OCTOBEE 14, 1910] 



SCIENCE 



521 



paring samples of those -whicli fall from the 

 tree with those ■which ripen, one can judge 

 whether elimination has been in any degree 

 dependent upon the number or arrangement 

 of ovules in the locules. Naturally conclu- 

 sions to be valid must be' based upon a very 

 large series of countings, and to be quite sure 

 that the differences are not obscured by 

 heterogeneity of material, the ovaries of each 

 individual should be treated separately. In 

 the spring and summer of 1908, about 7,000 

 ovaries (involving the opening and counting 

 of 21,000 locules) were taken from twenty- 

 eight individuals in the North American tract 

 of the Missouri Botanical Garden. These 

 were in three series ; a sample of flowers which 

 fell from the tree when it was shaken gently 

 and which had therefore ceased to develop 

 and were ready to fall from the tree, a sample 

 of those which remained, and, finally, a collec- 

 tion of the matured fruits later in the sum- 

 mer. 



The second collection represents probably 

 most nearly the condition in the original pop- 

 ulation of pods; it contains some which would 

 have developed to maturity and some which 

 would have fallen from the tree later. The 

 most critical comparison for the detection of 

 selective elimination is that of the eliminated 

 with the matured ovaries. This is the com- 

 parison which will be chiefly employed in these 

 pages.' 



The conviction that there must be a select- 

 ive elimination of ovaries came to me through 

 an extensive biometric study of fertility in 

 various kinds of fruits. The immediate sug- 

 gestion for the detailed investigation begun 

 in 1908 was furnished by a small series of 

 developing ovaries of Staphylea collected, for 

 quite a different purpose in the spring of 

 1906. These fell into three length groups, 

 5-10, 11-15 and 16-20 mm. If selective elim- 

 ination really occurs one would expect the 

 third series, which has most nearly reached 

 maturity, to differ sensibly from the second 

 and especially from the first. The results 



' All comparisons are worked out in the original 

 memoir, which must be consulted for details. 



from this series are in general agreement with 

 those for the 1908 collections, although the 

 method in which they were made prevents their 

 being strictly comparable. The difference in 

 method emphasizes the soundness of the con- 

 clusions drawn. 



Changes in Mean due to Selective Elimi- 

 nation. — In any investigation of natural se- 

 lection the first step is to ascertain whether a 

 difference in the size of organs or in the num- 

 ber of parts can be demonstrated between 

 those individuals which are eliminated and 

 those which survive. Concretely, for our 

 present problem, are ovaries with many or 

 ovaries with few ovules best fitted to become 

 functional ? 



Diagram 1 makes very clear the differences 

 in the average number of ovules per locule 

 for the 1908 series. The arrows show that in 

 27 out of the 28 individuals the result of the 

 elimination has been to raise the average 

 number of ovules per locule by the elimina- 

 tion of those with lower numbers. The 

 amount of difference in the mean of elimi- 

 nated and matured ovaries is shown by the 

 length of the shaft for the individual shrubs 

 and by the two transverse lines for the com- 

 bined collections. The broken line shows the 

 mean for all the eliminated, and the solid line 

 the mean for all the matured, ovaries. The 

 difference between the two is pronounced. 

 Arithmetically it is 



Average for eliminated ovaries = 7.2355 ± .0092 



Average for matured ovaries ^= 7.7474 ± .OOSO 



Difference = ^iTg'iTOm 



Absolutely the difference is only half an 

 ovule, but the number of observations on 

 which this average is calculated is so large 

 that the probable error of the difference is 

 small and its trustworthiness very great. 

 Kelatively the difference represents an in- 

 crease of no less than seven per cent, in the 

 number of ovules per fruit. 



Looking at the diagram again, we note that 

 individuals differ widely among themselves in 

 the lengths of the arrow shafts — the amount 

 of the difference between the eliminated and 



