What is memory in the brain? How is that different from the way a computer stores information? How do you put together billions of specialized cells and you store your home address in their activity? And what does any of this have to do with the happy Birthday song? Or with squirrels hiding acorns, or with bards memorizing epic poems, or with people who cannot forget any of the events in their life. Welcome to Inner Cosmos with me David Eagelman. I'm a neuroscientist and an author at Stanford and in these episodes we sail into our three pound universe to understand why and how our lives look the way they do. Today's episode is about memory. What is memory? How does it work? How do details get stored in your brains such that if I say, what was the name of your fifth grade teacher? You can retrieve that name even if you haven't thought about it in decades? And how is this totally different from the way that computers store memory? So in this episode, I'll give you a foundation into understanding the landscape of memory. Now, this is the first of a three part series about the way that our brains constantly unhook from the here and now, and they go somewhere else. We are time travelers, and only because we do this so constantly, we don't even notice how amazing this is. I mean, everything you might study in a textbook about the brain has to do with Okay, here's how vision works. Here's how the visual quartet analyzes photons captured at the retina and makes an assessment of what's in front of you. And yet we're able to travel back to previous times. You can put yourself back in that fifth grade classroom, or the first house you grew up in, or the moment of your first kiss. You can simulate the sights and sounds and smells. You can remember how you felt in certain moments. You can recreate and reexperience that. You may be able to recollect lots of the details that were around you. To do that, you just time traveled across years or decades to place yourself in another era. Your visual system is not simply looking at the photons in front of you. Instead, it's now involved in running a simulation of what had transpired and an earlier time wherever you physically are right now listening to this podcast, You just time traveled. When you thought about your fifth grade experience. Now it turns out that brains don't just time travel backwards, but they can also move forward. So in the next episode, Part two, we're going to talk about prediction, and then in part three we're going to pull pieces of the puzzle together to unlock some wild surprises about how we experience the world emotionally given our time traveling talents. So today we'll start with memory. And I just want to say this is a huge topic that I could teach you over the course of years. So for today's podcast, we're going to take a high level ride to get a feel for the landscape. So one way to approach any big topic is to look at the extremes. So a lot of memories come easily to us the day we moved into our new home, or the day we got the job offer, or the time that the phone rang with news of somebody's death. But imagine having a clear and immediate memory for all of the events of your day to day life, such as the dinner that you had on November thirtieth, twenty twenty three, or the friend who visited you on May twentieth, twenty nineteen, or the swimming pool party you went to on July eighth, twenty seventeen. Imagine having a vivid, movie like recollection of what you did on today's day ten years ago. Now imagine that these recollections dominate your thoughts. They run in parallel with the events of your waking life, as you commute to work, as you talk with your friends, as you shop for groceries, as you spend time with your family. You would be right that other people would marvel and probably envy your exceptional powers of memory, but you might find the process uncontrollable and overwhelming. Some memories are painful, or traumatic, or just annoying. Then you might wish that you could be rid of them, and exhausted by their presence, you might finally contact a neuroscientist, hoping that they could help you understand what's happening. And that is the real life story for a woman named Jill Price, who is a mneminist, which means she has an extraordinary memory and she can't forget things. So some years ago she went to the University of California at Irmine and described her experiences to two memory researchers there my colleagues Magaw and Cale, and at first they were skeptical, but they investigated her memory with a series of careful tests and interviews over a period of five years, and the results showed that Price indeed had a rare and astonishingly powerful memory, at least for certain types of information. Given a particular date, she could recall within seconds the day of the week that was, and the details of what she did on that day, and the news were the events that occurred on that day, without any preparations. She easily recalled the dates of every Easter from nineteen eighty to two thousand and three and her activities on those days. She also gave correct dates and personal anecdotes for randomly selected news events like the start of the First Gulf War, or the bombing of the Atlanta Olympic Games, or the death of Princess Diana. Now, while her recollections were detailed and consistent, they were generally specific to events of high personal relevance, like illnesses or relationships, or big news stories. Interestingly, her general intelligence was average, and her memory for other types of information was really no better than anyone else's. She never excelled in school and wasn't particularly good at memorizing dates and history books and at one point her interviewers asked her to close her eyes and describe their clothing, and she drew a blank. So her memory abilities were so powerful, but they were also strangely limited. So when these researchers published their first report on Jill Price, she became a news item. Her story drew headlines around the world, She was a guest on Oprah, and she published her memoirs. But by the way, Jill is not the only mneminist. The first official report ofneminism that I know about was a book from the Russian neuropsychologist named Luria about a newspaper reporter named Sharyshevski who had the gift of an untaxable memory, essentially total recall for all the moments of his life. Shiashevski first came to him in the nineteen twenties because he