History of Neuroscience: Francis Crick

I was interested in science from very early years. I can't exactly say when, but my parents gave me some bound volumes of a periodical called the Children's Encyclopedia, and that covered all sorts of things. Myths, literature, but also a lot about science, astronomy, and so on. I was fascinated especially by the scientific part, and that's when I really decided I wanted to be a scientist, because there's so many interesting things have been discovered, and obviously there were other things yet to be discovered. In fact I was afraid at some stage by the time I grew up they'd all be discovered, but that isn't the case. So I had this very early interest in science, and then of course my career scientifically was rather odd, because I started to do, I got a degree in physics and mathematics, and then started to work on a very dull problem which was interrupted by the war. Then I had to do war work as an engineer essentially during the Second World War, and the really big decision came after the war. I was by then had a permanent job as a scientific civil servant, but I didn't want to go on making weapons. So I had to decide what I was interested in, and by what I call the gossip test, finding out what people gossip about, which is what they're really interested in, I found I was interested in two main things, which seemed to be at that time both deeply mysterious, the borderline between the living and the non-living, which we should now call molecular biology, and how the brain works, and in particular consciousness. So I narrowed it down to those two things, although there are other interesting things like the origin of life and the origin of the universe, but I went for those two, because it was widely believed by non-scientists that scientists couldn't explain them, and that was what really motivated me, and then I decided I would work on what we now call molecular biology. I was offered a job to work on the eye, which I firmly refused on the grounds that I decided I was going to do molecular, which was a very fortunate decision, and then of course I had to find a way to get going, but fortunately the people I'd worked with during the war were helpful, and the Medical Research Council enabled me to get going, and they decided I wanted to know about the borderline between living and non-living things, like viruses, and in particular about proteins, and that's how I eventually joined the group at the Cavendish, the physics lab, which were working on that, but of course I also decided the really important thing were genes, what the genes were made of, how they replicated, and how they acted. That was my major interest. I wasn't so clear at that time that DNA wasn't necessary, the genetic material, because a lot of protein, after all, in chromosomes, so that's how I got into that, but I always retained an interest in the brain, and I took that up again when I read about the work of Hubel and Wiesel, and when the talk started, I tried to promote a meeting so that we might get work going here, so that I've had a long-term interest in the subject, but I wasn't actually working on it. I was just, you know, just seeing what was going on. When I moved here in 1976, well, by the time I was 60, I decided if I'm going to change, it'll be now or never, so I announced that I was going to go into neurobiology, and I decided I better learn various things, starting with neuroanatomy, and what I would try and do was to go and see people. I must say they were all very helpful. I did take the precaution of, before I went to see them, reading some of their papers first, so I knew roughly what they were doing, but people were really, really very helpful. On the other hand, I realized I had an enormous amount to learn. Everybody is in the situation of saying that they're flooded with information. Mind you, that's not a new thing. Cajal, in about the 20s, said that there was now more information than he could digest. You see, from our point of view in those days, it was very little, so everybody has always complained as they get older, there's more information, but I think it has been more dramatic in the last five or ten years and more. Now, it's much worse when you study something like consciousness, because basically you don't really know what you're looking for. If we had a really good theory of consciousness and knew what experiments to focus on, we could probably ignore a lot of what's being done, but as it is, we have to keep an eye on all sorts of things, all sorts of psychology, brain damage, funny things, gap junctions between neurons in the cortex, and all sorts of levels we have to look at, and this is very demanding indeed. So we have a special problem. I did have the attitude right from the beginning that the only way to solve the problem was to approach it from a large number of different positions. That's commonplace nowadays, but it wasn't really so commonplace when I started. What I mean by different positions is you have to look at it from the level of down to the behavior of the animals, down to looking at the neurons and how they behave, the neuroanatomy, looking at the molecules, all that thing. Every level you have to keep an eye on, because in general, although you want to explain things at different levels and they're different type of explanations, if you're doing things at one level, the clinch really comes usually by some experiments at the lower level. If you wanted to prove a theory in visual psychology, if you can show that it's strongly purported by the way the neurons fire, that really often clinches it. And equally, if you want to know how the neurons fire, and it depends on certain connections or molecules or this, that, and the other, again, you can easily disprove theories by going down to a lower level. So each level are going to have a different way of looking at it, but all the levels will have to interlock, and often you'll want to go down to a level or two to prove decisively, as opposed to just saying, well, this is a possible way of doing it, to prove decisively, or as far as you can in science, that something's happening. So that's why you have to keep an eye on multiple levels. First of all, let's say when we come to consciousness, one of the things we try to avoid doing is an exact definition. What we're trying to do, well, the reason we do it is because this is what Seymour Benson says to me, he says, if you want to work on consciousness, how do you define it? And I said to him, well, you define a gene. Well, he started to give me a definition, and of course he soon found out he couldn't. And I said, well, if you can't define something which is so well understood, how do you expect to define something which you've only got a