Support for this show comes from the Working Forests initiative. The Working Forest industry is committed to planting more trees than they harvest. More than 1 billion seedlings are planted in US working forests every year. From biologists to GIS analysts, hiring managers, accountants, working forest professionals have dedicated their focus towards sustainability, using their expertise to help ensure a healthy future for America's forests. They say they don't just plan for the future. They plant it. You can learn more at Working ForestsInitiative.com. If you picture yourself as a little baby, right? You are born with billions of neurons pooping around in your head. But in order for those neurons to do much of anything, they have to connect to each other so they can transfer signals across the brain. And so in your baby brain, you form connections. These things called synapses, like little bridges from one neuron to another neuron. And as a neuroscientist named Alison Barthold made, in the first few years of your life, your brain is basically like a connection forming machine. Neurons form synapses quite exuberantly. Trillions of them. They're building a ton of bridges. And like, this is not surprising, right? Sure. It makes sense. Of course, babies and toddlers are making lots of connections in their brain. But then the weird part is, as kids get older, instead of continuing with this building spree, the brain goes on kind of a synaptic demolition derby. It's an incredibly rapid rate of destruction. After you form all of these connections, you start to get rid of them quite aggressively. As you grow up and go through your teens, the evidence suggests that you start pruning back a lot of the bridges that you built in those first few years of your life. Eventually, not all those connections are needed. And metabolically, it's really expensive to keep all these connections there. So you want to get rid of the ones that you don't need, the ones that don't carry important information, or you only use very, very rarely. This process is known as synaptic pruning. It seems like a lot of it happens when we're kids and teenagers, and then it eventually slows down. And it plays a part in how we learn and how we become ourselves. And that was already both fascinating and kind of counterintuitive to me. Like, why would your brain build stuff up just to tear it down again? But then, as Allison was talking to me about synaptic pruning, she also mentioned that she had worked with some other scientists to try and replicate something like synaptic pruning outside the brain. So this is unexplainable. I'm Bird Pinkerton. And today on the show, the second in our two-part series on teen brains, we're going to talk about why young brains develop the way they do and what we might learn from that. Okay, you've already met Allison. The other person that you have to know for this story is a guy named Socket Navlica. I'm an associate professor at Cold Spring Harbor Laboratory, and I run a lab studying what we call algorithms in nature. Basically, you can think of Socket as poking around in the natural world in search of literal life hacks. A lot of biological systems have to solve problems to survive, and the mechanisms that they've evolved over millions of years of evolution can be viewed as a strategy or as an algorithm to solve this problem. And so if you think about a species, as what is a species, it's just a collection of really interesting algorithms. Right? That. I mean, and also some other things, but yeah. A couple of other things. But from a computer science perspective, it's just a bunch of algorithms. I mean, a human is basically the same. I'm not saying they're different. Socket has been interested in borrowing algorithms from nature for a while now. And as a post grad, he wound up joining the lab of this guy at Carnegie Mellon, who also worked on this kind of stuff, which is how Socket met Allison, actually, through his advisor. So one day he gets a phone call from Allison Barth in neuroscience. And I had some ideas about some interfacing between neuroscience and computer science that I thought would be really exciting. And he agreed to sit down and have coffee with me. As she was trying to explain her idea to Socket's advisor, she kind of casually mentioned that brains prune down their connections. And he kept stopping me. And he's like, wait, they prune? Wait, why do they prune? Socket pruning is well established in neuroscience. Like, the first paper about it came out all the way back in the 1970s. And I kept saying, yeah, they prune. And then there's these really exciting neurons that seem to kick off all this activity. I think it would be really great. And he interrupt me again. He'd be like, what do you mean they prune? Are you sure that they prune? I'm like, yes, yes, they prune. And so it was from that point that he decided that this would be a really interesting topic to pursue with his new postdoc. A postdoc named Socket Navlica. Socket told me the way he thought about networks back at that time was more like, say, a Facebook kind of thing. We mostly go along adding connections over time. And it's very little pruning, right? Like, okay, maybe I might unfriend somebody, but that's a far small likelihood event in the process of creating new connections, right? So pretty much everything