My recent research has focused on the limits of visual memory. You can read a full account of this research in the journal Science: "Dynamic shifts of limited working memory resources in human vision" by Paul M Bays and Masud Husain. For non-specialists, the main conclusions are summarized below.
Our ability to remember what we see, even for a second, is much poorer than you might imagine. As a demonstration, the picture below changes every couple of seconds — see if you can spot the difference...
If you really can't find it, click here to see the answer.
It often takes a surprisingly long time to spot a change like this (a phenomenon known as 'change blindness' — you can see more examples on Ronald Rensink's site). This difficulty arises because visual working memory, the memory that we hold for a few seconds after looking at a scene, is remarkably limited. In fact, for many years the dominant theory has been that visual memory can hold at most four items at any one time.
However, my recent experiments have revealed a different story. Visual memory, while indeed limited in capacity, is considerably more flexible than the four-item theory would suggest. Instead of storing a fixed number of items, visual working memory consists of a limited resource that must be shared out across an image. The more of this resource is allocated to an object, the more precisely that object will be remembered.
So why was it so difficult to spot the change in the photo above? It turns out that your brain allocates the majority of memory resources to the most important features of a scene: in this case, these would probably be the faces of the runners, and also the baton — a key object in the image. Visually arresting details — such as one of the runners' pink hair — will also be allocated more memory, and so be remembered in more detail. This leaves little memory for less important details, such as the athletes' footwear, which will at best be remembered only very approximately. This principle is illustrated below.
Surprisingly, you don't actually have to look directly at an object to store it in memory. Once you know which part of the photo is changing, that region becomes significant, and it should now be impossible not to see the change — even if you're looking elsewhere in the image. However, where you look does have an important influence on what you remember: my research has shown that making an eye movement towards an object causes working memory resources to be redistributed — with a greater proportion allocated to the target of the eye movement, and so less available to store other features of the scene. In fact, at the time you move your eyes, you hold a much more precise memory for the object you're about to look at than you do for the object you've just been looking at.
To conclude, my research supports the idea that our experience of seeing is an illusion — in that what we actually know about what's in front of our eyes is far less than it seems. That we remain unaware of this fact is most likely due to the flexibility with which the brain makes use of its very limited resources: the information we consider important is always available to us, so we don't notice how much else is not.