If then the two following propositions are admitted (and I think the first can hardly be doubted), viz. that variation of structure takes place at all times of life, though no doubt far less in amount and seldomer in quite mature life{476} (and then generally taking the form of disease); and secondly, that these variations tend to reappear at a corresponding period of life, which seems at least probable, then we might _a priori_ have expected that in any selected breed the _young_ animal would not partake in a corresponding degree the peculiarities characterising the _full-grown_ parent; though it would in a lesser degree. For during the thousand or ten thousand selections of slight increments in the length of the limbs of individuals necessary to produce a long-limbed breed, we might expect that such increments would take place in different individuals (as we do not certainly know at what period they do take place), some earlier and some later in the embryonic state, and some during early youth; and these increments would reappear in their offspring only at corresponding periods. Hence, the entire length of limb in the new long-limbed breed would only be acquired at the latest period of life, when that one which was latest of the thousand primary increments of length supervened. Consequently, the foetus of the new breed during the earlier part of its existence would remain much less changed in the proportions of its limbs; and the earlier the period the less would the change be. {476} This seems to me to be more strongly stated here than in the _Origin_, Ed. i.