rather vague idea about? But it is important to have a general idea of what it is, even if you don't have a definition. And most people get the idea that when they're under a deep anesthetic, or in a very deep, slow wave sleep, they're unconscious, and when they're, as we are today, they're conscious. But as you get into it, you realize there are various degrees. And also, I think it's important not just to talk about consciousness, but also to say, what is the problem? The main problem is what philosophers call qualia, the redness of red, the painfulness of pain. There are people who say it is impossible in principle to explain how you see colors, or shapes, or movement, or whatever it may be, or hear things, any of the things that you might say you're conscious of, in terms of neurons firing. So that is the basic problem of consciousness, how you're going to explain qualia. Now, the position that Christophe Koch and I have adopted is not to attack that problem head on. The argument is, let's first of all realize that much of what goes on in your brain is unconscious. When you see some things, a lot of unconscious activity leading up to your being conscious of it. And then it has lots of things you do, like learning to swim, you don't know how to do it. Almost hard to speak, you hardly know how to do, forming your will's instances. So there must be an enormous amount of neural activity in the brain, which is unconscious. Equally, we postulate there must be some sort of activity, whatever it may be, neurons firing in groups, or whatever, or firing in a special way, and so on, which is conscious. So the question is, key question is, what is the difference between them? So what we're looking for in the first instance is what we call the neural correlate of consciousness. Now remember, a correlate is not causation. So once you've found a neural correlate, you've obviously got to then think of a cause, how the thing actually works, why this produces that, rather than this going together with that. And the hope is that when you've done that, you have a much better position to adjust to what's called sometimes the hard problem, the problem of qualia. So that's the way we approach the whole thing. If I say to you, you know, look at things here for a moment, and then close your eyes and imagine it. So imagine you can imagine this sitting me. Well, you can do that. But the image that you have of me, at least I imagine, is very much more vague and diffuse than if you open your eyes and look at me. On the other hand, I think you'd have to say there was qualia there. So not all qualia have the same strength. Some of them are vivid, some are vaguer. Then there are problems as to whether you think that there are qualia for emotions or for self-consciousness and so on. The view we've taken is, let's go for the simple straightforward ones, strong qualia in the visual system, and see if we can explain that, bearing in mind that there are other ones. And there are also other probably transient forms. When you're driving along a road, you don't think there's nothing there. You've got the impression there's something there. And if there's a big bang or a flash or something, your eyes would turn in that direction. So you may be having very transient qualia for what's going along there. Normally, the point about being conscious is that in the visual system, what you have to say is the following. When you're conscious of something, it means the brain has worked out the best interpretation of what's coming in, which is a very complicated job it has to do in the light of previous experience. It's presented that result presumably to the front of the brain for sufficient time for it to maybe be able to make decisions and think about it and so on. If it was only there for a fleeting instant, you wouldn't be able to think about it. So the argument is there's some form of activity which persists for a certain time, at least one or two tenths of a second or something like that. So you can describe various features that you expect to find with consciousness and which have qualia, and those are the ones that you then try and find the neural correlate. Well, that's what we try and do. At the moment, we can't say we've had any striking success, but at least we feel we're getting sorted to understand the problem. What we think is roughly that in the visual system, you do start off with the back of the brain with area V1 and move forward in two streams. We'll just talk about the lower stream, and probably quite a bit of those neurons in those areas, except possibly the first one, do contribute to consciousness, and they send their message to the front part of the brain, which has to, as it were, decide what to do about what's coming in, what actions to take, what thoughts to have, and so on. And I think we broadly think that's true. I've been coming to the conclusion, although we haven't dared to put it in print, that really the way our brain makes may correspond to the way everybody thinks it works. See, there's an enormous illusion that here you are looking at some screen or something like that. Admittedly, it's the outside world on the one hand, and that you are looking at it. It seems to me remarkable if that isn't going to be reflected in the structure of the brain. Now, people say, well, what's wrong with that is that you're going to have a homunculus, and you're going to have an infinite regress, essentially, because who's conscious of the homunculus? But I don't think that's right, because I don't think the homunculus is doing the same sort of thing as what you're looking at. Moreover, you could argue, as Jakindoff would do, that the homunculus is largely unconscious. Personally, I don't think that's right now. I think it's more likely that it's conscious in a different way. It's a different sort of qualia which are involved in that. So the idea would be that you have complicated neural networks at the back and complicated neural networks at the front. And in some sense, these ones at the front are looking at the ones at the back. Of course, they are interacting. So obviously, you have to have some general picture like that, but then you've got to see how it maps onto the activity of all the neurons. And when you do that, you find you either haven't got much evidence, or it's rather complex. Well, I think you have to ask, basically, how it is you see something. And the first thing to say is that the information coming into your eyes is not sufficient by itself for you to work out what it is out there in the world. Your brain has got to do a lot of interpretation as to what comes in, and it does that in the light of past experience, either its own past experience or some of the wiring which has been laid down because