 I don't know, more than 20 hours of flying or something. So then we have the STP price ceremony. As you can see, this is one of the major events for STP every year. It started in 1982 towards the achievements of young scientists from developing countries, working in developing countries, that have made major contributions. The selection committee is always struggling to select the best candidates in different fields and this year they found excellent scientists who have made very important contributions from his home country, Argentina. So we have here Emilio Croff. He's a neuroscientist from Argentina working at the CONICET, which is the National Scientific and Technical Research Council, and an institute of, I would say, in Spanish, Instituto de Investigaciones Bioquímica de Buenos Aires, EIBBA, Le Loire Institute. And yes, the president recognizes his contributions in neuroscience. I will read some minor, some small details about Emilio's career. He started in Argentina, but then he came to CISA, his PhD from CISA, so it's almost a local. Moreover, he's also an ICDP associate. So it's a person that, as you know, we have this program for associates and so now he can come here and visit for some time. It's perfectly adjusted for the timing because we are starting this new group in quantitative life sciences and so he comes here as an associate from the quantitative life sciences group. His main work is in neuroscience. He has addressed several aspects of memory and spatial recognition, combining both experimental and theoretical approaches. Most notably, Emilio discovered speed cells in the brain's enthoronal cortex, neurons that encode a high precision measurement of speed. The discovery was crucial as it provides the missing link in our understanding of how path integration, the mechanism contributing to special orientation based on self-motion rather than sensory cues, is implemented in the brains of rats. So the study was featured in Nature in 2015. Emilio's research builds on that by May Britt Moser and Edward Moser, who shared a 2014 Nobel Prize in physiology or medicine, along with John O'Keefe, for the discovery of cells that constitute a positional system in the brain. The cells discovered by the Mosers, called grid cells, make up a coordinated system in the brain's hippocampus that is caused for a special navigation. A crop who's interested in memory and special navigation took the Mosers' work one step further by investigating how the brains of rats process the speed at which they travel to update their internal location map. So before starting the stock, he will give a talk today with the title, Space Time, Speed and Acceleration in the Brain's GPS. But before he starts his talk, I will be honored to give him the award, the ICDP award at the moment. So since we are late, the ceremony was quite quick, and also Emilio promised to give a relatively shorter version of his talk. So let's welcome Emilio. Hello. I hope I don't start crying after all this heroic traveling and everything. I want to thank a lot the selection committee, and I want to thank you who have endured all these changes in schedule and so on to be here and listen to some things I have to tell about space, time and speed and acceleration. So a very simple scheme. There's a rat that is going to run forward. And perhaps the rat itself or we want to understand and to be able to predict how it's going to move. So we can plot the displacement in space and time, and it will look like something like this. If we know the initial speed of the rat, then we can estimate its trajectory simply by multiplying that initial speed by the time that the rat spent running. Now this is a very rough approximation because if you look at this point in time, for example, it's going to be not a very good approximation, right? There's going to be a huge error between the actual displacement and the one that we are predicting. So what can do better than this, for example, by if we knew the original acceleration of the animal? And then we can have a better approximation still. The error is quite big. We want to know exactly the trajectory of the animal. We need to have infinite information. So we need to reconstruct the whole Taylor series. So have all this information about its initial movement. This is impossible for us, and it's impossible for a brain to calculate like infinite terms of a sum. So you need to cut this series at some point. And a lot of what I want to talk about is about how we think that the brain is doing this truncation of the series. Now, given that you are stuck with these two terms that are the first two, there's still something that you can do, which is to make this delta T lower. So to look at lower time spans. And then these two, the relative strength of the importance of these two terms in the whole infinite sum will be much more. And then your error will be also much smaller. So this is the outline of my talk, because what I originally wanted to talk about three different stories that all circle around this equation. Unfortunately, I'm only going to talk about two of these stories, the last two. And first I'm going to give a brief introduction of the systems, although we already have a very brief one. In the hippocampus, which is the area of the brain that I study, there are play cells. And these have been discovered in 1971 by O'Keeffe. And that's the main reason why he got the half of the Nobel Prize in 2014. What I'm showing here is the trajectory of a rat running in a square environment filmed from the ceiling. So you can see in black how the rat moved. And every time this particular cell is active, we are going to draw a color dot in that position. The cell activates, and we are drawing a color dot. If we do that for a long time, we'll see that there are lots of color dots, and they all accumulate in a region of space. So this is the place field of this cell, and this cell is a play cell. It learns to be electrically active in one confined region of space. And this is an example of another play cell. And you can also draw these maps as firing rate maps. So you look at one