522 



THE AMERICAN NATURALIST [Vol. XLVI 



to develop, thus bringing about a reduction in variability 

 which is independent of any change in the mean. The 

 results given in the final column of the comparison table 

 V, evidence even more strongly than the standard devia- 

 tions in favor of such a selective elimination. They are 



lower for the seeds which develop into mature plants in 

 21 out of the 28 cases. This is a deviation from equality of 

 7 ± .67449 V.5X.5X28 = 7 ± 1.79. 

 The averages are: 



For positive deviations, -f .325 

 For negative deviations, — .712 

 For all deviations, — .453 



Diagram 3 makes the reason for these differences clear 

 to the eye. With regard to their probable errors, the six 



