*THE POPULAR SCIENCE MONTHLY.*

mean number of ovules is materially increased. For a small series of developing ovaries taken at the Missouri Botanical Garden in 1906, the oldest—the group from which the most elimination had taken place—had about eight per cent, more ovules per fruit than the youngest. In large samples taken in 1908, the fruits which matured had about seven per cent, more ovules than those which were eliminated. The same result is seen if the material is split up into twenty-eight individual pairs of samples, each from a separate tree. This is made clear by Fig. 5. The solid dots connected by broken lines show the percentage excess of the matured fruits over the fallen ovaries in the number of ovules. In twenty-seven cases out of twenty-eight the number is larger in the ovaries which mature!

Fig. 6. Percentage of Ovaries which are perfectly radially Symmetrical in youngest and oldest Collections, Missouri Botanical Garden, 1906.

The ovary of *Staphylea* is three-celled. If each cell contains the same number of ovules,*e. g.,* 8-8-8, it may be regarded as radially symmetrical, while if the numbers differ from locule to locule, for instance 8-7-8 or 9-10-8, the ovary may be described as radially asymmetrical. If this radial asymmetry be expressed by a statistical constant such that a perfectly symmetrical fruit shall have a degree of asymmetry of 0, while the coefficient increases as the ovaries become more irregular, one can compare asymmetries in the ovaries which do and those which do not develop to maturity as easily as he can the means.

Fig. 6 shows the percentage of perfectly symmetrical ovaries in the youngest and oldest series of the 1906 collection. The conclusion that the conspicuously higher percentage of perfectly symmetrical ovaries in the oldest collection is due to a selective mortality by which the more irregular ones are weeded out is fully substantiated by the statistics of 1908.

The average asymmetry for the matured fruits for 1908 is about seventeen per cent, lower than the mean for the eliminated sample. For the individual trees the results are somewhat more irregular than they were for the mean number of ovules, but Fig. 5 shows that in