My parisian interview with György Buzsaki
by Oleg Senkov
Published "Scientific American" magazine, Russian Ed., October, 2008
O.S. What kind of experiments you are doing right now; what are you looking for? G.B.Most fascinating things for me of course is to figure out what are mechanisms that underlie our cognitive behaviors, and this is most difficult thing in the world, e.g. if I ask you to tell me about your life in Russia, I ask a question and you keep going down forever, and if somebody ask you the same question, you may tell the same story with slightly different variations, I can ask you what happens in your brain, and the answer is very simple, there’s a light bulb that lights up somewhere, and then it kick you back in time and space, and recreates that special-temporal content, in neurophysiological language it creates cell assembly that is in the right context in time and space, and then it leads in to another part of the story, and it keeps going forever, and that story can be continued basically infinitely, interesting thing is that without any external input in the brain, the brain is capable of generating of these self-generating cell assemblies continuously, so to speak, without any investment from the environment, we all know that there must be something like this, because, otherwise there would not be any planning, any recall, any interesting things in the brain, the brain is not responding to stimuli creating an output without a meaning, but it is done according to plans, these plans are based on the past and somehow this long internal activity must be born, and these are the questions we are trying to attack right now, you know there is spontaneous activity in the brain, many people consider it as problem, it is a noise, that has to be deleted from signal responses that are used to interrogate the brain, for me, the noise is most interesting part, because it is only potential source of cognition, so, I can tell you specific experiments we are trying to do, they are very complicated and difficult, because, you know, to answer all these questions, you have to record from large number of these cell assemblies, in order to do that you have to do it in animals, and animals won’t tell you about their episodic information, they do not tell you much about planning, and somehow you have to design experiments, where, at least you can see, that this evolution of activity, let’s call it, trajectories, you can imagine that in such complex interactive environments such as neocortex, or the hippocampus, my favorite structure, this is n-dimensional space, where activity can go in any tortuous way, but there are links, and an initial condition, an initial trigger, e.g., when I asked you about your days in Leningrad or St.Petersburg, that triggered the cell assembly that the that determines based on your past which way your trajectories travel, so, what we are trying to understand is how these trajectories are formed, and first of all, whether they exist in a such a small brain of a rat.
O.S. So, basically, it is a question of where memories are stored. G.B.It is a little bit more because these cell assemblies are not necessary taking you back to the past, as episodic memories, but it is an essential part of planning forward in time.
O.S. Understandable, so, basically, in our brain, there’s no quest for past, it always somehow undergoes of special activity, which is created by current state, right? G.B.Brain does not have current state, past state or future state, brain does not have time or space, what is done nicely, and we have specific experiments that demonstrates that, that, at least in the structure we are dealing in the hippocampus, space can be translated into time, because organization of brain activity is fundamentally in time, and that timing is translated into synaptic strength, there’s nothing else in the brain but activity of neurons and connections between these neurons, and these connections can very as a function of time, as a function of pattern, and the can vary in a short scale, and much longer scale, so, you have to face this issue, you have to recreate time from synaptic changes, from interactions of neurons, and then time again recreates space.
O.S. OK, it is very interesting, but how you experimentally achieve it? G.B.OK, let’s be more concrete, I can explain you quite beautiful experiments; hippocampal cells are place cells, this was discovered by John O’Keefe at the university of London, in 1971 or so, the discovery was about hippocampal cells fire at particular parts of the environment, that they explicitly determine xy-coordinates where the animal is, so, it is very fascinating observation, if a mouse or rat is running a simple track, nothing doing else, just running forth and back for food reward, if you record from large number of neurons, every single segment of the track is represented by one or several neurons, and these neurons have very wide representations, neuron or neuronal assembly picks at particular point in the place, but it has representation a little bit almost everywhere else, in short, neurons have particular representation of the “here”, but also they contribute to representation of the past and a future, the past position of animal, and a future position of the animal. Second interesting part of this large distribution of place cells, is that each neuron represents about half a meter or a meter of representation, that every single part of environment is represented by multiple neurons, so now, the outcome of this is that if you take a time window anywhere of 100 ms, which is a time window of famous of hippocampal theta rhythm that a lot of neurons will fire in that time, question is what is their temporal relationship between each other? E.g., if an animal runs from left to right, neuron A will fire first, following by neurons B, C, D and so on, and their representations will be a certain distances from each other, let’s assume that neurons A and B have representations which 30 cm apart, now you would ask what is a temporal relationship of these neurons, and then you can say that at every speed of the rat, 30 cm of a track takes about a half of a second of the rat’s run, so, the temporal distance between these two neurons, when they fire maximally, is a half of the second, that’s not that so exiting, exiting is, what is temporal relationship within a hundred of milliseconds time window, and the answer is that a neuron A always will fire before a neuron B, with the fixed temporal relationship, let’s say, if their place representations 30 cm, they fire at 10 ms time offset, now a distant representation between a neuron B and C was 40 cm, then you at a hundred millisecond time window, time difference between B and C, and you will find, aha, this is 12 ms, a bit longer because the distance representation is a bit longer, and if you do all these measurements, and do the measurements with a ruler in the real world, and then you look at the temporal relationship of these neurons, and you plot them, then you will find a nice relationship that the time will perfectly represent distances, and these distances of future of distance representation can be read out from a single cycle of the theta oscillation several second earlier, so if you have a large number of neurons recorded simultaneously, and you have just a snapshot of a hundred millisecond time, and you look at the sequences of these neurons within this small time window, at the theta cycle, the temporal sequences of the neurons will provide you an intelligent guess where in the future those neurons will fire maximally, and how far their representations from each other in the future will be.
