As You Snooze, Your Brain Reviews

JOHN DANKOSKY: This is Science Friday. I’m John Dankosky, and I’m sitting in for Ira Flatow. Ira’s fine, he’s just having a long-planned “staycation” week. Later this hour we’re going to talk about a public health tool called contact tracing, and take a geologic tour of the moon.

But first, imagine some sort of repetitive action that you’ve had to learn to do over and over again– maybe fishing for horse mackerel in Animal Crossing. Hell, you do this to the point where you, say I feel like I’m doing this in my sleep. Well, maybe you are.

Writing this week in the journal Cell Reports, a team of researchers studying two people with neural implants say that it appears that during sleep, peoples’ brains replay parts of what they’ve been learning that day. Joining me now to talk about the study is one of the authors of that report, Beata Jarosiewicz. She was a Research Assistant Professor at Brown University working on the BrainGate project when this research happened. Now she’s a Senior Research Scientist at Neuropace, a company in California. Doctor, welcome to Science Friday. Thanks for being here.

BEATA JAROSIEWICZ: Thank you so much for having me.

JOHN DANKOSKY: First, tell us about these study participants. Why do they have this neural implant installed in the first place?

BEATA JAROSIEWICZ: They were two different gentlemen, one that had ALS and one that had a brain stem stroke, I believe, and they were enrolled in the BrainGate pilot clinical trial, which– the main purpose of the BrainGate clinical trial is to try to develop brain-computer interfaces that will help people with paralysis. But our participants are also happy to participate in other basic neuroscience-type research studies, and this was an example of that.

JOHN DANKOSKY: So before we get to the study, I do want to ask a little bit more about this brain control interface. So this is essentially allowing them to do everything from typing an email, to maybe composing music on a keyboard. Can you tell us a bit more about how exactly it works?

BEATA JAROSIEWICZ: Sure. So the area that we’re recording from is the motor cortex, and specifically the hand and arm area of motor cortex. In this brain area, individual neurons have what are so-called firing rates. For example, if the person is imagining moving their hand upward, there might be a subset of neurons that increase their spiking rate above baseline– meaning above their steady state firing rate level. And then when the person moves their hand rightward, there might be another set of neurons that increase their firing rate.

And it turns out even in people with paralysis, neurons still have these sorts of patterns. So if we know, for example, all of the different directions about the neurons that we’re recording from and we know, at each moment in time, what their firing rates are relative to baseline, whether they’re increasing or decreasing their rates, we can look at that pattern across all of the recorded neurons and figure out, based on that, the direction in which the person wants to move their hand– for example, to control a mouse on a mouse pad or their finger on a track pad. And then we can use that information to move a computer cursor in that same direction.

JOHN DANKOSKY: So talk about the physical implant itself. How tiny are these electrodes?

BEATA JAROSIEWICZ: Each of the microelectrode arrays is a 4 by 4 millimeter square array, and it’s got 100 electrodes on it, arranged 10 by 10. Each one of them is about a millimeter and a half long, and we have usually two of these arrays implanted in each participant.

JOHN DANKOSKY: Tell me more about how you are interpreting what these electrodes sense. I mean, how exactly do you know that the activity in a certain patch of neurons means go up and to the as, opposed to go up and to the right? Can you explain that a little bit further?

BEATA JAROSIEWICZ: Sure. So we present a task on a computer screen where we give the participant a cursor that they control and a target somewhere on the screen, and we ask them to imagine that they’re moving their hand to control a mouse, for example, to move the cursor towards that target. And then we do this a few times, and after a few iterations we have enough data to start creating a model that maps each neuron’s activity to particular intended movement directions.

Once we have that model, we can allow the person to start controlling that cursor with their brain activity, and then we can further refine our estimate of that mapping by continuing to collect data as they’re doing this task with the presented targets. And then we can turn off that presented target task, and then allow them to do something practical or fun with their brain-computer interface– like, for example, play the Simon game, or, as you mentioned earlier, type emails or chat messages to their friends and so forth.

JOHN DANKOSKY: So let’s get to the Simon game. So in this study that we’re talking about, the participants had to do this game, which some people might remember as a memory computer game called Simon. Maybe you can describe the game and exactly what it was they had to do as part of the study.

BEATA JAROSIEWICZ: Sure. So the game has four colored wedges arranged in a circle around a cursor that appears initially in the center. And four of the targets light up in a particular sequence, and there’s an associated sound with each target. And the person tries to remember that sequence and then replicate it by moving the cursor to those same targets in that same order.

And most of the time we played the same sequence of four targets in a given session, where each session consisted of rest, and then playing the game, and then another rest period. And then we interleaved some other random sequences that were more rare that occurred only twice each that were different from this repeated sequence, and it was this repeated sequence we were looking for replay of in post-task rest.

JOHN DANKOSKY: So tell us what you found. What did you observe here?

BEATA JAROSIEWICZ: We found that the repeated sequences that were replayed frequently– first of all, the person was able to remember them more easily, because they were faster at repeating those sequences during the game than the control sequences. And secondly, we found that those repeated sequences were replayed during rest after the task more than those control sequences where when you compare them relative to the pre-task rest.

