30 



THE AMERICAN NATURALIST [Vol. LIII 



account for the facts, it is evident that, assuming all the 

 factors involved to have equal value, we must have an 

 type that is very nearly intermediate as regards its per- 

 centage condition. This we have not observed in any of 

 our series. The type, while being generally inter- 

 mediate, is not so as regards its hulled condition, for it 

 always contains fewer hulled kernels than hull-less. 

 Thus, so far as the percentage relation is concerned, we 

 do not have a strict intermediate. To be sure, there is a 

 reduction of the multiple-flowered spikelet and other 

 changes which cause the type to appear as an inter- 

 mediate. 



With the usual multiple factor hypothesis assuming 

 ordinary segregation, there must be a larger number of 

 individuals ranging from 30 to 70 per cent, than we have 

 at the extremes. With series 379 and 202 we do not have 

 any indication of such a condition. On the other hand, 

 there is a slight suggestion that series 382 does tend 

 more nearly to a frequency distribution such as would 

 usually be expected with the ordinary multiple factor 

 hypothesis. When the third generation distribution of 

 series 202 is observed (Correlation Table V) it is ap- 

 parent that there is more of a tendency to pile up nearer 

 the lower values. When the size of the classes is doubled 

 a decided skew curve is obtained with the mode at class 

 0-9.9, As stated above, the seed sown to obtain the plants 

 used in this distribution was selected from plants of 

 high, low, or medium value, and this may influence to 

 some extent the type of distribution. Yet, when one ex- 

 amines the percentages of the plants used as parents, it 

 is apparent that they are fairly evenly distributed. If, 

 as suggested above, the nearly dominant primary factor 

 pair influences the hulled or hull-less condition and the 

 other factors influence the hulled condition of plants het- 

 erozygous for the primary factor, then we would expect 

 a piling up near the lower values. 



With series 382 there is a tendency for both the third 

 and fourth generation percentage distributions to be 



