Besides the sterility of hybrids appeal has often been made to the phenomenon of incompatibility, in its several stages of completeness, as distinguishing species. No one doubts that incompatibility may arise from a variety of causes of most diverse degrees of importance, but though sometimes referred to as an extreme case of interspecific sterility, it is really a very different matter. In regard to one phase of this incompatibility, that associated with self-sterility, some progress has been made, and we are not wholly without experimental evidence of its being within the range of contemporary variation. Given the outline of Mendelian teaching as to gametic differentiation and the classification of individuals in a mixed population, it seemed highly probable that what we call self-sterility must mean that the species really consisted of _classes_, some of which are capable of interbreeding with others while others are not. According to the received account every individual, though incapable of fertilising itself, was supposed to be able both to fertilise and to be fertilised by any other individual. This notion has always seemed to me a self-evident absurdity, for it would imply that there can be as many categories as individuals. Such experiments, however, as I made did certainly give results consistent with that belief. I first tried Cinerarias, which are usually self-sterile, but I found no incompatible pairs of plants. Whether I was deceived by the consequences of apogamy, or whether the pollen of certain plants may belong to more than one class I do not know. The results were confused in various ways. Usually the self-fertilised plants set little or nothing, and cross-fertilised they set fully with such uniformity that the few failures could plausibly be attributed to mistakes in manipulation or to other extraneous causes. Later de Vries announced[6] (without giving particulars) that he had proved the existence of such classes in _Linaria vulgaris_; but on making experiments with that species I again got no positive results, and I came to the conclusion that in spite of inherent improbability the conventional belief must be substantially true. At last, however, the work of Correns, lately published,[7] does definitely show that in one species, _Cardamine pratensis_, classes of individuals exist such that individuals of the same class are incapable of fertilising themselves or each other, but fertilisation made between the classes is usually completely effective. Many complications were encountered and some contradictory evidence is recorded, but the general bearing of the results was positive and indubitable.