No. 549] ON DIFFERENTIAL MORTALITY 



515 



here, for I am confident that the plantings were true 

 random samples of the general population of seeds 

 weighed. With regard to the second, we note merely that 

 its influence will make the detection of a selective mor- 

 tality more difficult. If we find indications that the 

 chances for development of a seed are conditioned by its 

 weight we must therefore consider that the influence of 

 weight is probably even stronger than is indicated by our 

 evidence. Indeed, there might be a selective death rate 

 scarcely detectible by the methods necessarily used in 

 this study, but if these methods do indicate a -elective 

 elimination, we may have considerable confidence in its 

 reality. 



Studies of viability are generally limited to the capa- 

 city for forming a growing seedling, but there is no rea- 

 son why tests should not be made more stringent by 

 extending them to the capacity of the embryo for devel- 

 oping into a fertile plant. This has been done in these 

 experiments. 



Presentation of Data 



The seriations of weights of seeds for the general 

 populations are given in Table I; those for the sub-series 

 of seeds which actually developed into fertile plants in 

 Table II. The seriations for the grand populations are 

 designated by the key letters of the crops of plants which 

 produced them, those for the viable sub-samples by the 

 key letters of the crops into which they developed. 1 The 

 weight- arc recorded in unit- of .<>_.") urani range; class 1 

 being 0-.025 gram, class 2, .025-.050 gram, etc. The 

 constants are also expressed in terms of these units, but 

 any one desiring to do so may easily transmute them into 

 fractions of grams. 



The biometric constants for the grand populations are 

 given in Table III. Those for the sub-samples which 

 actually produced fertile plants appear in Table IV. 



