[00:00:05]
>> Everyone, thank you for tuning in to our continuing series of seminars from local speakers that are not only New faculty members, but established faculty who get a chance to tell US about the research that's going on in their Labs, which they don't often get a chance to do so it's actually pretty good to revisit these so today we have here at Stanley probably everyone on the call knows garrets readers and a lot of introduction is one of the co-directors of the Georgia Tech neural engineering.

[00:00:46]
Along with, you know, was our speaker last week Garrett and we have been very instrumental in pushing the momentum for Neuro science and Neuro across campus and across College of science and College. When you're really, really good to have them at the helm for all of US that has been driving the momentum the last few years.

[00:01:12]
So there is a, he's an actual engineer who like a lot of US, he trained in mechanical engineering, both undergraduate and PH. D. and then he did his I was talking Euro science and that's how we got introduced into using engineering tools as applied to the nervous system.

[00:01:32]
And down at BERKELEY where he studied the visual system. And since then, he has moved on to another sensory system, which he's going to tell US about the very end road. Instead, it's about a sensory system using that pathway to investigative pieces of neural coding. He coding motioning to behavior.

[00:01:52]
So here, thank you. Taking away OK, we think before I share screen, it's want to put my face out there and let you know this is not a recording. I'm actually really here. So I'm going to go ahead and share screen and see. All right, well thanks very much.

[00:02:11]
It is really kind of special and fine to speak locally and the walls, right? We don't get a lot of time chances to do that. So you know, I did when I got this slot, I did recognize that it was going to be just post-election. So I knew that it was either going to be a really great idea or really terrible idea.

[00:02:31]
People would either be in it Rotten or maybe a slightly good mood. And so it looks like I gambled and probably mostly one but, you know, I also, I want to just say before I start that, you know, during all of the last several months with the pandemic and everything, it's been really tough on everybody.

[00:02:55]
I think, you know, just having this Domino series and recognition that the community still together and so on. And the law and SIMON have done a wonderful job of keeping this going. And it's really been for those who are New to the you know, the landscape here, you might think that this is just always been the case, but it's this event, this seminar series kind of represents the culmination of lots of years with a lot of people working pretty hard to kind of launch this and so I think we've successfully done that and it's been just a really wonderful thing.

[00:03:29]
And what also is the dream is that people from all different areas of research come together and come to a seminar, whether someone's talking about, you know, something at the cellular level or a human behavior or whatever. And I think that's important. It's not that easy to establish such a thing.

[00:03:46]
So I think we've done that and I really like the fact that people come every week regardless of whether it's lined up with their own work. So. So again, I just want to recognize the, the community, the g.t. Neuro community in their own genes that are just like all the good stuff that's happening during this craziness that feels like a little bit of a constant.

[00:04:08]
And then when I started to put the this talk together, I'll tell you it was, it's challenging because you think of all the things you could talk about and you know, I thought of about 15 different versions of seminars. I wanted to give and ultimately have settled on one that I thought was, you know, maybe would reach the most people and try to connect.

[00:04:31]
And so I said, sat down thought about the goals for a seminar like this. You know, one is connecting more broadly with people in our community. You know, gives me a chance to talk about some old stuff that people may not know about, but then connect it to the newer work.

[00:04:46]
That's always opportunity. And then, you know, like all the rest of the other p.r.'s that are local, you know, I'm really proud of our lab and want to make sure you know who the lab members are and what they're doing. And then finally, scientifically, I want to convince you that timing is important.

[00:05:04]
And so the title of the talk is timing. Timing is everything. And you know, I think it's time, you know, maybe another aspect of this, another angle is that everyone I've talked, you know, in the pen Demick seems to think the time is doing something close to going fast.

[00:05:21]
It's a little unclear and I think we've all kind of sort of hijacked in that sense. But, so I think timing is an appropriate thing to talk about. Well, so what I'm going to do is I'm going to talk about the, some of the work of our lab and across a bunch of different relevant timescales, and ask the questions about what are the relevant timescales for information processing in the brain.

[00:05:47]
Both from a fast millisecond timescale of action potentials and coordinated activity to you know, this kind of medium timescale, seconds to minutes. We study adaptation, other types of phenomena that operate on these timescales. And then finally, to a slower time scales of learning over days to weeks. And the other aspects of time have to be aware of is that this is kind of a New Assembly of slides.

[00:06:13]
So I have no idea how long it's going to take, but I'll do my best to get it on the fly. So this is our laboratory. And so we have a group of really talented people in the lab that are really excited to work with. And I also want to give a shout out to the lab because during the pandemic, it's been really great to have a group of people to interact with who helped keep the lab going, the research ramp back up and so on.

[00:06:39]
And it's just nice to have that, and I really appreciate all the folks in the lab. So that's 11. Good thing through all of this. So I'm going to start out on the fast time scale a millisecond time scale. And you know, a lot of our work really rolls around this and, you know, and as the law all alluded to, I'm going to, I'm going to dip back into the archives a little bit and talk about some vision related work when I 1st started working in science and so I started working in the early visual pathway, became really, really excited and interested in studying signal processing.

[00:07:17]
And I started studying the early visual pathway and became really excited about just this signal processing. It was, you know, total engineer. So as you know, light enters the eye through the lens and fall, and the photons fall on the back of the eye, the retina, and the photo receptors transducer, the photons into electrical signals that travel through the layers of the retina through the optic nerve to a really deep region in the brain called the 1000 is where the specific region that I worked on was the lateral Geneva the nucleus with the l.g. and then projects to the primary visual cortex.

