4 June 2007

The eyes have it

Some people believe that they perceive the world as it actually is. There are many experiments that can disprove this notion. For instance, take a look at Edward H Adelson’s checker shadow illusion. To me, the two squares A and B look so obviously different that if I didn’t know it was an optical illusion, I would never pause to think that they might be exactly the same color.

Similar experiments can demonstrate that your hearing is just as subjective. There are tone mixes that can be played that some people will hear as ascending tones, some as descending tones. The sense of touch can be fooled too. I’ve not heard of any demonstrations of the subjectiveness of smell-based perception, but I don’t doubt that it could be done.

Nevertheless, we like to think that we see things as they are. Sure, maybe the colors aren’t always right, or the angles look distorted, but the basic details are correct, or so we assume. You were probably taught, like me, that the eye works like a camera—the scene before it is focused by the lens onto the retina, and the signals from the rods and cones are transmitted to the brain where they are processed into an image, right?

Wrong, it turns out.

In April’s Scientific American magazine there was a fascinating article that described how the eye actually works. Further details from the same researchers were published in Nature.

It turns out that the rod and cone cells connect to 10 different kinds of neurons known as bipolar cells. The bipolar cells have long axons which extend into one of 10-12 different layers of what’s known as the plexiform layer. Also connecting into those layers are 12 different types of ganglion cells, which are the cells that actually transmit to the optic nerve. There are also at least 27 types of amacrine cells, which can affect signal transmission between the layers and change propagation of signals within a layer.

Basically, the eye isn’t at all like a camera. It’s more like a chunk of brain tissue that has been wrapped around the inside of the eyeball. (Stranger still, all the sensor cells are on the outside, and all the wiring is on the inside, so light has to pass through everything else to get to the sensors. Intelligent design? I don’t think so.)

Anyway, the sensory data from the rod and cone cells is processed by this retinal tissue into 12 separate streams of information, all of which are sent to the brain in parallel.

One signal stream consists of only the edges detected in the scene; another is only the moving edges. One has just the shadows. One detects and emphasizes highlights. One seems to notice changes of brightness with respect to time. One detects large uniform areas. Another seems to detect backgrounds around central figures. And so on.

The 12 streams of data are sent to different parts of the brain. The brain then somehow uses all of these special purpose signals to work out a mental model of the external world, which it uses to “color in” the photo-like perception we imagine we have.

This is, of course, why optical illusions work. Given the nature of the signals sent to the brain, the bigger mystery is how the brain mostly does such a good job of fooling us into thinking we have cameras for eyes.

© mathew 2017