Dylan Wiliam on How the Brain Learns | Literacy and the Science of Learning

If I were to tell you that there was a drug that did what education does, few would believe it possible. With more education, people are healthier, they live longer, they children do better at school, and they are more likely to make positive contributions to society. With these kinds of outcomes, it's no wonder we keep looking for education's holy grail, the designer drug for better learning. But how do children learn? What conditions need to be in place for learning to happen? Welcome to season three of the Knowledge Matters podcast, Literacy and the Science of Learning. I'm Dylan William, co-author of Developing Curriculum for Deep Thinking, The Knowledge Revival. Over the next six episodes, Dublin Off, Natalie Wexler and I will look at the latest research on what students need to learn effectively, to become strong readers, writers and thinkers. So the question I just posed, what do children need in order to learn? For the last half century or so, the most common answer to this question has been that we need to make our children more skilled, more skilled at reading, more skilled at doing mathematics and science, and more skilled in understanding social affairs. More recently, to this list has been added what are sometimes called 21st century skills, such as creativity, critical thinking and problem solving. Now, I don't think there are many people who would say these are not valuable. However, the mistake we've made is to think of the best way to get our children to be more creative is to give them practice at being creative. The best way to get them to think critically is to give them practice at critical thinking and the best way to make our children into better problem solvers is to give them practice at solving problems. This series will explore why these ideas are inconsistent with what we now know about how humans learn and what we need to do to ensure that all students leave K-12 education and ready to flourish and thrive. In the first two episodes, I will look at the latest research on what students need from their schooling and the best ways to help them learn effectively. In the remaining four episodes, Douglas Moff and Natalie Waxler will look in more detail at how to make students better readers, better writers and better learners. Before we decide how schools should teach, we need an understanding of how our minds actually work. Over the years, psychologists have come up with many theories about this, but one particularly important idea is that our minds are basically huge stores of information. When we get better at something, more often than not, it is because we have become more knowledgeable about it. Take chess players. You have probably heard that expert chess players can play many games simultaneously and win most if not all of them. How exactly are they able to do this? This was the question that motivated William Chase and Herbert Simon to conduct experiments with chess players in the 1970s. I talked with Daisy Cristodulu, the author of many books, including Seven Nets About Education. We discussed how this study, called Perception and Chess, proves the importance of knowledge. So what Chase and Simon did is they had three chess players, an expert master chess player, an average club player and a beginner. They gave them all various tasks to do with memorising a chess board and then reproducing it. They took a real chess board from middle of a chess game that had about 25 chess pieces on it. They gave each player five seconds to look at the chess board, remember the chess board, and then that ball was taken away, then they had to reproduce it. So when they do that the first time around with the real chess game, the results are really striking. The chess master is just much better at that task. So the chess master of the 25 pieces in that chess master on average across a different range of boards, they could reproduce 16 pieces on average. Whereas the beginner could only reproduce four. So that's a huge difference. If you just stop there, you might say, well that proves that the chess master simply has a better memory. But Daisy explained that in a follow-up experiment, things get even more interesting. The experiment is the same except for one difference. Instead of the chess pieces being in a configuration for an actual chess game, the 25 pieces are placed at random. When they did that, all three players were equally bad and they could all only place two or three pieces on their first go. So this is incredibly striking and there are a lot of enormously important implications from that. Because what it really shows is that expertise is not about some general purpose memory ability or even indeed some kind of general purpose ability or skill. The chess master skill in this case is really, really tightly tied to very specific knowledge and patterns that they have stored in their long term memory. And even if you change the configurations, even quite a little bit, their skill breaks down. So what it's telling us is that expertise is dependent on schema and chunks and information that you have stored in long term memory. Those expert chess players have spent years playing chess and they have very good recognition of typical chess pieces and positions. And that's why they're really good at reproducing the task when it's taken from real actual chess games with patterns that make sense to them. And that's why they're very bad when it's just a random configuration of chess pieces. In education, there is a common belief that, again, good or something, students just need to practice it over and over and over. But what this chess study suggests is that actually the expert knowledge of the master chess player is based on knowledge. Knowledge of game patterns that are in their long term memory. The type of practice and repetition matter quite a lot to becoming an expert. In cognitive science, this is known as deliberate practice, a term that was coined by psychologists K. Anders Erickson. In deliberate practice, there is a distinction between the expertise of something and the practice of it. For example, he wrote that the best musicians don't only play a concert piece over and over. They also rehearse other songs, scales and musical drills. Daisy and I discuss the implications of this. And you can extend this, I think, to academic subjects, to a lot of other performance subjects, to all kinds of things. The analogy I always like to use is marathon money. That if you want to run a marathon and train to run a marathon, you would not