The Science of Everything Podcast - Episode 47: Vision Part 3

Episode Date: April 7, 2013

In part 3 of our journey through the visual system, we discuss the structure and function of the Primary Visual Cortex, including an analysis of occular dominance columns, orientation columns, and the six cortical layers. I also cover higher cortical regions involved in visual processing, including the V2, V3, V4, and IT areas, and how these regions are divided into distinct 'what' and 'where' processing pathways.

You're listening to The Science of Everything podcast, episode 47, Vision, Part 3. And I'm your host, James Fodor. So, in our previous two parts in our vision series, we talked about the structure of the eye, we talked about the photoreceptors, we talked about rhodopsin, and how that is involved in the transduction of a photon signal into an electrical impulse that the brain can interpret. We talked about the ganglion cells and the bipolar cells in the retina and their receptive fields. We talked about the optic tract and how their signals are sent out from the retina and how they cross over at the optic chasm. We talked about the lateral geniculate nucleus

with its six different layers and the different types of ganglion cells synapsing with different layers in the lateral genicular nucleus. We then talked about how the lateral genicular nucleus sends its input in turn to area V1 of the primary visual cortex. And so it's now time to take a look in more detail at area V1 and how it works and its structure and function and so on. And so in this episode, we'll be talking about that. We'll also be talking about the higher cortical areas like V2 and V3, the IT and the MT areas and some of the other areas in the parietal and temporal lobes that relate to vision. And I'll also be talking about the dual streams hypothesis of visual processing whereby visual processing is basically broken up into the what

and the where pathways that process different types of information. Recommended pre-listened for this episode? Well, you guessed it, vision, parts one and two, and therefore the prerequisites to those as well. You really do have to have listened to those before you listen to this, otherwise it's unlikely you're going to be able to follow much of what I'm saying. Okay, so let's get started.

So the problem in visual cortex is probably, well, it's definitely the best studied visual area in the brain. I might even go so far as to say that it's the best studied area of the brain, at least of the neocortex, which is like the outside of the brain that does a lot of the higher, that's responsible for processing a lot of higher cognitive functions and sensory input-like vision. So in all mammal study, it's located at the posterior pole of the occipital cortex. So the occipital cortex is just the back of the head, posterior pole, just like a really far back part.

So basically it's the very back of the head. So if you touch the far back of your head, you're around the middle, your visual cortex is somewhere around there. Primary visual cortex, I should say, somewhere around there. Because the visual cortex as a whole refers to a whole bunch of areas that are responsible for visual processing. It takes up like a massive chunk of the back of the cortex in total. The primary visual cortex, I was much smaller, and certainly a much smaller area. But that's the first region that the information goes after coming from the lateral geniculate nucleus. So V1, or the primary visual cortex, is defined functionally.

It's defined as the area that is responsible for the first lot of processing of visual input, basically, well, the first lot after the lateral genucle nucleus. There's also a region called the striate cortex, which basically is a region of the occipital cortex. remember the back region of the head, well, of the brain, which is visible to the naked eye, actually, as a result of a bunch of its myelinated axons that have looked like a bunch of stripes since the striated, like the striped cortex.

That's an anatomically defined area. So basically, people looked at the brain and saw, oh, there's this interesting area that looks a bit different. We'll call it the striate cortex. And then people talked about this functional area of the brain called V1 that does primary visual processing. And essentially, it was later realized that these two were the same thing. So in other words, the striate cortex is V1, but they didn't originally refer to the same thing,

so they defined slightly differently, but they turn out to be the same thing. The reason we know this is from a bunch of different types of evidence. So first of all, most LGN lateral geniculate nucleus axons terminate in V1, that is they synapse with neurons located in V1. All V1 cells respond to visual stimuli exclusively, so they don't do other things, but they do respond strongly to visual stimuli. Removing V1 results in blindness, so that's a very strong indication that it's involved in visual processing. And electrical stimulation of V1 elicits visual sensations.

So if you can electrically stimulate, for example, transcranial magnetic stimulation, you electrically stimulate artificially the V1 cells. You elicit visual sensations in the subject, even though they're not actually receiving any visual input, which is pretty cool. Okay, so striat cortex and V1, basically the same thing, and I'm now just going to talk about V1, but I think it's important to talk about the two different names, because sometimes some books will use one and some will use the other,

but they do actually mean the same thing, essentially. They just defined, one is anatomically defined, one is functionally defined. So, like basically all regions of the cortex, remember, the cortex is the outermost layer of the brain, basically. The cortex is very thin. It's only like a millimeter or two thick, which is really bizarre if you think about. Like all the cool stuff is happening in such a small space. But anyway, like basically all the regions of the cortex, V1 is comprised of six layers.

So the LGN, remember, had six layers. These six layers in the cortex don't really have anything to do with the six layers in the LGN. It basically just so happens that both have six layers. There's no particular reason as far as I know that nothing magical about on number six. And as you'll see, it's a bit of a stretch, actually, to say that V1 has six layers, because one of the layers is divided up into three sub layers, and one of those sublays in turn is divided up into two sublays.

So it's more like nine layers or something in total. But anyway, conventionally we say the cortex has six layers, and so therefore the V1 also has six layers. Okay, so first of all, let's talk about the different layers in V1. So layer 4 is by far the biggest layer. Layers are generally referred to by number using Roman numerals. So layer 4 is the layer that is subdivided into three sub-layers, A, B, and C. And most of the axons from the LGN terminate in layer 4.

