Aaron is a cognitive neuroscientist, working on volition and consciousness. Aaron and I met in 2016 in Paris when I did my MSc thesis in his lab at Neurospin on decoding planned and spontaneous movements, using M/EEG. Aaron has since moved to California where he is Assistant Professor at Chapman University.
In this conversation, we talk about Aaron's work in trying to understand the readiness potential. We talk about the classic interpretation, Aaron's interpretation, and how Aaron's interpretation can be applied to non-movement contexts, including the stock market and meteorology.
Time stamps
0:00:40 We don't really know what the readiness potential is
0:01:52 The classic interpretation of the readiness potential
0:16:39 Aaron's interpretation of the readiness potential
0:31:04 The origin of Aaron's interpretation
0:42:33 Applying Aaron's model to non-movement contexts: the stock market and meteorology
0:54:40 - Aaron's plans for studying the readiness potential in the next few years: breathing, individual differences, anticipation
Links
Podcast links
Website: https://bjks.buzzsprout.com/
Twitter: https://twitter.com/BjksPodcast
Aaron's links
Chapman profile: https://www.chapman.edu/our-faculty/aaron-schurger
Google Scholar: https://scholar.google.de/citations?user=lBSjfagAAAAJ&hl=en&oi=sra
Twitter: https://twitter.com/schurgerlab
Ben's links:
Website: https://bjks.page/
Google Scholar: https://scholar.google.de/citations?user=-nWNfvcAAAAJ&hl=en&oi=ao
Papers mentioned
Fried, I., Mukamel, R., & Kreiman, G. (2011). Internally generated preactivation of single neurons in human medial frontal cortex predicts volition. Neuron
Kagaya, K., & Takahata, M. (2010). Readiness discharge for spontaneous initiation of walking in crayfish. Journal of Neuroscience
Kornhuber, H. H., & Deecke, L. (1965/2016). Hirnpotentialänderungen bei Willkürbewegungen und passiven Bewegungen des Menschen: Bereitschaftspotential und reafferente Potentiale/Brain potential changes in voluntary and passive movements in humans: readiness potential and reafferent potentials. Pflüger's Archiv
Libet, B., Gleason, C. A., Wright, E. W., & Pearl, D. K. (1983). Time of conscious intention to act in relation to onset of cerebral activity. Brain
Schotanus, P., & Schurger, A. (2020). Spontaneous Volatility: Fooled by Reflexive Randomness. Journal of Behavioral Finance
Schurger, A., Sitt, J. D., & Dehaene, S. (2012). An accumulator model for spontaneous neural activity prior to self-initiated movement. Proceedings of the National Academy of Sciences
Schurger, A., Mylopoulos, M., & Rosenthal, D. (2016). Neural antecedents of spontaneous voluntary movement: a new perspective. Trends in Cognitive Sciences
Schurger, A. (2018). Specific relationship between the shape of the readiness potential, subjective decision time, and waiting time predicted by an accumulator model with temporally autocorrelated input noise. Eneuro
-
[This is an automated transcript that contains many errors]
Benjamin James Kuper-Smith: [00:00:00] Mainly like, I guess, intention sense.
Aaron Schurger: Yeah. Yeah. In fact, we just had a, um, the other day we had a, uh, seminar. It was, it was online, but we had a seminar about the VRP with, uh, myself and Idina Ross. Um, we, we talked about, you know, how, what are some different ways you might interpret this thing and, and why it's been a challenge to interpret it.
And, um, that was, we're working on a paper about that actually right now.
Benjamin James Kuper-Smith: Oh, okay. Like, uh, about that question, specifically
Aaron Schurger: about that question. Yeah. Why it's been such a challenge to interpret it and how, what are, you know, on, on assumption that we haven't really cracked it yet? We don't really know yet what it is.
What it represents for sure. Um. On that assumption then we're, we're just looking at different possible ways of interpreting it and [00:01:00] seeing what are the implications for each one of those and what's the evidence for and against each one of those different interpretations of it. Um, so I, I think we're at a, we're at really at a point.
I mean, anyone who I, I think, who says, oh, well we, we know what that is. You know, that's, I, I think that that's not, that's not honest. Um, we, we don't really know. So, um, this is, it. It remains, we have a lot of data that, you know, some from coming from this direction, some coming from another direction. And, you know, it's hard to make sense of it all.
Um, it's gonna take some work to, I think, make sense of all the different data that we have and, um, come up with a, a really clear interpretation that we convince ourselves is the right one.
Benjamin James Kuper-Smith: Yeah. So maybe shall we just start there then? So it seems to me at least that there was one predominant interpretation of the readiness potential.
Aaron Schurger: [00:02:00] Yeah. What it's, what we call the classic interpretation. So what the readiness potential really is, um, is a buildup in neural activity, a relatively slow buildup in neural activity that appears in the, in the last, roughly one second or so before you perform what I call a spontaneous voluntary movement.
Uh, which means some, some people use the phrase self-initiated movement. So it's a movement that you make without a, a cue, without a sensory cue telling you, you know, when to move. Um, so when, when you perform a movement kind of on your own time, uh, you get this, you know, if you, if you then. Uh, look, if you, if you time lock some brain data, for example, electroencephalography data or EEG data, [00:03:00] um, if you time lock to the moment when you make the movement and you look backwards in time, you see this buildup happening, uh, before the movement.
So part of the definition of what the readiness potential is, involves the way it's, it's measured. Um, that I say it is almost by definition. It is something that you observe in the average, after taking many movements, many data epochs that are timelock to the onset of the movement, taking many of those and averaging them together, and then you see this buildup, um, it's not something that you can really readily see by just looking at one instance of a movement.
Um, certainly not with EEG.
Benjamin James Kuper-Smith: With, so there are, if I remember correctly, some studies single cell recordings in humans and in animals. Yeah. Um, do you see in readiness potential like signal in single cells there [00:04:00] or not?
Aaron Schurger: Yes. You do. You do. In fact. So, um, there has been one study, one very nice study, uh, looking at single unit, uh, activity in areas in pre-motor areas.
Um, so these are areas like the supplementary motor area and the pres supplementary motor area, um, that, uh, those recordings were made from epilepsy patients. They were being monitored, so they had e electrodes in their brain, uh, implanted in their brain directly. Um, obviously those are there for medical reasons.
And the, the scientific, uh, study is something that's just done, uh, in addition, right? So the patients are asked, well, while you're, while you're in the hospital being monitored for epilepsy, may we observe the activity and, and you're, uh, uh, from those [00:05:00] electrodes while you're performing this or that task.
Um, so yes, there, there are those data showing that when, when, uh, humans perform a, a self-initiated movement, uh, you, you see this buildup in firing rate of individual neurons in, in pre-motor areas. Um, you also see the same sort of buildup, uh, in advance of self-initiated movements in primates. And
Benjamin James Kuper-Smith: also, sorry, just briefly, uh, the, oh, sorry.
