J. A. Harris 
457 
proportion of the ovaries develop to maturity ; Series A doubtless contains many 
which are non-selectively eliminated, while B includes many which a little later 
would have fallen from the tree and been classified as A. 
It seems worth while, however, to carry out the comparison between Series A 
and B. The critical tests are those between B and C and A and C, the most 
significant being the latter. 
In Staph i/lea the normal type of ovary is 3-merous. Occasionally individuals 
are found which produce a large proportion of 4-merous fruits. Such a one is 
shrub 36, which was omitted in all these comparisons because I wanted to deal 
with as nearly a "normal" population of 3-raerous ovaries as possible. I feared 
that the 3-merous ovaries produced on a shrub which bore a very high percentage 
of 4-merous fruits might be in some characteristics — morphological or physio- 
logical — different from those formed by the growing points of an individual with 
almost exclusively trimerous tendencies. 
Shrub 28 bore no flower in 1908. 
In dealing with the raw data it has seemed desirable to treat the material 
for each individual separately, and for the following reason. Rather extensive 
experience in dealing with biometric constants calculated from the fruits of indi- 
viduals has shown that there may be very real differences between those of 
different plants. If each individual did not contribute the same quantity of 
material to the three collections studied, it is quite possible that in dealing with 
lumped samples differences in constants might arise quite independently of any 
influence of selective elimination. I think that with 28 individuals the danger 
of error from this source is not great, but conclusions based on the results from 
28 individual small samples will be stronger than those resting upon merely one 
large composite series. 
2. The 1906 Collections. 
As indicated above these collections were not made primarily for a study of 
selective elimination at all, but for a study of some questions of fertility and 
fecundity. Constants calculated from them indicated the necessity for the detailed 
study attempted in 1908. The results are treated here as supplementary to those 
from the more detailed investigations made later. 
The collection comprises 270 inflorescences* bearing partly developed fruits, 
none of which had reached a length greater than 20 mm., and was made primarily to 
determine the change in the correlation between the number of ovules formed and 
the number developing into seeds at different stages in the development of the 
fruit. The fruit increases in size very rapidly during its early stages of develop- 
ment, and only a rough measure of its size was considered worth while. Three 
classes, 6 — 10 mm., 11 — 15 mm. and 16 — 20 mm. in length, were recognizedf. The 
* The correlation between the number of flowers formed and tLe number of fruits still continuing to 
develop at the time the inflorescences were examined is discussed in Biometrika, Vol. vi. pp. 440, 441, 
1909. 
+ These are exclusive of the short styles. 