Again, however, like in the lateral genucleiculate nucleus, different types of input are kept separate in V1. So the input from M cells, remember the M-gangelin cells, the magnocellular ganglion cells, projects into layer 4CA. So layer 4 is divided up into A, B, and C, and layer 4C in turn is divided up into A and B, or I think it might be alpha and beta actually,

but we'll just call it A and B. So M goes to 4CA, B cells go to 4CB, but again, it doesn't really matter which particular layers are. The point is that M cells synaps with one layer and P cells synaps with a different layer. And remember the third type of cells,

the K cells, actually go to a different spot entirely in, basically in layers 2 and 3, and they synaps with particular structures which are called Blobs, and yes, that is actually their name Blobs, and we'll talk a bit more about them later. So each of these different types of ganglion cells, synapses with a different part, a different layer in V1. So I've talked about layers 2 and 3, that's where the K ganglion cells go to, and layer 4, which is where the M and P cells synapse with. Layer 1 is just basically a bunch of axons coming from all the other, layers, so that's not especially interesting. There aren't any actual cell bodies there.

So remember I said before that in addition to receiving input from, you know, directly from the optic tracts, the lateral genuculate nucleus also receives a bunch of input back from V1. Well, that input comes from cells in layer six, which is the bottom-most layer, so layer one at the top and closest to the top of the head, basically, and then you go down, layer six is at the bottom of the stack. The layer six neurons project back to the lateral genucleineucleus. layer 5 cells, so just above that, project two various different parts of the brain, including the midbrain and the ponds, which again are those sort of subcortical areas, which are responsible mostly for essentially subconscious functions, and we're not terribly interested in those.

So don't worry too much about the cells in layer 5. Cells and layer 6 go back to the LGN, as said before. Layers 2, 3, and 4 are the main ones we're interested in because they are the ones that receive input directly from the LGN, and the different types of input from the M, from the M, P, and K cells go to different layers, the V1. Now, there's a few interesting principles of how the information is stored or processed in the primary visual cortex,

which also apply to some other regions of the brain as well. Now, I've already mentioned one of them, which is that different types of information from the different type of ganglion cells go to different layers in the primary visual cortex. This applies not just to the primary visual cortex, but also some later cortical areas, and it also applied, as we saw before,

to the lateral genucin nucleus. Different layers than the LGN, got input from different types of ganglion cells, different layers in the visual cortex, got input from different types of ganglain cells. There is another very interesting principle which is called retinotopic mapping, or it's also called spatiotopic mapping, various few of terms. This basically means that if I picked out a particular neuron, this could be in V1 or in V2 or V3, high-cordiclaris,

or it could just be in the LGN, so any of these, if I just picked a neuron there at random, and then figured out which photoreceptor cell or cells provided the input that corresponded directly to this neuron, and I would find some because there's a sort of a one-to-one linear mapping that the information is carried through directly through the LGN to V1. So I could find particular neurons that corresponded directly to this given cell in V1. Now, suppose that I move, I don't know, a few micrometers to the left or something like that,

or to the right, whatever. If I moved a few micrometres to the left and looked at a cell in V1 just to the left of the first cell that I looked at, and then I examined the original photoreceptor neurons that provided input to this second V1 area cell, I would find that the cells on the retina were right next to, particularly on what I say, the right or the left, whichever side of the photoreceptor cells that provided input to my first neuron.

So said in a different way, hopefully a clearer way, if you have two neurons that are next to each other in V1, or in the LGN, or in many regions of the visual cortex, they will receive input from groups of neurons that are next to each other on the retina. and so the further away the neurons are in, say, V1 or in the LGN, the further apart their original photoreceptor neurons will be on the retina. So this is retina topic mapping.

In other words, there seems to be a direct spatio mapping of the cells of the retina onto the LGN and onto V1 and other areas as well. Obviously, it has to be flattened out and altered so that the shapes of the different structures correspond, and it's not perfect, but it's a fairly close mapping. And it's very cool. So literally, like if you were to draw, I don't know, a smiley face on the retina somehow directly, like a pattern of activation of the photoreceptor neurons that look like a smiley face, then you would also observe a pattern of activation

that look like a smiley face in LGN and in V1. It would probably look squashed because it would be shifting in various ways because the structures have some of different shapes, you know, the LGN is like a bent knee, whereas the back of the retina's the internal part of a sphere. But the basic point remains that you would see that smiley face carry on as you went up through LGN and through to V1 and to high. areas. So this is retinotopic mapping, and it's very cool. So it's another example of how the brain preserves information. It doesn't just preserve information about which eye the input came from.

It doesn't just preserve information about which type of ganglion cell it came from. It also preserves the information about where in the visual field it came from in the form of where in the LGN or in the V1, that cell, the corresponds to the given neuron, is located. So the location of the cells carries information as well. So that's retinotopic maps. Another principle, this one only applies to v1, it's not relevant to the lateral geniculate nucleus. Ocular dominance columns. So ocular dominance columns are sort of strips of neurons in the visual cortex that respond preferentially to input from one eye or the other. There are ocular dominance cells for both eyes, and generally they occur in pairs, and one for the left eye, one for the right eye. Now this doesn't apply to the

LGN, because remember there are two LGNs, one on each side of the head, whereas there's only one visual cortex, essentially because it's in the back of the middle of the brain, so, I mean, You've got left and right sides to it, but there's only one visual cortex. The inputs from both LG and both on each side of the brain converge in the single V1 area at the back of the head, of the back of the brain. But the inputs are still kept separate by these ocular dominance columns. So you'll have a bunch of input that goes on. You know, the layers are still all the same. The different types of, you know, the M cells and the P cells and the K cells, they all still project to the same layers.

