458 
Selective Elimination in Staphylea 
shape of the fruit varies considerably, and the length affords only a rough measure 
of the size, but it is ample for present purposes. 
Fruits which have reached a length of 11 — 15 mm. may be considered to be 
a selected class as compared with those 6—10 mm. in length — if selection occurs 
at all — and those of 16 — 20 mm. in length may be regarded as a selected class as 
compared with both of the lower classes. 
III. Statement of Problems and Methods of Investigation. 
In this memoir five characters of the ovary are considered. They are : 
(1) Number of ovules per locule. 
(2) Total number of ovules per ovary. 
(3) Radial asymmetry of the ovary with respect to the number of ovules 
per locule. 
(4) Number of locules per ovary with an "odd" number of ovules. 
(5) Number of locules per ovary. 
To demonstrate that there is a selective elimination for any or all of these 
characters, it is necessary to show that there are differences between the constants 
for the ovaries which develop into fruits and those which do not, of such an order 
of magnitude that they cannot reasonably be referred to the probable errors of 
random sampling. In these pages I follow the rather common example of statis- 
ticians in regarding differences of at least 2'5 times their probable errors as 
significant. In addition I have demanded reasonably constant results from the 
series of individual plants. 
The working hypothesis of the existence of a selective elimination for any 
character whatever may be tested by determining the difference between the 
constants for : 
(a) A sample from an original population before elimination and a sample 
of the eliminated individuals. 
{h) A sample from an original population before elimination and a sample 
of the individuals remaining after elimination. 
(c) A sample of ehminated individuals and a sample of individuals unaffected 
by the elimination. 
Here "individual," "sample" and "population" are used in their commonly 
accepted and convenient statistical sense. By individual we understand a par- 
ticular ovary, by sample the collection of ovaries with which the limitations of 
time permitted us to deal, and by population the total ovaries produced by a 
given shrub, or the total produced by all the shrubs investigated, as the case 
may be. 