of the experience of your ancestors due to evolution. So it has to do that. And of course, it does change very much as you go from a baby to a mature person as to what you will actually see and what you will be able to recognize. For example, learning to read. If you look at something in Arabic, it doesn't mean anything to you. If you look at something, if you learn to read in English, you don't have to think about it. It just simply pops out. So it's clear that a lot of visual processing has to be learned. And it's also, now, the brain has to, one of the most important things it has to do, it has to do things extremely quickly, especially as it's got very slow components compared with transistors. Consequently, it can't, as it were, work out everything in the greatest detail with the greatest assurance. It has to have mechanisms where it can make a good quick guess. In point of fact, it probably makes a very good, very quick guess, and then it refines the guess, is what we think broadly happens. And because it does that, it therefore sometimes gets the wrong answer. And you have mechanisms which are helping it, which on some circumstances will give you the wrong things. For example, the well-known waterfall illusion. You look at a waterfall, the water's going down. You look there for a minute or two, look to one side. Briefly, the rocks move upwards, you see. Essentially, what we think it is that the motion detects is what's called recalibrated, and so on. And then, so that another way to say it is they fatigue going down, whether it's the balance of backwards. But I think it's more a recalibration. There is a very difficult problem as to whether attention is necessary for consciousness. And the general feeling, I think, is that in most cases it is. You can have some very remarkable things happening. For example, if you tell people that they're going to look at a point here, in the straight ahead, and that they're to pay attention to something over here, and they say you're to pay attention, something's going to go there. And if you put something there, they won't necessarily see it, because their attention is switched over there. It's quite astonishing. They miss it half the time. And so, now what does attention do? Well, what we think attention does is that normally you've got this, say, visual input or whatever it may be, and there are a lot of different ways of interpreting it. And they're all competing. And the one that you're going to see is the one that's going to be the winner of that competition. It's going to push the other interpretations down. Just as in rivalry, you push them down alternatively. But when you see something, there are all sorts of ways of imagining that hand would be. And all the ones which don't really fit, the firing of those neurons is pushed down, we believe, by the effect of the ones that are fitting very well. What we think attention does is to bias that competition. And there is evidence of looking at the way neurons behave, without going into in detail by Desimone people, to suggesting that's the case. It's very difficult to prove that it's the case, but it's compatible with the observation. So attention should be thought of a way, especially top-down attention, of biasing the competition between different interpretations of what you're looking at or what you're thinking about. Well, when you come to memory, you have to realize, of course, we have very many different sorts of memories, which happen to some extent in different parts of the brain. As we said before, there's a memory of how you learned how to ride a bicycle. There's the remembrance of the meaning of words. There's the memory of what you did half an hour ago. Then there's another sort of memory, which is involved in remembering a telephone number. So that's called working memory. Most people can easily remember about seven digits, which is why telephone numbers are seven digits. And you can remember them for a longer time if you, what's called, rehearse them, if you just go on repeating them. Now, as far as consciousness is concerned, it looks, if you had no memory at all of what was happening, it was very difficult to imagine you would be conscious. These arguments were made in the 19th century. But it isn't clear exactly which sort of memory you need. It looks as if you need a fairly short-term memory, at the very least, lasting for the order of one or two or three hundred milliseconds or like that. Whether you really need the working memory is not so clear. See, you have people who have a working memory. You can count how many digits you can remember. Some people can remember nine. Some can remember, say, five. But there are people who only remember one or two. But they seem to be fully conscious. So it looks, and moreover, when you think about how you remember something in working memory, it doesn't seem to be quite as vivid as when you actually see or hear the thing. It looks more like imagining to some extent. So I suspect there is a form of continual neural activity, which I think is what memory is. Because some of the memories, as we said before, are simply stored in your synapses. You don't have to be latent. I think there's probably some form of continual neural activity which correlates with consciousness and which has to be thought of as a very short-term memory. What it is exactly, I don't know. When I first got interested in the brain and when I first learned about the brain, it was from talks given by Horace Barlow, whom I knew personally in Cambridge, about the retina. And so I did learn about motion detectors in primitive retinas and so on. Then, of course, when I learned about Hubel and Wiesel, I learned about the striate cortex, V1 as we call it, a little bit about V2, and also about the lateral geniculate nucleus. And I think about that time, people began to realize that there were at least two streams of information going from the retina to the LGM, from the big cells and the small cells, the magno and the parvo, probably other ones as well possibly, those two. And they had different characteristics. But as time went on, I began to realize, I didn't originally when I was working with David Marr and Tommy Poggio, that was a lot more to the visual system. And it wasn't enough just to think about V1. And about that time, John Orman and colleagues and Semezeki were beginning to look at further regions, which we now call V5 or MT or V4 and things like that. And as now we know, there's a whole hierarchy of visual areas, and there are general rules about them. But as time went on, one realized that isn't enough, because the information has to go somewhere to be of some use. So nowadays, we have to also worry about what happens in the front of the brain. For example, it's