Chapter 4. The Neural Retina Basic Organization At one level the retina works like film in a camera: it records the optical image that is focused onto it by the cornea and lens, and relays this information to the rest of the brain. But the retina is itself officially a part of the brain it s classified as central, not peripheral, nervous system and within its thin layers it contains some remarkably sophisticated processing systems. The retina packages and formats the raw signal from 100 million photoreceptors and sends it off on 1 million ganglion cell axons, and does so with the least possible loss of important information. The retina also needs to adjust to lighting levels, emphasize areas of contrast over areas of uniform brightness, and translate between the analog signals of the photoreceptors and the pulse-code of the ganglion cells which is the only aspect of retinal activity (and hence, the only information about the visual world) that the rest of the brain can know. In one sense the retina may be regarded as perfect, for it is pretty much the only part of the central nervous system that does not get feedback connections from some other part of the brain. The retina takes image data, processes it, sends it off to the brain, and the only way that the rest of the brain can affect the retina is by moving the eyes, changing lens focus or otherwise altering the physical image. The primate/human brain has no significant direct neural input to the retina. The retina knows what it s doing. Figure 4-1 shows the Figure 4-1. Layers of the Neural Retina. basic layers of the neural retina in cross-section. Note that, somewhat counter-intuitively, light has to go through all the neurons of the retina before it gets to the photoreceptors: the neurons are mostly transparent, so this is not a problem, and at the fovea the neural elements are displaced to the side for maximal acuity. Also, with respect to the eyeball, outer always refers to the outside of the eyeball, and inner always refers to the inside of the eyeball. Light from the cornea and lens traverses the neural retina, and is absorbed by the photoreceptors, which causes the photoreceptors to hyperpolarize. This signal is relayed via chemical synapses to the bipolar cells, which then transfer the signal to the bipolar cells, which in turn transfer the signal to the ganglion cells. The ganglion cells provide the output signal from the eye to the rest of the brain, and their axons travel in the nerve fiber layer until they eventually form the optic nerve that exits the eyeball.
This pathway, from photoreceptor to bipolar cell to ganglion cell, is sometimes called the vertical organization of the retina, because if you lay the retina flat on a table the signal flow is in the vertical direction. The retina also has a horizontal organization whereby adjacent vertical paths can interact. The aptly named horizontal cells provide this connection in complex synaptic structures involving photoreceptors and bipolar cells, and amacrine cells via complex synaptic structures involving bipolar cells and ganglion cells. It is the horizontal structure of the retina that causes the output of the ganglion cells to be more than just a simple image transfer from the photoreceptors to the brain. Figure 4-2 illustrates the basic stages of signal transfer in the vertical pathway of the retina. Every photoreceptor, rod or cone, hyperpolarizes in response to light. This is the last part of the visual system where this is the case, for in order to give equal sensitivity to luminance increments and decrements, the connection from cones to bipolar cells is parceled out into an ON pathway, that depolarizes in response to light, and an OFF pathway, that hyperpolarizes in response to light (and depolarizes in response to darkness). Photoreceptors all use glutamate as their neurotransmitter, but by using different classes of glutamate receptor the synapses can be Figure 4-2. Information transfer from photoreceptors to ganglion cells. either sign-conserving or sign-inverting. The sign-conserving connection to the off bipolar cell means that this cell will follow the response of a cone, and hyperpolarize in response to light. Another sign-conserving connection to the OFF ganglion cell means that this cell also hyperpolarizes in response to light. The sign-inverting glutamate receptor that the ON cone bipolar cell uses to connect to a photoreceptor means that this cell depolarizes in response to light, a property that is passed to the ON ganglion cell via a more standard sign-conserving synapse. The rods, which are mostly only active in dim light, have only one type of bipolar cell, the ON rod bipolar. The rod signal is piggybacked onto the cone pathway via the aii ( a-two ) class of amacrine cells. By using a sign-conserving electrical synapse onto the ON pathway, and a sign-inverting chemical synapse onto the OFF pathway, the signals coming from both ON and OFF ganglion cells can seamlessly carry the same information to the brain regardless of the lighting level, sparing the necessity of a separate expensive parallel pathway to the brain.
