Torsten Wiesel

Brain Mechanisms of Vision

Category: Lectures

Date: 3 July 1990

Duration: 31 min

Quality: HD MD SD

Subtitles: EN

Torsten Wiesel (1990) - Brain Mechanisms of Vision

Torsten Wiesel only participated and lectured once at the Lindau meetings during the 20th century. But during the present one, he is a more regular participant in the meetings. The lecture he gave at the 40th meeting gave an overview of his results concerning how the brain treats the electrical impulses coming from the eyes

Thank you very much for introduction and I appreciate to be invited to participate in this meeting. I'm also very grateful to Jack Eccles who has made it possible for me to just make a few comments about the vision system. It’s true that by working on a system like the vision system you may learn some of the principles that are used also by other areas of the brain. And for those of us who are brain scientists it often pays off to focus your attention on one particular subject which I have done for about thirty years. That is on the primary visual cortex in the cat and the monkey. I'd like here to present work that first David Hubel and I did in the ‘60s and ‘70s and then discuss work that Charles Gilbert and some other colleagues have done more recently at Harvard and the Rockefeller University. I will sort of try to, for the students’ sake, show, illustrate something about how cells in the brain respond to visual stimuli and give you a sense of how it is to be a neurobiologist. With the hope obviously to try to seduce you to the fact that the brain is a very interesting structure and we know something about it and there’s a great deal more to learn. Why I'm using this thing is that the first slide (could I have the first slide and the light down) is by the same painter as Jack Eccles showed that means that in the field we think alike! So this is a painting of Seurat, maybe Jack has made a choice on this but the point I like to make in addition to the fact that these painters used small points to make their images. It’s also if you think about this painting when it falls on your retina it’s going to stimulate single photoreceptors which absorb the light and then convert that into electrical energy. And then it’s sent, being processed by the eye and by the higher visual centres. There are about 200 million receptors in your eyes and only one million optic nerve fibres, so already in the eye there is some processing occurring. So, let’s first take a picture, (could I have the next slide, I don’t know if I can control it myself). So this is a picture then of the human brain with the eye and outlined here is the visual pathway. As you know the retinal ganglion cells project into the thalamus, the nucleus called the lateral geniculate nucleus which in turn synapses there and then send their fibres up to the primary visual cortex. And Jack Eccles pointed out these cells in the primary visual cortex send fibres to higher visual areas V2, V4 and V5 etc. I'm only going to discuss what happens in the one, it’s a complicated enough structure for me to try to understand. There are many people now Zaki and many others who are recording from higher visual centres, both in anesthetised animals and also wide awake monkeys with implanted electrodes. So you can record single cell activity in the behaving monkey and correlate the behaviour of the monkey and response what the monkey sees with single cell activity. The fundamental assumption here in this work is that by recording from single cells and studying their response properties you can indeed learn something about how the brain works. When I left Sweden to go to the United States to Stephen Kuffler’s laboratory I was sworn by my colleagues who will be unnamed that this may be a useful task to try to understand the structure with billions and billions of cells by looking at one cell at a time. But as you will see this has over the last thirty years in many laboratories turned out to be a fruitful way of looking at things. And it’s still profitable and will be for many more years I believe be useful approach to try to probe out the secrets of the brain because a large part of this structure or knowledge is still very primitive. Now if you look at the visual pathway, (next slide) this is looking at the human brain from underside and just to make the point that the projection from the two eyes and the crossing here, so each left side of the brain projects to the left hemisphere and the right to the right. This makes it possible, and then each hemisphere receiving input from the contralateral visual field. This crossing makes possible binocular vision, that’s perception, fusion of the image etc, which I won’t have time to discuss today. The other important fact of the organisation of this visual system is called topography and that is, it’s a very orderly projection of each half retina onto that geniculate and onto the visual cortex. So that the peripheral part of the retina or the visual field projects to this part and the most central part here. So if you record from cells in this part of the brain you have to stimulate this part of the retina. If you record more on this part you have to stimulate more peripheral parts. So this is one of the fundamental organisations of all our sensory system is topography, the laying out of here the visual field in a very highly orderly fashion. The next slide shows the processing in the eye, here we have the eye with all this beautiful optics and a piece of the retina has been enlarged and here you can see the photoreceptors, the bipolar cells, the second order neurons and the retinal ganglion cells that project centrally into the central nervous system. Now there are also fibres, cells, horizontal cells, amacrine cells that make horizontal interconnect these cells lines and they are often inhibitor neurons. So this circuit here is possible to interact spatially, excitatory and inhibitory influences. And my mentor Stephen Kuffler was the first to show in the mammalian retina what happened. He showed that if you stimulate photoreceptors in the centre here then there’s a direct axonal pathway for some cells into the retinal ganglion cells into the brain. Then he stimulated the receptors on the side, the cell was inhibited rather than excited. So that was then a spatial separation, these cells receptors excitatory and these inhibitory. And if you could imagine that you are looking down on the retina from above in the next slide you will have then the area over which the cell responds, this is a centre. You stimulate here you excite the cell, if you stimulate the receptors