Transcript of second of the three part "Charlie Rose Brain Series," titled "Visual Perception," broadcast 11/24/2009, and streamed at http://www.charlierose.com CHARLIE ROSE: This series is a journey to the most exciting frontier of science, the brain, everybody we`re learning more of how this miraculous organ works. Last month was our introduction. We showed you the anatomy of the brain and the frontiers where brain science was taking us. Tonight we look at one of the most instructive areas, visual perception. One quarter of our brain activity is about visual perception, so the brain is devoting about 25 percent of its power to vision. Think about it -- what the visual system tells us is at the core of our impressions of the world, what is real and what is beautiful, what is permanent and what is changing. Vision is one of our five special sense, including hearing, smell, tough, and taste. We focus on vision because so much of our world is visual. Also, we understand more about vision than the other senses. Most of us think what we see is only determined by our eyes. But that is only the beginning. The extraordinary happens in the brain as it takes the inverted image from the retina and passes it to the cortex where the brain creates a perception of the seen world. It is in the brain where seeing happens. Patterns of light and color and translated into object and events, things that are meaningful for us and our survival. So the question is how does the brain produce our perception of what the eye has captured? That`s the difficult and the exciting part. How does this computation work? How does the brain organize the patterns of neural information into the visual perception we have? How does it make inferences about the world around us? What clues does it use? And how does it compensation for changes in our visual environment? What assumptions does it make? How much of its activity is passive and inferential, and how much of it is the active collective information? We know from the study of the development of that visual experience in the very first years of life when neurons and their connections are being forged is crucial for normal vision. We`re also learning what happens if during these early years one is deprived of visual experience. There is now encouraging news that the visual system is surprisingly plastic and able to recover from injury and to learn anew. We will ask what we know who about blindness in various forms and we will learn more of the relationship between vision and other sensory processes. This evening we`re joined by a remarkable group of scientists who are devoting their life to understanding how the brain creates visual perceptions. They are Tony Movshon. He`s interested in the way the brain encodes and decodes visual information and the mechanisms that use the information to ensure the control of behavior. He analyzes neurons in the visual areas of monkeys and is on the faculty of New York University. He has a doctorate in neurophysiology and psychophysics from Cambridge Ted Adelson is professor of vision science in the Department of Brain and Cognitive Sciences at MIT where he focuses on topics in human and machine vision motion analysis. He is an expert in many things, including the gestalt theory of perception. He has a Ph.D. in experimental psychology from the University of Michigan. Nancy Kanwisher is professor at the McGovern Institute at MIT. Her lab has identified several regions of the brain that play specialized roles visual perception, especially in face recognition. Pawan Sinha is associate professor of neuroscience at MIT, where he leads the Sinha Lab. He has done amazing work in India as a part of Project Prakash where he works with children who suffered injury or disease to the eye. And in answering these questions, once again, my co-host, my guide, my professor, is Dr. Eric Kandel. He is, as you know from our program, a noted brain scientist. He is a Nobel laureate. He`s affiliated with Columbia, the Howard Hughes Institute. He has written a remarkable book about the search for memory, and he is for this program a great friend. So I am pleased to have his wisdom and his direction guiding as we try to learn more in this multipart series about the brain. Welcome again. ERIC KANDEL: Thank you, Charlie, always a pleasure to be here. CHARLIE ROSE: So we looked at the introduction in our first episode. Tonight you`ve chosen visual perception. Why do we start there? ERIC KANDEL: We chose this together, Charlie. We chose it because we understand the visual system better than any other system of the brain and because it is a model for how the brain works. If we understand how vision works in the brain, we have a very good understanding of how the other sensory systems work. CHARLIE ROSE: You have suggested four ways we ought to look at this, four themes, the eye is not a camera. Sensory functions are localized, visual computations are hierarchical, and plasticity in the brain is pervasive and crucial. Take me quickly and summarize what we mean by the eye is not a camera. ERIC KANDEL: You`ve made some of these points. The eye takes incomplete information from the external world. It is not absolutely faithful to the external reality. It takes things that are important, it throws other things away, it emphasizes certain parts of the image, discards others. The brain is the creative organ that makes sense out of this, that gives you a feeling of three dimensional space, of faces looking at you, and landscapes. That is completely a reconstruction done by the brain. The brain does this in an orderly fashion. Each sensory system, the visual system is a perfect example, has a specific spot in the brain. So there are a series of relays that are localized in everybody`s brain in exactly the same way, and the visual system has its own space, the touch system has its own space, the smell system has its own space, every sensory system has its specific location. And it is organized in a hierarchical fashion. There are a series of relays that operate on the information in progressively more complex fashions. So we`ll see the early relays, the retina and in the thalamus, see whether or not light is present or absent. So they respond to small spots of light. As we get out to the cortex the cells in the cortex begin to respond to edges, to bars, whether they`re horizontal or vertical. Later on we get to more complex images, and ultimately, as