O.S. It looks like that in the hippocampus, and because of the hippocampus this information about space and xy-distances can be compressed and then encoded? G.B.First of all, there’s the first version is that a temporal representation at the long scale, somehow, these long-term temporal scale representations must come together at the time scale where neurons operate, it means that in order to connect two events in the brain, you have to bring them in time of synaptic plasticity, there is no mechanism in the brain that can bring together events that are several seconds apart or longer, synaptic integration occurs in neurons that are observers of presynaptic activity, their time-sense is best in tenth of milliseconds range, and it decays off after hundred milliseconds range, after about a time, two neurons send a message to common down neuron 1 second apart will be decoded by this neuron like two independent events, only when you bring them in time of short intervals, which drops into a time of integration of postsynaptic membrane of the neurons, in technical terms, the downstream neuron will say, aha, there’s a relationship between these two neurons, so, the key in these experiments that I describe to you is two oscillations, it is a little bit complicated process how it is, I am just telling you the end result, these long events are equals to distributed of tails of distribution, they can bring together neurons in much shorter time window, and once they are brought together in much shorter time window, they can be bound together representing something interesting about the world.
O.S. You basically answered the very question met in many scientific journals and popular science magazines namely how it is possible that our every day behavior which consists of quite a long events, in scale of second, minutes even hours, can be represented in the brain where neurons in the neocortex or the hippocampus operate at speed of milliseconds, how it is possible that our time in reality is compressed and represented in the brain? G.B.Yes, I can give you one example, that is very understood process, of course, there are other ways to bridge time, even beyond of one second time scale, this is usually termed as short-term synaptic plasticity, e.g. when you’d like to keep in mind something like a telephone number, this is the best known example of working memory, or short-term memory, oh, voluntarily I keep focused on particular things and try to distract all other incoming information, and this is done perhaps by persistent ongoing firing of a set of neurons, in my prefrontal cortex, this is a very old and convincing idea, that there are neurons in the prefrontal cortex, that keep firing as long as you keep something in mind, most of these experiments were done in monkeys, but this mechanism is always effective, it has specific components, meaning that it is very difficult to understand how a few neurons can represent so much of information, the other alternative is that in this cell assembly, which is cycling, reverberating and self-sustaining, you just change synaptic weights transiently for a minute or less, there are ways to do that, we know that from in vitro experiments, some synaptic contacts will be weaker, other will be stronger, and this changing of connecting strength will determine the path of the trajectory, so, if you have n-number of neurons, which can be connected in n-times possible ways, you can change synaptic weights in any direction, let’s simplify, let’s take just two neurons which are connected to each other, we can keep information about short-term memory by making one neuron to fire persistently, when the other neuron, target neuron will fire at particular frequency, that means indigently, this is very expensive method, the reason why this so expensive method is because the neurons should fire for a long time with many many action potentials, and sustaining action potentials are energetically very expensive, as well as it is not know how it is reduced, how it is possible, the other alternative is just to change synaptic weight between the neuron A and neuron B, that is the synaptic strength, at the time scale the same that require the neuron to fire persistently, and this change, of the synapse, will, can represent the same bit of information, this is something that people think very-very little, the information can be stored not only in the long-term, but also in the short-term, in synaptic weights, the reason for that is in behaving animals measuring this is very difficult, but people who do slice experiments, they know that there’s enormous capacity, because its ability is as good as firing of neurons, but it costs much less (this was published last week in Nature Neuroscience).