You mentioned firing rates, so that means. I mean, so in a, the, the red, the normal rate is potential as we think of it. It's a, you're an aggregate in, um, electrical potential, right? But so firing rate means it just gets more, more quick or like less and less time between action potentials, then
Aaron Schurger: that's right.
That the normal
Benjamin James Kuper-Smith: Okay.
Aaron Schurger: Begins to
Benjamin James Kuper-Smith: fire
Aaron Schurger: faster and faster.
Benjamin James Kuper-Smith: A similar kind of exponential increase or,
Aaron Schurger: um, that it, it, it's, it's hard to say if it's exponential or, or linear, but it's a [00:06:00] ramp. It's, it ramps up in advance of movement. Yeah. Um,
Benjamin James Kuper-Smith: mm-hmm.
Aaron Schurger: And, uh, so yeah, that, that we, we see that similar kind of ramping behavior again, in, in other kinds of, in other species like, like primates, like rodents and, and, and even in invertebrates, like a crayfish.
Um, so they also have this ramping behavior in, in, I mean, they don't have crayfish doesn't have. Cortex. Um, so you can't say it's in their motor cortex, but it's, it's, it's in a, a, a, a, a motor nucleus, um, yeah,
Benjamin James Kuper-Smith: yeah.
Aaron Schurger: Right. Of the, of the crayfish. So I, I think that's a very interesting and exciting data point, um, that the, the, the fact that crayfish exhibit this tells us that it's not a cortical phenomenon because the crayfish doesn't have cortex.
Um, and it suggests that it's, it's a much [00:07:00] more primitive phenomenon. Um, it, it's, it's not something that uniquely evolved in, in, uh, vertebrates or mammals.
Benjamin James Kuper-Smith: So I think we haven't actually finished the explanation of what the redness potential then represents.
Aaron Schurger: Right, right. So, um, it. I like to call it. So we, when you measure it with EEG, we call it the readiness potential because it's an electrical potential, uh, measured at the scalp when you record it.
Uh, when you record from single neurons and you observe that their firing rate is getting faster and faster, we call that a readiness discharge. Um, and when it's recorded, for example, with MAGNETOENCEPHALOGRAPHY or MEG, um, we call it readiness field because it's a, it's a change in the magnetic field. So the name we give it sort of depends on the, the, the way that you're recording brain activity.
[00:08:00] Um, I call it the pre movement buildup. So that just covers all of those. It's the same phenomenon, right? We're just measuring it in different ways. Um, it's this pre movement buildup. Um, but you know, for, just so that you know, that we understand that, I mean, uh, we can call it the readiness potential, that's fine.
The rp. Um, but we're talking about something that's not, that's more, uh, general than that. It's, it's, right. Um, it's not specific to EEG. So it, um, it, it appears right before self-initiated movements, not, we don't observe it. So when there is a cue to move and you're asked to respond quickly, you don't, obviously you don't see this, um, readiness potential.
Um, it, uh, was, it's, it's something that we've been, we've known about since, about the mid [00:09:00] 1960s. So 1965 was really first reported by, by, uh, Korn, Hubert and Deca. And I think they were in, I, I hope I get this right in Fry Borg at the time. Um, I
Benjamin James Kuper-Smith: think so. Yeah. I can't remember exactly.
Aaron Schurger: Yeah. Um, when they discovered it and their interpretation of what, what this meant.
I mean, it was at the time it was something completely radical. Um, you, you, you, before that, uh, seminal study, I don't think there were any examples of experimenters looking at brain activity, uh, when there, when subjects perform movements without a cue, without a stimulus. So the, the whole, uh, uh, edifice of, of psychology research was built around the stimulus response paradigm.
So you provide a stimulus, you observe the response. [00:10:00] Um, and here were these two scientists who said, well, let's see what happens when there's just no stimulus. Um, 'cause you can act without a stimulus. Um, and, and so they discovered that when they did that, lo and behold, they discovered this slow buildup before the onset of the movement.
And they interpreted that to reflect the brain's, uh, planning and, and preparation for movement. Um, which is a sensible conclusion to come to. Looks like the brain is getting ready to move. Um, you lock the data to the time of movement, uh, and you look at what's happening beforehand. On average, you see this ramping behavior, um, natural conclusion to come to as well.
This is the brain, you know, again, quote unquote planning and preparation for movement. Um, and that's, there, there, there hasn't really been any kind of formal model of what the readiness potential represents. [00:11:00] Um, we just, but we call that the classic model. Of the readiness potential. Um, it's, it's the, the brain is gearing up to move for some reason.
The brain needs a little bit of time to gear up to move before you move. Um, and that's what this ramp represents,
Benjamin James Kuper-Smith: but only for self-initiated movements, but Right.
Aaron Schurger: Specifically for self-initiated movements. Yeah. Um, that became very controversial, right, in the 1980s when Libert did his, uh, seminal and, and very famous or infamous, uh, studies where he recorded the readiness potentials, uh, while subjects reported, uh, when in time they felt their urge to move.
And, uh, you know, that, that, that of course is, is now, [00:12:00] uh, a classic. Study that's been very controversial and is very well known. But, um, to make a long story short, I mean he, Libby found that the time that people judge to have been conscious of their urge to move, uh, happened very close in time to the movement itself, maybe two tenths of a second before the movement.
Um, whereas the readiness potential is apparent quite a while before that 500 milliseconds, half a second or, or even a second, um, before the movement. So if you take the readiness potential to reflect the brain's gearing up to move, uh, then the conclusion you come to is, well, wait a minute, the brain seems to be gearing up to move before I'm even conscious of my own decision to move.
Um, and that's what I call lipids paradox. It is. Just at face [00:13:00] value and, and it just seems kind of paradoxical, right? We, we all sort of assume that, uh, you know, first you decide and then your brain does its thing.
Benjamin James Kuper-Smith: Yep.
Aaron Schurger: Right? I mean, that's just, that's makes sense. I mean, yeah.
Benjamin James Kuper-Smith: That's not that controversial.
You
Aaron Schurger: would think. That's not so controversial. I mean, that, that's what, that, that's what you'd expect. And so this was pretty radical, right? It was like, well, no, it looks like your brain does its thing and then you decide, or then you feel like you decided. Um,
Benjamin James Kuper-Smith: yeah. And I think Wagner's, Daniel Wagner's, I guess way of putting it together, was that the, but then you say that basically the brain initiates two things separately, once the actual movement and separately also this conscious, uh, intention or urge to move.
Aaron Schurger: Yeah.
Benjamin James Kuper-Smith: Um, and that then basically. Um, because the intentional urge still comes before the action, and because we [00:14:00] can't, you know, internally feel our readiness potential,
Aaron Schurger: right.
Benjamin James Kuper-Smith: Um, we think the urge caused it,
Aaron Schurger: right? We feel it's, it's, it's sort of by, according to that view, it's just something that the brain tacks on after the fact.
Um, maybe just to make us feel good, to make us feel like we're in control. Yeah. We're consciously doing these things. I mean, li Bit's conclusion was that well, movements are, they must be initiated unconsciously. Um, and then the conscious decision comes on as, as a, as an afterthought or, um, as after the fact.
