Perception of Visual Space by Colin Blakemore

I'm Colin Blakemore I'm professor of neuroscience and philosophy at the School of Advanced Study at the University of London and I'm going to talk about the perception of visual space people who study vision are normally interested in how we see objects things other people faces and so on but I'm going to be talking about how we see nothing how we see the space in between and in particular how we understand the 3rd dimension of space so important for for any animal whether you're a predator or prey you need to know how far away things are so I'm going to talk about the mechanisms in the brain that tell us something about how far away things are that's partly to do with the fact we've got two eyes so we can use our eyes like simultaneous range finders if you like binocular vision stereoscopic vision 3d vision but we also learn a lot just from the information that's in the image particularly from perspective and artists of course when they discovered that when crazy about it because it allowed them to create pictures on a flat surface looked as if they were three-dimensional so I'm going to talk about the interplay really between the discoveries of artists and the techniques that they use architects as well because they make solid buildings that fill space wanting them to look nice to us with our imperfect mechanisms for understanding space so I'm going to talk about the interplay between science and art and architecture so good evening and welcome to the final Darwin College lecture series this year this year this our lecturers have covered a wide range of aspects of vision the theme we've gone from hallucinations to our vision and understanding of the universe from computer vision and the development of artificial intelligence to technology and visions of the future from the evolution of eyes and the many and varied ways that organisms see including even scallops to the subjectivity of our interpretation of color so how do we see how is our perception of position and distance so accurate the compound eyes of a fly always managed to warn that the fly SWAT is descending the dog catches the tennis ball perfectly the rugby player gathers the unexpected past that's high and wet and in the wrong place and even more astonishingly the cricketer in the slips has just milliseconds to catch a spinning rocketing ball well of course that's perfect goalies don't always stop the ball but understanding of that third dimension is astonishingly good whether it's flies dogs or or humans we know where we are relative to other objects how does it happen how does our stereo vision work even how is it that we can see depth in the mystery of a 2-dimensional Turner landscape or in this painting so our lecturer this evening is clearly going to address just that so Colin Blakemore is the professor of neuroscience and philosophy in the University of London's a school of an advanced study was formerly chief executive of the Medical Research Council and president of the British Science Foundation Association he's worked on many aspects of vision and science has written books for general readership including mechanics of the mind and the mind machine and his awards for the public communication of science include through our society's michael faraday medal so this evening please welcome Colin Blakemore to speak on perception of the dual space [Applause] thank you very much Mary well this is the last talk in the series on on vision and those of you who have come to the preceding lectures will have heard a lot I'm sure about the processes in the in the brain in the eye and the brain that enable us and other animals to recognize objects to know what things are to know about people recognize faces and so on that's obviously very important if you want to find your way around in the world and respond appropriately to things objects are crucial but this in this last lecture I'm going to talk about the recognition of bits between the objects the space in which objects sit which is equally important as marius said knowing how far away an object is is crucial whether you're a predator or whether you're the potential prey of that object we need to know about the distances with things around us and I'm gonna talk about the mechanisms in the brain for achieving that and this story as with so much of the description of vision and the and research the history of a science of vision he's intermingled with the history of art and architecture because most artists and architectures are also crucially concerned with with space how to use it how to form it how to copy it and represent it so as as Mary said actually talking about artists is a good starting point because artists at least in the history of Western art since the Renaissance have made a valiant job of convincing the viewer of a picture in a gallery that they're looking at something that is not a picture in a gallery is a flat object it's a piece of canvas with some oil on it or whatever an etching or engraving or an drawing whatever and yet simultaneously you can entertain the view that that is indeed what you're looking at you're in a gallery you know perfectly well this thing in a flame frame is flat and yet you are convinced in parallel that it represents to you in a more or less convincing way another world looking through a window into a three-dimensional world as you can see in this in this example so I'm gonna talk about it about how that is is achieved the reasonable starting point to pose the problems for a scientist or an artist is is with day cut the cut was not the first person to observe directly the form of the image in the eye China had done that 30 years or so before but this is from the oblique in 1637 where Descartes not only repeated the experiment but also speculated on what it might mean he obtained a fresh eye an ox eye from the local abattoir fresh meaning still transparent and clear cut a window in the back of the eye taking away the white at the back of the eye but leaving the jelly and inside intact put a piece of paper against the jelly and then held the Ox eye up to his room and found I think even though he'd knew about Chinese experiments must still have been astounded to see a tiny inverted perfect image of his room where everywhere he pointed the eye now - a - of a philosopher interested in histology in how we gain knowledge of the world this must have been a really remarkable experience to see a bit of an animal in fact a piece of a dead animal capturing a perfect representation if you like of the outside world even though at that stage of course the representation is just light it's just a distribution of light doesn't mean anything in terms of the understanding of the animal as Descartes clearly saw but to see the image raised for him an important question which is really the guiding question through this talk and that is the fact that the image is is essentially two-dimensional it has depth in focus of course but at any moment the image in the you are in the eye is just a projection an optical projection onto the plane of the retina of the world outside which is almost always three-dimensional and the three-dimensionality is lost in this projection it might look three-dimensional if you look in a painting or even look at a retinal image but the fact that you see it in three dimensions is something of a miracle because it means that you must be interpreting the image itself to derive that sense of the third dimension and the retinal images it's not just that the information is directly given because all of the information the Rattle images is in principle completely ambiguous if you were to see against a blank field a round circle that is what you see out there in the world it could indeed represent a circle hoop or something like that facing you directly the froggen onto your line of sight but it it might represent some other object some oval object turn to such an angle that the projection of