No. 564] STUDIES OX DIFFERENTIAL MORTALITY 



745 



Ideally, to obtain results valid for individuals attain- 

 ing reproductive maturity one should take a small pro- 

 portion of the seeds germinating normally and a much 

 higher proportion of those giving rise to abnormal seed- 

 lings and combine them with the seeds failing to germi- 

 nate. There is.no possible way of estimating the pro- 

 portion of A and B which should be classed with ( . If 

 one wishes to make the comparison which shall bo at the 

 opposite extreme of that in which all seeds germinating 

 at all are compared with those failing to germinate, he 

 ma5 r combine the seeds germinating abnormally with 

 those which do not germinate. Thus {A +B) — C and 

 A — (B -f- C) will give us the upper and lower possible 

 measures of the influence of mortality of abnormal seed- 

 lings on seed weight. 



Turning to the comparison of means for A — (B + C) : 



Or restricting still further to those which are at least 4E: 



Certainly, there is in these figures no trustworthy indi- 

 cation of an increase of mean weight as a result of 

 selective mortality. 



