 All right, okay, so it's my pleasure to introduce Dr. Pascal Fries this morning who's joining us from his lab in Frankfurt. He's currently the director of the Ernst Strungman Institute for Neuroscience in Frankfurt. And his work is primarily, or the work of his laboratory is primarily focused on understanding the mechanisms that underlie rhythmic activity in the cerebral cortex, the functional consequences of this oscillatory temporally structured activity in local and long range networks. And the functional consequences of the downstream or cognitive cycle of the structured activity. So Dr. Fries completed both his medical degree and his PhD at Johann Wolfgang Götze University in Frankfurt. He did his PhD work in the laboratory of Wolfsinger at the Max Planck Institute for Brain Research also in Frankfurt. He did a short residency in the Department of Psychiatry at Johann Wolfgang Götze and a short postdoctoral fellowship with Wolf before moving to the US to work with Bob Desimone in his old lab at the laboratory of neuropsychology at the National Institute of Health. After his work in the US, he started his own lab in the Netherlands at the FC Donder Center at Radbaud University where he was until 2009. I believe when he took up his current position as director of VSI. So as I mentioned before throughout his career, Pascal's research has been the rhythmic synchronized activity observed throughout the cerebral cortex from some nice early work on the relationship between synchronized oscillations and interocular ovary in cats through investigations more recently. I think primarily in nonhuman primates, macaques and marmosets more recently and a little bit of work in humans looking at the role of attentional feedback on coherent oscillations and synchronizing oscillations between visual areas and then more recent work in which he's combined some functional anatomy along with high density recordings of general activity across many cortical structures looking at distinct oscillatory signatures of feedforward and feedback pathways. And I think that's a little bit of what he's going to talk to us about today. So without further ado, I'm going to hand the stage over to you, Pascal. Thank you very much, Andrew, for this very kind introduction. It's nice to be reminded of all the things that I have done over time. Thank you very much for the invitation. It's a pleasure to present our work to you. I'm sorry that I can't be in Salt Lake City right now, but it's at least nice to present the routine in this format and please interrupt me at any time if you have a question. The Zoom format shouldn't hold back a lively discussion. So as Andrew was saying, we are primarily interested in brain rhythms and particularly in the potential functional consequences of brain rhythms. So we asked the question, how brain rhythms might subserve cognition? And in the first few slides, I'm going to present to you the main brain rhythm that is playing the key role, but not the only role, that's the gamma rhythm. Later on, we're going to also talk about the beta rhythm. So here first comes the gamma rhythm. So this trace here is a wrong local field potential recording from the primary visual cortex of an awake macabre monkey. So we've simply placed a little one millimeter platinum disc in an electrocotic ocrophy recording on the primary visual cortex of an awake macabre monkey. And he first, for a few hundred milliseconds, looks at a great screen. And then we present a relatively natural stimulus, and at the same time, the monkey is allowed to then freely explore that relatively natural scene. And you can see that within a few tens of milliseconds, a very pronounced rhythm emerges, despite the fact that this local field potential is not filtered, right? So it's kind of very, very broadly filtered between something like 1 Hertz and maybe 200 Hertz. So the oscillation is really generated by the brain. And in order to quantify that, we calculate power spectra during the free exploration of the natural stimulus and express that in relation to power spectra that have been acquired through the crystal is baseline. And then you see that the power change when moving from the baseline to the free exploration is primarily in that gamma frequency range here shown as a function of the frequency, right? So it's in this classical gamma range. And in this monkey, it amounts for the stimulus to almost 7000%. And we find that was essentially to varying decrease of strength, but we find that clearly rhythmic activity for all natural scenes and objects that we presented to two Macaulay monkeys. And so we are primarily wondering, what, I mean, maybe this is good for nothing, but maybe it is, maybe it has function consequences. So our working hypothesis is that such rhythms will have consequences for the processing of neurons, particularly for the communication between neurons. And we try to basically come up with simple ideas of what those consequences could be. And then we try to test them. And here I show you the first experiment. So the most basic idea we have here is that these oscillations are important for the basically the information transmission. And in order to study that, we turn in this experiment actually to cat experiments. And those, this is the only data I show you that is obtained around this TZR. So this is lightly isotheraen anesthetized cat visual cortex area 17 indicated here is the primary visual cortex of the cat. And we first inject a virus, an adenos-associated virus, pseudotype 9, under a cum-tinase 2-alpha-chromotor, which restricts the expression of the product to excitatory neurons. And we express channel rodent C2, together with enhanced yellow fluorescent protein, as just as a marker. So we wait for a number of weeks, about eight weeks. And then we come back with an electrode together with a laser. And the laser can be entwined either in this blue frequency range or in the yellow control frequency range. And when we give, for example, two-second laser illumination, and we look at the multi-unit activity, so the spike rates of the recorded units. This is the average over dozens of recording sites. You see that with blue light, there is a beautifully sustained enhanced firing rate activity. And with yellow light, this is not present, they are only short photoelectric artifacts at the one-second