The visual system

Connected to the 100 million or so light-sensitive photoreceptor cells on the retina are roughly two layers of nerve cells, with various kinds of cross-connections, out of which come the million fibers that form the optic nerve. After essentially one stop, most of these go to the primary visual cortex at the back of the brain, which itself contains more than 100 million nerve cells. Physical connections between nerve cells have usually been difficult to map. But starting in the 1950s it became possible to record electrical activity in single cells, and from this the discovery was made that many cells respond to rather specific visual stimuli. In the retina, most common are center-surround cells, which respond when there is a higher level of light in the center of a roughly circular region and a lower level outside, or vice versa. In the first few layers of the visual cortex about half the cells respond to elongated versions of similar stimuli, while others seem sensitive to various forms of change or motion. In the fovea at the center of the retina, a single center-surround cell seems to get input from just a few nearby photoreceptors. In successive layers of the visual cortex cells seem to get input from progressively larger regions. There is a very direct mapping of positions on the retina to regions in the visual cortex. But within each region there are different cells responding to stimuli at different angles, as well as to stimuli from different eyes. Cells with particular kinds of responses are usually found to be arranged in labyrinthine patterns very much like those shown on page 427. And no doubt the processes which produce these patterns during the development of the organism can be idealized by simple 2D cellular automata. Quite what determines the pattern of illumination to which a given cell will respond is not yet clear, although there is some evidence that it is the result of adaptation associated with various kinds of test inputs. Since the late 1970s, it has been common to assume that the response of a cell can be modelled by derivatives of Gaussians such as those shown below, or perhaps by Gabor functions given by products of trigonometric functions and Gaussians. Experiments have determined responses to these and other specific stimuli, but inevitably no experiment can find all the stimuli to which a cell is sensitive.

The visual systems of a number of specific higher and lower organisms have now been studied, and despite a few differences (such as cross-connections being behind the photoreceptors on the retinas of octopuses and squids, but in front in most higher animals), the same general features are usually seen. In lower organisms, there tend to be fewer layers of cells, with individual cells more specialized to particular visual stimuli of relevance to the organism.