we`ll hear from Nancy Kanwisher, we`ll actually see how the brain responds selectively to faces or to landscapes. So this is a progressive processing operation that the visual system performs. CHARLIE ROSE: Tell me about illusions. ERIC KANDEL: The brain makes guesses. As a result, it can be deceived, and we can sometimes be presented with a two dimensional form that it sees as a three dimensional form and vice versa. And one can actually see the creativity of the brain at work, because in solving an illusion, you know, figuring out that there is a dog here or there`s a face here, there`s almost a high experience, like a sense of creativity. In a very primitive sense we see in the visual system of everyone you, me, and the person on the street, the creative process that probably is specialized in genuinely creative people in do great science, that do great art. So we see individual systems, the elementary forms of this creative process, it`s built into the brain. CHARLIE ROSE: We begin this conversation talking about how exciting this was. You`ll see that now and we go to the conversation that Eric and I had with a very, very distinguished panel. Here it is. Let`s begin by looking at the anatomy. Our anatomical expert is with us again. Welcome back. TONY MOVSHON, NEW YORK UNIVERSITY: My job is to tell you where things are. CHARLIE ROSE: Exactly right. So we have pictures. TONY MOVSHON: We have pictures, and having started with the idea that the eye is not a camera and that`s not how the visual system works, we will start by showing you that the eye, in fact, does form images like a camera does. So this is a cross section of the eye taken as if through this plane, in other words, if somebody has taken my eye and divided it in half along the mid-line of my face. The front of the eye is here and what you see an image being formed of a point in the world. The light rays are focused just as any other system focus it is rays onto the light sensitive part of the eye, the retina, which lines the whole back of the eyeball. The optical apparatus at the front of the eye works like other optics we understand. So there`s a cornea which provides most of the power for forming the image, and the lens which is the adjustable part of the system that lets us focus on near and far objects. Now, what we`re mostly concerned about here, though, is what happens once the information reaches the retina and forms an image on the retina. And at that point, the photoreceptive cells that line the back of the eye collect the information from the image and funnel it through this single, rather narrow channel, the on optic nerve. Now, the eye is doing something very dramatic right away in terms of neural function and neural circuitry. The retina is a part of the brain. The fact that it happens to live in the eye is a matter of anatomical convenience. But the retina contains the kind of circuits we find in the central nervous system. And those circuits perform some very elaborate computations, most of which is we won`t be talking about in detail today. But one very striking thing that the retina does is something we`re all familiar with in terms of, for example, how a digital camera works or how a video camera works. We take a lot of information and compress it down so it will fit conveniently on to our tapes or on to our hard drives. The retina has about a hundred million photoreceptor cells which sense the intensity and color of the light at each point of the image. And the information from those receptors is compressed down so it comes out of the eye along the optic nerve in about a million fibers. So there`s about 100 to one compression of the information performed by this circuit in the eye so that what transfers into the brain can be, arguably for physical reasons, carried along this optic nerve. One of the things about the eye that is very striking, a striking aspect of its design, is that it`s a visual resolution, its acuity is only very good at the very center of gaze exactly where you`re looking. And you can become aware of this yourself. If you look carefully at Eric Kandel`s face, as I am, and try to work out what color shirt Pawan Sinha on my left is wearing, you can`t do it. The reason is visual acuity falls off as you move away, and that`s because your retina is specialized for detail in the middle and its sensitivity falls off to the periphery, and as a result you have to move your eye around. So the way you make up your picture of the world is by moving your eye from place to place and capturing multiple snapshots and assembling them into a percept. Now it has to be said that all of us here work on the central nervous system, and most of what we`ll talk about today has to do with what the central nervous system does. And so here, which what we`re doing is looking at a view of the brain from the back, this same brain that we have in the middle of the table as a sort of colorful representation is here rendered. And what you can see here is a cone of light rays, light rays forming an image, entering the eye, which is visible in shadow on the left-hand side here. The optic nerve passes here, as you can see, into a relay nucleus in the core of the forebrain called the thalamus. This nucleus is called the lateral body, but it`s a nucleus. The thalamus is a part of core of the brain that actually connects specifically to all parts of the cerebral cortex, that sheet of cells that we`ve discussed before that do most of the interesting computations the brain does. Signals from the visual thalamus pass to the visual areas of the cerebral cortex which are marked here, some of them, in different colored zones. ERIC KANDEL: It also shows very beautifully, as Tony outlined, is this characteristic of other sensory systems. There`s a hierarchy, there`s a series of relays that process information in a progressively more complex fashion, and that`s why this is such an instructive system to study. CHARLIE ROSE: Ted, what is the most amazing part of this for you? TED ADELSON, MIT: Well, the most amazing thing is the fact that the visual system is so successful in putting this information together. We know from our research in machine vision where we try to make robots replicate the functions of human vision, we know this is much more difficult than it seems because it turns out to be extremely difficult to get the computers and the robots to do anything like what human vision does. So the problem