STUDIES ON THE COMPAEATIVE ANATOMY OF SPONGES. 23 



another from my very next page and putting it forward as 

 though it were original. 



Sollas's membrane, however, and the intercellular substance 

 which exists between the bases (and, so far as I have seen, 

 between the bases only) of the collared cells have nothing 

 whatever to do with one another. The former is endodermal 

 in origin, and is separated by a wide empty space from the 

 probably mesodermal ground-substance between the collared 

 cells. In the case of Stelospongos, if the mesodermal 

 ground-substance really filled up the whole of the intervals 

 between the collared cells it would be at once recognisable 

 by its highly granular appearance ; but it does not, as a glance 

 at my figures will show ; it stops at the bottom of the neck. 



I must now describe the condition of things in Grantia 

 labyrinthica. All the collars and flagella of the collared 

 cells are retracted in my preparations. This is not to be re- 

 garded as a purely artificial and post-mortem condition, but 

 probably rather as a periodically recurring phase in the life- 

 history of the cells. Carter has shown long since (18) that the 

 individual collared cells may become amceboid, and probably 

 in the living sponge they often spontaneously retract their 

 collars and flagella and enjoy a period of rest. 



In this retracted condition the collared cells (figs. 32, 33) of 

 Grantia labyrinthica are somewhat pyramidal bodies, poly- 

 gonal in transverse section, and with the narrow end pointing 

 towards the lumen of the chamber. They measure about 

 0'0048 mm. in height and about the same in breadth at the 

 base. The nucleus is situate in the apex of the pyramid (fig. 33). 

 This position of the nucleus appears at first sight a little curious, 

 but it is interesting to observe that it agrees with the position 

 of the nucleus in the long prismatic cells of the embryo (fig. 38) 

 from which the collared cells of the adult are admittedly derived. 

 Thus the collared cells in a state of rest revert more or less to 

 their embryonic condition, the chief distinction being that they 

 are now very much shorter. 



Even at their bases the collared cells appear to be separated 

 from one another by distinct intervals (fig. 32), but these may 



