In the second place, we have no warrant for the assumption that the phantom seen, even though it be somehow _caused_ by a deceased person, _is_ that deceased person, in any ordinary sense of the word. Instead of appealing to the crude analogy of the living friend who, when he has walked into the room, _is_ in the room, we shall find for the ghost a much closer parallel in those hallucinatory figures or phantasms which living persons can sometimes project at a distance. But experience shows that when--as with these _post-mortem_ phantoms--the deceased person has gone well out of sight or reach there is a tendency, so to say, to _anthropomorphose_ the apparition; to suppose that, as the deceased person is not provably anywhere else, he is probably here; and that the apparition is bound to behave accordingly. All such assumptions must be dismissed, and the phantom must be taken on its merits, as indicating merely a certain connection with the deceased, the precise nature of that connection being a part of the problem to be solved. And in the third place, just as we must cease to say that the phantom _is_ the deceased, so also must we cease to ascribe to the phantom the motives by which we imagine that the deceased might be swayed. We must therefore exclude from our definition of a ghost any words which assume its intention to communicate with the living. It may bear such a relation to the deceased that it can reflect or represent his presumed wish to communicate, or it may not. If, for instance, its relation to his _post-mortem_ life be like the relation of my dreams to my earthly life, it may represent little that is truly his, save such vague memories and instincts as give a dim individuality to each man's trivial dreams.