ON VARIATIONS. 369 



and so are placed in the -p3 per cent, column, 

 and 4 are diminished by 2 per cent., and so are placed 

 in the 1 per cent, column. Four more are placed 

 in the + 5 per cent, column, and 4 in the 3 per 

 cent, column, and so on. A similar process was ap- 

 plied to all the other measurements, as far as possible, 

 the 6 individuals 7 per cent, larger than the average 

 being, for instance, increased and diminished by 4, 

 6, and 8 per cent. The curve of distribution of the 

 324 individuals now takes the form of the lower figure. 

 "We see that it is much more flat-topped, indicating that 

 the range of variation is much greater than before. 

 This is, in fact, more than doubled, the arithmetical 

 mean error being increased from 3.2 per cent, to 

 6.8 per cent. The individuals now vary in size by 

 17 per cent., so Natural Selection can act with much 

 greater celerity and certainty than before. Thus no 

 less than 32 of the 324 individuals are now 11 per cent, 

 or more larger than the average, and so offer a very ap- 

 preciable handle for Selection to work upon. The dis- 

 tribution of the dots and crosses shows us, also, that all 

 the extremely large individuals are also individuals 

 which were larger than the normal before the variable 

 environment was brought to bear on them. Many of 

 the larger individuals were rendered smaller than the 

 average by the action of an unfavourable environment, 

 and many of the smaller rendered larger by a favour- 

 able environment i. e., there has been a good deal of 

 mixing of the individuals as originally distributed but 

 the fact remains that the extremely large individuals, 

 which Natural Selection would be especially likely to 

 favour, and the extremely small ones, which it would be 