Note that most of the signaling within the retina uses graded potentials, i.e., a cell s membrane potential gradually varies between the depolarized and hyperpolarized states. More depolarized means a higher continuous rate of neurotransmitter release, and more hyperpolarized means a lower continuous rate of neurotransmitter release. This is possible because the distances within the retina are short enough that graded potentials can travel the length of (for example) a bipolar cell without much loss of electrical signal strength. This is not, however, the case for ganglion cells, whose axons must travel many centimeters from eyeball to brain, a distance far too great for graded potentials to traverse. Thus, ganglion cells convert their degree of depolarization into a rate at which action potentials are generated (see Figure 4-3. Time course of response of photoreceptors, bipolar cells, and ganglion cells to a short pulse of light. AP: Action Potentials. Figure 4-3). More depolarized means more action potentials/second, and more hyperpolarized means fewer action potentials/second. The action potentials are selfrenewing chain reactions, like a string of firecrackers, and can thus make it all the way to the brain without attenuation. The connection that the cones make to bipolar cells and horizontal cells has been called the single most complex synapse in the entire brain. When you consider that the photoreceptor is the only cell type in the visual system that itself responds to light, it would seem to be a priority for the visual system to squeeze as much information from this source as possible. Figure 4-4 shows a cartoon of the connections from photoreceptors to bipolar cells. Rods connect to bipolar cells at spherules, and cones connect to bipolar cells Figure 4-4. Basic Synaptic Contacts of Rods and Cones.
through pedicels. Both the rod and cone ON cells (that depolarize, or become more activated by light and then release more neurotransmitter at their own synapses with ganglion cells) use sign-inverting synapses at specialized synaptic structures called triads. Here the processes of a single invaginating bipolar cell come together with the processes of two horizontal cells. The ribbon synapse is a specialized structure seen in a few other places in the nervous system, such as the auditory system, and seems to be specialized for high speed and continuous release of neurotransmitter. There are no rod OFF bipolars, and the cone OFF bipolars use the standard signconserving glutaminergic synapse. The OFF bipolar cells are not directly associated with triads, but rather Figure 4-5. The basic flavors of retinal bipolar cells. There is only one type of rod bipolar, it is always of the ON type, and connects to many rods. There are many types of cone bipolar cells. The midget bipolars are found in the central primate retina, they are of both ON and OFF types, and connect a single cone to a single midget ganglion cell (not shown). The diffuse bipolar cells (there are several types) are also of ON and OFF types, but have a more diffuse dendritic arborization that gets inputs from multiple cones. OPL: Outer Plexiform Layer. INL: Inner Nuclear Layer IPL: Inner Plexiform Layer GC: Ganglion Cell layer. make flat connections with cone pedicels. As usual these receptors bind glutamate, and then directly open ion channels in the post-synaptic membrane to cause depolarization. Exactly what is going on at the invaginating synapse is still not clear. Glutamate is the neurotransmitter, and it acts through a metabotropic receptor on the bipolar cell, which uses a second-messenger system to link separate receptor and ion channels, to reverse the normally excitatory effect of glutamate. The processes of horizontal cells are more problematic: as of this writing it s not clear how they work. Both chemical and electrical synapses have been implicated. The bottom line is that horizontal cells sum the amount of light over a broad region, and subtract it from the local response of a single cone. As we shall see later in this chapter, this emphasizes the responses to regions of an image where light intensity varies and de-emphasizes areas of uniform illumination. The use of + and - can be confusing. Here + means sign-conserving, and - means sign-inverting. However, in other contexts + and - can refer to whether the cell is excited or inhibited by light in some area in its receptive field (such as the center), which is not the same thing. This is one of those situations where you have to figure it out from context: we shall endeavor here to indicate what + and - mean in each figure, but other texts are not always so clear. As shown in Figure 4-5, each cone can contact several ON and OFF bipolar cells through its pedicel. Additionally, each diffuse bipolar cell typically gets input from many
