450 



THE AMERICAN NATURALIST [Vol. LII 



differences in the mean condition of the snout which dis- 

 tinguish the various local races from one another are 

 hereditary. 



Here, too, I am aware that we could argue, with flaw- 

 less logic, that the piginentless condition was due to the 

 dropping out of some single factor, without which the for- 

 mation of snout pigment in any quantity was impossible. 

 Each member of the graded series of pigmented snouts 

 we might suppose to be conditioned by the presence of 

 this color factor, together with one to many cumulative 

 factors determining the degree of its manifestation. 



Johannsen (1913), Morgan (1915) and others have 

 made much of the increased range of variability which 

 has frequently been met with in the F 2 and subsequent 

 generations, even when appearances otherwise pointed to 

 a permanent blending of types. Kecently several writers, 

 particularly MacDowell (1916) and Little (1917), have, 

 analyzed some of Castle's data and have reached conclu- 

 sions directly opposed to his. All of these authors (Cas- 

 tle excepted) hold that increasing range of variability in 

 successive hybrid generations is strong evidence for the 

 hypothesis of multiple factors, and we must grant that a 

 pretty good case can be made out along these lines. The 

 theory runs smoothly until we encounter the awkward 

 class of facts which Johannsen has called by the name of 

 "transgressive splitting," i. e., the ultimate extension of 

 the range of hybrid variability beyond that of both of the 

 parent races combined. These facts would seem to prove 

 too much, despite the ingenious explanation which has 

 been offered by the pan-Mendelians to account for them. 



An analysis of my quite limited data furnishes no evi- 

 dence of an increased variability in the F„ generation, 

 except where it pretty plainly results from an increase 

 m the amount of actual abnormality, due to the conditions 

 of captivity. In the largest, as well as the most normal 

 of the series, the range of variation actually diminishes 

 when we pass from the to the F 2 generation. I do not, 

 however, offer the present evidence as conclusive, even 

 for the single case of subspecific hybridization in Pero- 

 mi/scus. It should be confirmed by data derived from 



