398 MR. 8, T. DUNN: A REVISION 
regarded as a section of Clethra in Eriaceæ, and placed it next to Actinidia.. 
The placing together of the two genera, though undoubtedly correct, furnished 
in reality an argument against his recently made decision and provided a new 
reason for the inclusion of both in Ternstræmiaceæ, for Clematoclethra with 
its united styles, few loculi, few seeds, with smaller albumen and larger 
embryo is very different from all Dilleniaceæ. 
In 1893 Gilg contributed to Engler and Prantl’s * Pflanzenfamilien ? his. 
account of the Dilleniaceæ, making therein а separate section for Actinidia 
between Dillenia and Saurauja. The distinguishing characters between 
Dilleniacer and Theaceæ (Ternstræmiaceæ) are stated to be the aril, the 
abundant albumen, and the minute embryo. Оп two of these characters, the- 
first and last, Actinidia and Saurauja would be excluded from Dilleniaceæ, and 
as the quantity of albumen is variable in Ternstræmiaceæ (absent, for instance, . 
in Æurya) their inclusion in the former Natural Order is difficult to follow. 
From what has been said it is clear that some considerable divergence: 
exists between the opinions of the numerous eminent botanists who have 
discussed the systematic position of Actinidia and the group of genera to 
which it belongs. A review of the facts must now be taken. In locating a 
group of genera possessing admitted affinities with two Natural Orders. 
the systematist must follow one of three courses according to the strength, in 
his judgment, of the affinities of the members of the gronp (1) inter ze, 
(2) with one of the Natural Orders, (3) with both. In the first case, the affinities 
inter se being strong and those with the two Natural Orders weak it would 
be proper to make a new natural order. In the second сазе the group would 
naturally be attached to the natural order with which its affinity predominated.. 
In the third case there being strong affinities with both natural orders the 
boundary between the two for practical purposes would be broken down 
and some systematic compromise would be necessary to avoid their fusion into: 
one. І hope to be able to show that the second solution is the one applicable- 
to thisicase— that is to say, that Saurauja is inseparable from Ternstroemiacere, . 
while the remaining two, Actinidia and C lematoclethra, can be properly 
excluded from Dilleniacee. 
Sauranja differs from all Dilleniaceze in Из coherent petals, versatile 
anthers, united styles, numerous small seeds, and in the moderate (not minute) 
embryo. All these characters are common to а considerable number of the- 
genera of Ternstræmiaceæ, some, indeed, to all of them.  Clematoclethra 
closely agrees with Sawrauja. The only point, on the other hand, in which 
they agree with Dilleniaceæ rather than with Ternstrœæmiaceæ is in the. 
presence of raphide-bearing cells, and this anatomical character will probably 
not be considered of so much systematic importance as those of floral 
structure. lt now remains to be seen whether Actinidia, which undoubtedly 
possesses close affinities with Dillenia, can properly be excluded from 
ОШешасею. Actinidia agrees with Dilleniaceæ in its many-celled ovary, its. 
