46 (238) DOMINANCE. REVERSED DOMINANCE RECESSIVENESS OF 
means decreased intensity of variation and vice versa.” Also CASTLE 
(1919b, p. 2). | 
From the different mendelian investigations it appears, that it is in 
general accepted, that homozygotes are less variable than heterozy- 
gotes. A large variability of heterozygotes is often mentioned. It is 
therefore in agreement with mendelian conceptions, to determine, that, 
the polymery-theory containing that extreme specimens are combina- 
tions of many homozygous factors, it follows, that one must require, 
that the progeny of extreme specimens are less variable than of aver- 
age specimens (cf. 1920). It is a deduction from this theory. 
So it follows from the polymerytheory, that the offspring of extreme 
specimens have a smaller variability than those of average ones, whilst 
one must:still consider, that there are two kinds of average specimens: 
Ist. those, where the formula of heredity contains many heterozygotes 
and 2nd those, where this formula exists of the presence of some homo- 
zygous factors and the absence in homozygous form of the remaining 
factors. The progeny of the first group of average specimens only will 
have a large variability. NILSSON-EHLE has shown through the experi- 
ment, Cultivating F,, the presence in F, of such, different hereditary 
types. We have repeatedly indicated, — see f. i. the curves of fig. la 
(p. 186) — that parents with the same indices can have series of child- 
ren with very different variability of the indices. 
The influence of the different hereditary constitution of the indices 
of parents with high, average and low indices shows thus itself in the 
different variability of the indices of the children. It remains to be 
inquired, if this influence may be demonstrated by statistical research; 
namely among (amidst) other influences. It might bef. i., that extreme 
variations have a large non-hereditary variability (see also JENNINGS, 
1908, p. 511, 522). 
The standarddeviation for the indices of children belonging to clas- 
ses of parents with a different high index has been computated from 
the tab. XV, 1—8. These computations have also been done for the 
head-length and headbreadth as a completion of the tables 3 and 4 
(1920a, p. 134). | 
Tab. XIV has been composed from different groupings of the head 
lengths, in which the parents are grouped in falling succession accord- 
ing to the value of the headlength. The headlengths of both parents 
differ 0, 0.1, 0.2 up to 0.5 cM. From tab. XIV we see, that with the 

