Latest / The Mastery Project Podcast / Perception, Chunking & Expertise
Transcript
- Dr. T-Michael Dougherty, Ed.D.: I want to describe an experiment for you and I want you to guess the result before I give it to you. You take a chessboard in the middle of a real game that's being played and maybe put twenty five pieces on it. You show it to somebody for five seconds. And then you take it away and ask them to try to rebuild the board from memory. Now, a beginner can place about four pieces correctly, which sounds about right. Five seconds is not all that long. But a chess master can place about 16. Sometimes they can place the entire board. Now, the obvious explanation is that chess masters have exceptional memories. And that's what the researchers who ran this experiment thought, too. So they ran it again with one change. Same number of pieces, same five seconds, but this time the pieces were scattered randomly. No real game, no structure, it was just chess pieces on random squares. In this case, the master's advantage almost completely disappeared, down to roughly the same three or four pieces as the beginner. So it was never about memory. something else is going on and understanding what tells you more about how expertise works than almost anything else in the research literature. Let's go back earlier than that study to the Netherlands with a man who was both a chess master and a psychologist, and his name was Adrian de Groot. Adrian de Groot did the original work in the 1940s. And he published a book about it that got translated into English in the 60s. And he wanted to answer what seems like the obvious question: What do grandmasters do that ordinary players do not? Now, his method was simple. He showed players a position and asked them to think out loud while deciding on a move. Everything going through their head, every line they considered, he wrote all of it down. And he expected to find what everybody expects to find. That the grandmasters were calculating deeper or looking further ahead or seeing more moves. That is what chess looks like from the outside. It looks like a calculation contest. But that's not what de Groot found. The Grandmasters weren't searching dramatically deeper than the strong club players were. They weren't considering more options. In some cases, they considered few, actually. What separated them was that within the first few seconds, the moves that they were considering at all were better moves. The weaker player searched a wide field of mostly poor options and eventually found a halfway decent one. The master, on the other hand, looked at the position and a small number of strong candidates simply seemed to present themselves. DeGroot's conclusion was that the difference lived in perception, not in calculation. Grandmasters weren't thinking harder about the same board, they were seeing a different board, which raises the obvious question, seeing it how? And that's where William Chase and Herbert Simon come in. In 1973, in the journal Cognitive Psychology, a paper called Perception and Chess was published. And it's one of the most influential papers in the whole field. And it's the one I described at the top of this episode. They ran the five-second reconstruction test across players at different levels. And then they added the random board condition, And the result was the one you already heard. There was enormous advantage on real positions. Advantage was nearly gone on random ones. but they did something else that mattered even more. They watched how people placed their pieces. And they noticed that reconstruction didn't come out of one piece at a time, evenly spaced. It actually came in bursts. Four or five pieces placed rapidly, almost together. Then a pause And then another burst. The masters weren't recalling 20 individual pieces, they were recalling maybe four or five different groups. Chase and Simon called those groups chunks. And a chunk is worth defining carefully because the word kind of gets thrown around loosely. A chunk is a set of elements that's been used together so many times that it has stopped being a set of individual elements and become one thing. The master didn't see a knight on f3 or a pawn on g2 or a king on g1. He saw a castled kingside. One item, and this is why the random board destroyed the advantage. Random pieces never form something like a castled kingside. They don't form any structure that thousands of hours of chess has taught you to recognize. When you strip out the meaning, the master is back to placing pieces one at a time, same as everybody else. Now, I should give you one correction on that because the version that people repeat often is slightly too simple. Fernand Gobay and Herbert Simon went back to this study in nineteen ninety six and they found that on random boards, masters do retain a small but reliable advantage. Not a big one, but it's real. And the likely explanation is that a randomly generated board still accidentally contains a few familiar fragments, and the master's eye still catches them. So the honest version is that the advantage collapses, not that it vanishes. Now, let me tell you why chunking matters so much, because this connects to a limit you live inside every day without even noticing it. In 1956, George Miller published a paper with one of the great titles in the history of psychology The Magical Number Seven, Plus or Minus Two. Miller's argument was that. Human working memory can hold about seven items at once. That number has since been revised. Nelson Cowan and others have argued that the real capacity is closer to four, which is worse than you'd like and probably worse than it actually feels. But Miller's real insight wasn't the number. It was that the limit applies to items and that what counts as an item isn't fixed. Four is four, whether it's four letters or four words or four sentences or four chess formations. So expertise does not raise the ceiling, it changes what fits underneath it. And here's the analogy that I find to be most useful. Consider reading. When you were five, you processed letters C A T