is I add, I add, I add connections. It's never like I add a bunch and then I move a ton. So Socket also remembers being perplexed by Allison's talk of developing brains that prune themselves back. What do you mean you like put down all of these connections and then you lose half of them? Or like, think about building a road network in this way, right? We would never over build roads knowing in advance that we're going to shut down 50%. And of the road center there, right? It's just a completely wasteful thing. It doesn't make any sense. And yet, it had to make some kind of sense because all these mammalian brains evolved to do it. And socket was so intrigued by the idea that we might learn stuff from nature. He thought this might be worth looking into, asking some more questions about. So the first question is why, right? Like, why does the brain do this thing? Is there any values, quote unquote, to this process compared to doing the little ad by ad slowly? As he talked with Allison, Socket realized that there could be value in the brain doing things this way. For one thing, it wouldn't make sense for us humans to be born with an exact map of all the trillions of connections in the brain, right? Because flexibility is so important for us. But the genome can specify our very rough rules about what is going to be important for your environment. And some of these are, you know, walking, breathing, smelling, right? These things, beautiful. You need to do those right, you know, soon, maybe not walking, but breathing and smelling these are important things to do right out of the gate. But then there's a lot of stuff that the genome cannot possibly know about. Like, if I'm born in environment A versus B, and stuff is more important in A, it's less important in B. The genome has no way of knowing that. So how does it deal with this uncertainty is by giving you all of this potential? And it's actually a great idea because you can kind of let everything shake itself out. And then based upon how those connections are used, you can strengthen, you can fortify those connections, and then just get rid of the other ones that weren't used very much or don't work well. For example, let's say, as you're growing up, you spend a lot of time playing the fiddle. Skill is going to require signals to travel across different parts of your brain, right? They're going to cross your little synaptic ridges over and over as you practice. And the more you use your fiddle connections, the more those connections will be reinforced by the brain. Those synapses get kind of cemented in. Except, let's say you actually don't practice or use those synapses very much. Let's say you take one fiddle lesson and then you decide that playing instruments is just not for you. Your brain might not reinforce the synapses that connect things up in a fiddle kind of way. It'll reinforce the synaptic pathways you're using for whatever you are doing, for speaking French maybe, or for juggling. So you start off with the potential to wire your brain in all these different ways, and it's only eventually that it seems like the brain kind of prunes back the ones that you don't use much or, or at all. Use it or lose it. If you don't use it, maybe you're just going to get rid of it. So as you spoke with Allison, the pruning approach did start to make more sense to suck it. And then one thing that intrigued him about it was that it seemed to be a way for the brain to teach itself what to do. So he explained this idea to me by talking about one of the studies he'd read through this work that helps some researchers win the Nobel Prize actually. I will say though before we start that it is a little disturbing. So here's what they did. They took a cat. And under sort of typical conditions, there would be some neurons that respond to the left eye and some neurons that respond to the right eye. And these are different neurons. So what they did was very early in development. They took one of the cat's eyes. Let's say it was the left eye and they sutured it shut. Oh, yeah. Okay. Okay. Not optimal from the cast perspective. But the left eye gets sutured shut. So no activity now is entering that eye. Okay. So now they wait a little while. They present some stimulus to the cat into the right eye. The stimulus, of course, it goes and their neurons on the right side respond as usual. The question is what happens to those neurons on the left side? What happened was the neurons that would normally have been connected to the sutured eye were instead connected up to the open eye. That's great. Right? And it's good that this kind of rerouting happens. You don't want your neurons to just sit there not doing anything. And this is the kind of flexibility that we were talking about before. But to suck it, what was especially interesting was the idea that pruning was probably what made this process work. Because how does the brain know to do this kind of rerouting? Your brain is not like the movie inside out. There's no crew of tiny animated people in there sitting in a control room pulling lovers. Saying, hey, you guys are not acting so now go connect over there because that's where we're seeing some action. So this is where pruning comes in. A sucket understood it. If a neuron has a whole bunch of connections at the beginning, if it's connected to both eyes, for