584 NATURAL SCIENCE. oct.. 



least appears probable that if such variability did arise, a new 

 selection would again limit the variation, and reduce the species 

 again to comparative constancy, though perhaps the new species so 

 formed might differ radically from the old one. That such influences 

 may have helped in the comparatively rapid transformation of species 

 in nature, if not proven, at any rate appears possible. 



Whether such direct action of disturbing influences actually has 

 had a share in producing the transformation of some domesticated 

 animals or not, it is difficult to determine ; but the indiyect effect of 

 artificial selection is beyond doubt, and the comparative rarity of the 

 " natural " coloration among rats bred in captivity, and apparently 

 not even subjected to artificial selection, lends some support to my 

 view that no character will remain constant for an indefinite number 

 of generations unless that character be the immediate or remote 

 object of some kind of selection. The phenomena described as 

 resulting from " panmixia " are numerous enough, and sufficiently 

 well known to justify me in stating that what my theory would 

 indicate a priori is in accordance with observed facts. If heredity be 

 a limitation of variation by the indirect action of selection, then the 

 suspension of that limitation, as a consequence of the suspension of 

 selection — in other words, increased variation under domestication — 

 is comprehensible. 



Variation having increased under domestication, a new selection, 

 either natural or artificial, is found, if carried far enough, to produce 

 new limitations of that variation ; that is, to render new groups of 

 characters hereditary, groups of characters which, in the case of 

 domestic " breeds," distinguish each breed from others, as well as 

 from wild congeners. It is thus that we have seen new breeds of 

 poultry produced, breeds which have already become sufficiently 

 constant to entitle them to the distinctive names they have received. 



This constancy, however, is only relative : even the most con- 

 stant of domestic breeds is probably less constant than almost any 

 species evolved under the direction of natural selection. The 

 " heredity " is not only not complete, but it is less so than in natural 

 species. The average frequency of variation in the direction of 

 ancestral natural species has been progressively diminished, but has 

 not been reduced to zero. The stocks which most frequently gave 

 rise to animals possessing the desired characters have been selected : 

 those stocks which reproduced the ancestral characters either more 

 markedly or more frequently have been exterminated, and thus both 

 frequency and distinctness of the obnoxious ancestral characters have 

 been reduced : but the characters do nevertheless still occur, and we 

 call the recurrence of them " atavism." It is, in a certaiii sense, 

 the progressive elimination of atavism by selection which constitutes 

 the formation of a new species or new race. 



Tliough thus offering an explanation of true atavism, I would not 

 for a moment be understood to regard the great bulk of so-called 



