It is obvious that, in an ordinary case, you perceive a table because (in some sense) the table is there. That is to say, there is a causal chain leading backwards from your perception to something outside your body. This alone, however, is hardly sufficient as a criterion. Suppose you smell peat smoke and think of Ireland, your thought can equally be traced to a cause outside your body. The only real difference is that the outside cause (peat smoke) would not have had the effect (images of Ireland) upon every normal person, but only upon such as had smelt peat smoke in Ireland, and not all of them. That is to say, the normal cerebral apparatus does not cause the given stimulus to produce the given effect except where certain previous experiences have occurred. This is a very vital distinction. Part of what occurs in us under the influence of a stimulus from without depends upon past experience; part does not. The former part includes images, the latter consists of pure sensations. This, however, as we shall see later, is inadequate as a definition. Mental occurrences which depend upon past experience are called “mnemic” occurrences, following Semon. Images are thus to be included among mnemic occurrences, at least so far as human experience goes. This, however, does not suffice to define them, since there are others, _e.g._ recollections. What further defines them is their similarity to sensations. This only applies strictly to simple images; complex ones may occur without a prototype, though all their parts will have prototypes among sensations. Such, at least, is Hume’s principle, and on the whole it seems to be true. It must not, however, be pressed beyond a point. As a rule, an image is more or less vague, and has a number of similar sensations as its prototypes. This does not prevent the connection with sensation in general, but makes it a connection with a number of sensations, not with one only. It happens that, when a complex of sensations has occurred at some time in a person’s experience, the recurrence of part of the whole tends to produce images of the remaining parts or some of them. This is association, and has much to do with memory.