76 HEREDITY AND VARIATION 



a. If we attend to the nature of the change, we may 

 distinguish " meristic " variations — e.g. in the number and pro- 

 portions of parts, from "substantive" variations of a qualitative 

 sort — e.g. change in colour. 



[3. If we attend to the direction of the change in successive 

 generations, we may distinguish " definite " variations, which 

 occur along one line (like stages in normal development), from 

 "indefinite" variations, which "fluctuate hither and thither 

 with no uniformity in the course of generations." 



Many evolutionists have maintained that there is good 

 reason for believing in definite or determinate variation along 

 particular lines, as if certain organisms had an inherent bias to 

 change in certain parts and not in others, in certain directions 

 and not in others, just as certain inorganic substances can 

 crystallise in different forms but only within strict limits. It 

 is possible to arrange a series of species A, B, C, D, E, F, in such 

 a way that they suggest progressive definite variation along 

 a particular line, and it seems not unlikely that this kind of 

 evolution may sometimes occur. Moreover, along quite different 

 lines of evolution we find evidence that the same kind of step 

 has been taken independently, over and over again. This 

 suggests that the possibilities of variations may be limited and 

 defined by deep-rooted constitutional conditions or physio- 

 logical alternatives. But the weakness of the argument lies 

 in the almost insuperable difficulty of deciding whether the 

 apparent definiteness is not the result of the primary action of 

 selection which eliminates divergent variants at early stages — 

 nipping idiosyncrasies in the bud — or which may have estab- 

 lished a bias in previous generations. In conditions of rigid 

 elimination the lines of variation will naturally tend to become 

 more and more restricted. 



y. If we attend to the amount of the change from one generation 

 to the next, we may distinguish minute fluctuations about 

 a mean, which are connected by intergradcs, from sudden 



