The difference between physics and psychology is analogous to that between a postman’s knowledge of letters and the knowledge of a recipient of letters. The postman knows the movements of many letters, the recipient knows the contents of a few. We may regard the light and sound waves that go about the world as letters of which the physicist may know the destination; some few of them are addressed to human beings, and when read give psychological knowledge. Of course the analogy is not perfect, because the letters with which the physicist deals are continually changing during their journeys, as if they were written in fading ink, which, also, was not quite dry all the time, but occasionally got smudged with rain. However, the analogy may pass if not pressed. It would be possible without altering the detail of previous discussions, except that of Chapter XXV, to give a different turn to the argument, and make matter a structure composed of mental units. I am not quite sure that this is the wrong view. It arises not unnaturally from the argument as to data contained in Chapter XXV. We saw that all data are mental events in the narrowest and strictest sense, since they are percepts. Consequently all verification of causal laws consists in the occurrence of expected percepts. Consequently any inference beyond percepts (actual or possible) is incapable of being empirically tested. We shall therefore be prudent if we regard the non-mental events of physics as mere auxiliary concepts, not assumed to have any reality, but only introduced to simplify the laws of percepts. Thus matter will be a construction built out of percepts, and our metaphysic will be essentially that of Berkeley. If there are no non-mental events, causal laws will be very odd; for example, a hidden dictaphone may record a conversation although it did not exist at the time, since no one was perceiving it. But although this seems odd, it is not logically impossible. And it must be conceded that it enables us to interpret physics with a smaller amount of dubious inductive and analogical inference than is required if we admit non-mental events. In spite of the logical merits of this view, I cannot bring myself to accept it, though I am not sure that my reasons for disliking it are any better than Dr. Johnson’s. I find myself constitutionally incapable of believing that the sun would not exist on a day when he was everywhere hidden by clouds, or that the meat in a pie springs into existence at the moment when the pie is opened. I know the logical answer to such objections, and _qua logician_ I think the answer a good one. The logical argument, however, does not even tend to show that there are _not_ non-mental events; it only tends to show that we have no right to feel sure of their existence. For my part, I find myself in fact believing in them in spite of all that can be said to persuade me that I ought to feel doubtful.