Mark Changizi had an interesting insight on how some illusions work [1].
What we see is a reconstruction by the brain interpolated from our sensors. The idea of Mark is that the image is not only an interpolation of the present, but actually also an extrapolation of what the image will be in the next tenth of a second. For tasks such as catching a ball, this would allow us to compensate for the delay of the signal between our brain and our muscles.
Based on this idea, he wrote a classification of many illusions [2].