got in trouble at a meeting for not taking any notes when there was a speaker, and Sharishevski responded to his boss by recalling the speech word for word, and this blew everyone away there, and it also blew Sharyshevski away that the other people couldn't do that, and so this started this thirty year relationship with Luria, who tested him in all kinds of ways and had him memorize math formulas or huge blocks of numbers, or even poems and languages he didn't speak, and in all these cases, Sharyshevski could memorize these in a matter of minutes with perfect accuracy. Luria tested and observed him for three decades and then wrote a book called The Mind of a Mneminist. So, although they are rare, neminists exist, and as scientists continue to search for the source of their remarkable abilities, it's worth asking why their stories capture our imaginations so strongly. Perhaps it's because we'd all love to have supercharged memories, Or maybe we see how failing memory is so debilitating in old age and we hope to escape the fate of our grandparents. Or maybe we recognize how little we actually retain from our vast, rich set of life experiences that shape us. So in today's episode, we're going to dive into where memory comes from and how it works, and this is going to give us insight into both remarkable memories and bad memories. We'll see what memory is, how it functions, how it fails, and what purpose it serves. So the first thing I want to tackle is the mistake we make by calling the storage of zeros and ones in computer memory, because computer memory is actually nothing like human memory. Computers store in their memory exactly what you give them. So if I write a document, I expect my computer memory to function perfectly such that it gives me exactly that document back. I don't want something similar or a degraded version of it. But human memory is a completely different ballgame because you're not storing things in zeros and ones. Let's imagine that you speak two languages and I tell you a joke in English and you turn and tell it to someone else in your other language. So the joke is obviously not stored as zeros and ones. You're not memorizing my exact words and inflections. Instead, the joke is stored in a conceptual space in your mind, the way that the characters and events fit together, the gist of the action. It doesn't matter the language with which you output it. And note that this is the same thing with a tune. When you remember a tune, it's not about the individual notes. Instead, it's about the relationship between the notes, and that's why you can transpose the tune to any other key and you still have no trouble recognizing the tune. It's not about the exact notes, but about the relationship between them. And this is the same thing when you're recognizing your friend's face. What has been stored in your brain is the relationship between their nose length and lip thickness, and distance between their eyes and shape of their ears, and where their hairline is and so on. It's not about the individual features, but the relationships between them. And by the way, this is the only way we could have good recognition memory because other details change. For example, imagine that you learn a picture of the face where the lighting is on the left, and later you see the same face, but it's lit from the other side. If you are memorizing pixel by pixel, then the pattern is totally different and you would never be able to recognize it or account for all the lighting differences that might happen in the world, much less the size differences if you see a big picture or a small picture. So understanding the relationship between things is the only way to have a successful recognition and literally, for decades, computer scientists were chasing the wrong ideas by trying to get pixel by pixel recognition until they realized that it's all about the relationships. And this is not just about vision, but this applies to everything we learn. Like facts, your brain encodes new ideas with respect to other things you have learned. Just imagine two people looking at a list of important dates in Mongolian history. If one of them already has a richly developed Mongolian history timeline, then new facts are readily incorporated into that person's network of knowledge. We don't simply memorize facts. We fit new data into the relationship lattice of our internal model. Everything we've learned before based on what was relevant to us. And I talked about this in episode twenty seven. Everything you learn is represented in terms of what you already know. So when you listen to this podcast, the points made would be meaningless unless you all already we're living a human life and knew what a brain was and understood the particular language I am speaking. We do not encode things as little files of zeros and ones. What's happening in the human brain is much different. So that's the first thing. And there's another reason why human memory is not at all like computer memory. And this is because over the past several decades, we've come to realize that human memory is not one thing, but instead it's made up of lots and lots of different subsystems. There are many things going on under this one umbrella term that we call human memory. So for the next few minutes, let's dive into the nuts and bolts of that landscape. The first thing to know is that memory is divided between short term and long term. So short term memory is about in information that decays in a really short time if you don't use it, like seconds to minutes. This is also called working memory by neuroscientists. The idea is that short term memory uses information to address immediate situations. So let's say I say, oh, hey, your verification code to log into LinkedIn is six one, nine, five, three to two. You can remember that for some seconds or minutes, but you're probably not going to remember that number next week, or probably even in an hour from now. There's no strict time limit here, but short term memory typically lasts just for the duration of the task at hand. Now, there are lots of ways that your brain helps itself long with short term memories. Sometimes you do this by talking to yourself like six one nine five three two six one nine five three two, or sometimes you might hold on to information by visualizing something in front of you. But how whoever you do it, working memory has a limited capacity for holding