O.S. Another very known phenomenon in learning and memory is synaptic plasticity, namely, LTP and LTD, people believe that exactly as you said, that our memories are stored in different weights of synapses, but you, you are telling that for memory organization, for memory consolidation, your oscillations have to occur in theta or gamma frequencies, so, how one can put these two paradigms together and create some kind of mega-paradigm? G.B.In order to make a synapse strong we need to have strong relationship between sets of neurons, most effective way is to make a presynaptic neuron or cell assembly fire in a short time window and repetitively, in other words, synchrony is all essence you would like to do, because, one neuron does not make a lot of difference on the target neuron, but when neurons come together in time, their impact on postsynaptic neuron is very powerful, and then they come together they can change synapses of individual partners, so, this is where oscillations come very handy for particular reason, the essence of brain activity is you have to bring together elements cohesively that represent the same part of information, we call them cell assemblies, at the same time, you have to segregate them from the all other representations as competitive ones, integration and segregation are most important things in the nature, what is similar, what is different is always a matter of context, because what is not similar, they are not identical, therefore they are different, in the brain, it is also very critical, because, what are those neurons which represent similar aspects of something, and different aspects of something, and how you segregate these two sets of information, now, separation can be done very effectively in time, anything what we call information has a beginning and the end, it is like a DNA which codes a protein, in order to has a plan, you have to have a start signal and stop signal, and now applying this idea to the brain, if brain activity would be just continuous, without any beginning or end, you would never know, which one is beginning, which one is a end, now, one step forward to your question is we know that timing of pre- and postsynaptic neurons is very critical whether a synapse will be strengthened or weakened, so, the sequence of activity, and the beginning and the end of the massage you would like to code is very critical, the simplest way of packaging the information is to have a stop and start signals…
O.S. How it is possible to do in the brain, where everything is so massy? When information comes into the brain from so many modalities… G.B.No, the brain is not massy at all, more I look in the brain more I realize that it is fantastic engineering device. When you have oscillations, they are done on half of inhibition, oscillations also mean that excitability of the network, of the neurons which are involved in representing something goes up and down, goes up and down, at the extreme case, at the end of the cycle, they are completely silent, or at least they decrease their activity, that a perfect start’s point, it is a perfect way how they speak, doesn’t matter which language you speak, Russian, English or whatever, you have to punctuate they have a speed 2 Hz the way they speak, if I would use all words continuously, my speech would be incomprehensible, the same is applied in the brain, start and stop points and oscillations, they provide this “speaking” of neurons very effectively, what inhibition is good about, it gives you windows of opportunities when events can occur together, and once cell assemblies they miss this opportunity in particular time window, they can join only in another time window, and they are segregated in time, therefore, they will represent something different from those neurons that had happened oscillate in one time window, now, information can come in short packages such as wolves and consonants, longer ones, such as words, even longer ones that are part of the sentence, and so on, and so on, and they did not interrupt multiple oscillators, so, the brain invented a genius mechanism of connecting these oscillations in very logical and persuading ways, namely that local computations usually involves high frequency oscillations, which can be coordinated by global much slower but larger amplitude events, e.g., if your computation is going in your parietal or visual cortex, or in prefrontal cortex independently, that’s not very useful to combine those attributes that they represent, but if there’s a coordinating mechanism to combine those representation by slow oscillations, that fluctuate membrane potential of those neurons, and use possibility of spiking or not spiking simultaneously, that their common targets where their interactions will be already determined, so, oscillations generate synchrony almost for free, this is why the brain chose this very effective mechanism, repeatedly, it allows cell sequences to be formed, there’s nothing else, but excitability, a recruitment of a cycle from a single oscillatory period, those elements which are most excited, start fire first of one that less excitable, so on, and so on…, so what you need, synaptic strengths and oscillations and you already know much how the brain operates.
