No. 564] SIT DIES OX DIEEEREXTIAL MORTALITY 



755 



This apparently contradictory result finds an explana- 

 tion when the means are taken more fully into account. 

 Consider these, A-B: 



Eestricting to those at least 2.5E : 



The mean weights are higher in series B than in series 

 A. The relative variabilities for B are, therefore, re- 

 duced by the higher values of the means. Thus when we 

 take the comparison A-B for relative variabilities, the 

 reduction which we noted in dealing with the absolute 

 values does not appear. 



Consider now the result of combining all seeds which 

 germinated at all (whether normally or abnormally) and 

 comparing their coefficients of variation with those of the 

 seeds which failed to germinate, (A+B) —C: 



striding to differences at least 2.')E : 



Thus we have for all the material a deviation of 

 13 ± 2.38 cases from the equality to be expected if there 

 were no selective mortality tending to reduce variability. 

 In 38 cases out of 50 the variability of the seeds which 

 germinated is lower than that of those which failed. 