example, it doesn't need a central planner to say, hey, connect up here, don't connect there, whatever. It can figure things out for itself by being like, well, I am not getting any signal from over here. There is no point in really remaining connected. I am getting a signal from over here. So let me strengthen that connection. Neurons, right? They're just sitting there with very minimal ability to know what else is happening in the brain. And so the only way, it's sort of like building a sculpture, right? You start with the full brick and then you slowly sculpt your way down. And that's sort of the strategy that's evolved. This was exciting to suck it because he thought, wait, there are lots of times when it would be cool for a network to teach itself this way, to learn from the ground up instead of from the top down. At the time, he was thinking about power grids, for example, or like wireless and mobile networks. And so we started thinking, okay, could this be a better way of building networks compared to the alternative of starting very sparse and then slowly adding connections over time? Like sure. It does have a big upfront cost of making connections and then destroying them. But could this ground up approach, nevertheless, actually produce good results? Could we learn something from how our brains learn? After the break, suck it, learns the answer to that question. I'd like to shape some impressionable young minds. So, Socket and Allison had been working together. Allison had taught him a little bit about how our brains developed into themselves. And Socket wanted to see, could pruning help him build a more efficient network that could work out in the world outside of an animal brain context? And he decided to do essentially, some compare and contrast work. He built a bunch of model networks that you can kind of imagine like airplane networks. So nodes in this network are cities, Cincinnati, San Francisco, Oklahoma City, and all these cities. And just like in a normal airline network, in Socket's model networks, the goal was basically to get people from point A to point B efficiently. So Socket built some models that looked kind of like the networks he was used to, things that start small and then add and add connections over time. But he also made some networks that were inspired by the brain. So they had an overabundance of connections from the get go. And in these models, what we asked is, okay, what if like you could fly from any city to any other city directly non stop? What if you could fly from Providence, Rhode Island to Walla Walla Washington, right? Or from Atlantic City to Bemidji? Okay, but I know that I can't maintain this cost. So ultimately, I need to prune myself down to just offering a very small number of direct routes. But let's let's say that I was completely naive. Like the brain is to begin with, I don't know what routes are going to be popular. I have to figure that out over time based on the data that I see based on who's buying tickets, based on what neurons are firing and which ones are leading to activation. So the question was, would this kind of self pruning approach lead to a more efficient network than the approaches that socket was used to the kind of build it up as you go ones? And to work that out, he tested these different versions of the networks that he built. And he found that in fact, his pruned brain like networks did perform better on average. They were efficient, so they provided short energy saving routes from like an Oklahoma city to a New York or whatever. And they were also robust. So if you had a connecting flight that went through Atlanta, say, and then there was a terrible storm that knocked out Atlanta, you still had a good alternative option to get you to your destination. It actually went so well that a socket and the bar Joseph, sockets advisor, they wound up publishing a paper about it along with, of course, Allison. It was a really beautiful example of something that for them was not intuitive at all. And I think that the brain had the same problem and had worked on the problem and had solved the problem. And that you could kind of glimpse this way of nature that was a better way was really fun. Now, if you look up the words network and pruning and computer science today, you will find a bunch of papers talking about pruning and artificial neural networks. So that's in AI. I don't know if I would say that our work was the inspiration for all this. Probably not. I think these things might have evolved a little bit separately, but I think it just goes to show that if you're tuned into what the brain is doing, there could be a lot of useful things that come out of it. Socket is definitely still tuned into what the brain is doing. He has continued to explore ways to borrow from the brain, recently published a paper about this. And he and Allison also put together kind of a list of some other weird things the brain does that might be worth exploring. So there's one idea that Allison told me about, for example, that is so weird. It's actually a little hard for me to wrap my brain around. But she says that one of the ways that our neurons communicate is to excite each other, right? So to make it more likely that a neuron will pass a signal along. But another way they seem to interact is by inhibiting each other. Inhibition