HOMOTYPOSIS IN THE VEGETABLE KINGDOM. 
295 
My third series consists of 26 leaves from each of LOO trees in Monmouthshire, 
young trees being included, but the same rules as to gathering l)eiiig observed. The 
whole of the labour of gathering, counting and deducing the constants is due to 
Miss Mildred E. Barwell, of Girton College.* 
The general numerical results are given in the table below :— 
Besemblaiice of Ash Leaves from same Tree. 
Scries. 
Number. 
Mean No. of 
leaflets. 
S. D. of 
leaflets. 
Correlation. 
Trees. 
Leaves. 
Pairs. 
1 
Buckinghamshire 
I Dorsetshire . 
i Monmouthshire . 
1 
109 
120 
100 
28.34 
3120 
2600 
708.50 
78000 
65000 
10-1295 + -0214 
9-7260 + -0239 
9-8766 ± -0265 
1-6891 + -0151 
1- 9759 + -0169 
2- 0058 + -0188 
-3743 
•3964 
•4047 
+ -0109' 
■± -0102' 
A -Ollf 
t 
Mean of series . 
— 
— 
— 
9-9107 
1-8903 
•3918 
Thus a total of 329 trees, giving 8,554 compound leaves counted and yielding 
213,850 pairs on the symmetrical tables, were dealt with in all. 
It will be seen at once that the mean number of pinnce to the leaf, and their 
variability differs considerably with the locality, the age and the environment of the 
tree, but the correlation of leaves from the same tree is fairly alike for the tliree 
series, and almost identical with the theoretical value ‘4000 given by the Law 
of Ancestral Heredity for the degree of resemblance between pairs of brothers. 
The following is the actual distribution of the frequency of the leaflets :—■ 
Number of Pinnm on Leaves. 
Series. 
3. 
4. 
5. 
6. 
7. 
8. 
9. 
10. 
11. 
12. 
13. 
14. 
15. 16. 
Total. 
Buckingh’shire 
3 
0 
16 
21 
201 
67 
879 
156 
1140 
68 
257 
9 
17 0 
2834 
Dorsetshire. 
4 
0 
84 
30 
396 
115 
959 
228 
911 
72 
280 
11 
29 ' 1 
3120 
Monmouthshire 
1 
5 
42 
24 
279 
55 
836 
143 
896 
83 
216 
6 
13 j 1 
2600 
T otal . 
8 
5 
142 
75 
876 
237 
2674 
527 
2947 
223 
753 
26 
59 1 2 
1 8554 
* Tliis was a iemarl<;ably satisfactory piece of worLv, carried through from the collecting of the raw 
material down to the calculation of the constants, the statistical methods having to lie mastered and 
applied during the course of the investigation. 
t In determining the probable errors of the correlation coefficients, there is some question as to whether 
we should use the number of individuals, which seems too small, the number of pairs, which seems too large, 
or the number of individual organs dealt with. I have used the latter, but enclose the probable errors in 
brackets to mark the doubt. 
