82 Arber. — A Revision of the Seed Impressions of the 
task, yet it is on these grounds that a classification should be based. In 
my opinion one of the most fundamental distinctions among Coal Measure 
seeds should be founded on the symmetry. We know from the study of 
petrified specimens that most of these fossils belong to two distinct types, 
the Radiosperms and the Platysperms, as F. W. Oliver 1 pointed out in 1903. 
The former are symmetrical in more than two planes, the latter in only 
two or in a single plane. At the same time a classification based solely 
on symmetry is not ideal. Certain seeds, such as Trigonocarpus P arkinsoni, 
are radiospermic from one point of view, and platyspermic from another. 
Although the number of such cases is few at present, an intermediate class 
may have to be recognized eventually. 2 The main difficulty as regards the 
recognition of the type of symmetry among impressions arises, of course, 
from the compression which many seeds have undergone before or during 
fossilization. From an isolated example it may be sometimes impossible to 
determine whether the seed in question is radiospermic or platyspermic, 
especially if it be a new or rare type. In such circumstances it is best 
referred to the genus Carpolithus . But it will be found that, when a con- 
siderable number of specimens of the same seed from different localities are 
compared, it is usually fairly easy to settle this question, at least pro- 
visionally. It is extraordinary how, in many cases, both radiospermic and 
platyspermic seeds have escaped compression, or have been flattened only 
to a slight degree. Even such seeds as Cordaicarpus , the nucellus of which 
in the natural state was probably narrowly elliptical in transverse section, 
and which, one imagines, would only require comparatively slight pressure 
perpendicular to the major axis to become quite flattened, have often 
escaped with but little distortion of the original form. 
At the present time there is considerable confusion as to the limits 
of the generic terms in common use. This arises partly because the subject 
has not undergone revision for more than half a century, and partly because 
of the extremely unfortunate practice of including impressions in genera 
founded on structure material, or vice versa , without at the same time 
affording satisfactory proof of the correlation of the impression with the 
petrifaction or vice versa. This is quite unjustifiable. It was perhaps less 
glaring a decade ago, when the diversity in anatomical structure to be met 
with in Palaeozoic seeds had not been so fully recognized. New types of 
seed structure should receive temporary names as they are discovered, until 
the time arrives when a definite proof can be afforded of their generic 
correlation with impressions, the names of which have been in existence for 
a century. Further, a diagnosis of a genus such as Cardiocarpus should 
take into account the external form of the seed in addition to the details 
of its anatomical structure, for this name has been in use for impressions 
since 1828. 
1 Oliver (’ 03 ), p. 453. 2 Oliver and Salisbury (’ll), p. 44. 