think that the only way you can do that is to run a marathon every weekend. And not only would you not think that if you're an obvious, but elite marathon runners do not do that either. You do not just think, well, I just go out and run a marathon every weekend. That's the only thing I can do. You recognize that there are things you have to do, in this case, to adapt your body, to get better at running those long distances. That you have to build things up, that you have to start with shorter, slower distances. And even when you're getting very good, you're doing things like speed work, where you run very hard and then very slow and very hard and very slow. And that's building up the adaptations. And I think that's a really nice metaphorical analogy for a lot of things in learning as well. And the learning example, I will always come back to you because it's now what I spend a lot of my time doing. But if you want to learn to write well, so writing an essay is a kind of complex end goal that we often want our students to work towards. There's a lot of things you need to do to get good at that that do not look like essay writing. So for example, one, a very obvious one is to write a good essay. You have to have quite quite a good vocabulary. So I can envisage a lesson where a teacher is teaching a lesson on maybe a few new words and perhaps some word roots and prefixes and suffixes. And there's a lesson where the students never write anything, they never even pick up a pen. But I would still argue that that lesson could be very effective in making them a better writer in the long run. So I think that's the tension between performance and learning. There are things that will in the long run make you better at something that don't look like the end goal. That I think has major implications for the kinds of things we do in schools because what you're suggesting is that asking students to function like scientists, for example, in a science classroom will not be effective if they don't have the background knowledge that the scientist does. Ask them to be like historians in the history lesson doesn't work unless they also have these incredibly densely interconnected networks of facts and information that historians do. So it seems that this really has major implications for how we organize teaching in perhaps all school subjects. Absolutely, and I think there's a lot of a sort of heuristics that can be very helpful for experts that are not so helpful for novices who lack the expert that the background knowledge. So a few examples. If you say to someone with some background knowledge about a historical era to think about and it topic from both sides, they might be able to do so because they've got the background knowledge. But if you give that piece of general advice to a novice who's starting out, if they don't have the background knowledge to know what the sides are, it's not going to be very helpful. What you see when you look at an expert scientist and look at what they're doing, perhaps they're constructing hypotheses and they're thinking about the weight of evidence and whether the evidence proves the hypothesis and carrying out experiments. We're asking a seven-year-old to copy that. That's not how that scientist became the expert scientist. What you need to do is look at what that scientist was doing to get to that point where they're able to do that. If becoming an expert involves increasing knowledge, how do we do this? It turns out building knowledge depends on what kind of knowledge it is. David Geary is a professor of psychological science and neuroscience at the University of Missouri. He says that a two types of knowledge, biologically primary, which humans learn naturally, an biologically secondary, which needs to be taught. Primary knowledge is knowledge that has a deep evolutionary history and would involve things like language abilities, spatial navigation abilities, ability to recognize facial expressions, and so forth. These are the types of knowledge bases and basic cognitive competencies that we see throughout the world. People everywhere have these basic skills, assuming they have typical developmental experiences. Secondary knowledge is more recent and a cultural invention rather than an evolved competency. These would include the sorts of things that you typically learn in school, reading, writing, and arithmetic would be the standard basics that kids are expected to learn during schooling. Understanding the distinction between primary and secondary skills is critical in understanding how students learn. Take speaking, for example, our brains have regions that directly support language development. We are evolutionally preset to be able to learn it. For the primary skills, there are built-in, intentional systems, perceptual systems, and brain and cognitive systems that allow you to process words, for example, in language sounds and so forth. For those systems to become fully mature and kind of fleshed out and adapted to local conditions, kids have to engage in species typical activities. So for language, you would be talking to their parents or their friends or even talking to themselves or having some minimal exposure to it. As long as kids have adequate exposure to language, this system will develop normally. There's no need for any type of instruction whatsoever. You just have to be in the right environment and engage in sufficient social interactions involving language to develop these types of skills. In a similar vein, if children see faces around them, almost all children will learn to recognize faces. However, most of the things we need children to learn in school are not like speaking, listening, and recognizing faces. There are things that our evolution has not prepared us to learn naturally, things like reading, writing, and mathematics. Our brains don't have the built-in scaffolding for these skills. Most children will only learn this biologically secondary knowledge if they are formally taught through instructional materials, classroom activities, teachers, and so forth. If kids aren't explicitly instructed on, say, phonetic decoding, phonetic awareness that A is associated with an A, for example, then they don't pick it up from their environment in the same way they pick up language. The distinction is critically important in my view. One example is in the reading wars, and the argument there that's been going on for decades is that reading is just part of a general language or literacy competency. And learning to read, understanding what words mean, the ability to sound them out, and so forth, is