bin, and you count how many of these dots you have, and you divide it by the time that the rat has penned in that bin, and then you have your firing rate, which means that in the warmest parts of this plot, there have been 23 of these dots per second, and in the coolest parts there have been zero. These dots are also called spikes for action potential of a neuron. It's whatever happens in the neuron that is important and is communicated to other neurons. So let's do a thought experiment, which is to go to, for example, the dining room of a house and then select, we're going to have a lot of play cells describing this environment, and we can select four of them so that they are aligned here. I'm just choosing them arbitrarily. Now we go to a different place, like the garage, and we're going to find this. So some of these cells were firing in the other room and are still firing here, like the violet one and the yellow one. But their firing has nothing to do with what they were doing in the previous room. While other cells, like the red one, were firing in the dining room, but not in the garage, and vice versa. Some cells were not firing here, but also here. What I'm just telling you is that it looks like these maps, the spatial maps that there are in the hippocampus, are really describing differently different environments, and there is no structure preserved when you go from one environment to the other. So what this hippocampus seems to be good for is to store details about the places we visit. I'm going to speak about some of the sides I cut. So this is actually a rat in its brain. The hippocampus is this area here that I'm showing. And I'm going to talk about another area that you can't really see in this slide, but it's in this position here. It's an enthrinal cortex. An enthrinal cortex is the main gateway of information towards the hippocampus and from the hippocampus to the rest of the brain. And I'm going to show this video. So we have places in the hippocampus, and I'm going to show a video of what this cell in the enthrinal cortex does. This is a typical experiment that we find. This video is in YouTube, so probably a lot of you people have seen it. Those of you who have not seen it, you can enjoy not having a clue of what this cell is doing and then suddenly having a high effect and seeing that this cell is actually doing something. So every time that this cell is active, we are drawing a dot. Here my friend Jonathan is throwing chocolate so that the rat has something to look for in this environment. What you can see is that the cell apparently was not a play cell. It was not firing in a confined position, but it does know very well where the rat is. And it's firing in very specific locations in space. So this is called the grid cell, and these locations where it fires are the nodes of an hexagonal grid. You can have the spike version of the map or the firing rate version of the map. So this discovery is the main reason why the Moses got the other half of the Nobel Prize in 2014. Note that the grid cells have a typical spacing and from the very early discovery, people thought of them as Cartesian axes. They have a periodicity and they have a spacing which is a matrix of space. In the same way as Cartesian axes in a map have a periodicity and this typical spacing. So the nice thing about Cartesian axes is that you use them to describe all environments and the same thing applies to grid cells. If we did this experiment with the hippocampus, we found that there was no structure when we passed from one room to another one. But now if we have grid cells, this is slightly different because one single cell fires in many places but let's forget about that and just choose one firing field of the grid cell and then choose other three grid cells that are going to fire. We choose them arbitrarily so they are aligned. When we go to the garage what we see is the structure is preserved. So these are really like Cartesian axes that apply equally to all environments that the rat visits. So what is it good for? The hippocampus is good for storing details about the places we visit. The entranial cortex is probably good for doing spatial operations. And the most salient of these spatial operations is path integration. So a good metaphor for path integration is sailing in the sea in the old days. So if you were in the middle of the sea you would look around and none of the sensory information that you get I mean you only see water, right? So none of the sensory information you get is helping you to locate yourself. What you can do is to wait until it's night and then you can see the stars and locate yourself but of course the problem is that it could be cloudy and it could be cloudy for many days and many weeks so the sailors had to develop a different strategy. And the strategy was first to have a metric of space so a place where they could place their boat and measure distance and so on. Then to have a compass to know in which direction they were sailing and finally to measure their speed. So after a cloudy night they could estimate that they would move a given amount in a given direction and their final position would be this one. So you need three elements. The first of these elements is a metric of space and this is what we think it's grid cells. They were discovered in 2004. The only difference is that instead of having a square geometry as Cartesian axes they have an hexagonal geometry but we should not worry about that because the nature is full of examples and with hexagonal geometry seems to be more stable than the square one. The second element is a compass and I'm not going to speak about this but it was a discovery that came two years after the grid cell discovery it's called head direction cells and it's about knowing in which direction you're going. The third element was speed and speed was not so easy to solve right away. So right from the beginning there were several candidates that exhibited a significant correlation with speed but a very low one and these candidates were a frequency of theta oscillations. I missed that part of the introduction so I'm