cones. As you can imagine, the outer plexiform layer is a complicated place with lots of connections. There are at least four different types each of ON and OFF diffuse bipolar cells, it s not clear what the purpose of this is but they seem to have different response speeds. The ON bipolars make synapses with ganglion cells in the inner part of the inner plexiform layer, and the OFF bipolars make synaptic connections in the outer part of the inner plexiform layer. At the primate fovea detailed form perception requires that each bipolar cell get input from just one cone. These specialized bipolar cells make direct contact to just one ganglion cell, forming the midget pathway of small cells that can carry extremely detailed information to the brain. In the rest of the retina there are far too many photoreceptors for each one to get its own private line to the brain, as there are about 100 million photoreceptors but only 1 million ganglion cell axons per human eye. There is only one type of rod bipolar cell, of the ON type. As the rods operate under low light conditions where fine detail vision is not possible, the rods have no midget pathway. As shown in Figure 4-2, the rod bipolars piggyback onto the cone ON and OFF pathways via the AII amacrine cells. As we will discuss shortly, most cones in the primate fovea are of the red and green types: blue types are comparatively rare. There is still a lot that is not clear about how the blue cone signal gets through the retina, but it seems as if the blue cones are not typically part of the midget pathway, and the diffuse bipolars have been suggested to be somewhat similar to rod bipolars, in that they are only of the ON type. Regardless, as of this writing it seems as if there is something different about the way that the primate retina handles blue cone signals. Horizontal Cells As mentioned before, horizontal cells make connections between where the photoreceptors and the bipolar cells join (see again Figure 4-1). Looked at from the top (assuming the retina is spread flat on a table) horizontal cells have two star-like sections separated by a long thin axon. One section consists of the cell body and its dendrites, the other is the terminal arborizations of the axon. Because horizontal cells do not Figure 4-6. Schematic horizontal cell. There is a zone of local electrical processing clustered around the cell body (left), another around the terminal axonal arbors (right), connected by a long thin axon that, because it does not conduct action potentials, causes these two processing zones to be functionally isolated: the axon is here purely for the metabolic support of the axonal terminals. generate action potentials, the axon plays no role in transmitting information (passive electrical activity decays rapidly in long thin axons). Thus, each horizontal cell is essentially two distinct local processing zones. There are currently thought to be two different types of horizontal cells, named H1 and H2, the different ends of these classes
preferentially contact different proportions of red, green, and blue cones, and rods, but they are otherwise similar. As mentioned before, functionally what the horizontal cells do is to subtract out the average local activity from the response of each individual photoreceptor. In other words, the effect of the horizontal cells is to change the signal at the photoreceptor, which is only about the absolute amount of light at a particular location on the retina, into a bipolar cell signal about the difference between the signal carried by the photoreceptors directly above the bipolar cell and the local average. This serves to accentuate changes in the image and downplay regions of the image that have uniform brightness. This is an example of lateral inhibition, a process common to essentially all sensory systems where activity in an input tends to suppress the activity in neighboring ( lateral ) inputs channels, thus emphasizing differences or changes in sensory input. We can also say that horizontal cells help to create the center-surround receptive field organization of the bipolar cells. To understand what this means we need to carefully define what is meant by a receptive field. The Receptive Field The concept of the Receptive Field (RF) is a key one in visual neuroscience. Consider a single photoreceptor. It responds to the amount of light falling upon it. Because of the focusing of light by the lens, this means that the photoreceptor will respond to the amount of light in a specific area of the visual scene. This area will be roughly circular, and its extent will be determined by the size of the photoreceptor and the optics of the eye (see Figure 4-7). The edges of this area will not be perfectly sharp because of optical aberrations, but nevertheless there will be a relatively well-defined region of a visual scene where changing the amount of light affects the response of the photoreceptor, and outside of this region changing the amount of light has no effect. This region of a scene where you can affect the activity of a neuron in any way is called the classical receptive field (CRF), though this is usually just referred to Figure 4-7. The receptive field (RF) is that part of a visual scene where changing the image can affect a neuron s response. The receptive field of a photoreceptor is just a single small circular region. as the receptive field by default. It often happens that changing an image outside the classical receptive field by itself has no effect, but can modulate the response of the neuron to a visual stimulus in the classical receptive field. The region outside of the classical receptive field where changing an image can modulate the response to the image inside the classical receptive field is called the non-classical receptive field.
