The first proper step to vision - i.e., creating that picture in your mind - lies in the retinal ganglion cells.

The retinal ganglion cells' importance lies in the coding of visual information. The retina’s photoreceptor cells are the first stage information coding as they begin the signalling of visual information by altering their firing rate depending on the intensity, and wavelength, of light. Each retinal ganglion cell receives input from multiple photoreceptor cells and in doing so can mediate more complex coding than the simple responses of photoreceptors.

Before discussing the role of ganglion cells in visual perception, it would be helpful to discuss their physiological nature. Ganglion cells exit the retina at the optic disc (the "blindspot") and form the optic nerve, which leads to other higher visual processing areas. Ganglion cells transmit coded information in the form of action potentials1. However, they do not just transmit in response to stimulation – ganglion cells have a base rate of firing known as “maintained discharge”. There are three opposing views as to the nature of this maintained discharge. Firstly, there is the view suggested by Rodieck (1967), and supported by Barlow and Levick (1969), that it is caused by the random and spontaneous firing of receptor cells which can happen even in dark conditions. However, Hughes and Maffei (1965) had previously suggested that the maintained discharge was something that originates in the ganglion cell itself rather than its inputs. There is, naturally, a third view that the maintained discharge arises somewhere between the receptors and ganglion cells.

The arrangement and type of receptor inputs to the ganglion cells allows the coding of different types of visual information in the receptive field2 of a ganglion cell. The basic arrangement of ganglion receptive fields are centre/surround, with a lateral antagonism mechanism. In other words, they have a central area and a surrounding area which serve to inhibit each other's responses. There are two types of these cells – ON-centre cells, and OFF-centre cells. ON-cells have a central spot of excitatory input with a surround of inhibitory input. The OFF-cells are the opposite way around. The ON- and OFF-cells are an example of parallel processing in the visual system. The ON-cells can tell us how light a stimulus is, while OFF-cells can say how dark something is. These two systems are roughly mirrors of each other. There is a third basic type of ganglion cell – an ON-OFF cell, which can signal any change in light. It increases its response with both the onset and offset of a stimulus. ON-cells only increase response with the onset of a stimulus, and decrease with its offset. Again, OFF-cells are the other way around.

The function of the centre/surround organisation of retinal ganglion cells is to allow selective sensitivity to light contrasts. The lateral antagonistic nature of the centre/surround system means that ganglion cells fire selectively in response to contrast rather than total illumination (as receptor cells do). Ganglion cells have the property where they will always code for the same relative contrast despite the general light conditions due to the effects of dark/light adaptation. These ganglion cells are particularly useful for detecting boundaries in our visual field.

There are several illusions that demonstrate the existence of lateral antagonistic receptive fields, and also can show how they detect boundaries. Mach bands, for instance, are an effect of having receptive fields with lateral inhibition, as too are the spurrious dark spots in the seen in Hermann grid illusion. Other illusions demonstrating the effects of these ganglion cells are luminance steps and the Craik-O’Brien illusion. These illusions show how boundaries are made explicit by the lateral antagonism system.

The function of edge detection is a useful one as it reduces the amount of redundant information coded. An organism will use the boundaries of objects to make basic definitions of the visual scene. Minor changes in light intensity over certain areas are of little significant consequence to the organism’s function.

But not all ganglion cells function to detect edges. There are some more specialised types in organisms, although the majority in animals with chambered eyes are of the lateral antagonistic centre/surround type. The more specialised code more specific information about the visual scene.

An interesting case study which provides some examples of specialised ganglion cells is that of the visual system of the frog, studied by Lettvin (1959). Frogs have a fairly unique visual system due to the nature of its ganglion cells. It has four different specialised types. Firstly, they have “sustained contrast detectors” which have no response to diffuse light across the receptive field, as with normal ganglion receptive fields, but respond greatly when a lighted edge passes through it. These cells are sometimes direction dependant. Frogs also have “net convexity detectors” which serve to respond whenever there is a dot in its receptive field, darker than the background. This is an efficient mechanism for the frog as it allows it to detect potential food sources (such as flies) with very little analysis of its visual field. Frogs also have cells similar to the ON-OFF cells mentioned above, and “net dimming detectors” which serve to give a prolonged response to the reduction of a diffuse light stimulus. This could be used be used to detect potential predators.

