( xi) 
The question might still be raised whether these were, after 
all, genuine cases of reversion, or whether they were not 
simply to be explained as the like effects of like causes, pro- 
duced de novo in both ancestor and descendant. The latter 
supposition, no doubt, was sufficient to account for some of 
the phenomena observed; but there was a residuum, com- 
prising the more special reversionary features, which could 
hardly be so explained. What was the bearing of these 
latter on the general doctrine of reversion? Current expla- 
nations of atavism as a result of disturbance were inadequate, 
inasmuch as they gave no real reason why the more recently- 
established features should be less stable than those with a 
longer ancestral history behind them. As to the two more 
definite explanations afforded by the theories of Darwin and 
Weismann, there was no doubt that if Darwin’s hypothesis 
of centripetal gemmules were granted, the most usual cases of 
atavism (those following hybridisation) could be explained 
under the theory of pangenesis. The present cases, however, 
stood on a quite different footing, as the new conditions 
determining atavism were only applied at an advanced 
period in the life of the individual, and had no reference to 
the ovum from which that individual originated. On the 
other hand, it seemed that if Weismann’s theory of centri- 
fugal carriers of heredity were assumed, the present instances 
could be explained as being due to the critical influence of 
abnormal temperature-conditions on what Weismann called 
‘the struggle of the ids in ontogeny’’; the new external 
conditions favouring some of the ancestral determinants 
(which ex hypothesi exist in the germ plasm) at the expense 
of those more proper to the species. 
Certain observations seemed to show that some, at least, of 
these features might be hereditary ; and it would be most 
desirable to ascertain whether this were so with all or most 
of them. Their transmission, though not their first appear- 
ance, could be accounted for by pangenesis; but under the 
rival hypothesis it would be necessary—in these cases of 
heredity—to postulate, as Weismann now does, a direct effect 
upon some of the determinants wherever they occur, even in 
