1013 
F, P, 
Length of petal 10.0 mm 8.5—11.2 mm 
Breadth 2 91 ke §.3—10.1 ye 
Colour sl | 5 gef 
Length of petal 9.5 mm 8.2—10.0 mm 
Breadth pe 6. - G0=27 2 
Colour Te hair 1—2 
The above proves that there is still another relation between 
the characters in the plants studied in addition to ordinary 
correlation. The whole phenomenon is only superficially like such 
a correlation. 
Just as any single character which is based on several genes, gives 
in the second generation a pseudo-curve which shows itself as a 
eurve of fluctuating variability but in which the fluctuating variability 
plays only a more or less subordinate role, so also here in 4’, an 
inter-relation of different characters may appear that simulates ordinary 
correlation, but that is in reality a completely different phenomenon 
which is only slightly affected by this correlation. I point this out 
because it seems to me that in studying correlative variability it is 
of the highest importance to investigate only pure homogeneous 
material. Since JOHANNSEN has made known to us the “pure lines”, 
it has become clear that much that was formerly thought to be pure 
material, is a mixture of several forms perhaps also of hybrids. It 
is possible that the correlation found in such material is not a pure 
correlation between the fluctuating variability of the characters but 
is wholly or in part a different correlation phenomenon, This is also 
the case here. We must assume that here a genetic relation exists 
between the groups of factors for the different characters. This relation 
is such that in the formation of gametes in /’, definite combinations 
of factors occur preferentially. In general a tendeney exists to 
make the proportion in the number of factors for the various 
characters such as it was in the original forms or at least to ap- 
proximate to these. This explains that in /’, mere forms arise 
in which the characters all deviate in the same direction from 
the average than should be the case according to the laws of 
probability. 
In the crossing mentioned above the groups of factors for the 
various characters behave with respect to one another differently 
from the way in which the factors for one single character behave 
mutually; for my earlier investigations have shown, that for each of 
66 
Proceedings Rayal Acad. Amsterdam. Vol. XY. 
