No. 3.] AMPHIOXUS AND THE MOSAIC THEORY. 605 



worked out with extreme ingenuity, and especially Roux's 

 admirably clear and pointed discussion has the merit of placing 

 the whole question upon a well-defined basis for further investi- 

 gation. Both, however, are based upon two leading assump- 

 tions, which are not only without foundation in observed fact, 

 but are, as I believe, unnecessary for the interpretation of the 

 known phenomena ; and these assumptions I must briefly 

 consider, since Roux has endeavored to interpret some of my 

 own observations, as well as those of Driesch, in accordance 

 with them. 



The first assumption relates to the causes of histological differ- 

 entiation. It is assumed that in normal development differ- 

 entiation is primarily determined by the nature of cell-division, 

 karyokinesis being conceived as qualitative in character in such 

 wise that cells of different prospective value receive correspond- 

 ingly different forms of idioplasm at the moment of their 

 separation. Every cell, therefore, has an independent power 

 of self-determination (" Selbstdifferenzierung ") inherent in the 

 structure of its idioplasm, and this in turn owes its character 

 to the nature of the mitosis by which the cell-nucleus arose. 

 The entire ontogeny is, therefore, compared by Roux to a 

 mosaic-work ; it is essentially a whole arising from a number 

 of independent self-determining parts, though Roux admits 

 that the self-determining power of the cell is capable in some 

 measure of modification through interaction with its fellows. 



The second of the Roux-Weismann assumptions is logically 

 necessitated by the first in view of the phenomena of regenera- 

 tion. Obviously these phenomena are inexplicable under a 

 theory of strictly qualitative division. Both Weismann and 

 Roux, therefore, assume that during cell-division each cell may 

 receive, in addition to its specific form of idioplasm, a portion 

 of unmodified idioplasm afforded by purely quantitative division. 

 This unmodified idioplasm ("accessory idioplasm" of Weismann, 

 or in some cases "germ-plasm" ; "post-generation or regen- 

 eration idioplasm " of Roux), remains latent in normal develop- 

 ment which is controlled by the active specific idioplasm. 

 Injury to the ovum — e. g., mechanical separation of the 

 blastomeres — acts as a stimulus to the latent idioplasm, which 