going to show some of these theta oscillations. These oscillations are variations of the local feed potential that you can record as well as activity of neurons in the brain but also the firing rate of grid cells or the firing rate of this compass head direction cells but to many of us none of this was very convincing and there is a reason for this. Actually in these experiments as the ones that I've showed you just wander around looking for food there are very few episodes of high speed. If we did this experiment here and I'm saying I have hidden like a thousand euros somewhere in this room and whoever finds it can keep it you would not be running I would not get your highest speeds like running from one side of the room to the other you would be carefully looking for this price. So it's the same thing here with the chocolate that we throw to the rest. Eventually some of you may say okay no one looked up there so I'm going to go fast and look up there but in those moments in which there is high speed there is also a lot of other things happening so there is a correlation with other behavioral and cognitive variables such as a change in attention a strategy switch, acceleration so there are lots of other variables going on there so it's really a very dirty experiment to look how things correlate with speed in one of these experiments so we wanted to find a clearer experiment in which we could control the speed of animals and for a physicist the obvious solution is to just seat the animal on a car so it's fixed there and we can move the car using a motor that is controlled by a PC and we can do it in a reliable way in a repetitive way and repeat it a million times and see really what's the response of different neurons to this the problem is that this doesn't really work we know from literature that when animals are passive their GPS networks sort of detach they start doing other things that are not really coding for space but luckily we had other sources of literature that gave us the right the point is in the right direction which was to remove the floor of this car so now we have what we call the Flintstone car we still can move this car using this motor from one side to the other at whatever speed we want it to move but now the animal has to follow the movement of the car and they do it very willingly it's a very easy task to do because they know that when they reach the end of the track they're going to get a treat they're going to get their chocolate and this is just a graph to show you how well we could control the speed of animals what I'm showing here is the track that had four meters so this is position in the track and I'm showing speed in the y-axis and this is the average speed that we tracked from the animals after several repetitions of this protocol and the whole protocol was about making animals run very fast on the left half of the track and very slow on the right half of the track and on their way back again very slow on the right half and very fast on the left half and see how there are long periods of running at very regular speed in both the fast and the slow and there are very fast transitions between the speeds so to our surprise actually because we were not really looking for that we found neurons and the firing rate copied this exact speed profile and this is when we said Eureka there are cells that are actually specialized in coding speed among other cells in the internal cortex neighboring to grid cells and head direction cells and so on and these are two examples this is one putative pyramidal cell or principal cell that you can see that fires a lot when the animal is running fast and then there is a very fast transition and fires very little when the animal is running slow and this is a cell that fires much more it's a putative fast-piking cell and again it has the exact same behavior and these are these two cells were recorded together so to these cells we call them speed cells in the immediate internal cortex and so one question was okay they are doing this in the car but there might be many reasons why they do this in the Flintstone car after all you have to be following a car that might be stressful there might be many reasons for cells to do this so we went to see what these cells were doing in the regular open field experiment as the one I showed you before and what I'm showing here is in grid the speed of the animal and in color the firing rate of the cell and you can see that there is quite a nice correspondence between the variations of one and the other and this is true also for other examples of cells so these cells had been here all the time we just had not stopped to look for them so okay what we are saying then is that an animal can estimate its movements because it knows its speed okay but this estimation is always as we said it's not very good could it have a better estimation by for example adding acceleration and we did this study but we found no acceleration cells in the internal cortex so we really think that acceleration is not something that is available for this neurons to predict the position of the animal and this is the opening to the last story that I wanted and I want to tell you about it it's still unpublished so please don't take pictures of this but it has to do with these oscillations that I spoke to you about this is called the theta oscillation it has a frequency of around 8 to 10 hertz so and what it it's a very complex mechanism that creates these oscillations but what you have to think of is that neurons tend to be quiet during the peak of theta and they tend to activate in the trough and then be quiet again and activate and quiet and activate so these oscillations that one can measure from electrodes in the brain mainly tell us about how these populations of neurons get synchronized so they all tend to fire at the same moment more or less than silence and theta oscillations were important because they were one of the candidates that people had suggested as codes for speed so it has been known for a long time that when rats run faster their theta frequency increases and people suggested that