56 



THE AMERICAN 



[Yol.XLIV 



here a positive correlation, amounting to 16 per cent. 

 This is a weak correlation. The regression of the average 

 height of culms is very slight. Selection of any particu- 

 lar weight of seed would not imply that any particular 

 heights of plants were involved. 



It is interesting to note that the correlations given of 

 winter wheat hold about the same relations to each other 

 as do the corresponding correlations of the oats. 



Relation of Size to Volume of Grain 

 But little accurate data on this point are available. 

 We should expect to find a close correlation. The data 

 given in Fig. 6 are taken from Bulletin No. 78 mentioned 

 previously. It is seen that there is an almost perfect 

 correlation existing between the average weight of kernel 

 and the average volume per kernel. The three outstand- 

 ing individuals suggest errors rather than extreme 

 variations. 



If oat populations in general show constants similar 

 to those given, then the experimenter selecting the large 

 grains is not selecting from what is commonly considered 

 the best plants, and vice versa. If the plants from large 

 grains produce a better yield, then they must do so by 

 virtue of the increased vigor of the embryo and the in- 

 creased amount of food supply. If we consider that the 

 size and yield of the mother plant have an influence upon 

 the size and yield of the daughter plants, then we must 

 consider that this influence is decidedly less than the in- 

 fluence exerted by the size of the seed. If the size and 

 yield of the mother plant have no effect upon the off- 

 spring, then we might expect the yields from different 

 weights of seed to be somewhat in proportion to the 

 weights of seed. 



Despite all that has been written on the foregoing 

 points we have very little accurate knowledge pertaining 

 thereto, especially relative to the influence of selection 

 upon the close pollinated cereals. 



