The corresponding experiment could be made with plants of a pure sugar-race by pollination with hybrid pollen. The spikes would show exactly the same mixture as in the above case, but now this may be considered as conclusive proof that half the pollen-grains represent the quality of one parent and the other half the quality of the other. Another corollary of Mendel's law is the following. In each generation two groups return to purity, and one-half remains hybrid. These last will repeat the same phenomenon of splitting in their progeny, and it is easily seen that the same rule will hold good for all succeeding generations. According to Mendel's principle, in each year there is a new hybridization, differing in no respect from the first and original one. If the hybrids only are propagated, each year will show one-fourth of the offspring returning to the specific character, one-fourth assuming the type of the variety and one-half remaining hybrid. I have tested this with a hybrid between the ordinary nightshade with black berries, and its variety, _Solanum nigrum chlorocarpum_, with pale yellow fruits. Eight generations of the hybrids were cultivated, [299] disregarding always the reverting offspring. At the end I counted the progeny of the sixth and seventh generations and found figures for their three groups of descendants, which exactly correspond to Mendel's formula. Until now we have limited ourselves to the consideration of single differentiating units. This discussion gives a clear insight into the fundamental phenomena of hybrid fertilization. It at once shows the correctness of the assumption of unit-characters, and of their pairing in the sexual combinations. But Mendel's law is not at all restricted to these simple cases. Quite on the contrary, it explains the most intricate questions of hybridization, providing they do not transgress the limits of symmetrical unions. But in this realm nearly all results may be calculated beforehand, on the ground of the principle of probability. Only one more assumption need be discussed. The several pairs of antagonistic characters must be independent from, and uninfluenced by, one another. This premise seems to hold good in the vast majority of cases, though rare exceptions seem to be not entirely wanting. Hence the necessity of taking all predictions from Mendel's law only as probabilities, which will prove true in most, but not necessarily in all cases. [300] But here we will limit ourselves to normal cases.