[00:07:51]
And, you know, spent a lot of my time thinking about the Spike in activity of neurons in this particular part of the pathway and how that represents different parts of the visual scene. And, and at this kind of millisecond time resolution. And in particular, as it relates to the natural visual environment, those might me getting my feet wet in the field of neuroscience.

[00:08:15]
And one of the 1st things that, that I then worked on when I started my own lab was, is captured in this slide. So as I mentioned the following, this is this really strategic part of the, of the, of the visual pathway and almost all of our sensory pathways.

[00:08:33]
Somewhere in between the periphery of our spur for all sensors and the sensory cortices. And it's not just a simple part of the brain, and it's really, really complicated from an anatomical and cellular perspective. And cells in the, the doll in this actually have lots of inputs that are coming not only from the periphery, but from other brain regions as well.

[00:09:01]
As lots of projections that are coming from the cortex back to SOLOMON. But one of the things we locked onto early on was the fact that cells in this part of the brain have a particular type of channel t. type calcium channel that is normally in activated. But when inputs come into these cells that inhibit them, word or Hyper polarized them or drop the membrane voltage, then, then it d.n.a. activates this mechanism.

[00:09:31]
So when the excited for input comes along, there's a burst of x., potentially very rapid sequence of action. The tensions running on top of the calcium wave. And the important thing is that you know, those happen. But they induce a really strong impact on the downstream neurons that these, that these cells project to and in particular the sensory cortex is our target of a focus.

[00:09:55]
Now if you, if you Google this or look at, in the literature, most of the time people talk about this in the context of sleep and slow wave sleep and generating oscillations in the cortex and so on. And in fact, it's also implicated in generating certain types of epileptic behavior.

[00:10:14]
But Francis Crick came along and articulate. An interesting hypothesis was that it's not just about that, but potentially it serves a role in gauging the information flow to the, from the periphery to the cortex. And that's something that really resonated with me, this kind of idea and it was a bit of a, you know, it's a bit of a Crazy idea.

[00:10:37]
And maybe this mechanism that's typically associate with sleep is actually associate with some kind of signaling in the brain. And so we focused on that early on in my lab, and in particular we looked at the family activity in the early in the, in the fall of most of the anesthetize cat.

[00:10:57]
And this shows that indeed, this is not just reflective of sleep, but instead this shows the response of thalamic neuron to presentation of a natural visual scene where this is from Indiana Jones, where individual frames of a movie a movie has been presented to the visual pathway. This and this cat and as the leg moves into the visual field by about frame 5 or so between 3 and 5, there's this sudden burst of activity that's highlighted with these Red spikes here.

[00:11:30]
And that these are units of lead or suspected to be related. This kind of calcium bursting, and we show that it was actually at least correlate it with detection of change in the visual scene. So it could be very reliably driven by aspects of the visual scene. And we later went on to show that this is highly synchronized across the polemic.

[00:11:54]
And that it, it, it's really good at detecting change in the environment. Ok, so we became really in their mid with timing just as a whole and started to dig a little bit more deeply into this. And so specifically we started to look at Viking activity in these the Lamb ignorance in response to both natural movies as well as these this kind of white noise.

[00:12:23]
So it's a spatial uniform noise which is this flickering. Imagine a flickering computer monitor that every frame is changing the light level of different Gray scales, but it's moving really fast. And what we see, and what we saw at that time was that the neurons were individually when we look at one neuron across multiple repetitions of this, we're really, really precisely locked to the stimulus in this kind of this kind of noise stimulus.

[00:12:53]
But when we have natural scenes, which are a little bit more sort of flowing, we see these individual spiking activity was quite a bit more variable across time. So this is time on the horizontal axis. And so we see that it's not as locked to the vision in this natural movie case as it was with the not white noise.

[00:13:16]
And this is the work that I started a collaboration with the laboratory of Jose Manuel Alonzo in Sunni, in the College of Optometry and probably in the early 2000. And what we found in this work was that the, when we looked at the time scale there, and there's a lot of details, I'm going to sort of go the route of trying to cover some ground here and talk about high level concepts at least for some of the work, but some put folks may not be happy with the details I leave out, but the details are in a paper that's now you know, 13 years old, whatever.

[00:13:53]
But when we quantified the time scale of the visual stimulus versus the time scale of the response and the details are in the paper, what we found is that the, the, the time scale of the stimulus was much longer than that of the response. Or that the other way to talk about it is the knowledge were much more precise than the time scale of the visual stimulus would, would suggest it should be, or at least match to be.

[00:14:23]
So when we look at this, this ratio between the 2, we found that on average, the response of the neurons were about 3 to 4 times more precise then the inputs. When we look at the natural scenes, you'll notice that it's the same kind of trend when you plot one versus the other, the stimulus time scale, versus the response time scale.

[00:14:44]
We see that the stimulus time scales much longer than the response time scale there, all the points are below the diagonal here. The absolute numbers are much larger over here, but the, but it turns out that the proportion of these, the ratio of the time scale of the, of the input to the time scale in response was the same.

[00:15:03]
So the time scale response was again about 3 times more precise than the time scale of the input. So Despite the fact that these neurons get kind of sloppy in their little bit, they're significantly less more temporally loxton stimulus. They the ratio of the time scale of the input to the time scale.

[00:15:22]
The output was about the same between these 2 conditions. And we thought that was kind of interesting. And so what we showed in the rest of the paper was that when we considered this from the perspective of being able to reconstruct the visual input from the activity of the neuron, that we found that if the neurons were really, really precise, you could actually reconstruct the visual really well versus when the neurons were not that precise, then it's kind of ends up being sloppy.

