Lecture 3: The Functions of Neural Oscillations

Amplitude = strength Frequency = speed Objectives ========== Function of neural oscillations ------------------------------- 1. 2. 3. a. 4. The above functions of neural oscillations are not mutually exclusive - they\'re just examples of some of the things that we believe, in theory, os...

Amplitude = strength Frequency = speed Objectives ========== Function of neural oscillations ------------------------------- 1. 2. 3. a. 4. The above functions of neural oscillations are not mutually exclusive - they\'re just examples of some of the things that we believe, in theory, oscillations do, their functional significance in terms of cognition and brain activity. ALWAYS ASKED IN THE EXAM\* The function of neural oscillations can be divided into 4 different functions. The first is integration of activity in different brain areas - ie integrating brain activity across different spatial scales. The second is the involvement of spatial processing, like navigation, for instance; we also think theta oscillations might be involved in temporal coding. There\'s also the idea of the oscillation as a background reference signal for cell discharge. Broadly speaking, an oscillation can be thought of as alternating periods of excitation and inhibition, and the general idea is that the brain does something on the excitatory phase and does less on the inhibitory phase. Number 3: timing or co-incident activity; then finally the idea of item integration or segregation - the integration of activity that codes for a memory, for example, how do we integrate cell firing that\'s coding for a particular memory, and how do we segregate (ie separate) that from the cell firing that\'s involved in a completely separate memory. Spatial integration of brain activity ===================================== The general idea of this image is that you can think of the brain as consisting of local connections, global connections, and then between those, regional connections. Local connections, for example, might be connections in the visual cortex here (Q1 and Q2): one little bit of the visual cortex might communicate with another one and vice versa, over a very local (i.e. short) distance. On the other end of the scale you have global communication, where the frontal lobe, for instance, might communicate with the occipital lobe and vice versa, etc - i.e. global communication across wide areas of the brain. Somewhere in between you might have anterior temporal lobe connecting with posterior temporal lobe, across the temporal lobe region - so sort of intermediate distance of communication.  So that\'s the general idea. We know that there are connections that can support local communication, regional communication and big longitudinal fasciculi that can support global communication. The image on the right is a real diffusion image, showing those connections. So we have the architecture that supports that sort of general idea.  So oscillations come in different bands. Originally, these bands and their band widths and the definition of them were somewhat arbitrary, but over time, they\'ve been sort of adjusted slightly. And it does seem that a different band will occur during a particular sort of task. So while we feel we\'ve got a bit of a handle on these things, the bands themselves are still somewhat arbitrary, and tend to blur into each other. Delta and theta might in fact be doing much the same thing, just at slightly different scales, perhaps. Ward (2003) You can see what kind of oscillation you have in your EEG waveform by performing a fast fourier transform. Fast fourier transform will give you an idea of what frequencies are making up a particular waveform. Most often EEG waveforms aren't pure - they're most often composed of a mixture of different frequencies. But these here are predominantly theta, gamma, and alpha respectively. **Astrid von Stein, Johannes Sarthein** **Different frequencies for different scales of cortical integration: from local gamma to long range alpha/theta synchronisation** **International Journal of Psychophysiology 38 301-313, 2000.** **Lawrence M. Ward** **Synchronous neural oscillations and cognitive processes.** **Trends in Cognitive Sciences Vol. 7 No. 12 December 2003** These readings have been put up on canvas, and the top paper is basically the idea behind today\'s lecture. Different oscillation frequencies integrate activity over different distances --------------------------------------------------------------------------------------------------- So the general idea is this; though this is just taking 2 waveforms as an example. So the general idea is that relatively slow waves will tend to integrate areas of the brain that are quite far apart, that are spatially distant, certainly in terms of the amount of time it takes a neuron to conduct information from one part of the brain to another. Theta, for instance, 5Hz, a relatively slow oscillation, will coordinate activity in areas of the brain that are quite far apart, certainly in terms of the conduction velocity of the neurons joining them. A much faster rhythm (example, gamma, 40 Hz), will coordinate activity in areas much closer together, and join by fibres that are much faster conducting. So shorter fibres that are faster. 40 Hz will tend to integrate visual processing - you\'ll see gamma when we engage in visual processing. When we engage in working memory tasks, for instance, we will see activity in the frontal cortex and in the parietal, and we\'ll see theta rhythm. *This is basically the gist of it!* This is basically a summary of the above general idea. Basically in the neocortex and the visual cortex (ie back of the brain), when we engage in visual processing we have areas of the visual cortex that communicate with each other, and they\'ll do so by transmitting information between different areas pretty quickly. So an 18 millisecond conduction time between one area and another would be common; essentially if pink circle on the left is communicating with pink circle on the left, it might take 18 ms to get there, and it might take 18 ms to get back. What you see in this situation is this oscillation occurring, and the di8stance between one excitatory peak of oscillation and another will be, if we have gamma at 56Hz, about 18ms. So basically one area of the brain gets excited and cells in that area will discharge, will send information to another, and by the time that information has reached the other area, the oscillation will have gone from excitation down to inhibition and back to excitation again. In that way, the two areas become locked. Brains impose an