An idea about how imagery works… (this post is motivated by Eric’s post ‘Can You Directly Will Sensory Experiences?’)
The perception is full with noise. Imagery then consist in making part of the noise more salient, and pushing other part of the noise in the background.
So, it would be similar to how when you look at the clouds you can “see” different faces, animals and objects. Or how when you look at a wall with rough bumpy surface, you can also “see” different things.
If this is true, imagery would be in fact real – a specific part of the present noise. (Of course by real, I don’t mean that image of banana would be a banana, just that it is part of a real noise, and not something produced by the mind – it would be just putting attention on something, while ignoring other things).
The consequence is that if we can affect this noise then it would result also with affecting our imagery.
Per Perky subjects reported that bananas they imagined on a screen started to rotate when in fact a real, but very dim image of banana was projected on the screen. Also, the subjects apparently didn’t notice the image of banana, though it was clearly visible to the other observers. I think both of this things can be explained by the assumption that imagery is produced by selective focus. Because selective focus depends on what is in perception, the imaginary banana was affected, and because the attention was on specific part of the noise, people failed to see the dim projection of the banana on the screen.