abilities monitor much more information than central conceptual abilities can pro-cess; and second, central abilities always have plenty of unfinished business. The key problem for efficient short-term information processing is thus to achieve an optimal allocation of central processing resources. Resources have to be allocated to the processing of information which is likely to bring about the greatest con-tribution to the mind’s general cognitive goals at the smallest processing cost. Some information is old: it is already present in the individual’s representa-tion of the world. Unless it is needed for the performance of a particular cogni-tive task, and is easier to access from the environment than from memory, such information is not worth processing at all. Other information is not only new but entirely unconnected with anything in the individual’s representation of the world. It can only be added to this representation as isolated bits and pieces, and this usually means too much processing cost for too little benefit. Still other informa-tion is new but connected with old information. When these interconnected new and old items of information are used together as premises in an inference pro-cess, further new information can be derived: information which could not have been inferred without this combination of old and new premises. When the pro-cessing of new information gives rise to such a multiplication effect, we call it relevant. The greater the multiplication effect, the greater the relevance. Consider an example. Mary and Peter are sitting on a park bench. He leans back, which alters her view. By leaning back, he modifies her cognitive environ-ment; he reveals to her certain phenomena, which she may look at or not, and describe to herself in different ways. Why should she pay attention to one phe-nomenon rather than another, or describe it to herself in one way rather than another? In other words, why should she mentally process any of the assumptions which have become manifest or more manifest to her as a result of the change in her environment? Our answer is that she should process those assumptions that are most relevant to her at the time. Imagine, for instance, that as a result of Peter’s leaning back she can see, among other things, three people: an ice-cream vendor who she had noticed before when she sat down on the bench, an ordinary stroller who she has never seen before, and her acquaintance William, who is coming towards them and is a dread-ful bore. Many assumptions about each of these characters are more or less man-ifest to her. She may already have considered the implications of the presence of the ice-cream vendor when she first noticed him; if so, it would be a waste of processing resources to pay further attention to him now. The presence of the unknown stroller is new information to her, but little or nothing follows from it; so there again, what she can perceive and infer about him is not likely to be of much relevance to her. By contrast, from the fact that William is coming her way, she can draw many conclusions from which many more conclusions will fol-low. This, then, is the one truly relevant change in her cognitive environment; this is the particular phenomenon she should pay attention to. She should do so, that is, if she is aiming at cognitive efficiency. Our claim is that all human beings automatically aim at the most efficient information processing possible. This is so whether they are conscious of it or not;