offset of the laser. And if we just give that constant octogenetic drive to the excitatory neurons of the network, without any temporal structure in the drive, and we look at the local synchronization that is measured by the multi-unit spiking versus LFP phase locking metric, PPC, as a function of the frequency in the low frequency range and in the high frequency range, you see that the blue light stimulation compared to the yellow control induces a very pronounced gamma bond with it, very similar to the natural stimuli. And if we zoom in on the lower frequencies, note that this scale is much strongly zoomed in, so it's like four-fold the scale that we have here, then we see that there's also a certain reduction in the alpha frequency range. So this shows you that this gamma bond is actually a very, very simple phenomenon. It exists for all types of natural stimuli under natural viewing conditions for viewing conditions, but if you just drive a local network in primary visual cortex, even under anesthesia just through this optogenetic direct drive to the excitatory neurons, we get a very similar phenomenon. There's nothing, there's no big magic about this gamma region itself. But it might still have extremely interesting consequences. And one very interesting insight was gained in this experiment, I think, that was spearheaded by Christopher Lewis and published last year. So Chris had the idea to use the setup that I just introduced to you to present optogenetic white noise. White noise is interesting because it contains all frequencies with equal strength. And white noise is interesting because it is unpredictable in both directions. So the white noise at one moment is not predictive of the past and not predictive of the future. So if we now record a normal activity and we see that, for example, the spiking is dependent on the past of the white noise input that we assimilated in optogenetic drive, then this is not due to the correlation inside the white noise, it's something that's generated by the network of neurons. So Chris used this preparation, he recorded local feed potentials and he recorded spikes while he drove the local excitatory network with white noise. And this white noise basically emulates excitatory synaptic input to a local population of excitatory neurons. And then he took the individual spikes that you see here and he used the spikes to align the averaging of the laser power density. And he found that the spike is caused by a gamma-rhythmic component in the input. So remember that the input was white noise. There is no frequency preferred. And if anything, the optogenetic drive goes through the synapse and to the dendrites so there is a certain low-pass filtering happening that has been shown in Slyseware and Kautcherberg. But nevertheless, we find that spikes, when we drive the network with white noise or frequencies equal, the spikes are caused by the gamma component, the preceding gamma component. And we can directly quantify that by calculating the granger causality or the granger cause of influence from the light to the spike. And we see that it's the gamma component that granger causes the spikes. And just as a control, of course, we calculate the granger causality from spike to light, which must not be there, but just as a sanity check in detail is close to zero. This system can directly be modeled extremely well in standard biophysically realistic models, as have been spearheaded particularly by Nancy Copell, where biophysically realistic excitatory neurons are coupled, I mean, this stands for a local population of excitatory neurons, they are coupled excitatory to inhibitory neurons which couple back, and the inhibitory neurons are self-inhibiting. So if such a network is driven, for example, in this constant excitation, it generates a gamma group. Now Chris Dr. Lewis, he now used exactly the model that had been published in a book from a close collaborator of Nancy, Christopher Burgers, he took exactly the model without any modification, without any parameter tuning, and drove the excitatory neurons with the white noise, just like we did in the optogenetic experiments. And he used the spikes in the model to then trigger the averaging of the simulated white noise. And he found a remarkably similar response as what we had found experimented, except for a delay here that we expect because in the experiment there is an optogenetic delay in the optogenetics itself and in the dendritic delay. So here we have that and in the model now, we can now directly look into the intracellular membrane potentials of the excitatory neurons and we find that the input is most affected if it is coherent with the ongoing excitatory gamma range fluctuations in the network. So in other words, the cortical gamma band resonance selectively and preferentially transmits the gamma component in the input specifically the gamma component in the input that is phase coherent with the gamma in the driven network. And interestingly, this has been known before from the modeling work of propellant burgers, the excitation is always preceding the inhibition by a few milliseconds. And you can see that by being coherent with the excitation, it is close to exactly hitting into the minima of the inhibition and avoiding the maximum of the inhibition so it's maximally effective basically in driving the cycle. So, with this in mind, we have proposed a hypothesis which we call the communication through coherence hypothesis. Some of you may have heard of it but for those who have not heard about it, I will introduce it very very quickly in this slide. The idea is as follows. If a stimulus activates a group of neurons, for example in a lower region area, then this gamma region emerges, which is characterized by a sequence of excitation, which then triggers inhibition and when the inhibition decays, the excitation rises again because the stimulus is still present, which triggers inhibition and so on and the cycle repeats, and mainly the decay times of the inhibition set the cycle times of the gamma. Importantly, this then, as I'm going to show you in a minute, entrains a very similar gamma rhythm in the target structure. So the lower visual area neurons they will project to higher visual areas, and the gamma rhythm in the lower visual area will then train and synchronize the gamma