Mr Gardiner, On the Unit of Classification , etc. 423 
On the Unit of Classification for Systematic Biology. A Reply 
to Mr Bernard. By J. Stanley Gardiner, M.A., Fellow of 
Gonville and Caius College. 
[ Read 5 May 1902.] 
In a paper on the above subject ( Proc . Camb. Phil. Soc. Yol. xi. 
p. 268, 1901) Mr Bernard proposes a method of classification on 
a novel system by localities. The specimens of a collection are 
to be divided into the forms for each locality, and each form is 
to be termed X. loc. 1, 2, 3, 4, 5 etc. in accordance with the number 
of forms in that locality. 
I may say at the outset that Mr Bernard’s scheme has many 
merits, which have been very ably put forward, but I must at the 
same time remark that it is open to several objections, so serious 
that, in my view, its ultimate adoption is impossible. It is, however, 
eminently desirable that the method should be tested on a few 
genera of corals, and it is hoped that the authorities of the British 
Museum with their immense collections will publish their next 
few catalogues of corals in this form. In any case such catalogues, 
considering variation in a thorough and unique manner, as it is 
necessary they should, cannot fail to be of the highest scientific 
interest and importance. 
In the first place no attempt has been made by Mr Bernard 
to define accurately what is meant by the term form. Each form 
is said to be “ an aggregate of structural characters regarded in the 
abstract.” Now form is taken as the unit of classification, and 
for scientific purposes, as Mr Bernard proposes to use it, must 
be capable of exact definition. Is it, however, more fitted for 
accurate determination than species or variety ? So far as I have 
seen in corals the specimens of any collection have three classes of 
characters. Supposing that fifteen characters — a purely arbitrary 
number — may be perceived in the skeleton of any specimen, 
about five will belong to the genus, including in this class the 
family and group characters as well, five to the species and five 
to the specimen. The generic characters can admittedly be 
