Hatai, Spinal Ganglion Cells of Rat. 475 



section to section. In other words, some sections may contain 

 relatively more cells having larger or smaller diameters, therefore 

 the nine sections selected might not give a proper representation 

 of the entire cell population and therefore the 1108 cells measured 

 might not constitute a real random sampling of the entire popula- 

 tion. 



2. The disagreement between the theory and observation 

 may be due to an improper selection of the unit for grouping the 

 variates. If so, it may be improved by taking some other unit. 



3. The two curves may agree more closely if the uncorrected 

 or raw moments about the mean were used in determining various 

 analytical constants. 



4. The spinal ganglion cells may not represent a homoge- 

 neous population, but a mixture of various groups of elements. 

 If so, dissection of the frequency curve into several components 

 should give a better agreement. 



The question contained in point i has been answered in the 

 following way: the nine sections were divided into three series, each 

 represented by three sections. For the series i, one section was 

 taken from the middle and one from the midway between the 

 middle and the extremes on both sides, while for the series 2, the 

 three sections which lie to the right of the three sections of the series 

 I and for the series 3 those which lie toward left. The percentage 

 values of these three series just mentioned were plotted separately, 

 the same unit of course being used for each series. Comparison 

 shows that these three curves agree with each other in every minor 

 detail, and therefore with the original curve, too. This means 

 that the cell-bodies of different sizes are uniformly distributed, 

 otherwise the three curves should not agree so closely. There- 

 fore the 1 108 cells here measured can be regarded as giving a 

 true representation of the entire cell-population. The disagree- 

 ment found between the theoretical and observed curves is con- 

 sequently not due to a lack of uniformity in distribution. 



The question contained in point 2 was also answered by tak- 

 ing three different units for the cell-bodies; i, 1.8, 2 micra; and 

 one different unit for the nuclei, i micron, and comparing the 

 new results with those already obtained. It was found that these 

 variations in the units did not make any significant alteration in 

 curve. This proves then the difficulty is not due to an improper 

 choice of the unit. 



