The Central Visual Pathway

The Lateral Geniculate Nucleus

The main visual circuit, known as the central visual pathway, make its first stop in the brain in the lateral geniculate nucleus (LGN). To get an understanding of what the LGN is, you need to know a bit about the thalamus. The thalamus is a mass of different cell bodies - gray matter - located quite deep in the center of the brain, and it is a ‘paired’ structure (each hemisphere has its own). Each thalamus looks like an oblong egg, and the two together straddle the top of the brainstem, which is the stalk that runs up from the base of the skull. The thalami are connected by a subtle strip of cells (the interthalamic adhesion), but it's not known to what extent, if at all, they communicate.

The thalamus performs a myriad of functions, but it is best known for being a ‘sieve’ of neural information. Every single sensory pathway (except for smell) going upward to the cortex from our sensory detection organs relays through it. To accomodate all these pathways, the thalamus is segmented into numerous sections. Each section is defined by the particular set of neuronal cell bodies - or the ‘nucleus’ - that acts as a relay station in the course of a given pathway. In the image below, only the nucleus we're concerned with - the LGN - is labelled; but every little section that is cordoned off from the others is itself a specialized nucleus.

The central visual pathway makes its first stop at the back portion of the thalamus, passing through the lateral geniculate nucleus. In latin, ‘genu’ means knee, and some guy poking around in some brains a while ago was convinced this side nucleus on the back end of the thalamic egg looked just like a bent knee. In terms of thalamic anatomy, it's actually one of the easier structures to identify because it projects outward slightly, forming a little ball-like bump.

When tested, the neurons of the LGN behaves in a similar way to RGCs. They also have receptive fields, which are also responsive to circular areas of light contrast falling onto the retina. The geniculate neurons aren’t merely replicas of RGC neurons, however. This is because they also receive descending input from higher regions of the brain (mostly from the V6 layer of the visual cortex). This input modulates their signals. There’s no comparable top-down interference in the ganglionic cells.

But let’s return to the start, and follow the path of RGC axons - the optic nerve - on their way to the LGN. The nerve exits the back of the eye through the optic canal, a little hole in the sphenoid bone which forms the back wall of the eye socket. Just beyond this wall, the two optic nerves (one from each eye) meet, and in doing so form the optic chiasm, a place where some of the RGC axons 'cross over'.

This crossing over is formally called a decussation, and it occurs in a specific way. The retina of each eye has a field of space which it can see. One half of this visual field, closer to the nose, is the nasal hemifield. The other, closer to the temple, is the temporal hemifield. Because light rays travel in straight lines and the pupil is of a limited width, the nasal neurons in a retina receive photons from the temporal hemifield of the visual field. Similarly, neurons making up the temporal half of the retina receive photons coming in from the nasal eye hemifield.

Only those neurons in the retina receiving signals from the temporal visual fields - the nasally positioned neurons - will decussate. Looking at the picture below, you should be able to see that this property causes the right visual field information to travel to the left half of the brain at the optic chiasm. On the opposite side, the left visual field is projected into the right hemisphere.

This property in humans - that like visual fields fuse at the optic chiasm and travel to the contralateral part of the brain - is not just an evolutionary quirk. It's seen in different vertebrates, too, and is thought to be a contributing factor which allows for depth perception. The technical term for depth perception is stereopsis. The principle feature that makes stereopsis possible is two eyes placed in slightly different positions on the head. The two eyes are close enough so that each can see the same image, but have it horizontally shifted relative to the other. Higher centers of the brain are able to receive and synthesize these dissimilar images into one main image, giving the animal an idea of the relative depth each object has in the visual scene during the process.

To make things clearer, the name ‘optic nerve’ is dropped when referring to the string of nerves traveling through the separate halves of the brain after they decussate at the optic chiasm. Each string is now an ‘optic tract’ instead, and it is the optic tract that arrives to synapse on the LGN. When an optic tract connects to its LGN, it doesn't just insert itself randomly. In fact, when you dissect and analyze this nucleus, 6 clearly defined receiving layers are apparent.

In Latin, ‘magno’ means large and ‘parvus’ small. The upper 4 layers have neurons with small cell bodies compared to the bottom 2 layers; they're called the parvocellular layers. The two lowest layers are the magnocellular layers. The koniocellular layers are sandwiched between the M and the P, forming the third channel coursing through the LGN. Konis means ‘dust’; the K cells are tiny. It’s part of the reason why they are only a recent discovery and still only poorly understood.

The ipsilateral eye with respect to one of the two LGNs will project information from its temporal retina into layers 2,3, and 5 of that LGN. Contralateral information, coming in from the nasal retina of the opposite eye, will flow into layers 1,4, and 6.

