478 
Selective Elimination in Staphylea 
This measure is, in short, the standard deviation of the locules of the fruit 
around their own mean. According to this measure the asymmetry of a fruit 
with the formula 7 — 8 — 7 bears the ratio to one of the formula 7 — 8 — 6 of 
•471 : '817 instead of 2 : 4 as indicated by the interlocular difference method. 
Since the locules of a fruit are undifferentiated it does not matter how we 
arrange them for convenience of treatment. I have always copied the results of 
original countings off in a descending series. The two constants for a few 
illustrative ovaries are : 
Formula 
Interlocular 
Coefficient of 
Difference 
Asymmetry 
11—11—11 
0 
■0000 
11—10—10 
2 
•4714 
11—10— 9 
4 
•8165 
11— 9— 9 
4 
•9428 
11— 9— 8 
6 
1-2472 
11— 8— 8 
6 
1-4142 
11— 9- 7 
8 
1 -6330 
11_ 8— 7 
8 
r6997 
10— 6— 6 
10 
1-8856 
10— 7— 5 
10 
2-0548 
11— 7— 6 
10 
2-1602 
These illustrations show that while the interlocular difference does not 
distinguish between the amount of irregularity of certain fruits, the coefficient of 
asymmetry, as I have called the standard deviation of the locules of a fruit 
around their own mean, does. By mere inspection I am quite unable to decide 
whether a fruit of the formula 10 — 6 — 6 is more irregular, or radially asymmetrical, 
than one of the formula ]0 — 7 — 5. Both have the same mean number of ovules 
per locule and the interlocular difference is the same for both, but the coefficient of 
asymmetry is slightly higher for the formula 10 — 7 — 5. 
I have adopted the coefficient of asymmetry in this paper for the following 
reasons. 
(a) It is merely the standard deviation — so universally employed in modern 
statistical work — of the number of ovules per locule in an individual fruit. 
(6) It differentiates — whether with quantitative accuracy or not — between 
degrees of asymmetry not distinguished by the interlocular difference method. 
(c) From data tabled in coefficient of asymmetry classes the interlocular 
differences may be obtained by the use of a table. The converse is not true. 
There is one patent objection to the use of the coefficient of asymmetry as 
defined here : it is not independent of number of seeds per locule. There are 
reasons for regarding fruits of the formula 10 — 6 — 8 and 5 — 3 — 4 as equally- 
asymmetrical, but according to our coefficient their asymmetries bear to each 
other the ratio of the square roots of 8/3 and 2/3. 
The simplest method of freeing our constant from the influence of the absolute 
number of ovules would seem to be to take the ratio of the coefficient of 
asymmetry to the mean number of ovules per locule. The reason that I have 
