THE AMERICAN NATURALIST [Vol. XL VII 



NHD-J. . . . 

 NHD-M. . . 

 NHH-J . . . . 

 Mill M . . 

 NHHH-J. . 

 Mill II M. 

 NHDD-J. . 

 MIDI) M 

 NDH-D . . . 

 NDH-E .... 

 NDD-D .... 

 NDD-E. . . . 



NDDD-E. , 

 NDHH-D. 



M)iin /•;. 



FSS-I 



FSS-L 



FSH-C. . . . 



FSD-C 



FSHII ('... 

 F8DD r 



'.17 ;m : 

 14 93 : 

 Vi 72 : 



smaller seeds; we shall have negative differences for 

 standard deviations and coefficients of variation if there 

 is n mortality of both the larger and smaller seeds — thus 

 increasing the variability in the eliminated sample and 

 decreasing variability in the surviving population. 



Hence, regarding the abnormal germinations (tenta- 

 tively) as intermediate between normal development and 

 failure, we take our differences: 



(A) -(C), or normally germinating less failed, 



(B) -(C), or abnormal less failed, 



(A)-{B), or normally germinating less abnormal. 



Since the number of individual experiments is fairly 

 large, the comparisons may be made by merely noting 

 the sign of the differences— i. e., by taking the gross re- 

 sults of the individual experiments. Or one may treat 

 the data from a more numerical view-point, taking aver- 

 ages of the actual differences. Both methods will of 

 course be used. 



In considering the differences between the constants of 

 the three classes of seeds dealt with for the whole experi- 



