SELECTION EXPERIMENTS. 265 



individuals, the actual percentage of specimens showing heritable variations 

 was probably not far from 4 or 5 per cent. 



In plate 26 I have shown the results obtained from three experiments, A, B, 

 and C, each starting from a single pair of beetles where extreme and non- 

 transmissible variations were the parents of each generation and were selected 

 most rigorously, all others being exterminated. In these experiments with 

 general color characters there was no influence exerted upon the progeny by 

 selection of the parents. In general color characters exactly the same results 

 were obtained as in the selection of unit color characters. In plate 27, how- 

 ever, are seen quite different results from those portrayed in plates 25 and 26. 



Starting from a single pair of albinic beetles in which the selected character 

 was transmissible, and following the line of descent from generation to gener- 

 ation, the fact is graphically shown that the particular variation of the parent 

 was not only preserved, but carried close to the limit of normal variability 

 of the species, and that by selection the race was changed from one which 

 was variable to one which was relatively invariable that is, selection resulted 

 in the production of a race of albinic beetles of low variability, which, no 

 doubt, it would have been easy to maintain for a long period of time. From 

 the third generation a selection was made for the parents of the fourth 

 of the most and least albinic individuals, A still being the albinic race, and B 

 the divergent race tending toward the opposite extreme. These two lots of 

 parents gave in the fourth generation two distinct polygons which overlapped 

 only in the slightest extent. By continued selection the polygons in the fifth 

 generation did not overlap, and in this generation further division was made 

 of B into B and C. These two lines were continued for several generations, 

 diverging from the A line, but not far nor rapidly. In the second generation 

 there arose two distinct groups separated by a wide gap, A and D, the latter 

 being the exact opposite of the A race. This D race was propagated, and by 

 selection produced the result shown in the polygons along the line of descent 

 D, giving in the last generation of the race a group of beetles of almost uni- 

 form condition. In all the lines of descent, A, B, C, and D, artificial selection 

 did just what it was found to do in the elements of coloration, namely, it 

 created a race of low variability about the standard chosen, which it main- 

 tained as long as selection was practiced ; but it did not carry the race beyond 

 the normal range of variation of the species. That is, artificial selection can, 

 as De Vries points out, produce races and maintain them, but its power to 

 develop these races beyond the natural range of variability is yet to be demon- 

 strated. 



From the series of cultures represented in plate 27 it is shown that it is 

 easy by selection to create races from a species, which would, as long as the 

 artificial selection lasted, breed true to the ideal chosen. Such an experiment 

 was made and two races breeding true were produced. Their history is rep- 