that oval object made a perfect circle on your retina you have the task of distinguishing of deriving of guessing of projecting of making some kind of hypothesis about what's out there in the world and what its shape in its distance is on the basis of this two-dimensional image a transcendent step in evolutionary terms in this process of trying to understand the third dimension is the development of binocular vision most animals in the world of of all classes and groups have eyes pretty much at the sides of their heads which is an immense advantage if you were a rabbit imagine for a moment being a rabbit you would have an entire 360-degree view of this room including what's behind your your head to be very little overlap in the fields of vision of the two eyes at the front and also at the back but you'll be able to see everything in between very useful you'd think if your primary concern is avoiding predators well amongst all classes of species some animals have evolved to bring their eyes full was pointing forwards sacrificing the huge advantage of panoramic vision for what for having the fields overlapping each other sharing a view of the world between the two I saw they had they see the same thing there might be very minor advantages in that in terms of the ease of detection of very weak stimuli discrimination of fine detail perhaps but it's quite clear that the major advantage is the ability to use the differences the tiny differences between the two retinal images to compute information directly compute information about the distances of objects what we call stereoscopic vision that was most directly revealed by the work of a Wheatstone in 1830s 1838 he announced the invention of the stereoscope a device which allowed him to project different images to the two eyes so he could essentially show to the two eyes any combinations of images including pictures that represented the views that the eyes would actually gain when looking at a three-dimensional object so here for instance in this illustration from his paper if you see at the top there the two prisons at the front through which the eyes will look or two mirrors through which the eyes were looking and separated looking at the two screens on the side look what's on the screen two to two drawings of rectangles with diagonals in between they represent the views gained by the left eye on the left the right are on the right looking down on top of a truncated pyramid as if you're looking down a pyramid from above because your left eyes further to the left you see more of the left face your right eyes further to the right you see more of the right face and the brain can use that information these minut differences in the retinal image to give a compelling and immediate impression of depth at least if you're amongst the ninety six seven percent of people who have stereoscopic vision this is stereo it's what you put your glasses or colored glasses on for in the 3d movie it's what gives you this amazing compelling impression of things jutting out and stick OUTFRONT surfaces so we are able to understand the these minut differences in the positions of the images of parts of individual objects simply resulting from the slightly different viewing points of the two eyes so here pursuing the argument of the truncated pyramid here's the left eye view the left the right eye view on the on the right if you were able to fuse those two images in a stereoscope you would immediately see the top surface sticking out towards you well how's that done well interesting that was Newton who first recognized an essential step in that process was to bring information from the two eyes together within the brain previously anatomist s' calm very competent anatomist even working with Newton actually in Cambridge had suggested that the two eyes as they project into the brain are kept completely separate yes indeed they come very close to each other and they touch each other in this structure called the optic chiasma just hear me behind the top of the nose but the suggestion was that the fibers from the eyes that simply project in to the same side of the brain well in that case they're not they're not talking to each other they're not communicating directly it was Newton who actually in the optics wrote the illustrations not from optics he wrote very clearly a single paragraph towards the end of the optics beautiful description of what actually happens heaven knows where he got the information from just an idea perhaps but it was a very clever idea his idea was then he chose in in this sketch from his Diaries discovered after his death that he was suggesting that fibers from one half of one eye project inwards to the same side of the brain but the fibers from the nasal the nut nose side of the eye go to the opposite side of the brain and the process repeated on the other sides of the fibers from the outer half of the eye cross so don't cross over the air to the same side and those from the nasal side cross over that means that this structure the optic tract contains fibers from both eyes and contains fibers coming from corresponding half retinas the nasal retina of the opposite eye the temporal retina of the eye on the same side and because the retinal image is inverted that part of each of the retinas looks out towards the opposite side of space so this half of the brain that's the right half of the brain is receiving information through both eyes about the opposite half of the visual field it seems curious but it fits with a general pattern of organization of the vertebrate brain that it's it's a crossed structure but each half the brain seems to be concerned with the opposite half of the world whether it's coming through the skin or through the ears or through the eyes so Newton said that this was necessary in order for the brain to be able to understand information from and compare information from the two eyes this shows a little more precisely and accurately how it is done newton actually thought that the nerve fibers from the right and left eye combined as soon as they meet in the optic tract they fuse that's not true they stay separate as shown in this wonderful illustration by the great anatomist cahal the fibers again he draws it correctly here fibers crossing over joining fibers that don't cross over and they stay separate here in in this structure the nerve cells receiving the incoming fibers that's a so-called lateral geniculate nucleus part of the thalamus brain crush cluster of nerve cells that sends information up to the cerebral cortex and it's here cahal suggested without physiological evidence that the information from the from corresponding nerve fibers left eye and right eye nerve fibers in the thalamus combines unto single points in the cortex and he was he was virtually exactly right so the prediction would be that if you could if you could record activity from nerve cells in this region the so called primary visual cortex the back of the the head a big area of cerebral cortex if you could record from nerve cells in that area you would expect them to respond to visual stimulation in either the left or the right eye and moreover if it's precisely organized it should be in the same position in space whether it's coming through the left eye or the right eye so that individual nerve cells could fumes inflammation could combine information about single objects in space and well that's that's that's true and and I was fortunate enough a long time ago to be involved in some of the early work on this topic in the early 1960s David Hubel and Torsten Wiesel then at Johns Hopkins moving a little later to Harvard did work that won the Nobel Prize for them 20 years later and they were they were the the first at least correctly to describe the properties of nerve cells in that part of the brain recording from anesthetized cats surprisingly large parts of the cerebral cortex particularly the sensory areas of the cortex still respond in anesthetized animals