476 'Journal of Comparative Neurology and Psychology. 



As to point 3, I have treated the data for the cell-bodies, using 

 the uncorrected moments, but without effecting any improvement 

 on the results. 



Finally I have tried to split the observed frequency curve (point 

 4) into two normal curves according to the method given by 

 Pearson ('94). After laborious calculations it was found that 

 the present data can not be split. The reason for this conclusion 

 is omitted since it needs an elaborate mathematical presentation. 

 The result shows however that there is not the slightest indica- 

 tion of separate groups in the cell-population. . 



Therefore the cause of disagreement must depend on other 

 conditions than those already enumerated. 



After failing to obtain in this way a reasonable explanation for 

 the considerable deviations between the observed and theoretical 



Fig. 3. Diagram of the spinal ganglion cell containing nucleus and nucleolus. 



curves, it occurred to me that the explanation might be found in 

 the method of sectioning and measurement. In order to make 

 clear the relations existing within the ganglion let us suppose that 

 8000 spinal ganglion cells of various forms (from spherical to 

 oblong) are thoroughly mixed in an ovoid receptacle. This is 

 then cut into 80 slices of equal thickness. The entire series is 

 sampled by taking three slices from the middle and three shces 

 from the midway between the middle and extremes on both sides. 

 Thus nine slices are selected for examination. The slices of the 

 ganglion which we have examined in this way contained 1 108 cells. 

 Under these conditions the knife cuts the individual ganglion cell 