information. In nineteen fifty six, a cognitive psychologist wrote a paper called the Magical Number seven plus or minus two. Now, why that weird title is because he found that most people can hold about seven items of information in short term memory, sometimes a little more so as a less It doesn't matter whether that's shapes or locations or colors or numbers or whatever. You can hold about seven items. And by the way, this has been proposed to explain why telephone numbers in most countries use seven digits, because back in the day people often needed to memorize phone numbers and if they were like sixteen digits, no one would be able to do it anyway, So that's short term memory. I ask you to memorize a combination lock or a log in code or a phone number for some task and you say, cool, got it. That lives in your short term memory. And that memory is built of the activity of millions or billions of neurons holding onto a loop of spikes that run around and around across giant swaths of territory. The activity is kept alive, and that represents the thing you are remembering now. Often, something you hold in your short term memory tends not to transfer to your long term memory. But long term memory is the really interesting thing in the brain. So that's what we're going to concentrate on from here on out. Because with long term memory, it's not about active neurons. Instead, there are changes to the physical structure of your brain such that you can pull up information that happened to you a long, long time ago. You're actually changing the stuff of your brain. I mentioned in a previous episode that if a rock hits your windshield and leaves a you could say poetically that your car remembers the rock even long after the rock is gone. There is a physical mark on your car, and that's the way to think about it. With the brain, an event happens to you, and it leaves a physical mark, and that mark can be read out later. So as opposed to short term memory, long term memory involves brain systems that in code and store and retrieve information over long periods of time, anywhere from minutes to a lifetime, and the capacity of these systems is much greater than short term memory, as we see with someone like Jill Price or Shashevski. Now, amazingly, long term memory is not one thing. Instead, we divide it into two groups. On the one hand, we have implicit memory and on the other explicit memory. Now, implicit memory is the stuff that you can't express or articulate or consciously recall, like riding a bike. You have no idea how you do it, but you learned it and you remember it. You just can't articulate how you do it. It is implicit. And this category of implicit memory also includes unconscious emotional memories, like an aversion towards snowboarding years after an accident, even if the accident itself is long forgotten. Okay, so all of that is implicit memory. On the other hand, explicit memory is the type of information that can be consciously recalled and described, like facts and events. So think about something that happened to you that you can recall and describe, like this is where you went to college, or this is the name of that great movie you saw, or that's what your neighbor said to you last week. Explicit memory also includes memories of facts like the fact that Amazon dot Com sells books or that penguins can't fly. If you can say some recollection or fact out loud, then that is explicit memory. Now, I wanted to clearly lay out these two sides of implicit memory and explicit memory before taking a deeper dive to see why these are understood to be separate systems. So now let's return for a closer look at this first one, implicit memory. So think about memories involved in how to perform skills or habits like reading or typing, or whimming, or juggling or playing piano. You acquire these kinds of memory slowly through a lot of practice and repetition. The way you learn to perfect these actions is from your brain changing its structure. It is remembering. So even though you might not typically think of things like reading and typing and walking and so on as memory, they are indeed a form of remembrance, just one that you cannot articulate, And there are many forms of implicit memory. When you might have heard of is called priming. That's what happens when some past experience influences your future response. So imagine I show you the letters S, blank, P, and I ask you to fill in the missing letter. If you have just had a shower, you'll probably put an A in there to make the word soap. If you've just seen a steaming hot bowl at the table, you'll probably put the letter you in there to make the word soup. Your brain just recently experienced something, and that memory influences your behavior in the next moment. Now what's fascinating is that people with amnesia, which means they can't write down new explicit memories, can nonetheless show these kinds of priming effects. So they have no conscious recollection of having seen this steaming hot bowl at the table, and they'll deny that they ever saw that, but nonetheless they'll choose soup. I'll come back to this point in just a bit, but first I want to finish painting the different types of implicit memory. So another type of implicit memory is one you might have heard of called classical conditioning. This was first described by the Russian physiologist Ivon Pavlov at the end of the nineteenth century, so you generally remember this story. Pavlov saw that dogs respond to food by drooling or salivating, so he said Okay, I'm going to teach the dog through experience that the ringing of a bell predicts the food. The bell itself is totally arbitrary, but it comes to tell the dog that some food is coming. Pavlov could just as easily have used a pat on the head to predict the food, or a purple dot flashing three times or whatever. It's just some random thing that you link to food arriving a moment later. So he teaches the dog to associate the ringing of the bell with the impending delivery of food, and after setting up the relationship, the dogs salivate when he rings the bell, even though there's no food there yet, because the bell now is as good as seeing the food. And again, this is implicit memory because you don't have to be able to tell the story of a connection consciously in order to form this kind of learning, this memory. So that's classical conditioning, and there's another form of implicit memory, which is