the information is being received and processed so they were able to record activity from these nerve cells in cats that were released the ties while they moved patterns around on the screen in front of the animal's eyes and what they found was two things principally first by comparison with the properties of the incoming fibers which just respond to spots of light or spots of darkness the neurons that they terminate on respond very selectively to lines or edges at particular orientations so there's a transformation of information and the very first neurons start the process of describing the forms of things in the retinal image in terms of the angles of the boundaries of edges in in the retinal image the second really important thing at least from from my point of view reading this this work when I was a medical student in in Cambridge is that the that individual nerve cells in this area often respond to stimulation of either I close one eye and the cell responds let's say to a vertical line close that eye and you find that the cell responds to the same vertical line roughly the same part of the visual field with the other eye so when both eyes are open and this is what you blue bezel said these neurons must be responding simultaneously through both eyes and therefore fusing the world making a single world well the problem of course is because the eyes are separated because the views gained by the two eyes are very slightly different as which Stern said that can't happen for every nerve cell all the time because the images aren't in the corresponding positions all of the time depends on on the third dimension so when I went to to Berkeley to work with them Horace Barlow great-great grandson of Darwin or in this context worth them well worth saying now at Cambridge when I went to work with Horace Bala who was then at Berkeley to do a PhD I was very interested in this characteristic that have been described I wanted to work on binocular processing and we started by looking at asking the question could these individual neurons be encoding and representing the 3rd dimension of space and the this is true of us that published the first paper on on this Jack Pettigrew Australian medical student at the time here is Horace what we did was to indeed confirm what Hubel and Wiesel had said we're looking now at responses here these are action potentials impulses from an individual one nerve cell in the cat cortex and you can see that through the left eye with the right eye covered up this cell responded to an oblique line moving across a blank screen when it crossed a particular position of the screen or the receptive field but the response was very weak in this case just a couple of impulses the same cell if the right if the left eye was covered in the right eye was opened now responded to exactly the same stimulus pretty much exactly the same stimulus in a roughly corresponding part of the field with the strong response and when they're put together the responses is enhanced and facilitated as long as the alignment of the images is precise so this must mean that if the animal is looking at a scene in which there are objects in different distances only some nerve cells can risk can be can have their receptive fields in the right position to respond to a boundary or a contour at a particular distance so we thought if different nerve cells have slight different differently positioned receptive fields different cells would respond to objects at different distances automatically simply by virtue of where their receptive fields were on the retina and the distances the scatter is in minut it's it's it's a few thousandths of a millimeter variation in relative position so something can easily imagine just just happening by chance in the wiring between the retina and the brain so we didn't we did that we took great precautions to ensure that the eyes of the animal didn't move the initiative we worked on cats the anesthetized cat didn't move because of course that would contaminate the results and then calculated what the distances of objects would have to be in order to produce the optimal response from each cell we recorded and here's a result from a typical animal this is that a sample of neurons these are each number describes the number of the cell in the sequence as we were recording in in the brain and you can see a sample of neurons here from a from one particular cat's lucky cat 13 for us and the properties these neurons are reconstructed as if the cat was fixating was and its eyes converged to fixate at 50 centimeters away and each of these points then represents where a particular stimulus would have to be placed in space to generate the optimum response from each of these neurons each neuron responds best to a particular position and there's a whole scatter of them covering a range of visual space just what you need to decode the information about the third dimension there's a little problem here there are several problems one is it's the signals from any one cell aren't telling you anything precise signals from if two different cells are firing let's say some cell nine and so for fire off simultaneously the brain knows that there are two objects which are not in the same plane there are different distances from the eyes but the separation of them which is surely the crucial thing will vary with the viewing distance of the animal for the same neurons and that is because the sure queue is simply relative position on the retina and it's shown in this diagram here if you think about looking at two objects separated by a particular distance always the same distance but the nearer object and is that different distances from you the so what you would like to find in the brain is a mechanism for encoding that the relative distance linear distance between two objects so you could say from the singing in your brain yes that's two objects separated by let's say 10 centimeters whatever the viewing distance but the problem is that the disparity the angular disparity generated between these two points varies hugely with viewing distance a very long nonlinear function accelerating very rapidly as you look further away so this this separation here which is the same as depression here has a tiny angular difference in the in the left eye and in the writer I compare with this one same at the same linear distance but a huge angular separation so if the neurons in the brain are just looking at positive position on the retina they will not be capable of signaling this linear distance in in space and indeed there's a lot of evidence in the literature that we are not very good at recognizing relative relative distance in space that actually to summarize a whole lot of experimental work the brain seems to default to an assumption that everything is basically 80 centimeters away which is interesting because 80 centimeters is sort of extended arm's length it's not a bad default if you haven't got any way of compensating for viewing distance compensating the disparity information that's coming in directly to the cells in your cortex with knowledge about how far away you're viewing which which could in principle allow to correct it there's quite a big literature saying that at least for individual points and objects we are very bad recognising the linear separation and we just assume that everything is about age two meters away and and compute the separation on that basis whatever the actual viewing distance is well this would be unfortunate but it occurred to me that you know we don't live in a world of points visible psychologists and even physiologists are fascinated with how the brain responds to blank fields of light of different color and individual points of the world it's not like that we live in a world of surfaces the surfaces of objects and shapes so I thought maybe the brain does better at represent representing and encoding the angles and depths of surfaces than it does with points nice picture by Hockney here