oscillation on an area that will then result in particular subregions becoming part of an active network, and others being excluded. The same argument applies to a theta rhythm - much slower, much longer distances between areas - for example, longer distances between the frontal cortex and the hippocampus, or the temporal lobe - and the conduction time between temporal lobe and frontal lobe might be 200ms. So at the red circle, you have a cell in the hippocampus sending information out to some other cell somewhere else, and by the time the information reaches that cell, at the right green circle, the oscillation will have gone from excitation to inhibition back to excitation: red circle = excitation, fire; right green circle = excitation, receive and fire, and so on. And by imposing a theta rhythm in various structures, you can mesh or integrate those structures together in a functional neural network (or at least that\'s the idea). Did demonstration in person but point: different distances require different frequencies. But you could also impose a rhythm. Like a subcortical structure could send a fairly slow rhythm to other structures, and that would tend to pick out that slow communication, or slow oscillation, between frontal lobe and occipital cortex (examples used in demonstration). Or it could send a much faster rhythm. Background reference signal for cell discharge ==================================================================== (eg hippocampal theta) So that was integration across spatial scales. As previously mentioned, these are all somewhat related concepts - you\'ll see them overlapping a bit. The second idea is that oscillations provide a background reference signal for cell discharge. We\'ll be using the hippocampus as an example. It\'s a very interesting structure, involved in memory processing and a variety of other things, but it\'s also quite an old structure in the brian, and it\'s highly conserved throughout the evolutionary process. Here is a rat brain, and they have a relatively massive hippocampus. The second image is a petromyzon, lizard, then an opossum, then a human. Hippocampal Anatomy ------------------- So it\'s a highly conserved structure. The green image is a hippocampus stained for various proteins. You can see in both the colour and black and white staining just how well ordered the hippocampus is. The black line in the black and white stained image is a layer of pyramidal cells, and the smaller black line is the dentate gyrus. It's quite conveniently ordered, because you can put an electrode into an animal (including human) and record cell activity and ongoing oscillations. And you can directly record unit cellular activity, cell discharge, and simultaneously record local EEG, which is essentially the dendritic excitation and inhibition, mainly - so you can record different aspects of cell behaviour.   Hippocampal EEG and Behaviour ----------------------------- When you put an electrode into a behaving animal (here, rat), you\'ll see a very high amplitude regular oscillation, somewhere between 4 and 11 Hz (theta). It\'s quite easy to see with the naked eye, and you generally see it correlated with REM sleep, for instance, and with movement (certainly in a rat). The other sort of activity you\'d record is the non-rhythmic large irregular activity that happens when the animal isn\'t moving, basically. In this image, you can see regular theta activity when a rat swims, or jumps. Essentially, when a rat moves around its environment, you see theta. Oscillations as reference 2 Place ------------------------------------------------------- You can also record **place cells**. The idea here is that you can record populations of cells in a rat (or human) hippocampus that code where the animal is in the environment. So you can record place cells in the hippocampus that will fire when an animal approaches and is in a particular place in a learned environment. Look at the blue area in the place map. The rat has had a bunch of cells that would code for that area of space, and as the rat gets closer to that area, that population of cells will start to fire faster. **When an animal is in a specific place within their environment, a hippocampal place cell will fire, and different cells code for different areas of space.** Phase Precession ---------------- ### a) Coding in space What does all this have to do with oscillations? You combine the theta rhythm in the hippocampus and the cell firing that\'s coding for where you are in the environment. And there\'s a phenomenon known as **phase precession** - in which the combination of the place cell firing relative to the background theta activity in the hippocampus tells us how far away we are from the place that\'s been coded by that cell. And it does that by **precessing** (firing slightly earlier on the theta wave). So as an animal moves towards the centre of the place being coded for, the cell fires slightly earlier on the ongoing theta cycle (A). As the rat moves towards the centre, the cell fires earlier and earlier on the background theta wave (1, 2, 3, 4, 5). As the rat moves away from that centre, the cells still fire earlier on the wave. So in that way, it\'s argued that the place cell and the ongoing background theta activity can code for both places directly, and can code relatively how close you are to a particular place. It\'s complicated though: there\'s a lot of different cells that make up a particular place code, and there\'s lots of other different cells that make up different place codes. And in an environment, they fire quite close together, and all precess against the same background activity in different sorts of ways.  #### Precession in human hippocampus and entorhinal cortex We\'ve seen precession in humans as well. If you record from a human hippocampus and entorhinal cortex, we also have place cells that precess against background theta activity; they precess more when we have a preferred goal that we're moving towards relative to when we don\'t. So there\'s more coding going on than just space. ***Question: do place cells still fire when you\'re navigating through virtual space?*** ***Answer: yes!