O.S. OK, and this is how our memories are stored? So, if we take these basic principles then one can assume that certain memories can be stored in certain loops in the brain, which are constant active or they become active at certain target, or how it is organized? Could it be that in our brain all the time, exist millions of such a small loops which represent certain aspects of memories, or, as many people believe they are stored in protein-synthesis-dependent mechanism, via synaptic plasticity changes or other forms, but not in reverberating circuitries, like small constellations in the whole universe, so that they represent certain aspects of memories or events, or whatever you may call them? What is current paradigm for this? G.B.Current thinking of course, that long-term memories are represented in changes synaptic weights, that is communication strength between neurons, that can be understood in various ways, there’s enormous amount of very useful and very powerful molecular biology, that is trying to investigate that, the bottom line of all these things is that whatever these protein synthesis does, because we want these memories last for very long time, in order to do that the most effective way is to change the strength, or weight between neurons in the synapse, so this alone is not good for two reasons, one is no matter how information is stored, in what format it is stored, in order to be useful, it has to be played back in common currency in the brain, which is spikes, so somehow it has to be translated back common synaptic communication into spiking communication, even if the proteins are very important, they are subset of the process that is essential electrical, another issue which e.g. your memories of St.Petersburg if they would be stored in proteins, they are already gone from your brain, because proteins do not live for a long time, they have to be recycled, they have to be recreated, we do not know for how long synaptic life exist, some people estimate for a month, some go down to a week or even days, so, if the information is not present in your brain repeatedly, it may vanish forever, that’s sounds very paradox to most people because, some information comes back even you never thought it yet exist, giving us a forth of impression that there are certain permanent molecules in the brain that do this. We know that it is not a solution; there must be something else, there are ways to reconsolidate this information, strengthen it over and over, first one is to re-remember those events, remember them consciously that we remember them remembered, and we know that there are other memories which we never recall, there must be some unconscious way of recreating this information, and this is where spontaneous unconscious information comes very handy, all is a wonder, why one third of our life we sleep, what is the use of a sleep, because it is so expensive energetically, because the brain uses enormous amount of energy, 20% of whole our energy is devoted for the brain, and during sleep these processes can be repeated even though if you are not aware of them, in such brain areas like prefrontal and entorhinal cortex and the hippocampus, particular in the hippocampus, because we know that this structure is critical for storing a particular type of memory, personal information about events in our life, so-called, episodic memory, that are only yours or mine, and they are meaningful only in your life context, and these are also conscious memories, this particular type of memories are very different from those how to play in tennis, or do something, this is a area the most studied and perhaps most fascinated.
O.S. So, in other words, if we take in future one memory of one person, and put such a memory into a head of other person, in my understanding that such memory will not work at all, because, everything in the brain is organized in experience dependent manner of certain individual, so there’s no a universal code so to speak… G.B.Why not, everything in the brain has to be calibrated, this is a frame in my book I am using over and over, that the brain are useful as their calibrations, in other words, calibration is engineering term, and in most used as learning, you have to teach the brain, if you raise a brain in artificial environment, let’s say, in a dish, even though all sensors, eyes, ears, nose, skin, everything is connected in the brain, you can not sense anything, all the information can come in your brain, but you can not move, you cannot move your eyes, you do not have any outputs in your brain, that brain will have the same activity as mine, if you have the same number of synapses as I do, if you cut off a small portion from e.g. visual cortex you cannot tell any difference at microscopic or even at electron microscopic level, but that brain is totally useless, that brain can watch TV, your mama can caress you every day, it is can go forever, that brain never learn to see, to hear, it never learn to fear, it has no idea about of the shape of the body, or anything, once we connect all effectors, if all of a sudden of magic you would connect all muscles, that brain could do nothing useful, it wouldn’t generate an output which will be useful, in other words, a thought, a thought – it is a main output of the brain, it goes not through the muscles but some indirect way, because such a brain is not calibrated, you cannot impose information into the brain, whenever you take a book, starting with British Emprises, it starts saying aha, the number thing for us is perception, I am visiting a professor now, Rose, who is the head of one of the largest departments, it is called, perception and action, I keep teasing, it is a wrong term, it is very out-fashion term, it should be a department of action and perception, because action should come first, actions is a way of how a brain generates an output, the output has consequences, this perceived by the sensors, which is meaningful, but without an output there’s no meaning from an input to the brain, absolutely zero, if it is just misleading British Emprises, a Pavlovian paradigm, that supposes that only you need is to impose an association in the brain without a contribution an output from the brain, it is totally fool, an useless paradigm, I am not saying that is totally wrong, I do admire what he has done, the way how it can be taken to the extreme that any brain will be learning something, and it is true, you can make associations of anything at some particular neuron circuit level, a visual signal, an auditory signal, will be paired in this imaginary experiment where the brain has no output, but in will have consequences in that part of the brain, where this association is already made, there’s no consequence of translation to an output which will be an useful action, so, we can approach it differently, let’s say you would like to make a robot, that would fool me, let’s say you built that robot and put on the shelf of a turtle, my task is to tell a difference which is a real turtle which is not, a way how you can make and anyway did object, from an unanimated object to make an animated one, is to give to it a purpose, if you build a robot which has a goal, a first thing it has to do is to produce an output, in order to reach something like electrical outlet, because it has to charge its batteries, this is what the robot learns, for a real turtle it is food, a reword, in this case, every single part of the robot, should be built in a purpose of nothing else but for the purpose of this behavior, if electricity has no smell, then there’s no point the robot has smell receptors, if electricity, doesn’t have any visual component, there’s no point for building a visual system, unless, the electrical outlet has some shape and form, it is always from an output side, that we have to look at the animated world to see how they are organized, but not other way around, this is exactly how evolution works, and this is why, it is so catchy and difficult to look at the effects and causes, when we talk about goals, we reach an interesting territory of religion, and purpose, and all these things, but the goal is reconstructed after the fact, it is not a drive, it is a circular causation which is so powerful rejected by European thinkers, but in fact, if you go back to many philosophers, it turns out that, cause and an effect only present in European, it is started from Aristotle, in eastern philosophies, coincidence is what so important, not a cause and an effect, and synchrony is so important.