been shown there are neurons in the front of the brain in one or two places which respond to faces. It isn't clear what's different about those neurons compared with the face neurons in the so-called infratemporal region. Moreover, the whole organization of the front of the brain is less clear. When you come to the visual hierarchy, it's not only that you extract more complicated features, but the bit of the visual field that you're looking at gets bigger and bigger as you go further forward. So when you're looking at a line, it'll only be at a line that's a small place in the visual field. When you're looking at a face, the face can be here or here or tilted in various ways. And moreover, the impression I strongly get is that the neurons in the initial stages are firing quite a bit of the time, because there's always something going on coming into V1. When you get onto the front of the brain, the face neurons are not likely to fire very much unless you're looking at a face, so that you do get these different properties. Then, as we know, there are two streams. The one I've been talking about is the ventral stream, but there's a dorsal stream, which is probably concerned either with online things, knowing how to grasp things, things of that sort, and also it may be involved in attention in some way, and that's rather obscure. So essentially, what you find is that there's very elaborate structures of the brain, and any account of how you're going to see things has got to explain what all those structures are doing. For example, the area MT, or V5 as it's sometimes called, is mainly interested in moving things. It's not interested in color, and maybe some success in depth. Now, there is a lady in Germany, a patient of Zill's, who has bilateral damage. The net result is she can see colors and shapes, and she's perfectly intelligent, can talk about everything, but she can't see movement. If you pour out a cup of coffee, she doesn't see it moving, she just sees a gleaming thing. If there's a car in the street that is moving, she sees it as static, and then it moves it to some other place. People find that very surprising. They say, if the car is moving, and she can see the car, why can't she see it move? Well, the answer is, it's done in different parts of the brain. So this leads to a very serious problem, is how do the different parts of the brain work together to make a unified concept? Take a cup, for example. I can see it move, I can see its shape, I can see bits of its color, and they're all bound together in some way, so that if I had two cups, and this was like this one, and this was a green one, for example, I wouldn't get the two mixed up. This one wouldn't look green. I get the right shape with the right color. And essentially, that's what we call the binding problem. How is it that the activity representing a single object in different parts of the brain all come together and give you a unified concept of it? Now, it so happens, if you do things, you can trick it sometimes, but if you do things, very brief glimpses of things, you can get what's called force conjunctions. You can get things mixed up, but you hardly ever see this in real life. In real life, the brain is terribly good and puts everything together, the movement, the shape, and the color. The usual way of saying the binding problem is if you have a red square and a green triangle. How is it that you don't see it as get the colors mixed up with the shapes? That's basically the problem. Some people think that you do it because you attend first to one and then tend to the other. So it's not at all clear that there is a problem. Maybe you just see the one that you attend to, and you'll switch your attention very quickly. Other people believe there's partial binding when the things are very close together. Other people believe there has to be a special mechanism. There's essentially an unsolved problem. David Maher's view of the system was that it was hierarchical, but he thought it was largely a feed-forward hierarchy, whereas we now think that that's less likely to be the case. Consider the following thing. Suppose you're suddenly shown, for a very brief thing, a snapshot of what's there, say. You can say probably, oh, there's a bookshelves and some books in it. You may not see it very clearly, but you get the general idea. You say it was in a room, you're not in a forest, and so on. In other words, you get the gist of it. But the gist is a high-level thing. You get the gist before you see the details. So this certainly suggests, and other evidence as well, that the wave of activity comes in from the back and goes up to the higher levels, not quite by far, and then starts to send signals back to, as it were, pull the whole thing together in some way. And we wonder, when you do things in a stereotyped way, without consciousness, whether it's a purely feed-forward process, just like picking up this cup or something like that, and when you're looking at something, whether it has to go to the top levels and then come back again, and maybe do it more than once. So at the moment, we haven't applied the techniques which might help to answer that question. We do know that when you look at the way neurons behave in V1, they respond. At first, you show them like this, as a person in front of the visual field, the thing starts firing. But within about 100 milliseconds, it will also take notice of what's in the neighborhood here. So that if you put a number of things like this, it will fire less. In other words, it fires immediately as to what's immediately there, and later on, it fires according to the context. And that can either be the thing spreading sideways in V1, or things coming back from a higher level. So we have observations. We know there's something funny going on. We don't know exactly what it is. And obviously, if we're going to understand consciousness, we have to work out all the dynamics of the way these signals move rapidly backwards and forwards from one different area to another. And we do know that it happens very rapidly. You can be shown pictures at intervals of 150 milliseconds and asked to say, is it an animal? And it may be a lion, or a bird, or a scorpion, or whatever, a snake, or something. And of course, there'll be some wish hammer. And you're terribly good at it. You can say, yes, that was, you know, you can press a button and so on. Well, to do that, it has to go up to the higher parts of the thing, of the system. And yet, it doesn't get to the, it doesn't get to here till about 40 milliseconds. It only has about 10 or 15 milliseconds to go through each stage. There's hardly any time for a neuron to fire. It can only be, it can't be taking a big average of firing to do that. So we know the system is