ME. G. J. EOMANES OS PHTSIOLO&ICAL SELECTIOlSr. 349 



the average mean of all individual variations, any con- 

 siderable departure from this average being, hovrever, checked 

 by intercrossing. 



But now, should intercrossing by any means be prevented, 

 there is no reason why unuseful variations should not be per- 

 petuated by heredity quite as well as useful ones when under 

 the nursing influence of natural selection — as, indeed, we see to 

 be the case in our domesticated productions. Consequently, if 

 from any cause a section of a species is prevented from inter- 

 crossing with the rest of its species, we might expect that new 

 varieties — for the most part of a trivial and unuseful kind — • 

 should arise within that section, and that in time these varieties 

 should pass into new species. And this is just what we do find. 

 Oceanic islands, for example, are well known to be extraordinarily 

 rich in peculiar species ; and this can best be explained by con- 

 sidering that a complete separation of the fauna and flora on 

 such an area permits them to develop independent histories of 

 their own, without interference by intercrossing with their origi- 

 nally parent forms. We see the same principle exemplified by 

 the influence of geographical barriers of any kind, and also by 

 the consequences of migration. For when a species begins to 

 disperse in different directions from its original home, those 

 members of it which constitute the vanguard of each advancing 

 army are much more likely to perpetuate any individual varia- 

 tions that may arise among them, than are the members which 

 still occupy the original home. Not only is the population much 

 less dense on the outskirts of the area occupied by the advance 

 guard ; but beyond these outskirts there lies a wholly unoccupied 

 territory upon which the new variety may gain a footing during 

 the progress of its further migration. Thus, instead of being 

 met on all sides by the swamping efi"ects of intercrossing with its 

 parent form, the new variety is now free to perpetuate itself 

 with comparatively little risk of any such immediate extinction. 

 And the result is that wherever we meet with a chain of nearly 

 allied specific forms so distributed as to be suggestive of migra- 

 tion with continuous modification, the points of specific difference 

 are trivial or non-utilitarian in character. Clearly this general 

 fact is in itself enough to prove that, given an absence of over- 

 whelming intercrossing, independent variability may be trusted 

 to evolve new species. The evidence which I have collected, and 



