We've just seen how one agency could exploit another one by focusing its attention on some object in the outer world. Thus Thirst can make Get reach for a cup — provided there's a cup in view. But what about that fantasy about Professor Challenger, in which there was no real villain on the scene, but just a memory? Apparently one agency can activate another merely by imagining a stimulus! One way to do this would be for Anger to somehow construct an artificial picture that other agencies like See could see. If this were done with enough detail, the other agents couldn't tell that the image wasn't genuine. However, to construct the sorts of images we see on television screens, we'd have to activate a million different sensory nerves, which would require a huge amount of machinery. Besides, we could do more with less: A fantasy need not reproduce the fine details of an actual scene. It need only reproduce that scene's effect on other agencies.
Fantasies usually depict occurrences we've never seen. They need no detailed, realistic images — since the higher levels of the mind don't really see things anyway! Instead, they deal with summaries of signals that come from sensory experience and are condensed at several levels along the way. In the fantasy of Professor Challenger, there was no need to see any of the actual features of Challenger himself; it was enough to reproduce some sense of how his presence once affected us.
What kind of process could reproduce the effect of an imaginary presence? Although scientists don't yet know the fine details of how our vision-systems work, we can assume that they involve a number of levels, perhaps like this:
First, rays of light excite sensors in our retinas. Then, other agents detect boundaries and textures. Then, yet other agents describe regions, shapes, and forms. Then, some memory-frames recognize familiar objects. Next, we recognize structural relationships among those objects. Finally, we relate these structures to functions and goals.
Accordingly, it would be possible to produce illusions by operating at any of these levels. Most difficult of all would be to construct a picture-image by arousing, from inside the brain, the million lowest-level sensor-agents involved in real-world vision. Perhaps the simplest way of all would be to force just the highest-level vision-agents into whichever states would result from seeing a certain scene: this would only require some suitable K-lines. Let's call this a simulus — a reproduction of only the higher-level effects of a stimulus. A simulus at the very highest levels could lead a person to recollect virtually no details about a remembered object or event, yet be able to apprehend and contemplate its most significant structures and relationships while experiencing a sense of its presence. A simulus may have many advantages over a picture-image. Not only can it work more swiftly while using less machinery, but we can combine the parts of several simuli to imagine things we have never seen before — and even to imagine things that couldn't possibly exist.