rhythm in the higher visual area. So here comes the key idea that when this entrainment has happened, the inputs from the lower area come right at the times when the inhibition is decaying and when the post-synaptic neurons in the higher visual area are most susceptible to the input. So through the entrainment, basically a protocol of effective communication is set up, where one bound of input is effective, is then locally triggering inhibition, and the inhibition is just then going away when the next part of input is coming. And at the same time, potentially competing inputs that do not entrain the post-synaptic neurons, they will have a much harder time getting their foot into the doorstep. They will, as is illustrated here in a maybe slightly extreme case, consistently hit against maximal inhibition. I'm going to show you that this is the case that is more easy to illustrate here. What we actually find is that the non-attended or these irrelevant inputs, they simply have no systematic phase relation, have no coherence, no entrainment, and they probably come then at random times. So they are just very ineffective, but that's very hard to illustrate in three cycles. So I just kind of very quickly show you some slightly older birth from 2012. So okay, just very quickly, we call this the communication through coherence hypothesis because the effective communication here is through the coherence or entrainment. I just got very quickly now show you this paper from 2012, where we directly experimentally tested it, because also then the next slides will build on that. So here, we have used a microelectrochotic ocular field rate with about 250 recording sites over one hemisphere of two white macaque mungies. And I show you specifically two recording sites over the primary visual cortex, and one recording sites over a higher visual cortex before. So this is the first condition for which we always show the data I've read. One stimulus is shown in one part of the visual field, and it induces a nice gamma one activity inside one, and not in the, in the site to because it's written the topic the organized primary cortex. The gamma rhythm inside one is then Granger causing basically in training synchronizing the gamma rhythm. If we present a stimulus at a different location equal eccentricity equal contrast everything equal, but, but the precise position. Then this stimulus the blue condition entrains side to not side one, and then this government entrains the gamma and before. So we can come to the interesting condition where we present the two stimuli simultaneously, and we direct the monkeys attention to the stimulus that is shown here is a simple halo. This is not, of course, visible to the monkey is just indicates to you where the attention of the monkeys key by standard key. And then you see that the two stimuli into two gamma reasons inside one inside two, but only the government inside one manages to entrain and synchronize the government is in the form exactly as predicted by the communications hypothesis. And this radically switches, when we switch the attention so everything stays the same same monkey same recording same same stimuli only the attention switch over. And then the pattern of entrainment here is radically switch exactly as predicted by the by this hypothesis that the behavior you relevant information should be rooted through by entrainment and irrelevant information should be blocked out by an inability to entrain. So this is exactly what we had predicted and what we also found and it was actually in the same year, few months later, it was replicated. And more or less the same result was published in Journal of neuroscience by the lab of Andreas criteria. And the interesting. So this was hypothesized, the interesting, not hypothesized finding was here in the one, namely that the precise frequency of the government in the one is systematically higher. And the respective stimulus is attended. You see here for side one, that's the red condition and for side two, that's the blue condition. And when we talk to theoreticians in the area of dynamic system they tell us well, a slightly higher frequency gives you an advantage of driving a receiving oscillator. So maybe this higher frequency here is actually the reason for the selectable trend. So now, a few years later, Gustavo whole had, I think, a brilliant idea of analyzing these data in a different way. So he was his argument was basically that, if this is true, that this is also important for the cognitive analysis of attention. Then we should be able to directly relate the gamma synchronization in a single trial to the behavioral benefit in that trial. So, and to understand this how we did the analysis I quickly present how the trial structure was here. And in the data that he analyzed, what he was fixating here. This was the joint receptive field of a simultaneously recorded the one and the foresight. Okay, so we look at pairs of the one and the four primary products and I have recorded that have a joint receptive field here, and that are co activated by this stimulus and we have a controlled stimulus and the other field. They are randomly color, you are yellow, the switches of a trance randomly and the fixation point randomly assumed the color of one of them, which always points the attention to the, to the stimulus of the same color. And then either the attended like here or in other tribes the intended would change its form and then if you see the slide defamation here. If the distractor changes the shape monkey has to just wait long until the target changes the shape, but if the target changes the shape first as happened here. Then the monkey is supposed to release the bar to get a reward. And they do this in 90 whatever percent of the trials and Gustavo Ron Paul. Okay, let's look at the 200 milliseconds when everything's still well controlled here same stimuli, just before the behaviorally relevant go she was given and see whether we can predict the speed of the bar response, and he could so this is first for the two monkeys. And this is the synchronization between one and before to show you that it's very narrow back in both of them, but at slightly different frequencies at 75 herds and much heard, and at 65 herds and much. So, Gustavo aligned it to the individual gamma frequency here, plus minus 20 yards. And