The first perception of an object leaves an imprint in the substance of the intellect. A second perception partially resembling the first revives the first to the extent at least of the resemblance. Supposing this is done by a hundred similar objects it is plain that the resembling properties will have been experienced a hundred times, whereas the distinguishing attributes may have been felt a few times only, in some cases only once. Unless we have special reasons for observing the differences and so deepening the impressions of them, they will fade from our memory at a rate corresponding to the paucity of experiences. The most general idea will last longest because _there_ the impression has been very deep. Our idea of Man or Animal will on this principle, as it is found to do in fact, outlast our memory of many concrete men and animals. The objects that contribute to form a general idea or Class are commonly said to 'belong to,' or to 'inhere in,' or to be 'brought under' the idea or class. All these metaphors are wrong and occasion mistakes. Generalisation is nothing but condensed or epitomised recollection; it is practised by ourselves for our own convenience, and does not imply any essential or extra-personal relation between the objects. We are free to classify things in any order we find useful. A farmer's classification of some animals into cattle, game, fowls, birds, and vermin, is perfectly legitimate, for each species is based on a different utility for him. We should distinguish general ideas which we ourselves have drawn from our primary experience, from the ideas suggested by verbal definitions of general ideas formed by other minds. Supposing the objects in question to be quite unknown to us, the definitional idea is more like a particular or imaginary idea than a general idea. It is a single thin rigid idea, utterly unlike the flexible suggestive thought evolved from a large mass of personal experience. Definitional general ideas are as unsatisfactory as described objects, but we are sometimes compelled to use both when personal experience is totally wanting. It is a common error to suppose that general ideas cannot exist in the intellect without words by which to name them. Words and other modes of marking ideas are useful in all departments of thought, but not more necessary in general thought than in any other. An active intellect makes thousands of observations and scores of general ideas which it may have no means or wish to express in language. Generalisation is very like the operation called composite photography. A number of persons are posed in the same attitude and partially photographed on the same plate. The result is an average or mean likeness of the whole group, but not an exact portrait of any individual. So general ideas are 'means' or 'averages' of many resembling but slightly differing objects.