492 
Selective Eliminatio7i in Staphylea 
As the reader will see, the problems are very complex. Even the relatively 
large series of data available for this research are, I fear, insufficient for a detailed 
study of the questions of the interdependence of these two characteristics of the 
fruit and their fitness. Later I hope to publish extensive data already on hand 
that may throw some light on these questions. For the present I shall only show 
the correlation between the coefficient of asymmetry and the number of "odd" 
locules per fruit. The data for the three series appear in Tables XLI — XLIII, 
and the constants by the product moment method* are given in Table XXV. 
Clearly there is a definitely significant, though not very large, interdependence 
between the two characteristics. 
It may be possible, therefore, that one of these characteristics has compara- 
tively little independent significance in determining the fitness of an ovary for 
continuing its development. If only one of the two characteristics is thus potent 
while the influence of the second is only apparent because of its dependence upon 
the first, we cannot determine at present which is independent and which is 
dependent. 
Personall}', I believe that the coefficients of correlation between them are too 
low to account entirely for the results for elimination that have been secured, and 
that therefore they are both concerned in determining whether or no any ovary 
shall develop. My reasons are as follows : 
If the elimination of the ovaries of one characteristic, a, of the two here 
considered be not primarily due to their possessing their first characteristic but 
merely apparently connected with it because it is itself correlated with the second 
characteristic, h\, we should expect no selective elimination with respect to the 
dependent character within the subgroups of the independent character. Con- 
cretely, if the elimination of ovaries with a larger number of " odd " locules is due 
solely to the fact that these ovaries are also more radially asymmetrical, we would 
expect to find no elimination with regard to number of " odd " locules when we 
work within the same asymmetry class. Similarly for elimination with respect to 
radial asymmetry. Tersely : for constant a is there an elimination depending on h ; 
for constant h is there an elimination depending on a ? 
* Calculated without any combination or modification of classes for the coefficients of asymmetry, 
t Pearson, K. : Phil. Trans. A., Vol. cc. pp. 18, 19, 1902, has termed this indirect selection. 
TABLE XXV. 
Correlation and 
Probable Error 
Series A 
Series B 
Series C 
•2247 ±-0140 
•3296 ±-0121 
•3282 ±-0116 