operant conditioning. Here you learn to associate your own behavior like pressing a lever, with something rewarding like food, or something aversive like an electric shock, and you gradually do more or less of that behavior as a result. Of remembering the consequences. Now, I'm not going to go into the details of the brain regions that are involved in these different forms of implicit learning. If you're curious, to go to my textbook called Brain and Behavior. But I'll just mention that learning any of these types of responses depends on some brain areas like the amygdala and cerebellum, and some structures in the brain stem. The key thing I want you to appreciate is that all of these forms of memory I just described are forms in which your brain changes itself. It changes its detailed internal structure in response to things that it has experienced so that it be behaves differently in the future. And all of these things I just told you about, that's just implicit memory. Now let's switch back again to explicit memory, the things that you can consciously articulate, like events and facts. Even explicit memory can be divided into two different forms. The first is called episodic memory. This is memory for past events or episodes that you've experienced, like a birthday party or a surfing trip you took. These are called autobiographical memories. And the key point is that, unlike implicit memories, these Episodic memories can be consciously recalled and described. They can have a sort of cinematic quality to them, and they usually have a particular context, like in the living room on New Year's Eve. You're also generally able to pick out specific objects and features and the surroundings, like oh, yeah, the sofa was over here my left, and there was a television set over there, and my uncle was sitting over here on my right, And you can recall very particular sequences of actions like oh, we were wearing party hats and the wine was poured and then it clinked glasses with everyone. So those are episodic memories. Now there's another type of explicit memory, not episodic, but semantic memories, and these are things that your brain remembers about the outside world, like the fact that a sheep has four legs and it makes a noise that sounds like bah, and it has a wooly coat. Semantic properties are more general than single events that you saw or experienced, and because they're independent of any one particular kind of sensory input, they're generally useful for organizing the world into categories of related things Like these are vegetables and these are animals, and these are tools, these vehicles and so on. So that's semantic memory. Now, I just told you about a whole bunch of different subtypes of memories. But why do neuroscientists divide memory into all these different flavors. Well, it's because clinicians have observed for many, many decades that sometimes a person can damage one part of their brain and lose a very specific sort of memory while not losing other sorts. So, for example, you might see a patient with severe amnesia for personal experiences. She can remember essentially nothing about her personal life, but she retains a good general knowledge for facts that she learned before her brain injury. So that's how we know that episodic memory is different from semantic It's actually underpinned by different structures in the brain, and by the way, you see this often in forms of dementia like early Alzheimer's disease, where a person's episodic memory suffers dramatically. They can't remember the details of their own life, but this is long before their semantic knowledge begins to fail. And on the flip side, in a rarer illness known as semantic dementia, episodic memory is preserved, while even basic forms of semantic knowledge are lost, like what a sheep is or what sound it makes. I won't go into details here, but if you want to look up more, episodic memory generally depends on the medial temporal lobes, while semantic memory involves the anterior temporal lobes. Okay, so now how did people start to figure out what brain regions were involved in what functions. Well, like many things, the answer is that this generally happens when a person gets damage to a very specific part of the brain. So let's zoom in on a particular example. In nineteen fifty, a particular area of the brain called the hippocampus came to the center stage in the neuroscience of memory because there was a man named Henry Malaison, who, by the way, was known for decades in the literature as HM because the patient's real name is never used while he's still alive. So anyway, Henry had suffered a head injury at the age of nine, and after that he had epileptic seizures and these got more frequent and severe, and the doctors tried to control the seizures, but eventually they couldn't control them anymore. So at the age of twenty seven, Henry went in for a neurosurgery because of where the seizures were originating. The surgeon removed the hippocampus on both sides of his brain, on the right and the left, as well as some of the regions that surrounded the hippocampus. Now, after the surgery, the seizures were all better, but Henry became one of the most famous cases in the medical literature because while he was otherwise fine, he had a severe amnesia that means he couldn't remember things, and specifically, he had an antaro grade amnesia, which means he couldn't form new episodic memories. And so that meant he would function just fine if you were talking with him, but after say fifteen minutes, he couldn't recall anything about the conversation because he wasn't converting that into new long term memory. So just imagine being one of the people working with him. You walk into the room and you introduce yourself, and you bring a giant Saint Bernard dog with you. Then the dog leaves and you chat with him for about five minutes and you ask him, hey, when I came in here, did I have anything with me? And he says, yeah, you had a giant Saint Bernard dog with you. So you continue to have a nice conversation about things. He's a small guy. Now you distract him for about another fifteen minutes and you ask him, hey, when I came in here, did I have anything with me? And he draws a total blank. He says, I don't think So now you leave the room and you come back twenty minutes later and you say, hey, when I walked in the room before, did I have anything with me? And what does he say? He says, who are