and straighten be the powerful cue provided by surfaces so we did and when I say we here what I mean is the postdoc who did most of the work did you know it's the kind of Royal we this is this is work Yong Yong Zhao um we asked people to look at a screen with two lines on it this is a projection screen TV screen if you like tilted at a particular angle this is what's on the screen simply a couple of lines viewed through an aperture so you didn't know what the true angle of the screen was no information about the angle except the the depth and separation of the the lines and we asked we you know neuroscience these days is extremely sophisticated with massively expensive equipment this is our version of it yes the subject looked at a protractor and rotated the protractor to match what they thought was the angle of this surface they were looking at so they just saw two lines through an aperture using binocular vision stereoscopic vision they saw an apparent surface between the two lines NAB said if I was looking on down on top of that what would the angle be they set this now it sounds crazy judgment it is remarkably easy to do and the data are very compelling so the question is if you do this at different distances can the brain compensate for viewing distance and tell you accurately what the angle the surface cheese and what led me to think about that was that you know if you look at let's say a piece of paper a book or something like that and move it backwards and forwards to different distances it doesn't seem to change its angle as it should following the function that relates disparity to relative distance it seems to be a book until you get out to perhaps several meters when it gets really hard to tell that there's only depth at all okay so we were able to look at the effects of distance and here's a typical example these are for three different viewing distances 39 centimeters 70 856 and this is the angle of the actual stereo surface on the computer screen and this was the matching angle set on the protractor and you do pretty well whatever the viewing distance no defaulting to eighty centimeters and this shows again another example the results this is showing the perceived angle the peseta that is the protractor setting if you like for a real 45 degree angle of the screen for several different viewing distances this is the function you'd expect if it's just following pure disparity on the retina and it's defaulting to 80 centimeters this these are the actual data doing pretty well out to about 6 meters then collapsing very very long viewing distances so look so the brain does have a mechanism for compensating for viewing distance so it's accurately interpreting the angles of surfaces correcting the disparity information by knowledge about viewing distance well it's a good start and in fact him it was an essential start for other experiments which I'm going to describe in a moment so animals with binocular vision have the luxury of stereoscopic vision it must have been a really important evolutionary advantage because it's been selected for despite the loss of parameters it occurred quite late in each in each group and easting million animals particularly and there's very good evidence that it evolved the interpretive mechanisms in the brain for understanding stereoscopic vision evolves separately from mac mechanisms that might be based on single retinal images just the retinal image to find them out to compute what the world might be like now if you look at the organization of the the cerebral hemispheres of a primate this is in fact a monkey and old world monkey the human brain doesn't look that different and it's organized very similarly this is the back of the brain here this area called v1 shown diagrammatically here is the primary visual cortex that's the area that Barlow and Pettigrew and I had looked at and discovered that neurons can respond to particular disparities in space there's then a great tangle of other areas a network perhaps 30 40 of them in you stretching forwards occupying maybe a third of your entire cerebral cortex all pretty much devoted to recomputing the visual image all the information comes into here into v1 there it is but then he gets shunted off to these different regions who do their own computing job you know some of them more concerned with vision others more concerned with faces or object form and so on more others more concerned with movement certainly and interestingly the further analysis of stereoscopic distance seems to have occurred in what's called the dorsal stream the stream that's mainly concerned with recognizing movement and providing information to the motor system for the control of movement quite separate from the so called ventral stream going down here into the temporal though which is certainly more concerned with recognizing objects faces houses forms and things interesting because of course you know recognizing faces and objects itself depends on stereoscopic vision to some extent the way that you're able to see the full form of an object try looking at object in closing one eye it looks really rather different but it means them that that stereo is being handled by a separate part of the visual pathway that's shown in this picture here this is an imaginary flattened view of the human cerebral cortex with different regions corresponding to monkey areas here and you can see hotspots these two yellow hotspots are generated in the people in the scanner in the MRI scanner when a random-dot surface suddenly generated a stereo increment a bit that bobbed out towards the viewer in the looking at it on the screen in the scanner so the appearance of a pure stereo three-dimensional pattern elicited an activity principally in this area here and also in the sub region here these are parts v3 and and MT are parts of the dorsal stream here V 3 V 3 and MT here heading up towards the parietal cortex distinct regions which seem to be devoted to the analysis of stereo vision right that's fine for binocular animals but you know rabbits mice and so on do pretty well that flies at a dealing with the third dimension even though they don't have stereo vision and people who that the 3% or so of people without normal stereo vision also do pretty well introduced and well-known sports by the way and if you close one eye I mean just try it just look at this room with both eyes open and if you have stereo vision yes it has this sort of graphic quality of depth as you close one eye it isn't as if the whole thing just collapses and you don't know whether these people here are closer than those people there you still have a sense of depth but it's it's almost as if the colors the the stereo color if you like has gone out of it some additional quality to the depth is its lost if you loo stereo vision but you can still see relative distances and in some circumstances it's an essential skill I mean his written example this happens to be a photograph of the British ambassador's residence in Chile just showing off my adventures in the last year or so but now I thought it was a really nice example of the power of monocular cues you are looking now there's no stereo in this you're looking at a picture of photograph on the screen and yet it's full of depth information that just absolutely no doubt but this thing you know is further away than this thing and it's provided of course by all kinds of computable information in the image itself particularly by perspective by the fact that the lines here converge towards a particular point parallel surfaces have a share a single vanishing point all the rule the classic rules of perspective your brain is apparently able to use them or use something approximating to them as you'll see in a second the rather there's other information as well the texture in the carpet gets smaller the size this chair is smaller than this chair so if you know that chairs are always roughly of the same size then you can compute from that there's