[00:15:54]
And so when we quantify that, in terms of the information that the neural activity was revealing, were in coding about the visual input. We found that the neurons, if the neurons got more and more precise, that indeed the information would go up that's on plot, on the vertical axis here.

[00:16:14]
But that it plateaued at some point. And becoming more and more precise did not make. It didn't use the information in terms of this reconstruction. And so this was a very signal processing kind of sampling theory perspective that showed that there was some kind of potentially under a lot rhyme or reason to the timing perception of these individual neurons.

[00:16:41]
And what we found overall, well, I should say one more slide before I tell you the punchline of all of this was we then looked across neurons that were recorded from simultaneously in a pool. So we had multiple lecturers that were implanted into the US again and the nest ties cat.

[00:17:01]
And we used a bunch of different approaches and a whole range of studies. But we found a similar kind of relationship in that the timing, the tiny across the lines was precise. And did relevant timescale here seem to be on the order of about 10 to 20 milliseconds in the context of natural scenes.

[00:17:19]
Both in the context of what individual neurons did, and they're trying precision, as well as across pairs of the neurons within the pool of neural recorded simultaneously. So when responds to natural scenes, this 10 to 20 millisecond time range seemed to be really relevant. So we then started to open up the question a little bit more and say OK, well, you know, what does this really mean?

[00:17:48]
And why do I think timing is really important enough to talk about here by my bringing up some of my time talking about this. So you open up any undergraduate textbook and you'll see some figure that looks like this, where it shows the sort of concept of spatial and temporal, some Haitians.

[00:18:04]
So inputs are coming into a particular neuron here. And these priests and active neurons, the timing of action potentials in these priests and applique neurons matters in terms of what it does to the downstream not. So this one shows an individual actually attention or, you know, has some kind of effect on the downstream post napping.

[00:18:23]
Now on with that slightly depolarize that this is an excited story project in US. And then when 2 x. potentials occur close together, there's this kind of some Ation. So there's the 1st one and then the 2nd one has an impact. And of course, if you drive it hard enough, you potentially could reach threshold and fire and actually potentially downstream.

[00:18:45]
Similarly, across neurons, we have neuron one neuron to the timing of these matters when they're not so close together or not synchronous in time. Their impact is kind of diffuse, but when they're coordinated and synchronized across these neurons, that impact summits. And what's really important here is that it's, you know, it's more than just the simple thing.

[00:19:06]
It's, it's been shown in a number of really nice studies that in particular, from the 1000 most to the cortex, the elegy end of the one here. This intersection here is really sensitive to the time to these inputs. Neurons in the, in the cortex, in the brain, in general, are not really driven by single synaptic inputs, but take the coordinated activity of lots of neurons.

[00:19:30]
It's been estimated that about 50 to 100 sort of primary in puts, in this particular part of the brain are responsible for our a really sort of driving the output of driving the activity of these individual cortical neurons. And this 10 to 15 millisecond integration window is really, really critical.

[00:19:52]
So when there is coordinated firing, either rapid succession of action potentials from one neuron or coordinated fire across the rounds. This is a, this is an abscess, very, very sensitive timing. And that's really what gets through here, and it's a very Super linear process where it's very, very synergistic in driving the cortex and what the cortex responds to.

[00:20:15]
So I'm making a Big deal about this because this is what, this is what our visual pathway and other sensory pathways are dealing with. So we open this problem up and started thinking about pools of neurons. This is a group of neurons recorded simultaneously across actually several, several clusters of neurons recorded independent separately, but clusters recorded simultaneously in the elegy and in response to this synthesize scene.

[00:20:42]
And you can see this kind of coordinated activity as, as the, as we move through this kind of videogame environment. And we started thinking really care for the, about the coordination of activity across pools of neurons and what impact that would have downstream. So of course, what thing that you might immediately start to think about is the fact that in one thing that emerges in going from thought to cortex is that of orientation selectivity.

[00:21:15]
So one of the primary, emergent features in primary visual cortex in the 1000000 visual pathway is that of many Taishan tuning that's, that was discovered by he believes, all in the, their work in the Earth, in the Fifty's and sixty's. And of course, in this case, neurons in the, in the visual cortex are sensitive to the orientation of bars, of light that are presented into the, in the visual and the visual scene receptive field of these neurons.

[00:21:42]
And it's captured here in this kind of conventional to curve where the neuron is most sensitive to 0 degrees and less so to other angles. But, but in real life, things are much Messier and you get in this kind of polar plots. So orientation tuning, for example, at a particular angle.

[00:22:01]
Sometimes it's a little bit broad, sometimes it's really sharp. And other times you get direction selectively. So it responds to motion in one direction, but not. And the proposal from human Weasel was that this was wired up. This is just an, a construct of the anatomy that neurons in the retina in the 1000 masel g.n. don't have this property, but the way the neurons project to, to the primary visual cortex is how you get this, that there's an alignment of projections across particular axes.

[00:22:31]
That give this kind of orientation tuning and the reason this is important, this feature of the visual pathway is it was thought at the time to be a precursor to the segmentation of different objects and elements of the scene to give US think higher level things like feature detection and object recognition but it turns, and it turns out that the anatomy doesn't really support this kind of, of construct.

[00:23:01]
All of the models that have been constructed from about visual orientation tuning have relied on this really sort of extreme constraint of wiring coming from the retina to the thought most of the cortex. So these are some of the prominent models that we sort of to look at, and they all rely on this fact on this idea that the neurons that are giving inputs to the cortex are along this kind of very long axis of projections.

[00:23:26]
But in fact, in reality, it's not really wired up that way and also direction. I mean, direction selectively was proposed to arise from the fact that neurons that are receiving these kinds of inputs to be these long sort of delays that would cause the delay of this input here to arrive at about the same time as the input here.