you? As far as he knows, he has never seen you before in his life. So the researchers who studied Henry for years had to reintroduce themselves each time they walked in the room, even if they'd only left the room briefly. So the issue was that Henry could not form new explicit memories. But here's the interesting part. His implicit memory was fine. He could practice and learn new tasks like tracing a five pointed star viewed through a mirror, but he had no recollection of the practice in learning, and he expressed surprise at how well, he could perform this task that, as far as he knew, he had never seen before. Also, his short term memory was fine, well within the range of seven plus or minus two items. And so the tragic outcome from Henry's surgery helped to define the taxonomy of memory systems that we know today. And by the way, we've learned these same lessons from hundreds of other patients with amnesia since then. For example, somebody who can't form any new explicit memories can nonetheless learn an implicit task like the video game Tetris, and they can get better and better at the game. But each time you place them in front of the game, they claim genuinely that they've never seen this before. And what's fascinating is that if you wake them up when they're sleeping, if you catch them in the middle of a dream, they'll say that they were just dreaming about colorful falling blocks, but they have no idea why, because they have no conscious memory of ever having seen that before. By the way, if you saw the movie Memento, you know that was about a man who had lost his explicit memory. He had had a head injury that gave him amnesia. In his case, both in the forward and backward direction and tarot grade and retrograde, and so the only way that he could keep track of his goals through time was to tattoo information onto his skin. And I'll be talking more about amnesia in a future episode, but all I want to say for now is that for most of us, we are lucky enough to tattoo the information directly into our vast neural forests, and we can do this again and again throughout every day of our lives, and we never run out of room. Now, we know from Henry and many other patients that normal autobiographical memory depends on the integrity of the hippocampus. But I mentioned in episode number one of this podcast that there's another important player in the medial temporal lob memory system, and that's the amygdala, which is an almond sized structure that's just in front of the hippocampus, and this is involved in emotional memory. The amygdala assigns value positive or negative to things that it sees or hears or smells based on past experiences. And because it's so well connected, it can coordinate all the different prongs of an emotional response to something. It can coordinate the autonomic responses like an increased heart rate, and the endocrine responses like the secretion of stress hormones, and behavioral responses like fear and avoidance. So it turns out that while normal memories are just taken care of by the hippocampus, emotional situations like something very stressful or dangerous, those kick the amigdala into gear and memories get written down on what is essentially a secondary memory pathway. And that's why emotional events are more likely to be remembered, because in a sense, those are the most important memories. When something really emotionally important happens, that's what you want to write down. That's why memory exists to keep a record of the really important stuff. By the way, I spoke in the first episode about time seeming to run in slow motion when you're in danger, when you're in fear for your life, And if you're interested, you can go back and listen to the experiments that my lab did. But the punchline is that when you're in a very stressful situation, your brain writes down denser memories than normal, and so when you think back on what just happened, your brain pulls up much more detail than you would normally have, and you interpret that as that must have taken a long time because I hit the brakes, but because of all the ice, I couldn't get traction, and so my car slid into the intersection. And I saw the blue Toyota coming and I saw the expression on the other driver, and she hit the front of my car, and I saw the hood crumple, and the glass spider web, and the rear view mirror fall off, and so on and so on, and so when you think what just happened, what just happened, your brain assumes, Wow, that must have taken a long time. That must have taken many many seconds for all of that to happen in order for me to see that much detail. But in fact, time does not actually run in slow motion for you, as we demonstrated with experiments. Instead, it's a trick of memory. You just have more memory that you're drawing up. As I mentioned in the first episode. You can convince yourself of this because if you had a passenger on the car seat next to you, you don't actually remember his voice as saying, which would have to be the case if time were actually slowed down. So it's a trick of memory, and the retrospective illusion happens because the amygdala has come online because something very salient is happening, and it wrote down denser memories than normal. So that's what I wanted to say about the amygdala. But now we're going to return to the hippocampus, which underlies most normal day to day memory. And there was a very interesting discovery which won a Nobel Prize some years ago, and that is that the hippocampus is involved in encoding your location, your position in space. In other words, it's involved in saying, I'm in this spot, and the room I'm in connects to that hallway which leads to the door of that lobby, which has an elevator to go up to that other room. Your sense of where you are relies on your hippocampus. Your hipp campus, in fact, has these specialized neurons that we call place cells, and these neurons become active they fire off spikes only when you're in a particular spot in your environment. And there are other neurons known as grid cells, and these have multiple receptive fields that are arranged in a grid pattern that covers your local environment anyway. So all these cells work together to give you a very precise sense of your position in a room. And the hippocampus is not only about your current position, but it's also crucial for spatial memory. For example, imagine that you're in a labyrinth and you have to go down