O.S. So, let’s talk about abnormal oscillations, if current paradigm says that for memory formation theta frequency has to appear at some brain areas to facilitate this memory formation and consolidation, and for the certain perceptual tasks, gamma frequencies has to be fully increased, so, in this respect how one can say why at certain abnormal states of the brain, like schizophrenia or epilepsy, gamma and theta frequencies are increased, so the brain operates at different frequencies at much higher than during normal states, it looks like oscillations not always serve for good, they can be at certain conditions bad features of brain operations? G.B.Well, it is quite obvious, in fact one of the reasons, why oscillations brought in front of neuroscience is because people who had ability to watch brain work in real time, during EEG recordings, they concluded that if anything interesting going in the brain, such as cognitive behavior, then what you get from the scalp is just flat EEG, flat activity of the brain, every single time there was something we call prominent oscillations, because either no conscious state such as sleep, you had Parkinson’s tremor, or starts of epileptic seizures, which characterized by synchronous activity, only the past 15 years or so, mostly for the therapeutic reasons, neurosurgeons began to insert subdural electrode grids of the human brain of patients so they understood that what we see from the scalp and from the brain is so different as to see in a foggy day in 30 miles from a center of Paris, you don’t see any details, and I make conclusions that Paris is blurry place without any distinct colors or anything. But once you start seeing elements of it, the richness of color, a brain activity, then you realize, aha, maybe this is very-very important. Now, what we just discussed, information should be transmitted into packages, because this is how observers, neurons, can make an intelligent sense of this information, the brain was tuned over the millions of years of evolution, to be in a particular time scale, once you break that time scale, all of the sudden, a lot of things are lost. Lost can be very subtle, you can lose your consciousness temporally, like when you have seizures, or you have really little changes, when local computation cannot be coordinated globally, and then you have schizophrenia or many-many other diseases. So, recognition is becoming very clear, both, in clinical neurology, and psychiatry, that these oscillations are very important phenotypes, they can be measured, they can be characterized, and many-many drugs effect them in predictable way. I am invited on many-many meetings on drug discoveries, because, people have realized that understanding the brain doing research on drug that affect particular receptors is very difficult, at least we know that they affect particular mechanisms that are in turn are important for something, then we can classify newly synthesized drugs effecting thalamocortical system, limbic system, brain stem and things like that…
O.S. But you cannot create some drugs which will abolish any kind of oscillations, can you? B.G.Of course, you can! Every single anesthetic is changing the oscillations, the way how you it measure, like a physiological event when anesthesia kicks in, there are machines that are available now, conscious meters, if you want, because they monitor the level of consciousness and pain threshold and things like this, based on measuring the brain oscillations, so, this is a good example, as you just said, that if you look at firing of neurons, the firing pattern of the neurons doesn’t change dramatically when you are anesthetized, in fact it is so dam difficult, just to see when anesthesia kicks in just looking at the brain, not the behaviors, it is easy to ask you your name a hundred of times, and when I ask you at seventy times and you do not respond I know you lost consciousness, but when I look at your brain, grossly, especially, if I look at such elements like action potentials, it is very difficult, but if I look at overall state of oscillations, not only one, but everything together, then it becomes very clear, aha, there’s a threshold, which evolves how far a coherence of particular group of neurons in space can collaborate with each other, at what level of anesthesia, or drug effect you break down these long range of communications, this is the time when you lose consciousness. Alcohol is a very good example, alcohol is effecting GABA-A synapses, which are inhibitory very powerful synapses from inhibitory interneurons, they are very important sources of oscillations. Alcohol enhances these synapses, it is GABA-A agonist. So, it simply slows down these oscillations, gamma frequency oscillations, which are normally between 30 and 100 Hz in your brain, now goes from 25 to 50 on it. And that a little time shift is enough that you have a different brain. Another example, if somebody undergoes open-heart surgery, in such places like Moscow, or in Colombia University in New York, such surgery is very complicated, then they cool down your brain, because they don’t want to damage the brain by the long procedure, the blood flow is not perfect, you have to stop the blood, you have to stop the circulation for a quite long time, while you are playing with the heart, the way you do it you cool the brain, and once you cool the brain, first of all, you alter the oscillations, and if you cool it a bit further, you can block brain activity completely. But there’s another story which I would like you to tell as an illumination, is that you are deaf for two hours, without any brain activity whatsoever, the operation is successful, the brain temperature is elevated, the brain activity comes up, and you are looking at the mirror and you say I am Oleg, there’s no information lost whatsoever, the brain rebooted itself, this is what fascinate for quite a lot, for quite some time because, it is the same thing with anesthesia, like many of us, we had surgeries many times before, when I was done with anesthesia, I realized that after anesthesia also many funny things happened in my brain, yet, when it gets off everything comes to normal, how it is possible, actually this rebooting happens hundred and hundred times during sleep in every single individual, so when you go into deep sleep, there’re delta waves, or very slow oscillations in your brain, it has been know for a long time, that happens very synchronously over neocortex, now, when you measure what happens with neurons individually, we know that there are times when oscillations change excitability during a particular phase on this cycle, neurons are completely silent, they are silent from prefrontal cortex to occipital cortex, for about 200-300 ms, no body fires whatsoever, but on this total silence on several hundred milliseconds, the brain reboots itself, and it is completely normal. I will tell you another similar fascinating example, let’s talk about experiments which we do in real with animals. Just imagine if I could turn off your hippocampus in the middle when you telling me a story, I just stop any activity in this structure for a minute; the question is, if we turn the activity back, what will happen. Will you forget completely that you are telling me a story? Will you continue the story where you took off? Or, will you continue the story at the point where you would have been without the turning off the hippocampus?
