Lecture 1 - Psychology 2005 (2) - Transcript PDF

SPEAKER 0 Hello everyone, and welcome to the introduction to this series of lectures about visual perception. So if you were in class on Tuesday, you will notice that this is not the recording that we took on that day. Um, so I had to rerecord the whole lecture because there was an issue with the r...

SPEAKER 0 Hello everyone, and welcome to the introduction to this series of lectures about visual perception. So if you were in class on Tuesday, you will notice that this is not the recording that we took on that day. Um, so I had to rerecord the whole lecture because there was an issue with the recording that we did in class. So there might be some things that are slightly different. So before we proceed, I just want to introduce the part that I will be teaching on. And this is, um, a part about perception. Um, these will span over four lectures for a total of eight hours, which will cover very different topics, all about visual perception. My name is Domenica Pineiro and my office is room A13. If you want to talk to me, my office hours are on Tuesdays from 10 to 11. And one other final thing that you need to know is that all the sources that I use for the lectures will be listed at the end. As on the very last slide. Um, and one other thing is you will see that I am suggesting a few things to read throughout the lecture, and so you will find some links on the slides. And also there's some suggestions on the Moodle page. I do not expect you to read all of it. Um, I will just indicate to you which, um, book chapter or which paper, if any, is necessary to, um, review these lectures. So one thing that I taught, uh, this year might be useful is to have a Padlet. So if you have any question that comes to mind during these lectures, um, and you don't want to post on the forum, there is this kind of informal way of asking me questions. If I see questions on this Padlet. I will prepare, um, my answers for the next lecture. So use this Padlet if you have anything to ask during my, um, lecture time. So over these eight hours that we have together. Otherwise afterwards, you can use the forum. Okay. So now we start with the actual lecture. Um, you know that psychology, um, and psychologists are usually interested in understanding what's the link between how the brain works and behaviour. So we want to understand this relationship. And a key aspect to understand this is to understand how we perceive and interact with the world around us. So the question here is then how do we access the physical world? Um, so all of our experience, if you think about it, always will, um, will come to us and we reach our brain through information gathered by our sensory organs. Um, and then the information that is coming from our sensors will be processed by the brain. So this idea was proposed by Aristotle a few years back. Now, um, and he was the first one to recognise that we do not have direct access to the world, but rather we gather information through our senses. And it was also the first one to suggest that we have five senses. Now we know that in reality we have more than five. But for now we will stick to these classic, uh, distinction. So, um, if you are a psychology student and you have done cognitive psychology in the first year, you might know already that sensation and perception are two different things. Um, so sensation is strictly related to our senses. Um, so it's the process of taking information through receptors. Uh, receptors are cells that are able to translate information from the outside world into a signal that the brain can interpret and manipulate. And then from these receptors, then we can interpret these signals as images, sounds, smells and so on and forth. So this is an example to show you the difference between um sensation and perception. So in this image if you haven't seen it already you just see a bunch of dots. So you see black patches and white patches. Um, in this case our receptors in our eyes will detect that there are different, um, details in this image. So you will experience the presence of black and white parts. Um, so, um, if you look at this, all you see is dots. But if I tell you that there is a dog and there is a Dalmatian dog in this picture, your sensation will become slightly different. Um, so let's see. Um, so you can see here Now, um, that if I give you this information, it's clear that in this, um, picture, there is actually a dog. So this is just to demonstrate that we go from a bunch of dots to a full perceptual experience of a dog. So this is to say the perception involves a little bit of an extra step. The for perception involves the interpretation and understanding of our sensation. And the funny thing about this picture is that from now on, you will be, um, always seeing these Dalmatians. So there is no way that you can't, uh, see the Dalmatian in this configuration of dots. So the first question we have to ask ourselves is, um, is our perception of the world, uh, an accurate reflection of reality? So, philosophically speaking, we might tend to say, yes, um, at least most of the time, this must be the case. Otherwise we would not be surviving in our environment. So for example, if you think about crossing the road, um, it feels like we are accurate enough to survive these experience so we can cross without being run over. And also, um, if we think about everyday activities like recognising people from their faces, um, we feel like we rarely make mistakes. So we are quite good in our perception. Um, and also the other point is, um, that usually we tend to agree with other people. So if we're talking with someone else about what we are perceiving, we usually are perceiving the same thing. So we kind of find an agreement. However, our perception, which is a subjective experience of the world, uh, is not always matching the reality. Um, and the very famous example that we can, um, look at is the example of visual illusions. I know that you are very familiar with visual illusions, and they're are there to remind us that sometimes our senses and our perception is not so accurate. So let's see this very famous