[00:23:51]
Action is just a click on the cell and it would go in the opposite direction, wouldn't cause the same kind of effect. And this is a simple reichardt detector, but there's really little evidence for this in the 1000000 visual pathway. And it's interesting how the field is kind of moved on past this idea.

[00:24:07]
So we took a shot as we started to look at the Slam ignorance and the timing of the Linux activity. And so in particular, we looked at coordinated spiking across individual neurons to neurons. Recorded simultaneously, that had highly overlap, perceptive fields, and simply considered the synchronous activity of these neurons.

[00:24:27]
These what our neurons likely project to a common core of target. And by considering the synchronous activity you're looking at, the activity is really going to drive the downstream neurons that these neurons are projected to. And so when you toss this kind of polar plot, we see that neurons, when you consider the synchronous activity, you can get orientation selected out this lack of orientation selected to be a circle here and so, elongated in a long axis, years orientation selective.

[00:24:57]
And what we found is that you're the Big law moving around recorder from simultaneously in all the pair Wise combinations of synchrony and the kind of tuning that we would see. And we saw a very, very rich sort of substrate of tuning Pretties across these kinds of neurons Despite the fact that really all on top of each other.

[00:25:18]
So it's not this nice organized thing that we're led to believe in the textbooks. I don't have a lot of time to go into the details of that, but please take a look at this paper if you're interested in that kind of thing. And what we found, sort of the punch line here is that we could use this kind of approach to talk about a lot of features of early visual processing.

[00:25:39]
So like morning Taishan tuning temper frequency tuning contrast and variance and direction selectively. This kind of emerged out of this very simple timing argument. Ok. So again, or maybe I should say that the timing that we considered in that case was on the 10 to 15 millisecond time window to define this level of synchrony.

[00:26:01]
So bringing this fast forward into the more recent experiments we've dug a little bit deeper, the tools have changed and we now have the ability to look at little things that move and and we use put in particular, the virus a system of the road. And we've used both rats and mice, and I'll kind of present a little bit of each here.

[00:26:25]
But the basic idea is that this pathway is very, very similar to the visual pathway in terms of the basic anatomy of going from the periphery in, in this case, through the brain stem to the following. This and particular region called the pm, which is analogous to the algae and, and to the primary, some medicine, 3 cortex and this case called Barrow cortex which is very much like the one.

[00:26:50]
And so I'm certain a lot of neuroscientist would be sort of rolling over in their graves right now to hear that, that I'm calling these pathways. Very similar but, and from some perspective, they are similar in the process. And so it turns out that mechanic receptor is in the face or around the individual hairs, or the whiskers are transduced the mechanic Bishan whiskers come in contact some objects and signals through the pathway in a very similar way that I describe the visual pathway.

[00:27:24]
So I want to time to go into a lot of detail, but more recently, a former p.s.u. student, Peter Borden, and current post-doc Caleb Wright, have been working on digging into these kind of timing things in the both Almos and cortex in mouse using a combination of tools that are sort of bringing, letting US look at this in a lot more detail.

[00:27:50]
Though, for example, this includes wide field goal to diminishing as well as a cortical limiter program courting, as well as recording in the fall mist, and then using up the genetics to manipulate the phallus. And again, I don't have a lot of time to go into the details, but in this particular case, and this is in a way that mouse had fixed mouse by Hyper poor rising.

[00:28:13]
The 1000 most with this kind of light input using an L E D. And the Hyper polarizing option pillar of Upson. We're able to push the 1000 less into this kind of burst regime. And so along comes a sensory input, in this case a Whisper input. And we can show this kind of bursting activity I talked about in the 1st few slides.

[00:28:37]
So there's a couple things to note here. One is that the bursting does exist in the awake brain. Even the control condition is reduced compared to what we have served under anesthesia historically, but it's there. And you can really sort of turn the Knob on this thing just by providing this kind of hope.

[00:28:57]
I proposed rising inputs. This touch on the thought, this type of polarizing the subsequent sensory input comes along. It's this Big burst of activity and you get this boost of signaling at the level of the thought of US. Surprisingly what we see in the cortex and we record downstream with voltages aging is we see in attenuation, a slight attenuation of the cortical response.

[00:29:20]
Even though there's a boosting of this, well I make input and that really sort of surprised US. But what we also see is that there's a spatial sharpening of this, and so it's, this is been normalised. What has the same peed, but what we see is a much more spatially restricted region of cortex that's being activated.

[00:29:40]
So there's a lot more details to this that I don't have time to talk about. What we think is happening is that it's a combination of the level of synchronization that it, that's induced at the level of the fellas in this precise timing across not the fellow must, as well as the sensitivity of the synapses in the cortex and the differentials of sensitivity across different neural neural subtypes in the cortex.

[00:30:11]
And this induces a synchronization of the inhibitory neurons and cortex that causes this kind of startling. So again, that's kind of a mouthful and probably only for the specialists in this area. But something we're really excited about, this is kind of not yet published. Work that we're pretty excited about.

[00:30:28]
So I'm going to, in this part on the fast timescales, just by mentioning that, you know, the downstream downstream brain structures are really sensitive. I mean, 10 to 20 milliseconds is, is a really relevant time scale that it's not reflecting synaptic integration Windows, but more network integration into those that rely on the interaction between excited for inhibitory subpopulations neurons.

[00:30:54]
But think about this, there's a window, it gets your spikes in, but then the door Slam shut. And this seems to be the relevant time scale, 10 to 20 seconds. And that this kind of timing provides this timing mechanism, provides US this dynamic engaging that we think controls information flow from the periphery to the cortex.

