Mr N. N. Sherwood in his useful account of the origin and races of peas (_Jour. R. Hort. Soc._ XXII. 1899, p. 254) alludes to the great instability of this class of pea. To Laxton, he says, “we are indebted for a peculiar type of Pea, a round seed with a very slight indent, the first of this class sent out being _William the First_, the object being to get a very early blue-seeded indented Pea of the same earliness as the Sangster type with a blue seed, or in other words with a Wrinkled Pea flavour. This type of Pea is most difficult to keep true on account of the slight taint of the Wrinkled Pea in the breed, which causes it to run back to the Round variety.” But in Mendelian cases we are concerned primarily not with the product of gametes of blended character, but with the consequences of the union of gametes already discontinuously dissimilar. The existence of pure Mendelian gametes for given characters is perfectly compatible with the existence of blended or mosaic gametes for similar characters elsewhere, but this principle enables us to form a comprehensive and fruitful conception of the relation of the two phenomena to each other. As I also pointed out, through the imperfection of our method which does not yet permit us to _see_ the differentiation among the gametes though we know it exists, we cannot yet as a rule obtain certain proof of the impurity of the gametes (except perhaps in the case of mosaics) as distinct from evidence of imperfect dominance. If however the case be one of a “mule” form, distinct from either parent, and not merely of dominance, there is no _a priori_ reason why even this may not be possible; for we should be able to distinguish the results of breeding first crosses together into _four_ classes: two pure forms, one or more blend or mosaic forms, and “mule” forms. Such a study could as yet only be attempted in simplest cases: for where we are concerned with a compound allelomorph capable of resolution, the combinations of the integral components become so numerous as to make this finer classification practically inapplicable. But in many cases--perhaps a majority--though by Mendel’s statistical method we can perceive the fluctuations in the numbers of the several products of fertilisation, we shall not know whether abnormalities in the distribution of those products are due to a decline in dominance, or to actual impurity of the gametes. We shall have further to consider, as affecting the arithmetical results, the possibility of departure from the rule that each kind of gamete is produced in equal numbers[164]; also that there may be the familiar difficulties in regard to possible selection and assortative matings among the gametes. [164] In dealing with cases of decomposition or resolution of compound characters this consideration is of highest importance.