319 
IN THE TROPICS. 
50: 50 in each cob. The actual percentages were as 
follows :— 
Percentage of 
Plants. 
Percentage of 
Plants. 
Yellow Gfrains. 
Yellow G-rains. 
43 
1 
50 
20 
44 
1 
51 
12 
45 
3 
52 
9 
46 • 
3 
53 
9 
47 
5 
54 
6 
48 
11 
55 
. — 
49 
14 
56 
1 
The curve resulting from these figures is somewhat 
irregular, owing to the small number of cobs examined. 
Nevertheless it shows some resemblance to a normal curve of 
variability. By Galton’s method its probable error is 
approximately 1*6. The question arises whether this amount 
of probable error depends on anything besides the size of 
the samples taken, viz., the number of grains in the cobs ; a 
number which varied from 271 to 862, the average being 560. 
Now according to figures quoted by de Vries (70) from 
Rodewald, the probable error of a number of random samples 
taken from a mixture of two things in equal numbers is :— 
1*69 when the samples include 400 individuals each 
1-52 „ „ 500 
1*38 „ „ 600 
For samples of 560, therefore, the probable error is about 
1*43, but in the present case the samples are not all equal ; 
and as the smaller samples give rise to a proportionately 
greater error, the theoretical value cannot be very far from 
1*6, which is the value got by experiment in the present case. 
The statement therefore appears justified that the difference 
from exact equality of yellow and white grains observed in 
any particular cob depends simply upon the size of the 
sample which such a cob represents, or in other words is 
strictly a matter of chance. For the case, therefore, of the 
female germ cells of a particular variety of maize, it is 
shown that there is a segregation of the allelomorphic 
characters yellow and white in equal numbers ; the result of 