The receptive field of a photoreceptor is locked to the retina, so if the eye moves one degree to the left, the position of that specific photoreceptor s receptive field also moves one degree to the left. Much of the visual system consists of neurons with this property, but as we shall see in later chapters there are indeed neurons with receptive fields that are not strictly tied to a retinal frame-of-reference. As signals progress away from the photoreceptors, the visual system increasingly processes the information so that the receptive fields are increasingly complex and abstract. The receptive fields of retinal bipolar and ganglion cells have a center-surround organization (see Figure 4-8). There is a center region, which is excited by light in the case of ON cells ( + ), and excited by dark in the case of OFF cells ( - ). However, there is also a surrounding donutshaped region that has antagonistic effects. Hence, these cells tend not to respond very strongly to uniform illumination, as the + and - regions will tend to cancel each other out. For bipolar cells, activation means depolarizing, and suppression means hyperpolarizing. For ganglion cells, activation means firing more spikes/second, and suppression means firing fewer spikes/second. The different regions of the receptive field are usually schematized as being distinct (see Figure 4-8 left), but in reality the borders between excitatory and inhibitory regions are not sharp but rather vary smoothly. These receptive fields are often modeled as a difference of Figure 4-8. Receptive fields of retinal bipolar and ganglion cells have a center-surround organization. Gray rectangle on bottom: calculating the sensitivity to light and dark along a narrow strip passing through the center of the receptive field. Figure 4-9. Response of a center-surround receptive field near a luminance-defined edge. Point C is the Null Position.
Gaussians (see Figure 4-8 right). If you have a narrow Gaussian for the center, and subtract away a broader but shallower Gaussian for the surround, you get a Mexican hat shaped function. The center needs to have a higher peak if it is to balance the surround because it covers less area. Consider the implications of this arrangement for image processing. Figure 4-9 shows how an on-center off-surround cell would respond around an edge defined by a difference in luminance. In uniform regions of light (A) or dark (B) the center and surround mostly cancel out, and activation or suppression is weak. The balance between center and surround is usually imperfect, with the center having a smaller net greater effect than the whole of the surround. At point B excitation is maximal, because all of the + region is in light, but only part of the region is in light: the part of the - region in dark acts like a + region in light. At point D suppression is maximal, because all of the + region is in dark and some of the region is in light. Point C is sometimes called the null position, it is the perfect point of balance where everything cancels out. Cells whose receptive fields have a null-point are sometimes referred to as being linear (or quasilinear), meaning that the effects of + and regions are opposite and equal and sum linearly. Not all retinal ganglion cells will have a null-position. The visual system has to analyze the entire image at once. We say that the visual field is tiled with receptive fields. Neurons of any specific class must tile the entire visual field if they are to successfully analyze the image. Obviously, all of the photoreceptors on the retina provide coverage of the entire visible field of view. Bipolar and ganglion cells of both polarities also tile the entire visual field. Figure 4-10 illustrates the concept. The left panel is an unprocessed image. The right panel is an activity map of the image after it has been processed with on-center off-surround receptive fields centered at each point. The nervous system does not actually make a new image: this is a map where white indicates more activation and dark indicates less activation. Note how regions of high contrast (i.e., changes between black and white) are emphasized and regions of uniform luminance are de-emphasized. The center-surround receptive field organization exists to emphasize regions of an image that contain interesting bits (lines, edges, high-contrast patterns) and de-emphasize boring bits (regions of constant luminance). Amacrine Cells Figure 4-10. Left: Unprocessed image. Left: Activity map of the image after being processed by center-surround receptive fields located at all regions in the image.