And by replayed, I mean if you look at the patterns of neural activity over time and just try to correlate that with the repeated sequences that we saw during the game, we saw higher correlations in general with the repeated sequences than with the control sequences. So what this means is that the brain seemed to be replaying these learned sequences in rest more often than we expected by chance.

JOHN DANKOSKY: And it was replaying the sequence as though they were playing the game. I mean, if you’d actually turned on the game, would their brain have been hitting the patterns? Would they have been activating the various colored blocks in the correct order?

BEATA JAROSIEWICZ: That’s an interesting question. So the correlations were not quite high enough that I think we would see too many of those repetitions too robustly. Probably most of the time you’d see the cursor kind of wandering around aimlessly, and then occasionally hitting that same sequence of four targets.

Another difference is the timescale. We saw the repetitions, during post-task rest, of these repeated sequences either happening quite a bit faster– like, on the order of 10 times faster than during waking– or a little bit slower– like 1 and 1/2 to 2 times slower. It’s possible that some of the replay events happened during specific physiological events in the brain that map onto more consolidation in one of these timescales than the other, but we’re not quite sure, and that could be a good subject for follow-up studies.

JOHN DANKOSKY: So the shorthand that I was thinking about whenever we were reading about this study was something that I’m sure we’ve all experienced, which is you do something over and over again, and then you feel like you dream about it at night. Were these participants dreaming, or was this a different level of sleep?

BEATA JAROSIEWICZ: Well, they definitely didn’t enter REM sleep– Rapid Eye Movement sleep– where the vivid, story-like dreams occur when you’re sleeping, because a 30-minute nap period isn’t really enough time for them to enter that dream state. But it’s possible to have hallucinatory-type dreams when you’re first drifting off into sleep.

We think it’s possible that the patients were experiencing these replays as memories, or as these hallucinatory dreams. We didn’t actually study that specifically in this study, but just anecdotally, when we did happen to ask the participant, did you feel like you were thinking about the game or something else, maybe? The participant would answer us, it’s none of your business what I was thinking about.

So I’m guessing it wasn’t the game. So yeah, and the sleep state that they entered– that we did see a little bit of in every session– is called non-REM stage I, which is the first stage of sleep that you enter as you’re drifting off into sleep.

JOHN DANKOSKY: So what would the purpose of this replay be? Is it the brain going through a checklist of things that it did? Is there a purpose in terms of learning? What do you think it is?

BEATA JAROSIEWICZ: We didn’t directly address that in this study, once again, but we think that what’s happening, based on decades of research and throughout different labs in the country and modeling work and so forth– we think what’s happening is a process called memory consolidation. And this is a process by which memories acquire more permanence in their neural representation.

And what seems to be happening is that this brain structure called the hippocampus– which is very plastic, meaning it can learn things quickly, but its memories also can degrade quickly as a result– the hippocampus can rapidly store new memories that are, for example, emotionally salient or important for some other reason to the person, and take a snapshot of the neocortical brain activity that’s happening during the original processing of that to-be-remembered event.

And the neocortex is that wrinkly outer layer of your brain, and that’s the brain area that seems to be responsible for taking in sensory information, processing it, integrating it across different modalities, helping you make decisions, creating voluntary movements. All of these different things happen in different brain areas, and the hippocampus takes a snapshot of all of the neurons that were simultaneously active during this event– the pattern of activity over time– and then it gradually feeds that information into the neocortex to help it incorporate meaningful information into its existing knowledge base in a slow enough way that it doesn’t disrupt previous memories that have been stored there, or previous mechanisms of processing that everyday sensory information.

JOHN DANKOSKY: So if these electrodes had been placed in a different part of the brain, like a language center or some other part of the brain that is in charge of doing something other than moving your hand about, do you think that you’d see the same sort of replay effects?

BEATA JAROSIEWICZ: Yeah, I think so. We have seen studies come out from other laboratories showing replay of other kinds of cognitive tasks– for example, visual memory tasks, and even just from blips that were observed during the day. But the replay in these previous studies was on a larger scale. It was the averaged activity of thousands of neurons.

JOHN DANKOSKY: So does this research tell you anything that can help to improve your brain-computer interface, anything that will help you make this work more efficiently or effectively for the people you’re trying to help?

BEATA JAROSIEWICZ: That’s an excellent question. As far as this particular study, I think all it really speaks to is the issue of how memories might be consolidated in the human cortex. It might have applications down the road to memory prosthetics, or ways that we could potentially help people with memory disorders like Alzheimer’s or hippocampal dysfunction. But as for motor-brain-computer interfaces, not that I can think of right now.

JOHN DANKOSKY: Beata Jarosiewicz is a Senior Research Scientist at Neuropace in California. She previously was a Research Assistant Professor at Brown University working on the BrainGate project. Doctor, thank you so much for sharing this really interesting research with us. I really appreciate it.

BEATA JAROSIEWICZ: Thank you so much for having me.

Copyright © 2020 Science Friday Initiative. All rights reserved. Science Friday transcripts are produced on a tight deadline by 3Play Media. Fidelity to the original aired/published audio or video file might vary, and text might be updated or amended in the future. For the authoritative record of Science Friday’s programming, please visit the original aired/published recording. For terms of use and more information, visit our policies pages at http://www.sciencefriday.com/about/policies/