then did first a very, very simple analysis he just split the trials according to whether they resulted in slow reaction times, in which he got a moderate gamma by extension, or fast reaction times, in which he got a much stronger and importantly, this effect was entirely absent. When the stimulus in the joint we want to set the field was not attended. In other words, when it was activity there was going to have an activity, but it was behaviorally irrelevant. So, the control is extremely important because it shows that this reaction time predictive effect is specific. It's not driven by unspecific changes in the overall arousal of the market. But Gustavo wanted to go a step further and he said okay, that's good and nice, you know I can split my 500 trials into slow and fast but this is still like whatever 250 trials for each condition. Okay, but I want to test whether I can predict the reaction time on a trial by trial basis. So, he looked at this 200 millisecond window and looked at the, he made these plots where on the polar axis, there is the V1 V4 gamma phase relation in the given trial and on the radial axis, there is the reaction time on that and then you can see that there is actually a systematic relationship between the V1 V4 gamma phase relation and the reaction time. And this can be quantified now really on the trial by trial level as a circular linear correlation coefficient. Okay, which is significant here for the attend in condition and again importantly the control condition is absent. So this was nice but then really came the most important contribution from Gustavo because he and it took me like probably two days to realize his point but from that moment on I was completely enthusiastic about it. His point was, well that's the really important thing is that now we can test one crucial prediction namely that V1 entrains before at the phase that is optimal for the transmission of the information, right, because look at this. Here, we see that at this phase relation reaction times are fast and that this phase relation reaction times are slow. But we think that the one entrains before on average at the phase relationship that is good for communicate information. So that means it's in training before at the phase relation that's good for commutes for signaling changes and the stimulus from the one to be for which probably leads to short reaction. So we see that the V1 before gamma phase relation typically occurs at this phase where the reaction times are shortest. And then he looked at the distribution of all phase relations and they're extremely noisy it's also difficult to estimate the gamma bond activity in the end, not a very strong signal. We have a very short 200 millisecond window to estimate it, but you see the histogram here. So each dot is the gamma phase relation on a single trial the brown histogram is the distribution of these relations and yellow vector is the mean phase relation, and exactly at the phase which we want entrains before on average it for this particular side pair, we find the fastest reaction time so indicating the fastest communication from we want to be for. So this is one, the one before side pair. So what Gustavo did he he then rotated this mean phase relation to zero. So it's the same data just rotated, but this allowed him to then average. All those data for all the one before face relations it's it's pretty. I think it's about 100, we want me for face related, we want me for side pairs from two macaques, and you see that at the phase relation at which we want to transfer is the most effective communication and any moment to moment deviation from that mean phase relation leads to a worsening of the signal from the one for us indexed by reaction. And so we took the cosine of this deviation we call it goodness of phase relation, and then we could calculate the correlation between this goodness of phase relation and the reaction time. And indeed, this is negatively correlated in the gamma range because negatively because reaction times get shorter. So if the phase relation is better so closer to the mean phase relation reaction times get shorter. And again, not in the control condition. And the effect sizes actually remarkably so when we look at the reaction time during good phase relation trials minus the reaction time during that phase relation we get a 25 millisecond effect, which is very much in the range of attentional effects on reaction times certainly over trained macaques. And again, no effect in the control condition. So, I think that all of this taking together gives some experimental support to this idea that the gamma rhythm in the lower area might entrain in another area and through entrain that it will make the communication effective because it also entrain at the phase relation that is optimal for the information transmission. And at the same time, it, which is probably equally important, it, it excludes competing stimuli in a natural visual world, it is full of competing stimuli. And we are vitally dependent on having one object rooted forward and interpreted and acted on rather than being confused by all the objects simultaneously present. Right. So, now, recently, a very smart PhD student in my lab Benjamin Stauch, he said, Okay, I mean, if this is really true, I would want to do a causal experiment, right. So, we found before that the gamma, so basically here what we show is that the rhythm that comes first is the one that is winning, and the one that comes a bit later is the one that's essentially losing out. So, he said, Okay, if I now simply take a grating stimulus, and I give them a little phase, it might, there's actually a stimulus, so he modulates the intensity of creating stimuli on the monitor in an oscillatory fashion, and thereby, he basically dictates he entrains the gamma-rhythm in the visual cortex. And now he can experimentally determine which one of those two stimuli has a phase lead. And he predicts of course that the one that is experimentally given a phase lead is the one that will be more effectively communicated to higher areas. And to do that, he built a dedicated custom setup with two dedicated LED projectors, where the LEDs were driven by sound cards. And each one had a slide in it with a grating. One grating was tilted leftwards, and the other grating was tilted rightwards, and