NO. 564] STUDIES ON DIFFERENTIAL MORTALITY 753 



Only 8 are 4 or more times their probable error. Of 

 these, 3 are positive, averaging + 3.34, while 5 are nega- 

 tive, averaging — 4.45. 











Milll] l||J|| 























The differences for A-C, all material, are shown in 

 Diagram 5. Note by comparison with Diagrams 3 and 4 

 that the evidence for selective mortality becomes stronger 

 when variabilities corrected for size of the means are 

 substituted for absolute values. 