It's just that in one little region of V1, all of the input basically comes from, well, not all of the input, but most of the input comes from the left eye. and in an adjacent region, most of the input comes from the right eye. But the retinotopic mapping is still preserved, so that essentially you've just got the same error of the visual field represented, but in two adjacent regions of the cortex, one section will respond to input in a given area of the visual field from the left eye. The other area of the visual cortex will respond to the same region of visual field. It's just that it will only respond to input from the right eye.

So same visual field errors, the retinotopic mapping still applies, but responding only to input from one eye or the other. So these are called ocular dominance columns. Another principle of maintaining separation of information. And there's yet another fascinating principle of the primary visual cortex, probably the most interesting one. These are called orientation columns. These are a relatively recent find, only a few decades old, I think.

They don't exist in the LGN. Essentially, we think because the level of processing in the LGN is not as rich, sort of as more sophisticated as the level it's reached by the time we get to the primary visual cortex. Because each, basically, the higher you go up the hierarchy in a sense, the further away you get from the eyes, the more complex the processing is, and the more highly the signals being processed, and so the more complex the responses tend to get. These orientation columns are essentially located in the ocular dominance columns.

That is, each ocular dominance column has its own bunch of orientation columns. So they're located within these things. Basically, these are neurons, or small groups of neurons, which respond preferentially to bars of light or lines, basically, but only of a particular orientation. So a given cell might respond only to a line that points vertically. And then another cell next to it might only respond to a line that points, you know, that's got like it's bent 15 degrees to the left,

and then one next to that, a further 15 degrees to the left, and so on. And so there's cells that correspond to all different orientations, I think it's like a 15 degree intervals going all around, all the different possible combinations of orientations. These orientation columns span multiple layers in the cortex, So they don't just apply to some of the inputs, so it applies to all the different layers. So that means it applies to all three different types of the ganglion cell inputs, M, P, and K. So it's all of those layers, and in all of the ocular dominance columns,

so from both eyes, from all types of gangling cells in those eyes, you find these orientation columns. Even more fascinating, the geometry of these orientation columns is arranged, such that basically as you move around, the orientation of the line that the cell is responsive to, changes by a small amount. suppose I find a cell that responds only to vertical lines. If I then move to a cell right next to it, that cell won't respond to, say, horizontal lines. It will respond to lines that are just a little bit off vertical. And then if I move further along again, it'll be a little bit off that and so on.

And so if I want to get to a horizontal line, I'll have to move all the way around. So there's literally a progression of orientations going around from vertical to horizontal and back around again. So you can get any orientation you like, but only as you go correctly sort of around the order. These orientation columns, there are many of them, So it's not like there's just one orientation column in each layer or something like that.

There's many sections of a given chordical, a given layer of V1. A given region will be like one full set of orientation neurons. So it'll be able to basically respond to the full 360 of different line orientations. And then a region next to that might be another set of orientation column neurons, except it might be from a different eye or it might be the same eye, but it's just a different set. So there's many, many of these orientation columns that are all sort of laid out in like a pinwheel sort of structure. The columns move radially outwards from a single point, basically. So if you move along

the radius, the preferred orientation of the line doesn't change, but if you move around the circle, it does. And then if you move from one orientation column to another, well, you sort of start all over again. And as I said, before, the ocular dominance columns are incorporated into these orientation columns, such that you have, maybe you've got two ocular

dominance columns, one from the left eye, one from the right eye, within each of those. Each of the ocular dominance columns has a full set of orientation cells, orientation columns, within it, so, you know, the full 360 degrees, and then, so you've got those two sit next each other, and then you've got another one, and then you've got another one, you've got many of these sets of orientation columns and ocular dominance columns all throughout V1, and again, all throughout the different layers of V1 as well. Now, it's thought that these orientation columns are possible, because basically all you'd have to do is take a couple of ganglion cells,

say from the same type of ganglion cells, so both a few M cells or a few P cells or something like that, and just synapse all of them with a single neuron in V1. And as long as you selected the cells appropriately such that they were all in a line, because remember, we've got retinotopic mapping. So if we pick three neurons in a line, say three ganglion cells in a line, we will get input from the visual field that corresponds to an actual line in the visual field. There won't be three random points on the visual field.