From previous sections, it's clear to us that the optic tracts consist of multiple, parallel but separate, streams of information. The central visual pathway combines some of these streams into three secondary streams, as is evinced by the three distinct layers of the LGN. Each LGN layer emerges because the RGCs that form it respond to a different mix of information, way back in the inner plexiform layer of the retina. So, which RGC types form which of the LGN layers?

The midget cell, or P cell, is the RGC type which connects to the parvocellular layers of the LGN, forming the P pathway. As I mentioned in a previous section, most RGCs - around 80% - are midget cells. They're located mostly in the foveal region of the retina and therefore transmit electrical information of a high definition. Because of this, midget RGCs are characterized by tiny receptive field centers and high spatial resolution, as well as sensitivity to color.

The parasol RGC, with its characteristically large dendritic tree that looks somewhat like an umbrella, connects to the magnocellular layers of the LGN, forming the M pathway. M cells have much larger receptive field centers than P cells, formed from inputs stemming from a larger number of photoreceptors in the retina. M cells, to give you a preliminary idea of their function, communicate contrast and the awareness of movement to the brain.

The K cells are also known as bistratified - their structure expands into two layers (strata) in the retina. They project into the interlaminar (koniocellular) areas of the LGN, forming the K pathway. Although it's been recorded that K cells respond mainly to short wavelength (blue) light, their specific function in visual perception is at present unclear.

When an RGC axon reaches its layer in the LGN, whether this is an M, P, or K layer, it branches extensively to feed many different target neuron cell bodies housed within that layer. This is an example of information divergence. The targets are not randomly chosen; the axon terminals of RGCs coming from the retina never overstep the borders of their layers inside the LGN. The specific contralateral and ipsilateral segregations made in the layered LGN, described above, are maintained when the LGN itself projects into the visual cortex. This pattern at the level of the LGN forms the basis for ocular dominance columns in the cortex - of which more will be said in the sections below.

The Visual Cortex

Many of the geniculate neurons leaving the LGN travel through the optic radiation, the majority of which terminates in the visual cortex found in the occipital lobe in the back of the brain. In the brain, the term 'radiation' is used to describe a collection of axons traveling together. It is a 'white matter' structure, because the fat enveloping the axons causes it to appear white in color on a brain chemically prepared for dissection. It can be broken down into two streams. The upper stream terminates in a part of the upper half of the visual cortex called the cuneus, found above the calcarine fissure. The lower stream has a special name - the Meyer's loop - and it terminates in part of the lower half of the occipital lobe, the lingual gyrus.

Like the LGN, the visual cortex is made up of multiple layers. The most important layer, for our analysis, is the primary visual cortex, abbreviated V1. In the images above, it's represented using dots. Visual areas 2,3,4, and 5 (V2, V3, V4, and V5) make up the extrastriate cortex and are called ‘higher order visual cortices’, since they’re involved in more complex aspects of vision. Looking from an external viewpoint onto one hemisphere of the brain, you can see that V2 and V3 seem to form concentric rings around V1 in the occipital lobe. The other visual areas are scattered around.

The primary visual cortex is also known as the striate (striped) cortex. This is because of the line of Gennari, a stripe you'd see running through and defining the boundaries of V1, if you were to cut out a transverse section of occipital cortex at the level of the calcarine fissure. This has been done in the occipital lobe pictured below. The line is made up of the myelinated (fat-encircled) axons coming into V1 from the LGN. Almost all the output of the LGN terminates in V1, and the line of Gennari is the chief termination site.

The two streams described above as forming the optic radiation were first mapped out to account for what's known as the retinotopic plan of the V1 area. When light rays imprint an image onto a retina, it's necessarily inverted. The 'retinotopic plan' of area V1 is a reference to the way that such an inverted image has its orientation preserved as information travels backward through the brain and ends up in the V1 cortical area. The upper stream of the optic radiation connecting to the cuneus, located superiorly, receives the bottom half of an image. The lingual gyrus, inferior to the calcarine fissure, receives the lower stream, which transfers the upper half of the image to it.

Some fascinating experiments were done when figuring out the existence and features of this retinotopic plan. In one experiment, macaque monkeys were injected by a radioactive biological label called C-2-deoxy-d-glucose. This labelled form of glucose is broken down for energy by active cells in the brain just as normal glucose is. If you kill - 'euthanize' is the academic term - some poor monkey a short while after subjecting it to some kind of stimulus, you can tell which areas of the brain were particularly active in the period between injection and death by using a procedure called autoradiography. This is a technique which employs X-ray film to make the distribution of 'hidden' radioactive glucose visible.