O.S. So, what is your answer? What will happen? G.B.We know the answer, because we have done such experiments in rats. It is not lesions, we just could turn the hippocampus off for two theta cycles in the same paradigm that we discussed before, the rat is running in the track, and there are place cells, and then we stop completely electrical activity of the hippocampus bilaterally, this is done with stimulation and then a strong GABA-A and GABA-B inhibition which lasts for two theta cycles, and when the activity came back, the phase the oscillations, the neurons that resume their activity at the phase, where they would have been without their silent.
O.S. It is very fascinating indeed. G.B.The reason why it is so fascinating is, first of all we saw the paradox, but then we realized that aha, we didn’t manipulated the rest of the brain, only the hippocampus, and the rat is running in the same environment all the time, there are updated available about the position of animal in the track every single time. The conclusion is that the hippocampus can be updated, rebooted, instantaneously. Now, implication of these experiments of course is very important, namely here, we had environment signals available, but in our imaginary experiment with you, there would be no clues where to continue your story. The prediction would be that you will continue where you took off. Why do I say that? Herbert Jasper, from Montreal University, you may have heard the name, he was one of the greatest neurologists of the past century, He was a professor at Montreal Neurological Department for many years, I gave a couple of talks there, and every single time I said something about oscillations, he was one of those people who were commissions, who were saying everything the time I am seeing oscillations like epilepsy or something else is pretty bad, so he had some clinical observations, you ask a question of a patient “what is your name?” and you hear “my name is Juri” and then you pose for 40 sec on a minute because you have epileptic seizures, and then, at the end when the seizure is gone you say “Buzsaki”, not being ware at all that you have been absent from the world for very long time. Now, if you record from the brain during these epileptic seizures then you say how on earth it is possible that the activity didn’t erase the information and the brain rebooted itself instantly and set you back where you started from. Herbert Jasper had asked this over and over again, all I could do is just wave my hands. Now I have an answer. The answer is in the system such as a brain network, activity patterns depend only on two conditions, the connectivity of the system and initiating condition. And if you impose on that network a predictable signal such as periodic oscillatory epilepsy, when it is over, that you remove that constant predictable signal, then you get back nothing else but an initiating condition. Who controls this? It was a trigger, a question which should lead to neuronal assembly because in order to responding in real time, it is a trivial thing, this is what cerebellum can do, the cerebellum instantaneously online monitors everything, but once the input is over, the effect is over. Neocortex and hippocampal systems are very different, you give an input and can lunch a long-lasting effect which can be a few hundreds milliseconds or up to a year, at least it goes on till you fall asleep, but even after your sleep it can come back, without any external input. This is an essential thing about the brain, that it is spontaneously active, there’s a lot of spookiness in this.
O.S. People say that the hippocampus is trisynaptic loop, everything what gets into this loop starts reverberate… G.B.You can say that to any parts of the brain. The brain is a small network, you can go anywhere to anywhere else, at least in the neocortex, just through 5-7 synapses, the hippocampus is an amplifier, it amplifiers powerfully any signal, this is why it is so bad in temporal epileptic seizures, the hippocampus is the strongest amplifier in the brain. What I mean by that it produces so powerful synchrony that no other structure can do. And this is what we discovered when I was a postdoc in the hippocampal sharp waves. In your hippocampus, for example, right now, less than 1% of neurons cooperate in about time window of 100 milliseconds, it is very small number of neurons, however, when you fall asleep, about 15% of neurons which come together in this short time windows in these hippocampal sharp waves. This synchronous activity occurs at various stages of the hippocampus, subbiculum, presubbiculum, parasubbiculum, enhtorinal cortex, and in the rat, from 50 000 to 80 000 neurons discharge. So the hippocampus takes a very high risk of doing this, and this is for something useful, we think this pattern is extremely useful because it is responsible for transfer of information out of the hippocampus to neocortex during sleep, this critical pattern for what we call memory consolidation in the subconscious brain. But because it produces very high synchrony, it is a very risky device, because it can do even more synchrony than necessary leading to temporal epilepsy.