A line of ganglion cells corresponds to a line in the visual field. So if a given neuron in V1, synapses directly with all. all three of those, or four or whatever, of those ganglion cells, and it requires, say, all, input from all of those in order to fire, so one or two ganglion cells alone couldn't get the V1 neuron to fire, but it needed all input from all of them to fire, then what you would have is a V1 neuron that selectively fired only in response to a line of that orientation, such that, you know, the ganglion cells that gave an input were lined up in the right direction. Even if a bunch of ganglion cells fired and were synapsed with this hypothetical V1 neuron

that we're talking about, it still wouldn't fire because they would need to be oriented in the correct orientation. Essentially, because imagine you've got four, imagine you've got four ganglion cells in a sort of a horizontal line and four in a vertical line. Four of them in the vertical line synapsed with one V1 neuron. Four of them in the horizontal line synapses with another V1 neuron. If the four horizontal ones fire, basically because you see a horizontal line somewhere in the visual field, then only the V1 neuron that synapses with those four horizontal lines will fire. The other one synapses with the vertical gangling cells, even if one or two of the cells that in the horizontal line are also in the vertical line,

it still weren't fired because it doesn't have enough input from the vertical neurons. Of course, if all of the four vertical and all four horizontal neurons all fight at the same time, then both of the neurons in V1 would find because they both have the input of the right amount. So basically this just means that the orientation columns in V1 are possible because you elongate or stretch out the input, the visual input fields of the, of the, neurons in V1 by synapsing them with a bunch of adjacent ganglion cells upwards. So I mentioned before Blobs, I'm going to talk about them a bit more now. Blobs are particular groups of cells which look like blobs when you stain them in a certain way

and view them under a microscope, so hence their name. They're found in layers 2 and 3 and also layers 5 and 6 in V1. They're not found in layer 4. Remember, layer 4 is where most of the input from the LGN comes from, and so they don't tend to be found there for whatever reason. But blobs are, they're not direction sensitive. So the direction sensitive cells seem to be mostly like in layer four. They're sort of in other regions as well, but they're not in the blobs. The blobs seem to be very important in color vision, although their exact role is still fairly poorly understood. But remember, before I said that the K-type ganglion cells, their axons go directly

to the blobs in layers two and three. Well, that's significant because we know that K-type ganglion cells seem to be more responsive to colour because, you know, they synaps with it. Their input ultimately comes from cones, not rods, and it seems that they are mostly carrying information about colour, so the fact that the blobs are also doing something, some type of processing relating to colour, therefore makes sense. I should just mention a few additional complexity. So not all of the cells in V1, or even in layer 4 of V1,

are directionally sensitive. Many of them are, and that's what's, they're some of the coolest ones, so they tend to get a lot of attention, but not all of them are. There are also simple cells, they're called, that have receptive fields that are just the same as those found in the LGN and ganglain cells directly. So they're just, you know, the normal simple centre, on-center or off-center, and then the surround donut receptive fields. So you find some of those as well.

And you also find some other complex cells who have much larger receptive areas, so larger regions of the visual fields, and some that are sensitive to motion, some are sensitive to colour, some aren'ts, and so there's a lot of variety. But I mentioned the ocular dominance columns, the orientation columns, and the blobs, because they're some of the most interesting structures. that we find in V1, and they seem to be very crucial in processing. The crucial thing to understand is that basically the input comes from the ganglion cells, all three different types, projects to the LGN,

and in turn the information projects from there to V1, where it undergoes further processing. One other thing I wanted to mention is that you might have sort of thought that the whole LGN thing is kind of pointless, because the input just goes there and then go straight from there to V1, and it's not until you get to V1 that you find interesting things like the orientation columns. So is it true that the LGN is just like a way station? Basically, the information just goes there and then is passed straight on.

It's like a train station, basically. You just sort of get on, get off one train, get on another train, and a carriage the next station. You don't really do anything there. There's no real processing there. That may be true to some extent. It's definitely true that the main processing doesn't occur until V1 and later cortical errors. And the thalamus, where the LGN is located,

is sort of like a way station because a lot of other sensory input passes through there as well. In fact, I think all sensory input passes through the thalomomers. before it's processed by the cortex. However, it does seem like that the LGN at least does some processing, so it's not just a waste station. The information doesn't just pass through there. The LGN actually does something to it. And one piece of evidence that we have for this is essentially that the output patterns that we observe in the LGN, that is, you know, when we stick in electrodes and figure out what the neurons in the LGN respond to, it's not exactly the same as what the ganglion cells respond to back in the retina, and it's also not the same, not exactly the same, as

what neurons in V1 respond to. There's a lot of similarities, like we still have the retinotopic mapping, for example, and we still have separation of input from the different eyes, like, you know, you've got the ocular dominance columns in V1, you've got the different hemifield views going to different layers in the LGN, but although there are similarities, they're not exactly the same. So this indicates that there is some processing going on in the LGN. Also, remember, there's a lot of, in fact, most of the input to the LGN actually comes back from V1. So input goes from, it comes from the retina, from the ganglion cells, goes into LGN, goes to V1,

and then a whole bunch of it comes back into the LGN. So there's a lot of feedback there. The retina doesn't have that, and V1 doesn't have that either, because it doesn't have feedback from itself. So that feedback from V1 is going to change the output of LGN. So it's not, in other words, the output of LGN is not simply a product of the neurons that come in from the ganglion cells. It's also a product of the neurons that come back, or the axons from the neurons that come back from V1. So it's a combination of those things. And it's not exactly clear how those into play, but it is clear that some processing happens in the LGN

and further processing happens in V1. So the LGN is necessary for visual processing, even though it's not exactly clear what it does. I mean, it's not exactly clear what any part of the visual system does. It's probably clearest what V1 does because we know about the orientation columns and the ocular dominance columns and a little bit about the blobs too. But exactly how it all breaks down, it is still not very clear.