in the surrounding area the ganglion cell we record from is inhibited. Now all these recordings are done with microelectrode extra-cellularly and you can then record action potential. I will show you an example in a film in a few minutes of a cell like this, excited in the centre and is inhibited when you stimulate the surround. These cells are built, this is very sophisticated type of processing but now cells don’t only respond to light that falls on the retina but the pattern of light. The contrast is very important in order to optimally activate the cells. The bar here is to show that this is a circular symmetric thing and this is just to make the contrast with the way cells individual cortex respond. The size of these varies in a fovea that can only be a minute of all, that is very, very small and this is when you go to the eye doctor. The smallest distance between the bar of the E for example and then in the periphery that could be a degree or more to the centre area. So the higher acuity region is small and in the periphery larger. There it’s larger because the sum of the larger area and your high sensitivity whereas in the centre you have high acuity and lower sensitivity. These types of cell organisation, Kuffler also showed that you had the reverse arrangement in also about half of the cells have inhibitory centres and excitatory surrounds. In the film I will only show this kind of cell. Now the cells in the visual cortex which receive projection from lateral geniculate cells which have exactly the same organisation, the centre symmetric surround is quite different. It’s a major transformation that occurs in the way cells respond to visual stimuli. And this is illustrated in the next slide which is a diagrammatic illustration. This is a visual field, here’s a fovea and this is the cell that is then in the left hemisphere with visual field in the right. This is the size of the area over which the cell responds, called the receptive field. This is the small square form and here in the centre we are moving the bar across the receptive field in different orientations and we get the best response at the one o’clock orientation moving to the right. If you change orientation the response declines. And you can plot this sensitivity to a different orientation in the tuning curve shown here. The remarkable thing and I will show you, you see that in the film is that here you stimulate the same area, the same receptors in the retina and the only difference we make is that the change orientation of the stimulus and the cell doesn’t see that. It only sees the cell, the bar of a particular or a contour of a particular orientation. Now I'd like to show you the film of a few cells just to get you a feeling and the first cell is then a recording from the lateral geniculate body in the cat. The animal is asleep and is like you looking onto the screen, onto which we project spots of light and map the receptive field. You see the small area in the centre where you get a response and we will then stimulate the surround and the whole field. After that there will be two cortical cells with properties quite different. So this change from the first cell into the cell sensitive to orientation of contour is through specific wiring in the cortex, specific circuitry that makes it possible to generate these kinds of. So maybe we can have the film now? This is the centre of the receptive field (we could increase the sound a little bit more, that would be better). This is stimulating the whole field centre and surround cell response. This is a surround only, you see the response when you turn the light off. You are not eliminating the light that is in the centre. When you turn it off, you release inhibition and the cell fires. Also when you have a big spot and you contract it, this is now the centre only, big spot contracted you remove the inhibition and the cell fires. So again it’s quite powerful inhibition in this. This is just to illustrate that these cells are symmetric, orientation of the bar is of no consequences. We can move this orientation. This is very important because this is not seen at the cortical level. So now we have a cortical cell, we first map the receptive field, map the area of the individual field over which a cell responds. For the animal, it’s looking at the screen, is anesthetised but the sensors still active even if this animal is asleep and you then try to determine the area over which you can evoke the ordered influences, these charges. As you know cells communicate to each other through an action potential which is like a Morse code in a way, frequency tells you about the intensity or effectiveness of a stimulus. This is a rough map of the receptive field. This is in the area in which the cell responds. As you can see the orientation isn’t quite right and the field is a little smaller than it ought to be because its ... This cell is best to left but also to the right. So now we are stimulating the same receptive area in the eye or the visual field and there’s no response. So again you have to understand that it’s through specific wiring in the retina and in the cortex that this happens. This is a very robust thing, you can take an ordinary paint brush and move it across in different orientations and you will see. You can see cells like moving stimuli, just like we do. Our eyes are constantly moving so if you stabilise the visual image which is possible to do technically you go blind in three to five seconds. Then if you move the stimulus of the eye the whole visual scenes come back. These cells similarly stop firing if you don’t move the stimulus after a while. Now this is another cortical cell which I show because it has similar properties to the one you just saw but still an additional quality that is again an elegant demonstration or a demonstration of an elegant wiring, I should say, in the cortex. I didn’t know that field. So we have a map here and then we have a very directional cell, it varies, some cells are very directional, some cells responds to both direction, the first one…And this is the punch line of the cell. Here what happens is that you have inhibitory reflex here, you get no response here but there are reflex here that make it possible, that inhibit the cell from firing. And it’s only when you stimulate the centre alone. This is what we now call an end inhibitor cell, I like to in this talk to try to give you some feeling. So this I hope gives you a feeling for the, (its fine leave the slide on) now it turns out the organisation of the cortex, what we call the functional architecture is highly specific. You can keep