Amacrine cells connect between the junctions of bipolar and ganglion cells in the inner plexiform layer, somewhat like the horizontal cells connect between the junctions of the photoreceptors and bipolar cells in the outer plexiform layer. However, while the horizontal cells are mostly concerned with creating the centersurround receptive field profile, amacrine cells are an incredibly diverse lot. Currently we don t know what most of them are doing, though it seems likely that it s pretty important or they wouldn t be there. We have already covered the AII Figure 4-11. The dyad junction between bipolar cells, and processes of amacrine and ganglion cells, in the inner plexiform layer. amacrine cell, which piggybacks the rods onto the cone pathway. There are dozens of other classes of amacrine cells in the mammalian/primate retina. One common characteristic of every class of amacrine cell, however, is that the receptive fields of each separate class tiles the retina. This makes perfect sense: whatever specific function each amacrine cell performs, it must perform it across the entire image for it to be useful in analyzing the image. Amacrine cells are unlike all other interneurons in the retina in that some of them sometimes generate action potentials as well as graded potentials. We don t know why this is so. They use a variety of neurotransmitters, however, they appear to all be inhibitory, using neurotransmitters like GABA or glycine (the AII amacrines make an additional excitatory connection using a sign-conserving electrical synapse). Amacrine cells typically make a characteristic dyad junction in the inner plexiform layer. The bipolar cell has a synaptic ribbon, which as with the photoreceptor is the characteristic synapse of neurons that signal through constant release of neurotransmitter (rather than the brief pulses of neurotransmitter with each action potential characteristic of most chemical synapses). The bipolar cell makes contact with both ganglion cells and amacrine cells. There can be reciprocal synapses between an amacrine cell and a bipolar cell, where the bipolar cell talks to the amacrine cell process and the same amacrine cell process talks back to the bipolar cell. Once again, it s not known what this does, but is presumably Figure 4-12. Some ganglion cells exhibit motion selectivity, i.e., they respond more strongly (i.e., used in some sort of regulatory fire more spikes) to motion in one direction across (feedback) mechanism. their receptive fields (the preferred direction) than One the things that amacrine cells to motion in the other (null) direction. are implicated in is motion selectivity. Bear in mind that nearly every neuron in the visual system is sensitive to moving stimuli,
i.e., a visual stimulus that moves around a lot will elicit a larger response than a stimulus that sits still. This is no big deal: pretty much everything in the nervous system responds more to things that change than to things that stay the same. However, certain amacrine cells are implicated in creating selectivity to the direction of motion in certain classes of ganglion cells, i.e., they will respond more to a visual stimulus moving in one direction than in the opposite. Presumably this is useful in helping to create our sensation of motion. Ganglion Cells: Magno and Parvo; X and Y As mentioned before, retinal ganglion cells are the only way that information gets transmitted from the eye to the brain. They get inputs from bipolar and amacrine cells in the inner plexiform layer, and their axons travel down the optic nerve to various retinal recipient nuclei in the brain. There are many different categories of ganglion cells, and as of this writing we still do not completely understand exactly how many. Because biological systems are so complicated, the separate parts of them are typically first studied by different groups of researchers, and often in different species. Only later, after each subgroup of researchers has firmly established their own terminology, does it become apparent what aspects of the different subsystems correspond. Hence, biological nomenclature will for the foreseeable future remain inconsistent. The researchers studying the retina typically reference the dendritic field size