Okay, so we've now covered the first four stages of the visual system, the eye, the retina, moving into the brain, including the lateral genuculate nucleus, and the primary visual cortex, we're now going to move on to the fifth and final of our five stages, which is high-level processing in cortical areas beyond V1. Okay, so first of all, it's important to understand a very basic concept in our sort of current model of high-level visual processing, that of the two-stream hypothesis, essentially. The two-stream hypothesis basically says is that there are two distinct pathways of neural information that flow from both of these pathways begin in the V1 area, so that, you know, they take all the same up to that point, they take all the same inputs, but the neural pathways diverge after V1 and proceed through different parts of the brain. There's different names for these.

There's the, well, the proper names are the dorsal stream and the ventral stream, but we won't use those slightly confusing anatomy terms. They're also known as the where and the what pathways. So basically the idea is that the dorsal, or that just means basically the upper stream, or the upper path of information, goes around sort of the top of the brain across the parietal lobe, which is on the top of the head, and it focuses on processing mostly information about where objects are in the visual field, motion, things like that.

So that's why it's called the where pathway. The lower pathway, also called the ventral stream, goes sort of down along the bottom of the brain, loosely speaking, and is referred to as the what pathway. So it focuses more on object recognition and the finer details. So basically, again, we've got the upper stream or dorsal stream, which is where and the lower or the ventral stream, which is about what. So the what and the where pathways.

This is a very sort of high level of analysis. Obviously, there's a lot of detail and complexity in there, and we'll look at some of that. But there's a number of pieces of evidence that point to this sort of dual stream hypothesis or dual pathway hypothesis, including some interesting research into, lesion studies or deficit studies where basically we look at people who have particular rare or unusual neurological disorders and we can use that to infer how the functionality of the brain works. So there is a disorder called optic ataxia which is associated with damage in the

oxibital parietal cortex. Basically that's exactly the site of the wear pathway. So that's along the top of the brain. And it results in a lack of coordination between, lack of hand-eye coordination, basically. So people with optic ataxia can still name objects and they can still read and they can still perceive more or less normally, but they are unable to use visual information to coordinate hand motor movements. So for example, they could see, I don't know, a pencil or a spoon or something like that and say what it is, but when they were asked to use it, they'll be unable to do so properly. They wouldn't know how to grip it or be able to use it in the right way, because they're having trouble incorporating the information basically about object recognition

with actual use of the object and object's spatial location. And the reason for that, or the inference based on that, is that the occipital parietal cortex or the upper region of the brain we're talking about, is associated with where. It's associated with object location, hence it's called the where pathway. Even more interestingly, there's essentially the complete opposite of that condition, where the patient is capable of grasping and using objects, but they're not capable of naming it. So basically, in optic ataxia, you can see a pencil or a fork or something and say, yes, that's what it is, but can't use it properly.

In the other condition, the patient is able to grasp the pen or the fork and use it properly, but if you ask them what it is, they wouldn't be able to say so. It's rather fascinating how one is able to sort of conduct a motor task without actually be able to verbalize what one is doing, but that is possible because it's observed in these conditions. So these, it's called double dissociation, that is you can have one without the other in both cases. The double dissociation of these two conditions allows us to infer that there are two separate regions of the brain,

or streams of processing, one for object location and sort of motor function, and the other one focusing on object recognition. So the what and the where pathways. And, you know, there are other piece of evidence, too, like functional neuroimaging showing that different regions of the brain are active during different types of tasks. But we won't go into details of that. Suffice it to say, there are various convergent lines of evidence

that indicate that there are two, broadly speaking, there are two streams of processing. The bottom is the what pathway, and the top is the where pathway. Okay, so now that we've got that broad idea of the two-stream hypothesis, we're now going to look at the higher cortical areas in more detail, including those in the what and the where pathway. But before we get to either of those, we're going to talk about areas V2 and V3.

These are the cortical regions that are basically come just after V1. So information goes to V1, and then directly after that, it goes to V2 and or V3. So regardless of whether information is processed through the where or the what pathway, it will always go through V1 and V2 and V3, or it might go straight from V1 to V3, but it always goes through that initial V1, V2, V3. And only after V3 does it diverge between the WOT and the WARE pathways. Again, this is all broadly speaking.

There are going to be exceptions, but broadly speaking, this is the case. Area V2, it's the region of the visual cortex that essentially comes right after V1, and so hence it's called V2. It's very similar to V1, basically, not actually that much is known about it. We observe cells that are tuned to simple properties such as orientation or the spatial frequency of the light that it observes and also color. So similar things that we observe in V1. There are, however, some more complicated properties of neurons found in V2. So we refer to these as complex neurons that respond to complex stimuli as opposed to simple neurons that respond to a very simple stimuli,

like just light in this given region of the visual field or something like that. So in V2 we start to observe the first really complex neurons, although there are some in V1, such as orientation due to illusory contours, which is a particular type of optical illusion, cells that respond to binocular disparity of vision, and also cells that respond differentially depending on whether the stimulus is part of a figure in the foreground or sort of a part of the ground. These types of things, illusory contours, the binocular disparities, etc.