lots of rich computational information in a single image that could tell you about relative distance and there's been almost no research previously on this aspect of depth perception even though it is by far the most important one to head to animals that don't have stereoscopic vision and presumably in us it has to be integrated somehow with the stereo information so that's an interesting computational challenge artists as well of course have been intrigued with the same issues because they have the inverse problem the problem of wanting to create the impression of distance on a flat surface so in a way the painting is like a retinal image if the painting the flat painting can provide you with you in your eye with a retinal image which is identical to the image that would be created by looking at a three-dimensional scene and the painters hope is that you will see that scene with its three dimensionality well you know the story of course perspective was supposed to have been discovered at the beginning of the 15th century in Florence Brunelleschi in particular in fact it was a longer and more gradual process than that with some empirical steps towards understanding the rules of perspective and bit longer for Brunelleschi here though in the 13th century this is over this is absolutely characteristic example of the effort to try to represent the third dimension without an understanding of perspective so even the relative sizes of things are wrong the distant ones aren't consistently smaller than the near ones there's no perspective from information it's just a best guess to try to interpret this image principally in terms of overlap this head is definitely closer than the couch or whatever it is because it's overlapping it so they're accuse but not very accurate or convincing Giotto so you know so so far ahead of his time in so many respects but Giada was getting close even at the beginning of the 14th century if this in this it's example Christ before Caiaphas if you perform a perspective analysis showing the lines of convergence here you see that he's got the roof pretty right the vaulted roof with a an almost convergent single vanishing point things go a bit wrong here these should be converging too but they're not so it's a it's a mixture of half accurately tenants and just my eye I presume a perspective without clearly without understanding of the geometry and it was supposed to be in Brunelleschi V our architect of the Duomo in Florence and of the baptistry and and a painter of some repute who was supposed to have solved the problem in the sense of deriving some kind of formulaic approach that formulaic approach isn't recorded there's no record of any mathematics or any geometric construction all that we know is that it probably in about fourteen thirteen I think the estimate is he presented a demonstration of the fact that he could produce paintings which were a perfect replica of the projections of solid objects and the solid object in particular that he was concerned with was his baptistry that he built so here's a sketch of the baptistry as viewed from the entrance the porch of the duomo which is the point at which Brunelleschi painted a picture of the baptistry so he stood in the in the porch of the duomo painted on a wooden panel painting of the of the baptistry and then used that painting to demonstrate that he got it right geometrically and the way that he's supposed to have done it which can't actually be true but but this is what he's supposed to have done you can see the principle is he demonstrated that if you now stood where the painting had been painted turned the painting around so the picture was pointing towards the baptistry and peeked through a hole drilled in the panel so you could see the baptistry through the painting of about but then held a mirror up in front of you to look at the painting alright that everything matched you could move the mirror around and look at the painting it would precisely align with the boundaries the baptistry there are two problems with that one is it depends on the distance of the mirror really very critically you could you could make it match whatever it's of its appearance just by moving the mirror little bit secondly the image is mirror reversed well it doesn't matter with a Baptistery because it's symmetrical it wouldn't work with anything else it wasn't a terribly good experiment people got terribly excited and he was admired even at the time with the discover of a remarkable method that artists could use to represent three-dimensional form and there must have been rules because his in the fourteen 20s his masaccio also in in florence painting in the bronco chapel and clearly understanding the rules of perspective so here's a perspective reconstruction of this building here on the right and not only is it correct the vanishing the there's a single vanishing point all the lines converge but it is in the head of Christ and arches were already discovering that they could use the power of the imagery of perspective to emphasize features obscene and this became a favorite motif in in Renaissance painting to make the vanishing points of apparently irrelevant features of a scene like buildings draw the eye as they imagined and towards a crucial feature of the of the scene this is 1440 25 or there abouts the from the formulation of the rules and the publication of the rules is attributed to this man Albert II who published two books him in Italian and in Latin just a year apart 40 35 in which he described very precisely and the geometry of projection it's simply it's simply a matter of optics and geometry if patterns in a three-dimensional scene are imagined passing through a flat surface and you could record where each ray of light hit that flat surface that would be a perfect picture of that scene and if you were to view it then it should reproduce in your eye exactly the same image that you have if you look at a real three-dimensional scene that's the notion so he described all of the rules including things like how the spacing between equally spaced objects like tiles on a floor and so on and this propagated incredibly quickly through the painting painting and community well one of the questions that has not been addressed or had not been addressed is how the brain under the brain understands perspective clearly we can see form from perspective alone but how do we do it imagine that you look at a computer screen like like this with some lines drawn on it these thin black lines nothing else gray light gray screen with thin black lines you look at it through an aperture around aperture so you know nothing about the true form of the surface you immediately see an apparently angled surface generated by the geometry of the lines if indeed the geometry matches what it should be for a particular angled surface so you can compute what these lines would be if they were on if you imagine lines on a real surface which is rotated away from you the angles between the lines define the rotation of that surface with respect to you well how good are people at knowing the angle of that imaginary surface purely from a very simple pattern of lines that's a question that we asked the experiment we did was this and the only interesting feature of it is that we devised a way of measuring very simply quantitatively the perceptual impressions created by monocular images including just sets of lines on the screen what we did was to show the this is a large computer screen here and one part of it here it's viewed through a stereoscope so one part of it here is viewed by the left eye and on it is a pattern of lines creating an apparently rotated surface pure perspective information viewed monocular Lee no disparity no stereo but in place on top of the top top of it are three little dots there are three little dots on the other half of the screen which is viewed by the right eye and these dots are organized so that they form a stereo red they're seen