O.S. OK, if oscillations in our brain, their organization are so important for functioning, remembering, recalling and even thinking, then how one can manipulate with all these oscillations mentally or physically? I have seen some papers, not too many yet, but they proved that if a person listens to certain type of music, which has theta component of frequency, then it can boost up his or her memory capabilities, or if you do something again rhythmically, in certain frequency for a long time, you can enhance your thinking capabilities, e.g. physical exercises, like endured walking, jogging or running. How in everyday life we can help ourselves to benefit from these brain oscillations? G.B.There are many examples, I discuss some of them in my book, one of the older area is called “alpha oscillations and relaxation”, and this is based on very simple observation that when we are relaxed and we close our eyes, then from 10 to 12 Hz oscillations in you occipital cortex. The conclusion, then if you can produce such oscillations then you make the brain relaxed, but this is a little problematic, you can generate oscillations wishfully in some parts of the brain working very hard on that with other parts. So, if you sit at front of the screen looking at your feedback from your brain activity, saying “I want to do it, no matter what but theta oscillations” then it will exhaust you more than relax you. These movement, I call it alpha movement, was a total failure, but it was successful for thousands of years in the middle east, I mean in India for example, where many data wishfully and purposefully exclude the rest of the world, they do it in many ways but they also change their brain activity, we know that because now in many yoga’s brains we looked in. The first time I was faced with this was when I attended my first international EEG meeting in Amsterdam, there was a poster from some institute from India, and I looked at these signals from the yogis, this is very similar pattern of activity what I learned in neurology – epileptic activity, in other words, in practice you can produce rhythmic activity which gives you windows of opportunities to simply exclude the brain. What you do with brain oscillations, of course, it is self generated activity timing of which is independent from outside of the world, so, the outside of the world can perturb the activity only in small time windows, and if you do it right then you are excluded from outside of the world, and it is depending on the purpose of the brain training, you can use the rest of the brain capacity for internal use, and it is the first goal of meditation of course. Now in more practical things, now many of us experienced that when you are jogging, when you are doing some activity which is not social, when you do not have to interact with your partners, like swimming, for me it is the best time when I can think when I swim, it is not a hundred meters swim, it must be much longer period of time, and I do not have a proof, I have nothing, but in fact, there’s one reason, perhaps doing that you produce a pattern that is oscillatory, therefore it excludes the rest of the environment, and other things which come with isolating yourself. So, this is one way, but there’s a better way to do so. The good news is come from a research group in Germany, Jan Born group from Lübeck, I had a postdoc working in my lab, now she is working with Jan Born, she published the paper in Science journal which is the following. These slow oscillations during sleep facilitate the occurrence of these sharp waves in the hippocampus that we think are very important for memory consolidation. The neocortex produces the slow oscillations in stage of deep sleep, the argument; the hypothesis was that if this is true that if you can enhance the slow oscillations then they could produce more sharp waves which you can record in the hippocampus, but of course they did not record, because the subjects were students, but what they could do, they put electrodes on the surface of the brain and entrained the slow oscillations via electrical fields, in other words, they enhanced the magnitude of slow oscillations of the human brain, and the measure they used to prove that they did something useful was episodic memory. So, they gave episodic memory for the students at nights when their brains were stimulated and when their brain was not stimulated and they showed that there was an extra gain as a result of the stimulation, in other words, the brain remembered better if the entrainment of the slow oscillations was enhanced. Now, this is an experiment where I see a lot of potentials, first of all because it is extremely simple, it doesn’t require any sophisticated things, I can even imagine that some entrepreneur would sell such machines. And the reason why it is so important is because these slow oscillations occur in deep stage of sleep that I have very little, because it is a first pattern which goes way when you get not so young, after forty years of age, a proportion of your slow wave sleep decreases tremendously and this parallels very nicely with memory lose in the brain. One potential way to restore it is stimulation, so now it was 10% gain in a young person, who does relatively normal stage of slow wave sleep, but what if you could apply to a person who has very little of it? Can you improve it in further? So, this is a real experiment, it is not trans cranium magnetic stimulation, you just put EEG electrodes, you use electromagnetic field, just below of the threshold of producing muscle stimulation or any kind of other effects, some students said they felt something, but they still could have normal sleep pattern, and it is slow oscillations, it is not like a pulse, it is slow frequency, it is quite safe, and the equipment was really simple, you can build it for a hundred euro. The critical thing that you somehow have to monitor the activity of the brain to apply the stimulation in right time, this is something for future research, what if I have it for 24 hours in my brain, maybe it will be detrimental, it must occur at the time when state of brain is about to get in that stage of sleep, so, I have to monitor my state of sleep. But this opens up very interesting application, we have to think about.