is in order to put the information together, you get all kinds of little bits of information which you need to put together in into a coherent whole. And all of the bits, each of those bits, each of the little edges or lines or whatever you get, is completely interpretable by itself. So you have to figure out how to bring it all together, and that turns out to be very a very difficult thing to do. CHARLIE ROSE: What do we know about how the brain makes these inferences about these signals that it`s getting? TED ADELSON: Well, we know that there is this hierarchy that these multiple levels, each level of analysis putting together information that it`s getting from the previous level and combining that with information that we have stored, information based on our experience in the past with other kinds of similar images and objects. We don`t know in detail how that happens, but we know that that`s the critical thing. ERIC KANDEL: We also have some very good physiological experiments that indicate how the processing steps occur. And we have very good insight of what happens at the earliest stages and even at the later stages. And they give you clues as to how the more complete image is put together. TONY MOVSHON: Let me show you some examples. I`m actually going to show you some 50-year-old home movies. ERIC KANDEL: A classic! A classic! TONY MOVSHON: ... who revolutionized this field and who won the Nobel prize in 1981 for their work on this field. And the work they did was primarily based on a system that`s shown in this diagram. The brain is the same brain that we`ve become familiar with. The eye, I hope, is now also the same eye that we`re familiar with. And this simply shows the experimental setup that they used to make recordings from brain cells in animals while they were actually viewing targets on a screen. And so you`re going to see three video clips. And in each case what you`re going to do is be looking at the screen that`s shown over here as if from the animals` eyes point of view. You`re going see the images projected on the screen. You have to bear in mind that this is home movie quality, so this is a 50-year-old home movie. It does not look like it was made in the studio. These images will move around. The information from the retina passes into the brain along the pathways we`ve described, and it causes the activity of nerve cells in the cerebral cortex or the thalamus to change. Nerve cells communicate with one another, as we discussed in the first program of the series, by firing these trains of impulses. This is the nature of the signal that cells use to communicate over long distances. ERIC KANDEL: Morse code. TONY MOVSHON: It`s a Morse code. It`s a digital code. Because of the constraints of building a brain, the brain can`t use the kind of code that a computer would use. It has to use a code which involves the transmission of information along these long axons which form the nerves, for example, the optic nerve. And the transmission of that information is by these impulses, and these impulses are what we hear when we make these brain recordings. So in these videos, what you`re going to see is the picture will be the image on the screen and the sound will be the recording coming from the brain cell. Now, in the first image, in the first film, what we`re going to see is a set of stimuli that are causing changes in the activity of a cell recorded from primary visual cortex, and the first segment of the film is simply going to show the map of what we call the receptive field, which is simply the region of the retina within which visual stimuli can influence the firing of the cell. Every time you hear a burst of static, you`re hearing the brain`s activity. ERIC KANDEL: Notice how specific the response is. TONY MOVSHON: It responds only in this particular region. The point of the next segment of the film is to show that not only is the firing of the cell specific to where on the retina the image falls, but this particular cell will only be activated when the target has the right orientation. The cell will respond to a line, a line moving in one direction and of a particular contour orientation. Very little activity all through here. And all of a sudden... ERIC KANDEL: Beautiful. TONY MOVSHON: Very powerful striking activity. CHARLIE ROSE: This is one? TONY MOVSHON: One nerve cell? CHARLIE ROSE: Wow. ERIC KANDEL: This is really the conversation that that neuron is carrying out with its makers. TONY MOVSHON: So if you reflect back for a moment to the question of whether the eye is a camera, a digital camera captures the light and color of one pixel of an image. Clearly this cell is doing something completely different from that, because it is telling you not about the light or color of the pixel, but it`s telling you if a line of a particular orientation moving in a particular direction is present in that place in the visual field. So it`s giving you a very specific piece of information about a component of the image. Now, you might ask whether -- where the cell gets that information from. And it turns out that we know from a number of experiments, including the ones that Hubel and Wiesel did, that the transformation of the information takes place between the thalamus, this relay in the middle of the brain, and the cerebral cortex. So if we show images of the kind I`ve just shown you to a neuron recorded in the thalamus -- that`s what the next video clip will show -- that neuron will have very little specificity. And in fact, if you look at this video, what you`ll be seeing basically is a cell who`s firing -- again, you`ll hear it in the audio track -- is really vigorous whenever there`s light in the middle of the screen. This cell just cares about light. ERIC KANDEL: I just want to elaborate a little on what Tony said because it`s so important. This transformation that you see from the cells in the thalamus to the cortex, primary visual cortex, was a spectacular find because it made one realize that the cortex does a lot more than the lateral geniculate. So the lateral geniculate, as Tony showed, responds to the small spots of light. In the cortex it responds to bars, to edges. So we begin to see edge detection, and different cells will respond to edges of different orientation, vertical horizontal, or a V. So there`s a tremendous specificity, and the whole transformation from having circular receptive fields to linear is a major operation that