520 E. A. MINOHIN. 



radiate system the first sign of activity is the division of the 

 nucleus in a neighbouring pore-cell. An outgrowth from the 

 pore-cell containing one of the nuclei then extends over the 

 triradiate system, and in this extension of the cell the gastral 

 ray arises close to the nucleus. Finally the outgroveth be- 

 comes nipped off from the pore-cell to form the actinoblast. 

 We see that the gastral ray may make its appearance when 

 the actinoblast has become all but separate from the pore-cell 

 (PI. 42, figs. 49 and 50), or long before this event takes place 

 (PI. 39, figs. 18 and 18a). The amount of separation between 

 the two cells probably depends upon how far distant the pore- 

 cell originally was from the triradiate system. 



The origin of the gastral rays iu the oscular rim has not 

 been followed by me in the same detail, but only sufficiently to 

 make their origin clear, which I think is done by PI. 41, fig. 38. 

 Here we see a triradiate system with its six cells lying under 

 the coarse granular epithelium of porocytes ; one of the latter, 

 which already contains two nuclei, is engaged in the formation 

 of the fourth ray. The figure shows well the great difference 

 between the basal formative cells and the gastral actinoblast. 

 PI. 41, fig. 41, shows an almost identical state of things in a 

 section of a decalcified specimen of the sponge. I have not 

 seen the earliest stages of the gastral actiuoblasts in this 

 region, so cannot say whether they represent the whole of a 

 porocyte, or whether, as is more probable, a porocyte gives off 

 a cell to form a gastral ray, as in the cases already described. 



Having, as I think, made sufficiently clear the origin of the 

 gastral actinoblasts, it only remains to follow their subsequent 

 history, which I have done in the case of Clathrina contorta, 

 CI. reticulum, and CI. cerebrum. Their behaviour in the 

 different species presents some points of interest, and is most 

 conveniently studied in sections, which should not be too thin 

 — 10 ju at least. 



To begin with contorta. We have seen that the nucleus 

 of the actinoblast divides into two (PI. 41, figs. 38 and 41 ; 

 PL 43, figs. 51 — 53). The cell does not, however, divide 

 Each nucleus soon divides again (PI. 43, fig. 54). Finally 



