110 STUDIES IN EVOLUTION 



also in not having an operculum. From this and all other 

 arthropods they are distinguished by having compound eyes 

 on free-cheek pieces which apparently represent the pleura 

 of a head segment that is otherwise lost, except possibly in 

 some forms of stalked eyes and in the cephalic neuromeres of 

 later forms. The most recent discussions as to the affinities 

 of trilobites are to be found in the papers by Bernard,^- ^' ^' i° 

 Kingsley,"'^*' Woodward,^* and the writer,^ where from the 

 facts presented their intimate relationships with the Crus- 

 tacea follow as a necessary corollary. 



Previous Classifications. 



The various schemes of classification that have been applied 

 to the trilobites since that of Brongniart,^i in 1822, have 

 been enlarged and revised by various authors, until at the 

 present time no particular arrangement of the families or 

 genera can be said to be endorsed. The one which is gener- 

 ally recognized as manifestly faulty, that of Barrande,^ is the 

 one most commonly found in text-books and special memoirs. 

 Barrande's definitions and limitations of the generic and 

 family groups were natural and accurate, showing a most 

 complete knowledge of generic and specific values, but in the 

 grouping and arrangement of the families he selected char- 

 acters of secondary rank. 



Of all the investigators who have attempted any classifica- 

 tion of the families, J. W. Salter ^^ seems to have had the 

 clearest insight into the important value of certain characters, 

 and to have approached nearest to a natural system. In 

 zoological research the study of ontogeny and the principles 

 of morphogenesis were then scarcely recognized as having 

 any direct application. It is quite remarkable, therefore, 

 that Salter, as early as 1864, should have singled out, as 

 the basis of his sub-divisions, the characters which are the 

 dominant variants in ontogeny. 



It is not necessary in this place to discuss all the classifi- 

 cations which have been proposed. Barrande ^ gives a com- 

 plete r^sum^ down to 1850, and shows in a very satisfactory 



