It's easy to think of the
eye as a dumb
camera that just sends
images to the
brain where all the
real work is done. As it happens, this is not the case. The
eye itself has built into it many clever tricks to make it more
intelligent. One of these tricks is lateral inhibition. If the eye were
electronic, this would be one of the things referred to by hackers as
a truly cool hack.
In nature, edges are important. It's important not to walk into trees, it's important to see the horizon, it's important to not walk off the edge of a cliff (unless you're a lemming of course). So the eye has special hardware for detecting edges. It works like this. At the back of the eye is the retina, which has a whole bunch of light receptors. These light receptors are connected to neurons that take the image to the brain. If they were connected straight through, that would pretty much be a dumb camera. But they're not. Each neuron is connected to the light receptor (let's call it a rod) directly in front of it, which gives a positive signal when light falls on it. But it's also connected to the ones next to it (to the left and right). The trick is that this connection (the lateral connection) is inhibitive; that is that it gives a negative signal. The positive signal from the receptor directly above and the negative signals from the lateral receptors are added together to send the signal to the brain. A diagram may help. Let's say we have a sharp edge from dark (D) to light (L) falling on the retina on receptors R1 to R7.
DDDDDDDDDDDDDDDDDDDDLLLLLLLLLLLLLLLLLLLLLLLLLL
----------------------------------------------- Retina
R1 R2 R3 R4 R5 R6 R7
N1 N2 N3 N4 N5 N6 N7
To explain the original idea, the value at N4
isn't the value just at R4, but it might be something like
R4-0.5*R3-0.5*R5
So what? Well, now say you have a sharp change from say, dark to light. Now consider N3. R3, the receptor above N3, is in dark, so we get the straightforward darkness effect. But, R4 is in light, so it's going to have a huge positive value (because the receptors activate when exposed to bright light). But it feeds negatively into N3, so N3 will be more negative and look darker than it actually is. Alternatively, consider N4. R4 (the positive connection) is in light, so it will be positive. But R3 is in darkness (i.e will make a negative signal) but there's an inhibitive effect, and a double negative is positive, so it will actually increase the measure at N4, making it look brighter than it actually is.
The net effect is this: The lateral inhibition makes the eyes detect the sharp change in brightness seem even sharper by making the contrast even stronger. It's like an edge-detector built straight into our eyes!
This effect is the basis of many optical illusions, for example, the Mach banding effect.