SOS 
The seeds of sowings 2 and 5 ‘were obtained after free pollination 
at a time when no other Canna’s were in flower. 
Here therefore the offspring of a single individual is split accord- 
ing to different ratios, whilst there is some suggestion of periodi- 
city. Accidental variations are pretty well excluded on account of 
the considerable numbers employed. The mean error for segregation 
according to 9: 7 is for 1000 individuals 0.2510 per 16 *). Here 
with 1168 plants, it is 0.4384. For the separate sowings, such as 
those which separated into 101: 101 and 107: 108, the differences 
from the mean errors are much larger still. 
There is vet another objection. In the crossing of varieties of 
C. indica it was found that one of the factors for the red leaf 
margin, C, might be separately visible *), and indeed in a segrega- 
tion according to 27: 37 exactly as required by the theory, in 
37—16 = 21 of the 64 individuals. Such plants with a very narrow 
red margin were, however, always present in too small a number 
when the segregation deviated from 27: 37. In the cross with 
C. glauca this shortage extends so far that among the 543 green 
seedlings not « single one was found to have a narrow red leaf margin. 
For the sowings segregating according to 9:7 on the other hand 
an explanation is not readily available, for C may be completely 
coupled to A or B. The ratio 9: 7 also points to the complete 
joining of two of the three factors. 
In the cases of segregation according to 1: 1 we are not concerned 
with a mixture, formed by segregation according to two different 
ratios (viz. 9: 7 and 27: 37) for in that case some of the green 
seedlings would nevertheless have shown at a later stage that they 
possessed the factor C. 
If we adhere to the assumption that, in this case also, C is 
completely coupled to A and B, we cannot attribute the displace- 
ment of the ratio red: green in favour of green to the coupling of 
A and B, for in that case the number of individuals with red leaf 
margin would be inereased; we may, however, attribute the changed 
ratio to repulsion. This repulsion would not even have to go so 
for as to cause the ratio red: green to approximate to 1: 1, for 
AB — (Ab + aB + ab) = (2n? + 1) — [@? — 1) + (n? —1) +1] =2. 
As soon as n, half the sum of the numbers expressing the ratio 
of the gametes, is 5, 6 or more, the difference 2 is, proportionally, 
very small. But then also the Ad or af individuals, amounting to 
1) JOHANNSEN, W., Elemente der: exakten Erblichkeitslehre, 
2) le. b. 18, 
SE — a 