So that's, you know, that's been controversial ever since it appeared in, in the early 1980s. Um, and people are still arguing and debating to this day, still arguing and debating, uh. S experiment. Um, and why it was, I mean, it, it was, you know, one of the, just [00:15:00] for, uh, for example, I mean, one of the criticisms was that the task is just completely strange and weird, and nobody does anything like that in, in real life.
Um, you know, you're asked, I mean, basically, I mean, livid asked his subjects to be spontaneous, which is a weird instruction, right? Um, asked him, well just, you know, sit there and wait a little while and when you feel ready, just move your finger. Um, it is a weird instruction. Um, it is kind of goofy. Um, I mean, I've done his experiment many times with subjects and, um, the fact is, as goofy as that task is, uh, subjects can do it.
Um, and, uh, they do it just fine at first. They kind of look a little confused and, and, and they're not sure what to do. But you, you explain it, you say just, you [00:16:00] know, one thing you tell I often tell them is I say, well, kind of try to surprise yourself if you can do that. Don't.
Benjamin James Kuper-Smith: Oh, okay.
Aaron Schurger: Don't think in advance.
You know, don't think, okay, I am gonna, I'm, I think I'm gonna move now. Don't, don't do that. Um, don't count. Don't say, all right, I'm gonna move in five seconds. Don't do that either. Just, um, try to just keep your mind blank and, and at some random moment, just surprise yourself and lift your finger and subjects do it.
So they, they get the hang of it and then after a while they're fine. Um, and then they can do that for an hour.
Benjamin James Kuper-Smith: I'm curious with these, if, so let's say you have a, a participant and they do, you know, I dunno, a hundred trials an hour, whatever. Um, do you find that they have some sort of normal distribution as about around a certain.
Average waiting time or something like that, that they basically reinterpret the confusing task into something like, [00:17:00] um, you know, wait six, six seconds plus minus, uh, a few seconds delay or,
Aaron Schurger: right. What, what's interesting is that the, the distribution of their waiting times, um, is pretty well fit by a gamma distribution, uh, not a normal distribution.
So a gamma distribution looks, looks like a normal distribution, except that it has this, it's skewed a little bit, um, and, uh, it's, it, it, it skewed a little bit and has this long tail on the, on the right. Um, so, so you, you have, you know, you have a tendency to make movements around a certain, uh, you know, let's say five seconds, six seconds, mark.
But then you have a, a bunch of, you know, occasional movements that are 10 seconds or even 15 seconds. Um, but ever fewer of those. Um, so, and, and what's [00:18:00] interesting about that is that it's the very same distribution that is commonly used to fit reaction times when there is a stimulus. Uh, of course the time scale is much shorter.
Um, but the distribution, the shape of the distribution is, is very much the same.
Benjamin James Kuper-Smith: Does that tell us anything or is, I mean, does
Aaron Schurger: it to coincides or No, I, I absolutely think it think it does. Um, so I mean, if, if, if we've covered, uh, if you think we've covered enough of the basics so far, then um, you know what, what we proposed, um, back in 2012 was that.
The, uh, the brain uses the same mechanism for doing this kind of, uh, task like lipids task. Uh, just, you know, spontaneous movement uses the same mechanism there that it does, uh, in a, in a [00:19:00] regular perceptual decision making task where you do have a stimulus, um, except that in this case there is not much in the way of a signal.
Uh, there is no sensory signal, right? There's no stimulus. Um, there's just noise. Um, so when you model, uh, when you model decision-making in that kind of a task, in a perceptual decision making task where you have a stimulus, um, you include signal in that, uh, model. So that would be the stimulus itself. Um, and noise.
Which can reflect noise in the stimulus or noise in the brain, um, or a combination. Um, so we thought, well, maybe the readiness potential can be accounted for using this same sort of mechanism. Um, [00:20:00] and in fact, uh, that might, that might explain why, for example, the, the distribution of waiting times has the same shape as the distribution of, of reaction times in a perceptual decision making task.
Um, and it has the same roughly linear relationship between mean and mean and variance as well. So, um, we, we, we used that kind of a model, um, to try and represent or to, to, to, to model the readiness potential. Um, the idea, uh, and this gets to a really Im important point. Um, is that in a task like lipitz, when you think about it in the, in those terms, right?
In terms of, uh, evidence, um, and noise and a threshold, right? In, in these kind of models, you have a decision threshold. [00:21:00] Um, so you have a combination of evidence and noise that build up, and when they reach that threshold, that's the moment where you decide and you, you, you execute the movement. Um, so in, in that kind of, uh, situation, nor normally the evidence is, is quite dominant, you have a stimulus that's quite visible.
Um, in the case of the livid task, um, the evidence is, well, there isn't really any evidence, right? You're asked to just move, just. For the heck of it. So what evidence is there really? Um, I'll come back to that. There is a, I think a tiny bit of evidence, um, or a tiny bit of, of something that you can consider to be like evidence.
Um, but for all practical purposes, let's say there's just noise. Um, one way to [00:22:00] achieve that task, one way to, to comply with those instructions is to, um, bring the noise floor up closer to whatever your decision threshold is, um, and then just wait and it wait until the threshold is crossed at, at a random moment, right?
Um, when we, when we use the model in that way, so we have, we have mostly noise, uh. Climbing towards a threshold and then getting close to the threshold, and then waiting until it randomly crosses. Um, if you take those trajectories, when I say trajectory, I mean how that decision variable climbs up towards the threshold.
Um, and you time lock to the moment of [00:23:00] threshold crossing and you average them together, uh, they fit the shape of the readiness potential remarkably well. So you, you have this sort of, you know, exponential looking ramp up to a threshold. We, we were able to fit, literally fit the shape of a real empirically measured readiness potential using that kind of model.
So the, the idea is that when the, when the evidence or the imperative. To make a movement is weak. Then the precise moment at which that threshold is crossed at which the decision threshold is crossed, leading to movement, um, is largely determined by sub-threshold random fluctuations in, in, in the, in the, uh, neural activity in the motor system.
And so if you time lock to the moment of movement [00:24:00] initiation, which we presume to be the moment of threshold crossing, then you're gonna recover. Um, not the brain gearing up to move. No, you're gonna recover the noise. You're gonna recover the, the, the characteristics of that noise, um, which we know in, in brains and in other natural systems tend to be dominated by lower frequencies.
So you're gonna recover a slow, gradual buildup to the threshold. Um, so we were able to fit then the shape of the readiness potential, uh, with that. And very interestingly, um, if we took the distribution of first crossing times in the model, so the first moment at which the model, uh, uh, the decision variable crossed the threshold in the model and looked at that distribution, it fit the distribution of waiting times, [00:25:00] um, for that, that subjects exhibited doing this task.
Benjamin James Kuper-Smith: So this kind of gamma, uh,
Aaron Schurger: this kind of gamma distribution and, and fit very, in a, in a very similar way, exhibited the same sort of, uh, linear relationship between the variance and the mean. Um, so we were able to fit both the behavioral aspects of this task and the brain data using the same model, with the same parameters in both cases.
