That Neuroscience Guy - Vision Part 2 - The Ventral Stream

Episode Date: May 9, 2021

Sight is one of our most detailed sensations. To finalize our two-part series, I'm going to discuss the basics of how we process visual information for the perception of the world around us. 

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. On the last episode, we talked about the dorsal visual stream, the processing of visual information by the posterior parietal cortex to build the world around us and place everything in visual space. But as I mentioned, the dorsal visual stream doesn't provide meaning. It just provides location. So on today's episode, we're going to continue talking about the neuroscience of sight and specifically the ventral visual stream. So, on today's episode, we're going to continue talking about the neuroscience of sight, and

specifically, the ventral visual stream. If you recall from the last episode, visual information in the form of photons of light enters through the eyes and hits receptors on the back of the eye called rods and cones. Information then travels through midbrain regions, like the superior colliculus, to the primary visual cortex at the back of the head. In the primary visual cortex and other early visual areas, there's a buildup of information. So, if you remember in V1, for instance, we're detecting lines and edges. In V2, we're adding a bit more complexity and beginning to create the 3D world.

And by the time we get to areas V3 and V4, we're identifying shapes and colors. And if you also remember, as you leave these early visual areas, there's a split in the flow of the information. Some of the information flows up to the posterior parietal cortex in what's called the dorsal visual stream, or vision for action, the part of the visual process that builds the world around us. But the visual information also flows down to the inferior temporal cortex, or what's called the ventral visual stream, and this is vision for perception, or identification. As with the early visual regions, again, there's a buildup of information.

As with the early visual regions, again, there's a buildup of information. Studies using functional magnetic resonance imaging have found parts of the ventral visual stream that are sensitive to the processing of faces, to tools, even to houses. Yes, within your brain, there is a house processing area. For instance, in terms of a house, imagine a rectangle with a triangle on top of it. It's not much of a house, but that buildup of information is identified as a house. Obviously, it's a little bit more complex than that, but you get the idea. In face processing regions, it's looking for something that looks roughly circular with a couple of circles that are the eyes and something that looks like a mouth. So there's this continual buildup of information, the ventral visual stream, where more and more features are added as the image gets more and more complex.

This is what we call bottom-up processing. And by the time this buildup of information reaches the end of the inferior temporal cortex, in principle, you have object identification. But it's a little bit more complex than that, of course. Visual processing isn't just bottom-up in nature, this build-up until we reach identification. It's also top-down. What neuroscientists mean by that is that while there's this build-up of information through the ventral visual stream, a copy of the early stage of processing is sent to frontal parts of the brain. And what's happening there is your working memory system is sorting through what that image is.

Imagine you see a dog. An early visual representation of that dog is sent to the frontal parts of the brain, and they're sorting, basically going through memory, trying to figure out what is this thing? What does this match within my memory banks? And it might come up with the notion of dog. So this is top-down because the frontal part of the brain is then going to suggest this idea, if you will, to the ventral stream. I think this is a dog. And concurrently, there's a buildup of information, the bottom-up process. If it suggests dog as well, you've got identification. The bottom-up buildup of information aligns with the top-down processing

or suggestion of what object is in front of you. So, a buildup of information that's bottom-up in nature, gradually adding features until identification occurs, which is confirmed by a top-down process that's taking a copy of the early visual representation and trying to match it with templates and memory. Let's look at some examples. Perhaps the most series of famous experiments in neuroscience involving visual processing come from Mel Goodale and Brendan Milner. They found a patient who goes by the initials DF who had damage to their ventral visual stream. And DF is really interesting. Just like they did with RV in the last episode,

you can imagine showing DF some shapes, for instance, a circle or a triangle, you can imagine showing DF some shapes, for instance, a circle or a triangle, and DF cannot identify them. If you hold up a circle in front of them, they don't know what it is and just effectively guess randomly. If you show DF a circle, she doesn't know what it is and just guesses randomly. And the same is true if you show her a triangle. But what's really interesting is if you get DF to reach out for that thing that's in front of them, they actually grasp it perfectly. As opposed to thinking of circles and triangles, think of a glass. If you put a glass in front of DF, she wouldn't know what it was. But if you got her to reach out and try to pick it up,

her grip would actually form to perfectly scale to the glass and she would grasp it correctly. Dr. Goodale and Dr. Milner highlighted this in what's now called the posting task. The posting task is based on putting a piece of mail into a mail slot. And what they would do is give DF a piece of mail and there would be a mail slot in front of them. Their first step in the experiment was to try to align the piece of mail and there would be a mail slot in front of them. Their first step in the experiment was to try to align the piece of mail with the mail slot. DF wasn't able to do this because this is a perceptual judgment. So aligning the mail with the mail slot is perceptual. And that involves the ventral visual stream. And DF wasn't able to do this. But when they said post the male, as DF began to reach