acting very rapidly. And we also know that it goes to a higher level very quickly. Beyond that, I think it's a big question mark. Christoph Koch and I had the, put forward the hypothesis that we weren't aware of what goes on in V1, the strike, the first visual area, the striate cortex. And the reason we said this was the following. We said, what, when, what you're aware of surely has to be sent directly to the front of the brain. That's the one that's going to do the thinking and the judging and the action and so on. And moreover, we said it ought to go directly. Because when you go from one stage to another, like V1 to V2, V3 and V4, in each stage, you recode the information so that it's different. So it isn't that V1 isn't important. Everybody would agree that you can't see if you don't have a retina. But most people would agree that what's going on in the retina, you're not directly aware of. It's quite easy to show that. Neurons are firing in a way which don't correspond to what you see. It all has to be processed. And what you see is higher up. So that was the prediction. Now, it happened just about the time we had this idea that we had Don McLeod to dinner. And he was telling me about some result that he got. And I said, but that's exactly fits in what we say. And he would got a very elaborate arrangement of looking at very fine patterns, very high frequency. And he could produce those. You have to sort of cheat the eye a little bit to do it. And this does produce effects in V1, or at least one presumes it's in V1. But you're not aware of them. You can't tell whether it's smooth or very finely granulated. And since then, there have been a number of other experiments. But the problem is the following. First of all, if our eye hypothesis were true, we'd have to show that everything in V1 has this nature, that it doesn't correspond directly. It only leads on to what you're aware of. That's very difficult indeed. And equally, on the other hand, if you do find something that does correlate, you have to show that the real correlate isn't one of the next stages. So it's a very difficult hypothesis to prove directly. On the other hand, I do think it's along the right lines. I wouldn't like to say, I think the real question is, if V1 is, if you're not directly aware of V1, what's V1 doing? That's the really important question. And I don't think we're very clear about that. It's probably you may have to project things back to V1. Now, it isn't true that everything is recoded. For example, if you have two colors, a dark red and a light red, you can easily say whether one color is between the other or the difference. You can compare two things or you can compare three things. That seems to go through from one stage to another. So that we can, I can explore your visual system. I see if you can make the same judgment that I can make. Even if we don't know what red looks like, we can do that. That seems to go through from one stage to another. So I think it's a reasonable hypothesis, but we just have to wait and see how it turns out. The point about the zombie is that you have to go from a visual input, assuming we're talking about the visual system, a visual input to an appropriate motor output. For example, if there was a cup on the table and we were talking and I did that, I would scarcely be aware of what I was doing. You see, it's a stereotype reaction, but it's not completely stereotyped because my hand would adjust itself to the actual cup, you see. And there are cases a lady has a certain type of brain damage, which is carbon monoxide poisoning. She can see color, she can talk about color, she can describe it, she can distinguish color and texture, but she can't see shape. She can't see the difference between that shape and that shape. She can't see the difference between this line and that line, okay. You give her a letter with a slot, an inclined slot, and she posts it. She can't tell you what the slot is inclined to, but she can post it. On the other hand, if you do it by doing it to the computer screen, so there's a delay, you show it to her, and then there's two seconds delay, and then ask her to do it, she can't do it. She has to do it online. So the assumption is there's a mode of the brain, you have to say it's a mode of which you can go right through from the visual input to the motor output, which means going far beyond V1, in a way that you're not conscious. But of course, normally, there's a slower process going on, of which you are conscious. And what's the point of that? Well, you have to ask, why can't you do everything with zombie systems? Why can't you do, you know, everything we do? The answer, you'd have to have an enormous number of them. Our behavior is far too flexible to be encapsulated by that. So we think it would be a great selective advantage if you had these two modes, one of which was very fast and fairly stereotyped and fairly automatic, which we were initially not conscious of, and a slower mode coming along about 100 milliseconds later, which takes a much broader view of the scene, can assess it better, and also stays longer. It isn't just online, so you have time to think what to do, you see. So that's basically our approach, and that explains, I think, to some extent, what we feel about V1. It's very interesting to ask, what happens in a person when they have damage in V1? It's usually on one side. And, of course, the first thing that happens is, if the damage is on the right side, they can't really see anything on the left side of the visual field. However, it was discovered quite early, and then exploited by Weisskranz, that, in fact, something does get through in some way. That is to say, if you tell them you're going to shine a light like this, or like this, or like that, and ask them to integrate with their hand which direction the light has been going, they do fairly well. They say they don't see it, but they go like this, and it's usually right. That's called blindsight. That's the jargon for it. It's arguable exactly where it happens. It's probably a different pathway going through the colliculus, and then to the pulmonary, and up to the brain, one would think, at least for motion. Now, recently, although this hasn't been published yet, the people of Oxford, Alan Cowan and his colleagues, decided to do it again with a delay. So they asked the man, we're going to show you something, and then, as I say, a little buzz. I don't know how they exactly do it. Then show us which way. And then it was very poor, very feeble indeed. And, well, that's what we'd expect. If it's a stereotype reaction, it has to be done online, and it's unconscious. So that's what we think it is. We think it's a sort of zombie reaction. But until we've seen the experimental data, I