Page:Bergson - Matter and Memory (1911).djvu/245

way, into definite parts, and the cutting up is not the same when we pass from one copy to another, each of them being in fact characterized by the particular kind of dominant memories on which the other memories lean as on supporting points. The nearer we come to action, for instance, the more contiguity tends to approximate to similarity and to be thus distinguished from a mere relation of chronological succession: thus we cannot say of the words of a foreign language, when they call each other up in memory, whether they are associated by similarity or by contiguity. On the contrary, the more we detach ourselves from action, real or possible, the more association by contiguity tends merely to reproduce the consecutive images of our past life. It is impossible to enter here into a profound study of these different systems. It is sufficient to point out that these systems are not formed of recollections laid side by side like so many atoms. There are always some dominant memories, shining points round which the others form a vague nebulosity. These shining points are multiplied in the degree in which our memory expands. The process of localizing a recollection in the past, for instance, cannot at all consist, as has been said, in plunging into the mass of our memories as into a bag, to draw out memories, closer and closer to each other, between which the memory to be localized may find its place. By what happy chance