for the male slot with the male, her arm would almost magically rotate until the male was aligned with the slot. This is the dorsal visual stream kicking in because now it's all about action. Kind of cool. In another study, they put DF in an fMRI scanner to look at brain activity. In another study, they put DF in an fMRI scanner to look at brain activity. And what was very interesting is that when they showed DF a picture of an object versus static random noise, the visual response was the same. In other words, DF just wasn't seeing it. This is very different from URI. If you put us in an fMRI scanner and you showed us a picture of an object, you would see a lot of visual activity in the ventral visual stream as we try to figure out what the object is. What DF has is what's called visual formagnosia, an

inability to see objects and identify them. But remember at the same time she can still see. Her dorsal visual stream is intact as evidenced by the posting task. Now visual formagnosia is a broad condition in which you have a complete inability to identify and see objects. But there's specialized forms of this. There's a gentleman named Philip in the United Kingdom who has what's called prosopagnosia, face blindness. Prosopagnosia is a very specialized form of visual formagnosia where the only thing that the person can't see is the face of a person. In Philip's case, it's a little bit more than strict procybognosia. Philip also has trouble with all animate objects.

For instance, Philip can't really tell the difference between the face of a cartoon character or that of a real person. And if you take Philip to the zoo, he can't identify animals. He tries to make some guesses because he can count the number of legs and get a feeling, get a general sense of what the animal size is, but he's effectively guessing. And he's very bad with faces. He cannot identify specific faces. But he's really good with inanimate objects. If you hold a fork in front of Philip, he can tell you it's a fork, and if you hold up a spoon, he knows it's a spoon. But if you show him a picture of a cat, he literally doesn't know what it is. Again, Philip's not blind. He can see. In fact, Philip has a driver's license. He just has this deficit in being able to see animate objects.

True prosopagnosia, which Philip partially has, is just face blindness. The only thing you can't see is the face of the person. Again, these conditions are hard for us to visualize if we don't have them, but it's almost like they're not paying attention to the face. That part of the visual world is just missing to them. In a lot of cases, they can see the hairstyle of the person, they can see their clothes, they just can't see the face. Hard to imagine, but it's a real condition. Another thing to think about when we think about visual processing is in the initial stages of what's now called dual process theory or the action perception model of vision. The original proposal was that the dorsal stream and the ventral stream were relatively independent of each other.

Visual processing of the dorsal stream was only to do with action and building up the visual world around us. And visual processing in the ventral stream was all about perception and identification. But in recent years, it's become pretty clear that there's crosstalk. These two systems talk to each other. The ventral visual stream does support vision for action, and the dorsal stream does help with identification. I'll give you an example with a research study from when I was a master's student. There's a famous visual illusion called the Tishner circles. You might have seen it and if you haven't seen it, you can google it. It appears pretty quickly in google images. In the case of the Tishner circles, there's two layouts. In one layout, there's a series of large

circles that surround a small circle in the middle. And in the other layout, there's two layouts. In one layout there's a series of large circles that surround a small circle in the middle and in the other layout there's a large circle in the middle surrounded by a series of small circles. And when people evaluate the size of these objects when a small circle is surrounded by large circles people generally tend to think that that small circle is smaller than it actually is. And when a large circle is surrounded by small circles people generally tend to think that the large circle is larger than it actually is. Now what you can do is you can actually have people reach out and grab these objects. So imagine you're seeing a small circle surrounded by large circles.

If we got you to make a perceptual judgment for instance, the instruction was hold up your fingers and position them so that if you reached right now, you could grab the circle in the middle, your fingers would actually scale to the illusory object size. What I mean is that they'll be scaled a little bit smaller than the actual size of the middle circle because your ventral visual stream is tricked by the illusion. However, something interesting happens when you reach for that middle circle. If we now say reach out and grab the middle circle, your fingers will actually open up a little bit as the dorsal stream kicks in and says, no, hang on ventral stream. That

circle's not that size. It's actually a bit bigger. And you can see this in real time with people reaching out to grab these things. So, let's just do a quick review of the human visual stream. Visual information comes in through the eyes to a receptive field on the back of each eye. It's then captured by the rods and cones on the back of the eye that fire, encoding where the photons of light struck on the visual field. That information goes through some midbrain nuclei and then is transmitted to the back of the brain in the primary visual cortex. There's a buildup of visual information through the primary visual cortex

where the image gets more and more complicated, and there's a split. Information is sent up to the posterior parietal cortex, or what's called the dorsal visual stream, where the visual world around us is built up subconsciously. Your brain identifies where everything is in space. And at the same time, information is sent down to the inferior temporal cortex, or the ventral visual stream, vision for perception or identification. This is where we add meaning to the visual picture and we identify objects and people. This is where we add meaning to the visual picture and we identify objects and people. I hope you've learned something about visual processing in this quick summary of the dorsal and ventral visual stream. My name is Olof Kregolsen and I'm that neuroscience guy. You can check out my website at www.olofkregolsen.com or follow me on Twitter at that neuroscience guy.

Thanks for listening. I'll see you on the next episode.