in mind what Jack Eccles talked about dendrons. That is, cells close to each other have common properties. And what you find when you record from the visual cortex make a perpendicular penetration and record from cell after cell after cell is, it goes through all the cells in a given path we have the same orientation preference. Prefer the same orientation of a contra-crossing a receptive field. This is what we hope you find in orientation columns. That is columns of cells in the same orientation preference. Now, this really is a term common that we use, a term from some years ago but it turns out there are no real columns or had the structural architecture showed as circular but they are long narrow bands going through. Now, if you make many penetrations then you can get a feeling for the organisation of the orientation columns in the cortex and that’s illustrated in the next slide. Here is a visual cortex surface, white matter and here is one penetration perpendicular penetration here, all the cells had more or less vertical orientation preference and here’s another penetration also the cell had the same vertical orientation. The receptive fields you record from ten or so cells in this penetration all had overlapping receptive fields. Now because of the topography as you move from this point to this point there is a shift in the receptive field position. So all the cells we recorded here had their receptive fields here. In fact you have to move about one to two millimetres in order to get the fields not to overlap anymore. Now the main point here is to show that as you make an oblique penetration, so you go and record from several columns and these are steps about 50 micrometre steps, there is a shift in orientation preference. As you can see here from contra-clockwise back to after, actually in real life it’s about eighteen shifts, but it’s hard to draw from one vertical orientation column to the next. The highly orderly sequence on this changed orientation is shown here as you go from vertical counter-clockwise over many steps back to the vertical again. This is a distance of cortex about between half a millimetre, that is a chunk of cortex that deal with a small part of the visual field in terms of analysing orientation of contours, what David Hubel and I have called a hyper-column type of organisation. Now when we had come to this stage, Charles Gilbert and I started to collaborate and we were interested to understand the wiring of the cells within a given column. The approach we used was to record intracellularly with micropipette electrodes which were filled with a dye horseradish peroxidase, HRP, and as the next slide shows a cell filled with horseradish peroxidase, this is a cell here, the dark is a pyramidal cell you can see the apical dendrite and basal dendrite and also axons. This HRP stain is wonderful because it fills axons over long distances and also maleimited the fibres which the Golgi method does not do. It’s a counter stain to show all the other cells in the surrounding area. It’s to see how big a single cell is, you may think these are small cells but in fact most of them have this spread over several hundred My with their dendrites. Now in order to get this cell reconstructed you have to do serial section and then ... tracing (inaudible, 20.02) and a two-dimensional reconstruction is shown in the next slide of a pyramidal cell. These then are very beautiful cells, cell body, basal dendrite, epical dendrite and then axon leaving the cortex. This cell is in layer six, as Jack said there are six layers, this is layer six. And it leaves the cortex and goes back to the lateral geniculate body but it also has a very important projection up to layer four, to have a very specific function. Now Charles Gilbert and I we have recording from hundreds of cells in different layers and this has led us to do a circuit diagram of the connections that cross the cortex and that’s shown in the next slide. I don’t want you to pay too much attention to the details but it’s a nice picture that you get which very much confirms the classical histology with some additions. The projection is primarily into the middle of the cortex of layer four where geniculate fibres with a circular symmetric field make contact with spiny stellate cells in this layer and they have the simple cell properties. The simplest properties you find in the cortex. These cells in turn project to superficial layers, pyramidal cell for example which have more complex properties than larger receptive fields and these cells are the ones that project to higher visual centres. They leave the cortex and go to V2 and V3 and V4 etc. But they on the way out have a very important projection down to cells in layer five, where we have these beautiful pyramidal cells that Jack mentioned. And these are cells that project subcortically to the superior colliculus for example which is an important structure for control of eye movement and things of that sort. But they also on their way out give collateral to cells in layer six. Now the receptive fields in these cells are large, again as you go from step to step the fields get larger and larger. And in layer six cell that receives input then from layer five cells they are particularly large as you will, I will come back to in a little bit and then this recurrent projection. In each layer about 80% of the cells in the cortex are either pyramidal or spiny stellate cells and about 20 – 25% are inhibitory GABAergic cells. And these are illustrated here, I don’t want to go into the details of that. You have to keep in mind that at each stage and at each layer there is an interaction again between excitatory input and inhibitory circuitry. Now the one discovery that we made with this method of intercellular injections with HRP is illustrated in the next couple of slides. This is a cell in superficial, (could I have the next slide) this is a pyramidal cell in layer two and three, epical dendrite and basal dendrite. It looks messy but if you do a three-dimensional reconstruction with a computer graphic system of the dendrite axonal projection you can see what is shown in the next slide. And here is more or less axonal projection in the same plane which we saw before and this is rotated 90 degrees. Look in my hand, in one orientation and in the other. This is very interesting, I have a cluster of axon collaterals around the cell body here and then there’s a distance and then there’s another cluster of endings at a distance away and here too. And in the projection in layer five you have the same thing, you have a cluster of endings here and here and