That Neuroscience Guy - The Neuroscience of Top-Down Processing

Episode Date: April 7, 2024

In today's episode of That Neuroscience Guy, we take a high-level (pun intended) look at how your brain biases basic functions like attention and perception based on previous experience, otherwise known as top-down processing. 

Hi, my name is Olof Kregolsen, and I'm a neuroscientist at the University of Victoria. And in my spare time, I'm that neuroscience guy. Welcome to the podcast. You know, one of the big themes for the podcast, the way I see it, has been trying to explain how the brain works, which seems obvious enough for a neuroscience podcast. And I was looking back at our episodes and thinking about things that were there and not there. And I realized one concept that's absolutely crucial that I haven't talked about. And I've hinted at it. I mentioned a bit of it a long time ago when we talked about vision, but it's the concept of top-down processing and what that means within

the brain. A lot of our brain is bottom-up processing. What I mean by that is your eyes, for instance, if you think of visual perception, they take in an image that goes to the back of the head, the primary visual cortex, and then the dorsal stream and the ventral stream work together to build that up into something that you see. But it's more complex than that. One of the things that happens in the brain is your brain is current always, like all the time, making guesses about the world. It's making guesses about what you're looking at, about outcomes of decisions, even about motor actions you might be wanting to do. So this is called top-down processing, and that's what I'm going to talk about on today's podcast. So if we think of the perception of an object, see we're looking at something, we see a

fish. You know, that perception is instantaneous and it's effortless. It happens, you know, right away. So you might think from a bottom-up processing angle that the brain is constructing this image in the ventral stream and the dorsal stream, which we did talk about. But there's something else happening at the same time, which is the prefrontal cortex is searching your memory to try to guess what this object is. And the way it does this, at least one of the beliefs about how it does this,

is your brain does one of two things. It basically builds a template. And what the template is, it's a memory that you hold on to, which is basically the information you need to decide if something is a fish or not. All right, a tail, a fin, you know, scales potentially. And that template isn't an exact image of a specific type of fish. It's basically your brain's concept of a fish in general. So one way to think about this, if you're pondering what does this really mean,

let's go with dogs. We'll move it to dogs for the example. What would your template of a dog be? Well, it would be that it has four legs. It would be that it barks. It would be that it potentially has a tail. But imagine that that template is like the average of all the dogs you've ever seen. And this is the theory about how this works, is your brain has these templates. And then what happens is when you see something,

it's interesting because there's some cool research behind this. As the ventral stream and the dorsal stream are building up this image, trying to identify it, at a very early stage of visual processing, probably the second stage or maybe the third, what happens is a very low-level resolution image is sent to the front of the brain, all right, to the prefrontal cortex, into your working memory system. And basically that system then uses that image and tries to match the template to it. So think of you're looking at a dog, and when I talk about a low-resolution image,

it's basically just the outline of it. You're missing all the sharp contrasts and details. And that image is forwarded to the prefrontal cortex. And then the prefrontal cortex compares that to the templates you have. Is this a cat? Is this a dog? All right. Is it a fish? And when the brain gets that match, all right, the template matches the low level image, we can successfully and rapidly identify the object. And that's top-down processing. This influence of the prefrontal cortex on these lower systems, basically saying, listen, you can build up the image, and that's important to do, but we're going to start making guesses. And that's a key part of what the prefrontal cortex is doing.

There's another theory about this as well. It's called the geon model. It's not as popular. But basically, what's a Gion? A Gion is just a geometric shape like a cylinder or a square. And the idea is that your templates are just these things pieced together, right? So if you think of a house, might be a better example. It's basically a rectangle with another rectangle that's got a triangle side on the top of it, possibly with a chimney and possibly with a square for a door and some windows. And that would be a geon model. So it could be the template model, which is the average of all the images, or it could be the

geon model. No one's particularly sure. But what is cool about this is research has shown that this is what happens. There are direct pathways from areas V2 and V4. This is very early in your visual identification system. And that image then, like I've said, is sent up to the ventrolateral prefrontal cortex. And that's where you try to match templates or geons, if the case may be. And this sets up an expectation. Then the orbital prefrontal cortex is basically sitting there going, I think this is a dog. And this sets up an expectation. Then the orbital prefrontal cortex is basically sitting there going, I think this is a dog. And it's going to send that expectation back to the lower level visual system to help with identification. This is also how you get rapid emotional responses

because the template is also sent to the amygdala and that triggers your very rapid emotional response to something where if the template is something that might be scary or something you don't want to see, you get this emotional response very quickly. So it's this complex process where the back of the brain is working with the front of the brain with this top-down processing idea. And what's cool about it is some of the research that's been done with this basically shows that the activation in the orbital frontal cortex, that's right at the very tip in the front of the brain, that activity actually predicts whether someone can identify an object or not. If the correlation between the