example. Um, so in this case, we have a misperception because we think, um, or we tend to think that these lines do not have the same length. So if I ask, um, if I ask you to judge the length of these two lines, you will tend to say that this one on the right is longer than the one on the left. In reality, however, if we take away the edges or these v um, or yes, these V's that are around the end of the lines, you will clearly see that these lines are actually the same length. So this is just to say that we perceive does not always correspond exactly to the physical reality. Okay. There's also um, some variability in perception. And if you are, um, if you follow any sport, this is, um, something that happens all the time. So there is disagreement sometimes about what has happened. So for example, whether the footballer has scored or not, this might be one case that has been um, under discussion here in this picture. So disagreement about something that might have happened or not highlights the fact that there is variability in the perceptual outcome. Um, and this is um, not necessarily due to the fact that some people might be more or less skilled in judging a situation or, um, might be paying more or less attention. So even if there's the same amount of attention, if we are skilled the same, there's still some kind of variability. So in general, what we can say is that under easy conditions, when stimuli are clearly presented to us, perception is reliable. But where conditions are not ideal, responses can become probabilistic. Um, so this is to say that perceptual judgements are not always perfect. So is there anything to explain about perception. So perception, for example, vision is not just recording information from the outside world. So it's just not like our eyes are a camera. It's rather an active process where the brain must interpret sensory information. And this interpretation is what guide us, um, in interacting with our, our environment. But because usually we do things quite automatically, so our perception appears to be very easy. There's no effort. Um, people might believe that, uh, there is nothing else to explain. However, as pointed out by Gregory in 1977, uh, if we just think about vision, this is just a little bit of a miracle because we start with a bunch of photons or some light, and we take an image that is projected, um, at the bottom of the eye, and we make up a full, uh, perceptual experience. So this is not something, um, easy. So perception feel easy feels easy because we have so much, um, of our brain that is specialised and devoted to processing information. And this is just an example that is about visual cortical areas. So you can see here this is the the cortex. So the top part of human brain, um what you can see is that there is a series of areas. So these names it's not important that you remember these names, but this is just to tell you that there are so many different areas that are devoted to process visual information. So in this case, if we take the visual brain, we have more or less, Um, 30 areas, which is, uh, almost 50% of the cortex. So we have millions of photoreceptors in each retina to perceive light. We have millions of cells in the cortex, and all these cells communicate with each other with thousands of connections. So perception feels easy because we have specialised machinery to make these happens. So this is just to show you how complicated it is. Even if we just look at visual perception. So all these squares are representing one area that is more or less or somehow at some level, um, involved in visual perception. And you can see that there is an intricate pattern of communication across all these different areas. So this is just to say that our wiring is extremely complicated and this is what makes, um, perceiving so easy, because there's so much of this, um, machinery that we we have. So why do we need to study perception? This is an excellent question. So the first thing that I would say is that perception is where everything starts. So if you are, um, interested in any cognitive process or memory decision making, um, everything will start with perception anyway. So if we study perception, then we will better understand any other cognitive function. And we can understand how sensory inputs are processed and transformed in what we perceive, um, as a conscious experience. Studying perception can be very useful also to improve how we interact as humans with technology. So for example, if you think about designing an app, if you know how we, um, we perceive things, you might come up with very user friendly solutions. Um, rather than just guessing what it is that is best in terms of, um, usability. And then we can also enhance medical diagnosis and treatment because there are disorders of perception that can be better understood and treated. If we know how perception works both in normal and abnormal conditions. Another point is that if we know how perception works, we can also develop, um, artificial intelligence systems. Um, so you can think about Alexa or Siri, for example. They're all based on how humans perceive. Another reason for doing this is that if we know how we perceive that, we can improve educational learning. So educational techniques can be optimised if we know how people perceive and process information. So, for example, learning materials could be designed to align with how the brain process visual information, and these should make the material more accessible to, um, to learning. And finally, we can improve our everyday experiences. So what I mean by that is that, um, there are studies that have practical application, um, in areas like ergonomics, product design, marketing. So by understanding how people perceive things, we can, um, optimise things for better usability, safety and even, um, aesthetics. Okay. So let's see just a couple of practical applications. So discovery and perception have been used um, in real life. So for example all colours can be seen by mixing three basic uh primary colours um, and the television uses exactly the same principle. And also, as I already mentioned, artificial intelligence