this hallway and come back to the center, and then you have to remember that you've been there, and then you go down a different one and you come back, and then you have to figure out a different one and go back and so on. It's really easy for you to do this because even in a brand new environment, you have a spatial memory, as in, I've already been down that hallway. But if you have damage to your hippocampus on both sides, then you can't do that task. You need your hippocampi for spatial memory. Now, interestingly, you find hippocampus like structures throughout the animal kingdom, not only in mammals, but also in birds and in goldfish, and in all cases this is involved in spatial memory. And across the kingdom you see that the size of the hippocampus is related to the demands of the territory mapping and the spatial memory. For example, squirrels spend the autumn months hiding literally thousands of seeds and nuts in different locations throughout their territory so they can have a steady food supply through the winter. And during this period the volume of their hippocampus increases by fifteen percent. Or as another example, some species of birds hide their food and these species have larger hippocampi than the birds that don't do this, and when they're doing their food hiding, that stimulates growth of their hippocampus. And this is what you find in humans too. Our hippocampi give us map like spatial codes, and these regions guide navigations. So consider London taxi drivers. I don't know if they do this anymore, but pre GPS, they had to memorize an unbelievably detailed map of London with thousands of destinations encompassing something like twenty five thousand streets, and they had to be able to verbally recite the most efficient roots between any two points, and just working from memory, say points of interest along the way, like the names of all the theaters that they'd pass in sequential order. And when neu oh imaging studies were done on these cab drivers, it could be seen that their hippocampi had grown compared with novice taxi drivers. And interestingly, this doesn't happen for physicians, who have to master a similarly large body of knowledge. But it's not spatial. Okay, So I just told you a lot about how the hippocampus is involved in understanding this spatial layout of things. But why do I mention the hippocampus encoding space in an episode on memory? While first, the world is full of spatial things, and traditionally our memories had to care about that, just like the squirrels remembering where they had buried their seeds and nuts. Also, you may have noticed that we are generally very good at remembering spatial information, like all the important rooms in a very large building. And many people, when they're trying to memorize something like a list of items, will use what is called a memory palace, which is where you associate these different items with different locations. So, for example, let's say I had to memorize a long list of words like apple, baby, clock, dennis, exhibition, flaggirl, horse, ice, jester, ladder, machine, noos, ocean, pigeon, radio, sheep, theater, and so on. What I might do is I might picture my house and I visualize the first item apple at the front door, and then I imagine walking in and I visualize the second item, let's say a baby in this case in a crib just inside the door. And then just past the baby, I picture a giant pendulum clock on the wall, and on the couch over there, I visualize a dentist doing his work on a patient, and so on and so on. I leverage location. I take advantage of these cells in my hippocampus that care about place to tie other information to them. And when I need to remember this list much later, I simply imagine myself walking through the house, and I can recall an enormously long list of arbitrary objects this way. This is, in fact, the oldest memory technique called a mnemonic device, that is on record. The ancient Greek and Roman bards used to tell their epic tales this way. This was before the invention of the printing press, so they had to memorize these things. But happily this was after the evolution of the hippocampus. Now, although some people purposely leverage these kind of techniques to memorize for a small fraction of the population. This happens naturally. In episode four, I talked about synesthesia, and one of the most common forms of synesthesia involves imagining things with spatial locations. For example, you see the days of the week in a circle around you physically, like Monday is over here, and then tuesdays there, and then Wednesday is up here, and then Thursday's over there, and Friday off to the side, and so on. And for each person. This is idiosyncratic, meaning that it's a different pattern for everyone, and a cinnasthete might see the months of the year in spatial locations, or the years historical in future, like nineteen seventy one is over here, and then it goes up and up through nineteen ninety, and then it's flat over to two thousand and five, and then it goes a little up and down through twenty thirteen, and then cuts to the left at twenty twenty, and then it suddenly dives down and then curves around, and then the future goes off behind you. And it's different for each cynisthete, and I gave many more examples in episode four, but the point I want to make here is that This form of synesthesia is closely tied to memory because all of the things that I just mentioned, like weekdays and months and years, these are all sequences that you have to learn, you have to memorize, and so in many people these sequences get tied irreversibly to spatial location. It helps the brain to remember something by tagging it with a place. So some people suggest that the way to understand the hippocampus is as a cognitive map. The idea is that the hippocampus originally created and stored territory maps for orientation and navigation and finding resources, but eventually the system adapted to create and store episodic memories. Also, because events usually have a particular setting in space, and because movement cross locations also involves movement in time, the hippocampus would be naturally well suited to capture sequences of events in time, in other words, episodes. And I want to mention that there's a cousin theory about the hippocampus that proposes what it's doing is it's storing the associations between elements of the events, just like what we talked about with