One of the means by which the artificial races of animals and plants have been formed by man is selection. The breeder picks out individuals having a certain peculiarity, and allows them to breed together. He hopes to find among the offspring, not only individuals like the parent forms, but also some that have the special peculiarity even more strongly developed. If such are found, they are isolated and allowed to breed, and in the next generation it is hoped to find one or more new individuals that show still more developed the special character that is sought. This process, repeated through a number of generations, is supposed to have led to the formation of many of our various forms of domesticated animals and plants. This heaping up as a result of the union of similar individuals cannot for a moment be supposed to be the outcome of the addition of the two variations to each other. Such an idea is counter to all the most familiar facts of inheritance. For instance, when two similar forms unite, we do not find that the young show all the characters of the mother plus all those of the father, _i.e._ each peculiarity that is the same in both, increased twofold. On the contrary, the young are in the vast majority of cases not essentially different from either parent. A more thorough examination of the facts shows that the problem is by no means so simple as the preceding general statement might lead one to suppose, for our experience shows that it is not always possible to increase all variations by selection, and, furthermore, there is very soon found a limit, even in favorable cases, to the extent to which the process can be carried. The most important point appears to be the nature of the variations themselves which may arise from different causes, and which have different values in relation to the possibility of their continuation.