514 FRANK SPRINGER 



tht anal structure, while they will differ from each other in the inter- 

 brachials, and partly in the form. In this way the known genera 

 of this group — with the exception of some transition forms whose 

 place is difficult to assign — will fall into three principal family divi- 

 sions, which differ from each other in various degrees, according as 

 one or the other character is given the greater importance. Each 

 of them contains further groupings of genera upon some of the other 

 characters, which might be given subordinate designations according 

 to our notions of their value. 



This might be considered an imperfect attempt to work out the 

 resultant of the several modifications which I have mentioned, and 

 it necessarily encounters difhculties which can be evaded only by 

 some arbitrary — and perhaps temporary — disposition of the dis- 

 turbing elements. As to these no scheme will ever be perfectly 

 satisfactory, and there will always be some shifting of opinion by 

 different observers, and even by the same observer from different 

 points of view. For instance, as to some of the characters, we cannot 

 always give them the same order of precedence in the tables. If we 

 had genera showing every possible combination of the modifications 

 we have discussed, it might be practicable to construct a table with 

 some uniform order of sequence. We do not, however, find all such 

 combinations, and it is quite conceivable that they were never all 

 accomplished. But it is also to be confidently expected that some 

 additional ones will yet come to light, and we can readily point out 

 some vacancies to be filled by future discoveries. 



These three families are not so very different from those hereto- 

 fore proposed, although the grounds upon which they are defined 

 are considerably changed. In my former paper, above cited, I 

 arranged the genera into two main family groups, based upon the 

 difference in the habitus and general form; and I stated that the 

 second family might perhaps be divided into two subgroups. There 

 is no very material difference in the general arangement I now pro- 

 pose, except that I carry this suggestion a step farther, and erect the 

 two subgroups into families of equal rank with the first, thus bringing 

 in an additional family — Sagenocrinidae — between the other two; 

 while also restricting the first group within somewhat narrower limits. 



The definitions of a number of the genera differ considerably 