the spatial relations between objects. So the hippocampus stores the temporal relationship between events like this happened after this, or it happened close in time or far in time, and it stores the relationships between pieces of information. It stores the relationships rather than the specific items. So this gives us a way to see how the original spatial function of the hippocampus could have evolved into episodic memory. Okay, so now I want to turn to what's going on at the very tiny level to ask how does activity in the brain cause lasting changes such that we have learning and memory. So almost all theories of brain plasticity, that's what we call it, when the brain reorganizes itself. These use the idea that the strength of connections between cells, the strength of the synapses, can be modified by previous activity. Okay, but how does that work? Well, First of all, if you imagine one hundred years ago, microscopes didn't have the magnification power to actually see neurons in their inner connections. So a century ago, scientists believed that brain tissue was a continuous network like the blood vessels. Blood vessels are a system of tubes where you have stuff running along it, and people assumed that was what was happening in the brain. But this idea was overturned by a Spanish neuroscientist named Santiago Ramonicjl. He spent a lot of time at his microscope trying to look at brain tissue, and because he was a photographer also, he had all these chemicals in his workshop, these stains that he put on to thin slices of brain tissue to see if that would allow him to see anything better. And some of his stains leaped into a few of the neurons and turned them all black, and he was able to visualize them under the microscope that way. So he put forward this idea, which turned out to be right, that the brain, instead of being a subway system of tubes, was a massive collection of billions of discrete cells. He called this the neuron doctrine, and this was a massively important step in neuroscience, and it eventually won him the Nobel Prize. Now, this idea that the brain is made up of lots of individual cells, this ushered in an important new concept because people began to realize that these separate cells they have to influence each other through these little connections between them, these synapses. And Ramonic Cahol was the first to suggest that learning and memory might occur by changing these synapses, so several decades later, in nineteen forty nine, the neuroscientist Donald Hebb made a specific proposal for how synapses should adjust to underlie memory. He said that if one cell consistently participates in exciting another cell, then the connection between them is strengthened, and if the first cell consistently fails to excite the second cell, then the connection between them is weakened. So this rule is often as cells that fire together wire together, and most models of memory formation employ this kind of rule. Now, when Hebb proposed this hypothesis, there was no experimental evidence to support it, and it took until nineteen seventy three for two researchers to discover that, in fact, that seems to be what happens between neurons. So you stimulate some neuron over here, and you measure the very tiny electrical response that it causes in this neuron over here. Then you blast the first neural with a bunch of electrical stimulation for thirty seconds, and then you try that first experiment again, where you give this guy a little zapp and you look at the electrical signal that it causes in the second neuron, and what you find is that you now have a larger signal. The synapse has strengthened, and that strengthening lasts. The connection strength between these two neurons has changed, and that change is held on to through time. So the synaptic connection can be modified as a result of the cell's history of activity. Now, a typical neuron has ten thousand connections, and what's fascinating is that each individual connection can strengthen or weaken according to its history. So the way that activity flows through a network of billions of neurons can be completely changed. By dialing the strength of this connection here, in that connection there, and so on for all the connections across the brain. By dialing the strengths up and down and holding on to those, you can store information in the system. And just to give you a sense of the size of the parameter space here, the large language model GPT four has one point seventy six trillion connections, but the human brain has about one hundred times that. So there are a lot of parameters that can be tweaked in the brain to store information. Now, by the way the birth of artificial neural networks like GPT, for is rooted in these discoveries from the brain from the past century. You have a bunch of units, and you have connections between these units, and you change the strength of those connections. That's how large language models work because making small, subtle changes in the way that units communicate a network can change the entire network's output behavior. By tuning the parameters of the network just right, or in fact, letting a network adjust its own connections according to some algorithm, a network learns to associate inputs and remember what it has learned. This is how we build artificial learning and memory, and modern artificial neural networks are extraordinarily impressive. But I think it's really important to note that although these connections between neurons have gotten all the attention, both theoretically and experimentally, there are lots of other possible ways to store information in the brain. For the last century, people have assumed that synapses are the key to memory, but the story has been complexified by recent decades of research. Where it stands now is that synaptic changes are necessary for learning and memory, but we really don't know if they are sufficient. It's still unclear whether these changes that the synapse is will be the only or even the most important mechanisms involved in memory, or perhaps whether they're involved at all, because the fact is that a dense net of intertwined cells has to orchestrate a careful balance between excitation and inhibition, otherwise the whole thing blows up into epilepsy or it sinks down into non activity. And it could be that all these synaptic changes we see are just to keep the system away from epileptic overload or synaptic shutdown, and