220 
E. M. EAST 
were a constant error it would be considerable. But it must be 
remembered that it belongs to the class of accidental errors and that 
in the long run the minus errors are compensated by the plus errors. 
Such compensation can be clearly seen and the accuracy of the 
method perhaps most clearly demonstrated by comparing frequency 
distributions of the same pure line, daughters of the same plant, during 
successive seasons. In a number of cases populations of sister plants 
were grown for two and three years. The seed in each case came from 
single 1909 or 1910 plants, and since the percentage germination 
remained practically constant, the different populations are in the 
nature of duplicate and triplicate determinations. If then the fre- 
quency distributions are sufficiently alike that they may be presumed 
to be random samples of one population, the method is accurate enough 
for genetic purposes. A sample of the result is shown in Table III. 
Table III 
Random Samples of the Same Population Grown in Different Seasons 
Name 
Class Centers in Millimeters 
Means 
34 
37 40 
43 
8s 
88 
94 
97 
100 
N. longiflora, var. A, 1911. 
" " " " 1912. 
" " 1913. 
N. longiflora, var. B, 191 1 
I 
13 
4 
4 
80 
28 
32 
32 
16 
I 
40. 46 ±. 1 1 
40.61 ±.19 
39.76rt.i2 
93.22±.i6 
93.37zt.20 
92.12zb.37 
6 
2 
5 
22 
16 
7 
49 
32 
10 
II 
6 
2 
I 
" " 1912 
1 
When one takes into consideration the difference in size of corolla 
among Nicotiana species and varieties that will cross and give fertile 
hybrids — i. e. N. langsdorffii 21 mm. and N. alata grandiflora 85 mm., 
it is scarcely necessary to enter into a biometrical argument on the pre- 
cision of the method. Here are two small samples of the same popu- 
lation of N. langsdorffii grown in 191 1 and 19 14: 
Designation 
Class Centers in Millimeters 
21 22 
23 
191 1 plants 
3 
9 
12 I 
33 i 7 
2 
I 
1914 plants 
I 
Can it be doubted that the phenotype for corolla length to which 
N. langsdorffii belongs is shown here with an accuracy much greater 