the brain performs. You can imagine at later stages you can put lines together to have corners, ultimately to have faces. So this is the beginning of how the brain reconstructs a visual image. CHARLIE ROSE: Let me turn to Nancy. Tell me how you have begun to understand and focus on the localization of function. NANCY KANWISHER, MIT: Well, our work builds on lots of prior work using behavioral measures. So we`re interested in face perception, and one of the reasons we chose to work on face perception is that there were lots of reasons to think that the brain would have special machinery for processing faces. So if we show that movie you`ll see in the display that when a face stimulus is presented upside down, you can`t tell what it is. And this is characteristic of faces in particular. You don`t see it for other kinds of stimuli. It`s perfectly easy to recognize a chair or a dog or a tree if you see it upside down, whereas face recognition is severely impaired when you see a face upside down. So that`s been known for decades. And because of that work and other work like it, there was reason to think that the brain has special mechanisms that it uses when it recognizes faces. And so the idea was that at the higher level stages where from where Tony has been talking about, at the higher stages of the visual system there might be special machinery in there for face recognition. So we`ve looked at that with brain imaging methods, where you pop a subject in a scanner and show them faces and you can see that part of the brain turn on in brain imaging methods when a subject looks at faces. So right now my face area, which is right there in my brain, is active because I`m looking at your face, and now it`s off, and now it`s on, and now it`s off. And I know that because I`ve scanned myself and hundreds of other people looking at faces and objects and you can just see it turn on and off. CHARLIE ROSE: The fact that face recognition has a localized function is it because it`s more difficult, more, what? NANCY KANWISHER: That`s a great question. I`d love to know which mental functions get their own private piece of real estate in the brain and which don`t. So in my lab we found a few. We found regions that are specialized for face recognition right here. I`ll show you on the brain. I can show you where it is. That`s the front of the brain. It`s looking at you. And if we turn it upside down so you can see the bottom of the temporal lobe right here, so on me that`s right there, that face area is right about here in the brain. And we`ve also found other specialized regions. So we expected to find that face region because of the work that Eric just mentioned on people with brain damage. When they had damage in this general region, they tended to have this difficulty recognizing faces. So that wasn`t too much of a surprise, although it`s really fun to discover. But since then, we`ve found several other regions that we didn`t expect in advance at all. So one is a region that responds when you look at places and landscapes, and it`s right next door, a little further forward, right about there in the brain on both sides. CHARLIE ROSE: Let me just ask you this question. If, in fact, I could reach inside of somebody`s brain and remove the part of the brain which recognizes landscapes, and then that person would look at a landscape, what would they see? NANCY KANWISHER: Well, we`ve tested such a person. Years ago when we first discovered this place area here, I was dying to know what life would be like for somebody who didn`t have that region, but it`s actually on both sides of the brain, and so the chance of getting brain damage on both sides is very unlikely. But I was at a conference once, and I was walking through these rows of posters, dozens, hundreds of posters, and I saw a picture of a brain of somebody with brain damage right there on both sides. And I talked to the person whose poster it was, and I said "Who is that person? I need to test them." So I tested this guy who had been an artist who painted these beautiful paintings, very tragic. And he no longer painted, no longer took joy in looking at things. Interestingly, he could see where he was going, he could recognize faces, he can read, he can get around in the world. But as he told us, he never knows where he is. He has no idea where he is in the world. CHARLIE ROSE: He can see it, but he can`t put in the any context. NANCY KANWISHER: He can see it but he doesn`t know where it is. ERIC KANDEL: This is also true for the face area. CHARLIE ROSE: What happens in the face, if that`s removed? NANCY KANWISHER: So if you lose the ace area selectively, just that region, the typical finding is that people can recognize all kinds of other things. They can recognize landscapes and words and objects, but they can`t recognize faces, even their own face. They know it`s a face, so they can see a face. CHARLIE ROSE: They don`t associate it with any other -- connect it to a face they`ve seen before? NANCY KANWISHER: That`s right. Even more astonishingly, there`s one or two patients who have the opposite situation. So they have more diffuse damage and they can`t recognize objects, but one guy in particular is very impaired, he can`t read at all, can`t recognize objects, he`s functionally blind except that he`s 100 percent normal at face recognition. And that tells you even more strongly than the loss of face recognition that you can have selective preservation. It`s not a special fancy system that builds on top of the object recognition system. It`s a really very separate processing halfway. ERIC KANDEL: I think this shows two things very beautifully that principles that govern how the brain functions. One, it`s a hierarchy. You don`t see this at the earlier stages of the visual system. They don`t respond to faces. So this is a really very complicated perceptual function, sort of the end stage of the visual system, number one. And number two, this tremendous localization of function. So the brain is largely organized because different kinds of representations -- touch, pain, vision, hearing, smell, taste, localized to different regions of the brain. So this is a key, two key organizing principles -- localization functioning and hierarchy. Now, how do you set this up, one of the clues whereby the brain is able to put objects together? This is things that Ted and Pawan has been studying and they can tell us about how you get to the point