XVIII THE PROBLEM OF UTILITY 389 



ditions which have produced a definite variation in the 

 whole offspring of all the individuals subjected to it. But 

 it must do more than this. For it must produce a 

 variation so exceptionally stable that it constantly recurs 

 in all the offspring of successive generations, even though 

 those offspring are subjected to considerable change of 

 conditions, as are the individuals of all species except the 

 rarest or the most local. Only with such constancy and 

 stability of inheritance could a useless character become 

 fixed in every individual of a species, which it must be to 

 be a " specific " character. It must, therefore, from the very 

 first have been invariable. But this feature of in- 

 variability without selection has not been found to 

 characterize any variation, whether occurring among wild 

 or domesticated organisms. Such an occurrence would 

 necessarily have forced itself upon the attention of 

 breeders and horticulturists. For if the theory is true 

 that the majority of specific characters are of this useless 

 kind, their occurrence as permanent and unchangeable 

 variations must be a common phenomenon, and we ought 

 to find that foreign plants when first cultivated very often 

 present new characters, not sporadically but appearing in 

 every individual, and which cannot be got rid of, since 

 they do not vary, and selection would therefore be power- 

 less to eliminate them. Has any indication of a phe- 

 nomenon of this kind ever been noted ? 



Causes Alleged to Produce Useless Specific Characters. 



Let us come now to the actual causes said to produce 

 useless specific characters. According to Mr. Romanes 

 they are five in number : Climate, Food, Sexual Selection, 

 Isolation, and Laws of Growth. Let us consider how 

 these are known to act or are alleged to act. Climate 

 and Food undoubtedly produce modification in the 

 individual, but it has not yet been proved that these 

 modifications are hereditary. If this could be proved the 

 whole discussion on the heredity of acquired characters 

 would be settled in the affirmative. The supposed proof 

 that these causes produce definite changes which are 