In the experiment I mentioned, experimenters would first lightly paralyze a macaque monkey, in order to prevent its eyes from moving around too much. They would then use electrophysiological recordings - micro-electrodes stuck into V1 cortical cells - to map out the area of one eye's visual field corresponding to the fovea of the retina. They did the same to identify the peripheral regions. This enabled them to construct an appropriate visual stimulus, like the one pictured below.

The visual stimulus was then projected onto a screen, to which one eye of the paralyzed monkey was fixed. The projection was flashed in counterphase (white going to black, in cycles of 3 Hz) intensely for a period of time. This made the cells of the central visual pathway take up labelled deoxy-glucose intensively. Once the brain was taken out of the monkey, the area they chose to examine - a laterally-positioned area of V1 cortex - could be unfolded and rolled out, rather like a rolled-up poster. You might think that this procedure seems likely to distort the cortex somewhat, but other experiments had proven that this distortion was quite small. The unfurled piece of cortex was then analyzed via autoradiography. This yielded images like this one:

It's a beautiful visualization of the retinotopic map. It highlights several key features, too. One is the variation of 'striate magnification factor'. This is a cumbersome phrase which only means that the cortex gives some regions of the retina a larger expanse of physical area, which is assumed to mean more importance. It's measured using 'mm of cortex/degree of visual angle'. As could be expected, the fovea is the most magnified region; it has the most cortical cellular machinery devoted to it. A conservative estimate of the area assigned to it in the macaque monkey is around 15 mm of cortex/degree of visual angle, which is a pretty sizable amount.

Another key feature of cortical organization determined in this set of experiments is the location of the foveal region. It was confirmed as placed along the V1/V2 border, at the very top of the V1 region.

The last interesting fact I'd like to mention about the V1 area of primates is that the inferior visual field is overrepresented, at least relative to the superior field. In fact, one entire extrastriate area, area V3, seems to be entirely devoted only to the lower half of the world we see. This has led to some speculation that seeing the ground below us has been more useful to our historical survival as a species, presumably in helping us detect snakes and other predators. I mention snakes in particular because there has been a lot of interesting research describing the importance of the pressures exerted on us by these reptiles in shaping the manner in which our brains have come to be wired.

The Cells of V1

As the discussion above implies, V1 has been the most studied area of the visual system. It was first broken down into 6 anatomical layers, which was a mistake, because there are more layers. To make up for it, layer 4, where most LGN input terminates, was further broken down into 4 sub-layers: 4A, 4B, 4Cɑ and 4Cβ.

The two LGN streams, coming into V1 from the thalamus of each hemisphere, eventually get mixed up in the visual cortex. This represents a shift from monocular to binocular information processing; but more on that later. For now, it is good to know that the receiving sublayer of V1, called sublayer 4Cɑ, receives mostly magnocellular input - input from layers 1 and 2 of the LGN. Sublayer 4Cβ receives parvocellular input from LGN layers 3-6.

Given this rough image of V1 layers, let’s continue on down to the cellular level. In the visual cortex, all the cells respond to visual stimulation. Amongst a myriad of different cell types, the two chief ones it contains are pyramidal and stellate cells. The axon of the pyramidal cell is long and can leave the cortex. The stellate cell, on the other hand, normally won't leave the cortex. The distinctive feature of all the neurons and glial cells making up the cortex is their orientation; they run upward, vertically through its thickness.

When they were first tested for response activity with micro-electrodes, in the same manner as RGCs and geniculate neurons, it was found that V1 neurons don’t respond to concentric rings of light anymore. This was a change which confused researchers for some time, until it was found out by accident that they respond instead to bars of light. Specifically, V1 neurons presiding over a spot of space in the visual field may respond to:

1) the orientation of a bar of light

2) the direction in which it is moving

3) both of these characteristics simultaneously

Depending on their response to light, cortical neurons were first broken up, by Hubel and Wiesel, into simple cells and complex cells.

Let's start with a discussion of the former. A simple cell is 'simple' because the pattern of responses evoked by light shone inside its receptive field can be fairly easily accounted for by excitatory and inhibitory regions working together in a defined spatial relationship. Here are its characteristics:

• illumination of part or all of an excitatory region in the receptive field increases the firing of the cell over the baseline rate of firing.

• illumination of part or all of an inhibitory region suppresses the firing rate. An electrical discharge is observed when this illumination is switched off.

• A large spot of light confined within an excitatory or inhibitory region evokes more response than a smaller spot within the same region, indicating summation of small signals.

