Now consider for a moment the amazing consequences of an hypothesis of this kind. TheIf auditory images, for example, were really stored in the brain, there would be thousands of images for each single word: and then they would be useless. auditory image of a word is not an object with well-defined outlines; for the same word pronounced by different voices, or by the same voice on different notes, gives a different sound. So, if you adopt the hypothesis of which we have been speaking, you must assume that there are as many auditory images of the same word as there are pitches of sound and qualities of voice. Do you mean that all these images are treasured up in the brain? Or is it that the brain chooses? If the brain chooses one of them, whence comes its preference? Suppose, even, that you can explain why the brain chooses one or the other; how is it that this same word, uttered by a new person, gives a sound which, although different, is still able to rejoin the same memory? For you must bear in mind that this memory is supposed to be an inert and passive thing and consequently incapable of discovering, beneath external differences, an internal similitude. You speak of the auditory image of a word as if it were an entity or a genus: such a genus can, indeed, be constructed by an active memory which extracts the resemblance of several complex sounds and only retains, as it were, their common diagram. But, for a brain that is supposed—nay, is bound—to record only the materi-
