[huntsman] the ASCIDIAN FAMILY CAESIRIDAE 217 



of their intergrading. Nevertheless, it cannot be expected that a 

 character distinguishing a classificatory division, no mattter how high 

 the grade of tiie latter may be, shall be absolutely constant and in- 

 variable in that division, and shall absolutely fail to intergrade with 

 the opposed character distinguishing a neighbouring division. A 

 division into groups should not be condemned merely on account of 

 intergrading in the characters used in making the division, although 

 it is much better if clearly cut distinctions between natural groups 

 can be shown. The important thing is to define groups that are 

 natural. We may objectively define a natural group as one containing 

 units of a lower grade which can be shown to be similar on the basis 

 of as many and as invariable characters as possible. It is important, 

 therefore, to consider as many characters or systems as possible in 

 an attempt to produce a natural classification. 



The principal characters that have been used for distinguishing 

 genera in this family are the following: presence or absence of definite 

 pharyngeal folds, number of folds, arrangement of stigmata, condition 

 of siphonal lobes, position of intestinal canal, number and distribution 

 of gonads. The only features to which sufficient attention seems 

 not to have been paid, and which we wish to emphasize, are the 

 structure, orientation and situation of the gonads. In structure, the 

 chief differences consist in the extent to which the testicular portion 

 is separated from the ovarial part in each gonad, and in the arrange- 

 ment of the testicular lobes in relation to the ovary. At one extreme, 

 the testicular lobes are closely massed around the ovary, and almost 

 enclose the latter, leaving only parts of the inner and outer surfaces 

 free, and the eft'erent ducts unite into a common vas deferens, which 

 is closely applied to the oviduct, the two opening together. At the 

 other extreme, the testicular lobes are more or less distinct from the 

 ovary, and are divided into groups, each with its own vas deferens 

 opening separately into the atrial cavity. In orientation, we have in 

 one small group of species, the gonad reversed from the usual condition 

 and opening at its anterior end instead of its posterior. The situation 

 of the gonad varies considerably on the left side and in relation to 

 the intestinal loop, being either in the latter or above it. In the latter 

 case it may be more or less enclosed by a secondary bending of the 

 intestinal loop around it, or it may be above such secondary bending. 

 A new and clearer order is apparent if we take into account these 

 gonadic characters, and we, therefore, propose to make them the 

 basis of the main divisions in the family, indicating at the same time 

 how they are reinforced by other characters. 