O.S. How Mr. Buzsaki can make an one-sentence updated definition of "memory", what memory is? G.B.The dictionary definition of memory refers to the brain’s ability to recover information about past events. These days many neuroscientists equate memory with synaptic plasticity. While some plastic process is surely needed, there is a large gap between the synaptic-molecular level description of plasticity and our ability to travel back in time mentally a recreate a series of remembered events in a proper spatio-temporal context. Psychologists tend to distinguish many types of memories, the two major categories being consciously declarable and those which we do not ‘feel’ or be aware of. Such distinction is quite problematic when it comes to research in animals, with the goals of identifying brain mechanisms of such conceived boundaries. Instead, I try to identify mechanisms in a given brain structure (hippocampus) and see what aspects of the activity relate to phenomena best described with words as memory. The most interesting aspect of memory and the mechanisms behind it for me is that it is an unending, unlimited, self-generated process. If I ask you to tell me about your upbringing in Russia, with this single cue the brain is capable of hopping from one event to the next virtually forever without any further external cue. The mechanisms which allow this internally organized linking is most fascinating for me.
O.S. What Mr. Buzsaki thinks of how neuroscience will be changed in 50 years, 100 years? When major mental diseases will be combated? When neuroscience will become more practical in making our everyday life happier, healthier and more productive? G.B.I prefer to refrain from making sweeping predictions. Scientific ideas mature very unevenly. Things that ‘must be solved’ appear to resist, other times unexpected discoveries pave the way to many many explanations. Understanding the mechanisms of episodic memory is most interesting (and challenging) because it has so many ramifications. At its core, episodic memories are what makes us as individuals. This calibration process by the environment and other brains requires plasticity. Memories are affected by sleep and are affected by virtually all cognitive diseases. The implications for understanding memory will likely have implications beyond brain sciences. The hippocampus is the best known ‘search engine’ capable of creating wholes from fragmented information. If I give you a few details of a book or movie that you have read or seen, very quickly your brain can recreate the entire story. In comparison, Yahoo and Google are very ‘primitive’ search engines because they have not been ‘calibrated’ properly for human use. Human memory has ‘externalized’ extensively over the past few centuries and this externalizing proceeds at an unprecedented pace these days with the WEB, thus it is of utmost importance that we not only deposit the knowledge of mankind but also make it possible to use the stored information flexibly by individuals.
O.S. If Mr. Buzsaki has got once some unlimited sources of funding, on which projects, problems..etc in neuroscience, he would focus first of all his efforts? 1, 2, 3 by priority... G.B.It is not only the unlimited funding but also unlimited time what is critical. Without limits in time and manpower resources, the most effective strategy is to systematically explore everything. This is the approach taken eg. by the human genome project or cataloguing galaxies. However, when resources and time are limited the best strategy to discover the unknown is random search – no matter how counter-intuitively it sounds. Thus, the main issue is what an individual thinks is most interesting and solvable because this is the topic on which a motivated scientist wants to work on. Scientists always like to work on their own ideas. Brain oscillations and memory are the topics I would like to understand or at least keep working on. Every step of the process of understanding is entertaining, rewarding as well as useful for others even if I never get to the end.
O.S. What other hobbies besides the neuroscience Mr. Buzsaki has been possessed by so far? :)) G.B.I used to be an amateur ham radio in my teenage years. This is perhaps my first source of inspiration for communication, coding (I used to be very good reading the Morse code) and oscillations (I built most of my equipment). My main interest today is contemporary architectures. I am particularly intrigued by scalable structures since scalability of functions is also a key problem for the brain.
O.S. There are constantly circulating rumours in the field that it is much easier to reach the Moon than to become a postdoc or a PhD student in Buzsaki's lab :)) Actually is it possible at all for a young brainy student from some Russian university? What requirements Mr. Buzsaki has for such potential newcomers? :)) G.B.Two Russian scientists (Anatol Bragin and Anton Sirota) worked with me for more than 8 years and I had several short-term visitors. Rustem Khazipov (currently in Marseille) is the most outstanding example. Currently, we speak 12 languages in the lab. Russians tend to be well-educated in math, physics and engineering and such 'hard science' background is almost a must these days in systems neuroscience.