wouldn't like to comment on it in great depth. When you're thinking about the working of the brain, you have to think about the interaction in the neural networks. After all, the neural networks are very dense. Every neuron makes the order of 1,000 or 10,000 connections to other neurons and receives them. It's a very dense arrangement, even though, of course, not everything is connected to everything else. What it looks, and moreover, in addition to that, you've got a number of quite distinct type of inhibitory neurons, which must be doing special jobs that we don't know of. Now, we do know in certain cases, like binocular rivalry, that you clearly get competition between what the interpretation of what you're looking at. You put this into one eye and this into the other eye, and you see one or the other. They're clearly competing. The argument is that competition is a feature of all parts of neural processing. It's the fact that the brain doesn't immediately know what it is it's looking for. It's got to try out, as it were, simultaneously a lot of different hypotheses of whether that's a cup or something else or whatever it may be. And there's going to be a competition between them. And the best one is going to be what becomes conscious, pushing down the other interpretations. Now, that's the way we like to think about it. And we sometimes find it helps people to make a political analogy. Namely, it's rather like what happens when you have an election. All that struggling as to what you're going to see is all this stuff going on before the election. And the winner of the election, of course, is what becomes conscious and then has the ability to exercise power over the system for a limited period until it's challenged because things have changed and it loses its things and a new one comes up. So we think this way of looking at what we call coalitions, which are competing with each other, is very important. And it is an example, of course, of a winner-take-all situation. But it's only a temporary winner. In general, you're constantly conscious of new things. And the reason is that if you move your eyes, for example, for example, I was looking at you and then I moved my eyes to the cup, the coalition which did for your face has now got to yield to the cup because I'm not getting as much coming in for your face as I am coming from the cup. So the cup will push down the face. The face won't disappear any more than the opposition disappears when it's lost the election. It's trying to get there next time. What we would like to see is actually how the neurons do this competition, how they interact, how widespread the coalitions are. When I'm looking at you, it's all very vivid and it's probably a coalition of things in all sorts of parts of my visual system with the possible exception of V1. When I close my eyes and look, it's probably only the higher parts that are getting activated. The activation from the front of my brain isn't getting down to the details, you see. So not all coalitions are the same, just as not all qualia are the same. So we think in terms of when the time has come right to think in terms of coalitions and the dynamics of coalitions. It does look at once a coalition is formed, it stays for a certain time. There's a certain what's called hysteresis, times of the order of 50 to 200 milliseconds. It probably stays. And the other curious thing is you say, well, that's really like a static snapshot. But surely things are moving. But the answer is I don't think when things are moving, they're changing in the NCC. What's showing that there's something moving is that the neurons which symbolize movement are firing in that coalition. If like that later you lose those neurons, you can't see the movement. So we have this idea of a series of static snapshots. What we don't know is how widespread they are, how simultaneous they are, and things of that sort. And with a lot of psychophysical things showing that you do average what comes into your brain. You don't see everything. You take a chunk of it and average to some extent. But working that out in detail is going to be complicated. But that's the way we're thinking at the moment. Of course, the whole thrust of what Christof Koch and I are saying is that we should try and look for experiments which will be relevant to the tentative hypothesis we have. Now, there are a number of experiments which are worth doing. And mainly they're ones in which what's coming into your eye doesn't correspond exactly to what you see. And then you can ask, looking at the neurons, is it following what's coming into your eye or is it following what you see? Let's give a simple example of a visual illusion, the waterfall effect essentially, where if you look at the waterfall sometime, look at the rocks, they move upwards. Well, then you can look into the monkey's brain or to some extent even with people's brains by scanning and see where you get the illusion that the rocks are moving upwards. The rocks are clearly not moving upwards. It's not coming into your eye. It must be in your brain. The most dramatic one is binocular rivalry or what's sometimes called flash suppression. So I'll go through the flash suppression one, which is mainly the work of Nikos Logothetis and his colleagues. You find a neuron which is interested in the monkey, say, in faces. And so you put a face in there and the neuron fires happily away. Then you find something else like a sunburst. It doesn't fire to sunburst. So you put up sunburst, nothing happens. Now what happens if you put up a face and then in the other eye put up the sunburst? This one takes precedent and you don't see the face and simultaneously the neurons in infratempol, the face neurons, stop firing. That's very striking and it's moreover, it's very robust. When you've got things changing in irregular times, you can't tell. The monkey has to tell you when it's changing. With flash suppression, you know that it's going to change when you flash the extra thing. What happens in the different areas with binocular rivalry or flash suppression is that in V1, most of the neurons follow what's coming in, only a small number of the percept. V4, you get a bit more. And when you get to infratempol, most of them are following the percept. But we still haven't looked in other parts of the brain. We haven't looked in the frontal areas to see what's happening. We haven't looked in parietal areas. And Nicklaus-Lager's thesis is setting up to do that by having quite a number of electrodes in the infratempol areas and in the prefrontal areas. In other words, we can't say just because it happens in this part of the brain, we understand it. We have to look at a lot of different parts of the brain, including the front of the brain as