It was recognized, of course, that many embryonic stages could not possibly represent ancestral animals. A young fish with a huge yolk sac attached (fig. 6) could scarcely ever have led a happy, free life as an adult individual. Such stages were interpreted, however, as _embryonic_ additions to the original ancestral type. The embryo had done something on its own account. In some animals the young have structures that attach them to the mother, as does the placenta of the mammals. In other cases the young develop membranes about themselves--like the amnion of the chick (fig. 7) and mammal--that would have shut off an adult animal from all intercourse with the outside world. Hundreds of such embryonic adaptations are known to embryologists. These were explained as adaptations and as falsifications of the ancestral records. [Illustration: FIG. 7. Diagram of chick showing relations of amnion, allantois and yolk. (After Lillie.)] At the end of the last century Weismann injected a new idea into our views concerning the origin of variations. He urged that variations are germinal, i.e. they first appear in the egg and the sperm as changes that later bring about modifications in the individual. The idea has been fruitful and is generally accepted by most biologists today. It means that the offspring of a pair of animals are not affected by the structure or the activities of their parents, but the germ plasm is the unmodified stream from which both the parent and the young have arisen. Hence their resemblance. Now, it has been found that a variation arising in the germ plasm, no matter what its cause, may affect any stage in the development of the next individuals that arise from it. There is no reason to suppose that such a change produces a new character that always sticks itself, as it were, on to the end of the old series. This idea of germinal variation therefore carried with it the death of the older conception of evolution by superposition. In more recent times another idea has become current, mainly due to the work of Bateson and of de Vries--the idea that variations are discontinuous. Such a conception does not fall easily into line with the statement of the biogenetic "law"; for actual experience with discontinuous variation has taught us that new characters that arise do not add themselves to the end of the line of already existing characters but if they affect the adult characters they change them without, as it were, passing through and beyond them. [Illustration: FIG. 8. Diagram of head of chick A and B, showing gill slits, and aortic arches; and head of fish C showing aortic arches. (After Hesse.)] [Illustration: FIG. 9. Human embryo showing gill slits and aortic arches. (After His; from Marshall.)]