[00:31:17]
And I do want to emphasize it's really not enough just to look at firing rates of neurons. That the timing matters a lot. And so the majority of studies really just tend to look at activity and see an increase in activity or decrease activity and sort of call it a day.

[00:31:32]
But it's not really enough actually. And so what makes it through to the next downstream brain structures really has a lot to do with the timing. And this could really be a fundamental principle signaling. And I also really want to emphasize that revolving through all this work we're sort of woven through this is right or wrong, some theories about how signaling is happening.

[00:31:56]
So in the tonic burst to case I didn't mention tonic is the non bursting activity. This kind of tonic burst activity that you either transmitter detect, detect inputs in the sensory field, this kind of dynamic regulation of timing. The emergence of 2 New properties to feed forward wiring and so on.

[00:32:18]
These are just theories of how of what might be happening but are and might be right, might be wrong, but are important in terms of sort of motivating the studies that we conduct. And it's interesting to watch a lot of the field emerge with the technology that's enhanced and exploding if you like, we sort of lost a lot of these kind of theories about what might be happening.

[00:32:41]
It seems like we're just collecting a lot of data as a field, but I think it really needs to be framed by underlying theories and hypotheses of what we think is actually happening. And the other thing is that what, what I started to notice in the middle of all this work and it really isn't theme is like, there's often what people believe in the field, scientists believe, versus what's been proven.

[00:33:06]
And it's kind of interesting to notice what people sort of think is true about the nervous system, or maybe this is distributed in general in life versus what's asked actually factual. And so I would just encourage you to keep an eye on that very closely because a lot of things that people believe in their, of science and maybe others.

[00:33:25]
Well, it's just something that gets repeated, but there's actually when you dig in, there are not really, there's not really evidence for this. And so for example, you often hear bursting activity is doesn't really occur to one a person doesn't really occur in the way brain and so on.

[00:33:40]
But you're actually really hard to find definitive studies that have actually known that down. And I'll give another example of that in a moment. So, I wanted to mention before I move on to the next section that we have a couple of projects that are ongoing right now that relate to this.

[00:33:55]
And in particular, in Leiden, one lives a PH, d. student, a lab, and she's working on the same circuit, but looking very closely at the projections from the cortex back to the fall of US. And using a particular transgenic mouse line that lets US target neurons that are specifically projecting from cortex back to following this.

[00:34:18]
And being able to manipulate those. And we're really excited about this because this is an often overlooked fact that there's a lot, a large majority of inputs coming to the 1000 most are actually not from the sensory periphery, but coming back from the cortex. And we largely just don't know what this does actually in the field, so look forward to, to being able to report on that.

[00:34:40]
And then are we Palos post-doc in the lab who's working on looking it's integration of sensory signals across different sides of the Bidens case. Different sides of the face of the whiskers and what impact that the timing have on the getting information flowing that comes through again, you know, be on the lookout for that.

[00:34:59]
We hope to be reporting on that. And then a couple of New members Jacqueline and Adriana are New lab members and are just getting started this year. And Jacqueline is working on a New project that really is about looking at dynamics across different spatial and temporal scales using voltage imaging as well as electric recording.

[00:35:21]
And then age Miano is working on close with feedback control of neural circuits. Something we've been working on the lab service for the last several years and controlling things on fast timescales. Ok, so I'm going to quickly move on to medium timescales, in our lab work a lot on adaptation.

[00:35:45]
And processes that occur on seconds to minutes. And I'll try to move, I'll try to move faster and I think I'm around that time. A couple things to say about adaptation. So, you know, it's a ubiquitous feature of the brain. And we all kind of use that term, right?

[00:36:03]
Colloquially, we talk about things like adapt diet, right? We're, human beings are very adaptive as are all organisms at changing environment. And you have to be right. But that term is used very sort of broadly to describe how you might change in response to a changing environment. And in the brain, it turns out there's many different forms about a taste in many different ways in which the brain actually changes in response to the environment over lots of different time scales.

[00:36:31]
People who, who talk about mechanisms would see these all very different things. But if you think a lot about function, you might see them as similar kinds of functional functional things that operate on different time scales. And that I probably fall into that camp. But in particular, we've studied a lot REQ rapid.

[00:36:47]
Since we had a patient operating a milliseconds to seconds, time scales. And so if you're interested in this topic, in general, we wrote a review article a few years ago on this that it's pretty exhaustive on it. So before way, before I share this, I want to say I'm going about to show this.

[00:37:05]
So what is adaptation? We give you an example of a rapid adaptation. The what I want you to do is Stare at the middle of this spiral. And this is a really interesting version of something called the waterfall illusion that goes way back to the police there at this and just fixate in the Center just for a few seconds.

[00:37:25]
It doesn't take very long and I leave it on for just before seconds. If a watch is too much, it really freaks me out, but then I'll turn it off and if you just Stare at the same spot, you'll see it has this kind of lingering aftereffect, right?

[00:37:43]
And it's really dramatic, and it's, you know, it's something that's been studied a lot. It perceptually, this kind of adaptation on very short timescales. We can show how to different ways this one in particular is thought to be due to the differential adaptation of neurons in the visual pathway that are coding for different types of motion orientations.

[00:38:07]
But it's, I'd like to argue that this is more than just a parlor trick. It's more than just a trick to impress your friends at cocktail Party. Other I have found it's been very useful over the years and various social gatherings. But I'd like to argue that it's something more than that.

[00:38:23]
And you know, of course, I'm not the 1st people to think about this and a couple of people that I really paid attention to their thoughts and work on one is hard. Barlow who actually just died this summer who was a long time champion of talking about adaptation and processes in the brain and giving some reasons behind organization function.