The, the only extra ingredient that I haven't mentioned so far, which I said I would come back to, is this imperative or evidence, right? So there's no stimulus in this task, so to speak, but there is an imperative to move a weak one, um, which is the, which is, uh, the demand characteristics of the task. Um,
Benjamin James Kuper-Smith: you mean the kind of like you're supposed to make a movement at [00:26:00]some point,
Aaron Schurger: right?
The demand, the, the word, the phrase demand characteristics, right? Refers to the kind of unspoken, uh, unspoken imperatives in, in, you know, so the, the subject is coming to a, into the lab and, uh, they're coming into the lab to do an experiment about movement. They're being asked to do a task where they're asked to move their finger at a random moment.
So even though you tell them, well, you know, you can wait, kind of, you can wait as long as you want to lift your finger. It's just up to you. Just whenever you feel ready, just spontaneously move your finger. But there's no, there are no rules about when you have to move your finger. Um, you're, you're effectively telling subjects that they can, they could wait an hour if they wanted to, or they could just sit there and never move.
And then you, after a couple hours you say, okay, thank you. And send them home. Um, they could do that. You've given them full [00:27:00] permission to do that. Um, but they never, I've
Benjamin James Kuper-Smith: heard of people who actually do that
Aaron Schurger: guy.
Benjamin James Kuper-Smith: Yeah. Well,
Aaron Schurger: you, you get an occasional
Benjamin James Kuper-Smith: it three times in an hour.
Aaron Schurger: Yeah. You get an occasional smart ass.
But, uh, um, most people, the vast majority of people, they move. After, you know, at the very most, you know, 30 seconds maybe. Um, and, and very rarely that long. So, so without having to tell the subjects explicitly, there's this sort of unspoken, uh, agreement that, look, I'm telling you to move whenever you want to, but, you know, move, okay.
Benjamin James Kuper-Smith: Yeah.
Aaron Schurger: Move sooner or later. Right.
Benjamin James Kuper-Smith: It would be good if you actually felt this urge at some
Aaron Schurger: point. It would. Yeah, exactly. It would, it would, I'd really appreciate it if you did feel the urge, you know, before too long. So, so the, there is a, an imperative to move. Um, and, and then, and, and that's captured in the model.
So the model does have this term that usually gets [00:28:00] assigned the sensory evidence, the weight of the sensory evidence. Um, for example, if you were looking at a grading, a visual grading on the screen, um, you know, the, the, the contrast in that grading might. Be modeled as the, the weight of the evidence. Um,
Benjamin James Kuper-Smith: sorry, what does that mean?
Uh, I'm not, I'm not that familiar with these perceptual tasks.
Aaron Schurger: A grading is a, grading is just a, a, a, a a, a, an, an area on the screen where you have alternating black and white stripes. Um, so, and then you can, you can, you can, you know, have those at different shades of gray rather than black and white. And so you can, you know, black and white would be maximum contrast.
Right? And then as you go towards, you know, a darker and a lighter shade of gray, and you keep reducing the difference in luminance between the darker and the lighter, uh, uh, stripes, that's [00:29:00] lowering the contrast, um, until it's, until there's no contrast. And then you can't release, you can't see it at all.
So that you could call that the, the, the evidence or the weight of the evidence. In the, in the case of a task like Lipitz, well, we said, well, the, there's no sense, there's no stimulus, there's no sensory stimulus. But the quote unquote evidence, well that's coming from these demand characteristics. Um, and it's weak.
It's not, you know, again, there's not a strong imperative to move. You, you, you know, can move anytime you want reasonably soon. So, okay.
Benjamin James Kuper-Smith: Can't you model that or can't you infer that based on how often people actually pressed during the experiment? Like someone who pressed like lots of times would, you might assume, would have a perceived higher perceived imperative than someone who only pressed a few times
Aaron Schurger: E Exactly.
Yeah. Yeah. So that's, that's, I mean, the, when we, [00:30:00] when we, uh, when you model those data, um, yeah, you can take each individual's, um. Distribution of waiting times and model that separately. And some have a, you know, uh, their mean waiting time is, is shorter than others. Um, that would correspond in the model to them having a, a slightly stronger imperative.
Um, so, so there is this weak imperative to move. And, and in the model, that's what sort of helps to drive the system up a little bit closer to threshold. Um, so that eventually a movement is produced or eventually the threshold is crossed. Um, so again, I mean the, what the model says then is when that imperative to move is weak, um, then the precise moment when the threshold is crossed is largely determined by sub-threshold, random fluctuations in the, [00:31:00] in the motor system, in the signal.
Benjamin James Kuper-Smith: How did you, um, s. Like how, kinda, how did you get to start this project? Because if I remember correctly, you did in your PhD I think you did mainly consciousness or something like that. It seems to Yeah,
Aaron Schurger: I did perception research
Benjamin James Kuper-Smith: and
Aaron Schurger: consciousness research.
Benjamin James Kuper-Smith: I was perception. Okay. Yeah, because then, I mean like, the one thing that I, I found slightly, um, just somewhat surprising is that from, as far as I can tell, neither you nor Yako nor Stan had done anything with volition or had published anything on volition.
And then you did, and then you offered like a very new take on the readiness potential from a field that it seems like neither of you three were even doing at the time
Aaron Schurger: that, that's
Benjamin James Kuper-Smith: true. So how did, how did you come to actually do this project and kind of have this perspective on what the readiness potential might be?
Aaron Schurger: Um, I knew about the history. I, I knew about Korn, Huber and, and Decas [00:32:00] study. I knew, of course I knew about Libby's. Study. It was very famous in the field. And I think for some reason, I mean, I've, uh, well, for some reason that particular area of research surrounding, you know, li bits, uh, uh, studies in the 1980s, um, has f for whatever reason has found its way into this crowd that I kind of hang out with.
Yeah. Um, crowd of, of colleagues and researchers who, who, um, are, I mean, mostly a, a lot of what we have in common is, is, uh, uh, uh, exhibited by our membership in the A SSC, the Association for the Scientific Study of Consciousness, um, which is a, before the, before COVID was a, a group that would meet once a year and it's great.
It's, it's a, it's a great, it's a great meeting. Um, it's a great meeting of the minds. Happens once, once a year, and, [00:33:00] um. Colleagues from a lot of different disciplines, but who are, you know, have a serious interest in consciousness research, um, come together and, and in those circles, for whatever reason, even though, well, I mean, lis li bits, uh, study has this element, right, uh, about consciousness.
I mean, the idea that you're conscious of your decision after the brain seems to start gearing up for the movement. Um, so I had been exposed to, you know, Libby's study and the controversy surrounding it a lot throughout the years. And for me, this was, it was, this was one of those things where there came a point where I wondered, I felt like there was something wrong with that, that with that explanation, but I couldn't quite put my finger on it.