going to emerge in the same way that natural language skills emerge. And based on that assumption, kids are just put in an environment that has books, teachers might read to them, so forth, but they don't have any explicit instruction on how to sound that words, for example, decoding. And we know that kids who don't get this type of explicit instruction are typically poor readers. But let's take this a step further. David says that our brains do have a reason for conceptual learning. It's called the temporal cortex. Through exposure to everyday things, we learn things like, for example, the difference between dogs and cats. If you're exposed to 10 dogs and 10 cats, you learn what a dog looks like and what a cat looks like just through exposure. David says applying the same concept of mathematics is important to understanding the debate on how kids learn math. And the crux of the argument is that math teachers are biased toward student discovery learning and have kids figure things out for themselves without a lot of structure. And other folks argue that, well, kids need a lot of structure and a lot of practice and so forth. And the math ed folks will say, well, that's kind of kill and drill or drilling, you drill them over and over again, and you kill their motivation for learning. And so the math ed argument that, you know, kids conceptual learning is important and you need to let them discover for themselves kind of an old PhD, and type of thing actually goes against how we know the conceptual system works. It is repeated exposure across a variety of contacts, which would be the drill part of it leads to inferences and conceptual understanding about what this thing means like the mathematical equal sign. That's really interesting because one of the problems we see is that even when teachers do expose students to the equal sign, they do it in a kind of stereotypical way. And many children end up believing that equals means makes. So if you see three plus two equals, they think it's five. But if you give them five equals three plus blank, it doesn't make any sense to them because five doesn't make anything. And so it's not just being exposed multiple times, but it's being exposed to well crafted educational settings that will lead to the right kind of concept formation. That's exactly right. They need to be exposed multiple times in the various contacts in which, in this case, the equal sign can be used. That would be the standard stereotypical problems and then the non-standard problems, as you just mentioned, but also with algebra problems, calculus problems, whatever it is, the equal sign means the same thing. And they won't come to that conclusion unless they're exposed to all these different contacts repeatedly. So, if we have to systematically teach biologically secondary knowledge to children, how do we do it effectively? Australia's psychologist John Sweller explains that are basically two approaches to teaching biologically secondary knowledge. The randomness as Genesis principle and the borrowing and reorganizing principle. In the randomness as Genesis principle, you solve problems by simply trying things out. So you've got a problem and you're attempting to solve it because solving it will give you information that you need. You have to make a move. And in the first instance, you're going to long-term memory and you see, do I know what move to make here? And if you know what move to make here, you can make the move and it goes along smoothly. But sometimes, if it's a real problem that you've not come across before, you're going to be in a situation where you really have no knowledge as to what move to make. So what are you going to do? Well, the only possible thing to do and I need to emphasise this, the only possible thing to do is to randomly generate a move and see what effect it has. It doesn't get you closer to the goal of the problem. If it gets you closer to the goal of the problem, that's okay. You've got a new problem state and you can go through the procedure again from that new problem state. What move do I make? If you have some information on what move to make, then you can make it. If you don't have that information, you've got to randomly generate again. This random generation happens over and over until you get all the information you need. Eventually, you may solve the problem. John Swerr says that the randomness as Genesis principle is not teachable, but something we've evolved to do. The other way of obtaining bodily secondary information is by learning it from someone else, either by copying them or being taught it. This is the borrowing and reorganising principle and it is a defining feature of humans. We are the only species that is able to transmit large amounts of information to other members of the species. We're unique in that respect. We're really, really good at it and we do it all the time. In effect, that's what we're doing right now. Transferring information between people. And we can transfer enormous amounts of information far more efficiently than we can work it out ourselves using problem solving. Problem solving has to be done by somebody, but if the same problem solving process is done by everybody, then a problem which may take several years, literally several years to solve, everybody's going to have to take several years to solve it. That's hopeless. That same problem once one person has solved it, that information can be transferred to another person, literally in a few seconds or a few minutes. That's all that's needed. That second procedure transmission is really, really important. We have had a difficulty in about the last two generations in the people who have suggested, oh no, no, we need to teach people how to solve problems. We don't need to teach people how to solve problems. We've evolved to solve problems. We do need to give people information because of it's an enormously efficient way of learning something. You can get a feel for it if your car breaks down. Who are you going to hire to fix it? Are you going to hire somebody who's been certified as being a terrific problem solver? Or you'll go to hire a car mechanic who actually knows what the problem is. We need to teach people by providing them with information. There is one further feature of human learning that is important to understand. That is that our minds can only process a limited amount of information at any one time. For us to learn anything, it has to get through a kind of bottleneck. Imagine that we have two kinds of memory. Long-term memory contains all the things we remember including