Let's make this concrete. Your eye looks at a fish. This causes your brain to form a visual image of a fish. So far, your primary representation 'fish' still has accurate truth relations with the outside world. The real fish has fins, eyes and gills, and so does your image of the fish. Or your eye looks at a woman, and this causes your brain to form a visual image of the woman. Now you not only have a primary representation of a fish, but you also have a primary representation of a woman. This image, like the one of the fish, is also truthful. The woman you looked at has long hair and an alluring smile, and so does your primary representation of the woman.
In Leslie's important theory, to create such images or primary representations, the only hardware needed is a visual system that starts with an eye and ends in the visual cortex of the brain. But recall that that is only the first of his three steps. To move beyond imagery to imagination, to progress to steps two and three, one now needs an extra, special neurological mechanism. This extra mechanism can take each of the two primary representations (fish, and woman), and make copies of them. Whereas our brain previously just had two primary representations, it now has two second-order representations as well. So that was step two accomplished.
Finally, enter step three. This same special mechanism can now introduce modifications to the second-order representations at whim. It can for example delete some features on each of these second-order representations. Let's delete the head of the fish and delete the legs of the woman. And whilst we're at it, let's delete her long hair. Clearly these second-order representations are no longer veridical, that is, they no longer refer to anything in the outside world truthfully. But that's precisely the point. The brain is there as an evolved organ to represent what is going on in the outside world veridically. If there's a lion out there, the brain needs to know the image created by the visual system is accurate, so it can take the necessary action (fight or flight). But the human brain (whilst not wishing to sacrifice this important survival function of imagery) can be ratcheted up to do more than just represent the outside world veridically, and modifying second-order representations opens up a world of new possibilities. It allows the brain to think about the possible, the hypothetical, about currently-untrue states of affairs.
He ends up arguing that this ability to form "meta-representations" of things is directly linked to our theory of mind. In other words, we evolved an imagination so that we could imagine what other people are thinking. Austic individuals, he argues, are characterized by their absence of imagination:
Children with severe or classic autism may end up with an exclusive interest in the real world, with no interest at all either in mind-reading, pretending, or fiction. They may enjoy making patterns with real objects, or watching how real objects behave, but not even spare a thought for how someone else might be feeling or what they might be thinking, or understand why a mermaid or a unicorn is a fun idea.
My main problem with Baron-Cohen's version of the imagination is that it seems unnecessarily confined. He has reduced our imagination to a limited sphere of social cognition (and a specific nub of brain), since it's specifically designed to form "meta-representations" of other people's thoughts. But this limited view misses the way our imagination is embedded into the most basic aspects of cognition and perception. Wallace Stevens said it best: "Reality is a product of the most august imagination."
This is a romantic idea, but it's neurologically accurate. A glance at any of our senses quickly demonstrates that much of what we think of as being out there - in our sensations of the outside world - actually came from in here, from the top-down projections inside the mind. To take a single neural example: in the laternal geniculate nucleus (LGN), the thick nerve that connects the eyeball to the brain, ten times more fibers project from the cortex to the eye than from the eye to the cortex. As William James wrote in Pragmatism: "A sensation is rather like a client who has given his case to a lawyer and then has passively to listen in the courtroom to whatever account of his affairs the lawyer finds it most expedient to give." The inner lawyer in our brain is our imagination.
So if I one had to define the imagination (and I'm not sure the imagination needs a definition), I'd go with the druggy notions of Coleridge, not the adaptationist conjectures of Baron-Cohen. I think the scientist underestimates the way our imagined meta-representations of the world are constantly filtering down and influencing our actual representations of the world. "The imagination dissolves, diffuses, dissipates, in order to recreate," Coleridge wrote. "It is essentially vital, even as all objects (as objects) are essentially fixed and dead." The poet is right.
I think that imagination works exactly opposite to the characterization in your post. Everything that we perceive and know, we know by our imaginations. What makes us "in touch" with reality is some strange and mysterious ability to test our imaginitve fantasies for vallidity in the real world, some sort of feedback mechanism, and then our ability to assign some degree of reality to some of what we imagine, and reserve the rest as recognized fantasy.
hmmmm Not sure this makes as much sense as I thought it would. (I guess you had to be there, in my brain).