they were superimposed on the translucent screen behind the aperture, and he said basically the two gratings were seen at the phobia, at the fixation point. There was a laser-based fixation point. They were superimposed. Both were oscillating at frequencies between 5 and 50 Earths roughly, and the phase relation between them was experimentally controlled. In a given trial, it's an example trial, for example, the two might, so the two have the same, they are black and white, they are just shown in color here to make the point, right? So they oscillate, and for example here, the green one has a phase lead over the colored one. And then subjects were asked to push a button, so subjects all reported that one of them was brighter, was more intense, particularly was flickering more intensely. And 20 out of 20 human participants, when they were just prompted to tell us what they would see, every single one of the 20 subjects said one of them is somehow brighter, more intense, oscillating more intensely. And then said okay, tell us whether it's the one that's left tilted or right tilted, it turned out it was always the one that was phase leading, at least 90% of the trials. We wanted to know okay, how strong is this effect? And we measure this through the method of adjustment. So we basically gave them buttons that they could push to increase the actual physical intensity of one and at the same time decrease the actual physical intensity of the other one. And we asked them to do that until the two stimuli appeared equally bright. And they very consistently did that in the end, indicating that they had to increase the physical brightness of the lagging one until they perceive the two equally bright. And I'm going to show you these results. So here, these are the results from one example subject. We had a direct success to see the different frequencies that we've tested. And about here is the critical flick of human frequency when this effect is stopping. And here, the left stimulus was the one that had a phase lead. And they adjusted, they upregulated the bright and physical brightness of the right tilted stimulus. I mean left tilted stimulus was had a phase lead and they upregulated the physical brightness of the right tilted stimulus. And to these levels, until they said they are equally bright in our perception. So they needed to upregulate it by 0.3 of the total luminance that was on the screen or 60% of the average luminance of the two stimuli in order to overcome this phase lead induced illusion. And they were physically, or written in identical intensity. Only there was a phase lead of one of them that gave such a strong illusion immediately in all the subjects that the lagging stimulus had to be upregulated by 60% of like 30% of the sum of the two stimuli, 60% of the average of the two stimuli to overcome that phase lead induced illusion. And that was the same thing that we made the right tilted stimulus leading than the other stimulus again had to be upregulated. So if we combine the two conditions and all the 20 subjects, you get this picture was very, very consistent. And then if we don't ask the question, okay, what is the effect size like how much did they have to upregulate the phase lagging stimulus, but in which percentage of the trials was the phase lagging stimulus and giving higher brightness to overcome the illusion. This wasn't 90% of the times. So almost all the trials that the participants saw this illusion, which I guess is another support now from a very different type of study for the communications for coherence. All right. So let me see, Andrew, how long approximately should I talk, how many minutes do I still have for the presentation. I think it's planned that the entire meetings plan for an hour but I guess it should be about 10 minutes or so for discussion. Okay. Yeah, that's correct. That's correct. Yeah, give us about 10 minutes for discussion so finish about 10 to or so. Okay, so again, I going to be a little bit quick on the next few slides. I guess this is also published to come to a small other presentation with traveling waves recent interest of us. So here, I showed you a lot of the we want to be for interaction. So here this is the one to be for Granger causality. As I showed before, dominated by gamma. But when we look at the parietal seven a area 7821 Granger causality it's not dominated by gamma it's dominated by beta. And both are enhanced by the tension to the control of anything. He asked the question or Greg Richter asked the question, whether we can find the direct influence of this top down beta on the bottom of gamma. And I think he did this very diligently so he selected trials where the we want to be for Granger causality in the gamma range happened to be weak. And then he looked okay. I'm sorry. Sorry, I confused it. So he, he first looked at trials where the top down parietal cortex seven a to be one beta Granger causality happens to be weak. And he found that okay, the bottom up the one to be for Granger causality with moderate strength when he selected the trials where the top down beta influence happened to be strong so this was not due to attention or anything was just due to spontaneous fluctuations. And this, but the trials were selected exclusively on this condition. So when top down beta was strong, bottom up gamma was much stronger. And it got even more impressive when you look the other way around. So now you looked at the strength of the top down beta, but he first selected the trials, when the bottom up gamma happened to be weak. And interestingly, then the top down beta was just absent. When he selected the trials where the bottom up gamma happened to be strong top down beta was very strong. I find it highly interesting so I think bottom up gamma when there is a strong visual stimulus can probably not be absent in a wake animal, but top down beta can be absent as we can see here, and we see how strongly the top down beta is modulating across trials, and we see it is revealed by sorting the trials based on the bottom up gamma. So, then very quickly I browse over the next few slides, maybe most of many of you know this. So, we were impressed by this asymmetry that we generally found from a lower area we want to a higher area but something doesn't create a sort of like a karate mission area. We found that gamma was