activity and the object is low, then the person isn't able to identify the object. But if the activity in the orbital frontal cortex is highly correlated with the process, then the person isn't able to identify the object. But if the activity in the orbital frontal cortex is highly correlated with the process, then the person is able to identify the object successfully. I can give you another example of top-down processing, which is imagine you saw half a dog. Now, what do I mean by half a dog? It's not being cut in half, but imagine you see a dog that's behind a doorway and you can only see half of the dog's body. Well, another example of top down processing is your visual representation there is of half a dog because that's what you see. But your top down system is able to guess that there is another half to the dog that you simply can't see at that point in time.

So your top-down system is also making guesses about the reality of the world. And you could imagine trying to fake this system out, right? Where there is only half a dog. I don't know how that would work. But your top-down system is making these guesses that help you understand how the world is going to work. And it's doing this all the time. And I'll give you some more examples when I get to the decision-making bit on top-down processing. But what we do is we call this filling in. Because you actually have a blind spot in your eye.

Now, the reason you can't actually see the blind spot in the sense that you don't have the circle in your vision is you have two eyes. And when something is in the blind spot of the left eye, the right eye fills in and says, well, this is what's actually in that space and vice versa. But there's any number of tricks where this can happen, right? There's a couple of illusions out there that rely on the filling in process to make you see something that isn't there. One of the classic ones, if you go to Google Images, it's just a black line with a circle in the middle where the line is missing. But if you actually move this image around, you'll see the complete line and you won't see the missing piece because your brain is filling it in, if you can visualize that. There's some other weird offshoots to this as well. There's something

called Charles Bonnet syndrome, but I'm actually going to come back to that on a neuroscience bite, but it's got to do with failures of top-down processing. Now I mentioned that top-down processing plays a role in a bunch of different cognitive systems. Well, in decision making, we use top-down processing all the time. And where that comes in is, let's say we're trying to decide between ordering from pizza place A and pizza place B. And in previous episodes, I've talked a lot about values and the explore-exploit dilemma and things like this. But a part that I didn't really cover is that your brain has these expectations or predictions or beliefs. Your brain can search your working memory system and say,

okay, well, what have I heard about this thing? Or it can use information, like you might read a review in a paper of a restaurant, or it might be something that a friend tells you. And what's happening there is it's a form of top-down or it might be something that a friend tells you. And what's happening there is it's a form of top-down processing. Your brain is making guesses on what the quality of the pizza will be based on the information at hand. And this is your prefrontal cortex. It's doing all this math for you. And this is considerably different than say animals that don't have a large prefrontal cortex where they literally just work in a world where everything is bottom up. What they see is what they believe to be true. The information they have is just the information that's physically there in front of them. And they

can't make these top-down judgments. Another example where we do this all the time, if you think of driving as an example or riding a bike, if you approach an intersection, your top-down system is the system in the brain that's basically saying, okay, what do I believe that car is going to do based on the information at hand? Is it going to stop at the stop sign? Is it going to roll through it? And your top-down system is making these guesses to allow you to function in the world, all right? And without those guesses, it's incredibly hard to make decisions. And this top-down idea extends into other things like learning processes where our learning can be biased by our working memory system.

Again, just like decision-making, because our learning systems are using top-down information as opposed to just trial and error learning or repetition learning. And it even extends to the motor system. You know, when you go to generate a movement pattern, there is a buildup of information where you piece it together, but the motor system is going to make this top-down evaluation and say, listen, this is what I think I need to do. If you come up with a similar motor program, then guess what? That's what you end up doing because the buildup of your movement aligns with this top-down guess. So the prefrontal cortex, massive role is top-down processing. It's constantly making

guesses about the world, whether it's the reality of the perceptual world that you're looking at, or it's doing it to support decision-making or learning or your motor movements. All right. I hope you've found that interesting. Don't forget the website, thatneuroscienceguy.com. We've got a couple more episodes and we're going to take a break for a bit and we'll come back in the summer, but we want to know what you want to know about the neuroscience of daily life. So you can find links to our email address, thatneuroscienceguyatgmail.com, or on threads or X, you can just DM me and say, hey, here's an idea. I want to know more about this because a lot of the ideas we're getting now are coming from you. When we started this, it was all us and what we thought we should talk about. Now I'd say 80% of our episodes are generated by ideas from you, the listener. So please send them in. And of course,

thank you so much for listening and please subscribe. My name is Olof Kregolsen and I'm that neuroscience guy. I'll see you soon for another neuroscience bite.