34 IMPROVING THE QUALITY OF WHEAT. 



It is therefore quite evident that a sample of wheat taken from 

 spikes of different sizes when separated into lots of light and heavy 

 kernels would have both the larger spikes and smaller spikes repre- 

 sented in each lot of kernels, but doubtless the proportion of kernels 

 from large heads would be greater in the lot of heavy kernels. 



It would appear from these results that the evidence was over- 

 whelmingly in favor of large or heavy wheat kernels for seed. Most 

 of the experimenters selected seed of different kinds each year without 

 reference to previous selection. If large seed or small seed represent 

 plants of different characteristics and if these properties are hered- 

 itary, the results of selection of large or small seeds for several 

 years may be quite different from what they would be the first 3'ear. 

 It is only those experiments in which selection of the same kind of 

 seed has been continued for several generations that ma}^ be relied 

 upon to indicate the value of continuous selection of large kernels 

 for seed. 



Such experiments have been conducted by Sanborn, by Desprez, 

 and by Soule and Vanatter. The work of Desprez indicates that the 

 size of the kernel is a hereditary qualit} . That being the case, it is 

 evident that the small seed of the first separation may be composed 

 partly of seed that is small on account of immaturity and parth^ of 

 seed that is small hj inheritance, but which is perfectly normal. 

 When such seed is planted the immature seed will be largely elimi- 

 nated in the crop, but the naturally small seed will have reproduced 

 itself and will compose most of the crop. When the seed is again 

 separated a much smaller percentage of small seed will be immature, 

 and in consequence a larger number of kernels will produce plants. 

 It would appear from Desprez's experiments, however, that those 

 plants producing small kernels are not so prolific as those producing 

 large kernels. 



Sanborn's results make a ver}^ good showing for the small kernels, 

 but, as before stated, the extreme irregularity would lead to the 

 belief that the soil on the plots lacked uniformit}', or that some other 

 errors had influenced the results. To offset this the tests cover a 

 period of four ^^ears, which should help to rectify mistakes, and in 

 consequence the good showing made by the small kernels is entitled 

 to some consideration. 



Soule and Vanatter's results fulfill exactly the conditions of the 

 hypothesis that the small seed would the first j^ear contain a much 

 larger proportion of immature kernels than it would in subsequent 

 years, and hence yield more poorly the first year. Their results with 

 heavy kernels as compared with ordinary seed offer little encourage- 

 ment to the continuous selection of large kernels. 