of the ganglion cells, such as midget or diffuse/ parasol, and also whether ON or OFF. Other categorizations include whether a ganglion cell is motion-selective (sometimes referred to as Direction Selective, DS ), or has a response that is more transient or more sustained as a function of time. However, researchers studying the region of the brain where most of the ganglion cell axons go (the Lateral Geniculate Nucleus, or LGN: we will cover it in more depth in the next chapter) Figure 4-13. Cat X ganglion and LGN cells respond emphasized different aspects. strongly to a sinewave grating flashed on and off when the They initially partitioned the LGN centers of the receptive field (indicated by dashed circles) are in the peak white crest or black trough of the sinewave. neurons into magnocellular However, when the receptive field centers are positioned ( magno ) and parvocellular on the half-way point, either white/black or black/white, ( parvo ) cells, based on the you get the null position and the X cells do not respond. Y relative sizes of their cell bodies. ganglion and LGN cells, however, will respond to all Extending this research into the phases: there is no NULL position. rest of the brain, it was found that the magno system is relatively specialized for rapid motion, and parvo for fine detail
and color. Magno and parvo are, however, not typically used when discussing retinal ganglion cells, even though the properties of the magno and parvo pathways originate with the properties of the ganglion cells which connect to them. Be warned: the terminology used to describe retinal ganglion cells does not often line-up in any simple way with the terminology used in the rest of the brain. One partial exception to this comes from studies done in the cat, where instead of parvo and magno (the standard primate terminology) we have both X and Y ganglion and LGN cells. Classically the cat X ganglion and LGN cells are said to have linear receptive field structures with a null position (see again Figure 4-9C, and also Figure 4-13), and the cat Y ganglion and LGN cells do not. Cat X and Y cells are homologous to primate parvo and magno cells, that is, they likely evolved from a common ancestor, and they do share some similar properties, but they are not identical, even as a bird wing is not identical to a human arm. X and Y are nevertheless sometimes used in older texts as if they were parvo and magno. In the primate about 80% of the ganglion cells can be classified as parvo, and are most typically identified as being of the midget classification. The large number of parvo cells comes about because they carry fine detail information, which requires a large number of axons. About 10% of the ganglion cells can be classified as magno, and are most typically classified as being of the parasol morphological class. The remaining 10% are often classified as other, i.e, not currently well understood, and likely contain many different classes of numerically small but still functionally important cell types. The fact that most primate ganglion cells are parvo does NOT mean that the other categories are unimportant. The magno cells have few axons not because their role is unimportant, but because their function of carrying information about rapid changes or fast motion does not require a large number of axons. Photosensitive Ganglion Cells Recently an exception has been found to the rule that only the rods and cones are sensitive to light. There exists a relatively rare class of directly photosensitive ganglion cells that can respond directly to light. These ganglion cells do get direct inputs from the rods and cones, but also use the pigment melanopsin in an extended dendritic tree to also respond directly to light. There are thought to be only a few thousand such ganglion cells per human eye. They project primarily to locations such as the suprachiasmatic nucleus (SCN), a retinal recipient region dedicated to maintaining circadian rhythms, and the pretectum, an area important for the pupillary light reflex. These areas will be discussed in more detail later, but for now it suffices to say that they only require information about the total amount of light in a scene and little else, which is what these cells provide.