[00:38:46]
And this adaptive nature of the brain. The idea that it maximizes information transmission and sort of how much question, how much information is being transmitted is somehow being regulated by these kinds of processes in the nervous system. And then GEORGE run back and she is another person who influenced me a lot.

[00:39:07]
Did won a Nobel Prize for his work in the Coakley a and spent a lot of time talking about all sorts of things where you go back to some of the old white guys. And they talked about all sorts of things and speculate wildly. While here in a lot of their work, and one of the things that really triggered me in his work was discussion about the fact that these adaptive processes could be switching, changing what is being towed.

[00:39:34]
So that's something that we really focused on a lot. Again, in the context of the by a verse by verse a system and in particular we started looking at the fact that when we navigate the World, we detect, we detect things, did something happen or not? Did something come into my visual scene?

[00:39:48]
Did something touch me or not? Did I hear sound or not? Yes or no versus discrimination. What was it? What was it that I just saw? What was it that I just heard? What was it that I just felt? And so there's kind of this interesting dichotomy here, and these 22 things are not necessarily dissociate will, but we can study them in and we can study them separately.

[00:40:10]
And this is a study of things like this have been studied a lot in psychophysical literature, but the mechanisms underlying these things have been have been really elusive. And so I'm going to just quickly mention the study that we did a few years ago where it's from cheong, where he recorded from cortical neurons and thalamic neurons in the rat Madison's or pathway or tactile pathway.

[00:40:35]
Using these computer controlled inputs and simply asked from, by looking at the neural activity that we record in these different brain regions. Could we, in the simple detection task, could we determine what was signal versus noise was something there or not? In the case where he either adapts the pathway with some kind of persistent input or not.

[00:41:01]
Similarly, he asked, well, what about if I wanted to discriminate between different inputs? So if I have ended from inputs here, in this case they were just different velocities of whispering inflections on the, I'm a rat. Can I discriminate between these different things? In the case where I adapt the pathway versus not and what he found was that the adaptation makes the detection worse.

[00:41:26]
So you get worse of saying Yes or no, I felt something, but you get actually better at the ability when you can detect it to be able to discriminate. Was it one with a stimulus 1234 or N? And again, there's tons of details here, and I love to talk about this and the sleep offline.

[00:41:43]
But one of the things that he did was to say, well, OK, this is, that was in cortex, where is this coming from? So we recorded from 1000. This also the neurons that project directly to the cortex. And what we found is that the neurons in the thousands did not do the same thing.

[00:42:01]
Ok, that's interesting, right there, providing the inputs to this part of the brain saying, well OK, maybe it's just going along the pathway here. You found this and when he recorded these neurons, and also importantly when he recorded from neurons that were connected to each other. So we could actually record simultaneously from the ones in the balance and neurons in the cortex that were connected to each other, at least with high degree of confidence that they're connected to each other.

[00:42:24]
And they just didn't exhibit the same features in the fall. As with the cortex. So there's a lot, again, a lot of details to this, but what he did find is that the foul on US, in response to the gap in stimulus decreased the level of synchronization. So this is just a measure of synchrony.

[00:42:43]
So in response to the napping stimulus, the neurons would adapt. They did adapt in terms of their overall mean fine rate, but they also disenfranchise. And that turned out to be the really key thing. And it's key because as I told you on some of the 1st few slides, the cortex is really, really sensitive to the timing across the neurons that, that project to that these layer forecourt going on.

[00:43:11]
And so the timing, the relative timings of these matters. So when they become, when they're really, really precise, they're very potent drug and cortex. And when they become the synchronized, they're less potent. And so there's kind of a complicated technical argument here. But the idea is that when the system is not adapted, everything is so simple, none of that everything goes through.

[00:43:34]
And you can't actually tell the difference between any of these different inputs. However, when the neurons become disinclined, when they're dabbed, you only the really strong inputs are produced, the synchronous response and the weaker ones are be synchronized. And so they're actually filtered out. So what we found is that this, that this process, we showed this through modeling that, that was at least sufficient in a model that captures the solemnity activity to the cortical activity, to produce this kind of effect at the level of cortex.

[00:44:07]
So the disinclination of these neurons actually produced an increased discriminate ability in the cortex. This kind of cool, so please have a look at that paper if you're interested. We say we also did analogous experiments in behavior, and this is working double the Renshaw and showed that indeed in a head fixed rat trained to do I detect in task the adapting stimulus.

[00:44:34]
When the animals asked to say, did something or not did something touch my whiskers or not the animal is there, performance is degraded or, or decreased so they can't do it as well with an adapting stimulus. But when they're asked to discriminate between 2 different whisker inputs, they actually get better.

[00:44:53]
I'm not sure when the primary data here, but again, all the details are in this paper. So they get better at discriminating at the expense of detection. So it's really interesting question about what's being in code that we think this is actually a general kind of principle here, maybe same exact kind of inputs, different context.

[00:45:12]
Then finally in this, in this, at this time scale, I just wanted to say that, right? Again, post-doc in the lab has been doing the experiments in the awake head, fixed mouse, and showing 1st of all that. Again, the Court of the adaptation is pretty robust. This shows the control case versus when, when these cells have been adapted significantly decreased in their response for both the excited Torrie cells, inhibitory cells in the cortex.

[00:45:44]
And take a look at this by archive article of this year kind of thing. But almost more importantly, in that, the, the lambic activity that there is 1st of all bursting activity in the thalamic, in the, in the balance in the way mouse, but that it's attenuated or decreased in this adaptation.

