In reply to john arran:
Optical illusions are great because they reveal a little bit of what the brain is doing to create visual perception. What we perceive has little to do with the images on the retinas: this is data that goes into an unbelievably complex (compuational?) process and the output of the process is "displayed" on the "inner screen" of our consciousness. The "inner screen" is obviously a daft analogy, after all, someone (a homonculus) would need to be watching the screen...philosophical minefield altert!
In general, the data comes in from the retina and the brain asks "what kind of object in the external would generate retinal data like that?" and with a whole lot of very well-educated guesswork, we perceive a good representation of the object that caused the retinal data. In the case of this illusion, we get the data of the angle of the squares and conclude that the most likely object to cause that retinal data is a spiral pattern, without looking too closely into whether that's actually correct (we generally only look at a small bit of the visual field at any one time - the bit hitting the macula). This illusion gives us an interesting insight into the order of processing (pattern recognition) going on. We've decided there are spirals on the basis of the detailed (macular) data - the angle of the squares - and that's overridden the data that's come from the wider field (which tells us there are concentric circles). As has been pointed out, if you can the whole image to fall on the macula by viewing from a distance, the illusion disappears.
The 12 dots illusion is a similar macular versus peripheral vision job.
What I find amazing is that considering how little data actually comes in through the eyes and how our perception is mainly guessing what is out in the world on the basis of experience, how accurate it usually is. For example it's very rare for me to think that say, there's a bit of litter blowing from the pavement into the road in front of my car only find out later that it was toddler.