And so that kind of stewed for, for. A long time, a couple of years, just this feeling that [00:34:00] there was something not right, uh, about not quite right. I, I couldn't put my finger on it. And then I dis I remember then distinctly at a, at a lecture, I was at a lecture where someone was lecturing about the readiness potential.
And, um, he point was making a point about the brain getting ready for movement and here was this sign. And I just, I, I, at some point, I, I raised my hand and asked the question, well, wait a minute. I mean, what if the readiness potential doesn't mean that you're getting ready to move? And, and then, and, and then I was suddenly I realized, yeah, what if it doesn't?
I mean, the, the question just came to me and I think those couple of years of, of that sort of stewing under the surface, eventually I asked the obvious question, like, well, wait a minute.
Benjamin James Kuper-Smith: You questioned the assumption.
Aaron Schurger: I questioned the assumption, even though it seems so obvious. Um, and so I started thinking [00:35:00] more about it and I, I really can't say how the, the idea finally came to me.
I was, I was messing around with auto correlated data, so data that has this auto correlated noise, sorry. Um, it's noise that has this slow, it's dominated by the lower frequencies and has, it's, it's pink noise. What, what, what is often called pink noise. Um, and I thought, well, you know, what if the readiness potential doesn't mean that the brain is getting ready to move?
Um, what could it mean? And I thought, well, let's, let's, let's take a step back and, and let's say I was an engineer and I was asked to build a robotic arm. Could do Li's task, how would I do it?
Benjamin James Kuper-Smith: Okay. [00:36:00] That's a good question. Plan.
Aaron Schurger: How would I do it? Um, 'cause 'cause I, I was, I, I was interested in the fact that, in you, in spite of the fact that the task is really weird, li's task is, um, again, subjects do it.
So that begs the question, how does the brain do it? I mean, there are probably more than one ways of solving that problem, but, okay. It's a fair question, right? How does the brain do it? Um, so I thought if I had to build a robotic arm that does the task, how might I do it? I thought, well, I, the, the, the, the, the first and and simplest answer that came to my mind was, you know, I would take the actuator and turn up the gain until it was close, very close to the threshold for actuating a movement of the arm.
And then just wait. And, and sooner or later it'll just make a movement. And, and in fact, [00:37:00] assuming that the, you know, that the robotic arm, that the, um, actuator and the robotic arm doesn't have direct access to its own internal noise, it'll be spontaneous even from the point of view of the arm itself, the arm, that whole system will be basically surprising itself, right?
Well, that is a solution to the problem of doing Libby's task. Um, it may, you know, may, is it the solution that the brain uses? I don't know, but it's a solution. Um, so that was really the starting point. And I thought, well, what happens if I take auto correlated noise and I set a, a threshold that's not too high?
You know, or I just bring the baseline of that noise up closer to threshold either way, and just wait. And then if I timelock to those [00:38:00] threshold crossings, and I look at the average shape of the noise as it's approaching the threshold. Um, and when I first did that, I was just, I, that just totally blew me away.
I said, but that's the readiness potential.
Benjamin James Kuper-Smith: Yeah.
Aaron Schurger: Um, I, I, I, I was just at that point, I was just messing around and, and then I, I saw the readiness potential on my screen, uh, and I had gotten it just by playing around with this really silly, simple model. Um, so that then it, it all started to come together. I thought, well, you know, that then maybe the readiness meant maybe my, you know, this gut feeling I had, that there's just something not quite right about all that.
Um, it, it's, it's exactly that. The, the readiness pencil doesn't necessarily. Have to mean that your brain is getting ready for something. Um,
Benjamin James Kuper-Smith: yeah. I mean, in a way you have a much simpler explanation, right? Just random fluctuations that cross the [00:39:00] threshold and then using a bit of statistics makes it looks like this kind of, uh, ramping up, even though that's just a artifact of the way we analyze the data.
Aaron Schurger: Yeah, yeah, exactly.
Benjamin James Kuper-Smith: Although then my question is, so, and this now goes back to my question earlier about single cell recordings, um, in humans and animals, so that you don't average across many trials, right? I think you said you have this increase of neuro activity. Is it then more something like the, the input than
Aaron Schurger: has this No.
You, you also average over many trials. Oh, you do have single units.
Benjamin James Kuper-Smith: Oh, sorry. Okay. I thought it meant like, ah, I see.
Aaron Schurger: I mean, you do, if you look at a single trial, you can vaguely see a, a, a, some, uh, slight. Increase in firing rate leading up to the movement. Um, but if you look at a, if you look at a ster plot, you can see it's very noisy.
Um,
Benjamin James Kuper-Smith: uh, I guess [00:40:00] somehow I thought as soon as soon as it was single cell, it was also single trial somehow made. There's like two assumptions at the
Aaron Schurger: same. No, no. It's, you, you know, those, you still average together many trials to get this nice ramping shape on any one given trial. It's still quite, quite noisy.
Um,
Benjamin James Kuper-Smith: well that answers my question. I was gonna ask like how you can combine these, like how your theory can be correct if, or your model could be correct even on these single trial levels. But if that's not what you meant then
Aaron Schurger: right. It it applies.
Benjamin James Kuper-Smith: That answers it. Yeah.
Aaron Schurger: Yeah. It the same explanation can account for.
Increase, you know, the buildup in single unit firing rate. Um, but there are some assumptions that have to hold and, and, and, you know, for example, there would have to be a sort of baseline firing rate in those neurons that was above zero so that there's room for it to fluctuate up and down. [00:41:00] Um, if the firing rate of those neurons is just basically pegged at zero all the time, and then suddenly it starts to ramp upwards, um, then it's, it's more, it's, right, it's difficult to argue that that's just noise that's spontaneously, you know, randomly crossing a threshold.
Um, but you do have, you do typically have a non-zero firing rate, um, in, in, well, most neurons, in fact, in the brain. Um, but. Uh, that's something that hasn't been formally tested yet. So that's a, I mean, that's one way that you could challenge the model, um, is, is, is to, to, to show that. Um, so I mean, I think that's one of the strong points of the model is that it's, it's something that you could empirically test and you could disprove it.
Um, but I, I, I think the important thing about it is that, um, it in [00:42:00] my mind is much more parsimonious than the classical model. Uh, and so I think the, the, um, the say, the burden of proof rests with first disproving, this more parsimonious explanation before you go onto an explanation that involves the brain taking a second to get ready for a movement,
Benjamin James Kuper-Smith: which is also not an explanation at all.
I mean,
Aaron Schurger: I, I guess it's not a really an explanation. But's a description. Yeah.
Benjamin James Kuper-Smith: So one question I had or I have right now is that it seems to me that your model is much more, um, it seems to me that that applies to many more situations than just a movement. Um, so that in any kind of decision making where you basically, I mean, you could say you have no evidence for, so you have two options.
You have to choose between a OB, let's, um, but you could also say maybe you have no preference, like you don't care which one you take. So you need some sort of, [00:43:00] I mean, it seems to me almost like it's just like a random number generator where you just, uh, that you use like to flip a coin between two options.
