These plants were again self-fertilised, and the offspring of each plant separately sown. It next appeared that the offspring of the recessives _remained pure recessive_, and in subsequent generations never produced the dominant again. But when the seeds obtained by self-fertilising the dominants were examined and sown it was found that the dominants were not all alike, but consisted of two classes, (1) those which gave rise to pure dominants, and (2) others which gave a mixed offspring, composed partly of recessives, partly of dominants. Here also it was found that the average numerical proportions were constant, those with pure dominant offspring being to those with mixed offspring as one to two. Hence it is seen that the 75 per cent. dominants are not really of similar constitution, but consist of twenty-five which are pure dominants and fifty which are really cross-breds, though, like the cross-breds raised by crossing the two original varieties, they only exhibit the dominant character. To resume, then, it was found that by self-fertilising the original cross-breds the same proportion was always approached, namely-- 25 dominants, 50 cross-breds, 25 recessives, or 1_D_ : 2_DR_ : 1_R_. Like the pure recessives, the pure dominants are thenceforth pure, and only give rise to dominants in all succeeding generations studied. On the contrary the fifty cross-breds, as stated above, have mixed offspring. But these offspring, again, in their numerical proportions, follow the same law, namely, that there are three dominants to one recessive. The recessives are pure like those of the last generation, but the dominants can, by further self-fertilisation, and examination or cultivation of the seeds produced, be again shown to be made up of pure dominants and cross-breds in the same proportion of one dominant to two cross-breds. The process of breaking up into the parent forms is thus continued in each successive generation, the same numerical law being followed so far as has yet been observed. Mendel made further experiments with _Pisum sativum_, crossing pairs of varieties which differed from each other in _two_ characters, and the results, though necessarily much more complex, showed that the law exhibited in the simpler case of pairs differing in respect of one character operated here also. In the case of the union of varieties _AB_ and _ab_ differing in two distinct pairs of characters, _A_ and _a_, _B_ and _b_, of which _A_ and _B_ are dominant, _a_ and _b_ recessive, Mendel found that in the first cross-bred generation there was only _one_ class of offspring, really _AaBb_. But by reason of the dominance of one character of each pair these first crosses were hardly if at all distinguishable from _AB_. By letting these _AaBb_’s fertilise themselves, only _four_ classes of offspring seemed to be produced, namely,