like Siri, Alexa, or Google Assistant will process auditory information in a way that is very similar to how humans perceive speech. So the most studies studied of the five sensory modality is vision. And I guess if you think about how much of each of your senses you're using, you will realise that we as humans, we heavily rely on visual information more than sound, for example, or smell. And so if you imagine that you are in a place where you have never been before in your life, and you are asked to choose one of the senses to explore and become familiar with these new spaces, I'm pretty sure that if you stop and think about it, you will choose to use your vision. So this, um, next, eight hours will focus primarily only actually on visual perception. But many of these principles of information processing will be, um, still valid for all other sensory modalities. So there are different approaches to studying perception. So what I'm going to do now is I'm going to review some of them. And the reason why I picked these is because you will encounter these, um, approaches when you are studying for this module. Um, and when you're doing your reading. So the first approach I would like to discuss is psychophysics. So in general of course studying perception because it's so complicated is something that will require people with very different backgrounds. So for example, psychologists, physiologists, people who are um expert in anatomy, engineering and also computer science. Um, so let's focus now on psychophysics, which is the study of the relationship between our strongest sensory experiences, which is the cycle part, and how strong the, um, the stimulus is. So we want to relate the strength of the stimulation that is coming from the outside world physics to our sensory experience. So this is the oldest branch of experimental psychology, which was developed in the 1800s by Weber and then some of the techniques by Fechner. So subjects usually in this case normal um, human um, are asked to report whether they see the presence of, they see um, a visual stimuli. So they're asked to detect the presence of a sensory stimulus. Um, and this is a way of measuring perceptual performance of the, um, human brain. So the central idea is that we want to measure, uh, The limits of perception or uh, as psychophysics would say, um, thresholds. So there are different thresholds. One is detection threshold, which is the weakest stimulus that reliably evokes a sensation in the observer. And then the other one is discrimination threshold. So in this case we present to stimuli to a, um, our participant and we ask them to discriminate. So the question is how much different, uh, do they need to be in order to notice the difference? Um, so in this case we, we call this discrimination threshold. There's just not noticeable difference or J. And so the idea here, what we want to do is trying to understand perceptual mechanisms by looking at how these thresholds might change. So in typical experiments what will happen is that we measure threshold um in neutral condition. So let's say at baseline. And then there would be some kind of modulation that comes from dark adaptation, for example, of motion perception. And we see how this threshold is changing. So there are two main methods that are used. Um, and they're still coming from Fechner. So from the 1800s. The first one is the methods of limits. Um, so what we do is we ask participants to adjust the visibility or the intensity of the stimulus, um, until they can no longer perceive the stimulus. So we look at the limit of their perception. Um, and the other um method is called methods of constant stimuli. In this case, we present participants with a number of stimuli, um, that are slightly different in a random order. And every time that we present the stimulus, we ask them to report whether they can see it or not. So here on the right you can see a psychometric function. What you can see on the x axis is how strong the stimulus is. And on the y axis how many times participant we say yes, we can see something for each of these stimuli. So you can imagine, for example, that you have a grey background and you are presenting something like a dot. The dot could be of the same colour or very similar colour as the background, so grey and grey it would be really difficult to detect. So in this case the stimulus strength is really small. Let's say one. And then you can increase the visibility by going from something that is very similar to the grey background to something that is black. So at some point you will have this black on grey. And in this case the strength of the stimulus would be um, Higher. Um, somewhere here. So nine times compared to the first semester was almost the same as the background. So what you will see is that the more, um, the stronger is the stimulus, say full black on a grey background. Um, the higher the probability that the participants are perceiving the stimulus. So this is what is plotted in the psychometric function. I will um, post some some examples on the Moodle page. So when we do this experiment it's a very good idea to have a forced choice task, which means that every time we present something we ask participants to say yes or no. Uh, we don't just say press whenever you see something. Um, or we can ask participants to report something about the stimulus. So, for example, we can show, um, a line, and we can ask them to report whether the line is still to the left or right on each of the trials. What we will do is we present stimulus stimuli, um, with a random interval, so participants cannot predict when in time the stimuli are presented. After the we present each stimulus many, many times, we can then calculate the amount of correct responses and therefore the threshold. So in this case, for example that we have. On the right is example we have um eight different possible stimulus stimuli from one to um sorry nine. So from 1 to 9 um. And for each of them we present each of these stimuli many times. And what we can see here, for example when this trend is one. So imagine a greyish tone, a grey background is really hard to tell whether