October U, 1910] 



SCIENCE 



527 



2E+0 E+20 

 DiAGBAM 5. 



by the elimination. Diagrams 5 for 1908 and 

 6 for 1906 constructed on the same plan as 

 diagrams 3 and 4 make this perfectly clear. 



The two characters, radial asymmetry and 

 locular composition, are not independent. The 

 more asymmetrical fruits are, in the long run, 

 somewhat more likely to have more " odd-" 

 locules than the average. The intensity of the 

 relationship is shown by a correlation of about 

 .300. In consequence of this condition it is 

 not possible to say without further analysis 

 whether both characters are of selective value, 

 or whether the elimination of ovaries with one 

 of the characters — either the presence of 

 " odd " locules or radial asymmetry — is due 

 merely to the fact that they also possess, in 

 some degree, the other. 



2E + E + 20 

 Diagram 6. 



If we divide our material up into groups 

 according to both asymmetry and locular com- 

 position, and then determine whether within 

 these groups there is a change in the mean 

 value of the other character as we pass from 

 the eliminated ovaries to those which have 

 developed to maturity, we shaU, I believe, be 



TABLE m 



1908 Series. Selective Elimination of " Odd " 

 Locules within the Chief Asymmetry Classes 



Kadial Difference in Xumberof Odd Locules 



Asymmetry in Eliminated and Matured Ovaries. 



.0000 — .335 



.47U — .106 



.8165 —.115 



.9428 — .454 



1.2472 — .157 