three items. Your working memory was almost full at that point. But you're not doing that now. Right now, you're taking in entire phrases at a glance, and you even have room left over to disagree with me while you do it. Same brain, same capacity limit, but completely different unit size. And that's what happened to the chess masters. They learned to read the board. And here's a version that you might have experienced more recently. Listening to a language that you don't speak. It arrives as a wall of sound. You can't even tell where one word ends and the next one seems to begin. But a native speaker hears discrete words with gaps between them. The gaps aren't in the audio, they're in the listener. Now, the single best demonstration that chunking is what is happening came out of Anders Erickson's lab in 1980. Erickson took an undergraduate identified in the literature as SF, who was just a normal student with a normal memory, and they tested his digit span or his capacity to hold a string of digits in his head. They would read him a string of numbers and then have him repeat it back to them. And he managed about seven of them, which is completely average. Then they had him practice. An hour a day, several days a week for more than a year. And by the end of their experiment, his digit span was 79. Let me repeat that in case you missed it. 79 digits read out loud and repeated back correctly. Now, how? SF was a competitive runner, and he started hearing strings of digits as running times. 352 wasn't three digits, three, five, and two. It was a near-world record mile, three minutes, 52 seconds. It was one thing to him. He built a system and then a hierarchy of systems, and his capacity for retaining numbers in his head went up by a factor of more than 10. And here's the result that makes this study worth the whole episode. They tested his memory for letters, and he got about six. It was completely unchanged. He hadn't improved his memory. He had built a very large, very specific set of chunks for one kind of material, and it transferred to absolutely nothing else at all. Now, I want to pause for a second on that because it sort of destroys an entire category of promises that we've all been sold. Expertise is not a general capacity that you accumulate over time and then point at whatever you want. It's a library of patterns in one specific domain, but outside that domain, you're just an ordinary person. Let me take this outside of the laboratory. In 1975, Harold Koundel and Calvin Nodein published a study in the journal Radiology. And they showed experienced radiologist chest x-rays for 200 milliseconds. That's two-tenths of a second. And that's roughly the time that it takes to blink. It's far too short to move your eyes around an image. You basically get one fixation and that's it. Now the radiologists identified abnormal films at around 70% accuracy during this study. From a flash or a blink before any conscious search could even occur. Something in their trained visual perception flagged the abnormality that they were seeing before a deliberate process even had time to begin. And I want to bring this into my own field for a moment, which is physical therapy, because this is where I see it happen frequently. I can put an experienced clinician and a first-year student into the same room and watching the same patient walk the same 20 feet in front of us. The student will see a person walking. They'll see knees bending, hips moving, ankles, et cetera. Now the clinician sees a compensated Trendellenberg gate pattern and is already thinking about how to strengthen their hip abductors and external rotators before the patient even reaches the wall in front of them. The student isn't paying less attention. In fact, usually the student's trying even harder than the trained clinician. The information reaching both sets of their eyes is identical. One of them has the chunks. Now Gary Klein found the same thing in a much different setting. Klein studied how experienced people make decisions under pressure firefighters, military commanders, nurses in ICU. And he expected to find that experts weighed their options, compared alternatives, and then chose the best one among them. But they didn't do that. Experienced fireground commanders typically generated one course of action. mentally ran it forward in their head and if it worked, they executed it. No comparison at all. They weren't deciding in the way that we usually mean. They were recognizing. The situation in front of them matched a pattern, and the pattern came with a response attached. Klein tells a story about a lieutenant at what looked like a routine kitchen fire. The crew there put water on it, but the fire didn't respond in the way that it should have. Something felt off to the commander, And he couldn't have told you what it was exactly. So he ordered everybody out immediately. Moments later, the floor collapsed. The fire had been in the basement, underneath them. He called it a sixth sense. And what it almost certainly was is a large library of house fires in his head and a pattern that didn't match any of them. Now, so far this is a good news story. Build enough chunks, and you can see what others struggle to see. Now, let me tell you the cost though, because there is one and it's significant. In two thousand eight, a group of researchers published a study with a really interesting title, Why Good Thoughts Block Better Ones. And in this study, they gave strong chess players a position containing a familiar winning pattern. It was a well-known checkmate sequence that almost every serious chess player would immediately recognize. And it worked. It took five moves. The position also contained a different solution that took three moves. It was faster, it was cleaner, and it was right there. the experts found the familiar one almost immediately, and then most of them never found the better one. Now, here's the part that makes this study kind of remarkable. The researchers tracked where the player's eyes went. The players reported sincerely that they were searching the board