they are beautifully lined up. Now these sort of projections and this distance here in this case is the distance you would expect from going from one column to the next. So of this cell with vertical orientation preference our hypothesis from seeing pictures like these would be that they are projecting to cells in the same orientation columns, which is the thesis of this talk. Now I like to show you the three-dimension of this film and if the sound is off we can have the last part of the film, unfortunately there is no sound for this. This is a cell in the superficial layers and the cell body is here and then these are axonal processes. It’s just to show you the complex organisation of the cell and how they project not across the cortex but along the cortical surface. And they have these very typical clustered endings visually then our connection to cell with the same orientation preference as the cell here. And you can see this is then looking across a layer and you can see how it’s like a flat sheet within the cortex. This is a cell that you saw before, the dendrites and here are the axons and as you rotate it you can see the, it’s a little bit like a ship, majestic ship sailing. You can see again how the clusters are lined up very beautifully in a columnar fashion. So if you make a penetration through here and record the cells they will all have the same orientation preference. So maybe that’s enough of the film, just to give you a sense. So this is a general scheme which we have evidence and I always like to present evidence particularly since students are here. But time is not making it possible to go to those slides. But this is the scheme we ended up with, here are the highlighted cells you may be able to see from the back, cells all with the same orientation preference, in this case vertical orientation. And the proposal is that these cells are interconnected. They don’t connect with cells with different orientation preference. We can show that by physiological means by recording from this as a reference cell determining its orientation and then record from cell life to cell life and correlate their firing. And we find there are only cells that are correlated with our cell are the same or very close to the same orientation preference. The other method we have used is an anatomical one. You can inject a reticulated tracer, you inject the tracer in a region with a cell with vertical orientation and the reticulated tracer will then go back and fill cells that project to this area. We have shown that the cells that project to this area by this anatomical method are all within the vertical orientation, or the same orientation as the cell to which they project. So we believe that we have good evidence for this very highly precise horizontal connection. So the next slide then shows the general diagram, (could we have the next slide). So this is then the vertical connections that I talked about first between the different lamina. And then there is a horizontal connection which makes it possible for cell with the same orientation preference to interconnect. Now these connections go over many, many millimetres and it makes possible then to integrate the visual area not only to have an atomistic, very small representation but that cells connect and talk to each other over wide regions. Now we like to end up with a demonstration of how we try to illustrate the importance of the circuitry and the next slide is just the same (could I have the next slide). This is a cell what we showed on the film, the last one is end-stopped. It gives a good response to a short bar moved across back and forth across the receptive field. The long bar gives a very poor response. So this is then what we call an end-stopped cell. The next slide shows the circuit that we proposed to explain, and that is as I mentioned there’s a projection from layer six cells back to layer four and we have shown by EM studies that 80% of connections are within GABAergic inhibitory neurons and they in turn project to the spiny stellate cells. Now the spiny stellate cells respond very well to short bar as shown here and poorly to a long bar whereas the cell in layer six and this inhibitor neuron responds very well to a long bar because it needs summation over this area but very little at all to short bar. The experiment to test this circuit is shown in the next slide, in which we have the cortical layers, we have recording electrode one from spiny stellate cells here and another recording electrode in layer six which also have a GABA-electrode, so we can inject GABA and GABA is an inhibitory substance. So you can silence this area of the cortex without interfering directly with this cell. And if you do such an experiment which is shown in the next slide you have then the same cell, an inhibited cell, before injection of the GABA you get poor response to long bar, good response to short bar. And when we inject the GABA and activate layer six inhibitory you get as good a response to long bar as to a short bar. And then it recovers within a few minutes. The next slide shows the role of cells of this type to detect curvatures. A cell with no end inhibition, see no difference between a short bar, a long bar or a curve bar whereas a cell without inhibition responds well to a long bar, to a short bar and not at all to a long bar and reasonably well to a curve bar because the curvature here, these end cells here are the same orientation and sensitivity as the centre region. The late David Moore used these sort of cells as a building block (could we have the next slide) to think about how our perception works. This is from a teddy bear and you imagine that individual cells are in the primary visual cortex. Here cells are vertical, respond in the vertical contour will be activated and here the horizontal contours and in this way you could more or less do a complex picture like a teddy bear quite well. So this then is the concept, I like to think that cells of the kind that I’ve shown you are important for form vision and that they are building blocks. The last slide is (could I have the last slide), is a picture that ... Young, a well-known British neuroscientist, sent to me when David Hubel and I published original papers on oriented cells, cells that are in oriented lines and he said this is all very interesting, your work, but nothing new. It’s clear that Van Gogh knew about the importance of oriented line. So, I interpret this to mean that perhaps visual, that artists have a deeper, more profound understanding of the mind than individual neuroscientists who record from single cells. Thank you.