this could be a method to code for speed and that path integration could be done based on this code and these are examples from many different decades including last year so this is when Eric Carmichael a master student I had in Norway enters he did an experiment in which there were he did an experiment in the cart in the Flintstone cart where there were four possible speeds and at every lap he would randomly choose three of them so you would have for example in the first lab you would have 28 and 14 and then 21 and the second lab the animal would start running at 7, then 14 and then 28 and so on so this experiment is very nice because it allows you to compare what happens in the same piece of track but at different speeds comparing different labs and it also allows you to see what happens in the transitions between speeds and since you have four possible speeds you have 16 possible transitions between speeds so these are moments of acceleration and when we analyzed the data of this experiment the first thing that we were very surprised about was that when we looked at this episodes of constant running speed and that was very controlled because the car is moving at constant speed when we compared what was happening at different speeds we found no traces of this dependence of theta frequency on speed so for us in our data theta frequency was constant at all speeds so this was quite surprising and the next surprising thing was what we found when we went to look at acceleration episodes and what I'm going to show you here is this same protocol that I showed you before where the animal runs slow and then suddenly jumps and starts running fast so this is represented here it's running slow with zero acceleration it ends up running fast again with zero acceleration and on the way back it runs fast first and then ends up running slow and again fast with zero acceleration so in the middle I'm not showing this right now but that's where you have the acceleration so what happens with theta frequency again from what we saw in the last slide it does not vary with the running speed so it's at baseline around 8 hertz also when the animal is running slow or fast so this is true for the entorhinal cortex and for the hippocampus so what's interesting is to see what happens during acceleration and very interestingly we found that acceleration would go up together with theta frequency so and this is a very very tight control of theta frequency in the sense that these acceleration episodes lasts less than half a second so theta frequency is controlled over a period of 10 milliseconds or something and this kind of dynamics of theta frequency had never been shown before so this was surprising and even more surprising was what we saw when the animal had negative acceleration because with negative acceleration acceleration went down but theta frequency stayed at baseline all the time so theta frequency is somehow representing acceleration in a rectified way when it's positive it goes up but when it's negative it does nothing so to our initial surprise we added some more surprise and of course here I'm talking about instantaneous theta frequency which is kind of a tricky methodological thing but I want to really convince you that this is the case but just showing you oscillations theta oscillations so I'm going to show two oscillations one that corresponds to acceleration at zero and the other one corresponds to deceleration at zero and I chose these two oscillations I paired them to be very similar in this first half of the plot so when we do that they come together and then suddenly the accelerating one starts oscillating faster and this is true for other examples as well so this is very clear it doesn't have anything to do with how you define instantaneous theta frequency it's just a phenomenon that is there and now we can go back to the Eric's experiment where we had four possible speeds so 4x4 is 16 possible acceleration values and they were chosen in this way all multiples of 7 because in this way we can arrange these accelerations according to their initial and their final speed and it turns out that when you arrange them this way along these diagonals acceleration is the same so if you go from 7 to 7 centimeters per second you're having zero acceleration and if you go from 28 to 28 you're also having zero acceleration but the speed is very different so acceleration is the same across the whole diagonal the speed changes and this is true also for the other diagonals and on the upward diagonals we have constant average speed and varying acceleration so now we can really see what happens with theta frequency at different speeds and accelerations and what we saw is this every time that acceleration is negative you just see a blue color meaning that theta frequency is at baseline there's no significant variations there now when acceleration is positive you see an increase in theta frequency that only depends on what diagonal you're standing at so we knew from what I showed you previously that when acceleration is zero then theta frequency does not depend on speed but now we are saying more than that we'll say even when there is acceleration going on speed is really not a variable that is affecting theta frequency so you can do an average across all these diagonals and see that when acceleration is positive you have a very nice linear relationship between theta frequency and acceleration and no relationship at all when acceleration is negative so why have people seen this effect for so many decades I'm not trying to say that people are stupid actually people are very smart and they do very thorough experiments but there must be something that has mislead as all so partly one of the things that I can think of is that all these experiments that have been published before most of them were done on animals that were freely behaving and when animals are freely behaving they don't run at constant speed they are looking for their chocolate so they tend to accelerate and decelerate constantly so what you would observe if you look at speed is peaks of speed one after the other one so a way