particularly strong in the bottom up direction, whereas in the top down direction. Gamma was weaker, whereas beta was stronger in the top down direction, then in the bottom up direction in theta again was a distinct peak was stronger in the bottom up. So, we captured these asymmetries for each frequency in the so called directional influence asymmetry. It's basically branched out into one direction by the other direction divided by the sum. So we did this frequency by three. And then we teamed up with Henry Kennedy. One of the world experts in a lot of internal projections and the feed forward or bottom up versus top down character of these projections. And I don't have the time to go into details here but this anatomical characterization of a projection being of a bottom up type or feedback or top down type. It also has to do with the layers of origin in the cortex where that comes from superman in the neurons or in front of us when we cross correlate this anatomical gold standard of whether a product projection bottom up or top down. We cross correlate it with our direction influences symmetry index across many pairs of areas. We find that in the gamma range. The direction influences are stronger in the feed forward direction. In the data range, the Granger causality interactions are stronger in the feedback direction. And in the data range, the Granger causality interactions are systematically stronger in the bottom up direction. Okay, so just again is a bit complicated but to explain it to you once more. So we have recordings from several visual areas to be precise of eight visual areas we want it to be for to do as a prelunate seven a to frontal I field areas. And this gives many pairs of areas. And, and for each of those pairs of areas in two monkeys we could measure the Granger causality interactions, and we have different data set the information whether the projections, the anatomical projections are bottom up type, or top down time. We found that systematically gamma is stronger in the anatomically defined bottom up direction, beta is stronger in the anatomically defined top down direction and beta is stronger in the anatomically defined bottom up direction. And then actually triggered by a reviewer. We. So based on this, we could build a hierarchy of visual areas. So the hierarchy, the hierarchical level of these areas have been determined before based on these anatomical projections. And then we based it on the Granger causality interactions. And you see that the hierarchy can never be determined by Granger causality is very, very similar to the Granger causality based on anatomy. So very, very quickly, we basically repeated this result from monkeys in the human. So we, we did magnetic encephalography in the relatively large cohort of a few dozen human participants source projected the energy data calculated Granger causality, for example, we are between four and seven a, and we found a very similar pattern that Granger causality in the bottom up direction. In the government is much stronger in the bottom up direction, whereas in the beta band here is stronger in the top down direction, something that we didn't have in the monkey data which we saw in the energy test. That is, there's not only this beta peak here, which is stronger in the top down direction, but there's also a distinct alpha peak that's also stronger in the top down direction. And in the energy data we did not find the data, but we believe that there's nothing to do with the difference between humans and monkeys but the fact that here in the energy, the stimulus was in the formula. And due to other results we have to believe that you see that feature effect only when there's active sampling of the peripheral students, but that is, I don't have the time to go into it. So, then we could do the same correlation of the human Granger data with the monkey anatomically data, and we find the pattern in the human that is conspicuously similar to the pattern in the monkey. Gamma Granger causality, stronger in the bottom up direction, a little gap here between gamma and beta, then the beta Granger causality is stronger in the top down direction and then in the monkey specifically the theta being strong in the bottom up direction. So, also in the human, we could then build hierarchical level, which again was highly correlated to the level that was derived from the anatomy, and then we could extrapolate to all the areas that were described in the human. And, you know, without having the anatomically data for all of those areas because in the humans we cannot do tracer experiments. We see that the Granger causality places arranges the areas exactly as we would expect it from many other pieces of evidence. So, B1 is followed in the hierarchy level by B2, B3, B4, natural occipital B5, B6, B7, intra parietal areas, and then broadman area 46 and parietal cortex at the top of the hierarchy. So, to conclude this part of the talk, the free viewing of natural images into pronounced gamma and primate area B1. Optogenetics love to generate depolarizing currents and pyramidal neurons, emulating excitatory synoptic inputs under precise time control. CAT visual cortex transforms this constant excitation into gamma, revealing a critical gamma resonance. Optogenetic white noise input sequences enable causal analysis of network transmission and reveal that spikes are caused preferentially by the input component in the gamma band. This model that I show this in the literature typically referred to as PIM, which stands for pyramidal intramural network gamma, particularly when combined with an M current, so it's PIM plus M. The PIM plus M model shows that correlated gamma resonance selectively transmits input components that are coherent to the ongoing gamma light. I showed you that when two stimuli induce two local gamma reasons in macaque B1, then only the gamma induced by the attendance stimulus entrains the gamma light. And B1 entrains before at the gamma phase relation that leads to the shortest reaction times, any deviation from this phase relation leads to longer reaction times. And this suggests that B1 gamma, B1 before gamma entrainment improves actually the behavior before. So it's directly functionally relevant. I showed you in these human psychophysics experiments that when two stimuli have oscillating intensities