In species with separate sexes there is the further difficulty that two individuals must be chosen for each mating, and superficial examination of them does not insure that they belong to the same group--their germ plasm cannot be inspected. Hence selection of biparental forms is a precarious process, now going forward, now backwards, now standing still. In time, however, the process forward is almost certain to take place if the selection is from a heterogeneous population. Johannsen's work was simplified because he started with pure lines. In fact, had he not done so his work would not have been essentially different from that of any selection experiment of a pure race of animals or plants. Whether Johannsen realized the importance of the condition or not is uncertain--curiously he laid no emphasis on it in the first edition of his "Elemente der exakten Erblichkeitslehre". It has since been pointed out by Jennings and by Pearl that a race that reproduces by self-fertilization as does this bean, automatically becomes pure in all of the factors that make up its germ plasm. Since self-fertilization is the normal process in this bean the purity of the germ plasm already existed when Johannsen began to experiment. HOW HAS SELECTION IN DOMESTICATED ANIMALS AND PLANTS BROUGHT ABOUT ITS RESULTS? If then selection does not bring about transgressive variation in a general population, how can selection produce anything new? If it can not produce anything new, is there any other way in which selection becomes an agent in evolution? We can get some light on this question if we turn to what man has done with his domesticated animals and plants. Through selection, i.e., artificial selection, man has undoubtedly brought about changes as remarkable as any shown by wild animals and plants. We know, moreover, a good deal about how these changes have been wrought. (1) By crossing different wild species or by crossing wild with races already domesticated new combinations have been made. Parts of one individual have been combined with parts of others, creating new combinations. It is possible even that characters that are entirely new may be produced by the interaction of factors brought into recombination. (2) New characters appear from time to time in domesticated and in wild species. These, like the mutants in Drosophila, are fully equipped at the start. Since they breed true and follow Mendel's laws it is possible to combine them with characters of the wild type or with those of other mutant races.