SELECTION EXPERIMENTS. 263 



limits of variation, if selection is the powerful evolutionary factor its advo- 

 cates hold it to be. It is possible that my experiments were not conducted 

 over a sufficient number of generations to bring about greater changes. 



From this series of experiments it is evident that in the character chosen 

 some variations are transmitted, others not; and, as far as my experience 

 goes, there is no middle ground. To all appearance these variations are 

 alike, and are distinguishable only by their behavior in pedigree breeding. 



The results of the above experiment are paralleled by similar conditions 

 found in nature. Thus, from selected pairs of parents found in nature, only 

 about 4 per cent pass on their extreme characters to their progeny, showing 

 how small a percentage of the variations are transmissible. In nature, where 

 crossing is going on, such heritable variations are swamped as soon as they 

 arise a fact that I have proven experimentally by allowing such variations 

 to breed freely with the general population. It is only when we practice 

 artificial selection or isolation that they are preserved. With such charac- 

 ters as the spots used in the above series of experiments, which are absolutely 

 neutral and could not by any conceivable process come within the domain 

 of selection, and are yet good, specific characters, it is certain that the 

 character as it exists in the different species could never have come about as 

 the result of selection and extinction. It is possible that it may be correlated 

 with some other characters of direct selective value, but I have not been able 

 to establish any such connection. As far as the experiments go, the variations 

 are either transmissible or not. and there appears to be no intermediate state 

 of partial or weak inheritance. Experiments were also tried with other 

 spots on the pronotum, c, d, and e, and with spots upon the epicranium and 

 upon the legs and ventral surface of the abdominal segments, with exactly the 

 same results. 



When we combine variations and treat them statistically, we get results 

 that are false. Thus, on the basis of statistical method the variations of 

 B, A, A', B' are continuous and of the same kind, while in reality they are 

 different ; and by no method of biometry could we discover the two kinds. 

 These can be discovered only by the observation of individual cases and by 

 breeding. When we examine the pairs possessing heritable variations, it is 

 found that in the population of any generation they stand apart, separated 

 from the mode of the species, and that intermediate states are absent. It is 

 true that when many of these individuals are brought together, "lumped," 

 they form a regular polygon of distribution. The reason for this is that 

 while each individual represents a distinct deviation from the mode of the 

 species, these deviations are not all of the same size some are large, others 

 small, and the latter are far more numerous than the former. Moreover, we 

 do not find many of the larger variations in any given generation. Consid- 

 ered statistically, we should recognize these as continuous variations; exam- 

 19 T 