where you recognize faces. CHARLIE ROSE: Pawan, go ahead first, and then come back to Ted. PAWAN SINHA, MIT: It`s a whole big question, and we still are at the starting box in our understanding of how this process unfolds. But in some of the work that we`re doing, we are trying to approach the cushion by adopting a somewhat unique approach of working with children who have been blind for several years but they have treatable blindness. So if you have, say, a ten-year-old who has never seen form until that point, and you`re able to surgically initiate vision in such a child, then you have a remarkable opportunity of then following the progress of this child and studying what kinds of mistakes do they start out making. CHARLIE ROSE: So what do you discover? PAWAN SINHA: So what we discover is that the initial stages of vision in time are fairly disordered. The world to such a child -- in fact, let me show you some images. So here`s a child who has a cataract, a dense cataract which would be kind of like if you were to be wearing ping-pong balls that are cut in half, one on each eye, that`s the level of vision that such a child would have. After such a child gets vision by removing the cataract and implanting a clear lens, their world is very unlike -- or their percept is very unlike our percept. So if you show to such a child an object that to us is instantly recognizable and you ask them "Do you recognize this object?" they do not. And then if you ask them, well, even if you can`t name the object, just point to where the objects are, they point to every little region of the different color or illuminants. So the world to them is this patchwork of different colors, different brightness. CHARLIE ROSE: And intensity and color? PAWAN SINHA: Yes. Yes. So the world is greatly broken up into many different regions. In fact, as I think Tony will mention about edges, edges are a key constant in our visual world, and we are very easily -- we as mature visual observers, we are easily able to tell that this edge is due to shadow, this edge is due to a depth of continuity. But to a child who`s just starting to see edges, all edges seem to have equal salience. So for such a child, even the shadow on this ball becomes an important edge. And the world gets fragmented into all these pieces and they`re unable to glue it together and see coherent objects. TONY MOVSHON: The act of going from light and color on the retina to objects and events in the world is a matter of assigning the edges properly. And actually this is something that Ted has done a lot of work on and has nice examples to show. TED ADELSON: Well, I think this example is a good one. This is a famous old example of the principles of gestalt psychology with which are the principles by which we organize information. And when you first look at this picture, it`s very hard to figure out what it is. In fact, it just looks like a bunch of light and dark splotches, and you may look at it for a while and just see light and dark splotches. And that actually is the information that your eye is getting from this picture and your brain has to figure out what to do with it. Now, I can show you what you should do with it, because we`ve here drawn some lines on top to show you there`s a dog there, it`s a Dalmatian dog, and the Dalmatian dog has these light and dark spots on it. And if we go back to the original picture now... CHARLIE ROSE: Unbelievable. ERIC KANDEL: Isn`t it unbelievable? That`s fantastic. TED ADELSON: Now that you`ve seen what it really is, now you can hold that image together and you can see how to organize it. And now normally in our normal vision everything seems so automatic we don`t realize that this is really what`s going on all the time. Our eye gives us this light and dark information, but it doesn`t come in an organized form. And the problem is light and dark can come from many different sources. It could be a white or dark spot because the fur has got black or white pigment in it, or it could be some shadow being cast causing it to be light or dark, or it could be the edge of the dog where the dog stops and the background starts. So this information that you`re given at the level of the eye is very ambiguous. And so there`s a sort of detective problem, a problem-solving task that the brain has to deal with, which is how you piece all these bits of information, each piece being ambiguous, how do you piece it all together into a single coherent story that tells you about what`s really in the world. CHARLIE ROSE: Suppose with shadows, and we took a tree in the morning, at full midday sun, the afternoon, and then at dusk. The tree would remain the same in our vision even though shadows and the light and intensity would change. TED ADELSON: Yes, that`s one of the amazing things that vision has to do, because the brain is designed pull out the information that`s stable and important and meaningful and to throw away the information that`s sort of accidental. We have another illustration this. This is a -- this is a picture of a cylinder casting a shadow on a checkerboard. And you`ll see there`s two checks. There`s a dark check labeled "a" and a light check labeled "b." And it`s quite obvious when you look at it that one of them is dark and one of them is light. But the fact is that the ink on the page is exactly the same. If you were to be a light meter and you would measure the gray level of the ink for "a" and the ink for "b," it`s exactly same ink. So it seems like a failure, like the visual system is not managing to do this very simple thing, to tell you what color the ink is on the page. But that`s because the visual system`s job is not to tell you about ink on a page. It`s to tell you about what`s out there in the world. TONY MOVSHON: That what`s so striking is that although this process can be written down and described the way Ted describes it, you would have thought you could unpack it. You would have said, OK, now that I know that "a" and "b" actually have the same color on the page I should be able to see them as well as they are the same. But you can`t. It`s absolutely automatic. It`s built into a low level of your visual system to tell you about the checkerboard and not about the shadow. CHARLIE ROSE: When does most of this development in our brain about visual perception take place? PAWAN SINHA: I would say that it`s through our entirety of life. Initially I started out with the worry that a child