CROSSINGS IN MELANIUM-VIOLETS 277 



showing coupling — a correlation between the length of tiic upper 

 petals, the spur-petals, and the lahellum must exist in F. and later 

 generations. 



Although the number ot individuals in /•':. was almost too small, 

 the variation is arranged in correlation tables. Table 12 shows the 

 correlation between the upper petals and the spur-petal; tables 14 and 

 15 between the petals and the labellum, and table 13 between the upper 

 petals and the spur-petal of the same individuals as in table 12. The 

 calculation of tlu' coefficient of correlation was made with aid of 

 Galton's method described in Johannsen's Elemente^ (19131. 



As seen in tables 12 — 15 there is a positive correlation between 

 the length of the petals and the size of the labellum, and, further, 

 between the length of the upper petal and the si)ur-petal. These facts 

 may indicate the existence of only one inhibiting factor or several 

 coupled ones. 



The coefficient of the correlation between the upper petals (r = 0,sr, 

 and 0,74) and the spur-petal is much larger tluin between the petals 

 and the labellum (r = 0,55 and (),J2l. The reason of this may be 

 explained as follows: the upper petals and the spur-j)etal have joint 

 factors with positive effect while the petals have only the inhibiting 

 factor in connnon with the labellum. 



However, the number of individuals investigated is rather small and 

 the values of the coefficient of correlation are therefore rather uncertain 

 — although they are more than trice their standard error. For a 

 positive settling of the correlation it would be necessary to investigate 

 i\ much larger number of individuals as well as the Fs-generation. 

 It is possible that the correlation observed only depends on a mere 

 chance distribution of the values, on account of the small number of 

 individuals; or the value may depend on a correlative modification. 

 1 had no occasion to make such an investigation. However, I consider 

 this correlation — not the values of the coefficient of correlation — 

 rather firmly established. It is not probable that the correlations 

 presented in tables 12 — 15 should be the results of a mere chance. 



THE INHERITANCE OF QUANTITATIVE CHARACTERS 

 OF THE VEGETATIVE PARTS. 



The inheritance of some vegetative characters has also been in- 

 vestigated, most of them only in Fi. Table 16 shows the results ob- 

 tained. The greater size of the particular organ has always been 