is turning things off. A neuron will kind of tamp down another neuron and make it less likely to fire. And there are some insights on why it might be a way to make a signal in the brain clearer, right? To quiet down the surrounding noise like a one person talking at a time kind of thing. But Allison says there are circumstances where she doesn't fully understand why this happens always. How can you just have networks of neurons where I'm telling you to shut up and you're telling someone else to shut up? I mean, right, right, it's just like, well, after a while, what if nobody's talking? I mean, how do you even do that? But these networks of inhibition of inhibition of inhibition are really common. Definitely feels strange and counterintuitive to me. But as Allison and Socket showed, counterintuitive things might be the things we have the most to learn from. It is funny though, because working on this episode, I felt like I was learning. Allison and Socket were talking about things they had learned. And yet, the thing we were learning and talking about was teenage and developing brains. And this idea that the developing brain is uniquely good at learning. That this synaptic pruning process is an incredible superpower that specifically young brains have for vacuuming up skills and knowledge. I obviously no longer have a teenage brain. And so I kept wondering what was going on inside my head, as I was listening to researchers talk about this and reading through papers. Now that the kind of developmental bonanza of synaptic pruning in my head is over, or at least has trailed off. Am I just cooked? Am I too fixed in my cerebral ways to grow? But as I was wondering, slash spiraling about this, I learned that Allison had recently picked up the fiddle. I love the idea of practicing something and seeing it get better. I can see that I am getting better. When I'm playing the fiddle, I'm learning music by heart. And I can play, I don't know, like, 100 tunes or something. And I kind of know them. And I didn't know them a couple years ago. You know, that's really cool. It's very satisfying. And yes, her daughter did also pick up the fiddle. So she's playing fiddle now. And it's kind of frustrating. She's super fast. You know, she picks things up really quickly. And I'm okay. You know, I pick things up, but I see she can do things that I don't think I'm going to be able to do. So our kids brain probably has an advantage on hers. But clearly, adult brains can still pick up new skills. Learning might be slower or harder. But the research suggests that even in adults, there's still a version of this flourishing of new connections and then pruning them back as we learn. And by the way, can I just say, I am learning how to play the violin. And I'm much better than I was even a couple years ago, right? So I definitely, there's progress here. It's definitely happening in adults. I like the idea that as I read through papers about synaptic pruning or as Allison and socket talked to each other, the neurons in our brains were forming new connections and pruning them back, helping us cement all these ideas in our heads. And I really appreciate that, at least hopefully, there's still plenty of learning left for all of us to do. This episode was produced by me, Bird Pinkerton. It was edited by Sarah Kate Kramer and by Julia Lungoria, Jorge Just, Joanna Salatarov, and Meredith Hadnott helped keep the wheels on this project. Christian Ayala did the mixing and the sound design. No, I'm Hassan Feldd, wrote the music. Melissa Hirsch checked our facts. Sally Helm and Amy Padula are the fact that polar bear fur is transparent. And I am always, always, always grateful to Brian Resnick for co-founding the show with me and Noam. I owe a lot of people, a lot of thanks this week. So a huge thank you to Anna Hutton Locker for talking to me about her dad, Peter Hutton Locker, who discovered synaptic pruning. She wrote a book about her father called From Lost to Memory. Thank you so much to Beatrice Luna, B.J. Casey, and David Greskin for helping me better understand the science of synaptic pruning and answering my many questions. And thank you to Nurej Cha and Martin Hoffman for talking to me about pruning in the context of artificial intelligence. Thanks also to Hannah Choi for recording Fiddle for this episode. And to Alison for sending me a recording of her playing the fiddle with her banjo partner, she did not Austin. The series was made possible by support from the Annie E. Casey Foundation. And if you have thoughts about this episode or about teen brains in general, please do not hesitate to reach out at unexplainableadfacts.com. We would love to hear from you. If you would like to support the show and the journalism that Vox does, we would love it if you would become a member. This holiday season, your membership actually goes further. So when you join Vox as an annual member, we will gift a complimentary membership to a reader facing financial barriers. You can read more about all of that at Vox.com slash members. If you cannot join our membership for whatever reason, it would also mean a lot if you would leave us a nice review on your podcast platform or just tell someone in your life that they might want to listen to our show. Unexplainable is part of the Vox Media Podcast Network and we will be back next week.
