AS WE HAVE seen, there are some memory-surfaces which do not just passively record what is put onto them but actually alter the material. A simple and rather interesting model of such a memory-surface can be made with polythene and pins.
A number of pins are stuck just far enough into a white board for them to stand firmly upright. On the heads of the pins is laid a thin polythene sheet which covers the whole board. This is the memory-surface. The input to this memory-surface consists of drops of coloured water which are sprayed from above onto the polythene sheet. These are just as much an input as the rays of light which fall on a film to give the memory-trace known as a photograph. The memory-trace on the polythene sheet is the pattern made by the coloured water wherever this settles.
For the first experiment the coloured water is sprayed quite randomly over the whole surface without any attempt to produce a pattern. But even though the spraying is quite haphazard, at the end of it definite pools of coloured water will be found to form a pattern against the white board; something like the patterns shown here. It appears that the memory-surface has been so active that it has in fact created a pattern out of a random input.
Fig. 13
It is unlikely, however, that the input could ever have been truly random, no matter how hard one might have tried to make it so. The water may have been sprayed a little more heavily in some areas than in others; some areas may have been sprayed earlier than other areas; the polythene may not have been exactly level to start with; the pins may not have been evenly spaced. All these are very small differences; but just enough for the water to collect in one spot rather than in another. The weight of the water would then depress the polythene sheet in that spot. As the polythene sheet is depressed, the water runs off the surrounding areas into that depression, so making the depression deeper. The deeper the depression the more water it collects. And so the process goes on until at the end there is a definite pattern of pools of water instead of an even layer of water over the whole surface.
Even though this rather special type of memory-surface does not actually create patterns out of nothing, it does greatly amplify small differences. It takes hesitant little suggestions of pattern and builds them up into bold definite patterns. The nature of the system makes it an amplifier and a definer of patterns.
The interesting thing is that the memory-surface does not actually do anything. It is quite passive. In the case of the corrugated surface onto which marbles were dropped it was the surface which moved the marbles about and so processed the pattern. But with the polythene and pins model it is the water and not the polythene which organizes the pattern. The system is a self-organizing and a self-maximizing one. The polythene and pins arrangement merely allows the input to organize itself. The memory-surface carries out no active processing, it passively provides an opportunity for the information to be self-organizing. This difference between active processing and merely allowing information to process itself is a very fundamental one and a very important one.
In the second experiment the water is not sprayed randomly but made to fall in a definite pattern. This can be done by puncturing the bottom of a large can with a certain pattern of holes, through which the water drips onto the polythene surface. The pattern that forms on the surface resembles the pattern of holes in the bottom of the can. There may be some distortion, for a pool of water cannot form on top of a pin but only between the pins, but on the whole the pattern will be recorded as it has been received.
If the same pattern is put onto a polythene surface which already holds another pattern that has been put on it before, then the result may be different. When the new pattern does not overlap the established one then the new pattern is accepted as if it were on a virgin surface. But if the new pattern does fall across part of the established pattern then the incoming pattern is likely to be much altered. The pattern already on the surface will have changed the smooth surface into a contoured one. There will be valleys and pools where the water lies, and ridges and hills where the water can no longer settle. Where the incoming pattern falls onto these contours it has to follow them, even if this means a shift from the position it ought to occupy. The water will run off the ridges and into the valleys. Thus if one were to try and impose a straight line pattern athwart a similar pattern that was already established on the surface, then the new pattern would not be recorded but the old one would be reinforced (see below).
Fig. 14
This behaviour of the memory-surface shows how old patterns come to interact with new ones. It shows that the old patterns can actually determine how the new ones are received. This can mean that new information may only be received in terms of the old patterns.
The way a new pattern which is close to an old pattern flows into this and reinforces it can be very useful from a practical point of view. It means that a pattern does not have to be repeated in exactly the same way for it to become well established. If the different patterns, or presentations of the same pattern, are similar enough, then the special properties of the surface will lead to their being treated as identical. This leads to the development of definite patterns even if the input shows much variation. Here again the behaviour of the surface is directed to defining and establishing patterns. It is not the surface itself which is processing the incoming pattern but the previous patterns which have themselves altered the surface. It is the memory of previous patterns that processes the new one.
An obvious advantage of this type of system is that things go on building up with a useful continuity. The disadvantage of the system is that it becomes very difficult to change the old patterns. The established patterns are like the streams and rivers that are formed on the surface of the land by the action of rain. Once the streams and rivers have become established they tend to become self-perpetuating as they drain off into themselves whatever rain falls on the land.
The polythene and pins memory-surface does not process incoming information but offers an opportunity for the information to process itself. The general effect is to define, amplify, build up and establish fixed patterns out of information which may be rather confused. This effect is achieved by allowing incoming information to organize itself, and by allowing already established patterns to guide this organization. Above all, the system is a self-organizing and a self-maximizing one.
This type of surface could be called a preferability surface. Water sprayed randomly on an ordinary type of surface would wet it evenly all over. But on a preferability surface a few deep holes would develop and water would collect in these. In effect, certain areas of the surface come to be preferred by the incoming information.