Um, so I mean, do you think that this model then therefore makes the readiness potential much more general, that you also might find in non-movement situations?
Aaron Schurger: Y yes, so, sure. I mean, you should basically, this the, the general idea applies. Um, any, in any situation where some sort of. Event, uh, is determined by a threshold crossing, and that happens in the presence of auto correlated noise.
Uh, then, uh, in any such system, if you time lock to the moments of the threshold crossing, uh, you'll see this kind of buildup in advance. So that means that you could, what, maybe [00:44:00] you could find something like a readiness potential in the stock market. Um,
Benjamin James Kuper-Smith: yeah, that's one question I had. I saw you had like a, a paper on that, uh, which I was just really surprised to see that can actually, can you briefly talk about that?
Aaron Schurger: Yeah. So we had, I, I worked with uh, uh, someone who's, who's, um, in finance. Um, and he came up to me a after a talk I gave at a conference and was very interested in, in. In this idea. Um, and, and then I thought, you know, well, I had thought about it for some time as well, that in, in kind of any system, it doesn't really be a brain, any system that exhibits, um, auto correlated noise, uh, and where some decision is, is, uh, dependent on a threshold crossing, then you'll get, if you, if you time lock to the threshold crossing, you'll recover the, the characteristic, uh, uh, say quote unquote shape of the noise, [00:45:00] um, which will look like a buildup.
And so we thought, well, you know, maybe, maybe we could look at stock market data and see if there's, uh, you know, we tried, we tried to, to think of what's the analogous situation. There's an event. Um, the event could be, uh, uh, you know, uh, a, a buy, um, or a sell or, or, um. Or it could be, uh, what we looked at as events.
In fact, were when the, uh, when some index, when some indicator in the market crosses its own 50 day moving average, um, which for some reason, uh, that I even don't understand, is something that people look at to help make decisions about when to buy and sell. Um, I think even, well, even Patrick agrees and, and others agree that this is not more than just reading the tea leaves.
Um, and, and people [00:46:00] still do it all the time. Uh, there are traders who, you know, they, they, they look at these things and they say, oh yeah, when it, you know, when it crosses the, you know, the, the, the 50 day moving average, that means something that's a special and that means you should do X, Y, or Z. And, um, so you know, whether it means something or not, I don't know, but it's, it is something that people use.
Um, so it, it's a trigger for a lot of traders. Um, so we looked at that and we tried to, to look for, um, you know, a buildup, uh, in something called excess volatility, um, leading up to those, uh, events. Um, the, we, we did find some evidence, uh, for it, but I would say it's, it's not conclusive. Um, but there are so many ways that you could look at stock market data, uh, that I think we've [00:47:00] only just kind of just scraped the surface a little bit.
Um,
Benjamin James Kuper-Smith: is this a project you're
Aaron Schurger: continuing? Oh, I, I mean, not right at the moment, but I'll probably, yeah. Continue looking at, um, ways in which this same sort of idea might appear in different contexts. Like, like, like weather forecasting, for example. Um. So, um, you, you might,
Benjamin James Kuper-Smith: what would, yeah, can you ex, uh, what's the word?
Yeah,
Aaron Schurger: continue search. Yeah, sure, sure. I mean, um, you have
Benjamin James Kuper-Smith: elaborate, that's the
Aaron Schurger: word. Elaborate. Can elaborate, elaborate, yeah. Mean you have, well, you have, you know, again, trying to, to, to take some situation and, and break it down into an event. Um, and then some variable, some underlying variables that determine the onset time of that event.
Um, and where, where the, the kind of fluctuations in those variables are [00:48:00] auto correlated. Um, then there should be a, something like a buildup in advance. So in, in, in, uh, meteorology you have, you know, variables like barometric pressure, for example, that, uh, or temperature. Uh, these are all very much auto correlated.
Um. So, you know, temperature, for example, is very, very auto correlated. If you look at, um, uh, temperature over a, a a 30 day period, or 60 or 90 day period, um, you can see that the best predictor of the temperature tomorrow is the temperature today. Um, right. And so if you ask, you know, if you ask a child, a young child, what's the weather gonna be like tomorrow?
If they just say, same as today, they'll actually be right a lot more often than they'll be wrong. Um, so, so there is those, you know, those variables are auto correlated. So, you know, it [00:49:00] would be interesting to see if some event, like the onset of rainfall, for example, um, then we would predict that it would be preceded by this same shape, this sort of, um, exponential looking buildup.
Benjamin James Kuper-Smith: Uh, so what are, what are we basically, what's on our x and y axes right now? If we imagine, like, is it, I'm just trying to figure out, like are you saying like how much rain? Sorry. Yeah. What's the threshold? And
Aaron Schurger: so let's, let's, let's say, what's the event basically? I mean, this, I, I haven't, I should qualify that.
I haven't actually done any of this yet. This kind of work, this is just a, a, a, a half-baked idea at this point. But, um, uh, you know, I've gotten as far as downloading the database of second by second weather data, which took an awful long time because it's a lot of data to download. Um, the idea would be, for example, something like this, and again, I, I, you know, I might, I might be off on [00:50:00] this, but, um, for example, you might take the onset of rainfall as your event and your variable might be say, barometric pressure.
Um, and that's, you know, fluctuating about, um, we know that. That barometric pressure tends to drop, uh, before, before the onset of rainfall. Um, the question is can that, can you model that using the same sort of model where, um, right, where barometric, well, I mean, the key is that sometimes maybe get barometric pressure drops, even though it doesn't rain.
Um, that, that would be the, the, the important point be you have, you have this variable that's fluctuating randomly, uh, in a, in,
Benjamin James Kuper-Smith: yeah, I think I get what you
Aaron Schurger: mean, right. Temporal auto temporally. Auto correlated. Um, [00:51:00] and then there's an event, right? Should be, well, it wasn't raining and now it is. That's an event.
Um,
Benjamin James Kuper-Smith: yeah. Yeah.
Aaron Schurger: So, you know, you. I'm curious really more than anything else to see if this really, if, if it doesn't appear in just other systems, um, in nature.
Benjamin James Kuper-Smith: Yeah. This actually, um, relates to a question I wanted to ask in general, which is one thing I always liked about how you approach, I guess, science in general is that it seems to me you always take a very interdisciplinary aplo approach and kind of, um, it seems to me, at least compared to many other people, I speak to you how your inspirations seem to be a more broader or broader.
Um, and I'm curious, like, for example, like the, I think that the, at least to me it seems like the natural thing to do when you have like this model of the redness potential is to look at [00:52:00] other, you know, decision making. Like you move one small step, um, you know, you slightly change the task or something, uh, and then you get like, you know, something's very similar, but you now went like.
Not even outside of psychology. You went outside of like living, being right, right. You went right into meteorology. Um, so is that, was that like a, um, how should we say, was that like an intentional thing to see like, how far can I take this? Or like, for example, like, yeah, why, like why, why study the weather and
Aaron Schurger: Well, any, I mean,
Benjamin James Kuper-Smith: you know, something else,
Aaron Schurger: but it's, it just, I mean, that sort of thing occurred to me by, you know, doing the kinds of simulations I was doing and, and you know, working with, um, auto correlated noise and, and the threshold.
