In the case of retrogressive changes the visible modification is due, at least in the best known instances, to the reduction of an active quality to a state of inactivity or latency. Now if we make a cross between a species and its variety, the differentiating character will be due to the same internal unit, with no other difference than that it is active in the species and latent in the variety. In the hybrid these two corresponding units will make a pair. But while all other pairs in the same hybrid individuals consist of like antagonists, only this pair consists of slightly unlike opponents. This conception of varietal crosses leads to three assertions, which seem justifiable by actual experience. First, there is no reason for a diminution of the fertility, as all characters are paired in the hybrid, and no disturbance whatever ensues in its internal structure. Secondly, it is quite indifferent, how the two types are combined, or which of them is chosen as pistillate and which as staminate parent. The deviating pair will have the same constitution in both cases, being [278] built up of one active and one dormant unit. Thirdly this deviating pair will exhibit the active unit which it contains, and the hybrid will show the aspect of the parent in which the character was active and not that of the parent in which it was dormant. Now the active quality was that of the species, and its latent state was found in the variety. Hence the inference that hybrids between a species and its retrograde variety will bear the aspect of the species. This attribute may be fully developed, and then the hybrid will not be distinguishable from the pure species in its outer appearance. Or the character may be incompletely evolved, owing to the failure of cooperation of the dormant unit. In this case the hybrid will be in some sense intermediate between its parents, but these instances are more rare than the alternate ones, though presumably they may play an important part in the variability of many hybrid garden-flowers. All of these three rules are supported by a large amount of evidence. The complete fertility of varietal hybrids is so universally acknowledged that it is not worth while to give special instances. With many prominent systematists it has become a test between species and varieties, and from our present point of view this assumption is correct. Only the test is of little use in practice, as fertility may be diminished [279] in unbalanced unions in all possible degrees, according to the amount of difference between the parents. If this amount is slight, if for instance, only one unit-character causes the difference, the injury to fertility may, be so small as to be practically nothing. Hence we see that this test would not enable us to judge of the doubtful cases, although it is quite sufficient as a proof in cases of wider differences.