134 Shull. 



a measure of genotypic constitution is apt to lead to quite factitious 

 conclusions; consequently, such use of the correlation table should be 

 most guai-ded and genetic inferences should be drawn from it with the 

 utmost reserve. If two different size - characters , such as number of 

 leaves and total area of leaves, show increased variability in the Fg, 

 this increase in each case may be interpreted logically as the result 

 of segregations among plural Mendelian determiners which affect these 

 characters, but a low statistical correlation between such characters 

 does not necessarily indicate that the series of determiners which affect 

 the one character is in large measure distinct from the set of genes 

 which modify the other character. Both may owe their gi-eater F2 

 variability to exactly the same set of segregating genes, the low corre- 

 lation being due to the fact that the independent fluctuation of the 

 two characters is much gi'eater than the modifications produced by each 

 of the several hereditary factors which affect them. Such indei)endent 

 fluctuations are readily comprehensible when it is remembered that the 

 number of leaves and the area of each leaf are determined at different 

 times, and concei\able undei' the influence of very different elements 

 of the environment. 



Neither does a high coefficient of correlation between the quanti- 

 tative variations of two characters prove that any of the Mendelian 

 genes which affect those characters are coupled, as assumed by Tammes 

 (1913). She found by careful measurements of length and breadth of 

 seeds, and length and breadth of petals, and by estimations of the in- 

 tensity of pigmentation of the petals, in her Fä Linum-hybrids, that 

 all of these characters are positively correlated; the larger the petals 

 borne by one of these F2 hybrids, on the whole, the deeper i)lue is 

 the color of its flowers, and the larger its seeds. All of these charac- 

 ters exhibit increased variability in the F^, thus indicating that they 

 are jirobably controlled or affected by segregating Mendelian determiners 

 (Tammes 1911). The obvious basis for the inference that genetic coup- 

 ling exists among some of the plural factors which affe(;t these charac- 

 ters in flax, is the assumption that they are specific determiners for 

 the particular quantitative character under investigation, and therefore 

 essentially duplicate. I believe I have made it sufficiently clear by the 

 foregoing discussion that nothing in the observed facts warrants such 

 an assumption. Such correlations arc readily understood if we keep in 

 mind the fact that the various unit-characters are always compound, — 

 the result of the combined action of a g(me and the rest of the geno- 

 type, under limiting conditions supplied by both th(> internal and the 