I mean, I saw it in front of me on my computer screen, uh, and what I saw at some point, I said, well, that's just auto correlated noise. Um, that doesn't have to be brain noise. Uh, or it doesn't have to be [00:53:00] EEG noise. Um, it would, it, it would seem that anywhere that, I mean, this ought to. Come up anywhere that you have oughta correlated noise in a threshold, and you ought to, you ought to see this.
Um, and, but, but it is true. I've always liked to think, um, and, and, you know, work with people who are working in other disciplines and, and think kind of broadly and get a little crazy. Um, and, and, um, I, I don't really know why. I just, uh, always liked that. Um, and, uh, I mean, we got, I mean, the, the, the measurement that we used in, in some studies on, um, brain activity, in, in perception, uh, came from a, a, a method that was used to measure the persistence of ocean currents.
Um, and, uh, you know, it, it, [00:54:00] it, I mean, it turns out, it, it mathematically it works extremely well to measure. Other things as well, like, uh, patterns of brain activity. Um,
Benjamin James Kuper-Smith: so it's just something you Yeah. Just kind of naturally think about or, uh, drawn
Aaron Schurger: towards? Uh, I, I guess, yeah. I mean, it, it, um, you have to be, I have to be careful not to get too distracted.
Um, 'cause at the end of the day, I'm doing neuroscience and I, I have to stick with what I know and, you know, not just go off on some tangent. Um, I would never get anything done. Right.
Benjamin James Kuper-Smith: Because I guess we have like then like 50 minutes left or something. 20
Aaron Schurger: minutes. Yeah.
Benjamin James Kuper-Smith: Um, yeah. Um, shall we maybe then just talk about like what you are doing now or in the next, or want to do in the next few years or something?
I guess some of the stuff we mentioned probably. Will be part of that. But um, as you mentioned, you, you also do psychology, [00:55:00] neuroscience, so there, there's, there is still that, so what are you, what are you planning to do there or are doing right now?
Aaron Schurger: Yeah. Well, I mean, some of these, so like I said, I think, and, and, and this came up in, in a talk that I, I just gave with, with Adina Roski, um, that, you know, the bottom line is I think when it comes to the readiness potential and this buildup, um, what we need I think right now is just intellectual humility.
Just, we, we don't really know what this is. Um, you know, we have, there's a, there's this classic model, which, which, uh, has a lot of problems. And, and, and it, it's, it, there are lots of reasons to criticize it. There's this stochastic model that we introduced in 2012 that, um, maybe is more parsimonious, but it's not perfect either.
Um, you know, who knows? Maybe the reality is that, you know, there's, there's, you know, [00:56:00] uh, some fraction of the variance is accounted for by stochastic fluctuations, but then there's some other, other stuff mixed in there. Um, and, and, and, you know, it could be that some subjects do the task in a different way than others.
I mean, uh, it, it almost certainly there are other things mixed in. I mean, if you're doing a task like Lipitz and you say, well, you know, on every trial you start over again and you say, ready, set, go. And then you wait as long as you wanna wait, and then you lift your finger, right? Um, when you do the task, and I can just, you know, I, speaking for myself when I do the task or have done the task in the past, you start thinking, well, you know, on the last couple of trials I waited kind of a long time, so maybe this time I should not wait so long.
Right?
Benjamin James Kuper-Smith: Yeah.
Aaron Schurger: Try to be random. Trying to be random. So, so there's that kind of, um, strategizing. That that can go on and that can factor into it. And we found in our data that that, that, you know, you're watching a clock while you do this, you're monitoring [00:57:00] a fast moving clock. Um, and the clock makes one once, it goes once around every three seconds.
Um, so it's 50 milliseconds per second on the clock. Um, and when, when you're, when you're doing that, it turns out for, for whatever reason, subjects tend to have a preference for the bottom of the clock. So, you know, there's, there's a, on average, there tend movements tend to be made more often around the bottom of the clock and the top who knows why.
Um, and, and there's some persistence in, in, in these behavioral data, so subjects on, you know, subjects, uh, if, if you waited a long time on this trial, you're more likely to wait 'em long time on the next trial. So there's some long, you know, long range. There's some sort of, you know. Memory, if you wanna call it that.
Um, and, and, and so you know that there, there are all these different factors, right? Um, that they're [00:58:00] all playing into this mix together. I mean, there's a recent study showing that when people do this task, uh, they tend overwhelmingly to move in phase with their own breathing. Um, so that you tend, and, and in particular you tend, you, you tend to plant your finger.
You tend to make the movement when you're towards the end of your exhalation cycle. Um, and by no means do you have to instruct people to do this. You don't ever say anything about breathing before the experiment starts. Um, subjects just naturally do that. Um, and when you think about it, it sort of feels natural, like breathe in.
Breathe out press, it kind of feels,
Benjamin James Kuper-Smith: yeah. Who'd press like at the beginning of breathing out.
Aaron Schurger: Yeah. That's weird. So, um, you know, it, it, it, it could be that the readiness potential is, is, uh, reflects a, a many different things that are all also [00:59:00] sort of cobbled together. But the, the, the important point is that we don't know.
We don't know what it reflects. And, and I think that, you know, if the, if the work we published in 2012 had any effect, it was to just sort of throw a wrench in the works and to say, you know, look, we, we, we don't really know what this is. Let's Wes set about trying to come up with some different explanations and, and try and figure it out.
Um,
Benjamin James Kuper-Smith: mm-hmm.
Aaron Schurger: And so one of the things we're, one of the things we're looking at, for example, is that, um, you, again, the readiness potential is something that you observe on the average after you average together many trials. And in fact, it's, it's, it's something that. Really takes shape. When you take the average across many subjects, many participants, if you look at individual participants in the experiment, the shapes of those readiness potentials is, is as, you know, unique as the subjects are themselves.
I mean, everybody's is a little different. And, um, what's [01:00:00] interesting is this canonical sort of exponential looking ramping shape that you see in the average across many subjects. It's hard to find in individual subjects who has a readiness potential with that shape. Um, there are, you know, you'll have in any group, there'll be a few, um, but then they just, uh, you've got every kind of crazy shape and, and all the way to the, to the extreme of where you'll have some subjects who just don't have a readiness potential at all.
They still do the task just fine,
Benjamin James Kuper-Smith: just kind of a
Aaron Schurger: flat, flat line. And then all of a sudden at the very end there's a spike, which is, you know, obviously the motor cortex at some point. Becomes engaged, you're always gonna have that. Right. Um, but yeah, so, you know, one thing that bears explaining, I think is why some subjects don't even have a readiness potential.
Uh, and, and what does that mean? Um, so explaining the variance across subjects, uh, is, is I think that's very worthwhile. And I think that whatever comes of that, [01:01:00] it will help us to better understand what, what the RP means, um, right. Um, and trying to look at in general the factors that influence the shape and amplitude, uh, an onset, um, of the rp.