for alternatives. Their eyes kept returning to the squares that were involved in the familiar solution. They believed they were looking, but they weren't looking. the pattern that they had already found was pulling their attention back to itself. And I want to be clear about what that means because it's uncomfortable. the same process that makes an expert fast makes an expert selectively blind. Recognition is a filter. A filter that catches what it was built to catch is also a filter that discards what it was not. In my field, this has a name and we call it premature closure. The first plausible explanation arrives, it feels right, and quietly the door closes on everything that comes after it. And I've watched this happen to genuinely excellent clinicians, which is the whole point. It's not a novice error. Novices don't have strong enough patterns to be captured by them. This particular failure requires expertise. So what do you do with all of this? You do four things. First, stop trying to remember and start trying to organize. If you're studying something and it feels like a list of completely unrelated facts, that's a signal that you don't yet have the structure that the facts belong to. More repetition isn't going to fix that, and more repetition of unstructured material is just going to build a longer list. Go find the organizing principle first, and the facts get easier after that. Second. Second, get volume of varied cases, not volume of hours. Chunks come from repeated exposure to real patterns, not from rereading a description of the pattern. If you want to recognize a movement dysfunction, you need to have watched several hundred people move. If you want to recognize a good argument, you need to have read several hundred bad ones. There is no substitute for case volume, and description isn't exposure. Third, Name what you see. When you observe something, put words on it out loud. That sounds trivial, but it isn't. Labeling a pattern is a large part of how it becomes a single unit in your head rather than just a collection of disparate details. Every field's jargon exists for this very reason. It's compression. Fourth, and this is the one for the people who already have some degree of expertise. Build in a second look. Once you've arrived at an answer, the recognition machinery in your head is already fired, and it will now quietly work against you finding a better one, as we said earlier. The countermeasure isn't to try harder. The chess players tried harder and their eyes still went back to the wrong squares. The countermeasure is procedural. Instead, ask specifically, what would I be seeing if this were something else? Not, am I sure? You're always sure. That's the problem. A specific alternative when it's named out loud can force attention somewhere that the pattern was never going to send it to begin with. Now, I want to mention three books, and the links to these books will be in the show notes, and they're affiliate links, so I earn a small commission if you end up purchasing them. I'm telling you this because you should always know when somebody has a stake in something that they're recommending to you. So the first book that I want to recommend is a book called A Mind for Numbers by Barbara Oakley. Chunking is the spine of this book. And it's the most practical treatment of the idea that I know of. Oakley came to mathematics late after a career as a linguist, and the book is written by somebody who remembers what not understanding feels like. And that's rarer than it should be in today's world. Where I think it overreaches, though, is that it leans on the focused mode and diffuse mode distinction more heavily than the underlying neuroscience actually supports. It's useful as a metaphor, so just treat it as one. The second book is Sources of Power by Gary Klein. And this is where the firefighter research that we described today came from. If you make decisions under time pressure with incomplete information in front of you, this is the most useful thing I think that you can read. Now it overreaches a bit because Klein is fairly generous toward expert intuition, and he and Daniel Kahneman spent years arguing about when it should be trusted and when it shouldn't. They eventually co-authored a paper together, narrowing the conditions under which it works. So read the book, then read that paper. And then finally, Thinking Fast and Slow by Daniel Kahneman. The other half of that argument, the fast recognition system that made the radiologist accurate in two-tenths of a second, is the same system that produces systematic predictable errors. Now, where this book overreaches is that some of the specific studies that are cited, particularly in the chapter on priming, haven't replicated well over time and Kahneman himself publicly acknowledges this. Here's this entire episode in one thought. Expertise is not a faster version of what you do now. It's a different perception of the same information. The novice and the expert receive identical input and extract different amounts of it, because the expert has spent years building the units that input arrives in. Which means the goal of practice isn't to get quicker at processing details, it's to stop having to process them altogether. And it means that the day you notice, you can see something that used to be invisible to you, like a pattern in someone's gate, or a weakness in an argument, or a shape in a set of numbers, that's not a small thing. That is the process working. That's what progress actually looks like from the inside. And it rarely announces itself any more clearly than that. In the next episode, we're going to look at the other side of this: what happens when patterns stop improving? The plateau, why competence can become a trap, and what the research says about people who spend 20 years in a job but only get better for three of them. Thank you for listening to this episode of the Mastery Project Podcast. I'm Dr. T. Michael Dougherty and if you found this information helpful, please share it with a colleague who you think might benefit from it as well.