Distribution of Photoreceptors Across Retina The photoreceptors are not uniformly distributed across the retina, but distributed in a specific manner that optimizes functional vision (see Figure 4-14). The cones, which are most important for detail and color vision, have a maximum distribution in the center of the field of view: at the very center of the fovea there are no rods at all, only cones. The rods, which operate in the dark, have their maximal density about 20 degrees eccentric from the fovea. Across most of the retina, the cones have a larger radius than the rods, typically about 8 µm for the cones and 2 µm for the rods. This might seem odd as the cones are for fine detail and the rods are supposed to be simple light buckets less concerned with fine detail, but as the rods have a very high convergence onto their bipolar cells (i.e., the signal from a given rod is never sent separately to the brain but always thrown together with the signals from many other rods) they do indeed function more like low-resolution light buckets than the cones. In the central fovea the cones are miniaturized for the finest possible vision, being a little more than 2.5 µm in diameter. In the primate there are three different cones that differ only in the color of light that they are maximally sensitive to. They are blue (short-wavelength, S ), green (medium-wavelength, M ), and red (longwavelength, L ), see figure 4-15A. A digital video recorder arranges the three different colored sensors in a regular grid arrangement (Figure 4-15B). However, the arrangement of cones in the human retina is quite different. As shown in figure 4-15C, there are few blue cones and they are scattered widely. The red and green cones are more common, but they are generally found in Figure 4-14. Distribution of photoreceptors across the retina (redrawn from McIlwain 1996). The blind spot is where the optic nerve exits the eye, there are no photoreceptors there. Nasal is towards the nose and temporal towards the temporal bone i.e. the outside of the face. The blind spot is nasal retina, hence it lies in the temporal visual field for each eye. clumps and different people will have different proportions of red and green cones. This might at first appear to be chaotic, but in fact it is a very efficient design. Consider that color vision has a low spatial resolution, so to determine the color of a patch of the retina we need only have the relative activations of the three classes of cones over a large chunk of the retina: it does not matter if the numbers of S, M, and L cones are unequal. For spatial vision notice that the absorption spectra for the red and green cones are very similar. Hence, for spatial tasks you can treat the red and green cones as being the same, and get spatial resolution at the level of the photoreceptor spacing. In contrast, the rigid
Figure 4-15. A. The normalized absorption characteristics of the three classes of primate cones. B. Arrangement of colored sensors in a typical video camera. C. Arrangement of cones on human fovea one degree eccentric from the fovea (redrawn in approximate schematic format to emphasize distribution from Roorda and Williams, 1999. Real photoreceptors are not of course perfectly uniform circles). grid of the digital camcorder means that spatial resolution is only one-third that of the actual sensor resolution, because you need to clump a red, green, and blue sensor together to get whiteness/brightness at one point. Color One thing is certain about color: it is the relative activations of red, green, and blue cones in a patch of retina that gives rise to the ability of the nervous system to discriminate colors. However, the actual method by which the retina transfers color information to the rest of the brain is still not clear. For midget ganglion cells, the receptive field center could indeed contact a single cone of one color type, but the surround would likely be influenced by cones of different color types. Outside of the fovea, ganglion cells are connected to multiple cones in a given area: averaging the signals from different color cones would destroy color information. Thus, in some manner, the retina must construct receptive field classes that preserve information about the relative balance of red, green and blue cones without transmitting the signals from individual cones to the brain. How does the retina do this? We are still not clear on this point. A big part of the problem is that even though a given class of ganglion cells may carry color-related information in its response, the brain might not actually be using that information. As shown in Figure 4-16, there are currently thought to be two different classes of color-selective ganglion cell in the primate retina. The Type I have a center-surround organization, but the center and surround have different wavelength sensitivities. For example, the R+ center/g-
surround is excited by red light in the center, and inhibited by green light in the surround. Nearly every combination of red and green, and blue and yellow (yellow is red and green) can be found for the type I ganglion cells, except for B- center/y+ surround (possibly because of the quasi-rod like properties of blue cones, and the apparent absence of blue-off bipolar cells). Type I cells can clearly transmit information about color to the brain, however, they confound color information with information about form (because of the center-surround organization). The unsolved mystery is how, or whether, this information is actually Figure 4-16. The two major classes of color-selective retinal ganglion cells. R+ means that red light causes increased firing, and R- that red light suppresses firing. G refers to green, B to blue, and Y to yellow, or red+green. The only possible class of type I ganglion cell that you never find is the B- center/y+ surround (indicated with red cross-out). used by the brain. Remember, the neural retina is just the start of the process of perceiving colors. In contrast, type II ganglion cells have no obvious center-surround organization (see Figure 4-16, right half). The G-/R+ type II cell is inhibited by green light anywhere in its receptive field, and excited by red light anywhere in its receptive field. The entire issue of color vision is surprisingly complex, and we can expect our understanding of these issues to change in the near future.