well as the back of the brain. So that's the direction that this sort of work will go. It will be looking at new cases where the input is different from what you perceive on the one hand, and looking at more different, more cortical areas, preferably in parallel of these things to try and build up some picture so we can start to think about how it all works. One of the things you can ask is what can you learn by brain damage in humans? Because after all, humans can tell you what it is that they're seeing or not seeing. And we know a number of cases. We've already mentioned the one of the lady who can't see shape but can see color. It's well known that you can have damage in another part of the brain, the one which is active when you see color in humans, which you then only see in black and white. And quite typically, if the damage is on this side, you'll see black and white on this side and color on that side. Then another part of the brain where the neurons which respond to faces, you can get two sorts of damage. One is damage such that you can recognize a face as a face, but you can't say whose face it is. You can't recognize your wife's face. You can't even recognize your own face in the mirror. Although if you wink, of course, it will wink back. But you can recognize your wife by her voice and you know all about her. So it isn't anything like that. It's just you can't see faces. A more serious form of damage is you can't recognize a face is a face, but that means you probably see that there's little bits and pieces of it. So it's clear that when you think of what you're conscious of, some of it must be local. On the other hand, it's very unlikely if just the neurons were firing by themselves and nothing else in the brain for, say, looking at whether face neurons, that you'd see it. In other words, although you have to have local effects, it must be the way they interact together globally. So now you can say, well, possibly for every aspect of what you see, there's what you might say in the network, an essential node. Essential is a little strong because you've got them on two sides, but let's forget about that for the moment. And the argument would be if you don't have the essential node for that, you'll be able to see lots of other things, but you won't be able to see that particular aspect like color or movement or faces or so on. And that's an assumption because the examples we have may be rather special. Moreover, it doesn't say how big the essential node is. The node we look at is quite big. But we speculate, well, what is the smallest amount of cortical tissue which might be considered to be an essential node for that particular aspect of what you might be seeing? And we think it might be a column, in other words. And what is there is the property of the column. You recall that when you look in cortex in a lot of places and you go down this way, the neurons all have a similar property. They all respond to orientation, which is roughly the same, or they all respond to movement in another area, to movement, which is roughly the same, or they all respond to something like faces. So you've got a columnar property. So the argument is that what makes an essential node is its columnar property. So what's the point of that? Well, the point is that all those neurons in that column, maybe a mini column, let's call it a column for a moment, are all handling the same sort of thing. But they're sending the information to a lot of different places. You don't have one neuron which is sending your information everywhere. In that column, some will be sending it forwards, some backwards, some sidewards, some down to the thalamus, some down to the claustrum, and so on. They're sending it in a lot of places. Now, if you want to be the winning coalition, you want as much support as you can get. So if you send the same information to a lot of places, as it were, that'll help you to become a member of the coalition. And that's the way we tend to think about it. And we come to this from brain damage. But of course, we can't create brain damage on people. And we can't get the monkey to tell us so easily what it's seeing. So it's not so easy to test the ideas. I don't think in the long run, we will ever understand exactly how a person's brains work from moment to moment. That's much too complicated. What we're trying to understand is all these different things of the self and will and consciousness, to explain them in sufficiently general terms, supported by enough evidence to say, it's working in this sort of way, even if we can't predict exactly. It's just like the weather. We can say all the rules for the weather, largely anyway, what happens when air expands, when it goes upwards, and so on. But there's a limit to what you can actually predict. And the more details you have, of course, the more you can predict. But even then, it's very difficult to say whether a spot of rain is going to fall on a certain pot over there. We'll always have that limitation with people. Now, when you come on to thoughts, or emotions and things of that sort, people very much differ. Some people think that the only content of your thoughts that you're directly aware of are the sensory ones associated with them. Speech, unspoken speech, visual imagery, other sorts of imagery, and so on. And a number of people have said this, including Freud, incidentally. And there are people like Jakintoff who defend that. So what you're really saying is, I think, do we get the same sort of qualia when we see something and feel pain as we do when we have a thought? I think the answer is you don't. And I think the reason is you're dealing with a different sort of coalition. You're dealing with the coalitions at the front of the brain, which we don't fully understand. And they must have qualia, which I think, you can either say they don't have qualia, they certainly don't, I think, have sensory qualia. So it isn't clear what it is you want to say. And one way of saying it is, well, are there essential nodes for any of these things, or aspects of these things? And we don't know enough about the front of the brain to say that. So it's a very big question. And philosophers tend to take the view it's obvious what qualia are, but it isn't true at all. And a typical thing was emotion. Are you really aware of emotion? Or are you aware of, you know, you're getting flushed on this, that, and the other, and then you know the word angry, and things of that sort. Is it more than that? Or is there something underlying it, which you have to think about? And it's unclear to me how you should go there. Obviously, one can ask, from my point of view, where are things going, which would interest me? And I think they have to go in two directions. The first