No. 564] 1NHEIHTAXCI-: OF LEFT IF 1 XDEDXES: 



749 



of the 33 negative differences, only 18 or roughly of) per 

 cent, are not statistically significant. The deviations 

 from equality are 8 ± 2.38 for the whole material and 

 5 ± 1.51 for the 20 series which are more probably statis- 

 tically significant. 



Only 12 individual differences are over four times 

 their probable error: 



These results can leave no doubt as to the reduction in 

 the absolute variability when the seeds which produce 

 normal seedlings are selected out from those which fail 