to model that is to think of a Gaussian speed peak as I'm showing here and these are arbitrary units I'm going to show very different units in this plot so you can calculate the acceleration from this Gaussian peak and now we can play with different models of how theta frequency could look like depending on acceleration and see what is their correlation with speed so first of all if theta frequency followed acceleration with correlation equal one which is not, we know it's not what's happening but if it was this if this was the case then the correlation with acceleration would be one and the correlation with speed would be zero and you can very easily calculate this that because acceleration is the derivative of speed then the correlations of these two are zero now if we have a rectified signal so that theta frequency changes with positive acceleration but does not change with negative acceleration then the correlation with speed goes up and the correlation with acceleration goes down there is a huge difference here the correlation with speed is half of what it is with acceleration but there is a third element and in the figures I've shown you there is no the curve of theta frequency is rather smooth it's not like it doesn't have this edge here so the third element that we will add to this is that the decay of theta frequency when acceleration goes down is smooth this is slow and if you add this element then now the correlation with speed and with acceleration are at the same level so this might really be the source of the confound and you can this is just a model but you can actually go and see what's happening in the data so these are peaks of speeds that I identified and averaged and I also averaged the acceleration and the theta frequency around them and you can see that this picture looks very similar to our model and the same thing happens if you look at peaks of acceleration instead of peaks of speed so this is one of the reasons why we think that there has been a great confound between what is controlling theta frequency it's not speed, it's acceleration but then things get correlated in the way we have some more proof but unfortunately it wouldn't fit into this version of the talk so I'm going to end up my talk trying to speculate about why we see an asymmetry between what's happening at positive acceleration and what's happening at negative acceleration why would biology care about positive and not negative acceleration and the way I'm trying to sell you this is trying to think about errors that are either conservative or dangerous, this means if you have a budget for your holidays and by the end of your holidays you end up having still 100 euros on your hand then you made an error your budget was not correct but that's a conservative error because no one is going to blame you for having money in your pocket on the opposite side if you are red on 100 euros you might spend one day in prison so that's the same kind the same size of error but it's a very dangerous one so I told you that the GPS has a signal the speed signal that is available to calculate displacement but very probably it doesn't have acceleration as a signal to estimate displacement so whenever there is acceleration the system is going to make an error right so we come back to this representation where the rat moves forward and if it's running at 20 centimeters per second at time zero then the animal is going to estimate with a linear prediction that he's going to be here after a while now let's think of a case in which at the same time he's having positive or negative acceleration if acceleration is negative then the correct prediction would be the animal ends up here so he thinks he's here but he's actually here now this is a conservative error it's a big error but it doesn't put the animal into risk if this was a safe trajectory from here to here then being in between the initial and the predicted position is also a safe place the opposite things happens when we have positive acceleration and the prediction is more far away and we don't know how much far away it is because we cannot calculate this term here so now this is a dangerous error of the same size of the other one but this one is dangerous because the animal could bump against a wall or even if he's very unlucky fall off a cliff or something like this so this kind of error you would really like to be able to make it smaller you could live without making it smaller and a way to make this error smaller is you don't have an idea of your acceleration is, as I said in my introduction to make smaller this delta t and there are reasons to believe that this delta t again this is something that didn't fit in this version of the talk this delta t is the period of theta frequency this network is doing a calculation and they are turning off and then the network is doing a calculation and then they are turning off again and the network is doing a calculation and turning off again so that resembles a lot what we would expect from this delta t so by diminishing this delta t what we are doing is to reduce the theta period so we are having more and shorter predictions when these errors are dangerous and sorry and reducing the theta period means to increase the theta frequency ok so it's very possible that the reason why we have seen that theta frequency increases with positive acceleration is that you want to make this error smaller by making the prediction smaller now this is speculation but if this was true it's a door to something that is sort of weird which is that we are saying that since the system doesn't have acceleration to compute displacement then acceleration is sort of affecting the spatial matrix of the system the animal thinks that it's in a different position of where he actually is in a way what I'm saying is that negative acceleration would contract the matrix while positive acceleration would expand it and we have very little but we do have evidence for this and I'm going to show you an example of a grid cell so this grid cell is again measured in the Flintstone car in a situation in which the animal was constantly