direction I think they have to do, you have to get round to the details of what the neurons are doing. It's a very peculiar situation at the moment. People record from, say, MT, and they'll find some neurons do this, and some neurons do that, and some the other. But they've no idea where those neurons are sending the information. They don't even usually know which layer they're in. And yet it's perfectly clear they are, although the jobs are similar, they are doing somewhat different jobs. So the first thing we have to do is to get down to the details of how all the neurons are interacting. Now you must realize you've got a very complex system. And how do you unscramble this complex system when it's certainly highly tuned by evolution? You've got to interfere with it delicately. And to interfere with it delicately, what you'd like to do is to turn off, reversibly, certain groups of neurons. Now we can do this for a little area. We can give GABA, or GABA analog, and since all the neurons respond to GABA, they'll all go silent for a time. So you can turn off the bottom layers of V1 and see what happens with the rest of the system. You can already do that. What we don't have the tools is to take a particular type of neuron, the large neurons in layer 5 which project to the colliculus or something like that, and turn them on and off reversibly. These tools are going to come from molecular biology. First of all, we need to distinguish what's different between the different types of neurons, from pyramidal cells, different inhibiting ones, in terms of genes that are working in some and not in others. And then we used to use those things to do tricks with them. And that will certainly be coming. And moreover, molecular biologists are very interested in the brain. And they're very good at producing new techniques and solving things. So you can reckon within the next 5 or 10 years that we're going to be some very striking techniques of interfering delicately with parts of the brain. Now some of it can only be done in a mouse. As I'm not talking now, I'd be able to do things on a marmoset. Much more difficult to do them on a macaque, because you normally want a cheap animal with a short breeding time. But nevertheless, we have the problem in humans, how can we alter their genes for medical reasons. And methods, the vectors, as they're so called, which are used for that, can also probably be used for the macaque monkeys. That's one way of looking at it. The other way of looking at it, we must have ways of looking at lots of neurons simultaneously on a very fast time scale. And the sort of thing I'm thinking of is you would take an animal with a smooth cortex, like a marmoset, which is a primate, and put lots and lots and lots of electrodes on its head, perhaps one every millimeter. That's a big engineering problem. And pick up what is called the local field potentials, so you get a sort of average view of what's going on in that area. Then you have this enormous amount of information, what are you going to do with it? What I think you do is you put it on a television screen, and you speed it up or slow it down, so that you can use your visual system to see all the waves that are moving around and all the activity. So I think you have to approach it from both these things. You have to look at the dynamics of the system as a whole on the one hand, and the way the little bits are doing on the other. And those are the two major directions, I think. Of course, it'll be supported all the time by brain damage, brain scans, looking at different, using different sorts of visual psychology, and so on. Those will go on anyway. But those are the things I would like to see. People work in different ways. Dirac, for example, the physicist, was really pretty solitary. I think he once had a graduate student or something like that, and he would do things almost entirely on his own. Personally, I prefer to work with one close colleague. I like to have other colleagues around, but I like to have one close colleague. And that first happened with Jim Watson, and then over 20 years with Sidney Brenner, to some extent with Leslie Orville, and more recently with Christoph Koch. Now, in those cases, it's a very curious thing. Of course, when you do it, you and the other person bring different backgrounds. But the important thing is that you shouldn't be too specialized in saying, I'm doing this and he's doing that. You should both of you try to understand in some detail what the other's point of view and what the other's point of view. And if you look at what would happen in the double helix, for example, the things that Jim Watson discovered were the things that I might have discovered, but it didn't go that way. So it's very important to interact. The other thing, of course, is the nature of the interaction, and that is it should be very candid. In other words, it must be a close interaction. When the DNA structure came out, and especially when we got a Nobel Prize, there was a question in my mind of whether I should become what might be called a celebrity. That, in my mind, is somebody made by the media, and I was very much against that. When we had the prize, they wanted to do a television program of the five of us who had got the prize, and I was against it. Hemsworth, the head of the MRC, persuaded me. He said, you know, you really must go forward. But I've always tried to keep out of the limelight in that sense. But the fact is that I've had an extremely good press agent, which is Jim Watson, and he's the one who's been promoting what this myth. I mean, Aaron Klug said to me one day, Francis, you must realize you've become a myth. So I've had to live with being a myth, but I haven't exactly encouraged it, and that's why I got a reputation of not being very friendly to the press, because I didn't want to be one of these people they consult on any issue just because they're well known and things of that sort. But the fact remains that partly due to the way that the subject has developed on the one hand, and partly due to Jim and Jim's book and his thing, that one has become what I would say totally unreasonably well known. Well, it helps in a way. I did find when I started in neuroscience, it did help me that if I wanted to go and talk to somebody, they were happy to see me, although I did do, as I explained, some homework beforehand. And it certainly helps with bureaucrats and things of that sort. But otherwise, I find it a bit of a nuisance. Of course, it does mean that I'm financially more secure than I would be otherwise, so I can't complain about that. But I'm not one to want to pontificate on a lot of issues which are important for society, for example, so I try and keep out of that. Transcribed by https://otter.ai