• The inhibitory and excitatory regions are mutually antagonistic. This means simultaneous illumination of both regions by the same degree will tend not to change the baseline firing rate, as the two 'cancel' each other out.

Simple cells are not all alike. They come in different forms, with inhibitory and excitatory regions of differing strengths, positions and sizes. What allowed Hubel and Wiesel to group these cells was that their responses could be theoretically accounted for by simple arrangements of these antagonistic regions, something that's not so easy to do with complex cells. Here are some 'simple' receptive fields which they described in their 1961 paper:

In some simple cells, the direction of movement of the bar of light was also found to be an important factor in the extent of electrical response produced. A bar moved from left to right and then right to left might produce different signals. This discrepancy can normally be attributed to the flanking inhibitory regions, which may not always be symmetrical. Consider the image below:

FIGURE 2E from Hubel and Wiesel

Simple cells are also sensitive to the rate of movement across the receptive field. Some respond to slow movement, which is quantified as 1 degree/second or lower. Some simple cells prefer fast movement, of 10 degrees/second or more. A degree is a measure of space traversed in the retina. If you hold your index finger away from yourself at arm's length, and look at the area your fingernail takes up, you'll have a pretty good approximation of the space that one degree in your retina encompasses.

The most widely accepted manner by which the receptive field of a simple cell is thought to be formed is when the circular fields of inputting LGN neurons align, as in the picture below.

As its name betrays, a complex cell is far more complicated than the simple type that's described above. Complex cells respond to various shapes of light, both stationary and moving. What defines them as a class is that the light stimulus they react to can't be predicted by simple combinations of excitatory and inhibitory regions of space. Even when regions of space that do respond in a familiar excitatory or inhibitory manner can be located, the properties of summation and mutual antagonism that make simple cells easy to deal with don't hold anymore.

I'll give an example of a complex cell's receptive field in the picture below. This cell, described by Hubel and Wiesel, responds to a horizontal bar of a certain width. If this bar is anywhere in the upper part of the receptive field, it's inhibitory. The exact position it occupies in this upper half doesn't seem to matter to it. In the precise center of the field, the cell gives both an excitatory and an inhibitory signal. However, once the bar is moved into the center or the lower half, the reaction is only excitatory.

To make things even more peculiar, widening the bar of light in the upper or lower halves, so that a given half is covered to a greater extent, produces no signal at all, rather than the increased signal you might expect after studying the summation principle exhibited by simple cells. This behavior squarely makes away with this principle.

This one example doesn't begin to cover the individuality shown by different complex cell receptive fields. Compared to simple cells, all complex cells are very sensitive to orientation. The exact position within the receptive field of the light bar is not as crucial. On average, they have larger receptive fields. Some complex cells are very sensitive to movement, in a given direction, at a given rate - but across the entire receptive field, and not just a narrow portion of it. Some react to the most effective stimulus with both excitatory- and inhibitory-type responses.

Complex cells receive some input directly from the LGN; but they also get input from many simple cells. The manner in which simple cells unite their receptive fields to make up the wide field of a complex cell is not yet understood.

Cellular Organization in V1

We've covered the types of cell present in the primary visual cortex, and a few examples of the receptive fields they exhibit, in the subsections above. These cortical cells can be driven by one eye, or by both. When they're considered as a whole, formative the V1 area, several different organization principles emerge.

The first such manner of organization is that of the ocular dominance columns. This phrase describes columns of cells found in layer 4 of V1 which receive input preferentially from one eye. In fact, alternative columns, one responding to the left and one to the right eye, span the striate cortex, creating a strikingly crisp striped pattern.

This pattern is produced by incoming LGN neurons. When a primate is born, these LGN neurons terminate in layer 4 of the striate cortex with their dendrites unfurled and overlapping. As the organism matures and passes its critical period of development, the dendrites are pruned so that those emerging from one eye each govern over a clearly defined column of cortical area.

The second way in which the cells of V1 are organized is by orientation columns. These are made up of complex cells, and run straight downward through the cortical surface, perpendicular to the ocular dominance columns. If you dug down into the cortex, testing cell after cell, you'd find that each complex cell in an orientation column responded to the same kind of edge. These edge-specific columns form a pinwheel pattern, if you look down upon the flattened cortex.

The third and final organizational feature of V1 is the cytochrome oxidase blob. These regions of cells contain high levels of the protein cytochrome oxidase within them. This is a respiratory enzyme that's widely employed throughout different biological systems, since it conserves energy produced by the reduction of oxygen. The blobs take their name from the enzyme because it was first used as a chemical stain to discover their existence, but they're actually cylindrical groups of neurons sensitive to particular colors.