[00:46:04]
And there's a distinction is Ation of these, the lining neurons with the adaptation. And so, you know, again, it's kind of a long story in the, underneath all of this. But it's consistent what we observed in Lee's anesthetized animal experiments. And I don't have time to talk about this, but there's a whole range of controls that he does.

[00:46:23]
Did that all of these modern tools are allowing US now to do to really zoom in on the relative roles of these different elements of the circuit combined with the modeling network modeling that led US to conclude that this adaptation is, is that we have certain the cortex is primarily due to adaptations of synchronous Atlantic firing.

[00:46:47]
And the way that this synchronous activity, differentially engages the cortical network. So it's really, again all about the time. So I'm going to summarize this little bit that, that we believe that this is again, a fundamental principle that there's this adaptive getting through this timing. If you're not really looking at this carefully, if using Big bins and looking spiking activity over very long Windows, you're not going to see it.

[00:47:14]
And we believe the shapes not only how much information is being transmitted, but the what. And I think the, what part of the most more interesting that different aspects are being actually coded, depending on the context. And this could be this fundamental principle. And again, you know, this, this theory is really guiding US here to think about that.

[00:47:35]
It's not just something we observe, but there may be some reason for these things existing. And so I framed this up with, you know, the efficient coding hypothesis, or the idea of like that, that how much of what kind of information is being, is being regulated here, but with these kind of prophecies.

[00:47:53]
And that this adaptive Slam again, maybe sort of the fundamental principle. And again, I just want to comment on the what people believe versus what's been proven. I guess we had several years of information problems in life. But that it was also the case that was echoed by lots of people that know well, this kind of adaptation stuff just doesn't exist in the wake brain that all of these and that's the tie studies didn't really represent what's happening.

[00:48:23]
But yet the experiments really hadn't been done for this. And so it's kind of interesting again that I found a lot of by interest was kind of an Echo of these things when you couldn't actually identify papers that showed that, that we just went in and did them and show that it's actually indeed the case just want to link to a current lab number you is a PH d. student lab and she's been working on ways of techniques to record simultaneously across brain structures with large scale recording Pro Chiz and a den of fine connected neurons when, when, when possible.

[00:48:59]
And so this is something she's been working on, and this is an onion because I like to think about this is kind of peeling the onion. We think that there are multiple adaptive prophecies that may be happening. Not just the sort of rapid things, but maybe at multiple times scales and that, that kind of get obscured and we're working on because peeling the onion of this kind of adaptation right now.

[00:49:20]
Ok, so finally I just want to quickly mention this last little bit very quickly and go on to this slower time scale of learning. And see OK. So, you know, one of the things that we're, we're interested in and think anyone who studies the brain is interested in certain sensory processing is that, you know, obviously behavior is the most important thing.

[00:49:48]
And that if we're sensory systems, we think about perception. And it's really kind of an invasive thing, where is it? Where is perception? And it's generally assumed that it emerges somewhere at the cortex. But pinning it down is a really tough. And there's just a ton of old literature with all sorts of techniques, like lesion ING, reactivation protocols, that support all sorts of different stuff.

[00:50:12]
And it's confusing if you grab, if you get your head around all this literature, you will find it really, really confusing. And sometimes contradictory, and there's recently richer that is really question whether cortex is necessary for simple sensory tasks. And this was seemingly contradict lots of all older literatures.

[00:50:32]
And I think what's, what I would say people in our lab would think is that all of these different brain regions is really dynamic. And it may be that what, the reason this literature is really confusing is that, that maybe the role of different brain regions, not really some static fixed thing, but it may be changing depending on complexity of tasks or changing over time.

[00:50:55]
And I just want to quickly go over the last couple slides. This is worth the price. Why we were was a post-doc in the lab. Welcome back, Chris, from Germany being banished for several months. So. So what he did is trained to fix mice, but also did voltage imaging in the superficial layers of cortex while the mice were doing this kind of tasks.

[00:51:21]
This just shows expression of this genetically engineered voltage indicator called Jeff, which is particularly arclight in this case. When the animals were trained to do a simple go, no go detection task, though the Whisper stimulus is delivered. And if they Lick the water pipes spout they correctly, then they get a reward.

[00:51:42]
If they possibly Lick the, can have a false alarm. If they Miss the stimulus, it's a Miss. Ok. And what Chris, we published on this something on the basic paradigm of behavior and 2019. But I'm going to talk about some more recent stuff here. And I'm going to emphasize looking at different time scales, the early during training, amateur level versus experience.

[00:52:04]
And we try to do this quickly. Sorry, sorry, Chris, on this Justice. But what he looked at is a process Sions of training multiple weeks here, that the animal just gets better and better and better at this task. This purple line here is the hit rate. So how well can they do it?

[00:52:20]
You know, how well do they get hit versus the false alarm rate starts? It really kind of stays pretty Pons. The false alarm, it does not work. Now interestingly, when he looked at the cortex, it's pretty constant across this is just the voltage imaging in response to an individual whisker reflection.

[00:52:38]
And so Despite the fact that their own, like to be shows this kind of growth in performance, it's the Prime measurement. The neural activity, pretty flat. Ok, so not much happening early. Ok, now you manimal is being trained up. Nothing happening in the last one. Just seems like a static thing.

[00:52:58]
However, he showed in the previous paper and more recently in the head, fix mouth that when the elbow was challenged with either with a stimulus, in this case, drawn from a random distribution, a statistical distribution. So in this case, velocity of different inputs from a high range versus a lower range than one case, the animals train trained to detect inputs from this.

