§ 89. Thus far we have contemplated the process under its simplest aspect. While we have assumed the two parents to be somewhat unlike, we have assumed that each parent has a homogeneous constitution--is built up of physiological units which are exactly alike. But in no case can such a homogeneity exist. Each parent had parents who were more or less contrasted--each parent inherited at least two orders of physiological units not quite identical. Here then we have a further cause of variation. The sperm-cells or germ-cells which any organism produces, will differ from each other not quantitatively only but qualitatively. Of the slightly-unlike physiological units bequeathed to it, the reproductive cells it casts off cannot habitually contain the same proportions; and we may expect the proportions to vary not slightly but greatly. Just as, during the evolution of an organism, the physiological units derived from the two parents tend to segregate, and produce likeness to the male parent in this part and to the female parent in that; so, during the formation of reproductive cells, there will arise in one a predominance of the physiological units derived from the father, and in another a predominance of the physiological units derived from the mother. Thus, then, every fertilized germ, besides containing different _amounts_ of the two parental influences, will contain different _kinds_ of influences--this having received a marked impress from one grandparent, and that from another. Without further exposition the reader will see how this cause of complication, running back through each line of ancestry, must produce in every germ numerous minute differences among the units. Here, then, we have a clue to the multiplied variations, and sometimes extreme variations, that arise in races which have once begun to vary. Amid countless different combinations of units derived from parents, and through them from ancestors, immediate and remote--amid the various conflicts in their slightly-different organic polarities, opposing and conspiring with one another in all ways and degrees; there will from time to time arise special proportions causing special deviations. From the general law of probabilities it may be concluded that while these involved influences, derived from many progenitors, must, on the average of cases, obscure and partially neutralize one another; there must occasionally result such combinations of them as will produce considerable divergences from average structures; and, at rare intervals, such combinations as will produce very marked divergences. There is thus a correspondence between the inferable results and the results as habitually witnessed.