there is something presented Participants will perceive will say, yes, I've seen it. Around 51% of the times. But if it's a full black dot on a grey background, they will say, yes, I've perceived 100% of the time. So by convention, the threshold is defined as the stimulus intensity. That will give, um a perception. So when participants say yes, I've seen I've detected the stimulus 75% of the time in this case is, um, the stimulus strength of five. So let's take a look at this psychometric curve. Is there anything that you can notice? Something, anything at all that is weird. So one thing that I want you to think about is that there is no stimulus value at which there is a step change in response from no to yes. So the the line here is what happens in reality. The dashed line I just superimposed to show you how this would look like if we had one single, um, stimulus trend that would go from no to yes. So in this case, in this hypothetical scenario, when the stimulus trend is from 1 to 4 participants, let's say always say no, and then all of a sudden all, um, of a sudden that we start just saying yes all the time after five. So there is instead, in reality, if you look at this line, a very smooth transition from no perceptual experience, I mean, this case is 50% means just chance, uh, to perceptual experience to 100%. So clear perception. So what it means is that if we look at the same stimulus, say stimulus trend number five, Sometimes participants will respond correctly and say, yes, I've seen something and some other times they will not identify the stimulus. So the transition from chance performance here won to full perfect. Performance under percent is gradual and is not abrupt. Why is this the case? So I want to think about it, um, in a slightly different way. So here we have if you focus on stimulus trend number five, it would be always the same stimulus. But sometimes people will say yes. So we'll say yes. We have perceived it. Some other times they will not be able to perceive. So let's see why. So one possible explanation um, very popular explanation is offered by the signal detection theory. So the signal detection theory divides the sensory process in multiple stages. So in the first one um It's acknowledging that the stimulus is producing a sensory response, that we will depend strongly on the intensity of the stimulus. So super bright light is very easy to perceive. A super black dot on a white background is super strong. Um, so its intensity is high and therefore will generate a nice response in the, uh, in the visual system, however, the response will not only depend on the strength of the stimulus, but will also depend on internal noise. This is because our sensory system are imperfect. They're they're noisy. And as we will see in the next lecture, generally speaking, every cell as a, um, generate some noise. So even if a cell is not necessarily involved in a particular, um, in particular computation or processing, it will fire every now and then right randomly. And this is generating noise. So internal noise interferes with our perception decisions about the world. And this happens when the stimulus is weak. So in our psychometric function there was this would be the case most of um for most of stimuli strength until we get to stimulus strength number nine. However, there's also another important concept that comes from signal detection theory. And it's the bias. So, uh, subjects are biased in the sense that they will have a criterion that they use to decide whether there is a stimulus or not. So highly motivated subjects, for example, my, uh, may adopt a very low criterion. This means that they will tend to accept the presence of weak stimulus. So they're, um, ready to take a risk and say yes. Most of the time when they are not sure, um, whether something has been presented. However, subjects were lacking confidence. They might want to be extra sure, so they might want to have a very strong stimulus before they say yes. I've seen something so they are biased towards accepting only, uh, intense stimuli. So this variability explained by the noise and the bias will lead to incorrect decisions some of the time. Um, and this is the reason why thresholds are probabilistic and not absolute measures of performance. Another way of studying perception comes from electrophysiology or neurophysiology, is based on recording electrical activity of cells in sensory pathways. Um, one of the main um methods is single cell recordings. This is, um, an invasive technique that has been used extensively in animal studies. So what we do is we, um, record extracellular activity that is coming from one single cell in the sensory system, uh, and in the sensory area, for example, visual area of the brain that we are interested in. So, um, in this case, what we want to know is whether there is a stimulus that will activate a specific cell. Um, and this activation is measured in terms of action potential. Um, so what we have here is very tiny little electrodes, micro electrodes that are placed into the area of the visual system under study, for example, in the retina, the thalamus or the cortex. The tip of these electrodes is positioned next to the axon of the cell that we want to study. Um, and from there we will pick up the action potential. The electric signal that we record is very, very small and the form must be amplified. And sometimes this is attached to a speaker so that the experimenters can hear the selectivity. So every time that the neurone is firing there would be a noise. And this is a method that has been used extensively to study the receptive fields of a cell. And we will talk about receptive fields in the next lecture. So this is what it looks like. So in this case we have the stimulus. Um so you will become familiar with these because this is just a light. Um the light is off when this line is here. And then as the light goes, uh, on, also the, the visual representation of the stimulus goes up and it stays, and then it stays up until the light. Um, it switched off and then it goes down again. So this is when the light is on. So there is this squared, um, shape. So here the light