Schröder's second example is not more convincing, in my opinion, though Semon regards it as one of the most important pieces of evidence. It concerns a leaf-rolling moth, _Gracilaria stigmatella_, the larva of which is said normally to make its house by bending over the _tips_ of the sallow leaves on which it feeds. Schröder placed larvae on leaves from which the tips had been cut, and these larvae made their houses by rolling over the _sides_ of the leaves. Their offspring were again fed on leaves without tips, and as before, they rolled in the leaf-margins either on one side or both. The offspring of this second generation were then fed on entire leaves. There were 19 houses made by these (?19) larvae, and of them 15 were normal, made by folding down the tips of the leaves, while 4 were abnormal, made by rolling in the leaf-margins. Schröder says that in nature he has only twice seen abnormal houses; but it is clearly essential not only that the frequency of such variability in nature should be thoroughly examined, but also that we should know whether when the species is bred in captivity these irregularities of behaviour do or do not occur when the larvae are fed on uninjured leaves.