memory maybe is stored in an entirely different manner. Our science is still young, and it's possible that we're still missing the core mechanisms of memory, because after all, there are adjustable parameters throughout the brain. You change something here and the network behaves differently there. So there are many other possible mechanisms involved in memory in the brain. For the cognitionanty, this could involve neurogenesis or changes in the excitability of the neuron, or the distribution of ion channel or the shape of dendritic trees and their spines, where the phosphorylation states of intracellular proteins, where the epigenetic codes and on and on. Okay, this is all in my tech book and we're not going to go into that here, but I do want to say that with so many degrees of freedom in biological systems, the number of possible ways to storm memory in a brain is vast. So why do we look at the connection strength of synapse as well? It's partially because that's what we can measure most easily. It's extremely difficult, essentially impossible currently to measure in a living animal changes in channel distribution or dendritic spines or epigenetic codes in single neurons. So as a result, almost the whole field of neuroscience just measures changes at the connections between cells. And that might turn out to be right, But if you could read the textbooks one hundred years from now, it might turn out to be that we were like the drunk, looking for our keys under the street light because that's where the lighting was better. So while artificial neural networks are awe inspiring and they're rapidly changing the world, we don't know for sure that they're doing the same things algorithmically that the brain is doing. They in a sense simplify everything. They are a clever step backwards from biology. In other words, you take the complexity of a single cell with the entire human genome in it and millions of proteins trafficking around, and you just imagine that it's a unit with simple connections to other units. And as I said, that has turned out in large language models to be shockingly effective. But we really have no way of knowing right now how much more we could get out of artificial neural networks if we included the rich complexity of actual biology. Maybe it wouldn't add anything, but maybe it would unlock in entirely new levels of function, making artificial neural networks more like a human with a sense of what information is relevant to learn and incorporating needs and goals and strategies and enjoying the experience of consciousness. The thing to keep in mind is that Mother Nature has had billions of years to try quadrillions of experiments, and we've only been making artificial neural networks for a few decades. So if I were a betting man, I would say there's probably a lot more to be discovered. It is certainly possible that we have not yet found memories Rosetta Stone and I want to point to one more thing about real brains that's definitely not captured in artificial neural networks, and that is the concept of forgetting. Remember I mentioned at the beginning about the mneminist named Sharyshevski, It turns out that Sherishevski's enviable memory went hand in hand with a surprisingly handicapped personality. He was crippled by the fact that he could not forget. Because when you have a memory like his, all the moments in life are retained. Past vandettas and things people did to slight you, and embarrassing moments and situations you've outgrown, and heartbreaks you would rather let slip into the past. All of these remain present and emotionally salient for you. And remember I mentioned the mnemenist Jill Price, who also had an untaxable memory. Just think about what drove her to seek help in the first place. Her memories were deeply emotional and sprang up throughout her day. Jill is constantly pouring over her past, an obsessive detailing mementos from childhood onward and becoming distressed by changes like moving to a new home. So for both Sharashevski and Jill Price. Having a perfect memory impaired them as much as equipped them, and this unmasks one of the great values of the way memory typically exists in most people. We don't retain everything, but instead items fade. One of my favorite quotations is from the French novelist and playwright Honore de Balzac, who wrote, memories beautify life, but the capacity to forget makes it bearable. So when it comes to thinking about human memory, remember that of all the things we talked about that differentiate us from digital computers, a very important one is our capacity to not remember everything. And even that which we do remember doesn't last too long. So let's wrap up. We've seen that human memory is not quite the same as a computer's memory. We don't store zeros and ones, but instead we hold memories about the gist of things, the relationships between things, and we forget through time and memory for us is stored in lots and lots of subsystems. We have different mechanisms for short term memory and for long term memory, and within each of those categories we have subcategories. For example, implicit memories that you can't articulate, like how to ride a bike and explicit memories like facts and events that you can talk about. And in all of these cases you can lose some of these types of memory without any effect on the other types. And now that we've seen an overview of memory in the brain, we will now be ready to turn to the main reason we have memory, and that is to predict the future. We retain information in the detailed configuration of our forest of billions of cells and trillions of connections, and the point is to develop a better understanding of the world so that we can know what will happen next. And for that, please join me in the next episode where we look at prediction how we simulate possible future worlds. This is one of the main jobs of brains, and this is the reason they retain memory. So right now, if you're imagining tuning into the next episode and feeling the emotional joy of what you will learn and what you will see there, this is your brain doing what it was meant to do, simulating the future. I will see you there. Go to Eagleman dot com slash podcast for more information and to find further reading. Send me any questions to podcast at eagleman dot com and I'll be making episodes in which I address those until next time. I'm David Eagleman, and this is Inner Cosmos.