is on. Um, and then the this is the micro electrode. And what we can see here is that this micro electrode is recording activity. And in particular, what we want to know is how many action potential this cell is firing. So the micro electrode is placed next to the axon of a neurone. What we want to do is we want to find the optimal stimulus that causes a change in firing rate. So what will happen is that, um, there is a screen where stimuli are presented in various positions. So imagine like here, here, here and then is slowly moved. There would be a point on the screen, um, that will make the cell fire. And this is because the cell would be responsible for that tiny part of the space. So imagine that, um, receptive field, as we will see soon of this cell. Is this part of the space, let's say this plus sign here. If we put, um, a stimulus, for example, a bit of light in any of these, um, part of the space, the cell will not respond. But when we get here, it will respond. And this is because that cell is responsible for processing all the information coming from this small part of the space. So by using this technique, we will be able to have this Paris stimulus time histogram, which is a plot that shows how the firing rate of a cell. So action potential per second will change, um, when the stimulus is presented within the receptive field. So in this case we have, um, a stimulus that is accreting, um, and is moving from left to right. And what you can see is that, um, on the x axis is the time on the y axis is how many spikes per second. So the firing rate every single vertical cell represents one, uh, action potential. So what you can see here is that when the grating is really clear, so there's clear vertical stripes white and black. And it's moving from right to left. We have a lot of action potentials. This means that the cells really likes this configuration and this movement from right to left. As you will see when the same grating becomes less clear. So for example here the white and black stripes are not as visible. We are still moving from left to right. You can see there's still a lot of action potentials, but a little bit less compared to this situation here. If we go here, we are still moving right to left, but the gratings of these stripes are not as visible anymore. There's just a few action potentials. Now we take exactly the same gratings and we move them from right to left. In this case, you can see that it doesn't matter how clear these stripes are, there's always, um, less action potentials compared to this situation here when we are moving from right to left. So single Cell recordings is a reductionist approach to perception, because what we are trying to do is we are trying to reduce a very complex problem. So how we perceive to a set of, um, more basic problems that can be studied in a lab. So we just reduce everything to action potentials, limitations. Of course, this tells us little about perceptual processes, which instead rely on the activity coming from many neurones and in networks in the brain. So it is a little bit like we want to study the TV, how it works, um, by measuring properties of single components or chips, resistors of capacitors. So we miss the bigger picture. However, it's also true that cells are the basic function, um, functional unit of the brain. And therefore we must start from the single cells. And we have gained incredible insights in how visual perception works by looking at single cell recordings, a different way of using electrical activity, um, in humans without any invasive procedure is by using EEG or electroencephalography. And in this case, we position this EEG cap on people's heads. Um, and there's electrodes. So these electrodes are able to pick up the activity generated by many neurones, um, from the scalp, so the electrical activity is very small as well and needs to be amplified and then recorded. The EEG will record the activity of a large group of neurones, so it's not as precise as a single unit recording where we were looking at the activity of one single cells. However, um, EEG is very good when it comes to, uh, information in time. So we can tell what happens in the brain at any given millisecond. So very tiny. Uh, time. Um, so what we will do, typically we present stimuli many times and we evaluate how the brain processes them. Another neuroimaging technique is, um, MRI, in particular functional magnetic resonance imaging or fMRI. Uh, when we want to know which area is involved in a task, for example, face recognition, we can use this technique. So fMRI will generate a picture of which area or set of areas are more active during doing a specific task. So one thing is that it takes quite some time to take a picture. Um, two three seconds. So we lose information on when things happen, but we gain information on where things are happening. So in summary, all this neuroimaging technique will measure um, when and where things are happening, but will not tell us how things are happening. Also, these techniques offer correlation and information. One last approach I would like to discuss with you is neurophysiology. Neuropsychology. Sorry. Um, which is the study of patients that have suffered some some kind of injury to the brain and therefore will offer us some information about how perception works because of brain damages. Um, so this has been an exceptionally useful field, uh, to study perception and understanding how colour vision works motion perception or face recognition. Because basically what we can see is that when a specific area is disrupted, then there is a specific problem in behaviour related to, um, a perceptual deficit. However, brain injuries are never, um, restricted to a specific area. So sometimes it's really hard to tell which area produces which deficit. Um, and also, um, what happens is that the brain is extremely good in compensating. So there might be other areas that take over the task. That used to be the task of the damaged area. So sometimes it's really hard to have information from patients. So these are the sources that I've used for this lecture. Um, and that's all for me. SPEAKER 1 Thank you.