who has been blind for the first several years would probably have lost the ability to acquire vision. CHARLIE ROSE: The brain would then shut down? PAWAN SINHA: Shut down, right. So there`s a dogmatic view that the first view years, maybe three or four years are the critical periods for learning vision. If that were to be true, then the work that we are doing in the Project Prakash would not be serving a purpose either for the child or for science. But I`m happy to report that even children as old as 14 or 15 -- the oldest we worked with is 29 -- even with individuals as old as that, when you restore sight, you see significant improvements in vision, the acquisition of visual function, which goes to suggest that the programs for learning, the programs for acquiring vision can be initiated even late in life. NANCY KANWISHER: You can also see this in certain brain areas. So each of us learns to read. And when you learn to read, there`s a particular part of the same visual area of the brain that I was talking about before that comes to respond selectively to words and letters presented visually. And it`s right in there. And it`s very small, but you can find it in almost every subject. And since people have only been reading for a few thousand years, humanity hasn`t been reading very long, that piece of brain cannot be the product of natural selection. So it must be that each of us in our lifespan wires up the circuits based on experience to make that region selectively responsive to the orthographies of the language as we know. A colleague of mine has recently shown that that region can develop even if you don`t learn to read until well into adulthood. So he found, first of all, he found that region in Chinese subjects when they look at Chinese characters. That`s sort of expected. But then he found a bunch of Chinese illiterates, and he scanned them and he did not find them that region, and then he taught them to read, and then he scanned them again, and there it was. And some of these people were 40 when they learned to read and that region still developed. So some of these regions are extremely plastic and can develop late in life. CHARLIE ROSE: And usage makes them come alive, so to speak. NANCY KANWISHER: That`s right. TONY MOVSHON: It is what -- the plasticity that people like Pawan and Nancy and others have shown in adult brains is very impressive, and it is unexpected. But it`s worth bearing in mind I think that Pawan`s critical periods, as he mentioned, do exist, and they exist for particular functions. There are some aspects of visual development and other kinds of development that do end and do end at the age of three or four or five. And one example in vision that`s very clear is that if you grow up with your eyes misaligned so they don`t point in the same direction so you have a squint, the parts of your brain that are involved with binocular vision, the perception of depth in space based on comparing the images from the two eyes, break down and they lose the ability to have that function. And once you`ve lost the ability to do binocular vision because of this, if that is not repaired by the age of three or four, then it`s gone for life. So there are some functions for which plasticity is not a remedy. And so one of the interesting things, the challenges that Pawan`s work has presented us, is we thought for a while that pretty much everything was done by the age of five. Now we realize that there`s a whole variety of things, some of which remain plastic for quite long periods in life. ERIC KANDEL: There are two clinical lessons that are really important for this. Even in Pawan`s work, your work shows visual acuity is, in fact, compromised. So obviously the earlier one starts a corrective procedure, the better off one is. The fact that Pawan can rescue an amazing amount of vision at age 12 doesn`t mean it wouldn`t have been better to begin at age six months. And what he has done and why his work has been so revolutionary -- these cataracts are quite common in certain populations in India. This has been a major public health effort that, as he pointed out, is not just for the kids. But imagine having a child who`s blind. You feel that their life is really tossed away. And now you can restore vision, they can recognize faces. It`s a fantastic medical achievement. CHARLIE ROSE: I once read somewhere, this is a little bit off course, but if you take a kitten and put it in the dark from the moment of birth, and then later... TONY MOVSHON: So there`s an extensive literature of visual deprivation. And if you indeed taken an experimental animal and raise it with absolutely no experience of light or vision, its visual system seems to be permanently and profoundly... CHARLIE ROSE: Permanently. TONY MOVSHON: Permanently and profoundly disrupted. There is some slow recovery, but the recovery you get in those animals is much less striking than the recovery Pawan sees in his kids. CHARLIE ROSE: And the difference is what? PAWAN SINHA: There are some important differences. In dark rearing a kitten, you`re depriving the visual system of all input. So there is no light reaching the retina. There`s no stimulation reaching the visual optics. With children that we`re working with, they have cataracts. There`s some light, and maybe there`s even some rudimentary amount of motion. If you wave your hand in front of their eyes, they can tell something is darkening and lightning. ERIC KANDEL: But you may want to show how kids recover this visual capability and how similar it is to the dog illusion that we saw before. PAWAN SINHA: Absolutely. Can I preview a little video of Project Prakash. So if we would roll the video. So Prakash is an effort which starts out by outreach, indentifying children who need treatment. So in this video you see us working in the school for the blind, and we are screening the children to see which children might actually have a treatable condition. So because this particular video comes from a school for the blind, most of the children we encountered there have permanent conditions. There were several chance we found who has light sensitivity, which is an initial encouraging sign that the condition might be treatable. So what you see in the video is the child responding to light, even saying where the light`s coming from. We then bring the children to the hospital for a more thorough ophthalmic exam, and the children who we find are, in fact, treatable