520 



SCIENCE 



[N. S. Vol. XXXII. No. 824 



of so great complexity we should keep con- 

 stantly in mind the fact that it is idle to 

 attempt to untangle the whole snarl at once. 

 Eather we should try to study each factor in- 

 tensively, isolating it when possible from 

 others and measuring its force. If in doing 

 this we find that variations do occur and in 

 abundance, we have demonstrated an impor- 

 tant physiological fact, but a fact without 

 significance in Darwinian evolution unless the 

 variations be both heritable and some of them 

 of such superior utility that they have an 

 advantage over others in the struggle for ex- 

 istence. Again the demonstration of the in- 

 heritance of any character does not yield con- 

 clusive evidence on Darwinian evolution 

 unless accompanied by a proof of its selective 

 value. Finally it is quite conceivable that a 

 stringent selection should recur every genera- 

 tion without effecting any change in a 

 species — such being necessary to maintain the 

 type in its present condition or without sig- 

 nificance because acting upon characters not 

 inherited. 



In the face of these difficulties only one 

 course is open to the naturalist: to spend 

 much time in the potting shed and the breed- 

 ing pen, to be strenuous in the use of the eye 

 piece micrometer, the calipers, the color scale, 

 the statistical tables and the calculating ma- 

 chine; to believe that ten times the conven- 

 tional number of observations are desirable; 

 to repeat his experiments and to make new 

 series of measurements; and to believe that a 

 few gourds full of statistical constants with 

 tabulated data from which they may be veri- 

 fied are more to be desired than an artesian 

 well of personal opinion based on non-quanti- 

 tative observations. 



The results outlined in this essay are drawn 

 from recent contributions to these gourds full 

 of quantitative data. The tabulated observa- 

 tions and the detailed analysis from which 

 others may verify these statements if they 

 choose are to appear in a forthcoming num- 

 ber of Biometrika. 



By a selective elimination one understands 

 that the members of a population do not die 



at random, but that some individuals are, be- 

 cause of innate physical, physiological or 

 psychical peculiarities, much more likely to 

 die' than others. 



Theories may be spun concerning the rela- 

 tionships of any character whatever to natural 

 selection, but for purposes of scientific in- 

 vestigation one must limit his attention to 

 those which are directly or indirectly meas- 

 urable. Illustrations of the directly measur- 

 able characters are to be found in Bumpus's 

 sparrows, Weldon's crabs, Crampton's moths 

 and Weldon's and di Cesnola's snails. 



The characteristics of an individual are the 

 sum of the characteristics of several organs: 

 probably it is the fitness of these organs which 

 largely determines whether the individual 

 shall be able to survive and leave the average 

 number of offspring or more. But suppose 

 that each individual produces a great number 

 of organs, only a small fraction of which be- 

 come matured and functional. Might it not 

 be possible to determine from some measur- 

 able character of such organs whether failure 

 to develop to maturity is due to any char- 

 acteristic of the organ — in short, whether 

 there is a selective elimination of organs? 

 Ideal material for such investigation is found 

 in some of the flowering plants. A large num- 

 ber of ovaries are formed, of which only a 

 small per cent, develop to maturity. There is 

 a large elimination: to determine whether 

 this is a selective elimination, whether those 

 which survive to maturity differ in any meas- 

 ure from those which die and fall from the 

 plant, is our problem. 



The American bladder nut, Stapliylea tri- 

 foliaj has a fruit with three cells in each of 

 which from four to a dozen ovules are formed. 

 The number of the ovules can be counted in 

 the ovary of the opened flower and in the ma- 

 tured fruit. Only a small proportion of the 

 ovaries formed reach maturity and by com- 



^ In sexual selection the elimination would occur 

 as a failure to mate. In reproductive selection it 

 would occur as a relatively lower capacity for 

 producing oiTspring. Here only elimination in- 

 volving death — in our present material the death 

 of an organ — is taken into account. 



