author (1907 p. 
12) terms a formation is: 
“The series of plant associations which begins its history 
as an open or unstable association, and eventually becomes 
a closed or stable association.” 
Even if the Desert-plain be not closed, it is in itself 
stable, a terminal sub-formation, and the definition of for- 
mation given by Moss thus seems to be applicable to the 
Sand-desert as a whole. The definition of formation by Moss 
applied in this way, is employed here as a means of illustra- 
ting the unity of the sand-desert and to elucidate its meta- 
morphosis. It must be emphasised, however, that I have 
grouped the different types of sandy desert in the same for- 
mation not because they constitute what might be called a 
historical series derived from each other in a definite sequence, 