[00:53:21]
This is the psychometric function, the probability of response of the hit hit rate as a function of the strength of the input force stronger input. They're better at it. But when they are trained, when they switch to the low range condition, the animal actually improves their performance. And so notice that some of the stimulator exactly the same.

[00:53:42]
So there's overlap, so this 8 degree or 4 degree, they're both part of this stimulus that but they're in a different context. And so what we found, it again, was not timed to show this is that the animal seems to be changing their behavior to maintain reward or reward expectation.

[00:54:01]
So he showed this beautifully in the 21000 paper. And more recently, in this paper work, we're finishing up that they change their behavior in a way as to maintain the reward in the face of this changing stimulus to, to sticks and it's reversible. So he changes back, they change back.

[00:54:20]
Ok. So when he looks at the experienced animal, what he sees is that the voltage imaging shows that same exact input, this 8 degree stimulus, is stronger in this low range condition when the, when the task is harder. Ok. And that's pretty interesting, right? To same exact stimulus on a particular trial, there's a stronger response in the cortex to this particular input and shown summarized here for the time series of this voltage signal.

[00:54:56]
So what we did then was we looked at the relationship with, Here's a psychometric function. And Here's what she just demonstrated at the level of cortex, the amplitude of the input versus the voltage imaging at the level of cortex as a function of the amplitude. And this explains some of the behavior but not all of it.

[00:55:18]
And so when you actually attach a 2nd function here that represents what's happening downstream of primaries, Madison Street cortex. Together these 2 functions are constrained to represent the psychometric function, which is overall the strength of the input to the behavior of the animal. Then we do this, what it gives is the ability to fit different parts of things to different datasets.

[00:55:43]
And what it showed is that the hiring condition is that with this kind of function versus the lower range. So there's some kind of shift in the US. However, it doesn't completely explain the behavior and so the downstream neurons, whatever is receiving inputs from primary Samantha's and the cortex is actually doing the rest of the work.

[00:56:04]
Ok, again, I can tell, I'm not doing this just isn't making it too fast, but I want to give the punch line here, which is that it turns out that early in training in the amateur phase, that a large majority of what's happening and what explains the behavior is captured by the downstream neurons, whereas it only a little bit of it is happening in as one.

[00:56:29]
But when he looks later at the more experienced animal, a larger percentage of the explanatory power lies in s. more. So this is something that emerges later in the animal training. So if you were not aware of this, you're looking at this thing a different time frames. You might actually conclude different things.

[00:56:51]
And so that ties into my overall sort of thought that that time scale matters a lot and it suggests the higher level, high level decision decision making could be pushed upstream or earlier in the sensory motor arc. And we have some that's a type of there. We have some additional evidence that we see cognitive decision making signals as early as the thoughtless the doesn't really get a causality, but there is a and there's a ton of work to do.

[00:57:16]
But it's Super exciting for US. So again, when you look is critical. So last slide, you know, again my goal is to try to sort of throw in something for that community and let you know the kinds of things we're working on. And tie the old stuff to the New stuff and make sure you know who our lab is and in case you see them wandering around the halls and hopefully convinced you that timing matters a lot at all scales.

[00:57:41]
The key that is that everything is that we can adaptive and that the theories are really important, sort of guiding US in this kind of technology or technique phase of neuroscience. So again, like to thank my lab and I'll stop at that point and you thanks there. One minute to spare.

[00:58:05]
Yeah, sorry I was a brand New sort of talk, so I jammed, I know the g.m. stuff and you know, maybe we may be limited to one questions, hope you won't have to kind of hang around. So there was a question in the chat from Tim Cole. Tim, are you still here?

[00:58:21]
You want to ask your question? Sure, Yeah, here is a great talker. Thank you. Yes. So I'm wondering why you're excluding integration of the synoptic level in the 20 millisecond time scale. You know, I, P S P's can be tens of milliseconds long, and you wouldn't necessarily need a network explanation or that time scale.

[00:58:44]
It seems to me, Yeah, OK. So maybe that's me just with knee jerk reaction because we, you know, I really, I think it's, I think it's probably a combination of the synaptic properties as well as you know, the excited Tory inhibitory interactions in the future forward a bit and have to inhibition at the level of cortex, it's just that others in the field have kind of had this response that, well wait, how can it possibly be on the order of 10 to 15 milliseconds when synapses are only integrating over a couple of milliseconds.

[00:59:16]
But I agree with you in that I don't think even the synapses are that simple, right? I think that these prolonged inhibitory inputs can, can actually have long lasting the facts on it. And so I don't think the 10 to 50 milliseconds is maybe unique to the network at I agree a maybe Yes, maybe overstated that was a kneejerk reaction to the community.

[00:59:37]
It's OK and thanks for recognising that because people are now in auditory cortex and served in spinal cord or recognising that inhibitory synapses are maybe misnamed. They do have the op, they do have the effect of suppressing neural activity. If they are, if there's temporal dispersion, when they're synchronized, they Act like they Act more as a clock than they do as a suppressor of activity.

[01:00:04]
So they went 100 OK. Well, I'd love, I'd love to talk with you more offline about that sounds pretty exciting. Thanks. And again, just I know Bill is going to cut me off, but I'm happy to, you know, my main job was trying to tell you about our lab.

[01:00:21]
I'm happy to talk with anyone who's interested in these pieces of it offline. Ok, maybe that's a good point to transition. People need to go. But Yeah, definitely reach out to take care of any of the locals because as I mentioned last time, good. Follow up on the interactions that we can happen in our virtual World these days.

[01:00:48]
So thanks Garrett for the tour of your kind of overview of how you've already got those range yet. And thanks to you and sidemen for, you know, keeping this seminar series going to be such a great job during all this craziness it's something all right. Thank you.