It is, however, not only the immediate ancestors, that is to say, the parents, that have to be taken account of in this mingling of hereditary contributions, but also those more remote. Not a few characters in the child do not occur in either parent, but were present in the grandparents, and their reappearance is called 'atavism' or 'reversion.' Let us consider this phenomenon in more detail, and try to find out whether and how far it can be interpreted by means of our theory. The simplest and clearest cases are again found among plant-hybrids. It may happen, for instance, that a hybrid between two species, when dusted with its own pollen, gives rise to descendants, some of which resemble only one of the ancestral forms: thus we have reversion to one of the grandparents. The explanation of this lies in the different modes in which the reducing divisions are effected; if they take place in such a manner that all the paternal ids of the hybrid are separated from the maternal ids, then the result is germ-cells which are like those of the grandparents, that is, those of the parent species, and these, if they happen to combine in amphimixis, must give rise to a pure seedling of one or other of the two ancestral species. This case occurs less rarely than was formerly supposed, and than it could do if absolutely free combination of the idants took place at reduction. If combination were quite unrestricted, all other possible combinations would be likely to occur as frequently as these. But recent experiments have shown that, in many plant-hybrids, the germ-cells of the hybrids which are fertilized by their own pollen are either purely paternal or purely maternal. There cannot, therefore, be free combination of the idants at the reducing divisions; the idants of the two parent-forms separate from one another and do not combine. It is doubtful, however, whether the same thing occurs within a race, for instance, in the case of reproduction within a human race. In Man reversion to a grandparent occurs not infrequently, and we may explain it thus: the id-group which controlled the type of the grandfather was also contained in the germ-cell which gave rise to the existence of the father, but it did not dominate the type in that case because a more powerful id-group was opposed to it in the germ-cell of the grandmother. When, later on, at the reducing divisions of the germ-cells of the father, this id-group again arrived in the sperm-cells of the father, it would predominately control the type of the child, that is, of the third generation, provided that the egg-cell with which it combined contained a weaker id-group.