accelerating and this plot shows for each line is a lap that the car does and then where you see the dots activities, action potentials and you can see that there is a periodicity in the firing of this grid cell in general the periodicity in one dimension is never as good as it is in two dimensions but you can still see it here and this is with positive acceleration now let's look at the same cell with negative acceleration and what you see is this so it's very easy to see that there are more of these peaks and that they are sort of smaller but positive acceleration is contracting the matrix or positive acceleration is expanding it or both things are happening at the same time we have some more evidence for this but again to really test this we are designing a whole set of new experiments to test this very precisely so this is the end of my talk the conclusions of the first one I didn't really talk about it but then I told you that we found speed cells and we characterized them and they are the ideal candidates to participate in path integration we don't know if they are actually doing it no one has found a way to for example kill all speed cells or turn them off and we don't know what is the mechanism of the communication between speed cells and grid cells when do they talk to each other and how do they do it what's the message that they are passing we don't know that but the data oscillation is controlled with unprecedented precision by positive acceleration is this a mechanism to selectively attenuate dangerous errors this is what we are proposing but we don't really know that it could be that we can test it and the way of testing this is to study systematic errors produced by acceleration this is the tool we are going to start applying systematically in the lab to study these issues so I want to thank a lot these two institutions from Trieste that have helped me a lot throughout my career ICDP and CISA I want to thank collaborators that participated in all the results that I told you about and funding agencies from the past present and future and I'm leaving you with this picture of my daughter Julia the first time she went to visit Mariloccia and this is a visit to the Physics Institute in Mariloccia and since it was the first time she was visiting you can imagine what kind of map she had to use since she had no previous knowledge of the place thanks a lot okay well thank you very much it was a great talk and very interesting questions yes hi Emilio thanks for a great talk quick question so it occurred to me that acceleration and deceleration obviously they're quite different notorically right as in when you're accelerating the motor system is acting quite differently versus deceleration you might even just thinking of in terms of effort accelerating versus decelerating given that theta is related at least in the rodent to movement have you thought about perhaps the relationship between theta frequency and acceleration the dissociation between acceleration and deceleration the effect it has on theta frequency could have something to do with the fact that notorically those things are really quite different usually I think it's an interesting idea it's hard to think how to test it typically when we think about the entorhinal cortex and the hippocampus we're talking about information that is reaching and it's very elaborated so you don't expect to have signals that respond to a given color in the visual system or a given muscle that is moving as opposed to another one this is just an argument it's not proving the point of course there are lots of things that change when you have acceleration typically if you are escaping from a predator you're probably accelerating and not decelerating so this might also be an evolutionary reason why you treat it differently but I think that still the fact that I mean I think that we could still exploit that to show what's happening with grid cells and if they actually expand or not so to do self-motion update one needs not only linear velocity but one needs axial velocity so we know that in the monkey there are cells of course that respond in the hippocampus to axial velocity of movement and I was interested to know whether you've seen anything like that in the rat it's a whole body rotation we have not seen anything like this but we have not looked for anything like this what I can tell you is that about the talk that I have not seen but I have heard of Mike Haselmo in Boston is claiming that speed cells are not modulated by axial velocity so if anything these are two separate codes it should be two separate codes so I have two short questions so you showed us magnificent observations the first observation is what you call speed cells okay but we could interpret it in a totally different way what if these antiretinal cells are just receiving inputs from the muscles and the fact that you accelerate then you get proprioception that is different that is reflecting the contraction of the muscles which could just be this and not speed that could be a totally different interpretation my second question is related to acceleration so we know that theta oscillations that are recorded in the brain are contaminated by theta breathing and when you accelerate you increase your breathing rate so when you increase your breathing rate the theta breathing will also increase maybe this increase of theta is just reflecting the fact that the animal is breathing faster and it is totally unrelated to the coding of acceleration or speed itself because these are two totally different mechanisms so those are mean questions so the first one what was it the first one the first one is is it really speed or just a muscle signal we are talking about the speed signal in the antiretinal cortex and we don't really know where it comes from typically there should be three sources of knowledge about your own movements one is sensory flow the other one is actually your motor cortex which is moving the muscles and might send you a carbon copy of that signal so you can know in advance what you are going to do and then there is a vestibular system and we don't