WE have seen what Darwin meant by natural selection, and we understand that this process really implies a transformation of organisms by slow degrees, in the direction of adaptive fitness--a transformation which must ensue as necessarily as when a human selector, prompted by conscious intention, tries to improve an animal in a particular direction, by always selecting the 'fittest' animals for breeding. In nature, too, there is selection, because in every generation the majority succumb in the struggle for life, while on an average those which survive, attain to reproductive maturity, and transmit their characters to their descendants, are those which are best adapted to the conditions of their life--that is, which possess those variations of most advantage in overcoming the dangers of life. Since individuals are always variable in some degree, since their variations can be inherited by their progeny, and since the continually repeated elimination of the majority of those descendants is a fact, the inference from these premisses must be correct; there must be a 'natural selection' in the direction of a gradually increasing fitness and effectiveness of the forms of life. We cannot, however, directly observe this process of natural selection; it goes on too slowly, and our powers of observation are neither comprehensive nor fine enough. How could we set about investigating the millions of individuals which constitute the numerical strength of a species on a given area, to find out whether they possess some variable character in a definite percentage, and whether this percentage increases in the course of decades or centuries? And there is, furthermore, the difficulty of estimating the biological importance of any variation that may occur. Even in cases where we know its significance quite well in a general way, we cannot estimate its relative value in reference to the variation of some other character, though that other may also be quite intelligible. Later on, we shall speak of protective colouring, and in so doing we shall discuss the caterpillars of one of the Sphingidæ, which occur in two protective colours, some being brown, others green. From the greater frequency of the brown form we may conclude that brown is here a better adaptation than green, but how could we infer this from the character itself, or from our merely approximate knowledge of the mode of life of the species, its habits, and the dangers which threaten it? A direct estimation of the relative protective value of the two colours is altogether out of the question. The survival of the fittest cannot be proved in nature, simply because we are not in a position to decide, _a priori_, what the fittest is. For this reason I was forced to try to make the process of natural selection clear by means of imagined examples, rather than observed ones.