of the same frequency, but different phases, then the phase leading stimulus is perceived as brighter. This holds for phase leads. I didn't show you that, but it holds the phase leads of about one to two milliseconds in each participant in almost each trial. So it's really constitutes a vision, a robust visual illusion based directly on the CGC hypothesis. The Granger causality between macaque visual areas for steta and gamma is stronger in the feed forward or bottom up direction and for beta it's stronger in the feedback or top down direction. And this defines a functional hierarchy. And these results are highly similar for human magnetoencephalocracy, with gamma for the forward and alpha and beta strong back, again defining a meaningful functional hierarchy, all the described human visual areas. And we saw that attention enhances top down beta. And this then again enhances bottom up. I'd like to thank the people who did the work, particularly Conrad Bosman, who trained and recorded these ECOG monkeys. Nick donated the natural food viewing experiments, Andre Bustos discovered the gamma beta direction of symmetry. Chris Lewis together with John Wong me did the optogenetic experiments together with more people from the team but they were leading it. Andre Bustos discovered the gamma beta symmetry in the monkey, the August mission areas, founded also in the human energy. Julian was only played a crucial role because he was the mastermind between linking it to the anatomy. And that correct reader showed that the top down beta directly enhances the bottom up. Gamma and Gustavo Rohncourt showed that the phase relation gamma phase relation on the tri-by-tri basis predicts reaction times, and Benjamin Stau did this human cycle presence. Thank you very much. I have another mini talk about traveling waves but I reserve this if there is time after questions. Thank you very much for your attention. If no one has a question, is it all right if I go ahead and ask a quick question? If you ask me, yes, sure. Sure. So that was a great talk. You know, everything's very, very clear. I thought the human psychophysics data, the unpublished work was really interesting. I had a quick question. Do you think that, or have you thought about whether that frequency, like that fall off frequency, beyond which you don't see the effect. Is that generated just by, say, earlier stages of the visual pathways, like some, you know, the determined flicker fusion and that kind of thing, like it's just that it's past their past band, basically. So you point to the fact that I can just go back quickly. So your question is concerned with the fact that the effect is reliable up to the critical flicker fusion of 40 hertz and then it goes away. So we don't know exactly why this is the case. The one possibility is that the neurons in visuals in the visual cortex at least simply do not follow the visual stimulus at frequencies beyond that. This is the case or not. It's actually not really known. So there are studies that investigate the effect of different stimulus parameters on the critical flicker fusion frequency so for example it can be shifted higher by more peripheral stimulation actually because in the periphery we have more rods to your cones and also by more intense stimulation. And the same is found for the entrainment of interquartedly recorded visual cortex neurons. So if you go peripheral if you have high intensity, also their highest entrainment frequency shifts up. And whether there is a one to one correspondence has not been investigated and probably require the training of macaque monkeys to report whether they see flicker or not and so on and of course this has not been systematically done. So that's the simplest possibility right so the neurons simply entrain up to 40 hertz and beyond that they don't entrain so we don't get the effect as we would expect. There is of course also the possibility that the neurons higher up do still train, but for some other reason the effect isn't there maybe because the visual illusion is really linked to this perception of more intense oscillation of the stimulus. So the first order of the study has actually looked at so many of those trials and he can see the effect up to 80. Even though he doesn't anymore see a flicker, but when he does a tool alternative for us choice, which stimulus is on higher brightness. He doesn't know it because the computer selects it so it is a properly controlled test. Then he is significantly above chance up to 80 hertz even though he of course doesn't see any, any flicker at that frequency. But that is something that is an end of one remark so just discussion. Please go ahead. If anybody has a question, please go ahead. I think Brian one to one. You're still muted. There we go. Sorry, I couldn't unmute. So the origin of the electroencephalograph waveforms, but whichever waveform you happen to you happen to be looking at is is an emergent phenomena from multiple sources in cortex and sub cortex. Right. What happens if you play back frequency and amplitude that mimics electroencephalographic waveforms into these same structures. Can you elicit the same behavioral response. You mean when we when you play it back optogenetic optogenetically or with biotic implants or electrode implants or strips or grids or very good question. Thank you. So we have not tried this with electric stimulation. I think that if we would do this so we would just use electrical stimulation in the gamma frequency range. The neurons would clearly in train to that. So we use optogenetics to do them that is in our in one of our papers from last year. When we use optogenetics, and we present basically optogenetics sine waves. It's in the paper from last year. When we optogenetically stimulate a different sine wave frequencies, then the neurons are very strongly in train to these frequencies when we when we stimulate with a five hertz frequency. They in train and on the crown of that theater in a very systematically show a little internally generated camera. In train at higher frequencies 10 and 20 Hertz and 40 Hertz and 80 Hertz, the fidelity of in train becomes better and better. So actually the neurons follow that in post rhythm with with increasing fidelity. So basically the, the, the, the, if you will the ultimate test that we did on that was to not present