really know now speed cells to which of them they really respond to but we have reasons to believe that it is actually responding to the motor cortex and we have seen that speed cells actually anticipate the actual speed of the animal by very little something like 200 milliseconds but you can still see these and there is no way that you can anticipate speed by looking at sensory flow or even the vestibular system it has to be that you are preparing a movement and you are getting a copy of that before you execute the movement because execution of movements takes a long time I mean it takes more or less this kind of time I mean whether it is real speed that is being measured or it is the speed that I know that I am going to have it is irrelevant you can still do all these calculations so I actually think that it is not real speed it is the muscle information that you are getting not really from the muscles but from the secondary motor areas that actually project very strongly to internal cortex and now the other one is a contamination by things like respiration and so on so we were worried about all this and one thing we did was to look at the modulation of neurons by theta so if you have a contamination of a signal that is actually coming from some other place you would not expect that the neurons that are modulated by theta would follow this faster pace that we are observing and we actually have shown that all of the neurons that are modulated by theta in the hippocampus and internal cortex when the animal accelerates they are modulated by a faster pace and when the animal decelerates they are just at baseline so they follow the same sort of behavior that as a theta wave well so how exactly would you suggest to separate these two signals well it could be so that could be a good idea that is a good idea that is another question here basically it is a similar point that acceleration will require more energy and that will require more flow of blood to your brain and that may increase your frequency so it may be just simple enough like an energy requirement yes but we have seen for example this actually might also be an answer to the previous question so we have done an experiment we did the animal on the car instead of having no floor we put a platform on the car and we just move the animal around and then we they have this step of positive acceleration and then the animal is not spending any more energy or anything like this and this effect and it is not breathing faster either unless it gets sort of nervous but in this experiment we also see the increase in theta frequency so it really looks I think the best candidate in this case is the vestibular system it looks like the vestibular system should be sending some signal that makes theta run faster another question thanks a lot for a great talk I was wondering so you do this nice analysis of theta frequencies and how it correlates to acceleration so how about theta amplitude that is at least in the hippocampus known to be very well correlated with speed and you can actually use it to predict animal position on a linear track by integrating it yes so now this is the case but I have to say that we have seen theta amplitude it was not so interesting for us from the beginning because it's more variable somehow I mean you don't see this very clear you see that in some cases it goes up in some other cases it doesn't amplitude is a local variable amplitude can change a lot from one place to another of the hippocampus or the internal cortex but frequency cannot change frequency has to be the same everywhere another question so you have speed cells that code for a speed but you can compute the acceleration from the speed signal so in principle but you fail to find these accelerating cells but then you find the accelerated signal in the theta waves so this is meaning that probably you are computing the accelerated signal from the speed cells and for that what the only thing not necessarily so the theta waves typically are generated from an external source so there is an external force which is the medial septum so it's kind of a different origin of a signal so what we are thinking of a system is telling the medial septum to go faster and then the medial septum imprints this faster theta oscillation on the system so you don't think that an internal operation local operation in the micro-circuit could compute the speed from acceleration in the internal cortex I would be happy with that idea but we just didn't find it at least in the internal cortex because it's not in the firing of the cell so what is the primary potential well again I think so we have found several differences between in these experiments such as the one in the platform that I was talking previously we have found that it seems to be that there are two roots of information of self-motion information one is more related to oscillations and the other one is more related to neural firing and for example when you see the animal on the platform and move it around you still have this increase in theta frequency but you lose all your spatial maps and so on we didn't specifically look to speed cells but I'm sure there are other people who have looked at this and so you lose the speed signal as well so it really seems there are two different roots of information and that you are profiting from that I mean on the other side a linear operation is a very convenient thing to have because you are just multiplying by delta t so you are just integrating time and that is something very easy to do in any biological system but now delta t square which is a non-linear operation it's much harder to implement so there might be reasons why you don't want to use acceleration and you'd rather use acceleration to correct your errors eventually but not to estimate your displacement just because it's much harder to do ok so it has been a long day and I know that Emilio has had a big odyssey and I know how your speed cells were excited at some point to arrive here on time but I think we all enjoyed your presentation and thank you very much again and congratulations for the ICDP press