facts, events, experiences and skills. Short-term or working memory is a kind of mental scratch pad for things we are currently doing like remembering a six-digit number from your phone long enough to enter it into a website with multi-fact authentication or working on a math problem. To get into long-term memory, things have to get through working memory, but working memory is limited. Here's John Swiner again. We can't remember more than about seven items of information in working memory. That's novel information in working memory. We can't process, by process I mean, handle, deal within some way, combine in some way, more than about maybe three or four items of information. So working memory is extremely limited in capacity, but it's got another limitation. It's also limited in duration. Working memory can hold material for about 18 seconds. After that, it's pretty much all gone. You can keep something in working memory for indefinite periods, but only if you keep rehearsing it. Now, how do we function intellectually given those limitations? Well, once the information is processed, if we think we're going to need it again, we can transfer it into long-term memory. Now, long-term memory is dramatically different from working memory. It presumably has capacity and duration limitations, but we have no idea where they are, what the limits are. It's enormous. So, even though long-term memory is enormous, what we can get into long-term memory has to pass through working memory, which is extremely limited. Thus, the amount we can learn at any one time is limited. Therefore, we have to be careful when trying to teach anyone anything, but the limited amount of working memory must be used carefully. So, what does this mean for teaching? John Sweller, in a series of experiments with school-aged students, showed that students often fail to learn what they are taught because their working memory is overloaded. Sometimes, the material being taught is too complex for the students to process, but sometimes it is because the material is presented in a way that makes things harder than it needs to be. This is called the split attention effect. Take learning geometry, for example. Okay, so you've got a geometric diagram, and you're likely to have a series of statements below that diagram. The statements might say things like angle, A, B, C equals angle, X, Y, Z, and the reason they're equal is for whatever the theorem is that makes them equal. Now, consider what you've got to do to understand that statement. Angle, A, B, C equals angle, X, Y, Z. Okay, well, where's angle, A, B, C? Let me look at the diagram. Oh, there we are, there's angle, A, B, C. Now, why did I look for angle, A, B, C? Oh, okay, I've got to find angle, X, Y, Z as well. Where's that? Oh, there it is. Now, that process of going from one to the other, holding as much of the diagram in your working memory as you can when you're reading the statement, that imposes a heavy working memory load, a heavy cognitive load. It's split attention. You've got to split your attention between the diagram and the statements, and that act of splitting your attention requires working memory resources to switch from one to the other accurately. You can reduce the load on working memory by instead of having angle, A, B, C in a statement below the diagram, put the statement angle, A, B, C, right where angle, A, B, C actually is. So that the learner doesn't have to search where the world is angle, A, B, C, it's here somewhere, and while you're doing that search, everything else is not out of working memory, and you're not going to learn a great deal. In later episodes, we'll explore other implications for teaching of what we know about how our minds work, but there is one other issue that we need to address here. That is whether the capacity and duration of short term memory can be increased. The simple answer is that it can't. However, increases in long term memory can make our use of short term memory more effective. Let's do an experiment. I'm going to say 10 digits. I want you to remember as many of them as you can. Ready? 9, 0, 4, 9, 6, 4, 5, 0, 2, 7. OK, now repeat back what you remember. If you weren't able to recall more than a few digits, you are not alone. Most people would struggle to repeat back these 10 digits because very few people can hold 10 numbers in their head. But if you were able to remember 7 to 10 of those digits, you probably live in northeast Florida. Those digits are actually the telephone number for the City of Stark office in Florida, which is about halfway between Jacksonville and Gainesville. If you live in the area, you will know that approximately half of the telephone numbers in Stark have a Gainesville area code, 352, and half have a Jacksonville area code, 904. If you know these two area codes, then you do not have to hold all 10 of the digits in your head. All you need to do is to hold the last 7 digits in your head, and remember whether the area code is 352 or 904. People living in Stark, or near Jacksonville, could reproduce this 10 digits string better than people who live in other parts of the world, not because they have better short term memory, but because those area codes are in their long term memory. The contents of long term memory are always and instantly influencing what we can do in our short term memory. And because we can't really increase the capacity or duration of short term memory, increasing the capabilities of our students involves increasing the content of long term memory. This is why knowledge matters. The way to make our students smarter is not to give them practice in thinking, but to give them more to think with. In the next episode, we will look at research on how human memory works, and discover how we can make K-12 education more effective for all our children. The knowledge matters podcast is co-hosted by me, Dylan William, along with Doug Lemoff and Natalie Wexler. To learn more about my work, go to www.dylanwilliam.org. That's William with 1L. To download a free copy of my book, Developing Curriculum for Deep Thinking, The Knowledge Revival, visit bit.ly/bit.ly/knowledgerevive. This podcast is produced by the Knowledge Matters campaign. Learn more about this episode and their work at Knowledge MattersCampaign.org. There, you can find curriculum resources, blogs and sign up for their newsletter. To catch all of the Knowledge Matters podcast season three, literacy for the science of learning, make sure you subscribe to the Knowledge Matters podcast on Apple Podcasts, Spotify or wherever you listen to podcasts. Thanks for listening.