we do an ultrasound of the eyes to make sure that the posterior segment of the eye is all fine. Those children are then provided treatment. And we then monitor their progress. So in the video, you see a child who has congenital cataracts in both eyes. So until this point, the child has lived the life of a blind person with very few prospects for prospects for vision later in life. But this child was then given surgery. Clear acrylic lenses were implanted into his eyes, and what you see in the video is him three weeks postoperatively. And he is now responding to visual clues, catching objects. CHARLIE ROSE: A lot of what we`ve learned in the understanding of development of the brain has come from what we`ve discovered from injuries to the brain, correct? PAWAN SINHA: Absolutely. NANCY KANWISHER: Absolutely. That`s a major source of scientific insights. But unfortunately we`re a long way from being able to fix that region right there, if you look at brain damage. ERIC KANDEL: One of the reasons you can tell that is both of you have had experience with computer vision. You might just sort of discuss how difficult it is to even come close to the way the normal visual system -- we can recognize each other`s face with enormous facility. Computer vision has enormous difficulty doing that. Perhaps you`d elaborate on that. TED ADELSON: Well, it certainly is true. And a lot of very smart people have spent a lot of years with very powerful computers. We get better at it year after year but it`s still true that the ability of computers to do any kind of simple recognition is still very primitive. And as computers get more powerful just in terms of their processing speed and their memory getting more powerful, they get better. But it`s clear we`re still missing some fundamental insights about how this needs to happen, because all I can say is the computer vision systems even as they`re getting more advanced, they still fall very far short of what human vision can do. CHARLIE ROSE: What human computation can do? PAWAN SINHA: Just to give you one specific example of that. So a computer vision system for face recognition, the cutting edge computer vision system would require a facial image that was at least 100 x 150 pixels in its resolution. The human visual system can work with images as degraded as this. These are maybe about 12 x 14 pixel images. And we can achieve recognition rates on these images that are superior to a computer vision system working with images that have a hundred times more resolution than this. So just as Ted said, it`s not -- it doesn`t seem like we can get to this level of performance just by making incremental improvements. We need to have a qualitatively different... TONY MOVSHON: I understand the logic of systems, but one of the things that`s key to visual systems is what Ted was talking about, which is the ability to throw away information that`s incidental. So to throw away where the light happens to come from, how the face happens to be posed, where the shadows happen to fall. CHARLIE ROSE: And do we know how that happens? TONY MOVSHON: Well, we know the cylinder on the checkerboard what some of the principles are, but we do not know exactly how that happens in detail. If we did, we would have written a computer program to do it and the computers would be as good as we do. And a classic example is a cube. This is not Necker cube. This is just a cube. And if you have a wireframe cube you don`t normally have any difficulty interpreting it as a cube. But if you have a wireframe cube of which you have made an image on a piece of paper, as, for example, in the case of the images here, there is a bi-stability to the image. You can see that image in two different ways. So if you look at the cube on this side, what most of you will do is see this as a cube with this face near and this face, the one that`s behind, far. The cube on this side which has information encoded in a slightly different way in the image, will represent itself with this face near and the other one far. Now if you look at the cube in the middle, by a simple inspection, either by passive inspection or maybe by an effort of will, you can actually see that cube in either form. CHARLIE ROSE: Absolutely. TONY MOVSHON: You can`t see it both ways. You have to see it one way or the other, right? Now, one of the striking things about this Necker cube is addition to its slipping, is that the one thing you don`t see actually in many ways the simplest thing you could see. So here`s a Necker cube image again. But this Necker cube image as you can see here as I hold it up is, in fact, a flat object, which isn`t a cube at all. It`s in many ways the simplest possible interpretation of the image. It`s just a series of lines on a piece of paper. But this interpretation is the one you never see. And the question of why the visual system constructs this three dimensional representation out of the information is one of the questions that we`ll have to answer if we`re going to answer questions about, for example, how complex objects are represented and recognized. CHARLIE ROSE: One broad question, Eric, is the notion between genetics and environment as it influences everything we talk about here. ERIC KANDEL: Well, the genes determine the basic line diagram of the visual system. If you were to see the genes involved in development of the visual system, you will not have normal functioning vision. But given that so that the basic neural circuit is worked out by genetic and developmental processes, plasticity can occur at every one of those relays, particularly in higher cortical areas, to modify how we use it. So we first of all have evolved to live in a certain world, and the brain of human beings has evolved to live in the world we live in. It`s different than snake, who have a very visual spectrum that is wider than we have. They see a different world than we see. So we have -- our capability of seeing the world is in part determined by this genetic program. But we learn all the time. We learn how to recognize objects and we make those associations the next time we see an object like that. So this involves alterations in the brain, and that continues as long as we live. We continue to encounter new images, new people, and we acquire that information and store it in the brain. So both are involved. NANCY KANWISHER: So a lot of new information just in the last few years about the relative roles