binocularly and you can adjust the spacing of the dots in the eyes so the two outer dots can be rotated around to different positions in other words forming a stereo surface so you've got in the in the person's head two sources of information about the surface they're looking at the information from the perspective lines seen only by the left eye and superimposed on top of it the information from the stereo dots which can be rotated around in space until they match so you can use the stereo setting to measure the apparent perception of this perspective surface it turns out to be very easy to do reproducible and return the angle of the surface in any way you want to and the subject can be the person can very easily just set these dots too to match it it depends the principle depends on stereo sensations being accurate but we know that they are from the preceding experiment from the stereo constancy experiment so we can use stereo as a veridical readout for the way in which people are perceiving information from sets of lines and here's a typical example this is the angle of the surface represented by the perspective lines ranging from 70 degrees tilted in one direction to 70 degrees tilt in the other direction and here the angle of the stereo surface the three dots that matched it what you'd expect if everything is perfect if your perception from perspective is perfect you'd expect the stereo and the perspective depth impressions to line out of precisely along the dotted line they don't but they do follow a nice function it's always less than the expected value of the stereo settings were always less tilted than the geometry of the perspective surface and would predict this means we under perceive we under perceive the till the three-dimensional rotation of surfaces when the only information we have available about them is perspective this is true in this room I can reach you in a second is true anyway buildings rooms whatever we under perceive the at the the tilt of the slope of surfaces the degree of the you can express the error as a gain the the magnitude of the tilt and angular gain the magnitude of the tilt of the stereo surface needed to match a perspective surface a computed perspective surface and the gain varies from person to person and it very interesting ways these are I don't know half a dozen or more different observers this is the gain and you can see it doesn't reach one for anybody always the gain is less than one they're always under perceiving the perspective surface and this here is the number of lines on the screen providing the perspective information now we thought that surely no more confirming evidence you put in like an artist board the more lines you put on the surface of a building surely the more powerful the perception of the tilt and the depth would be and it's not two lines is quite enough to line two little thin black lines on the screen which are not parallel to each other immediately generate the full impression of depth you're ever going to get from perspective it's not the impression you get is not the true impression but it's certainly not improved by adding more now I think this tells you something about what's happening in the brain I think it must mean that the brain the mechanism wherever it is and whatever it is depends on the brain doing a computation based on neighboring lines and only neighboring lines not taking into account others I'm just saying hey there are two lines there that are not the same orientation I'm going to assume that those two lines belong to contours which are really parallel to each other in the real world like you know the top and the bottom of this piece of paper if I make that assumption and I can see that there's a difference in angle between them if I know how far away it is that I'm view and I know the angular separation of the lines I can think compute what the surface is that's just geometry well and the brain sort of does that it gets it wrong it under calculates but that's the that must be the computation that's that's going on I think and it doesn't depend on additional lines doesn't help now what about what artists do and what about the real world like this mean well we have much more than just their lines telling us about form we have people of different sizes and we have texture on the floor and all the other things how do they integrate do they actually enhance and add to the impression of depth are we computing perhaps in a kind of Bayesian we're using the best possible individual features to nudge Assad towards the most accurate interpretation and we decided to look at that using identical techniques so what we did was to ask whether artists might be able to improve the gain of monocular depth perception by throwing in more cues in their paintings but taking into account that when paintings are actually viewed in a gallery their view with both eyes open normally and they are actually flat so stereo information in a gallery is always contradicting what the monocular information says the painting says you'll hear the senior at some piazza and their buildings here your stereo system you says now don't be fooled by that this is just a flat painting how's that solved is there a conflict okay so what we did was to repeat our experiments but using bits of paintings we found parts of paintings with reasonably flat sloped surfaces we could compute from the geometry what that surface represented in depth what the angle of that rotation the geometry actually represented and we could ask people what they saw by giving a stereo comparison and just asking them to rotate the stereo lines until it matched and read out from the stereo what they were seeing did that were the number of them well-known paintings or all kinds of examples of pictorial representation with very strong monocular features of slanting and sloping surfaces photographs for instance and so on some of the experiments were done monarch at least only one eye sore the painting but here's the stereo readout below in this case it's below the the the image of the person rotated in depth these three lines to make them match the apparent surface slope of the surface and this is done monocular but we could also do it binocularly as a viewing a painting in a gallery show the two pictures identical no stereo information about depth it's just flat and but let's see whether the stereo matches that they give down here are not flat if they're not flat it means that the monarchical information must be overriding the stereo impression created by the binocular view of a flat surface and here the and we compared that with just our standard lab experiment and the same people in the same experimental run with just lines on the screen now asking them again to match with stereo this is typical result this shows the the slant the virtuals of the slant the rotation calculated from the geometry of the monarchy of the of the monocular cues of the perspective cues the rotations of surface in the painting the person was looking at and this shows the matching the stereo match to it and the oak crosses here show the results for our simple lab demonstration with just lines on the screen and it shows the typical nice variation you match the stereo to the perspective but you under match it the gain is less than one I've described that before but now the dots the colored dots refer to paintings or photographs rather than lights and they're viewed in two conditions monarch you delete the orange dots where you have no conflicting binocular information tell you it's flat and green when it's viewed binocularly and what you can see is immediately viewing it binocularly reduces the gain even further it makes it look flatter but not completely flat so there must be some interaction some quantitative interaction between disparity information perspective information and neither one completely wins but if you look at these the orange dots these are viewing monocular only where you're not just viewing