Uh, and, and you know, there are still, you know, there are, there are more predictions that the, that the stochastic model makes that we could test, um, to see, for example, uh, it. It predicts that the readiness potential should be, um, uh, cha have a different shape depending on how long you waited before you moved.
Um,
Benjamin James Kuper-Smith: you should be able to analyze that with
Aaron Schurger: existing paper we have. So we have looked at it in one study. Uh, I, I, that's published actually, we did find a difference. Um, but that bears,
Benjamin James Kuper-Smith: oh, sorry, which
Aaron Schurger: paper is that? That's in 2018. [01:02:00] Uh, that's a 2018 paper in eau and, um, we did find that, but that bears replicating, I think, and, and, and, you know, looking into some more.
Um, so, so I think there's a, there's a lot to do. There's a lot that, uh, can be done. But I think what we should be focusing on is just trying to explain what this, what this ref, what this reflects. Um, one thing I sh I guess I should say one thing that we've been looking at ref recently. Is, um, the role that anticipation might play in the readiness potential.
So there is, uh, to make things, you know, more, not less confusing. There is another cortical potential called the SPN, the stimulus preceding negativity, which is another sort of shallow, slow buildup, um, that looks very much like the readiness potential, uh, at [01:03:00] least in the earlier, uh, a little bit earlier in time.
Uh, towards the, as you get closer to the event, the readiness potential becomes more lateralized and the PN doesn't become so lateralized towards the end, but the early, the early phase of it, they're most indistinguishable. The PN is something you get when you're just anticipating the arrival of a stimulus.
Benjamin James Kuper-Smith: So you mean like, let's say you see something every two seconds or something?
Aaron Schurger: Yeah. Or, or if you see something two or three seconds after a a, a specific queue. So let's say the fixation cross changes color, and then you know that in, in exactly two seconds, something's gonna appear on the screen. So for those two seconds, you're anticipating it, you're waiting for it, and then it appears you get this, this, this buildup, um, that looks very much like the readiness potential.
So, you know, we wondered, well, maybe, I mean, if there's any, if there's anything fundamental in all of this, um, [01:04:00] it seems that an it's anticipation that, that all of these contexts have in common. So when you're performing a movement, even in Lis task, um, you are anticipating and you can't help it. I, I doubt that the brain could help but anticipate your own movement.
It, the, your, you know, your no one can predict probably, uh, what your brain is gonna do next, better than the, your brain. Uh, so it's probably anticipating your own movement. And, and not just that, but it's anticipating the proprioceptive, uh, consequences of that movement. Feeling the button. Well, the, and, and it's, yeah.
The, I should say the sensory and proprioceptive consequences of that movement. So you're anticipating the click the button is gonna maybe click and, and maybe something's gonna change on the screen when you press that button. So there's all these sensory [01:05:00] consequences. But yeah, there's proprioceptive con consequences.
You know, the skin is gonna stretch on your finger, muscle's gonna flex, um, joint positions will be different. So I, I, I, I think it, it's a very reasonable assumption that your brain is anticipating that con, that that confluence of events. Maybe, maybe the, you know, the early part, at least of the readiness potential reflects that anticipation.
So we did an experiment where we tried to control for anticipation. So we had subjects do this task where they're watching a slideshow and, um, just pretty pictures, you know, and, and, and they, uh, on some trials, for some pictures, they have to press a button to advance to the next photo in the slideshow.
Uh, and sometimes they, they didn't, sometimes they just had to wait and the slide would advance by itself after several seconds. Um, and the time that would elapse on those automatic trials was [01:06:00] drawn from their own distribution of waiting times on the trials where they advanced the slide themselves so that by the end of the experiment, you had these epochs in time that either ended with a movement or ended without a movement.
But they were well matched in terms of how much time elapsed. Um, and so when we, when we look at those data and we try then using some very powerful machine learning techniques to distinguish between those movement and those new movement epoch, uh, we can't until the very, until the very moment, until roughly the moment that the movement is about to start.
Benjamin James Kuper-Smith: So when you actually get like the motor cortex,
Aaron Schurger: when the motor cortex really starts to become engaged, then you can start to tell them apart, but not before that. Um, as if, you know, that anticipation was accounting for most of the [01:07:00]variants potentially in, in, in both cases. Um, and you might say, well, that's just a null result.
You couldn't tell the difference until the time of movement. But there two things speak against that. One is that at. At the time of movement, you can tell them apart exceedingly well, like near, near perfect, uh, which is rare in machine learning. Um, and also when we don't control for anticipation, when we do the analysis in a slightly different way where we compare the signal to itself earlier in time.
So we, we, we take a From
Benjamin James Kuper-Smith: the same trial or,
Aaron Schurger: yeah. So we take a piece of signal from early in time and we consider, we say, well, this is far, this is three seconds before any movement happened. So we'll call this a no movement piece of data. And then we will take other windows of data that are progressively closer to the time of movement and compare them u using the same machine learning [01:08:00] technique.
Um, and so when we do it that way, so here we're just looking at the movement trials. Um, then, then we can tell, we, we, uh, can distinguish these, uh, movement and no movement bu extremely well, even well in advance. Um, even, you know, a full second in advance, um, when we don't control for anticipation. So it's not like the machine learning algorithm doesn't work,
Benjamin James Kuper-Smith: it just, it really seems to struggle between those two conditions,
Aaron Schurger: right?
So that suggests maybe that, um, anticipation is, you know, uh, is, is a big part of, of the readiness potential. Um, maybe the early part. Who knows? Remember all we know maybe the early part of the readiness potential is, is, uh, dominated is some sort of anticipation. Same signal,
Benjamin James Kuper-Smith: which is quite ironic, isn't it?
Because the whole point of the readiness potential is that it's [01:09:00] something that's self, like, it's kind of a voluntary thing that isn't, it isn't Q driven, right? But it seems to me almost you're saying. Driven to a queue you're about to expect almost even if that is self-generated,
Aaron Schurger: even if that is your, your own movement.
Yeah. Yeah. And, and there could, I mean, that doesn't, that doesn't preclude the stochastic explanation because it, it, it could simply be that you're, um, the, this, the level of anticipation in your brain, uh, is waxing and waning, uh, kind of randomly. Um, and that it, it covar with, with the actual probability that you will make a movement, um, from a sort of predictive coding standpoint.
Right. Um, so it remains to be seen, but we're looking at this from whatever angles. We can, including angles that might ultimately, [01:10:00] um, uh, uh, cast doubt even on what we suggested in 2012. Um, you know, the point is not to be right. The point is to get it right. Um, and, and so that's what we're, that's what we're trying to do.
Benjamin James Kuper-Smith: Okay. Cool. I think, um, we're pretty much running out of time.
Aaron Schurger: Yeah.
Benjamin James Kuper-Smith: Uh, so yeah. Thanks again for
Aaron Schurger: You're welcome.