Nobel prize winning studies by Hubel and Wiesel in the 1960s showed that the primary visual cortex is organized in relation to features in the visual field of the retinas. Neurons that respond to vertical lines -- trees or telephone poles in the visual field, for instance -- are grouped together in patches called orientation columns. Other orientation columns respond to horizontal lines, 45¡ lines, etc. Other neuron groupings form according to whether they respond predominantly to the right or left eye.
Orientation columns in a patch of monkey visual cortex. Colors from red to violet indicate the orientation preference of the neurons, from 0 to 180 degrees. Overlaid white lines show the orientation preference of each area. (From research by Blasdel, 1992.)
Since the mid-70s, experiments have demonstrated that these patterns aren't genetically fixed, but develop from visual experience. "An organism," explains Miikkulainen, "seems to store a set of learning rules, genetically programmed procedures expressed in the behavior of each neuron, so that information from the outside world organizes the connections from neuron to neuron. Because the organizing mechanism is unsupervised and local to each neuron, it is referred to as self-organization."
Experiments have shown that connections among neurons are initially more or less randomly scattered. With visual experience, the neurons appear to divide their labor, in effect saying, "You take this kind of input. I'll take that kind of input." "Neurons that reinforce each other," says Miikkulainen, "wire together, and unnecessary connections are eliminated. Such filtering eliminates redundancies and allows the brain to efficiently process the massive amounts of visual information impinging on the retinas. This trimming down process is what we're modeling."
go back to the main page