These have been demonstrated in various studies that they've done. There will be either neuroimaging studies or placing basically microelectrodes in that region of the brain. And generally this is done with monkey brains rather than human brains for rather obvious reasons. So basically they're either putting microelectrodes in the brain and seeing when the neurons activate or they're using neuroimaging to see where does this region of the brain, when does this region of the brain activate when we present various stimuli. And from those sorts of studies, we've found neurons that respond to these particular somewhat complex situations.

But it's not very clear exactly how all that fits together. So the basic picture from V2 is that it's similar to V1, but maybe it's a little bit high level again because we seem to have a few neurons that respond to somewhat more complex properties. Area V3, very much the same story. Again, it's not very well investigated. There's been observed some selectivity of neurons to color, direction, and various patterns, but we don't really know very much about it. So the broad picture seems to be consistent that as we move from V1 to higher cortical regions, by the way, when I say higher, I don't necessarily mean like literally higher in the brain,

as in closer to the top of the head. That's not really what I'm saying. I'm just mean higher in the sense of the, if you think about levels of processing, the processing begins at the retina and then moves on to the LGN and then to V1 and then it moves on to subsequent regions that deal with a higher and higher level of abstraction. And that's what I mean by higher cortical regions. So as you move to higher cortical regions, the field of vision, sorry, the receptive field of any given neuron seems to generally increase. That is, a given neuron will respond to stimuli over a larger portion of the total visual field. And also it seems to be the case that more and more neurons, as you go higher up in the visual cortex hierarchy, more and more of the neurons that you find

will respond to complex stimuli, like, for example, the direction of lines, or particular patterns, or binocular disparities and things like that, rather than just simple, like lines of a given orientation, or simple colours, or even simpler than that, just light in a particular area of the visual field. So the higher up you go, the more complex

the neuronal responses get. And that begins really, particularly in V1 and V3, sorry, V2 and V3 after coming out from V1. Okay, so V2 and V3 are the first areas that come after v1. Then we observe a sort of bifurcation of the information or of the neural pathways into two, the what and the wear pathways, as I mentioned before.

So we'll cover the where pathway first. This is the dorsal pathway, so it comes sort of along the top of the brain. The first part of, or the first region that we identify within the where pathway is called V5. You might wonder what happened to V4. Well, that turns out that's in the ventral or the watt pathway,

so we'll get to that. But area V5 is also referred to as area MT. Now, this area receives input from areas V2 in V3, and also a little bit directly from V1. So it's not necessarily purely linear. It doesn't have to go from V1 to V2 to V3, but some of it does, some of it goes directly from V1 to V5 and so on. So it can skip stages, but generally there is that forward progression.

There are also backward projections as well, so there's some projections from V5 back to V1 and so on, just as we saw before, there were backward projections from V1 to the LGN. MT neurons respond very strongly to motion, much more so than earlier regions. We talked about some motion-sensitive neurons in, say, V2 and 3, but V5 really seems to specialize in this type of, in recognizing this type of activity. Interestingly enough, we also have movement columns, which are very similar to the orientation

columns found in V1. So remember, back in V1, we had our orientation columns where if you moved from one cell to the next cell, it would respond to lines in a particular, in a slightly different orientation, so the further way you move from your initial cell, the more the orientation changes that the neuron responds to. Well, we observe basically the same thing in V5, except instead of just the orientation of the line, it's now the direction of motion. So a given neuron will only respond to motion up and down.

The one next to it, the neuron next to it will respond only to motion sort of, you know, 15 degrees to the right of up or something like that. And the one to the next neuron along will respond to motion maybe 30 degrees to the right of up and so on. So you're kind of going around in all the different directions. So we observe orientation columns but for motion this time, which is very interesting. And in fact, scientists have been able to reliably alter the direction of motion perception of monkeys, generally macaque monkeys, by so.

selectively stimulating parts of V5. So that is basically, they do something like presenting a screen of dots just moving down or moving sideways or whatever, and they can train the monkey to basically tell them what direction the dots are traveling in, give them food rewards if they answer it correctly, and you can train monkeys to do this quite reliably. Then what the scientists do is that they'll selectively stimulate just the particular neurons in V5 that are sensitive to motion in a given direction. And so, and by doing that, they're able to change the monkey's perceptions.

So the monkey will start giving wrong answers. So say the dots are going up and down, but if the neuroscientists start stimulating all of the side-to-side motion-sensitive neurons, then the monkey will start more often saying that the dots are moving side-to-side. Or they'll give some sort of hybrid response, whereby some of their neurons are telling them it's up and down, some of them says it side-to-size, and maybe they'll say it's like diagonal or something like that.

Yeah, scientists have been able to reliably do this in monkeys, which I think is pretty cool. Also, there's at least one case of a human stroke victim who had a localized legion, oh, sorry, lesion, to area V5, which basically means that just that region of the brain was damaged, but no other regions. And they reported the inability to perceive continuous motion. Otherwise, their vision seemed to be unaffected, but the only thing that was wrong was that they could not perceive continuous motion. Instead, what they saw was a series of disjointed images.

