The epibolic gastrula of Mammalia (_vide_ pp. 215 and 291) is a still more striking case of a secondary embryonic process, and is not directly derived from the gastrula of the lower Chordata. It probably originated in connection with the loss of food-yolk which took place on the establishment of a placental nutrition for the foetus. The epibolic gastrula of the Scorpion, of Isopods, and of other Arthropoda, seems also to be a derived gastrula. These instances of secondary gastrulæ are very probably by no means isolated, and should serve as a warning against laying too much stress upon the frequency of the occurrence of invagination. The great influence of the food-yolk upon the early development might be illustrated by numerous examples, especially amongst the Chordata (_vide_ Chapter XI.). If the descendants of a form with a large amount of food-yolk in its ova were to produce ova with but little food-yolk, the type of formation of the germinal layers which would thereby result would be by no means the same as that of the ancestors of the forms with much food-yolk, but would probably be something very different, as in the case of Mammalia. Yet amongst the countless generations of ancestors of most existing forms, such oscillations in the amount of the food-yolk must have occurred in a large number of instances. The whole of the above considerations point towards the view that the formation of the hypoblast by invagination, as it occurs in most forms at the present day, can have in many instances no special phylogenetic significance, and that the argument from frequency, in favour of invagination as opposed to delamination, is not of prime importance. A third possible method of deciding between delamination and invagination is to be found in the consideration as to which of these processes occurs in the most primitive forms. If there were any agreement amongst primitive forms as to the type of their development this argument might have some weight. On the whole, delamination is, no doubt, characteristic of many primitive types, but the not infrequent occurrence of invagination in both the Coelenterata and the Porifera--the two groups which would on all hands be admitted to be amongst the most primitive--deprives this argument of much of the value it might otherwise have. To sum up--considering the almost indisputable fact that both the processes above dealt with have in many instances had a purely secondary origin, no valid arguments can be produced to shew that either of them reproduces the mode of passage between the Protozoa and the ancestral two-layered Metazoa. These conclusions do not, however, throw any doubt upon the fact that the gastrula, however evolved, was a primitive form of the Metazoa; since this conclusion is founded upon the actual existence of adult gastrula forms independently of their occurrence in development. [FIG. 205. DIAGRAM SHEWING THE FORMATION OF A GASTRULA BY DELAMINATION. (From Lankester.)