of genes and experience in setting up the face system. And we still don`t the whole picture, but it`s getting very tantalizing. So one clue comes from the fact that even babies who are one to three days old have a pretty good ability to distinguish one face from another, even if there`s no hair or external features shown, just the internal part of the face, quite similar faces they can discriminate. And they can do it for up right faces, not inverted faces like adults. So it`s possible that that`s learned in the first one to three days. But it seems more likely that part of that face system may be wired in and waiting for experience to embellish it and fine tune it. CHARLIE ROSE: Let me just go around the table, as I often do, and say what`s the most important thing you want to know? TONY MOVSHON: So my interest actually lies somewhere where the Hubel and Wiesel left us and where Nancy left us. CHARLIE ROSE: They did their work in the early `60s. TONY MOVSHON: They did their work in the `60s through the `80s. And what they did is they basically described the early processing of visual information that brings information to the visual cortex. Now, what Nancy has described is a lot of work that has to do with the highest levels of visual pathway processing information about faces and places and other objects. My own interest and the challenge that I and my colleagues would like to solve has to do with how the information from this single representation and primary visual cortex gets channeled through this whole set of visual areas which there are at least 30 and maybe more until it finally reaches these high level representations where things like faces and places get processed. There is a great deal of what we often call mid-level vision which has a representation in the cortex in many different places and areas. And so the challenge I think that we face is basically to bridge what the Hubel and Wiesel told us and what Nancy tells us to find out how we get the whole process worked out from begin to end. CHARLIE ROSE: Ted, what would you most like to know? TED ADELSON: I would like to know what the computations are that the visual system does in order to tell the difference between light and shadow, between light paint and dark paint, very simple things. Things that seem trivially simple to us but which apparently involve very sophisticated computations. CHARLIE ROSE: Is that a mathematical formula or something? TED ADELSON: Yes, well, someday we`ll figure out what it is. But the way you take all these numbers that the eye is given you, light and dark and color, how do you shuffle those numbers, recombine them into something sensible that tells you about what`s in the world? That`s the problem of, a theoretical problem of human vision and the practical problem of computer vision. CHARLIE ROSE: Pawan? PAWAN SINHA: I want to understand how -- not just how the mature visual system works, but how it gets there. What`s the process of learning? So starting with the seed, what the principles, the learning, what`s the scaffolding that supports the later flowering of all these visual skills that we have? NANCY KANWISHER: I want to know why we have these special regions for faces and places and bodies and some others I didn`t mention, one for thinking about what other people are thinking, possibly regions up here -- oops, over here -- especially involved in language. Why do we have special brain regions for those functions and apparently not other ones? And how do those functions land so systematically in the same place in every normal subject? CHARLIE ROSE: Eric? ERIC KANDEL: I`m interested in sort of two interrelated questions. One is autistic children don`t look at other people directly in the eye. They have a difficult time processing the social interaction. What is going on and to what degree this is -- it reflects aspects of visual perception or to what degree it reflects other aspects of social interaction and how these interconnect is very interesting. Also, visual perception, of course, is so important for the enjoyment of art. And I would think that as we understand more and more why it is that exaggerated images of people have such a powerful affect on us, we`re going to have a better understanding of how we respond to certain works of art. And I can see in the long run dialogues between people sitting around this table and not only art historians but artists each informing each other about how the brain works, because what artists are really doing, they`re doing experiments with visual perception all the time, and they`re finding out better and better ways in order to get a positive or negative effect on people looking at their works of art. CHARLIE ROSE: So there was this panel that I enjoyed enormously and learned a lot, but what should we take away? What do you want the people at home who see this to come away with? ERIC KANDEL: I think the important thing to learn is how synthetic the brain is, how it lives in a world from which it extracts limited information and how much of what we know about the world is reconstructed in our brain. And this not only holds true for vision, it holds true for all senses. We see the complete picture even though we get fragmentary information. So it makes us realize how magical the brain is. We also realize how remarkably blast tick brain is, that throughout our life we`re constantly modifying our view of the world as we learn more about it. And in case of injury we can disrupt function of vision, but under many circumstances there`s the capability of recovery as we saw in those Indian kids. CHARLIE ROSE: Next month we`ll do what? ERIC KANDEL: Next month is a natural extension of sensory systems. We`re going to discuss action, movement. The reason sensation is important, the reason we want to build up an internal representation of the outside world is we want to act. I want to interact with you. I want to be able to shake your hand. How do I reach out and see where your hand is located? These are again really magnificent computational tasks that the brain accomplishes this time in terms of movement. And that`s what we`ll take up with another outstanding group of specialists. CHARLIE ROSE: I look forward to it. If you want in to know more about our brain series, go to my web site CharlieRose.com, and you`ll get a sense of what we`re doing, some additional reading, and what`s coming up. Thank you for joining. See you next time.