** Just not as well. When you wheel a rat around its environment, the place cells still fire, just not as compellingly as if the animals self propelling through the environments. This suggests we possibly need a little bit of motor activity as well (maybe also vestibular activity)*. Phase Precession ------------------------------------- ### b) Coding in time **Bar press = reward after a certain time since last reward** The other thing about the background wave idea is that we don\'t only code for space, we can code for time too. Here\'s an experiment: a rat has been trained to bar press for reward, but they\'ll only get the reward after a certain amount of time. So basically they have to estimate time since their last reward, before they can press again and get a reward. Found: cells in the hippocampus will code for this passing time by precessing on the background theta wave. Co-incident activity. Theta phase and LTP ----------------------------------------- Another thing we can do is look at co-incident activity, like oscillation (theta phase, in this example) and some process, like LTP. Remember that an oscillation is an excitation and an inhibition → more happens in the excitatory phase. So if input is co-incident with excitation, things happen; if input is co-incident with inhibition, less happens. LTP: the general idea is that we form memories by modifying populations of synapses, and the way we modify the synapse is basically co-incident pre- and post-synaptic excitation. So if you excite one cell that\'s connected to another one repeatedly, and strongly, the synapses (i.e. connections between those two cells) strengthen - so it lays down a sort of memory trace in a network. The basic idea is that you can stimulate the pre- and post-synaptic cell in the hippocampus of an experimental animal, and if you stimulate it at a relatively low rate, you\'ll get a certain output in a postsynaptic cell (the baseline in the graph); then if you increase the (input) stimulation to about 200Hz (about as fast as any cell will fire), this excites the pre- and postsynaptic elements, and you get an increased postsynaptic response that can last hours (even days or months) - this means that this sort of synaptic modification, this increase, is a very good model for memory. What does this have to do with oscillations? If you put pulsed input solely on the excitatory phase (peak) of theta in the hippocampus, you get LTP. If you put the same input at the inhibitory phase (trough) of ongoing oscillatory activity, you get no LTP - sometimes you get the reverse, a deep depression of the synaptic efficacy. **So the oscillation allows you to code and do something with an input, depending on when, on that oscillation, that input arrives.**  #### Hippocampal cell discharge correlations (Wilson and McNaughton, 1994) In a more practical sense, combine those 2 ideas, oscillations and LTP. The image below shows some hippocampal cells that are relatively uncorrelated in the precondition. In the run condition, an animal is learning an environment, and being rewarded for doing so. You can see here that different hippocampal cells become better connected with each other. This requires theta activity while the animal is exploring the environment. This post condition here is a subsequent sleep condition - probably for memory consolidation. You can see here that the established increased correlations between cell discharges in the learning condition are maintained in the post learning condition, suggesting a memory trace has been laid down. Item integration/segregation (segmentation) Theta/Gamma Nesting & cell binding/segregation =============================================================================================================== I did mention that we have many different places in our environment, and many different networks of place cells coding for those almost simultaneously - we don\'t just have one place that were interested in, we\'re constantly looking at where we are in the environment, and different subpopulations of hippocampal cells (and many other cells across quite a large network) are involved in this process. So how do we, first of all, integrate those cells that should be considered together as coding for a particular space; and how do we segment/segregate/separate populations that are coding for different places in our environment? The idea is: we'll never see a pure theta wave. So in the hippocampus, we see theta, and on top of the theta wave, particularly on the excitatory phase, we'll see gamma activity - a much faster waveform → ie **gamma is nested on the theta wave**. When there are cells coding for a particular place that's precessing on the wave, as the animal moves closer to the place field, all those cells tend to fire on one gamma cycle. You\'ll also have cells coding for a completely different place, that tend to fire on a different gamma cycle. The ongoing theta wave can handle a multitude of separate spatial coding networks. It can precess or separate them, and separation is due to them firing on separate gamma waves.  Evidence in rats navigating through an environment is reasonably convincing - the gamma, and theta precession on separate gamma waves. But it\'s harder to get an in humans, because recording from signal cells in the human hippocampus is more difficult. Usually we can only do it when we have indwelling electrodes that are monitoring potential seizure activity. But the evidence is accruing. There\'s certainly theories that working memory in humans is basically controlled in much the same way. The general idea is that the separate phases of the ongoing gamma rhythm hold different memory components, and all of these are lumped together on a phase of the theta wave.  **From Ward, 2003** If you\'re holding something in short term memory, like letters. For example: HBOCTPE → if you're holding those separate items in working memory, they tend to get held in a sequential stack that is coded for by different phases of the ongoing gamma rhythm. So each memory item gets stored in a separate phase of the ongoing gamma, and all of these are lumped together sequentially and repeated across subsequent theta phases. When you read them out, the idea is that they precess in the way that we\'ve seen for spatial memory. In the same way a rat will separate out different place fields coded for by different place cells, we separate out different items of working memory by the same gamma-theta nest (gamma rhythm nested on theta wave). When I give you a working memory task - remember 3759 - 3 goes in one gamma cycle, 7 goes on another gamma cycle, 5 goes on the next, 9 goes on the next. Rehearsing the numbers is held and repeated on subsequent theta waves. This is just a theory!!! There is less evidence for this, as how working memory works. But if this was true, one of the consequences would be, for people who have greater working memory, the relative frequency of gamma and theta might be different. People with large working memory would have faster gamma and slower theta, so you can fit more stuff (fit more items - gamma cycles - on the theta wave).