Of all parts of those variable objects which we call organisms, the most variable is the reproductive system; and the variations may carry with them functional changes, which may be either in the direction of increased or of diminished fertility. Consequently variations in the way of greater or less fertility frequently take place, both in plants and animals; and probably, if we had adequate means of observing this point, we should find that there is no one variation more common. But of course where infertility arises--whether as a result of changed conditions of life, or, as we say, spontaneously--it immediately becomes extinguished, seeing that the individuals which it affects are less able (if able at all) to propagate and to hand on the variation. If, however, the variant, while showing some degree of infertility with the parent form, continues to be as fertile as before when mated with similar variants, under these circumstances there is no reason why such differential fertility should not be perpetuated. Stated in another form this suggestion enables us to regard many, if not most, species as the records of variations in the reproductive systems of their ancestors. When variations of a non-useful kind occur in any of the other systems or parts of organisms, they are, as a rule, immediately extinguished by intercrossing. But whenever they arise in the reproductive system in the way here suggested, they tend to be preserved as new natural varieties, or incipient species. At first the difference would only be in respect of the reproductive systems; but eventually, on account of independent variation, other differences would supervene, and the variety would take rank as a true species. Now we must remember that physiological isolation is not like those other forms of isolation (e.g. geographical) which depend for their occurrence on accidents of the environment, and which may therefore take place suddenly in a full degree of completeness throughout a large section of a species. Physiological isolation depends upon distinctive characters belonging to organisms themselves; and it would be opposed to the whole theory of descent with progressive modification to imagine that absolute sterility usually arises, in a single generation between two sections of a perfectly fertile species. Therefore evolutionists must believe that in most, if not in all cases--could we trace the history, say of any two species, which having sprung from a single parent stock on a common area, are now absolutely sterile with one another--we should find that this mutual sterility had been itself a product of gradual evolution. Starting from complete fertility within the limits of a single parent species, the infertility between derivative or divergent species, _at whatever stage in their evolution this began to occur_, must usually at first have been well-nigh imperceptible, and thenceforth have proceeded to increase stage by stage.