No. 517 J HEAVY AND LIGHT SEED GRAIN 49 



ance of evidence in favor of the large seed. The errors 

 are as apt to tell against the heavy seed as against the 

 light seed. In fact, where error has heen most carefully 

 eliminated, as in the experiment of Zavitz, the large 

 seed gives the most striking positive results. 



'Idle result, empirically derived, while of great prac- 

 tical importance, does not throw much light on transmis- 

 sion. The majority of the experimenters have paid no 

 attention to the plants from which the large or small 

 grains have come. Bolley selected large and small grains 

 from the same heads of wheat and found that the large 

 grains generally produced the largest yields. Lyon has 

 stated that both large and small grains in a lot of wheat 

 must represent both large and small spikes, and if only 

 large grains are sown one is not necessarily selecting 

 from the best plants. 



If, according to Johannson.- •' In a population contain- 

 ing only one single type, the selection of fluctuations has 

 no action at all," then it would make no difference, as far 

 as transmission is concerned, if all sorts of plants were 

 represented in the seed, so long as we are dealing with a 

 pure line. Most American breeders, however, would 

 prefer to select for seed the best plants from the field 

 each year, even if working with a pure strain. This prac- 

 tice is doubtless based on opinion at the present time 

 rather than on well-grounded experimental knowledge. 

 AVe ought to he willing to acknowledge our ignorance of 

 the possibilitv of changing the type by selecting fluctua- 

 tions of close pollinated cereals. Cntil more and accurate 



Correlation Data of Oats 



The writer, in securing some statistical data on oats 

 preparatory to breeding, noticed that the data were in- 

 teresting in connection with the question of light and 

 heavy seed. Measurements were taken on 1,000 oat culms 

 grown at Dickinson, Xorth Dakota, under field conditions. 

 In nearly all cases, each head hearing culm measured rep 



^Rpt. Third Int. Conf. on Genetics, London. 1906. 