Torsten Wiesel only participated and lectured once at the Lindau meetings during the 20th century. But during the present one, he is a more regular participant in the meetings. The lecture he gave at the 40th meeting gave an overview of his results concerning how the brain treats the electrical impulses coming from the eyes. The tape recording cannot really give full credit to his fascinating lecture, which among other things included a film, in which sound played an important role. But since I am lucky enough to have heard Wiesel lecture on a similar subject at the Royal Swedish Academy of Sciences, I believe that I understand what is going on. By inserting fine electrodes as antennas into the brain of, e.g., a cat and showing the cat different objects, the signals from the antennas give important clues on how the brain treats the information from the eyes. As I remember it, showing, e.g., a triangle, the signals from the antennas show that the triangle is actually projected on the surface of the brain! It is a strange coincidence that another film was shown at the meeting, on the day before Wiesel gave his talk. This was because of the 40th anniversary of the meetings and the film was a documentary with the title “Nobel brought them together”. In German this becomes “Nobel führte sie zusammen”, which is also the title of a book by Alexander Dées de Sterio from 1975 (2nd edition 1985). This is a very interesting and useful book for someone interested in the history of the Lindau Meetings. It has a complement in Ralph Burmester’s bi-lingual book from year 2000, “Science at First Hand” (“Wissenschaft aus erster Hand”), which tells the story of the Lindau meetings up to 1999/2000. A large part of the information in these two books is, of course, also available on the Lindau web site. Try writing “Wiesel” in the search engine and read more about the background and present activities of this Nobel Laureate! Anders Bárány