band limited sine waves but I guess you still see the presentation screen right, but we presented this white dots. Simultaneously contains all the frequencies, and then we found that the neural activity as measured by spikes was selectively driven through the gamma component in the input is which we drove. So they optogenetic drives contain all frequencies strictly equally. The spikes were driven specifically or caused specifically in this case we can really say that of course, that of course specifically by the gamma component in the input. And, and is this is, is this response correlated with behavior or is this an evoked potential to you think. And in these experiments we cannot say anything about behavior this was in the cat this wasn't the anesthetized cat. I, if you ask me I would bet that if we depend and we are now preparing to do that. So if we would do that in a wake my car. In the one, the animal would report it and actually we've done that in a wake mama sets, sorry, I didn't show those data, but in a wake mama sets after some training, they can report a gamma risk simulation but they can also report a sustained simulation except what happens if you give sustained simulation with optogenetics the cortex transforms into a kind of, right. But it's known that if you stimulate a primary cortex, these macaques, they apparently see this as a possible and they can report the stimulus through a second. And this has been shown a few times now. And, and we have found similar things for a wake mama sets we didn't in the mama sets stimulate and primary vision cortex, there it seems to be most immediate, you stimulate it in the eye vision area, and then it took them one training to be able to report. So, so yeah, that sort of gets to the root of my question which was, you know, how, how different would this be from say when you present full field flash to retina and record visual about potentials from occipital cortex so you can, you can you can you can drive cortex and do visual about potentials that that are are recordable in the electrophysiology but have no real behavioral or conscious component to them right so so you can drive cortex and and the person will report seeing flashes so there's no, there's no additional sort of behavioral component from the visual of potentials. And then and then you change frequency and they can be blocked. Right. The other behavioral outcome is, is you can induce, but it can use epilepsy and some folks, but yeah, okay, this is, this is cool. Thank you. Thank you very much for the questions. Hi, Dr. freeze. One question I have is going back to the human psychophysical experiments you did, and you had a phase lead versus a phase lag on those two moving in gradings. And obviously there's a frequency effect but I didn't maybe you talked about it but did you vary the phase lead versus phase lag and was there something characteristic about that lead lag that that made this effect maximal and does that have any. Very good point. Let's see. So, so I don't have the slides here with me but in this particular data that I showed you, they were the 90 degrees phase relation. But we also systematically investigated different phase relations, and of course different phase relations correspond to different time relations. Sorry, I don't have these slides here. The actual of it is that we find the effect significant down to slightly less than two milliseconds. And the other insight is that we find the effect to increase for larger phase legs up to the maximum tested of 90 degree. And when we thought about it why is that the case why does the effect get larger for larger phase offsets. We thought about excitation inhibition, you know so one stimulus wave will induce excitation, full load by the self generated inhibition of the cortex. So, one stimulus gives one way for input, which is then followed by the inhibition which will hit the other stimulus right the other stimulus that is coming in. We asked, when is this inhibition to excitation ratio reduced by the first stimulus actually maximum. And it is maximal at 90 degrees after the peak of the first stimulus, pretty irrespective of the. Maybe I can quickly show this to pretty irrespective of the, of the precise pregnancy, give me a second. So here, can you see this. So these are different phase offsets corresponding to different time offsets and shown for these three frequencies. Okay, and you see that the effect here is still significant. The effect here is still significant for 1.95 milliseconds at 32 hertz. So that's what I meant when I said slightly below two milliseconds the effect is significant. But you also see the effect is getting larger for larger phase relations. Okay. And now why, why is that. Maybe because of this reason so look at this so one stimulus induces about excitation followed by inhibition. And when is the one is the inhibition to excitation ratio maximum, we're always at 90 degree maximal inhibition minus excitation is close to 90 degree pretty irrespective of the frequency at which you drive it and pretty irrespective of the delay between excitation. I can't hear you. Oh, I'm sorry. It's interesting that it's always at 90 degrees irrespective of the phase you would I mean I would think ahead of time that it would have to be some sort of characteristic synaptic timescale. Right. So, so the excitation is intricately triggering inhibition with a delay of two to four milliseconds right that's known that's physiology that has been shown in many studies. But the question is then when is there the maximum inhibition related to the excitation. And if you actually look at it. This is always of course at 90 degrees after the peak. At 90 degrees after the P inhibition is maximum relative to the excitation. And if you now think of another stimulus coming in, it will be maximally inhibited if it hits right into that moment when the inhibition is maximum relative to excitation. And that's exactly what we find. And the second stimulus when the phase lagging still is 90 degrees phase lagging is maximum. So it's physically brightness needs to be maximally upgraded to overcome the illusion. Okay, well I think if there are no more questions we've gone a little bit over time so that was a great talk. And thanks a lot. Thank you again very much for the attention and for the interesting questions.