When we enter a room, we seem to see the entire scene at a glance. But, really, it takes time to see — to apprehend all the details and see if they confirm our expectations and beliefs. Our first impressions often have to be revised. Still, one wonders how so many kinds of visual cues can lead so quickly to consistent views. What could explain the blinding speed of sight?
The secret is that sight is intertwined with memory. When face to face with someone you've just met, you seem to react almost instantly — but not as much to what you see as to what that sight reminds you of. The moment you sense the presence of a person, a whole world of assumptions is aroused that are usually true about people in general. At the same time, certain superficial cues remind you of particular people you've already met. Unconsciously, then, you will assume that this stranger must also resemble them, not only in appearance but in other traits as well. No amount of self-discipline can keep those superficial similarities from provoking assumptions that may then affect your judgments and decisions. When we disapprove of this, we complain about stereotypes — and when we sympathize with it, we speak of sensitivity and empathy.
It's much the same with language, too. If someone said, It's raining frogs, your mind would swiftly fill with thoughts about the origins of those frogs, about what happens to them when they hit the ground, about what could have caused that peculiar plague, and about whether or not the announcer had gone mad. Yet the stimulus for all of this is just three words. How do our minds conceive such complex scenes from such sparse cues? The additional details must come from memories and reasoning.
Most older theories in psychology could not account for how a mind could do such things — because, I think, those theories were based on ideas about chunks of memory that were either much too small or much too large. Some of those theories tried to explain appearances only in terms of low-level cues, while other theories tried to deal with entire scenes at once. None of those theories ever got very far. The next few sections describe what seems to be a useful compromise; at least it has led to some better results in some projects concerned with Artificial Intelligence. Our idea is that each perceptual experience activates some structures that we'll call frames — structures we've acquired in the course of previous experience. We all remember millions of frames, each representing some stereotyped situation like meeting a certain kind of person, being in a certain kind of room, or attending a certain kind of party.