So this would sort of be like, you know, picking out like every 15th frame from a movie or something and just watching those. It would be very disjointed, one thing, and then, you know, the person's over there and then they're over there. So, for example, this person had great difficulty crossing the street because they say things like, well, the car's far away, but then the next moment it's close. I'm like, I don't know how fast it's moving, and I'm not sure if it's safe and so on. So obviously that would be a hard condition to live with, but it is interesting from a neuroscience perspective because it's another clear piece of evidence that area V5 does seem to focus very strongly on detection of motion, and motion in particular directions. Okay, so that's the first region within the wear or the

dorsal pathway, area MT or V5, same thing. The next area long, the main one that we're going to talk about is area MST. So this is the, medial superior temporal area. Again, we won't worry too much about what all that means, just MST. It receives most of its inputs from area MT, the medial temporal area. So basically, you've got V5, and then that sends most of its input to, most of the output of V5 goes to MST, which is the next one along. Both of these are in the wear pathway. And again, MST is involved also in the detection of motion. Particularly, it seems to be involved with detecting slightly different types of motion. For example, optical flow has been associated with, detection of optical

flow has been associated with area MST. Optical flow, I'm referring to the apparent smooth motion of objects, especially when the object aren't necessarily moving, there might just be spinning. So, like, if you watch a tire spinning around, that might be optical flow, because it looks like it's moving around continuously. Again, these are just sort of isolated studies that have found these sorts of things. So it's a bit up in the air exactly what the difference between MST and MT is, but they're both definitely involved in motion and perception of, yeah, the movement and location. of objects in the visual field. I should also mention that all of the areas we've been talking about so far,

so V1, V2, V3, MT, MST, and the ones we'll be talking about later on. Pretty much all of them have the spatiotopic mapping, that is, so if you pick a neuron, it responds to output from a given region of the visual field, and then you move to a neuron next to it, it will respond to input from a neighboring region of the visual field, so that you have a direct mapping of retinal input

onto these different areas, V2, V3, MST, and so on. The size of the receptive fields have given neurons tends to increase as you go up the cortical areas, as I mentioned before, but it still maintains generally the same type of spatiotopic mapping. Now we'll move on to talking about the what or the ventral pathway, which runs along the bottom of the brain. Again, this receives input directly from areas V1 and V2 and V3. The first region within this pathway is called area V4, so I promise we'd come back to that. This is area V4. This seems to be responsible mostly for the detection of color, and,

color vision, and so this area has received a fair bit of attention to try and work out how that works exactly. It also responds to various simple geometric shapes and patterns of particular type. So what scientists generally try and do is that they'll try and elicit the maximum response from neurons, say, in area V2 or V4 or whatever, by presenting different shapes to the visual field of that particular neuron. So they'll present a vertical line to it, and they'll get, I don't know, a small amount of response. Then they'll try presenting a square to it, a square of light or whatever, of a given color maybe, and they'll get maybe a bit more response,

and then they try a spiral, and they get a bit more response, and so they just try and keep varying the shapes to get more and more a higher and higher level of response. The trouble with this method is that it's very hard to know, actually, when you've found the thing that the neuron is really specialized for, because generally neurons will respond. They have basically a graded response, whereby the closer you get to what they're truly specialized for,

the more they'll respond, but they'll always respond at least to some degree to pretty much any visual input. So you can't really know if you've found the true thing that the neurons specialised for, because all you can say is, well, this particular stimuli has given us more activation in this neuron that we're looking at than anything else that we've tried to present it with. But maybe there's something we haven't tried yet that gives even greater activation. So it's very hard to know what these neurons are truly specialized for, but some of them do seem to be specialized to respond to geometric shapes of particular types,

like a square or something like that, or a spiral. But also it does respond to color. And there's an interesting condition called achromatopsia. I mentioned this before, in which patients do not see any color at all. They just see the world in gray. And it's not clear if this condition is related to V4, but it does lend support to the notion of specialized color processing in one particular part of the brain or one particular pathway.

Because apart from the lack of color perception, achromatopsia patients seem to have normal vision. So it seems like you can selectively disrupt just that part of visual perception. Okay, so that's area V4. Moving on now to the next area, which is area IT, or the inferior temporal cortex. This region takes input from V4, so that the previous region in the ventral pathway. And this region is one of the more interesting higher cortical regions, because it's known to respond very specifically to quite highly abstract images.

So this is especially where you start to get selective responses to quite particular shapes, even more so than area V4. In particular, there's one sort of sub-part or sub-region within the inferior temporal cortex, which is called the fusiform face area. And it's so-called because it exhibits particularly robust activation in response to faces being presented to its visual field. Now, again, the caveats that I mentioned before apply here. So it's not clear whether the fusiform face area is truly specialized just to view faces, or if it's specific to viewing, to responding to things that kind of look like faces, or maybe it's something different entirely, but it does seem to respond

quite specifically to faces. And basically what scientists try and do is just present it with parts of faces or things that are neely faces and see how much it responds there. So like he can present it with a smiley face and it seems to respond quite well, not quite as well as an actual picture of a face, but quite well. So it seems to have a sort of an abstract idea of what a face should look like and respond mostly to that. But again, it's not completely clear exactly what role it plays. And again, there's a very interesting neurological disorder called prosopagnosia, which is caused by legions to particular parts of the inferior temporal and oxibital cortexes, so including the fusiform face area, and in prosopagnosia, basically you can't recognize faces.