The ganglion cells of the frog are a good example of how ganglion cells are adapted to suit the visual requirements of the organism in question. Frogs do not receive anything at all resembling the kind of ganglionic input we humans have, or indeed other primates and higher mammals. But then, higher mammals have different needs for their visual systems.

There are some specialised ganglion cells in higher mammals. A common type are direction sensitive cells (Barlow & Hill, 1963). These cells only respond primarily to a light stimulus moving in a certain direction. This is achieved by a fairly complex system of delayed inhibition in a certain direction (the “null” direction), thus causing the excitation of the light stimulus reaching a receptor to be cancelled out as it arrives. Light moving in the other direction (the “preferred” direction) always precedes this inhibition. Such cells have not been confirmed to be in primates (including humans), but there are other orientationally selective cells which we know we do have. Such cells are selectively sensitive to the orientation of a bar of light.

The ganglion cells that are found in humans and other primates and higher mammals seem to be geared towards providing information to the visual system that is used to construct a picture of the visual world. Ganglion cells seem to act as “feature detectors”, providing “building blocks” from which the visual system can build up a picture of the world and the objects within it. The ganglion cells responsive to form features give us information to construct the forms of objects, while the cells responsive to movements, obviously tell us how these objects are moving.

There are two important visual streams of information in the visual system. There is the magnocellular system and the parvocellular system. The main input to the parvocellular system are “X-cells” which have small receptive fields concentrated in the centre of the retina at the fovea. They produce fairly sustained responses. The main input to the magnocellular stream comes from the ganglion cells categorised as “Y-cells” which have larger receptive fields and are evenly spread across the whole retina. Their responses are more transient than those of X-cells and the signals are transmitter more quickly due to the thicker axons of Y-cells.

It seems that Y-cells are useful for detecting and orienting to new, potentially important stimuli that could appear anywhere in the visual field. It seems reasonable that X-cells would be useful for providing more detailed information about stimuli after they have been fixated. This corresponds to the organisation of parallel processing of visual information in the brain. It has been found that the parvocellular stream leads mainly to the ventral visual processing stream where form is processed. This would be compatible with the small receptive fields of the X-cells and their ability to detect finer features of objects. So too, has it been found that the magnocellular system leads mainly to the dorsal visual processing stream, which is responsible for detecting motion. This type of processing is suited to input from cells with large receptive fields, which are of course very sensitive to movement.

So, it would seem that the signals from ganglion cells are relayed to visual processing areas to perform specific processing tasks. This could be cited as evidence in support of a direct visual perception system since information is relayed into areas that perform specific processes, that a stimulus elicits an automatic perceptual sensation. The importance of optic flow detection3 could be mediated by the Y-cell’s movement sensitivity. The third grouping of ganglion cells – W-cells – could support a direct view of visual perception. These cells project directly to the brainstem rather then the lateral geniculate nucleus of the thalamus. It would appear that these cells have some kind of automatic response function.

However, the indirect view of visual perception is not ruled out. After all, the ganglion cells signal information about individual features of objects which the perceptual system must somehow construct into something meaningful. In other words, the visual information is not mere impulses, but it is built into something to which we ascribe some meaning relevant to ourselves.

Still, with either view, it is the exact nature of the ganglion outputs that determines what we perceive. The ganglion cells have a direct bearing on what is perceived as they provide the first stage of coding that goes beyond simple intensity responses, and go some way to coding the light incident on the retina in a form useful to us.

1 - Action potentials are sharp spikes of electrical impulse singalling the stimulation of a cell.

2 - The receptive field is the area over which the cell recieves input stimuli.

3 - My write-up "2D retina, 3D perception" explains how optic flow is important.