lines neutral lines on the screen you're viewing a rich painting or a photograph and the gain increases so yes there must be some integrative process which is combining monocular cues to enhance the perception of distance stereo can contradict that process but not completely eliminated and summarizing those results here's the pure perspective image just the screen with the lines on it here's the effect of monocular viewing of real paintings in increased gain and here's the effect of viewing them flat and binocular reducing the game right so artists could enhance the perception of depth by not only by adding more information about distance but also by deliberately breaking the rules any art student and I would guess any art student from 14 35 onwards has learnt the basic rules of perspective they all know how to construct a perfect kinetically perfect construction with vanishing points and song breaking the rules would mean disobeying what you've learnt and recognizing that by by altering the geometry of the image you could produce an even more enhanced impression of depth and I've found a few examples of artists doing that but to do it just remember they have to overturn all this at the amazingly simple way of constructing very compelling images which Brunelleschi's work and other said had taught them so let me give you one example this is a Google Earth view of the Spanish Steps in Rome yeah I'm here the steps Keats lived here the last year of his life and there's a Keats Museum there now and Keats was a keen amateur painter and draw and he artists and he drew a rather nice compelling picture of the Spanish Steps and the PR above from this point here just in front of the fountain now here's a photograph 35 millimeter camera correct proportions photograph taken from that point and you can see here it is and I think you know that you'll see what I mean they do that can I ask you do you think that these surfaces are actually parallel to each other does this look as if those two surfaces there are actually precisely parallel to each other now to my eye and my gain is about 0.5 they don't look parallel they look tilted in which which is just what I'd predict from my my gain less than one well interestingly this is Keats and there's a perspective construction by the way to show you that they are actually parallel to each other this is a photograph of two roughly parallel surfaces the single viewing point they don't look it this is Keith's drawing and I think you'll see that what he's done is to tilt the buildings upwards this slope here is more extreme than this this is coming this appears to be converging in words that's sloping more outwards and this one too so if you do a perspective reconstruction on Keats it's not correct I think he's just done it by eye because it looks better that looks more parallel than this does even though this is real and that synthesized okay so has any artist deliberately use that to make his pictures look more real I think this is this is a nice example and by the way this isn't wonderful van Eyck painted in Northern Europe one year after I bet his book already the information had spread everywhere everyone use rules so if you look at the wonderful floor here it's drawn in perfect perspective our beti perspective and you can show from the reconstruction of the geometry of the perspective how precise it is with a vanishing point in the womb of the Virgin immediately symbolic use of the vanishing point but look at the vanishing point for the canopy above her head that's in her and perhaps in our heart it's not sharing the same position as the vanishing point of the floor which it geometrically should do if the canopy is parallel to the floor which presumably it's supposed to be I think what van Eyck has done is to shift deliberately shift apart the vanishing points to make it look more right to make it look more compelling breaking the rules of perspective which must even then have been hard for an artist to do because the rules of perspective are so amazing and saying generatively amazing okay I'm nearly finished so now the final question is if we don't perceive perspective quite correctly in paintings or on computer screens do we see correctly in the real world where everything is combined all the informations combined accurate the stereo the relative sizes the perspective and everything and if we don't see a perspective does it interfere with you know with the with the aesthetics if you like of our experiences well let me and this is my starting point when I noticed this feature this is perhaps the most famous Piazza in whether it's the company doll you the original northern entrance just north of the forum - classical Rome and the pattern Hill and it was a Michelangelo was given given the task of creating it there were two existing buildings this existed the palazzo here and the library biblioteca here already existed this space was free but he was constrained by the fact that the church sitting here on the edge of the hill so many have argued you know when looking at this view it's very odd why did why these lines converging so stupidly you know it's supposed to be a nice Piazza and people are wars argue well it had to do with the fact there's another building here he couldn't squeezy thing he had to rotate it but I think it's more than just that and there are some wonderful features of the design of this that suggest very strong that it was all deeply thought out the fact for instance that this amazing pattern in the courtyard at the Piazza which wasn't by the way completed to Michelangelo's design until the Second World War misil Eenie commissioned the completion of the floor but this shape here forms a perfect circle when viewed from the viewpoint of the top of the pallets are absolutely perfect well what about this oddness here imagine you're coming up the steps the so called them coordinate are coming from the road below coming up for the first time to see the Magnificent view the palazzo ahead of you what do you see you're standing here you look upwards and you see this this is the view and I would posit that these two buildings these two surfaces do not look as if they're converging towards you I mean to my eye they look pretty nicely straight in parallel if on the other hand if you're coming out leaving the view that you have is this obviously you see the strong angular convergence because the building's really are converging in that direction and it's adding to your perceptual misinterpretation but going in everything looks aesthetically much more satisfactory it turns out the angle of this which is about 18 degrees exactly compensates for the average perceptual under-under perception of from perspective you might say that was just chance but look at this this is the Vatican the the Piazza San Pietro in front of the Vatican Michelangelo had worked on the dome with the Vatican he died he died actually 20 years before the completion of the Campidoglio and Bernini took over the restoration of the interior of st. Peter's and the design this amazing design of some pita squares I'm showing many of you beautiful very dramatic innovative oval design the plan was prepared in the year and after the opening of the Campidoglio the last building was completed big celebrations and panini them worked on the design this will look at the entrance the colonnade entering the approaches and Peter's it's angled and actually it's even a slightly asymmetrical like the Campidoglio and so therefore when you're you're here and you're looking towards some Peters the view you have of this and I think those look pretty parallel to me if you're leaving if the Pope's viewers you're leaving is like this the the angle is exaggerated but what matters is when you're approaching really you know when you're leaving you've been there done that you're not so concerned about the aesthetics all right this is the Campidoglio which was could this building here the Palazzo morville was