baccata , with Remarks on the Antiquity of the Taxineae . 123 
If this has really been the course of events in the history of Torveya , 
the seed of this genus must belong to a truly isolated type, and would seem 
to be the result of specialization along a restricted side line. Nevertheless, 
the fact that the three known genera of Taxales so naturally fall into line 
with each other in most other respects justifies the expectation that it should 
be possible also to make the peculiar seed-organization of Torreya intelligible 
in terms of Taxus and Cephalotaxus . An attempt to do this is described 
in the following pages, which at the same time appear to corroborate in 
some detail the ingenious theory advanced by Professor Oliver sixteen 
years ago. 
The two foramina in the basal disc of the sclerotesta of Taxus naturally 
r recall those well known to occur in Torreya at a much higher level. Since 
it is tolerably certain that the paired ovular strands in both genera have 
been derived by a splitting of an original central main supply, it would 
seem that Taxus is nearer to the ancestral type in so far that in the 
former genus the ‘ chalazal foramina 5 are nearer the middle line than in the 
latter. 
But in such a comparison between Taxus and Torreya there is one 
apparently insuperable difficulty, namely the fact that whereas in Torreya 
the strands of the ‘ outer system ’ are entirely outside the stone, in Taxus 
they are entirely inside the seed-cavity, for, as is well known, in this genus 
the outer flesh (if this name is at all applicable) is represented by only 
a thin membrane of unlignified cells covering the stone. 
How, then, is it possible to derive these two apparently divergent 
types from the same ancestral form ? 
We will, in the first instance, fix our attention upon certain well- 
preserved platyspermic seeds of Cordaitean affinity, described long ago by 
Ad. Brongniart (1874) from the Carboniferous of St. Etienne, and subse- 
quently (1881) refigured to show the details of their structure. The 
original diagnoses have to some extent been recently modified by Professor 
C. E. Bertrand (1907, 1 , 2 , 3 , 4 , 1908, i, 2 , 3 ), who undertook a revision of 
Brongniart’s types with the help of Renault’s later preparations. More 
recently Mrs. Agnes Arber (1910) has made an interesting contribution to 
our knowledge of these ancient seeds, which exhibit a remarkable diversity 
of form without departing from the essentials of the plan upon which they 
are built . 1 
1 The further literature is referred to in Mrs. Arber’s paper. The attribution of the seeds in 
question is, with the exception of C ardiocarpus , still a matter of conjecture, but there is a strong 
presumption in favour of the view that like Cardiocarpus they all belonged to members of the 
phylum Cordaitales. Apart from their agreeing with Cardiocarpus in their general plan of structure, 
they all possess a more or less well developed ‘ tent-pole ’—an organ which appears to be of con- 
siderable importance in questions of relationship. In the case of one of these seeds (. Mitrospermum ) 
Dr. Scott has recently (1918-1919) discovered some indirect evidence to show that it was probably 
borne upon the highly specialized axillary shoots of the Cordaitean species Mesoxylon multirame . 
