NATURAL SELECTION 299 



appears much more probable in fact that differences in the 

 genitalia are usually the result rather than the cause of the 

 cessation of interbreeding. If this be true, then the differentia- 

 tion of the genitalia cannot have taken place under the action 

 of selection (at any rate in this particular way), since, if the 

 forms have already ceased to interbreed, there is no advantage 

 in developing mechanical difficulties to crossing. 



Actually the specific differences in the genitalia appear to 

 be an excellent illustration of the non-adaptive nature of 

 specific characters. There is a general mechanical co-adapta- 

 tion of the sexes, sometimes (but apparently by no means 

 always) very close, but there is no evidence for adaptation in 

 the extraordinary specific diversity. 



There is a considerable difficulty to be met in connection 

 with the co-adaptation of the genitalia in the sexes. This 

 difficulty is much greater for those who believe in the ' lock- 

 and-key ' theory, but is still of some magnitude even if the 

 genitalia are not regarded as the most important means of 

 isolation. Any change, in one sex, of a character (whether 

 structural, physiological or habitudinal) directly connected 

 with pairing appears to necessitate a correlated change in the 

 other sex. Thus a new development in the male genitalia 

 requires, in so far as the male and female structures are co- 

 adapted, a corresponding development in the female. Simi- 

 larly, if certain females start to produce a sex-scent of a new 

 character, the male perceptor-organs must be able to perceive 

 the new scent and the males must react to it in the appropriate 

 way. It will be suggested that this parallel evolution would not 

 be very difficult if, at all stages, the amount of change at any 

 one step was very small ; but this gradual evolution is very 

 difficult to explain as an adaptation. For these changes 

 would be adaptive (in the course of the fission of a species into 

 two or more locally adapted races) only in so far as they 

 tended to stop interbreeding and therefore, ipso facto, required 

 correlated change in the other sex ; if the changes were too 

 small to require correlated variation, then they would appear 

 to have no adaptive value in the promotion of fission. It is 

 probable that the division between adaptive and non-adaptive 

 changes in sexual characters may not be quite so sharply 

 marked as has been suggested above ; yet there does appear to 

 be a real difficulty. Petersen (1909, p. 308) has attempted to 



