752 THE AMERICAN NATURALIST [Vol. XLYII 



For cases at least 2.5E, the results are : 



Plus differences 5 +.352 +22.21 



Mini . lift, ,,<„,-,., it; -..V_>1 -2(>.2f, 



Thus the treatment of the abnormal germinations does 

 not materially affect the general results for reduction in 

 variability. 



It seems unnecessary to consider both absolute and 

 relative variabilities for the individual varieties. The 

 results will be summarized for the coefficients of variation. 



Comparison of Relative Variabilities 

 As demonstrated in the preceding sections, mortality is 

 so related to seed weight that absolute variability is 

 reduced in passing from seeds which fail to germinate to 

 those which produce seedlings. Possibly, too, there is a 

 change in type. Such changes in mean, even if due only 

 to the errors of sampling, may somewhat affect absolute 

 variabilities. It is desirable, therefore, to reduce all 

 these to relative terms — to express them as ratios of the 

 absolute variabilities (X100) to the means. 



The coefficients of variation, being already in relative 

 terms, give only one set of means to consider. 

 For A-C, all' series, the results are: 



Thus we have a deviation from the 25:25 ratio of 

 13 ± 2.38 which must certainly be regarded as significant. 

 For cases at least 2.5 times their probable error: 



