due to variability, but as a return to the former state of some ancient progenitor. Independently of the appearance of new characters from crossing, a few words may be added to what has been said in former chapters on the unequal combination and transmission of the characters proper to the two parent-forms. When two species or races are crossed, the offspring of {267} the first generation are generally uniform, but subsequently they display an almost infinite diversity of character. He who wishes, says Koelreuter,[643] to obtain an endless number of varieties from hybrids should cross and recross them. There is also much variability when hybrids or mongrels are reduced or absorbed by repeated crosses with either pure parent-form; and a still higher degree of variability when three distinct species, and most of all when four species, are blended together by successive crosses. Beyond this point Gaertner,[644] on whose authority the foregoing statements are made, never succeeded in effecting a union; but Max Wichura[645] united six distinct species of willows into a single hybrid. The sex of the parent-species affects in an inexplicable manner the degree of variability of hybrids; for Gaertner[646] repeatedly found that when a hybrid was used as the father, and either one of the pure parent-species, or a third species, was used as the mother, the offspring were more variable than when the same hybrid was used as the mother, and either pure parent or the same third species as the father: thus seedlings from _Dianthus barbatus_ crossed by the hybrid _D. chinensi-barbatus_ were more variable than those raised from this latter hybrid fertilised by the pure _D. barbatus_. Max Wichura[647] insists strongly on an analogous result with his hybrid willows. Again Gaertner[648] asserts that the degree of variability sometimes differs in hybrids raised from reciprocal crosses between the same two species; and here the sole difference is, that the one species is first used as the father and then as the mother. On the whole we see that, independently of the appearance of new characters, the variability of successive crossed generations is extremely complex, partly from the offspring partaking unequally of the characters of the two parent-forms, and more especially from their unequal tendency to revert to these same characters or to those of more ancient progenitors. * * * * *