Vision is otherwise normal, but you have trouble distinguishing faces. I mean, you can see that that's someone's face, but you just can't recognize one person from another, at least not based on their face. You might be able to recognize them as on other characteristics, but not their face, necessarily. So, again, it's not clear that the only cause of prosopagnosia is a lesion to the fusiform face area, but it does seem to be one potential cause of prosopagnosia, and it also, the condition generally leads support to the idea of specific regions in the brain responsible for processing faces. And I think this is very reasonable, evolutionarily speaking, because humans are exceptionally

good at recognizing faces and distinguishing one face from another. I mean, you think about, if you look at a gorilla or a monkey, even things that are quite similar to us and try to recognize them by their face. You'll have a great deal of difficulty. The reason is because they're so very similar. Human faces are also so very similar, for the most part, but we don't perceive it that way. We perceive different faces as being very different. That's not because they objectively are very different. You know, just get computers to try and recognize faces, and you'll essentially discover that. It's just because we're very, very closely attuned to very subtle differences in the structure and spacing and so on of different faces. And that's obvious, that has obvious

evolutionary advantage in terms of recognizing your relatives or recognizing people you're cooperating with and things like that. And there's a lot of work in evolutionary psychology, and so on about this, which is quite interesting. But the point I'm making here is that it would not at all be surprising if we had a particular region of the brain that was strongly specialised to facial recognition, given how important it is for primate, especially human social interactions. Okay, so moving on from the IT area to the third and final part of the Watt

or ventral pathway, which is called the intraparital sulcus. This has many sub-regions, which you might have heard of or seen in diagrams before, including the lateral and the ventral intraparital. sororietal sulcus and there's the medial intraparital sulcus and the anterior intraparital sulcus and so on. So they have various acronyms like VIP and LIP and AIP and so on. I just mention this in case you've seen some various diagrams of the high cortical regions. You'll see all these acronyms on them. Basically they're just referring to various parts of the intraparital sulcus.

By the way, you may have noticed that when we were talking about regions like inferior temporal cortex and now the intraparital sulcus, we've moved out from the occipital lobe, where V1, V2 and those main visual areas, are located. We've now moved into more anterior, that is more forward regions of the brain, including the parietal and the temporal lobes. Don't worry if you don't know exactly where those are, but the point is we've just moved out of the very back region of the brain where we started with with V1, because basically as you move into higher and higher cortical areas, the information tends to also move forwards towards the front of the brain, where the prefrontal cortex is located to the very basically behind your forehead, and that's where higher conscious functioning occurs, that gives rise to, you know, awareness and attention and so on.

And eventually the visual information is going to have to get to the prefrontal cortex in order for us to be aware of it and perceive it. So it's not at all surprising that we sort of move from back to the front of the head and get more abstract and higher order as that occurs. The principal functions of the various sub-regions of the intraparietal sulcus seem to be related to motor coordination, directing eye movements, visual attention, things like that. So, again, quite high-level functions that are sort of on the way towards, you know, conscious visual perception.

So we won't say too much more about that, but I just wanted to highlight. highlight those areas. Again, it's still very up in the air about exactly what these different sub-regions are for, but they do seem to have distinct functionalities and relate to attention and awareness and coordinating visual perception with motor function and stuff like that. Okay, so that's the where and the what pathways, and that completes our essentially tour of the higher cognitive regions of the brain and the different functions that they carry out in terms of visual processing. And that's about it for this episode. In the four fourth and final installment of our vision series, we'll be talking about, I'll be talking about

the more abstract level of computational analysis that we suspect the visual system may be engaged in in terms of how it detects edges and how it recognises objects and detects motion and things like this. I may not make that the very next episode because I'm thinking of sort of separating that app from the main body of the three vision episodes because it's a little bit different in the sense that it's not talking about the anatomy and the biology so much as the sort of computer science, more abstract side of things. So I might not make that part of the official series, but it will be unofficially part of the vision series because I will be covering some of the important topics that I haven't talked about so far.

So, hopefully you enjoyed this podcast. If so, then please jump on to iTunes and give me a favorable review. Five-star rating is good, or an actual comment, an actual rating, you know, with words and a description of why you like the show is even better. I'd love to hear from you if you have a suggestion for a future topic or feedback about a previous episode or just if you want to tell me who you are and where you listen to the show and how you found out about it and whatever else. I'd love to hear that as well.

My email address is Fods12 at gmail.com. That's F-O-D-S-12 at gmail.com. You can also find the podcast on Facebook if you type in The Science of Everything podcast in a Facebook search. You should be able to find our page. Give us a like. By joining the Facebook page, you'll be helping to spread the word about the podcast. You'll also get updates about future episodes that I'm working on,

and I also post visual materials to accompany the podcast and provide an additional learning tool there. The visual material that I post is actually quite high quality, I think, because I go to a fair bit of effort to make sure that I find the best images and diagrams and someone available. Obviously, you can just do a Google search and find your own, but I try and find ones that are most relevant to the episode. So check those out. I think they're a good resource. Again, if anyone has any ideas or feedback about what they would like me to talk about, what

special topic they would like me to talk about for episode 50, I'd love to hear from you. I haven't heard of any feedback yet, so I suppose if no one says anything, maybe I'll just do an ordinary episode. I don't know. I just thought it might be interesting to do something a bit different for episode 50 as a sort of a special occasion. Talk about the physics of time travel or movie mistakes or something like the common science movie mistakes, something like that. You know, if anyone has any preferences on that or ideas, send me an email or post on the Facebook page. So, that's enough of rambling for me.

Thanks for listening again, and I'll talk to you next time.