The Best Illusion of the Year Contest took place last week. As always, there were some cool, new illusions among the finalists—a feast for illusion-afficionados like myself. I particularly like the Honeycomb Illusion by Marco Bertamini and Nicola Bruno. Take a moment to watch (my rendition of) their illusion in its beautiful simplicity (important: for best effect, watch in full screen and HD quality):

This illusion is nothing but a honeycomb grid with little stars (or barbs) at each node. You can see this in close-up on the right.

The stars are clearly visible when you look directly at them. But, and this is where things get interesting, you don't see any stars in the parts of the grid that you don't directly look at. In other words, you get the impression that the little stars follow your eyes around, as you scan the grid with your eyes. I personally find this very compelling.

So what's going on here?

The key to this illusion is that, at any one moment, you only have a clear view of a very small part of the world: the part that falls onto the central part of your retina (the fovea). This part is about the size of your thumb at arm's length. Your peripheral vision, the things that you see from the corner of your eye, is much less sharp, and color blind. This, among other things, is why you move your eyes: You successively direct your central vision at things that you want to see, or that grab your attention. (See also this post.)

This partly explains the Honeycomb Illusion: The stars are simply too small to see with your peripheral vision, so you only see them when you look at them directly. But this is not all; the full explanation is, I believe, a bit more subtle. After all, the Honeycomb Illusion gives the impression that things appear and disappear as you move your eyes. And this is not the impression that you usually have when you move your eyes.

Consider what happens when you see a girl from the corner of your eye, and prepare an eye movement to look at her face. Even though you cannot make out any facial features with your peripheral vision, you know what a face looks like: there are two eyes, with one nose below, and one mouth below that. So when you make an eye movement toward a face, you, or rather your brain, makes a coarse prediction of what you should see. And this prediction is usually accurate.

But sometimes it isn't.

I'm kind of a day dreamer, and I don't pay much attention to my surroundings. So when I walk through the city with my girlfriend, I don't always keep track of where she is. I just assume that the nearest person is her. (I'm not in general a neglectful boyfriend though; or at least I don't think so.) And when I look at this person and discover that it's actually someone else, I get a bit of a shock. This shock is, you might say somewhat fancifully, the feeling of my brain's prediction being violated.

And this may be what happens in the Honeycomb Illusion as well. For some reason, your brain assumes that the the grid is a regular honeycomb—without stars. You can see that some nodes, those that you look at, have little stars on them. But your brain doesn't appear to conclude from this that therefore all nodes must have little stars on them. And when you shift your eyes to another node, and see (again, yet unexpectedly) little stars, you get a little shock; your brain didn't see that coming.

So I believe that this illusion tricks your brain into making an incorrect prediction about what the grid looks like. When you make eye movements, this prediction is violated over and over again, which subjectively gives the impression of stars appearing and disappearing.

Or maybe not. But it's a cool illusion in any case.