completed to Michelangelo's design one year before the design of this different slightly different scale this is bigger but virtually precisely aligns in proportions and in angles even the asymmetry actually this is less tilted than that there were there were previous examples of trapezoidal piazzas preceding michelangelo there was not the entire entirely original development and all every trapezoidal piazza had been able to find a record of has the same as it were distortion it's always turned inwards towards the viewer as they look at the principal building it can't just be chance and that's the view have this church in piensa and we're still doing it I went to a meeting in the South of France last year and visited this amazing if you have the chance to go do just north of Marseille there's a beautiful vineyard which has been meticulously developed by an Irish family that owns hotels in their name but anyway it's it's not just a wonderful vineyard it's scattered with the most extraordinary works of art they commissioned major architects and sculptors and so on designers to produce pieces of art which are scattered around the vineyard including the pavilion de exposición bill designed by Renzo Piano the designer of the shard and it's an exquisite building it doubles up as exhibition space and as a wine store you don't see the wine store it's underground on each side but you enter you go down this path down the low ground normal ground level and as you approach this exhibition space you go through glass doors four panes of glass you look through into the space and you see four panes of glass at the other end going out onto a balcony and you look out across the vineyard right now again I don't know what your eyes will tell you but my eyes tell me that this is a rectangular space those two walls are parallel to each other therefore these four panels are the same width absolute width as those four in fact these panels are more than twice as big as these panels twice as wide and if you view the building from above you see why here it is you're at this point here looking in there's the design you're here looking in its and notice by the way the asymmetry same asymmetry is in the Campidoglio here it is so campy Dario Renzo Renzo Piano superbe I've been trying to get in touch with his piano he was in London until three months ago and I couldn't reach him to see whether this is deliberate or again just just chance but I think this is a device and it's an aesthetic trick and with the architects understand better than we do the perceptual base of the trick I don't know but it seems to be very unreal right one one minute more because this just shows the power of art schools you learn this wonderful these wonderful rules of perspective you apply them and you make beautiful pictures but of course if what you're drawing is not rectilinear ly regular the rules are wrong so here's a very nice example this is Canaletto this can let those painting of the Campidoglio so we know the Campidoglio is not rectangular right we know that this building is not parallel to that building there sloping inwards total of about eighteen degrees but he has painted it as if they are and here also interestingly the viewing point and you can see why he did this it might have made the mistake because the imaginary viewing point from which he painting this picture is sixteen meters above the road below he could not have been at this level painting this picture he must have drawn sketches come back to the studio and reconstructed it following the rules he learns at art school but he reconstructed and an imaginary and false Piazza not the real thing all right just to conclude there's something very interesting about perspective it's compelling it's wonderful artists love it but I think that it must be something we learn because the natural world is not full of parallel lines there are no parallel lines in the natural world and yet the whole the algorithm for computing perspective is based on the assumption but when the brain sees two lines in the retinal image that are not the same angle they are derived from contours that are actually parallel to each other therefore the angle tells you about the rotation so it only works in a world that has real parallel lines in other words a carpenter's world I must say man person made world it's a world that we've put together and based of course on the fact that you know because of what John but what gravity does it's much easier to make things that are parallel and straighten that stand up with straight lines and it is curvy things but we all have to learn it I think the children have under some evidence for this learn to use perspective very gradually from learning from their image well that's fine as long as our experience is accurate in telling us the rules and the problem is and architects are detaching themselves so efficiently from the constraints of gravity this lipscomb building in Singapore this extraordinary building are on City Road if you if you know the m building everything's wrong I mean this it's ridiculous this point is very much higher in reality than that point and it creates the impression of the building being much longer it isn't look at when the windows are there all over the show look at this these there's a to building side by side those surfaces are actually angled equally with respect to each other so mind-blowing for the perspective processing and not to mention this this is Frank Gehry and now and that's the interior of that they're building so I've tried to give you a taste of the way in which actually still very simple methodology can still provide information in interesting information about vision vision is such a rich resource of of questions and answers and I'm also to show as I said at the beginning the way in which the history of art in architecture are interwoven really with the history of the science of the perception of space and finally just to acknowledge my collaborators and sources of funding and so on thank you very much for your attention [Applause] so Colin thank you very much indeed that was just fascinating now as regular visitors to the attendees that these lectures will know usually at this point I announce highlight the following leak week's speaker but you know all good things come to an end and we've come to the end of this this year's series but I can highlight the fact that each week's lecture is filmed and then stream do you can access it from the Darwin College website this one will be should be available early next week also each series of eight lectures the lecturers right write that up and it's edited in a book so here I have read a book launched last week last evening this is the the book from the extremes lecture series with with least asset and David Runciman and Rawls savage and Nassim Nicholas Taleb in so on table outside in the Friday's there are copies of the books from the from extremes development and games lecture series if you're interested in those but what else do I have to do I mean we've had annex fascinating a superb set of lectures this year and I'm sure that the series next year will be - and I know that some of you have been puzzling as to what next year's series might be I'm trying to extract information from those of us who know but there's no need to wait any longer because I can give you a preview hopefully these are previous titles of lecture series none of these what is it starts on the 17th of January every Friday in the Lent term you have the titles and the speakers what's the theme in Ignace so we'll see some of you many of you perhaps 17th of January next year thank you very much [Applause] but oddly just thank many thanks to Colin Blakemore again you you

Channel: Darwin College Lecture Series

Views: 7,648

Keywords: darwin college lecture series, 2019 - Vision, Colin Blakemore, Neuroscience, Art History, Architecture

Id: 3OqKIEekk8w

Length: 72min 32sec (4352 seconds)

Published: Tue Mar 26 2019