Page:Bergson - Matter and Memory (1911).djvu/170

 ality of the sounds perceived, there must be, of one and the same word, thousands of distinct images. Uttered by a new voice, it will constitute a new image, which will simply be added to the others.

But there is something still more perplexing. A word has an individuality for us only from the moment that we have been taught to abstract it. What we first hear are short phrases, not words. A word is always continuous with the other words which accompany it, and takes different aspects according to the cadence and movement of the sentence in which it is set: just as each note of a melody vaguely reflects the whole musical phrase. Suppose, then, that there are indeed model auditory memories, consisting in certain intra-cerebral arrangements, and lying in wait for analogous impressions of sound: these impressions may come, but they will pass unrecognized. How could there be a common measure, how could there be a point of contact, between the dry, inert, isolated image and the living reality of the word organized with the rest of the phrase? I understand clearly enough that beginning of automatic recognition which would consist, as I have said above, in emphasizing inwardly the principal divisions of the sentence that is heard, and so in adopting its movement. But, unless we are to suppose in all men identical voices pronouncing in the same tone the same stereotyped phrases, I fail to see how the words we hear are able to rejoin their images in the brain.